M-

U'X.'^,
dt**-^
c^.
"V,.
Presented
to the
LIBRARY
of
the
UNIVERSITY
OF TORONTO
by
Mr. J. R. McLeod
A
TREATISE
ON
THE
MATHEMATICAL
THEORY
OF
ELASTICITY
CAMBRIDGE UNIVERSITY PRESS
WAREHOUSE,
C. F.
CLAY,
Manager.
EonSon:
FETTER
LANE,
E.G.
SUugoia: M.
WELLINGTON STREET.
Itipjifl:
F. A. BR0CKHAU8.
fiiitt gotk:
THE MACMILLAN COMPANY.
JSombag
an* ealtulta: MACMILLAN AND
CO.,
Ltd.
[^{2 Right* reterved.]
A TREATISE
ON THE
MATHEMATICAL THEORY
OF
ELASTICITY
BY
A. E. H.
LOVE, M.A., D.Sc,
F.R.S.
FORMERLY FELLOW OF ST JOHN'S
COLLEGE,
CAMBRIDGE
HONORARY FELLOW OF QUEEN's
COLLEGE,
OXFORD
SEDLEIAN PROFESSOR OF NATURAL PHILOSOPHY IN THE UNIVERSITY OF OXFORD
SECOND EDITION
Cambridge
:
at the
University
Press
1906
CTambrtliBe:
PBINTED BY JOHN
CLAY,
M.A.
AT THE CNIVEBSITY PRESS.
PREFACE.
rpHIS
book is a second edition of one with the same title which was
-*-
published by
the
Cambridge University
Press in two volumes
bearing
the dates 1892 and 1893. At the
time,
about five
years ago,
when it first
became
necessary
to think
seriously
about a new
edition,
a number of friends
had sent me criticisms of
particular
sections of the book and
suggestions
for
improvement
in matters of detail.
Among
these friends Prof W. J. Lewis
and Prof. W. M'^F. Orr must be named with
especial gratitude.
I knew then
that two or three
Chapters ought
to be
rewritten,
and that the results of
several new researches
ought
to be
incorporated,
but I did not
contemplate
a
very
extensive revision. The task of
rearranging
the old
matter,
with some
considerable additions and a few
slight
omissions,
became so
distasteful,
and
the result
appeared
so
unsatisfactory,
that at
length
I abandoned the
attempt,
and wrote a new book
containing
some extracts from the old one. The
science of
Elasticity

the mechanics of solid bodies as

they really
are

is
so
important
in
itself,
and the
physical
notions and
analytical processes
belonging
to the
theory
are .so
widely
used in other branches of
Physics,
that no
apology
seems to be
necessary
for the course that has been
pursued.
In the
selection,
and the mode of
presentation,
of the matter three
objects
have been
kept
in view : to make the book useful to
engineers,
or
others,
whose aims are
chiefly practical,
to
emphasize
the
bearing
of the
theory
on
general questions
of Natural
Philosophy,
to afford a
reasonably complete
picture
of the state of the science as it is
to-day.
The desire to be useful
has led me to undertake some rather laborious arithmetical
computations,
physical
interest has
prompted something
more than a
passing
reference to
several matters which lie outside the strict
scope
of the mechanical
theory,
completeness
has
required
the inclusion of some rather
long analytical
investigations.
At the same
time,
purely
technical
matters,
such as
descrip-
tions of
apparatus
and calculations
relating
to
particular
structures,
have
been excluded
;
related
subjects,
such as the
production
of strain
by unequal
VI
PREFACE
heating,
the
rendering
of
glass doubly refracting by
strain,
the
theory
of the
luminiferous medium
regarded
as an elastic
solid,
have received but a
slight
measure of attention
;
detailed discussion of
problems
of which the interest
is
mainly
mathematical has been
kept
within rather narrow bounds.
Numerous references to authorities on
these,
as well as on
other,
matters
have, however,
been introduced.
One
change
which has been made
may perhaps require
a word of defence.
The notation for
components
of stress and
components
of strain is different
from that
adopted
in the first edition. A wish for this
change
was
expressed
to me in several
quarters,
and I have
myself
been much
impressed
with the
advantages
of a notation which
conveys
its own
meaning. Although
I still
think that Kelvin and Tait's
notation,
which was
adopted
before,
has
many
merits,
yet
I did not feel that I should be
justified
in
neglecting
the
repre-
sentations that had been made to me.
The student to whom the
subject
is new is advised to turn as
early
as
possible
to
Chapter
V,
where he will find a condensed
recapitulation
of the
most essential
parts
of
previous Chapters,
some indications of the kind of
problems
which can be treated
mathematically,
and of methods of
dealing
with
them,
and a number of results of which the
verification,
or direct
investigation,
will be useful to him as exercises.
It remains to
attempt
to
express my
thanks to those who have
helped
me with this book. Three friends have laid me under
especially heavy
obligations
: Prof. J. Larmor and Prof. H. Lamb have read most of the
proofs,
and have sent me
many kindly
criticisms and
many helpful suggestions
in
regard
to matters of
principle
;
and Prof. H. M. Macdonald has read all the
proofs,
and his
vigilance
has detected
many misprints
and errors of detail.
Dr A.
Timpe,
who is
translating
the book into
German,
has also
kindly
called
my
attention to a few
passages
which needed correction
;
and the
scrupulous
care which he has bestowed
upon
the translation leads me to
hope
that few serious errors remain. To the
Syndics
of the Press
my
thanks
are due for their kindness in
acceding
to
my proposal
to
print
the new
edition in a
single
volume,
and the readiness with which the staff of the
Press have met all
my
wishes in
regard
to
printing
and
diagrams
deserves
more than a word of
recognition.
A. E. H. LOVE.
OxFOBD, December,
1905.
CONTENTS.
Historical Introduction.
PAGE
1
Scojje
of
Hi.story.
Galileo's
enquiry.
Enunciation of Hooke's Law. Mariotte's in-
vestigation.s.
The
problem
of the elastica. Euler's
theory
of the
stability
of
struts. Researches of Coulomb and
Young.
Euler's
theory
of the vibrations of
bars.
Attempted theory
of the vibrations of bells and
plates.
Value of the
researches made before 1820. Navier's
investigation
of the
general equations.
Imi)ulse given
to the
theory by
Fresnel.
Cauchy's
first memoir.
Cauchy
and
Poi.sson's
investigations
of the
general equations by
means of the "molecular"
hypothesis.
Green's introduction of the
strain-energy-function.
Kelvin's
appli-
cation of the laws of
Thermodynamics.
Stokes's criticism of Poisson's
theory.
The
controvei-sy concerning
the number of the "elastic constants." Methods
of solution of the
general problem
of
equilibrium.
Vibrations' of solid bodies.
Propagation
of waves. Technical
problems.
Saint-Venant's theories of torsion
and flexiu-e.
Equipollent
loads.
Simplifications
and extensions of Saint-Venant's
theories. Jouravski's treatment of
shearing
stress in beams. Continuous beams.
Kirchhoff's
theory
of
springs.
Criticisms and
applications
of Kirchhoff's
theory.
Vibrations of bars.
Impact. Dynamical
resistance. The
problem
of
plates.
The
KirchhofF-Gehring theorj".
Clebsch's modification of this
theory.
Later
researches in the
theory
of
plates.
The
problem
of shells. Elastic
stability.
Conclusion.
Chapter I. Analysis ok strain.
ART,
1. Extension 32
2. Pure Shear
3.3
3.
Simple
Shear
33
4.
Displacement
3.5
5.
Displacement
in
simple
extension and
simple
shear
3,5
6.
Homogeneous
strain
36
7. Relative
displacement
37
8.
Analysis
of the relative
displacement
38
9. Strain
corresponding
with small
displacement
39
10.
Components
of strain
40
11. The strain
quadric
41
12. Transformation of the
components
of strain
42
a 5
viii
CONTENTS
ART.
I'AGK
13. Additional methods and results
-13
14
Tyjies
of strain,
{a)
Uniform
extension, (6) Simple
extension, (t) Shearing strain,
\d)
Plane strain
44
15. Relations
connecting
the
dilatation,
the rotation and the
displacement
. . 46
16. Resolution of
any
strain into dilatation and
shearing
strains . . . .47
17. Identi&il relations between
components
of strain
49
18.
Displacement corresponding
with
given
strain
50
19. Cur\ilinear
orthogonal
coordinates
51
20.
Comjwuents
of strain i-eferred to curvilinear
orthogonal
coordinates . . .53
21. Dilatation and rotation referred to curvilinear
orthogonal
coordinates . . .54
22.
Cylindrical
and
polar
coordinates
56
Appendix to Chapter I. The general theory of strain.
23.
Introductory
57
24. Strain
corresponding
with
any displacement
57
25. Cubical dilatation
59
26.
Reciprocal
strain
ellipsoid
.60
27.
Angle
between two curves altered
by
strain 60
28. Strain
ellipsoid
61
29. Alteration of direction
by
the strain 62
30.
Application
to
cartography
63
31. Conditions satisfied
by
the
displacement
63
32. Finite
homogeneous
strain 64
33.
Homogeneous pure
strain 65
34.
Analysis
of
any homogeneous
strain into a
pure
strain and a rotation . . .67
35. Rotation 67
36.
Simple
extension 68
37.
Simple
shear 68
38. Additional results
relating
to shear 69
39.
Comiwsition
of strains 69
40. Additional results
relating
to the
composition
of strains 70
Chapter II. Analysis of stress.
41.
Introductory
. 72
42. Traction across a
plane
at a
point
72
43. Surface tractions and
body
forces 73
44.
Equations
of motion 74
45.
Equilibrium
76
46. Law of
equilibrium
of surface tractions on small volumes 75
47.
Specification
of stress at a
point
75
48. Mea.sure of stress
.77
49. Transformation of
stress-components
78
50. The stress
quadric
79
51.
Tyi)e8
of streas.
(a) Purely
normal
stress, (6) Simple
tension or
jiressure,
(c) Shearing stress, {d)
Plane stress 79
CONTENTS
IZ
AllT.
52. Resolution of
any stress-system
into uniform tension and
shearing
sti-ess
53. Additional results
54. The
stress-equations
of motion and of
equilibrium
55. Uniform stress and
uniformly varying
stress
56. Observations
concerning
the
stress-equations
57.
Graphic representation
of stress ....
58.
Stress-equations
referred to curvilinear
orthogonal
coordinates
59.
Special
cases of
stress-equations
referred to curvilinear
orthogonal
coordinates
PAGE
81
81
82
84
85
86
87
89
Chapter III. The elasticity of solid bodies.
60.
Introductory
90
61. Work ;iud
energy
90
62. E.xistence of the
strain-energy
-function
. . .92
63. Indirectness of
e.xperimental
results . .
.^__^__,-
. ... 94
64. Hooke's Law
/"T
95
65. Form of the
strain-energy-function
96
66. Elastic constants
97
67. Methods of
determining
the stress in a
body
98
68. Form of the
strain-energy-function
for
isotropic
solids 99
69. Elastic constants and moduluses of
isotropic
solids 100
70. Observations
concerning
the stress-strain relations in
isotropic
solids . . 101
71.
Magnitude
of elastic constants and moduluses of some
isotropic
.solids . . 103
72. Elastic consfamts in
general
. . . 103
73.
Moduluses
of
elasticity
'
. . . 104
74. Thermo-elastic
equations
106
75. Initial .stress
107
Chapter IV. The relation between the mathematical
theory of elasticity and technical mechanics.
I
76. Limitations of the mathematical
theory
77. Stress-strain
diagrams
.
78. Elastic Hmits ....
79. Time-eflfects.
Plasticity
.
80.
Viscosity
of solids
81.
jEolotropy
induced
by permanent
set
82.
Repeated loading
....
83.
Hypotheses concerning
the conditions of
rupture
84.
Scope
of the mathematical
theory
of
elasticity
.
110
111
113
114
116
116
lie
117
119
Chapter V. The
equilibrium
of isotropic ela.stic solids.
85.
Recapitulation
of the
general theory
122
86.
Uniformly varying
stress,
(a)
Bar stretched
by
its own
weight, (b) Cylinder
immersed in
fluid, (c) Body
of
any
form immersed in fluid of same
density,
(rf)
Round bar twisted
by couples
123
X CONTENTS
AKT. I'AGE
87. Biir bent
by couples
124
88. Discussion of the .solution for the
bending
of a bar
by
terminal
couple
. .127
89. Saint-Venant's
principle
129
90.
Bectaugular plate
bent
by couples
129
91.
Equations
of
equilibrium
in terms of
displacements
130
92.
Equilibrium
under surface tractions
only
132
93. Various methods and results 133
94. Plane strain and
plane
stress 134
95.
Bending
of narrow
rectangular
beam
by
terminal loa<l 136
96.
Equations
referred to
orthogonal
cm-vilinear coordinates . . . . . 138
97. Polar coordinates 138
98. Radial
displacement. Spherical
shell under internal and external
pressure.
Compression
of a
sphei-e by
its own
gravitation
139
99.
Displacement symmetrical
about an axis 140
100. Tube under
pressure
141
101.
Application
to
gun
construction . . . . .143
102.
Rotating cylinder. Rotating
shaft.
Rotating
disk 143
Chapter VI.
Equilibrium
of .eolotkopic elastic solid bodies.
103.
Symmetry
of structure 146
104. Geometrical
symmetry
147
105. Elastic
symmetry
148
106.
Isotropic
solid 152
107.
Symmetry
of
crystals
152
108. Clas.sification of
crystals
154
109.
Elasticity
of
crystals
156
110. Various
tyjws
of
symmetry
157
111. Material with three
rectangular planes
of
symmetry.
Moduluses . . .158
112. Extension and
bending
of a bar 159
113. Elastic constants of
crystals.
Results of
experiments
160
114. Curvilinear
asolotropy
161
Chapter VII. General theorems.
115. The variational
equation
of motion 163
116.
Applications
of the variational
equation
. 164
117. The
general problem
of
equilibrium
166
118.
Uniqueness
of solution 167
119. Theorem of minimum
energy
. 168
120. Theorem
concerning
the
i)0tential energy
of deformation . . . . . 170
121. The
reciprocal
theorem 170
122. Determination of
average
strains 171
123.
Average
strains in an
isotropic
solid
body
172
124. The
general problem
of vibrations.
Uniqueness
of solution .... 173
125. Flux of
energy
in
vibratory
motion 174
126. Free vibrations of elastic solid bodies 175
127. General theorems
relating
to free vibrations . . . . . . .177
128. Load
suddenly applied
or
suddenly
reversed 178
CONTENTS
XI
Chapter VIII. The transmission of force.
ART.
FAQS
129.
Introductory
180
130. Force at a
point
180
131. First
type
of
simple
solutions
182
132.
Typical
nuclei of strain
183
133. Local
perturbations
186
134. Second
type
of
simple
solutions
187
135. Pressure at a
point
on a
plane boundary
188
136. Distributed
pressure
189
137. Pressure between two bodies in contact. Gleometrical
preliminaries
. . 190
138. Solution of the
problem
of the
pressure
between two bodies in contact . . 192
139. Hertz's
theory
of
impact
195
140.
Impact
of
spheres
197
141. Effects of nuclei of strain referred to
polar
coordinates . . . . . 198
142. Problems
relating
to the
equilibrium
of cones
200
Chapter IX. Two-dimensional elastic systems.
143.
Introductory
201
144.
Displacement corresponding
with
plane
strain 201
145.
Displacement corresponding
with
plane
stress 203
146. Generalized
plane
stress 206
147. Introduction of nuclei of strain 205
148. Force
operative
at a
point
206
149. Force
operative
at a
point
of a
boundary
207
150. Case of a
straight boundary
208
151. Additional results
(i)
the stress
function, (ii)
normal tension on a
segment
of
a
straight edge, (iii)
force at an
angle
208
152.
Typical
nuclei of strain in two dimensions 209
153. Transformation of
plane
strain 211
154. Inversion
212
155.
Equilibrium
of a circular disk under forces in its
plane, (i)
Two
opposed
forces at
points
on the rim.
(ii) Any
forces
applied
to the rim.
(iii) Heavy
disk on horizontal
plane
. . . . . . . . . . .213
156.
Examples
of transformation ..........
216
Chapter X. Theory of the integration
of the
equations
OF
equilibriu.\i
of an isotropic elastic solid body.
157. Nature of the
problem
....
158. Re-sume of the
theory
of Potential
159.
Description
of Betti's method of
integration
160. Formula for the dilatation ....
161. Calculation of the dilatation from surface data
162. Formula: for the
components
of rotation .
217
218
220
221
223
SS4
xil
CONTENTS
ART.
PAGE
163. Calculation of the rotation from surface data 224
164.
Body
bounded
by plane

FormulsB for the dilatation 225

165.
Body
bounded
by jilane

Given surface
displacements
227
166.
Body
bounded
by plane

Given surface tractions 228

167. Historical Note 230
168.
Body
bounded
by plane

Additional results 231

169. Formula) for the
displacement
and strain . . 232
170. Outlines of various methods of
integration
234
Chapter XI. The
equilibrium
of an elastic sphere and
RELATED PROBLEMS.
171.
Introductory
236
172. Solution in
spherical
harmonics of
positive degrees
236
173. The
sphere
with
given
surface
displacements
238
174. Generalization of the
foregoing
solution,
(i) Integration by
means of
poly-
nomials,
(ii) Body
force
required
to maintain a state of strain with zero
surface
displacement, (iii)
General method for
integrating
the
equations
by
means of series 239
175. The
sphere
with
given
surface tractions 240
176. Conditions
restricting
the
prescribed
surface tractions 243
177. Surface tractions directed
normally
to the
boundary
244
178. Solution in
spherical
harmonics of
negative degrees
245
179.
Sphere subjected
to forces
acting through
its volume. Particular solution . 246
180.
Sphere
deformed
by body
force
only
247
181.
Gravitating incompressible sphere
248
182. Deformation of
gravitating incompressible sphere by
external forces . . 250
183.
Gravitating nearly spherical body
263
184.
Rotating sphere
263
185. Tidal deformation. Tidal effective
rigidity
of the Earth 266
186. Plane strain in a circular
cylinder
257
187.
Applications
of curvilinear coordinates 259
188.
Symmetrical
strain in a solid of revolution
260
189.
Symmetrical
strain in a
cylinder
263
Chapter XII. Vibrations of spheres and cylinders.
190.
Introductory 265
191. Solution
by
means of
spherical
harmonics

. . 266
192. Formation of the
boundary
conditions for a
vibrating sphere
.... 268
193.
Incompres.sible
material
271
194.
Frequency equations
for
vibrating sphere
271
195. Vibrations of the finst class
272
196. Vibrations of the second class
273
197. Further
investigations
of the vibrations of
spheres
274
198. Radial vibrations of a hollow
sphere
. 274
199. Vibrations of a circular
cylinder 276
200. Torsional vibrations
276
201.
Longitudinal
vibrations
276
202. Transverse vibrations
278
CONTENTS
Xiii
Chapter XIII. The
propagation of waves in elastic
SOLID
MEDIA.
ART.
p^^OE
203.
Introductory
281
204. Waves of dilatation and waves of distortion
281
205. Motion of a surface of
discontinuity.
Kinetnatical conditions
.... 283
206. Motion of a surface of
discontinuity. Dynamical
conditions
.... 284
207.
Velocity
of waves in
isotropic
medium
285
208.
Velocity
of waves in
Eeolotropic
solid medium .
. . . . . 286
209. Wave-surfaces
287
210. Motion determined
by
the characteristic
equation
289
211.
Arbitrary
initial conditions

291
212. Motion due to
body
forces
293
213. Additional results
relating
to motion due to
body
forces
294
214. Waves
propagated
over the .surface of an
isotropic
elastic solid
body
. . 296
Chapter XIV.
Torsion.
215. Stress and Strain in a twisted
prism 298
216. The torsion
problem 299
217. Method of solution of the torsion
problem
301
218.
Analogies
with
Hydrodynamics 302
219. Distribution of the
shearing
stress
303
220.
Strength
to resist torsion
304
221. Solution of the torsion
problem
for certain boundaries
305
222. Additional results
306
223.
Graphic expression
of the results
308
224.
Analogy
to the form of a stretched membrane loaded
uniformly
. . . 310
225.
Twisting couple
310
226. Torsion of
eeolotropic prism
312
Chaffer XV. The bending of a beam by terminal
transverse load.
227. Stress in a bent beam
314
228. Statement of the
problem
315
229.
Neces.sary type
of
shearing
stress 316
230. Formulae for the
displacement
318
231. Solution of the
problem
of flexure for certain boundaries :
(a)
The
circle,
(6)
Concentric
circles, (c)
The
ellipse, (d)
Confocal
ellipses, (e)
The
rectangle,
(/)
Additional results 320
232.
Analysis
of the
displacement
:
(a)
Curvature of the strained
central-line,
(b)
Neutral
plane, (c) Obliquity
of the strained
cro.ss-sectious, (rf) Deflexion,
(e) Twist, (/)
Anticlastic
curvature, (ff)
Distortion of the cross-sections into
curved surfaces 323
233. Distribution of
shearing
stress
327
234. Generalizations of the
foregoing theory
:
(a) Asymmetric loading, (6)
Combined
strain, (c) .<Eolotropic
material 328
XIV
CONTENTS
ART.
PAQB
235. Criticisms of certain methods :
(a)
A method of
determining
the
shearing
stress in the case of
rectangular sections, (6)
Extension of this method to
curved
boundaries, (c)
Form of
boundary
for which the method
gives
the
correct
result, (d)
Defectiveness of the method in the case of an
elliptic
section, (e)
Additional deflexion described as
"
due to
shearing," (/)
Defective
method of
calculating
this additional deflexion
331
Chapter XVI. The bending of a beam loaded uniformly
ALONG ITS LENGTH.
236.
Introductory
334
237. Stress uniform
along
the beam 334
238. Stress
varying uniformly along
the beam .
336
239.
Uniformly
loaded beam. Reduction of the
problem
to one of
plane
strain . 339
240. The constants of the solution 342
241. Strain and stress in the elements of the beam 343
242. Relation between the curvature and the
bending
moment 345
243. Extension of the central-line 347
244. Illustrations of the
theory

(a)
Form of solution of the related
problem
of
plane
strain.
(6)
Solution of the
problem
of
plane
strain for a beam of
circular section bent
by
its own
weight, (c)
Correction of the curvature in
this case,
(o?)
Case of narrow
rectangular
beam loaded
along
the
top
treated
as a
problem
of
"generalized plane
stress."
(e)
Narrow
rectangular
beam
supported
at the ends and loaded
along
the
top
347
Chapter XVII. The theory of continuous beams.
245. Extension of the
theory
of the
bending
of beams 350
246. The
problem
of continuous beams 353
247.
Single span, (a)
Terminal forces and
couples. (6)
Uniform load.
Supported
ends,
(c)
Uniform load. Built-in ends,
{d)
Concentrated load.
Supported
ends,
(e)
Concentrated load. Built-in ends 355
248. The theorem of three moments,
(a)
Uniform load.
(6) Equal spans.
(c)
Uniform load on each
span, (d)
Concentrated load on one
span
. 357
249.
Graphic
method of solution of the
problem
of continuous beams . . . 360
250.
Development
of the
graphic
method
362
Chapter XVIII. General theory of the bending and
twisting of thin rods.
251.
Introductory . 355
252. Kinematics of thin rods
386
253. Kinematical formula)
366
254
Equations
of
equilibrium 37O
255. The
ordinary approximate
theory
. 372
256. Nature of the strain in a bent and twisted rod .
'
373
257.
Approximate
formulae for the strain
378
258. Discussion of the
ordinary approximate
theory
. . .
'
. . . . 377
259. Rods
naturally
curved
379
I
CONTENTS
XV
Chapter XIX. Problems concerning the
equilibrium
of
THIN RODS.
ART.
PAGE
260. KirohhoflF's kinetic
analogue 3P2
261. Extension of the theorem of the kinetic
analogue
to rods
naturally
curved . 383
262. The
problem
of the elastica
384
263. Classification of the forms of the elastica.
(a)
Inflexional
elastica, (6)
Non-
inflexional elastica
385
264.
Buckling
of
long
thin .strut under thrust . . . . . , . 388
265.
Computation
of the
strain-energy
of the strut
389
266. Resistance to
buckling 390
267. Elastic
stability 392
268.
Stability
of inflexional elastica
392
269. Rod bent and twisted
by
terminal forces and
couples
394
270. Rod bent to helical form
395
271.
Theory
of
spiral springs 396
272. Additional results,
(a)
Rod
subjected
to terminal
couple, (6) Straight
rod
with initial
twist, (c)
Rod bent into circular
hoop
and twisted
uniformly,
(d) Stability
of rod
subjected
to
twisting couple
and
thrust, (e) Stability
of
flat blade bent in its
plane 398
. 402
. 403
. 406
. 405
273. Rod bent
by
forces
applied along
its
length
274. Rod bent in one
plane by
uniform normal
pressure
275.
Stability
of circular
ring
under normal
pressure
276.
Height
consistent with
stability
Chapter XX. Vibrations of rods. Problems of dynamical
RESISTANCE.
277.
Introductory
407
278. Extensional vibrations 408
279. Torsional vibrations 409
280. Flexural vibrations 409
281. Rod fixed at one end and struck
longitudinally
at the other . . . .411
282. Rod free at one end and struck
longitudinally
at the other .... 415
283. Rod loaded
suddenly
416
284.
Longitudinal impact
of rods 418
285. Problems of
dynamical
resistance
involving
transverse vibration . . . 420
286. The
whirling
of shafts 421
Chapter XXI. Small deformation of naturally curved rods.
287.
Introductory
423
288.
Specification
of the
displacement
423
289. Orientation of the
principal
toi-sion-flexurc axes 424
290. Curvature and twist 425
291.
Simplified
formute 426
292. Problems of
equilibrium, (a) Incomplete
circular
ring
bent in its
plane.
(6) Incomplete
circular
ring
bent out of its
plane
427
293. Vibrations of a circular
ring, (a)
Flexural vibrations in the
plane
of the
ring.
(b)
Flexural vibrations at
right angles
to the
plane
of the
ring, (c)
Torsional
and extensional vibrations
430
xvi
CONTENTS
Chapter
XXII. The stretching and bending of plates.
ART.
PAE
294.
Specification
of stress in a
plate
434
295. Transformation
of stress-resultants
and
stress-couples
435
296.
Equations
of
equilibrium
436
297.
Boundary
conditions
437
298. Relation between the flexural
couples
and the curvature 442
299. Method of
determining
the stress in a
plate
444
300. Plane stress
446
301. Plat stretched
by
forces in its
plane
446
302. Plat* bent to a state of
plane
stress 449
303. Generalized
plane
stress
450
304. Plate bent to a state of
generalized plane
stress 452
305. Circular
plate
loaded at its centre
454
306. Plate in a state of stress which is uniform or varies
uniformly
over its
plane
. 454
307. Plate bent
by pressure
uniform over a face 456
308. Plate bent
by pressure varying uniformly
over a face 458
309. Circular
plate
bent
by
uniform
pressure
and
supported
at the
edge
. . . 460
310. Plate bent
by
uniform
pressure
and
clamped
at the
edge
461
311. Plate bent
by uniformly varying pressure
and
clamped
at the
edge
. . . 463
312. Plate tent
by
its own
weight
464
313.
Approximate theory
of the
bonding
of a
plate by
transverse forces . . . 465
314. Illustrations of the
approximate theory, (a)
Circular
plate
loaded
symmetric-
ally. (6) Application
of the method of inversion,
(c) Rectangular plate
supported
at a
pair
of
opposite edges, {d)
Transverse vibrations of
plates,
(e)
Extensional vibrations of
plates
466
Chapter XXIII. Inextensional deformation of curved
PLATES or shells.
315.
Introductory
471
316.
Changes
of curvature in inextensional deformation
472
317.
Typical
flexural strain
474
318. Method of
calculating
the
changes
of curvature
476
319. Inextensional deformation of a
cylindrical
shell,
(a)
Formulas for the dis-
placement. (6) Changes
of curvature
477
320. Inextensional deformation of a
spherical
shell,
(o)
Formulae for the
displace-
ment.
(6) Changes
of curvature
479
321. Inextensional vibrations,
(i) Cylindrical
shell,
(ii) Spherical
shell . . 486
Chapter XXIV. General theory of thin plates and shells.
322. Formulae
relating
to the curvature of surfaces ....
323.
Simplified
formulae
relating
to the curvature of surfaces .
324. Extension and curvature of the middle surface of a
plate
or shell
325. Method of
calculating
the extension and the
changes
of curvature
326.
.Formulee
relating
to small
displacements
488
490
491
492
494
CONTENTS
xvii
ART.
PAGE
327. Nature of the strain in a bent
plate
or shell
....... 498
328.
Specification
of stres.s in a bent
plate
or shell . . . . . . .501
329.
Approximate
formulse for the
strain,
the
stress-resultants,
and the stress-
couples
502
330. Second
approximation
in the case of a curved
plate
or shell
.... 506
331.
Equations
of
equilibrium 507
332.
Boundary
conditions
610
333.
Theory
of the vibrations of thin shells
511
334. Vibrations of a thin
cylindrical
shell,
(a)
General
equations. (6)
Extensional
vibrations,
(c)
Inextensional vibrations,
(d)
Inexactness of the inexten-
sional
displacement, (e)
Nature of the correction to be
applied
to the
inextensional
displacement
516
335. Vibrations of a thin
spherical
shell
622
336. Problems of
equilibrium
525
337. Problems of
stability, (a) Buckling
of a
rectangular plate
under thrusts in its
plane. (6) Collapse
of a tube under external
pressure
528
NOTES.
A.
Terminology
and Notation
531
B. The notion of stress
533
C.
Applications
of the method of
moving
axes
536
D. Lines of stress due to
pressm-e
between two bodies
541
E. Stress in a beam loaded
uniformly
. . . . 541
F. Extensional vibrations of a
plane plate
541
INDEX.
Authors cited 643
Matters treated
.... 546
ADDITIONS AND CORRECTIONS.
p. 10,
footnote 37. Add
"
Reprinted
in Stokes's Jifath.
andPhys. Papen,
vol. i.
(Cambridge
1880), p.
75."
p. 27,
line 1. For
"
M.
Phillips"
read "E.
Phillips."
p.
66,
line 16. Insert
"square
of the
"
before
"central radius vector."
p. 112,
lines 3 and 4 from foot. Delete "In
particular,
there is no
yield-point
under thrust."
For some
examples
of the determination of the
yield-point
under thrust
{Quetsch-
grenze)
see
Bauschinger, Mittheilungen,
xiii.
p.
133. The
equation
numbered
"
(29 bis)
"
should be numbered
"
(30 bis)."
p. 143, equations (61).-
For
''
,(^-^ -^y
read
"^(^-'|)."
p.
273. After
equation (39)
add
"
The
displacement has,
in
general,
both transver.se and
radial
components,
but the rotation has no radial
component."
p.
402,
lines. For
"
R{yi-y)"
read
"E{ji-y)."
line 20. For
"
dyj,j?s
"
arid
"
3</)/a
"
read
"
df/ds
"
and
"
dcji/ds."
p.
428.
Fig.
64
suggests
that the
displacement
is determined so tliat the ends of the rod
move
along
the line of action of the forces R. To secure the .satisfaction of this
condition an additional
displacement,
which would be
possible
in a
rigid body,
must be
superposed upon
the
displacement given
in the text. I am indebted for this correction
to Mr G. C.
Calliphronas.
p. 470,
line 2. For
"
M.
Barthdldmy
"
read
"
A.
Bai-thdldmy."
HISTORICAL INTRODUCTION.
The Mathematical
Theory
of
Elasticity
is
occupied
with an
attempt
to reduce to calculation the state of
strain,
or relative
displacement,
within
a solid
body
which is
subject
to the action of an
equilibrating system
of
forces,
or is in a state of
slight
internal relative
motion,
and with endeavours
to obtain results which shall be
practically important
in
applications
to
architecture,
engineering,
and all other useful arts in which the material
of construction is solid. Its
history
should embrace that of the
progress
of our
experimental knowledge
of the behaviour of strained
bodies,
so
far as it has been embodied in the mathematical
theory,
of the
develop-
ment of our
conceptions
in
regard
to the
physical principles necessary
to
form a foundation for
theory,
of the
growth
of that branch of mathematical
analysis
in which the
process
of the calculations
consists,
and of the
gradual
acquisition
of
practical
rules
by
the
interpretation
of
analytical
results.
In a
theory ideally
worked
out,
the
progress
which we should be able to
trace would
be,
in other
particulars,
one from less to
more,
but we
may say
that,
in
regard
to the assumed
physical principles, progress
consists in
passing
from more to less. Alike in the
experimental knowledge
obtained,
and in the
analytical
methods and
results,
nothing
that has once been dis-
covered ever loses its value or has to be discarded
;
but the
physical principles
come to be reduced to fewer and more
general
ones,
so that the
theory
is
brought
more into accord with that of other branches of
physics,
the same
general
dynamical principles
being ultimately requisite
and suflficient to
serve as a ba.sis for them all. And
although,
in the case of
Elasticity,
we find
frequent retrogressions
on the
part
of the
experimentalist,
and errors on
the
part
of the
mathematician,
chiefly
in
adopting hypotheses
not
clearly
established or
already
discredited,
in
pushing
to extremes methods
merely
approximate,
in
hasty generalizations,
and in
misunderstandings
of
physical
principles, yet
we observe a continuous
progress
in all the
respects
mentioned
when we
survey
the
history
of the science from the initial
enquiries
of
Galileo to the conclusive
investigations
of Saint-Venant and Lord Kelvin.
L. E.
1
2
HISTORICAL INTRODUCTION
The first mathematician to consider the nature of the resistance of solids
to
rupture
was Galileo'.
Although
he treated solids as
inelastic,
not
being
in
possession
of
any
law
connecting
the
displacements produced
with the
forces
producing
them,
or of
any physical hypothesis capable
of
yielding
such
a
law,
yet
his
enquiries gave
the direction which was
subsequently
followed
by many investigators.
He endeavoured to determine the resistance of a
beam,
one end of which is built into a
wall,
when the
tendency
to break
it arises from its own or an
applied weight;
and he concluded that the
beam tends to turn about an axis
perpendicular
to its
length,
and in the
plane
of the wall. This
problem,
and,
in
particular,
the determination of
this axis is known as Galileo's
problem.
In the
history
of the
theory
started
by
the
question
of
Galileo,
un-
doubtedly
the two
great
landmarks are the
discovery
of Hooke's Law in
1660,
and the formulation of the
general equations by
Navier in 1821.
Hooke's Law
provided
the
necessary experimental
foundation for the
theory.
When the
general equations
had been
obtained,
all
questions
of the small
strain of elastic bodies were reduced to a matter of mathematical calculation.
In
England
and in
France,
in the latter half of the I7th
century,
Hooke
and Mariotte
occupied
themselves with the
experimental discovery
of what
we now term stress-strain relations. Hooke^
gave
in 1678 the famous law
of
proportionality
of stress and strain which bears his
name,
in the words
"
Ut tensio sic vis
;
that
is,
the Power of
any spring
is in the same
proportion
with the Tension thereof."
By "spring"
Hooke
means,
as he
proceeds
to
explain, any "springy body,"
and
by
"tension" what we should now call
"extension," or,
more
generally,
"strain." This law he discovered in
1660,
but did not
publish
until
1676,
and then
only
under the form of an
anagram,
ceiiinosssttuu. This law forms the basis of the mathematical
theory
of
Elasticity,
and we shall hereafter consider its
generalization,
and its
range
of
validity
in the
light
of modern
experimental
research. Hooke does not
appear
to have made
any application
of it to the consideration of Galileo's
problem.
This
application
was made
by
Mariotte",
who in 1680 enunciated
the same law
independently.
He remarked that the resistance of a beam
to flexure arises from the extension and contraction of its
parts,
some of its
longitudinal
filaments
being
extended,
and others contracted. He assumed
that half are
extended,
and half contracted. His
theory
led him to
assign
the
position
of the
axis,
required
in the solution of Galileo's
problem,
at one-half
the
height
of the section above the ba^e.
In the interval between the
discovery
of Hooke's law and that of the
general
differential
equations
of
Elasticity by
Navier,
the attention of those
mathematicians who
occupied
themselves with our science was
chiefly
'
Galileo
Oalilei,
Vhcorgi e Vimostrazioni
mateviatiche, Leiden,
1G38.
'
liobert
Hooke,
De Potentia
restitutiva, London,
1678.
'
E.
Mariotte,
I'raite du mouvement lies
eaiix, Paris,
1680.
HISTORICAL INTRODUCTION 3
directed to the solution and extension of Galileo's
problem,
and the related
theories of the vibrations of bars and
plates,
and the
stability
of columns.
The first
investigation
of
any importance
is that of the elastic line or elastica
by
James Bernoulli* in
1705,
in which the resistance of a bent rod is assumed
to arise from the extension and contraction of its
longitudinal
filaments,
and
the
equation
of the curve assumed
by
the axis is formed. This
equation
practically
involves the result that the resistance to
bending
is a
couple
proportional
to the curvature of the rod when
bent,
a result which was
assumed
by
Euler in his later treatment of the
problems
of the
elastica,
and
of the vibrations of thin rods. As soon as the notion of a flexural
couple
proportional
to the curvature was established it could be noted that the
work done in
bending
a rod is
proportional
to the
square
of the curvature.
Daniel Bernoulli'
suggested
to Euler that the differential
equation
of the
elastica could be found
by making
the
integral
of the
square
of the curvature
taken
along
the rod a minimum
;
and
Euler",
acting
on this
suggestion,
was
able to obtain the differential
equation
of the curve and to
classify
the
various forms of it. One form is a curve of sines of small
amplitude,
and
Euler
pointed
out' that in this case the line of thrust coincides with the
unstrained axis of the
rod,
so that the
rod,
if of sufficient
length
and vertical
when
unstrained,
may
be bent
by
a
weight
attached to its
upper
end. Further
investigations'
led him to
assign
the least
length
of a column in order that
it
may
bend under its own or an
applied weight. Lagrange,"
followed and
used his
theory
to determine the
strongest
form of column. These two writers
found a certain
length
which a column must attain to be bent
by
its own
or an
applied weight,
and
they
concluded that for shorter
lengths
it will
be
simply compressed,
while for
greater lengths
it will be bent. These
researches are the earliest in the
region
of elastic
stability.
In Euler's work on the elastica the rod is
thought
of as a line of
particles
which resists
bending.
The
theory
of the flexure of beams of finite section
was considered
by
Coulomb'". This author took account of the
equation
of
equilibrium
obtained
by resolving
horizontally
the forces which act
upon
the
part
of the beam cut off
by
one of its normal
sections,
as well as of
the
equation
of moments. He was thus enabled to obtain the true
position
*
Bernoulli's memoir is
entitled,
'Vc^ritable
hypothese
de la resistance des
solides,
avec la
demonstration de la oourbure des
coi'ps qui
font
ressort,'
and will be found in his collected
vforks,
t.
2, Geneva,
1744.
^
See the 26th letter of Daniel Bernoulli to Euler
(October, 1742)
in
Fuss, Correspondmice
mathfmatique
et
physique,
t.
2,
St
Petersburg,
1843.
"
See the Additamentum
'
De curvis elasticis
'
in the Methodtis inveniendi tineas cunMS maximi
miiiiriuve
proprietate gaudentes,
Lausanne,
1744.
'
Berlin,
Histoire de
I'Academie,
t. 18
(1757).
"
Acta Acad.
Petropolitance
of
1778,
Pars
prior, pp.
121

193.
"
Miscellanea
Taurinensia,
t. 5
(1773).
1"
'
Easai sur une
application
des
regies
de Maximis et Minimis a
quelques
Probl^mes de
Statique,
relatifs a
I'Architecture,'
Mem..
..par
divers savans,
1776.
12
4
HISTORICAL INTRODUCTION
of the "neutral
line,"
or axis of
equilibrium,
and he also made a correct
calculation of the moment of the elastic forces. His
theory
of beams is
the most exact of those which
proceed
on the
assumption
that the stress in a
bent beam arises
wholly
from the extension and contraction of its
longitudinal
filaments,
and is deduced
mathematically
from this
assumption
and Hooke's
Law. Coulomb was also the first to consider the resistance of thin fibres
to
torsion",
and it is his account of the matter to which Saint-Venant refers
under the name I'ancienne
thdorie,
but his formula for this resistance was
not deduced from
any
elastic
theory.
The formula makes the torsional
rigidity
of a fibre
proportional
to the moment of inertia of the normal section
about the axis of the fibre. Another matter to which Coulomb was the
first to
pay
attention was the kind of strain we now call
shear,
though
he
considered it in connexion with
rupture only.
His
opinion appears
to have
been that
rupture'"
takes
place
when the shear of the material is
greater
than
a certain limit. The shear considered is a
permanent
set,
not an elastic strain.
Except
Coulomb's,
the most
important
work of the
period
for the
general
mathematical
theory
is the
physical
discussion of
elasticity by
Thomas
Young.
This naturalist
(to adopt
Lord Kelvin's name for students of
natural
science)
besides
defining
liis modulus of
elasticity,
was the first to
consider shear as an elastic strain'^. He called it
"detrusion,"
and noticed
that the elastic resistance of a
body
to
shear,
and its resistance to extensicm
or
contraction,
are in
general
dififerent
;
but he did not introduce a distiuct
modulus of
rigidity
to
express
resistance to shear. He defined
"
the modulus
of
elasticity
of a substance"" as "a column of the same substance
capable
of
producing
a
pressure
on its base which is to the
weight causing
a certain
degree
of
compression,
as the
length
of the substance is to the diminution
of its
length."
What we now call
"Young's
modulus" is the
weight
of this
column
per
unit of area of its base. This introduction of a definite
physical
concept,
associated with the coefficient of
elasticity
which
descends,
as it were
from a clear
sky,
on the reader of mathematical
memoirs,
marks an
epoch
in
the
history
of the science.
Side
by
side with the statical
developments
of Galileo's
enquiry
there were
discussions of the vibrations of solid bodies. Euler" and Daniel Bernoulli"
"
nittoire de I'Acadgmie for
1784, pp. 229269, Paris,
1787.
'2
See the iutroduction to the memoir first
quoted,
Mem..
..par
divers
mvatis,
1776.
"
A Course
of
Lectures on Natiiral
Philosophy
and the Mechanical
Arts, London, 1807,
Lecture xiii. It is in Kelland's later edition
(1845)
on
pp.
105 et
seq.
'*
Loc. cit.
(footnote 13).
Tlie definition was
given
in Section ix of Vol. 2 of the first
edition,
and omitted in Kelland's
edition,
but it is
reproduced
in the Miscellaneous Works
of
Dr
Young.
15
'De vibrationibuB...laminarum elasticarum . . .
,
'
and 'De souis multifariis
quos
laminae
ela8ticse...edunt...'
published
in Commentarii Academim Scientiarum
Imperialis Petropolitance,
t. 18
(1751).
The reader must be cautioned that in
writings
of the 18th
century
a "lamina"
means a
straight
rod or curved
bar, supposed
to be out out from a thin
plate
or
cylindrical
shell
by
two normal sections near
together.
This
usage lingers
in
many
books.
HISTORICAL INTRODUCTION
5
obtained the differential
equation
of the lateral vibrations of bars
by
variation
of the function
by
which
they
had
previously expressed
the work done in
bending'^ They
determined the forms of the functions which we should
now call the "normal
functions,"
and the
equation
which we should now call
the
"period equation,"
in the six cases of terminal conditions which arise
according
as the ends are
free,
clamped
or
simply supported.
Chladni"
investigated
these modes of vibration
experimentally,
and also the
longi-
tudinal and torsional vibrations of bars.
The success of theories of thin
rods,
founded on
special hypotheses, appears
to have
given
rise to
hopes
that a
theory might
be
developed
in the same
way
for
plates
and
shells,
so that the modes of vibration of a bell
might
be
deduced from its form and the manner in which it is
supported.
The first
to iittack this
problem
was Euler. He had
already proposed
a
theory
of the
resistance of a curved bar to
bending,
in which the
change
of curvature
played
the same
part
as the curvature does in the
theory
of a
naturally
straight
bar'*. In a note "De Sono
Campanarum'""
he
proposed
to
regard
a bell as divided into thin
annuli,
each of which behaves as a curved bar.
This method leaves out of account the
change
of curvature in sections
through
the axis of the bell. James Bernoulli""
(the younger)
followed. He assumed
the shell to consist of a kind of double sheet of curved
bars,
the bars in
one sheet
being
at
right angles
to those in the other.
Reducing
the shell
to a
plane plate
he found an
equation
of vibration which we now know to
be incorrect.
James Bernoulli's
attempt appears
to have been made with the view of
discovering
a theoretical basis for the
experimental
results of Chladni
con-
cerning
the nodal
figures
of
vibrating plates".
These results were still
unexplained
when in 1809 the French Institut
proposed
as a
subject
for
a
prize
the
investigation
of the tones of a
vibrating plate.
After several
attempts
the
prize
was
adjudged
in 1815 to Mdlle
Sophie
Germain,
and her
work was
published
in 1821^-. She assumed that the sum of the
principal
curvatures of the
plate
when bent would
play
the same
part
in the
theory
of
plates
as the curvature of the elastic central-line does in the
theory
of
rods,
and she
proposed
to
regard
the work done in
bending
as
proportional
to the
'
The form of the
energy-function
and the notion of
obtaining
the differential
equation by
varying
it are due to D. Bernoulli. The
process
was carried out
by
Euler,
and the normal
functions and the
period equations
were determined
by
him.
"
E. F. F.
Chladni,
Die
AkuMk, Leipzig,
1802. The author
gives
an account of the
history
of his own
experimental
researches with the dates of first
publication.
"
In the Methodns invenieiidi...
p.
274. See also his later
writing
'
Genuina
principia...
de statu
sequilibrii
et motu
corporum...,'
Nov. Comm. Acad.
PetropoUUnue,
t. 15
(1771).
'"
Nov. Comm. Acad.
Fclropolilana;,
t. 10
(1766).
'"'
'Essai
thdorique
sur les vibrations des
plaques elastiques...,'
Nov. Acta
PetropoUtaiue,
i. o
(1789).
'I
First
published
at
Leipzig
in 1787. See Die
Akustik, p.
vii.
**
Recherches snr la theorie des
surfaces Elastiques. Paris,
1821.
6
HISTOBICAL INTRODUCTION
integral
of the
square
of the sum of the
principal
curvatures taken over
the surface. From this
assumption
and the
principle
of virtual work she
deduced the
equation
of flexural vibration in the form now
generally
admitted.
Later
investigations
have shown that the formula assumed for the work
done in
bending
was incorrect.
During
the first
period
in the
history
of our science
(1638

1820)
while
these various
investigations
of
special problems
were
being
made,
there was
a cause at work which was to lead to wide
generalizations.
This cause was
physical speculation concerning
the constitution of bodies. In the
eighteenth
century
the Newtonian
conception
of material
bodies,
as made
up
of small
parts
which act
upon
each other
by
means of central
forces,
displaced
the
Cartesian
conception
of a
plenum pervaded by
"
vortices." Newton
regarded
his
"
molecules
"
as
possessed
of finite sizes and definite
shapes^,
but his
successors
gradually simplified
them into material
points.
The most definite
speculation
of this kind is that of
Boscovich''*,
for whom the material
points
were
nothing
but
persistent
centres of force. To this order of ideas
belong
Laplace's
theory
of
capillarity'"
and Poisson's first
investigation
of the
equi-
librium of an
"
elastic surface
'","
but for a
long
time no
attempt
seems to
have been made to obtain
general equations
of motion and
equilibrium
of
elastic solid bodies. At the end of the
year
1820 the fruit of all the
ingenuity
expended
on elastic
problems might
be summed
up
as

an
inadequate theory
of
flexure,
an erroneous
theory
of
torsion,
an
unproved theory
of the vibrations
of bars and
plates,
and the definition of
Young's
modulus. But such an
estimate would
give
a
very wrong impression
of the value of the older
researches. The
recognition
of the distinction between shear and extension
was a
preliminary
to a
general theory
of strain
;
the
recognition
of forces
across the elements of a section of a
beam,
producing
a
resultant,
was a
step
towards a
theory
of
stress;
the use of differential
equations
for the
deflexion of a bent beam and the vibrations of bars and
plates,
was a fore-
shadowing
of the
employment
of differential
equations
of
displacement;
the
Newtonian
conception
of the con.stitution of
bodies,
combined with Hooke's
Law,
offered means for the formation of such
equations
;
and the
generalization
of the
principle
of virtual work in the
Micanique Analytique
threw
open
a
broad
path
to
discovery
in this as in
every
other branch of mathematical
physics. Physical
Science had
emerged
from its
incipient stages
with definite
methods of
hypothesis
and induction and of observation and
deduction,
with
the clear aim to discover the laws
by
which
phenomena
are connected with
each
other,
and with a fund of
analytical processes
of
investigation.
This was
the hour for the
production
of
general
theories,
and the men were not
wanting.
**
See,
in
particular, Newton,
Optickn,
2nd
Edition,
London, 1717,
the 31st
Query.
^
B. J.
Boscovich,
Theoria
Philosophia
Naturalis redacta ad unicam
legem
virium in natura
exUtenlium, Venice,
1743.

Micanique Cileste, Supplement

au W
Livre, Paris,
180G.
"
Pari$,
Mf,m. de
I'Imtitut,
1814.
HISTORICAL INTRODUCTION
7
Navier" was tlie first to
investigate
the
general equations
of
equilibrium
and vibration of elastic solids. He set out from the
Newtonian
conception
of the constitution of
bodies,
and assumed that the elastic reactions arise
from variations in
the.intermolecular
forces which result from
changes
in the
molecular
configuration.
He
regarded
the molecules as material
points,
and
assumed that the force between two
molecules,
whose distance is
slightly
increased,
is
proportional
to the
product
of the increment of the distance and
some function of the initial distance. His method consists in
forming
an
expression
for the
component
in
any
direction of all the forces that act
upon
a
displaced
molecule,
and thence the
equations
of motion of the molecule.
The
equations
are thus obtained in terms of the
displacements
of the molecule.
The material is assumed to be
isotropic,
and the
equations
of
equilibrium
and
vibration contain a
single
constant of the same nature as
Young's
modulus.
Navier next formed an
expression
for the work done in a small relative dis-
placement by
all the forces which act
upon
a molecule
;
this he described as
the sum of the moments
(in
the sense of the
Mecanique Analytique)
of the
forces exerted
by
all the other molecules on a
particular
molecule. He
deduced,
by
an
application
of the Calculus of
Variations,
not
only
the
differential
equations pievioiisly
obtained,
but also the
boundary
conditions
that hold at the surface of the
body.
This memoir is
very important
as the
first
general investigation
of its
kind,
but its
arguments
have not met with
general acceptance. Objection
has been raised
against
Navier's
expression
for the force, between two
"
molecules,"
and to his method of
simplifying
the
expressions
for the forces
acting
on a
single
"
molecule." These
expressions
involve
triple
summations,
which Navier
replaced by integrations,
and the
validity
of this
procedure
has been
disputed^'l
In the same
year,
1821,
in which Navier's memoir was read to the
Academy
the
study
of
elasticity
received a
powerful impulse
from an un-
expected qviarter.
Fresnel announced his conclusion that the observed facts
in
regard
to the interference of
polarised light
could be
explained only by
the
hypothesis
of transverse vibrations'". He showed how a medium
^'
Paris,
Mem. Acad.
Sciences,
t. 7
(1827).
The memoir was read in
May,
1821.
"^
For criticisms of Navier's memoir and an account of the discussions to which it
gave
rise,
see Todhunter and
Pearson, History of
the
Theory of Elasticity,
vol.
1, Cambridge,
1886,
pp. 1,S9,
221,
277: and cf. the account
given by
H. Burkhardt in his
Report
on
'
Entwickelungen
nach oscillirenden Functionen
'
published
in the Jahresbericht der Deiitschen Mathenuitiker-
Vereinigung,
Bd.
10,
Heft
2, Lieferung
3
(1903).
It
may
not be
superfluous
to remark that the
conception
of molecules as material
points
at rest in a state of stable
equilibrium
under their
mutual forces of attraction and
repulsion,
and held in
slightly displaced positions by
external
forces,
is
quite
different from the
conception
of molecules with which modern
Thermodynamics
has made us familiar. The "molecular" theories of
Navier,
Poisson and
Cauchy
have no
very
intimate relation to modern notions about molecules.
=
See E.
Verdet,
(Exwres
completes d'Augustin
Fresnel,
t.
1, Paris, 186G, p.
Ixxxvi,
also
pp.
629 et
seq.
Verdet
points
out that Fresnel arrived at his
hypothesis
of transverse
vibrations
in 1816
(loc.
cit.
pp. lv, 385, 394).
Thomas
Young
in his Article 'Chromatics'
(Encycl.
ISrit.
Supplement, 1817) regarded
the luminous vibrations as
having relatively
feeble transverse
com-
ponents.
8 HISTORICAL INTRODUCTION
consisting
of
"
molecules
"
connected
by
central forces
might
be
expected
to
execute such vibrations and to transmit waves of the
required type.
Before
the time of
Young
and Fresnel such
examples
of transverse waves as were
known

waves on
water,
transverse vibrations of
strings,
bars,
membranes and
plates

were in no case
examples
of waves transmitted
through
a medium
;
and neither the
supporters
nor the
opponents
of the
undulatory theory
of
light appear
to have conceived of
light
waves otherwise than as
"
longi-
tudinal
"
waves of condensation and
rarefaction,
of the
type
rendered familiar
by
the transmission of sound. The
theory
of
elasticity,
and,
in
particular,
the
problem
of the transmission of waves
through
an elastic medium now
attracted the attention of two mathematicians of the
highest
order:
Cauchy*"
and Poisson''

the former a
discriminating supporter,
the latter a
sceptical
critic
of Fresnel's ideas. In the future the
developments
of the
theory
of
elasticity
were to be
closely
associated with the
question
of the
propagation
of
light,
and
these
developments
arose in
great part
from the labours of these two savants.
By
the Autumn of 1822
Cauchy''
had discovered most of the elements
of the
pure theory
of
elasticity.
He had introduced the notion of stress at
a
point
determined
by
the tractions
per
unit of area across all
plane
elements
through
the
point.
For this
purpose
he had
generalized
the notion of
hydrostatic pressure,
and he had shown that the stress is
expressible by
means of six
component
stresses,
and also
by
means of three
purely
normal
tractions across a certain triad of
planes
which cut each other at
right angles

the
"
principal planes
of stress." He had shown also how the differential
coefficients of the three
components
of
displacement
can be used to estimate
the extension of
every
linear element of the
material,
and had
expressed
the
state of strain near a
point
in terms of six
components
of
strain,
and also in
terms of the extensions of a certain triad of lines which are at
right angles
to each other

the
"
principal
axes of strain." He had determined the
equations
of motion
(or
equilibrium) by
which the
stress-components
are
connected with the forces that are distributed
through
the volume and with
the kinetic reactions.
By
means of relations between
stress-components
and
strain-components,
he had eliminated the
stress-components
from the
equa-
tions of motion and
equilibrium,
and had arrived at
equations
in terms of the
displacements.
In the later
published
version of this
investigation Cauchy
^
Cauchy
'a studies in
Elasticity
were first
prompted by
liis
being
a member of the Commission
appointed
to
report upon
a memoir
by
Navier on elastic
plates
which was
presented
to the Paris
Academy
in
^ugust,
1820.
'I'
We have noted that Poisson had
already
written on elastic
plates
in 1814.
"
Cauchy's
memoir was communicated to the Paris
Academy
in
September
1822,
but it
was not
published.
An abstract was inserted in the Bulletin ties Sciences a lit Society
philo-
yiiathique, 1823,
and the contents of the memoir were
given
in later
publications,
viz. in two
Articles in the volume for 1827 of
Cauchy's
Kxercices de
vmthdmatique
and an Article in the
volume for 1828. The titles of these Articles are
(i)
'
De la
pression
ou tension dans un
corps
solide,'
(ii)
'
Sur la condensation et la dilatation des
corps solides,'
(iii)
'Sur les
Equations qui
expriment
les conditions
d'^quilibre
ou les lois de mouvement int^rieur d'un
corps
solide.'
The last of these contains the correct
equations
of
Klasticity.
HISTORICAL INTRODUCTION
9
obtained his stress-strain relations for
isotropic
materials
by
means of two
assumptions,
viz. :
(1)
that the relations in
question
are
linear,
(2)
that the
principal planes
of stress are normal to the
principal
axes of strain. The
experimental
basis on which these
assumptions
can be made to rest is the
same as that on which Hooke's Law
rests,
but
Cauchy
did not refer to it.
The
equations
obtained are those which are now admitted for
isotropic
solid
bodies. The methods used in these
investigations
are
quite
different from
those of Navier's memoir. In
particular,
no use is made of the
hypothesis
of material
points
and central forces. The
resulting equations
differ from
Navier's in one
important respect,
viz. : Navier's
equations
contain a
single
constant to
express
the elastic behaviour of a
bodj',
while
Cauchy's
contain
two such constants.
At a later date
Cauchy
extended his
theory
to the case of
crystalline
bodies,
and he then made use of the
hypothesis
of material
points
between
which there are forces of attraction or
repulsion.
The force between a
pair
of
points
was taken to act in the line
joining
the
points,
and to be a function
of the distance between them
;
and the
assemblage
of
points
was taken to be
homogeneous
in the sense
that,
if
A, B,
C are
any
three of the
points,
there
is a
point
D of the
assemblage
which is situated so that the line CD is
equal
and
parallel
to
A,
and the sense from C to D is the same as the sense
from A to B. It was assumed further that when the
system
is
displaced
the relative
displacement
of two of the material
points,
which are within each
other's
ranges
of
activity,
is small
compared
with the distance between them.
In the first memoir** in which
Cauchj'
made use of this
hypothesis
he formed
an
expression
for the forces that act
upon
a
single
material
point
in the
system,
and deduced differential
equations
of motion and
equilibrium.
In
the case of
isotropy,
the
equations
contained two constants. In the second
memoir**
expressions
were formed for the tractions across
any plane
drawn in
the
body.
If the initial state is one of zero
stress,
and the material is
isotropic,
the stress is
expressed
in terms of the strain
by
means of a
single
constant,
and one of the constants of the
preceding
memoir must vanish. The
equations
are then identical with those of Navier. In like
manner,
in the
general
case
of
seolotropy, Cauchy
found 21
independent
constants. Of these 15 are true
"
elastic
constants,"
and the
remaining
6
express
the initial stress and vanish
identically
if the initial state is one of zero stress. These matters were not
fully explained by Cauchy.
Clausius'*", however,
has shown that this is the
^
Exercices de
jnathematique,
1828,
'
Sur
I'dquilibre
et le mouvement d'un
syst^me
de
points
materielB soUicites
par
des forces d'attractiou ou de
ri^pulsion
mutuelle.' This memoir follows
immediately
after that last
quoted
and
immediately precedes
that next
quoted.
**
Exercices de
viaMmatique, 1828,
'
De la
pression
ou tension dans un
systtoe
de
points
mat^riels.
'
^ '
Ueber die
Veriinderungen,
welche in den bisher
gebriiuchlichen
Formeln fiir das
Gleichge-
wicht und die
Bewegung
elastischer fester
Korper
durch neuere
Bcobachtungen notbwendig
gewordeu sind,'
Ann.
Phys.
Chem.
{Poggendorff)
,
Bd. 76
(1849).
10 HISTORICAL INTRODUCTION
meaning
of his work. Clausius criticized the restrictive conditions which
Cauchy imposed upon
the
arrangement
of his material
points,
but he
argued
that these conditions are not
necessary
for the deduction of
Cauchy
's
equations.
The first memoir
by
Poisson"
relating
to the same
subject
was read
before the Paris
Academy
in
April,
1828. The memoir is
very
remarkable
for its numerous
applications
of the
general theory
to
special problems.
In
his
investigation
of the
general equations
Poisson,
like
Cauchy,
first obtains
the
equations
of
equilibrium
in terms of
stress-components,
and then estimates
the traction across
any plane resulting
from the
"
intermoleoular
"
forces.
The
expressions
for the stresses in terms of the strains involve summations
with
respect
to all the
"molecules,"
situated within the
region
of "molecular"
activity
of a
given
one. Poisson decides
against replacing
all the summations
by integrations,
but he assumes that this can be done for the summations
with
respect
to
angular space
about the
given
"
molecule,"
but not for the
summations with
respect
to distance from this
"
molecule." The
equations
of
equilibrium
and motion of
isotropic
elastic solids which were thus obtained
are identical with Navier's. The
principle,
on which summations
may
be
replaced by integrations,
has been
explained
as follows
by Cauchy'';

The
number of molecules in
any
volume,
which contains a
very large
number of
molecules,
and whose dimensions are at the same time small
compared
with
the radius of the
sphere
of sensible molecular
activity, may
be taken to be
proportional
to the volume.
If, then,
we make abstraction of the molecules
in the immediate
neighbourhood
of the one
considered,
the actions of all the
others,
contained in
any
one of the small volumes referred
to,
will be
equiva-
lent to a
force,
acting
in a line
through
the centroid of this
volume,
which
will be
proportional
to the volume and to a function of the distance of the
particular
molecule from the centroid of the volume. The action of the
remoter molecules is said to be
"
regular,"
and the action of the nearer
ones,
"
irregular
"
;
and thus Poisson assumed that the
irregular
action of the
nearer molecules
may
be
neglected,
in
comparison
with the action of the
remoter
ones,
which is
regular.
This
assumption
is the text
upon
which
Stokes'' afterwards founded his criticism of Poisson. As we have
seen,
Cauchy
arrived at Poisson's results
by
the aid of a different
a.ssumption'*.
Clausius'' held that both Poisson's and
Cauchy's
methods could be
presented
in
unexceptionable
forms.
"
'Mimoire sur
I'^quilibre
et le mouvement des
corps elastiques,'
Mem. Paris
Acad.,
t. 8
(1829).
"
'On the Theories of the...
Equilibrium
and Motion of Elastic
Solids,' Cambridge
Phil.
Soc. Traim. vol. 8
(1845).
"
In a later memoir
presented
to the
Academy
in 1829 and
published
in ./. de t'Kcnh;
pohj-
technique,
t. 13
(18.S1),
Poisson
adopted
a method
quite
similar to that of
Cauchy (footnote 34).
Poisson extended his
theory
to
ieolotropic
bodies in liis
'
M^moire sur
I'^quilibre
et le mouve-
ment des
corps cristallis^es,'
read to the Paris
Academy
in 1839 and
published
after his death in
M^m. de I'Acad. t. 18
(1842).
HISTORICAL INTRODUCTION
11
The
theory
of
elasticity
established
by
Poisson and
Cauchy
on the then
accepted
basis of material
points
and central forces was
applied by
them
and also
by
Lame and
Clapeyron*"
to numerous
problems
of vibrations and
of statical
elasticity,
and thus means were
provided
for
testing
its con-
sequences experimentally,
but it was a
long
time before
adequate experi-
ments were made to test it. Poisson used it to
investigate
the
propagation
of waves
through
an
isotropic
elastic solid medium. He found two
types
of
waves
which,
at
great
distances from the sources of
disturbance,
are
practically
"
longitudinal
"
and
"
transverse,"
and it was a
consequence
of his
theory
that
the ratio of the velocities of waves of the two
types
is
\/3
: 1**.
Cauchy"
applied
his
equations
to the
question
of the
propagation
of
light
in
crystalline
as well as in
isotropic
media. The
theory
was
challenged
first in its
application
to
optics by
Green
*^
and afterwards on its statical side
by
Stokes". Green
was
dissatisfied with the
hypothesis
on which the
theory
was
based,
and
he
sought
a new foimdation
;
Stokes's criticisms were directed rather
against
the
process
of deduction and some of the
particular
results.
The revolution which Green effected in the elements of the
theory
is
comparable
in
importance
with that
produced by
Navier's
discovery
of the
general
equations. Starting
from what is now called the
Principle of
the
Conservation
of Energy
he
propounded
a new method of
obtaining
these
equations.
He himself stated his
principle
and method in the
following
words :

"
In
whatever
way
the elements of
any
material
system may
act
upon
"
each
other,
if all the internal forces exerted be
multiplied by
the elements
"
of their
respective
directions,
the total sum for
any assigned portion
of the
"
mass will
always
be the exact differential of some function. But this
"
function
being
known,
we can
immediately apply
the
general
method
"given
in the
Mecanique Analytique,&nA
which
appears
to be more
especially
"
applicable
to
problems
that relate to the motions of
systems composed
"
of an
immense number of
particles mutually acting upon
each other. One
"of the
advantages
of this
method,
of
great importance,
is that we are
"
necessarily
led
by
the mere
process
of the
calculation,
and with little care
"on our
part,
to all the
equations
and conditions which are
requisite
and
"sufficient
for the
complete
solution of
any problem
to which it
may
be
"applied."
'''
'Memoire sur
I'equilibre
int^rieur des
corps
solides
homogenes,'
Paris,
Mem.
par
divers
savants,
t. i
(1833).
The memoir was
published
also in J.
f.
Math.
(Crelle),
Bd. 7
(1831)
;
it
had been
presented
to the Paris
Academy,
and the
report
on it
by
Poinsot and Navier is dated
1828. In
regard
to the
genera! theory
the method
adopted
was that of Navier.
*
See the
addition,
of date November
1828,
to the memoir
quoted
in footnote 36.
Cauchy
recorded the same result in the Exercices de
mttthematique,
1830.
"
Exercices de
Mathenmtique,
1830.
*'- '
On the laws of reflexion and refraction of
light
at the common surface of two non-
crystallized media,'
Cambridge
Phil. Soc. Trans, vol. 7
(1839).
The date of the memoir is 1837.
It is
reprinted
in Mathematical
Papers of
the late
George G-reen,
London, 1871,
p.
245.
12 HISTORICAL INTRODUCTION
The function here
spoken
of,
with its
sign changed,
is the
potential
energy
of the strained elastic
body per
unit of
volume,
expressed
in terms
of the
components
of
strain;
and the differential coefficients of the
function,
with
respect
to the
components
of
strain,
are the
components
of stress.
Green
supposed
the function to be
capable
of
being expanded
in
powers
and
products
of the
components
of strain. He therefore
arranged
it as a sum of
homogeneous
functions of these
quantities
of the
first,
second and
higher
degrees.
Of these
terms,
the first must be
absent,
as the
potential energy
must be a true minimum when the
body
is unstrained
; and,
as the strains
are all
small,
the second term alone will be of
importance.
From this
principle
Green deduced the
equations
of
Elasticity, containing
in the
general
case 21 constants. In the case of
isotropy
there are two
constants,
and the
equations
are the same as those of
Cauchy's
first memoir*".
Lord Kelvin
**
has based the
argument
for the existence of Green's strain-
energy-function
on the First and Second Laws of
Thermodynamics.
From
these laws he deduced the result
that,
when a solid
body
is strained without
alteration of
temperature,
the
components
of stress are the differential
coefficients of a function of the
components
of strain with
respect
to these
components severally.
The same result can be
proved
to hold when the strain
is effected so
quickly
that no heat is
gained
or lost
by any part
of the
body.
Poisson's
theory
leads to the conclusions that the resistance of a
body
to
compression by pressure
uniform all round it is two-thirds of the
Young's
modulus of the
material,
and that the resistance to
shearing
is two-fifths
of the
Young's
modulus. He noted a result
equivalent
to the first of
these",
and the second is
virtually
contained in his
theory
of the torsional vibrations
of a bar**. The observation that resistance to
compression
and resistance to
shearing
are the two fundamental kinds of elastic resistance in
isotropic
bodies was made
by
Stokes",
and he introduced
definitely
the two
principal
moduluses of
elasticity by
which these resistances are
expressed

the
"
modulus of
compression
"
and the
"
rigidity,"
as
they
are now called. From
Hooke's Law and from considerations of
symmetry
he concluded that
pressure
equal
in all directions round a
point
is attended
by
a
proportional compression
without
shear,
and that
shearing
stress is attended
by
a
corresponding
proportional shearing
strain. As an
experimental
basis for Hooke's Law
he cited the fact that bodies admit of
being
thrown into states of isochronous
vibration.
By
a method
analogous
to that of
Cauchy's
first
memoir^'-',
but
resting
on the above-stated
experimental
basis,
he deduced the
equations

Sir W.
Thomson, Quart.
J.
of
Math. vol. 5
(1855), reprinted
in Phil.
Mag. (Ser. 5),
vol. 5
(1878),
and also in Mathematical and
Physical Papers by
Sir William
Thomson,
vol.
1, Cambridge,
1882, p.
291.
**
Annales de Chimie et de
Physique,
t. .'iG
(1827).
*
This
theory
is
given
in the memoir cited in footnote 36.
"
See footnote 37. The distinction between the two kinds of
elasticity
had been noted
by
Poncelet,
Introduction h la
Micunique industrielle,
physique
et
experimentale, Metz,
1839.
HISTORICAL INTRODUCTION
13
with two constants which had been
given by Cauchy
and Green.
Having
regard
to the
varying degrees
in which different classes of bodies

liquids,
soft
solids,
hard solids

resist
compression
and
distortion,
he refused to
accept
the conclusion from Poisson's
theory
that the modulus of
compression
has to the
rigidity
the ratio 5 : 3. He
pointed
out
that,
if the ratio of these
moduluses could be
regarded
as
infinite,
the ratio of the velocities of
"
longi-
tudinal
"
and
"
transverse
"
waves would also be
infinite,
and
then,
as Green
had
already
shown,
the
application
of the
theory
to
optics
would be facilitated.
The methods of
Navier,
of
Poisson,
and of
Cauchy's
later memoirs lead to
equations
of motion
containing
fewer constants than occur in the
equations
obtained
by
the methods of
Green,
of
Stokes,
and of
Cauchy's
first memoir.
The
importance
of the
discrepancy
was first
emphasized by
Stokes. The
questions
in
dispute
are these

Is elastic
seolotropy
to be characterized
by
21 constants or
by
15,
and is elastic
isotropy
to be
characterized
by
two
constants or one? The two theories are
styled by
Pearson^' the
"
multi-con-
stant
"
theory
and the
"
rari-constant
"
theory respectively,
and the
controversy
concerning
them has lasted almost down to the
present
time. It is to be
understood that the rari-constant
equations
can be included in the multi-
constant ones
by equating
certain
pairs
of the
coefficients,
but that the
rari-constant
equations
rest
upon
a
particular hypothesis
concerning
the
constitution of
matter,
while the
adoption
of
multi-constancy
has been
held to
imply
denial of this
hypothesis. Discrepancies
between the results
of the two theories can be submitted to the test of
experiment,
and it
might
be
thought
that the verdict would be
final,
but the
difficulty
of
being
certain that the tested material is
isotropic
has diminished the credit of
many
experimental investigations,
and the
tendency
of the multi-constant elasticians
to
rely
on
experiments
on such bodies as
cork,
jelly
and india-rubber has
weakened their
arguments.
Much of the discussion has turned
upon
the value
of the ratio of lateral contraction to
longitudinal
extension of a bar under
terminal tractive load. This ratio is often called
"
Poisson's ratio." Poisson^
deduced from his
theory
the result that this ratio must be
^.
The
experi-
ments of Wertheim on
glass
and brass did not
support
this
result,
and
Wertheim**
proposed
to take the ratio to be
^

a value which has no

theoretical foundation. The
experimental
evidence led Lame in his treatise**
to
adopt
the multi-constant
equations,
and after the
publication
of this
book
they
were
generally employed.
Saint-Veuant,
though
a firm believer
in
rari-constancy, expressed
the results of his researches on torsion and
flexure and on the distribution of elasticities round a
point"
in terms of the
"
Todhunter and
Pearson,
History of
the
Theory of Elasticity,
vol.
1, Cambridge, 1886, p.
496.

Annales de
Chimie,
t. 23
(1848).
*^
Lefons
sur la theorie
mathcmalique
de I'elasticite des
corps
solides, Paris,
1852.
*"
The memoir on torsion is in Mem. des Savants
ctrangers,
t. 14
(1855),
that on flexure is in
J. de Math.
(Liouvilk), (S6r. 2),
t. 1
(1856),
and that on the distribution of elasticities is in J. de
Math.
{LiouvilU), (S6r. 2),
t. 8
(1863).
14
HISTORICAL INTRODUCTION
inulti-constant
theory.
Kirchhoff"'
adopted
the same
theory
in his in-
vestigations
of thin rods and
plates,
and
supported
it
by experiments
on
the torsion and flexure of steel
bars"";
and Clebsch in his treatise" used the
language
of bi-constant
isotropy.
Kelvin and Tait* dismissed the
controvei-sy
in a few words and
adopted
the views of Stokes. The best modern
experi-
ments
support
the conclusion that Poissou's ratio can differ
sensibly
from the
value
J
in materials which
may
without cavil be treated as
isotropic
and
homogeneous.
But
perhaps
the most
striking experimental
evidence is that
which
Voigt"
has derived from his
study
of the
elasticity
of
crystals.
The
absence of
guarantees
for the
isotropy
of the tested materials ceased to be a
difficulty
when he had the
courage
to undertake
experiments
on materials
which have known kinds of
ffiolotropy''^
The
point
to be settled
is, however,
more remote.
According
to Green there
exist,
for a material of the most
generally
aeolotropic
character,
21
independent
elastic constants. The mole-
cular
hypothesis,
as worked out
by Cauchy
and
supported by
Saint-
Venant,
leads to 15
constants,
so
that,
if the rari-constant
theory
is
correct,
there must
be 6
independent
relations
among
Green's 21 coefficients. These relations
I call
Cauchy
's relations".
Now
Voigt's experiments
were made on the
torsion and flexure of
prisms
of various
crystals,
for most of which Saint-
Venant's
formula} for
aeolotropic
rods hold
good,
for the others he
supplied
the
required
forraulfe. In the cases of
beryl
and rocksalt
only
were
Cauchy
's
relations even
approximately
verified
;
in the seven other kinds of
crystals
examined there were
very
considerable differences between the coefficients
which these relations would
require
to be
equal.
Independently
of the
experimental
evidence the rari-constant
theory
has
lost
ground through
the
widening
of our views
concerning
the constitution
of matter. The
hypothesis
of material
points
and central forces does not
now hold the field. This
change
in the
tendency
of
physical speculation
is
due to
many
causes,
among
which the
disagreement
of the rari-constant
theory
of
elasticity
with the results of
experiment
holds a rather subordinate
position.
Of much
greater importance
have been the
development
of the
atomic
theory
in
Chemistry
and of statistical molecular theories in
Physics,
the
growth
of the doctrine of
energy,
the
discovery
of electric radiation.
It is now
recognized
that a
theory
of atoms must be
part
of a
theory
of the
aether,
and that the confidence which was once felt in the
hypothesis
of central
!^'
J./.
Math.
(Crelle),
Bd. 40
(18.50),
and 13d. 56
(1859).
52
Ann.
Phyg.
Chem.
(Poygendorff)
,
Bd. 108
(1859).
"'
Theorie der Elasticitat
fester KSrper, heipzig,
18C'2.
M
Thomson and
Tait,
Natural
Philosophy,
1st edition Oxford
1867,
2nd edition
Cambridge
18791883.
i*
W.
Voigt,
Ann.
Phys.
Chem.
(Wiedenumn),
Bde 31
(1887),
34 and 35
(1888),
38
(1889).
"*
A certain
assumption,
first made
by
F. E. Neumann,
is involved in the statement that
the
SBolotropy
of a
crystal
as
regards elasticity
is known from the
crystallographic
form.
"
They appear
to have been first stated
explicitly by
Saint-Venant in the memoir on
torsion of 1855.
(See
footnote
50.)
HISTORICAL INTRODUCTION
16
forces betweeu material
points
was
premature.
To determine the laws of
the
elasticity
of solid bodies without
knowing
the nature of the
aithereal
medium or the nature of the
atoms,
we can
only
invoke the known laws of
energy
as was done
by
Green and Lord
Kelvin;
and we
may place
the
theory
on a finn basis if we
appeal
to
experiment
to
support
the statement
that,
within a certain
range
of
strain,
the
strain-energy-function
is a
quad-
ratic function of the
components
of
strain,
instead of
relying,
as Green
did,
upon
an
expansion
of the function in series.
The
problem
of
determining
the state of stress and strain within a
solid
body
which is
subjected
to
given
forces
acting through
its volume
and to
given
tractions across its
surface,
or is held
by
surface tractions so
that its surface is deformed into a
prescribed figure,
is reducible to the
analytical problem
of
finding
functions to
represent
the
components
of dis-
placement.
These functions must
satisfy
the differential
equations
of
equi-
librium at all
points
within the surface of the
body
and must also
satisfy
certain
special
conditions at this surface. The methods which have been
devised for
integrating
the
equations
fall into two classes. In one class of
methods a
special
solution is
sought
and the
boundary
conditions are satisfied
by
a solution in the form of a
series,
which
may
be
infinite,
of
special
solu-
tions. The
special
solutions are
generally expressible
in terms of harmonic
functions. This class of solutions
niay
be
regarded
as
constituting
an
extension of the methods of
expansion
in
spherical
harmonics and in
trigonometrical
series. In the other class of methods the
quantities
to be
determined are
expressed by
definite
integrals,
the elements of the
integrals
representing
the effects of
singularities
distributed over the surface or
through
the volume. This class of solutions constitutes an extension of the
methods introduced
by
Green in the
Theory
of the Potential. At the
time of the
discovery
of the
general equations
of
Elasticity
the method of
series had
already
been
applied
to astronomical
problems,
to acoustical
problems
and to
problems
of the conduction of heaf*
;
the method of
singu-
larities had not been invented^'. The
application
of the method of series to
problems
of
equilibrium
of elastic solid bodies was initiated
by
Lamd and
Clapeyron*'. They
considered the case of a
body
bounded
by
an unlimited
plane
to which
pressure
is
applied according
to an
arbitrary
law. Lame""
later considered the
problem
of a
body
bounded
by
a
spherical
surface and
deformed
by given
surface tractions. The
problem
of the
plane
is
essentially
that of the transmission into a solid
body
of force
applied locally
to a
small
part
of its surface. The
problem
of the
sphere
has been
developed
^
See
Buikbardt, 'Entwickelungen
nach oscillirenden
Fanclionen,'
Jahiesbericlit der DeuUchen
Mathemutiher-VeieiniyuiKj,
Bi3.
10,
Heft 2.
"*
It was invented
by Green,
An
Essay
on the
Application of
Mathematical
Analysis
to the
Theories
of Electricity
and
Magnetism, Nottingham,
1828.
Reprinted
in Malliematical
Papers
of
the late
Georye Green, London,
1871.

J. de Math.
{Liouville),
t. 19
(1854).
16 HISTORICAL INTEODUCTION
by
Lord
Kelvin'",
who
sought
to utilize it for the
purpose
of
investigating
the
rigidity
of the
Eaith'^,
and
by
G. H. Darwin in connexion with other
problems
of cosmical
physics"'.
The serial solutions
employed
are
expressed
in terms of
spherical
harmonics. Solutions
of the
equations
in
cylindrical
coordinates can be
expressed
in terms of Bessel's
functions"'', but,
except
for
spheres
and
cylinders,
the method of series has not been
employed very
successfully.
The method of
singularities
was first
applied
to the
theory
of
Elasticity by
E.
Betti",
who set out from a certain
reciprocal
theorem of the
type
that is now familiar in
many
branches of mathematical
physics.
From
this theorem he deduced
incidentally
a formula for
determining
the
average
strain of
any type
that is
produced
in a
body by given
forces. The method
of
singularities
has been
developed chiefly by
the elasticians of the Italian
school. It has
proved
more effective than the method of series in the
solution of the
problem
of transmission of force. The fundamental
particular
solution which
expresses
the
displacement
due to force at a
point
in an in-
definitely
extended solid was
given by
Lord Kelvin**. It was found at a
later date
by
J.
Boussinesq*^ along
with other
particular
solutions,
which
can,
as a matter of
fact,
be derived
by synthesis
from it.
Boussinesq's
results
led him to a solution of the
problem
of the
plane,
and to a
theory
of
"
local
perturbations,"
according
to which the effect of force
applied
in the
neigh-
bourhood of
any point
of a
body
falls off
very rapidly
as the distance
from
the
point
increases,
and the
application
of an
equilibrating system
of forces
to a small
part
of a
body produces
an effect which is
negligible
at a con-
siderable distance from the
part.
To estimate the effect
produced
at a
distance
by
forces
applied
near a
point,
it is not
necessary
to take into
account the mode of
application
of the forces but
only
the statical resultant
and moment. The direct method of
integration
founded
upon
Betti's
reciprocal
theorem was
applied
to the
problem
of the
plane by
V. Cerruti"*. Some
of the results were found
independently by
Hertz,
and led in his hands to a
theory
of
impact
and a
theory
of hardness".
A different method for
determining
the state of stress in a
body
has
81
Phil. Tram.
Roy.
Soc,
vol. 153
(1863).
See also Math, and
Phys. Papers,
vol. 3
(Cambridge,
1890), p. 351,
and Kelvin and
Tait,
Nat.
Phil.,
Part 2.

Brit. Assoc.
Hep.
1876,
Math, and
Phys. Papers,
vol.
3, p.
312.

Phil. Trails.
Roy. Soc,
vol. 170
(1879),
and vol. 173
(1882).

L.
Poclihammer, J.f.
Math.
(Crelle),
Bd. 81
(1876), p.
33.
<
n Nuovo Cimento
(Ser. 2),
tt. 610
(1872
et
seq.).
"*
Sir W.
Thomson, Cambridge
and Dublin Math.
J., 1848, reprinted
in Math, and
Phys,
Papers,
vol.
1, p.
97.
^
For
Boussinesq's
earlier researches in
regard
to
simple
solutions,
see
Paris,
C.
R.,
tt. 86

88
(18781879)
and tt. 9396
(1881

1883).
A more
complete
account is
given
in his
book, A2>pli-
cations des
potcntiels
a I'etude de
Viquilibre
et du monrement des solides
elastiques, Paris,
1885.
""
Rome,
Ace.
Lincei,
Mem.
fis. mat.,
1882.
"*
J.
f.
Math.
(Crelle),
Bd. 92
(1882),
and
Verhandlungen
des Vereins zur
Vejiirderung
des
Gewerbefleisses, Berlin,
1882. The memoirs are
reprinted
in Ges. Werke von Heinrich
Hertz,
Bd.
1, Leipzig,
1895,
pp.
155 and 174.
HISTORICAL INTRODUCTION
17
been
developed
from a result noted
by
G. B.
Airy.
He observed
that,
in the case of two
dimensions,
the
equations
of
equilibrium
of a
body
deformed
by
surface tractions show that the
stress-components
can be
expressed
as
partial
differential coefficients of the second order of a
single
function. Maxwell"' extended the result to three
dimensions,
in which case
three such "stress-functions" are
required.
It
appeared
later that these
functions are connected
by
a rather
complicated system
of differential
equations'^
The
stress-components
must in fact be connected with the
strain-components by
the stress-strain
relations,
and the
strain-components
are not
independent
;
but the second differential coefficients of the strain-
components
with
respect
to the coordinates are connected
by
a
system
of
linear
equations,
which are the conditions
necessary
to secure that the strain-
components
shall
correspond
with a
displacement,
in accordance with the
ordinary
formulae
connecting
strain and
displacement".
It is
possible by
taking
account of these relations to obtain a
complete system
of
equations
which must be satisfied
by stress-components,
and thus the
way
is
open
for a direct determination of stress without the intermediate
steps
of
forming
and
solving
differential
equations
to determine the
components
of
displace-
ment". In the case of two dimensions the
resulting equations
are of a
simple
character,
and
many interesting
solutions can be obtained.
The
theory
of the free vibrations of solid bodies
requires
the
integration
of the
equations
of
vibratory
motion in accordance with
prescribed boundary
conditions of stress or
di.splacement.
Poisson^'
gave
the solution of the
problem
of free radial vibrations of a solid
sphere,
and Clebsch""
founded the
general theory
on the model of Poisson's solution. This
theory
included the
extension of the notion of
"principal
coordinates" to
systems
with an infinite
number of
degrees
of
freedom,
the introduction of the
corresponding
"
normal
functions,"
and the
proof
of those
properties
of these functions
upon
which
the
expansions
of
arbitrary
functions
depend.
The discussions
which had
taken
place
before and
during
the time of Poisson
concerning
the vibrations
of
strings,
bars,
membranes and
plates
had
prepared
the
way
for Clebsch's
generalizations.
Before the
publication
of Clebsch's
treatise
a different
theory
had been
propounded by
Lame".
Acquainted
with Poisson's
discovery
of two
types
of
waves,
he concluded that the vibrations
of
any
solid
body
must fall
into two
corresponding
classes,
and he
investigated
the vibrations
of various
bodies on this
assumption.
The fact that his solutions
do not
satisfy
the
conditions which hold at the boundaries
of bodies free from surface traction
'"
Brit. Assoc.
Rep. 1862,
and Phil. Trans.
Roy.
Soc,
vol. 163
(1863), p.
49.
1
Edinburgh Roy.
Soc.
Trans.,
vol. 26
{1870)
=
Scientific Papers,
vol.
2, p.
161.
"
W. J.
Ibbetson,
An
Elementary
Treatise on the Mathematical Theory
of
perfectly
ElaMc
Solids, London,
1887.
"
Saint-Venant
gave
the identical relations
between strain-components
in his edition of
Navier's Resume des
Lemons
sur
I'application
de la
iM^canique,
Paris, 1864, 'Appeudice
3.'
"
J. H.
MicheU,
London Math. Soc. Proc. vol. 31
(1900),
p.
100.
h. E,
18 HISTORICAL INTRODUCTION
is a sufficient
disproof
of his
theory
;
but it was
finally disposed
of when
all the modes of free vibration of a
homogeneous isotropic sphere
were deter-
mined,
and it was
proved
that the classes into which
they
fall do not
verify
Lam(?'s
supposition.
The
analysis
of the
general problem
of the vibrations
of a
sphere
was first
completely given by
P.
Jaerisch"'',
who showed that the
solution could be
expressed by
means of
spherical
harmonics and certain
functions of the distance from the centre of the
sphere,
which are
practically
Bessel's functions of order
integer'
+
^.
This result was obtained
indepen-
dently by
H.
Lamb^",
who
gave
an account of the
simpler
modes of vibration
and of the nature of the nodal division of the
sphere
which occurs when
any
normal vibration is executed. He also calculated the more
important
roots
of the
frequency equation.
L. Pochhammer" has
applied
the method of
normal functions to the vibrations of
cylinders,
and has found modes of
vibration
analogous
to the known
types
of vibration of bars.
The
problem
of
tracing, by
means of the
equations
of
vibratory
motion,
the
propagation
of waves
through
an elastic solid medium
requires
investi-
gations
of a different character from those concerned with normal modes of
vibration. In the case of an
isotropic
medium Poisson and
Ostrogradsky"''
adopted
methods which involve a
synthesis
of solutions of
simple
harmonic
type,
and obtained a solution
expressing
the
displacement
at
any
time in
terms of the initial distribution of
displacement
and
velocity.
The investi-
gation
was afterwards conducted in a different fashion
by
Stokes^,
who
showed that Poisson's two waves are waves of irrotational dilatation and
waves of
equivoluminal
distortion,
the latter
involving
rotation of the
elements of the medium.
Cauchy*'
and Green*' discussed the
propagation
of
plane
waves
through
a
crystalline
medium,
and obtained
equations
for
the
velocity
of
propagation
in terms of the direction of the normal to the
wave-front. In
general
the wave-surface has three sheets
;
when the medium
is
isotropic
all the sheets are
spheres,
and two of them are coincident.
Blanchet*'' extended Poisson's results to the case of a
crystalline
medium.
Christoffel** discussed the advance
through
the medium of a surface of
discontinuity.
At
any
instant,
the surface
separates
two
portions
of the
medium in which the
displacements
ai-e
expressed
by
different formulae
;
and Christoffel showed that the surface moves
normally
to itself with a
'
J.f.
Math.
(Crelle),
Bd. 88
(1880).
"
Loudon Math. Soc.
Proc,
vol. 13
(1882).
"
J.
f.
Math.
(Crelle),
Bd. 81
(1876), p.
324.

Paris,
M6m. de
I'Acad.,
t. 10
(1831).
'9
St
Petershurg,
Mgm. de
I'Acad.,
t. 1
(1831).
* '
On the
Dynamical Theory
of
Diffraction,' Cambridge
Phil. Soc.
'Trans.,
vol. 9
(1849).
Beprinted
in Stokes's Math, and
Phys. Papers,
vol. 2
(Cambridge, 1883).
*'
Cambridge
Phil. Soc.
Tram.,
vol. 7
(1839). Reprinted
in Green's Mathematical
Papers^
p.
293.
"B
J. de Math.
(Liouville),
t. 5
(1840),
t. 7
(1842).
8
Ann. di Mat.
(Ser. 2),
t. 8
(1877).
HISTORICAL INTRODUCTION
19
velocity
which is
determined,
at
any point, by
the direction of the normal to
the
surface,
according
to the same law as holds for
plane
waves
propagated
in that direction. Besides the waves of dilatation and distortion which can
be
propagated through
an
isotropic
solid
body
Lord
Rayleigh**
has investi-
gated
a third
type
which can be
propagated
over the surface. The
velocity
of waves of this
type
is less than that of either of the other two.
Before the
discovery
of the
general equations
there existed theories of
the torsion and flexure of beams
starting
from Galileo's
enquiry
and a
suggestion
of Coulomb's. The
problems
thus
proposed
are
among
the most
important
for
practical applications,
as most
problems
that have to be dealt
with
by engineers
can,
at
any
rate for the
purpose
of a
rough approximation,
be reduced to
questions
of the resistance of beams.
Cauchy
was the first
to
attempt
to
apply
the
general equations
to this class of
problems,
and his
investigation
of the torsion of a
rectangular prism*", though
not
correct,
is
historically important,
as he
recognized
that the normal sections do not
remain
plane.
His result had little influence on
practice.
The
practical
treatises of the earlier half of the last
century
contain a
theory
of torsion
with a result that we have
already
attributed to
Coulomb, viz.,
that the
resistance to torsion is the
product
of an elastic
constant,
the amount of
the
twist,
and the moment of inertia of the cross-section.
Again,
in
regard
to
flexure,
the
practical
treatises of the time followed the Bernoulli-
Eulerian
(really Coulomb's) theory, attributing
the resistance to flexure
entirely
to extension and contraction of
longitudinal
filaments. To
Saint-Venant
belongs
the credit of
bringing
the
problems
of the torsion
and flexure of beams under the
general theory. Seeing
the
difficulty
of
obtaining general
solutions,
the
pressing
need for
practical purposes
of some
theory
that could be
applied
to the
strength
of
structures,
and the im-
probability
of the
precise
mode o'f
application
of the load to the
parts
of
any
apparatus being
known,
he was led to reflect on the methods used for the
solution of
special problems
before the formulation of the
general equations.
These reflexions led him to the invention of the semi-inverse method of
solution which bears his name. Some of the habitual
assumptions,
or some
of the results
commonly
deduced from
them,
may
be
true,
at least in a
large
majority
of
cases;
and it
may
be
possible by retaining
some of these
assumptions
or results to
simplify
the
equations,
and thus to obtain solutions

not indeed such as

satisfy arbitrary
surface
conditions,
but such as
satisfy
practically important types
of surface conditions.
The first
problem
to which Saint-Venant
applied
his method
was that
of the torsion of
prisms,
the
theory
of which he
gave
in the famous memoir
on torsion of 1855'". For this
application
he assumed the state of strain
to consist of a
simple
twist about the axis of the
prism,
such as is
implied
8*
London Math. Soc.
Froc,
vol. 17
(1887)
=
Scientific Papers,
vol.
2, Cambridge,
1900,
p.
441.
*"*
Exercices de
matltematiques,
4me Annfe,
1829.
22
20 HISTORICAL INTRODUCTION
in Coulomb's
theory,
combined with the kind of strain that is
implied by
a
longitudinal displacement
variable over the cross-section of the
prism.
The
effect of the latter
displacement
is manifested in a distortion of the sections
into curved surfaces. He showed that a state of strain
having
this character
can be maintained in the
prism by
forces
applied
at its ends
only,
and that
the forces which must be
applied
to the ends are
statically equivalent
to
a
couple
about the axis of the
prism.
The
magnitude
of the
couple
can
be
expressed
as the
product
of the
twist,
the
rigidity
of the
material,
the
square
of the area of the cross-section and a numerical factor which
depends
upon
the
shape
of the cross-section. For a
large
class of sections this
numerical factor is
very nearly proportional
to the ratio of the area of the
section to the
square
of its radius of
gyration
about the axis of the
prism.
Subsequent investigations
have shown that the
analysis
of the
problem
is
identical with that of two distinct
problems
in
hydrodynamics,
viz.,
the flow
of viscous
liquid
iu a narrow
pipe
of the same form as the
prism**,
and the
motion
produced
in frictionless
liquid filling
a vessel of the same form as the
prism
when the vessel is rotated about its axis*'. These
hydrodynamical
analogies
have resulted in a considerable
simplification
of the
analysis
of
the
problem.
The old theories of flexure involved two
contradictory assumptions:
(1)
that the strain consists of extensions and contractions of
longitudinal
filaments,
(2)
that the stress consists of tension in the extended fila-
ments
(on
the side remote from the centre of
curvature)
and
pressure
along
the contracted filaments
(on
the side nearer the centre of
curvature).
If the stress is
correctly given by
the second
assumption
there must be
lateral contractions
accompanying
the
longitudinal
extensions and also
lateral extensions
accompanying
the
longitudinal
contractions.
Again,
the
resultant of the tractions across
any
normal section of the bent
beam,
as
given by
the old
theories, vanishes,
and these tractions are
statically equi-
valent to a
couple
about an axis at
right angles
to the
plane
of
bending.
Hence the theories are
inapplicable
to
any
case of
bending by
a transverse
load. Saint-Venant**
adopted
from the older theories two
assumptions.
He
assumed that the extensions and contractions of the
longitudinal
filaments
are
proportional
to their distances from the
plane
which is drawn
through
the line of centroids of the normal sections
(the
"
central-line
")
and at
right
angles
to the
plane
of
bending.
He assumed also that there is no normal
traction across
any plane
drawn
parallel
to the central-line. The states of
stress and strain which
satisfy
these conditions in a
prismatic body
can be
maintained
by
forces and
couples applied
at the ends
only,
and include two
cases. One case is that of uniform
bending
of a bar
by couples applied
at its
8
J.
Boussinesq,
J. de Math.
{Liouville), {Sir. 2),
t. IG
(1871).
^
Kelvin and
Tait,
Nat.
Phil.,
Part
2, p.
242.
^
See the memoirs of 1855 and 1856 cited in footnote 60.
HISTORICAL INTRODUCTION
21
ends. In this case the stress is
correctly given by
the older theories and the
curvature of the central-line is
proportional
to the
bending couple,
as in those
theories;
but the lateral contractions and extensions have the effect of
distorting
those
longitudinal
sections which are at
right angles
to the
plane
of
bending
into anticlastic surfaces. The second case of
bending
which is
included in Saint-Venant's
theory
is that of a
cantilever,
or beam fixed in a
horizontal
position
at one
end,
and bent
by
a vertical load
applied
at the
other end. In this case the stress
given by
the older theories
requires
to
be corrected
by
the addition of
shearing
stresses. The normal tractions
across
any
normal section are
statically equivalent
to a
couple,
which is
proportional
to the curvature of the central-line at the
section,
as in the
theory
of
simple bending.
The
tangential
tractions across
any
normal section
are
statically equivalent
to the terminal
load,
but the
magnitude
and
direction of the
tangential
traction at
any point
are
entirely
determinate and
follow rather
complex
laws. The strain
given by
the older theories
requires
to be corrected
by
the addition of lateral contractions and
extensions,
as in
the
theory
of
simple bending,
and also
by shearing
strains
corresponding
with the
shearing
stresses.
In Saint-Venant's theories of torsion and flexure the
couples
and forces
applied
to
produce twisting
and
bending
are the resultants of tractions
exerted across the terminal
sections,
and these tractions are distributed in
perfectly
definite
ways.
The forces and
couples
that are
applied
to actual
structures are seldom distributed in these
ways.
The
application
of the
theories to
practical problems
rests
upon
a
principle
introduced
by
Saint-
Venant which has been called the
"principle
of the elastic
equivalence
of
statically equipollent systems
of load."
According
to this
principle
the
effects
produced by
deviations from the
assigned
laws of
loading
are un-
important except
near the ends' of the bent beam or twisted
bar,
and near
the ends
they produce merely
"
local
perturbations."
The condition
for
the
validity
of the results in
practice
is that the
length
of the beam should
be a considerable
multiple
of the
greatest
diameter of its cross-section.
Later researches
by
A. Clebsch'" and W.
Voigt^'
have resulted in con-
siderable
simplifications
of Saint-Venant's
analysis.
Clebsch
showed that
the
single assumption
that there is no normal traction across
any plane
parallel
to the central-line leads to four cases of
equilibrium
of a
prismatic
body,
viz.,
(1) simple
extension under terminal
tractive
load, (2) simple
bending by couples, (3)
torsion,
(4) bending
of a cantilever
by
terminal
transverse load.
Voigt
showed that the
single
assumption
that the stress
at
any point
is
independent
of the coordinate
measured
along
the bar led
to the first three
cases,
and that the
assumption
tliat the stress is a linear
function of that coordinate leads to the fourth case.
When a
quadratic
function is taken instead of a linear
one,
the case of a beam
supported
at
89 '
Theoretische Studien iiber die ElasticitatsverhiiUnisse
der
Krystallc,'
OotUngen
Abliaml-
lungen,
Bd. 34
(1887).
22 HISTORICAL INTRODUCTION
the ends and bent
by
a load which is distributed
uniformly along
its
length
can be included"". The case where the load is not uniform but is
applied
by
means of surface tractions
which,
so far as
they depend
on the coordinate
measured
along
the
beam,
are rational
integral
functions,
can be reduced to
the case where the load is uniform". It
appears
from these theories
that,
when lateral forces are
applied
to the
beam,
the relation of
proportionality
between the curvature of the central-line and the
bending
moment,
verified
in Saint-Venant's
theory,
is no
longer
exacf^-. Unless the conditions of
loading
are rather
unusual,
the modification that
ought
to be made in this
relation
is, however,
of little
practical importance.
Saint-Venant's theories of torsion and of
simple bending
have found
their
way
into technical
treatises,
but in most current books on
applied
Mechanics the
theory
of
bending by
transverse load is treated
by
a method
invented
by
Jouravski"" and
Rankine",
and
subsequently developed by
Grashof. The
components
of stress determined
by
this method do not
.satisfy
the conditions which are
necessary
to secure that
they
shall cor-
respond
with
any possible displacement".
The distribution of stress that is
found
by
this method
is, however,
approximately
correct in the case of a
beam of which the breadth is but a small fraction of the
depth**.
The most
important practical application
of the
theory
of flexure is that
which was made
by
Navier'' to the
bending
of a beam
resting
on
supports.
The load
may
consist of the
weight
of the beam and of
weights
attached to
the beam.
Young's
modulus is
usually
determined
by observing
the deflexion
of a bar
supported
at its ends and loaded at the middle. All such
applications
of the
theory depend upon
the
proportionality
of the curvature to the
bending
moment. The
problem
of a continuous beam
resting
on several
supports
was at first
very
difficult,
as a solution had to be obtained for each
span by
Navier's
method,
and the solutions
compared
in order to determine
the constants of
integration.
The
analytical complexity
was
very
much
diminished when
Clapeyron"'
noticed that the
bending
moments at three
consecutive
supports
are connected
by
an invariable
relation,
but in
many
particular
cases the
analysis
is still formidable. A method of
graphical
"
J. H.
Miohell, Quart.
J.
of Math.,
vol. 32
(1901).
"
E.
Almansi, Home,
Ace. Lincei Rend.
(Ser. 6),
t. 10
(1901), pp. 333,
400. In the second of
these
papers
a solution of the
problem
of
bending by
uniform load is obtained
by
a method which
differs from that used
by
Michell in the
paper just
cited.
'^
This result was first noted
by
K.
Pearson, Quart.
J.
of Math.,
vol. 24
(1889),
in con-
nexion with a
particular
law for the distribution of the load over the cross-section.
^
Ann. des
pants
et
chaussees,
1856.
**
Applied Mechanics,
1st
edition, London,
1858. The method has been retained in later
editions.
"
Elasticitiit and
Festigkeil,
2nd
edition, Berlin,
1878. Grashof
gives
Saint-Venant's
theory
as well.

Saint-Venant noted this result in his edition of Navier's

icfOJis, p.
394.
"'
In the second edition of his
Letjons (1833).

Paris,
C.
R.,
t. 45
(1857).
The
history
of
Clapeyron's
theorem is
given by
J. M.
Heppel,
Proc.
Ray. Soc, London,
vol. 19
(1871).
HISTORICAL
INTRODUCTION
23
solution
has, however,
been invented
by
Mohr* and it
has,
to a
great
extent,
superseded
the calculations that were
formerly
conducted
by
means of
Clapeyron's
"
Theorem of Three Moments."
Many
other
applications
of the
theory
of flexure to
problems
of frameworks will be found in such books as
Miiller-Breslau's Die Neueren Methoden der
Festigkeitslehre
(Leipzig, 1886),
Weyrauch's
Theorie Elastischer
Korper (Leipzig, 1884),
Ritter's Anwend-
ungen
der
graphischen
Statik
(Zurich, 1888).
A considerable literature has
sprung up
in this
subject,
but the use made of the
Theory
of
Elasticity
is
small.
The
theory
of the
bending
and
twisting
of thin rods and wires

in-
cluding
the
theory
of
spiral springs

was for a
long
time
developed,
independently
of the
general equations
of
Elasticity, by
methods akin to
those
employed by
Euler. At first it was
supposed
that the flexural
couple
must be in the
osculating plane
of the curve formed
by
the central-line
;
and,
when the
equation
of moments about the
tangent
was introduced
by
Binet"*,
Poisson'" concluded from it that the moment of torsion was
constant. It was
only by
slow
degrees
that the notion of two flexural
couples
in the two
principal planes sprang up,
and that the measure of
twist came to be understood. When these elements of the
theory
were made
out it could be seen that a
knowledge
of the
expressions
for the flexural and
torsional
couples
in terms of the curvature and twist
^"^
would be
sufficient,
when combined with the
ordinary
conditions of
equilibrium,
to determine
the form of the curve assumed
by
the
central-line,
the twist of the wire
around that
line,
and the tension and
shearing
forces across
any
section. The
flexural and torsional
couples,
as well as the resultant forces across a
section,
must arise from tractions exerted across the elements of the
section,
and
the correct
expressions
for them must be
sought by
means of the
general
theory.
But here a
difficulty
arises from the fact that the
general equations
are
applicable
to small
displacements only,
while the
displacements
in such
a
body
as a
spiral spring
are
by
no means small. Kirchhoff'"" was the first
to face this
difficulty.
He
pointed
out that the
general equations
are
strictly
applicable
to
any
small
portion
of a thin rod if all the linear dimensions of
the
portion
are of the same order of
magnitude
as the diameters of the cross-
sections. He held that the
equations
of
equilibrium
or motion of such a

'Beitrag
zur Theorie des
Fachwerkg,' Zeitschrift
des Architekten- uiid
Ingenieur-Vereim
zu
Hannover,
1874. This is the reference
given by
MuUer-Brcslau.
L^vy gives
an account of the
method in his
Statique Graphique,
t.
2,
and attributes it to Mohr. A
slightly
different account
is
given by
Canevazzi in Memorie delV Accademia di
Bologna (Ser.
4),
t. 1
(1880).
The method
has been extended
by
Culman,
Die
graphisrhe
Statik,
Bd.
1, Ziirich,
1875. See also Bitter,
DU
elastische Linie und ihre Auwemlumi intf
den continuirl ichen ISalken, Ziirich,
1883.
"*'
J. de I'Ecole
polytechnique,
t. 10
(1815).
""
Correspondance
sur I'Ecole
poll/technique,
t. 3
(1816).
!
They
are due to Saint-
Venaut, Paris,
G.
It.,
tt.
17,
19
(1843, 1844).
103
'tjber das
Gleiehgewieht
und die
Beweguug
eines unendlioh diinuen elastisohen Stabes,"
J.
f.
Math.
{Crelle),
Bd. 56
(1859).
The
theory
is also
given
in Kirohhoff's Vorlemngen
iiber math.
Physik,
Mechanik
(3rd edition, Leipzig, 1883).
24 HISTORICAL INTRODUCTION
portion
could be
simplified,
for a first
approximation, by
the omission of
kinetic reactions and forces distributed
through
the volume. The
process
by
which Kirchhoff
developed
his
theory
was,
to a
great
extent,
kine-
matical. When a thin rod is bent and
twisted,
every
element of it
undergoes
a strain
analogous
to that in one of Saint-Venant's
prisms,
but
neighbouring
elements must continue to fit. To
express
this kind of con-
tinuity
certain conditions are
necessary,
and these conditions take the form
of dififerential
equations connecting
the relative
displacements
of
points
within a small
portion
of the rod with the relative coordinates of the
points,
and with the
quantities
that define the
position
of the
portion
relative to the
rod as a whole. From these differential
equations
Kirchhoff deduced an
approximate
account of the strain in an
element
of the
rod,
and thence
an
expression
for the
potential energy per
unit of
length,
in terms of the
extension,
the
components
of curvature and the twist. He obtained the
equations
of
equilibrium
and vibration
by varying
the
energy-function.
In
the case of a thin rod
subjected
to terminal forces
only
he showed that the
equations by
which the form of the central-line is determined are identical
with the
equations
of motion of a
heavy rigid body
about a fixed
point.
This
theorem is known as
"
Kirchhoff's kinetic
analogue."
Kirchhoff's
theory
has
given
ri.se to much discussion. Clebsch'"
proposed
to
replace
that
part
of it
by
which the flexural and torsional
couples
can be
evaluated
by
an
appeal
to the results of Saint-Venant's theories of flexure
and torsion. Kelvin and Tait"
proposed
to establish Kirchhoff's formula
for the
potential energy by general reasoning.
J.
Boussinesq'"'' proposed
to
obtain
by
the same kind of
reasoning
Kirchhoff's
approximate expression
for the extension of a
longitudinal
filament. Clebsch'"
gave
the modified
formulae for the flexural and torsional
couples
when the central-line of the
rod in the unstressed state is
curved,
and his results have been confirmed
by
later
independent investigations.
The discussions which have taken
place
have cleared
up many
difficulties,
and the results of the
theory,
as
distinguished
from the methods
by
which
they
were
obtained,
have been
confirmed
by
the later writers.
The
applications
of Kirchhoff's
theory
of thin rods include the
theory
of the elastica which has been
investigated
in detail
by
means of the theorem
of the kinetic
analogue,
the
theory
of
spiral springs
worked out in detail
by
Kelvin and
Tait",
and various
problems
of elastic
stability. Among
the
latter we
may
mention the
problem
of the
buckling
of an elastic
ring
sub-
jected
to
pressure
directed
radially
inwards and the same at all
points
of the
circumference"".
'*
J. de Math.
(Liouville), (S^r. 2),
t. 16
(1871).
"
See,
for
example,
A. B.
Basset,
London Math. Soc.
Proc,
vol. 23
(1892),
and Amer. J.
of
Math.,
vol. 17
(1895),
and J. H.
Michell,
London Math. Soc.
Proc.,
vol. 31
(1900), p.
130.
i"
W.
HesB,
Math.
Ann.,
Bde. 23
(1884)
and 25
(1885).
""
This
problem appears
to have been discussed first
by Bresse,
Cours de
m4canique appUquSe,
Premiirt
partie, Paris,
1859.
HISTORICAL
INTRODUCTION
28
The
theory
of the vibrations of thin rods was
brought
under the
general
equations
of
vibratory
motion of elastic solid bodies
by
Poisson*. He
regarded
the rod as a circular
cylinder
of small
section,
and
expanded
all
the
quantities
that occur in
powers
of the distance of a
particle
from the
axis of the
cylinder.
When terms above a certain order
(the
fourth
power
of the
radius)
are
neglected,
the
equations
for flexural vibrations are
identical with Euler's
equations
of lateral vibration. The
equation
found
for the
longitudinal
vibrations had been obtained
by
Navier"*. The
equation
for the torsional vibrations was obtained first
by
Poisson'*. The chief
point
of
novelty
in Poisson's results in
regard
to the vibrations of rods is that the
coefficients on which the
frequencies depend
are
expressed
in terms of the
constants that occur in the
general equations;
but the deduction of the
generally
admitted
special
differential
equations, by
which these modes of
vibration are
governed,
from the
general equations
of
Elasticity
constituted
an advance in method. Reference has
already
been made to L. Pochhammer's
more
complete investigation''.
Poisson's
theory
is verified as an
approxi-
mate
theory by
an
application
of Kirchhoff's results. This
application
has
been extended to the vibrations of curved
bars,
the first
problem
to be
solved
being
that of the flexural vibrations of a circular
ring
which vibrates
in its own
plane.
An
important problem arising
in connexion with the
theory
of
longitudinal
vibrations is the
problem
of
impact.
When two bodies collide each is thrown
into a state of internal
vibration,
and it
appears
to have been
hoped
that
a solution of the
problem
of the vibrations set
up
in two bars which
impinge
longitudinally
would throw
light
on the laws of
impact.
Poisson"" was the
first to
attempt
a solution of the
problem
from this
point
of view. His
method of
integration
in
trigonometric
series
vastly
increases the
difficulty
of
ded>icing general
results, and,
by
an unfortunate error in the
analysis,
he
arrived at the
paradoxical
conclusion
that,
when the bars are of the same
material and
section,
they
never
separate
unless
they
are
equal
in
length.
Saint-Venant"' treated the
problem by
means of the solution of the
equation
of vibration in terms of
arbitrary
functions,
and arrived at certain
results,
of which the most
important
relate to the duration of
impact,
and to the
existence of an
apparent
"
coefficient of restitution
"
for
perfectly
elastic
bodies"% This
theory
is not confirmed
by experiment.
A correction
sug-
gested by Voigt"^
when worked
out,
led to little better
agreement,
and it
'"S
Bulletin des Sciences a la Soeiiti
philomathique,
1824.
109
E.
Hoppe, J.f.
Math.
(Crelle),
Bd. 73
(1871).
"
In his Traite de
Mecanique,
1833.
1"
'
Sur le choc
longitudinal
de deux barres
Mastiques.
.
,'
J. de Math.
{LiouvilU),
(Sir. 2),
t. 12
(1867).
"-
Of.
Hopkinson,
Messenger of
Mathematics,
vol.
4,
1874.
113
Ami.
PUys.
Chem.
{Wiedeinann),
Bd. 19
(1882).
See also
Hausmaninger
in the same
Annalen,
Bd. 25
(1885).
26 HISTORICAL INTRODUCTION
thus
appears
that the
attempt
to trace the
phenomena
of
impact
to vibra-
tions must be abandoned. Much more successful was the
theory
of
Hertz'",
obtained from a solution of the
problem
which we have named the
problem
of the transmission of force. Hertz made an
independent investigation
of
a
particular
case of this
problem

that of two bodies

pressed together.
He
proposed
to
regard
the strain
produced
in each
by impact
as a local statical
effect,
produced gradually
and
subsiding gradually
;
and he found means to
determine the duration of
impact
and tlie size and
shape
of the
parts
that come
into contact. The
theory yielded
a
satisfactory comparison
with
experiment.
The
theory
of vibrations can be
applied
to
problems concerning
various
kinds of shocks and the effects of
moving
loads. The inertia as well as
the elastic reactions of bodies come into
play
in the resistances to strain
under
rapidly changing
conditions,
and the resistances called into action
are sometimes described as
"dynamical
resistances." The
special problem
of the
longitudinal impact
of a massive
body upon
one end of a rod was
discussed
by
Sebert and
Hugoniot"'
and
by Boussinesq"^
The conclusions
which
they
arrived at are tabulated and illustrated
graphically
by
Saint-
Venant"'. But
problems
of
dynamical
resistance under
impulses
that tend
to
produce
flexure are
perhaps practically
of more
importance.
When a
body
strikes a rod
perpendicularly
the rod will be thrown into
vibration,
and,
if the
body
moves with the
rod,
the
ordinary
solution in terms of
the normal functions for the vibrations of the rod becomes
inapplicable.
Solutions of several
problems
of this
kind,
expressed
in terms of the normal
functions for the
compound system consisting
of the rod and the
striking
body,
were
given by
Saint-Venant"'.
Among problems
of
dynamical
resistance we must note
especially
Willis's
problem
of the
travelling
load. When a train crosses a
bridge,
the strain is
not identical with the statical strain which is
produced
when the same train
is
standing
on the
bridge.
To illustrate the
problem
thus
presented
W^illis""
proposed
to consider the
bridge
as a
straight
wire and the train as a
heavy
particle deflecting
it.
Neglecting
the inertia of the wire he obtained a
certain differential
equation,
which was
subsequently
solved
by
Stokes'*.
Later writers have shown that the effects of the
neglected
inertia are
very
"*
'Ueber die
Beriihrung
fester elastischer
Korper,'
J.
f.
Math.
[Grelle),
Bd. 92
(1882).
1"
Paris,
V.
11.,
t. 95
(1882).
""
Applications
des Potentiels . . .
, Paris,
1885. The results were
given
in a note in Paris C.
R.,
t. 97
(1883).
"'
In
papers
in
Paris,
G.
R.,
t. 97
(1883), reprinted
as an
appendix
to his Translation
of Clebsch's Treatise
(Paris, 1883).
"*
In the 'Annotated Clebsch'
just cited,
Note du

fil. Cf. Lord
Rayleigh, Tlicory of
Sotitid,
Chapter
VIII.
""
Appendix
to the
Report of
the Commissioners... to
enquire
into the
Application of
Iron to
Railway
Structures
(1849).
'2"
Cambridge,
Phil. Soc.
Trans.,
vol. 8
(1849)
=
Stokes,
Math, and
Phijs. Papers,
vol. 2
(Cambridge, 1883), p.
178.
HISTORICAL INTRODUCTION
27
important.
A more
complete
solution has been obtained
by
M.
Phillips'"
and Saint-
Venant'**,
and an admirable
precis
of their results
may
be read in
the second volume of Todhunter and Pearson's
History (Articles
373 et
seq.).
We have seen
already
how
problems
of the
equilibrium
and vibrations of
plane plates
and curved shells were
attempted
before the
discovery
of the
general equations
of
Elasticity,
and how these
problems
were
among
those
which led to the
investigation
of such
equations.
After the
equations
had
been formulated little advance seems to have been made in the treatment
of the
problem
of shells for
many years,
but the more
special problem
of
plates
attracted much attention. Poisson'-^ and
Cauchy'^*
both treated this
problem, proceeding
from the
general equations
of
Elasticity,
and
supposing
that all the
quantities
which occur can be
expanded
in
powers
of the
distance from the middle-surface. The
equations
of
equilibrium
and free
vibration which hold when the
displacement
is directed at
right angles
to
the
plane
of the
plate
were deduced. Much
controversy
has arisen con-
cerning
Poisson's
boundary
conditions. These
expressed
that the resultant
forces and
couples applied
at the
edge
must be
equal
to the forces and
couples arising
from the strain. In a famous memoir Kirchhoflf^ showed
that these conditions are too numerous and cannot in
general
be satisfied.
His method rests on two
assumptions
:
(1)
that linear filaments of the
plate
initially
normal to the middle-surface remain
straight
and normal to the
middle-surface after
strain,
and
(2)
that all the elements of the middle-
surface remain unstretched. These
assumptions
enabled him to
express
the
potential energy
of the bent
plate
in terms of the curvatures
produced
in its
middle-suri'ace.
The
equations
of motion and
boundary
conditions were then
deduced
by
the
principle
of virtual
work,
and
they
were
applied
to the
problem
of the flexural vibrations of a circular
plate.
The
problem
of
plates
can be attacked
by
means of considerations of the
same kind as those which were used
by
Kirchhoff in his
theory
of thin rods.
An
investigation
of the
problem by
this method was made
by Gehring''^"
and
was afterwards
adopted
in an
improved
form
by
Kirchhoff'". The work is
very
similar in detail to that in Kirchhoff's
theory
of thin
rods,
and it leads to
an
expression
for the
potential energy per
unit of area of the middle-surface
'2'
Paris,
Aim. des
Mines,
t. 7
(1855).
122
lu the 'Annotated
Clebsch,'
Note du

Gl.
'-'
In the memoir of 1828. A
large part
of the
investigation
is
reproduced
in Todhunter
and Pearson's
History.
''^
In an Article 'Sur
I'fiquilibre
et le mouvement d'une
plaque
solide' in the Kxercices de
mathematiques,
vol. 3
(1828).
Most of this Article also is
reproduced by
Todhunter and Pearson.
'-
J.
f.
Math.
(Crelle),
Bd. 40
(1850).
'2
'De
^quationibus
differentialibus
quibus ajquilibrium
et motus lamina;
orystaUinsB
definiuntur'
(Diss.),
Berlin,
18(>0. The
analysis may
be read in Kirchhoff's VorUsuiigen
iiber
viath.
Pliys.,
Mechanik,
and
parts
of it also in Clebsch's Treatise.
127
Vorlesungen
iiber math.
Phys.,
Mechanik.
28 HISTORICAL INTRODUCTION
of the
plate.
This
expression
consists of two
parts
: one a
quadratic
function
of the
quantities defining
the extension of the middle-surface with a coefficient
proportional
to the thickness of the
plate,
and the other a
quadratic
function
of the
quantities defining
the flexure of the middle-surface with a
coefficient
proportional
to the cube of the thickness. The
equations
of small motion
are deduced
by
an
application
of the
principle
of virtual work. When the
displacement
of a
point
on the middle-surface is
very
small the flexure
depends only
on
displacements
directed at
right angles
to the
plane
of the
plate,
and the extension
only
on
displacements
directed
parallel
to the
plane
of the
plate,
'and the
equations
fall into two sets. The
equation
of normal
vibration and the
boundary
conditions are those
previously
found and dis-
cussed
by
Kirchhoff"^.
As in the
theory
of
rods,
so also in that of
plates,
attention is directed
rather to
tensions,
shearing
forces and flexural
couples,
reckoned across
the whole
thickness,
than to the tractions across elements of area which
give
rise to such forces and
couples.
To fix ideas we
may
think of the
plate
as
horizontal,
and consider the actions exerted across an
imagined
vertical
dividing plane,
and on this
plane
we
may
mark out a small area
by
two
vertical lines near
together.
The distance between these lines
may
be called
the
"
breadth
"
of the area. The tractions across the elements of this area
are
statically equivalent
to a force at the centroid of the area and a
couple.
When the
"
breadth
"
is
very
small,
the
magnitudes
of the force and
couple
are
proportional
to the
breadth,
and we estimate them as so much
per
unit
of
length
of the line in which our vertical
dividing plane
cuts the middle
plane
of the
plate.
The
components
of the force and
couple
thus estimated
we call the "stress-resultants" and the
"stress-couples."
The stress-resultants
consist of a tension at
right angles
to the
plane
of the
area,
a horizontal
shearing
force and a vertical
shearing
force. The
stress-couples
have a
component
about the normal to the
dividing plane
which we shall call the
"
torsional
couple,"
and a
component
in the vertical
plane containing
this
normal which we shall call the
"
flexural
couple."
The stress-resultants and
stress-couples depend upon
the direction of the
dividing plane,
but
they
are known for all such directions when
they
are known for two of them.
Clebsch"^
adopted
from the
Kirchhoff-Gehring theory
the
approximate
account
of the strain and stress in a small
portion
of the
plate
bounded
by
vertical
dividing planes,
and he formed
equations
of
equilibrium
of the
plate
in terms
of stress-resultants and
stress-couples.
His
equations
fall into two
sets,
one
set
involving
the tensions and horizontal
shearing
forces,
and the other set
involving
the
stress-couples
and the vertical
shearing
forces. The latter
set of
equations
are those which relate to the
bending
of the
plate,
and
they
have such forms
that,
when the
expressions
for the
stress-couples
are known
in terms of the deformation of the middle
plane,
the
vertical
shearing
forces
can be
determined,
and an
equation
can be formed for the deflexion of the
HISTORICAL
INTRODUCTION
29
plate.
The
expressions
for the
couples
can be obtained from
Kirchhoff's
theory.
Clebsch solved his
equation
for the deflexion of a circular
plate
clamped
at the
edge
and loaded in an
arbitrary
manner.
All the
theory
of the
equations
of
equilibrium
in terms of
stress-resultants
and
stress-couples
was
placed beyond
the reach of criticism
by
Kelvin and
Tait". These authoi-s noticed
also, that,
in the case of uniform
bending,
the
expressions
for the
stress-couples
could be deduced from
Saint-Venant's
theory
of the anticlastic flexure of a bar
;
and
they explained
the union of
two of Poisson's
boundary
conditions in one of Kirchhoff's as an
example
of the
principle
of the elastic
equivalence
of
statically equipollent systems
of
load. More recent researches have assisted in
removing
the
difficulties which
had been felt in
respect
of Kirchhoff's
theory'^.
One obstacle to
progress
has been the lack of exact solutions of
problems
of the
bending
of
plates
analogous
to those found
by
Saint-Venant for beams. The few solutions
of this kind which have been obtained''"' tend to confirm the main result
of the
theory
which has not been
proved rigorously,
viz. the
approximate
expression
of the
stress-couples
in terms of the curvature of the middle-
surface.
The
problem
of curved
plates
or shells was first attacked from the
point
of view of the
general equations
of
Elasticity by
H. Aron'*". He
expressed
the
geometry
of the middle-surface
by
means of two
parameters
after the
manner of
Gauss,
and he
adapted
to the
problem
the method which Clebsch
had used for
plates.
He arrived at an
expression
for the
potential energy
of
the strained shell which is of
th^
same form as that obtained
by
Kirchhoff
for
plates,
but the
quantities
that define the curvature of the middle-surface
were
replaced by
the differences of their values in the strained and unstrained
states. E. Mathieu"'
adapted
to the
problem
the method which Poisson had
used for
plates.
He observed that the modes of vibration
possible
to a shell
do not fall into classes characterized
respectively by
normal and
tangential
displacements,
and he
adopted equations
of motion that could be deduced
from Aron's formula for the
potential
energy by retaining
the terms that
depend
on the
stretching
of the middle-surface
only.
Lord
Rayleigh'"
proposed
a different
theory.
He concluded from
physical reasoning
that
the middle-surface of a
vibrating
shell remains
unstretched,
and determined
the character of the
displacement
of a
point
of the middle-surface in accord-
ance with this condition. The direct
application
of the
Kirchhoff-Gehring
iss
See,
for
example,
J.
Boussinesq,
J. de Math.
{Liouville), (Sr. 2),
t. 16
(1871)
and
(S6r. 3),
t. 5
(1879 ) ;
H.
Lamb,
London Math. Soc.
Proc,
vol. 21
(1890) ;
J. H.
Miohell,
London
Math. Soc.
Proc,
vol. 31
(1900),
p.
121
;
J.
Hadamard,
Trans. Amer. Math.
Soc,
vol. 3
(1902).
'29
Some solutions were
given by
Saint-Venant
in the 'Annotated
Clebsch,' pp.
337 et
seq.
Others will be found in
Chapter
XXII of this book.
'
J.f.
Math.
(CrelU),
Bd. 78
(1874).
'"1
J. de I'Ecole
polytechnique,
t. 51
(1883).
1^
London Math. Soc
Proc,
vol. 13
(1882).
30 HISTORICAL INTRODUCTION
method^*" led to a formula for the
potential energy
of the same form as
Aron's and to
equations
of motion and
boundarj'
conditions which were
difficult to reconcile with Lord
Rayleigh's theory.
Later
investigations
have
shown that the extensional strain which was thus
proved
to be a
necessary
concomitant of the vibrations
may
be
practically
confined to a narrow
region
near the
edge
of the
shell,
but
that,
in this
region,
it
may
be so
adjusted
as
to secure the satisfaction of the
boundaiy
conditions while the
greater part
of the shell vibrates
according
to Lord
Rayleigh's type.
Whenever
very
thin rods or
plates
are
employed
in constructions it
becomes
necessary
to consider the
possibility
of
buckling,
and thus there
arises the
general problem
of elastic
stability.
We have
already
seen that
the first
investigations
of
problems
of this kind were made
by
Euler and
Lagrange.
A number of isolated
problems
have been solved. In all of them
two modes of
equilibrium
with the same
type
of external forces are
possible,
and the
ordinary proof"
of the
determinacy
of the solution of the
equations
of
Elasticity
is defective. A
general theory
of elastic
stability
has been
proposed by
G. H.
Bryan'**.
He arrived at the result that the theorem of
determinacy
cannot fail
except
in cases where
large
relative
displacements
can be
accompanied by very
small
strains,
as in thin rods and
plates,
and in
cases where
displacements differing
but
slightly
from such as are
possible
in a
rigid body
can take
place,
as when a
sphere
is
compressed
within a
circular
ring
of
slightly
smaller diameter. In all cases where two modes of
equilibrium
are
possible
the criterion for
determining
the mode that will be
adopted
is
given by
the condition that the
energy
must be a minimum.
The
history
of the mathematical
theory
of
Elasticity
shows
clearly
that
the
development
of the
theory
has not been
guided exclusively by
con-
siderations of its
utility
for technical Mechanics. Most of the men
by
whose
researches it has been founded and
shaped
have been more interested in
Natural
Philosophy
than in material
progress,
in
trying
to understand the
world than in
trying
to make it more comfortable. From this attitude of
mind it
may possibly
have resulted that the
theory
has contributed less to
the material advance of mankind than it
might
otherwise have done. Be
this as it
may,
the intellectual
gain
which has accnied from the work of these
men must be estimated
very highly.
The discussions that have taken
place
concerning
the number and
meaning
of the elastic constants have thrown
light
on most recondite
questions concerning
the nature of molecules and
the mode of their interaction. The efforts that have been made to
explain
optical phenomena
by
means of the
hypothesis
of a medium
having
the same
physical
character as an elastic solid
body
led,
in the first
instance,
to the
understanding
of a concrete
example
of a medium which can transmit
i*>
A. E. H.
Love,
Phil. Trans.
Roy.
Soc.
(Ser. A),
vol. 179
(1888).
"*
KirohhofI, Vorleiiungen
iiher math.
Plitjs.,
Mechanik.
'
Cambridge
Phil. Soc.
Pruc,
vol. C
(1889), p.
199.
HISTORICAL INTRODUCTION
81
transverse
vibrations, and,
at a later
stage,
to the definite conclusion that
the himiniferous medium has not the
physical
character assumed in the
hypothesis. They
have thus issued in an essential
widening
of our ideas
concerning
the nature of the aitber and the nature of luminous
vibrations.
The methods that have been devised for
solving
the
equations
of
equilibrium
of an
isotropic
solid
body
form
part
of an
analytical theory
which is of
great
importance
in
pure
mathematics. The
application
of these methods to the
problem
of the internal constitution of the Earth has led to results which
must influence
profoundly
the course of
speculative thought
both in
Geology
and in cosmical
Physics.
Even in the moj'e technical
problems,
such as the
transmission of force and the resistance of bars and
plates,
attention has been
directed,
for the most
part,
rather to theoretical than to
practical aspects
of
the
questions.
To
get insight
into what
goes
on in
impact,
to
bring
the
theory
of the behaviour of thin bars and
plates
into accord with the
general
equations

these and such-like aims have been more attractive to most of

the men to whom we owe the
theory
than endeavours to devise means for
effecting
economies in
engineering
constructions or to ascertain the conditions
in which structures become unsafe. The fact that much material
progress
is
the indirect outcome of work done in this
spirit
is not without
significance.
The
equally significant
fact that most
great
advances in Natural
Philosophy
have been made
by
men who had a first-hand
acquaintance
with
practical
needs and
experimental
methods has often been
emphasized
; and,
although
the names of
Green, Poisson,
Cauchy
show that the rule' is not without
important exceptions, yet
it is
exemplified
well in the
history
of our science.
k
CHAPTER I.
ANALYSIS OF STRAIN.
1 . Extension.
Whenever,
owing
to
any
cause,
changes
take
place
in the relative
positions
of the
parts
of a
body
the
body
is said to be
"
strained." A
very
simple example
of a strained
body
is a stretched bar. Consider a bar of
square
section
suspended vertically
and loaded with a
weight
at its lower
end. Let a line be traced on the bar in the direction of its
length,
let
two
points
of the line be
marked,
and let the distance between these
points
be measured. When the
weight
is attached the distance in
question
is
a little
greater
than it was before the
weight
was attached. Let
l^
be the
length
before
stretching,
and I the
length
when stretched. Then
{I

lo)jlo
is a number
(generally
a
very
small
fraction)
which is called the extension of
the line in
question.
If this number is the same for all lines
parallel
to the
length
of the
bar,
it
may
be called
"
the extension of the bar." A steel
bar of sectional area 1
square
inch
(=
6'4515
cm.')
loaded with 1 ton
(=
1016'05
kilogrammes)
will
undergo
an extension of about 7 x 10~'. It
is clear that for the measurement of such small
quantities
as this rather
elaborate
apparatus
and refined methods of observation are
required*.
Without
attending
to methods of measurement we
may
consider a little
more in detail the state of strain in the stretched bar. Let e denote the
extension of the
bar,
so that its
length
is increased in the ratio
1+e
:
1,
and
consider the volume of the
portion
of the bar contained between
any
two
marked sections. This volume is increased
by stretching
the
bar,
but not
in the ratio 1
+ e : 1. When the bar is stretched
longitudinally
it contracts
laterally.
If the linear lateral contraction is
e,
the sectional area is dimi-
nished in the ratio
(1

e')':
1,
and the volume in
question
is increased
in the ratio
(l
+
e)(l e')^
: 1. In the case of a bar under tension e' is a
certain
multiple
of
e,
say
are,
and a is about
^
or
|
for
very many
materials.
If e is
very
small and e* is
neglected,
the areal contraction is
2a-e,
and the
cubical dilatation is
(1

2a-)
e.
I
*
Se,
for
example, Ewing, Strength of
Materials
(Oambridge, 1899), pp.
73 et
teq.
1-3]
EXTENSION AND SHEAR 33
For the
analytical description
of the state of strain in the bar we should
take an
origin
of coordinates
x,
y,
z on the
axis,
and measure the coonlinate z
along
the
length
of the bar.
Any particle
of the bar which has the co-
ordinates
X,
y,
z when the
weight
is not attached will move after the
attachment of the
weight
into a new
position.
Let the
particle
which was
at the
origin
move
through
a distance
z^,
then the
particle
which waa at
{x,
y, z)
moves to the
point
of which the coordinates are
x{\-ae), y(l-cre),
z^
+
{z
-
z) (I
+
e).
The state of strain is not
very simple.
If lateral forces could be
applied
to the bar to
prevent
the lateral contraction the state of strain would be
very
much
simplified.
It would then be described as a
"
simple
extension."
2. Pure shear.
As a second
example
of strain let us
suppose
that lateral forces are
applied
to the bar so as to
produce
extension of amount
e,
of lines
parallel
to the axis of x and extension of amount
e^
of lines
parallel
to the axis of
y,
and that
longitudinal
forces are
applied,
if
any
are
required,
to
prevent
any
extension or contraction
parallel
to the axis of z. The
particle
which
was at
(x, y, z),
will move to
{x
+
e^x, y
+
e^y, z)
and the area of the section
will be increased in the ratio
(1
+
e,)
(1
+
e^)
: 1. If
e^
and
e,
are related
80 that this ratio is
equal
to
unity
there will be no
change
in the area
of the section or in the volume of
any portion
of the
bar,-
but the
shape
of the section will be distorted. Either
e,
or
e,
is then
negative,
or there
is contraction of the
corresponding
set of lines. The strain set
up
in the
bar is called
"pure
shear."
Fig.
1 below shows a
square
ABCD distorted
by pure
shear into a rhombus A'B'C'D' of the same area.
3.
Simple
shear.
As a third
example
of strain let us
suppose
that the bar after
being
distorted
by pure
shear is turned
bodily
about its axis. We
suppose
that
the axis oi x is the direction in which contraction takes
place,
and we
put
2

i
=
2 tan a.
Then we can show
that,
if the rotation is of amount a in the sense from
y
to
X,
the
position
reached
by any particle
is one that could have been
reached
by
the
sliding
of all the
particles
in the direction of a certain line
through
distances
proportional
to the distances of the
particles
from a certain
plane containing
this line.
Since
(1 +,)(!
+
{3)
=
1,
and
f2-i
=
2tan
a,
we have
1
+j
=
sec a

tan
a,
l
+
f2=seca
+
tan a.
By
the
pure shear,
the
particle
which was at
(x, y)
is moved to
{x^, y^),
where
i=^(sec
a

tana), ^j =y (sec
a
+
tan
a) ;
and
by
the rotation it is moved
again
to
(a-^, ^2)1
where
.r2=^iCos a+yj
sin
a, yi= x^
sin
a+yi
eoso;
L. K.
3
34
so that we have
SPECIFICATION OF STRAIN
[CH.
I
x^x+Uktia {

a'C0Sa+y(l
+
8ina)},
yj=y
+
tan a
{

X
(1

siu
a) +y
cos
a).
Now, writing
/3
for
^ir

a,
we have
ar.2
=
^
+
2 tan a cos
^/3 (
-
x sin
^/3+y
cos
i^),
y2=y
+
2tan
asini/3( ^sin^jS+ycos^jS)
;
and we can observe that
and that

:!'2sin ^j3+y2Cos^/3= -a^sin^^+ycos^^,

Xjcos Jj3+yjSin^0
=
xcos^^+3/sin
i^
+
2 tan
a(-
A' sin
^/3+y
cos
J^).
Hence,
taking
axes of X and Y which are obtained from those of x
and
y by
a rotation
through ^tt

^a
in the sense from x towards
y,
we see
that the
particle
which was at
(A", Y)
is moved
by
the
pure
shear followed
by
the rotation to the
point (X^, Y,),
where
Z,
=
X+2tana.F, Y,=
Y.
Thus
every plane
of the material which is
parallel
to the
plane
of
{X, z)
slides
along
itself in the direction of the axis of X
through
a distance
proportional
to the distance of the
plane
from the
plane
of
(X,
z).
The
kind of strain
just
described is called a
"simple
shear,"
the
angle
a is the
"
angle
of the
shear,"
and 2 tan a is the
"
amount of the shear."
Y
/y
Fig.
1.
3-5]
BY MEANS OF DISPLACEMENT
85
Fig.
1 shows a
square
ABGD distorted
by pure
shear into a rhombus
A'B'C'D' of the same
area,
which is then rotated into the
position
A"B"G"D".
The
angle
of the shear is
A'OA",
and the
angle
AOX is half the
complement
of this
angle.
The lines
AA", BB", CC",
DD" are
parallel
to OX and
propor-
tional to their distances from it.
We shall find that all kinds of strain can be desciibed in terms of
simple
extension and
simple
shear,
but for the discussion of
complex
states
of strain and for the
expression
of them
by
means of
simpler
strains we
require
a
general
kinematical
theory*.
4.
Displacement.
We
have,
in
every
case,
to
distinguish
two states of a
body

a first
state,
and a second state. The
particles
of the
body pass
from their
positions
in the first state to their
positions
in the second state
by
a
displace-
ment. The
displacement may
be such that the line
joining any
two
particles
of the
body
has the same
length
in the second state as it has in the first
;
the
displacement
is then one which would be
possible
in a
rigid body.
If the
displacement
alters the
length
of
any
line,
the second state of the
body
is described as a
"
strained
state,"
and then the first state is described
as the
"
unstrained state."
In what follows we shall denote the coordinates of the
point occupied
by
a
particle,
in the unstrained state of the
body, by
x,
y,
z,
and the co-
ordinates of the
point occupied by
the same
particle
in the strained state
by
x-Vu,
y
+
V,
z
+
w. Then
u, v,
w are the
projections
on the axes of
a vector
quantity

the
displacement.
We must take
u, v,
w to be con-
tinuous functions of
x,
y,
z,
and we shall in
general
assume that
they
are
analytic
functions.
It is clear
that,
if the
displacement (m,
v, w)
is
given,
the strained state
is
entirely
determined
;
in
particular,
the
length
of the line
joining any
two
particles
can be determined.
5.
Displacement
in
simple
extension and
simple
shear.
The
displacement
in a
simple
extension
parallel
to the axis of x is
given by
the
equations
u
=
ex,
v
=
0, w=0,
where e is the amount of the extension. If is
negative
there is contraction.
The
displacement
in a
simple
shear of amount
(
=2 tan
a), by
which lines
parallel
to
the axis of a: slide
along
themselves,
and
particles
in
any plane parallel
to the
plane
of
.{x, y)
remain in that
plane,
is
given by
the
equations
u
=
sy,
11
=
0,
w
=
0.
*
The
greater part
of the
theory
is due to
Cauohy (See
Introduction).
Some
improvements
were made
by
Clebsch in his treatise of
1862,
and others were made
by
Kelviu and
Tait,
Nat. Phil.
Part I.
3-2
36
SPECIFICATION OF STRAIN
[CH.
I
In
Fig. 2,
.45 is a
segment
of a line
parallel
to the axis of
x,
which subtends an
angle
2a at and is bisected
by Oy. By
the
simple
shear
particles lying
on the line OA are
displaced
so as to lie on OB. The
particle
at
any point
P on A B is
displaced
to
Q
on AB
80 that
PQ
=
AB,
and the
particles
on OP are
displaced
to
points
on
OQ.
A
parallelogram
svich as OPNM becomes a
parallelogram
such as
OQKM.
Fig.
2.
If the
angle
xOP=6 we
may prove
that
2 tan a tan''' 6
^^POQ=-^
taxi
xOQ
=
tand
1
+
2 tan a tan 6
'
"sec^fl +
atanatan^'
In
particvilar,
if
d=^jr,
cot
xOQ
=
s,
so
that,
if is
small,
it is the
complement
of the
angle
in the strained state between two lines of
particles
which,
in the unstrained
state^
were at
right angles
to each other.
6.
Homogeneous
strain.
Id the cases of
simple
extension and
simple
shear,
the
component
dis-
placements
are
expressed
as linear functions of the coordinates. In
general,
if a
body
is strained so that the
component displacements
can be
expressed
in this
way,
the strain is said to be
homogeneous.
Let the
displacement corresponding
with a
homogeneous
strain be
given
by
the
equations
M=ajia;+ a]2?/+ Uia^,
v=
a^x
+
a^y
-k-a^z,
w
=
OiiU;
+
a,y
+
a-^z.
Since
x,
y,
z are
changed
into x
\u,
y
+
v,
z
+w,
that
is,
are transformed
by
a linear
substitution,
any plane
is transformed into a
plane,
and
any
ellipsoid
is
transformed,
in
general,
into an
ellipsoid.
We infer at once
the
following
characteristics of
homogeneous
strain :

(i) Straight
lines,
remain
straight, (ii)
Parallel
straight
lines remain
parallel, (iii)
All
straight
lines in the same direction are
extended,
or
contracted,
in the
same ratio,
(iv)
A
sphere
is transformed into an
ellipsoid,
and
any
three
orthogonal
diameters of the
sphere
are transformed into three
conjugate
diameters
of the
ellipsoid. (v) Any ellipsoid
of a certain
shape
and
orientation is transformed into a
sphere,
and
any
set of
conjugate
diameters
5-7]
BY MEANS OF
DISPLACEMENT
87
of the
ellipsoid
is transformed into a set of
orthogonal
diameters of the
sphere, (vi)
There is one set of three
orthogonal
lines in the unstrained
state which remain
orthogonal
after the strain
;
the directions of these lines
are in
general
altered
by
the strain. In the unstrained state
they
are the
principal
axes of the
ellipsoid
referred to in
(v) ;
in the strained
state,
they
are the
principal
axes of the
ellipsoid
referred to in
(iv).
The
ellipsoid
referred to in
(iv)
is called the strain
ellipsoid;
it has
the
property
that the ratio of the
length
of a
line,
which has a
given
direction in the strained
state,
to the
length
of the
corresponding
line in
the unstrained
state,
is
proportional
to the central radius vector of the
surface drawn in the
given
direction. The
ellipsoid
referred to in
(v) may
be called the
reciprocal
strain
ellipsoid
;
it has the
property
that the
length
of a
line,
which has a
given
direction in the unstrained
state,
is increased
by
the strain in a ratio
inversely proportional
to the central radius vector
of the surface drawn in the
given
direction.
The
principal
axes of the
reciprocal
strain
ellipsoid
are called the
principal
axes
of
the strain. The extensions of lines drawn in these
directions,
in the unstrained
state,
are
stationary
for small variations of
direction. One of them is the
greatest
extension,
and another the smallest.
7. Relative
displacement.
Proceeding
now to the
general
case,
in which the strain is not
necessarily
homogeneous,
we take
(
+
x,
y
+y,
z +
z)
to be a
point
near to
(x, y,
z),
and
(m
+
u,
v
+
v,
w
+
w)
to be the
corresponding displacement.
There
will be
expressions
for the
components
u, v,
w of the relative
displacement
as series in
powers
of
x, y, z,
viz. we have
du du
,
du
,
\
u=Xr-.
+
y.-
+z^+...
dx
dy
02
dv dv dv
~
dx
dy
dz
'"
,
dw dw dw
W=X:5-+y^ +Zr-+...,/
dx
oy
02
'
where the terms that are not written contain
powers
of
x, y,
z above the
first. When
x, y,
z are
sufficiently
small,
the latter terms
may
be
neglected.
The
quantities
u, v,
w are the
displacements
of a
particle
which,
in the
unstrained
state,
is at
(a;
+
x,
y
+
y>
z
>(
z),
relative
to the
particle
which,
in the same
state,
is at
{x, y,
z).
We
may accordingly
say
that,
in a
sufficiently
small
neighbourhood
of
any point,
the relative
displacements
are linear functions of the relative coordinates.
In other words,
the strain
about
any point
is
sensibly homogeneous.
All that we have said about the
effects of
homogeneous
strain
upon straight
lines
will remain
true for linear
elements
going
out from a
point.
In
particular,
there
will be one set of
(1)
38 IRROTATIONAL DISPLACEMENT
[CH.
I
three
orthogonal
linear
elements,
in the unstrained
state,
which remain
orthogonal
after the
strain,
but the directions of these lines are in
general
altered
by
the strain. The
directions,
in the unstrained
state,
of these linear
elements at
any point
are the
"
principal
axes of the strain
"
at the
point.
8.
Analysis
of the relative
displacement*.
In the discussion of the formulae
(1)
we shall confine our attention to
the
displacement
near a
point,
and shall
neglect
terms in
x, y,
z above the
first. It is convenient to introduce the
following
notations :

\
du
^^'-dlo
7-9]
EXTENSION OF A LINEAR ELEMENT
39
displacements
is described as a
"pure
strain." We learn that the relative
displacement
is
always compounded
of a
displacement involving
a
pure
strain and a
displacement represented by
such
expressions
as

w^y
+
Wy%.
The
line-integral
of the latter
displacement
taken round a closed curve does
not vanish
(cf
Article
15,
infra).
If the
quantities
Tsy^, w,,, cr^
are
small,
the
terms such as

in-zy
+
'"'y^ represent
a
displacement
that would be
possible
in a
rigid bodj',
viz. a small rotation of amount
VCof/
+
^y'
+
^z')
about an axis
in direction
{btx
:
Wy
:
ra-^).
For this reason the
displacement corresponding
with a
pure
strain is often described as
"
irrotational."
9. Strain
corresponding
with small
displacement*.
It is clear that the
changes
of size and
shape
of all
parts
of a
body
will
be determined when the
length,
in the strained
state,
of
every
line is known.
Let
I, m,
n be the direction cosines of a line
going
out from the
point
(x, y,
z).
Take a
very
short
length
r
along
this
line,
so that the coordinates
of a
neighbouring point
on the line are x +
lr,
y
+
mr,
z
+
nr. After strain
the
particle
that was at
{x,
y, z)
comes to
{x-\-u, y
+
v,
z
+
w),
and the
particle
that was at the
neighbouring point
comes to the
point
of which the
coordinates are
, /,3m
du
,
du\ \
x+
Ir +u +
r
11^
+
7/1
^
+
M
^
1
,
/,dv
dv
,
dv\
provided
r is so small that we
may neglect
its
square.
Let
r,
be the
length
after strain which
corresponds
with r before strain.
Then we have
(5)
'l>
+
S
)-i-g"-{'^'(-i)-sF
When the relative
displacements
are
very
small,
and
squares
and
products
of
such
quantities
as
|^*,
... can be
neglected,
this formula
passes
over into
r,
=
r
[1
+
e^xl'
+
e.njm''
+
ewH'
+
e,j^mn
+
enl
+
ea:yli]
(7)
where the notation is the same as that in
equations
(2).

In the
applications
of the
theory
to strains iu elastic soUd bodies,
the
displacemenU
that
have to be considered are in
general
so small that
squares
and
products
of Brst differential
coefficients of
, f,
w with
respect
to
x,
y,
z can be
neglected
in
comparison
with their first
powers.
The more
general theory
in which this
simplification
is not made will be discussed
in the
Appendix
to this
Chapter.
*0
SPECIFICATION OF STRAIN
[CH.
I
10.
Components
of strain*.
By
the formula
(7)
we know the
length
r^
of a line
which,
in the un-
strained
state,
has an
assigned
short
length
r and an
assigned
direction
{I,
m,
n),
as soon as we know the values of the six
quantities
exx,
^yy,
^iz>
^yz,
^zx,
^xy-
These six
quantities
are called the
"
components
of strain." In the
case of
homogeneous
strain
they
are constants
;
in the more
general
case
they
are variable from
point
to
point
of a
body.
The extension e of the short line in direction
{I,
m,
n)
is
given
at once
by (7)
in the form
e
=
Bxxl^
+
Byyin^
+
ean^
+
ey^mn
\-
e^xnl
+
Bxylm, (8)
so that the three
quantities
exx,
Syy,
e^^
are extensions of linear elements
which,
in the unstrained
state,
are
parallel
to axes of coordinates.
Again
let
(li, tWj, rii)
be the direction in the strained state of a linear
element
which,
in the unstrained
state,
has the direction
(I,
m, n),
and let e
be the
corresponding
extension,
and let the same letters with accents refer
to a second linear element and its extension. From the formulae
(5)
it
appears
that
with similar
expressions
for
m,, iii.
The cosine of the
angle
between the
two elements in the strained state is
easily
found in the form
lili
+
minii
+
n,ni'
=
{W
+
mm'
+
nn') (1

e

e')
+
2
{exxH'
+
Byymm'
+
Bann)
+
Byi {mn'
+
m'n)
+
e^x {nV
+
n'l)
+
Bxy {hn
+
I'm) (9)
If the two lines in the unstrained state are the axes of w and
y
the cosine
of the
angle
between the
corresponding
lines in the strained state is
e^y-
In
like manner
Cy^
and
e^e
are the cosines of the
angles,
in the strained
state,
between
pairs
of lines
which,
in the unstrained
state,
are
parallel
to
pairs
of
axes of coordinates.
Another
interpretation
of the
strain-components
of
type exy
is afforded
immediately by
such
equations
as
_dv
du
from which it
appears
that
exy
is made
up
of two
simple
shears. In one of
these
simple
shears
planes
of the material which are at
right angles
to the

When the relative

displacement
is not small the strain is not
specified completely by
the
quantities ^j.,
...
ey^,
.... This matter is considered in the
Appendix
to this
Chapter.
Lord Kelvin
has called attention to the
unsymmetrioal
character of the
strain-components
here
specified.
Three of
them,
in
fact,
are extensions and the
remaining
three are
shearing
strains. He has
worked out a
symmetrical system
of
strain-components
which would be the extensions of lines
parallel
to the
edges
of a tetrahedron. See
Edinburgh,
Froc.
Boy.
Soc,
vol. 24
(1902),
and
Phil.
Mag. (Ser. 6),
vol. 3
(1902), pp.
96 and 444.
10,
11]
BY MEANS OF EXTENSION
41
axis of X slide in the direction of the axis of
y,
while in the other these axes
are
interchanged.
The strain denoted
by e^
will be called the
"
shearing
4strain
corresponding
with the directions of the axes of x and
y."
The
change
of volume of
any
small
portion
of the
body
can be
expressed
in terms of the
components
of strain. The ratio of
corresponding very
small
volumes in the strained and unstrained states is
expressed by
the functional
determinant
du
42 GEOMETRICAL CHARACTER
[CH.
I
or
minimum,
are the
principal
axes of the
quadrics (4).
These axes are
therefore the
principal
axes of the strain
(Article 7),
and the extensions
in the directions of these axes are the
"
principal
extensions." When the
quadrics
are referred to their
principal
axes,
the left-hand member of
(4)
takes the form
e,X'-\-e,Y'
+
e,Z%
wherein the coefficients
d, ej, 63
are the values of the
principal
extensions.
We now see
that,
in order to
specify completely
a state of
strain,
we
require
to know the directions of the
principal
axes of the
strain,
and the
magnitudes
of the
principal
extensions at each
point
of the
body.
With
the
point
we
may
associate a certain
quadric
surface which enables us to
express
the strain at the
point.
The directions of the
principal
axes of the strain are determined as follows:let
I, m,
n
be the direction cosines of one of these
axes,
then we have
I
~
rti
~
n
'
and,
if e is written for either of these three
quantities,
the three
possible
values of e are
the roots of the
equation
^xx
~
^
2 '^xy 2
^xz
^xy ^yy
~
^
2^v*
^^
"

these roots are

real,
and
they
are the values of the
principal
extensions
e^, e^, e^.
12. Transformation of the
components
of strain.
The same state
<^f
strain
may
be
specified by
means of its
components
referred to
any system
of
rectangular
axes
;
and the
components
referred to
any
one
system
must therefore be determinate when the
components
referred
to some other
system,
and the relative situation of the two
systems,
are
known. The determination can be made at once
by using
the
property
of
the strain
quadric,
viz. that the
reciprocal
of the
square
of the radius vector
in
any
direction is
proportional
to the extension of a line in that direction.
We shall take the coordinates of a
poiut
referred to the first
system
of axes
to
be,
as
before, x,
y,
z,
and those of the same
point
referred to the second
system
of axes to be
x',
y',
z
,
and we shall
suppose
the second
system
to be
connected with the first
by
the
orthogonal
scheme
11-1'^]
OF
COMPONENTS OF STRAIN
43
operation
of
rotation*. We shall write
e^,;
e,.,,,
e,,.
e^,, e,^.
e,^
for the
components
of strain referred to the second
system.
The relative
coordinates of
points
in the
neighbourhood
of a
given point
may
be denoted
by
x, y,
z in the first
system
and
x", y',
2' in the second
system.
These
quantities
are
transformed
by
the same
substitutions as
X,
y,
z and x'
,
y',
z'.
When the form
exz^''
+
eyy'
+
ea2= +
Byzyz
-t-
ei^zx.
+
e^xy
is transformed
by
the above
substitution,
it becomes
ea:':t'X''
+
gyy
y
+
e//z'2
+
gyvy'z'
+
e^^z'it +
e^^xfj'.
It follows that
e^'x'
=
exxli
+
eyyttii'
+
e^n^'
+
ey^m^n^
+
e^^?,/,
+
exyhnh
,
gyv
=
2eia;4;3
+
ieyyTHim,
+
'2.6^^71^713
+
ey2{nii7i3
+
m^Ti^
\
(H)
+
e^x
{nA
+
n-j4)
+
e^y (1^7)1,
+
/jj/ij),
These are the formulae of
transformation of
strain-components.
13. Additional methods and results.
(a)
The fomiulaj
(] 1) might
have been inferred from the
interpretation
of
e^.^-
as the
extension of a Hnear element
parallel
to the axis of
x',
and of
e,.,,
as the cosine of the
angle
between the
positions
after strain of the linear elements which before strain are
parallel
to
the axes of
y'
and /.
(6)
The formulae
(11) might
also have been obtained
by introducing
the
displacement
{u', 1/, w')
referred to the axes of
{xf, y', /),
and
forming
hi'jdx'
The
displacement being
a
vector, u, v,
w are
cogredieut
with
x, y, 2,
and we have for
example
This method
may
be
applied
to the transformation of
sj^, m,
or,.
We should find for
example
n7j:.
=
?in7a;
+
m,cr,
+
,ir,,
(12)
and we
might
hence infer the vectorial character of
(oj^, ar, ar,).
The same inference
might
be drawn from the
interpretation
of
nr^,
07^,
w,
as
components
of rotation.
(c) According
to a well known theorem t
concerning
the transformation of
quadratic
expressions,
the
following quantities
are invariant in
respect
of transformations from one
set of
rectangular
axes to another :
j,.^
+
fezx+'':TO-i(^+''
+
V).
[
(13)
fxxyyez,
+
l{ey,!xexy
-
ixxey^-^yyej
-e^^).
I
The first of these invariants is the
expression
for the cubical dilatation.

This restriction makes no difference to the relations between the

oomponenta
of strain
referred to the two
systems.
It affects the
components
of rotation
w,., TS, W,.
t Salmon,
Oeometrij of
three
dimensions,
4th
ed., Dublin, 1882, p.
66.
44 INVARIANTS OF A STRAIN
[ClI.
I
(d)
It
may
be shown
directly
that the
following quantities
are invariants :

(i) arj.2
+
OT/+ar,2^
(ii) e:cz'!^/+ei,'aj'^
+
ew,^
+
ey,tiix!T,
+
e^7iTj^x
+
e:ci/^x'!^!/'<
and the direct verification
may
serve as an exercise for the student. These invariants
could be inferred from the fact that
^^, m^, ar,
are
cogredient
with
.r, y,
z.
(e)
It
may
be shown also that the
following quantities
are invariants* :
..... lew dv div dv\ /9m dw du dto\ /dv du dv du\
^'"^
V^
8i
~

^J
*
Vai
8S
~
8^-
J
*
\di dy 'd^dij'
(iv) e:
+
,2
+
e2
+
i(e,2
+
e^8+ V)
+
2
(nr/
+
tir,,s+ar,2).
(/)
It
maybe
shown
+
also
that,
in the notation of Article
7,
the invariant
(iv)
is
equal
to
(x2+y2+z2)rfxc?yrfz
///'
where the
integrations
are taken
through
a
very
small
sphere
with its centre at the
point
(^, y, )
(ff)
The
following
result is of some
importance
t
:

If the strain can be

expressed by
shears
e,
e^, only,
the
remaining components being zero,
then the strain is a
shearing
strain
- ;
and the
magnitude
of this
shear,
and the direction of the axis x' in the
plane
of
X, y,
are to be found from
e
and
e,, by treating
these
quantities
as the
projections
of a
vector on the axes of .v and
y.
14.
Types
of strain.
(a) Uniform
dilatation.
When the strain
quadric
is a
sphere,
the
principal
axes of the strain are
indeterminate,
and the extension
(or contraction)
of all linear elements
Issuing
from a
point
is the
same;
or we have
where A is the cubical
dilatation,
and the axes of
a;, y,
z are
any
three
orthogonal
lines.
In this case the linear extension in
any
direction is one-third of the cubical dilatation

a result which does not hold in

general.
(6) Simple
exteTision.
We
may exemplify
the use of the methods and formulsa of Article 12
by finding
the
components,
referred to the axes of
x, y,
z,
of a strain which is a
simple
extension,
of
amoimt
e, parallel
to the direction
(I, m, n).
If this direction were that of the axis of x'
the form
(4)
would he ex"^
;
and we have therefore
A
simple
extension is
accordingly equivalent
to a strain
specified by
these six com-
ponents.
It has been
proposed
to call
any
kind of
quantity,
related to
directions,
which is
equivalent
to
components
in the same
way
as a
simple
extension,
a tensor.
Any
strain
in,
*
The invariant
(iji)
will be useful in a
subsequent investigation (Chapter VII.).
+
E.
Betti,
/; Niiovo CimeiUo
(Ser. 2),
t. 7
(1872).
X
Cf.
Chapter
XIV.
infra.

W.
Voigt, Gottingen
Nachr.
(1900), p.
117. Cf. M. Abraham in
Eiicy.
d. math. Wins. Bd.
4,
Art. 14.
13,
14.]
TYPES OF STKAIN
46
as we have
already seen, equivalent
to three
simple
extensions
jMirallel
to the
principal
axes of the strain. It has been
proposed
to call
any
kind of
(luantity,
related to
directions,
which is
equivalent
to
components
in the same
way
as a
strain,
a
tetuor-triad. The
discussion in Articles 12 and 13
(6) brings
out
clearly
the
distinction between tensors and
vectors.
((')
Shearing
strain.
The strain denoted
by ^
is called "the
shearing
strain
corresixinding
with the direc-
tions of the axes of x and
./."
We have
already
observed that it is
equal
to the cosine of
the
angle,
in the strained
state,
between two linear elements
which,
in the unstrained
state,
are
parallel
to these
axe.s,
and that it is
equivalent
to two
simple shears,
consisting
of the relative
sliding, parallel
to each of these
directions,
of
planes
at
right angles
to the
other. The
"
shearing
strain
"
is measured
by
the sum of the two
simple
shears and is inde-
pendent
of their ratio. The
change
in the
length
of
any
line and the
change
in the
angle
between
any
two lines
depend upon
the sum of the two
simple
shears and not on the ratio
of their amounts.
The
components
of a
strain,
which is a
shearing
strain
corresjionding
with the direc-
tions of the axes of xf and
i/,
are
given by
the
equations
where s is the amount of the
shearing
strain. The strain involves no cubical dilatation.
If we take the axes of x' and
y'
to be in the
piano
of
x, y,
and
suppose
that the axes of
X, y,
z are
parallel
to the
principal
axes of the
strain,
we find that
e vanishes,
or there is
no extension at
right angles
to the
plane
of the two directions concerned. In this case we
have the form
s:dy' equivalent
to the form
e^x^-^-e^^y"^.
It follows that
e^=
-e= \t,
and that the
principal
axes of the strain bisect the
angles
between- the two directions
concerned. In other words
equal
extension and contraction of two linear elements at
right angles
to each other are
equivalent
to
shearing strain,
which is
numerically equal
to
twice the extension or
contraction,
and
corresponds
with directions
bisecting
the
angles
between the elements.
We
may enquire
how to choose two directions so that the
shearing
strain
corresponding
with them
may
be as
great
as
possible.
It
may
be shown that the
greatest shearing
strain
is
equal
to the dift'erence between the
algebraically greatest
and least
principal extensions,
and that the
corresponding
directions bisect the
angles
between those
principal
axes of
the strain for which the extensions are the maximum and minimum extensions*.
{d)
Plane strain.
A more
general type,
which includes
.simple
extension and
shearing
strain as
particular
cases,
is obtaineil
by assuming
that one of the
principal
extensions is zero. If the corre-
sponding principal
axis is the axis of
i,
the strain
quadric
becomes a
cylinder, standing
on a conic in the
plane
of
x, y,
which
may
be called the strain conic
;
and its
equation
can
be written
ex2-(-eyy^-)-iXy
=
const.
;
so that the
shearing
strains
e^,
and
e^ vanish,
as well as the extension
e.
In the
jiarticular
case of
simple extension,
the conic consists of two
parallel
lines;
in the case of
shearing strain,
it is a
rectangular hyperbola.
If it is a
circle,
there is extension or con-
traction,
of the same
amount,
of all linear elements
issuing
from the
point {x, y, i)
in
directions at
right angles
to the axis of z.
*
The theorem here stated is due to W.
Hopkins, Cambridge
Phil. Soc.
Tram.,
vol. 8
(1849).
46 THEOREMS CONCERNING
[CH.
I
The relative
displacement corresponding
with
plane
strain is
parallel
to the
plane
of the
strain;
or we have w=
const.,
while u and v are functions of x and
y only.
The axis of the resultiint rotation is normal to the
plane
of the strain. The cubical
dilatation, A,
and the
rotation, cr,
are connected with the
displacement by
the
equations
3m 3i>
_dv
du
dx
dy'
dx
dy'
We can have states of
plane
strain for which both A and or vanish
;
the strain is
pure
shear,
i.e.
shearing
strain combined with such a rotation that the
principal
axes of the
strain retain their
primitive
directions. In
any
such state the
displacement components
V,
u are
conjugate
functions of x and
y,
or
v+m
is a function of the
complex
variable x
+
ly.
15. Relations
connecting
the
dilatation,
the rotation and the dis-
placement.
The cubical dilatation A is connected with the
displacement {u, v, w) by
the
equation
3tt Sv dv>
dx
dy
dz
'
A scalar
quantity
derived from a vector
by
means of this formula is described as the
divergence
of the vector. We write
A
=
div.
{u, V, w) (14)
This relation is
independent
of
coordinates,
and
may
be
expressed
as follows :

Let
any
closed surface S be drawn in the field of the
vector,
and let N denote the
projection
of the
vector on the normal drawn outwards at
any point
on
S,
also let dr denote
any
element of
volume within
S,
then
1 1
NdS=
1 1 1
Adr, (15)
the
integration
on the
right-hand
side
being
taken
through
the volume within
S,
and that
on the left
being
taken over the surface S*.
The rotation
(or,, nr,,
isr,)
is connected with the
displacement (u, v, w) by
the
equations
dw dv 9m dw dv du
A vector
quantity
derived from another vector
by
the
process
here indicated is described
as the curl of the other vector. We write
2(ar;t, oTy, ar2)
=
curl
(m, v, w) (16)
This relation is
independent
of
coordinatesf,
and
may
be
expressed
as follows :

Let
any
closed curve be drawn in the field of the
vector,
and let
any
surface S be described
so as to have the curve s for an
edge
;
let T be the resolved
part
of the vector
(m, v, w)
along
the
tangent
at
any point
of
,
and let
2zii,
be the
projection
of the vector 2
(orj,, ur,, cjj)
on the normal at
any point
of
S,
then
JTds=ff
2-sXydS, (17)
*
The result is a
particular
case of the theorem known as
'
Green's tlieorem.' See
Kncy.
d.
math. Wiss. ii. A
2,
Nos. 4547.
+ It is assumed that the axes of
x, y,
z form a
right-handed system.
If a transformation to
a left-handed
system
is admitted a convention must be made as to the
sign
of the curl of a vector.
14-16]
DILATATION AND ROTATION
47
the
integration
on the
right being
taken over the surface
S,
and that on the left
being
taken
along
the curve s*.
16. Resolution of
any
strain into dilatation and
shearing
strains.
When the strain involves no cubical dilatation the invariant
,
+
!/!/
+

vanishes,
and it is
possible
to choose
rectangular
axes of
x',
y',
z so that the
form
exx?^
+
eyy]f-
+
e^z^
+
ey^yz
+
ej^zx
+
e^xy
is transformed into the form
e^iV'z'
+
ei^z'ad
+
eji^x'y',
in which there are no terms in
x"',
y'^,
z''. The strain is then
equivalent
to
shearing
strains
corresponding
with the
pairs
of directions
(y'.z), {z',x'), (x'.y').
When the strain involves cubical dilatation the
displacement
can be
analysed
into two constituent
displacements,
in such a
way
that the cubical
dilatation
corresponding
with one of them is
zero;
the strains derived from
this constituent are
shearing
strains
onlj',
when the axes of reference are
chosen
suitably.
The
displacement
which
gives
rise to the cubical dilatation
is the
gradientf
of a scalar
potential (</>),
and the
remaining part
of the
displacement
is the curl of a vector
potential
(F, 0,
H),
of which the
divergence
vanishes. To
prove
this statement we have to show tliat
any
vector
(u, v,
w)
can be
expressed
in the form
(m, t),w)
=
gradient
of
<^
+
curl
{F,
0,
H), (18)
involving
the three
equations
of the
type
M=5^
+
-5---^
, (19)
dx
oy
(iz
in which
F, G,
H
satisfy
the
equation
^l+f^^f^O
(20)
dx
oy
oz
In the case of
displacement
in a
body
this resolution must be valid at all
points
within the surface
bounding
the
body.
There are
many
different
ways
of
effecting
this resolution of
(m, v,
ivY^.
*
The result is
generally
attributed to Stokes. Cf.
Ency.
d. math. Wiss. ii. A
2,
No. 46. It
implies
that there is a certain relation between the sense in which the
integration along
dt is
taken and that in which the normal v is drawn. This relation is the same as the relation of
rotation to translation in a
right-handed
screw.
fd<t> d<t> d<b\
t
The
gradient
of
<p
is the vector
( 5^ .
^
.
g;
I

+
See, e.g.,
E.
Betti,
II Nuovo Cimento
(Ser. 2),
t. 7
(1872),
or P. Dnhera,
J. de Math.
(Lioiwille), (S6r.
5),
t. 6
(1900).
The resolution was first effected
by
Stokes in his memou: on
Diffraction.
(See
Introduction,
footnote
80.)
48
DILATATION AND ROTATION
[CH.
I
We observe that if it is effected the dilatation and rotation will be
expressed
in the forms
A
=
V<^,
2Br^
=
-V^, 2'sr,,
=
-V2G', 2t!r^
=
-V^/r,
(21)
the last three
holding good
because
dFjdx
+
dGjdy
+
dH/dz
=
0. Now solu-
tions of
(21)
can be written in the forms
^
=
-^///^'^"'^2''^'''
^=^///t-^^''^2/W
<22)
where r is the distance between the
point (x', y\ z)
and the
point {x, y, z)
at which
<^, F,...
are
estimated,
A' and
{rsj, ra-j,', in-/)
are the values of A
and
(tr-t, Iffy, isTj)
at the
point {x', y', z),
and the
integration
extends
through
the
body.
But the solutions
given
in
(22)
do not
always satisfy
the
equation
div
{F,
G, H)
=
0. A case in which
they
do
satisfy
this
equation
is
presented
when the
body
extends
indefinitely
in all
directions,
and the
displacements
at infinite distances tend to zero in the order i'"' at least. To see this we
take the
body
to be bounded
by
a surface
S,
and write the first of
equations
(22),
viz.
in the
equivalent
form
<f>
=
T-~
1
1
~
{"'
^^^
(* ")
+ '"' <^^
(y- ")
+
'*'' c<>s
(z,
v)}
dS
^
1
[[[{
,37-' ,9r-i
,ar-n , ,, ,. ,
and omit the
surface-integral
when S is
infinitely
distant. In the same
case we
may put
^=
"
^
//] ("''
w
~
"'
^l^')
^'^y'^"'
dy'
or,
since
d7'~^/dx'
=

dr~^/dx,
... we have
with similar forms for G and H. From these forms it is clear that
div
(F, G,H)
=
0.
The
expressions
into which the
right-hand
members of
equations (22)
have been transformed in the
special
case are
possible
forms for
<f>,
F, G,
H
in
every
case,
that is tio
say
one mode of resolution is
always given by
the
equations
'
16,
17]
CONDITIONS OF
COMPATIBILITY
*=-'J//("'''-''"+l;>'''H dy
49
(23)
where the
integration
extends
throughout
the
body;
for it is clear that
these make div
{F,
G,
H)
=
and also make
d<l>
dH dG 1
, f[fii ,,,,,,
17. Identical relations between
components
of strain*.
The values of the
components
of strain
Bxx,

6,,^,
... as functions of
x,
y,
z
cannot be
given arbitrarily; they
must be
subject
to such relations as will
secure that there shall be functions
m, v, w,
which are connected with them
by
the six
equations
9m dw dv ....
f=^-r
+
5^.
(24)
^xx

dx
'
dy^dz'
The relations in
question may
be obtained
by taking
account of the three
equations
n
dw dv
ay
dz
for all the differential coeflBcLents of
u, v,
w can be
expressed
in terms of
...
e,
yzt
We have in fact three
pairs
of
equations
such as
du
=
hex
dv

2^X1/
+
'
and the conditions that these
may
be
compatible
with the three
equations
such as
dtijdx
=
exx,
are nine
equations
of the
type
dy
1
dexy
''
2 dx dx
and these
equations express
the first differential coefficients of
vr^,
vr^,
Wj
iu
terms of those of
Bxx,
e,jz,
.... If we write down for
example
the three
*
These relations were
given by
Saint-Venant in his edition of Navier's
Lerong, Appendix
in.
The
proof
there indicated was
developed by
Kirchhoff, Mechanik, Vorlesung
27. The
proof
in the
text is due to
Beltrami, Paris,
C.
R.,
t. 108
(1889),
cf.
Koenigs,
Lerons dt
Cinfmatique,
Paris
1897, p.
411.
L. E.
50
DISPLACEMENT
[CH.
I
equations
that contain
nr^
we can see at once how to obtain the conditions
that
they may
be
compatible.
These three
equations
are

9^a,
_
de^i
_
dexy
dx
dy
dz
2

=
^-' -
2
^
dy dy
dz
'
2
^
=
2
- ^
dz
dy
dz
'
and from the set of nine
equations
of this
type
we can eliminate
Wx,
ra-j,,
ot,
and obtain the six identical relations between the
components
of strain.
They
are
L..(25)
S^^w
9^ez2
d\z
dz^
dy^ dydz'
17-19]
DETERMINED BY STRAIN
51
would not
express
the strain in the
body correctly,
and the terms of
(27)
that
contain
p, q,
r would not
represent
a
displacement possible
in a
rigid body.
Bearing
this restriction in
mind,
we conclude
that,
if the six
components
of
strain are
given,
the
corresponding displacement
is
arbitrary
to the extent of
an additional
displacement
of the
type expressed by (27); but,
if we
impose
six
independent conditions,
such as
that,
at the
origin,
the
displacement
(w,
V,
w)
and the rotation
{ctx, ^y, ^z)
vanish,
or
again
that,
at the same
point
.
=
0,=0.;
=
0,
|=0,g
=
0,
1
=
(28)
the
expression
for the
displacement
with
given
strains will be
unique.
The
particular
set of
equations (28)
indicate that one
point
of the
body (the
origin),
one linear element of the
body (that
along
the axis of z
issuing
from the
origin)
and one
plane-element
of the
body (that
in the
plane
of
z,
x
containing
the
origin)
retain their
positions
after the strain. It is mani-
festly possible,
after
straining
a
body
in
any way,
to
bring
it back
by
trans-
lation and rotation so that a
given point,
a
given
linear element
through
the
point
and a
given plane-element through
the line shall recover their
primitive positions.
19. Curvilinear
orthogonal
coordinates*.
For
many problems
it is convenient to use
systems
of curvilinear co-
ordinates instead of the
ordinary
Cartesian coordinates. These
may
be
introduced as follows :

Let
f{x, y, z)
=
a,
some
constant,
be the
equation
of a surface. If a is allowed to
vary
we obtain a
family
of surfaces. In
general
one surface of the
family
will
pass through
a chosen
point,
and a
neighbouring point
will in
general
lie on a
neighbouring
surface of the
family,
so that a is a function of
x,
y,
z, viz.,
the function denoted
by/.
If
a -f- da is the
parameter
of that surface of the
family
which
passes through
{x
+
dx,
y
+
dy,
z
+
dz),
we have
da=^:^dx
+
^dy
+
^dz
=
;^dx
+
^dy-\-i^dz.
dx
dy
dz ox
dy
^
dz
If we have three
independent
families of surfaces
given by
the
equations
f,(x,y,z)
=
a,
f(x,y,z)
=
^,
f,(x,y,z)=y,
so that in
general
one surface of each
family passes through
a chosen
point,
then a
point may
be determined
by
the values of
a, /3, 7
which
belong
to the surfaces that
pa.ss through
itf,
and a
neighbouring point
will be
*
The
theory
is due to Lamfi. See his
Lemons
sur les coordonnget
airviligneg,
Paris,
1859.
t
The determination of tlie
point may
not be free from
ambiguity, e.g.,
in
elliptic
coordinates,
&n
ellipsoid
and two confocal
hyperboloids pass through any point,
and
they
meet in seven other
points.
The
ambiguity
is removed if the
region
of
space
considered is
suitably
limited, e.g.,
in
the case of
elliptic
coordinates,
if it is an octant bounded
by principal planes.
42
52
STRAIN REFERRED TO
[C'H.
I
determined
by
the
neighbouring
values a + da, /3
+
rf/3,
7
+
^7.
Such
quan-
tities as
a, /3, 7
are called
"
curvilinear coordinates
"
of the
point.
The most convenient
systems
of curvilinear coordinates for
applications
to the
theory
of
Elasticity
are determined
by
families of surfaces which cut
each other
everywhere
at
right angles.
In such a case we have a
triply-
orthogonal family
of surfaces.
It is well known that there exists an infinite
number of sets of such
surfaces, and,
according
to a celebrated theorem due
to
Dupin,
the line of intersection
of two surfaces
belonging
to different
families of such a set is a line of curvature on each*. In what follows we
shall take
a, /8, 7
to be the
parameters
of such a set of
surfaces,
so that the
following
relations hold :
dx dx
dy dy
dz dz
'
dy
da
dy
da
dy
da
_
dx dx
dy dy
dz dz
'
8a
a^ 8aa^
aaa/3^^'
dx dx
dy dy
dz dz
The
length
of the
normal, drii,
to a surface of the
family
a
intercepted
between the surfaces a and a
+ da is determined
by
the observation that
the direction-cosines of the normal to a at the
point
{x, y, z)
are
1 8a 1 8a 1 8a
,j,>
h,di'
h,d^'
h,dz'
^ '
where
Aj
is
expressed by
the first of
equations
(31)
below.
For,
by projecting
the line
joining
two
neighbouring points
on the normal to
o,
we obtain the
equation
dn.
=
U'ldx
+
'idy^%dz)
=
f
(30)
hi
^dx
dy
dz
) Ih
In like manner the elements
dwa, dn^
of the normals to
/S
and
7
are
cZ/S/A^
and
dyjh-i,
where
, ,
/8a\^ /8a\-''
^=(8-Jn8-J
+
dy]
\dz
1)
\dy.
dx)
\dyj
\dz
y
(31)
;T-a
\dxj
\dyj
\dzj
The distance between two
neighbouring points being
{dn^
+
dn,/
+
dn^)^,
we
have the
expression
for the
"
line-element," ds,
i.e. the distance between the
points (a, (8, 7)
and
(a
+
da, /3
-t-
d^, 7
-I-
dy),
in the form
{dsf
=
{dajKY
+
(c^/S//'.)'-^
+
{dyjKf .(32)
In
general
hi, //j, h^
are
regarded
as functions of
a, /3, 7.
*
Salmon,
Geometry of
three
dimeitsiom,
4th
ed., p.
209.
19,
20] CURVILINEAR
COORDINATES
68
20.
Components
of strain
referred to
curvilinear
orthogonal
co-
ordinates*.
The
length
in the
unstrained state of the line
joining
the
points (a, /9, 7)
and
(a
+
da, /3
+
rf^,
y+dy)
is
given by (32);
we seek the
length
in the strained
state of the line
joining
the same
pair
of
particles.
Let
.,
u^, u^
be the
projections
of the
displacement
of
any particle
on the normals to the surfaces
a,
^, 7
that
pass
through
its
position
in the
unstrained state. When the
displacement
is small the coordinates of the
point occupied by
a
particle
are
changed
from
a, ^, 7
to o +
A.m., /3
+
huf, .y
+
fi^Uy.
The
a-coordinate
(a
+
da)
of a
neighbouring point
is
changed
into
a
+ dcL+
|/,,M,
+ da
~
{h,u^)
+
d^^
(h,iu)
+
dy~ (A.m.)I ,
and similar
changes
are made in the
/3-
and
7-coordinates (/S
+
d^S
and
7
+
^7).
Again,
the values of
A,,
... at a
displaced particle
differ from those
at its
undisplaced position.
For
example, 1/A,
is
changed
into
It follows that
da/hi
must be
replaced by
da
|l
+
^
(h,u.)
^
+
d^~
{Ku.)
+
dy^
(k.u.)
and,
when
products
of
quantities
of the order
?,
are
neglected,
this is
J
+ da
f5
+
h.u,
^(i_)
+
h,u,
^i
(
1
)
J
+d^~~iKu,)
+
dy^^~
{Ku.) (33)
Similar
changes
must be made in
d^jh^
and
dyjh,.
The
length
of the line
joining
two
particles
in the strained state is found
by forming
the
square
root of the sum of the
squares
of the three
expressions
of the
type (33).
Let ds be the
length
of the linear element in the unstrained
state,
and let
/, m,
n be the direction-cosines of it referred to the normals to the surfaces
a, /3,
7
at a
point,
so that
da/h^
=
lds,
.... Also let
ds{l +e)
be the
length
of the
corresponding
linear element in the strained state. Then e is
given
by
the
equation
{l+ey

The method here

given
is due to
Borchardt,
J.
f.
Math.
{Crelle),
Bd. 76
(1873).
Another
method is
given
in the
'
Note on
applications
of
moving
axes
'
at the end of this book.
54 STRAIN REFERRED TO
[CH.
I
Neglecting squares
and
products
of
u.^,
u^, Uy,
we
may
write the result in
the form
e
=
eaJ'' +
e^^ni'
+
Cyyii-
+
e^yVin
+
e^^nl
+
e,3
Im,
(35)
in which
e..
=
A,
|f
+
AAu,|(-^l)
+
/.3V,^(i),
epy--
.(36)
The
quantities
e^a,

e^y
^^"6 *^^ ^i^
components
of strain refen'ed to
the
orthogonal
coordinates. In fact
e^
is the extension of a linear element
which,
in the unstrained
state,
lies
along
the normal to the surface a
;
and
e^y
is the cosine of the
angle
between the linear elements
which,
in the
unstrained
state,
lie
along
the normals to the surfaces
/3
and
7.
21. Dilatation and Rotation referred to curvilinear
orthogonal
coordinates.
The results of Art. 15 can be utilized to
express
the cubical dilatation
A,
and the
component
rotations
nra, w/j,
Wy
about the normals to the three
surfaces,
in terms of the
components
Ua, wpi My
of the
displacement.
To obtain the
expression
for A we form the surface
integral
of the normal
component
of
the
displacement*
over the surface of an element of the
body
bonided
by
the three
pairs
of
surfaces
(a, a-'rda), (3, /3
+
rf/3), (y, y
+
dy),
the normal
being
drawn
away
from the interior
of the element. The contributions of the faces of the element can be
put
down in such
forms as
contribution of
d^dry
, ,
d^dy
_
,
d
o
lit
Oa
'
v^/H d'
and,
on
adding
the six
contributions,
we obtain
this must be the .same as
Adad^dylhih^h^.
We therefore have
-vaSG5->*(S;)*^05;))
<="
This result is the same as would be found
by adding
the
expreasious
for
t'ao, ejS/S, (^yv
in
(36).
*
This method is due to Lord Kelvin.
(Sir
W.
Thomson,
Math, and
Phys. Papers,
Vol.
1,
p.
25. The date of the
investigation
is
1843.)
20,
21]
CURVILINEAR COORDINATES
55
To obtain the
expression
for
2^^
we form the line
integral
of the
tangential component
of the
displacement along
the
edge
of the element in the face
y
+
dy.
The contributions
of the four
portions
of the
edge
can bo written down
by help
of
Fig.
3 as follows :

contribution of RP^
da
yj'=-.|_rf^|(,g,
QR=-u,f,
"
^'^'=
^^^t^<(:-ti-
On
adding
these
contributions,
we obtain
''<-km-
This must be the same as
H-urydad^jh^h^,
and we have thus an
expression
for
OTy
which
is
given
in the third of
equations (38)
;
the other
equations
of this set can be obtained in
the same
way.
The formula;* are
The formula3
(38),
as also
(36)
and
(37),
are due to Lame. The method here used to
obtain
(38),
and used also in a
slightly
more
analytical
form
by Cesilro,
Introduzione
alia Uoria
matematica della
Elasticity,
(Turin,
1894), p. 193,
is familiar in
Electrodynamics.
Cf. H.
Lamb,
Phil. Tram.
Roy. Soc,
vol. 178
(1888), p.
150,
or J. .T.
Thomson,
lieceiit Researchen
in
EUclricity
and
Magnetism, Oxford, 1893, p.
367. The
underlying physical
notion
is,
of course,
identical
with the relation of
'
circulation
'
to
'
vortex
strength
'
brought
to
light
in Lord Kelvin's memoir
'
On Vortex
Motion,' Edinburgh, Roy.
Soc. Trail.'!.,
vol. 25
(18G9).
56 CURVILINEAK COORDINATES
[CH.
I
22.
Cylindrical
and
polar
coordinates.
In the ciise of
cylindncal
coordinates
r, 6,
z we have the
line-element
{{drf
+
r''{def-{-{dzf}\
and the
displacements it,, ug, u^.
The
general
formulae take the
following
forms :

(1)
for the strains
3Mr
1
one Ur SUj
-l?!^a.?^
_9Mr
3mj
_due
tie
1
du^
(2)
for the cubical dilatation
A
=
(3)
for the
components
of rotation
^ 18,
.
,
1
8m
, dui
_1
8mj 39
r or^
'
r OS
In the case ot
polar
coordinates
r, 6, (j),
we have the line-element
{{dry
+
r^{dff)^
+
r^ sin^ 6
{d<l>ff,
and the
displacements m,, lie, u^.
The
general
formula; take the
following
forms :
(1)
for the strains
dur Idue
^
Ur
I
du4,
ue .. Ur
cr
'"'
r do r
^^
rsin6
d<j>
r r
'
ee^
^y^^
"*<=*V 'sin^a.^'
'^--rsiK^
8^
+
"S^
~
7
' ''"-
di^ "f ^rTO
'
(2)
for the cubical dilatation
(.3)
for the
components
of rotation
1 (du. 8
,
.
^,1
^'^''=.-^i^i8>-87(*'"^)}'
The verification of these formulae
may
serve as exercises for the student.
APPENDIX TO CHAPTER I.
GENERAL THEORY OF STRAIN.
23. The
preceding part
of this
Chapter
contains all the
results,
relating
to
strains,
which are of
importance
in the mathematical
theory
of
Elasticity,
as at
present developed.
The discussion of strains that
correspond
with
displacements
in
general,
as
opposed
to small
displacements,
is an interest-
ing
branch of
kinematics;
and some account of it will now be
given*.
It
may
be
premised
that the
developments
here described will not be
required
in the remainder of this treatise.
It is
customary,
in recent books on
Kinematics,
to base the
theory
of
strains in
general
on the
result,
stated in Article
7,
that the strain about a
point
is
sensibly homogeneous,
and to
develop
the
theory
of finite strain in
the case of
homogeneous
strain
only.
Frcmi the
point
of view of a
rigorous
analysis,
it
appears
to be desirable to establish the
theory
of strains in
general
on an
independent
basis. We shall
begin
with an account of the
theory
of the strain
corresponding
with
any displacement,
and shall after-
wards
investigate homogeneous
strain in some detail.
24. Strain
corresponding
with
any displacement.
We consider the effect of the
displacement
on
aggregates
of
particles
forming given
curves in the unstrained state.
Any
chosen
particle
occupies,
in the unstrained
state,
a
point {x, y, z).
The same
particle occupies,
in the
strained
state,
a
point
{x
+
u,
y
+
v,
z
+
w).
The
particles
which lie on a
given
curve in the first state lie in
general
on a different curve in the second
state. If ds is the differential element of arc of a curve in the first
state,
the direction-cosines of the
tangent
to this curve at
any point
are
^, ^y, ^f-.
If
ds,
is the differential element of arc of the
corresponding
ds ds ds
*
Eeference
may
be made to
Cauchy,
Exercices de
matMmatiques,
AnnSe
1827,
the Article
'Sur la coudensation et la dilatation des
corps
solides';
Green's memoir on the reflexion of
light quoted
in the Introduction
(footnote 42)
;
Saint-Venant,
'
M^moire sur
I'^quilibre
des
corps
sohdes.quand
les
d6plaeeraents...ne
sont
pas
trSs
petits,'
Paris,
C. R.,t-
2*
(1847)
;
Kelvin and
Tait,
Nat.
Phil,
Part i.
pp.
115144;
Todhunter and Pearson, History,
vol.
1,
Articles 1619
1022
;
J.
Hadamard, Le(:ons
mr la
propagation
des ondes,
Paris 1903, Chapter
vi.
58 GENERAL THEORY
[CH.
I. APP.
curve in the second
state,
the direction-cosines of the
tangent
to this
ci
(a;
+
m) d(y
+
v) d{z+w)
. ,
curve are

^ , ^ ,
-^-v
.
Herein,
for
example,
rfs, dsi dsi
d
(x
+
m)
_
ds
(dx
dudx du
dy
du dz
+
m)
_
ds fax
Si dsi \ds dsi dsi \ds
'
dx ds
dy
ds dz
with similar formula} for the other two.
).
(1)
Let
I, m,
n be the direction-cosines of a line in the unstrained
state,
Ij, nil, i
the direction-cosines of the
corresponding
line in the strained
state, ds, ds^
the differential elements of arc of
corresponding
curves
having
these lines
respectively
as
tangents.
In the notation used above
. dx
~
ds'
24, 25]
OF STRAIN
59
The state of strain is
entirely
determined when we know the
lengths
in
the strained and unstrained states of
corresponding
lines*. The
<iuantity
ds . . .
j-'

1 is the extension of the linear element ds. This is determined by

the
as
formula
(3).
We observe that the extensions of linear elements
which,
in
the unstrained
state,
are
parallel
to the axes of coordinates are
respectively
V(l
+
26^)-], x/(l
+
2e^)-l, V(l+26)-l,
where the
positive
values of the
square
roots are taken. We thus obtain an
interpretation
of the
quantities
e^,
yy,
.
We shall
presently
obtain an
intei-pretation
of the
quantities 6j,j,
zx,
(xy,
in terms of the
angles,
in the
strained
state,
between linear
elements, which,
in the unstrained
state,
are
parallel
to the axes of coordinates. In the
meantime,
we observe that the
strain at
any point
is
entirely
determined
by
the six
quantities
exx,
fyy,
f,
eyz,
e^x,
fxy-
These
quantities
will be called the
components of
strain. The
quantities
exx,---
which were called
"components
of strain" in
previous
Articles are
sufficiently
exact
equivalents
of
xx,
- when the
squares
and
products
of such
quantities
as
dujdx
are
neglected.
25. Cubical Dilatation.
The ratio of a differential element of volume in the strained state to the
corresponding
differential element of volume in the unstrained state is
equal
to the functional determinant
d(x +u,
y
+
v,
z +
w)
or it is
=
(7)
60
GENERAL THEORY
[CH.
I. APP.
26.
Reciprocal
strain
ellipsoid.
The ratio
dsi
:
ds,
on which the extension of a linear element
issuing
from a
point depends,
is
expressed
in the formula
(3)
in terms of the
direction-cosines of the
element,
in the unstrained
state,
and the
components
of strain at the
point.
The formula shows
that,
for
any
direction,
the ratio
in
question
is
inversely proportional
to the central radius
vector,
in that
direction,
of an
ellipsoid
which is
given by
the
equation
(1
+
2e^^)
*-
+
(1
+
2yy) y-
+
(1
+
2e)
z- +
2eyz yz
+
2e^^
zx +-
26^^ xy
=
const.
(6)
This is the
reciprocal
strain
ellipsoid already
defined
(Article 6)
in the case
of
homogeneous
strains. Its axes are called the
principal
axes
of
the strain
;
they
are in the directions of those linear elements in the unstrained
state
which
undergo stationary
(maximum
or minimum or
minimax)
extension.
The extensions of linear elements in these directions are called the
principal
extensions, e^, e^, e,.
The values of 1 +
e^, l+e^,
I +
e^
are the
positive
square
roots of the three values of
k,
which
satisfy
the
equation
1
+
^Sxx

1^,
^xy>
^xz
xz >
Sj/2>
i
+
^^zz
~
The invariant relation of the
reciprocal
strain
ellipsoid
to the state of
strain
may
be utilized for the
purpose
of
transforming
the
components
of strain from one set of
rectangular
axes to
another,
in the same
way
as the
strain
quadric
was transformed in Article 12. It would thus
appear
that
the
quantities exx,---xy
are
components
of a "tensor triad." Three in-
variants would thus be
found,
viz. :
^xx
+
Syy
+
^zz,
^yy^zz
+
^zz^xx
+
^xx^yij
~
\ i^'yz
+
^'zx
+
^'xyh\
. .
.(g)
^xx^yy^a
"r
J (^yz^zx^xti ^xx^'yz ^yy^'zx ^zz^'xy)-
'
27.
Angle
between two curves altered
by
strain.
The effect of the strain on the
angle
between
any
two linear
elements,
issuing
from the
point (.r, y, z),
can be calculated. Let
I, m,
n and
I', ni,
n'
be the direction-cosines of the two lines in the unstrained
state,
and 6 the
angle
between them
;
let
l^, vh, n^
and
i/, ?/, ?</
be the direction-cosines
of
the
corresponding
lines in the strained
state,
and
6^
the
angle
between
them. From the formulae such as
(2)
we find
ds ds'
, ,
cos
^,
=
,
-^
{cos
-f
2
{exxll'
+
eyymm
+
e^^nn')
-H
ey^
(nin
+
m
n)
+
e^(nr
+
n'l)
+
exy{lm'
+
l'm)},
(9)
where
dsjds
and
ds/jds'
are the ratios of the
lengths,
after and before
strain,
of
corresponding
linear elements in the two directions.
26-28]
OF STRAIN
61
We observe
that,
if the two
given
directions are the
positive
directions of
the axes of
y
and
z,
the formida becomes
iyz
=
VRl
+
^e^x) (1
+
2yj,)}
cos
e
(10)
and we thus obtain an
interpretation
of the
quantity tyi.
Similar inter-
pretations
can be found for
e^^
and
e^y.
From the above formula it
appears
also
that,
if the axes of
x,
y,
z are
parallel
to the
principal
axes of the strain
at a
point,
linear
elements,
issuing
from the
point,
in the directions of these
axes continue to cut each other at
right angles
after the strain.
We
may
show
that,
in
general,
this is the
only
set of three
orthogonal
linear
elements,
issuing
from a
point,
which remain
orthogonal
after the
strain. For the condition that linear elements which cut at
right augles
in
the unstrained state should also cut at
right angles
in the strained state is
obtained
by putting
cos 6 and cos
0^
both
equal
to zero in
equation (9).
We
thus find the
equation
{(1
+
ie^x)
I
+
e^yin
+
e^zn]
I'
+
[e^yl
+
(I
+
2ey,,)
m
+
ey^n]
vi
+
{exzl
+
eyzin
+
(1
+
2)?i)
n'
=
0,
wherein W
+
mm'
+
nn
=
0. This
equation
shows that each of two such
linear
elements,
(besides being
at
right angles
to the
other),
is
parallel
to
the
plane
which is
conjugate
to the other with
respect
to the
reciprocal
strain
ellipsoid. Any
set of three such elements must
therefore,
(besides
being
at
right angles
to each
other),
be
parallel
to
conjugate
diameters
of this
ellipsoid.
The formulae so far obtained
may
be
interpreted
in the sense that a
small element of the
body,
which
has,
in the unstrained
state,
the
shape
and
orientation of the
reciprocal
strain
ellipsoid, corresponding
with that
point
which is at the centre of the
el'ement, will,
after
strain,
have the
shape
of a
sphere,
and that
any
set of
conjugate
diameters of the
ellipsoid
will become
three
orthogonal
diameters of the
sphere.
28. Strain
ellipsoid.
We
might express
the ratio
dsi
: ds in terms of the direction of the linear
element in the strained state instead of the unstrained. If we solved the
equations
of
type (2)
for
I, m,
n we should find that these are linear functions
of
?i, m,, Ml
with coefficients
containing ds^jds
as a
factor; and,
on
squaring
and
adding
and
replacing
I- + m- +
n-
by unity,
we should find an
equation
of the form
rJ'\ =
(uj,
+
b,Jih
+
c,n,y
+ {a,,l,
+
hvh
+
Coi)'
+
(O'Ji
+
h^,
+
c^ihY,
where
,,...
depend
only
on
^^
,
^,
...
g^-.
62
GENERAL THEORY
[CH.
I. APP.
The
ellipsoid represented by
the
equation
(a^x
+
b^y
+
Cizf
+
{a^x
+
h^y
+
c^zf
+
{a^x
+
h^y
+
c^zy
=
const,
would have the
property
that its central radius
vector,
in
any
direction,
is
proportional
to the ratio
dsi
: ds for the linear element
which,
in the strained
state,
lies
along
that direction. This
ellipsoid
is called the strain
ellipsoid.
The
lengths
of the
principal
axes of this
ellipsoid
and of the
reciprocal
strain
ellipsoid
are inverse to each
other,
so
that,
as
regards shape,
the
ellipsoids
are
reciprocal
to each
other;
but their
principal
axes are not in
general
in the same directions. In fact the
principal
axes of the strain
ellipsoid
are in the directions of those linear elements in the strained state
which have
undergone
stationary (maximum
or minimum or
minimax)
extension. The
simplest way
of
finding
these directions is to observe that
the
corresponding
linear elements in the unstrained state are
parallel
to the
principal
axes of the
strain,
so that their directions are known. The formulae
of
type (2) express
the
direction-cosines,
in the strained
state,
of
any
linear
element of which the
direction-cosines,
in the unstrained
state,
are
given.
The direction-cosines of the
principal
axes of the strain
ellipsoid
can thus be
found from these formulae.
29. Alteration of direction
by
the strain.
The
correspondence
of directions of linear elements in the strained and
unstrained states can be made clearer
by
reference to the
principal
axes of
the strain. When the axes of coordinates are
parallel
to the
principal
axes,
the
equation
of the
reciprocal
strain
ellipsoid
is of the form
(1
+
e.yx"
+
(1
+
e^yy^
+
(1
+
63)''^''
=
const.
where
ei, e^, e-j
are the
principal
extensions. In the formula
(9)
for the
cosine of the
angle
between the strained
positions
of two linear elements we
have to
put
l
+
2e^={l+e,y,
l
+
2eyy={l+e,y,
1
+
26= (1
+
e,)^
e^=
6:
=
6^^=
0.
Let the line
{I',
m', n')
of the formula
(9)
take
successively
the
positions
of the three
principal
axes,
and let the line
(/, m,
n)
be
any
chosen line in
the unstrained state.
We have to
equate ds'/ds/
in turn to
(1
+
e,)"', (1
+
e,)"',
(1
+
3)"',
and we
have to
put
for
dsjds^
the
expression
[(1
+
6,y
l^
+ (l+
eO'^m^
+
(1
+
e,)^?^^]-^.
The formula then
gives
the cosines of the
angles
which the
corresponding
linear element in the strained state makes with the
principal
axes of the
strain
ellipsoid. Denoting
these cosines
by
\,
fx, v,
we find
<\, (I, v)
=
[(1
+
e,y
i
+
(I
+
e,ym'
+
(1
+
e,yn^-i {(1
+
,)l,(\+ e,) m,
{l
+
,)n}.
(11)
28-31]
OF STRAIN
63
By solving
these for
I, m,
n we find
(I, m,
n)
=
-}
l+e,'l+e/l+e,j----^^^^
Here
I, m,
n are the direction-cosines of a line in the unstrained state
referred to the
principal
axes of the
strain,
and
X, fj.,
v are the direction-
cosines of the
corresponding
line in the strained state referred to the
principal
axes of the strain
ellipsoid.
The
operation
of
deriving
the second
of these directions from the first
may
therefore be made in two
steps.
The
first
step*
is the
operation
of
deriving
a set of direction-cosines
(\, fi, v)
from the set
{I, m, ?i) ;
and the second
step
is a rotation of the
principal
axes
of the strain into the
positions
of the
principal
axes of the strain
ellipsoid.
The formulte also admit of
interpretation
in the sense that
any
small
element of the
body,
which is
spherical
in the unstrained
state,
and has
a
given point
as
centre,
assumes after strain the
shape
and orientation of the
strain
ellipsoid
with its centre at the
corresponding point,
and
any
set of
three
orthogonal
diameters of the
sphere
becomes a set of
conjugate
diameters of the
ellipsoid.
30.
Application
to
cartography.
The methods of this
Chapter
would admit of
application
to the
problem
of
coustnicting
maps.
The surface to be
mapped
and the
plane map
of it are the
analogues
of a
body
in
the unstrained and strained states. The theorem that the strain about
any jioint
is
sensibly homogeneous
is the theorem that
any
small
portion
of the
map
is similar to one
of the
orthographic projections
of the
corresponding portion
of the
original
surface. The
analogue
of the
properties
of the
strain-elliiwoid
is found in the theorem that with
any
small circle on the
original
surface there
corresponds
a small
ellipse
on the
map
;
the
dimensions and orientation of the
ellipse,
with its centre at
any point, being known,
the
scale of the
map
near the
point,
and all distortions of
length,
area and
angle
are deter-
minate. These theorems form the foundation of the
theory
of
cartography. [Cf. Tissot,
Memoira sur la
representation
des
surfaces
et les
projections
des cartes
g^ographiques,
Paris,
1881.]
31. Conditions satisfied
by
the
displacement.
The
components
of
displacement
u, v,
w are not
absolutely arbitrary
functions of
x,
y,
z. In the
foregoing
discussion it has been assumed that
they
are
subject
to such conditions of
differentiability
and
continuity
as will
secure the
validity
of the "theorem of the total differential
+."
For our
purpose
this theorem is
expressed by
such
equations
as
du
_
du dx du
dy
du dz
ds
~
dx ds
dy
ds dz ds
'
Besides this
analytical
restriction,
there are others
imposed by
the
assumed condition that the
displacement
must be such as can be conceived
to take
place
in a continuous
body.
Thus,
for
example,
a
displacement, by
*
This
operation
is one of
homogeneous pure
strain. See Article
33, in/ra.
t Cf.
Harnack,
Introduction to the
Calculus,
London, 1891, p.
92.
64 GENERAL THEORY OF STRAIN
[CH.
I. APP.
which
every point
is
replaced by
its
optical image
in a
plane,
would be
excluded. The
expression
of
any component displacement by functions,
which become infinite at
any point
within the
region
of
space occupied by
the
body,
is also excluded.
Any analytically possible displacement, by
which
the
length
of
any
line would be reduced to
zero,
is also to be excluded. We
are thus concerned with real transformations
which,
within a certain
region
of
space,
have the
following properties
:
(i)
The new coordinates
{x
+
u,
y
+ V,
z
+
iv)
are continuous functions of the old coordinates
(a:, y,
z)
which
obey
the
theorem of the total differential,
(ii)
The real functions
u, v,
w are such
that the
quadratic
function
(1
+
2e^)
l'
+
{l
+
2ey,j)
m- +
(1
+
26^,)
Ji= +
2e,jim)i
+
2e^nl
+
2e^ylm
is definite and
positive, (iii)
The functional determinant denoted
by
1
+ A
is
positive
and does not vanish.
The condition
(iii)
secures that the strained state is such as can be
produced
from the unstrained
state,
by
a continuous series of small real
displacements.
It can be shown that it includes the condition
(ii)
when the
transformation is real. From a
geometrical point
of
view,
this amounts to
the observation
that,
if the volume of a variable tetrahedron is never reduced
to
zero,
none of its
edges
can ever be reduced to zero.
In the
particular
case of
homogeneous
strain,
the
displacements
are linear
functions of the coordinates. Thus all
homogeneous
strains are included
among
linear
homogeneous
transformations. The condition
(iii)
then ex-
cludes such transformations as involve the
operation
of reflexion in a
plane
in addition to transformations which can be
produced by
a continuous series
of small
displacements.
Some linear
homogeneous transformations,
which
obey
the condition
(iii), express
rotations about axes
passing through
the
origin.
All others involve the strain of some line. In
discussing homogeneous
strains and rotations it will be convenient to
replace {x
+
u,
y
+
v,
z
+
w) by
(i, 2/i. ^i)-
32. Finite
homogeneous
strain.
We shall take the
equations by
which the coordinates in the strained
state are connected
with the coordinates in the unstrained state to be
X,={1+ )
*
+
i..2/
+
i:|2^, 1
2/i
=
0^1
a;
+ (1
+
022) y
+
a-aZ, \
(13)
^1
=
a^^x
+
a-^y
+
(1
+
a^,)
z.
J
The
corresponding
components
of strain are
given by
the
equations
exx
=
n
+
i (an'
+
'2.'
+
W31'),
j:

Ct'oz
f
^23
"T"
^I'i'^is
t
^^22^23
"r
'^3-j^':i:i)
.(14)
31-33]
FINITE
HOMOGENEOUS STRAIN
(J5
The
quantities e:,^, ...,
defined in
Article
8,
do not lose their
importance
when the
displacements
are not small. The
notation used
here
may
be
identified with that of Article 8
by
writing,
for the
expressions
ii, 22. 3si da
+
(hi, 3i
+
i3. Oia
+
Oji, ttaa
-
a^, a,s
-
(J,,, Oj,
-
a,j,
the
expressions e^^,
eyy,
e,
ey
e,
e^
%m^,
2rsy,
2ct^.
Denoting
the radius vector from the
origin
to
any point P,
or
{x, y, z),
by
r,
we
may
resolve the
displacement
of P in the direction of
r,
and consider the ratio
of the
component displacement
to the
length
r. Let E be this ratio. We
may
define E to be the
elongation
of the material in the direction of r. We find
^
=
J
|(^>-^)^
+
(y.-2/)f
+
(..-.)^};
(1.5)
and this is the same as
Er""
=
e^ar"
+
Cyyy-
+
e^z''
+
Byiyz
+
e^^zx
+
e^yxy.
(16)
A
quadric
surface obtained
by equating
the
right-hand
member of this
equation
to a constant
may
be called an
elongatim quadric.
It has the
property
that the
elongation
in
any
direction is
inversely proportional
to
the central radius vector in that direction. In the case of
very
small dis-
placements,
the
elongation quadric
becomes the strain
quadric
previously
discussed
(Article 11).
The invariant
expressions
noted in Article 13
(c)
do
not cease to be invariant when the
displacements
are not small.
The
displacement expressed by (13)
can be
analysed
into two constituent
displacements.
One constituent is derived from a
potential, equal
to half
the
right-hand
member of
(16);
this
displacement
is
directed,
at each
point,
along
the normal to the
elongation quadric
which
passes through
the
point.
The other constituent
may
be derived from a vector
potential
-^[^Af+n ^!/(^'+n ^z(^+f)] (17)
by
the
operation
curl.
33.
Homogeneous pure
strain.
The direction of a line
passing through
the
origin
is unaltered
by
the
strain if the coordinates
x,
y,
z of
any point
on the line
satisfy
the
equations
(1
+
Un)
X
+
Oity
+ (h^
_
a-gX
+
(1
-I-
Oa) y
-|- O-gZ
X
~
y
_a
^a:-faMy
+
(l-t-a,)^
jg^
z
If each of these
quantities
is
put equal
to
\,
then \ is a root of the cubic
equation
1
-|- a,i

X
rt,a a,3
U.J1
1
-1-
a.^

\
iha
asi a^
1
-t-
0.33
-
X
=
(19)
L. K.
66 GENERAL THEORY OF
[CH.
I. APP.
The cubic has
always
one real
root,
so that there is
always
one line of which
the direction is unaltered
by
the
strain,
and if the root is
positive
the sense
of the line also is unaltered. When there are three such
lines,
they
are not
necessarily orthogonal;
but,
if
they
are.
orthogonal, they
are
by
definition the
principal
axes of the strain. In this case the strain is said to be
pure.
It
is worth while to
give
a formal
definition,
as follows :

Pure strain is such

that the set of three
orthogonal
lines which remain
orthogonal
retain their
directions and senses.
We
may prove
that the sufficient and
necessary
conditions that the strain
corresponding
with the
equations (13), may
be
pure,
are
(i)
that the
quad-
ratic forai on the left-hand side of
(20)
below is definite and
positive, (ii)
that
nTx,
^y,
nTz
vanish. That these conditions are
sufficient may
be
proved
as
follows:

When
CTj;,
Wy,
sr^ vanish,
or
a^
=
a33, ...,
the
equation (19)
is the
discriminating
cubic of the
quadric
(1
+
Ou)
x''
+
(l+ a^) y^
+
(l
+
a,,)
z^
+
la^yz
+
la^^zx
+
^.a^^xy
=
const.
; (20)
the left-hand member
being positive,
the cubic has three real
positive
roots,
which determine three real directions
according
to
equations (18);
and these
directions are
orthogonal
for
they
are the directions of the
principal
axes of
the surface
(20).
Further
they
are the
principal
axes of the
elongation quadric
a^^a?
-^
a^y"^
-f-
a^z^
-t-
^a^yz
-(-
la^^
zx
+
'ia^^xy
=
const.,
(21)
for this surface and
(20)
have their
principal
axes in the same directions.
The
vanishing
of
rs^,
vTy
and
ot^
are
necessary
conditions in order that the
strain
may
be
pure.
To
prove
this we
suppose
that
equations (13) represent
a
pure
strain,
and that the
principal
axes of the strain are a set of axes of
coordinates
f, rj, f.
The effect of the strain is to transform
any point (f
, 77, f)
into
(fi, 77,, f,)
in such a
way
that
when,
for
example, 7;
and
f vanish, -Vi
and
5i
also vanish. Referred to
principal
axes,
the
equations (13)
must be
equivalent
to three
equations
of the form
fi
=
(l+Of ^.
=
(1+6,)'?, ?,
=
(!+ 63)?, (22)
where
e-^, e^, e,
are the
principal
extensions. We
may express
the coordinates
f, 77, f
in terms of
x,
y,
z
by
means of an
orthogonal
scheme of substitution.
We take this scheme to be
33-35]
HOMOGENEOUS STRAIN
67
We should find the same
expression
for
a,,,,
and in the same
way
we should
find identical
expressions
for the
pairs
of coefficients
a.a, Wsa
and
a,,, Ou.
It
appears
from this discussion that a
homogeneous pure
strain is
equivalent
to three
simple
extensions,
in three directions
mutually
at
right
angles.
These directions are those of the
principal
axes of the strain.
34.
Analysis
of
any homogeneous
strain into a
pure
strain and a
rotation.
It is
geometrically
obvious that
any homogeneous
strain
may
be
produced
in a
body by
a suitable
pure
strain followed
by
a suitable rotation. To
determine these we
may proceed
as follows:

When we have found the

strain-components corresponding
with the
given
strain,
we can find the
equation
of the
reciprocal
strain
ellipsoid.
The
lengths
of the
principal
axes
determine the
principal
extensions,
and the directions of these axes are tho.se
of the
principal
axes of the strain. The
required pure
stiain has these
principal
extensions and
principal
axes,
and it is therefore
completely
determined. The
required
rotation is that
by
which the
principal
axes of
the
given
strain are
brought
into coincidence with the
principal
axes of
the strain
ellipsoid. According
to Article
28,
this rotation turns three
orthogonal
lines of known
position respectively
into three other
orthogonal
lines of known
position.
The
required angle
and axis of rotation can there-
fore be determined
by
a well-known
geometrical
construction.
[Cf.
Kelvin
and
Tait,
Nat. Phil. Part I.
p. 69.]
35. Rotation*.
When the
components
of strain
vanLsh,
the
displacement expressed by (13)
of Article
32 is a rotation about an axis
passing through
the
origin.
We shall take ^ to be the
angle
of rotation
and shall
suppose
the direction-cosines
I, m,
n of
the axis to be taken so that the rotation is
right-
handed.
Any point P,
or
{x, y, z\
moves on a circle
having
its centre
(C)
on the
axis,
and conies into a
position P[,
or
(j;,, yi, z,).
Let
X, /x,
be the direc-
tion-cosines of CP in the sense from C to
P,
and let
X,, ^1, v,
be those of
CP^
in the sense from C to
/*,.
From
/"i
let fall
P^N perpendicular
to GP. Tlie di-
rection-cosines of
iVPj
in the sense from iVto
P^
aret
mv
-
M/i,
X

Iv, Ifi

jnX.
Let
I, I), f
be the coordinates of C. Then these
Fig.
4.
satisfy
the
equations
1
=
1
=
^, l(^-a:)
+
m{,,-2/)
+
n{(-z)=0,
i m n
so that
^
=
l(Lv
+
my-\-nz)
with similar
expressions
for
v, C-
*
Cf. Kelvin and
Tait,
A'af. Phil. Part i.
p.
09,
and
Minohin,
Statics.
Third
Kdn.,
Oxford
1886,
vol.
2,
p.
103.
t
The coordinate axes are taken to be a
right-handed system.
52
68
GENERAL THEORY OF
[CH.
I. APP.
The coordinates of
P,
are obtained
by equating
the
projection
of
Cl\
on
any
coordinate
axis to the sums of the
projections
of CN and
NP^. Projecting
on the axis of x we
find,
taking p
for the
length
of CP or
CP^,
Xi p
=
Xp
cos 6 +
(otv

np.) p
sin
6,
or
Xi-i
=
{.r-^)coa6+{m{z-()-n{y-r,}}iime,
or
x\
=
x
+
(mz-7iy)iiine- {x-l{l.v+mi/
+
m)} (l-coaS) (23)
Similar
expressions
for
i/^
and
z,
can be written down
by symmetry.
The coefficients of the linear transformation
(13)
become in this case
o=-(l-;2)(l-cos5),
aj2=
-
n aia 6+
Im
{I
-cos
6),
a,3=
msin 6 +
In
{\
-COS
6),
.(24)
and it
appears,
on
calculation,
that the
components
of strain
vanish,
as
they ought
to do.
36.
Simple
extension.
In the
example
of
simple
extension
given by
the
equations
Xi
=
{l+e)x; y,=y, !^=z,
the
components
of
strain,
with the
exception
of
t^x vanish,
and
The invariant
property
of the
reciprocal
strain
ellipsoid may
be
applied
to find the
comix)nent8
of a strain which is a
simple
extension of amount e and direction
I, m,
n.
We should find
li---

2mn~-
^^
The same
property may
be
applied
to determine the conditions that a strain
specified by
six
components may
be a
simple
extension. These conditions are that the invariants
n,fK+...
+
...-i(V+

+
).
**x*yy'"r4 \^yM^xx*xy~
*xz*
yt~
'~
")
vanish. The amount of the extension is
expressed
in terms of the
remaining
invariant
by
the formula
s/{l
+2
{(xx
+
fyt+^n)}
-
h
the
jMsitive
value of the
square
root
being
taken.
Two roots of the cubic in
k, (7)
of Article
26,
are
equal
to
unity,
and the third is
equal
to
1+2
(fii
+
fj,,
+
jj).
The direction of the extension is the direction
{I, m, n)
that is.
given by
the
equations
ifjj
+
f^j^n
+
f^n
f^/
+
2f^OT
+
, _f^+m
+
2f^_
37.
Simple
shear.
In the
example
of
simj)le
shear
given by
the
equations
xi=x
+
si/, i/j=y, z^=z,
the
components
of strain are
given by
the
equations
^xx
^
f

~
"j ^yz^^
'z
^^
") *yy^ 2^J ^xy
^
*
By putting

=
2 tana we
may prove
that the two
principal
extensions which are not zero
are
given,
as in Article
3, by
the
equations
l
+
fj
=
seca-tana,
1
+
fj
=
seca +
tan a.
35-39] HOMOGENEOUS STRAIN
69
AVc
may prove
tlwit the area of a
figure
in tlie
piano
(,f
.v, y
i.s
unaltered
by
the shear and
that the
difference of the two
principal
extensions is
equal
to the
amount of the shear
Furtlier we
may
show that the
directions of tlie
principal
axes of the strain are the
bisectors of the
angle
AOx in
Fig.
2 of Article
5,
and that the
angle
through
which the
prmcipal
axes are tin-ncd is the
angle
n. So that the
simple
shear is
equivalent
to a
"
pure
shear
"
followed
by
a rotation
through
an
angle a,
as was
explained
before.
By using
the
invariants noted in Article
26,
we
may prove
that the
conditions that
a strain
with
given components e,
...
way
be a
shearing
strain are
and that the amount of the shear is
^/{2 (f
+
*j
+
))
38. Additional results
relating
to shear.
A
good example
of shear* is
presented by
a
sphere
built
up
of circular cai-ds in
parallel
planes.
If each card is shifted in its own
plane,
so that the line of centres becomes a
straight
line inclined
obliquely
to the
planes
of the
cards,
the
sphere
becomes an
ellipsoid,
and the cards coincide with one set of circular sections of the
ellipsoid.
It is an instructive
exercise to determine the
principal
axes of the strain and the
principal
extensions.
We
may
notice the
following
methodst of
producing any homogeneous
strain
by
a
sequence
of
operations
:

(a) Any
such strain can be
produced by
a
simple
shear
parallel
to one axis of
planes
I)erpendicular
to
another,
a
simple
extension in the direction at
right angles
to both
axe.s,
an uniform dilatation and a rotation.
(i) Any
such strain can be
produced by
three
simple
shears each of which is a shear
parallel
to one axis of
planes
at
right angles
to
another,
the three axes
being
at
right angles
to each
other,
an uniform dilatation and a rotation.
39.
Composition
of strains.
After a
body
ha.s been
subjected
to a
homogeneous
strain,
it
may again
be
subjected
to a
homogeneous
strain
;
and the result is a
displacement
of the
body,
which,
in
general,
could be effected
by
a
single homogeneous
strain.
More
generally,
when
any aggregate
of
points
is transformed
by
two homo-
geneous
linear transformations
successively,
the
resulting displacement
is
equivalent
to the effect of a
single
linear
homogeneous
transformation. This
statement
may
be
expressed by saying
that linear
homogeneous
transforma-
tions form a
group.
The
particular
linear
homogeneous
transformations with
which we are concerned are
subjected
to the conditions stated in Article
31,
and
they
form a continuous
group.
The transformations of
rotation,
described
in Article
35,
also form a
group
;
and this
group
is a
sub-group
included in
the linear
homogeneous group.
The latter
group
also includes all homo-
geneous strains;
but these do not
by
themselves form a
group,
for two
successive
homogeneous strains;): may
be
equivalent
to a rotation.
*
Suggested by
Mr B. K. Webb. Cf. Kelvin and
Tait,
Nat. Phil. Part i.
p.
122.
t
Cf. Kelvin and
Tait,
Nat. Phil. Part i.

178 et
seq.
t
A transformation such as
(13)
of Article
32, supposed
to
satisfy
coudition
(iii)
of Article
31,
expresses
a rotation if all the
components
of strain
(U)
vanish. In
any
other case it
expresses
a
homogeneous
strain.
.(26)
70
GENERAL
THEORY OF
[CH.
I. APP.
The result of two successive
linear
homogeneous
transformations
may
be
expressed
conveniently
in the notation of matrices. In this notation the
equations
of transformation
(13)
would be written
(^1, yi. ^i)
=
(
l+Oii i2 (h, )(, 2/. ^). (25)
Osi 1+022 ajs
Ctsi ^32
1
+
(ISS
and the
equations
of a second such transformation
could in the same
way
be
written
(2, 2/s. ^2)
=
(
1
+
^11 ^1-^ ^"
) (^'> 2/1
' ^i)-
fcji
1
+
b.
h.a
I 631 ^32
1
+
^3!
By
the fir.st transformation a
point {x, y, z)
is
replaced by (^1, y-^,
z^),
and
by
the second
{x^,
y^, z^)
is
replaced by {x^, y^, z.^).
The result of the two
operations
is that
{x,
y, z)
is
replaced by {x^, y^, z.,);
and we have
(xi, y^, Z;,)
=
(
I
+
Cn c,2 c,3 )(x,y,z),
(27)
C21
1
+
C22 C23
C31 C32
1
+
C3J
Cii
=
bn
+ an
+
bnttn
+
^^0^
+
bi^Chi,
Cm
=
b,2
+
ai2
+ 61, 012
+
612022
+
by^a^i,
where
.(28)
In
regard
to this
result,
we notice
(i)
that the transformations are not in
general
commutative
; (ii)
that the result of two successive
pure
strains is
not in
general
a
pure
strain
;
(iii)
that the result of two successive trans-
formations,
involving very
small
displacements,
is obtained
by simple super-
position,
that is
by
the addition of
corresponding
coefficients. The result
(ii) may
be otherwise
expressed by
the statement that
pure
strains do not
form a
group.
40. Additional results
relating
to the
composition
of strains.
When the transformation
(26)
is
equivalent
to a rotation about an
axis,
so that its
coefficients are those
given
in Article
35,
we
may
show that the
components
of strain
corresponding
with the transformation
(27)
are the same as those
corresponding
with the
transformation
(25),
as it is
geometrically
evident
they ought
to be.
In the
particular
case where the transformation
(25)
is a
pure
strain referred to its
principal axes, [so
that
ii
=
^i> ''22
=
^2) "33

'3;
^^^^ ^^^
remaining
coefficients
vanish],
and
the transformation
(26)
is a rotation about an
axis, [so
that its coefficients are those
given
in Article
35],
the coefficients of the resultant strain are
given by
such
equations
as
l+c
=
(l+f,){l-(l-^-^)(l-cos0)},
Cj2
=
(l
+
<2){-isin
d +
lm{\
-cos
5)},
39,
40]
HOMOGENEOUS STRAIN
71
The
quantities ra^,
sTj,,
w,,
corresponding
with this strain are not
components
of
rotation,
the
displacement
not
being
small. We should find for
example
2n'x=C32- "23
=
2^ sin ^ +
((4+63)
i sin 6
+
{(^-13)
mn
(1
-cos
fl).
We
may
deduce the result
that,
if the
components
of strain
corresponding
with the
transformation
(27) vanish,
and the condition
(iii)
of Article 31 is
satisfied,
the rotation
expressed by (27)
is of amount 6 about an axis
{I, m, n)
determined
by
the
equations
^*
=
2 8in5.
I m
We
may
show that the transformation
expressed by
the
equations
represents
a
homogeneous
strain
compounded
of uniform extension of all lines which are
at
right angles
to the direction
(htj.
:
nr
;
OTj)
and rotiition about a line in this direction.
The amount of the extension is
v'(l
+
n'x^
+
t5'/
+
a''j^)

1,
and the
tangent
of the
angle
of
rotation is
^(sij^
+
m/+-s!i^).
In the
general
case of the
composition
of
strains,
we
may
seek
expressions
for the
resultant
strain-components
in terms of the
strain-comixments
of the constituent strains
and the coefficients of the transformations. If we denote the
components
of strain
corresponding
with
(25), (26), (27) respectively by (fxx)ai 'x,x,t

{fxx)c>

>
we find such
formulcB as
(fix)c
=
(f)<<
+
(1
+
ll)^
f
:r:,I,
+
\l
f
Hill
+
%1 1.1
(f!,)c= (f.)a
+
2ai213':m
+
2(1 +022) 23*H.
+
2
(1 +33) 32*...,
+
{(l-|-a22)(l+a3.l)
+
2332}f..
+
{(l+'33)12
+
3213}*,a;i
+
{(l+22)"l3
+
<1223}'n,-
CHAPTER II.
ANALYSIS OF STRESS.
41. The notion of stress in
general
is
simply
that of
balancing
internal
action and reaction between two
parts
of a
body,
the force which either
part
exerts on the other
being
one
aspect
of a stress*. A familiar
example
is that
of tension in a bar
;
the
part
of the bar on one side of
any
normal section
exerts tension on the other
part
across the section. Another familiar
example
is that of
hydrostatic pressure.
At
any point
within a
fluid,
pressure
is
exerted across
any plane
drawn
through
the
point,
and this
pressure
is
estimated as a force
pec
unit of area. For the
complete specification
of the
stress at
any point
of a
body
we should
require
to know the force
per
unit of
area across
every plane
drawn
through
the
point,
and the direction of the
force as well as its
magnitude
would be
part
of the
specification.
For a
complete specification
of the state of stress within a
body
we should
require
to know the stress at
every point
of the
body.
The
object
of an
analysis
of
stress is to determine the nature of the
quantities by
which the stress at
a
point
can be
specified-}-.
In this
Chapter
we shall
develope
also those
consequences
in
regard
to the
theory
of the
equilibrium
and motion of
a
body
which follow
directly
from the
analysis
of stress.
42. Traction across a
plane
at a
point.
We consider
any
area 8 in a
given plane,
and
containing
a
point
within
a
body.
We denote the normal to the
plane
drawn
it;
a
specified
sense
by
v,
and we think of the
portion
of the
body,
which is on the side of the
plane
towards which p is
drawn,
as
exerting
force on the
remaining portion
across
the
plane,
this force
being
one
aspect
of a stress. We
suppose
that the
force,
which is thus exerted across the
particular
area
S,
is
statically equi-
valent to a force
R,
acting
at in a definite
direction,
and a
couple
0,
about
a definite axis. If we contract the area S
by any
continuous
process,
keeping
*
For a discussion of the notion of stress from the
point
of view of Bational
Mechanics,
see Note B. at the end of this book.
+ The
theory
of the
specification
of stress was
given by Cauohy
in the Article
'
De la
pression
ou tension dans nn
corps
solide
'
in the volume for 1827 of the Exercices de
matMmatiques.
41-43]
TRACTION ACROSS A PLANE
78
the
point always
within
it,
the force R and the
couple
tend
towards zero
limits,
and the direction of the force tends to a
limiting
direction
{I, m,
n).
We assume that the number
obtained
by
dividing
the number of units of
force in the force R
by
the number of units of area in the area 8
(say RjS)
tends to a limit
F,
which is not
zero,
and that on the other hand
OjS
tends
to zero as a limit. We define a vector
quantity by
the direction
{I, m,
n),the
numerical measure
F,
and the dimension
symbol
(mass)
(length)-' (time)"".
This
quantity
is a force
per
unit of area
;
we call it the traction across the
plane
v at the
point
0. We write
X^, Y,, Z,
for the
projections
of this vector
on the axes of coordinates. The
projection
on the normal v is
Xy
cos
(*', v)
+
Yy
cos
(y, v)
+
Z
cos
{z,
v).
If this
component
traction is
positive
it is a tension
;
if it is
negative
it is a
pressure.
If dS is a
very
small area of the
plane
normal to v at the
point
0,
the
portion
of the
body,
which is on the side of the
plane
towards which v is
drawn,
acts
upon
the
portion
on the other side with a force at the
point
0,
specified by
{XydS, Y,dS,
Z,dS);
this is the traction
upon
the element
of
area dS.
In the case of
pressure
in a fluid at
rest,
the direction
(/, m, n)
of the
vector
(Xy,
Yy, Z^)
is
always exactly opposite
to the direction v. In the cases
of viscous fluids in motion and elastic
solids,
this direction is in
general
obliquely
inclined to v.
43. Surface Tractions and
Body
Forces.
When two bodies are in
contact,
the nature of the action between them
over the surfaces in contact is assumed to be the same as the nature of the
action between two
portions
of the same
body, separated by
an
imagined
surface. If we
begin
with
any point
within a
body,
and
any
direction
for
V,
and allow to move
up
to a
point
0' on the
bounding
surface,
and
V to coincide with the outward drawn normal to this surface at
0',
then
Xy, Yy, Zy
tend to
limiting
values,
which are the
components
of the
surface-
traction at
0';
and
X^BS, Y^BS, Z^BS
are the forces exerted across the
element B8 of the
bounding
surface
by;^
some other
body having
contact
with the
body
in
question
in the
neighbourhood
of the
point
0'.
In
general
other forces act
upon
a
body,
or
upon
each
part
of the
body,
in
addition to the tractions on its surface. The
type
of such forces is the force
of
gravitation,
and such forces are in
general proportional
to the masses of
particles
on which
they
act, and, further,
they
are determined as to
magnitude
and direction
by
the
positions
of these
particles
in the field of force. If
X, Y,
Z are the
components
of the
intensity
of the field at
any point,
m the
mass of a
particle
at the
point,
then
inX, mY,
mZ are the forces of the field
74
EQUATIONS
OF MOTION OF BODIES
[CH.
II
that act on the
particle.
The forces of the field
may
arise from the action of
particles forming part
of the
body,
as in the case of a
body subject
to its own
gravitation,
or of
particles
outside the
body,
as in the case of a
body subject
to the
gravitational
attraction of another
body.
In either case we call them
body forces.
44.
Equations
of Motion.
The
body
forces,
applied
to
any portion
of a
body,
are
statically equi-
valent to a
single
force,
applied
at one
point, together
with a
couple.
The
components, parallel
to the
axes,
of the
single
force are
jjlpXdxdydz, jjlpYdxdydz, jjjpZdccdydz,
where
p
is the
density
of the
body
at the
point (x, y, z),
and the
integration
is taken
through
the volume of the
portion
of the
body.
In like
manner,
the
tractions on the elements of area of the surface of the
portion
are
equivalent
to a resultant force and a
couple,
and the
components
of the former are
\\x,dB,
jJY.dS, jjz^dS,
where the
integration
is taken over the surface of the
portion.
The centre of
mass of the
portion
moves like a
particle
under the action of these two sets
of
forces,
for
they
are all the external forces
acting
on the
portion.
If then
{fx, fy, fi)
is the acceleration of the
particle
which is at the
point (x, y, z)
at time
t,
the
equations
of motion of the
portion
are three of the
type*
1 1 Ipfxdxdydz
=
1 1
IpXdxdydz
+ I
jX^dS
(1)
where the
volume-integrations
are taken
through
the volume of the
portion,
and the
surface-integration
is taken over its surface.
Again
the
equations,
which determine the
changes
of moment of momentum
of the
portion
of the
body,
are three of the
type
jjjp
{yfz
-
rfy) d^dydz
=
jjjpi
yZ
-/F)
dxdydz
-H
[j"(
^/Z,
-zY:)dS;
(2)
and,
in accordance with the theorem
f
of the
independence
of the motion of
the centre of mass and the motion relative to the centre of
mass,
the
origin
of the coordinates
x,
y,
z
may
be taken to be at the centre of mass of the
portion.
The above
equations (1)
and
(2)
are the
types
of the
general equations of
motion
of
all bodies for which the notion of stress is valid.
*
The
equation (1)
is the form assumed
by
the
equations
of the
type
Smi'
=
SA',
of
my
Theoretical
Mechanics, Chapter
VI.
;
and the
equation (2)
is the form assumed
by
the
equations
of the
type
2m
{yz
-
zy)
=
2,{yZ -zY)
ot the same
Chapter.
t Theoretical
Mechanics,
Chapter
VI.
43-47]
SPECIFICATION OF STRESS
76
45.
Equilibrium.
When a
body
i.s at rest under the action of
body
forces and surface
tractions,
these are
subject
to the conditions of
equilibrium,
which are obtained from
equations (1)
and
(2) by
omission of the terms
containing f^, /,, /,.
We have thus six
equations,
viz. : three
of the
type
I I lpXdvdydz+ j j
A\dS=0,
(3)
and three of the
type
llp(j/Z-zr)dxdi/d!+jj(j/Z^-z}\)dS=0
(4)
It follows that if the
body
forces and surface tractions are
given
arbitrarily,
there will
not be
equilibrium.
In the
particular
case where there are no
body forces, equilibrium
cannot be maintained
unless the surface tractions
satisfy
six
equations
of the
types
I j
XydS=0,
and 1 1
{yZ.-zY,)dS=0.
46. Law of
Equilibrium
of surface tractions on small volumes.
From the forms alone of
equations (1)
and
(2)
we can deduce a result of
great importance.
Let the volume of
integration
be
very
small in all its
dimensions,
and let P denote this volume. If we divide both members of
equation
(1) by
P,
and then
pass
to a limit
by diminishing
I
indefinitely,
we
find the
equation
lim.
i-='//Z,cZS
=
0.
Again,
if we take the
origin
within the volume of
integration,
we obtain
by
a
similar
process
from
(2)
the
equation
lim.
;-// {yZ,
-
z
Y,)
dS
=
0.
1=0
The
equations
of which these are
types
can be
interpreted
in the statement :
The tractions on the elements
of
area
of
the
surface of any portion of
a
body,
which is
very
small in all its
dimensions,
are
ultimately,
to a
first approxima-
tion,
a
system offorces
in
equilibrium.
47.
Specification
of stress at a
point.
Through any point
in a
body,
there
passes
a
doubly
infinite
system
of
planes,
and the
complete specification
of the stress at involves the know-
ledge
of the traction at across all these
planes.
We
may
use the results
obtained in the last Article to
express
all these tractions in terms of the
component
tractions across
planes parallel
to the coordinate
planes,
and to
obtain relations between these
components.
We denote the traction across
a
plane

=
const,
by
its vector
components
{Xx,
Yx, Zx)
and use a similar
notation for the tractions across
planes y
=
const, and z
=
const. The
capital
letters show the directions of the
component
tractions,
and the suffixes the
planes
across which
they
act. The sense is such that
X^
is
positive
when it
is a
tension,
negative
when it is a
pressure.
If the axis of x is
supposed
76
SPECIFICATION OF STRESS
[CH.
II
Fig.
5.
drawn
upwards
from the
paper (cf. Fig.
5),
and the
paper
is
placed
so as to
pass through
0,
the traction in
question
is exerted
by
the
part
of the
body
above the
paper upon
the
part
below.
We consider the
equilibrium
of a tetrahedral
portion
of the
body, having
cue vertex at
0,
and the three
edges
that meet at this vertex
parallel
to the
axes of coordinates. The
remaining
vertices are the intersections of these
edges
with a
plane
near to 0. We denote
the direction of the normal to this
plane,
drawn
away
from the interior of the tetra-
hedron,
by
V,
so that its direction cosines
are cos
(a;, i^), cos(y, i^), cos(^, v).
Let A
be the area of the face of the tetrahedron
that is in this
plane
;
the areas of the
remaining
faces are
A cos
(w, v),
A cos
{y, v),
A cos
{z, v).
For a first
approximation,
when all the
edges
of the tetrahedron are
small,
we
may
take the resultant tractions across the face
V to be
X^A, ...,
and those on the remain-
ing
faces to be

Xa;Acos(a;, v)
The
sum of the tractions
parallel
to x on all the faces of the tetrahedron can be
taken to be
Z^
A
-
XxA
cos
(, v)

XyA
cos
(y, v)-
XjA
cos
(z, v).
By dividing by
A,
in accordance with the
process
of the last
Article,
we
obtain the first of
equations
(.5),
and the other
equations
of this set are
obtained
by
similar
processes
;
we thus find the three
equations
X,
=
Xx
cos
(w, v)
+
Xy
cos
{y, v)
+
X^
cos
{z, v),
\
F=
YrcCos{x, v)+
YyCos(y, v)+ Y^cos{z, v),
I
(.5)
Z
=
Zx
cos
{x, v)
+
Zy
cos
{y, v)+ Z^
cos
{z, v).
j
By
these
equations
the traction across
any plane through
is
expressed
in
terms of the tractions across
planes parallel
to the coordinate
planes. By
these
equations
also the
component
tractions across
planes, parallel
to the coordinate
planes,
at
any point
on the
bounding
surface of a
body,
are connected with
the tractions exerted
upon
the
body,
across the
surface,
by any
other
body
in
contact with it.
Again,
consider a
very
small cube
(Fig.
6)
of the material with its
edges
parallel
to the coordinate axes. To a first
approximation,
the resultant
tractions exerted
upon
the cube across the faces
perpendicular
to the axis
of X are
AXx, AYx,AZx,
for the face for which x is
greater,
and
AXx,
AYx, AZx,
for the
opposite
face,
A
being
the area of
any
face. Similar
expressions
hold for the other faces. The value o{
Jf(yZ^

zY^)dS
for the
47,
48]
BY MEANS OK SIX
COMPONENTS
77
cube can be taken to be /A
{Zy
-
Y,),
where I is the
length
of
any edge. By
the
process
of the last Article we obtain the first of
equations (6),
and the
other
equations
of this set are obtained
by
similar
processes;
we thus find
the three
equations
^y^z>
^z
=
Zx, Vx
=
Xy (6)
Y*
9
Fig.
6.
By equations (6)
the number of
quantities
which must be
specified,
in
order that the stress at a
point may
be
determined,
is reduced to
six,
viz.
three normal
component
tractions
X^,
Yy,
Zz,
and three
tangential
tractions
Fj, Zx,
Xy.
These six
quantities
are called the
components of
stress* at the
point.
The six
components
of stress are sometimes written
xx,
yy,
zz,
yz,
zx,
xy.
A notation of this kind is
especially
convenient when use is made of the
orthogonal
curvilinear coordinates of Article 19. The six
components
of
stress referred to the normals to the surfaces
a,
;S, 7
at a
point
will hereafter
be denoted
by
aa,
^B, 77, /37, 7a, a/3.
48. Measure of stress.
The state of stress within a
body
is determined when we know the values
at each
point
of the six
components
of stress. Each of these
stress-components
is a traction of the kind described in Article
42,
so that it is measured as a
force
per
unit area. The dimension
symbol
of
any .stress-component
is ML~^T~^.
*
A
symmetrical
method of
specifying
the stress is worked out
by
Lord Kelvin
(Article
10 foot-
note).
The method is
equivalent
to
taking
as the six
components
of stress at a
point
the tensions
per
unit of area across six
planes
which are
perpendicular respectively
to the six
edges
of a
chosen tetrahedron.
78
GEOMETRICAL CHARACTER
[CH.
II
A stress
may accordingly
be measured as so
many
"
tons
per square
inch,"
or so
many
"
dynes per square
centimetre,"
or more
generally,
as so
many
units of force
per
unit of area.
[One
ton
per square
inch
=
1'545 x 10'
dynes
per square
centimetre.]
For
example,
the
pressure
of the
atmosphere
is about 10
dynes per square
centimetre. As
exemplifying
the stresses which have to be allowed for
by
engineers
we
may
note the statement of W. C. Unwin* that the
Conway
bridge
is
daily subjected
to stresses
reaching
7 tons
per square
inch.
49. Transformation of
Stress-components.
Since the traction at a
given point
across
any plane
is determined when
the six
components
of stress at the
point
are
given,
it must be
possible
to
express
the six
components
x)f
stress,
referred to
any system
of
axes,
in terms
of those referred to another
system.
Let the
components
of stress referred
to axes of
od,
y,
z' be denoted
by
X'^,
. . .
;
and let the new coordinates be
given
in terms of the old
by
the
orthogonal
scheme of transformation
48-51]
OF COMPONENTS OF STRESS
79
50. The stress
quadric.
The formulas
(9)
show
that,
if the
equation
of the
quadric
surface
X^ofi+Yyy-
+
Z;:Z'+2Yiyz
+
2Z^zx+2XyXy
=
c6nBt
(10)
is transformed
by
an
orthogonal
substitution so that the left-hand member
becomes a function of
x',
y
,
z'
,
the coefficients of
x''^,
...
2y'z',
... in the left-
hand member are X'
r^,
...
FV,
The
quadric
surface
(10)
is called the stress
quadric.
It has the
property
that the normal stress across
any plane through
its centre is
inversely pro-
portional
to the
square
of that radius vector of the
quadric
which is normal
to the
plane.
If the
quadric
were referred to its
principal
axes,
the
tangential
tractions across the coordinate
planes
would vanish. The normal tractions
across these
planes
are called
principal
stresses. We learn that there
exist,
at
any point
of a
body,
three
orthogonal planes,
across each of which
the traction is
purely
normal. These are called the
principal planes of
stress. We also learn that to
specify completely
the state of stress at
any
point
of a
body
we
require
to know the directions of the
principal planes
of
stress,
and the
magnitudes
of the
principal
stresses
;
and that we
may
then
obtain the six
components
of
stress,
referred to
any
set of
orthogonal
planes, by
the
process
of
transforming
the
equation
of a
quadric
surface from
one set of axes to another. The stress at a
point may
be
regarded
as a
single quantity
related to directions
;
this
quantity
is not a
vector,
but has
six
components
in much the same
way
as a strain*.
51.
Types
of stress.
(a) Purely
normal stress.
If the traction across
every plane
at a
point
is normal to the
plane,
the terms contain-
ing products yz,
zx, xy
are
always
absent from the
equation
of the stress
quadric,
however
the
rectangular
axes of coordinates
may
be chosen. In this ovse
any
set of
orthogonal
lines
passing through
the
point
can be taken to be the
principal
axes of the
quadric.
It
follows that the
quadric
is a
sphere,
and thence that the normal
stress-components
are all
equal
in
magnitude
and have the same
sign.
If
they
are
positive
the stress is a
tension,
the same in all directions roimd the
iwint.
If
they
are
negative
the stress is
pressure,
with the like
property
of
equality
in all directionst.
(6) Simple
tension or
pressure.
A
simple
tension or
pressure
is a state of stress at a
point,
which is such that the
traction across one
plane through
the
point
is normal to the
plane,
and the traction across
any peri)endicular plane
vanishes. The
equation
of the stress
quadric
referred to its
principal
axes would be of the form
A'';j.a;'2=
const.
so that the
quadric
consists of a
pair
of
planes
normal to the direction of the
tension,
or
pressure.
The
components
of stress referred to
arbitrary
axes of
x, y,
z would be
X^=^X'^l\ Yy
=
X'^.m\ Z.= X'^,n\ Y.^X'^.mn, Z^=X'^nl, X,=^X':,dm,
where
{I, m, n)
is the direction of the
tension,
or
pressure,
and
A":^'
is its
magnitude.
If the
stress is tension
A''^^.
is
positive
;
if the stress is
pressure
X'.^'
is
negative.
*
In the
language
of
Voigt
it is a tensor-triad. Cf. Article 14
(b) supra.
i
This is a fundamental theorem of rational
Hydrodynamics,
cf.
Lamb,
Hydrodynamics, p.
2.
It was
proved
first
by Caucby,
see
Eiicy.
d. math,
il'iss.,
Bd.
4,
Art.
15, p.
52.
80 TYPES OF STRESS
[CH.
II
(c) Shearing
Stress.
The result
expressed by equations (6)
is
independent
of the directions of
the axes of
coordinates,
and
may
be stated as follows:

The
tangential
traction,
parallel
to a line
I,
across a
plane
at
right angles
to a line
I',
the
two lines
being
at
right angles
to each
other,
is
equal
to the
tangential
traction,
parallel
to
I',
across a
plane
at
right angles
to I. It follows that the
existence of
tangential
traction across
any plane implies
the existence of
tangential
traction across a
perpendicular plane.
The term
shearing
stress is
used to
express
the stress at a
point specified by
a
pair
of
equal tangential
tractions on two
perpendicular planes.
Wc
may
use the
analysis
of Article 49 to determine the
corresponding |)rincii)al
.stresses and
principal planes
of stress. Let the stress
quadric
be
2A'''.r'^'
=
const.,
so
that there is
tangential
traction
j>arallel
to the axis x' on a
plane y
=
const.,
and
eqnal
tangential
traction
parallel
to the axis
y'
on a
plane
y
=
const. Let the axes of
x, i/,
z he
the
principal
axes of the stress. The form
iX'yix'y'
is the same as
and this
ought
to be the same as
r,|(2ia-Y mi V2 ;
We therefore have
Z,=0, X^=

yy=X\';
and we find that the
shearing
stress is
equivalent
to tension across one of the
planes,
that
bisect the
angles
between the two
perpendicular planes concerned,
and
pressure
across the
other of these
planes.
The tension and the
pressure
are
equal
in absolute
magnitude,
and
each of them is
equal
to either
tangential
traction of the
shearing
stress.
iiiiUllil
y
Fig.
7.
The
diagram (Fig. 7),
illustrates the
equivalence
of the she--'-
"
'"es and the
principal
stresses.
Shearing
stress
equivalent
to such
princin.-'.
"|'<'^
' th"se
shown in
the
left-hand
figure maybe expected
to
pro<luce
shear ,= ^'".*
"'which
pianos
of the
material
that arc
iierpendicular
to the axis of
y
befor/
'-il-PHcation of the
stress
slide in a
direction
parallel
to the axis of
y,
and
plane.
.^'"1'"
'"'!':
*"
"^" '*"'" "^ ^'
'^'"^^ '" '"^
direction
parallel
to the axis of
y. T^- T^^^J ^.'f^;f ^^'
t^P"
^y '^^Y
^
expected
J V.

i

fii, t
'" (^ee
Article 14
(c).
to
produce shearing
strain of the
m'
'
51-53]
MEAN TENSION OB PRESSURE
81
{d)
Plaiie Stress.
A more
general type
of
stress,
which includes
simple
tension and
shearing
stress as
particular cases,
is obtained
by assuming
that one
principal
stress is zero. The stress
quadric
is then a
cylinder standing
on a conic as
base,
and the latter
may
be called the
stress conic
;
its
plane
contains the directions of the two
principal
stresses which do not
vanish.
Taking
this
plane
to be at
right angles
to the axis of
z,
the
equation
of the stress
conic is of the form
JCxX^
+
F,y*
+
2X,m/
=
const.
and the
shearing
stresses
Z,
and
V,
are
zero,
as well as the tension
Z,.
In the
particular
case of
simple
tension the stress conic consists of a
pair
of
parallel lines,
in the case of
shearing
stress it is a
rectangular hyperbola.
If it is a circle there is tension or
pressure
the same in all directions in the
plane
of the circle.
52. Resolution of
any stress-system
into uniform tension and
shearing
stress.
The
quantity
X^
+
Yy
+
Z^
is invariant as
regards
transformations from
one set of
rectangular
axes to another. When the
stress-system
is uniform
normal
pressure
of amount
p,
this
quantity
is

3p.
In
general,
we
may
call
the
quantity ^ (X^
+
Yy
+
Z^)
the
"
mean tension at a
point
"
;
and we
may
resolve the
stress-system
into
components
characterised
respectively by
the
existence and non-existence of mean tension. For this
purpose
we
may
put
X,
=
i{X,
+
Yy
+
Z,)
+
lX,-:^{Yy
+
Z,).
Then the
stress-system expressed by |Zj;

^ ( Fj,
-I-
Z^),
... involves no mean
tension. This
system
has the
property
that the sum of the
principal
stresses
vanishes
;
and it is
possible
to choose
rectangular
axes of coordinates
x',
y,
z
in such a
way
that the normal tractions
X'^,
Y'^,
Z'^,
corresponding
with
these
axes,
vanish.
Accordingly, stress-systems,
which involve no mean
tension at a
point,
are
equivalent
to
shearing
stresses
only,
in the sense that
three
orthogonal planes
can be found across which the tractions are
purely
tangential.
It follows that
any stress-system
at a
point
is
equivalent
to
tension
(or pressure),
the same in all directions round the
point, together
with
tangential
tractions across three
planes
which cut each other at
right
angles.
53. Additional results.
The
proofs
of the
following
results*
may
serve as exercises for the student :
(i)
The
quantities
X,+ Yy
+
Z YZ,+Z,X,+X,Yy-
Y^-Z^^-X,\
X,Y,Z,+2Y,Z,Xy-X,
17- Y,Z,^-Z,X,'^
are invariant as
regards orthogonal
transformations of coordinates.
*
The results
(i) (v)
are due to
Cauchy
and Lam(i.
L. K.
6
82
EQUATIONS
OF MOTION ANU OF
EQUILIBRIUM
[CH.
II
(ii)
If
Xj,
}',
Z,
are
principal stresses,
the traction across
any piano
is
proportional
to the central
i)erpendicular
on the
parallel tangent plane
of the
ellipsoid
3^lX^^-y^\ r2+0V2;2^
const.
This is Lame's
stresa-ellipsoid.
The
reciprocal
surface was discussed
by Cauchy
;
its
central radius vector in
any
direction is
inversely proportional
to the traction across the
plane
at
right angles
to that direction.
(iii)
The
quadric
surface
x^lX^+y^lYy
+
z^lZ^^consX. (in
which
X^,
... are
principal
stresses),
called Lamp's stress- director
quadric,
is the
reciprocal
of the stress
quadric
with
respect
to its centre
;
the radius vector from the centre to
any point
of the surface is in the
direction of the traction across a
plane parallel
to the
tangent plane
at the
point.
(iv)
The
planes
across which thei-e is no normal traction at a
point envelope
a cone of
the second
degree
which is the
reciprocal
of the
asymptotic
cone of the stress
quadric
at
the
point.
The former cone is Lamd's cone
of shearing
stress. When it is
real,
it
separates
the
planes
across which the normal traction is tension from those across which it is
pressure ;
when it is
imaginary
the normal traction across all
planes
is tension or
pressure according
as the mean tension
J (A'j.+ Yy-\-Z^
is
positive
or
negative.
(v)
If
any
two lines x and x' are drawn from
any point
of a
body
in a state of
stress,
and
planes
at
right angles
to them are drawn at the
point,
the
component parallel
to x' of
the traction across the
plane perpendicular
to x is
equal
to the
component parallel
to x of
the traction across the
plane perpendicular
to x'.
This
theorem,
which
may
be
expressed by
the
equation x\=Xx'i
is a
generalization
of
the results
(6)
of Article 47.
(vi)
Maxwell's electrostatic
stress-system*.
Let V be the
potential
of a
system
of electric
charges,
and let a
stress-system
be
determined
by
the
equations
^.-i{"-&7-}.
If
/3r\2
/3Fy /3T^^1
'~4n-
by
dz
'
8ffl \dxj '^\dyj \dzJi'
"'4^ dz dx'
I
dVdV
'
"
iir dx
dy'
It
may
be
shown, by taking
the axis of x to be
parallel
to the normal at
(x, y, z)
to the
equipotential
surface at the
point,
that one
principal plane
of the stress at
any point
is the
tangent plane
to the
equipotential
surface at the
point,
and that the traction across this
plane
is tension of amount
IP/Sn,
while the traction across
any perjiendicular plane
is
pressure
of the same
amount,
R
being
the resultant electric force at the
point
so that
"-^(%)'--
(vii)
If
u, V,
w are the
components
of
any
vector
quantity,
and
X^,
. . . arc the
components
of
any stress,
the three
quantities
X^u+X^v+Z^w, XyU
+
VyV+Y.w,
ZxU+Y,v+Z,iv
are the
components
of a
vector,
i.e.
they
are transformed from one set of
rectangular
axes
to another
by
the same substitution as
u, v,
w.
54. The
stress-equations
of motion and of
equilibrium.
In the
equations
of the
type (1)
of Article
44,
we substitute for
X,
...
*
Maxwell, EUetricity
and
Magnetism,
2nd
Edn., Oxford, 1881,
vol.
1,
ch. 5.
53,
54]
EXPRESSED IN TERMS OF
STRESS-COMPONENTS
83
from
equations (5).
We then
have,
as the
equation
obtained
by resolving
all
the forces
parallel
to the axis of
x,
1 1
{pfxdxdydz
=
1 1
\pXdxdydz
+
j|{A'';tCOs(a;,
v)+Xy(ios{y, v)
+
X^co6{z,
v)]dS. ...(11)
We
apply
Green's transformation* to the
surface-integral,
and
transpose,
thus
obtaining
the
equation
jj/r#^f'+f'+'"^-''/.)'^<'j'*-<'
02)
In this
equation
the
integration may
be taken
through any
volume within
the
body,
and it follows that the
equation
cannot be satisfied unless the
subject
of
integration
vanishes at
every point
within the
body.
Similar
results would follow
by transforming
the
equations
obtained
by resolving
all
the forces
parallel
to the axes of
y
and z. We thus obtain three
equations of
motion of the
type
pX
=
pf, (13)
>
+
^^*
-t-
^~
+
pz
0,
.(14)
dXx dXy , dX;i
da;
dy
dz
If the
body
is held in
equilibrium, /j;,/y,yi
are
zero,
and the
equations of
equilibrium
are
^^X
_|_
^Xy
_^
dZ^
^
-^
^
Q
due
dy
dz
^
'
ex
dy
dz
ciZx dTt
dZjc
dx
dy
dz
wherein
Y^, Z^,
Xy
have been written for the
equivalent Zy,
Xz, F^.
If the
body
moves so that the
displacement (u,
v,
w)
of
any particle
is
always very
small,
we
may put
s^t 3^ a^
'W dt'' dt-
instead oi
fx, fy,
fz,
the time
being
denoted
by
t;
the
equations of
small
motion are therefore
dXx
, dXy^dZ^^ Y
9'"
"
dz
dY
dz
dt'
dx
^
dy
^
^^
^^
P^f"
\
dx
dy
3X . dYu
^ ^j.K
dt^
d^w
The transformation is that
expressed by
the
equation
.(15)
dZxdY.dZ,
dx
dy
dz
J
62
84
EQUATIONS
OF MOTION AND OF
EQUILIBRIUM
[CH.
II
Other forms of
equations
of
equilibrium
and of
motion,
containing
fewer
unknown
quantities,
will be
given
hereafter. We
distinguish
the above
forms
(14)
and
(15)
as the
stress-equations.
55. Uniform stress and
uniformly varying
stress.
We observe that the
stress-equations
of
equilibrium (14)
hold within a
body,
and
equations (5)
hold at its
boundary, provided that,
in the latter
equations,
v is the direction
of the normal to the
bounding
surface drawn outwards and
X,,
... are the surface tractions.
The
equations may
be used to determine the forces that must be
applied
to a
body
to
maintain a
given
state of stress.
When the
components
of stress are
independent
of the
coordinates,
or the stress is the
same at all
points
of the
body,
the
body
forces vanish. In other
word.s, any
state of
imiform stress can be maintained
by
surface tractions
only.
We shall consider two cases :
(a) Uniform pressure.
In this case we have
^x=^v^z=Pt YiZx=Xy=0,
where
p
is the
pressure, supposed
to be the same at all
points
and in all directions round
each
point.
The surface tractions are
equal
to the
components
of a
pressure p
exerted
across the surface of the
body,
whatever the
shape
of the
body may
be. We
may
conclude
that,
when a
body
is
subjected
to constant
pressure p,
the same at all
points
of its
surface,
and is free from the action of
body forces,
the state of stress in the interior can be a state
of mean
pressure, equal
to
p
at each
point, unaccompanied by any shearing
stress.
(6) Simple
tension. Let T be the amount of the
tension,
and the axis of x its direction.
Then we have
Xx=T,
and the
remaining stress-components
vanish. We take 7' to be the
same at all
points.
The surface traction at
any point
is directed
parallel
to the axis of
x,
and its amount is
7'cos(^, v).
If the
body
is in the
shape
of a
cylinder
or
prism,
of
any
form of
section,
with its
length
in the direction of the axis of
.r,
there will be tensions on
its ends of amount T
per
unit
area,
and there will be no tractions across its
cylindrical
surface. We
may
conclude that when a bar is
subjected
to
equal
and
opposite
uniform
normal tensions over its
ends,
and is free from the action of
any
other
forces,
the state of
stress in the interior can be a state of tension across the normal
sections,
of the same
amount at all
points.
Uniform traction across a
plane
area is
statically equivalent
to a force at the centroid
of the area. The force has the same direction as the
traction,
and its
magnitude
is
measured
by
the
product
of the measures of the area and of the
magnitude
of the traction.
If the traction across an area is uniform as
regards
direction
and,
as
regards magnitude,
is
proportional
to
distance,
measured in a definite
sense,
from a definite line in the
plane
of the
area,
we have an
example
of
uniformly varying
stress. The traction across the area
is
.statically equivalent
to a
single
force
acting
at a certain
point
of the
plane,
which is
identical with the
"
centre of
pres.sure
"
investigated
in treatises on
Hydrostatics.
There
is an
exceptional case,
in which the line of zero traction
passes through
the centroid of the
area
;
the traction across the area is then
statically equivalent
to a
couple.
When the line
of zero traction does not intersect the
boundary
of the
area,
the traction has the same
sign
at all
points
of the area
;
and the centre of
pressure
must then lie within a certain curve
surrounding
the centroid. If the area is of
rectangular shape,
and the line of zero traction
is
jmrallel
to one
side,
the
greatest
distance of the centime of
pressure
from the centroid
is
ith
of that side. This result is the
engineers'
"rule of the middle third*."
*
Ewing, Strength of Materials, p.
104.
54-56]
EXPRESSED IN TERMS OF
STRESS-COMPONENTS
85
56.
Observations
concerning
the
stress-equations.
()
The
equations
of
type (13) may
be obtained
by applying
the
equations
of
type (1)
[Article 44]
to a small
parallelepiped
bounded
by planes parallel
to the coordinate
planes.
The
contributions of the faces x and x
+dxto i
jXydS
can be tiiken to be
-
X^dydz
and
{X^->r(bX^fiix)dx)dydz,
and similar
expressions
for the contributions of the
remaining
pairs
of faces can be written down.
(6)
The
equations
of moments of
type (2)
are
already
satisfied in
consequence
of
equations (6).
In fact
(2) may
te written
=
\\{p{yZ-zY)dxdydz
+
I I [y {2'x
cos
(^, i/)
+
Zcos(y, v)->rZ,CQ,9.{z, v))
-z{Yx
cos
{x, v)
+
Y
cos
{y, v)
+
}\
cos
(, v)}] dS,
by substituting
for
f^,
... from the
equations
of
type (13),
and for
F, ir
from
(5). By
help
of Green's
transformation,
this
equation
becomes
///
{Zy -Y,)dxdydz
=
0;
and thus the
equations
of moments are satisfied
identically
in virtue of
equations (6).
It
will be observed
that, equations (6) might
be
proved by
the above
analysis
instead of that
in Article 47.
(c)
When the
equations (14)
are satisfied at all
points
of a
body,
the conditions of
equilibrium
of the
body
as a whole
(Article 45)
are
necessarily satisfied,
and the resultant
of all the
body forces, acting upon
elements of volume of the
body,
is balanced
by
the
resultant of all the
tractions, acting upon
elements of its surface. The like statement is
true of the resultant moments of the
body
forces and surface tractions.
{d)
An
example
of the
application
of this remark is afforded
by
Maxwell's
stress-system
described in
(vi)
of Article 53. Wo should find for
example
dx
-(iy
dz 4^ ox
^
where
v^ stands for
dydx'
+
d'^ldy^
+
d^ldz\
It follows
that,
in
any region throughout
which
V^1'=0,
this
stress-system
is
self-equilibrating,
and
that,
in
general,
this
stress-system
is
1 /dV dV
dV\
in
equilibrium
with
body
force
specified by ~t-'^^^
\jr> 3~>
s~) P""
"*' volume.
Hence the tractions over
any
closed
surface,
which would be deduced from the formulae for
A'a-, ...,
are
statically equivalent
to
body forces, specified by
4"^*
T (
5-; ,
^
, ,,-
j
per
unit volume of the volume within the surface.
(e) Stress-functions.
In the
development
of the
theory
we shall be much
occupied
with bodies in
equilibrium
under forces
applied
over their surfaces
only.
In this case there are no
body
forces and
no
accelerations,
and the
equations
of
equilibrium
are
dx dx
dz^
dX^dY,dY,
^z.^^_r
hz.
j^-^-W
^ ^
8y
3^ s*'
9^
3^
while the surface tractions are
equal
to the values of
(A',
Y^, Z^)
at the surface of the
body.
The difierential
equations (16)
are three
independent
relations between the six
comjwnents
86 EQUATIONS
OF MOTION AND OF
EQUILIBRIUM
[CH.
II
of stress at
any point
; by
means of them we
might express
these six
quantities
in terms of
three
independent
functions of
position.
Such functions would be called "stress-functions."
So lono- as we have no information about the state of the
body,
besides that contained in
equations (16),
such functions are
arbitrary
functions.
One
way
of
expressing
the
stress-components
in terms of stress-functions is to assume*
and then it is clear that the
equations (16)
are satisfied if
r-^x.,^h2 yJlxi.^lM z-^'x^ +
^Im
Another
way
is to assumet
y
_3!i
r
-^2
y
_
?i3
^'~dydz'
'"dzdx'
'~dxdy'
'~
2dx\ dx'^dy^dzj'
"^
^dy\dx dy
Zz
)'
"'
'i.dz\dx'^ dy dzj'
These formulas
may
be
readily
verified. It will be observed that the relations between
the
X
functions and the
i/^
functions are the same as those between the
quantities
e,
...
and the
quantities e^^,
... in Article 17.
57.
Graphic representation
of stress.
States of stress
may
be illustrated in various
ways by
means of
diagrams,
but
complete
diagrammatic representations
cannot
easily
bo found. There are cases in which the
magnitude
and direction of the stress at a
point
can be determined
by inspection
of a
drawing
of a
family
of
curves, just
as
magnetic
force
may
be found
by
aid of a
diagram
of lines of force. But such cases are
rare,
the most
important being
the stress in a
twisted bar.
In the case of
plane stress,
in a
body
held
by
forces
applied
at its
boundary,
a
complete
representation
of the stress at
any point
can be obtained
by using
two
diagrams J.
The
stress is determined
by
means of a stress-function
Xt
so that
X
-^'-^
r
- -^
r
-
- ^
(17)
^^~dy^'
^"'dx'"
""
dxdy'
^ '
the
plane
of the stress
being
the
plane
of
x, y,
and
x being
a function of
x, y,
z. If the
curves ;,^
=
const, and 5r^
=
const, are traced for the same value of z and for
equidiflferent
ex
ay
values of the
constants,
then the tractions at
any point,
across
planes parallel
to the
planes
of
(x, z)
and
{y, z),
are directed
respectively along
the
tangents
to the curves =^
=
const, and
^esconst.
which
pass through
the
point,
and their
magnitudes
are
proportional
to the
closeness of consecutive curves of the
respective
families.
*
Maxwell, Edinburgh Roy.
Soc. Trans, vol. 26
(1870),
=
Scientific Papers,
vol.
2, p.
161.
The
particular
case of
plane
stress was discussed
by
G. B.
Airy,
Brit. Assoc.
Rep.
1862.
t
G.
Morera, Rome,
Ace. Lincei Rend.
(Ser. .5),
t. 1
(1892).
The relations between the two
Bystems
of stress-functions were discussed
by
Beltrami and Morera in the same volume.
X
3- H.
Michell,
London Math. Soc.
Proc,
vol. 32
(1901).
56-58]
EXPRESSED IN TERMS OF
STRESS-COMPONENTS
87
Partial
representations by graphic
means liave sometimes been used in cases where a
complete representation
cannot be obtained. Of this kind are
tracings
or models of the
"
lines of stress." These lines are snch tliat the
tangent
to
any
one of them at
any point
is normal to a
principal plane
of stress at the
point. Through any point
there
pass
three
such
lines, cutting
each other at
right angles.
These lines
may
determine a
trii)ly orthogonal
set of
surfaces,
but in
general
no such set exists. When such surfaces exist
they
are
described as
"
isostatic
surfaces*,"
and from a
knowledge
of them the directions of the
principal
stresses at
any point
can be infeired.
Distributions of stress
may
also be studied
by
the aid of
polarized light
The
method
+
is based on the
experimental
fact that an
isotropic transparent body,
when
stressed,
becomes
doubly refracting,
with its
optical principal
axes at
any point
in the
directions of the
principal
axes of stress at the
point.
58.
Stress-equations
referred to curvilinear
orthogonal
coordi-
nates^.
The
required equations may
be obtained
by finding
the transformed
expression
for
\\XydS
in the
general equation (1)
of Article 44. Now
we
have,
by equations (5),
Xy
=
Xx
cos
{x, v)
-)-
Xy
cos
(y, v)-\- X^
cos
(z, v),
and cos
{x, v)
=
cos
(a, v)
cos
{x, a)
+
cos
(/9, v)
cos
(x, /9)
-t-
cos
(7, v)
cos
(x, y),
so that
Xy
=
{Xx
cos
(x,
a) +
Xy
cos
(y, o)
+
X^
cos
{z, a)}
cos
(a, v)
+ two similar
expressions
=
Xa
cos
(a, v)
+
Xfi
cos
(/3, i')
+
Xy
cos
(7, v),
where,
for
example,
X^
denotes the traction in direction
x,
at a
point (a, /9, 7),
across the
tangent plane
at the
point
to that surface of the a
family
which
passes through
the
point. According
to the result
(v)
of Article 53 this is
the same as
cix,
the traction in the direction of the normal to the a surface at
the
point,
exerted across the
plane
x
=
const, which
passes through
the
point.
Further we
have,
by equations (5),
Ux
=
aoL cos
(a,
x)
-|-
a/3
cos
()S,
x)
+
7a
cos
(7, x).
Again,
cos
(a, v)
dS is the
projection
of the surface element
dS,
about
any
point
of
S,
upon
the
tangent plane
to the a surface which
passes through
the
*
These surfaces were first discussed
by Lam6,
J. de Math.
[Liouville),
t. 6
(1841),
and
Lefont
lur les coordonnees
curvilignes.
The fact that
they
do not in
general
exist was
pointed
out
by
Boussinesq,
Paris C.
R.,
t. 74
(1872).
Of.
Weingarten, J.f.
Math.
(CrelU),
Bd. 90
(1881).
t
The method
originated
with D.
Brewster,
Phil. Trans.
Ray.
Soc,
1816. It was de-
veloped by
F. E.
Neumann,
Berlin Abh. 1841,
and
by
Maxwell, Edinburgh Roy.
Soc.
Trans.,
vol. 20
(lSoS)
=
Scientific Papers,
vol.
1,
p.
30. For a more recent
experimental
investigation,
see
J.
Kerr,
Phil.
Mag. (Ser. 5),
vol. 26
(1888).
Reference
may
also be made to M. E.
Mascart,
Traits
d'Optique,
t. 2
(Paris 1891), pp.
229 et
set].
t
Other methods of
obtaining
these
equations
will be
given
in
Chapter
VII. and in the Note
on the
applications
of
moving
axes at the end of this book.
88
EQUATIONS
OF MOTION AND OF
EQUILIBRIUM
point,
and this
projection
is
d^dy/h^hs.
Hence
[CH.
II
d^dy
jiXydS
==
1
1
{aa
cos
(a, x)
+
a/3
cos
(/9, x)
+
7a
cos
(7, ))
y^^
+ 1 1
{a^cos (a, x)+ ^
cos
(/8, x)
+
^y
cos
(7, x)}
+
jj{ya.
cos
(a, x)
+
^y
cos
(yS, x)
+
77
cos
(7, x)}
-r-
When we
apply
Green's transformation to this
expression
we find
I
\X,dS
=
1
1
dad0dy
\^
7-7-
{aa
cos
(a, x)
+
ayS
cos
(0, x)
+
ya.
cos
(7, x)}
hsfii
dad0
+
go
ITT- {/3
cos
(a, x)+ 00
cos
(0,
x) +
0y
cos
(7, x)]
1
=-
rj-
{7a
cos
(o, x)
+
0y
cos
(0, x)
+
77
cos
(7, x)]
and,
since
{hihji^~^dad0dy
is the element of
volume,
we deduce from
(1)
the
equation
pfx
=
pX
+
hjiji^
da
\h2h3
{aa
cos
(a, x)
+
a0
cos
{0,
x)
+
7a
cos
(7, x)]
3 r 1
--^ ^-^ ^^
+
g^ U-^
{a;8
cos
(a, x)
+00
cos
(/8, x)+ 0y
cos
(7, a;)}
+
g- U-^
{7a
cos
(a,
a;)
+
0y
cos
(/3, )
+
77
cos
(7, x)}
j
(18)
The
angles
denoted
by (a, x),
... are variable with
a, 0,
7
because the
normals to the surfaces a
=
const.,
. . .
vary
from
point
to
point.
It
may
be
shown* that for
any
fixed direction of x the differential coefficients of
cos
(a, a;),
... are
given by
nine
equations
of the
type
^
cos
(a,
a^)
=
-h,^(j-y
cos
(0, x)
-
h,
^
(i)
. cos
(7, x),
A
cos
(a, ..)
=
^,
i
(i)
. cos
(0, X),
^
cos
(a, x)
=
h,
1
{)
. cos
(7. x).
We now take the direction of the axis of x to be that of the normal to
the surface a
=
const, which
passes through
the
point (a, 0, 7).
After the
differentiations have been
performed
we
put
cos
(a, x)=l,
cos
(0, x)
=
0,
cos
(7, x)
=
0.
We take
/
for the
component
acceleration
along
the normal to the surface
*
See the Note on
applications
of
moving
axes at the end of this book. In the
special
case of
cylindrical
coordinates the
corresponding equations
can be
proved directly
without
any difficulty.
58, 59]
EXPRESSED IN TERMS OF STRESS-COMPONENTS 89
a
=
const.,
and
F
for the
component
of
body
force in the same direction.
Equation
(18)
then becomes
The two similar
equations containing components
of acceleration and
body
force in the directions of the normals to
/3
=
const, and
y
=
const, can be
written down
by symmetry.
59.
Special
cases of
stress-equations
referred to curvilinear
co-
ordinates.
(i)
In the case of
cylindrical
coordinates
r, 0,
z
(cf.
Article
22)
the
stress-equations
are
drr IdrB drz rr-66
,
dr
+
r W
+
di
+
-^^
+
P^r=p/.,
drz 1 d6z dzz rz
r, j-
(ii)
In the case of
plane
stress referred to
cylindrical coordinates,
"when there is
equi-
librium under surface tractions
only,
the
stress-components,
when
expressed
in terms of
the stress-function
;(
of
equations (17),
are
given by
the
equations*
"-^29^-2+^37. ^*-g^2'
'''-
dr\rdd)'
(iii)
In the case of
polar
coordinates
r, 6, ^
the
stress-equations
are
drr 1 dr6 1 drd> 1
, c:: T", ^^i ^ ^ o j-
-^
+
--^
+
-^-z.,J+-{2rr-ed-<^(p
+
recote)
+
pFr
=
pfr,
or r 06 rmnd
d(f>
r^
^^
(iv)
When the surfaces
a, ft y
are isostatic so that
^y
=
ya=a^=0,
the
equations
can
be written in such forms+ as
da
P13 pn
where
p^^
and
pi,
are the
principal
radii of curvature of the surface a
=
const, which cor-
respond respectively
with the curves of intersection of that surface and the surfaces
/3
=
const, and
y
=
const.
*
J. H.
Michell,
London Math. Soc.
Proc,
vol. 31
(1899), p.
100.
t Lam^,
Coordonnees
curvilignes, p.
274.
CHAPTER III.
THE ELASTICITY OF SOLID BODIES.
60. In the
preceding Chapters
we have
developed
certain kinematical
and
dynamical
notions,
which are
necessary
for the theoretical discussion
of the
physical
behaviour of material bodies in
general.
We have now
to
explain
how these notions are
adapted
to elastic solid bodies in
par-
ticular.
An
ordinary
solid
body
is
constantly subjected
to forces of
gravitation,
and,
if it is in
equilibrium,
it is
supported by
other forces. We have no
experience
of a
body
which is free from the action of all external forces.
From the
equations
of Article 54 we know that the
application
of forces to
a
body
necessitates the existence of stress within the
body.
Again,
solid bodies are not
absolutely rigid. By
the
application
of suitable
forces
they
can be made to
change
both in size and
shape.
When the
induced
changes
of size and
shape
are
considerable,
the
body
does
not,
in
general,
return to its
original
size and
shape
after the forces which induced
the
change
have ceased to act. On the other
hand,
when the
changes
are
not too
great
the
recovery may
be
apparently complete.
The
property
of
recovery
of an
original
size and
shape
is the
property
that is termed
elasticity.
The
changes
of size and
shape
are
expressed by specifying
strains. The
"
unstrained state
"
(Article 4),
with reference to which strains
are
specified,
is,
as it
were,
an
arbitrary
zero of
reckoning,
and the choice of
it is in our
power.
When the unstrained state is
chosen,
and the strain is
specified,
the internal
configuration
of the
body
is known.
We shall
suppose
that the differential coefficients of the
displacement
{u, V, lo), by
which the
body
could
pass
from the unstrained state to the
strained
state,
are
sufficiently
small to admit of the calculation of the strain
by
the
simplified
methods of Article 9
;
and we shall
regard
the con-
figuration
as
specified by
this
displacement.
For the
complete specification
of
any
state of the
body,
it is
necessary
to
Ivuow the
temj)erature
of
every part,
as well as the
configuration.
A
change
of
configuration may,
or
may
not,
be
accompanied by changes
of
temperature.
61. Work and
energy.
Unless the
body
is in
equilibrium
under the action of the external
forces,
0,
61]
ENERGY OF ELASTIC SOLID BODY
91
it will be
moving through
the
configuration
that is
specified
by
the
displace-
ment,
towards a new
configuration
which could be
specified by
a
slightly
different
displacement.
As the
body
moves from one
configuration
to
another,
the external forces
(body
forces and
surface
tractions)
in
general
do some work
;
and we can
estimate the
quantity
of work done
per
unit of
time,
that is to
say
the rate at which work is done.
Any body,
or
any portion
of a
body,
can
possess
energy
in various
ways.
If it is in
motion,
it
possesses
kinetic
energy,
which
depends
on the distri-
bution of mass and
velocity.
In the case of small
displacements,
to which
we are
restricting
the
discussion,
the kinetic
energy per
unit of volume is
expressed
with sufficient
approximation by
the formula
*^{(IT-'-'}.
in which
p
denotes the
density
in the unstrained state. In addition to the
molar kinetic
energy, possessed by
the
body
in
bulk,
the
body possesses
energy
which
depends upon
its
state,
i.e.
upon
its
configuration
and the
temperatures
of its
parts.
This
energy
is called "intrinsic
energy";
it is to
be calculated
by
reference to a standard state of chosen uniform
temperature
and zero
displacement.
The total
energy
of
any portion
of the
body
is the
sum of the kinetic
energy
of the
portion
and the intrinsic
energy
of the
portion.
The total
energy
of the
body
is the sum of the total
energies
of
any parts*,
into which it can be
imagined
to be divided.
As the
body passes
from one state to
another,
the total
energy,
in
general,
is altered
;
but the
change
in the total
energy
is
not,
in
general, equal
to the
work done
by
the external forces. To
produce
the
change
of state it
is,
in
general, necessary
that heat should be
supplied
to the
body
or with-
drawn from it. The
quantity
of heat is measured
by
its
equivalent
in work.
The First Law of
Thermodynamics
states that the increment of the
total
energy
of the
body
is
equal
to the sum of the work done
by
the
external forces and the
quantity
of heat
supplied.
We
may
calculate the rate at which work is done
by
the external forces.
The rate at which work is done
by
the
body
forces is
expressed by
the formula
/.
/
P
^i*y%*'Ty'y^'
(')
where the
integration
is taken
through
the volume of the
body
in the
unstrained state. The rate at which work is done
by
the surface tractions
is
expressed by
the formula
/A^-l'-''-^^. '!)<
*
For the
validity
of the
analysi.s
of the
energy
into molar kinetic
energy
and intrinsic
energy
it is
necessary
that the dimensions of the
parU
in
question
should be
large compared
with
molecular dimensions.
92
ELASTICITY OF SOLID BODIES
[CH.
Ill
where the
integration
is taken over the surface of the
body
in the unstrained
state. This
expression may
be transformed into an
integral
taken
through
the volume of the
body, by
the use of Green's transformation and of the
formulae of the
type
Xy
=
Xx
cos
(x, v)
+
Xy
cos
(y,v)
+
X^
cos
(z, p),
We use also the results of the
type
Yz
=
Zy,
and the notation for strain-
components
e-Rt,
We find that the rate at which work is done
by
the
surface tractions is
expressed by
the formula
///[(
dX^_^^_Xy^dZ^\'du_^fdXy
.
dx
dy
dz ) dt \ dx
dy
dz
V
dx
dy
dy
dY,dZ,\dw
'^
dz J dt
aw
dt
+
III
X,
de^
dt
+
Xy
dxdydz
dexy
dt
dxdydz.
...(2)
We
may
calculate also the rate at which the kinetic
energy
increases.
This rate is
expressed
with sufficient
approximation by
the formula
lljPWdi
d^u du d'v dv d^tu dw
dt^Vt^m^W^^y^''
.(3)
where the
integration
is taken
through
the volume of the
body
in the
unstrained state. If we use the
equations
of
motion,
(15)
of Article
54,
we can
express
this in the form
///[("-
5^
, dXy
dx
dy
+
dZx\
du
17
...1,
+
...
+ ... \
dxdydz.
It
appears
hence that the
expression
///
9^
dt
de:
3e,
+
z,-;^
+
Y,-;^+z,
de.
dt dt w^^-
xy
"
dt
dxdydz (4)
represents
the excess of the rate at which work is done
by
the external forces
above the rate of increase of the kinetic
energy.
62. Existence of the
strain-energy-function.
Now let
Sr,
denote the increment of kinetic
energy per
unit of
volume,
which is
acquired
in a short interval of time ht. Let hUhe the increment of
intrinsic
energy per
unit of
volume,
which is
acquired
in the same interval.
Let 8
Wi
be the work done
by
the external forces in the
interval,
and let
hQ
be
the mechanical value of the heat
supplied
in the interval. Then the First
Law of
Thermodynamics
is
expressed by
the formula
jjj(B2\
+
BU) dxdydz
=
8W,
+
BQ (5)
61,
62]
EXPRESSED BY
STRAIN-ENERGY-FUNCTION
98
Now,
according
to the final result
(4)
obtained in Article
61,
we have
SW^-jjjST.dxdydz
=
I jj(XxBe^x
+
YyBeyy
+
Z,Be^,
+
Y^Sey^
+Z,8ea,
+
XyBe^y) dxdydz,
. .
.(6)
where
8e^,
...
represent
the increments of the
components
of strain in the
interval of time St. Hence we have
jjJBUdxdydz
=
BQ+jjj(X^Be^^
+
...)dxdi/dz (7)
The differential
quantity
BU is the differential of a function
U,
which is an
one-valued function of the
temperature
and the
quantities
that determine
the
configuration.
The value of this function
U,
corresponding
with
any
state,
is the measure of the intrinsic
energy
in that state. In the standard
state,
the value of U is zero.
If the
change
of state takes
place adiabatically,
that is to
say
in .such
a
way
that no heat is
gained
or lost
by any
element of the
body,
BQ vanishes,
and we have
BV'=
XxBexx
+
YyBeyy
+
Z2Beiz
+
YzBeyz
+
ZJie^x-'r
XyBe^y (8)
Thus the
expression
on the
right-hand
side
is,
in this
case,
an exact
differential
;
and there exists a function
W,
which has the
properties
ex-
pressed by
the
equations
dW dW
Z.
=
^-f
.... F.
=
|f
(9)
The function W
represents potential energy, per
unit of
volume,
stored
up
in
the
body by
the strain
;
and its
variations,
when the
body
is strained adiaba-
tically,
are identical with those of the intrinsic
energy
of the
body.
It is
probable
that the
changes
that
actually
take
place
in bodies
executing
small
and
rapid
vibrations are
practically
adiabatic.
A function which has the
properties
expressed by equations (9)
is
called a
"strain-energy-function."
If the
changes
of state take
place isothermally,
i.e. so that the
temperature
of
every
element of the
body
remains
constant,
a function W
having
the
properties expressed by equations (9)
exists. To
prove
this we utilise the
Second Law of
Thermodynamics
in the form
that,
in
any
reveirsible
cycle
of
changes
of state
performed
without variation of
temperature,
the sum of the
elements
BQ
vanishes*. The sum of the elements BU also
vanishes;
and it
follows that the sum of the elements
expressed by
the formula
S
{X^Be^
+
YyBeyy
+
ZMzz
+
Y.Bey^
+
Z^Be
+
XyBe,y)
*
Cf, Kelvin,
Math, and
Phys. Papers,
vol.
1, p.
291.
94 ELASTICITY OF SOLID BODIES
[CH.
Ill
also vanishes in a reversible
cycle
of
changes
of state without variation of
temperature.
Hence the differential
expression
XxBbxx
+
YyScyy
+
Z^Bciz
+
i'j
S^y^
+
Z^Bb^x
+
Xyhexy
is an exact
differential,
and the
strain-energy-function
W exists.
When a
body
is strained
slowly by gradual
increase of the
load,
and
is in continual
equilibrium
of
temperature
with
surrounding
bodies,
the
changes
of state are
practically
isothermal.
63. Indirectness of
experimental
results.
The
object
of
experimental investigations
of the behaviour of elastic
bodies
may
be said to be the
discovery
of numerical relations between the
quantities
that can be
measured,
which shall be
sufficiently
varied and
sufficiently
numerous to serve as a basis for the inductive determination of
the form of the intrinsic
energy-function,
viz. the function U of Article 62.
This
object
has not been
achieved,
except
in the case of
gases
in states that
ai-e far removed from critical states. In the case of elastic
solids,
the con-
ditions are much more
complex,
and the results of
experiment
are much
less
complete;
and the indications which we have at
present
are not
suflScient for the formation of a
theory
of the
physical
behaviour of a solid
body
in
any
circumstances other than those in which a
strain-energy-function
exists.
When such a function
exists,
and its form is
known,
we can deduce from it
the relations between the
components
of stress and the
components
of strain
;
and,
conversely,
if,
from
any experimental
results,
we are able to infer such
relations,
we
acquire thereby
data which can serve for the construction of the
function.
The
components
of stress or of strain within a solid
body
can
never,
from
the nature of the
case,
be measured
directly.
If their values can be
found,
it must
always
be
by
a
process
of inference from measurements of
quantities
that are
not,
in
general, components
of stress or of strain.
Instruments can be devised for
measuring average
strains in bodies of
ordinary
size,
and others for
measuring particular
strains of small
superficial
parts.
For
example,
the
average
cubical
compression
can be measured
by
means of a
piezometer;
the extension of a short
length
of a
longitudinal
filament on the outside of a bar can be measured
by
means of an extenso-
meter.
Sometimes,
as for
example
in
experiments
on torsion and
flexure,
a
displacement
is measured.
External forces
applied
to a
body
can often be measured -with
great
exactness,
e.g.
when a bar is extended or bent
by hanging
a
weight
at one
end. In such cases it is a resultant force that is measured
directly,
not
the
component
tractions
per
unit of area that are
applied
to the surface of
the
body.
In the case of a
body
under normal
pressure,
as in the
experi-
ments with the
piezometer,
the
pressure per
unit of area can be measured.
62-64]
EXPRESSED BY GENERALIZED HOOKE'S LAW
95
In
any experiment designed
to determine a relation between stress and
strain,
some
displacement
is
brought
about,
in a
body partially
fixed,
by
the
application
of definite forces which can be varied in
aniount.
We call these
forces
collectively
"
the load."
64. Hooka's Law.
Most hard solids show the same
type
of relation between load and
measurable strain. It is foand
that,
over a wide
range
of
load,
the measured
strain is
proportional
to the load. This statement
may
be
expressed
more
fully by saying
that
(1)
when the load increases the measured strain increases in the same
ratio,
(2)
when the load diminishes the measured strain diminishes in the
same
ratio,
(3)
when the load is reduced to zero no strain can be measured.
The most
striking exception
to this statement is found in the behaviour of
cast metals. It
appears
to be
impossible
to
assign any
finite
range
of
load,
within which the measurable strains of such metals increase and diminish in
the same
proportion
as the load.
The
experimental
results which hold for most hard
solids,
other than cast
metals,
lead
by
a
process
of inductive
reasoning
to the Genwalized Hookes
Law
of
the
proportionality of
stress and strain. The
general
form of the
law is
expressed by
the statement :

Each
of
the six
components of
stress at
any point of
a
body
is a linear
function of
the six
components of
strain at the
point.
It is
necessary
to
pay
some attention to the
way
iu which this law
represents
the
experimental
results. In most
experiments
the load that is
increa.sed,
or
diminished,
or
reduced to zero consists of
part only
of the external forces. The
weight
of the
body
subjected
to
experiment
must be balanced
;
and neither the
weight,
nor the force
employed
to balance
it, is,
in
general,
included in the load. At the
beginning
and end of the
experi-
ment the
body
is in a state of stress
;
but there is no measured strain. For the strain that
is measured is reckoned from the state of the
body
at the
beginning
of the
experiment
as
standard state. The strain referred to in the statement of the law must be reckoned from
a different state as standard or
"
unstrained
"
state. This state is that in which the
body
would be if it were freed from the action of all external
forces,
and if there were no internal
stress at
any point
of it. We call this state of the
body
the "unstressed state." Reckoned
from this state as
standard,
the
body
is in a state of strain at the
beginning
of the
experi-
ment
;
it is also in a state of stress. When the load is
applied,
the stress is altered in
amount and distribution
;
and the strain also is altered. After the
application
of the
load,
the stress consists of two
stress-systems
: the
stress-system
iu the initial
state,
and a
stress-system by
which the load would be balanced all
through
the
body.
The
strain,
reckoned from the unstressed
state,
is likewise
compouuded
of two strains : the strain from
the unstressed state to the initial
state,
and the strain from the initial state to the state
assumed under the load. The
only things,
about which the
experiments
axn tell us
any-
thing,
are the second
stress-system
and the second strain
;
and it is consonant with the
96
ELASTICITY OF SOLID BODIES
[CH.
Ill
result of the
experiments
to assume that the law of
proportionality
holds for this stress
and strain. The
general
statement of the law of
proportionality implies
that the stress in
the initial state also is
proportional
to the strain in that state. It also
implies
that both
the initial
state,
and the state assumed under the
load,
are derivable from the unstressed
state
by displacements,
of amount
sufficiently
small to admit of the calculation of the
strains
by
the
simplified
methods of Article 9. If this were not the
case,
the strains would
not be
compounded by simple superposition
: and the
proportionality
of load and measured
strain would not
imply
the
proportionality
of
stress-components
and
strain-components.
65. Form of the
strain-energy-function.
The
experiments
which lead to the enunciation of Hooke's Law do not
constitute a
proof
of the truth of the law. The law formulates in abstract
terms the results of
many
observations and
experiments,
but it is much
more
precise
than these results. The mathematical
consequences
which can
be deduced
by assuming
the law to be true are sometimes
capable
of
experimental
verification
; and,
whenever this verification can be
made,
fresh
evidence of the truth of the law is obtained. We shall be
occupied
in sub-
sequent chapters
with the deduction of these
consequences
;
here we note
some results which can be deduced
immediately.
When a
body
is
slightly
strained
by gradual application
of a
load,
and
the
temperature
remains
constant,
the
stress-components
are linear functions
of the
strain-components,
and
they
are also
partial
differential coefficients of
a function
(W)
of the
strain-components.
The
strain-energy-function,
W,
is therefore a
homogeneous quadratic
function of the
strain-components.
The known
theory
of sound waves* leads us to
expect
that,
when a
body
is
executing
small
vibrations,
the motion takes
place
too
quickly
for
any
portion
of the
body
to lose or
gain any
sensible
quantity
of heat. In this
case also there is a
strain-energy-function
; and,
if we assume that Hooke's
Law
holds,
the function is a
homogeneous quadratic
function of the strain-
components.
When the
stress-components
are eliminated from the
equations
of motion
(15)
of Article
54,
these
equations
become linear
equations
for the
determination of the
displacement.
The
linearity
of
them,
and the
way
in which the time enters into
them,
make it
possible
for them to
possess
solutions which
represent
isochronous vibrations. The fact that all solid
bodies admit of
being
thrown into states of isochronous vibration has been
emphasized by Stokesf
as a
peremptory proof
of the truth of Hooke's Law for
the
very
small strains involved.
The
proof
of the existence of W
given
in Article 62
points
to different
coefficients for the terms of W
expressed
as a
quadratic
function of strain-com-
ponents,
in the two cases of isothermal and adiabatic
changes
of state. These
coeflBcients are the
"
elastic
constants,"
and
discrepancies
have
actually
been
*
See
Eayleigh, Theory of
Sound, Chapter
XI.
t
See Introduction,
footnote 37.
64-66]
EXPRESSED BY GENERALIZED HOOKE'S LAW
97
found in
experimental
determinations of the constants
by
statical
methods,
involving
isothermal
changes
of
state,
and
dynamical methods,
involving
adia-
batic
changes
of state*. The
discrepancies
are
not, however,
very
serious.
To secure the
stability
of the
body
it is
necessary
that the coeflBcients of
the terms in the
homogeneous quadratic
function W should be
adjusted
so that the function is
alvs'ays p&sitivef.
This condition involves certain
relations of
inequality among
the elastic constants.
If Hooke's Law is
regarded
as a first
approximation,
valid in the case of
very
small
strains,
it is natural to assume that the terms of the second order
in the
strain-energy-function
constitute likewise a first
approximation.
If
terms of
higher
order could be taken into account an extension of the
theory
might
be made to circumstances which are at
present
excluded from its
scope.
Such extensions have been
suggested
and
partially
worked out
by
several writers
|.
66. Elastic constants.
According
to the
generalized
Hooke's
Law,
the six
components
of stress
at
any point
of an elastic solid
body
are connected with the six
components
of strain at the
point by equations
of the form
X
z
^
^ii^xx
T
Ci2^yy
+
Ca^zz
1
^u^yz
'
^is^zx
"I"
^te^xyi
.(10)
The coefficients in these
equations,
c,
... are the elastic constants of the
substance.
They
are the coefficients of a
homogeneous quadratic
function
2
W,
where W is the
strain-energy-function
;
and
they
are therefore con-
nected
by
the relations which ensure the existence of the function. These
relations are of the form
Cr,
=
c; (r,
s
=
l, 2,
...
6), (11)
and the number of constants is reduced
by
these
equations
from 36 to 21.
*
The
discrepancies appear
to have been noticed first
by
P.
Lagerbjelm
in
1827,
see Todhunter
and Pearson's
History,
voL
1,
p.
189.
They
were made the
subject
of extensive
experiments by
G.
Wertheim,
Ann. de
CIdmie,
t. 12
(1844).
Information
concerning
the results of more recent
experimental
researches is
given by
Lord Kelvin
(Sir
W.
Thomson)
in the Article
'Elasticity'
in
Ency.
Brit.,
9th edition, reprinted
in Math, and
Phys. Papers,
vol. 3. See also W.
Voigt,
Ann.
Phys.
Chem.
(Wiedemann),
Bd. 52
(1894).
t Kirchhoff, Vorlesungen
iiber...Mechanik, Vorlesung
27. For a discussion of the
theory
of
stabiUty
reference
may
be made to a
paper by
R.
Lipschitz, J.f.
Math.
(Crelle),
Bd. 78
(1874).
t
Reference
may
be
made,
in
particular,
to W.
Voigt,
Ann.
Phys.
Chem.
(Wiedemann),
Bd.
52,
1894, p.
536 and Berlin
Berichte,
1901.
L. E.
*
98
DETERMINATION OF STRESS
[CH.
Ill
We write the
expression
for 2 TT in the form
2W=
ce^ax
+
^Cuejcxeyy
+
^.c^^exxezz
+
^c^exx^yz
+
^c-^texxezx
+
-Cwexxexy
+
C3se^zz
+
2c:ejjeyz
+
^c^e^^e^x
+
2c36eea:j,
+
C44e'yz
-\'
^Ci!,eyze2x'^ ^^AS^yz^xy
(12)
The
theory
of
Elasticity
has sometimes been based on that
hypothesis
concerning
the constitution of
matter,
according
to which bodies are
regarded
as made
up
of material
points,
and these
points
are
supposed
to act on each
other at a
distance,
the law of force between a
pair
of
points being
that the
force is a function of the distance between the
points,
and acts in the line
joining
the
points.
It is a
consequence
of this
hypothesis*
that the co-
efficients in the function W are connected
by
six additional
relations,
whereby
their number is reduced to 15. These relations are
C2S
=
C44> ^31= C53, C12
=
Cos,
/10\
We shall refer to these as
"
Cauchy's
relations
"
;
but we shall not assume
that
they
hold
good.
67. Methods of
determining
the stress in a
body.
If we wish to know the state of stress in a
body
to which
given
forces are
applied,
either as
body
forces or as surface
tractions,
we have to solve the
stress-equations
of
equilibrium (14)
of Article
54,
viz.
dx
dy
dz
^
'
\
dx
^
dy
^
dz
^P^
"'
f
^^*^
dx
dy
dz
"
'
and the solutions must be of such forms that
they give
rise to the
right
expressions
for the surface
tractions,
when the latter are calculated from the
formulae
(5)
of Article
47,
viz.
X,
=
Xx
cos
{x, v)
+
Xy
cos
(y, v)
+
Zx
cos
{z, v),
.(15)
The
equations (14)
with the conditions
(15)
are not sufficient to determine
the
stress,
and a
stress-system may satisfy
these
equations
and conditions
and
yet
fail to be the correct solution of the
problem ;
for the
stress-compo-
*
See Note B at the end of this book.
^
66-68]
IN AN ELASTIC SOLID BODY
99
nents are functions of the
strain-components,
and the latter
satisfy
the six
equations
of
compatibility (25)
of Article
17,
viz. three
equations
of the
type
?"S' J. ?"
^v^
dy
df'~dydz'
and three of the
type
2
5^
=

(-
^
+
^-^
+
^-%"\
dydz
dx
V
9a;
dy
dz )
'
When account is taken of these
relations,
there are sufficient
equations
to
determine the stress.
Whenever the forces are such that the
stress-components
are either
constants or linear functions of the
coordinates,
the same is true of the
strain-components,
and the
equations
of
compatibility
are satisfied
identically.
We shall consider such cases in the
sequel.
In the
general
case,
the
problem may
in various
ways
be reduced to that
of
solving
certain
systems
of differential
equations.
One
way
is to
form,
by
the method described
above,
a
system
of
equations
for the
stress-components
in which account is taken of the identical relations between
strain-components.
Another
way
is to eliminate the
stress-components
and
express
the strain-
components
in terms of
displacements by using
the formulae
du dv
100
ELASTICITY OF ISOTROPIC SOLIDS
[CH.
Ill
which
depend
on the coefficients
d,
... in the formula
(12),
and also on the
quantities by
which the relative situations of the old and new axes are
determined. The results
might
be found more
rapidly by transforming
the
expression
2W
according
to the formulae of Article 12. The
general
result is that the elastic behaviour of a material has reference to certain
directions fixed
relatively
to the material.
If, however,
the elastic constants
are connected
by
certain
relations,
the formulae
connecting stress-components
with
strain-components
are
independent
of direction. The material is then
said to be
isotropic
as
regards elasticity.
In this case the function W is
invariant for all transformations from one set of
orthogonal
axes to another.
If we knew that there were no invariants of the
strain,
of the first or second
degrees, independent
of the two which were found in Article 13
(c),
we could
conclude that the
strain-energy-function
for an
isotropic
solid must be of
the form
i-^ K^xx
+
^yy
+
^zz)
+
5^ (fi yz
"^
^
zx"^
^
xy

^^yy^zz

^^zz^xx
~
^^xx^yy)-
We shall hereafter
(Chapter vi.) perform
the
transformation,
and
verify
that
this is the actual form of W.
At
present
we shall assume this form and deduce some of the
simpler
consequences
of it. It will be convenient to write
\+2fi
in
place
of A and
fi.
in
place
of B. We shall
suppose
the material to be
homogeneous,
so that
\ and
fi
are the same at all
points.
69. Elastic constants and moduluses of
isotropic
solids.
When W is
expressed by
the
equation
'
M
\yyz
I ^
zx
"T
6
xy ^^yy^zz
^^zz^xx
^^xx^yy)> \^' )
the
stress-components
are
given by
the
equations
Za;
=
XA-|-2/iea;a;, F,y
=
XA4-2/iey,
Z^
=
\A
-|-
2yae
JL
z
^
flCyz
, iitx
'^
M'^zx >
-^y^^ f^^xy
where A is written for
e^x
+
Syy
+
6^.
A
body
of
any
form
subjected
to the action of a constant
pressure p,
the
same at all
points
of its
surface,
will be in a certain state of stress. As we
have seen in Article
55,
this state will be
given by
the
equations
X,=
Yy=Z,
=
-p,
Y,= Z,
=
Xy=0.
According
to
equations (18),
the
body
is in a state of .strain such that
exx
=
eyy
=
e^=
-pl{S\
+
2fi),
^yz
^
^zx
'^
^xy^
"
The cubical
compression
is
p/{\
+
^/j,).
We write k
=
X+^fj,
(19)
Then k is the
quantity
obtained
by dividing
the measure of an uniform
.(18)
68-70]
EXPRESSED BY TWO CONSTANTS 101
pressure by
the measure of the cubical
compression produced by
it. It
is called the modulus
of compression.
Whatever the
stress-system may
be,
it can be
resolved,
as in Article
52,
into mean
tension,
or
pressure,
and
shearing
stresses on three
orthogonal
planes.
The mean tension is measured
by ^
(X^
+
Yy
+
Z^).
We learn
that the
quantity
obtained
by dividing
the measure of the mean tension
at a
point by
the measure of the cubical dilatation at the
point
is a constant
quantity

the modulus of
compression.
A
cylinder
or
prism
of
any
form,
subjected
to tension T which ia uniform
over its
plane
ends,
and free from traction on its lateral
surfaces,
will be in
a certain state of stress. As we have seen in Article 55 this state will be
given by
the
equations
Xx
=
J
,
Yy
=
Z2=Yl
=
Zx
=
Xy
=
0.
According
to
equations (18)
the
body
will be in a state of strain such that
We write E
=
''^p-'-\
(20)
X +
M
^
(21)
2(X,-I-
n)
Then E is the
quantity
obtained
by dividing
the measure of a
simple
longitudinal
tension
by
the measure of the extension
produced
by
it. It is
known as
Youngs
modulus. The number a is the ratio of lateral contraction
to
longitudinal
extension of a bar under terminal tension. It is known as
Poisson's ratio.
Whatever the
stress-system may
be,
the extensions in the directions of the
axes and the normal tractions across
planes
at
right angles
to the axes are
connected
by
the
equations
e^
=
A-'{Z,-(7(F,+.^,)!,
]
eyy
=
E-^{Yy-aiZ,+Xx)],
\
(2^)
e,,
=
E-^{Z,-<7{X,+
Yy)}.
Whatever the
stress-system
may
be,
the
shearing
strain
corresponding
with a
pair
of
rectangular
axes and the
shearing
stress on the
pair
of
planes
at
right angles
to those axes are connected
by
an
equation
of the form
Xy
=
^e^y
-(23)
This relation is
independent
of the directions
of the
axes. The
quantity
/m
is
called the
rigidity.
70. Observations concerning
the stress-strain
relations
in
isotropic
solids.
(a)
We
may
note the relations
Ea
E ,
E
,24)
'^
=
(T+^r^Ta)'
''
=
2a
+
<r)'
*-3(l-2<r)-
" '
102 ELASTICITY OF ISOTROPIC SOLIDS
[CH.
Ill
(6)
If <r were
>i,
i would be
negative,
or the material would
expand
under
pressure.
If (T were <

l,n
would be
negative,
and the function Wwould not be a
positive quadratic
function. We
may
show that this would also be the case if k were
negative*. Negative
values for o- are not excluded
by
the condition of
stability,
but such values have not been
found for
any isotropic
material.
(c)
The constant i: is
usually
determined
by e.xperiments
on
compression,
the
constant E sometimes
directly by experiments
on
stretching,
and sometimes
by experi-
ments on
bending,
the constant
fi usually by experiments
on torsion. The value of the
constant o- is
usually
inferred from a
knowledge
of two
among
the
quantities E, k, fiA
(d)
If
Cauchy's
relations
(13)
of Article 66 are
true,
\
=
^
and <t
=
J.
(e)
Instead of
assuming
the form of the
strain-energy-functiou,
we
might
assume
some of the relations between
stress-components
and
strain-components
and deduce the
relations
(18).
For
example J
we
may
assume
(i)
that the mean tension and the cubical
dilatation are connected
by
the
equation ^(X^i- Fy
+
Z,)='kA, (ii)
that the relation
A"y' =/iej.,.
holds for all
pairs
of
rectangular
axes of x" and
y.
From the second
assumption
we .should
find, by taking
the axes of
x, y,
z to be the
principal
axes of
strain,
that the
principal planes
of stress are at
right angles
to these axes. With the same choice of axes
we should then
find, by
means of the formulae of transformation of Articles 12 and
49,
that the relation
holds for all values of
^j,
... which
satisfy
the
equation
i[^2
+
li"*2
-(-
ni!-2
=
0.
It follows that we must have
JT;,
-
2^^^
=
r,
-
2,ie
=
if,
-
2^
.
Then the first
assumption
shows that each of these
quantities
is
equal
to
{k
-
fj)
A. The
relations
(18)
are thus found to hold for
principal
axes of
strain, and, by
a fresh
application
of the formulae of
transformation,
we
may prove
that
they
hold for
any
axes.
(/)
Instead of
making
the
assumptions just
described we
might
assume that the
principal planes
of stress are at
right angles
to the
principal
axes of strain and that the
relations
(22)
hold for
principal axes,
and we
might
deduce the relations
(18)
for
any
axes.
The
working
out of this
assumption may
serve as an exercise for the student.
{g)
We
may
show
that,
in the
problem
of the
compression
of a
body by pressure
uniform over its surface which was associated with the definition of
k,
the
displacement
is
expressed by
the
equations
n
_v _w_
p
X
y
z 3k'
(A)
We
may
.show
that,
in the
problem
of the bar stretched
by simple
tension T
which was associated with the definitions of E and
<r,
the
displacement
is
expressed by
the
equations
u_v_
,tT^
\T w T
^
(X+m)^
x~y~
E~
2^i(3X-f2^)'
z

"
/i (3X
-1-
2/1)
"
*
2ir
may
be written
(X
-(-
(e^
+
eyy
+
O"
+
?- { {'vv
-
e?
+
(e
-
e^^f
+
(e^
-
eyf}
+
/x
(,=
-I-
e,/
+
e^/).
t
Experiments
for the direct determination of Poisson's ratio have been made
by
P.
Cardani,
Fhy.
Zeitechr. Bd.
4, 1903,
and J.
Morrow,
Phil.
Mag. (Ser. 6),
vol. 6
(1903).
M. A.
Cornu,
Paris, C.R.,
t. 69
(1869),
and A.
Mallock,
Proc.
Hoy. Soc,
vol. 29
(1879)
determined a
by
experiments
on
bending.
J
This is the method of Stokes. See
Introduction,
footnote 37.

A
displacement
which would be
possible
in a
rigid body may
be
superposed
on that
given
in
the text. A like remark
applies
to the Observation
(h).
Cf. Article
18, supra.
70-72]
EXPRESSED BY TWO CONSTANTS 103
71.
Magnitude
of elastic constants and moduluses of some
isotropic
solids.
To
give
an idea of the order of
magnitude
of the elastic constants and
moduluses of some of the materials in
everyday
use a few of the results of
experiments
are tabulated here. The table
gives
the
density (p)
of the
material as well as the elastic-
constants,
the constants
being expressed
as
multiples
of an unit stress of one
dyne per square
centimetre. Poisson's
ratio is also
given.
The results marked
"
E
"
are taken from J. D. Everett's
Illustrations
of
the G.O.S.
system of
units, London, 1891,
where the authorities
for them will be found. Those marked "A" are reduced from results of more
recent researches recorded in a
paper by Amagat
in the Journal de
Physique
(Ser. 2),
t. 8
(1889).
It must be understood that considerable differences are
found in the elastic constants of different
samples
of
nominally
the same
substance,
and that such a
designation
as
"steel,"
for
example,
is far from
being precise.
Material
10*
ELASTICITY OF SOLIDS
[CH.
Ill
These
equations may
be
solved,
so as to
express
the
strain-components
in terms of the
stress-components.
If II denotes the determinant of the
quantities
c,
and
(7
denotes the minor determinant that
corresponds
with
Crg,
so that
Il=CnGri
+
CnCr2
+
Cr3Gr3
+
'CriCr,
+
CnGr.s
+
CnCr6, (26)
the
equations
that
give
the
strain-components
in terms of the stress-
components
can be written
'^l
\^xxi ^yy>
**>
^t/z>
^zxt
^xy)
=
(
Oil c7,2 (7,3 (7i4 l7io (7i6
){X-i,, Yy,
Z^, Y^, Zx,
Xy),
Cn
.(27)
where

=
C
(r,
s
=
1, 2,
...
6).
The
quantities ^Cn,

Ci,,
... are the coefficients of a
homogeneous
quadratic
function of
exx,
This function is the
strain-energy-function
expressed
in terms of
strain-components.
The
quantities ^GiJU,
...
Oij/II,
... are the coefficients of a
homogeneous
quadratic
function of
X^
This function is the
strain-energy-function
expressed
in terms of
stress-components.
73. Moduluses of
elasticity.
We
may
in various
ways
define
types
of stress and
types
of strain.
For
example, simple
tension
[X^], shearing
stress
[Fz],
mean tension
[J (Xx
+
Yy+ Z^)]
are
types
of stress. The
corresponding types
of strain are
simple
extension
[exx], shearing
strain
[e,jz],
cubical dilatation
[exx
+
^yy
+ ^
J.
We
may express
the strain of
any
of these
types
that
accompanies
a stress
of the
corresponding type,
when there is no other
stress,
by
an
equation
of the
form
stress
=
Mx
(corresponding strain).
Then M is called a
"
modulus of
elasticity."
The
quantities
H/G^ ,
II/C^
are
examples
of such moduluses.
The modulus that
corresponds
with
simple
tension is known as
Young's
modulus for the direction of the related tension. The modulus that corre-
sponds
with
shearing
stress on a
pair
of
orthogonal planes
is known as the
rigidity
for the related
pair
of directions
(the
normals to the
planes).
The
modulus that
corresponds
with mean tension or
pressure
is known as the
modulus
of compression.
We shall
give
some
examples
of the calculation of moduluses.
(a)
Modulus
of compression.
We have to assume that
Xx=
Yy=Z,,
and the
remaining stress-components
vanish
;
the
corresponding
strain is cubical
dilatation,
and we must therefore calculate
e^ +
eyy
+
e.
We find for the modulus the
expression
n/(6'-)-Cj2
+
C3,,-|-2C,3-(-2(73,
+
2C) (28)
72, 73]
WHICH ARE NOT ISOTROPIC 105
As in Article
68,
we see that the cubical
compression produced
in a
Ixxly
of
any
fonn
by
the
application
of uniform normal
pressure, p,
to its surface is
pjk,
where k
now.
denotes
the above
expression (28).
{b) Rigidity.
We
may suppose
that all the
stress-components
vanish
except Y
and then we have
^ey,= C^^Y^,
so
th^t 'n|C^^
is the
rigidity corresponding
with the
pair
of directions
y,
z.
If the
shearing
stress is related to the two
orthogonal
directions
{I, m, n)
and
{V, m', n'),
the
rigidity
can be shown to be
expressed by
nH-(Cn, C.^, ...Oi2,...)(2U', imm', 27m',
mn'
+
m'n,
nl' +
n'l,
In'
+ l'viY, ...(29)
where the denominator is a
complete quadratic
function of the six
arguments
2ll\
... with
coefficients
Cj,, C22,
...
(c) Younff's
modulus and Poisson's ratio.
We
may suppose
that all the
stress-components
vanish
except Xj,
and then we have
ne^x=CjiXx,
so that
Tl/Cn
is the
Young's
modulus
corresponding
with the direction x.
In the same case the Poisson's ratio of the contraction in the direction of the axis of
y
to the extension in the direction of the axis of .r is
-
C,2/C,,
. The value of Poisson's ratio
depends
on the direction of the contracted transverse linear elements as well as on that of
the extended
longitudinal
ones.
In the
general
ease we
may
take the stress to be tension
X'^
across the
planes
.r:*
=
const.,
of which the normal is in the direction
{I,
m,
n).
Then we have
X,=IKX',., Yy
=
mKX'^, Z.
=
nKX'r,
Yi=mnX':c', Z^='nlX'x', Xy
=
lmX'x',
and we have also
it follows that the
Young's
modulus E
corresixsnding
with this direction is
n-=-(C, Cj.^,
...
C,2, ...){P, m\ n\ mn, nl, Imf, (30)
where the denominator is a
complete quadratic
function of the six
arguments P,
... with
coefficients
Cj,,
....
If
(?', m', n')
is
any
direction at
right angles
to
x',
the
contraction, -<",,
in this
direction is
given by
the
equation
'
=
^xxl'^
+
euym"^
+et'^
+
ey^m'n'+e^^n'l'
+
e,/m',
and the
corresponding
Poisson's ratio a is
expressible
in the form
2(^|_ d{P) d{ni') (n^) d{nm) d{nl) (lm)J
where
cjt
is the above-mentioned
quadratic
function of the
arguments
P, ...,
and the
differential coefficients are formed as if these
arguments
were
independent.
It
may
be
observed that
a/E
is related
symmetrically
to the two directions
in which the
corresponding
contraction and extension occur.
If we construct the surface of the fourth order of which the
equation
is
(Ci(722,
...
C12, ...)(^^y^^^y^,
2.1-, ^y)'
=
const., (32)
then the radius vector of this surface in
any
direction is
proportional
to the
positive
fourth root of the
Young's
modulus of the material
corresponding
with that direction*.
*
The result is due to
Cauchy,
Exercices de
ilatMmatiques,
t. 4
(1829), p.
30.
106
STRESS DUE TO VARIATION OF TEMPERATURE
[CH.
Ill
74. Thermo-elastic
equations.
The
application
of the two fundamental
laws of
Thermodynamics
to the
problem
of
determining
the .stress and strain in elastic solid bodies when
variations of
temperature
occur has been discussed
by
Lord Kelvin*. The
results at which he arrived do not
permit
of the formulation of a
system
of
differential
equations
to determine the state of stress in the
body
in the
manner
explained
in Article 67.
At an earlier date
Duhamelf
had obtained a set of
equations
of the
required
kind
by developing
the
theory
of an elastic solid
regarded
as a
system
of material
points,
and F. E.
Neumann,
starting
from certain
assumptions :J:,
had arrived at the .same
system
of
equations.
These
assumptions may,
when the
body
is
isotropic,
be
expressed
in the follow-
ing
form :

The
stress-system
at
any point
of a
body
strained
by
variation
of
temperature
consists of two
superposed stress-systems.
One of these is
equivalent
to uniform
pressure,
the same in all directions round a
point,
and
proportional
to the
change
of
temperature
;
the other
depends upon
the
strain at the
point
in the same
way
as it would do if the
temperature
were
constant.
These
assumptions
lead to
equations
of the form
dXx
dXy
dZx
y
dd ,
.
-d^
+
W
'^
'^
^^S^'
^ ^
where
/9
is a constant coefficient and is the excess of
temperature
above
that in the unstrained state. The
stress-system
at
any point
has com-
ponents
-^9+x,,
-^e+Y,,
-^e
+
z,
(
Fj, , Zx ,
Xy,
I
in which
Xx,
... are
expressed
in terms of
displacements by
the formulae
(18)
of Article 69. The
equations
are
adequate
to determine the
displacements
when 6 is
given.
When 6 is not
given
an additional
equation
is
required,
and this
equation may
be deduced from the
theory
of conduction of
heat,
as
was done
by
Duhamel and Neumann.
The
theory
thus arrived at has not been
very
much
developed.
Attention
has been directed
especially
to the fact that a
plate
of
glass
strained
by
unequal heating
becomes
doubly refracting,
and to the
explanation
of this
effect
by
the
inequality
of the stresses in difTerent directions. The reader who
wishes to
pursue
the
subject
is referred to the
following
memoirs in addition
to those
already
cited:

C. W.
Borchardt,
Berlin
Monatsberichte, 1873;
*
See
Introduction,
footnote 43.
t Paru,
iUm....par
divers
savans,
t. 5
(1838).
i
See his
Vorlenungen
Uber die Theorie der Elatticitat der
festen K'orper, Leipzig,
188-5,
and
ct. the memoir
by
Maxwell cited in Article 57 footnote.
74,
75]
INITIAL STRESS
107
J.
Hopkinson,
Messenger of
Math. vol. 8
(1879);
Lord
Rayleigh,
Phil.
Mag. (Ser.
6)
vol. 1
{1%Q\)
=
Scimtific Papers,
vol.
4,
p. 502;
E.
Al'mansi,
Tonno
Atti,
t. 32
(1897);
P.
Alibrandi,
Giornale di matetn. t. 38
(1900).
It must be observed that the elastic "constants" themselves are functions
of the
temperature.
In
general,
they
are
diminished
by
a rise of
tempera-
ture
;
this result has been
established
by
the
experiments
of
Wertheim*.
Kohlrauschf
and Macleod and Clarke
J.
75. Initial stress.
The initial state of a
body may
be too far removed from the unstressed
state to
permit
of the stress and strain
being
calculated
by
the
principle
of
superposition
as
explained
in Article 64. Such initial states
may
be induced
by processes
of
preparation,
or of
manufacture,
or
by
the action of
body
forces. In cast iron the exterior
parts
cool more
rapidly
than the
interior,
and the
unequal
contractions that
accompany
the
unequally rapid
rates of
cooling give
rise to considerable initial stress in the iron when cold. If a
sheet of metal is rolled
up
into a
cylinder
and the
edges
welded
together
the
body
so formed is in a state of initial
stress,
and the unstressed state cannot
be attained without
cutting
the
cylinder open.
A
body
in
equilibrium
under the mutual
gravitation
of its
parts
is in a state of
stress,
and when
the
body
is
large
the stress
may
be enormous. The Earth is an
example
of
a
body
which must be
regarded
as
being
in a state of initial
stress,
for the
stress that must exist in the interior is much too
great
to
permit
of the
calculation,
by
the
ordinary
methods,
of strains reckoned from the unstressed
state as
unstrained state.
If a
body
is
given
in a state of initial
stress,
and is
subjected
to
forces,
changes
of volume and
shape
will be
produced
which can be
specified by
a
displacement
reckoned from the
given
initial state as unstrained state.
We
may specify
the initial stress at a
point by
the
components
^x >
-^
y
1 ^z ,
-J
z , ^x >
-^y
>
and we
may specify
the stress at the
point
when the forces are in action
by
Xa;"" +Xa;',
In like manner we
may specify
the
density
in the initial
state
by p
and that in the strained state
by p
+
p,
and we
may specify
the
body
force in the initial state
by (Xo,
F,, Z^)
and that in the strained state
by (Z
+
X', F
+
y, Z^
+
Z').
Then the conditions of
equilibrium
in the
initial .state are three
equations
of the
type
^+^^4^
+
-^
+PZ
=
(35)
ox
oy
oz
*
Ann. de
Chimie,
t. 12
(1844).
t
Ann.
Phys.
Client.
(Poggendorff),
Bd. 141
(1870).
t
A result obtained
by
these writers is
explained
in the sense stated in the text
by
Lord Kelvin
in the Article
'Elasticity'
in
Ency.
Brit,
quoted
in the footnote to Article 65.
108
INITIAL STRESS
[CH.
Ill
and three
boundary
conditions of the
type
Z^<>
cos
(x, V,)
+
Xy
<'
cos
(y,
v)
+
ZJ<
cos
(2,
,/)
=
(36)
in which
i',
denotes the direction of the normal to the initial
boundary.
The conditions of
equilibrium
in
the
strained state are three
equations
of
the form
l(Z..o.
+
z.')+lwo.
+
z,')
+
^(Z,m+Z,')+ip
+
p')(Xo
+
X')
=
(37)
and three
boundary
conditions of the
type
(Z,''
+
X^')
cos
(x, v)
+
(Z<
+
Xy')
cos
(y, v)
+
(Z^<>+^/)cos(^, i/)
=
Z .(38)
in which
{X^,
Y,,
ZJ)
is the surface traction at
any point
of the
displaced
boundary.
These
equations may
be
transformed,
when the
displacement
is
small,
by using
the results
(35)
and
(36),
so as to become three
equations
of
the
type
dXJ
j_
dXy
_
+
- ^
+
^
+
pZ'
+
p'A'
=
.(39)
.(40)
dx
dy
and three
boundary
conditions of the
type
Xx
cos
{x, v)
+
Xy
cos
(y, v)
+ Z^
cos
{z, v)
=
Xy

Xa;""
{cos
(x,
v)

COS
{x, I'o)}
-
Zj," {cos iy,
v)
-
cos
{y, v^)}
-
Za;<'
{cos {z, v)
-
cos
{z, I/)].
If the initial stress is not known the
equations (3-5)
and conditions
(36)
are not sufficient to determine
it,
and no
progress
can be made. If the
initial stress is known the determination of the additional stress
{XJ, ...)
cannot be effected
by
means of
equations (39)
and conditions
(40),
without
knowledge
of the relations between these
stress-components
and the dis-
placement.
To obtain such
knowledge
recourse must be had either to
experi-
ment or to some more
general theory. Experimental
evidence
appears
to be
entirely wanting*.
Cauchyf
worked out the
consequences
of
applying
that
theory
of material
points
to which reference has been made in Article 66. He found for
Xx\
expressions
of the form
'du dv
\dx
dy
X'
=
Z
<
aI)
+
^^.-|-^^-"s-^."'
"'
'y
dy^'^'
dz
'
dx^'^" dx^^" a^^^
.(41)
*
Beference
may
be made to a
paper by
F. H.
Cilley,
Amer. J.
of
Science
{Silliman), (Ser. 4),
vol. 11
(1901).
t
See Introduction and cf. Note B at the end of this book.
75]
INITIAL STRESS
109
where
{u, v, w)
is the
displacement
reckoned from the initial
state,
and
{Xx, ...)
is a
stress-system
related to this
displacement by
the same
equations
as would hold if there were no initial stress. In the case of
isotropy
these
equations
would be
(18)
of Article 69 with X
put equal
to
/i.
It
may
be observed that the terms o{
XJ,
... that contain
X,,''",
... arise from
the
changes
in the distances between
Cauchy's
material
points,
and from
changes
in the directions of the lines
joining
them in
pairs,
and these
changes
are
expressed by
means of the
displacement {u,
v,
w).
Saint-Venant* has obtained
Cauchy's
result
by adapting
the method of
Green,
that is to
say by
the use of the
energy-function.
His deduction has
been criticized
by
K. Pearson
+,
and it cannot be
accepted
as valid. Green's
original discussion^ appears
to be restricted to the case of uniform initial
stress in an unlimited elastic
medium,
and the same restriction characterizes
Lord Kelvin's discussion of Green's
theory.
*
J. de Math.
{Liouville), (Ser. 2),
t. 8
(1863).
t
Todhunter and Pearson's
History,
vol.
2, pp. 84,
85.
+
See the
paper quoted
in the
Introduction,
footnote 81.

Baltimore Lectures on Molecular

Dynamics
and the Wave
Theory of Light,
London, 1904,
pp.
228 et
seq.
CHAPTER IV.
THE RELATION
BETWEEN THE MATHEMATICAL THEORY OF
ELASTICITY AND TECHNICAL MECHANICS.
76. Limitations
of the mathematical
theory.
The
object
of this
Chapter
is to
present
as clear an idea as
possible
of the
scope
and
limitations of the mathematical
theory
in its
application
to
practical questions.
The
theory
is worked out for bodies strained
gradually
at a constant
temperature,
from an initial state of no stress to
a final state which differs so little from the unstressed state that
squares
and
products
of the
displacements
can be
neglected
;
and further it is worked
out on the basis of Hooke's
Law,
as
generalized
in the statements in
Article 64. It is known that
many
materials used in
engineering
structures,
e.g.
cast
iron,
building
stone, cement,
do not
obey
Hooke's Law for
any
strains that are
large enough
to be observed. It is known also that those
materials which do
obey
the law for small measurable strains do not
obey
it for
larger
ones. The statement of the law in Article 64 included the
statement that the strain
disappears
on removal of the
load,
and this
part
of it is
absolutely
necessary
to the mathematical
theory ;
but it is known
that the limits of
strain,
or of
load,
in which this condition holds
good
are
relatively
narrow.
Although
there exists much
experimental knowledge*
in
regard
to the behaviour
of bodies which are not in the conditions to
which the mathematical
theory
is
applicable, yet
it
appears
that the
ap-
propriate
extensions
of the
theory
which would be needed in order to
incorporate
such
knowledge
within it cannot be made until much fuller
experimental
knowledge
has been obtained.
*
Information
in
regard
to
experimental
results will be found in treatises on
Applied
Mechanics. The
following may
be mentioned :

W. J. M.
Bankine,
Applied Mechanics,
Ist
edition, London, 1858, (there
have been numerous later
editions) ;
W. C.
Unwin,
The
testing of
materials
of construction,
London, 1888;
J. A.
Ewing,
The
Strength of Materials, Cambridge, 1899;
Flamant,
Stabilite des
constructions,
liesistance des
matiriaux, Paris, 1896;
C.
Bach,
Elasticitdt
und
Festigkeit,
2nd
edition,
Berlin,
1894
;
A.
Foppl, Vorlesungen
Hber technische
Mechanik,
Bd.
3, Festigkeitslehre, Leipzig,
1900.
Very
valuable
experimental
researches were made in
recent times
by
J.
Bauschinger
and recorded
by
him in
Mittheilungen
aus dem mechanisch-
technischen Laboratorium...in
Miinchen;
these researches have been continued
by
A.
Foppl.
New
facts in
regard
to the nature of
permanent
set in
metals,
which are
likely
to
prove
to be
very
important,
have been
brought
to
light by
J. A.
Ewing
and W.
Roseuhain,
Phil. Tram.
Roy.
Soc.
(Ser. A),
vols.
193, 195, (1900, 1901).
76,
77]
LIMITATIONS OF THE THEOKY
111
The restriction of the
theory
to conditions in which the strain
disappears
on removal of the load is
usually expressed by saying
that the
body
must
be strained within the limits of
"
perfect elasticity."
The restriction to
conditions in which the measurable strain is
proportional
to the load is
sometimes
expressed by saying
that the
body
must be strained within the
limits of
"
linear
elasticity."
The
expression
"
limits of
elasticity
"
is used
sometimes in one of these senses and sometimes in the
other,
and the limits
are sometimes
specified by
means of a
"
stress
"
or a
"
traction,"
i.e.
by
a load
per
unit of
area,
and sometimes
by
the measurable strain.
When the strain does not
disappear
after removal of the
load,
the strain
which remains when the load is removed is called
"set,"
and the excess
of the strain which occurs under the load above the set is called
"
elastic
strain." The strain is then
compounded
of set and elastic strain. A
body
which can be strained without
taking any
set is sometimes said to be in
a "state of ease"
up
to the strain at which set
begins.
77. Stress-strain
diagrams.
One of the
greatest
aids to scientific
investigation
of the
properties
of matter
subjected
to stress is the use of these
diagrams. They
are
usually
constructed
by taking
the strain
developed
as
abscissa,
and the stress
producing
it as the
corresponding
ordinate. For most materials the case
selected for this kind of treatment is the extension of
bars, and,
in the
diagram,
the ordinate
represents
the
applied
traction,
and the abscissa the
extension of a line traced ou the bar
parallel
to its
length
and rather near
the middle. The extension is measured
by
some kind of extensometer*.
The load at
any
instant is
known,
and the traction is estimated
by assuming
this load to be distributed
uniformly
over the area of the cross-section of the
specimen
in the initial state. If
any
considerable contraction of the section
were to occur the traction would be underestimated. The
testing
machine,
by
means of which the
experiments
are
made,
is sometimes fitted with an
automatic
recording apparatusf by
which the curve is
drawn;
but this
cannot be done
satisfactorily
with some
types
of
machine^.
It is clear
that,
in
general,
the
quantities
recorded
by
such
arrangements
are the
traction,
estimated as
stated,
and the extension which it
produces
immediately. Special
methods of
experimenting
and
observing
are
required
if elastic strain is to be
distinguished
from
set,
and if the various etfects that
depend upon
time are to be calculated.
The
general
character of the curve for
moderately
hard metals under
extension,
is now well known. It is for a considerable
range
of stress
very
*
Several kinds of extensometers are described
by Ewing
and Unwin.
t Unwin,
loc. cit,
X Bausohinger, Mittheilungen,
xx.
(1891).
112 STRESS-STKAIN DIAGRAMS
[CH.
IV
nearly straight.
Then comes a
stage
in which the curve is
generally
concave
downwards,
so that the strain increases faster than it would do if it were
proportional
to the traction
;
in this
stage
the strain is
largely
a
permanent
set. As the traction increases there comes a
region
of well-marked dis-
continuity,
in which a small increase of traction
produces
a
large
increase
of set. The traction at the
beginning
of this
region
is called the Yield-Point.
After a further considerable increase of traction the bar
begins
to thin down
at some
section,
determined
apparently by
accidental
circumstances,
and there
it
ultimately
breaks. When this local
thinning
down
begins
the load is
usually
eased off somewhat before
rupture
occurs,
and the bar breaks with
less than the maximum traction. The maximum traction before
rupture
is
called the
"breaking
stress" of the
material,
sometimes also the "ultimate
strength
"
or
"
tenacity."
Fig.
8.
Fig.
9.
Figure
8 shows the character of the
diagram
for
"
weld iron." It is
reduced from one of
Bauschinger's
curves. Similar
diagrams
for mild steel
are drawn in
many
books. A is the limit of linear
elasticity ;
between A and
B the strain increases rather faster than between and A and at a
varying
rate,
B is the
yield-point
and D
represents
the maximum traction.
Fig.
9
is reduced from one of
Bauschinger's
curves for cast iron. There is no
sensible
range,
and so no
limit,
of linear
elasticity,
and no
yield-point.
Diagrams may
be constructed in the same
way
for thrust and
contraction,
but the forms of them are in
general
different from the above. In
particular,
there is no
yield-point
under thrust. In the case of cast iron it has been
verified that the curve is continuous
through
the
origin,
where there is an
inflexion*.
*
See
e.g. Ewing,
loc.
cit., p.
31.
77,
78]
LIMITS OF ELASTICITY
113
78. Elastic limits.
The
diagrams
do not show the limits of
perfect
elasticity
when these
are different from the limits of linear
elasticity.
These limits
usually
are
different,
and the former are lower than the latter*. The
numerical
measures of the limits for extension and contraction are
usually
different
for the same
specimen.
The limits are not
very
well
defined. The limit
of
perfect elasticity
for
any type
of stress would be determined
by
the
greatest
traction which
produces
no
set,
but all that
experiment
can tell us
is the smallest traction for which set can be measured
by
means of our
instruments. The limits of linear
elasticity
are shown
by
the
diagrams,
but
they
are liable to the same kind of
uncertainty
as the limits of
perfect
elasticity,
inasmuch as the determination of them
depends upon
the
degree
of
accuracy
with which the
diagrams
can be drawn.
The limits of linear
elasticity
can be raised
by
overstraint. If a bar of
steel,
not
specially
hard,
is
subjected
to a load above the elastic
limit,
and
even above the
yield-point,
and this load is maintained until a
permanent
state is
reached,
it is found afterwards to
possess
linear
elasticity up
to a
higher
limit than before. If the load is
removed,
and the bar remains for
some time
unloaded,
the limit is found to be raised still
further,
and
may
be
above the load which
produced
the overstrain.
On the other
hand,
the limits of
elasticity
can be lowered
by
overstrain
|.
If a bar of iron or mild steel is
subjected
to a load above the
yield-point,
and then unloaded and
immediately
reloaded,
its
elasticity
is found to be
very
imperfect,
and the limit of linear
elasticity very
low
;
but if the bar remains
unloaded for a few
days
it is found to have recovered
partially
from the effects
of the
previous overstraining,
and the
longer
the
period
of rest the more com-
plete
is the
recovery. Wrought
iron recovers much more
rapidly
thau steel.
In the case of cast
iron,
not
previously subjected
to
tests,
any
load that
produces
a measurable strain
produces
some
set,
and there is no
appreciable
range
of linear
elasticity.
After several times
loading
and
unloading,
the
behaviour of the metal
approaches
more
closely
to that of other metals
as
exemplified
in
Fig.
8. These results
suggest
that the set
produced
in
the first tests consists in the removal of a state of initial stress.
The
yield-point
also is raised
by
overstrain,
if the
original
load is above
the
original yield-point,
and the amount
by
which it is raised is increased
by allowing
a
period
of rest
;
it is increased still more
by maintaining
constant
the load which
produced
the
original
overstrain. This effect is described as
"
hardening by
overstrain."

Bausohinger, Mittheilungen,
xiii.
(1886).
t
Ibid.
t
See
e.g. Ewing,
loc.
cit., pp.
33 et
seq.,
and the tables in
Bausohinger's Mittheilungen,
xin.
L. E.
8
114 PLASTICITY AND
[CH.
IV
The
following
table*
gives
some
examples
of the limit of linear
elasticity
and the
yield-point
for some kinds of iron. The
results,
given
in
atmospheres,
ai-e in each case those for a
single specimen
not
previously
tested.
Metal
78-80]
VISCOSITY OF SOLIDS 115
of solids to flow under
great
stress is called
plasticity.
A solid is said to be
"
hard
"
when the force
required
to
produce
considerable set is
great,
"
soft
"
or
"
plastic
"
when it is small. A substance must be termed
"
fluid
"
if con-
siderable set can be
produced by any
force,
however
small,
provided
it is
applied
for a sufficient time.
In
experiments
on extension some
plasticity
of the material is shown
as soon as the limit of linear
elasticity
is exceeded*. If the load
exceeding
this limit is removed some set can be
observed,
but this set diminishes at
a rate which itself diminishes. If the load is maintained the strain
gradually
increases and reaches a constant value after the
lapse
of some time. If the
load is removed and
reapplied
several
times,
both the set and the elastic
strain increase. None of these effects are observed when the load is below
the limit of linear
elasticity.
The
possibility
of these
plastic
effects tends
to
complicate
the results of
testing,
for if two like
specimens
are loaded at
different
rates,
the one which is loaded more
rapidly
will show a
greater
breaking
stress and a smaller ultimate extension than the other. Such
differences have in fact been
observedf,
but it has been
shownj
that
under
ordinary
conditions of
testing
the variations in the rate of
loading
do not affect the results
appreciably.
80.
Viscosity
of Solids.
"
Viscosity
"
is a
general
term for all those
properties
of matter in virtue
of which the
resistance,
which a
body
offers to
any change, depends upon
the
rate at which the
change
is effected. The existence of viscous resistances
involves a
dissipation
of the
energy
of the
substance,
the kinetic
energy
of
molar motion
being
trausfornied,
as is
generally supposed,
into kinetic
energy
of molecular
agitation.
The most marked effect of this
property,
if it exists
in the case of elastic
solids,
would be the subsidence of vibrations set
up
in
the solid.
Suppose
a solid of
any
form to be
struck,
or otherwise
suddenly
disturbed. It will be thrown into more or less
rapid
vibration,
and the
stresses
developed
in it
would,
if there is
genuine viscosity, depend partly
on
the
displacements,
and
partly
on the rates at which
they
are effected. The
parts
of the stresses
depending
on the rates of
change
would be viscous
resistances,
and
they
would
ultimately
destroy
the
vibratory
motion. Now
the
vibratory
motion of elastic solid bodies is
actually destroyed,
but the
decay appears
not to be the effect of viscous resistances of the
ordinary type,
that is to
say
such as are
proportional
to the rates of strain. It has been
pointed
out
by
Lord
Kelvin
that,
if this
type
of resistance alone were
involved,
the
proportionate
diminution
of the
amplitude
of the
oscillations
*
Bauschinger, Mittheilungen,
xiii.
(1886).
t
Cf.
Unwin,
loc.
cit., p.
89.
t Bauschinger, Mittheilungen,
xx.
(1891).

Sir W.
Thomson,
Article
'Elasticity,' Ency.
Brit, or Math, and
Phys. Papers,
vol.
3,
Cambridge, 1890, p.
27.
82
116 CHANGES OF
QUALITY
[CH.
IV
per
unit of time would be
inversely proportional
to the
square
of the
period ;
but a series of
experiments
on the torsional oscillations of wires showed that
this law does not hold
good.
Lord Kelvin
pointed
out that the
decay
of vibrations could be accounted
for
by supposing
that,
even for the
very
small strains involved in
vibratory
motions,
the effects of elastic
after-working
and
plasticity
are not
wholly
absent. These
effects,
as well as viscous resistances of the
ordinary type,
are
included in the class of
hysteresis phenomena.
All of them show that the
state of the
body
concerned
depends
at
any
instant on its
previous
states as
well as on the external conditions
(forces, temperature,
&c.)
which obtain at
the instant.
Hysteresis always implies irreversibility
in the
sequence
of
states
through
which a
body passes,
and is
generally
traced to the molecular
structure of matter.
Accordingly,
theories of molecular action have been
devised
by
various
investigators*
to account for
viscosity
and elastic after-
working.
81.
iEolotropy
induced
by permanent
set.
One of the
changes produced
in a
solid,
which has received a
permanent
set,
may
be that the
material,
previously isotropic,
becomes
seolotropic.
The
best known
example
is that of a bar rendered
solotropic by permanent
torsion.
Warburgf
found
that,
in a
copper
wire to which a
permanent
twist
had been
given,
the elastic
phenomena
observed could all be
explained
on the
supposition
that the substance of the wire was rendered
ajolotropic
like a rhombic
crystal.
When a
weight
was
hung
on the wire it
produced,
in
addition to
extension,
a small
shear,
equivalent
to a
partial untwisting!
of
the wire
;
this was an elastic
strain,
and
disappeared
on the removal of the
load. This
experiment
is
important
as
showing
that
processes
of manu-
facture
may
induce considerable
aBolotropy
in materials which in the un-
worked
stage
are
isotropic,
and
consequently
that estimates of
strength,
founded on the
employment
of the
equations
of
isotropic elasticity,
cannot be
strictly interpreted .
82.
Repeated loading.
A
body
strained within its elastic limits
may
be strained
again
and
again
without
receiving any injury
;
thus a
watch-spring may
be coiled and
*
The
following maybe
mentioned:

J. C.
Maxwell,
Article 'Constitution of
Bodies,' Ency.
Brit, or
Scientific Papers,
vol.
2, Cambridge, 1890;
J. G.
Butcher,
London Math. Sac.
Proc,
vol. 8
(1877)
;
0. E.
Meyer,
J.
f.
Math.
(Crelle),
Bd. 78
(1874)
;
L.
Boltzmann,
Ann.
Phys.
Chem.
{Poggendorjf), Ergzgsbd.
7
(1878).
For a
good
account of the theories the reader
may
be
referred to the Article
by
F. Braun in Wiukelmann's Handbuch der
Physik,
Bd. 1
(Breslau, 1891),
pp.
321

342. For a more recent discussion of the

viscosity
of metals and
crystals,
see W.
Voigt
Ann.
Phys.
Chem.
(Wiedonann),
Bd. 47
(1892).
t Ann.
Phys.
Chem.
(Wiedemann),
Bd. 10
(1880).
t
Cf. Lord
Kelvin,
loc.
cit.,
Math, and
Phys. Papers,
vol.
3, p.
82.

Cf.
Unwin,
loc.
cit., p.
25.
80-83]
PRODUCED BY
OVERSTRAIN
117
uncoiled one hundred and
twenty
millions of times a
year
for several
years
without deterioration. But it is different when a
body
is
strained
repeatedly
by rapidly varying
loads which exceed the limits of
elasticity.
Wohler's*
experiments
on this
point
have been held to show that the resistance of
a
body
to
any
kind of
deformation can be
seriously
diminished,
by rapidly
repeated applications
of a load. The result
appears
to
point
to a
gradual
deterioration
t
of the
quality
of the material
subjected
to
repeated loading,
which can be verified
by
the observation that after a
large
number of
appli-
cations and removals of the
load,
bars
may
be broken
by
a stress much
below the statical
breaking
stress.
Bauschinger:[:
made several
independent
series of
experiments
on the
same
subject.
In these the load was reversed 100 times a
minute,
and
the
specimens
which endured so
long
were submitted to some millions of
repetitions
of
alternating
stress. In some cases these severe tests revealed
the existence of flaws in the
material,
but the
general
result obtained was
that the
strength
of a
piece
is not diminished
by repeated loading, provided
that the load
always
lies within the limits of linear
elasticity.
An
analogous property
of bodies is that to which Lord
Kelvin
has called
attention under the name
"
fatigue
of
elasticity."
He observed that the
torsional vibrations of wires subsided much more
rapidly
when the wires
had been
kept vibrating
for several hours or
days,
than
when,
after
being
at rest for some
days, they
were set in vibration and
immediately
left to
themselves.
Experimental
results of this kind
point
to the
importance
of
taking
into
account the manner and
frequency
of the
application
of force to a structure
in
estimating
its
strength.
83.
Hypotheses concerning
the conditions of
Rupture.
Various
hypotheses
have been advanced as to the conditions under which
a
body
is
ruptured,
or a structure becomes unsafe. Thus
Lamd|| supposed
it
to be
necessary
that the
greatest
tension should be less than a certain limit.
PonceletlT,
followed
by
Saint-
Venant**,
assumed that the
greatest
extension
must be less than a certain limit. These measures of
tendency
to
rupture
agree
for a bar under
extension,
but in
general they
lead to different limits
*
Ueher
Festigkeitsversuche
mil Eisen unci
St<M, Berlin,
1870. An account of Wohler's
experiments
is
given by Unwin,
loc.
cit., pp.
356 et
seq.
t A different
explanation
has been
proposed by
K.
Pearson, Messenger of
Math. vol. 20
(1890).
J Mitthcilungen,
xx.
(1891)
and xxv.
(1897)
edited
by FQppl.

Loc.
cit..
Math, and
Pliys. Papers,
vol.
3, p.
22.
II
See
e.g.
the memoir of Lamfe and
Clapeyron, quoted
in the Introduction
(footnote 39).
IT See Todhunter and Pearson's
History,
vol.
1,
art. 995.
**
See
especially
the
Historique Abre'ge
in Saint-Venant's edition of the
L^fOfw
de
Navier,
pp.
cxcix

ccv.
118 HYPOTHESES CONCERNING THE
[CH.
IV
of safe
loading*. Again,
Tresca followed
by
G. H.
Darwinf
makes the
maximum difference of the
greatest
and least
principal
stresses the measure
of
tendency
to
rupture,
and not a
very
different limit would be found
by
following
Goulomb'st
suggestion,
that the
greatest
shear
produced
in the
material is a measure of this
tendency.
An
interesting
modification of this
view has been
suggested
and worked out
geometrically by
0.
Mohr,
It
would enable us to take account of the
possible dependence
of the condition
of
safety upon
the nature of the
load,
i.e.
upon
the kind of stress which
is
developed
within the
body.
The manner and
frequency
of
application
of
the load are matters which
ought
also to be taken into account. The con-
ditions of
rupture
are but
vaguely
understood,
and
may depend largely
on
these and other accidental circumstances. At the same time the
question
is
very importaut,
as a
satisfactory
answer to it
might suggest
in
many
cases
causes of weakness
previously unsuspected,
and,
in
others,
methods of econo-
mizing
material that would be consistent with
safety.
In all these
hypotheses
it is
supposed
that the stress or strain
actually
produced
in a
body
of
given
form,
by
a
given
load,
is somehow calculable.
The
only
known method of
calculating
these effects is
by
the use of the
mathematical
theory
of
Elasticity,
or
by
some more or less
rough
and
ready
rule obtained from some result of this
theory. Suppose
the
body
to be
subject
to a
given system
of
load,
and
suppose
that we know how to solve the
equations
of elastic
equilibrium
with the
given boundary-conditions.
Then
the stress and strain at
every point
of the
body
can be
determined,
and the
principal
stresses and
principal
extensions can be found. Let T be
the
greatest principal
tension,
S the
greatest
difference of two
principal
tensions at the same
point,
e the
greatest principal
extension. Let
T,,
be the
breaking
stress as determined
by
tensile tests. On the
greatest
tension
hypothesis
T must not exceed a certain fraction of
T^.
On the stress-
difference
hypothesis
S must not exceed a certain fraction of
T^.
On the
greatest
extension
hypothesis
e must not exceed a certain fraction of
T^jE,
where E is
Young's
modulus for the material. These conditions
may
be
written
T<To/^,
S<To/<t>, e<TJ<PE
and the number $ which occurs in them is called the
"
factor of
safety."
*
For
examples
see Todhunter and Pearson's
History,
vol.
1, p.
550 footnote.
+
'
On the stresses
produced
in the interior of the Earth
by
the
weight
of Continents and
Mountains,'
Phil. Trans.
Boy.
Soc,
vol. 173
(1882).
The same measure is
adopted
in the account
of Prof. Darwin's work in Kelvin aud Tait's Nat. Phil. Part ii. art. 832'.
t
'Essai sur nne
application
des
regies
de Maximis
&c.,'
M4m.
par
divers
Savatis, 1776,
Introduction.
Zeitschrift
der Deutschen
Ingenieure,
Bd. 44
(1900).
A discussion
by Voigt
of the views of
Mohr and other writers will be found in Ann.
Phys. (Ser. 4),
Bd. 4
(1901).
'
83,
84]
CONDITIONS OF SAFETY
119
Most
English
and American
engineers adopt
the first of these
hypotheses,
but take <I> to
depend
on the itind of strain to which the
body
is
likely
to be
subjected
in use. A factor 6 is allowed for
boilers,
10 for
pillars,
G for
axles,
6 to 10 for
railway-bridges,
and 12 for
screw-propeller-shafts
and
parts
of
other machines
subjected
to sudden reversals of load. In France and
Germany
the
greatest
extension
hypothesis
is often
adopted.
Recently attempts
have been made to determine which of these
hypo-
theses best
represents
the results of
experiments.
The fact that short
pillars
can be crushed
by longitudinal pressure
excludes the
greatest
tension
hypo-
thesis. If it were
proposed
to
replace
this
by
a
greatest
stress
hypothesis,
according
to which
rupture
would occur when
any principal
stress
(tension
or
pressure)
exceeds a certain
limit,
then the
experiments
of A.
Foppl*
on
bodies
subjected
to
very great pressures
uniform over their surfaces would be
very important,
as it
appeared
that
rupture
is not
produced by
such
pressures
as he could
apply.
These
experiments
would also forbid us to
replace
the
greatest
extension
hypothesis by
a
greatest
strain
hypothesis.
There remain
for examination the
greatest
extension
hypothesis
and the stress-difference
hypothesis. Wehage's experiments"}*
on
specimens
of
wrought
iron
subjected
to
equal
tensions
(or pressures)
in two directions at
right angles
to each
other have thrown doubt on the
greatest
extension
hypothesis.
From
experiments
on metal tubes
subjected
to various
systems
of combined stress
J. J.
Guest:]:
has concluded that the stress-difference
hypothesis
is the one
which accords best with observed results. The
general tendency
of modern
technical
writings
seems to be to attach more
importance
to the limits
of linear
elasticity
and the
yield-point
than to the limits of
perfect elasticity
and the
breaking
stress,
and to
emphasize
the
importance
of
dynamical
tests
in addition to the usual statical tests of tensile and
bending strength.
84.
Scope
of the mathematical
theory
of
elasticity.
Numerical values of the
quantities
that can be involved in
practical
problems may
serve to show the smallness of the strains that occur in
structures which are found to be safe.
Examples
of such values have been
given
in Articles
1, 48, 71,
78. A
piece
of iron or steel with a limit of linear
elasticity equal
to
10^
tons
per square
inch,
a
yield-point equal
to 14 tons
per square
inch and a
Young's
modulus
equal
to 13000 tons
per square
inch
would
take,
under a load of 6 tons
per square
inch,
an extension 0"00046.
Even if loaded
nearly up
to the
yield-point
the extension
would be small
enough
to
require very
refined means of observation. The
neglect
of
squares
and
products
of the strains in iron and steel structures
within safe limits of
loading
cannot be tbe cause of
any
serious error. The fact that for loads much
*
Miltheihmgen (Miinchen),
xxvn.
(1899).
t
Mittheilmigen
der mechanisch-technischen
Venuchsamtalt
zu
Berlin,
1888.
t
Phil.
Mag. (Ser. 5),
vol. 48
(1900).
Mohr
(loc. cit.)
has oritioized Guest.
120 SCOPE OF THE MATHEMATICAL
[CH.
IV
below the limit of linear
elasticity
the
elasticity
of metals is
very imperfect may
perhaps
be a more serious cause of
error,
since set and elastic
after-working
are
unrepresented
in the mathematical
theory
;
but the sets that occur
within the limit of linear
elasticity
are
always extremely
small. The effects
produced by unequal heating,
with -which the
theory
cannot deal satis-
factorily,
are
very important
in
practice.
Some
examples
of the
application
of the
theory
to
questions
of
strength may
be cited here :

By
Saint-Venant's
theory
of the torsion of
prisms,
it can be
predicted
that a shaft
transmitting
a
couple by
torsion is
seriously
weakened
by
the existence of a dent
having
a curvature
approaching
to that in a reentrant
angle,
or
by
the existence of
a flaw
parallel
to the axis of the shaft.
By
the
theory
of
equilibrium
of a
mass with a
spherical boundary,
it can be
predicted
that the shear in the
neighbourhood
of a flaw of
spherical
form
may
be as
great
as twice that at
a
distance. The result of such theories would be that the factor of
safety
should be doubled for shafts
transmitting
a
couple
when such flaws
may
occur.
Again
it can be showia
that,
in certain
cases,
a load
suddenly applied
may
cause a strain twice* as
great
as that
produced by
a
gradual application
of
the same
load,
and that a load
suddenly
reversed
may
cause a strain
three times as
great
as that
produced by
the
gradual application
of the same
load. These results lead us to
expect
that additional factors of
safety
will be
requii-ed
for sudden
applications
and sudden
reversals,
and
they suggest
that
these extra factors
may
be 2 and 3.
Again,
a source of weakness in
structures,
some
parts
of which are
very
thin bars or
plates subjected
to
thrust,
is a
pos-
sible
buckling
of the
parts.
The conditions of
buckling
can sometimes be
determined from the
theory
of Elastic
Stability,
and this
theory
can then be
made to
suggest
some method of
supporting
the
parts by stays,
and the best
places
for
them,
so as to secure the
greatest strength
with the least ex-
penditure
of
materials;
but the
result,
at
any
rate in structures that
may
receive small
permanent
sets,
is
only
a
suggestion
and
requires
to be verified
by experiment. Further,
as has been
pointed
out
before,
all calculations of
the
strength
of structures rest on some result or other deduced from the
mathematical
theory.
More
precise
indications as to the behaviour of solid bodies can be
deduced from the
theory
when
applied
to obtain corrections to
very
exact
physical measurements"!-.
For
example,
it is
customary
to
specify
the tem-
perature
at which standards of
length
are correct
;
but it
appears
that the
effects of such
changes
of
atmospheric pressure
as
actually
occur are not too
small to have a
practical significance.
As more and more accurate instru-
ments come to be devised for
measuring lengths
the time is
probably
not far

This
point appears
to have been first
expressly
noted
by
Poncelet in his Introduction h
la
Mfxanique industrielle,
physique
et
expirimentale
of
1839,
see Todhunter and Pearson's
History,
vol.
1,
art. 988.
t Cf. C.
Ohree,
Phil.
Mag. (Ser. 6),
vol. 2
(1901).
84]
THEORY OF ELASTICITY
121
distant when the effects
produced
in the
length
of a standard
by
different
modes of
support
will have to be taken into account. Another
example
is afforded
by
the result that the cubic
capacity
of a vessel intended to
contain
liquid
is increased when the
liquid
is
put
into it in
consequence
of the excess of
pressure
in the
parts
of the
liquid
near the bottom of the
vessel.
Again,
the
bending
of the deflexion-bars of
magnetometers
affects
the measurement of
magnetic
force.
Many
of the
simpler
results of the
mathematical
theory
are
likely
to find
important applications
in connexion
with the
improvement
of
measuring apparatus.
CHAPTER V.
EQUILIBRIUM
OF ISOTROPIC ELASTIC SOLID BODIES.
85.
Recapitulation
of the
general theory.
As a
preliminary
to the further
study
of the
theory
of
elasticity
some
parts
of the
general theory
will here be
recapitulated briefly.
(a)
Stress. The state of stress at a
point
of a
body
is determined when the traction
across
every plane through
the
point
is known. The traction is estimated as a force
per
nnit of area. If v denotes the direction of the normal to a
plane
the traction across the
plane
is
specified by
means of
rectangular components A\, F, Z,
parallel
to the axes
of coordinates. The traction across the
plane
that is normal to v is
expressed
in terms of
the tractions across
planes
that are normal to the axes of coordinates
by
the
equations
Tf
=
XtCos(a;, v)
+
Xy0os(j/, v)
+
X,coa{z, v),
\
r=r^co8(.j:, i/)+reos(y, r)+r,cos(j, v),
I
,.(1)
Z
=Z^
COS
(x, v)
+
Zy
cos
{y, v)
+
Zj
cos
{z, v).
)
The
quantities Xj,
... are connected
by
the
equations
Y,

Zy, Zx=X Xy=Y^ (2)

The six
quantities X^, Yy Z Y Z^,
Xy
are the
"components
of stress." Their values
at
any point depend
in
general upon
the
position
of the
point.
(6)
Stress-equalions.
In a
body
in
equilibrium
under
body
forces and surface
tractions the
components
of stress
satisfy
the
following equations
at
every point
in
the
body
:

dX^
dXy dZ^
^
,
dXy dYy
dY
^
+
"3^
+
17+''^-'
dZ dY dZ
In these
equations p
is the
density
and
{X, Y, Z)
the
body
force
per
unit of mass.
The
components
of stress also
satisfy
certain
equations
at the surface of the
body.
If V denotes the direction of the normal drawn outwards from the
body
at
any point
of its surface and
{X,, Y, Zy)
denotes the surface traction at the
point,
the values of the
components
of stress at the
point
must
satisfy
the
equations (1),
in which
X,,,
... are
written for
X,,
.... x
.(3)
85,
86]
RECAPITULATION OF THEORY
128
(c) Displacement.
Under the action of the forces the
body
is
displaced
from the
configuration
that it would have if the
stress-components
were zero
throughout.
If
{x, y, z)
denotes the
position
of a
point
of the
body
in the
unstressed
state,
and
(x
+
n,
y+v, z-^w)
denotes the
position
of the same
point
of the
body
when under the action
of the
forces,
(k, v, w)
denotes the
displacement,
and the
components
of
displacement
M, y,
w are functions of
x, y,
z.
(d)
Strain. The strain <at a
point
is determined when the extension of
every
linear
element
issuing
from the
point
is known. If the relative
displacement
is
small,
the
extension of a linear element in direction
{I, m, n)
is
'!xx^^+eym--\-en^+ey,mn-\-e'nl-\-e^lm, (4)
where
e,
... denote the
following
:

:-g^.
^'"'-g^'
*"=a^'
I
_dw
dv
_du
dw dv du
(
^"~dy'^dz'
^'^'-Tz'^d^'
"""-^v^d^-)
The
quantities e,xj
ejia
ai'e the
"components
of strain."
The
quantities uij., nr^, tzr,
determined
by
the
equations
dw dv

Su dw - dv du
^'''=dy'dz^
^='=8.-ai'
2^'=ai-^
()
are the
components
of a vector
quantity,
the "rotation." The
quantity
A determined
by
the
equation
^^3m
3i) 3w
dx
dy
dz
^
is the "dilatation."
(e)
Stress-strain relations. In an elastic solid
slightly
strained from the imstressed
state the
comiwnents
of stress are linear functions of the
components
of strain. When the
material is
isotropic
we have
Xx='K^
+
'iiiexx, Yy
=
\i^
+
'iy.ey,j,
2,=XA+
2;ie,
]
>
(8)
Y,iiey Zx=iie, Xy=iiexy
;
)
and
by solving
these we have
exx
=
\,{Xx-<T{Yy
+
Z.)},
eyy=^{Yy-,r{Z.
+
Xx)},
e='^{Z,-,r{X,+
Yy)},
I
_2(l+.r)
_2(l+,r) ,_
2(l
+
.r)
("
where
-'f^'". "qxb
"'
The
quantity
E is
"
Young's modulus,"
the number o- is
"
Poisson's
ratio,"
the
quantity
H
is the
"
rigidity,"
the
quantity
X
+
f/i,
=k,
is the
"
modulus of
compression."
86.
Uniformly varying
stress.
We considered some
examples
of uniform stress in connexion with the definitions of
E, k,
etc.
(Article 69).
The cases which are next in order of
simplicity
are those in which
the
stress-components
are linear functions of the coordinates. We shall record the results
in
regard
to some
particular
distributions of stress.
124 DISPLACEMENT ACCOMPANYING
[CH.
V
(a)
Let the axis of 2 be directed
vertically upwards,
let aU the
stress-components
except
Z, vanish,
and let
Z,=gpe,
where
p
is the
density
of the
body
and
ff
is the acceleration
due to
gravity.
The
stress-equations
of
equilibrium (3)
are satisfied if
A'=0, r=0,Z= -g.
Hence this
state of stress can be maintained in a
body
of
any
form
by
its own
weight provided
that
suitable tractions are
applied
at its surface. The traction
applied
at the surface must be
of amount
jrpz
cos
(2, v),
and it must be directed
vertically upwards.
If the
body
is a
cylinder
or
prism
of
any
form of
cross-section,
and the
origin
is at the lower
end,
the
cylinder
is
supported by
tension
uniformly
distributed over its
upper
end. If I is the
length
of the
cylinder
this tension is
gpl,
and the resultant tension is
equal
to the
weight
of the
cylinder.
The lower end and the ciu-ved surface are free from traction.
The strain is
given by
the
equations
To find the
displacement*
we take first the
equation
3'
gpz
which
gives
^
ffP 2 ,
where
jcj
is a function of .v and
y.
The
equations ejj
=
ejx=0 give
3m
dwQ
dv
_ dw0
dz~ dx
'
dz~
dy
'
and therefore we must have
where

and
v,
are
fmictions of ^
an-,
T^e
equations
dx
dy
E
.
give
"
^-n 5''o_n SH '^ap dha,
ox
'
dy
'
8^2 ^.
I
a2
-
-^

The
equation
exy=Q gives
S^^Sa--"'
d^y=^-
The
equations
containing .
can be satisfied
only by
an
equation
of the form
where
a',
^'.7
are
constants. The
equations
containing , v,
show that
u,
is a functJpn
or
y say y',(y),
and

is a
function of
x, say F^{x\
and that these
functions
satisfy
t^e
equation
dy
"^
dx
~"'
and this
equation
requires
that
dP, {y)ldy
and
dF, {x)ldx
should be
con.stant.s,
V and
-
V
say
Hence we have
' ^ ""j'-
^1 LV)
=
y'y
+
a,
Fi (x)
=
-
y'x
4-
ft
*
The work is
given
at
length
as an
example
of
method.
86]
UNIFORMLY VARYING STRESS
125
where a and
^
are constants. The
complete expressions
for the
displacements
are
therefore
trap , ,
v^-^zy-l^z-y'x+^,
The terms
containing a, 13, y,
a', ^, y represent
a
displacement
which would bo
possible
in a
rigid body.
If the
cylinder
is not
displaced by
rotation we
may
omit
a', /S",
y'.
If it is not
displaced laterally
we
may
omit
a, fi.
If the
jwiiit (0, 0, I)
is not
displcMied
vertically,
we must have
)/=
- -
^^
. The
displacement
is then
given by
the
equations
<Tgpzx <rffpyz
E
'
^=l^-(^'+'^-^^+'^^-l^)
(11)
Any
cross-section of the
cylinder
is distorted into a
paraboloid
of revolution about the
vertical axis of the
cylinder,
and the sections shrink
laterally by
amounts
projwrtional
to
their distances from the free
(lower)
end.
(6)
A more
general
case* is obtained
by taking
^1=
!'= -p+9p'^,
^z=
-p+9 {p
-
p') l+gpi.
This state of stress can be maintained in a
cylinder
or
prism
of
any
form of
length
2i
suspended
in fluid of
density p
so as to have its axis vertical and the
highest point (0, 0, I)
of its axis fixed
;
then
p
is the
pressure
of the fluid at the level of
the
centre of
gravity
of the
cylinder.
The
displacement may
be shown to be
given by
the
equations
w=-'-^[{l-2a)p-g{p-p')l]
+
y{p-2crp')iz^-n
.(12)
(c) By putting
X,=^ =}!,= -p
+
cfpz,
r,=Z,
=
Xy=0,
we obtain the state of stress in a
body
of
any
form immersed in
liquid
of the same
density,
p being
the
pressure
at the level of the
origint.
The
displacement may
be shown to
be
given by
the
equations
.(.-P2/+9P^)>
1
^
1
U
=
;i^-;-s (
-
PX+gpZX),
V
=
3X-l-2fi
1
"3X +
2^
3X-l-2,x'
{-pz+ypiz-'-^-y^)}.
.(13)
*
C.
Chree,
Phil.
Mag. (Ser. 6),
vol. 2
(1901).
t
E. and F.
Cosserat,
Paris C.
P.,
t. 133
(1901).
126 THE BENDING OF BARS
[CH.
V
{d)
Let all the
stress-components except V,
and
Z^ vanish,
and let these be
given
by
the
equations
X
-y
'''
where t is a constant and
/x
is the
rigidity.
This state of stress can be maintained in a bar of circular section with its axis
coinciding
with the axis of z
by
tractions
applied
at its ends
only.
If a is the radius of the circle the
tractions on the terminal sections are
statically equivalent
to
couples
of moment
\na*fiT
about the axis of
z,
so that we have the
problem
of a round bar held twisted
by opposing
couples.
The
displacement may
be shown to be
given by
the
equations
u=ryz,
v
=
TZX,
w=0
(14)
so that
any
section is turned in its own
plane through
an
angle ,
which is
proportional
to the distance from a fixed section. The constant t measures the twist of the bar.
87. Bar bent
by couples*.
Our next
example
of
uniformly varying
stress is of
very great importance.
We take the
stress-component
Z^
to be
equal
to

ER~^x,
where R vs, &
constant,
and we take the
remaining stress-components
to vanish. If this
state of stress existed within a
body,
in the
shape
of a
cylinder
or
prism
having
its
generators
in the direction of the axis of
z,
there would be no
body
force,
and there would be no tractions on the
cylindrical boundary.
The resultant traction over
any
cross-section is of amount I
jZidxdy
;
and this
vanishes if the axis of z coincides with the line of centroids of the normal
sections in the unstressed state. We take this to be the case. Then the bar
is held in the
specified
state of stress
by
tractions over its terminal sections
only,
and the traction across
any
section is
statically equivalent
to a
couple.
The
component
of the
couple
about the axis of z vanishes. The
component
about the axis of
y
is I
\ER'~^a?dxdy,
or it is
EIjR,
where / is the
moment of inertia of the section about an axis
through
its centroid
parallel
to the axis of
y.
The
component
of the
couple
about the axis of x is
ER~^xydxdy,
and this vanishes if the axes of x and
y
are
parallel
to
principal
axes of inertia of tbe cross-sections. We shall
suppose
that this
is the case.
The
strain-components
are
given by
the
equations
du dv <TX dw
_
X
dx'^dy^R'
dz~~R'
dw dv
_du
dw
_dv
du
_
^
dy
dz dz dx dx
dy
'
*
The
theory
was
given by
Saint-Venant in his memoir on TorBiou of 1855. See
Introduction,
footnote 50 and
p.
20.
II
86-88]
BY TERMINAL COUPLES
127
and the
displacement may
be shown to be
given by
the
equations
;
=
^R-^z^
+
ax^
-
oy'),
v^aR-'xy,
w
=
-R-'xz
(15)
u-
This
example corresponds
with the
bending of
a bar
by couples.
The line
of centroids of the cross-sections is
displaced according
to the law u
=
^R-^z',
so that it becomes
very approximately
an arc of a circle of
large
radius
R,
in
the
plane (x, z),
which is the
plane
of the
bending couple EIjR ;
the centre
of the circle is at a;
=
i?,
^
=
0.
88. Discussion of the solution for the
bending
of a bar
by
terminal
couple.
The forces
applied
at either end of the bar are
statically equivalent
to
a
couple
of moment
EI/R.
This
couple,
called the
"bending moment,"
is
proportional
to the curvature
l/R.
When the bar is bent
by
a
given
couple
M the line of centroids of its
cross-sections,
called the
"
central-line,"
takes a curvature
MjEI
in the
plane
of the
couple.
The formulae for the
components
of strain show that the linear elements of the material
which,
in the unstressed
state,
are in the
plane
a;
=
undergo
no
extension or
contraction.
This
plane
is called the
"
neutral
plane
"
;
it is the
plane
drawn
through
the central-line at
right angles
to the
plane
of
bending.
The same
formulae show that linear elements of the material
which,
in the
unstressed
state,
are
parallel
to the central-line are contracted or extended
according
as
they
lie on the same side of the neutral
plane
as the centre of curvature
or on the
opposite
side. The amount of the extension or contraction of a
longitudinal
linear element at a distance x from the neutral
plane
is the
absolute value of
MxjEI
or
xjR.
The stress consists of tensions and
pressures
across the elements of the normal sections. It is tension at a
point
where
the
longitudinal
filament
passing through
the
point
is
extended,
and
pressure
at a
point
where the
longitudinal
filament
passing through
the
point
is
contracted.
The amount of the tension or
pressure
is the absolute value
of
Mxjl,
or
ExjR.
Fig.
10.
The formula for the
displacement
show that the cross-sections
remain
plane,
but that their
planes
are rotated so as to
pass
through
the centre
of
curvature,
as shown in
Figure
10. The formulae for the
displacement
128 THE BENDING OF BARS
[CH.
V
also show that the
shapes
of the sections are
changed.
If,
for
example,
the
section is
originally
a
rectangle
with boundaries
given by
the
equations

=
a,
y=^,
in a
plane
^^
=
7,
these boundaries will become the curves that are
given
respectively by
the
equations
00
+
a-
^rflR
-
^a- {a'
-
f)/R
=
0,
y
+
13
-
a-^x/R
=
0.
The latter are
straight
lines
slightly
inclined to their
original
directions;
the former are
approximately
arcs of circles of radii
R/cr,
with their
planes
Fig.
11.
parallel
to the
plane
of
(x, y),
and their curvatures turned in the
opposite
sense to that of the line of centroids. The
change
of
shape
of the cross-
sections is shown in
Figure
11. The neutral
plane,
and
every parallel plane,
is strained into an anticlastic
surface,
with
principal
curvatures of
magnitudes
Fig.
12.
R~^ in the
plane
of
(x, z)
and crR"^ in the
plane
of
{x, y),
so that the
shape
of the bent bar is of the kind illustrated in
Figure
12,
in which the front
face is
parallel
to the
plane
of
bending {x,
z).
The distortion of the
bounding
surfaces
x=a
into anticlastic
surfaces,
admits of
very
exact verification
by
means of the interference
fringes
which are
produced by light
transmitted
through
a
plate
of
glass
held
parallel
and
very
close to these surfaces of the
bent bar. Coniu* has used this method for an
experimental
determination of Poisson's
ratio for
glass by
means of the
bending
of
glass
bars. The value obtained was almost
exactly J.
*
Paru,
G.
R.,
t. 69
(18G9).
The method has been used for several materials
by
Mallock.
See Article 70
(e),
footnote.
88-90]
AND PLATES
129
It is worth while to calculate the
potential energy
of strain. The value of the strain-
energy-function
at
any point
is
easily
found to be
\E:i^llP.
The
potential energy
of strain
(if the
part
of the bar lietweeti two normal sections distant I
apart
is
i
{EIjlP) I,
so that
the
potential energy per
luiit of
length
is
hJil/K^.
89. Saint-Venant's
principle*.
In the
problem
of Article
87,
the
tractions,
of which the
bending
moment
EI/R
is the statical
equivalent,
are distributed over the terminal sections
in the manner of tensions and
pressures
on the elements of
area,
these tensions
and
pressures being proportional
to the distance from the neutral
plane.
But
the
practical utility
of the solution is not confined to the case where this
distribution of terminal traction is
exactly
realised. The extension to other
cases is made
by
means of a
principle,
first
definitely
enunciated
by
Saint-
Venant,
and known as the
"principle
of the elastic
equivalence
of
statically
equipollent systems
of load."
According
to this
principle,
the strains that
are
produced
in a
body by
the
application,
to a small
part
of its
surface,
of a
system
of forces
statically equivalent
to zero force and zero
couple,
are of
negligible magnitude
at distances which are
large compared
with
the linear dimensions of the
part.
In the
problem
in
hand,
we infer
that,
when the
length
of the bar is
large compared
with
any
diameter
of its
cross-section,
the state of stress and strain set
up
in its interior
by
the
terminal
couple
is
practically independent
of the distribution of the
tractions,
of which the
couple
is the
resultant,
in all the
portions
of the bar
except
comparatively
small
portions
near its ends.
90.
Rectangular plate
bent
by couplesf.
The
problem
solved in Article 87 admits of
generalization
in another
direction. A bar of
rectangular
section is a
particular
case of a
brick-shaped
body;
and,
when two
parallel
faces are near
together,
such a
body
is a
rectangular plate.
We have therefore
proved
that a
plate
can be
held,
so that its faces are anticlastic
surfaces, by couples applied
to one
pair
of
opposite edges,
and
having
their axes
parallel
to those
edges.
The ratio
of the
principal
curvatures is the number a. It is clear
that,
by
means
of suitable
couples simultaneously applied
to the other
pair
of
opposite edges,
the
plate
can be bent into a
cylindrical
form,
or the ratio of curvatures can
be altered in
any
desired
way.
It is most convenient to take the faces of the
plate
to be
given by
the
equations
2=h,
so that the thickness is 2h. The coordinate z thus takes the
place
of the
coordinate which we called x in the case of the bar. The
requisite
stress-
components
are
Xx
and
F,
and both are
proportional
to tlie coordinate z. If
*
Stated in the memoir on Torsion of 1855.
t
Kelvin and
Tait,
Nat.
Phil.,
Part
II, pp.
265,
266.
L. E.
9
130 UNIFORM BENDING OF PLATES
[CH.
V
we assume that all the
stress-components except
X^
and
F,, vanish,
and that
these are
given by
the
equations
X^
=
Eaz,
Yy==E^z, (16)
where a and
/3
are
constants,
we find that the
displacement
is
given by
the
equations
M
=
(a

(7/8)
xz,
v
=
{0

era) yz,
Hence
any
surface which in the unstrained state was
parallel
to the faces
becomes curved so that the curvatures in the
planes
of
(x, z)
and
{y, z)
are
respectively
cr/3

a and era

/8.
These are the
principal
curvatures of the
surface. If these
quantities
are
positive,
the
corresponding
centres of
curvature lie in the direction in which z is
positive.
Let
Hi
and
B^
be the
radii of curvature so that
1 dhv 13^
The state of curvature
expressed by
iJ,
and R, is maintained
by couples
applied
to the
edges.
The
couple per
unit of
length, applied
to that
edge
X
=
const, for which x has the
greater
value,
has its axis
parallel
to the axis
of
y,
and its amount is
r*
^
- .... 2 Eh' /I a
J_^.Z.d.,
which
is-g^_-^-^_
+
-^-
An
equal
and
opposite couple
must be
applied
to the
opposite edge.
The
corresponding couple
for the other
pair
of
edges
is
given by
/:
-Zi.dz,
which IS
-^-^^-^-+-^^
The value of the
strain-energy-function
at
any point
can be shown without
difficulty
to be
h'YS^^[{i^wJ-'^'-^i^if2'
and the
potential energy
of the bent
plate per
unit of area is
1 Eh'
[(i-/iy-^(^-'^)7rJ-
3 l-<r'
It is
noteworthy
that this
expression
contains the sum and the
product
of the
principal
curvatures.
91.
Equations
of
equilibrium
in terms of
displacements.
In the
equations
of
type
dx
dy
oz
90,
91]
GENERAL
EQUATIONS 131
we substitute for the normal
stress-components
Xx,...
such
expressions
as
XA
+
2f/,du/dx,
and for the
tangential
stress-components
Fj,
... such
expressions
as
/M(dw/dy
+
dvjdz)
;
and we thus obtain three
equations
of the
type
i\
+
/^)^^^+^V^u
+
pX
=
0,
(19)
,
. dii dv
,
div
_
d" d-' df
where A
=
;r- +
v + ^ >
V- =
^ +
j
.
These
equations may
be written in a
compact
form
^^+''K^'l^'S^+'^'^'^''''''"'^+^^^'^'^^=^
^^^^
If we introduce the rotation
{^x, ^y, ^z)
=
h
curl
{u, V,
w),
_
1 fdw dv du dw dv du\
2
\di/
dz
'
dz dx
'
dx
dy)
'
and make use of the
identity
V=
(u, V,
w)
=
(^g^
,
^,
g-
)
^
-
2 curl
(w^, CT, w,),
the above
equations (20)
take the form
(^
+
2/i)
(1^
,
g^
,
g^)
A
-
2m
curl
(t^r,, .^,
^,)
+
p (X,
Y,
Z)
=
0..
..(21)
We
may
note that the
equations
of small motion
(Article 54)
can be
expressed
in either of the forms
^^
+
'''*
[L
'
a|
'
I)
"^
+
^
^'
^"' '
"^^
+
^ ^^^'
^'
^)
=
P

("'
''' ^>'
or
g^
,
g-
,
g^J
A
-
2/x
curl
(tii^, -57,,, sr^)
-f
p
{X, Y,Z)
=
p
g-
{u,
v, w).
^
'
(22)
The traction
{X^, Y, Z,)
across a
plane
of which the normal is in the
direction
v,
is
given by
formulae of the
type
Z.
=
cos(..,.)(xA-t-2Mg+cos(y,.)/*g-H|)-fcos(z,.)Mg
+
g);
and this
may
be written in either of the forms
Z,
=
XA cos
(, I/)
-f
/i
jg-
4-
cos
{x,
v)y
+
cos
{y,
v)
^-V
cos
{z, v)
^>,..
.(23)
or
X^
=
XA cos
{x, i/)
4-
2/i
] ^
-
otj,
cos
(^, r)
-I- isr^
cos
(y, r);-
,
...
(24)
where
dv
=
"'^
(^' "^
d-x
+
''
^2/. -)
3^
+
cos
{z, v)
g^
.
92
132 SYSTEM OF
EQUATIONS
[CH.
V
If V is the normal to the
bounding
surface drawn outwards from the
body,
and the values of
A,
dujdx,
. . . are calculated at a
point
on the
surface,
the
right-hand
members of
(23)
and the similar
expressions represent
the com-
ponent
tractions
per
unit area exerted
upon
the
body
across the surface.
92.
Equilibrium
under surface tractions
only.
We record here some results deducible from the
displacement-equations
(i) By dififerentiating
the left-hand members of those
equations
with
respect
to
x, y, z,
and
adding
the
results,
we find
(X
+
2^)v2a=0, (26)
80 that A is an harmonic
function,
i.e. a function
satisfying Laplace's equation,
at all
points
within the
body.
(ii)
It follows from this and
(25)
that each of
u, v,
w satisfies the
equation
V^0
=
O
(27)
at all
points
within the
body.
All
components
of strain and of stress also
satisfy
this
equation.
(iii) Again, by differentiating
the left-hand member of the third of
equations (25)
with
resiject
to
y,
and that of the second with
respect
to
z,
and
subtracting
the
results,
we find
V^^x=0
(28)
Similar
equations
are satisfied
by icTj,
and
nr^,
so that each of the
components
of the
rotation is an harmonic function at all
points
within the
body.
(iv)
The
stress-component.s satisfy
a
system
of
partial
differential
equations.
In order to obtain them it is convenient to introduce a
quantity
0,
the sum of the
principal
stresses at
any point ;
we have
=
(3\
+
2/i)A;
(29)
thus is an harmonic function at all
points
within the
body.
Further we find*
-^--I^S=-
W
In like manner we find
^'^-'mB-o
<'
Similar formulae can be obtained for
V-Tj,,
V%, V-Z^,
V-Xy.
The
coefficient 2
{\
+
fi)/{S\
+
2fi)
is
1/(1
+
o").
*
The
equations
of
types (30)
and
(31)
were
given by Beltrami, liovie,
Ace. Lincei Rend.
(Ser. 5),
t. 1
(1892).
91-93]
SATISFIED BY THE STRESS-COMPONENTS
133
(v)
As an
example
of the
application
of these
formute,
we
may
observe that Maxwell's
stress-system,
described in
(vi)
of Article
53,
cannot occur in an
isotropic
solid
body
free
from the action of
body
forces and
slightly
strained from a state of no stress*. This
appears
at once on
observing
that
A'^+Y^
+
Z^,
as
given
for that
system,
is not an
harmonic function.
93. Various methods and results.
(i)
The
equations
of
types (30)
and
(31) may
also be
deducedt
from the
stress-equations
(3)
and the
equations
of
compatibility
of the
strain-components (Article 17).
We
have,
for
example,
e^
=
E-^
{{l
+
<r) X,-<re],
...
e,,=2{l+,r) E-n'-..
Thus the
equation
82'
8/'' di/dz
becomes
g5
i
+
'^) ^'^
-
"^ei-
+
3p
Ki
+
'^) ^^
-
"^e}
=
2
(1
+
<r)
|
.
3F, dZ, dZ.
Hi OW
^5
=

^5 -5 ,
?y
ox oz
dr._
dx,
djj,
OS ox
cy
82r._
s
/a^
azA
'^y,
s^z.
We have therefore
d'^X^ d'^Yy
d^Z,
dx^
3y2
3^2
(i-^.)[v^e-vX.-g]-.(v^e-g)=0;-
and,
on
adding
the three
equations
of this
type,
we find that V^e must
vanish,
and the
equation
reduces to
(l-l-o-)V2A-^+|^
=
(29 6w)
We
may
in like manner deduce
equation (31)
from the
equation
dydz
dx
\
ox
dy
dz
j'
(ii)
It
may
be
shownj
that the stress-functions
xi. X2> Xs
"^ Article 56
satisfy
thi'ee
equations
of the
type
^^^^'&^^-''
''''
and three
equations
of the
type
g|-,[(i+<^)V^x.-e]=o,
(33)
where 6 is written for
V'(Xi
+
X2+X3)-^^--^-a^
^**^
*
Minehin, Static*,
3rd edn.
Oxford, 1886,
vol.
2,
Ch. 18.
t
Miohell,
London Hath. Soc.
Proc,
vol. 31
(1900),
p.
100.
t Ibbetson,
Mathematical
-Theory 0/ Elasticity,
Loudon,
1887.
IS*
PLANE STRAIN
[CH.
V
It
may
be shown also that the stress functions
\/ci, i/'ji ^s
"f the same Article
satisfy
three
equations
of the
type
(^-)^^|S+S=
^^^)
and thi"ee
equations
of the
type
(-)-%4(S-t-*)+|l=.
<>
where 9 is written for
^ll
+
^lh.^^?
(37)
dydz
dzdx
'dxdy
(iii)
It
may
be shown* also
that,
when there are
body forces,
the
stress-components
satisfy equations
of the
types
and
vr,+
,-^|!^=-p^/-p3/
\
+
<TOyoz
'^
cy
'^
dz
^
The
equations
of these two
types
with the
equations (3)
are a
complete system
of
equations
satisfied
by
the
stress-components.
94. Plane strain and
plane
stress.
States of
plane
strain and of
plane
stress can be maintained in bodies
of
cylindrical
form
by
suitable forces. We take the
generators
of the
cylin-
drical
bounding
surface to be
parallel
to the axis of
z,
and
suppose
that the
terminal sections are at
right angles
to this axis. The
body
forces,
if
any,
must be at
right angles
to this axis. When the
lengths
of the
generators
are small in
comparison
with the linear dimensions of the cross-section the
body
becomes a
plate
and the terminal sections are its
faces.
In a state of
plane
strain,
the
displacements
u,
v are functions of
x,
y only
and the
displacement
w vanishes
(Article 15).
All the
components
of strain
and of stress are
independent
of
z;
the
stress-components
Z^, Y^ vanish,
and
the
strain-components e^x,
^yz,
e^^
vanish. The
stress-component
Z^
does not
in
general
vanish. Thus the maintenance of a state of
plane
strain
requires
the
application
of tension or
pressure,
over the terminal
sections,
adjusted
so as to
keep
constant the
lengths
of all the
longitudinal
filaments.
Without
introducing any
additional
complication,
we
may
allow for an
uniform
extension or contraction of all
longitudinal filaments,
by taking
w
to be
equal
to
ez,
where e is constant. The
stress-components
are then
expressed by
the
equations
*
Michell,
loc. cit.
93,
94]
AND PLANE STRESS
186
The functions
u,
v are to be determined
by solving
the
equations
of
equi-
librium. We shall discuss the
theory
of
plane
strain more
fully
in
Chapter
IX.
In a state of
plane
stress
parallel
to the
plane
of
(x, y)
the stress-
components
Zx, Fj, Z^ vanish,
but the
displacements
u, v,
w are not in
general independent
of z. In
particular
the
strain-component
e^
does not
vanish,
and in
general
it is not
constant,
but we have
:=S=-. h-it
+
~]--^^^
(40)
dz \
+
2(i
\dx
dyl
2/ii
The
maintenance,
in a
plate,
of a state of
plane
stress does not
require
the
application
of traction to the faces of the
plate,
but it
requires
the
body
forces and tractions at the
edge
to be distributed in certain
special ways.
We shall discuss the
theory
more
fully
in
Chapter
ix.
An
important generalization*
can be made
by supposing
that the normal
traction
Z^
vanishes
throughout
the
plate,
but that the
tangential
tractions
Zx, Yz
vanish at the faces z=
h
only.
If the
plate
is thin the deter-
mination of the
average
values of the
components
of
displacement,
strain
and stress taken over the thickness of the
plate may
lead to
knowledge nearly
as useful as that of the actual values at each
point.
We denote these
average
values
by
tl,
...
^xx^ ^x,
so that we have for
example
u
=
{2h)-'{\idz (41)
J -h
We
integrate
both members of the
equations
of
equilibrium
over the thick-
ness of the
plate,
and observe that
Zx
and
Y^
vanish at the faces. We thus
find
that,
if there are no
body
forces,
the
average stress-components
Xx,
Xy, Yy satisfy
the
equations
^^
+
^Ay
=
0,
^
+
^^
=
(42)
dx
oy
ax
dy
Since
Z^
vanishes
equations (40)
hold,
and it follows that the
average
dis-
placements
u,
V are connected with the
average stress-components
Xx,
Xy, Yy
by
the
equations
y ^Kfifdu
dv\ du .
.(43)
F,
=
^-^L^-f^)
+
2^|^,
"
\ +
2/jk
\dx
dyj '^dy
^^
=
''[dy+rx.
States of stress such as are here described will be termed states of
"generalized
plane
stress."
*
Cf. L. N. G.
Filoii,
Phil. Tram.
Roy.
Soc.
(Ser. A),
vol. 201
(1903).
13G BENDING OF NARROW BEAM
[CH.
V
95.
Bending
of narrow
rectangiilar
beam
by
terminal load.
A
simple example
of the
generalized type
of
plane
stress,
described in
Article
94,
is afforded
by
a beam of
rectangular
section and small breadth
(2h),
bent
by
forces which act in directions
parallel
to the
plane containing
the
length
and the
depth.
We shall take the
plane
of
(x, y)
to be the
mid-plane
of the beam
(parallel
to
length
and
depth);
and,
to fix
ideas,
we shall
regard
the beam as horizontal in the unstressed state. The
top
and bottom surfaces
of the beam will be
given by y
=
+ c,
so that 2c is the
depth
of the
beam,
and
we shall denote the
length
of the beam
by
I. We shall take the
origin
at one
end,
and consider that end to be fixed.
From the
investigation
in Article
87,
we know a state of stress in the
beam,
given by
X^

EyjR
;
and we know that the beam can be held in
this state
by
terminal
couples
of moment
^hc^E/R
about axes
parallel
to the
axis of z. The central-line of the beam is bent into an arc of a circle of
radius R. The traction across
any
section of the beam is then
statically
equivalent
to a
couple,
the same for all
sections,
and
equal
to the terminal
couple,
or
bending
moment.
Let us now
suppose
that the beam
is bent
by
a load W
applied
at the
end 00
=
1 as in
Fig.
13. This force
cannot be balanced
by
a
couple
at
any
section,
but the traction across
any
section is
equivalent
to a force W
and a
couple
of moment W
(I

x).
The
stress-system
is therefore not so
simple
as in the case of
bending
by couples.
The
couple
of moment
W
{I

x)
could be balanced
by
tractions
Xx,
given by
the
equation
w
w
Fig.
13. ^'^
=
-4fe^(^-
oo)y\
and the
average
traction
Xx
across the breadth would be the same as
Xx-
We seek to combine with this traction
X^
a
tangential
traction
Xy,
so that
the load W
may
be
equilibrated.
The conditions to be satisfied
by Xy
are
the
following
:

(i) Xy
must
satisfy
the
equations
of
equilibrium
dx
dy
'
dx
'
(ii) Xy
must vanish when
y=
c,
(iii)
2/t I
Xydy
must be
equal
to W.
J
-c
95]
BY TERMINAL LOAD
These are all satisfied
by putting
137
It follows that the load W cau be
equilibrated by
tractions
Xx
and
Xy,
with-
out
Yy, provided
that the terminal
tractions,
of which W is the
resultant,
are distributed over the end so as to be
proportional
to c"

y-.
As in
Article
89,
the distribution of the load is
important
near the ends
only,
if the
length
of the beam is
great
in
comparison
with its
depth.
We
may
show that a
system
of
average displacements
which would
correspond
with
this
system
of
average
stresses is
given by
the
equations
^,u=^iZc^y-f)-l^^^i^l.y+y^--Z,^y),
2fiv
^=
IS
{<''-^
+
il-^)f]
+
^t-o-:. Z^
{3^ {^^-f)-a^
+
Z^f}.
'
8/i<^
.(46)
3X
+
2/i4/ic3'
Since these are deduced from known
stress-components
a
displacement possible
in a
rigid
body miglit
bo
added,
so as to
satisfy
conditions of
fixity
at the
origin.
These conclusions
may
be
compared
with those found in the case of
bending by couples
(Article 88).
We note the
following
results :

(i)
The tension
per
unit area across the normal sections
(A'j.)
is connected with the
bending moment, W{l-.r), by
the
equation
tension
=
-
(bending moment) (y/I)
where
y
is distance from the neutral
plane,
and / is the
appropriate
moment of inertia.
3(\+/i) W{l-x)
(ii)
The curvature
{cPvldx^)y^fj
is
;
so that we have the
equation
4Ac3^i(3X+2;a)
curvature
=
(bending moment)/('/).
(iii)
The surface of
particles which,
in the unstressed
state,
is a normal section does
not continue to cut at
right angles
the line of
particles which,
in the same
state,
is the line
of centroids of normal sections.
The
cosine of the
angle
at which
they
cut when the beam
is bent is
{dvldx+duldy)y^(
and this is 3
TF/8/i/ic.
(iv)
The normal sections do not remain
plane,
but are distorted into curved surfaces.
A line of
particles which,
in the unstressed
state,
is vertical becomes a curved
line,
of
i
normal of central line
tangent
of central line
central
tangent
Fig.
14.
138
GENERAL
EQUATIONS
[CH.
V
which the
equation
is determined
by
the
expression
for it as a function of
y
when x is
constant. This
equation
is of the form
and the
corresponding displacement
consists of a
part ay
which does not alter the
planeness
of the section combined with a
part
which does. If we construct the curve
x=&y^
and
place
it with its
origin (=0, ^
=
0)
on the strained
central-line,
and its
tangent
at the
origin along
the
tangent
to the line of
jmrticles
which,
in the unstre.ssed
state,
is
vertical,
the curve will be the locus of these
particles
in the strained state.
Fig.
14 shows the form into which an
initially
vertical filament is bent and the relative
situation of the central
tangent
of this line and the normal of the strained central-line.
96.
Equations
referred to
orthogonal
curvilinear coordinates.
The
equations
such as
(21) expressed
in terms of dilatation and rotation
can be transformed
immediately by noticing
the vectorial character of the
terms. In fact the terms
[^
,
^
,
-
j
A
may
be read as
"
the
gradient
of
A,"
and then the
equations (22) may
be read
(X
-1-
2/4) (gradient
of
A)

2/i.(curl
of
w)
-t-
p (body force)
=
p
(acceleration) (47)
v^'here ts stands
temporarily
for the rotation
i^isx,
rsy, ct^),
and the factors
such as X
-I-
2/i
are scalar.
Now the
gradient
of A is the vector of which the
component,
in
any
direction,
is the rate of increase of A
per
unit of
length
in that direction
;
and the
components
of this
vector,
in the directions of the normals to three
orthogonal
surfaces
a, /3, 7 (Article 19),
are
accordingly
SA 8A 3A
We have
already
transformed the
operation
curl,
and the
components
of
rotation,
as well as the dilatation
(Article 21);
and we
may
therefore
regard
A and
ara,
wp, OTy
as known in terms of the
displacement.
The
equation (47)
is then
equivalent
to three of the form
(X
+
2.)
*.|^
-
tA*.
I (I;)
+
V...
l^
H-
PK
=
.
^-
()
where
F^,
F^, Fy
are,
as in Article
58,
the
components
of the
body
force in
the directions of the normals to the three surfaces.
97. Polar coordinates.
As an
example
of the
equations (48)
we
may
show that the
equations
of
equilibrium
under no
body
forces when referred to
polar
coordinates take the forms
(X
+
2,)sin^^-2,{||'-|^(,^,i,)|=0,^
(X
+
2^)
r sin
6-.--^^.
\^^
{^^
sm
6)
-
^]
=
0. I
95-98]
IN POLAR COORDINATES
139
We
may
show also that the radial
components
of
displacement
and rotation and the
dilatation
satisfy
the
equations
but that some solutions of these
equations correspond
with states of stress that would
require body
force for their maintenance*.
98. Radial
displacementt.
The
simplest applications
of
polar
coordinates relate to
problems involving purely
radial
displacements.
We
suppose
that the
displacements ue, u^ vanish,
and we write U in
place
of
Ur-
Then we find from the formulse of Articles 22 and 96 the
following
results :

(i)
The
strain-components
are
given by
_dU. _ _
C
(ii)
The dilatation and rotation are
given by
(iii)
The
stress-components
are
given by
(iv)
The
general equation
of
equilibrium,
under radial
body
force
li,
is
(v)
If
R=Q,
the
complete primitive
of the
equation just
written is
where A and B are
arbitrary
constants. The first term
corresponds
with the
problem
of
compression by
uniform normal
pressure [Article
70
(^r)].
The
complete primitive
cannot
represent
a
displacement
in a solid
body containing
the
origin
of r. The
origin
must
either be outside the
body
or inside a
cavity
within the
body.
(vi)
The solution in
(v) may
be
adapted
to the case of a shell bounded
by
concentric
spherical surfaces,
and held strained
by
internal and external
pressure.
We must have
'^'
cr r
(Pi
when r
=
ri.
where
jOq
is the
pressure
at the external
boundary (=?(,),
and
p^
is the
pressure
at the
internal
boundary {r=rj).
We should find
The radial
pressure
at
any point
is
Pi
73
r3-ri3
"^P"
73
r3_ri3'
'
Michell,
London Math. Soc. Proc. vol. 32
(1901), p.
24.
t Most of the results
given
in this Article are due to
Lam^,
Lei-ons sur la tMoric.de
Velasticite,
Paris 1852.
140 RADIAL STRAIN IN
[CH.
V
and the tension in
any
direction at
right angles
to the radius is
In case
P(,=0,
the
greatest
tension is the
superficial
tension at the inner
surface,
of amount
iPi ('o^
+
2''i')/(''o' ''i')
>
*"d the
greatest
extension is the extension at
right angles
to the
radius at the inner
surface,
of amount
(vii)
If in the
general equation
of
(iv)^= yr/r,),
where
^r
is
constant,
the surface r
=
r^
is free from
traction,
and the
sphere
is
complete up
to the
centre,
we find
r _ L
ffPV /5X
+
6^
_
'^\
This
corresponds
with the
problem
of a
sphere
held strained
by
the mutual
gravitation
of its
parts.
It is
noteworthy
that tlie radial strain is contraction within the surface
r=rov'{(3-<T)/(3
+
3or)},
but it is extension outside this surface.
The
application
of this result to the case of the Earth is beset
by
the serious
difficulty
which has been
pointed
out in Article 75.
99.
Displacement symmetrical
about an axis.
The conditions that the
displacement may
take
place
in
planes through
an
axis,
and be the same in all such
planes,
would be
expressed, by
reference
to
cylindrical
coordinates
r, 0, z,
by
the
equations
Ug
=
0,
dur/dd
=
duzldd
=
0.
It will be convenient to write U for
Ur,
and w for
u^.
The
strain-components
are then
expressed by
the
equations
du u dw
'^'^ ' '^^
!- (50)
_dU
dw
_
_n
^"
~
dz dr'
^'^
~
^*'
~
The cubical dilatation and the rotation are
expressed by
the
equations
dU U dw
a
^U dw ,.,.
^
=
^+7+8.'
''^^
=
^-'
-^
=
-^
=
(-1)
It will be convenient to write w for
or.
The
equations
of motion in terms of
displacements
take the forms
{\+2,M)^f^+2,.^^^+pFr
=
pfr,
and the
stress-equations
of
ecjuilibrium
take the forms
.(52)
drr drz rr

6d _
a-
+
ir
+ +
pF,.
=
Q,
or oz r
drz dzz rz
p

dr dz
r
.(53)
98-100]
CIRCULAR
CYLINDER
141
In case w
=
ez,
where e is
constant,
and
dU/dz=0,
we have a state of
plane
strain,
with an uniform
longitudinal
extension
superposed.
In this case
rz
=
0. In case
zz, rz, F^ vanish,
we have a state of
plane
stress.
100. Tube under
pressure.
In the case of
plane
strain,
under no
body forces,
the
displacement
U
satisfies the
equation
3
fdU ,
f^\
,,
a-r(87+7J=^
(54)
of which the
complete primitive
is of the form
U=Ar
+
B/r
(.55)
We
may adapt
this solution to the
problem
of a
cylindrical
tube under
internal and external
pressure,
and we
may
allow for an
uniform
longitudinal
extension e. With a notation similar to that in
(vi)
of
Article
98,
we should
find for the
stress-components
.
PiV-JvV
Pi-Po
n^ri'
ro'-n' r^'-r:-
^
zz
=
:
X
Pin'
-
PoTo''
^
iS\+2fJL)fJl.
and for the constants A and B in
(55)
Pi^i'
-
p.ro'
Xe
A-
+
/X
.(56)
'
/
A
=
p
_
(
Pi
~
Po)
^o'^i
dT^
2(\-|-/i)(r^-r,^) 2{\
+
fx)'
~
2/t (r^
-
r,^)
^""^
The constant e
may
be
adjusted
so that the
length
is maintained constant
;
then e
=
0,
and there is
longitudinal
tension zz of amount
X
Pfl^zPfl
X
+
fi r- -ri'
It
may
also be
adjusted
so that there is no
longitudinal
tension
;
then
zz =0 and
X(Pi>?-JOo?-o')
/i(3\
+
2/i)(r,r-?V)-
When
p
vanishes,
and e is not too
great,
the
greatest
tension is the circum-
ferential
tension, 66,
at the inner
surface,
r
=
rj,
and its amount is
The
greatest
extension is the circumferential
extension,
e^^,
at the same
surface.
142
RADIAL STRAIN IN
[CH.
V
If a closed
cylindrical
vessel is under internal
pressure p^
and external
pressure po,
the resultant tension tt
('"- n")
^ must balance the resultant
pressure
on the
ends,
and we must therefore have the
equation
71-
(V
-
n')
22
=
"
(ri-pi
-
n-po)-
This
equation gives
for e the value*
1
prrl-p^
(5S^
If we assume that the ends of the vessel are
plane,
and
neglect
the alteration
of their
shape
under
pressure,
the volume of the vessel will be increased
by
7rr,Z, (eri
+
2
?7i),
where
l^
denotes the
length
of the inside of the
cylinder,
and
Ui
is the value of U at r
=
7\.
With the above value of e this is
n-n'^i
3
joir,^
-
p,n'
_^
1
(p,-po)r,^
.(59)
2X+2/1
ro=-r,^ fi
n'-ri"
In like
manner,
if we denote
by Z
the
length
of the outside of the
cylinder,
and
neglect
the
change
of volume of the
ends,
the volume within the external
boundary
of the vessel will be increased
by
Trn%
Pirl-pfl
1
(pi-po)n'
'
.(60)
_3X
+
2/x
n^
-
r,- /i
ro=
-
r;'
The
quantity l^
differs from
l-^
by
the sum of the thicknesses of the ends.
In the case of a
long cj'linder
this difference is
unimportant.
The constant
3/(3\
+
2/i)
is
1/Z;,
the
reciprocal
of the modulus of
compression.
When the
difference between
Z
and
li
is
neglected,
the result accords with a more
general
result
f,
which can be
proved
for a closed vessel of
any
form under internal
and external
pressure,
viz. if
Fj
and
V^
are the internal and external volumes in
the unstressed
state,
then
V^

Vi
is increased
by
the amount
{piVi p^V^jk,
when internal and external
pressures p^, p^
are
applied.
In
obtaining
the
results
(59)
and
(60)
we have not taken
proper
account of the action of the
ends of the
cylinder,
for we have assumed that these ends are stretched in
their own
planes
so as to fit the distended
cylinder,
and we have
neglected
the
changes
of
shape
and volume of the ends
; further,
we have
supposed
that
the action of the ends
upon
the walls of the vessel is
equivalent
to a tension
uniformly
distributed over the thickness of the walls. The results will
provide
a
good approximation
if the
length
of the
cylinder
is
great
in com-
parison
with its radii and if the walls are
very
thin.
*
The
problem
has been discussed
by
numerous writers
including Lam6,
loc. clt. ante
p.
139.
It is
important
in the
theory
of the
piezometer.
Cf.
Poynting
and
Thomson,
Properties of
Matter,
London
1902, p.
116. The fact that c
depends
on
k(
=
\ +
liJ.)
and not on
any
other elastic
constant has been utilized for the determination of k
by
A.
Mallock,
Proc.
Roy.
Soc.
London,
vol. 74
(1904).
t
See
Chapter
vii.
infra.
100-102]
CIRCULAR CYLINDER
143
101.
Application
to
gun-construction.
In
equations (56),
tiie
stress-components
rr and 66 are
expressed by
formula) of the
tyjx)
where A and B are constants. These constants are determined
by
the internal and
external
pressures.
We have therefore a solution of the
stre-ss-equations
in a tube under
internal and external
pressure
which is
applicable
in other cases besides the case where
the material
would,
in the absence of the
pressures,
Ije in the unstressed state. The
solution has been taken to be
applicable
to states of initial
stress,
and has been
applied
to
the
theory
of the construction of cannon*. At one time cannon were constructed in the
form of a series of
tubes,
each tube
being
heated so that it could
slip
over the next interior
tube
;
the outer tube contracted
by cooling
and exerted
pressure
on the inner. Cannon so
con.structed were found to be
stronger
than
single
tubes of the same thicknesss.
If,
for
example,
we take the case of two tubes between which there is a
pressure P,
and
suppose
r' to be the radius of the common
surface,
the initial stress
may
be taken to be
given by
the
equations
'^
{ro>r>r')
and
The additional stress when the
compound
tube is
subjected
to internal
pressure p may
be
taken to be
given by
the
equations
'0-
r'
rJr/
r' r
66-p'^[^_
.(61)
The diminution of the
hoop
tension 66 at the inner surface
r=ri may
be taken as an
index of the increased
strength
of the
compound
tube.
102.
Rotating cylinderf.
An
example
of
equations
of motion is aflbrded
by
a
rotating cylinder.
In
equations
(52)
we have to
put/,.=
-
a-r,
where <a is the
angular velocity.
The
equations
for the
displacements
are
with the conditions
rr=rz=(i when ra or
r=a',
rz
=
zz=0 when z
1.
The
cylindrical bounding
surface is here taken to be
r=a,
and it is
supposed
that there
is an axle-hole
given by
r=-a';
the terminal sections are taken to be
given by
z=?,
so
that the
cylinder
is a shaft of
length 21,
or a disk of thickness 21.
*
A. G.
Greenhill, Nciture,
vol. 42
(1890).
Cf.
Boitzmann,
Wien Berichte,
Bd. 59
(1870).
+ See
papers by
C. Chree in
Cambridge
Phil. Soc.
Proc,
vol. 7
(1891, 1892), pp. 201,
283.
The
problem
had been discussed
previously by
several writers
among
whom Maxwell
(loc.
cit.
Article
.57),
and
Hopkinson,
Mess,
of
Math.
(Ser.
2),
vol. 2
(1871) may
be mentioned.
144
ROTATING
SHAFT
[CH.
V
C<ue
(a). Rotating shaft.
An
approximate
solution can be obtained in the case of a
long shaft, by treating
the
problem
as one of
plane
strain,
with an allowance for uniform
longitudinal
extension,
e.
We
regard
the
cylinder
as
complete,
i.e. without an axle-hole
;
and then the
approximate
solution satisfies the
equations
rz
=
throughout,
rr=0 when r
=
a,
but it does not
satisfy
zz=0 when
z=l.
The uniform
longitudinal
extension e can
be
adjusted
so that the tractions zz on the ends shall have no statical
resultant,
i.e.
/:
zzrdr=0
;
and then the solution
represents
the state of the shaft with sufficient exactness over the
greater part
of the
length,
but is defective near the ends.
[Cf.
Article
89.]
We shall
state the results in terms of E and <r. We should find
^=^'-w

r^^

""='"' ^^^^
where the constants A and e are
given by
the
equations
^
=
WT37' '--2E-
(^^>
The
stress-components
are
given by
the
equations
-2a- l-)-2<r \ \
,...(64)
--
w2p(a2-r2)3-2(r
Ci
(2p/3-2,7
l-)-2<r
zz

-
_
4 1 -<r
I
Instead of
making
the resultant
longitudinal
tension
vanish,
we
might suppose
that
the tension is
adjusted
so that the
length
is maintained constant. Then we should have
,=0,
.4^";^^(J-^-)(;+-)^-^-^-)
; (65)
the first two of
equations (64)
would still
hold,
and the
longitudinal
tension would be
given by
the
equation
2_
>V{(3-2.7)'-2>^
,T
,gg^
4 1-0-
Ca^e
(6). Rotating
disk.
An
approximate
solution can be obtained in the case of a thin
disk, by treating
the
problem
as one of
plane
stress. If the disk is
complete,
the
approximate
solution
satisfies the
equations
zz
=
Q,
rz
=
throughout,
so that the
plane
faces of the disk are free
from
traction;
but it does not
satisfy
the condition rr
=
Q when r
=
a. Instead of this it
makes I rrdz vanish at r
=
a,
so that the resultant radiiil tension on
any portion
of
the rim between the two
plane
faces
vanishes*;
and it
represents
the state of the disk
in the
parts
that are not too near the
edge.
*
A small
supplementary displacement corresponding
with traction
-
rr at the
edge
surface
and zero traction over the
plane
faces would be
required
for the
complete
solution of the
problem.
Se a
paper by
F. Purser in
Dublin,
Trans. R. Irish
Acad.,
vol. 32
(1902).
102]
AND ROTATING DISK
145
of
),
satisfies the
equation
(X+;*)8
(dU ^U\ .,
111 this case
U,
as a function of
r,
satisfies the
equation
X
and we also have
9w__
X
(^JJ,U\
3w_
dU
dz X +
g/iVs^"*"
r
j'
dr~~dz'
^^'
from which we
may
deduce the
equation
dz^
~
V(X
+
/x)
^^^>
These
equations,
with the condition that / rrdz vanishes when ?=
a,
determine U
and
', apart
from a
displacement
which would be
possible
in a
rigid body;
and we
may
impose
the conditions that U and w vanish at the
origin (r=0,
2
=
0),
and that
2nr,
which
is
equal
to
dU/dz -dw/dr,
also vanishes there. We should then find that
U,
w are
given by
the
equations
t^
=
"J|'
(1
-
<r) {(3
+
<r)
a^-
(1 +<r) r^]
+
%^^-
<r
(1 +<r) {P-Zz''),
from these
equations
we should deduce the
following expressions
for the
sti-ess-com-
ponents
:

2 1
.^
\
('!)
e(9
=
Y{(3
+
,r)a^-(l+3<r)r2}
+
Y<^I~(^'-322)-J
When there is a circular axle-hole of radius a' we have the additiona:l condition that
j
rrdz
=
when
r=a',
but now the
displacement may
involve terms which would be
infinite at the axis. We should obtain the
complete
solution
by adding
to the above
expressions
for U and w terms U' and
v', given by
the
equations
^, / [
(72)
and these
displacements correspond
with additional stresses
given by
the
equations
;?=ej(3
+
<r)(a'-;f),
r^=?^(3
+
.)(a'H^f);
(73)
these are to be added to the
expressions given
in
(71)
for it and 66.
10
CHAPTER VI.
EQUILIBRIUM
OF ^OLOTROPIC ELASTIC SOLID BODIES.
103.
Symmetry
of structure.
The
dependence
of the stress-strain relations
(25)
of Article 72
upon
the
directions of the axes of reference has been
pointed
out in Article 08. The
relations are
simplified
when the material exhibits certain kinds of
symmetry,
and the axes of reference are
suitably
chosen. It is
necessary
to
explain
the
geometrical
characters of the kinds of
symmetry
that are observed in
various materials. The nature of the
aeolotropy
of the material is not
completely
determined
by
its elastic behaviour alone. The material
may
be
leolotropic
in
regard
to other
physical
actions,
e.g.
the refraction of
light.
If,
in an
aeolotropic body,
two lines can be
found,
relatively
to which all the
physical
characters of the material are the
same,
such lines are said to be
"equivalent."
Different materials
may
be
distinguished by
the distributions
in them of
equivalent
lines. For the
present,
we shall confine our attention
to the case of
homogeneous
materials,
for which
parallel
lines in like senses
are
equivalent ;
and we have then to consider the distribution of
equivalent
lines
meeting
in a
point.
For some
purposes
it is
important
to observe that
oppositely
directed lines are not
always equivalent.
When certain
crystals
are
undergoing changes
of
temperature, opposite
ends of
particular
axes
become
oppositely
electrified
;
this is the
phenomenon
of
pyro-electricity.
When certain
crystals
are
compressed
between
parallel planes,
which are
at
right angles
to
particular
axes,
opposite
ends of these axes become
oppositely
electrified
;
this is the
phenomenon
of
piezo-electi'icity
*. We
accordingly
consider the
properties
of a material relative to
rays
or directions
of lines
going
out from a
point;
and we determine the nature of the
symmetry
of a material
by
the distribution in it of
equivalent
directions.
A
figure
made
up
of a set of
equivalent
directions is a
geometrical figure
exhibiting
some kind of
symmetry.
*
For an outline of the main facts in
regard
to
pyro-
and
piezo-electricity
the reader
may
consult
Mascart, Leijons
sur Vdectriciti et le
inagnetiume,
t.
1, Paris, 1896,
or
Liebisch,
Physikaliiche Krystallographie, Leipzig,
1891.
103,
104]
SYMMETRY
147
104. Geometrical
symmetry*.
When a surface of revolution is turned
through any angle
about the axis
of
revolution,
the
position
of
every point,
which is on the surface but not
on the
axis,
is
changed
;
but the
position
of the
figure
as a whole is un-
changed.
In other
words,
the surface can be made to coincide with
itself,
after an
operation
which
changes
the
positions
of some of its
points. Any
geometrical figure
which can be
brought
to coincidence with
itself,
by
an
operation
which
changes
the
position
of
any
of its
points,
is said to
possess
"symmetry."
The
operations
in
question
are known as
"covering operations";
and a
figure,
which is
brought
to coincidence with itself
by any
such
operation,
is said to
"
allow
"
the
operation.
The
possible covering operations
include
(1)
rotation,
either
through
a definite
angle
or
through any angle
whatever,
about an
axis,
(2)
reflexion in a
plane.
A
figure,
which allows a rotation
about an
axis,
is said to
possess
an
"
axis of
symmetry
"
;
a
figure,
which
allows reflexion in a
plane,
is said to
possess
a
"
plane
of
symmetry."
It can be shown that
every covering operation,
which is neither a rotation
about an axis nor a reflexion in a
plane,
is
equivalent
to a combination
of such
operations.
Of such combinations one is
specially important.
It
consists of a rotation about an axis combined with a reflexion in the
perpendicular plane.
As an
example,
consider an
ellipsoid
of semiaxes
a, b,
c
;
and
suppose
that it is cut in half
along
the
plane (a, b),
and
thereafter let one half be
rotated,
relatively
to the
other,
through ^tt
about
the axis
(c).
The
ellipsoid
allows a rotation of amount
tt
about each
principal
axis,
and also allows a reflexion in each
principal plane
;
the solid
formed from the
ellipsoid
in the manner
explained
allows a rotation of
amount
^tt
about the c
axis,
combined with a reflexion in the
perpendicular
plane,
but does not allow either the rotation alone or the reflexion alone.
A
figure
which allows the
operation
of rotation about an axis combined witii
reflexion in a
perpendicular plane
is said to
possess
an
"
axis of
alternating
symmetry."
A
special
case of the
operation just
described arises when the
angle
of rotation about the axis of
alternating symmetry
is ir. The effect of the
operation, consisting
of this rotation and reflexion in a
perpendicular plane,
is to
replace every ray going
out from a
point by
the
opposite ray.
This
operation
is known as
"
central
perversion,"
and the direction of the cor-
responding
axis of
alternating symmetry
is
arbitrary
;
a
figure
which allows
this
operation
is said to
possess
a
"
centre of
symmetry."
It can be shown that the effect of
any
two,
or
more,
covering operations,
performed
successively,
in
any
order,
is either the same as the effect of
*
The facts are stated in
greater
detail and the
necessary proofs
are
given by Schoenflies,
Krystalhysteme
und
Ki-ystalhtructur,
Leipzig,
1891. Eeference
may
also be made to H. Hilton.
Mathematical
Crystallography
and the
Theory of Groups of
movements, Oxford,
1903.
102
148 EFFECT OF SYMMETRY
[CH.
VI
a
single covering operation,
or else the first and last
positions
of
every point
of the
figure
are identical. We include the latter case in the former
by
introducing
the
"
identical
operation
"
as a
covering operation
;
it is the
operation
of not
moving any point.
With this convention the above state-
ment
may
be
expressed
in the form :

the
covering operations
allowed
by any
symmetrical figure
form a
group.
With
every covering operation
there
corresponds
an
orthogonal
linear
transformation of coordinates. When the
operation
is a rotation about an
axis,
the determinant of the transformation is
+
1
;
for
any
other
covering
operation,
the determinant is

1. All the
transformations,
that
correspond
with
covering operations
allowed
by
the same
figure,
form a
group of
linear
substitutio7is.
105. Elastic
symmetiry.
In an
isotropic
elastic solid all
rays going
out from a
point
are
equivalent.
If an
aeolotropic
elastic solid shows
any
kind of
symmetry,
some
equivalent
directions can be found
;
and the
figure
formed with them is a
symmetrical
figure,
which allows all the
covering operations
of a certain
group.
With
this
group
of
operations,
there
corresponds
a
group
of
orthogonal
linear sub-
stitutions
;
and the
strain-energy-function
is unaltered
by
all the substitutions
of this
group.
The effect of
any
such substitution is that the
components
of
strain,
refei'red to the new
coordinates,
are linear functions of the
components
of
strain,
referred to the old coordinates. It will be convenient
to determine the relations between elastic
constants,
which must be satisfied
if the
strain-energy-function
is
unaltered,
when the
strain-components
are
transformed
according
to such a substitution.
Let the coordinates be transformed
according
to the
orthogonal
scheme
104,
105]
ON
STRAIN-ENERGY-
FUNCTION
149
of
central
perversion.
The
corresponding
substitution is
given by
the
equations
x'
=
-x,
y'
=
-y,
z'=-z.
This
substitution does not affect
any component
of
strain,
and we
may
conchide that the elastic
behaviour of a material is in no
way dependent
upon
the
presence
or absence of central
symmetry.
The absence of such
symmetry
in a material could not be detected
by experiments
on the relation
between stress and strain.
It remains to
determine the conditions which must hold if the strain-
energy-function
is
unaltered,
when the
strain-components
are transformed
by
the
substitutions that
correspond
with the
following operations:

(1)
reflexion in a
plane, (2)
rotation about an
axis,
(3)
rotation about an
axis
combined with reflexion in the
plane
at
right angles
to the axis.
We shall take the
plane
of
symmetry
to be the
plane
of
x,
y,
and the
axis of
symmetry,
or of
alternating symmetry,
to be the axis of z. The
angle
of rotation will be taken to be a
given angle
6,
which will not in
the first instance be
thought
of as
subject
to
any
restrictions.
The conditions that the
strain-energy-function may
be
unaltered,
by any
of the substitutions to be
considered,
are obtained
by substituting
for
e^^j;,...,
in the form
c^-^e^^x'
+
,
their values in terms of
exx, ,
and
equating
the
coefficients of the several terms to their coefficients in the form
Cne-xx
+
The substitution which
corresponds
with reflexion in the
plane
of
{x, y)
is
given by
the
equations
x'^x,
y'
=
y,
z'
=
-z;
and the formulae
connecting
the
components
of strain referred to the two
systems
of axes are
The conditions that the
strain-energy-function may
be unaltered
by
this
substitution are
Ci4

C].,
=
C24
=
C25
=
Cjj
=
C35
=
Cj6
=
C5C
=
(a)
The substitution which
corresponds
with rotation
through
an
angle
about the axis of z is
given by
the
equations
x
=
X 0,0?, S
-\-
y
sva.
,
y'
=

xsm 6
+
y
co?,6,
z'
=
z; (3)
and the formute that connect the
components
of strain referred to the two
systems
of axes are
^odif
=
^xx
cos^ 8
-f
eyy
sin- 6
+
Cxy
sin 6 cos
6,
e,yy'
=
Bxx
sin^ Q
-I-
e^y
cos"

exy
sin d cos
d,
^^^=^-
_ \
(4)
ex/i
=
ey2
cos d

Bzx
i^in
6,
e^x!
=
Syz
sin +
e,x
cos
0,
e^^
=

lexx
in ^ cos -f
%eyy
sin ^ cos ^ -f
Sxy (cos=
Q
-
%\v? 0^.
150

EFFECT OF SYMMETRY
[CH.
VI
The
algebraic
work
required
to determine the conditions that the
strain-energy
-function
may
be unaltered
by
tliis substitution is more
complicated
than in the cases of central
perversion
and reflexion in a
plane.
The
equations
fall into sets
connecting
a small
number of
coefficients,
and the relations between the coefficients involved in a set of
equations
can be obtained without much
difficulty.
We
proceed
to sketch the
process.
We have the set of
equations
''ii
=
'^11
"^os* 5
+
20,2
sin^ 5 cos'' 8
+
Cjj
sin'' 5
-
4c,j
cos^ ^ sin 5
-
4^26
sin' 6 cos 6 +
4c^
sin^ 6 cos^
6,
Cjj
=
c,i
sin* tf +
2c,2
sin' 6 cos^
6+c^
cos* 6
+
4cj8
sin' 5 cos 6
+
4c2g
cos' 5 sin fl
-f
ic^
am^e cos^
6,
c,2
=
Cj,
sin^ 6 cos^ ^
-t-Cu (cos*
6
+
sin*
6)
+
c^^
sin''' 6 cos'-'
+
2
(c,g
-
c^J
sin 6 cos 6
(cos^
6
-
sin^
6)

4cgf,
sin^ 6 cos^
6,
Cjg
=
Cn
sin^ 5 cos''' 5

2c,2
sin'-* 5 cos'^ 5 +
C22
sin^ 5 cos'-'
+
2
(c,6
-
Cjg)
sin 5 cos 6
(cos^
5
-
sin^
6)
+
<^6 (cos'
5
-
sin'
dy^,
c,j
=
CjJ
cos' 6 sin
-
c,2
sin cos 6
(cos''
5
-
sin'
fl)

C22
sin' cos
+
Cj,,
cos' 5
(cos'
6-Z sin'
5)
+
C2fl
sin' 6
(3
cos' 5
-
sin'
6)
-
2%
sin 6 cos
(cos'
5
-
sin'
6),
%=<-'ii
sit'' 5 cos 6 +
("12
sin 6 cos 5
(cos'

sin'
6)
-
c^
cos' 6 sin
+
c^
sin' 5
(3
cos' 6

sin'
0)
+
C26
cos' 6
(cos'
0-3 sin'
6)
+
2cj8
sin cos 6
(cos'
-
sin'
0).
The
equations
in this set are not
independent,
as is seen
by adding
the first four. We
form the
following
combinations :

%
+
'^26
=
('^u
~
"22)
sin cos 6
+
(c^
+
Cjj) (cos*
-
sin*
0),
''ii
~
*22
=
(''ii
~
"^22) (cos*
-
sin*
0)

4
(Cjj
+
02^)
sin 6 cos
6,
from which it follows
that,
unless sin 6
=
0,
we must have
Cll
=
C22, C26= ~<'l6-
When wo use these results in
any
of the first four
equations
of the set of six we find
(cji
-
012
-
2C||5)
sin' 6 cos' 6
+
2cjj
sin 6 cos 6
(cos'
6

sin'
0)
=
0,
and when we use them in either of the last two
equations
of the same set we find
-
8c,5
sin' 6 cos'
+
(c,,
-
C12
-
2cjj)
sin cos 6
(cos'
6
-
sin'
0)
=
;
and it follows
that,
if neither sin 6 nor cos 6
vanishes,
we must have
<'ii(i
=
5(''n'~''i2)i '^\6

^-
Again
we have the set of
equations
c,3
=
c,3
cos'
+
C23
sin'
-
2c'3
sin cos
6,
C23
=
C13
sin'
+
C23
cos'-
+
2C35
sin cos
0,
^36 (''is
~
''23)
sin cos
+C38 (cos'

sin'
0) ;
from which it follows
that,
unless sin
=
0,
we must have
^I3~''23' ''36~'''
In like manner we have the set of
equations
C44
=
C44
cos'
+
Cjj
sin' 6 +
%c^^
sin cos
0,
Cj5
=
C44
sin'
+
C55
cos'
-
2C45
sin cos
0,
'^v,
=
-
(1^44
-
c^j)
sin cos +
C4J (cos'
-
sin'
0) ;
from which it follows
that,
unless sin
=
0,
we must have
In like manner we have the set of
equations
C34
=
C34
cos +
C35
sin
0,
c.jj
=
-
C34
sin
+
C35
cos
;
105]
ON
STRAIN-ENEROY-FUNCTION
151
from which it
follows,
since cos
6^1,
that we must have
Finally
we have the set of
equations
c,4
=
c,4
cos' ^ +
f,5
cos2 6 sin 6
+
c^
sin^ cos 5
+
c^
sin' 6
-
2c4
cos^ fl sin 5
-
2c5
sin^ 6 cos
0,
''is
= -
(^14
cos2 ^ sin fl
+
c,5
cos' 5
-
Csi4
sin' d
+
Cjjs
sin* tf cos fl
+
2e4a
sinM cos 5
-
2c5j
cos2 fl sin
5,
24
=
Cn
sin2 5 cos d
+
Ci5
sin' ^ +
C24
cos' d
+
c.^
cos^ ^ sin fl
+
2tf48
cos^ 6 sin 6
+
ic^^
sin^ fl cos
6,
25=-i4sin'^
+
Ci,sin2 5cos5-C24Cos2|9sin5
+
C25C08'5-2c4(,sin2
5co8d+2<!5,cos2tfsind,
4
=
^14
<^os2 5 sin fl
+
c,5
sin2 5 cos 5
-
C24
cos''' 6 sin 5
-
Cjj
sin^ 5 cos d
+
(C4
cos 5
+
C5
sin
6) (cos^
5
-
sin^^
6),
'50
=
-
i4
s'"' ^ cos i9
+
(^15
cos2 5 sin ^
+
C24
sin^ ^ cos 5
-
c.^
cos^ $ sin 5
-
(C4
sin 5
-
C5g
cos
^) (cos2
d
-
sin-
5).
From these we form the combinations
i4
+
C24=(Ci4+24)
COS (9
+
(cjs
+
C25)
sin
5,
C16
+
25
=
-
(Cl4
+
C24)
sill ^
+
(Cl5
+
%)
COS fl
;
and it
follows,
since cos
5^ 1,
that we must have
I4
+
C24
=
0. C,5
+
C'25
=
0.
Assuming
these
results,
we form the combinations
(Cl4
-
Cos)
=
(Cl4
-
m)
cos ^
-
(0,6
+
C46)
S'"
^.
(Ci6
+
^46)
=
(c,4
-
cj
sin 5 +
(c,5
+
C4,)
cos e
;
from which it follows that
Assuming
these
results,
we
express
all the coefficients in the above set of
equations
in terms
of
Cjij
and
c^ ,
and the
equations
are
equivalent
to two :

C4g (1
-
cos' ^ +
3 sin- 6 cos
6)
-
c^ (3
cos^ 5 sin 5
-
sin'
6)
=
0,
C4g (3
cos^ ^ sin 5
-
sin'
6)
+
c^(l-
cos' 5
-H
3 sin^ 5 cos
d)
=
0.
The condition that these
maj'
be
compatible
is found to reduce to
(1
-
cos
5) (1
-f
2 cos
6)'^
=
;
so
that,
unless cos d=
i,
we must have
'40
=
f
56
=
0-
We have thus found
that,
if the
strain-energy-function
is unaltered
by
a substitution which
corresponds
with rotation about the axis
z,
through any
angle
other than
tt,
^tt, |7r,
the
following
coeflicients must vanish :

^16) ^28, C35, C46, Csc, C45, Cij, Cnj, Ci5, Cns, 634, C35 , (O)
and the
following equations
must hold
among
the
remaining
coefficients :

Cii
=
C22, C13
=
Cj:), C44
=
C55, C|j8
=
-^ (Cji

Cia) (0)
When the
angle
of rotation is
tt,
the
following
coefficients vanish :

Cl4i ^24) *^15 *^25' ^4> ^5i> ^34* ^35 J \l )

no relations between the
remaining
coefficients are involved. When the
angle
of rotation is
^tt,
the
following
coeflScieuts vanish :

^3S> (^46' Cjg, C45, Ci4, C24, C15, Cjs, Cs4, C35, (,o)
and the
following equations
connect the
remaining
coefficients :

Cn
=
C.22, Ci3=C23, C44
=
C|S5, C.jt,
=
Cie, (9)
152 NATURE OF THE SYMMETRY
[CH.
VI
When the
angle
of rotation is
^tt,
the
following
coefficients vanish :

Ci6> Cj8i Css, C46, Cs4, C3J ;

(^^)
and the
following equations
connect the
remaining
coefficients :

"ii
^
^22) Cjs
=
Cja, Cu
=
Cm, C(ia
=
^ (Cji Cu),
.(11)
^14

^-24 ^56 > *?15

^25

^43. )
In like
manner,
when the axis of z is an axis of
alternating symmetry,
and the
angle
of rotation is not one of the
angles
tt,
^tt, ^tt,
the same
coefficients vanish as in the
general
case of an axis of
symmetry,
and the
same relations connect the
remaining
coefficients. When the
angle
is
ir,
we
have the case of central
perversion,
which has been discussed
already.
When
the
angle
is
^tt,
the results are the same as for direct
symmetry.
When
the
angle
is
^tt,
the results are the same as for an axis of direct
symmetry
with
angle
of rotation
7r.
106.
Isotropic
solid.
In the case of an
isotropic
solid
every plane
is a
plane
of
symmetry,
and
every
axis is an axis of
symmetry,
and the
corresponding
rotation
may
be of
any
amount. The
following
coefficients must vanish:

Cl4. Ci5, C16, C24, C25, C26, C34, C35, Css, C45, C46, Cjs
...(12)
and the
following
relations must hold between the
remaining
coefficients :

Cn
=
C22
=
C33, C23
=
C3i
=
C]2, C44
=
C55
=
Cgs
=
^^11
"
Cjo) (13)
Thus the
strain-energy-function
is reduced to the form
i
Cii
i^xx
+
e^y
+
e'',,)
+
C12
(Byye^^
+
Bi^exx
+
e^x
eyy)
+
1 (c
-
CiO
(eV
+
6\x
+
e'xy), (14)
which is the same as that assumed in Article 68.
107.
Symmetry
of
crystals.
Among aeolotropic materials,
some of the most
important
are
recognized
as
crystalline.
The structural
symmetries
of
crystalline
materials have been
studied
chiefly by examining
the
shapes
of the
crystals.
This examination
has led to the
construction,
in each
case,
of a
figure,
bounded
by planes,
and
having
the same
symmetry
as is
possessed
in common
by
the
figures
of all
crystals,
formed
naturally
in the
crystallization
of a material. The
figure
in
question
is the
"
crystallographic
form"
corresponding
with the
material.
F. Neumann*
propounded
a
fundamental
principle
in
regard
to the
physical
behaviour of
crystalline
materials. It
may
be stated as follows :
Any
kind of
symmetry,
which is
possessed by
the
crystallographic
form
of a
material,
is
possessed
by
the material in
respect
of
every physical
*
See liis
Vorlesungen
ilber die Theorie der
Elasticitat,
Leipzig,
1885.
105-107]
OF CRYSTALLINE MATERIALS
153
quality.
In other words we
may say
that a
figure consisting
of a
system
of
rays, going
out from a
point,
and
having
the same
symmetry
as the
crystallographic
form,
is a set of
equivalent rays
for the material. The law
is an induction from
experience,
and the evidence for it consists
partly
in
a
posteriori
verifications.
It is to be noted that a
cry.stal may,
and
generally does, possess,
in
respect
of some
physical qualities,
kinds of
symmetry
which are not
possessed by
the
crystallographic
form. For
example,
cubic
crystals
are
optically isotropic.
Other
e.xaraples
are afforded
by
results obtained in Article 105.
The laws
of
the
symmetry of crystals
are laws which have been observed to be
oljeyed
by crystallographic
forms.
They may
be
e.xpressed
most
simply
in terms of
equivalent
rays,
as follows:

(1)
The number of
rays, equivalent
to a chosen
ray,
is finite.
(2)
The numter of
rays, equivalent
to a chosen
ray, is,
in
general,
the .same for all
positions
of the chosen
ray.
We take this number to be N-
1,
so that there is a set of X
equivalent rays.
For
special positions, e.g.
when one of the
rays
is an axis of
symmetry,
the number of
rays
in a sot of
equivalent rays
can be less than N.
(3)
A
figure,
formed of N
equivalent rays,
is a
symmetrical figure, allowing
all the
covering oi)erations
of a certain
group. By
these
operations,
the N
equivalent rays
arc
interchanged,
so that each
ray
comes at least once into the
position
of
any equivalent ray.
Any figure
formed of
equivalent rays
allows all the
covering operations
of the same
group.
(4)
When a
figure,
formed of ^V
equivalent rays, possesses
an axis of
symmetry,
or an
axis of
alternating symmetry,
the
corresponding angle
of rotation is one of the
angles
TT,
%n, iff,
^TT*.
It can be shown that there are 32
groups
of
covering operations,
and no
more,
which
obey
the laws of the
symmetry
of
crystals.
With each of these
groups
there
corresponds
a claas of
crystals.
The
strain-energy-function corresponding
with each class
may
l>e
written down
by making
use of the results of Article
105;
but each of the forms which
the function can take
corresponds
with more than one class of
crystals.
It is
necessary
to describe
briefly
the
symmetries
of the classes. For this
purpose
we shall now introduce
a few definitions and
geometrical
theorems
relating
to axes of
symmetry
:

The
angle
of rotation about an axis of
symmetry,
or of
alternating symmetry,
is
27r/,
where n is one of the numbers :
2, 3, 4,
6. The axis is described as
"
M-gonal."
For
n
=
2, 3, 4,
6
respectively,
the axis is described as
"digonal," "trigonal," "tetragonal,"
"
hexagonal."
Unless otherwise stated it is to be understood that the
M-gonal
axis is an
axis of
symmetry,
not of
alternating symmetry.
The existence of a
digonal
axis,
at
right angles
to an
ra-gonal
axis, implies
the existence
of n such axes
;
e.g.
if the axis z is
tetragonal,
and the axis x
digonal,
then the axis
y
and
the lines tliat bisect the
angles
between the axes of x and
y
also are
digonal
axes.
The existence of a
plane
of
symmetry, passing through
an
n-gonal
axis, implies
the
existence of n such
jjlanes; e.g.
if the axis z is
digonal,
and the
plane
.r
=
is a
plane
of
.symmetry,
then the
plane y
=
also is a
plane
of
symmetry.
If the
n-gonal
axis is an axis of
alternating symmetry,
the two results
just
stated still
hold if n is uneven
; but,
if n is
even,
the number of axes or
planes implied
is
^n.
*
The restriction to tliese
angles
is the
expression
of the "law of rational indices."
154 CRYSTALS CLASSIFIED
[CH.
VI
108. Classification of
crystals.
The
symmetries
of the classes of
crystals may
now be described
by
reference to the
groups
of
covering operations
which
coi-respond
with them
severally
:

One
group
consists of the identical
operation
alone;
the
corresponding figure
has no
symmetry;
it will be described as
"asymmetric."
The identical
operation
is one of the
operations
contoined in all the
groups.
A second
group contains,
besides the identical
operation,
the
operation
of central
perversion only ;
the
symmetry
of the
corresponding
figure
will be described as "central." A third
group contains,
besides the identical
operation,
the
operation
of reflexion in a
plane only ;
the
symmetry
of the
corresponding
figure
will be described as
"equatorial."
Besides these three
groups,
there are 24
groups
for which there is a
"
principal
axis
"
;
that is to
say, every
axis of
symmetry,
other than
the
principal axis,
is at
right angles
to the
principal
axis
;
and
every plane
of
symmetry
either
passes through
the
principal
axis or is at
right angles
to that axis. The five
remaining groups
are characterised
by
the
presence
of four axes of
trigonal symmetry
equally
inclined to one
another,
like the
diagonals
of a cube.
When there is an
ra-gonal principal axis,
and no
plane
of
symmetry through it,
the
symmetry
is described as
"ji-gonal" ;
in case there are
digonal
axes at
right angles
to the
principal
axis,
the
symmetry
is further described as
"holoaxial";
in case there is a
plane
of
symmetry
at
right angles
to the
principal axis,
the
symmetry
is fiu'ther described as
"equatorial";
when the
symmetry
is neither holoaxial nor
equatorial
it is further
described as
"polar."
When there is a
plane
of
symmetry through
the
n-gonal principal
axis,
the
symmetry
is described as
"di-n-gonal";
it is farther described as
"equatorial"
or
"polar," according
as there
is,
or is
not,
a
plane
of
symmetry
at
right angles
to the
principal
axis.
When the
principal
axis is an axis of
alternating symmetry,
the
symmetry
is described
as
"di-m-gonal alternating,"
or
"ji-gonal alternating," according
as there
is,
or is
not,
a
plane
of
symmetry through
the
principal
axis.
The
appended
table shows the names* of the classes of
crystals
so far
described,
the
symbols
t
of the
corresponding groups
of
covering operations,
and the luimbers of the
classes as
given by Voigt|.
It .shows also the
grouping
of the classes in
systems
and
the names of the classes as
given by
Lewis
.
The
remaining groups,
for which there is not a
principal axis, may
be described
by
reference to a cube
;
and the
corresponding crystals
are
frequently
called
"
cubic,"
or
"tesseral," crystals.
All such
crystals possess,
at
any point,
axes of
symmetry
which are
distributed like the
diagonals
of a
cube, having
its centre at the
point,
and
others,
which
are
parallel
to the
edges
of the cube. The latter
may
be called the "cubic axes." The
symmetry
about the
diagonals
is
trigonal,
so that the cubic axes are
equivalent.
The
symmetry
with
respect
to the cubic axes is of one of the
types previously
named. There
are five cla8.ses of cubic
crystals,
which
may
be
distinguished by
their
symmetries
with
respect
to these axes. The table shows the names of the classes
(Miers, Lewis),
the
symbols
of the
corresponding groups (Schoenflies),
the
numbers
of the classes
(Voigt),
and
the character of the
symmetry
with
respect
to the cubic axes.
*
The names are those
adopted by
H. A.
Miers,
Mineralogy,
Oxford,
1902.
t
The
symbols
are those used
by
Schoenflies in his book
Krijstalhysteme
uiid
KryslaUstructur.
t Ilapportu prisenties
au
Congres
International de
Physique,
t.
1, Paris,
1900.

W. J.
Lewis,
Treatise on
Crystallograplig, Cambridge,
1899. The older classification in six
(sometimes seven) "systems"
as
opposed
to the 32 "classes" is
supported by
some modern
authorities. See V.
Goldschmidt,
Zeitschr.
f. Krystallographie,
Bde. 31 and 32
(1899).
108] BY THEIR
SYMMETRIES
155
System
15G EFFECT OF SYMMETRY
[CH.
VI
109.
Elasticity
of
crystals.
We can now
put
down the forms of the
strain-energy-function
for the
different classes of
crystals.
For the classes which have a
principal
axis
we shall take this axis as axis of z
;
when there is a
plane
of
symmetry
through
the
principal
axis we shall take this
plane
as the
plane (, z)
;
when there is no such
plane
of
symmetry
but there is a
digonal
axis at
right
angles
to the
principal
axis we shall take this axis as axis of
y.
For the
crystals
of the cubic
system
we shall take the cubic axes as coordinate axes.
The classes will be described
by
their
group symbols
as in the tables of
Article 108
;
we shall first write down the
symbol
or
symbols,
and then the
corresponding strain-energy-function;
the omitted terms have zero
coefficients,
and the constants with different suffixes are
independent.
The results* are
as follows :

Groups
0,, jSj

(21 constants)
^CiiC^xx
"r
(^n^xx^yy
"r
^is^xx^zz
'*'
^i-i^zx^yz
"f"
Ci56xx^2x
'
^le^xx^xt/
I
2^^^~yy
'
^^i^yy^zz
"r
(^n^yy^yz
+
C-25^yy^zx
i
^-^ti^yy^xy
+
5033622
-r
C'^e^z^yz
I
(^M^zz^zx
+
C-its^zz^xy
+
i^^u^'yz
+
^a^yz^zx
+
^m^yz^xy
"r
T^C^r>^2X
*T"
G^n^zx^xy
"T
2^i)6^~Xy'
Groups
S, C'i, (7/

(13) constants)
a^ll^asr
I
Gi^xx^yy
""
O^^Gxx^zz +
C\Q6xxGxy
+
i'^'a^'yy
+
C^sSyyCzz
-\-
C^s^yyexy
"T
"^OsqB
zz *i
C^Gzz^xy
+
i^C^e xy
Groups
V, G/,
F''
(9 constants)
2^nB'xx
"T
Ci2^xx^yy
i
^la^xz^zz
-f-
^C.,.,e'yy +0236^^,622
+
iCy^S'zz
+
iCu^'yz
+
^'^'xB'zx
+
jCesC xy
Groups
C3, /S'a

(7 constants)
aCiifi XX
+
CuexxSyy
+
0136^0:622
+
C]j6xx6y2
+
Cise^rx^zx
+
20ll6"yjy
-|-
Cl36j,y622 Cu6yye,ii C^iByye^x
+
5C336
22
+
^6446 yz

^IhByz^xy
T
iCuB zx
+
CuBziBxy
+
4 (Cii

Cjo)
Cxy
*
The results are due to
Voigt.
109,
110]
ON
STRAIN-ENERGY-FUNCTION
157
Groups
Da, Cs", /Sg"

(6 constants)
+
i(c>i
-
c,j) e^.
Groups
Cs*, A*, C's, A, Co*, ^^6", A''
(5 constants)
Groups
Cj, Si, C'/

(7 constants)
iCiiS'xx
+
Ci^^xx^yy
+
CijCxx^zz +
C^^^xx^xy
>
2^n^ yy
+
^136^2,622

^le^yy^xy
+
jCjje^jj.
+
j^Cne yi
\-
T^C^iB'zx
"^^^m^xy
Groups
A, CV, >S'4", A*
(6 constants)
Groups
y, 0, y'', T",
0''
(3 constants)
2^11 yfi'xx
+
S'yy
+
e'2z)
+
Cj2
(eyyCzz
+
^tz^xx
+
^xx^yy)
+
^"m (*"!/i
+
^'zx
"I"
^"xtf)'
110. Various
types
of
symmetry.
Besides the kinds of
symmetry
shown
b}' crystals
there are others which
merit
special
attention. We note the
following
cases :

(1)
The material
may possess
at each
point
three
planes
of
symmetry
at
right angles
to each other.
Taking
these to be the coordinate
planes
the formula for the
strain-energy-function
would be
^W^Ae^xx-^
Be'yy
-f-
Ce^z +
iFeyye^z
-f IGe^zCxx
+
SiTex^e^
^L&\2-\-Me\^-^N(i\y
(15)
This formula contains a number of those which have been obtained for
various classes of
crystals.
(2)
The material
may possess
an axis of
symmetry
in the sense that
all
i-ays
at
right angles
to this axis are
equivalent. Taking
the axis of
symmetry
to be the axis of
z,
the formula for the
strain-energy-function
would be
2 If
=
^
{e'xx
+
eV,,/)
+
Ce\2 +
2i?'
{Cyy
+
e^x)
e
-f-
2
(4
-
2iV) e^Cyy
-vL{e\2
+
e\x)
+
Ne\y
(16)
Bodies wliich show this kind of
symmetry may
be described as
"
transversely
158 MATERIAL HAVING THREE
[CH.
VI
isotropic."
It is to be noted that cubic
crystals
are not
transversely isotropic.
For a cubic
crystal
A=B=C,
F=G
=
H,
L
=
M=N,
but the relation
H
=
A-2N does not hold.
(3)
The material
may possess symmetry
of one of the kinds
already
discussed,
or of some other
kind,
but the axes of
symmetry may
be directed
differently
at different
points*.
In such cases we
may
be able to choose
a
system
of
orthogonal
curvilinear coordinates so that the normals to the
orthogonal
surfaces at a
point
become lines with reference to which the
strain-energy-function
is
simplified.
For
example,
formula
(15) might
hold
for axes of
x,
y,
z directed
along
the normals to the surfaces of reference
at a
point,
or the material
might
be
transversely isotropic
with reference
to the normals and
tangent planes
of a
family
of surfaces. This kind of
symmetry
of structure
may
be
possessed by
curved
plates
of metal. When
a
body possesses symmetry
in this
way
it is said to
possess
"
curvilinear
aiolotropy."
111. Material with three
orthogonal planes
of
symmetry.
Moduluses.
In the ca-ses where formula
(15) holds, Young's
modulus E for an
arbitrary
direction
(;,, !, Mj)
is
given by
the
equation
^3 ^1 '2 -"'a
where
E^, E^, E^
are the
Young's
moduluses for the three
principal directions,
and the
"8 and F'h are
given by
s\ich
equations
as
E-E,'^E,r'E;'^
F:
'^
F,
"^
Fr
.(17)
E,
BC-F'^
110-112]
ORTHOGONAL PLANES OF SYMMETRY 159
The
rigidities
for the
pairs
of axes at
right angles
to the
planes
of
ymmetry
are
L, M,
N.
With the same notation we could show that the Poisson's ratios for contractions
parallel
to the axes
ofy
and 2
respectively,
when the stress is tension across the
planes ^=con8t.,
are
fory, Ei{\li.N-\IF^\
and for
2,
Ey{\j2M-\lt\) (20)
The values for other
pairs
of directions can be written down without
difficulty (Article 73).
With the same notation we
may
show that the modulus of
compression
is the
reciprocal
of
1 11 222111
E^'^E^^^'Ei^Fi'^F^^F^~L~M~N
^^^^
In the case of cubic
crystals
we
may
show that the value of
E, Young's
modulus for
tension in direction
{I, m, n),
is
given by
the
equation*
i
=^i+{i~^^^^}
("*'"' +'''^+^''"'^ (^^)
Provided that the coefficient of the second term is
positive,
^ is a maximum in the
directions of the
principal axes,
and a minimum in the directions of lines
equally
inclined
to the three
principal axes;
further it is
stationary
without
being
a maximum or a
minimum in the directions of lines
bisecting
the
angles
between two
principal axes,
and
remains constant for all lines
given by
lmn
=
0.
112. Extension and
bending
of a bar.
As
examples
of distributions of stress in an
aeolotropic
solid
body,
we
may
take the
problems
of extension of a bar and
bending
of a bar
by
terminal
couples.
We shall
suppose
that the material
has,
at each
point,
three
planes
of
symmetry
of
structure,
so that
the
strain-energy-function
is
given by
the formula
(15) ;
we shall
suppose
also that the bar
is of uniform
section,
that the axis of z is the line of centroids of its normal
sections,
and
that the axes of x and
y
are
parallel
to
principal
axes of inertia of its normal
sections,
so
that the line of centroids and the said
principal
axes are at
right angles
to
planes
of
.symmetry.
(a)
Extension.
We
suppose
that all the
stress-components except Z^ vanish,
and take
Z^=Ee,
where t
is
constant,
and E is the
Young's
modulus of the material
corresponding
with tension
Z,.
We find the
displacement
in the form
M=-.(ri.r, v=-cr,iey,
w
=
ez, (23)
where
o-j
is the Poisson's ratio for contraction
parallel
to the axis of x when there is
tension
Z^,
and
o-^
is the
corresponding
ratio for contraction
parallel
to the axis of
y.
(6) Bending by couples.
We assume that all the
stress-components
vanish
except ^j,
and take
Z,=
-
Elt'^x,
where R is constant.
We find that the
displacement
is
given by
the
equations
u=^^R-\z'^
+
<T^x'^-(r^y^),
v
=
<r.,R-^xy,
w=-R-^xz, (24)
and that the traction across a normal section is
statically equivalent
to a
couple
about an
axis
parallel
to the axis of
y,
of moment
EI/R,
where /=
/ I a^dxdy,
the
integration being
taken over the cross-section.
The
interpretation
of the result is similar to that in Article 88.
*
A
figure showing
the variation of
IjE
with direction is drawn
by
Liebisch,
I'hysikalische
Kryitalloijraphie (Leipzig,
1891), p.
5G4.
160 voigt's results for crystals
[CH.
VI
113. Elastic constants of
crystals.
Results of
experiments.
Tlie elastic constants of a number of minerals have been determined
by
W.
Voigt* by
experiments
on the
twisting
and
bending
of rods. Some of his
principal
results are
stated here. The constants are
expressed
in terms of an unit stress of 10"
grammes' weight
jx;r square
centimetre.
For
Pyrites (cubic),
the constants are
c,i
=
3680, C44
=
1075, c,2
=
-
483,
and we have
Principal Young's modulus, ^=3530,
Principal Rigidity, C44
=
/i
=
1075 :
also
by
calculation we find
Principal
Poisson's ratio o-
=
-
1 nearly.
These results are
very
remarkable,
since
they
show that these moduluses of
pyrites
are
much
greater
than those of steel
t,
and further that a bar of the material cut in the
direction of a
principal
axis when extended
expands slightly
in a lateral direction
\.
The
modulus of
compression
is about 1070 x 10"^
grammes' weight per square centimetre,
which
is
considerably
smaller than that of steel.
The table shows the values of the constants for three other minerals for which the
energy-function
has the same form as for
Pyrites.
In this table
C44
is the
principal rigidity,
and E is the
principal Young's
modulus.
Material !
E
t
113,
114]
CUKVILINEAR iEOLOTROPY
161
Topaz
is a rhombic
crystal (of
the class
specified by
the
group K*)
whose
principal
Young's
moduluses and
rigidities
are
greater
than those of
ordinary
steel. The constauts
of formula
(15)
are for this mineral
^=2870, 5=3560, C=3000, F=90Q,
=860, ^^=1280,
Z
=
1100, i/=1350,
^=1330.
The
principal Young's
moduluses are
2300, 2890,
2650.
Barytes
is a
crystal
of the same
class,
and its constants are
.4=907,
fi
=
800, C=1074, F=273,
6'
=
275, fi'=468,
Z=122, J/=29.3,
iV=283.
These results show that for these materials
Cauchy's
reduction is not valid.
114. Curvilinear
seolotropy.
As
examples
of curvilinear
aiolotropy (Article 110)
we
may
take the
problems
of a tube
(Article 100)
and a
spherical
shell
(Article 98)
under
pressure,
when there is transverse
isotropy
about the radius vector*.
(a)
In the case of the tube we should have
8r
.(25)
r Or
I
^
, ^dU ^U \
zz
=
Ae
+
F
-^
+
H
-,
I
Or r
where H is written for A
-
2^. The
displacement
U is
given by
the
equation
C^lU^CW_AU^iF^,H)e
oH r or r^ r
of which the
complete primitive
is
U=ar''+^r-''+
^--^er,
(27)
n
being
written for
J {AjC),
and a and
/3 being arbitrary
constants The constants can be
adjusted
so that rr has the value
-jOj
at the outer surface
r=r,
and
-jo,
at the inner
surface r
=
rj.
The constant e can be
adjusted
so as to make the resultant of the
longitudinal
tension zz over the annulus
?->r>r,
balance the
pressure
n
(p^r^^
-
Pij>'o^)
on
an end of the
cylinder.
(6)
In the case of the
sphere
we should find in like manner that the radial
displacement
U satisfies the
equation
C<PU^2C<W_^ U^.^ (28)
ar' r ar r
MO that 6'
=
<ir-i +
;3r--4,
where "=
i
|l
+
8
^-t|^
,
*
Saint-
Venaut,
J. de Math.
(LiouvilU),
(S6r. 2),
t. 10
(1865).
L. E.
11
162 CURVILINEAR iEOLOTROPY
[CH.
VI
and we can find the formula
r,*--r,^\ (,n-i)C+2F
'^ ^
+
^'^0^1'
(n+i)C-2F
'^
*/'
^^^>
which
agrees
with the result obtained in
(vi)
of Article 98 in the case of
isotropy.
The cubical dilatation of the
spherical cavity
is the value of
3(I/r
when r
=
r^,
and
this is
'o**-'-!^! (n-i)C+2F
"^^'^o
(n
+
l)C-2F J
'-"^^
This result has been
applied by
Saint-Venant to the
theory
of
piezometer experiments,
in which a
discrepancy appears
to have been observed between the results obtained and
the dilatation that should
theoretically
bo found to occur if the material were
isotropic.
The solution in
(30)
contains 3
independent
constants and Saint-Venant held that these
could be
adjusted
so as to
explain
the
experiments
in
question.
CHAPTER VII.
GENERAL THEOREMS.
115. The variational
equation
of motion*.
Whenever a
strain-energy-function,
W, exists,
we
may
deduce the
equations
of motion from the Hamiltonian
principle.
For the
expression
of this
principle,
we take T to be the total kinetic
energy
of the
body,
and V to be the
potential energy
of
deformation,
so that V is the volume-
integral
of W. We
form,
by
the rules of the Calculus of
Variations,
the
variation of the
integral
\
{T

V)dt,
taken between fixed initial and final
values
{to
and
<i)
for t. In
varying
the
integral
we assume that the
displacement
alone is
subject
to
variation,
and that its values at the initial
and final instants are
given.
We denote the variation so formed
by
ij(T-V)
dt.
We denote
by
8TF,
the work done
by
the external forces when the
displace-
ment is varied. Then the
principle
is
expressed by
the
equation
sj(:T-V)dt+lBWidt
=
(1)
We
may carry
out the variation of I Tdt. We have
and therefore
/du dBu
*
Of.
Kirohhoff, Vorlesungen
uber...Mechanik, Vorlesung
11.
11-
1(J4.
VARIATIONAL
EQUATIONS [CH.
VII
Here

and
t,
are the initial and final values of
t,
and
Bu,
... vanish for
both these values.
The first term
may
therefore be omitted
;
and the
equation (1)
is then transformed into a variational
equation of
motion.
Further,
8F is
lUhWdxdydz,
and
hW^
is
given by
the
equation
hW^=
11
i
p
(XBu
+
Y8v +
ZSw) dxdydz
+
[J
{XM
+
Y^h)
+
Z,Bw)
dS.
Hence the variational
equation
of motion is of the form
-
fjlp
(XBu
+
YBv +
ZBw)
dxdydz
-
\\{XM
+ Y^Bv +
Z^Bw)
dS
=
0. ..
.(3)
Again,
6 )r IS ^

bexz
+
^

oeyy
+
. . .
+
r
oexy
,
where,
for
example,
he^x
is
dBu/dx.
Hence
liJB
W
dxdydz maybe
transformed,
by integration by parts,
into the sum of a surface
integral
and a volume
integral.
We find
jjJBWdxdydz
=
//[||
cos
(X, .)
+
1^
cos
(y, .)
+
1
cos
(.,
.)}
Bu
-III
+
...
+
..^dS
J
'
dx
dcxx
9y 9x!/
9^
^^zx
dxdydz. ...(4)
The coefficients of the variations
Bu,
... under the
signs
of volume
integration
and surface
integration
in
equation (3),
when transformed
by
means of
(4),
must vanish
separately,
and we thus deduce three differential
equations
of
motion which hold at all
points
of the
body,
and three conditions which
hold at the
boundary.
The
equations
of motion are of the
type
^-
X
- d_dW
ddW_
^
dt^"
dxdexx
^y^^xy
dzdezx'
and the surface conditions are of the
type
r

cos(ir, i')
+ ~

cos{y, v)
+
:^

cos
(z, v)
=
A
(6)
cxx
^^xy
oCzz
116.
Applications
of the variational
equation.
(i)
As an
example*
of the
application
of this method we
may
obtain the
equations
(19)
of Article 58. We have
tia
pp op
*
Cf. J.
Larmor, Cambridge
Phil. Soc.
Trans.,
vol. 14
(1885).
115,
116]
OF MOTION AND OF
EQUILIBRIUM
165
and, by
the formute
(36)
of Article
20,
we have also
Every
term of
j j j
'<"
I/L/i^
''^ ""^ ^ ^
transformed
by
the aid of the formulw of
the
type
jjj
dad^dy
=
/
j
AjAjIcos (a, v) dS,
and the
integral
will then be
transformed into the sum of a surface
integral
and a volume
integral,
in such a
way
that no
differential coefficients of
8m.,
bu^,
bu occur. We
may
collect,
for
example,
the terms
containing m.
in the volume
integral. They
are
Mji
/_!_ 3^\ _ 1_ 3^
nX
^_ 1_ ^
/
1_\
SIf
3a
\h^h^ dej h,
da
\/iJ
de^^
h^
da
\hj
de^
The
equations
in
question
can be deduced without
difficulty.
(ii)
As another
example,
we
may
obtain
equations (21)
of Article 91 and the second
forms of
equations (22)
of the same Article. For this
puriwse
we observe that
Hence the
strain-energy-function
in an
isotropic body may
be
expressed
in the form
Pf=i (X-t-2^)
A2+
2,x {^^^
+
vj'^
+
^,^)
+
2p.
r('|'?^
_
?^gUtwo
similar
terms].
and therefore the terms of the
type 2/i
( =- i-
-
^-^ )
in IF do not contribute
anything
to
the volume
integral
in the transformed
expression
for
j j
I
bWdxdydz.
Hence the
equations
of motion or of
equilibrium
can be obtained
by forming
the variation of
[ / f
[J (X
-H
2/x)
A2
+
2fi (07/
-I-
nr/
-I-
ts-/)] dvdffdz
instead of the variation of /
|
/
Wdxdydz.
The
equations (21)
and the .second forms of
equations (22)
of Article 91 are the
equations
that would be obtained
by
this
process.
The result here found is that the dift'erential
equations
of
vibration,
or of
equilibrium,
of an
isotropic
solid are the same as those of a
body possessing potential energy
of deforma-
tion
per
unit of volume
expressed by
the formula
\ {\
+
2,i)
A-^
+
2^ (nr/--or,''
+
sr,2).
166
GENERAL THEORY
[CH.
VII
The surface conditions are different in the two cases. In
MacCullagh's theory
of
optics*
it
was shown
that,
if the luminiferous tether is
incompressible
and
possesses potential energy
according
to the formula
2^i(iiTj.2+i3r/
+
nr,2),
the observed facts about reflexion and re-
fraction of
light
are accounted for
;
the surface conditions which are
required
to hold for
the
purposes
of the
optical theory
are
precisely
those which arise from the variation of the
volume
integral
of this
expression.
Larmor+ has described a
medium,
which
possesses
potential energy
in the
required manner,
as
"
rotationally
elastic." The
equations
of
motion of a
rotationally
elastic medium are
formally
identical with those which
govern
the
propagation
of electric waves in free iether.
117. The
general problem
of
equilibrium.
We seek to determine the state of
stress,
and
strain,
in a
body
of
given
shape
which is held strained
by body
forces and surface tractions. For this
purpose
we have to
express
the
equations
of the
type
dxXdexx)
Sy\dexy)
dzxdezx)
as a
system
of
equations
to determine the
components
of
displacement,
u, V, w;
and the solutions of them must be
adapted
to
satisfy
certain
conditions at the surface S of the
body.
In
general
we shall take these
conditions to
be,
either
(a)
that the
displacement
is
given
at all
points
of
S,
or
(b)
that the surface tractions are
given
at all
points
of S. In case
(a),
the
quantities
u, v,
w have
given
values at S
;
in case
(b)
the
quantities
of
the
type
dW
, .
dW , .dW . ,
A
,
=
^

cos
(x, v)
+
^

cos
(y, v)
+
^

cos
(z, v),
OCxx
^^xy
^^zx
have
given
values at S. It is clear
that,
if
any displacement
has been
found,
which satisfies the
equations
of
type (7),
and
yields
the
prescribed
values for
the surface
tractions,
a small
displacement
which would be
possible
in a
rigid
body may
be
superposed
and the
equations
will still be satisfied
;
the strain
and stress are not altered
by
the
superposition
of this
displacement.
It
follows
that,
in case
(6),
the solution of the
equations
is
indeterminate,
in the
sense that a small
displacement
which would be
possible
in a
rigid body may
be
superposed upon any displacement
that satisfies the
equations.
The
question
of the existence of solutions of the
equations
of
type (7)
which also
satisfy
the
given boundary
conditions will not be discussed here.
It is of more
importance
to remark
that,
when the surface tractions are
given,
the
equations
and conditions are
incompatible
unless these
tractions,
with the
body
forces,
are a
system
of forces which would
keep
a
rigid body
in
equilibrium. Suppose
in fact that
u, v,
w are a
system
of functions which
satisfy
the
equations
of
type (7).
If we
integrate
the left-hand member
of
(7)
through
the volume of the
body,
and transform the
volume-integrals
*
Dublin,
Tram. R. Irish Acad. vol. 21
(1839)
=
Collected Worku
of
James
MacCullagh,
Dublin,
1880, p.
145.
t Phil. Tram.
Roy.
Soc.
(Ser. A),
vol. 185
(1894).
116-118]
OF
EQUILIBRIUM
167
9
fdW\
of such terms as
g- [^

1
by
Green's
transformation,
we find the
equation
jj
X4S +
jjj
pXdxdydz
=
0.
(8)
If we
multiply
the
equation
of
type (7)
which contains Z
by y,
and that
which contains Y
by
e,
and
subtract,
we obtain the
equation
Jjj
L^
(dw \dej
^
dy \deyj
"^
dz
\de^
j]
^
\
dx
\dej
"^
dy \deyj
+
dz
\d^^ l\
+
p{yZ-zY)\ dxdydz
=
;
and,
on
transforming
this
by
Green's
transformation,
we find the
equation
fj{yZ,-zY,)dS+jjjp(yZ-zY)da;dydz^O
(9)
In this
way
all the conditions of statical
equilibrium may
be shown to hold.
118.
Uniqueness
of solution*.
We shall
prove
the
following
theorem :

If either the surface

displace-
ments or the surface tractions are
given
the solution of the
problem
of
equilibrium
is
unique,
in the sense that the state of stress
(and strain)
is
determinate without
ambiguity.
We observe in the first
place
that the function
W,
being
a
homogeneous
quadratic
function which is
always positive
for real values of its
arguments,
cannot vanish unless all its
argum,ents
vanish. These
arguments
are the
six
components
of strain
; and,
when
they
vanish,
the
displacement
is one
which would be
possible
in a
rigid body.
Thus,
if W
vanishes,
the
body
is
only
moved as a whole.
Now,
if
possible,
let
u, v',
w and
tt",
v"
,
w" be two
systems
of
displace-
ments which
satisfy
the
equations
of
type (7),
and also
satisfy
the
given
conditions at the surface 8 of the
body.
Then u'

u",
v

v",
w

w" is a
system
of
displacements
which satisfies the
equations
of the
type
,if|ZU|Y|^Ul:f?^W0 (10)
. dx
\dexxl
dy \dexyj
dz
\dez,
throughout
the
body,
and also satisfies conditions at the surface. Denote
this
displacement by {u,
v, w).
Then we can write down the
equation
///
]^x\dexx! dyXdexy)
dzxde^a
[dx\dexyj dy\deyyj dz\deyj
(d /dW\ d /dW\ d
/dW\]'
, , ,

Cf. Kirchhoff,
J.
f.
Math.
(Crelle),
Bd. 66
(1859).
168 GENERAL THEORY
[CH.
VII
and this is the same as
W
[{cos(^..)g
+cos(3,,
^)g
+
cos(.,
.)g^|
+ two similar
expressions
-///
^w_ aw; 8f;
aw air air
peme oeyy
06^
dey^
de^;
dexy
dS
dxdydz
=
0.
When the surface conditions are of
displacement
u, v,
w vanish at all
points
of
S;
and when
they
are of traction the tractions calculated from
u, v,
w
vanish at all
points
of S. In either
case,
the surface
integral
in the above
equation
vanishes. The volume
integral
is
j
I \2W
dxdydz;
and since Wis
necessarily positive,
this cannot vanish unless W vanishes. Hence
(, v, w)
is a
displacement possible
in a
rigid body.
When the surface conditions are
of
displacement
u, v,
w must
vanish,
for
they
vanish at all
points
of S.
119. Theorem of mlnimuin
energy.
The theorem of
uniqueness
of solution is associated with a theorem of
minimum
potential energy.
We consider the case where there are no
body
forces,
and the surface
displacements
are
given.
The
potential energy
of
deformation of the
body
is the volume
integral
of the
strain-energy-function
taken
through
the volume of the
body.
We
may
state the theorem in the
form :

The
displacement
which satisfies the differential
equations
of
equilibrium,
as well as the conditions at the
bounding
surface,
yields
a smaller value for
the
potential
energy
of
deformation than
any
other
displacement,
which
satisfies the same conditions at the
bounding
surface.
Let
(u,
V,
w)
be the
displacement
which satisfies the
equations
of
equilibrium
throughout
the
body
and the conditions at the
bounding
surface,
and let
any
other
displacement
which satisfies the conditions at the surface
be denoted
by {u
+
u',v
+
v',
w
+
w').
The
quantities
', v\
w' vanish at the
surface. We denote
collectively by
e the
strain-components
calculated from
u, V, w,
and
by
e' the
strain-components
calculated from
u', v, w';
we denote
by f(e)
the
strain-energy-function
calculated from the
displacements
u, v, w,
with a similar notation for the
strain-energy-function
calculated from the
other
displacements.
We write V for the
potential energy
of deformation
corresponding
with the
displacement (u,
v,
w),
and
F,
for the
potential energy
of deformation
corresponding
with the
displacement {u-\-u', v+v',
w
+
w').
Then we show that
V^
V must be
positive.
We have
V^-V^jjj{f{e
+
e')
-f{e)] dxdydz,
118,
119]
OF
EQUILIBRIUM
109
and this is the same as
^.'^i-'-'+zw
dxdydz.
because
/(e)
is a
homogeneous quadratic
function of the
arguments
denoted
collectively by
e. Herein
/ (e')
is
necessarily positive,
for it is the strain-
energy-function
calculated from the
displacement {u', v, w').
Also we
have,
in the
ordinary
notation,
,(^^,dv\dW^/du'
aw/\SF
/9j/
du\dW
\dy dzj'dey^
\dz dx )
dei-, \dx
dyldcxy'
We transform the volume
integral
of this
expression
into a surface
integral
and a volume
integral,
neither of which involves differential coefficients of
u', v',
w'. The surface
integral
vanishes because
u, v',
w' vanish at the
surface. The coefficient of u' in the volume
integral
is
ox
\dexxJ
oy \dexy/
oz
\dezx/
and this vanishes in virtue of the
equations
of
equilibrium.
In like manner
the coefficients of v' and w' vanish. It follows that
V,-V
=
jjjf(e')
dxdydz.
which is
necessarily positive,
and
therefore
F<
F,.
The couverse of this theorem has been
employed
to
prove
that there exists a sohition
of the
equations
of
equilibrium
which
yields given
values for the
displacements
at the
boundary*.
If we knew
independently
that
among
all the sets of functions
u, v,
w,
which
take the
given
values on the
boundary,
there must be one which
gives
a smaller value to
j j
I
Wdxdydz
than
any
other
gives,
we could infer the truth of this converse theorem.
The same
difficulty
occurs in the
proof
of the existence-theorem in the
Theory
of
Potential t. In that
theory
it has been
attempted
to turn the
difficulty by devising
an
explicit process
for
constructing
the
required
function
J.
In the case of two-dimensional
potential
functions the existence of a minimum for the
integral
concerned has been
proved
by Hilbert.
*
Lord KeWin
(Sir
W.
Thomson),
Phil. Trans.
Hoy.
Soc. vol. 153
{1863)
=Math, and
Phyt.
Papers,
vol.
3, p.
351.
t The
difficulty appears
to have been
pointed
out first
by
Weierstrass in his lectures on the
Calculus of Variations. See the Article
'
Variation of an
integral
'
in
F.ncy.
Brit.
Supplement,
[Ency.
Brit.,
10th
ed.,
vol. 33
(1902)].
X See, e.g.,
C.
Neumann, Unlersuchungen
iiber das
toffarithmische
und Newtoti'sche Potential,
Leipzig,
1877.

'
Ueber das Dirichlet'sehe
Princip,' (Festsclirift
zur Feicr des 150
jahrigen
Bestehens d.
Kimigl.
Ges. d. Wiss. zu
Gottinytn), Berlin,
1901.
170
BETTl'S THEOREM
[CH.
VII
120. Theorem
concerning
the
potential energy
of deformation*.
The
potential energy
of deformation of a
body,
which is in
equilibrium
under
given
load,
is
equal
to half the work done
by
the external
forces,
acting through
the
displacements
from the unstressed state to the state
of
equilibrium.
The work in
question
is
lljp
{uX
+vY+wZ) dxdydz
+
[[
{uX,
+
vY, +
wZ,)
dS.
The surface
integral
is the sum of three such terms as
and the work in
question
is therefore
equal
to
d dW d dW d dW\
1
, ,
J
4- :r- ;r h
5-
1-
^ ^
+... + ...
I dxdydz
dx
dexx
oy de^y
oz
de^xJ
J
////
dW dW dW dW dW
dW\^
- -
"
JJJ I'- air.
+
'''^
8i7.
*
'"
ail
"
'''^
3^.
""
'^
a^
"^
'^^
8^i
^'^^^'-
rhe first line of this
expression
vanishes in virtue of the
equations
of
equilibrium,
and the second line is
equal
to 2 1 1 1
Wdxdydz.
Hence the
121. The
reciprocal
theorem+.
Let
u, V,
w be
any
functions of
x,
y,
z,
t which are one-valued and free
from
discontinuity throughout
the
space occupied by
a
body
;
and let us
suppose
that
11, v,
w are not too
great
at
any point
to admit of their
being
displacements
within the
range
of "small
displacements" contemplated
in
the
theory
of
elasticity
founded on Hooke's Law. Then suitable forces
could maintain the
body
in the state of
displacement
determined
by
u, v,
w.
The
body
forces and surface tractions that would be
required
can be
determined
by calculating
the
strain-components
and
strain-energy-function
from the
displacement (m,
v, w)
and
substituting
in the
equations
of
the
types
1 /a]r\ a_
/dW\
d_ /dw^\
^
a^
P
"^
dx
[dexxl
^
dy [dexyj
^
dz
[de,x)
~^dt''
X
=
cos
{X,
v)
(g^J
+
COS
(y,
v)[^)+
cos
(., .)
^
.
*
In some books the
potential energy
of deformation ie called the
*'
resilience
"
of the
body.
t The theorem is due to E.
Betti,
II nuovo Cimento
(Ser. 2),
tt. 7 and 8
(1872).
It is a
special
case of a more
general
theorem
given by
Lord
Rayleigh,
London Math. Soc.
Proc,
vol. 4
(IWli)
=^
Scientific Papers,
vol.
1, p.
179. For a
general
discussion of
reciprocal
theorems
in
Dynamics
reference
may
be made to a
paper by
H.
Lamb,
London Math. Soc.
Proc,
vol. 19
(1889), p.
144.
120-122]
OF
RECIPROCITY
171
The
displacement
u, v,
w is one that could be
produced by
these
body
forces
and surface tractions.
Now let
(m, V,
w), (w',
v',
w')
be two sets of
displacements, (X, Y,
Z)
and
(Z', F', Z')
the
corresponding body
forces,
(Z Y
Z,)
and
{X' Y'
Z',)
the
corresponding
surface tractions. The
reciprocal
theorem is as follows :
The whole work done
by
the
forces of
the
first
set
(including
kinetic
reactions),
acting
over the
displacements
produced by
the second
set,
is
equal
to the whole work done
by
the
forces of
the second
set,
acting
over the
displace-
ments
produced by
the
first.
The
analytical
statement of the theorem is
expressed by
the
equation
+
lUx^u' +Y,v' +
Zw')
dS
+
11
(X\u+
Y\v
+
Z\w)dS.
(11)
In virtue of the
equations
of motion and the
equations
which connect
the surface tractions with
stress-components,
we
may express
the left-hand
member of
(11)
in terms of
stress-components
in the form of a sum of terms
containing
u', v',
w'
explicitly.
The terms in u'
are
jjj
\dw
\de^J
dy [de^yJ
dz
\deiJ]
"
V
"
+
\\
u
jcos
{X, V)
{^^]
+
cos
(y,
v)
(|^)
+ cos
(., .)
(g)|
dS.
It follows that the left-hand iliember of
(11) may
be
expressed
as a volume
integral
;
and it takes the form
///
'
,
dW
,
d_W
,
dW
,
dW
,
dW
,
dW
dxdydz
By
a
general property
of
quadratic
functions,
this
expression
is
symmetrical
in
the
components
of strain of the two
systems,
Cxxy-
and
e'xx,
It is there-
fore the same as the result of
transforming
the
right-hand
member of
(11).
122. Determination of
average
strains*.
We
may
use the
reciprocal
theorem to find the
average
values of the
strains
produced
in a
body by any system
of forces
by
which
equilibrium
can be maintained. For this
purpose
we have
only
to
suppose
that
u', v,
w
are
displacements corresponding
with a
homogeneous
strain. The stress-
*
The method is due to
Betti,
U>c. cit.
172 AVERAGE VALUES OF STRAIN-COMPONENTS
[CH.
VII
components
calculated from
u, vf,
w' are then constant
throughout
the
body.
Equation
(11)
can be
expressed
in the form
W
jiexxX'x
+
ByyY'y
+
B^Z'^
+
By^Y'
^
+
B.^Z'^
+
B^y
X'
y) (Ixdydz
=
{\{p (Z'
+
Yv'
+
Zw') dmdydz
+
\Uxai'
+
Y^v'
+
Z^)
dS. . .
.(12)
If
X'x
is the
only stress-component
of the uniform stress that is different
from zero the
corresponding strain-components
can be calculated from the
stress-strain
relations,
and the
displacements (u',
v', w)
can be found. Thus
the
quantity
1 1 1
exxdxdydz
can be
determined,
and this
quantity
is the
product
of the volume of the
body
and the
average
value of the strain-
component
Bxx
taken
through
the
body.
In the same
way
the
average
of
any
other strain can be determined. To find the
average
value of the
cubical dilatation we take the uniform
stress-system
to consist of uniform
tension the same in all directions round a
point.
123.
Average
strains in an
isotropic
solid
body.
In the case of an
isotropic
solid of volume V the
average
value of
e^x
is
^v\\\p
i^* -<^{Yy
+
Zz)] dxdydz
+
^jj
{X^-a( Y^ij
+
Z^z)]
dS
;
.
..(13)
the
average
value of
e^z
is
^jjjp{Yz
+
Zy)dxdydz+^^^lJpiY.z
+
Z.y)d8: (14)
the
average
value of A is
^
jjjp
(Xx
+
Yy
+
Zz) dxdydz
+
-^^jj{X,x
+
Y,y
+
Z,z)
dS. .
..(15)
The
following
results*
may
be obtained
easily
from these formulae :

(i)
A solid
cylinder
of
any
form of section
resting
on one end on a horizontal
plane
is
shorter than it would be in the unstressed state
by
a
length Wl/2E<o,
where W is its
weight,
I its
length,
a> the area of its cross-section. The volume of the
cylinder
is less
than it would be in the unstressed state
by WljQL
(ii)
When the same
cylinder
lies on its
side,
it is
longer
than it would be in the
unstressed state
by
a
Wh/Ea,
where h is the
height
of the centre of
gravity
above the
plane.
The volume of the
cylinder
is less than it would be in the unstressed state
by
Wh/3k.
(iii)
A
body
of
any
form
compressed
between two
parallel planes,
at a distance c
apart,
will have its volume diminished
by pcjSk,
where
p
is the resultant
pressure
on
either
plane.
If the
body
is a
cylinder
with
piano
ends at
right angles
to its
generators,
*
Numerous
examples
of the
application
of these
formulie,
and the
correspouding
formulse
for an
seolotropic body,
have been
given by
C.
Chree, Gambridyc
Phil. Hoc.
Trans.,
vol. 15
(1892), p.
313.
122-124]
THEORY OF VIBRATIONS
173
and these ends are in contact with the
compressing planes,
its
length
will be diminished
by pcjE<j>,
where o> is the area of the cross-section.
(iv)
A vessel of
any form,
of internal volume
Vi
and external volume
!',
when
subjected
to internal
pressure p^
and external
pressure />,
will be deformed so that the
volume
K F]
of the material of the vessel is diminished
by
the amount
(j^o ^'o "Pi 'i )/'''
124. The
general problem
of vibrations.
Uniqueness
of solution.
When a solid
body
is held in a state of
strain,
and the forces that
maintain the strain cease to
act,
internal relative motion is
generally
set
up.
Such motions can also be set
up by
the action of forces which
vary
with the
time. In the latter case
they may
be described as
"
forced motions." In
problems
of forced motions the conditions at the surface
may
be conditions
of
displacement
or conditions of traction. When there are no
forces,
and the
surface of the
body
is free from
traction,
the motions that can take
place
are
"
free vibrations."
They
are to be determined
by solving
the
equations
of
the
type
dx\dexx/
Oy\dexy)
dz\dea)
^dV-'
^ '
in a form
adapted
to
satisfy
the conditions of the
type
dW dW dW
cos(x,v)^
+
coa{ij,
v)^+cos{z, v)^
=
(17)
at the surface of the
body.
There is an infinite number of modes of free
vibration,
and we can
adapt
the solution of the
equations
to
satisfy given
conditions of
displacement
and
velocity
in the initial state.
When there are variable
body
forces,
and the surface is free from
traction,
free vibrations can coexist with forced
motions,
and the like holds
good
for
forced motions
produced by
variable surface tractions.
The methods of
integration
of the
equations
of free vibration will
occupy
us
immediately.
We shall
prove
here that a solution of the
equations
of free vibration which also satisfies
given
initial conditions of
displacement
and
velocity
is
unique*.
If
possible,
let there be two sets of
displacements
(', v,
w')
and
(u",
v",
w")
which both
satisfy
the
equations
of
type (16)
and the conditions of
type (17),
and,
at a certain
instant,
t
=
to,
let
(', v', w')
=
(u",
v",
w")
and
au' Iv'
dw'\
idu" dv" dw"-
I'du
dv ow\
_
\dt'
di'
~dt)~
dt
'
dt
'
dt
The
difference
(u'

u",
v'

v",
w'

w")
would be a
displacement
which would
also
satisfy
the
equations
of
type (16)
and the conditions of
type (17),
and,
at the instant t
=
t^,
this
displacement
and the
corresponding velocity
would
*
Cf. F.
Neumann, Vorlesungen
uber...Elatticitat, p.
126.
174
GENERAL THEORY OF
[CH.
VII
vanish. Let
(,
v,
w)
denote this
displacement.
We form the
equation
dv
{
?h)
dtVdt^
dx
\dexx
d_fdW
dx
\.9e~. 3
dy \deyy)
dz
KdCy^
dwi d-w d /dW\
d2\de:ix/.
d
(dW\
d /dW
d_
/dW\
_ ^
/'
dy\deyt)
dz\de.,
dxdyd2
=
0,..{18)
in which the
components
of
strain, exx---,
and the
strain-energy-function,
W,
are to be calculated from the
displacement (w,
v,
w).
The terms
containing p
can be
integrated
with
respect
to
t,
and the result is that these terms are
equal
to the kinetic
energy
at time t calculated from
dujdt,
...,
for the kinetic
energy
at time
t
vanishes. The terms
containing
W can be transformed into
a surface
integral
and a volume
integral.
The surface
integral
is the sum
of three terms of the
type
and this vanishes because the surface tractions calculated from
(u,
v,
w)
vanish. The volume
integral
is
dWdexx
dWdeyy
XX
dt
deyy
dt
de.
dWde.
de-.
_
_^dWdey, _^dWde^
^dWde^u'
dt
dCyi
dt
dezx
dt
dcxy
dt
dxdydz,
and this is the value of 1 1 1
Wdxdydz
at time
t,
for W vanishes at the instant
t
=
tt,
because the
displacement
vanishes
throughout
the
body
at that instant.
Our
equation
(18)
is therefore
and this
equation
cannot hold
unless,
at the time
t,
the
velocity (dn/dt,...)
and the
strain-energy-function
W vanish. There would then be no
velocity
and no
strain,
and
any displacement
(m,
v,
w)
that could exist would be
possible
in a
rigid body
and
independent
of the time. Since
(u,
v, w)
vanishes
throughout
the
body
at the instant t
=
tf
it vanishes
throughout
the
body
at
all
subsequent
instants.
125. Flux of
energy
in
vibratory
motion.
The kinetic
energy
T and
potential energy
V of the
portion
of the
body
within a closed
surface S are
expressed by
the formulee
T=
I j I
i p{u'
+
v^
+
w^)dxd2/dz,
V=
j I I
Wdxdydz,
124-126]
FREE VIBRATIONS
175
in which the dots denote differentiation with
respect
to
t,
and the
integration
extends
through
the volume within 8. We have at once
^,(7'+n=/f/{f(
+
i'-*+)
+
^5"
+
^S
+
;
dWdii dWdv ZWZw
dW/dw dh\
dW/du
dw\ dW fZv
'duW , , ,
+a7Aa7
^
a"J
^
U^rx)^
g^, fc
+
Ty)]
'^'='^y'^'- -(^o)
The
right-hand
member
may
be transformed into a volume
integral
and a surface
integral.
The terms of the volume
integral
which contain u are
///./
.. ddW 3 dW
ZdW\^ ^ ^
and the terms of the surface
integral
which contain ic are
f
f (dW
dW dW 1
i J
"
fc
"'
^'' "^
+
3i;
"'
^'"^^d^
"'
^''
"^J
'^^-
When there are no
body forces,
we deduce the
equation
{T+V)=
j j{uX
+vn+
wZy)dS (21)
This
equation may
be
expressed
in words in the form :

The rate of increase of the

energy
within (S is
equal
to the rate at which work is done
by
the tractions across S.
According
to the theorem
(vii)
of Article 53 the
expression

(uXy
+vY^ +
wZ,)
is the
normal
component
of a vector
quantity,
of which the
components parallel
to the axes are
-(,uX^
+
itXy
+
wZ^),
-
(a Xy-\-iYy
+
w
Y,),
-
{uZ^
+
vY,
+
wZ,).
This vector therefore
may
be used to calculate the flux of
energy.
126. Free vibrations of elastic solid bodies.
In the
theory
of the small oscillations of
dynamical systems
with a finite
number of
degrees
of
freedom,
it is shown that the most
general
small motion
of a
system,
which is
slightly
disturbed from a
position
of stable
equilibrium,
is
capable
of
analysis
into a number of small
periodic
motions,
each of which
could be executed
iudependently
of the others. The number of these
special
types
of motion is
equal
to the number of
degrees
of freedom of the
system.
Each of them is characterized
by
the
following properties
:

(i)
The motion of
every particle
of the
system
is
simple
harmonic.
(ii)
The
period
and
phase
of the
simple
harmonic motion are the
same for all the
particles.
(iii)
The
displacement
of
any particle
from its
equilibrium position,
estimated in
any
direction,
bears a definite ratio to the
displacement
of
any
chosen
particle
in
any specified
direction.
When the
system
is
moving
in one of these
special ways
it is said to be
oscillating
in a
"principal"
(or "normal")
mode. The motion
consequent
upon any
small disturbance can be
represented
as the result of
superposed
motions in the different normal modes.
176
tiENERAL THEORY OF
[CH.
VII
When we
attempt
to
generalize
this
theory,
so as to
apply
it to
systems
with infinite
freedom,
we
begin by seeking
for normal modes of vibration*.
Taking pjiir
for the
frequency
of such a mode of
motion,
we assume for the
displacement
the formulae
u
=
' cos
(pt
+
e),
v
=
v cos
{pt
+
),
^v
=
w cos
{pt
+
e),
. .
.(22)
in which
u', v,
w are functions
of
x,
y,
z,
but not of
t,
and
p
and e are
constants.
Now let W be what the
strain-energy-function,
W,
would become
if
u, v\
w' were the
displacement,
and let A'
'j, ,
... be what the
stress-components
would become in the same case. The
equations
of motion under no
body
forces take such forms as
'-+f'+'&
+
''^-''
=
^ <^^>
and the
boundary
conditions,
when the surface is free from
traction,
take
such forms as
cos
{x, v)
X'x
+
cos
(y, v) X'y
4-
cos
{z, v)Z'x
=
(24)
These
equations
and conditions suffice to determine
', v',
w as functions of
X,
y,
z with an
arbitrary
constant
multiplier,
and these functions also
involve/).
The
boundary
conditions lead to an
equation
for
p,
in
general
transcendental
and
having
an infinite number of roots. This
equation
is known as the
"
frequency-equation."
It thus
appears
that an elastic solid
body possesses
an infinite number
of normal modes of vibration.
Let
pi, Pt,..-
be the roots of the
frequency-equation,
and let the normal
mode of vibration with
period ^irjpr
be
expressed by
the
equations
U
=
ArUr
COS
{prt+r),
V
=
A
rVr
COS
{prt
+
6^),
W
=
A
rWr
COS
{prt
+
6,), ...(25)
in which
Ar
is an
arbitrary
constant
multiplier;
the functions
,, Vr, Wr
are
called
"
normal functions."
The result of
superposing
motions in the different normal modes would
be a motion
expressed by equations
of the
type
U
=
'%Ur4>r,
V
=
l.Vr(l)r,
rU
=
XWr(f)r,
(26)
in which
<j>r
stands for the function
Ar
cos
(pj;
+
e,).
The statement that
every
small motion of the
system
can be
represented
as the result of
super-
posed
motions in normal modes is
equivalent
to a
theorem,
viz.: that
any
arbitrary displacement (or velocity)
can be
represented
as the sum of a finite
or infinite series of normal functions. Such theorems
concerning
the
expansions
of functions are
generalizations
of Fourier's
theorem, and,
from
the
point
of view of a
rigorous analysis, they require independent proof.
Every problem
of free vibrations
suggests
such a theorem of
expansion.
*
See
Clebscb, Elatticitat,
or Lord
Bayleigh, Theory of Sound,
vol. 1.
126,
127]
FREE VIBRATIONS 177
127. General theorems
relating
to free vibrations*.
(i)
In the variational
equation
of motion
jjj
8
Wdxdydz
+jjj
p
(^^^
8u
+
^'~
Bv
+
'^
8w)
dxdydz
=
0. .
.(27)
let
u, V,
w have the forms
Ur<f)r, Vr<i>r> Wt4>r>
i^nd let
Bu, Bv,
Bw have the forms
<^, Vg^s, Wj^,,
where
0^
and
</>,
stand for
Ar
cos
(prt
+
e,)
and
.4,
cos
(p,t
+
e,),
and the constants
Ar
and
A,
may
be as small as we
please.
Let W become
Wr
when
Ur, Vr, Wr
are substituted for
u, v, w,
and become
W,
when
,, v,, w,
are substituted for
, v,
w. Let e denote
any
one of the six strain-com-
ponents,
and let
e,
and
e,
denote what e becomes when
m,, Vr, w,
and
u,, v,, w,
respectively
are substituted for
ii, v,
w. Then the variational
equation
takes
the form
\\\'S.i -X

e,
j
dxdydz
=
pr
Ijlp
("t
+
VrVg
+
WrW,) dxdydz.
The left-hand member is unaltered when
e^
and
e,
are
interchanged,
i.e. when
u, V,
w are taken to have the forms
UgC^,,
... and
Bu, Bv,
Bw are taken to have
the forms
Ur<f>r,---
and then the
right-hand
member contains
^/instead
of
/>r'.
Since
p^
and
pg
are
unequal
it follows that
I
p {UrUg
+
VfVg
+
WrWg)
dxdydz
=
(28)
This result is known as the
"
conjugate property
"
of the normal functions.
(ii)
We
may
write
<})r
in the forms
ArCos
prt
+
Br
sin
p^t,
and then the
conjugate property
of the normal functions enables us to determine the
constants
Ar, Br
in terms of the initial
displacement
and
velocity.
We
assume that the
displacement
at
any
time can be
represented
in the
form
(26).
Then
initially
we have
U
=
1Artlr, V
=
'lArVr, Wo
=
'2A
r^'r, (29)
U
=
^BrPrUr
, Vo
=
-BrPr^r
, W<,
=
^BrfrWr
(30)
where
(,
Wo,
v^o)
's the initial
displacement
and
(,
Vo, Wo)
is the initial
velocity.
On
multiplying
the three
equations
of
(29) by pUr, pVr, pw,
respectively,
and
integrating through
the volume of the
body,
we obtain the
equation
Arjjj
p {u/
+
V
+
Wr") dxdydz
=
I \\
p (!/,
-f
ot'r
+
WoWr) dxdydz.
. .
.(31 )
The other coefficients are determined
by
a similar
process.
(iii)
The
conjugate property
of the normal functions
may
be used
*
These theorems were
given by
ClebBch as a
generalization
ot Poisson's
theory
of the
vibrations of an elastic
sphere.
See Introduction.
L. E.
12
178
EFFECT OF SUDDEN APPLICATION
[CH.
VII
to show that the
frequency-equation
cannot have
imaginary
roots. If there
were a root
p,'
of the form a
+
1/8,
there would also be a root
p/
of the form
a

ij8.
With these there would
correspond
two sets of normal functions
iir, Vr, Wr
and
Mj, V,, w,
which also would be
conjugate imaginaries.
The
equation
I
p
{UrUg
+
VrVg
+
w^w,)
dxdydz

///'
could not then be
satisfied,
for the
subject
of
integration
would be the
product
of the
positive quantity p
and a sum of
positive squares.
It remains to show that
'pf
cannot be
negative.
For this
purpose
we
consider the
integral
///'
p (ur"
+
v/
+
w/) dxdydz,
J J J
which is
equal
to
---///{-(
dxdydz,
dx
dy
dz
where
Xa,""',
... are what
Xx,
... become when
Ur, Vr, Wr
are substituted for
u, V,
w. The
expression
last written can be transformed into
-Pr-'fj[ur
{cos
(x,
v)
Z^,"-' +
COS
{y, v) Xy<-'-^
+
cos
{z, v)Zx^'^\
+ ...+...
]dS
+Pr-'
III
2Wr
dxdydz,
in which the surface
integral
vanishes and the volume
integral
is
necessarily
positive.
It follows that
p/
is
positive.
128. Load
suddenly applied
or
suddenly
reversed.
The
theory
of the vibrations of solids
may
be used to
prove
two theorems
of
great importance
in
regard
to the
strength
of materials. The first of these
is that the strain
produced by
a load
suddenly applied may
be twice as
great
as that
produced by
the
gradual application
of the same load
;
the second is
that,
if the load is
suddenly
reversed,
the strain
may
be trebled.
To
prove
the first
theorem,
we observe
that,
if a load is
suddenly applied
to an elastic
system,
the
system
will be thrown into a state of vibration
about a certain
equilibrium configuration,
viz. that which the
system
would
take if the load were
applied gradually.
The initial state is one in which
the
energy
is
purely potential,
and,
as there is no elastic
stress,
this
energy
is due
simply
to the
position
of the elastic solid in the field of force
constituting
the load. If the initial
position
is a
possible position
of
instantaneous rest in a normal mode of oscillation of the
system,
then the
system
will oscillate in that normal
mode,
and the
configuration
at the end
of a
quarter
of a
period
will be the
equilibrium configuration,
i.e. the
127,
128]
OR REVERSAL OF LOAD
179
displacement
from the
equilibrium configuration
will then be
zero;
at the
end of a
half-period,
it will be
equal
and
opposite
to that in the initial
position.
The maximum
displacement
from the initial
configuration
will
therefore be twice that in the
equilibrium configuration.
If the
system,
when left to itself under the
suddenly applied
load,
does not oscillate in
a normal mode the strain will be less than twice that in the
equilibrium
configuration,
since the
system
never
passes
into a
configuration
in which
the
energy
is
purely potential.
The
proof
of the second theorem is similar. The
system being
held
strained in a
configuration
of
equilibrium,
the load is
suddenly
reversed,
and
the new
position
of
equilibrium
is one in which all the
displacements
are
reversed. This is the
position
about which the
system
oscillates. If it
oscillates in a normal mode the maximum
displacement
from the
equilibrium*
configuration
is double the initial
displacement
from the
configuration
of no
strain
; and,
at the instant when the
displacement
from the
equilibrium
configuration
is a
maximum,
the
displacement
from the
configuration
of
no strain is three times that which would occur in the
equilibrium
configuration.
A
typical example
of the first theorem is the case of an elastic
string,
to which a
weight
is
suddenly
attached. The
greatest
extension of the
string
is double that which it
has,
when
statically supporting
the
weight.
A
typical example
of the second theorem is the case of a
cylindrical
shaft held twisted. If the
twisting couple
is
suddenly
reversed the
greatest
shear can be three times that which
originally accompanied
the twist.
122
CHAPTER VIII.
THE TRANSMISSION OF FORCE.
129. In this
Chapter
we
propose
to
investigate
some
special problems
of the
equilibrium
of an
isotropic
solid
body
under no
body
forces. We
shall take the
equations
of
equilibrium
in the forms
^^
+
^KL
ly'
^>
+
mV^K-.-)
=
0.
0)
and shall consider certain
particular
solutions which tend to become infinite
in the
neighbourhood
of chosen
points.
These
points
must be outside the
body,
or in cavities within the
body.
We have a
theory
of the solution of
the
equations, by
a
synthesis
of solutions
having
certain
points
as
singular
points, analogous
to the
theory
of harmonic functions
regarded
as the
poten-
tials due to
point
masses. From the
physical point
of view the
simplest
singular point
is a
point
at which a force acts on the
body.
130. Force
operative
at a
point*.
When
body
forces
(X, Y,
Z)
act on the
body
the
equations
of
equi-
librium are
and the most
general
solution of these
equations
will be obtained
by adding
to
any particular
solution of them the
general
solution of
equations (1).
The
effects of the
body
forces are
represented by
the
particular
solution. We
seek such a solution in the case where
{X,
Y,
Z)
are different from zero
within a finite volume T and vanish outside T. The volume T
may
be that
of the
body
or that of a
part
of the
body.
For the
purpose
in hand we
may
think of the
body
as extended
indefinitely
in all directions and the volume T
as a
part
of it. We
pass
to a limit
by diminishing
T
indefinitely.
*
The results obtained in this Article are due to Lord Kelvin. See
Introduction,
footnote 66.
129,
130]
FORCE
OPERATIVE AT A POINT
181
We
express
the
displacement
by
means of a
scalar
potential <f>
and
vector
potential (F,
G,
H) (cf.
Article
16)
by
means of
formulae of the
type
d<b
,
dH
dO
dx
dy
dz
*
(3)
and we
express
the
body
force in like manner
by
means of
formulaj of
the
type
dx
dy
dz
.(4)
as
Since A
=
V^^,
. . .
,
the
equations (2)
can be written in such forms
and
particular
solutions can be obtained
by
writing
down
particular
solutions
of the four
equations
(i.V'G
+
pM=0,
ixV^H+pN=0. J
(^)
Now
X, Y,
Z can be
expressed
in forms of the
type (4) by
putting
'47r
_1_
47r
ox
dy
'
dz ')
dx'dy'dz',
.(7)
frrJJJX
Ox
where
X', Y',
Z' denote the values of
X, F,
Z at
any point {x, y', z)
within
T,
r is the distance of this
point
from
x,
y,
z,
and the
integration
extends
through
T. It is at once obvious that these forms
yield
the correct
values for
X, Y,
Z at
any point
within
T,
and zero values at
any point
outside T.
We now
pass
to a limit
by diminishing
all the linear dimensions of T
indefinitely,
but
supposing
that l i I X'd^'
dy'
dz' has a finite limit. We
pass
in this
way
to the case of a force
Xo
acting
at
(x', y, z')
in the direction
of the axis of x. We have to
put
p
jjjX'dx'dy'dz'^Xo
(8)
and then we have
^
=
--~x/^,
Z
=
(), J/=,
Zo^,
N
=
-j^ X,%-.
...(9)
^TTp
dx
i-n-p
dz
4nrp dy
182 EFFECT OF FORCE
[CH.
VIII
Now V''
(dr/dx)
=
2dr~^/dx,
and we
may
therefore
put
"P~87r(\
+
2/t)8a;'
'
87r/.a^' 87r/iay-
"^ ^
The
corresponding
forms for
u, v,
w are

=
-
^-^

7. . ^ V =r::;
+
v
= -
w
=
Sttm
(X
+
2/a)
9a;''
47r/ir
'
87r/A(X
+
2fi}dxdy'
(\+fi)Xo
d'r
.(11)
Stt/a (X
+
2/i)
3a; 8^;
'
More
generally,
the
displacement
due to force
(X^,
Yo, Z^)
acting
at the
point (a/, y
,
z),
is
expressed by
the
equation
,
, X
+
3/i
IX, Fo Z,\
^"'^'^>=
87r^(X
+
2^)
iy ^'
7j
X
+
/A
/a; -a;'
y-y
^-/N
Xp
(a;
-
a;')
+
Fp
(y
-
y')
+
Z,, (^r
-
g')
87ru(X
+
2/4)
V
r
'
r
'
r J
r'
(12)
When the forces
X, Y,
Z act
through
a volume T of finite
size,
particular
integrals
of the
equations (2)
can be
expressed
in such forms as
X+
u
7/ ^
r
V
8ir/M{\
+
2fi)
///{^
,
^
+Pi^-^')
^'^'-"'^^
y'(y^-y')+^'(^
--')
}
,.'dyw,
...(13)
where the
integration
extends
through
the volume T.
It
may
be observed that the dilatation and rotation
corresponding
with
the
displacement (11)
are
given by
the
equations
^
=
4^(X
+
2,4)
-8ar'
2-^
=
0'
2-.=
^"^-^
"g^
, 2-.
=
-^-^-9^.
...(14)
131. First
type
of
simple
solutions*.
When the force acts at the
origin parallel
to the axis of z we
may
write the
expressions
for the
displacement
in the forms
=^;t. -=^^.
-'=^\:p^
+
Y^rr)
^ ^
It
may
be verified
immediately
that these constitute a solution of
equations (1)
in all
space
except
at the
origin.
We
suppose
that the
origin
is in a
cavity
within a
body,
and calculate
the traction across the surface of the
cavity.
The tractions
corresponding
with
(15)
over
any
surfaces
bounding
a
body
are a
system
of forces in statical
equilibrium
when the
origin
is not a
point
of the
body [cf.
Article
117].
It follows
that,
in the case of the
body

The solution
expressed
in
equations (15)
has received this title at the hands of
Boussinesq,
Applicatioiu
det Potentiels
130-132] OPERATIVE AT A POINT
183
with the
cavity,
the resultant and resultant
moment of these tractions at the outer
boundary
of the
body
are
equal
and
opposite
to the resultant and resultant moment of
the tractions at the surfece of the
cavity.
The values of these tractions at the outer
boundary
do not
depend upon
the
shape
or size of the
cavity,
and
they may
therefore be
calculated
by taking
the
cavity
to be
spherical
and
passing
to a limit
by diminishing
the
radius of the
sphere
indefinitely.
In this
way
we
may verify
that the
displacement
expressed by (15)
is
produced by
a
single
force of
magnitude
8n^(\
+
2^)AI(\+,i) applied
at the
origin
in the direction of the axis of z.
We write
equations (15)
in the form
The cubical dilatation A
corresponding
with the
displacement (16)
is A
^^
^~
and
.,
,
,
X+it
6z
the
stress-components
can be calculated
readily
in the forms
The tractions across
any plane (of
which the normal is in direction
v)
are
given by
the
equations
X,
=
2^4|^3g^g-^
^+^
|C0S(.,
.)_-COS(^, .)
g-Jj,
n=2M[_3g-g-^^+^-^jcos(.,
v)-^-co8(y, v)-^]\,
and,
when v is the inwards drawn normal to a
spherical
surface with its centre at the
origin,
these are
Vte
%iLAyz 2^^
/ z2
^\
^^
,.4
.
J
--74-.
^'-i^\^^^xV^)
(17)
Whatever the radius of the
cavity may be,
this
system
of tractions is
statically equivalent
to a
single force, applied
at the
origin,
directed
along
the axis of z in the
positive sense,
and of
magnitude 8ir/i.4 {\-\-2fi)/{\+fi.).
Some additional results in
regard
to the state of stress set
up
in a
body by
the
applica-
tion of force at a
point
will be
given
in Article 140
infra.
132.
Typical
nuclei of strain.
Various solutions which
possess singular points
can be derived from that discussed in
Article 131. In
particular,
we
may suppose
two
points
at which forces act to
coalesce,
and obtain new solutions
by
a
limiting proce.ss.
It is convenient to denote the
displace-
ment due to force
(X^, Vg, Z^ applied
at the
origin by
184 DISPLACEMENT DUE TO
[CH.
VIII
so that for
example (m,, jj,, ?,)
is the
displacement
obtained
by replacing Xq by unity
in
equations (11).
We consider some
examples*
of the
synthesis
of
singularities
:

(a)
Let a force h'^P be
applied
at the
origin
in the direction of the axis of
x,
and let
an
equal
and
opposite
force be
applied
at the
point {h, 0, 0),
and let us
pass
to a limit
by
supposing
that h is diminished
indefinitely
while P remains constant. The
displacement
is
We
may
describe the
singularity
as a "double force without moment." It is related to
an
axis,
in this case the axis of
x,
and is
specified
as
regards magnitude by
the
quantity
P.
(a')
We
may
combine three double forces without
moment, having
their axes
parallel
to the axes of
coordinates,
and
.specified by
the same
quantity
P. The
resulting displace-
ment is
Now the result
(12)
shows that we have
3
=
W2, 'i=M3, U2
=
Vi, (19)
and thus
(18) may
be written
P(A,, Aj, A,),
where
Ai
is the dilatation when the
displace-
ment is
(m,, Vi, Wi),
and so on. Hence the
displacement (18)
is
4n{\+2^)\dx' dy'
dz
\
^ '
We
may
describe the
singularity
as a "centre of
compression";
when P is
negative
it
may
be called a
"
centre of dilatation." The
point
must be in a
cavity
within the
body ;
when the
cavity
is
spherical
and has its centre at the
point,
it
may
be verified that the
traction across the
cavity
is normal tension of amount
{^lP/{\+2^,)n}r-^
(6)
We
may suppose
a force A~'/* to act at the
origin
in the
positive
direction of the
axis of
X,
and an
equal
and
opposite
force to act at the
point (0, A, 0),
and we
may pass
to
a limit as before. The
resulting displacement
is
^V' ' /
We
may
describe the
singularity
as a "double force with moment." The forces
applied
to
the
body
in the
neighbourhood
of this
point
are
statically equivalent
to a
couple
of moment
P about the axis of z. The
singularity
is related to this axis and also to the direction of
the
forces,
in this case the axis of x.
(6')
We
may
combine two double forces with
moment,
the moments
being
about the
same axis and of the same
sign,
and the directions of the forces
being
at
right angles
to
each other. We take the forces to be h~^P and
-
/r'^P
parallel
to the axes of x and
y
at
the
origin,

h~^P
parallel
to the axis of x at the
point (0, A, 0),
and h'^P
parallel
to the
axis of
y
at the
point (A, 0, 0),
and we
pass
to a limit as before. The
resulting displace-
ment is
P jf^ _ ^\ (^ _ ^^ (^ _ ?^M
^\\dy dxj' {dy dxj' \dy dxjr
or it is
u^-
-^.
)
<-'
*
In most of these the
leading steps only
of the
analysis
are
given.
The results
(a')
and.
(6')
are due to J.
Dougall, Edinburgh
Math. Soc. Prnc. vol. 16
(1898).
132]
TYPICAL NUCLEI OF STRAIN
185
We
may
deacribe the
singularity
as a "centre of rotation about the axis of i." The forces
applied
to the
body
in the
neighbourhood
of this
point
are
statically equivalent
to a
couple
of moment iP about the axis of z
;
the
singularity
is not related to the directions
of the forces. In like manner we
may
have
singularities
which' are centres of rotation
about the axes of x and
y,
for which the
displacements
have the forms
P A 9r-i ar-i\
S^('
IT'
-ii)<
(22)
'^"'^
4;;;i(-ar'0. -^)
(23)
(c)
We
suppose
that centres of dilatation are distributed
uniformly along
a semi-
infinite line. The line
may
be taken to be the
portion
of the axis of z on which z is
negative.
The
displacement
is
given by equations
of the form
where S is a
constant,
and B?=x^
+y''
+
{z->rz'Y.
^^
jo
/^^
x^-k-v'^l
^\_
It
yr^-z-^y r)~r{r+zy
and the
displacement
is
given by
the
equations
u
=
B
-r^-,
v
=
B-r^-.,
w=-
(24)
r{z-\-r) r{z-\-r)'
r
^ '
These
displacements
constitute the
"simple
solutions of the second
type*."
The result
may
be
expressed
in the form
u,.,w)
=
bQ-^, |. J)log(e
+
r) (25)
A
singularity
of the
type
here described
might
be called a "line of
dilatation,"
and B
might
be called its
"strength."
If B is
negative,
the
singularity might
be called a "Une of
compression."
[d)
A line of dilatation
may
be terminated at both
ends,
and its
strength may
be
variable. If its extremities are the
origin
and the
point (0, 0,

Ic),
and its
strength
is
proportional
to the distance from the
origin,
we have
-^)z'dz:
where C" is constant. Now we have
Jo R'~JoVit^~ RV
r
B,
a^
+
f\Ri
r)
'
where
R{'
=
x^
+y'^
+
(z +1-)'^.
Tlie
integral
remains finite when k is increas
and we have
r"^_l _
/, A_J_
jo
R^ ~r
r2-zn
r)~z
+
r'
Again
we have
/*(2
+
')2' ,
k
, fdz'
i . z
+
k+R^
.(26)
*
Boussinesq,
loc. cit.
18G
DISPLACEMENT DUE TO
[CH.
VIII
This does not tend to a limit when k is increased
indefinitely.
Let C
{U, F, W)
denote
the
displacement (26)
; and,
in addition to the line of dilatation which
gives
rise to the
displacement ( V, F, IT),
let there be a line of
compression,
with the same law of
strength,
extending
from the
point (A, 0, 0)
to the
point (A, 0,
-
k).
We
pass
to a limit
by taking
h to diminish
indefinitely
and C to increase
indefinitely,
in such a
way
that C'h has a
finite
limit,
C
say.
The
displacement
is
given by
the
equations
^zu ^BK
ar
=^a^' ^'^^a^' ""^^a^-
3^
kx X
x_
-^""^
'Z^'
B^^ R^{z->rk^-R^)
r(z-^ry
and this has a finite limit when k is increased
indefinitely,
viz.
-.r/r(0
+
r).
The dis-
placement
due to such a semi-infinite double line of
singularities
as we have described here
is
expressed by
the
equations
=^(,-i^-r(^)2)'
"^-^T^r '"^'^rl^y
.(27)
or,
as
they may
be
written,
In like manner we
may
have
C' ^'
-)=-<a-^' I' a^.){^'s(^+'-)-^}
(^^)
(e)
Instead of a line-distribution of centres of
dilatation,
we
may
take a line-distribu-
tion of centres of rotation. From the result of
example (6')
we should find
where 2) is a
constant,
and the axes of the centres of rotation are
parallel
to the axis of x.
This
gives
u=0, =--,
w=D-^.
(30)
'
r
r{z
+
r)
In like manner we
may
have
=:?,
v
=
0,
,(,=
-7)-^
(31)
or,
as
they may
be written
(u, V, ')
=
i>(|,
0,
-^){log(.
+
r)} (32)
Other formula; of the same kind
might
be obtained
by taking
the line of
singularities
in directions other than the axis of z.
The reader will observe
that,
in all the
examples
of this
Article, except (a)
and
(6),
the
components
of
displacement
are harmonic
functions,
and the cubical dilatation vanishes.
The
only
strains involved are
shearing strains,
and the
displacements
are
independent
of
the ratio of elastic constants X :
/x.
133. Local Perturbations.
Examples (a)
and
(a')
of the last Article show in
particular
instances how
the
application
of
equilibrating
forces to a small
portion
of a
body
sets
up
strains which are
unimportant
at a distance from the
portion.
The
displacement
due to a distribution of force
having
a finite resultant for a
small volume varies
inversely
as the distance
;
that due to forces
having
zero
132-134]
TYPICAL NUCLEI OF STRAIN 187
resultant for the small volume varies
inversely
as the
square
of the
distance,
and
directly
as the linear dimension of the small volume. We
may
conclude
that the strain
produced
at a
distance,
by
forces
applied locally, depends
upon
the resultant of the
forces,
and is
practically independent
of the mode
of distribution of the forces which are
statically equivalent
to this resultant.
The effect of the mode of distribution of the forces is
practically
confined to
a
comparatively
small
portion
of the
body
near to the
place
of
application
of the forces. Such local effects are called
by Boussinesq
"
perturbations
locales *."
The statement that the mode of distribution of forces
applied locally
gives
rise to local
perturbations only,
includes Saint-Venant's
"
Principle
of the elastic
equivalence
of
statically equipollent systems
of
load,"
which is
used in
problems relating
to bars and
plates.
In these
cases,
the
falling
off
of the local
perturbations,
as the distance from the
place
of
application
of
the load
increases,
is much more
rapid
than in the case of a solid
body
of which all the dimensions are
large compared
with those of the
part
subjected
to the direct action of the forces. We
may
cite the
example
of a
very
thin
rectangular plate
under uniform torsional
couple along
its
edges.
The local
perturbations
diminish
according
to an
exponential
function of the distance from the
edgef.
134. Second
type
of
simple
solutions.
The
displacement
is
expressed by
the
equations given
in Article 132
(c),
viz. :

u=:B
,
^
,
,
v
=
B-^^^,
w
=
-, (24 6w)
r(z +
r)' r(z
+
r)'
r
'
or,
as
they may
be
written,
^^^^aiog(.J^) ^^s^Jog(zr)^
y^^B^^'^P^.
dx
oy
oz
It
may
be verified
immediately
that these
expressions
are solutions of the
equations (1)
at all
points except
the
origin
and
points
on the axis of z
at which z is
negative.
There is no
dilatation,
and the
stress-components
are
given
by
the
equations
*
Boussinesq,
loc. cit.
t
Kelvin and Tait,
Nat. Phil.,
Part ii.
pp.
267 et
seq.
188
EFFECT OF PRESSURK APPLIED
[CH.
VIII
At the surface of a
hemisphere,
for which r is constant and z is
positive,
these
give
rise to tractions
Z,
=
2^B
^,
F.
=
2^5^^L
Z,
=
^,
(33)
the normal
{v) being
drawn towards the centre.
135. Pressure at a
point
on a
plane boundary.
We consider an elastic solid
body
to which forces are
applied
in the
neighbourhood
of a
single point
on the surface. If all the linear dimensions
of the
body
are
large compared
with those of the area
subjected
to the
load,
we
may regard
the
body
as bounded
by
an infinite
plane.
We take the
origin
to be the
point
at which the load is
applied,
the
plane
^
=
to be the
bounding
surface of the
body,
and the
positive
direction
of the axis of z to be that which
goes
into the interior of the
body.
The
local effect of force
applied
at the
origin being very great,
we
suppose
the
origin
to be excluded
by
a
hemispherical
surface.
The
displacement expressed by (15)
could be maintained in the
body by
tractions over the
plane
boundary,
which are
expressed by
the
equations
X +
fji
r' X
+
fi
r
and
by
tractions over the
hemispherical boundary,
which are
expressed by
the
equations (17).
The resultant of the latter for the
hemispherical
surface
is a force in the
positive
direction of the axis of z of amount
4^irfiA{X
+
2ij.)l(\
+
/x).
The
displacement expressed by (24)
could be maintained in the
body by
tractions over the
plane boundary,
which are
expressed by
the
equations
Z,
=
-2m^,
F,
=
-2;u5j,
Z,
=
0,
(34)
and
by
tractions over the
hemispherical boundary,
which are
expressed by
the
equations (33).
The resultant
of the latter is a force in the
positive
direction of the axis of z of amount
^ttjiB.
If we
put
B
=

AfjiliX
+
/x),
the state of
displacement expressed by
the
sum of the
displacements (15)
and
(24)
will be maintained
by
forces
applied
to the
hemispherical
surface
only
; and,
if the resultant of these forces is
P,
the
displacement
is
given by
the
equations
P xz
U
=
V
47r/4
7-^ 47r
{\
+
ix) r{z
+
r)
'
l-TT/j.r' 'iir{X
+
fj,)r{z
+
r)'
^
^
P z' P(X +
2fi)
1
w
=

H
-^ ^
.
134-136]
TO PART OF A BOUNDING SURFACE
189
At all
points
not too near to the
origin,
these
equations express
the
displace-
ment due to a
pressure
of
magnitude
P
applied
at the
origin.
For the discussion of this
sohition,
it is convenient to
regard
the
plane
boundary
as
horizontal,
and the
body
as
supporting
a
weight
P at the
origin.
We observe that the
tractions across a horizontal
plane
are
so that the resultant traction
per
unit area exerted from the
ui)per
side across the
plane
at
any point
is a force directed
along
the radius vector drawn from the
origin
and of
magnitude f (P/jrr^)
cos^
tf,
where 6 is the
angle
which the radius vector drawn from the
origin
makes with the vortical drawn downwards. The tractions acro.ss horizontal
planes
are tlie siime at all
points
of
any sphere
which touches the
bounding plane
at the
origin,
and their
magnitude
is
%r\niy^
where D is the diameter of the
sphere.
These
expressions
for the tractions across horizontal
planes
are
independent
of the elastic constants.
The
displacement may
be resolved into a horizontal
component
and a vertical com-
ponent.
The former is
Psin^r
^
\
-\
cos 6

^r^^ jz :r, ;
X-K^ir |_
X +
/i (l+cos5)J'
it is directed towards or
away
from the line of action of the
weight according
as the
radius vector is without or within the cone which is
given by
the
equation
(X-(-/i)
cos 6
(1
+COS
ff)
=
ii.
When Poisson's ratio for the material is
\
the
angle
of the cone is about 68 32'. At
any
point
on the
bounding plane
the horizontal
displacement
is directed towards the axis and
is of amount
JP/7rf(\-(-/i).
The vertical
displacement
at
any point
is
-2^
P
/X +
S
47r/ir \\->r,
--|-cos^^
it is
always
directed downwards. Its
magnitude
at a
point
on the
bounding plane
is
^P(X-(-2/i)/jrr|i(X-(-/i).
The
initially plane boundsiry
is deformed into a curved siu-face.
The
parts
which are not too near the
origin
come to he on the surface formed
by
the
revolution of the
hyperbola
xz=\PQ<-^ ^li.)ln)i (X
+
^i)
about the axis of z.
136. Distributed
pressure.
Instead of
supposing
the
pressure
to be
applied
at one
point,
we
may
suppose
it to be distributed over an area on the
bounding plane.
Let
{x',
y, 0)
be
any point
of this
plane,
P' the
pressure per
unit of area at this
point,
r the distance of a
point {x, y, z)
within the
body
from the
point
(x'.
y',
0).
Let
-v/r
denote the direct
potential
of a distribution P' over the
area,
%
the
logarithmic potential
of the same
distribution,
so that
f=jjP'rdx'dy', x=//i"log(^
+
r)d^'rfy', (36)
where the
integrations
are taken over the area
subjected
to
pressure.
We
observe that
V'x
=
0,
V'^}r^2'^=2j|dx'dy'
=
2(f>,s&y, (37)
190 EFFECT OF LOCAL PRESSURE
[CH.
VIII
where
(f>
is the
ordinary
or inverse
potential
of the distribution P'. We
observe also that
^
=
zS.
oz
The
displacement
at
any point
of the
body produced by
the distributed
pressure
P' is
expressed by
the
equations
1
dx
1_
9^
47r
(X
+
/a)
dx
'iiTfi
docdz
'
1
dx
1
9'^^
4!7r
(\ +
fi)dy 4>7riu.dydz'
^
=
L_9X_J_8!+
X
+
2^
47r
(\
+
/x)
9^;
47r/i
9^*
47r/x (X
+
/i)
These
expressions
can be
simplified by introducing
a new function fl
determined
by
the
equation
n^-^-^-
(38)
and we have the
expressions*
for the
displacement
9fl dn dn
X+2fi , ,.--.
We observe that these
expressions
are finite and determinate for all
values of
{x, y, z), provided
z is
positive
;
and
that,
as the
point {x, y, z)
approaches any point (', y', 0), they
tend to definite finite limits.
They
represent
the
displacement
at all
points
of the
body,
bounded
by
the infinite
plane
z
=
0,
to which
pressure
is
applied
over
any areaf.
The normal com-
ponent,
w,
of the
displacement
at
any point
on the surface of the
body
is
(\
+
2/i) (/>/47r/i (X
+
fi).
137. Pressure between two bodies in contact.

Geometrical Pre-
liminaries.
Let two bodies be
pressed together
so that the resultant
pressure
between
them is P. The
parts
of the bodies near the
points
of contact will be com-
pressed,
so that there is contact over a small area of the surface of each.
This common area will be called the
compressed
area,
and the curve that
bounds it the curve
of compression.
We
propose
to determine the curve
of
compression
and the distribution of
pressure
over the
compressed areaj.
The
shapes,
in the unstressed
state,
of the two bodies near the
parts
that come into contact can be
determined,
with sufficient
approximation, by
equations
of the form
z,
=
A,a?
+
B,f
+
2H,xy,]
z^
=
A^x^
+
B,y^
+
2H^xy,\
*
These formula; are due to
Hertz, J.f.
Math.
(Crelle),
Bd. 92
(1881).
t
A number of
special
cases are worked out
by Boussinesq,
loc. cit.
X
The
theory
is due to
Hertz,
loc. cit.
136, 137]
PRESSURE BETWEEN TWO BODIES
191
the axes of
Zi
and
z^
being
directed
along
the normals drawn towards the
interiors of the bodies
respectively.
In the unstressed
state,
the bodies are
in contact at the
origin
of
(x, y), they
have a common
tangent plane
there,
and the distance
apart
of two
points
of
them,
estimated
along
the common
normal,
is
expressed
with sufficient
approximation by
the
quadratic
form
{Ai
+
A^
x^
+
{Bi
+
B) y^
+ 2
{Hi
+
H.^)
xy.
This
expression
must be
positive
in whatever
way
the axes of x and
y
are
chosen,
and we
may
choose these
axes so that
-ffi +
H.^
vanishes. Then
Ai-\- A^
and
jB,
+
B^
must be
positive.
We
may
therefore write
Ai
+
A,
=
A, Bi
+
B,^B, Hi
=
-H,
(41)
A and B
being positive.
If
-Ri, -R/
are the
principal
radii of curvature at the
point
of contact
for the
body (1),
and
R^, R^
those for the
body (2),
and if these have
positive signs
when the
corresponding
centres of curvature are inside the
bodies
respectively,
we have
2{A
+
B)=\IR,
+
IIR:
+
\IR,+ IIR^ (42)
The
angle (<)
between those normal sections of the two surfaces in which
the radii of curvature are
R^, R^
is
given by
the
equation
4(^-5)^=(^^--l.)
+
(l-A)%2(l-_i,)(i^-^,)cos2a,....(43)
The
angle
(a)')
between the
{x, z) plane,
chosen so that
H^
=
Hi,
and the
normal section in which the radius of curvature is
Ri
is
given by
the
equation
If we introduce an
angle
t
by
the
equation
cosT
=
j-^,
(45)
so that
2^cosec2iT
=
25sec=iT=l/i?,
+
l/E/+l//i2+l/i2,', (46)
the
shape
of the
"
relative
indicatrix,"
Ax"
+
By^
=
const,
depends
on the
angle
r
only.
When the bodies are
pressed together
there will be
displacement
of both.
We take the
displacement
of the
body (1)
to be
(mi,
v^,
w,)
relative to the
axes of
{x, y, z^,
and that of the
body
(2)
to be
{u^,
v^, w^)
relative to the
axes of
(, y, z^).
Since the
parts
within the
compressed
area are in contact
after the
compression,
we must
have,
at all
points
of this
area,
Zi
+ Wi
=
-
(z.,
+
w.,)
+
a,
192
PRESSURE BETWEEN TWO
[CH.
VIII
where a. is the value of
Wi
+
Wj
at the
origin*.
Hence within the
compressed
area we have
w^
+
'W^
=
a-Aai'-By\
(47)
and outside the
compressed
area we must have
w.
+
w^xx-Ax"-
By'' (48)
in order that the surfaces
may
be
separated
from each other.
138. Solution of the
problem
of the
pressure
between two bodies
in contact.
We denote
by
Xj,
/a,
the elastic constants of the
body (1),
and
by
X^, /x^
those of the
body (2).
The
pressure
P between the bodies is the resultant
of a distributed
pressure
(P' per
unit of
area)
over the
compressed
area.
We
may
form functions
^i, %i,
lij
for the
body (1)
in the same
way
as
0,
X'
^ "^ere formed in Article
136,
and we
may
form
corresponding
functions for the
body (2).
The values of
w^
and
w^
at the common surface
can then be
written,
Wj
=
^i^o,
W2
=
^2<^o
(49)
where
^i
=
{\,
+
2/:i,)/4'7rM, (Xi
+
Mi), ^2
=
(^2
+
^iJ.i)l^'Tixt (X^
+
jj..),
. .
.(50)
and
<f)o
is the value of
^,
or
(^2
at the
surface,
i.e. the value of the
convergent
integral
jiP'r-^dx'dy'
at a
point
on the surface. The value of
0
at
any
point
within the
compressed
area is determined in terms of the
quantity
o
and the coordinates of the
point by
the
equation
Ax^-Bif) (51)
This result
suggests
the next
step
in the solution of the
problem.
The
functions denoted
by <^i
and
^^
are the
potentials,
on the two sides of the
plane
z
=
0,
of a
superficial
distribution of
density
P' within the
compressed
area,
and the
potential
at a
point
of this area is a
quadratic
function of the
coordinates of the
point.
We recall the result that the
potential
of a homo-
geneous ellipsoid
at an internal
point
is a
quadratic
function of the coordinates

If the
points (Xj
,
j/j, Zj)
of the
body (1)
and
{x^, y.^, Zj)
of the
body (2)
come into
contact,
we
mast have
a^
+
j
=
X2
+
j, yi
+
f,
=
y2
+
f2, ^i
+
Wi= -(2
+
)2)
+
a;
and in
equation (47)
we
identify (x,, y,)
with
(xj, y^).
We
may
show
that,
without
making
this
identification,
we should have
Wj
+
Ml2=
a
-
^Xi''
-
Byi^
-
2
[^^X, (m,
-
1/2)
+
^2^1 ('i
-
"2)
+
-^2 {! ("i
-
"2) +^1 (i
-
2)}]-
In the result we shall find for
Wi+w^
an
expression
of the order
Aa'',
where a is the
greatest
diameter of the
compressed area,
and
Uj
,
u, . . . will be of the same order in a as
Wj
+
u>2 ;
thus the
terms
neglected
are of a
higher
order of small
quantities
than those retained. If the bodies are
of the same material we have
j
=
u^
^"^^
^1
=
"2
when
Xj
=
x^
and
)/i
=
j/2
>
B,ni thus the identification
of
(xj, j/i)
with
(xj, y,)
leads in this case to an exact result.
jTpabc
I
137,
138]
BODIES IN CONTACT 193
of the
point.
We therefore seek to
satisfy
the conditions of the
problem by
assuming
that the
compressed
area is the area witliin an
ellipse, regarded
as
an
ellipsoid very
much
flattened,
nnd that the
pressure
P'
may
be obtained
by
a
limiting process,
the whole mass of the
ellipsoid remaining
finite,
and
one of its
principal
axes
being
diminished
indefinitely.
In the case of an
ellipsoid
of
density p,
of which the
equation
referred to its
principal
axes is
x'ja-
+
y-/b''
+
z^/c-=l,
the mass would be
^irpabc;
the
part
of this mass that would be contained in
a
cylinder standing
on the element of area
dx'dy'
would be
2pdx'dy'c
V(l
-
a.'V'
-
y'V^').
and the
potential
at
any
external
point
would be
x^ _
2/
_
^= \
d^p'
(1^ +
^
b:
+
^|r c-+^lr) [{a''
+
'f){b'
+
f){c'+\ir)li'
where v is the
positive
root of the
equation
x^Ka^
+v)
+
fl(b- +v)
+
z'Kc"-
+
I/)
=
1.
At an internal
point
we should have the same form for the
potential
with
written for v. We have now to
pass
to a limit
by taking
c to diminish in-
definitely,
and
/3
to increase
indefinitely,
while a and b remain
finite,
in such
a
way
that
(i) fTT (pc)
ab
=
P,
(ii) 2(pc)s/{l-xya'-y''m
=
P',
(iii)
<t>,^-.ab{pc)j^
[i-^-^^^-^^^-^^j^^^^^^^^^^^^^,
the third of these conditions
being
satisfied at all
points
within the
compressed
area. Hence we have
/=.v/(-S-S
>-)
and
^^^(o^-Ax'--Bf)
J
^i_^l
yL^
^
(.53)
The
equation (52)
determines the law of distribution of the
pressure
P' over
the
compressed
area,
when the dimensions of this area are known. The
equation
(53)
must hold for all values of a; and
y
within this
area,
and it is
therefore
equivalent
to three
equations,
viz.
^
=
|P(^.
+
^..)/V^^,^^-^-^^^^-^^
(54)
B-^P {X
+
%)
j^
^^,
_^
ff{(a^-+ ^) ylr]i
13
/
194
PRESSURE BETWEEN TWO
[CH.
VIII
The second and third of these
equations
determine a and
b,
and the first
of them determines a when a and b are known. If we
express
the results in
terms of the
eccentricity (e)
of the
ellipse,
e will be determined
by
the
equation
d^
_ ,r
dx
B
i:
DMra +?)}*'
...(55)
rff
(i+r)Mr(i-e+r))*'
rf?
.(56)
(1
+
?)M?(
a will be
given by
the
equation
and a will be
given by
the
equation
sp r
"
a
=
^
(^,
+
%)
j^
j^^^
+
f) (1
_
e^
+
^)jJ

We observe that e
depends
on the ratio A : B
only.
Hertz has tabulated the
values of
bja, =(1

e-)-,
in terms of the
angle
t,
of which the cosine is
{B

A)I{B
+
A).
He found the
following
results:

.(57)
r

138,
139]
BODIES IN CONTACT
195
formula
(52), satisfy
all the conditions of the
problem.
When P' is known
the functions
<^,
^,
fi for each of the bodies can be
calculated,
and hence
we
may
determine the
displacement
and the distribution of stress in
each
body.
Hertz* lias drawn the lines of
principal
stress in the
{x, z) plane
for the case in which
X
=
2/i (Poisson's
ratio
=
J).
Near the centre of the
compressed
area the
principal planes
of stress are
nearly parallel
to the coordinate
planes,
and both tractions are
pres.sures.
As we
go
from the centre of the
compressed
area
along
the axis of
x,
the
component
traction that is
nearly parallel
to the surface falls to
zero, changes
to tension and increases
to a maximum near the
edge
of the
compressed
area
;
it then diminishes more
gradually
without
changing sign again.
The other
component
is
pressure,
which
continually
diminishes as we
go
into the interior of the
body along
a line of stress
starting
near the
oentre of the
compressed
area. The state of stress is illustrated in
Fig. 15,
in which is
the centre of the
compressed area,
A A' the trace of this area on the
plane
of
(.r, z);
lines
like those
ending
at P are lines of
pressure throughout,
lines like those
ending
at Tare
lines of tension
throughout,
the lines
ending
at R are lines of stress in which the traction
in the central
(dotted) portion
is
pressure,
and in the
remaining portions
is tension.
Hertz made a series of
experiments
with the view of
testing
the
theory.
The result
that the linear dimensions of the
compressed
area are
proportional
to the cube root of the
pressure
between the bodies was verified
very exactly
;
the
dependence
of the form of the
compressed
area
upon
the form of the relative indicatrix was also verified in cases in which
the latter could be determined with fair
accuracy.
139. Hertz's
theory
of
impact.
The results obtained in the last Article have been
applied
to the
problem
of the
impact
of two .solid
bodiesf.
The
ordinary theory
of
impact,
founded
by
Newton,
divides bodies into two
classes,
"perfectly
elastic" and
"imper-
fectly
elastic." In the case of the former class there is no loss of kinetic
energy
in
impact.
In the other case
energy
is
dissipated
in
impact, llany
actual bodies are not
very
far from
being perfectly
elastic in the Newtonian
sense. Hertz's
theory
of
impact
takes no account of the
dissipation
of
energy
;
the
compression
at the
place
of contact is
regarded
as
gradually produced
and
*
VerhantUangen
ties Vereins zur
Beforderung
dcs
Gewerbejleisses,
1882
=
Ges.
Werke,
Bd.
1,
p.
174.
t Hertz, J.f.
Math.
(Crelle),
Bd. 92
(1881).
132
196 hertz's theory
[ch.
viii
as
subsiding completely by
reversal of the
process by
which it is
produced.
The local
compression
is thus
regarded
as a statical effect. In order that
such a
theory may
hold it is
necessary
tliat the duration of the
impact
should
be a
large multiple
of the
gravest period
of free vibration of either
body
which involves
compression
at the
place
in
question.
A formula for the
duration of the
impact,
which satisfies this
requirement
when the bodies
impinge
on each other with moderate
velocities,
has been
given by
Hertz,
and the result has been verified
experimentally*.
At
any
instant
during
the
impact,
the
quantity
a is the relative
displace-
ment of the centres of mass of the two
bodies,
estimated from their relative
positions
at the instant when the
impact
commences,
and resolved in the
direction of the common normal. The
pressure
P between the bodies is the
rate of destruction of the momentum of either. We therefore have the
equation
iU-^)=-P,
(60)
dt\ mi
+ mj
where d stands for
da/dt,
and
mj, wtj
are the masses of the bodies. Now P is
a function of
t,
so that the
principal
semi-diameters a and b of the
compressed
area at
any
instant are also functions of
t,
determined in terms of P
by
the
second and third of
equations (54);
in fact a and b are each of them
pro-
portional
to pi.
Equation (57)
shows that a is
proportional
to
P^,
or that P
is
proportional
to a^
;
we write
P
=
ha^ (61)
where
Jo
[Ul-i-Oa-e'+O]^]
~h
(l-f-?)M?(l-e=
+
?))*
(62)
Equation
(60) may
now be written
a
=
-
A,l^^a^ (63)
where
ki
=
{m^
+
m^jmim2.
This
equation may
be
integrated
in the form
^{a^-V')
=
-lhha\
(64)
where v is the initial value of
d,
i.e. the
velocity
of
approach
of the bodies
before
impact.
The value of a at the instant of
greatest compression
is
(4)'^
; ;
(->
and,
if this
quantity
is denoted
by
a,,
the duration of the
impact
is
da
{iyk,^A{%
+
%y
*
Sclineebeli,
Arch, des sci.
phys.,
Geneva,
t. 15
(1885).
luvestigations
of the duration of
impact
in the case of
high
velocities were made
by Tait, Edinburgh lioy.
Soc.
'Trans.,
vols.
26,
27
(1890, 1892).
139,
140]
OF IMPACT
197
We
may express
a,
in terms of the
shapes
and masses of the bodies and
the velocities of
propagation
of waves of
compression
in them
;
let
F,
and
V,
be these
velocities*,
p^
and
p.^
the densities of the
bodies, o-j
and a. the values
of Poisson's ratio for the two materials
;
then
so that
where
^ =
__(l-^0'
^ ^
(1-0-.)'
""'
7rF,V,(l-'2cr,)'
-
7rr,V=(l-2<7,)'
amim^v"- St/A( (1
o-,)' (l-<r,)'
[4 (m,
+
m.?,
47r
( V'^p,
(1
-
-Za,)
^
V,^p,{l
-
2a,)]
dK

(l
+
OMr(l-e=+?)P'
J
d?
.(66)
.(67)
.(68)
{i:(i+f)(i-e+r))i.
It
appears
that the duration of the
impact
varies
inversely
as the fifth root
of the relative
velocity
of
approach
before
impact.
The order of
magnitude
of the
gravest period
of free vibration that would involve
compression
is
lM,Fi,
and thus the duration of
impact
bears to this
period
a ratio of which
the order of
magnitude
is
(F,/ii).
140.
Impact
of
spheres.
When the bodies are
spheres
of radii
r,,
r.,,
we have
J
=
B
=
HVn
+
l/'-2). ''=0,

=
6, \
^=T^^;,(^^+^^)^'
a
=
^{d,
+
d.^P;
.(69)
from which we find
h
=
:.
:?,r
\r,
+
rj 5,
+5,'

=
{a(r,,-2)/(r,
+
r2)}i,
\
.(70)
Hence the duration of the
impact
and the radius of the
(circular) compressed
area are
determined.
In the
particular
case of
equal spheres
of the same material the duration of the
impact
is
f2.Q432
^f^
(l^)n*_r_
....(71)
(2 9432...)
I
^
(i_2a)4
^{ yi'
^' >
where r is the radius of either
sphere,
cr is the Poisson's ratio of the
material,
and V is the
velocity
of
propagation
of waves of
compression.
The radii of the circular
patches
that
come into contact are each
equal
to
\V) \_\Q
1-2,
These results have been verified
experimentally
t.

F,2
is
(X,
+
2yii,)/ft
and
V^-
is
(\
+
2fi^)lp.:^
.
t
Schneebeli, Rep.
i1.
Phys.,
Bd. 22
(1886),
and
Hamburger, Tngeblatt
d. Nat. Vers, in
Wiesbaden, 1887.
-2<r
J
.(72)
198 DISPLACEMENT DUE TO NUCLEI OF STRAIN
[CH.
VIII
141. EflFeots of nuclei of strain referred to
polar
coordinates.
We
may
seek solutions of the
equations (1)
in terms of
polar coordinates,
the dis-
placement being
taken to be
inversely proportional
to the radius vector r. The
displacement
must
satisfy equations (49)
of Article 97. If wo take
v,.
and
u^
to be
proportional
to
cos
m<^,
and ?< . to be
proportional
to sin
?0,
we
may
show that*
cos
v4
f
^
^^^
_|_
^^^
^^
^^^^
^^B{n-
con
6)
cot- ~l
,
A
=
-
sin
Tvf)
2u/f^ 5

'
tan"
-
cos<^
(
X-f2/i
7) cot"
-'" ^--
+
Ctanf-|-/>cot'
cos
?j<^
2
'%
where
A, B, C,
D are
arbitrary
constants
;
and then we
may
show that
cos
n<p
(
'
? sin 6
\
X
+
3u . . d
(-^M+cos.f
\oos ntpj \
C tan" -+
D cot"
~
+
G'tan2+fi'cot'
'
ainndi ( X + 3u r^A
^j.
=

^
^ {"n r-
*
rsm ^
( 2/i cosm<^
cos 5
I Ctan";
-
7> cot"
^ 1
-
G-' tan"
-
+
fl" cot"
-
where G and H are
arbitrary
constants. In the
particular
cases where ?i
=
or 1 some of
the solutions
require independent investigation.
These cases include the first
type
of
simple
solutions for
any
direction of the
applied force,
the second
type
of
simple solutions,
and the solutions arrived at in Article
132, examples {d), (e).
We
give
the
exj^ressions
for
the
displacements
and
stress-components
in a series of cases.

J. H.
MicheU,
London Math. Soc.
Proc,
vol. 32
(1900), p.
23.
141]
EXPRESSED IN TERMS OF POLAR COORDINATES 199
(a)
The first
type
of
simple .solutions,
corresponding
with a force F
parallel
to the
axis of
z,
is
expressed by
the
equations
_
F
cosfl
_
X
+
3/i
F sinfl
^
the
stress-components
are
expressed by
the
equations
'^'^-
X
+
2^4^
r^' ^^-'t"f'-\
+
2^4n
^
'
The meridian
planes (<^
=
const.)
are
principal planes
of
stre.ss;
and the lines of
principal
stress,
which are in
any
meridian
plane,
make with the radius vector at
any point angles
^
determined
by
the
equation
tan
2f
=
-
{2/i/(3X
+
5/i)}
tan 6.
These lines have been traced
by Michell,
for the case where X
=
fi,
with the re.sult shown in
Fig. 16,
in which the central
point
is the
point
of
application
of the force.
O)
When the line of action of the force F' is
parallel
to the axis of
x,
the
displace-
ment is
expressed by
the
equations
_
F'
sin^cosc^ _
X-l-3^
/" cos 6 cos
_
X
+
3m
F'
sin(^
'''4'^
r
'
^~2"(X
+
2,x) 4fl-fi
r
'
"*~
"2
(XT2^ 4^
~V'
'
the
stress-components
are
expressed by
the
equations
^
3X-f-4u
i?" sinflcosd)
-TA Tl H-
F'
sinBcoacj}
"=
-
X
+
2^
Tn W~
'
^^
=
**
=
X
+
2^
ii ^
'
^
<-^ a F' sin rf>
^ f-
F' cos 6 cos d)
^ ' ^
X-f2/x4jr
r^
X-f2;i4n-
r*
(y)
The second
type
of
simple
solutions is
expressed by
the
equations
B B &m6
1-^'

rl-f-cos5'
''
'
the
sti-ess-components
are
expressed by
the
equations

B
-Tx
^
B coaO
ri ^
B 1
rr=-2u.~,,
66
=
2)1-^^-
., <>0
=
2fi-
-
.,
^
-->
^
B siad
^ ^
'
'^r'
H-cosfl
(8)
The .solution
(28)
obtained in Article 132
{d)
is
expressed by
the
equations
C
cos(^
_C
sin
''e=""r
l-fcosd'
"*""^
1-fcosfl'
the
stress-components
are
expressed by
the
equations
'
^ ^ -r-j o
C
(l-cosd)cos</)
I -fr '^
J-*
(1
-f-
cos
6)
sm fl
i-r
C
(l-ooa6)m\<f)
-7-

C
8in</)
q_
C
cos<^
^'^
=
^'*^(T+cos5)sinfl'
*''=-^''F"2 1-l-cosfl' '^-^T^ll^^fl-
200 FORCE APPLIED AT VERTEX OF CONE
[CH.
VIII
(f)
The solution
(31)
obtained in Article 132
{e)
is
expressed by
the
equations
B sin 6coad>
I)
^
D .
,
,
=
^,
Is ?
=

COS0,
M.
=
sind); '
r
l+cos^
'

,.
-f'

f
f'
the
stress-components
are
expressed by
the
equations
'^
r' l + cos5
^^
flrf)=-u

-^ ,
d>r=u-= 2--
.isinrf),
re= -iiA^-
~_ 2, cos0.
^
'^r2
l+cosd
' ^
'^7-2
\
l
+costf/
^'
'^J-'V l+cosfl/
142. Problems
relating
to the
equilibrium
of cones*.
(i)
We
may
combine the solutions
expressed
in
(a)
and
(y)
of the last Article so as
to obtain the distribution of stress in a
cone, subjected
to a force at its vertex directed
along
its
axis,
when the
parts
at a
great
distance from the vertex are held fixed. If 6=a
is the
equation
of the surface of the
cone,
the
stress-components
66, 6(^,
rd must vanish
when 6
=
a,
and we have therefore
X-)-2/i47r
H-COSa
The resultant force at the vertex of the cone
may
be found
by considering
the traction in
the direction of the axis of the cone across a
spherical
surface with its centre at the vertex
;
it would be found that the force is
27X^2")
^'^ ^-^
~
"'"*'
"^ "'''' ^^
~
''''*
"^ ^^
"^ **'
''^''
and,
when F is
positive,
it is directed towards the interior of the cone.
By putting
a
=
|7r
we obtain the solution for a
point
of
pressure
on a
plane boundary
(Article 135).
(ii)
We
may
combine the solutions
expressed
in
()3), (8), (e)
of the last Article so as
to obtain the distribution of stress in a
cone, subjected
to a force at its vertex directed at
right angles
to its axis. The conditions that the surface of the cone
may
be free from
traction are
^l-cosa n

n
'i.C -.- -Z)sma
=
0,
sm a
F'
2C
-i)(l-h2cosa)-
.
,, ,^
cosa(l-t-cosa)
=
0,
4n-(A-(-zp}
-
2(7 ^-^^-
-t-2i)sin
x-\--r-,^^-^.
sin a
(1
-1-cos
a)
=
0,
sma
4r(X-f-2;x)
/"(l-t-cosa)2
i^'(l-t-C08a)
8'^'"g
^
=
-
8.(X-H2;.)
'
^=
-
-4^(X
+
Vr

The resultant force at the vertex is in the
positive
direction of the axis of
i-,
when F' is
positive,
and is of
magnitude
i^'(2-f-cosa)X-(-2^i,,
.
By combining
the results of
problems (i)
and
(ii)
we
may
obtain the solution for force
acting
in a
given
direction at the vertex of a
cone;
and
by putting
a
=
|r
we
may
obtain
the solution for force
acting
in a
given
direction at a
point
of a
plane boundary.
*
Michell,
loc. cit.
CHAPTER IX.
TWO-DIMENSIONAL ELASTIC SYSTEMS.
143. Methods of the kind considered in the last
Chapter, depending upon
simple
solutions which tend to become infinite at a
point, may
be
employed
also in the case of two-dimensional elastic
systems.
We have
already
had
occasion
(Chapter V.)
to remark that there are various
ways
in which such
systems present
themselves
naturally
for
investigation. They
are further
useful for
purposes
of illustration. As in other
departments
of mathematical
physics
which have relations to the
theory
of
potential,
it
frequently happens
that the
analogues,
in two
dimensions,
of
problems
which cannot be solved in
three dimensions are
capable
of exact solution
;
so it will
appear
that in the
theory
of
Elasticity
a two-dimensional solution can often be found which
throws
light upon
some wider
problem
that cannot be solved
completely.
144.
Displacement corresponding
with
plane
strain.
In a state of
plane
strain
parallel
to the
plane {x, y),
the
displacement
w
vanishes,
and the
displacements
u,
v are functions of the coordinates
x,
y
only.
The
components
of rotation
vtx
and
lir,,
vanish,
and we shall write
or for
TO-^.
When there are no
body
forces,
the
stress-equations
of
equilibrium
show that the
stress-components
X^,
Yy, Xy
can be
expressed
in terms of a
stress-function
;\;,
which is a function of x and
y,
but not of
z,
by
the formulae
X
-^^
Y
-?X
X
--
^'^
^^^
""df
"~a='
""
dxdy
^^
The identical relation between
strain-components (Article 17)
dy'
dx^
~dxdy
^
takes the form
or
^
+ ^+2-^^
=
(3)
We shall denote the
operator 3-/9*'"
+
S'/f'^" by
V,=,
and then this
equation
is
Vi*;^
=
0. It shows that
^^x
is a
plane
harmonic function.
202 DISPLACEMENTS CORRESPONDING WITH
[CH.
IX
The
equations
of
equilibrium
in terms of dilatation and rotation are
From these we deduce that A and xs are
plane
harmonic
functions,
and
that
(\
+
2/i.)
A
+
t2/xtir
is a function of the
complex
variable x
+
ly.
The
plane
harmonic function
Vj^;;^
is
equal
to 2
(\
+
/x)
A. We introduce a new-
function
^+ it]
oi
x+
lyhy
means of the
equation
^+i,r)
=
||(\
+
2/x)
A (-
i2fj,-s7] d{x
+
iy) (5)
so that
.(6)
Then we have
Also we have
dy
dx
"
I
'^dx~dy' 2{X
+
ix)
'^
daf^dx'
"^f^dy'dx^
2{X
+
fM)
'^
df-dy-
"^
dy~ dxdy
Sxdy dy'
2
^
=
_ ^'X.
+
2 OT
=
- -^--
+

3a;
dxdy
dxdy
dx'
2/.w=-^
+
^, 2/.t;
=
-^^+7;
(7)
These
equations
enable us to
express
the
displacement
when the stress-
function
x
is known.
Again,
when A and w are
known,
we
may
find
expressions
for
u,
v. We
have the
equations
du dv dv du , .
Fx'-dy
=
^'
dx-dy^^"^
(^>
It follows that
These,
with
(6), give
"
dx
\20:+2n))
^
dy \2iJi)
dy\2{X
+
2fi))
dx{2fij^

in which
[Article
14
(d)]
v' +
m' is a function of a; +
ly.
We
may put
, df
.
df
8a!
dy
144,
145]
PLANE STRAIN AND PLANE STRESS
203
where
y
is a
plane
harmonic
function,
and then
m,
v can be
expressed
in
the forms
^
,
^ +
/^
J^^
\
(9)
2/*
'^
2/i (\
+
2/i)
^
ay
"^
8a;
'
_
V
X
+
M
^J? ,
3/
^~2(x+2/i)~2;.r(x+2;;Z)^9^"''a^"
;
We
may
show without
difficulty
that the
corresponding
form of
x
is
and we
may verify
that the forms
(7)
for
u,
v are identical with the forms
(9).
145.
Displacement corresponding
with
plane
stress.
In the case of
plane
stress,
when
every plane parallel
to the
plane
of
a.',
y
is free from
traction,
we have
X^
=
Fj
=
Zj
=
0. We wish to determine
the most
general
forms for the
remaining stress-components,
and for the
coiTesponding displacement,
when these conditions are satisfied and no
body
forces are in action. We recall the results of Article 92
(iv).
It was there
shown
that,
if
@=^Xx+
Yy
+
Z^,
the function is
harmonic,
and
that,
besides
satisfying
the three
equations
of the
type
aZa;
_^
3Z,,
_^
9^
_
^
(11)
dx
dy
dz
the
stress-components
also
satisfy
six
equations
of the
types
VZ,
+
,-^^-^
=
0, V=F,
+
-^
'
fi-=0
(12)
\
+
(7 dx- 1+0-
0^02
Since
X^, Fj, Z^
are
zero,
d^jdz
is a
constant, /3 say,
and we have
@
=
0
+
/3^ (13)
where
%
is a function of x and
y,
which must be a
plane
harmonic function
since is
harmonic,
or we have
Vr0o
=
O
(14)
The
stress-components
X^,
Fy, Xy
are derived from a stress-function
X'
which is a function of
x,
y,
z,
in accordance with the formulae
(1),
and we have
VrA:
=
^o
+
^^ (15)
The first of
equations (12) gives
us
S/
1
+
o- dx'
or,
in virtue of
(14)
and
(15),
W^^'j
1
+
cray^
"'
dy'
dy'
9'
/S'v
<^
^A A
204 DISPLACEMENTS CORRESPONDING WITH
[CH.
IX
In like manner the
remaining equations
of
(12)
are
It follows that --^ +
;; Bo
is a linear function of x and
y,
and this function
oz' 1
+ cr
may
be taken to be zero without
altering
the values of
X^,
Yy, Xy.
We
therefore find the
following
form for
^
:

X
=
Xo+X.^-il~^o^'.
(16^
where
'x^
and
;^i
are
independent
of z and
satisfy
the
equations
Vrx=o, V-^Xi
=
/3 (17)
We
may
introduce a
pair
of
conjugate
functions
|
and
t;
of a; and
y
which are such that
^ =
^
=
0, ^1^J_^^ (18)
dx
dy
"'
dy
dx'
and then the most
general
forms for
;;^
and
x^i
can be written
Xo
=
i^r+/ Xi
=
i/3(.T-^
+
2/=)
+
^. (19)
where
/
and F are
plane
harmonic functions. The
general
form for
% being
known,
formula; for the stress can be
found,
and the
displacement
can be
deduced.
The
displacement (u,
v, w)
must
satisfy
the
equations
g
=
-{X,-aYy\
g^
=
-^,(F,-crX,), '^'|=-|(Z,+
F,),
dw S**
_
rv
9"
,
9^
_
A
Sv 9m
_
2
(1
+
0-)

dy
dz
'
dz dx
'
dx
dy
E
"'
There is no
difficulty
in
obtaining
the formulae*
1
/^ o 1 ,3o\
1
+
0- a ,
,
=
-^i^^
+
^xz
+
^az-^j-
^,^-^--(x
+
^x.),
v=
-^(,
+
^y.
+
i<..-3^)-^^-g-(Xo+.X.).
;
(21)
w
=
-
;^
{i/3 {o:'
+
y'+ TZ')
+
o-^ol
+
-^-^
Xi-
Any
small
displacement possible
in a
rigid body may,
of
course,
be
super-
posed
on. this
displacement.

EquiraleDt
formula; were obtained
by Clebsoh, ElaUicitiit,
39.
145-147]
PLANE STRAIN AND PLANE STRESS
205
146. Generalized
plane
stress. .
We have shown in Article 94
that,
when the
stress-component Zg
vanishes
everywhere,
and the
stress-components
Z^
and
Fj
vanish at two
plane
boundaries z
=
h,
the
average
values of the
remaining
stress-
components
Xx,
Yy, X,,
are determined
by
the
equations
dX^ dXj,_ ^_Xj,
dYy_
dx
^
dy
~^'
dx
+
dy
"""'
^^^)
and that the
average
values of the
displacements
u,
v are connected with the
average
values of the
stress-components by
the
equations
fdu dv\
9
^''^'^iA)-'4A
<^^)
where V
=
2
V/(X
+
2/i) (24)
It follows that
u,
V are determined
by
the same
equations
as if the
problem
were one of
plane
strain,
provided
that \ is
replaced by
X'. The
quantities
Xx,
Y,j, Xy
are derived from a stress-function
exactly
in the same
way
as in
problems
of
plane
strain.
The
average
values of the
displacements
in
any problem
of
plane
stress
are
independent
of the
quantities y8
and F of Article
145,
and are the same
as if the
problem
were one of
generalized plane
stress. It
appears
from this
statement that the
investigation
of states of
plane
strain
may
be
applied
to
give
an account of the effects
produced by
some distributions of forces which
do not
produce
states of
plane
strain. The
problems
to which this method
is
applicable
are
problems
of the
equilibrium
of a thin
plate
which is
deformed in its own
plane by
forces
applied
in the
plane.
The actual values
of the stresses and
displacements produced
in the
plate
are not
determined,
unless the forces are so distributed that the state is one of
plane
stress,
but
the
average
values across the thickness of the
plate
are determined.
Any
such
problem
can be solved
by treating
it as a
problem
of
plane
strain, and,
in the
results,
substituting
X' for \.
147. Introduction of nuclei of strain.
We
may investigate
solutions of the
equations
of
plane
strain which
tend to become iniinite at
specified points.
Such
points
must not be in the
substance of the
body,
but
they may
be in cavities within the
body.
When
this is the
case,
it is
necessary
to attend to the conditions which ensure that
the
displacement,
rotation and strain are one-valued. When the
points
are
outside the
body,
or on its
boundary,
these conditions do not in
general
need
206
EFFECT OF FORCE OPERATIVE
[CH.
IX
to be
investigated.
The
displacement being
determined
by
certain functions
oi x+
ly,
the
singular points
are
singularities
of these functions. Without
making
an exhaustive
investigation
of the
possible singular points
and their
bearing upon
the
theory
of
Elasticity,
we shall consider the states of stress
that
correspond
with certain
simple types
of
singular points.
148.
Force
operative
at a
point.
The
simplest
singularity
is arrived at
by taking
(\
+
2/Lt)
A
+
t2/xCT
=
^
(.r +
2/)-\
(25)
so that the
origin
is a
simple pole. Equation (5)
becomes in this case
^+i7^
=
A
\og{x+iy)
=
A{\ogr
+
te), (26)
where
r,
6 are
polar
coordinates in the
plane
of
{x, y).
The
corresponding
formulae
for
u,
v are
2/.
lo" r
+
X +
fji
2fi (\
+
2/a)
\ +
fi
2(\
+
2/x) 2/j.{\+2fji)
A
^.,
+ u\
r-
A-i
+
V.
.(27)
To make v one-valued we must
put
A
V


; e,
2(\
+
2^)'"'- 2{\
+
2fi.)
The formulae
for
u,
v then become
2/x (X
+
2/^)
=
^
2/^ (X
+
2/.1)
v=
,
\ +
li
A%
.(28)
2ya (X
+
2/i)
r
The
stress-components
X^,
Yy, Xy
are
given by
the
equations
X
(
2\
+
:
3/<,
_
2
(X
-f
ya) y^\
'
X,=
A-
Yy-^l
X-I-2/X
X-l-2/i
--I-
X -1-
2/tt
r-J
2JX
-f
m)
y2
\
+
2(i
r-
Xy
=
A
y fjL
2
(X-^
fi)
X-
.(29)
U+
2/i,'*' X-l-2/i
The
origin
must be in a
cavity
within the
body;
and the statical resultant
of the tractions at the surface of the
cavity
is
independent
of the
shape
of
the
cavity.
The resultant
may
be found
by taking
the
cavity
to be bounded
(in
the
plane) by
a circle with its centre at the
origin.
The
component
in
the direction of the axis of x is
expressed by
the
integral
147-149]
AT A POINT OF A PLATE
207
which is
equal
to
-
2A7r. The
component
in the direction of the axis of
y
vanishes,
and the moment of the tractions about the centre of the
cavity
also
vanishes. It follows that the state of stress
expressed
by (29)
is that
produced by
a
single
force,
of
magnitude
2TrA,
acting
at the
origin
in the
positive
sense of the axis of x.
The effect of force at a
point
of a
plate may
be deduced
by
writing
X' in
place
of X and
replacing
u, Xx,
...
by
il, Xx,
149. Force
operative
at a
point
of a
boundary.
If the
origin
is at a
point
on a
boundary,
the term of
(27)
which
contains 6 can be one-valued
indepen-
dently
of
any adjustment
of
u',
v'. It
is
merely necessary
to fix the
meaning
of 0. In
Fig.
17,
OX is the initial
line,
drawn into the
plate,
and the
angle
XOT=a. Then 6
may
be taken to lie
in the interval
a
>
^
^
-
(tt
-
a).
We
may
seek the
stress-system
that
would
correspond
with
(27)
if ii' and v'
were
put equal
to zero. We should find
2(\
+
f,)
a?
Fig.
17.
^y~
X +
2^
7-*
' Z
=
2{X
+
/J-)
^
x-y
X-I-2M
.(30)
In
polar
coordinates the same
stress-system
is
expressed by
the
equations
^
=
0,
7^
=
(.31)
rr
2(X
+
/Li)
. cos 6
A
,
X-l-2/i
This distribution of stress is described
by
Michell* as a
"simple
radial
distribution." Such a distribution about a
point
cannot exist if the
point
is within the
body.
When the
origin
is a
point
on the
boundary,
the state of
stress
expressed by (31)
is that due to a
single
force at the
point.
We
calculate the resultant traction across a semicircle with its centre at the
origin.
The
.^-component
of the resultant is
/:
rr .cos
.rd6,
. . . X +
fji
or it IS

^ : ^-
TV.
X +
2/i
' The
2/-component
of the resultant is
rr . sin .
rd0,
Loxihrn Math. Soc.
Proc,
vol. 32
(1900), p.
Sa.
/:
208 FORCE AT A POINT OF A PLATE
[CH.
IX
or it is zero. Thus the resultant
applied
force acts
along
the initial line and
its amount is tt^
(\
+
ytt)/(X
+
2/[i) ;
the sense is that of the continuation
of the initial line outwards from the
body
when A is
positive.
This result
gives
us the solution of the
problem
of a
plate
with a
straight
boundary,
to which force is
applied
at one
point
in a
given
direction.
Taking
that direction as initial
line,
and F as the amount of the
force,
the stress-
system
is
expressed by
the
equations
rr^--F^^,
re
=
0,
66
=
0,
(32)
TT r
and these
quantities
are of course
averages
taken
through
the thickness
of the
plate.
150. Case of a
straight boundary.
In the
particular
case where the
boundary
is the axis of
x,
the axis
of?/
penetrates
into
the
plate,
and the force at the
origin
is
pressure
Fdirected
normally inwards,
the
average
stresses and
displacements
are
expressed by
the
equations
z^
u^ ^i/=--''':3' ''^i'=-z^^'
(33)
n r<
' "
77 r*
.(34)
_ F(\' +
2t,)
_JL^
[
"
27r^i(X'
+
^i)
^^''
277^1
>-2'
J
This solution* is the two-dimensional
analogue
of the solution of the
problem
of
Boussinesq (Article 135).
Since
u,
v do not tend to zero at infinite
distances,
there is
some
difficulty
in the
application
of the result to an infinite
plate
;
but it
may
be
regarded
as
giving correctly
the local effect of force
applied
at a
point
of the
boundary.
151. Additional results.
(i)
The stress-function
corresponding
with
(32)
of Article 149 is -ir~^Fr6md.
(ii)
The effect of
pressure
distributed
uniformly
over a finite
length
of a
straight
boundary
can be obtained
by integration.
If
p
is the
pressure per
unit of
length,
and the
axis of X is the
boundary,
the axis of
y being
drawn into the
body,
the stress-function
is found to be
^r~'jo{(r2^^2~''i^^i)}i
where
r^, 6^
and
r.^, 6^
are
polar
coordinates with
the axis of x for initial line and the extremities of the
part subject
to
pressure
for
origins.
It
may
be shown that the lines of stress arc confocal conies
having
these
points
as fooit.
(iii)
Force at an
angle.
The results obtained in Article 149
may
be
generalized by supposing
that the
boundary
is made
up
of two
straight edges meeting
at the
origin. Working,
as
before,
with the case
of
plane strain,
we have to
replace
the limit
-
tt + a of
integration
in the calculation of

Flamant, Paris,
C.
R.,
t.
114,
1892. For the verification
by
means of
polarized light
see
Mesnager
in
Rapports presentes
uii
congris
international de
physique,
t.
1,
Paris
1900, p.
348.
Cf. Carus
Wilaon,
Phil.
Mag. (Ser. 5),
vol. 32
(1891),
where an
equivalent
result obtained
by
Boussinesq
is recorded.
'
t Michell,
London Math. Soc.
Proc.,
vol. 34
(1902), p.
134.
149-152]
FORCE AT AN ANGLE
209
the force
by -y+a,
where
y
is the
angle
between the two
straight edges.
We find for the
x-component
of force at the
origin
the
expression
and,
for the
y-component
of force at the
origin,
we find the
expression
~
^
X+^
^^^^ y
sin
(2a
-
y)}.
The direction of maximum radial stress is
not,
in this
case,
that of the resultant force. The
former of these is the initial
line, making
angles
a and
y-a
with the
edges
;
the latter
makes with the same
edges angles <^
and
y

<j>,
where
tan d,
=
ysina-sinysin(a-y)
_
y
cos a
+
sin
y 008(0

7)
It follows that the
angle
a is
given by
the
equation
y
sin
(^
-
sin
y
sin
{y

<f>)
y
cos
(t>
-
sin
y
cos
(y-<p)'
When a
given
force ^ is
applied
in a
given
direction, <f)
will be
known,
and a can be
found from this
equation
;
and the constant
A can be determined in terms of the re-
tan a
=
force
iuitial liue
Fig.
18.
sultant force F. The conditions that the radial stress
may
be
pressure everywhere
are
"<o
'
7~<o
' ^^^'
'"^ ^^^ extreme case a
=
-,
we should have
tan
<^-
y
-
sm
y
COS
y
sin^
y
The solution is due to
Michell*,
who remarks that for values of
y
not
exceeding
-
,
the last
result is
nearly equivalent
to a "rule of the middle
third,"
that is to
say,
the extreme
value of
(/)
is
nearly equal
to
y.
If the line of action of the
appUed
force lies within the
middle third of the
angle,
the radial stress is
one-signed.
The stress is
given by (32)
so that the laws of transmission of stress from an
angle
are
(i)
that the stress is
purely radial, (ii)
that it is
inversely proportional
to the distance
from the
angle, (iii)
that it is
proportional
to the cosine of the
angle
made
by
the radius
vector with a certain line in the
plane
of the
angle.
152.
Typical
nuclei of strain in two dimensions.
(a)
The formulas
(28) express
the
displacements
in
plane strain, corresponding
with
a
single
force of
magnitude
24 tt
acting
at the
origin
in the
negative
direction of the axis
of X. We
may
obtain a new
type
of
singular point by supposing
that the
following
forces
are
applied
near the
origin
:

parallel
to the axis of
x,
-2An at the
origin
and 2.4jr at
{h, 0) ;
parallel
to the axis of
y,
-2Air at the
origin
and ZAtt at
(0, h) ;
loc. cit.
p.
207.
14
-rr=ee
=
^^ -
rd=0, (37)
210 TWO-DIMENSIONAL NUCLEI OF STRAIN
[CH.
IX
and we
may pass
to a limit
by supposing
that Ah remains
constantly equal
to B while h is
diminished without limit. The
resulting displacement
is
given by
the
equations
'"~^
2^. (X
+
2;.)
da;
^^8'"^
+
-
2^ (\ +
2^) \da:
r^
dy r^J
'
__
\
+
3/i
d .
\ n
X
+
M
/ d x^ d
m/\
""
2^(X-l-2^i)a^^^'"^"^ 2,i(X
+
2,t) V?y^
85^/'
('^''')
=
X^' D^^S'-.
(35)
This
displacement
is
expressible
in
polar
coordinates
by
the formulaj
^^
=
X^^
""^^^ ^^^^
it involves no dilatation or rotation. The stress is
expressed by
the formulse
2/x
B
"X
+
2/ir2
so that the
origin
is a
point
of
pressure.
If the
origin
is in a circular
cavity
there is
uniform
pressure
of amount
2fiBr~^j(\
+
2)i)
over the
cavity.
(6) Again
we
may
obtain a different
type
of
singular point by supposing
that the
following
forces are
applied
near the
origin
:

parallel
to the axis of
x,
24rr at the
origin,
-
2An at the
point (0, h),
parallel
to the axis of
i/,
2An at the
origin,
2An at the
point {h, 0) ;
and we
may pass
to a limit as in case
(a).
We thus obtain the
following displacement
:

_
X+3/X
8,, ,

X+^
fd
f
d
xy\

X +
3^
8
""
2,x{\
+
2^L)dx^
('''^)=f(-8y
)l''8'-
(3)
This
displacement
is
expressible
in
polar
coordinates
by
the formulae
M,
=
0, ue
=
Blfir; (39)
it involves no dilatation or rotation. The stress is
expressed by
the formulse
5^
=
^=0, 7e='-2Br-\ (40)
80 that the state of stress is that
produced by
a
couple
of
magnitude
inB
applied
at the
origin.
(c)
We
may
take
(X
+
2;j)
A +
i2/ii<7
=
Clog(:r
+
iy).
Since nr is not one-valued in a
region containing
the
origin,
we shall
suppose
the
origin
to be on the
boundary. Equation
(5)
becomes
^
+
ir)
=
C{x\ogr-y6-x)
+
iC{y\o%r-\-x6-y),
and the
displacement may
be taken to be
given by
the formulte
C
,
. ,
(2X
+
3;.)
C .
u=-{x\ogr-x)-^^^^^-^y6,
'
.
^
=
2-(X?2^("^-^)-27(X^2m)^^'''-
2^i(X-|-2^i)8y^
152,
153]
TRANSFORMATION OF PLANE STRAIN
211
The stress is then
given by
the formulae
We
may
take
n^d^O,
the axis of .v to be the
boundary,
and the axis of
y
to be drawn into
the
body.
Then the traction on the
boundary
is
tangential
traction on the
part
of the bound-
ary
for which x is
negative;
and the traction is of amount Cn-
(X
+
/x)/(X
+
2/i),
and it acts
towards the
origin
if C is
positive,
and
away
from the
origin
if C is
negative.
The most
important parts
of
v,
near the
origin,
are the tenn
containing logr
and
6,
and if x ia
negative
both these have the
opposite sign
to
C,
so that
they
are
positive
when C is
negative.
We learn from this
example
that
tangential
traction over a
portion
of a surface
tends to
depress
the material on the side towards which it acts*.
153. Transformation of
plane
strain.
We have seen that states of
plane
strain are determined in terms of
functions of a
complex
variable so
+
ly,
and that the
poles
and
logarithmic
infinities of these functions
correspond
with
points
of
application
of force to
the
body
which
undei-goes
the
plane
strain. If the two-dimensional
region
occupied by
the
body
is
conformally represented upon
a different two-
dimensional
region by
means of a functional relation between
complex
variables
x'+
ly'
and x
+
ly,
a new state of
plane
strain,
in a
body
of a
different
shape
from that
originally
treated,
will be found
by transforming
the function
(\
-1-
2/i)
A
-H
1
2/i'6T
into a function of
ar'+ty' by
means of the
same functional relation. Since
poles
and
logarithmic
infinities are conserved
in such conformal
transformations,
the
points
of
application
of isolated forces
in the two states will be
corresponding points.
We have found in Article 149
the state of
plane
strain,
in a
body
bounded
by
a
straight edge
and otherwise
unlimited,
which would be
produced by
isolated forces
acting
in
given
direc-
tions at
given points
of the
edge.
We
may
therefore determine a state of
plane
strain in a
cylindrical body
of
any
form of
section,
subjected
to isolated
forces at
given points
of its
boundary,
whenever we can effect a conformal
representation
of the cross-section of the
body upon
a
half-plane.
It will in
general
be
found, however,
that the isolated forces are not the
only
forces
acting
on the
body
;
in
fact,
a
boundary
free from traction is not in
general
transformed into a
boundary
free from traction. This defect of
correspondence
is the main
difficulty
in the
way
of advance in the
theory
of two-dimensional
elastic
systems.
We
may approach
the matter from a different
point
of
view,
by
con-
sidering
the stress-function as a solution of
V,*;^
=
0. If we
change
the
independent
variables from
x,
y
to x'
,
y',
where x' and
y'
are
conjugate
functions of x and
y,
the form of the
equation
is not
conserved,
and thus the
form of the stress-function in the
(', y) region
cannot be inferred from
its form in the
{x, y) region.

Cf. L. N. G.
FUon,
Phil. Tram.
Ray.
Soc.
(Ser. A),
vol. 198
(1902).
142
212 TRANSFORMATION OF INVERSION
[CH.
IX
154. Inversion*.
The transformatioa of io
version, x'+
i,y'
=
{p
-I-
ty)~',
constitutes ati
excep-
tion to the statement at the end of Article 153 It will be more convenient
in this case to avoid
complex
variables,
and to
change
the
independent
variables
by
means of the
equations
x
=
k^xjr^, y
=
Ic'y/r^,
in which k is the constant of
inversion,
and r' stands for x' +
y^.
We write
in like manner r'^ for x'^
+
y\ Expressed
in
polar
coordinates the
equation
Vj*;^
=
becomes
r
dr\
dr
1
^
lU'-'ih-M-o--
<)
and,
when the variables are
changed
from
r,
6 to
r', d,
this
equation may
be
shown to become
1 a
( ,
a
>(^'^
+
4,|,(r"x)
=
0.
...(42)
r dr
\
dr'
It follows
that,
when
%
is
expressed
in terms of
x,
y', r'^;^
satisfies the
equation
(aV^
+
^^+^aW^)<^''x)
=
^; (*-^)
and therefore
r"'x
i^ * stress-function in the
plane
of
{x, y').
The
stress-components
derived from
r'^'x^
are
given by
the
equations f
rV
=

(r"'v) -I-
-
(r'^'y)
ffd'
=

(r'^v) ^'
=
- -
(r'-'v)
dr' r d6
.(44)
where ff is the same as 6
;
and we find
f0'= r\e0
+
2(x-r'
r^'
=
-r\rd.
dr,
dr
.(45)
where
rr, 60,
rO are the
stress-components
derived from
x< expressed
in
terms of
r,
0. Thus the stress in the
(r', d') system
differs from that in
the
(r, 0) system by
the factor
r*,
by
the reversal of the
shearing
stress
r6,
and
by
the
superposition
of a normal traction 2
{p^

r(9x/a'")l>
^^^
.
same in all directions round a
point.
It follows that lines of stress are
*
Michell,
loc.
cit., p.
208.
t
See the theorem
(ii)
of Article 59.
154,
155]
APPLIED TO PLANE STRAIN
213
transformed into lines of
stress,
and a
boundary
free from stress is trans-
formed into a
boundary
under normal traction
only.
Further this normal
traction is
constant. To
prove
this,
we observe that the conditions of zero
traction across a
boundary
are
cos(..,.)p-cos(y,.)||
=
0,
and these are the same as
-cos(.,.)||+cos(y,.)g
=
0.
ds\dyj
'
ds\dxj
^'
where ds denotes an element of the
boundary.
Hence
dx/dx
and
d^l^y
are
constant
along
the
boundary,
and we have
dsV^ dr)
ds\X '^dx
^dyj~ds
ds dx ds
By
It follows that a
boundary
free from traction in the
(r, ff) system
is trans-
formed into a
boundary subject
to normal tension in the
(r', 6') system.
This tension has the same value at all
points
of the transformed
boundary,
and its effect is known and can be allowed for.
155.
Equilibrium
of a circular disk under forces in its
plane*.
(i)
We
may
now
apply
the transformation of inversion to the
problem
of Articles
149,
150.
ij'ig.
19.
*
The results of
(i)
and
(ii)
are due to
Hertz,
Zeit.
/.
Math. u.
Physik,
Bd. 28
(1888)
or Get.
Werke,
Bd.
1, p.
283,
and
Michell,
London Math. Soc.
Proc,
vol. 32
(1900), p.
36,
and vol.
84,
(1802), p.
134.
214
EQUILIBRIUM
OF CIRCULAR DISK
[CH.
IX
Let 0" be a
point
of a fixed
straight
line OfA
(Fig. 19).
If O'A were the
boundary
of
the section of a
body
in which there was
plane
strain
produced by
a force Fdirected
along
OffX,
the stress-function at P would bo
n'^FrBsmB,
where r stands for
O'P;
and this
may
be written
n'^Fdy,
where
y
is the ordinate of P referred to O'X. When we invert
the
system
with
respect
to
0, taking k=0(y,
P is transformed to
P',
and the new stress-
function is
-
ir-^r^F{6i
+
6^'kh/h'\,
where
6^
and
n-6^
are the
angles XOP\ XO'P',
and
we have written
rj
for
0P\
and
y'
for the ordinate of P' referred to OX. Further the line
(/A is transformed into a circle
through 0, 0',
and the
angle
2a which 00' subtends at the
centre is
equal
to twice the
angle
AO'X. Hence the function
-n'^F'y' {6^-^-6^
is the
stress-function
corresponding
with
equal
and
opposite
isolated
forces,
each of
magnitude
F', acting
as thrust in the line
00\ together
with a certain constant normal tension round
the
bounding
circle.
To find the
magnitude
of this
tension,
we observe
that,
when P' is on the
circle,
rj
cosec
5,
=
r,^
cosec
6-^==-k
cosec
(fi^-'r6^
=
'i.R,
where R is the radius of the circle.
Further,
the formulae
(1)
of Article 144
give
for the
stress-components
2/"
/cos^^i
X,=
/cos3
6, cos' 6X
^
2/" /cos 6, siu^ 6 cos
5,
sin'''
e\
_
^F'
/cos^^jsin^i
cos'
6^
sin
6^
Also the
angle {<\>
in the
figure)
which the central radius vector
(iJ)
to P' makes with
the axis of
x,
when P' is on the
circle,
is
^ir

a-)-25,,
or
Jtt-F^j-^j.
Hence the normal
tension across the circle is
Xj
sin^
{^2
-
^i)
--
Yy
cos2
(^2
-
^1)
+
2X
sin
(^2
-
6^
cos
(^2
-
^1).
and this is
-
{F'
sin
a)/7rR.
If the circle is
subjected
to the two forces F'
only
there is stress
compounded
of mean
tension, equal
at all
points
to
(F'
sin
a.)/nR,
and the
simple
radial distributions about the
points
and 0" in which the radial
components
are
-{2F'ooaei)lnri
and
-{2F'
00s
6^1nr^.
(ii)
Circular
plate subjected
to forces
acting
on its rim.
If the force F' is
applied
at in the direction 00'
(see Fig. 19)
and suitable tractions
are
applied
over the rest of the rim the stress-function
may
consist of the
single
term
_
n--i/"y'^,.
Let r and 6 be
polar
coordinates with
origin
at the centre of the circle and
initial
line
parallel
to 00'. The
angle (r, r^)
between the radii vectores drawn from the
centre and from to
any point
on the circumference is
^
n-
-
^2-
The
stress-system
referred
to
(rj, ^i)
is
given by
the
equations
r^j
=
-
(2/"
cos
6{)/{nri), d^i
=
0, r^,
=
;
and
therefore,
when referred to
(r, 6),
it is
given,
at
any point
of the
boundary, by
the
equations
^
2F' cos 6, sin'
6^ aa
^^' '^^^
^1
^^^
^2 ~a
^^' '^^
^1
^^
^2
^in
^2
rr=
,
6'S=
,
ro
=
,
or we have at the
boundary
^
/^' cos
5i
sin
^2
'^_i^'
cos
d, 008^2
"=~ir It
'
'*~ir R
'
and this is the same as

i^'sina F' .
,.
..
-^
F'cosa
,
F'
,a us
155]
UNDER FORCES IN ITS PLANE 215
where
a, =6^
+
$^,
is the acute
angle
subtended at a
point
on the circumference
by
the
chord OCy. Hence the traction across the
boundary
can be
regarded
as
compounded
of
(i)
uniform tension
-
^ {F'
sin
ajjirR
in the direction of the
normal,
(ii)
uniform
tangential
traction
^ (/"
cos
a)/nR,
(iii)
uniform traction
-
i
F'/rrR
in the direction Off.
Let
any
number of forces bo
applied
to various
points
of the
boundary.
If
they
would
keep
a
rigid body
in
equilibrium they satisfy
the condition 2/" cos a
=
0,
for sF'Rcoaa is
the sum of their moments about the centre. Also the uniform tractions
corresjxjnding
with
(iii)
in the above solution would have a zero resultant at
every point
of the rim.
Hence the result of
superixising
the
stress-systems
of
tyjie (32) belonging
to each of the
forces would be to
give
us the state of stress in the
plate
under the actual forces and a
normal tension of amount
-
2
(F'
sin
a)/2nR
at all
points
of the rim. The terms F' sin a
of this summation are
equal
to the normal
(inward) components
of the
applied
forces.
Mean
tension, equal
at all
points
to 2
{F'
sin
a)!2nR,
could be
superposed upon
this distri-
bution of
stress,
and then the
plate
would be
subject
to the action of the forces F'
only.
Fig.
20.
(iii) Heavy
disk*.
The state of stress in a
heavy
disk
resting
on a horizontal
plane
can also be found.
Let w be the
weight per
unit of
area,
and lot
r,
^ be
polar
coordinates with
origin
at
the
point
of contact A and initial line drawn
vertically upwards,
as in
Fig.
20.
The stress can be shown to be
compounded
of the
systems
(i) J^=i_w(y+R), 7y=-^w(j/-R), Xy=-\wa!,
(ii)
Vr=-'2.wR^r-^coae,
66
=
0,
f6
=
0.
The traction across
any
horizontal section is
pressure
directed
radially
from
A,
and is
of amount
^ior"' (i/Pcos^d-r^);
the traction across
any
section drawn
through
A is
horizontal tension of amount
^tv{2Rcos6-r).
The solution is due to
Michell,
loc. cit.
p.
207.
Figures showing
the distribution of stress
in this case and in several other
cases,
some of which have been discussed
in this
Chapter,
are
drawn
by
Michell.
216
TRANSFORMATION OF PLANE STRAIN
[CH.
IX
156.
Examples
of transformation.
(i)
The direct method of Article 153 will
lead, by
the substitution .v
+
iy=k^/{ii/+iy')
in the formula
(X
+
2;i)A
+
2j:a-
=
J(x
+
iy-^-)-i, (46)
to a
stress-system
in the
plane
of
(of, y),
in which
simple
radial stress at the
point (k, 0)
is
superposed upon
a constant
simple
tension
(Xj.)
in the direction of the axis a/. If the
boundary
in the
{x, y) plane
is
given by
the
equation y
=
{x-k)
tan
a,
the
boundary
in the
(y, y') plane
will be a
circle,
and the results
given
in
(i)
and
(ii)
of Article 155 can be
deduced.
(ii) By
the transformation
x+iy
=
ijif
+
i.j/y^
the
wedge-shaped region
between
y
=
and
y/x'
=
taur/w
is
conformally represented
on the half
plane y>0.
If we substitute
for x
+
iy
in
(46)
we shall obtain a state of stress in the
wedge-shaped region
bounded
by
the above two lines in the
plane
of
(x', y'),
which would be due to a
single
force
applied
at
(/(:"", 0),
and certain tractions distributed
_over
the boundaries. When m
=
2 the traction
over
y
=
vanishes and that on '
=
becomes tension of amount
proportional
to
(iii) By
the transformation 2
=
(e^-l)/(e^-l-l),
where z
=
x
+
iy
and z'=x'
+
iy',
the
strip
between
y'
=
and
y
=
7r is
conformally represented upon
the half
plane y>0,
so
that the
origins
in the two
planes
are
corresponding points,
and the
points (
+
1, 0)
in the
plane
of
{x, y) correspond
with the
infinitely
distant
points
of the
strip.
Let a
single
force F act at the
origin
in the
{x, y) plane
in the
positive
direction of the axis of
y.
Then the solution is
given by
the
equation
n \
+
fi
x
+
ty
Transforming
to
(x', y')
we find
,^
.

/^X-|-2u
siny-l-tsinha;'
(X
+ 2u
A-fi2uc7=
-^^

f-^
r,
^ r^' r-
^ \ +
fi
cosh.j;
cosy
and
^
+
tr,
=
-
YJ-J^ |f
2 tan"'
^
^'";^
^
-
y']
-
1
log (cosh
^'- cos
y')l
-fconst.
This solution
represents
the effect of a
single
force
2F, acting
at the
origin
in the
positive
direction of the axis of
y,
and
purely
normal
pressure
of amount
F/{1
-(-cosh
x') per
imit
of
length, acting
on the
edge y
=
jr of the
strip, together
with certain
tangential
tractions
on the
edges
of the
strip.
The latter can be annulled
by superposing
a
displacement
(u', v') upon
the
displacement
V2^
^
2^ (X -l-2/t)
-^
ay
'
2
(X
4-
2^t) 2^1 (X
-f
2,i)
-^
dy'J
'
provided
that
and this additional
displacement
does not affect the normal tractions on the
boundary.
"+'"^
2MX
+
2;x)
(''-'^^'
CHAPTER X.
THEORY OF THE INTEGRATION OF THE
EQUATIONS
OF
EQUILIBRIUM
OF AN ISOTROPIC ELASTIC SOLID BODY.
157. Nature of the
problem.
The chief
analytical problem
of the
theory
of
Elasticity
is that of the
solution of the
equations
of
equilibrium
of an
isotropic body
with a
given
boundary
when the surface
displacements
or the surface tractions are
given.
The case in which
body
forces act
upon
the
body may
be
reduced,
by
means
of the
particular integral
obtained in Article
130,
to that in which the
body
is held strained
by
surface tractions
only. Accordingly
our
problem
is to
determine functions
u, v,
w which within a
given boundary
are continuous
and have continuous differential
coefficients,
which
satisfy
the
system
of
partial
differential
equations
(X
+
ya)|^+/.V^M
=
0,
(\
+
^)|^-+/tiV^
=
0,
(X+M)^
+
/tV^w
=
0,
...(1)
, . da dv dw
,_.
"^^"^ ^
=
a.
+
a^+a?'
^'^
and which also
satisfy
certain conditions at the
boundary.
When the surface
displacements
are
given,
the values of
u, v,
w at the
boundary
are
prescribed.
We know that the solution of the
problem
is
unique
if
fi
and 3\ +
2^
are
positive.
When the surface tractions are
given
the values taken at the
surface
by
the three
expressions
of the
type
/ V (9w
du
,
. dv
,
. dw
/ s) /o\
XA
cos(a;, j/)
+
/i
j^
+5-
cos
(a:, 1')
+
^
cos(y,
v)
+
^
cos(^, v)> ...(3)
are
prescribed,
dv
denoting
an element of the normal to the
boundarj-.
We
know that the
problem
has no solution unless the
prescribed
surface tractions
satisfy
the conditions of
rigid-body-equilibrium
(Article 117).
We know
also
that,
if these conditions are
satisfied,
and if
/x
and 3X
+
2/i
are
positive,
the solution of the
problem
is
effectively unique,
in the sense that the strain
and stress are
uniquely
determinate,
but the
displacement may
have
super-
posed upon
it an
arbitrary
small
displacement
which would be
possible
in
a
rigid body.
218 RESUME OF THE THEORY
[CH.
X
158. Resuin6 of the
theory
of Potential.
The methods which have been devised for
solving
these
problems
have
a close
analogy
to the methods which have been devised for
solving
corre-
sponding problems
in the
theory
of Potential. In that
theory
we have the
problem
of
determining
a function U
which,
besides
satisfying
the usual
conditions of
continuity,
shall
satisfy
the
equation
V>U=0
(4)
at all
points
within a
given boundary*,
and either
(a)
shall take an
assigned
value at
every point
of this
boundary,
or
(6)
shall be such that
dU/dv
takes
an
assigned
value at
every point
of this
boundary.
In case
(6)
the surface-
integral
Ij -^dS
taken over the
boundary
must
vanish,
and in this case the
function U is determinate to an
arbitrary
constant
pres.
There are two main lines of attack
upon
these
problems,
which
may
be
described
respectively
as the method of series and the method of
singu-
larities. To illustrate the method of series we consider the case of a
spherical boundary.
There exists an infinite series of
functions,
each of
them rational and
integral
and
homogeneous
in
x,
y,
z and
satisfying
equation (4).
Let the
origin
be the centre of the
sphere,
let a be the
radius of the
sphere,
and let r denote the distance of
any point
from the
origin. Any
one of these functions can be
expressed
in the form
r"/S,
where
n is an
integer,
and
Sn,
which is
independent
of
r,
is a function of
position
on
the
sphere.
Then the functions
*S
have the
property
that an
arbitrary
function of
position
on the
sphere
can be
expressed by
an infinite series
OO
of the form 2
AnSn.
The
possibility
of the
expansion
is bound
up
with
=o
the
possession by
the functions
/S
of the
conjugate property expressed by
the
equation
'\8nSmdS
=
(.5)
//^
The function f/" which satisfies
equation (4)
within a
sphere
r
=
a,
and takes
on the
sphere
the values of an
arbitrary
function,
is
expressible
in the form
u=^aJ-s.
n=0
Q'
If the surface
integral
of the
arbitrary
function over the
sphere
vanishes
there is no term of
degree
zero
(constant term)
in the
expansion.
The
function U which satisfies
equation (4)
when r <
a,
and is such that 9
Ujdv
has
assigned
values on the
sphere
r
=
a,
is
expressed by
an
equation
of the form
*
A function which has these
properties
ia said to be
"
harmonic
"
in the
region
within the
given boundary.
158]
OF THE NEWTONIAN POTENTIAL
219
The
application
of the method of series to the
theory
of
Elasticity
will be
considered in the next
Chapter.
The method of
singularities depends
essentially upon
the
reciprocal
theorem,
known as Green's
equation,
viz. :
jjj(UV^V-W^U)dxdydz=jj(^U^^-V^-^yS
(6)
in which U and V are
any
two functions which
satisfy
the usual conditions
of
continuity
in a
region
of
space
;
the
volume-integration
is taken
through
this
region (or part
of
it),
and the
surface-integration
is taken over the
boundary
of the
region (or
the
part).
The normal v is drawn
away
from the
region (or
the
part).
The method
depends
also on the existence of a solution
of
(4) having
a
simple infinity (pole)
at an
assigned point ;
such a solution is
1/r,
where r denotes distance from the
point. By taking
for V the function
1/r,
and,
for the
region
of
space,
that bounded
externally by
a
given
surface S
and
internally by
a
sphere
S with its centre at the
origin
of
r,
and
by
passing
to a limit when the radius of
2 is
indefinitely
diminished,
we obtain
from
(6)
the
equation
^^^-Ikl^-^'C)''
m
so that U is
expressed explicitly
in terms of the surface values of U and
dUjdv.
The term that contains
dUjdv explicitly
is the
potential
of a
"simple
sheet,"
and that which contains U
explicitly
is the
potential
of a "double
sheet." In
general
the surface values of U and
dV/dv
cannot both be
pre-
scribed,
and the next
step
is to eliminate either U or
dUjdv

the one that is

not
given.
This is effected
by
the introduction of certain functions known
as
"
Green's functions." Let a function G be defined
by
the
following
con-
ditions:

(1)
the condition of
being
harmonic at all
points
within S
except
the
origin
of
r,
(2)
the
possession
of a
simple pole
at this
point
with residue
unity, (3)
the condition of
vanishing
at all
points
oiS. The function Q
may
be
called "Green's function for the surface and the
point."
The function G

l/r
is harmonic within S and
equal
to

1/?-
at all
points
on
S,
and we have the
equation
Since G vanishes at all
points
on /S we find that
(7) may
be written
47rf/=
-jju^dS.
(8)
Hence U can be
expressed
in terms of its surface values if G can be
found.
When the values of
dU/dv
are
given
at the
boundary
we introduce a
function F defined
by
the
following
conditions :

(1)
the condition of
being
220 METHOD OF INTEGRATING THE
EQUATIONS
[CH.
X
harmonic at all
points
within S
except
the
origin
of r and a chosen
point
A,
(2)
the
possession
of
simple poles
at these
points
with residues
+
1 and

1,
(3)
the condition that
dV/dv
vanishes at all
points
of S. We find for U the
equation
47r(fr-J7^)=jJr|^dS.
(9)
Hence U can be
expressed effectively
in terms of the surface values of
dU/dv
when T is known. The function F is sometimes called the "second
Green's function."
Green's function O for a surface and a
point may
be
interpreted
as the electric
potential
due to a
ix)int charge
in
presence
of an uninsulated
conducting
surface. The second
Green's function r for the
surface,
a
point
P and a chosen
point
A
may
be
interpreted
as
the
velocity potential
of
incompressible
fluid due to a source and sink at P and A within
a
rigid boundary.
The functions G and r are known for a few surfaces of which the
plane
and the
sphere*
are the most
important.
The existence of Green's functions for
any
surface,
and the existence of functions which
are harmonic within a surface and take
prescribed values,
or have
prescribed
normal rates
of
variation,
at all
points
on the surface are not obvious without
proof.
The efforts
that have been made to
prove
these existence-theorems have
given
rise to a mathematical
theory
of
great
interest. Methods have been devised for
constructing
the functions
by
convergent processes
t
;
and these
methods, although very complicated,
have been successful
for certain classes of surfaces
(e.g.
such as are
everywhere convex)
when some restrictions
are
imposed upon
the
degree
of arbitrariness of the
prescribed
surface values.
Similar existence-theorems are involved in the
theory
of
Elasticity,
but
comparatively
little
progress
has as
yet
been made with the
proof
of them.
159.
Description
of Betti's method of
integration.
The
adaptation
of the method of
singularities
to the
theory
of
Elasticity
was made
by
Betti
j,
who showed how to
express
the dilatation A and the
rotation
(-btx, tSy,
lu-j) by
means of formulse
analogous
to
(7)
and
containing
explicitly
the surface tractions and surface
displacements.
These formulae
involve
special systems
of
displacements
which have been
given
in
Chapter
viil. Since A is harmonic the
equations (1)
can be written in
such forms as
V2[M4-^(l-f X//i)a;A]
=
(10)
and thus the determination of
u, v,
w when A is known and the surface
values of
u, v,
w are
prescribed
is reduced to a
problem
in the
theory
of
Potential. If the surface tractions
{X,
Y
Z^)
are
prescribed,
we observe
that the
boundary
conditions can be written in such forms as
g-
=

X,
-
2"
A cos
{x, v)
+
TSy
cos
{z, v)
-
^z
cos
{y, v),
. .
.(11)

See
e.g. Maxwell, Electricity
and
Magnetism,
2nd
edition,
Oxford
1881,
and W. M.
Hicks,
Phil. Trans.
Roy.
Soc,
vol. 171
(1880).
t
See
e.g. Poincar^,
ThSorie du
potentiel Newtonien,
Paris 1899.
t
See
Introdtiction,
footnote 65.
158-160]
OF
EQUILIBRIUM
OF ISOTROPIC SOLIDS 221
SO
that,
when A and
Wx,
isry,
btj
are
found,
the surface values of
du/dv, dvldv,
dwjdv
are
known,
and the
problem
is
again
reduced to a
problem
in the
theory
of Potential.
Accordingly
Betti's method of
integration
involves the
determination of
A,
and of
sr^,
ay,
tsr^,
in terms of the
prescribed
surface
displacements
or surface
tractions,
by
the aid of
subsidiary special
solutions
which are
analogous
to Green's functions.
160. Formula for the dilatation.
,
The formula
analogous
to
(7)
is to be obtained
by
means of the
reciprocal
theorem
proved
in Article 121. When no
body
forces are in action the
theorem takes the form
Jl(X,u'+Y,v'+Zw')dS=IJ(X;u
+
Y,'v +
Z;w)dS, (12)
in which
(m,
v,
w)
is a
displacement satisfying equations (1)
and
X,, Y,, Z,
are the
corresponding
surface
tractions,
and also
(y!, v,
w')
is a second dis-
placement
and
XJ, Y,', ZJ
are the
corresponding
surface tractions.
Further,
the
integration
is taken over the
boundary
of
any region
within which
u, V,
w and
u', v',
w
satisfy
the usual conditions of
continuity
and the
equations
(1).
We take for u
, v',
w' the
expressions given
in
(20)
of
Article 132. It will be convenient to denote
these,
omitting
a
factor,
by
Mo. ^0. Wo.
and the
corresponding
surface tractions
by
X,"", F,<', Z.
We
write
(,.,;)
=^-^,
^,
-g^j
.
(13)
and then the
region
in
question
must be bounded
internally by
a closed
surface
surrounding
the
origin
of r. This surface will be taken to be a
sphere
2,
and we shall
pass
to a limit
by diminishing
the radius of this
sphere
indefinitely.
The external
boundary
of the
region
will be taken to
be the surface S of the
body.
Since the values of cos
{x, i/),
... at 2 are
-
xjr,
-
yjr,
-
zjr,
the contribu-
tion of
S
to the left-hand member of
(12)
is
which is
X
I
dy)
r\
'^dyj r'^\dy
dzl
\ dy
'^
-2
+
2/^
1^^
9^
+
';j
9^
+
^ 9^
ff^ [yz /9w dv\ ,
zx
(du ,
dw\
xy
(dv
du\
^lr-^\y(dy+dz)'-^\dz
+
d^)^^[d-x''d-y)
dt
222 METHOD OF INTEGRATING THE
EQUATIONS
[CH.
X
All the
integrals
of
type
I
lyzdX
vanish,
and each of those of
type
jjx^dl,
is
equal
to
^4Trr*,
and therefore the limit of the above
expression
when the
radius of 2 is diminished
indefinitely
is
47r(X.
+
1/^) (A),
where
{A\
denotes
the value of A at the
origin
of r.
Again,
since the values of
X^"", F"", ^""
are
expressed by
formulae of
the
type
cos
(x, ")
g^
+
cos
(y, ")
g-
+
cos
(z, v)
^ dx
the contribution of 2 to the
right-hand
side of
(12)
is
Now'such
integrals
as
jlxdl,
vanish,
and we therefore
expand
the functions
u, V,
w in the
neighbourhood
of the
origin
of r in such forms as
.
, fdn\
/du\
fdu\
and retain first
powers
of
x,
y,
z. Then in the
limit,
when the radius of
2
is
diminished
indefinitely,
the above contribution becomes
or

J^7r/t(A)o. Equation (12)
therefore
yields
the result
47r
(X
+
2/i) (A)
=
f f[(X^""
M
+
y.''
+
^.<'
w)
-
(X,M
+
Y,v,
+
Z^w,)]
dS.

(14)
The formula
(14)
is the
analogue
of
(7)
in
regard
to the dilatation.
This formula has been obtained here
by
a
strictly analytical process,
but
it
may
also be arrived at
synthetically* by
an
interpretation
of the
displace-
ment
(m, Vo,
Wo).
This
displacement
could be
produced
in a
body (held by
suitable forces at the
boundary) by
certain forces
applied
near the
origin
of r. Let
forces,
each of
magnitude
P,
be
applied
at the
origin
in the
positive
directions of the axes of
coordinates,
and let
equal
and
opposite
forces be
applied
in the
negative
directions of the axes of
x,
y,
z
respec-
tively
at the
points
(h, 0,
0), (0,
h,,
0), (0,
0,
h).
Let us
pass
to a limit
by increasing
P
indefinitely
and
diminishing
h
indefinitely
in such a
way
that lim P/i
=
47r
(X,
-f
2/i).
We know from Article 132 that the
displace-
ment
(mo,
Uo, Wo)
will be
produced,
and it is clear that the work done
by

J.
Dougall, Edinburgh
Math. Soc.
Proc,
vol. 16
(1898).
Betti's
reciprocal
theorem shows
that the work done
by
the tractions
Xy,
... on the surface
S, acting through
the
displacement
(Uj, r,,, iCo),
is
equal
to the work done
by
certain forces
applied at,
and near
to,
the
origin,
aeting through
the
displacement (it,
v,
w), together
with the work done
by
the tractions
Xi/("',
...
on the surface
S, acting through
the same
displacement.
160,
161]
OF
EQUILIBRIUM
OF ISOTROPIC SOLIDS 223
the above
system
of
forces,
applied
at,
and near
to,
the
origin, acting through
the
displacement (u,
v,
w)
is

47r
(\
+
Ifx) (A),.
161. Calculation of the dilatation from surface data.
(a)
When the surface
displacements
are
given
u, v,
w are
given
at all
points
of 8 but
X, Y^, Z
are not
given.
In this case we seek a
displace-
ment which shall
satisfy
the usual conditions of
continuity
and the
equations
(1)
at all
points
within
S,
and shall become
equal
to
(mo.
^o.
Wo)
at all
points
on 8. Let this
displacement
be denoted
by (w,,', v^,
Wo),
and let the corre-
sponding
surface tractions be denoted
by
.X"/'"'. I^/"", ^i-'*"'-
Then we
may
apply
the
reciprocal
theorem to the
displacements
(m, v,
w)
and
(<,', v,^, w,')
which have no
singularities
within
8,
and obtain the result
[j(Z;')
u
+
F/<'
V
+ ZJ^'^
w)
dS
=
[f(Z,Mo'+
YX
+
Z,w,')
dS
=
I
/(Z^Mo
+
Y^Vo
+
Z,w)
dS.
We
may
therefore write
equation (14)
in the form
47r
(X
+
2/x) (A)o
=
|T[(X,ii
-
Z;<>)
u
+
(F,'"*
-
Yy'^''>)v
+
{Z.^"^
-
.?;<');]
dS.
(15)
The
quantities
Z^""

X/"",
... are the surface tractions calculated from
displacements
u^

u^,
and
they
are therefore the tractions
required
to
hold the surface fixed when there is a
"
centre of
compression
"
at the
origin
of r. To find the dilatation at
any point
we must therefore calculate the
surface tractions
required
to hold the surface fixed when there is a centre of
compression
at the
point
;
and for this we must find a
displacement
which
(1)
satisfies the usual conditions of
continuity
and the
equations
of
equili-
brium
everywhere except
at the
point, (2)
in the
neighbourhood
of the
point
tends to become
infinite,
as if there were a centre of
compression
at the
point,
(3)
vanishes at the surface. The latter
displacement
is
analogous
to Green's
function.
(6)
When the surface tractions are
given,
we
begin by observing
that
Z^"", F"", ^''
are a
system
of surface tractions which
satisfy
the conditions
of
rigid-body-equilibrium.
Let
{u^\
vj',
tv")
be the
displacement produced
in the
body by
the
application
of these surface tractions. We
may apply
the
reciprocal
theorem to the
displacements (u,
v,
w)
and
(o", V'. w,,"),
which have no
singularities
within
8,
and obtain the result
[[(Z,('
u
+ F,<'
V
+ ^,"
w)
d8
=
jT(Z,"-h
F,t;"+ Z,w:')d8 ;
and then we
may
write
equation (14)
in the form
47r
(\
+
2/i) (A)
=
jj{X,
(m"- m)
+ F,
(vo"- V,)
+
Z,
{tv;
-
w,)]
dS. . .
.(16)
224 METHOD OF INTEGRATING THE
EQUATIONS
[CH.
X
To find the dilatation at
any point
we must therefore find the
displacement
produced
in the
body
when the surface is free from traction and there is
a centre of dilatation at the
point.
This
displacement
is
(uo"Uo,t>o"V(,,
Wo" Wo);
it is an
analogue
of Green's function.
The dilatation can be determined if the
displacement (mo",
Vo", w^")
can be
found. The
corresponding
surface tractions
being given,
this
displacement
is indeterminate in the sense that
any
small
displacement possible
in a
rigid
body may
be
superposed upon
it. It is
easily
seen from
equation (16)
that
this indeterminateness does not affect the value of the dilatation.
162. Formulae for the
components
of rotation.
In
applying
the formula
(12)
to a
region
bounded
externally by
the surface S of the
body,
and
internally by
the surface 2 of a small
sphere surrounding
the
origin
of
r,
we
take for
(u', v', w')
the
displacement given
in
(22)
of Article 132. It will be convenient to
denote this
displacement, omitting
a
factor, by (u^, Vf ii>^,*
and the
corresponding
surface
tractions
by
Z(*), r(*),
Z,^*).
We write
(3r
^
c'/'~'\
The contributions of 2 to the left-hand and
right-hand
members of
(12) may
be calculated
by
the
analytical process
of Article 160. We should find that the contribution to the left-
hand member
vanishes,
and that the contribution to the
right-hand
member is
Stt/i (n'i)oi
where
(tFj.)o
denotes the value of
otj.
at the
origin
of r. We should therefore have the
formula
87r,x(tirJ=
f
[{(A-.Mi-t- YvV^
+
Z,w^)-{X^Wu+ Y,(*)v-\-Z^mw)} dS, (18)
which is
analogous
to
(7).
The same result
may
be arrived at
by observing
that
(M4, v^, w^)
is the
displacement
due to forces
injxjh applied
at the
origin
in the
positive
and
negative
directions of the axes of
y
and z
respectively,
and to
equal
and
opposite
forces
applied
respectively
at the
points (0, 0, h)
and
(0, h, 0),
in the
limiting
condition when h is
diminished
indefinitely.
It is clear that the work done
by
these forces
acting
over the
displacement (^i, v, w)
is in the limit
equal
to
inii
l~
^ j
. Formulae of the same
type
as
(18)
for
nr
and
w,
can be written down.
163. Calculation of the rotation from surface data.
(a)
When the surface
displacements
are
given,
we introduce a
displacement
{u^, Vi, w^')
which satisfies the usual conditions of
continuity
and the
equations
of
equi-
librium
(1),
and takes at the surface the value
(M4, v^, w^)
: and we denote
by X,'(*), Y'W,
Z,'^*^
the
corresponding
surface tractions. Then
equation (18)
can be written
8ff^(w;,)=
j l{(Xy'm-Xym)u+(y,'m-
nm)v+(Zy'w-z,m)w}dS, (i9)
in which the
quantities
A^'W
-
A^f*),
... are the surface tractions
required
to hold the
surface fixed when a
couple
of moment
Snfi.
about the axis of x is
applied
at the
origin
in
such a
way
that this
point
becomes "a centre of rotation" about the axis of x. The
corresponding displacement {u^'

u^, v^-v^, w^-w^

is an
analogue
of Green's function.
*
This notation is
adopted
in accordance with the notation
(j, v-^, w-^,
... of Article 132 for
the
displacement
due to unit forces.
161-164]
OF
EQUILIBRIUM
OF ISOTROPIC SOLIDS
225
(i)
When the surface tractions are
given
we observe that the tractions
X,^*), Y,i*), ZJi*\
being statically eqviivalent
to a
couple,
do not
satisfy
the conditions of
rigid-body-equi-
librium,
and
that, therefore,
no
displacement
exists
which,
besides
satisfying
the usual
conditions of
continuity
and the
equations
of
equilibrium, gives
rise to surface tractions
equal
to
Ay*),
..,*. We must introduce a second centre of rotation at a chosen
point A,
so that the
couple
at A is
equal
and
opposite
to that at the
origin
of r. Let
m^-*', v^^*), w^-*)
be the
displacement
due to a centre of rotation about an axis at A
parallel
to the axis of
x,
so that
(^M,, ^M,,
,U,)=.(0,
^-^^,
^^),
(20)
where
i\i
denotes distance from A. Let
AV'(^), JV'*',
Zv"^*) denote the siuface tractions
calculated from the
displacement {u^-u^^^), v^-v^^^), w^-Wfi-^)).
The conditions of
rigid
body equilibrium
are satisfied
by
these tractions. Let
(u^", v^\ v^")
be the
displacement
which,
besides
satisfying
the usual conditions of
continuity
and the
equations
of
equi-
librium, gives
rise to the surface tractions
AV'(*>,
....
Then, denoting by (nr^)^
the value of
OTj:
at the
point A,
we find
by
the
process ahady
used to obtain
(18)
the
equation
SniJ. {(tzr,)
-
{v:,U}
=
jjl{A\
(w^
-
u,W)
+
...}- {A\"m u+...]]dS:
and from this
again
we obtain the
equation
+Zy
(Wi
-
1C^I^^)
-
<')}
dS.
(21 )
The
quantities u^

u^^'^'>

u^',
... are the
components
of
displacement produced
in the
body by equal
and
opposite
centres of rotation about the axis of x at the
origin
of r and
a
parallel
axis at the
point
A when the surface is free from traction. This
displacement
is an
analogue
of the second Green's function.
The rotation can be determined if such a
displacement
as
(m/', v^", w^')
can be found.
The indeterminateness of this
di.splacement,
which is to be found from surface conditions
of
traction,
does not affect the
rotation,
but the indeterminateness of
vjx
which arises from
tlie additive constant
{oi^a
is of the kind
already
noted in Article 157.
164.
Body
bounded
by plane

Formulae for the dilatation.

The
difficulty
of
proceeding
with the
integration
of the
equations
in
any
particular
case is the
difficulty
of
discovering
the functions which have been
denoted above
by
Mo'. Mo"> /',
These functions can be obtained when the
boundary
of the
body
is a
plane
f.
As
already
remarked
(Article 135)
the
local effects of forces
applied
to a small
part
of the surface of a
body
are
deducible from the solution of the
problem
of the
plane boundary.
Let the
bounding plane
be z
=
0,
and let the
body
be on that side of
it on which z>0. Let
{x, y', z')
be
any point
of the
body, {x, y', -/)
the
optical image
of this
point
in the
plane
z
=
0,
and let
r,
R denote the
distances of
any point {x, y, z)
from these two
points respectively.
For the
determination of the dilatation
when the surface
displacements
are
given
we
require
a
displacement
(', v^, Wo')
which,
besides
satisfying
the usual
J.
Dougall,
loc. cit.
p.
222.
t The
application
of Betti's method to the
problem
of the
plane
was made
by
Cerruti.
(See
Introduction,
footnote
08.)
L. E.
'
15
226 SOLUTION OF THE
[CH.
X
conditions of
continuity
and the
equations
of
equilibrium
(1)
in the
region
z>0,
shall at the
plane
z
=
have the value
{,
Vo,
w),
i.e.
(dr~^jdx, dr~^jdy,
dr-^/dz),
or,
what is the same
thing, (dR-'/dx, dR-^jdy, -dR-^jdz).
It can
be shown without
difficulty*
that the functions
m/, v', w^
are
given by
the
equations
dR-^
. ^X
+
fi
d"-R-' \
0
=
f
=
Wo=--
dx X
+
Sfi"
dxdz
'
dy
'^
\ +
Sfi
^
dydz
'
r
dR--"
,
X
+
M
d-R''
f-2; -^--
Z
.(22)
dz \
+
3/j.
dz-
The surface tractions
X"", F"", Z,,*"'
on the
plane
z
=
Q calculated from
the
displacement {u^,
v^, w^ are,
since
cos(^, i')
=
1,
given by
the
equations
X
I"' =
2a
3-'
dzdx
d'r-
2^
dzdx
'
d'R-'
dzdy dzdy
^('"
=
-2,
;/i
ay-'
dz'
-ifi
dz^-
'
.(23)
and the surface tractions
JT/*",
... on the
plane
z
=
calculated from the
displacement (u^, Vq, w^)
are
given by
the
equations
xy
=
-^i^
+
dug dw
'
dz dx
-lit.
"-"(V^a-^'x^
X
+
3jU.
dxdz
'
X +
/J,
d'R-'
3fjL dydz
'
'"-^-(:-s^w*^>'^^}"-''m/-^->
.(24)
*
If in fact we assume for
(,', v^', w^'
such forms as the
following
:
we find for
u', v',
w' the
equations
z
8F+'"''
<=-e^-+"''''
<=--^+^''.
f,^ ,
,3 /au'
,
dv' dw'\
, ,1 ,^ ,
dw'

Sd'
,
o'72-i
( 'Oz\Oj; cti/
dz
J
'^^
J
^
'^'V*'^ 02/ "'/
''^ ^'^
which are all satisfied
by
"
X +3/J
ex9z
'
" ~
X +
3/* 81/82
'
""
X +
3/i
8^2
'
for these functions are harmonic ard are such that =

1-^

[---
=
0.
ox
Oy
cz
164,
165]
PROBLEM OF THE PLANE
227
We observe that
Z/'"', T/'"', Z/""
are
equal respectively
to the
products
of
Z"", F^'"', ^^c'
and the numerical factor
-
(\
+
/t)/(\
+
3/t),
and hence that
(Mo", Vo",
Wo")
=
-
{(X
+
3m)/(\
+
m)} <, <).
It follows
that,
when the surface
displacements
are
given,
the value of A
at the
point {x', y', z)
is
given by
the
equation
A
=
-
TT
(K
+
3fi)
} ] \dxdz
"
'
dydz
the
integration extending
over the
plane
of
{x, y).
When the surface tractions
are
given
the value of A at the
point {x\ y\ z')
is
A
=
-
b7^)//(^''^'+^'w-*-^-->^^
(^>
27r(X
+
/i).
165.
Body
bounded
by plane

Given surface
displacements.
The formula
(25)
for the dilatation at
{x, y', z')
can be written
9
A
=
-
/"
\M>^y*hll>'''*M'>''V-''-'^
TT
(X
+
3/i)
dz
If we introduce four functions
L, M, N,
<f) by
the definitions
.(28)
dL dM dN
'^~dx''^
dy''^
dz"
these functions of
x',
y',
z' are harmonic on either side of the
plane
/=0,
and at this
plane
the values of
u, v,
w are
lim/=+o

^i-^^,, lim^ _+o

ir~ o~^ >
Zir oz ZTT oz
The
value of A at
{x', y', z)
is
^^ , ^ ^ ^,
,
and the
'^""^'=+-27r8/ TT
(\
+
3/i)
dz'
equations
of
equilibrium
can be written
V'2
u

\
+
/* ,90
w
27r
(\
+
3/i) 9<
X +
/i
,
9<^
27r
(\
+
3/x)
^
92/.
X +
/i
,
9(^
=
0,
=
0,
.(29)
27r
(X
+
;3/i)
9^'
where V'^
=
9V9a;''
4-
d-jdy''
+
9V9^''.
The three functions such as
M
-
{(X
+
/x)/27r (X
+
3/i,))
z
(d(f>ldx')
are harmonic in the
region
z'>0, and,
at the
plane
z'=
0,
they
take the values
152
228
SOLUTION OF THE
[CH.
X

^TT-' (dL/dz),
...
,
which are themselves
harmonic in the same
region.
It
follows that the values of
u, v,
w at
(', y', z')
are
given by
the
equations*
__l^dL
1 X
+
fJ. ,
dj>
"~
2Trdz' 2TrX +
'SfjL^
dx'
1
dM^
1
X+M ,3* ,
.cm
''-~2^W^27rX +
3fi^dy"
^
^"^
'
--JL^JL 1-
^
+
At '
^A
^~
2-77 dz''^2'7rX +
Sfj,^
dz'
'
The
simplest example
of these formula) is afforded
by
the case in which ti and v vanish
at all
points
of the
surface,
and w vanishes at all
points except
those in a
very
small area
near the
origin.
In this case the
only points (x, y, z)
that are included in the
integration
are close to the
origin,
and
<^
is the
potential
of a mass at the
origin.
We
may suppress
the accents on
xf, y,
z' and obtain the solution
^^. =^^.
'''=^
A
+
3^
1 z^\
which was considered in Article 131. In the
problem
of the
plane
this solution
gives
the
displacement
due to
pressure
of amount

47r/i
,
-^ exerted at the
origin
when the
plane
2=0 is held fixed at all
points
that are not
quite
close to the
origin.
166.
Body
bounded
by plane

Given surface tractions

-f.
It is
unnecessary
to
go through
the work of
calculating
the rotations
by
the
general
method.
The formula
(26)
for A can be
expressed
in the form
27r
{\
+
fi)
dz'
To eflfect this we introduce a function
y^
such that
dxl^z'
=
1/r
at z
=
0.
The
required
function is
expressed by
the formula
X
=
\og{z
+
z^R)- (31)
it is harmonic in the
space
considered and has the
property expressed by
the
equations
Now at the surface z
=
we have
dx dx dx'
~
dz'dx"
dy
~
dy'dz"
dz
~
d7-'
If therefore we write
F=jjx,X<i^dy,
G
=
jJY,xd^dy,
H
=
jjz^xd^dy,
^
dx'

dy'
"^
dz
'
*
The results are due to
Bonssinesq.
See
Introduction,
footnote 67.
t
The results are due to Cerruti. See
Introduction,
footnote 68.
...(33)
165,
166]
PROBLEM OF THE PLANE
229
the value of A at
{x, y', z)
is
given by
the
equation
^"27r(\
+
/.)97
<^*>
We observe also that the functions
F, G, H,
yjr
are harmonic and that the
values of
X^, F, Z^
at z'
=
0,
are
equal
to
,.
1 d"-F
,.
1 9^G
,.
1 d'H
hm,_+
-
2^ aZ^
.
i^"+o
-
2^ aZ'
'
^'"^'-^
-
2^ a?J
Now the third of the
equations
of
equilibrium
is
V'2 w
+
47r^
^'
9/
J
~
^'
and the third of the
boundary
conditions is
or
Hence at z'
=
dw 1 d^H
d^
dz'
47r/i
9s'-
47r/* (X
+
/it)
9/
'
11
,
J_
,9^|r]
9 f 1 9g 1

9s'
r 4'r/*
^
9s'
J
~
9/
t47rAt
9s'
"^
47r
(\
+
fi)
"^
,
It follows that w is
given by
the
equation
^
\ dH 1 1
,,9Vr
47r/i
9s' 47r
(A,
+
/it)
^
47r/t
9s'
"
Again
the first of the
equations
of
equilibrium
is
V'2
1
,d<r
"
+
1
^
5^
47r/ii
da;
_
and the first of the
boundary
conditions is
=
0.
Hence at s'
=
(du
-'^19?
du
.
9w
doc
J
.(35)
9s'
^
1
,9l|r
47r/u.
ox
1 9-f
1 9-5"
-,+
dyjr
.(36)
2^
9s^
4nrfJ.
dx'dz'
iirfi (\
-f
fi)
dx
and it follows that u is
given by
the
equation
^J^dF_
1 dH
X
9V^i
1
^^djr
2'7rfi
dz'
47rya
dx'
^irfi (\
+
/i)
dx
iirfi
dx"
where
^Iti
is an harmonic function which has the
property
d-^ildz'= ^.
Such
a function can be obtained
by introducing
a function
fl
by
the
equation
12
=
(s-)-s) log (s
+
s'--ii)-iJ.
(37)
230
SOLUTION OF THE
[CH.
X
Then fi is harmonic in the
space
considered and has the
property
an dn
If we write
dz dz'
^'
.(38)
F,=jjx,ndxdy, G,=jJY,ndxdy,
H,=
jjz.ndxdy/
'^'~
dx'
"^
dy'
^
dz'
[
...(39)
then all the functions
Fi, Gi, Hi,
yfr,,
are harmonic in the
space
considered
and
^Il
=
F
^^
=
(?
^^'
=
H
^
=
.>
(40)
dz'
'
dz'
^'
dz
'
dz'
^
^ '
In the same
way
as we found u we
may
find v in the form
_
\ dG \ dH
X
a^i
1
,a-f
,^j-
^~27r/xS/ 47r/i82/' 47r/L<,(X
+
/i)8y'
47r/x
^
8/
^
In the
special
case of a
iiressure
P
applied
at the
origin,
the tractions
Xv,
Y,
vanish
everywhere,
and
Z,
vanishes
except
in a small area
containing
the
origin,
but
\\Zu
dxdy=P.
In this case Fand O
vanish,
and
,
95" P
where r denotes the distance of
(V, y', z)
from the
origin.
Also
F^
and
Gj
vanish and
05i
V']
=
^=i'log
('+r). Suppressing
the accents we obtain the formulae
(35)
of Article 135.
167. Historical Note.
The
problem
of the
plane

sometimes also called the

"problem
of
Boussinesq
and
Cerruti"

has been the

object
of numerous researches. In addition to those mentioned
in the Introduction
pp. 15,
16 we
may
cite the
following
:

J.
Boussinesq,
Paris C.
R.,
1. 106
(1888), gave
the solutions for a more
general type
of
boundary conditions,
viz. : the normal
traction and
tangential displacements
or normal
displacement
and
tangential
tractions are
given.
These solutions were obtained
by
other methods
by
V. Cerruti Rome Ace. Lined
Rend.
(Ser. 4),
t. 4
(1888)
and
by
J. H.
Michell,
London Math. Soc.
Proc,
vol. 31
(1900),
p.
183. The
theory
was extended
by
J. H.
Michell,
London Math. Soc.
Proc,
vol. 32
(1901), p. 247,
to
ajolotropic
solid bodies which are
transversely isotropic
in
planes parallel
to the
boundary.
The solutions
given
in Articles 165 and 166 were obtained
by
a new
method
by
C.
Somigliana
in II Nuovo Cimento
(Ser. 3),
tt. 1720
(18851886),
and this
was followed
up by
G. Lauricella in II Nuovo Cimento
(Ser. 3),
t. 36
(1894).
Other
methods of
arriving
at these solutions have been
given by
H.
Weber,
Part.
Diff.-Gleiehungen
d. math.
Phydh,
Bd.
2,
Brunswick
1901, by
H.
Lamb,
London Math. Soc.
Proc,
vol. 34
(1902), by
O.
Tedone,
Ann. di mat.
(Ser. 3),
t. 8
(1903),
and
by
R.
Marcolongo,
Teoria
matematica dello
equilihrio
dei
corpi elastici,
Milan 1904. The extension of the
theory
to
the case of a
body
bounded
by
two
parallel planes
has been discussed
briefly by
H.
Lamb,
loc.
cit.,
and more
fully by
J.
Dougall, Edinburgh Roy.
Soc
Trans.,
vol. 41
(1904),
and also
by
0.
Tedone,
Rend. d. Circolo mat. di
Palermo,
t. 18
(1904).
166-168]
PROBLEM OF THE PLANE
281
168.
Body
bounded
by plane

Additional results.
(a)
In the calculation of the rotations when the surface tractions are
given
we
may
take the
point
^ of Article 163
(6)
to be at an infinite
distance,
and omit
i^^M), ...altogether.
We should find for
u^, v^', w^'
the forms
*
dxdy'dz X+fidxd)/'
,(,"--22^4.^^
S'X ,
Six
'"*
~
^'
dydz^
+
X
+,t dydz
^
dydz
'
and we
may
deduce the formula
In like manner we
may prove
that
1 r
X+2/X 8V.
8
/9G'_8^1
'^''~47r^L x+n Sx-'^ay V8*-' sy/J'
For the calculation of
or,
we should
require
a
subsidiary displacement
which would
give
rise to the same surface tractions as the
displacement {dr~^/di/, dr~^/dx, 0),
and this
displacement
is
clearly {-dR-^jdy, dR'^ftx, 0)
and we can deduce the formula
(h)
As an
example
of mixed
boundary
conditions we
may
take the case where
u, V, Zy
are
given
at 2=0. To calculate A we
require
a
displacement {u,
v', w')
which at

=
shall
satisfy
the conditions
'=, V'=Vf,,
Zy'=Z^V\
where
(AV, JVj ^y)
is the surface traction calculated from
(m', i/, w').
Then we
may
show that the value of A at the
origin
of r is
given by
the
equation
in
(X
+
2^.)
A
=
[
[{(AV")
-
AV)
u
+
(
r,()
-
IV) v-Z, {wo
-
1(^)} dxdy.
We
may
show further that
232 MODIFICATION OF BETTl'S
[CH.
X
(c)
As a second
example
we
may
take the case where
Xy, T^,
w are
given
at z=0.
To calculate A we
require
a
displacement (u", v", w")
which at 2=0 shall
satisfy
the
conditions
X/=Z(<'), IV'=iy),
w"
=
;,
where
X,", Ty", ZJ'
denote the surface tractions calculated from
(m", v", w").
We can
prove
that the value of A at the
origin
of r is
given by
the
equation
4JT
(X
+
2^i)
A
=
i
[{Xy {u"
-
Mo)
+ Tv
(v"
-
f
o)
+
(^<'''
-
Zy") w} dxdy,
and that
and then we can find for A the foimula
1 d
(dF ,dQ
dN\
2n{\
+
2,i)dz' \dx'
""
dff-
^>'
dz'
J
'
and for
(u, v, w)
the formulis
I
(X
+
2^)
^
3^
W'"3y
'*
32'
J
'
47r/i(X
+
2^)
'2n^
d^
^
2n-
3/
^
4n-,i (X+2p) 3y V3y
^
3y'
'^
/
I
3yV8'^
+
3y' ^'*3/J' 47r/i(X
+
2|i)
'
\dx' 3y
"
1 dJV
X+/i
'
8 /3i?' 3g
_^
*"
2n- 3/
4ff/x(X
+
2/x)^
32'
W"''3y
'^32'
169. Formulae for the
displacement
and strain.
By
means of the
special
solutions which
represent
the effect of force at a
point
we
may
obtain formula!
analogous
to
(7)
for the
components
of
displacement.
Thus let
(ui, Vi, Wj) represent
the
displacement
due to unit force
acting
at
{x', y, 2*)
in the
direction of the axis of
x,
so that
. .
X+^
/3V 3V 3V X +
2;i
1\ ,
and let
X^C), JV), ZJ^)
be the surface tractions calculated from
(u^, v^, it\).
We
apply
the
reciprocal
theorem to the
displacements (, v, w)
and
(i, v^, w,),
with a
boundary
consisting
of the surface S of the
body
and of the surface 2 of a small
sphere suiTounding
(y, y', z'),
and we
proceed
to a limit as before. The contribution of 2 can be evaluated as
before
by finding
the work done
by
the unit
force, acting
over the
displacement (ti, v, w),
and the same result would be arrived at
analytically.
If the
body
is
subjected
to
body
forces
(X, Y, Z)
as well as surface tractions
Xy, Yy, Zy,
we find the formulae*
()q=
I I
\p{Xuy+Yvy{-Zw^dxdydz
+
{
([{XyUl
+
YyVi+ZyWl)
-
(.!',(')+
Yy^V
+
ZyWlO)] dS, (43)

The formnlffi of this

type
are due to C.
Somigliana,
II Nuovo Cimento
(Ser. 3),
tt. 1720
(1885, 1886)
and Ann. di mat.
(Ser. 2),
t. 17
(1889).
168,
169]
METHOD OF INTEGRATION
233
where the volume
integration
is to be taken
(in
the sense of a
convergent integral)
throughout
the volume within S. We should find in the same
way
+
j j[{X,u^+
Y^v^JrZyW^)
-
(A\(2)u+ YJ.nv+Z,<^w)\ dS,
and
{w\=
I
I
\p{Xu^+Yv^-\-Zw^dxdydz
A method of
integration
similar to that of Betti has been founded
upon
these formulae*.
It should be noted that no
displacement
exists
which,
besides
satisfying
the usual con-
ditions of
continuity
and the
equations
of
equilibrium (1), gives
rise to surface tractions
equal
to
AV'), I'^C, Z^^^\
or to the similar
systems
of tractions
XyP),
... and
AV'), ...,
for
none of these satisfies the conditions of
rigid-body-equilibrium
t.
When the surface
tractions are
given
we must
introduce,
in addition to the unit forces at
(.t^,y, z'\ equal
and
opposite
unit forces at a chosen
point A, together
with such
couples
at A as
will,
with the
unit
forces, yield
a
system
in
equilibrium.
Let
(w/, Vj', w-{)
be the
displacement
due to
unit force
parallel
to x at
{x', y, ^)
and the
balancing system
of force and
couple
at
A,
and let
X/O, F'('), Z/O
be the surface tractions calculated from
(m,', v^, w{).
Also let
(Mi", fi", 10^')
be the
displacement which,
besides
satisfying
the usual conditions of
continuity
and the
equations
of
equilibrium (1), gives
rise to surface tractions
equal
to
X/C), Fy'l'), Zy'y^.
We make the
displacement precise by supposing
that it and the corre-
sponding
rotation vanish at A. Then we have
(''^^ I I
j
p{Xui'+Yvi+Zu\')dxdi/dz
+
jj{X,
(,'
-
O
-1- n
(v^'
-
v^")+Zy (?(;,'
-
<')}
dS.
(44)
The
problem
of
determining
u is reduced to that of
determining (itj", ;/', w/').
The dis-
placement (ii'

ttj", v^'

Vj^', ?('i'
-
lo^")
is an
analogue
of the second Green's function.
If,
instead of
taking
the
displacement
and rotation to vanish at
A,
we
assign
to .4 a
series of
positions very
near to
{x', y', 2'),
and
proceed
to a limit
by moving
A
up
to coin-
cidence with this
point,
we can obtain
expressions
for the
components
of strain in terms
of the
given
surface tractions
J.
In the first
place
let us
apply
two
forces,
each of
magnitude
h~^ at the
point {x^, y', z')
and at the
point (.r'-t-A, y, z'),
in the
positive
and
negative
directions
respectively
of the axis of x. In the limit when h is diminished
indefinitely
the
displacement
due to these forces is
[-^
,
^, jr)-
L*
("m %> ""u)
be the
displacement produced
in the
body by
surface tractions
equal
to those calculated
from the
displacement
( ^-
,
^
,
-^ )
Then the value of
(du/dx)
at the
point (x', y', z')
is
given by
the formula
In like manner formula3
may
be obtained for
cv/c'y
and
dw/dz.
*
G.
Laurieella,
Pisa
Ann.,
t. 7
(1895),
attributes the method to Volterra. It was
applied by
. C.
Somigliana
to the
problem
of the
plane
in II Nuovo Cimento
(1885, 1886).
+
J.
Dougall,
loc. cit.
p.
222.
J
G.
Laurieella,
loc. cit.
234 VARIOUS METHODS
[CH.
X
Again,
let us
apply
forces of
magnitude
h~^ in the
positive
directions of the axes of
y
and z at the
origin
of
r,
and
equal
forces in the
negative
directions of these axes at the
points {x', y',
z!
+
h)
and
{x', y'
+
h, z') respectively,
and
proceed
to a limit as before. This
system
of forces satisfies the conditions of
rigid-body-equilibrium,
and the
displacement
due to it is
/BMj 82 3^3 82 3^3
3m?.A
\oy
dz
'
dy
dz
'
dy
dz
)'
Let
(Mjsi %, Wja)
be the
displacement produced
in the
body by
surface tractions
equal
to
those calculated from the
displacement
\^
+
^^
)
)
Proceeding
as before we
obtain the
equation
(l+s)r-///'{^(l+?)*>-+5)(|-+w)}"''''
-//[-^'((+S')-)*^-{(
+
s)-'}*^-{(l'+t)-4]'"-<'
In like manner formulae
may
be obtained for
du/dz+dwjdx
and
dv/dx+du/dy.
170. Outlines of various methods of
integration.
One method which has been
adopted
sets out from the observation
that,
when there
are no
body forces, m^, ra^j, nr,,
as well as
A,
are harmonic functions within the surface of
the
body,
and that the vector
{ca-c, cjj,, ot^)
satisfies the circuital condition
dx
dy
dz
From this condition it
appears
that
axx, ^y,
cr^
should be
expressible
in terms of two in-
dependent
harmonic
functions,
and we
may
in fact write*
d(l> d^ 3;^
'^''
=
dy"^'di~^dz'
'
ez
dy
"
dx'
where
<f>
and
x
^^^ harmonic functions.
The
equations
of
equilibrium,
when there are no
body forces,
can be written in such
forms as
dy
dz dx
\dy^^ dz^J dxdy
dxdz
'
"
(^
a^"^"^ a^"*"^
a^ay
"
^
d^dz)
and it follows that
3
/
dl^y'dl/^'^
dz

Cf.
Lamb, Hydrodynamics (Cambridge, 1895), pp.
526528.
169,
170]
OF INTEGRATION 235
This
expression represents,
as it
should,
an harmonic
function;
and the
quantities
A, -m^, nr^, or,
are thus
expressible
in terms of two
arbitrary
harmonic functions
<^
and
x-
If now these functions can be
adjusted
so that the
boundary
conditions are satisfied
A and
(tB;j, or^, tjr,)
will be determined. This method has been
applied successfully
to the
problem
of the
sphere by
C. Borchardt* and V.
Cerrutif.
Another method
| depends upon
the observation
that,
in the notation of Article
132,
W2='i, W3=tCi, '2
=
4'3!
and therefore the surface traction
JT^C)
can be
expressed
in
the form
where
I, m,
n are written for cos
{x, v),
cos
(y, v),
cos
{z, v).
The surface tractions
XP), AV^)
can be written down
by putting
v and w
respectively everj'where
instead of u
in the
expression
for
X(^).
It follows that
(X('), XJ-^\ A'^P))
is the
displacement produced
by
certain double forces. In like manner
(lyi), r^(2), 1\^^))
and
{Z,m, Z^^\ Z^C))
are
systems
of
displacements
which
satisfy
the
equations (1) everywhere except
at the
origin
of
r.
On this result has been founded a method
(analogous
to that of C. Neumann
||
in
the
theory
of
Potential)
for
solving
the
problem
of
given
surface
displacements by
means
of series.
*
Berlin
Monatsber.,
1873.
t
Comptes
rendus de VAssociation
Fran^aise pour
I'avancement de
Science, 1886,
and Rome Ace.
Lincei Eend.
(Ser. 4),
t. 2
(1886).
t
G.
Lauricella,
Pisa Ann. t. 7
(1895),
and Ann. di mat.
(Ser. 2),
t. 23
(1895J,
and II Nuovo
Cimento
(Ser. 4),
tt.
9,
10
(1899).

The result is due to C.

Somigliana,
A7m. di mat.
(Ser. 2),
t. 17
(1889).
II Untersuchungen
ilber das
logarithmiscke
und Newton'sche
Potential, Leipzig,
1877. Cf.
Poincare,
loc. cit.
p.
220.
CHAPTER XI.
THE
EQUILIBRIUM
OF AN ELASTIC SPHERE AND RELATED PROBLEMS.
171. In this
Chapter
will be
given examples
of the
application
of the
method of series
(Article
158)
to the
problem
of the
integration
of the
equations
of
equilibrium
of an
isotropic
elastic solid
body.
Of all the
problems
which have been solved
by
this method the one that has attracted
the most attention has been the
problem
of the
sphere.
In our treatment
of this
problem
we shall follow the
procedure
of Lord
Kelvin*,
retaining
the
equations
referred to Cartesian coordinates instead of
transforming
to
polar
coordinates,
and we shall
give
his solution of the
problem.
The solution is
expressed by
means of infinite
series,
the terms of which involve
spherical
harmonics. We shall
begin
with a
general
form of solution
involving
such
functions.
172. Solution in
spherical
harmonics of
positive degrees.
We
propose
to solve the
equations
/3A aA
aA>
where
('^+'^>(at
^'
f)+MVH.^,-)=o.
(1)
. du dv dw ,,
^
=
9^
+
8y+a7'
('>
subject
to the conditions that
u, v,
w have no
singularities
in the
neighbour-
hood of the
origin.
Since A is an harmonic
function,
we
may express
it as a sum of
spherical
solid harmonics of
positive degrees,
which
may
be infinite in number. Let
A
be a
spherical
solid harmonic of
degree
n,
that is to
say
a rational
integral
homogeneous
function of
a;,
y,
z of
degree
n which satisfies
Laplace's equation;
then A is of the form
A
=
2A,
*
See
Introduction,
footnote 61. References will be
given
in the course of the
Chapter
to
other solutions of the
problem
of the
sphere,
and additional references are
given by
E.
Marcolongo,
Teoria matematica dello
equilibria
dei
corpi
elaatici
(Milan, 1904), pp. 280,
281.
171,
172]
SOLUTION IN SERIES OF SPHERICAL HARMONICS 287
the summation
referring
to different values of n. Take one term
A,
of the
series,
and observe that
d^njdx
is a
spherical
solid harmonic of
degree
n

1,
and
that,
if r denotes the distance of the
point {x, y, z)
from the
origin,
We see that
particular integrals
of
equations (1)
could be written in
such forms as
^
_
X
+
/i
r'
aA
"~
/x 2(2n+l)
8a;
'
and more
general integrals
can be obtained
by adding
to these
expressions
for
u,
...
any
functions which
satisfy Laplace's equation
in the
neighbourhood
of the
origin, provided
that the
complete expressions
for
u,
...
yield
the
right
value for A. The
equations (1)
and
(2)
are
accordingly integrated
in the forms
A
=
2A,
X
+
M^
r-
/9A
aA aA,A
^

(3)
<"'"'^)
=
-^7~^ 2(2
+
i)
l-a^'
-W'
-87J
+
^^^"'^'^"\
in which
f/, F, Wn
are
spherical
solid harmonics of
degree
n,
provided
that these harmonics
satisfy
the
equation
^^"
=
-'-7^^5S^^.
+
^(t
+
t-'#)
W
Introduce the notation
^
=
-T*-
+
^y"+-87-'
(")
then
y\rn
is a
spherical
solid harmonic of
degree
n,
and
equation (4) requires
that
A,i
and
-\/r
should be connected
by
the
equation
(2w
+
l)/x
The harmonic function
A
is thus
expressed
in terms of the
comple-
mentary
functions
tr+i,
...
;
and the
integrals (.3) may
be
expressed
as sums
of
homogeneous
functions of
degree
n in the forms
(,.,,.)
=
-Silfr=(^-^,
^-1;^,
'^^)+2(fr,F,Tr),
...(7)
where
f7, F, TF
are
spherical
solid harmonics of
degree
n, Jl/
is the
constant
expressed by
the
equation
^^"~2(n-l)\
+
(3ra-2)/i'
^ -*
and
tItji-i
is a
spherical
solid harmonic of
degree
n

\
expressed by
the
equation
238 SOLUTION OF THE
[CH.
XI
It
may
be observed that
equations (7)
also
give
us a solution of the
equations
of
equi-
librium when n is
negative,
but such a solution
is,
of
course,
valid
only
in
regions
of
space
which exclude the
origin.
As an
example,
we
may put
n=
-\,
and take
We should thus obtain the solution which was discussed in Article 131.
173. The
sphere
with
given
surface
displacements.
In
any region
of
space containing
the
origin
of
coordinates,
equations (7)
constitute a
system
of
integrals
of the
equations
of
equilibrium
of an
isotropic
solid
body
which is free from the action of
body
forces. We
may adapt
these
integrals
to
satisfy given
conditions at the surface of a
sphere
of radius a.
When the surface
displacements
are
prescribed,
we
may suppose
that the
given
values of
u, v,
w a.t r
=
a are
expressed
as sums of surface harmonics of
degree
n in the forms
(u,
V,
W)r-a
=
'^{^n, Bn,
Cn) (10)
Then
r^A-n, r^Bn, r"C
are
given spherical
solid harmonics of
degree
n.
Now select from
(7)
the terms that contain
spherical
surface harmonics
of
degree
n. We see that when r
=
a the
following equations
hold :

An

-iun+uCl'
o Cl-^n+1
^.
=
-il/+.a^^^+F,
I.
(11)
The
right-hand
and left-hand members of these
equations
are
expressed
as
spherical
solid harmonics of
degree
n,
which are
equal respectively
at the
surface r
=
a. It follows that
they
are
equal
for all values of
x,
y,
z. We
may accordingly
use
equations (11)
to determine
Tin, Vn, W
in terms of
"iij -"n; ^n-
For this
purpose
we differentiate the left-hand and
right-hand
members
of
equations
(11)
with
respect
to
x,
y,
z
respectively
and add the results.
Utilizing equation
(9)
we find the
equation
^-.=,^&>-)-a^(S>.)4.g''.)
(-)
Thus all the functions
^n
are determined
in terms of the
corresponding
An, Bn, Cn,
and then
Un,
.. are
given by
such
equations
as
TJ -i^A +M a'^^^^
'
~
a"
"*"''
dx
'
The
integrals
(7) may
now be written in the forms
(u,v,.)=^l^JAn,Bn,Cn)+%Mn,.ia^-r^)(^-^\
'^^,
'-t^y-im
172-174]
PROBLEM OF THE SPHERE
239
in which
"+"
2(n+l)X
+
(3n
+
4)/.'
and
t.-.:
=9-
t..
^n.=
J
+
9-
(^,
.,
)
+
^-^
(^^
C,,
)
.
By equations (13)
the
displacement
at
any point
is
expressed
in terms of
the
prescribed displacements
at the surface of the
sphere.
174. Generalization of the
foregoing
solution.
(i)
The
expressions (7)
are
genenxl integrals
of the
equations
of
equilibrium
arranged
as sums of
homogeneous
functions of
x,
y,
z of various
integral degrees. By selecting
a
few of the terms of lowest orders and
providing
them with undetermined coefficients
we
may
obtain solutions of a number of
special problems.
The
displacement
in an
ellipsoid
due to rotation about an axis has been found
by
this method*.
(ii)
If wo omit the terms such as
J
(r/a)"
from the
right-hand
members of
equations
(13)
we arrive at a
displacement expressed by
the
equation
(. ", >) =('*'-'')
(s^'si;'!)^'-
(14)
This
displacement
would
require body
force for its
maintenance,
and we
may
show
easily
that the
requisite body
force is derivable from a
potential equal
to
?[(H-l)X
+
(3
+
4),i]V.^
and that the
corresponding
dilatation is
-2(n+l) \/^ + j.
We observe
that,
if X and
y.
could be connected
by
an
equation
of the form
(jl-l-l)X
+
(3n
+
4);:*
=
0,
(15)
the
sphere
could be held in the
displaced configuration
indicated
by equation (14)
without
any body
forces,
and there would be no
displacement
of the surface. This result is in
apparent
contradiction with the theorem of Article 118
;
but it is
impossible
for X and
n
to be connected
by
such an
equation
as
(15)
for
any positive integral
value of
n,
since the
strain-energy-function
would not then be
positive
for all values of the strains.
(iii)
The results
just
obtained have
suggested
the
following generalization
t :

Denote
{\
+
li)lfi. by
T. Then the
equations
of
equilibrium
are of the form
ox
We
may suppose
that, answering
to
any given bounding surface,
there exists a
sequence
of
numbers, say r,, t^, ...,
which are such that the
system
of
equations
of the
type
T
C^-+t-'')-'---=.
'-.v^.)
*
C.
Chree, Qiiart.
J.
of
Math. vol. 23
(1888).
A number of other
applications
of the method
were made
by
Chree in this
paper
and in an earlier
paper
in the same
Journal,
vol. 22
(188G).
+
E. and F.
Cosserat,
Paris C.
R.,
tt. 126
(1898),
133
(1901).
The
generalization
here
indicated
is connected with researches on the
problem
of the
sphere by
E.
Almansi,
Rome Ace.
Lincei Rend.
(Ser. 5),
t. 6
(1897),
and on the
general equations by
G.
Lauricella,
Ann. di mat.
(Ser. 2),
t. 23
(1895),
and II Nuovo Cimento
(Ser. 4),
tt.
9,
10
(1899).
240
THE PROBLEM OF THE SPHERE
[CH.
XI
possess
solutions which vanish at the surface. Denote
8{7/9.r
+
3r,/8y4-3Tr,/32 by
A.
Then
A^
is an harmonic
function,
and we
may prove
that,
if k is different from
k,
Jl!
AkAk'
dxdi/dz

0, (16)
when the
integration
is extended
through
the vohime within the
bounding
surface. We
may suppose accordingly
that the harmonic functions
A,
are such that an
arbitrary
harmonic function
may
be
expressed,
within the
given surface,
in the form of a series
of the functions A
with constant
coefficients,
as is the case with the functions
V'+i
when the surface is a
sphere.
Assuming
the existence of the functions
t',
... and the
corresponding
numbers
t,,
we
should have the
following
method of
solving
the
equations
of
equilibrium
with
prescribed
displacements
at the surface of the
body
:

Let functions
Uq, Vq, lo^
be determined so as to
be harmonic within the
given
surface and to
take,
at that
surface,
the values of the
given
components
of
displacement.
The function
Mq,
for
example,
would be the
analogue
of
2

An
in the case of a
sphere.
Calculate from
?, Vq, Wq
the harmonic function
Aq
deter-
mined
by
the
equation
_3m
ao
3wp
Assimie for
u, v,
w within the
body
the
expressions
(, V, >)=(, t'o,
w)-r2--^((7,
F
If,), (17)
T

Tk
where the J's are constants. It
may
be shown
easily
that these
expressions satisfy
the
equations
of
equilibrium provided
that
2^,A,c
=
Ao.
The
conjugate property (16)
of the functions
A^
enables us to
express
the constants A
by
the formula
Ak
ill
{:^Kfdxdydz=
j j j
AA^dxdi/dz, (18)
the
integrations being
extended
through
the volume of the
body.
The
problem
is there-
fore solved when the functions
Uk,
...
having
the assumed
properties
are found*.
175. The
sphere
with
given
surface tractions.
When the surface tractions are
prescribed,
we
may suppose
that the
tractions
Xr, Yr> Z^
a,t r
=
a are
expressed
as sums of surface harmonics of
various
degrees
in the forms
{Xr, l-r, Zr)r^a

'^{Xn,
In,
Zn) (19)
so that
r^Xn, r'^Yn, i"'^Zn
are
given spherical
solid harmonics of
degree
n.
Now
Xr,
... are
expressed
in terms of
strain-components by
formulae of
the
tjrpe
*
E. and F.
Cosserat,
Paris C,
R.,
t. 126
(1898),
have shown how to determine the functions
in
question
when the surface is an
ellipsoid.
Some solutions of
problems relating
to
ellipsoidal
boundaries
have been found
by
C.
Chree,
loc. cit.
p. 239,
and
by
D.
Edwardes, Quart.
J.
of
Math.
vols. 26 and 27
(1893,
1894).
241 174,
175]
WITH GIVEN SURFACE TRACTIONS
and these are
equivalent
to formulsB of the
type
rXr
A.
A 9f
()u
=
-a;A
+
^+rr
u,
(I fi
dx dr
in which
^=vx
+
vy
+
ZVZ,
so that
f/r
is the radial
component
of the
displacement.
We have now to calculate
X,,
...
by
means of the formulae of
type (20)
from the
displacement expressed by
the
equations (7).
We know
already
that this
displacement
can be
expressed by
such formulfE as
.(20)
(21)
u
=
l
An-n
+
M^+,a;-
^^
^^"+'-rW""^"-
dx
(22)
We
proceed
to calculate
Xr, Yr, Z^
from these formulae. In the result
we shall find that
A^, Bn, C
can be
expressed
in terms of
X, F, Zn-
When these
expressions
are obtained the
problem
is solved.
We have at once
r=s (xAn
+
!/Bn
+
zC'n)
-
+
Mn+,a%n
+
l)-^n+,
-
Mr^
{n-l)yjr,
ni
The terms such as
xAnr"ja"
are
products
of solid
harmonics,
and we
transform them into sums of terms each
containing
a
single
surface harmonic
by
means of such identities as
xf{x,y,z)
=
In +
1 dx a2"+' dx
\r^+^ f)
.(23)
We obtain in this
way
the
equation
{xA
+
yBn^-zGr,)-
=
a"2n+lV'''"~'
a^+i*^-"-')'
.(24)
where
i^n-i
is
given by equation (12),
and
0_,i_2
is a
spherical
solid harmonic
of
negative degree

(n
+
2)
which is
given by
the
equation
Hence we have
+
Jf+j
a=
(n
+
1) ^n+i
-
M
r^
(n
-
1) -^n-i
I
L. E.
..(26)
16
242 THE PROBLEM OF THE SPHERE
[CH.
XI
in which the
expression
under the
sign
of summation is
homogeneous
of
degree
m +
1 in
x,
y,
z,
and we find
-
5^^
1^
-
..<-
')} f-l?-
SS 'L
(IS
^.-.)}]
(^')
where we have used an
identity
similar to
(23)
to transform
a;i|r,j_i.
Again
we have
A
=
s'4?J^i>/t_.
(28)
and therefore
'^^-\+^'^ ^^""X
dx a="-i 8*'
Vr--'
^"-VJ
^^^
Also we have
du
r M
=
Z
or
(n
-
1)
j4
^;
+
a-'M^,,
^-^
-
r-M
^"^"^
dx dx
...(30)
We
may
now select the terms of
degree
n in the
expression
on the
right-
hand side of
equation (20) ;
they
are
dx 2n + ldx
Va^"+'
<^-n-i
11
,, , ,x
4, 2\
dx
+
-:
that of r^+'
~
f ^I
IS
-
+
2_(n-l)
^
_
JX_^
I ^+,l(^zA
X in this
expressior
\(w
+
2)-/i(ri-3)
(2n-l)(2ri
The coefficient of
r^3i|r_,/9a;
in this
expression
is

2(?i

2)J!/,
and
Hence,
denoting
dxXr'^-^j
(2n
+
l){\(n-l)
+
Ai(37i-2)}
this coefficient
by E^,
we have the
expression
for the terms of
degree
n in
rXrlII
in the form
(
-
1
) ^,
^;
+
2n
il/+,a^
^fe -
2
(
-
2)
if,'^
^"^
a
^ ^
"
dx
2n
+
1 dx U-+' ^-""V
"
dx
[r^^-'J
'
' '
"^^^^
where

1
X(w
+
2)-;^(w-3)
"~
2+l\(n-l)
+
M(3w-2)'
.(32)
175,
176]
WITH GIVEN STJKFACE TRACTIONS 243
From the sum of all the terms in the
expression
for
rXrIfi
we select
those which contain
spherical
surface harmonics of
degree
n. The value
of the sura of these terms at the surface r
=
a must be the same as the
value of
r"X/a"~y
at this surface. We have therefore the
equation
which holds at the surface r
=
a. Since both members of this
equation
are
spherical
solid harmonics of
degree
n,
the
equation
holds at all
points.
There are two similar
equations
which are derived from the above
by
replacing
An
successively by

and
0, X
successively by
F
and
Z,
and
d/dx successively by d/dy
and
d/dz.
To obtain the
expressions
for
An,
... in terms of
X,
... we introduce two
spherical
solid harmonics
'^_i
and
<&_n-2
by
the
equations
I
dx
Va""">'
'
dy
-'-'
-H^'')4A^'-)4S'-)'
.(34)
Then we differentiate the
right-hand
and left-hand members of the
equations
of
type (33)
with
respect
to
x,
y,
z
respectively,
and add the results. We
obtain the
equation
K-l)
+
n(2+l)^)V-_,=
-^_,
."
(35)
Again
we
multiply
the
right-hand
and left-hand members of the
equa-
tions
(33) by
X,
y,
z
respectively
and add the results. We obtain the
equation
20__,=
-*__
(36)
By equations
(35)
and
(36) ^n-\
and
^__o
are
expressed
as the
products
of
'^_i
and
^t-n-a
and constant factors. On
substituting
in the
equations
of
type (33)
we have
An, Bn, On
expressed
in terms of
X, F, Z.
The
problem
is then solved.
176. Conditions
restricting
the
prescribed
surface tractions.
The
prescribed
surface tractions
must,
of
course,
be
subject
to the conditions that are
necessary
to secure the
equilibrium
of a
rigid body.
These conditions show
immediately
that there can be no constant terms in the
expansions
such as
2X. They
show also that
the terms such as
X^, Y^, Z^
cannot be taken to be
arbitrary
surface harmonics of the
first
degree.
We must
have,
in
fact,
three such
equations
as
//
162
244 THE PROBLEM OF THE SPHERE
[CH.
XI
where the
integration
is extended over the surface of the
sphere. Writing
this
equation
in the form
and
transforming
it
by
means of identities of the
type (23),
we find the
equation
For
any positive integral
value of
,
the
subject
of
integration
in the second of these
integrals
is the
product
of a
power
of r
(which
is
equal
to
a)
and a
spherical
surface
harmonic,
and the
integral
therefore
vanishes,
and the like statement holds
concerning
the first
integral except
in the case n
=
\. In this case we must have three such
equa-
tions as
and these
equations
show that
rX-^, rFj, rZ-^
are the
partial
differential coefficients with
respect
to
, y,
z of a
homogeneous quadratic
function of these variables. Let
Ay),
...
be the
stress-components
that
correspond
with the surface tractions
Xj
Then we
have such
equations
as
rZj
=
:rZ^(i)
+.yX,(i)
-I-
X,(i),
It thus
appears
that
X^,
... are
constants,
and the
corresponding
solution of the
equations
of
equilibrium represents
the
displacement
in the
sphere
when the material is
in a state of
uniform,
stress.
in. Surface tractions directed
normally
to the
boundary.
When the surface traction consists of tension or
pressure
at
every point
of the surface
we
may
take the normal traction to be
expressed
as a sum of surface harmonics in the
form
'S.Rn-
Then we have at the surface
rXr=x^^Rn,
rYr=yS.^^K,
rZ,
=
22~R.
Now the first of these
equations gives
for
rX^
at r
=
a the formula
p ~/x 2M-I-3L" B^U"-'' 7
a2"--3
8^V^
+
i'^"*Vj-
The
right-hand
member of this
equation
must therefore be the same as the left-hand
member of
equation (33),
or it must be the same as
By
the
processes already employed
we deduce the two
equations
[(+l)
+
(.^2)(2.
+
5)^.,]^.,
=
(^f|^)
Q-' ^,
'^'^
-2^y <^-'-^=2;^3Uj ir--
and then we can
easily
find the
.4's,
ffa and C's. In the case where
2iJ
reduces to a
176-178]
WITH GIVEN SURFACE TRACTIONS 245
single
term
M^+i,
the
only A's,
... which occur have sufiixeH n and
+
2,
and we
may
show
thcU
"*'
ft (2
+
3) [(+i)4.(
+
2)(2n
+
6) ^-.^.J [3*- V-*'^ "^'jjr.a'
with like
expressions
for the Ba and C's.
178. Solution in
spherical
harmonics of
negative degrees.
When the
space occupied by
the
body
is bounded
by
two concentric
spheres*
solutions
can be obtained in the same
way
as in Article 172
by
the introduction of
spherical
harmonics of
negative degrees
in addition to those of
positive degrees.
To illustrate the
use of harmonics of
negative degrees
we take the ca.se where there is a
spherical cavity
in an
indefinitely
extended mass.
Using,
as
before, U, V, \V
to denote
spherical
solid
harmonics of
positive integral degree n,
we can write down a solution of the
equations
of
equilibrium
in the form
where
^ -r^n.s[l(
_Un_\
.
l(
J^\
.
I
(JK)-]
and
Jka
=
2(M
+
2)X
+
(3re
+
5)^'
The function
^a + i
is a
spherical
solid harmonic of
degree
re
+
1,
and the dilatation
calculated from the above
expression
for the
displacement
is
given by
the formula
_2^2n+3)ji, ,^^i
The solution
expressed by
a sum of
particular
solutions of the above
type
can be
adapted
to
.satisfy
conditions of
displacement
or traction at the surface of a
cavity
r
=
a.
An
example
of some interest is afforded
by
a
body
in which there is a distribution of
shearing
strain t. At a
great
distance from the
cavity
we
may
take the
displacement
to
be
given by
the
equation
{u,v,w)
=
(si/,0,0),
where s is constant. In this
example
we
may
show
that,
if the
cavity
is free from
traction,
the
displacement
at
any point
is
expressed by equations
of the form
-=
4.(?)+<^al(?)'
where
A, B,
C are
constants,
and we
may
find the
following
values for
A, B,
C :

,
3X+8^
_ _
3(X
+
^)
, ^
3(\
+
f.)
^=9xTr4M"^'
^=9JH-U/'-'^'
^=-9X
+
i4;."*-
*
Lord Kelvin's solution is worked out for the case of a shell bounded
by
concentric
spheres,
.and includes the solution of this Article as well as that of Articles
172,
173, 175.
t
See Phil.
Mag. (Set. 5),
vol. 33
(1892), p.
77.
246 THE PROBLEM OF THE SPHERE
[CH.
XI
The value of the
shearing
strain
-^
+
~r
can he calculated. It will he found
that,
at
the
point
x=0, y=0, r=a,
it is
equal
to
-
a. The result shows that the shear in
the
neighbourhood
of the
cavity
can be
nearly equal
to twice the shear at a distance from
the
cavity.
The existence of a flaw in the form of a
spherical cavity may
cause a serious
diminution of
strength
in a
body subjected
to
shearing
forces*.
179.
Sphere subjected
to forces
acting through
its volume. Par-
ticular solution.
When the
sphere
is
subjected
to
body
forces we seek in the first
place
a
particular
solution of the
equations
of
equilibrium
of the
type
(\
+
fi)^
+
/^^' +pX=0,
and
then,
on
combining
this solution with that
given
in
(13),
we obtain
expressions
for the
displacement
which are
sufficiently general
to enable us
to
satisfy
conditions of
displacement
or traction at the surface of the
sphere.
If the
body
force
(X, Y,
Z)
is the
gradient
of a
potential
V which satisfies
Laplace's equation,
the
particular integral
can be obtained in a
simple
form;
for,
within the
sphere,
V can be
expressed
as a sum of
spherical
solid
harmonics of
positive degrees-
Let F=
2F,
where
F
is such an harmonic
function,
and consider the
equations
of the
type
(x
+
m)|^
+
mV^
+
/^"
=
o.... (-^7)
Particular
integrals
of these
equations
can be obtained
by putting
_
d^
d(j>
dip
dx
'
dy'
dz
'
if
{\+2fi,)V^(f>
+
pV
=
0-
and we
may
therefore take for the
particular integrals
of the
equations
of
type (37)
the
following
:

'd d d-
(""^>-Va*'ay'3.
r=F
.(38)
2
(2w
+
3)
\
+
2;t4
With a view to
satisfying special
conditions at the surface of the
sphere
we calculate the
corresponding
traction
(X^,
F,.,
Zr) by
means of the formulae
of
type (20).
We find for the radial
displacement f/r
the formula
^
2(2n+S)X+-2ij,
and for the dilatation A the formula
A-
P
V ^~
\
+
2/t'^"-
*
Cf. Article
84, supra.
178-180]
DEFORMED BY BODY FORCES 247
Hence we find for
Xr
the formula
rXr^
jO
and this can be reduced
by
means of the
identity (23)
to the form
'X
(
w
+
2 n
]
9 .
,^r
\
M
""
"
"^
XW^^
"^
2(2
+
3)|
ai
^"^
^^\
rXr
p
fi
X +
2/i
'
\ +
(n
+
l)f,
dV^_
X(2n
+
3)+2^(n
+
l)
9^
/
J^
V
_(2n+l)/i
aa;
(2n
+ 1
) (2?i
+
3) /i
aaiVr^'+V
(39)
The formulae for
Yr
and
^^
can be written down
by substituting 9/9y
and
d/de
successively
for
9/9*
in the
right-hand
member of
(39).
180.
Sphere
deformed
by body
force
only.
When the surface is free from traction the
displacement
is obtained
by adding
the
right-hand
members of
equations (13)
and
(38),
in the former
of which the functions
An,
... are to be determined in terms of
Vn,
...by
the conditions that the sum of the
expressions
for
rXrjf'
in the left-hand
member of
(33)
and the
right-hand
member of
(39)
must vanish. We take
the
potential
XVn
to consist of a
single
term
Vn+i,
in which n
> 1,
and then
we have three
equations
of the
type
'(n + l)A "^-E r^n+^l(t^\--L_li!^A,
X+
Ijjb
a+2
+'
9a;Vr"'+V
2n
-f
1 9a;
Va^+'
'
X
+
jn
+
Tj fi dVn+i
X
(2w-|-o)
+
2n(w-|-2
.)
(2/1
-f
3) /i
"'
9
(2w
-f
3) (2n
-I-
5) /i
(40)
which hold at the surface of the
sphere,
and
therefore,
in accordance with an
argument already employed,
hold
everywhere.
We notice
that,
if the material is
incompressible
so that the ratio
/x/X vanishes,
the
particular integrals expressed by (.38) vanish,
but the surface tractions
depending upon
the
particular integrals
do not vanish. The
right-hand
member of
(39)
becomes,
in
fact,
2?i
+
3
In this case the
equations by
which
A,
... are to be determined are the same as those
which were used in Article
177, provided that,
in the
latter, (r/a)"*' R + i
is
replaced by
pFn + j.
It follows that the
displacement produced
in an
incompressible sphere by body
force derived from a
potential F+,
is the same as that
produced by purely
normal
surface traction of amount
p Fn + ja"*'//-"*'*.
Returning
to the
general
case,
we
find,
as in Article
177,
that
^+i
and
<^__2
are the
only yjr
and
<^
functions that
occur,
and that the
only
-4's,
...
*
Chree, Cambridge
Phil. Soc.
Trans.,
vol. 14
(1889).
248 GRAVITATING INCOMPRESSIBLE SPHERE
[CH.
XI
which occur have suffixes n and + 2.
By
the
processes already employed
we obtain the
equations
-
2;m UJ
'^-"-^
=
^rT2^
(2n
+
.S);.
('^
+
1^ ^"-
The value of
f
at r=a is

+ 3
_p 2Tr ,
a'
,
o'
/''\"'+'
.
"
2
(2/1
+
5)
\
+
2m
"
"+^
"^
2n
+
5
^"
~
2^^^ Uj
'^-"-^'
and,
since
-|r+,
and
^__2
are
multiples
of
F+i,
the radial
displacement
at the surface of the
sphere
is a
multiple
of
F+,/?-"+\
and it is
found,
after a
little
reduction,
to be
equal
to
fi
\r)
2w
{2w'
+ 8n
+
9)\+{2n''
+ 6n
+
6)fji,
^ ^
In like manner we
may
show that the radial
displacement
at a distance r
from the centre is
(2n
+
3)
X
+
(2w
+
2) /u,
(2n^
+
8m
+
9)
\
+
(2n2
+ 6n +
6)/M
/ 4- 9^ "1
...(42)
+
'_Zi"'
(n
+
1)
(n
+
3
)X
+
(n
+
2) /.
r^
^
'{2n^
+
87i +
9)\+{2ii'+6n
+
6)fi
Since the radial
displacement
is
always proportional
to
F,i+i,
all the
spherical
surfaces concentric with the
boundary
are strained into harmonic
spheroids
of the same
type,
but these
spheroids
are not similar to each other. In the
case where n= 1 the
ellipticities*
of all the
principal
sections increase from
the outermost to the
centre,
the ratio of the extreme values
being
5\
+
ifj.
: 8X +
G/xf.
181.
Gravitating incompressible sphere.
The chief interest of
problems
of the kind considered in Article 179
arises from the
possibility
of
applying
the solutions to the di.scussion of
problems relating
to the Eai'th.
Among
such
problems
are the
question
of the
dependence
of the
ellipticity
of the
figure
of the Earth
upon
the
diurnal
rotation,
and the
question
of the effects
produced by
the
disturbing
attractions of the Sun and Moon. All such
applications
are beset
by
the
difficulty
which has been noted in Article
75,
viz. :
that,
even when the
effects of rotation and
disturbing
forces are left out of
account,
the Earth
is in a condition of
stress,
and the internal stress is much too
great
to
permit
of the direct
application
of the mathematical
theory
of
superposable
small
*
The
ellipticity
of an
ellipse
is the ratio of the excess of the axis
major
above the axis minor
to the axis
major.
+
Kelvin and
Tait,
Nat.
Phil.,
Part ii.
p.
433.
180,
181]
DEFORMED BY BODY FORCES 249
strains*. One
way
of
evading
this
difficulty
is to treat the material of
which the Earth is
composed
as
homogeneous
and
incompressible.
When the
homogeneous incompressible sphere
is at rest under the
mutual
gravitation
of its
parts
the state of stress
existing
in it
may
be
taken to be of the nature of
hydrostatic pressure
"f"; and,
if
po
is the
amount of this
pressure
at a distance r from the
centre,
the condition of
equilibrium
is
dpoldr
=
-gpr/a, (43)
where
g
is the acceleration due to
gravity
at the
bounding
surface
=
a.
Since
jj
vanishes at this
surface,
we have
p,= yp(a^-9^)la. (44)
When the
sphere
is strained
by
the action of external forces we
may
measure the strain from the initial state as "unstrained"
state,
and we
may suppose
that the strain at
any point
is
accompanied by
additional stress
superposed upon
the initial stress
p^.
We
may
assume further that the
components
of the additional stress are connected with the strain
by
equations
of the
ordinary
form
X^
=
\!,
+
2fj,exx Yi
=
/xey2,
....
in which we
pass
to a limit
by taking
\ to be
very great compared
with
fi,
and A to be
very
small
compared
with the
greatest
linear
extension,
in such a
way
that XA is of the same order of
magnitude
as
fiexx,

We
may put
lim. XA
=

p,
and then
Po+p
is the mean
pressure
at
any point
of the
body
in the
strained .state.
Let V be the
potential
of the
disturbing
forces. The
equations
of
equilibrium
are of the form
The terms
containing

p
and
-
gp
cancel each
other,
and this
equation
takes the form
dp ^,
dV _
The
equations
of
equilibrium
of the
homogeneous
incompressible sphere,
deformed from the state of initial stress
expressed by (44) by
the action of
external
forces,
are of the same form as the
ordinary equations
of
equilibrium
of a
sphere subjected
to
disturbing
forces,
provided
that,
in the latter
equa-
tions,
XA is
replaced hy
-
p
and
/aA
is
neglected.
The existence
of the
*
The
difficulty
has been
emphasized by Chree,
Phil.
Mag. (Ser. 5),
vol. 32
(1891).
t Cf. J. Larmor
'
On the
period
of the Earth's free Eulerian
precession,"
Cambridge
Phil. Soc.
Proc,
vol. 9
(1898), especially
13.
250 GRAVITATING INCOMPRESSIBLE SPHERE
[OH.
XI
initial stress
p^
has no influence on these
equations,
but it has an influence
on the
special
conditions which hold at the surface. These conditions are
that the
deformed
surface is free from traction. Let the
equation
of the
deformed surface be r
=
a +
eS,
where e is a small constant and S is some
function of
position
on the
sphere
r
=
a. The
^'
inequality
"
eS must be such
that the volume is unaltered. We
may
calculate the traction
{X,,
Y
Z,)
across the surface
=
a + eS. Let
V, vi',
n' be the direction cosines of the
outward drawn normal v to this surface. Then
Xy
=
V
{Xx
-
Po)
+
m'Xy
+
n'X^.
In the terms
Xx,
Xy,
X^,
which are linear in the
strain-components,
we
may
replace
I', m',
n'
by xja, yja, zja,
for the true values differ from these values
by quantities
of the order e
;
but we must calculate the value of the term

I'pn
at the surface r
=
a+ eS
correctly
to the order e. This is
easily
done
because
p^
vanishes at r
=
a,
and therefore at r
=
a-'r
eS it
may
be taken
to be e(S
[-^l
,
or
gpeS. Neglecting
e^,
we
may
write
-l'p^
=
-gpeS.
Hence the condition that
X,
vanishes at the surface r
=
a
+
eS can be
written
iXr)r.a
+
l9peS==0
(45)
The conditions that
Y^, Z^
vanish at this surface can be
expressed
in similar
forms and the results
rnay
be
interpreted
in the statement: Account can
be taken of the initial stress
by assuming
that the mean
sphere,
instead
of
being
free from
traction,
is
subjected
to
pressure
which is
equal
to
the
weight per
unit of area of the material
heaped up
to form the
inequality*.
182. Deformation of
gravitating incompressible sphere by
ex-
ternal forces.
Let the external
disturbing
forces be derived from a
potential satisfying
Laplace's equation
; and,
within the
sphere,
let this
potential
be
expressed
as
a sum of
spherical
solid harmonics of
positive degrees
in the form
SM^n.
Let the surface of the
sphere
be
deformed,
and let the
height
of the
inequality
be
expressed
as a sum of
spherical
surface harmonics in the
form
S6<S', 6
being
a small
quantity
which is at most of the order of
magnitude
of the
inequality.
The attraction of the
inequality
is a
body
force
acting
on the matter within the
sphere,
and at
points
within the
sphere
this force is derived from a
potential
of amount
47r7pa2 (2n
+
1)-'
e
{rla^
S,
*
This result is often assumed without
proof.
It
appears
to involve
implicitly
some such
argument
as that
given
in the text.
181,
182]
DEFORMED BY BODY FORCES
251
where
7
is the constant of
gravitation.
When the
potential
of all the
disturbing
forces is
expressed,
as in Article
179,
in the form
SF,
we
have
^"=^^"+2r+i^"S^'"
(*6>
in which
'iirypa
has been
replaced by
the
equivalent expression 3g.
The
displacement
within the
sphere
is
expressed by
formulae of the
type
P
+s
An~~,
+
Mn{a'-r^)
di}f
.(47)
a"
"
' '
dw
where
An,
... are unknown surface
harmonics,
and
Mn
and
i/^n-i
are
expressed
by
means of
equations (9)
and
(12).
To
complete
the solution we must
determine the harmonics
An, B,i, C', *S
in terms of the known harmonics
Wn
In the
process
we make such
simplifications
as arise from the
assump-
tion that the material is
incompressible.
The
boundary
conditions which
hold at the surface r= a are of two kinds. We
have,
in the first
place,
the
kinenmtical condition that the radial
displacement
at this surface is that
which has been denoted
by
26iS, and,
in the second
place,
the condition
that the surface
traction,
calculated from the
displacements
of
type (47),
is
equivalent
to a
pressure equal
to the
weight
of the
inequality.
The kinematical condition is
expressed by
the
equation
n + 2
p
'-L^"+2n+l'"a"'^"_
2
(2m
+
3)
X +
2ya
When we select from this
equation
the terms that contain surface harmonics
of order n
+
1,
and
simplify
the result
by
means of the condition of incom-
pressibility,
we find the
equation
f)i+i '^n+i
"
2(1+5'^""^'
(2n
+
l)a="+^
The traction at the surface is
expressed by equations
of the
type
.(48)
rXr dVn
2?i + 1 1 dx
/^zn-ri
d
F
dx r+'
,
n ^
^" "
+
2
;"
^^^"a
(n-l)(2+l)'
.sn+i ^
(
rn-
dx
\r"-
d
fr'
2n + ldx\a^+' Cf>-n-.)
.(49)
which are obtained
by simplifying
the
expressions
in
(33)
and
(39)
in accord-
ance with the condition of
incompressibility.
The conditions in addition to
252
GRAVITATING INCOMPRESSIBLE SPHERE
[CH.
XI
(48)
which hold at the surface are obtained
by equating
the
expression
on the
right-hand
side of
(49)
to

/i^'^fpa;
26S.
We thus find the
equation
99^
"2n
+ l
\dx\a'
K'-^y-
%
^
r r^
[8
/r \ r^+
6,
Ai
L (2+l)M3\a'
/[t
"
2n+
1
(
aa; 9a;
V^+\
-l-A|^(
i;^n^
(n-l)(2w+l)
S^Vr-^"-^
1 a /7-'+
2n+l
aS
W"+'
<^-
0, -.(50)
which holds at the surface r
=
a. When we select from this
equation
the
terms that contain surface harmonics of order n we find the
equation
gpa'
2n
i Z'-"!'
o
^
gp
2
(n-2)
r^+'
^ /S_j
(2n
+
3) /i
2w
+
3
^"""^
dx
U"+' '"^V
(2n
-
1
) /^
2h
-
1
^""'
u-i 9*
V
?"
pa'
1
aifn+i P
r'"+^ a
/Tf_.
/i~2n
+ 3 a
^2m-iaa;Vr^-i
+
^
i)^^
(n-l)(2n-l-l)
aa;Vr^"-V
-2nTTaSl^*-H=^'
^'^^
in which the left-hand member is a solid harmonic of order n. Since this
harmonic function vanishes at the surface r
=
a,
it vanishes for all values of
X,
y,
z. There are two similar
equations,
which are obtained
by considering
the tractions in the directions of
y
and z.
We differentiate the left-hand members of the three
equations
of
type (51)
with
respect
to
x,
y,
z
respectively
and add the results. We thus obtain the
equation
gp 2(w-2)w(2w+l)
r^
p n(2)i-H)

n(n +
2) ,
-
+
(-!) t_i
+
^^TT"
"^"-^
=
^
This
equation
holds for all values of n. When we
replace
?i
by
?i
-1-
2 it
becomes
gp
2
(n
+
2) (2
+
5)
r+'

p
(n
+
2)(2n
+
o)
^
/* (2n-l-3) ^+'a+'^"+'~/.
'
2n
+
3
^"+'
2(n
+
2y+l , ,.
,.-
182-184]
DKFORMED BY BODY FORCES
253
Again
we
multiply
the left-hand members of the three
equations
of
type (51) by
x,
y,
z
respectively
and add the results. We thus obtain
the
equation
gpa^
2n(n
+
1)
r"+'

_pa^
n
+
1
^
2n r^+'
fi.
(2n
+
3)-^
^"+'
a''+'
*"+'
~
'^
2ir+3
"^"+'
"
W+\ (i^S+5
9--
=
0.
(53)
The
equations (48), (52), (53)
determine
S+,,
-^^rn+i, <f)-n-a
in terms
of
W+i.
Hence all the functions denoted
by
S,
yfr, ^
with
various suffixes
are
determined,
and the
equations
of
type (51)
determine the functions
An,
When the
potential
of the external forces reduces to the
single
term
Wn+i, '^n+i
and
<f)-n-2
are the
only
functions of the
types yjr
and
(f>
that
occur,
and
Sn+i
is the
only
function of the
type
S. The
equations
of
type (51)
show
that,
of the functions
A, ...,
those which occur have
suffixes either n. or n + 2. The result that
Sn+i
is a
multiple
of
W+i
may
be
interpreted
in the same
way
as the
corresponding
result noted in Article 180.
183.
Gravitating body
of
nearly spherical
form.
The case of a
nearly spherical body
of
gravitating incompressible
material can be
included in the
foregoing analysis.
The surface conditions as
regards
traction are still
expressed by
such
equations
as
(50),
but we have not now the kineniatical condition
expressed by (48).
If the
equation
of the surface is of the form r
=
a
+
(n + i'S + i
the
values of
ir + i
and
<^-_2
are
given by putting
zero for
W^ + i
in
equations (52)
and
(53),
and the harmonic functions such as
A^
and
A+2
a.re determined
by equations
of the
type
of
(51)
from which the Wa are omitted.
G. H. Darwin has
applied analysis
of this
kind, without, however,
restricting
it to the
case of
incompressible
material,
to the
problem
of
determining
the stresses induced in the
interior of the Earth
by
the
weight
of continents*.
Apart
from the
difficulty concerning
the initial stress in a
gravitating body
of the size of the Earth

a
difficulty
which we
seem unable to avoid without
treating
the material as
incompressible

there is another
difficulty
in the
application
of such an
analysis
to
problems concerning compressible
gravitating
bodies. In the
analysis
we take account of the attraction of the
inequality
at
the
surface,
but we
negle(-t
the
inequalities
of the internal attraction which arise from the
changes
of
density
in the interior
; yet
these
inequalities
of attraction are of the same
order of
magnitude
as the attraction of the surface
inequality.
To illustrate this matter
it will be sufficient to consider the case where the
density pf,
in the initial state is uniform.
In the strained state the
density
is
expressed by pf,(l

A) correctly
to the first order in
the strains. The
body force, apart
from the attraction of the surface
inequalities
and
other
disturbing forces,
has
components per
unit of mass
equal
to
gx/a, gyja, gzja.
Hence
the
e.xpressions
for
pX,
... in the
equations
of
equilibrium ought
to contain such terms as
gpi^xa'^{\
-A),
and the terms of
type-gpQxAJa
are of the same order as the attractions
of the surface
inequalities
+.
184.
Rotating sphere.
In the case of the Earth the most
interesting problems
are those of the
ellipticity
of
figure
due to the diurnal rotation and of the tidal deformation
*
Phil. Tram.
lioy. Soc,
vol. 173
(1882).
Darwin's results have been discussed
critically
by
Chree, Cambridge
Phil. Soe.
Trans.,
vol. 14
(1889),
and Phil.
Mag. (Ser. 5),
vol. 32
(1891).
+
See a
paper by
J. H.
Jeans,
Phil. Tram.
Roy.
Soc.
(Ser. A),
vol. 201
(1903).
254 SPHEROIDAL FIGURE OF
[CH.
XI
produced by
the attractions of the Sun and Moon. The effect of the rotation
can be
represented
as due to
body
force of
magnitude <'(, y, 0),
where
u is the
angular velocity,
and the force at
any point may
be derived from a
potential
of
magnitude ^m^ {x^
+
y'^).
This
potential may
be
arranged
as the
sum of two terms
ia> {x^
+
y^+ z')
-
^w' (2z'
-
x"-
-
f),
of which the
former,
equal
to
^co-r", gives
rise to a radial force
f
eoV. This
term can be included in the term

gpr/a
of
equation (43) by writing
1
-
1 instead of
g.
Since,
in the case of the Earth
a-a/g
is a small
fraction,
equal
to about
^^,
we
may,
for the
present purpose, disregard
this alteration oi
g.
The term

^m'' (2z''

w'^

y")
when
expressed
in
polar
coordinates is

|a)V (f
cos^0

^),
so that it contains as factors r^ and a
spherical
surface harmonic of
degree
2. We
may
determine the effect of the
rotation from the results of Article 182
by putting
n
=
1 and
using
this
expression

^coV(^cos^6

^)
for
W^.
When w
=
1
equations (52)
and
(53)
become
19
,
21
p
"^
rh
-
^
P
a^'P-' 5
/.
and
equation (48)
becomes
,

/Si
=
a
2 r^
2 r^
W,~ige,'-S,
1 1 7-''
3 a'
Hence we find the
height
of the harmonic
inequality
in the form
5
F.a^'
/
/ 19
n\
gr'
I
\
'Z
gpa
It follows that the
inequality
is less for a solid
incompressible sphere
of
rigidity
/x
than it would be for a
sphere
of
incompressible
fluid in the ratio
19 a
1 : 1 4
-s-
-^
If the
sphere
has the same size and mass as the
Earth, this
2
gpa
"^
ratio is
approximately equal
to
^
when the
rigidity
is the same as that of
steel,
and
approximately equal
to
^
when the
rigidity
is the same as that of
glass.
The
eilipticity
of the
figure
of the Earth is about
^.
The
ellipticity*
of a
nearly spherical spheroid
of the same size and mass as the
Earth,
con-
sisting
of
homogeneous incompressible
fluid,
and
rotating uniformly
at the
rate of one revolution in 24
hours,
is about
t^^.
The
ellipticity
which would
be obtained
by replacing
the
homogeneous incompressible
fluid
by homogeneous
*
An
equation
of the form
r=a{l-fe(JcoB2S-^)},
represents,
when e is
small,
a
nearly spherical spheroid
of
ellipticity
e.
184,
185]
ROTATING SOLID BODY
/
255
incompressible
solid material of the
rigidity
of
glass,
to
say nothing
of
steel,
is too small
;
in the case of
glass
it would be
j^ nearly.
The result that a
solid of considerable
rigidity
takes,
under the
joint
influence of rotation and
its own
gravitation,
an oblate
spheroidal figure appropriate
to the rate of
rotation,
and
having
an
ellipticity
not
incomparably
less than if it were
fluid,
is
important.
It is
difficult, however,
to base an estimate of the
rigidity
of the
Earth
upon
the above numerical results because the deformation of a
sphere
by
rotation is
very greatly
affected
by heterogeneity
of the material. In
the case of the Earth the
average density
of surface rock is about half the
Earth's mean
density.
It is not difficult to see
that,
in the case of an
incompressible
solid stratified in
nearly spherical layers
of
equal density,
deficiency
of
density
in the
layers
nearest the surface
may
tend to increase
the
ellipticity
of
figure
due to rotation *. In our
equations
we have taken
the
density
to be
uniform,
but we
may
take account of variations of
density,
in a
roughly approximate
fashion,
by observing
that the
weight
of the
inequality,
and the
potential
of it at internal
points,
must be
proportional
to the mean
density
of the surface
layer.
Let
p
denote this
density.
The
rough approximation
referred to would be made
by writing p
for
p
in the firet
two lines of
equation (50).
The result would be
that,
instead of the
expression
19 u, .
1
+
^
in the denominator of the
right-hand
member of
(54)
we should
have

I-
^r
If
P
were
ip
the numbers which were A for steel and
P
2
gpa
f 2f s
f
for
glass
would become
|
for steel and
^
for
glass,
and the
ellipticity
of the
figure
would,
if this
rough approximation
could be
trusted,
be increased
accordingly.
185. Tidal deformation. Tidal effective
rigidity
of the Earth.
The tidal
disturbing
forces also are derived from a
potential
which is a
spherical
solid harmonic of the second
degree.
The
potential
of the Moon
at
any point
within the Earth can be
expanded
in a series of
spherical
solid harmonics of
positive degrees.
With the terms of the first
degree
there
correspond
the forces
by
which the relative orbital motion of the two bodies
is
maintained,
and with the terms of
higher degrees
there
correspond
forces
which
produce
relative
displacements
within the Earth.
By analogy
to the
tidal motion of the Sea relative to the Land these
displacements may
be
called
"
tides." The most
important
term in the
disturbing potential
is the
term of the second
degree,
and it
may
be written
{M'yr''lD^) {^
cos-d
-i^),
where M denotes the mass of the
Moon,
D the distance between the centres
*
This result was noted
by Chree,
Phil.
Mag. (Ser. 5),
vol. 32
(1891), p.
249. In the case of
a
fluid, deficiency
of
density
in the outer
layers may
tend to diminish the
ellipticity
of
figure
due
to the rotation. In
Laplace's
"law of
density
in the interior of the Earth" the
pressure
and
density
are assumed to be connected
by
a certain
law,
and the
density
of the
heterogeneous
fluid
is
adjusted
so as to make the
ellipticity
the same as that observed in the case of the Earth.
See
Kelvin and
Tait,
Nat.
Phil.,
Part ii.
p.
403.
256 TIDAL EFFECTIVE RIGIDITY
[CH.
XI
of the Earth and
Moon, 7
the constant of
gravitation,
and the axis from
which 6 is measured is the line of centres*. This is the
"tide-generating
potential
"
referred to the line of centres. When it is referred to axes
fixed in the
Earth,
it becomes a sum of
spherical
harmonics of the
second
degree,
with coefficients which are
periodic
functions of the time.
Like statements hold with reference to the attraction of the Sun. With
each term in the
tide-generating potential
there
corresponds
a deformation
of the mean surface of the Sea into an harmonic
spheroid
of the second
order,
and each of these deformations is called a
"
tide." There are diurnal
and semi-diurnal tides
depending
on the rotation of the
Earth,
fortnightly
and
monthly
tides
depending
on the motion of the Moon in her
orbit,
annual and
semi-annual tides
depending
on the motion of the Earth in her
orbit,
and
a
nineteen-yearly
tide
depending
on
periodic changes
in the orbit of the
Moon which are characterized
by
the revolution of the nodes in the
Ecliptic.
The
inequality
which would be
produced
at the surface of a
homogeneous
incompressible
fluid
sphere,
of the same size and mass as the
Earth,
or of an
ocean
covering
a
perfectly rigid spherical
nucleus,
by
the force that corre-
sponds
with
any
term of the
tide-generating potential,
is called the
"
true
equilibrium height
"
of the
corresponding
tide. From the results
given
in
Article 184 we learn that the
inequalities
of the surface of a
homogeneous
incompressible
solid
sphere,
of the same size and mass as the Earth and as
rigid
as
steel,
that would be
produced by
the same
forces,
would be about
^
of the true
equilibrium heights
of the tides.
They
would be about
f
of
these
heights
if the
rigidity
were the same as that of
glass.
It follows that
the
height
of the ocean
tides,
as measured
by
the rise and fall of the Sea
relative to the
Land,
would be reduced in
consequence
of the elastic
yielding
of the solid nucleus to about
|
of the true
equilibrium height,
if the
rigidity
were the same as that of
steel,
and to about
f
of this
height
if the
rigidity
were the same as that of
glass.
The name "tidal effective
rigidity
of the Earth" has been
given by
Lord Kelvin + to the
rigidity
which must be attributed to a
homogeneous
incompressible
solid
sphere,
of the same size and mass as the
Earth,
in order
that tides in a
replica
of the actual ocean
renting
upon
it
may
be of the same
height
as the observed oceanic tides. If the tides followed the
equilibrium
law,
the
rigidity
in
question
could be determined
by
observation of the actual
tides and calculation of the true
equilibrium height.
Now the
dynamical
theory
of the oceanic tides on a
rigid nucleus;]:
shows
that,
in the absence
of
friction,
no tide would follow the
equilibrium
law
;
even in the case of
very
*
See 6. H. Darwin's Article
'
Tides
'
in
Ency. Brit.,
9th edition.
t Sir W.
ThomBon,
Phil. Trans.
Roy. Soc,
vol. 153
(1863),
and Math, and
Phys. Papers,
vol.
8, p.
317.
X
G. H.
Darwin,
Proc.
Roy. Soc,
vol. 41
(1886), p.
837. Cf.
Lamb,
Hydrodynamics,
Chapter
VIH.
185,
186]
OF THE EARTH 257
long periods
the
heights
of the tides on oceans of such
depths
as
actually
exist would be less than half of the
equilibrium heights.
The friction of the
ocean bed would tend to make a tide follow the
equilibrium
law more
closely
the
longer
the
period.
We must therefore confine our attention to tides of
long period.
Of these the annual and semi-annual tides are
entirely
masked
by
the fluctuations of ocean level that are due to the
melting
of ice in the
polar regions.
The
nineteen-yearly
tide is too minute to be detected with
certainty.
From observations of the
fortnightly
tides which were carried out
in the Indian Ocean
*
it
appeared
that the
heights
of these tides are
little,
if
anything,
less than two-thirds of the true
equilibrium heights.
If the
fortnightly
tide followed the
equilibrium
law we could infer that the tidal
effective
rigidity
of the Earth is about
equal
to the
rigidity
of steel. The
friction of the ocean bed is not
likely,
however,
to be
great enough
to render
valid the
application
of the
equilibrium theory
to the
fortnightly
tides.
The fact that there are observable tides at
all,
and the above cited
results in
regard
to the
fortnightly
tides in the Indian
ocean,
have been held
by
Lord Kelvin to
disprove
the
geological hypothesis
that the Earth has a
molten interior
upon
which there rests a
relatively
thin solid crust
; and,
on this and other
independent grounds,
he has contended that the Earth
is to be
regarded
as
consisting mainly
of solid material of a
high degree
of
rigidity.
The evidence from tidal
phenomena
seems not to be
absolutely
conclusive in favour of this
viewf.
186. Plane strain in a circular
cylinderJ.
Methods
entirely
similar to those of Articles 173 and 175
may
be
applied
to
problems
of
plane
strain in a circular
cylinder. Taking
r and to be
polar
coordinates in the
plane {x, y)
of the
strain,
we
have,
as
plane
harmonics
of
integral degrees, expressions
of the
type
r"
(a
cos nd
+
^n
sin
n6),
in which
a,,
and
/3
are
constants,
and as
analogues
of surface harmonics we have the
coefficients of r" in sUch
expressions.
We
may
show that the
analogue
of the
solution
(13)
of Article 173 is
(,)
=
2(^-,-j-H2-(^^^3^2^^^(^^,^,
...(o5)
in which
An
and
-B
are functions of the
type

cosn^-f-
/S
sin
0,
and the
functions
v/r
are
plane
harmonic functions
expressed by equations
of the form
The
equations (55)
would
give
the
displacement
in a circular
cylinder
*
Kelvin and
Tait,
Nat.
Phil.,
Part. ii.
pp.
442460
(contributed
by
G. H.
Darwin).
t
The
question
has been discussed from a different
standpoint by
J.
Larmor,
loc. cit.
p.
249.
J
Cf. Kelvin and
Tait,
Nat.
Phil,
Part ii.
pp.
298300. The
problem
of
plane
itress in a
circular
cylinder
was solved
by
Clebsch, ElasticitUt,
42.
L. E.
17
258 EXAMPLES OF SOLUTION
[CH.
XI
due to
given displacements
at the curved
surface,
when the tractions that
maintain these
displacements
are
adjusted
so that there is no
longitudinal
displacement.
When the tractions
applied
to the surface are
given,
we
may
take
SX,
2
Yn
to be the
components, parallel
to the axes of x and
y,
of the tractions
exerted afcross the surface r
=
a,
the functions
X, Yn
being again
of the form
On
cos 110
+^n
sin nd. We
write,
by analogy
to
(25),
dx
.(57)
and we introduce functions
'^_i
and
$__,
by
the
equations
d
^ --(x
-
dy
, 9 / aN 9 / a\
.(58)
All these functions are
plane
harmonics of the
degrees
indicated
by
the
suffixes. The surface tractions can be calculated from
equations (55).
We
find two
equations
of the
type
*-"
^^"^"a"
2(n-l)U
X
+
Syaj"^
dx\r-^-^)
from which we
get
^n-i
=
2ft dx
\
cC'
X
+
3/X
g
2k(X,
+
/x) fi
fjLa"
.(59)
^n
and thus
.4, 5
can be
expressed
in terms of
X, F.
As
examples
of this method we
may
take the
following*
:

(i) X=acos25, F
=
0. In this case we find
_a_f\
+
2^,2_
1
axy
(ii)
X
=
acos2^, y
=
asin25. In this case we find
(iii) X
=
a cos
45, l'n=0.
In this case we find
.(60)
.=
"^.
*
The solutions in these
special
cases will be UReful in a
subsequent investigation (Chap. XVI).
186,
187]
IN SERIES
259
187.
Applications
of curvilinear coordinates.
We
give
here some indications
concerning
various researches that have been made
by starting
from the
equations
of
equilibrium expressed
in terms of curvilinear co-
ordinates.
(a)
Polar coordinates. Lamp's
original
solution of the
problem
of the
sphere
and
spherical
shell
by
means of series was obtained
by using
the
equations expressed
in terms
of
polar
coordinates*. The same
equations
were afterwards
employed by
C. W.
Borchardtt,
who obtained a solution of the
problem
of the
sphere
in terms of definite
integrals,
and
by
C.
ChreeJ
who also extended the method to
problems relating
to
approximately
spherical boundaries!, obtaining
solutions in the form of series. The solutions in series
can be built
up by
means of solid
spherical
harmonics
( F) expressed
in terms of
polar
coordinates,
and related functions
(U)
which
satisfy equations
of the form V^U=
V.
(5) Cylindrical
coordinates. Solutions in series have been obtained
|| by observing
that,
if
Jn
is the
symbol
of Bessel's function of order
n,
e''^'^"^
J{kr)
is a .solution of
Laplace's equation.
It is not difficult to deduce suitable forms for the
displacements
Ur,
Wg,
Mj.
The case in which
Ug
vanishes and
u^
and
u,
are
independent
of 6 will
occupy
us
presently (Article 188).
In the case of
plane strain,
when
m,
vanishes and
u^
and
Ug
are
independent
of
2,
u.se
may
be made of the stress-function
(cf.
Article 144
supra).
The
general
form of this function
expressed
as a series
proceeding by
sines and cosines of
multiples
of 6 has been
given by
J. H. MichelllT.
(c)
Plane strain in non-circular
cylinders.
When the boundaries are curves of the
family
a
=
con8t.,
and a is the real
part
of a function of the
complex
variable
."c-^iy,
we
know from Article 144 that the dilatation A and the rotation nr are such functions of x
and
y
that
(X-l-2jii)
A
+
iS/icr
is a function of
x+iy,
and therefore also of a +
tfi,
where
j3
is
the function
conjugate
to a. For
example,
let the elastic solid medium be bounded
internally by
an
elliptic cylinder.
We take
a;-|-ty=ccosh(a-l-i0),
so that the curves a
=
const, are confocal
ellipses,
and 2c is the distance between the foci.
Then the
appropriate
forms of A and or are
given by
the
equation
(X -l-2/x)
A +
tS/xor
=
2e
"
"'"
(4
cos
n^
+
jB
sin
.^).
If we denote
by
h the absolute value of the
complex quantity
d
{a
+
i^)/d (x
+
iy),
then
the
displacements
u^
and u~ are connected with A and vj
by
the
equations
In the case of
elliptic cylinders
ujh
and
m/A
can be
expressed
as series in
cosn^
and
sinm/3
without much
difficulty.
*
J. de Math.
(Liouville),
t. 19
(1854).
See also
Le(;om
aur Us coordonnies
curvilignes,
Paris,
1859.
t
Berlin
Monatsberichte, 1873.
t Cambridge
Phil. Soc.
Trans.,
vol. 14
(1889).

Amer. J.
of Math.,
vol. 16
(1894).
II
L.
Pochhammer,
J.
f.
Math.
[Crelle],
Bd. 81
(1876), p. 33,
and C.
Chree, Cambridge
Phil.
Soc.
Trans.,
yol 14
(1889).
IT London Math. Soc.
Proc,
vol. 31
(1900), p.
100.
172
260 SYMMETRICAL STRAIN
[CH.
XI
As an
example*
we
may
take the case where an
elliptic cylinder
of semi-axes a and b is
turned about the line of centres of its normal sections
through
a small
angle (j).
In this
case it can be shown that the
displacement produced
outside the
cylinders
is
expressed by
the
equations
h
(d)
Solids
of
revolution. If
r, 6,
z are
cylindrical coordinates,
and we can find a
and
/3
as
conjugate
functions of z and r in such a
way
that an
equation
of the form
a
=
const,
represents
the meridian curve of the surface of a
body,
we transform
Laplace's
equation
V^F=0 to the form
da
where J denotes the absolute value of
d{z
+
ir)/d{a
+
tfi).
If we can find solutions of this
equation
in the cases where T"^ is
independent
of
0,
or is
proportional
to sin nd or cos
n6,
we can obtain
expressions
for the dilatation and the
components
of rotation as series.
Wangerint
has shown how from these solutions
expressions
for the
displacements
can
be deduced. The
appropriate
solutions of the above
equation
for V are known in the case
of a number of solids of
revolution, including ellipsoids,
cones and tores.
188.
Symmetrical
strain in a solid of revolution.
When a solid of revolution is strained
symmetrically,
so that the dis-
placement
is the same in all
planes through
the axis of
revolution,
we
may
express
all the
quantities
that occur in terms of a
single
function,
and reduce
the
equations
of
equilibrium
of the
body
strained
by
surface tractions
only
to a
single partial
differential
equation. Taking
r, 6,
z to be
cylindrical
coordinates,
we have the
stress-equations
of
equilibrium
in the forms
drr drz rr

66 . drz dzz rz
/,x
-^
+
-^
+
=0,

--f
-
+ -=0
(61)
or dz r dr oz r
'
Writing
V,
w for the
displacements
in the directions of r and
z,
and
sup-
posing
that there is no
displacement
at
right angles
to the axial
plane,
we
have the
expressions
for the
strain-components
"=
g;:,
eee
=
-,
^zz=^,
^rz
=
^
+
^,
ere=e,o
=
0-
...(02)
We
begin by putting, by analogy
with the
corresponding theory
of
plane strain,
drdz'
Then the second of
equations (61) gives
us
*
The
problem
was
proposed by
R. R. Webb. For a different method of
obtaining
the solution
see D.
Edwardes,
Quart.
J. of
Math.,
vol. 26
(1893), p.
270.
+ Archiv
f.
Math.
{Grunert),
vol. 5.5
(1873).
The
theory
has been
developed
further
by
P. Jaerisch, J.f.
Math.
(Crelle),
Bd. 104
(1889).
The solution for an
ellipsoid
of revolution with
given
surface
displacements
has been
expressed
in terms of series of
spheroidal
harmonics
by
0.
Tedone,
Rome Ace. Lincei Rend.
(Ser. 5),
t. 14
(1905).
187,
188]
IN A SOLID OF REVOLUTION
261
no
arbitrary
function of r need be added for
any
such function can l)e included in A.
We observe that
e=g-;(re^),
and write down the
equivalent
equation
in terms of stress-
components,
viz. :
rr-(T66-<TZz=^
{{66
-
<Trr-a-'^)r},
and hence we obtain the
equation
(1 -t-<r) {n-- e6)=r
^
(66-
a- f?-
<t7?).
We introduce a new function H
by
tlie
equation
and then the first of
equations (61)
can be written
and we
may put
66
=
(rV'^(t>-R,
,
where
v^ denotes
d^/dr^+r-^d/dr
+
d^/dz^
the
subjects
of
oiieration being independent
of 6.
No
arbitrary
function of z need be added because
any
such function can be included
in
(ji.
All the
stress-components
have now been
expressed
in terms of two functions
<l>
and
H. The sum 9 of the
principal
stresses is
expressed
in terms of
<j) by
the
equation
e='rr
+ 6e
+
zz
=
{l
+
tT)v^(j>,
and,
since is an harmonic
function,
we must have
v*^=0.
The functions
(p
and li are not
independent
of each other. To obtain the relations
between them we
may proceed
as follows:

The
equation
U/r=e^
can be written
U=r
{66
-art-
-
<rTz)IE,
or
U=-{\+a-)rRIE;
and then the
equation rz=fier,
can be written
dr~ E drdz'*' E
^
dz
'
Also the
equation e^^= {zz-a-rr-
a-
66)/
E can be written
-^-E~W-^ri--''^V-
The
equations giving dw/dr
and
dwjdz
are
compatible
if
and,
if we introduce a new function Q
by
means of the
equation
die
8^
3Q
wo have
S2Q
822'
where,
as
before,
no
arbitrary
function of z need be added.
g^=(l-..)V=<^,
262 SYMMETRICAL STRAIN
[CH.
XI
The
stress-components
are now
expressed
in terms of the functions
(j)
and Q which are
connected
by
the
equation
last written. The
equations giving dw/dr
and
dw/dz become,
when Q is
introduced,
dw_l+a
d
rda_d^\
dw
_
\
+
a- d /dil
B(>\
dr~
'E~dr\dz dij'
d^~ E
dz\dz
dzj'
We
may
therefore
express
U and w in terms of Q and
<^ by
the formulae
^,^
1+o-/8q,8.^
{dr'^drj'
'""

\_dz dzj
From these formulie we can show that Q must be an harmonic
function,
for we have at
the same time
dr r dz
-l"[v-^<^
+
V^Q-2g]
=
^-J^[(l-2<.)v^<A-V^Q],
and A
=
l^e=^(l+a)v20.
It follows
that,
besides
satisfying
the
equation d^Q/dz^
=
(1
-
<t) v^(t>,
the function Q also
satisfies the
equation
V2Q
=
0.
Instead of
using
the two functions
(p
and J2 we
may express
the
stress-components
in
terms of a
single
function. To this end we introduce a new function
\jf by
the
equation
ylf=<t>
+
Q. Then we have
'"''"822 '^r dr'^r 3r
~^
'''
df^ dz^ d^~'^''
'*'
dr^
'
and we have also
The first of
equations (61)
would enable us at once to
express
rz in terms of a function
x
such that
^d^/dz.
We therefore
drop
all the
subsidiary
functions and retain
x only.
In accordance with the above detailed work we assume
--S?{(2-<.)T.x-gf}
(63)
Then the first of
equations (61) gives
us
r.;^
=
-rc7|-V^;t
+
-?^
+
??--?r. (64)
dr
'^
dr^ or- r dr
'
and the second is satisfied
by
this value of rz if
V^X
=
(65)
The
stress-components
are now
expressed
in terms of a
single
function
x
which satisfies
equation (65)
*.
*
A method of
expressing
all the
quantities
in terms of a
single function,
which satisfies a
partial
differential
equation
of the fourth order different from
(65),
has been
given by
J. H.
Michell,
Londtm Math. Soc.
Proc,
vol. 31
(1900), pp.
144146.
188,
189]
IN A
CYLINDER
263
The
corresponding displacements
are
easily
found from the stress-strain
relations in the forms
(l-2*)V^X
+
9;?
+
|.|j}.-(66)
E drdz'
^
E
189.
Symmetrical
strain in a
cylinder.
When the
body
is a circular
cylinder
with
plane
ends at
right angles
to
its
axis,
the function
^
will have to
satisfy
conditions at a
cylindrical
surface
r
=
a,
and at two
plane
surfaces z
=
const. It must also
satisfy equation (65).
Solutions of this
equation
in terms of r and z can be found
by
various
methods.
The
equation
is satisfied
by any
solid zonal
harmonic,
i.e.
by any
function
9"
of the form
{r^
+
z')"*^
^-^
(r^
+
z-)~^
,
and also
by
the
product
of such a
oz
function and
(r^
+
z'^).
All these functions are rational
integral
functions of
r and
z,
which contain even
powers
of r
only. Any
sum of these functions
each
multiplied by
a constant is a
possible
form for
)(.
The
equation (65)
is satisfied also
by any
harmonic function of the form
e***
Jci(A'r),
where k is
any
constant,
real or
imaginary,
and
Jq{x)
stands for
Bessel's function of zero order. It is also satisfied
by any
function of the
'I
form e***r
5- Joikr),
for we have
or
V^
|e**^r ^
Jo{kr)\
= -
2k'e^'''J,{kr).
When k is
imaginary
we
may
write these solutions in the form
J(iKr) (A
cos KZ +
Bsin
kz)
+ t^
5- /o(^) {(^
cos kz
+
Dsin
kz),
. .
.(67)
in which k is real and
A, B, C,
B are real constants.
Any
sum of such
expressions,
with different values for
k,
and different constants
A, B, C, D,
is a
possible
form for
^.
The formulae for the
displacements
U,
w that would be found
by
each of
these methods
have
been obtained otherwise
by
C. Chree*.
They
have been
applied
to the
problem
of a
cylinder pressed
between two
planes,
which are in
contact with its
plane
ends,
by
L. N. G.
Filonf.
Of the solutions which are
rational and
integral
in r and
z,
he
keeps
those which could be obtained
by
the above method
by taking )(^
to contain no terms of
degree higher
than
the
seventh,
and to contain uneven
powers
of z
only.
Of the solutions that
could be obtained
by taking ;^
to be a series of terms of
type
(67),
he
keeps
*
Cambridge
Phil. Soc.
Trans.,
vol. 14
(1889), p.
250.
t
Phil. Trans.
Roy.
Soc.
{Set. A),
vol. 198
(1902).
J'ilon
gives
in the same
paper
the solutions
of other
problems relating
to
symmetrical
strain in a
cylinder.
264 CYLINDER UNDKR TERMINAL PRESSURE
[CH.
XI
those which result from
putting
k
=
nirjc,
where n is an
integer
and 2c is
the
length
of the
cylinder,
and omits the cosines. He finds that these
solutions are
sufficiently general
to admit of the satisfaction of the
following
conditions :
(i)
the
cylindrical boundary
r
=
a is free from traction
;
(ii)
the ends remain
plane,
or w
=
const, when z=
c\
(iii)
the ends do not
expand
at the
perimeter,
or {7
=
when r
=
a
and
z=c;
(iv)
the ends are
subjected
to a
given
resultant
pressure.
He shows also how a correction
may
be made
when,
instead of condition
(iii),
it is assumed that the ends
expand by
a
given
amount. The results are
applied
to the
explanation
of certain
discrepancies
in estimates of the
strength
of short
cylinders
to resist
crushing
loads,
the
discrepancies arising
from the
employment
of different kinds of tests
;
and
they
are
applied
also to
explain
the
observation
that,
when
cylinders (or spheres)
are
compressed
between
parallel
planes, pieces
of an
approximately
conical
shape
are sometimes cut out at the
parts subjected
to
pressure.
CHAPTER XII.
VIBRATIONS OF SPHERES AND CYLINDERS.
190. In this
Chapter
we .shall illustrate the method
explained
in
Article 126 for the solution of the
problem
of free vibrations of a solid
body.
The free vibrations of an
isotropic
elastic
sphere
have been worked
out in detail
by
various writers*. In
discussing
this
problem
we shall
use the method of Lamb and record some of his results.
When the motion of
every particle
of a
body
is
simple
harmonic and of
period lirfp,
the
displacement
is
expressed by
formulae of the
type
i<
=
J.m'cos
(pi
+
g),
v
=
Av'cos{pt-\-e),
w
=
Aw' cos
{pt
-{
e), ...(1)
in which
u', v',
w' are functions of
x,
y,
z,
and A is an
arbitrary
small
constant
expressing
the
amplitude
of the
vibratory
motion. When the
body
is
vibrating freely,
the
equations
of motion and
boundary
conditions
can be satisfied
only
if
p
is one of the roots of the
"
frequency equation,"
and
u', v',
w' are "normal functions." In
general
we shall
suppress
the
accents on
u', v',
w
,
and treat these
quantities
as
components
of
displace-
ment. At
any stage
we
may
restore the
amplitude-factor
A and the
time-factor cos
{pt
+
e)
so as to obtain
complete expressions
for the dis-
placements.
The
equations
of small motion of the
body
are
(^
+
^Ha^'
Ty'
a^J+^^("'
''^)
=
Ka-
al^'
d^)'
^^^
, . du dv dw
,,
^^'''
^=aS
+
a^
+
^
(^)
When
u, v,
w are
proportional
to cos
(pt
+
e)
we obtain the
equations
(\
+
/i)(g^,
g-, ^)+fiV^(u.
V, w)
+
pp-{u,
V, tv)
=
(4)
*
Reference
may
be made to P.
Jaerisch,
J.
f.
Math.
{Crelle),
Bd. 88
(1880);
H.
Lamb,
London
Math. Sac.
Proc,
vol. 13
(1882) ;
C.
Chree, Cambridge
Phil. Soc.
Trans.,
vol. 14
(1889).
266 SOLUTION OF THE
EQUATIONS
OF VIBRATION
[CH.
XII
Differentiating
the left-hand members of these
equations
with
respect
to
w,
y,
z
respectively,
and
adding
the
results,
we obtain an
equation
which
may
be written
(V2+;0^
=
o, (5)
where
h'
=
p'pl{X
+
2fi) (6)
Again,
if we write
ic'
=
p'pIij-,
(7)
equations (4)
take the form
We
may suppose
that A is determined so as to
satisfy equation (5),
then one
solution
(z/i, v^,
W])
of the
equations
last written is
^8A
aA
aA>
dy
and a more
complete
solution is obtained
by adding
to these values for
1*1, V], Wi,
complementary
solutions
{u, v^, w^)
of the
system
of
equations
(V^+k')u,
=
0,
(V^+k'')v,
=
0, iV-+K')W,=::0, (8)
9ttj dv<, dw^
,.
and
^
+
^5-
+
T~
=*^
v^)
dx
ay
02
When these functions are determined the
displacement
can be written in
the form
(u, V, w)
=
A
{u,-^
+
u.i, Vi
+
Vi, 'w^
+
w^cos{pt+ e) (10)
191. Solution
by
means of
spherical
harmonics.
A solution of the
equation (V^
+
A^)
A
=
can be obtained
by supposing
that A is of the form
f{r)Sn,
where r'=ii^
+
f
+
2-,
and
/S
is a
spherical
surface harmonic of
degree
n. We write
i?
instead of
/{r).
Then
rRn
is
a solution of Riccati's
equation
1 /8A aA
dA\
(;a;-^=-^~^^)(^^)=o-
of which the
complete primitive
is
expressible
in the form
rRn
=
r+'
[- g-J
^
,
An
and
5
being arbitrary
constants. The function
?'"/S
is a
spherical
solid
harmonic of
degree
n. When the
region
of
space
within which A is to be
determined contains the
origin,
so that the function A has no
singularities
in
the
neighbourhood
of the
origin,
we take for A the formula
A
=
Sa,t(^0.
(11)
190,
191]
IN TERMS OF SPHERICAL HARMONICS 267
where
&>
is a
spherical
solid harmonic of
positive degree
n,
the summation
refers to different values of
n,
and
<|^n(*')
is the function determined
by
the
equation
*.=es"(^')
c^)
The function
yj/^ (.v)
is
expressible
as a
power series,
viz. :
(-)"(
a;2 X* 1
'''"
^''^"l.3.5...(2,+'l) r~2(2
+
3)
"^
2.4.(2re
+
3)(2n
+
5)~-J
'
-^^'^
which is
convergent
for all finite values of x. It is an
"integral
function." It
may
be
expressed
in terms of a Bessel's function
by
the formula
>/'(^)
=
(-)"iv/(27r)x-<+y^j(^)
(14)
It satisfies the differential
equation
(-^-<-?-"i-)*-("=^
(-)
The functions
\/r {x)
for consecutive values of n are connected
by
the
equations
The function
* (x)
determined
by
the
equation
which has a
pole
of order 2ji
+
l at the
origin,
and is
expressible by
means of a Besael's
function of order

{n
+
^),
satisfies
equations (15)
and
(16).
In like manner solutions of
equations (8)
and
(9)
which are free from sin-
gularities
in the
neighbourhood
of the
origin
can be
expressed
in the forms
U.,=
U'n'fn('Cr), Vi=Vny}fn(lcr), W2=WnirniKr), (17)
where
Z7, F, TF
are
spherical
solid harmonics of
degree
n,
provided
that
these harmonics are so related that
9^
+
|
+
9|^
=
{9Us)
One
way
of
satisfying
this
equation
is to take
Z7, F,,,
Wn
to have the forms
where
Xn
is a
spherical
solid harmonic of
degree
n
;
for with these forms
we have
9|^
+
^
+
^
=
and
xU
+
yVn
+ zWn=0.
ox
dy
oz
A second
way
of
satisfying equation (9 Us)
results from the observation that
268
DISPLACEMENT IN A
[CH.
XII
curl
(ms,
fa, Wa)
satisfies the same
system
of
equations (8)
and
(9)
as
(u^,
v^,
w.^.
If we take
Mj', v^, w^
to be
given by
the
equations
we find such formulae as
where
\|r'(/tr)
means
dy]fn(icr)ld{K7-). By
means of the
identity
and the relations between
i/r
functions with consecutive
suffixes,
the above
formula is reduced to the
following
:

dy
dz''2n
+
l
"^"^
'
""^^
dx %i
+
1
^"+' ^"^ ''
" '
3*
Vr+V
'
of which each term is of the form
Un-^niicr).
In like manner the other
components
of curl
{u^',
v.2, w^)
can be formed.
Hence,
taking t^
and
j>n+i
to be
any
two solid harmonics of
degrees
indicated
by
their
suffixes,
we have solutions of the
equations (8)
and
(9)
in
such forms as
-^:^t.(-)v="l(^;)]
(20)
The
corresponding
forms of
v^
and
Wa
are obtained from this
by cyclical
interchange
of the letters
x,
y,
z.
192. Formation of the
boundary
conditions for a
vibrating
sphere.
We have now to
apply
this
analysis
to the
problem
of the free vibrations
of a solid
sphere.
For this
purpose
we must calculate the traction across
a
spherical
surface with its centre at the
origin.
The
components
Xr,
Yy, Zr
of this traction are
expressed,
as in Article
175,
by
formula of
the
type
rXr
^
A .
9 ,
, ,
,
,
3m
^5,,,
-
=
-
a;A
+
5-
(ux + w+
wz)
+
r
;r
u
(21 )
fi
ti,
ox^
^
or
In this formula A has the form
given
in
(11),
viz.:
S<Bi/r (/-),
and
w, v,
w
have such forms as
1
3A^
"
=
-^3^+^
^"^
-^
V^
dz
dy
ax
lfn..i'cr)..-^'l{^:)
.(22)
191,
192]
VIBRATING SPHERE
2C9
We find
T ? A
wa;
+
?;y
+ m
=
-
^ g-
+ 2
(n
+
1) {Vr (r)
+
/eVf+,
(/er)} (^+.
,
or
MX' +
1)?/
+
w^
=
2
-
p
{V^n(Ar)+An/r'(/tr)la)-(+l)(2n+3)t+.(/w)<^+,l
.
(23)
This formula
gives
us an
expression
for the radial
displacement
{ux +
z)y
+
wz)lr.
In
forming
the
typical
terms of
a;A,
^{mc
+
vy
+
wz),
r^
u we make
continual use of identities of the
type (19)
and of the
equations
satisfied
by
the
yfr
functions. We shall obtain in succession the contributions of the several
harmonic functions
a),
^, x^
to each of the above
expressions.
The function

contributes to a;A the terms
_L_^.(,,)|^^._^|(^))
(,
and the functions
<^, Xn
contribute
nothing
to a;A.
The function
<
contributes to d
(ux
+
vy
+
wz)/dx
the terms
-
1,
\ny}rn (hr)
+
hrf^' (hr)}
-^
-
1
{(.
+
1) Art' (hr)
+
h^r^J' (hr)}
^^^
{^
-
^-
^
(^.)}
.
which reduce to
-{+.+-J^+.'}^i(^-)
(2a,
The function
<^
contributes to 3
(ux
+
vi/
+
wz)/dx
the terms
-
n
{{2n
+
1) >/r (r)
+
Krf' (r)}
^
+
nVr' (r)
r+
|; f^,)
. . .
.(26)
The function
^
contributes
nothing
to this
expression.
The function
lOn
contributes to u the terms
-
-h^
[{^"
(^')
+
2-Vi
'^^"'
('''>}
^
-
(2-^)
'^^' <'^^
^"
8^
(;^0]
(27)
270 DISPLACEMENT IN A
[CH.
XII
and it contributes to r= u the terms
or
fn
(hr)
+
^^
hrf.'
(kr)]
^
-
^-^
hr^.' (hr)
r-
^
{^)
1-2
which reduce to
_
1
h;-
('^
-
2)
-
2-^1
^ (^^)
+
r^
^^^' ('">
9a)
dx
-{t(A.).^V'(^2^
a^^C-:^)
(28)
The function
<f>n
contributes to
rdu/dr

u the terms
{(n
-
2)
t-i
(-)
+
'-^'-i (^)j
^
-^MV'+:(0
+
rt;+.(r)lr-+^(^,)
(29)
The function
Xn
contributes to the same
expression
the terms
{{n-l)ir^{Kr)
+
Krir'{^r)}{y^f^-z^-^)
(30)
Complete expressions
for the tractions
X^, Y^, Z^
can now be written
down in accordance with
(21),
and we
may express
the conditions that these
tractions vanish at the surface of a
sphere
r
=
a in forms of which the
type
is
P"
^
'dz^
dy)^
"
dx
^
^"'^
dx
V+'j
dx dx
\r+V
=
0,
.(31)
where
pn,
ctn, bn, Cn, dn
are constants. The values of these constants can be
found from the above
analysis.
When we write
K^jh'^

2 for
X/yti,
and use the
equations
satisfied
by
the
-^
functions,
we find the
following expressions
for
the constants
Pn
=
(W
-
1) v/^n {kO)
+
Ka^n ().
1
""
~
(2W+ 1)
A
i'^'''''*'^" ('*'*)
+
2
(n
-
1) ^|r_l {ha)],
1
(' , /J, N ,
2(n
+
2)
, ,,j
^
Cn
=
K'd'-'^n {ko)
+
2
(n
-
1) y^n-i {ica},
n {
, , X 2(n
+
2) , ,,
,1
.(32)
192-194]
VIBRATING SPHERE
271
There are two additional
equations
of the
type (31)
which are to be obtained
from the one written down
by cyclical
interchange
of the letters
x,
y,
z.
These
equations
hold at the surface r
=
a.
193.
Incompressible
material.
In the case of
incompressible
material we have to take A
=
and to
replace
\^
by
-0,
where n denotes a finite
pressure.
The
equations
of motion become three of the
tyixj
8n
,
ZH
in which
du/da;+dv/di/+dw/d!=0.
We find at once that n must be an harmonic
function,
and we
may put
n=
-fiS(a,
in which
a>
is a
spherical
solid harmonic of
degree
n. When
u, v,
w are
simple
harmonic
functions of t with
period 2jr/p,
the
equations
of motion become three
equations
of the
type
(v2+ic2)M-^i-i3n/ax=0,
and the
integrals
can be found in such forms as
where
u^
is
given by (20).
The formula for
rX^/fi
now liecomes
and the terms contributed to the
right-hand
member
by
are
2(?t- l)\
dm^ _ ^3 ^
/
a),
\
k2
J
dx 2re
+
l8a;V^
+
V' \2-t-l
while the terms contributed
by (^
and
Xn
are the same as before. The result of
assuming
((2 2
(
jt

1
)
incompressibility
of the material is therefore to
change
a
into
^^
^
-
and
6
into

r
,
without
altering
the
remaining
coefficients in the left-hand member of
(31).
194.
Frequency equations
for
vibrating sphere.
The left-hand members of the
equations
of
type (31)
are sums of
spherical
solid harmonics of
positive degrees,
and
they
vanish at the
surface r
=
a. It follows that
they
vanish
everywhere.
If we differentiate
the left-hand members of these
equations
with
respect
to
uc,
y,
z
respectively
and add the results we obtain the
equation
6co
-I-
c^,j<^
=
(33)
If we
multiply
the left-hand members of the
equations
of
type (31) by
X,
y,
z
respectively
and add the
results,
we
find,
after
simplification by
means
of
(33),
the
equation
ata-f
c<^
=
(34)
272 MODES OF VIBRATION
[CH.
XII
The
equations
of
type (31)
then show that we must have
^
(''#-'?)-.
i'-('|--4-)=''-
""(^l^-^^-)-"-
It follows that the vibrations fall into two classes. In the first class
&)
and
<f>n
vanish and the
frequency
is
given by
the
equation
Pn
=
(35)
where
j9
is
given by
the first of
(32).
In the second class
^n
vanishes
and the
frequency
is
given by
the
equation
adn- bnCn
=
(36)
where
a, 6, c, d
are
given by (32).
In the vibrations of this class
&>
and
<f>n
are connected with each other
by
the
compatible equations (33)
and
(34).
195. Vibrations of the first class*.
When the vibration is of the first class the
displacement
is of the form
(.,.,.)=4cos(p.+o^^(<.)(/f-|^
^l"-!-,
-|"-/i")-
-(37)
where
K^=p^plfi;
and the
possible
values of
p
are determined
by
the
equation
(?i-l)i^((ta)
+
(caV'(ica)=0
(38)
The dilatation vanishes. The radial
displacement
also
vanishes,
so that the
displacement
at
any point
is directed at
right angles
to the radius drawn from the centre of the
sphere.
It is also directed at
right angles
to the normal to that surface of the
family ;^=
const,
which
passes through
the
point.
The
spherical
surfaces determined
by
the
equation
^(Kr)=Oare
"nodal,"
that is to
say
the
displacement
vanishes at these surfaces. The
spherical
surfaces determined
by
the
equation
in which k is a root of
(38),
are
"anti-nodal,"
that is to
say
there is no traction across
these surfaces. If
kj, kj,
... are the values of k in
ascending
order which
satisfy (38),
the
anti-nodal surfaces
corresponding
with the vibration of
frequency (2rr)"'^(;i/p) k,
have
radii
equal
to
K^aJK,, K^aJK,,
...
ic,_ia/K,.
If n

\ we have
rotatory
vibrational.
Taking
the axis of z to be the axis of the
harmonic
xii
^^^
displacement
is
{u, V, w)
=
A cos
(pt
+
e) ^l^iiKr) {}/, -X, 0),
so that
every spherical
surface concentric with the
boimdary
turns round the axis of z
through
a small
angle proportional
to
i/^j ((cr),
or to
{kt)''^
cos Kr
-
{kt)-^
sin kt. The
possible
values of k are the roots of the
equation i/^j'(Ka)
=
0,
or
tan
Ka=3Ka/(3

kV).
*
The results stated in this Article and the
following
are due to H.
Lamb,
loc. clt.
p.
265.
+ Modes of vibration
analogous
to the
rotatory
vibrations of the
sphere
have been found
f^r any
solid of revolution
by
P.
Jaerisch, J.f.
Math.
{Crelle),
Bd. 104
(1889).
194196]
OF A SPHERE 273
The lowest roots of this
equation
are

=
1-8346, 2-8950, 3-9225, 4*9385, 5-9489, 6-9563,....
The number
tt/km
is the ratio of the
period
of oscillation to the time taken
by
a wave of
distortion-* to travel over a distjince
equal
to the diameter of the
sphere.
The nodal
surfaces are
given by
the
equation
tan (tr
=
icr,
of which the roots are
-=1-4303, 2-4590, 3-4709, 4-4774, 5-4818, 6-4844,....
TT
196. Vibrations of the second class.
When the vibration is of the second class the
components
of
displacement
are
expressed
by equations
of the
type
The
frequency equation (36)
cannot be solved
numerically
until the ratio
KJh
is known.
We shall consider
chiefly incompressible material,
for which
A/ic=0,
and material ful-
filling
Poisson's condition
(X^/i),
for which
k/A
=
v'3.
Radial vibrations.
Wlien n=0 we have radial vibrations. The normal functions are of the form
=^V'o'(M> =f
V'o'(A'-),
w
=
^,^o'(^'-)
(40)
and the
frequency equation
is
b^
=
0,
or
ylro{'ta)
+
^^haf'
{ha)
=
0,
(411
which is
. tan ha 1
~^a~~l-i(K2/A2)A2a2"
There
are,
of
course,
no radial vibrations when the material is
incompressible.
When
K^/h'
=
:i,
the six lowest roots of the
frequency equation
are
given by

=
-8160, 1-9285, 2-9359, 3-9658, 4-9728,
5-9774.
TT
The number
-ir/ha
is the ratio of the
period
of oscillation to the time taken
by
a wave of
dilatation t to travel over a distance
equal
to the diameter of the
sphere.
Spheroidal
vibrations.
When n
=
2 and
a.^
and
cjy^
are zonal harmonics we have what
may
be called
spheroidal
vibrations,
in which the
sphere
is distorted into an
ellipsoid
of revolution
becoming
alternately prolate
and oblate
according
to the
phase
of the motion. Vibrations of this
type
would tend to be forced
by
forces of
appropriate period
and of the same
type
as tidal
disturbing
forces. It is found that the lowest root of the
frequency equation
for free
vibrations of this
type
is
given by Kajtr
=
-848 when the material is
incompressible,
and
by
ko/tt
=
-840 when the material fulfils Poisson's condition.

For a
sphere
of the same size
and mass as the
Earth, supposed
to be
incompressible
and as
rigid
as
steel,
the
period
of
the
gravest
free vibration of the
ty{)e
here described is about 66 minutes.
*
The
velocity
of waves of distortion is
(tijp)
. See
Chapter
XIII.
t
The
velocity
of waves of dilatation is
^/{(^
+
2/n)/p}.
See
Chapter
XIII.
L. E.
18
274 RADIAL VIBRATIONS OF A SPHERICAL SHELL
[CH.
XII
197. Further
investigations
on the vibrations of
spheres.
The vibrations of a
sphere
that would be forced
by
surface tractions
proportional
to
simple
harmonic functions of the time have been
investigated by
Chree*. Free vibrations of a
shell bounded
by
concentric
spherical
surfaces have been discussed
by Lambt,
with
special
reference to the case in which the shell is thin. The influence of
gravity
on the free
vibrations of an
incompressible sphere
has been considered
by Bromwich}.
He
found,
in
particular,
that the
period
of the
"spheroidal"
vibrations of a
sphere
of the same size
and mass as the Earth and as
rigid
as steel would be diminished from 66 to 55 minutes
by
the mutual
gravitation
of the
parts
of the
sphere.
A more
general
discussion of the
effects of
gravitation
in a
sphere
of which the material is not
incompressible
has been
given by
Jeans
.
198. Radial vibrations of a hollow
sphere||.
The radial vibrations of a
sphere
or a
spherical
shell
may
be
investigated very simply
in terms of
polar
coordinates. In the notation of Article 98 we should find that the radial
displacement
U satisfies the
equation
dr'^ r dr r^
'
and that the radial traction rr across a
sphere
of radius r is
(X
+
2^)
1^+2X^^=0.
The
primitive
of the differential
equation
for fJ
may
be written
_
d /A sin
hr+B
cos hr\
dip, \
kr
)
'
and the condition that the traction rr vanishes at a
spherical
surface of radius r is
[(X
+
2/i) {(2
-
^V2)
sin hr
-
2ArjCos hr)
+
2X
{hr
cos hr
-
sin
hr)]
A
+
[(X
+
2^i) {(2
-
AV)
cos A?-
+
2Ar sin
Ar}
-
2X
(Ar
sin Ar
+
cos
Ar)]
5
=
0.
When the
sphere
is
complete up
to the centre we must
put Z?=0,
and the condition for the
vanishing
of the traction at r
=
a is the
frequency equation
which we found before. In
the case of a
spherical
shell the
frequency equation
is found
by eliminating
the ratio
A : B from the conditions which
express
the
vanishing
of ?v at r
=
a and at r
=
b. We
write
so that
2X/(X
+
2/i)=2-i',
and then the
equation
is
i'Aa+(AV- v)tan
Aa_
vkb
+
ih^}^ -v)ta,x\hb
{h^a^

v)

vha tan ha
(Ji?}?

v)

vhh tan hh
'
In the
particular
case of a
very
thin
spherical
shell this
equation may
be
replaced by
9 vAa
+fAW

i/)tan
Aa_
da
(A
V
-v)

vha tan ha
'
which is
A2a2 sec2 ha
{h'^a^
-
^
(.3
-
v)]
=
0,
*
Loc. cit.
p.
265.
t
London Math. Soc.
Proc,
vol. 14
(1883).
+
London Math. Soc.
Proc,
vol. 30
(1899).

Phil. Trans.
Roy.
Soc.
(Ser. A),
vol. 201
(1903).
II
The
problem
of the radial vibrations of a solid
sphere
was one of those discussed
by
Poisson
in his memoir of 1828. See
Introduction,
footnote 36.
197-199]
VIBRATIONS OF A CIRCULAR CYLINDER
and we have therefore
ha
=
^{v{3-v)}.
In terms of Poisson's ratio <r the
period
is
275
VVi+J'
199. Vibrations of a circular
cylinder.
We shall
investigate
certain modes of vibration of an
isotropic
circular
cylinder,
the curved surface of which is free from
traction,
on the
assumption
that,
if the axis of z coincides with the axis of the
cylinder,
the
displace-
ment is a
simple
harmonic function of z as well as of i*. Vibrations of these
types
would
result,
in an unlimited
cylinder,
from the
supeiposition
of two
trains of waves
travelling along
the
cylinder
in
opposite
directions. When
the
cylinder
is of finite
length
the
frequency
of free vibration would be
determined
by
the conditions that the
plane
ends are free from traction.
We shall find
that,
in
general,
these conditions are not satisfied
exactly by
modes of vibration of the kind
described,
but
that,
when the radius of the
cylinder
is small
compared
with its
length, they
are satisfied
approximately.
We use the
equations
of vibration referred to
cylindrical
coordinates
r, 6,
z. The
equations
are
=
(^
+
2/^)|^'
2jU 8otj
P^=(^
+
2/.)-^^--2^^
+
2M-g^ dt' r dd dz
.(42)
in which
and
2ct,=
r dr r dd dz
'
.(43)
1
dn^
r dd
dug dur du^
1
(d{rui)
duA
,.,.
so that
sTr, tn-a, sJ'z
satisfy
the identical relation
1 5
(rCTy)
1
dog disTi
_
r dr r dd dz
.(45)
The
stress-components
rr, rd,
rz vanish at the surface of the
cylinder
r
=
a. These
stress-components
are
expressed by
the formulae
dUr -^ fl dur
d
iug\]
^^
fdur
du.
rr
=
XA
+
2/tt
dr
!(?))
"-"'+&)
<>
*
The
theory
is
effectively
due to L.
Poohhammer,
/.
/.
Math.
(Crelle),
Bd. 81
(1876), p.
324.
It has been discussed also
by
C.
Chree,
loc. cit.
p.
265.
182
276 TORSIONAL AND LONGITUDINAL
[OH.
XII
In accordance with what has been said above we shall take
Ur, its, i<z
to
be of the forms
M^=fre'ir^+p, Mg=7e'i>'^+^, Mj= TTe'i^^+P", (47)
in which
U, V,
W are functions of
r,
Q.
200. Torsional vibrations.
We can obtain a solution in which U and W vanish and V is
independent
of 6. The first and third of
equations (42)
are satisfied
identically,
and the
second of these
equations
becomes
3-^-
+
-^ ,F+'n^=0,
(48)
where K''
=
p^pl/i

'f.
Hence V is of the form
BJi{K'r),
where iJ is a
constant,
and
Ji
denotes Bessel's function of order
unity.
The conditions
at the surface r
=
a are satisfied if k is a root of the
equation
da\
a
]
'
One solution of the
equation
is k'=
0,
and the
corresponding
form of V
given
by equation (48)
is V=
Br,
where is a constant.
We have therefore found a
simple
harmonic wave-motion of the
type
Ur
=
0, Ue
=
Bre'^y'+P^\ u,= 0, (49)
in which
7'
=
p''pjfi-
Such waves are waves of
torsion,
and
they
are
propa-
gated
along
the
cylinder
with
velocity ^(/x/p)*.
The traction across a normal section z
=
const, vanishes if
dug/dz
vanishes
;
and we can
have, therefore,
free torsional vibrations of a circular
cylinder
of
length
I,
in which the
displacement
is
expressed by
the formula
cos^-p V 7'
n
being any integer,
and the
origin being
at one end.
Me
mrz
=
cos

,
-
.
r I
\
t
'\/
p
201.
Longitudinal
vibrations.
We can obtain a solution in which V vanishes and U and W are
independent
of 6. The second of
equations (42)
is then satisfied
identically,
and from the first two of these
equations
we find
or' r dr
d'isg
1
9i3-e We ,
o^
+
-
^ ^
-f K
-CTe
=
0,
01'^ r or r'
where h'^
=
p'pl{X+ 2fi)-y\
K''
=
p'pliJi- y^ (52)
^
*
Cf. Lord
Rayleigh, Theory of
Sound, Chapter
VII.
.(51)
199-201]
VIBRATIONS OF A CIRCULAR
CYLINDER 277
We must therefore take A and
isr^,
as functions of
r,
to be
proportional
to
/o
(/i''')
and
Jx (k'v).
Then to
satisfy
the
equations
we have to take U and W to be of the forms
C/
=
J.^y(AV)
+
6Wi(V),
W
=
AiyJ,{h'r)
+
iC d
r dr
.(o3)
{rJ,(/cV)},
where A and G are constants.
The traction across the
cylindrical
surface r
=
a vanishes if A and C are
connected
by
the
equations
2/i
d'Jo (h'a)
p-p\
tia^
X+-1/J.
Jo
{h'a)
+
2f^Cy^-^^=0,
2Ay
dJ,{h'a)
+
(7(27^-^''')
/,('aWO
.(54)
da
\
'
/J,
On
eliminating
the ratio i4 : C we obtain the
frequency equation.
When the radius of the
cylinder
is small we
may approximate
to the
frequency by expanding
the Bessel's functions in series. On
putting
Jo
{h'a)
=
1
-
ih'^a'
+
^\h'*a\
/,
{x'a)
=
/c'a
-
^K^a^
the
frequency equation
becomes
^-2y).a(l-^
h''0-la^-h'^)^l/^(l-la^h'')
+
27V (1
-
f aV)
ah''
(1
-
ia^h'^)
=
0.
It is
easily
seen that no wave-motion of the
type
in
question
can be found
by
putting
k'
=
0.
Omitting
the factor x'a and the terms of order
a",
we find a
first
approximation
to the value of
p
in terms of
7
in the form
p
=
y^{Elp), (55)
where
E,
=
fi (3X
+
2fji,)l{X
+
fi),
is
Young's
modulus. The waves thus found
are
"
longitudinal
"
and the
velocity
with which
they
are
propagated along
the
cylinder
is
^/{Ejp) approximately*.
When we retain terms in
a^,
we find a second
approximation
-f"
to the
velocity
in the form
p
=
y^{E/p){l-\aY-a'') (56)
where
o-,
=^X/{X
+
p,),
is Poisson's ratio.
*
Cf. Lord
Rayleigh, Theory of Sound, Chapter
VII.
+ The result is due to L.
Pochhammer,
loc. cit.
p.
275. It was found
independently by
C.
Chree, Quart.
J.
of
Math.,
vol. 21
(1886),
and extended
by him, Quart.
J.
of
Math.,
vol. 24
(1890),
to oases in which the normal section of the
cylinder
is not circular and the
material is not
isotropic
;
in these cases the term
J<r-7^(i^
of the above
expression (56)
is
replaced
by ^<T-y^K-,
where k is the radius of
gyration
of the
cylinder
about the liue of centres of the
normal sections.
278 LONGITUDINAL AND FLEXURAL
[CH.
XII
When the
cylinder
is terminated
by
two
plane
sections z
=
and z
=
l,
and these sections are free from
traction,
zz and zr must vanish at ^;
=
and
at z
=
l. We find for the values of zz and zr at
any
section the
expressions
zz
=
zr
=
/it
dJ,
(K'r)
J,
{k't)
^
{P'P
xhn^
H
'^^^'''^
+
''^^^
f
^
+
gi(yj+p()
'2Ay^-^^^
+
c[2y^-^)j,Wr
/*
gilyz+pt)^
Now we can have a solution of the form
mrz
Ur
=
.
dJo (h'r)
WTT
r / / X
sin
-J-
cos
(pnt
+
e),
mr
,
^.^(AV)
+
Cf^V^)'
WTT^
COS-y-
C08{pnt
+
e),
...(57)
in which the ratio
An
:
(7
is known from the conditions which hold at r
=
a,
7
has been
replaced by mr/l,
and
pn
is
approximately equal
to
(mr/l) ^(E/p)
when a is small
compared
with I. This solution satisfies the condition
zz
=
at z
=
and at
=
i,
but it does not
satisfy
the condition zr
=
at
these surfaces.
Since, however,
zr
==
at the surface r
=
a for all values
of
z,
the traction zr is
very
small at all
points
on the terminal sections
z
=
and z
=
I when a is small
compared
with I.
If we take
m,
to contain
cos(n7r2/0,
and
Ug
to contain

sin(?i7r2^/^),
the
other factors
being
the same as
before,
we have a solution of the
problem
of
longitudinal
vibrations in a
cylinder
of which the centres of both ends are
fixed.
202. Transverse vibrations.
Another
interesting
solution of
equations (42)
can be obtained
by taking
Ur
and
Ui
to be
proportional
to cos
d,
and
u^
to be
proportional
to sin 0.
-'Modifying
the notation of
(47)
in Article
199,
we
may
write
Ur=U
cos e
e'(r^+P', Ue=V
sin
e'^'+P", u^=W
cos 8
e'lr^+J"), (.58)
where
U, V,
W are functions of r. Then we have
A= cos^e'<i'^+^"
(
+
-
+
\dr
r
^+cyW),
"
2w,
=
-
sin ^ e'<T'+P"
W
+
iyV
2^9
=
cos e &
'v^+i'"
{oyU-^j,
2ot,= 8in^e"T*+J">
/dV
.
V
[dr
r r
(59)
201,
202]
VIBRATIONS OF A
CIRCUI-AR CYLINDER 279
From
equations (42)
we
may
form the
equation
d'A 1 8A A
, ^
o^
+
-5
r + AM
=
(60)
where h'^ is
given by
the first of
equations (52);
and it follows that A
can be written in the form
A
=
--^'^
AJ,{h'r)
cos
de'^y'+p^^
(61)
where J. is a constant.
Again,
we
may
form the
equation
which,
in virtue of
(45),
is the same as
a^OTj
1
aiCTj 07, ,
-^-T
+
"
~^
j
+
-i!r2
=
0,
(02)
oH r or v
where '" is
given by
the second of
equations (52).
It follows that
2t!rj
can
be written in the form
2w,
=
K'''CJ,iK'r)
sin
ee^y'+P^\
(63)
where C is a constant.
We
may
form also the
equation
p^p
_
sTr
2 19/
9sre\
d
dw,
fj.
'^
r"
'
r^
dr\
d6
j
dr dz
which,
in virtue of
(45),
is the same as
7^l^^lM-5+"'"^+^^'^^=^
^^*>
In this
equation
2w,
has the value
given
in
(63),
and it follows that
2xb-,.
can
be written in the form
2^,=
|,^(7?^)
+
,5^'^"^>|sin^e"r^+^"
(65)
where B is a constant. The
equations connecting
the
quantities
U, V,
W
with
A, w,, CTj
can then be satisfied
by putting
dr dr r
'
r r dr
'
Tf
=
tJ
7
J,
(h'r)
-
iBk'
J, (K'r).
When these forms for
U, V,
W are substituted in
(58)
we have a solution
of
equations (42).
Since
m,
sin ^
+
,
cos ^ vanishes when r
=
0,
the motion
.(66)
280 FLEXURAL VIBRATIONS OF A CYLINDER
[CH.
XII
of
points
on the axis of the
cylinder
takes
place
in the
plane containing
the
unstrained
position
of that axis and the line from which 6 is measured
; and,
since
Uz
vanishes when r
=
0,
the motion of these
points
is at
right angles
to
the axis of the
cylinder.
Hence the vibrations are of a
"
transverse
"
or
"
flexural
"
type.
We could form the conditions that the
cylindrical
surface is free from
traction. These conditions are
very complicated,
but it
may
be shown
by
expanding
the Bessel's functions in series that, when the radius a of the
cylinder
is
very
small,
the
quantities ,p
and
7
are connected
by
the
approximate equation*
f
=
laY{E/p) (67)
where E is
Young's
modulus. This is the well-known
equation
for the
frequency p/27r
of flexural waves of
length 27r/7 travelling along
a
cylindrical
bar. The ratios of the constants
A, B,
G which
correspond
with
any
value of
7
are determined
by
the conditions at the
cylindrical
surface.
When the
cylinder
is terminated
by
two normal sections z
=
and z
=
l,
we write
mjl
for the real
positive
fourth root of
'^p^pjo'E.
We can obtain
four forms of solution
by substituting
for
47
in
(52), (58), (66)
the four
quantities
+
mjl
and +
imjl successively.
With the same value of
p
we
should have four sets of constants
A, B, C,
but the ratios A : B : C in each
set would be known. The conditions that the
stress-components
zz,
z6
vanish at the ends of the
cylinder
would
yield
sufficient
equations
to
enable us to eliminate the constants of the
types
A, B,
C and obtain an
equation
for
p.
The condition that the
stress-component
zr vanishes at
the ends cannot be satisfied
exactly
; but,
as in the
problem
of
longitudinal
vibrations,
it is satisfied
approximately
when the
cylinder
is thin.
*
Cf. Lord
Eayleigh, Theory of Sound, Chapter
VIII.
CHAPTER XIII.
THE PROPAGATION OF WAVES IN ELASTIC SOLID MEDIA.
203. The solution of the
equations
of free vibration of a
body
of
given
form can be
adapted
to
satisfy any given
initial
conditions,
when the
frequency equation
has been solved and the normal functions determined
;
but the account that would in this
way
be
given
of the motion that ensues
upon
some local disturbance
originated
within a
body,
all
points (or
some
points)
of the
boundary being
at considerable distances from the
initially
disturbed
portion,
would be difficult to
interpret.
In the
beginning
of the
motion the
parts
of the
body
that are near to the
boundary
are not
disturbed,
and the motion is the same as it would be if the
body
were of unlimited
extent. We
accordingly
consider such states of small motion in an elastic
solid
medium, extending indefinitely
in all
(or
in
some)
directions,
as are at
some time restricted to a limited
portion
of the
medium,
the remainder of
the medium
being
at rest in the unstressed state. We
begin
with the case
of an
isotropic
medium.
204. Waves of dilatation and waves of distortion.
The
equations
of motion of the medium
may
be written
,/SA
aA 8A\
, , , (dHi
dH
d-w\
If we differentiate the left-hand and
right-hand
members of these three
equations
with
respect
to
x,
y,
z
respectively
and add the
results,
we obtain
the
equation
(\+2^)V^A
=
p^
(2)
If we eliminate A from the
equations
(1) by performing
the
operation
curl
upon
the left-hand and
right-hand
members we obtain the
equations
ti^''Wx,'^y,'^^
=
P^^{'^x,'^y,-^zl
(3)
282 PROPAGATION OF WAVES THROUGH
[CH.
XIII
If A vanishes the
equations
of motion become
/iV2(M,
V,
w)
=
p^Ju,
V,
w) (4)
If
BTj;,
zTy,
1^2 vanish,
so that
(u,
v,
w)
is the
gradient
of a
potential <^,
we
may
put V'<^
for
A,
and then we have
/aA aA aA\
,a^'
dy'
a. 7
=
^^
("'"')
In this case the
equations
of motion become
''af
(\+2/t)V(,t;,w)
=
p^,(M,
?;,
w) (5)
Equations (2), (3), (4), (5)
are of the form
^-'*^
=)
for
A,
c^ has the vahie
(X
+
2/i)/p
;
for
ot^;,...
it has the value
filp.
The
equation (6)
will be called the
"
characteristic
equation."
If is a function of t and of one coordinate
only, say
of
x,
the
equation (6)
becomes
which
may
be
integrated
in the form
^=f{x-ct)
+
F{x+ct),
f
and F
denoting arbitrary functions,
and the solution
represents plane
waves
propagated
with
velocity
c. If
(/>
is a function of t and r
only,
r
denoting
the radius vector from a
fixed
point,
the
equation
takes the form
8^
=
7 p
('*).
which can be
integrated
in the form
/(r-cQ F{r
+
ct)

r r
'
and
again
the solution
represents
waves
propagated
with
velocity
c. A function of the
form
r~'^f{r-ct) represents spherical
waves
diverging
from a source at the
origin
of r.
We learn that waves of dilatation
involving
no rotation travel
through
the
medium with
velocity {(X
+
2yii)//3)i,
and that waves of distortion
involving
rotation without dilatation travel with
velocity [jj-jp]^.
Waves of these two
types
are sometimes described as
"
irrotational
"
and
"
equivoluminal
"
respectively
*.
*
Lord
Kelvin,
Fhil.
Mag. (Ser. 5),
vol. 47
(1899).
The result that in an
iaotropio
solid there
are two
types
of waves
propagated
with different velocities is due to Poisson. The
recognition
of
the irrotational and
equivoluminal
characters of the two
types
of waves is due to Stokes. See
Introduction.
204,
205]
AN ISOTROPIC SOLID MEDIUM 283
If
plane
waves of
any type
are
propagated through
the medium with
any
velocity
c we
may
take
u, v,
w to be functions of
Ix
+
mij
+
nz
+
ct,
in which
I, m,
n are the direction cosines of the normal to the
plane
of the
waves. The
equations
of motion then
give
rise to three
equations
of
the
type
pc"-u"
=
{\
+
fM)l (lu"
+ mv"
+
nw")
+
/j,{l-
+
m'
+
n?) u",
where the accents denote differentiation of the functions with
respect
to their
argument.
On elimination of
u", v",
w" we obtain an
equation
for
c,
viz.:
(X
+
2^-pcO(/t-pcT
=
0, (7)
showing
that all
plane
waves travel with one or other of the velocities found
above.
205. Motion of a surface of
discontinuity.
Kinematical conditions.
If an
arbitrary
small disturbance is
originated
within a restricted
portion
of an elastic solid
medium,
neighbouring portions
will soon be set in motion
and thrown into states of strain. The
portion
of the medium which is dis-
turbed at a
subsequent
instant will not be the same as that which was disturbed
initially.
We
may suppose
that the disturbed
portion
at
any
instant is
bounded
by
a surface S. If the medium is
isotropic,
and the
propagated
disturbance involves dilatation without
rotation,
we
may expect
that the
surface >S' will move
normally
to itself with
velocity {(X
+
2/x.)/p]i;
if it involves
rotation without
dilatation,
we
may expect
the
velocity
of the surface to be
{fj'/p}^.
We assume that the surface moves
normally
to itself with
velocity
c,
and seek the conditions that must be satisfied at the
moving
surface.
On one side of the surface S at time t the medium is disturbed so that
there is
displacement (u,
v,
w) ;
on the other side there is no
displacement.
We take the
velocity
c to be directed from the first side towards the
second,
so that the disturbance
spreads
into
parts
of the medium which
previously
were undisturbed. The
displacement
(u,
v,
w)
is
necessarily
continuous in
crossing
8,
and it therefore vanishes at this
moving
surface. Let the normal
to S in the direction in which c is estimated be denoted
by
v;
and let s
denote
any
direction in the
tangent plane
at a
point
of
S,
so that s and v are
at
right angles
to each other. Since u vanishes at
every point
of
S,
the
equation
du
,
. du till
/ \ r,
;r- COS
(X, S)
+ TT- cos
( W, s)
+
^
COS
(Z, S)
=
dx
^ '
dy
^^
dz
holds for all directions s which
satisfy
the
equation
cos
{x, s)
cos
{x, v)
+
cos
(y, s)
cos
(y, v)
+
cos
{z, s)
co.s
{z, v)
=
0.
284 PROPAGATION OF WAVES THROUGH
[CH.
XIII
It follows
that,
at all
points
of
8,
dujdx
_
du/dy
_
du/dz
_du
,
cos
(a;, I/) cos{y,i') cos{z,v)
dv
Again
m
=
is an
equation
which holds at the
moving
surface
S,
and this
equation
must be satisfied to the first order in Bt when for
x,
y,
z,
t we
substitute
a; + c cos
{x, v)
Bt,
y
+ c cos
(y, v)
Bt,
z
+
c cos
{z, v)
Bt,
t
+
Bt.
It follows that at
every point
of N we must have
g^
+ c
jcos
{x,
v):^
+
cos(y,v)^
+
cos
(z, i^)
g^l
=
(9)
On
combining
the
equations (8)
and
(9)
we find that the
following
equations
must hold at all
points
of S:

du/dx
_
du/dy
_
du/dz
_du _
Idu
cos
{x, v)
cos
{y, !>)
cos
{z, v)
dv c dt
Exactly
similar
equations
hold with v and w in
place
of u. In these
equations
the differential coefficients of
m,
...
are,
of
course,
to be calculated
from the
expressions
for
m,
. . . on that side of S on which there is disturbance
at time t.
206. Motion of a surface of
discontinuity. Dynamical
conditions.
The
dynamical
conditions which hold at the surface .S' are found
by
considering
the
changes
of momentum of a thin slice of the medium in the
immediate
neighbourhood
of 8. We mark out a small area B8 of
8,
and
consider the
prismatic
element of the medium which is bounded
by
S,
by
the
normals to 8 at the
edge
of Bs and
by
a surface
parallel
to /S at a distance cBt
from it. In the short time
Bt,
this element
passes
from a state of rest without
strain to a state of motion and strain
corresponding
with the
displacement
(m,
V,
w).
The
change
is effected
by
the resultant traction across the
boundaries of the
element,
that is
by
the traction across
B8,
and the
change
of momentum is
equal
to the
time-integral
of this traction. The traction in
question
acts across the surface normal to v
upon
the matter on that side of
the surface towards which v is
drawn,
so that its
components per
unit of area
are
X^,

F,

2.
The resultants are obtained
by multiplying
these
by
BS,
and their
impulses by multiplying by
Bt. The
equation
of momentum is
therefore
P^s.cul^^,.%.'-;^
=
-iX,.y..z,m,
from which we have the
equations
205-207]
AN ISOTROPIC SOLID MEDIUM 285
In these
equations du/dt,...
and
Xy,...
are to be calculated from the values
(if
!t,...
on that side of S on which there is
disturbance;
and the
equations
hold at all
points
of S.
In the case where there is motion and strain on both sides of the surface
S,
but the
displacements
on the two sides of S are
expressed by
different
formulae,
we
may
denote
them
by (?<i, Wj, w^)
and
(jtjj i'2> '^'-'i)-
A-t all
points
of <Si the
displacement
must be the
same whether it is calculated from the
expressions
for
u^,
... or from those for
Mj,
We
may prove
that the values at S of the dififerential coefficients of
m,,
... are connected
by equations
of the
type
du-i Smj 9mi 3jt2 ^^ ^2
cos{x, v)" cos{y,
v) cos(z, v)~8i/
dv~
c\dt
dt
J'
with similar
equations
in which u is
replaced by
v or
by
w. If we denote the tractions
calculated from
(iti, v^, w,) by A'^-O,
... and those calculated from
(mj, v.^, w.^) by XJ^),
...
we
may
show that the values at 5 of these
quantities
and of
dujdt,
... are connected
by
the
equations
207.
Velocity
of waves in
isotropic
medium.
If we write
I, m,
n for the direction cosines of
v,
the
equations (11)
become
three
equations
of the
type
dv
dv\
A
,
dw dw
f^iOV dv\ f.,,dw dw\] , ,
of which the
right-hand
member
may
also be written in the form
, ,
/9m 'bv
d'w\ \
du dv
,
dv
^^+2^)^9-^
+
9^
+
W
+
^r9-y
+
"^9-S-''9y
f du
,
dw
^,dw] ,,,
+
^rde'-"d.-^^Y^\
(13)
These
equations
hold at the surface
S,
at which also we have nine
equations
of the
tvpe
'-M
a*)
so
that,
for
example.
dx c dt'
J
dv
_
dv
_
Im dv
dy
doc c dt'
On
substituting
for
dujdx,
... from
(14)
in
(12),
we obtain the
equation
.du
,,, , , ,9m ,^ \fi
dv
,
dw\
,,
_.
pc
g^
=
K^
+
/x)
^^
+
/.^}
g^
+
(X
+
/i)
\lm^^
+ In
-^j
;
(15)
286 PROPAGATION OF WAVES THROUGH
[CH.
XIII
and,
on
eliminating dti/dt, dv/dt, div/dt
from this and the two similar
equations,
we obtain the
equation (7)
of Article 204. The form
(13)
and the
equations
of
type (14)
show that
equation (12) may
also be written
-pCg^
=
(X
+
2M)i(g-+g^+g^)-/xm^--^J+;x,z^^--j.
(16)
Hence it follows
that,
when the rotation
vanishes,
we have three
equations
of
the
type
from which we should Hnd that
pc^
=
X +
2/i ; and,
when the dilatation
vanishes,
we have three
equations
of the
type
pc|
=
M|(m''
+
n^)|-im|-k|^},
from which we should find that
pc"

fi.
These results show that the surface of
discontinuity
advances with a
velocity
which is either
{(X
4-
2/a)/pP
or
(/x/p)*,
and
that,
if there is no
rotation,
the
velocity
is
necessarily [(X
+
2fj.)/p]\
and,
if there is no
dilatation,
the
velocity
is
necessarily {nlp)^-
208.
Velocity
of waves in
seolotropic
solid medium.
Equations
of the
types (10)
and
(11)
hold whether the solid is
isotropic
or
not. The former
give
the six
equations
c c c
(
w
v\ {
u
,w\ f ,v
u
(17)
in which the dots denote differentiation with
respect
to
t,
and
/, m,
n are
written for cos
(x, v),....
The
equations (11)
can be written in such forms as
-pcu
=
l^+m^ +n^,
(18)
^^xx
^*^xy
^*^zx
where W denotes the
strain-energy-function expressed
in terms of the
components
of strain.
Now let
^, r), f
stand for
ujc, v/c, w/c. Equations
(17)
are a linear
substitution
expressing
exz,---
in terms of
^, rj, f.
When this substitution
is carried out W becomes a
homogeneous quadratic
function of
f, rj, ^.
Denote this function
by
IT. We observe
that,
since
e^
dependent
of
f,
we have the
equation
an
,dw
dw dW
-_.
= _
^ m. . n
5

,
Of
oexx
oBxy
oe
=
(21)
207-209]
AN iEOLOTROPIC SOLID MEDIUM 287
and we have similar
equations
for
dU/dt)
and
dTL/d^.
Hence the
equations
of
type (18)
can be written
P'^^JI' P'-^^-d^' P^'^-^J^
^^^^
Now
suppose
that IT is
given by
the
equation
n
=
i [\^^
+
\^v'
+
^,,^'+2\,v^+ 2X3,??+ 2\f,7], (20)
then the
equations (19)
show that c^ satisfies the
equation
A<]2 1 ^22
~~
P t 23
"13, Ai23,
\^TpC'
Since
^, r], ^
are connected with
exxy-
by
a real linear
substitution,
the
homogeneous quadratic
function 11 is
necessarily positive,
and therefore
equation (21) yields
three real
positive
values for c^ The coefficients of
this
equation depend upon
the direction
{I,
m,
n).
There are
accordingly
three real wave-velocities
answering
to
any
direction of
propagation
of waves*.
The above
investigation
is
effectively
due to E. B.
Christoffelt,
who has
given
the
following
method for the formation of the function n:

Let the six

components
of strain
e^x, By^,
...
e^y
be denoted
by x^, x^,
...
^o
;
and let
c^
denote the form
in which
Cj, Cjj
have no
quantitative meaning,
but
c^
is to be
replaced by c, c^c^ by c,2
and so
on, Cu, c,2,
...
being
the coefficients in the
strain-energy-function.
Then we
may
write
Again,
let
Xj, Xj, X3
be defined
by
the
symbolical equations
\
=
c-J,
+
Cgin-\-Ci^n, X2
=
Cgi!-t-C2OT-fC4., 'K^
=
Ci,l-\-c^m
+
C3n,
then we have
-Cx
=
'!<i^+\ri-'r\^^, W^Q^y^
+
^rj
+
X^^Y,
and therefore the coefficients
Xn,
... iu the function rt are to be obtained
by squaring
the
form
\k-'r\n
+
\Cy
'' ^^'^ ^^ye
\\

'^n I'+Cea^n^
+
''ss'^^
+
Scj^mM
-I-
2ci^nl
+
2cjj?ot,
209. Wave-surfaces.
The
envelope
of the
plane
lx+mi/
+
n2
=
e
(22)
in which c is the
velocity
of
propagation
of waves in the direction
(I, m, n)
is the
"
wave-
surface"
belonging
to the medium. It is the surface
bounding
the disturbed
portion
of
the medium after the
lapse
of one unit of
time, beginning
at an instant when the dis-
turbance is confined to the immediate
neighbourhood
of the
origin.
In the case of
isotropy,
c is
independent
of
I, m, n,
and is
given by
the
equation (7);
in the case of
*
For a
general
discussion of the three
types
of waves we
may
refer to Lord
Kelvin,
Baltimore
Lectures,
London 1904.
t
Ann. di Mat.
(Ser. 2),
t. 8
(1877).
288
WAVE-SURFACES
[CH.
XIII
wolotropy
c is a function of
I, in,
n
given by
the
equation (21).
In the
general
case the
wave-surface is
clearly
a surface of three
sheets, corresponding
with the three values of c^
which are roots of
(21).
In the case of
isotropy
two of the sheets are
coincident,
and
all the sheets are concentric
spheres.
Green* observed
that,
in the
general
case of
aeolotropy,
the three
possible
directions
of
displacement, answering
to the three velocities of
propagation
of
plane
waves with a
given wave-normal,
are
parallel
to the
principal
axes of a certain
ellipsoid,
and
are, therefore,
at
right angles
to each other. The
ellipsoid
would be
expressed
in our notation
by
the
equation (Xj,, Xjj,
...
Xjj) {x, y, 0)^
=
const. He showed
that,
when W has the form
iA{e^+eyyiref+\L{fi,.^-4^,,e^)
+
\M{eJ-iee^)-^^N{e^^-4e^eyy), (23)
the wave-surface is made
up
of a
sphere, corresponding
with the
propagation
of waves of
irrotational
dilatation,
and Fresnel's
wave-surface,
viz. : the
envelope
of the
plane (22)
subject
to the condition
<?-Llp
"
~<^'-MIp
"
^^Njp
^ '
The two sheets of this surface
correspond
with the
propagation
of waves of
equivoluminal
distortion. Green arrived at the above
expression
for W as the most
general
which would
allow of the
propagation
of
purely
transverse
plane waves,
i.e. of waves with
displacement
parallel
to the wave-fronts.
Green's formula
(23)
for W is included in the formula
(15)
of Article
110,
viz. :
2lf=(J, B, G, F, O, H){e^, e, e,,f
+
Ley,^^MeJ^Ne^\
which characterizes elastic solid media
having
three
orthogonal planes
of
symmetry.
To obtain Green's formula we have to
put
A
=
B=C, F=A-2L, 0=A-2M,
H=A-2N.
It is
noteworthy
that these relations are not satisfied in cubic
crystals.
Green's formula for the
strain-energy-function
contains the
strain-components only;
the notion of a medium for which
lF=2(Zarj,2-fi/^ii7,2+irar/) (25)
was introduced
by MacCuUaghf.
The wave-surface is Fresnel's wave-surface.
Lord
Rayleight, following
out a
suggestion
of
Rankine's,
has discussed the
propagation
of waves in a medium in which the kinetic
energy
has the form
///^[-(sy^-dT-^-]'^'^^*'
^^)
while the
strain-energy-function
has the form
appropriate
to an
isotropic
elastic solid.
Such a medium is said to exhibit
"aeolotropy
of inertia." When the medium is incom-
pressible
the wave-surface is the
envelope
of the
plane (22) subject
to the condition
2
+ ~,
-1-^
=0; 27)
Vi-M c^^2-;* cVs-J*
it is the fii'st
negative pedal
of Fresnel's wave-surface with
respect
to its centre.
*
'On the
propagation
of
light
in
crystallized media,' Cambridge
Phil. Soc.
Trans.,
vol. 7
(1839),
or Mathematical
Papers,
London
1871,
p.
293.
t
'An
essay
towards a
dynamical theory
of
crystalline
reflexion and
refraction,' Dublin,
Trans. R. Irish
Acad.,
vol. 21
(1839),
or Collected Works
of
James
MacCullagh,
Dublin
1880,
p.
145.
X
'On Double
Kefraction,'
Phil.
Mag. (Ser. 4),
vol. 41
(1871),
or
Scientific Papers,
vol.
1,
Cambridge
1899.
209,
210]
poisson's integral formula
289
The case where the
energy-function
of the medium is a function of the
components
of
rotation as well as of the
strain-components,
so that it is a
homogeneous quadratic
func-
tion of the nine
quantities
^,
5-,
^,
,
has been discussed
by
H. M. Macdonald*.
The most
general
form which is admissible if transverse waves are to be
propagated
independently
of waves of dilatation is shown to lead to Fresnel's wave-surface for the
transverse waves.
The still more
general
case in which there is
ajolotropy
of inertia as well as of elastic
quality
has been
investigated by
T. J. I'A.
Bromwichf.
It
appears that,
in this
case,
the
requirement
that two of the waves shall be
purely
transverse does not lead to the same
result as the
requirement
that
they
shall be
purely rotational, although
the two
require-
ments do lead to the same result when the
reolotropy
does not affect the inertia. The
wave-surface for the rotational waves is derived from Fresnel's wave-surface
by
a homo-
geneous
strain.
210. Motion determined
by
the characteristic
equation.
It
appears
that,
even in the case of an
isotropic
solid,
much
complexity
is
introduced into the
question
of the
propagation
of disturbances
through
the
solid
by
the
possible
co-existence of two
types
of waves
propagated
with
different velocities. It will be well in the first instance to confine our
attention to waves of a
single type

irrotational or
equivoluminal.
The
motion is then determined
by
the characteristic
equation (6)
of Article
204,
viz.
d'<l>ldf
=
c-V^<l>.
This
equation
was solved
by Poissonj
in a form in which the value of
(f>
at
any place
and time is
expressed
in terms of the initial values of
^
and
d<f)/dt.
Poisson's result can be stated as follows : Let
^0
and
^0
denote the initial
values of
</>
and
d(j>/dt.
With
any point {x, y, z)
as centre describe a
sphere
of
radius
ct,
and let
^0
and
(^
denote the mean values of
<^
and
<|>o
on this
sphere.
Then the value of
^
at the
point {x, y, z)
at the instant t is
expressed by
the
equation
^
=
j^{t^.)
+
t^o (28)
If the initial disturbance is confined to the
region
of
space
within a closed
surface
2o,
then
(f>o
and
^0
have values different from zero at
points
within
2,,
and vanish outside
So.
Taking any point
within or on
2
as
centre,
we
may
describe a
sphere
of radius ct
;
then the disturbance at time t is confined to
the
aggregate
of
points
which are on the surfaces of these
spheres.
This
aggregate
is,
in
general,
bounded
by
a surface of two sheets

an inner and
an outer. When the outer sheet reaches
any point,
the
portion
of the medium

London Math. Soc.

Proc,
vol. 32
(1900), p.
311.
t
London Math. Soc.
Proc,
vol. 34
(1902), p.
307.
J
Paris,
Mem. de
I'Institut,
t. 3
(1820).
A
simple proof
was
given by Liouville,
J. de Math.
(Lioinille),
t. 1
(1856).
A
symbolical proof
is
given by
Lord
Hayleigh, Theory of Sound,
Chapter
XIV.
L. E.
19
290 MOTION RESULTING FROM
[OH.
XIII
which is close to the
point
takes
suddenly
the small strain and
velocity
implied by
the values of
(j)
and
d<j)ldt
;
and after the inner sheet
passes
the
point,
the same
portion
of the medium returns to rest without strain*.
The characteristic
equation
was solved in a more
general
manner
by
Kirchhofff.
Instead of a
sphere
he took
any
surface
S,
and instead of the
initial values of
(f>
and
d<f>/dt
on S he took the values of
<f>
and its first
derivatives at
points
on S and at certain instants
previous
to the instant t.
If
Q
is
any point
on
S,
and r is the distance of
Q
from the
point (x, y, z),
the
values of
<^
and its first derivatives are estimated for the
point Q
at the instant
t

rjc.
Let
[^],
... denote the values of
<^,
... estimated as stated. Then the
value of
^
at the
point {x, y, z)
at the instant t is
expressed by
the
equation
where v denotes the direction of the normal to S drawn towards that side on
which
(x, y, z)
is situated.
Kirchhoff's formula
(29) may
be obtained
very simply {, by substituting t-r/c
for t in
fj) {x, y, z, t),
where r now denotes the distance of
{x, y, z)
from the
origin. Denoting
the
function
<^ {x, y, z,
t-
rjc) by -^ (x, y,
z, t),
we
may
show that when
(\) {x, y, z, t)
satisfies
the characteristic
equation (6), yjr
satisfies the
equation
,^"*-![(5l)4(5t)4(5t)>-
('
If this
equation
holds
throughout
the
region
within a closed surface S which does not
contain the
origin,
we
integrate
the left-hand member of this
equation through
the
volume within S and transform the volume
integral
into a surface
integral,
thus
obtaining
the
equation
J J \
ov r ov cr ov at
/
If now
[(^],
... denote the values of
(f),...
at the instant
t-r/c,
this
equation
is the
same
as
since,
as is
easily proved.
"When the
origin
is within the surface S we
integrate
the left-hand member of
(30) through
the volume contained between S and a small
sphere
2 with its centre at the
origin,
and
pass
to a limit
by contracting
the radius of 2
indefinitely.
We thus find for the value
of
<f)
at the
origin
the formula
(29),
and the same formula
gives
the value of
(j)
at
any point
*
Cf.
Stokes,
'
Dynamical theory
of
diffraction,' Cambridge
Phil. Soc.
Trans.,
vol. 9
(1819),
or Math, and
Phys. Papers,
vol.
2, p.
243.
t
Ann.
Phys.
Chem.
(Wiedemann),
Bd. 18
(1883).
See also
Kirchhoff, Vorlesungcn
iiher math
Physik, Optik, Leipzig,
1891.
X
Cf.
Beltrami, Rome,
Ace. Lincei Rend.
(Ser. 5),
t. 4
(1895).
210,
211]
GIVEN INITIAL CONDITIONS 291
and instcant. The formula holds for a
region
of
space
bounded
internally
or
externally by
a closed surface
S, provided that,
at all instants which como into
consideration,
<{)
and its
first derivatives are
continuous,
and its second derivatives are finite and are connected
by
equation (6),
at all
points
of the
region*.
In case the
region
is.outside
S, <j>
must tend to
zero at infinite distances in the order r-i at least. These conditions
may
bo
expressed by
saying
that all the sources of disturbance are on the side of S remote from
{x, y, z).
Kirchhofi^s formula
(29)
can be shown to include
Poisson'sf.
The formula
may
also be
written in the form
^-mm-mv'-
<'
where
^
(
'^-'
J
is to bo formed
by
first
substituting
t

ric for t in
<^
and then diflerentiat-
ing
as if r were the
only
variable
quantity
in
[<^]/r.
The formula
(31)
is an
analogue
of
Green's formula
(7)
of Article 158. It can be
interpreted
in the statement that the
value of
(j>
at
any point
outside a closed surface
(which
encloses all the sources of dis-
turbance)
is the same as that due to a certain distribution of fictitious sources and double
sources on the surface. It is
easy
to
prove,
in the manner of Article
124,
that the motion
inside or outside
S,
that is due to
given
initial
conditions,
is
uniquely
determined
by
the
values of either
(p
or
dfftfbv
at S. The theorem
expressed by equation (31)
can be deduced
from the
properties
of
superficial
distributions of sources and double sources and the
theorem of
uniqueness
of solution
J.
211.
Arbitrary
initial conditions.
When the initial conditions are not such that the disturbance is
entirely
irrotational or
equivoluminal,
the results are more
complicated. Expressions
for the
components
of the
displacement
which
arises,
at
any place
and
time,
from a
given
initial distribution of
displacement
and
velocity,
have been
obtained,
and the result
may
be stated in the
following
form :

Let
i;Ua,Va, w^)
be the initial
displacement, supposed
to be
given throughout
a
region
of
space
T and to vanish on the
boundary
of T and outside
T,
and
let
(m,i)o,Wo)
be the initial
velocity supposed
also to be
given throughout
T
and to vanish outside T. Let a and b denote the velocities of irrotational
and
equivoluminal
waves. Let
Si
denote a
sphere
of radius at
having
its
centre at the
point (, y, z),
and
8^
a
sphere
of radius ht
having
its centre at
the same
point.
Let V denote that
part
of the volume contained between
these
spheres
which is within T. Let r denote the distance of
any point
(', y', z)
within
V,
or on the
parts
of
Si
and S., that are within
T,
from the
point
{x,
y, z),
and let
q^
denote the initial
displacement
at
(oo', y
,
z'),
and
q^
the initial
velocity
at the same
point,
each
projected upon
the radius vector
*
For the case where there is a
moving
surface of
discontinuity
outside
S,
see a
paper by
the
Author,
London Math. Soc. Proc.
(Ser. 2),
vol. 1
(1904), p.
37.
t See
my paper just
cited.
t
Of. J.
Larmor,
London Math. Soc. Proc.
(Ser. 2),
vol. 1
(1904).

For references see

Introduction, p.
18. Reference
may
also be made to a
paper by
the
Author in London Math. Soc. Proc.
(Ser. 2),
vol. 1
(1904), p.
291.
192
292 MOTION DUE TO
[CH.
XIII
r,
supposed
drawn from
{x, y, z).
Then the
displacement
u at
{x, y, z)
at the
instant t can be written
1
//(
/ 8V-1 av-> 3V-'\ 8r->/
.
a^oA
1
/.. . 9wi)
-^{tu,
+ u,+
r^)^dS,
(32)
and similar
expressions
for v and w can be written down. The surface-
integrations
extend over the
parts
of
S^
and
S^
that are within T.
The dilatation and the rotation can be calculated from these
formulae,
and it can be shown that the dilatation is
entirely
confined to a wave of
dilatation
propagated
with
velocity
a,
and the rotation to a wave of rotation
propagated
with
velocity
b. If
Vi
and
r^
are the
greatest
and least distances
of
any point
of the medium from the
boundary
of
T,
the motion at
begins
at the instant t
=
r^/a,
the wave of dilatation ends at the instant
t
=
r-t/a,
the wave of rotation
begins
at the instant t
=
r^/b,
and the motion
ceases at the instant t
=
rjb.
If the wave of dilatation ends before the wave
of rotation
begins,
the motion between the two waves is of the character of
irrotational motion in an
incompressible
fluid*;
at a distance from T which
is
great compared
with
any
linear dimension of T this motion is
relatively
feeble.
The
problem
of the
integration
of the
equations
of small motion of an
isotropic
elastic
solid has been the
subject
of
very
numerous researches. Reference
may
be made to the
following
memoirs in addition to those
already
cited :

V.
Cerruti,
'Sulle vibrazioni de'
corpi
elastici
isotropi,' Borne,
Ace.
Idncei,
Mem.
fis.
mat., 1880;
V.
Volterra,
'Sur les vibra-
tions des
corps ^lastiques isotropes,'
Acta
Math.,
t. 18
(1894);
G.
Lauricella,
'Sulle
equazioni
del moto dei
corpi elastici,'
Torino Mem.
(Ser. 2),
t. 45
(1895);
O.
Tedone,
'Sulle
vibrazioni dei
corpi
solidi
omogenei
ed
isotropi,'
Torino Mem.
(Ser. 2),
t. 47
(1897) ;
J.
Coulon,
'Sur
I'integration
des
equations
aux d^rivdes
partielles
du second ordre
par
la mdthode des
caraotdristiques,'
Paris
(Thhe)
1902. Hadamard's
treatise. Lemons
sur la
propagation
des
ondes,
Paris
1903,
also
may
be consulted.
212. Motion due to
body
forces.
Exactly
as in Article 130 we
express
the
body
forces in the form
(X,
Y, Z)
=
gradient
of 3>
+
curl
{L, M, N),
and the
displacement
in the form
(m,
V, w)
=
gradient
of
<^
+ curl
{F,
G,
H).
'
*
Cf.
stokes,
loc. cit.
211,
212]
VARIABLE BODY FORCES
293
Then the
equations
of motion of the
type
can be satisfied if
(f>,
F, G,
H
satisfy
the
equations
and
particular
solutions can be
expressed
in the forms*
The values of
<I>, Z,
. . . are
given
in terms of
X, Y,
Z
by
the
equations (7)
of Article
130,
and the
integrations expressed
in
(33)
can be
performed.
Taking
the case of a
single
force of
magnitude x (0> acting
at the
origin
in the direction of the axis of
x,
we
have,
as in Article
130,
4nrp
.(33)
where R denotes the distance of
{x', y', z)
from the
origin.
We
may partition
space
around the
point
{x,
y, z)
into thin sheets
by
means of
spherical
surfaces
having
that
point
as
centre,
and thus we
may express
the
integrations
in
(33)
in such forms as
where d8 denotes an element of surface of a
sphere
with centre at
{x,
y, z)
and radius
equal
to r. Now
JJ{dR~^/dx')
dS is
equal
to zero when the
origin
is inside
S,
and to iirr'
(dr^'^/dx)
when the
origin
is outside
S, r^
denoting
the
distance of
(x, y, z)
from the
origin.
In the former case
?,,
<
?%
and in the
latter
r
> r. We
may
therefore
replace
the
upper
limit of
integration
with
respect
to r
by
r^,
and find
,
^
i__ drf2
r\.
L
_ ^\
^,. "
4fiTd-p
9a;
Jo
V
/
Having
found
^
we have no further use for the r that
appears
in the
process,
and we
may
write r instead of
r^,
so that r now denotes the distance
of
{x, y, z)
from the
origin.
Then we have
'^-^XfS^'-'^''
(^'>
47r/3
dx
J
*
Cf. L.
Lorenz,
J.
f.
Math.
(Crelle),
Bd. 58
(1861),
or (Euvres
Scientijiques,
t. 2
(Copenhagen,
1899), p.
1. See also Lord
Eayleigh, Theory of Sound,
vol.
2,
276.
294! MOTION DUE TO VARIABLE FORCES
[CH.
XIII
In like manner we should find
47rp
H
=
-^^
t'x{t-t')dt'.
Jo-
'
.(35)
W36)
The
displacement
due to the force
;;^; (t)
is
given by
the
equations*
1 av->
/'/ft
1 ar ar ( 1
/.
r\ 1 /
r\]
213. Additional results
relating
to motion due to
body
forces.
(i)
The dilatation and rotation calculated from
(36)
are
given by
the
equations
^
4-i|^(-0}
(^^)
(ii)
The
expressions (36)
reduce to
(11)
of Article 130 when
x{t)
is
replaced by
a
constant.
(iii)
The tractions over a
spherical cavity required
to maintain the
displacement
expressed by (36)
are
statically equivalent
to a
single
force
parallel
to the axis of x.
When the radius of the
cavity
is diminished
indefinitely,
the
magnitude
of the force
is
X (0-
(iv)
As in Article
132,
we
may
find the efiects of various nuclei of strain t. In the
case of a "centre of
compression"
we
have, omitting
a constant
factor,
..->
=
(8~.|.|){ix(-0)
<=)
(m,
representing
irrotational waves of a well-known
type.
In the case of a "centre of rotation
about the axis of z" we
have, omitting
a
factor,
(' ^'
-)=(!'
-h
) {^^(^-i)}'
^^'^
representing equivoluminal
waves of a well-known
type.
(v)
If we combine two centres of
compression
of
opposite signs
in the same
way
as two forces are combined to make a "double force without moment" we obtain
irrotational waves of the
type expressed by
the
equation
(?, V, W):
=
(^.' Ik- S){r-^('-0}
(')
*
Formulae
equivalent
to
(36)
were obtained
by Stokes,
loc. cit.
t
For a more detailed
discussion,
see
my paper
cited on
p.
291.
212-214]
SURFACE WAVES
295
If we combine two
pairs
of centres of rotation about the axes of x and
y
and about
parallel axes,
in the same
way
as two
pairs
of forces are combined to make a centre of
rotation,
we obtain
equivoluminal
waves of the
type
in which the
displacement
is
expressed by
the same formulffi as the electric force in the
field around Hertz's oscillator*. Lord Kelvin t has shown that
by superposing
solutions
of the
types (40)
and
(41)
we
may
obtain the effect of an
oscillating rigid sphere
close to
the
origin.
(vi)
When
;( {t)
is a
simple
harmonic function of the
time, say x (0
=
-4
cos^*,
we
find
/^\'x(^-Orf^'=p{cosp(.-g-cos^(.-g-fsinp(^-;-)+fsinp(^-^)},
and
complete expressions
for the effects of the forces can be written down
by (36) t.
In this case we
may regard
the whole
phenomenon
as
consisting
in the
propagation
of two
trains of
simple
harmonic waves with velocities
respectively equal
to a and
6;
but the
formulas
(36)
show
that,
in more
general cases,
the efiect
produced
at the instant < at a
point
distant r from the
point
of
application
of the force does not
depend
on the
magnitude
of the force at the two instants
t-rja
and
trjb only,
but also on the
magni-
tude of the force at intermediate insfcmts. It is as if certain effects were
propagated
with velocities intermediate between a and
6,
as well as the definite effects
(dilatation
and
rotation)
that are
propagated
with these velocities
.
(vii)
Particular
integrals
of the
equations
of motion under
body
forces which are
proportional
to a
simple
harmonic function of the time
(written e'''')
can be
expressed
in
the forms
-\.X>TJb
F=
h^JIlL"^,
-d^'dy-d^',
where
*=-i///(^'^-l-I"'j^
+
^''^)'^-W<^^.
IIIK^'
214. Waves
propagated
over the surface of an
isotropic
elastic
solid
bodyj|.
Among periodic
motions
special importance
attaches to those
plane
waves
of
simple
harmonic
type, propagated
over the
bounding
surface of a solid
*
Hertz,
Electric
Waves, English
edition, p.
137. For the discussion in
regard
to the result
see W.
Konig,
Ann.
Phys.
Cliem.
(Wiedemann),
Bd. 37
(1889),
and Lord
Bayleigh,
Phil.
Hag.
(Ser. 6),
vol. 6
(1903), p.
38.5.
t Phil.
Mag. (Ser. 6),
vols. 47 and 48
(1899).
J
For the effects of forces which are
simple
harmonic functions of the
time,
see Lord
Bayleigh,
Theory of Sound,
vol.
2, pp.
418 et
seq.

Cf.
my paper
cited on
p. 291,
and Stokes's result recorded on
p.
292.
II
Cf. Lord
Bayleigh,
London Math. Soc.
Proc,
vol. 17
(1887),
or
Scientific Papers,
vol.
2,
p.
441.
296 WAVES PROPAGATED OVER THE
[CH.
XIII
body,
which involve a disturbance that
penetrates
but a little distance into
the interior of the
body.
We shall take the
body
to be bounded
by
the
plane
z
=
0,
and shall
suppose
that the
positive
sense of the axis of z is directed
towards the interior of the
body.
We shall
suppose
that the
components
of
displacement,
besides
being proportional
to
e'p*,
are
proportional
to
e'^^^ov)^
so that
2Tr/^(f^ +g^)
is the
wave-length.
As in Article
190,
we denote
p'p/iX
+
2/i) by
h^ and
p^p/i^ by
'. The dilatation A satisfies the
equation
(V-
+
h^)
A
=
0,
and since it is
proportional
to e'
'f^+sy)
we must have
A
=
Pe-"+''^''+<'y+p'\
(42)
where P is a
constant,
and
r"-=P
+
9'-h' (43)
A
particular integral (mj,
v^,
Wj)
of the
equations
of motion is then
expressed by
the
equations
(mi,
Vi,
Wi)
=
(- t/,
-
ig, r)
/t-2Pe-+'(/^+w+pO,
(44,)
and a more
general integral
will be found
by taking {u,
v, w)
to be of the
form
(wi
+
M2, Vi
+
V2, vji
+
w^),
where
U2, v^, vo^
are
given by
the
equations
(ms,V2, W2)
=
(4,P,a)e-+''/^+^!'+p", (45)
in which
A, B,
C are constants connected
by
the
equation
ifA
+
i.gB-sC
=
0,
(46)
and
s'=f'
+
g'-K"- (47)
The surface z
=
being
free from
traction,
we must have the
equations
ic-
^\ r, ^
r-
^,-2JP-2^,P-2.(7=0,
the third of which can be written
[k' -2{p
+
g')]
P
-
2h'sG
=
0.
We
may
solve these
equations
so as to
express
A, B,
C in terms of P.
Writing
'-^
=
V(/^
+
f),
h'-'
=
h%p
+
g% (48)
we find
ic'^-2
_B_
K''-2 +
'irsj(f'
+
g')
~
2h'H
'
f'g''
2h'^
(1
-
k") {p
+
<7)
'
and,
on
substituting
in
(46),
we obtain the
equation
{K''-2f==4>rsl{p
+
g^) (49)
which
becomes,
on elimination of r and s
by
means of
(43)
and
(47),
'8_8/c'<'+24'''-16(l+/t'0^+16/i'==0
(50)
When the material is
incompressible,
or
h'-/K'''=0,
the
equation
for '"
becomes a cubic '" 8:'''+
24*;'^ 16
=
0,
which has a real
positive
root
214]
SURFACE OF A BODY 297
91275... and two
complex
roots
(3-5436...)
t(2-2301...).
Since
ic'li/^
+
g')
is finite and
h^JK^
=
0,
equation (43)
shows that r is real.
Equation (49)
shows that for the
complex
values of k
4rs/(/
+
5^0
=
-
(2-7431...)
t
(6-8846...) (51)
Since the real
part
of
s,
as
given by
this
equation,
has the
opposite sign
to
r,
there are no motions of the
type
under discussion which
correspond
with the
complex
values of k. But when we take the real
root,
viz. '=
=
-91275...,
we find
r'^f'
+
g\
s^
=
(-08724...)(/=+^=),
(52)
and we have a wave-motion of the
required type.
The
velocity
of
propagation
of the waves is
WV(/=
+
^0
=
C9554...)V(Wp)
(53)
which is a little less than the
velocity
of
equivoluminal
waves
propagated
through
the solid.
When the material satisfies Poisson's condition
(X
=
/i),
we have
K^jh''
=
3,
and then there is a wave-motion of the
required type,
in which
'=
=
-8453...,
r^={-7l82...)(r-
+
g%
s^
=
(-1546...)(/^
+
r=), (54)
and the
velocity
of
propagation
is now
(-9194...) V(m^)
(55)
Concerning
the above
type
of waves Lord
Rayleigh {loc. cit.)
remarked :
"
It is not
improbable
that the surface waves here
investigated play
an
important part
in
earthquakes,
and in the collision of. elastic solids.
Diverging
in two dimensions
only, they
must
acquire
at a
great
distance
from the source a
continually increasing preponderance."
The
subject
has
been
investigated
further
by
T. J. I'A. Bromwich* and H.
Lambf.
The
former showed
that,
when
gravity
is taken into
account,
the results obtained
by
Lord
Rayleigh
are not
essentially
altered. The latter has discussed the
effect of a limited initial disturbance at or near the surface of a solid
body.
He showed
that,
at a distance from the
source,
the disturbance
begins
after
an interval
answering
to the
propagation
of a wave of irrotational dilatation
;
a second
stage
of the motion
begins
after an interval
answering
to the
propa-
gation
of a wave of
equivoluminal
distortion,
and a disturbance of much
greater amplitude begins
to be received after an interval
answering
to the
propagation
of waves of the
type investigated by
Lord
Rayleigh.
The
importance
of these waves in relation to the
theory
of
earthquakes
has
perhaps
not
yet
been
fully appreciated.

London Math. Soc.

Proc,
vol. 30
(1899).
+ Phil. Trans.
Roy.
Soc.
(Ser. A),
vol. 203
(1904).
CHAPTER XIV.
TORSION.
215. Stress and strain in a twisted
prism.
In Article 86
(d)
we found a
stress-system
which could be maintained in
a
cylinder,
of circular
section,
by
terminal
couples
about the axis of the
cylinder.
The
cylinder
is twisted
by
the
couples,
so that
any
cross-section
is
turned,
relatively
to
any
other,
through
an
angle proportional
to the
distance between the
planes
of section. The traction on
any
cross-section
at
any point
is
tangential
to the
section,
and is at
right angles
to the
plane
containing
the axis of the
cylinder
and the
point
;
the
magnitude
of this
traction at
any point
is
proportional
to the distance of the
point
from the
axis.
When the section of the
cylinder
or
prism
is not
circular,
the above
stress-system
does not
satisfy
the condition that the
cylindrical boundary
is
free from traction. We seek to
modify
it in such a
way
that all the conditions
may
be satisfied. Since the tractions
applied
at the ends of the
prism
are
statically equivalent
to
couples
in the
planes
of the
ends,
and the
portion
of the
prism
contained between
any
cross-section and an end is
kept
in
equilibrium by
the tractions across this section and the
couple
at the
end,
the tractions in
question
must be
equivalent
to a
couple
in the
plane
of the
cross-section,
and the moment of this
couple
must be the same for all cross-
sections. A suitable distribution of
tangential
traction on the cross-sections
must be the essential feature of the
stress-system
of which we are in search.
Accordingly,
we seek to
satisfy
all the conditions
by
means of a distribution
of
shearing
stress,
made
up
of
suitably
directed
tangential
tractions on the
elements of the
cross-sections, combined,
as
they
must
be,
with
equal
tangential
tractions on elements of
properly
chosen
longitudinal
sections.
We shall find that a
system
of this kind is
adequate
;
and we can
foresee,
to some
extent,
the character of the strain and
displacement
within the
prism.
For the strain
corresponding
with the
shearing
stress,
which we
have
described,
is
shearing
strain which
involves,
in
general,
two
simple
shears at each
point.
One of these
simple
shears consists of a relative
sliding
215,
216]
THE TORSION PROBLEM
299
in a transverse direction of elements of different cross-sections
;
this is the
type
of strain which occurred in the circular
cylinder.
The other
simple
shear consists of a relative
sliding, parallel
to the
length
of the
prism,
of
different
longitudinal
linear elements.
By
this shear the cross-sections
become distorted into curved surfaces. The
shape
into which
any
cross-
section is distorted is determined
by
the
displacement
in the direction of
the
length
of the
prism.
216. The torsion
problem*.
We shall take the
generators
of the surface of the
prism
to be
parallel
to
the axis of
z,
and shall
suppose
that the material is
isotropic.
The discussion
in the last Article leads us to assume for the
displacement
the formulae
u
=

ryz,
v
=
TZX,
w
=
T<f>, (1)
where
^
is a function of x and
y,
and t is the twist. We work out the
consequences
of this
assumption.
The
strain-components
that do not vanish are
e^x
and
Cyz,
and these are
given by
the
equations
n_.y\
. -J^_
(I-')
'^)
\dx ^)'
^^^
'
\dy
The
stress-components
that do not vanish are
X^
and
Y^,
and
they
are
given
by
the
equations
^-'"(s-')'
''='^(1+-)
<^>
The
equations
of
equilibrium,
when there are no
body
forces,
are satisfied if
the
equation
S+P=
w
holds at all
points
of
any
cross-section. The condition that the
cylindrical
bounding
surface of the
prism
is free from traction is satisfied if the
equation
-^
=
y
cos
(, j^)

a; cos
(y, v) (5)
holds at all
points
of the
bounding
curve of
any
cross-section. The com-
patibility
of the
boundary
condition
(.5)
with the differential
equation (4)
is
shown
by integrating
the left-hand and
right-hand
members of
(5)
round the
boundary,
and
transforming
the
line-integrals
into
surface-integrals
taken
over the area of the cross-section. The
integral
of the left-hand member of
(5)
taken round the
boundary
is
equivalent
to the
integral
of the left-hand
member of
(4)
taken over the area of the cross-section
;
it therefore vanishes.
The
integral
of the
right-hand
member of
(5)
taken round the
boundary
also
vanishes.
*
The
theory
is due to Saint-Venant. See
Introduction,
footnote 50 and
p.
19.
300 THE TORSION PROBLEM
[CH.
XIV
The tractions on
any
cross-section
are,
of
course,
statically equivalent
to
a
single
force
(which may
be
zero)
at the
origin
of
(x, y)
and a
couple.
We
show that
they
are
equivalent
to a
couple only.
The axis of the
couple
is
clearly parallel
to the
generators
of the surface of the
prism.
We have to
show that
jjx.dxdij
=
0,
jJY.dccdy
=
0.
Now
I jX^dxdy
=
/xt
1 1
U^
-
y
j
dxdy,
and this
may
be
replaced by
-SH^^-^l,HI-)i:
dxdy,
by
the
help
of the differential
equation (4).
The
expression
last written
may
be transformed into an
integral
taken round the
bounding
curve,
viz.
fjLT
la; (-

2/
cos
(x, v)
+
x cos
(y,
vn
ds,
where ds is the element of arc of the
bounding
curve. This
integral
vanishes
in
consequence
of the
boundary
condition
(5).
We have thus
proved
that
UXidxdy

0,
and in a similar
way
we
may prove
that
ljYgdxdy
=
0. It
follows that the tractions on a cross-section are
statically equivalent
to a
couple
about the axis of z of moment
^,jj(^^.
+
y.
+
JA.y^jdxdy
(6)
dy
We have now
proved
that the
prism
can be held in the
displaced position
given by equations (1) by
means of
couples applied
at its
ends,
the axes of
the
couples being parallel
to the central-line of the
prism.
The moment of
the
couple
when the twist is t is a
quantity
Or,
where
C
=
^.fj{a^
+
f
+
x^^-ylf)dxdy
..(7)
dy
The
quantity
C is the
product
of the
rigidity
of the material and a
quantity
of the fourth
degree
in the linear dimensions of the cross-section. G is
sometimes called the "torsional
rigidity"
of the
prism.
The
complete
solution of the
problem
of
torsion,
for a
prism
of
any
form
of
section,
is effected when
(f>
is determined so as to
satisfy
the
equation
(4)
and the
boundary
condition
(5).
The
problem
of
determining 4>
for a
given
boundary
is sometimes called the
"
torsion
problem
"
for that
boundary.
The
function
(p
is sometimes called the
"
torsion-function
"
for the
boundary.
216,
217]
THE TORSION PROBLEM
301
In the above solution the
twisting couple
is
applied by
means of tractions
X V
which
are
expressed by (3).
The
practical utility
of the solution is not confined to the case where
the
couple
is
applied
in this
way.
When the
length
of the
prism
is
great compared
with
the linear dimensions of its
cross-section,
the solution will
represent
the state of the
prism everywhere except
in
comparatively
small
parts
near the
ends,
whether the
twisting
couple
is
applied
in the
specified way
or not.
[Cf.
Article
89.]
The
potential energy per
unit of
length
of the twisted
prism
is
and this is
equal
to
*"-//{(i-)'-(t-)'}-*
"t,-'^** "*(l)'}-'-
Now
(j) {y
cos
{x, v)-x
cos
{y, v)}
ds
ds
^^^<ixdy.
It follows that the
potential energy per
unit of
length
is
\CtK
217. Method of solution of the torsion
problem.
Since
^
is a
plane
harmonic
function,
there exists a
conjugate
function
i^
which is such that
(f)
+
i^fr
is a function of the
complex
variable x +
cy
; and,
if
yfr
can be
found,
<j>
can be written down
by
means of the
equations
d(j> d^lr 3<^
_
Bi/r
dx
dy
'
dy
dx'
Fig.
21.
The function
f
satisfies the
equation
-^
+
^.=^'^^
^^^
P^*^
'"^^^'^ *^
bounding
curve of the cross-section,
and a certain condition
at this
boundary.
We
proceed
to find the
boundary-condition
for
i/f.
302 ANALOGY OF THE TORSION PROBLEM TO
[CH.
XIV
Taking
ds for the element of arc of the
bounding
curve,
and
observing
that,
when the senses of s and v are those indicated
by
arrows in
Fig.
21,
cos
(x, v)
=
dyjds,
cos
(y, v)
=

dw/ds,
the condition
(5) may
be written
dyft dy 9\^
dx
_
dy
dx
dy
ds dx ds
"
ds ds
'
and it follows that at the
boundary,
>/r
-
-^ (=
+
2/2)
=
const
(8)
The
problem
is thus reduced to that of
finding
a
plane
harmonic function
which satisfies this condition.
Apart
from additive constants the functions
<f>
and
yp-
are
uniquely
determinate*.
218.
Analogies
with
Hydrodynamics.
(a)
The functions
<p
and
yjr
are
mathematically
identical with the
velocity-potential
and stream-function of a certain irrotational motion of
incompressible
frictionless
fluid,
contained in a vessel of the same
shape
as
the
prism
f.
This motion is that which would be set
up by rotating
the
vessel about its axis with
angular velocity equal
to

1.
(6)
The function
yjr

^ (of
+
y^)
is
mathematically
identical with the
velocity
in a certain laminar motion of viscous fluid. The fluid flows under
pressure through
a
pipe,
and the section of the
pipe
is the same as that
of the
prismJ.
(c)
The function
yfr

^{x'^
+
y^)
is also
mathematically
identical with the
stream-function
of a motion of
incompressible
frictionless fluid
circulating
with uniform
spin, equal
to
unity,
in a fixed
cylindrical
vessel of the same
shape
as the
prism.
The moment of momentum of the
liquid
is
equal
to
the
quotient
of the torsional
rigidity
of the
prism by
the
rigidity
of the
material. The
velocity
of the fluid at
any point
is
mathematically
identical
with the
shearing
strain of the material of the
prism
at the
point.
In the
analogy (a)
the vessel rotates as stated
relatively
to some frame
regarded
as
fixed,
and the axes of x and
y
rotate with the vessel. The
velocity
of a
particle
of the
fluid relative to the fixed frame is resolved into
components parallel
to the instantaneous
positions
of the axes of x and
y.
These
components
are
d<f)jdx
and
3^/3y.
The
velocity
of
the fluid relative to the vessel is utilized in the
analogy (c).
Wo
may
use the
analogy
in the form
(a)
to determine the effect of
twisting
the
prism
about an axis when the effect of
twisting
about
any parallel
axis is known. Let
^d
be the
torsion-function when the axis meets a cross-section at the
origin
of
{x, y)
;
and let
<^'
be
the torsion-function when the
prism
is twisted about an axis
parallel
to the
first,
and
meeting
the section at a
point {it/, -if).
Rotation of the vessel about the second axis is
equivalent
at
any
instant to rotation about the first axis combined with a certain motion of
*
The functions are determined for a number of forms of
boundary
in Articles
221,
222
infra.
t
Kelvin and
Tait,
iJat. Phil. Part
Ji., pjx
242 et
seq.
X
J.
Boussinesq,
./. de math.
(Liouville), (Ser. 2),
t. 16
(1871).

A. G.
Greenhill,
Article
'
Hydromechanics,' Ency.
Brit.,
9th edition.
217-219]
CERTAIN PROBLEMS IN HYDRODYNAMICS 303
translation,
which is the same for all
points
of the vessel. This instantaneous motion
of translation is the motion of the first axis
produced by
rotation about the second
;
and
the
component
velocities in the directions of the axes are

y
and
a/,
since the
angular
velocity
of the vessel is -1. It follows that we must have
(f>''=<f>g-xy'
+
i/x'.
The
component displacements
are therefore
given by
the
equations
^= "'^
(y ~y')2^j v=t{x

x')z,
w
=
t^';
and the stress is the same as in the case where the axis of rotation
passes through
the
origin.
The torsional
couple
and the
potential energy
also are the same in the
two cases.
219. Distribution of
shearing
stress.
The stress at
any point
consists of two
superposed stress-systems.
In one
system
we have
shearing
stresses
X^
and
F,
of amounts

/iry
and
fiTX
respectively.
In this
system
the
tangential
traction
per
unit of area on the
plane
z
=
const, is
directed,
at each
point, along
the
tangent
to a
circle,
having
its centre at the
origin
and
passing through
the
point.
There must
be
equal tangential
traction
per
unit of area on a
cylindrical
surface
standing
on this
circle,
and this traction is directed
parallel
to the axis of z. In the
second
system
we have
shearing
stresses
X^
and
Fj
of amounts
fnd^jdx
and
fiTdcfj/dy.
The
corresponding tangential
traction
per
unit of area on the
plane
z
=
const, is directed at each
point along
the normal to that curve of
the
family
=
const, which
passes through
the
point,
and its amount is
proportional
to the
gradient
of
<j).
There must be
equal tangential
traction
per
unit of area on a
cylindrical
surface
standing
on that curve of the
family
^fr
=
const, which
passes through
the
point,
and the direction of this traction
is that of the axis of z. These statements
concerning
the stress are inde-
pendent
of the choice of axes of x and
y
in the
plane
of the
cross-section,
so
long
as the
origin
remains the same.
The resultant of the two
stress-systems
consists of
shearing
stress with
components
Xz
and
Y^,
which are
given by
the
equations (3).
If we
put
^y-i(x^
+
f)
=
'^,
(9)
the direction of the
tangential
traction
(X^, Y^)
across the normal section at
any point
is the
tangent
to that curve of the
family
^
=
const, which
passes
through
the
point,
and the
magnitude
of this traction is
;aT9^/9r,
where dv
is the element of the normal to the curve. The curves "^
=
const,
may
be
called
"
lines of
shearing
stress."
The
magnitude
of the resultant
tangential
traction
may
also be
expressed
by
the formula
-l(g-)'-(|-)T
^^-^
and this result is
independent
of the directions of the axes of x and
y.
If
we choose for the axis of a; a line
parallel
to the direction of the
tangential
traction at one
point
P,
the
shearing
stress at P will be
equal
to the value
304 STKENQTH TO RESIST TORSION
[CH.
XIV
at P of the function
(xt {d(f>/da;

y),
and the
^-component
of the traction at
any
other
point
Q
will be
equal
to the value of the same function at
Q.
Now this
function,
being
harmonic,
cannot have a maximum or a minimum
value at P
;
there is therefore some
point, Q,
in the
neighbourhood
of
P,
at
which it has a
greater
value than it has at P. Thus the
^-component
of the
traction at some
point
Q
near to P is
greater
than the traction at P
;
and
the traction at
Q
must therefore be
greater
than that at P. It follows that
the
shearing
stress cannot be a maximum at
any point
within the
prism;
and therefore the
greatest
value of the
shearing
stress is found on the
cylindrical boundary*.
220.
Strength
to resist torsion.
The resultant
shearing
strain is
proportional
to the resultant
shearing
stress,
and the extension and contraction
along
the
principal
axes of the
strain at
any point
are each
equal
to half the
shearing
strain at the
point
;
and thus the
strength
of the
prism
to resist torsion
depends
on the maximum
shearing
stress. Practical rules for the limit of safe
loading
must
express
the
condition
that this maximum is not to exceed a certain value.
Some results of
practical importance
can be deduced from the form of
hydrodynamical
analogy [Article 218(c)]
in which use is made of a circu-
lating
motion with uniform
spin. Suppose
a shaft
transmitting
a
couple
to
contain
a
cylindrical
flaw of circular section with its axis
parallel
to that of
the shaft. If the diameter of the
cavity
is small
compared
witR that of
the
shaft,
and the
cavity
is at a distance from the surface
great compared
with its
diameter,
the
problem
is
very nearly
the same as that of
liquid
streaming past
a
cylinder.
Now we know that the
velocity
of
liquid streaming
past
a circular
cylinder
has a maximum value
equal
to twice the
velocity
of
the
stream,
and we
may
infer
that,
in the case of the
shaft,
the shear near
the
cavity
is twice as
great
as that at a distance. If the
cavity
is a
good
deal
nearer to the surface than to the
axis,
or if there is a semicircular
groove
on
the
surface,
the shear in the
neighbourhood
of the
cavity (or
the
groove)
may
be
nearly
twice the maximum shear that would exist if there were no
cavity (or groove)
f-
If the
boundary
has
anywhere
a
sharp
corner
projecting
outwards,
the
velocity
of the fluid at the corner
vanishes,
and therefore the shear in the
torsion-problem
is zero at such a corner. If the
boundary
has a
sharp
corner
projecting
inwards,
the
velocity
is
theoretically
infinite,
and the torsion of
a
prism
with such a section will be
accompanied by
set in the
neighbourhood
of the corner.
*
This theorem was first stated
by
J.
Boussinesq,
loc. cit. The
proof
in the text will be found
in a
paper by
L. N. G.
Filon,
Phil. Trans.
Roy.
Soc.
(Ser. A),
vol. 193
(1900). Boussinesq
had
supposed
that the
points
of maximum
shearing
stress must be those
points
of the contour
which are nearest to the axis
;
but Filon showed that this is not
necessarily
the case.
t
Cf. J.
Larmor,
Phil.
Mag. (Ser. 5),
vol. 33
(1892).
219-221]
SPECIAL SOLUTIONS OF THE TORSION PROBLEM 305
Saint-Venant in his memoir of 1855 called attention to the
inefficiency
of corners
projecting
outwards,
and
gave
several numerical illustrations of
the diminution of torsional
rigidity
in
prisms having
such corners as
compared
with circular
cylinders
of the same sectional area.
221. Solution of the torsion
problem
for certain boundaries.
We shall now show how to iind the function
(j)
from the
equation (4)
and
the condition
(5)
when the
boundary
of the section of the
prism
has one or
other of certain
special
forms. The
arbitrary
constant which
may
be added
to
(f>
will in
general
be
adjusted
so that
(f>
shall vanish at the
origin.
(a)
The circle.
If the
cylinder
of circular section is twisted about its axis of
figure,
<f)
vanishes,
and we have the solution
already given
in Article 86
(d).
If it
is twisted about
any parallel
axis
^
does not
vanish,
but can be determined
by
the method
explained
in Article 218. In the latter case the cross-sections
are not
distorted,
but are
displaced
so as to make an
angle differing slightly
from a
right angle
with the axis,
(6)
The
ellipse.
The function
i|r
is a
plane
harmonic function which satisfies the condition
i/r

Kar"
+
3/2)
=
const, at the
boundary a^/a"
+
y^/b''
=1. If we assume for
i^
a form A
{a?

y%
we find the
equation
{^-A)a^
=
{\
+
A)h\
It follows that we must have
,
la*-6,
, ox
,
a^-h-
,,^
^=2a^Ty^(^-2/)'
'^
=
-aMT^"y
^'^^
It is clear that this solution is
applicable
to the case of a
boundary
consisting
of two concentric similar and
similarly
situated
ellipses.
The
prism
is then a hollow
elliptic
tube.
(c)
The
rectangle*.
The boundaries are
given by
the
equations x=a,y=h.
The function
i/r
differs
by
a constant from
\ {y-
+
a?)
when
x=a
and
h>y>-h\
it
differs
by
the same constant from
| (a;-
+
h-)
when
y
=
b
and a > a; >
-
a.
We introduce a new function
yjr' by
means of the
equation
Then
yjr'
is a
plane
harmonic function within the
rectangle;
and we
may
take
-^jr'
to vanish on the sides
y=
b,
and to be
equal
to
tf
-
6" on the sides
x
=
a.
Since the
boundary
conditions are not altered when we
change
x
into X or
y
into
-
y,
we seek to
satisfy
all the conditions
by assuming
for
A|r'
a formula of the
type
2J.,
cosh m,r cos
my.
The conditions
which hold at
*
The
corresponding hydrodynamioal problem
was solved
by Stokes, Cambridge
Phil. Soc.
Trans.,
vol. 8
(1843)
=
J/a(/i. aiid
Phys. Papers,
vol.
1, p.
16.
L. E.
20
306 SPECIAL SOLUTIONS OF
[CH.
XIV
the boundaries
y
=
+b
require
tliat m should be
| {2n
+
1) Tr/b,
where n is an
integer.
If we assume
that,
when h>
y
>

h,
the function
y-

}f can be
expanded
in a series
according
to the form
t-h^^
2^...,.
cosh
<>^^
COS
(^-^^:tlW
.
we
may
determine the coefficients
by multiplying
both members of this
equation by
cos
{(2n
+
l)7ry/26j,
and
integrating
both members with
respect
to
y
between the extreme values

b and b. We should thus find
A^+,
cosh
^
5^^^
=
(-)"+'
46=
;
26
^ '
(2n
+
l)7r'
This
process suggests
that when b>
y>

b the sum of the series
^
^K-Trl^Vos^^^^^^
(13)
=o WJ
(2n
+
1/
26
is
y"

b^. We cannot at once conclude that this result is
proved by
Fourier's
theorem*,
because a Fourier's series of cosines of
multiples
of
Tryjib represents
a function in an interval
given by
the
inequalities
26 >
y
>

26,
and the
value
y-

U of the function to be
expanded
is
given only
in the interval
b>y>

b. If the Fourier's series of cosines contains uneven

multiples
of
'rry/2b only,
the
sign
of
every
term of it is
changed
when for
y
we
put
2b

y
;
it follows
that,
if the series
(13)
is a Fourier's series of which the sum is
y"

6= when b>
y>0,
the sum of the series when 2b
>y
>b is b"

(26

yy.
Now we
may
show that the Fourier's series for an even function of
y,
which
has the value
y''

when b>
y
>0,
and the value 6-

(26

yf
when
2b>y>b,
is in fact the series
(13).
We
may
conclude that the form
of
-^
is
(^n+2W
,7 w . ., f2V^ (-)"
26
(2n+l)7ry
and hence that
W =o
(2n
+
If
^,
(2n
+
1)
7ra
cosn
2^
.
,
(2n
+
l)7ra;
f2y^
(-)"
"
26 .
(2n+l)7ryf
d,
=
-wy
+
U-{-)
S
^

^
tt; ^r:

sin
^^
-,
'^ ^ '
.
...(14)
^ -^
WJ =o {in
+
1>^ _^
(2n
+
l)7ra
26
cosn
2^
222. Additional results.
The torsion
problem
has been solved for
many
forms of
boundary.
One method is to
assume a
plane
harmonic function as the function
\|/-,
and determine
possible
boundaries
*
Observe,
for
example,
that the Fourier's series of cosines of
multiples
of
7rj//2l
which has
the sum
y-
-
W
throughout
the interval
2b>y>
-2b is
,79
1662

(-) niry
TT''
_i
' 26
t
The
expression
for
ip
must be unaltered when x and
y,
a and
6,
are
interchanged.
For an
account of the identities which arise from this observation the reader is referred to a
paper by
F.
Purser, Messenger of Math.,
vol. 11
(1882).
221,
222]
THE TORSION PROBLEM
807
from the
equation y^-i {x^ +y^)
=
const. As an
example
of this method we
may
take
i\r
to
be ^
(^
-
Zxy^)
;
if we
put
A=
-
l/Qa,
the
boundary
can )je the
equilateral triangle,
of
altitude
3a,
of which the sides are
given by
the
equation
Other
examples
of this method have been discussed
by
Saint-Venant.
Another method is to use
conjugate
functions
|, >;
such that
|
+
i>)
is a function of
A'+iy.
If these functions can be chosen so that the
boundary
is made
up
of curves
along
which either
|^
or
r;
has a constant
value,
then
yjf
is the real
part
of a function of
^
+
ii),
which has a
given
value at the
boundary
;
and the
i)roblem
is of the same kind as the
torsion
problem
for the
rectangle.
We
give
some
examples
of this method :

(i)
A sector of a
circlet,
boundaries
given hy
r
=
0,r=a,
6=
3.

We find
where
^,,,
=
(_)...i[^__L
If we write
re'9=cu;,
then
where
j
x
] ^1,
and tan~'.r'^'' denotes that branch of the function which vanishes with x.
In case
7r/2/3
is an
integer greater
than 2 the
integrations
can be
performed,
but
when
jr/2^
=
2 the first two terms become
infinite,
and their sum has a finite
limit,
and we
find for a
quadrantal cylinder
>f'-i<^=
-
x^
log
X
+
ta.n
-
^
X-
+
^
(.r-
-A
log (\ +.%*)
I.
For a semi-circular
cylinder
^-''^-~[i--'-'{-+iyi^(r'-+x-^-'')'-s\^]-
(ii)
For a curvilinear
rectangle
bounded
by
two concentric circular arcs and two
radii,
we use
conjugate
functions a and
/3,
which are
given by
the
equation
x
+
ti/
=
ce'^'^^^
;
we take the outer
radius,
a to be ce"' and the
inner,
b to be ce"""
(so
that c is the
geome-
trical mean of the
radii),
and we take the
bounding
radii to be
given by
the
equations
i3=^o-
We find
'^^-^"^^'"ctll+^^^WlA*,
where
*
=
.
,
('2n
+
\)na
,
(2n
+
1
)
tto'i
-^^
^"0
,(2.+
l).ao
+
-"'^-
^"O
.
,
i2,J)J
'
and
An=-.
"{(2-|-l)7r-4^o} (2n
+
l)
^
{(2+ l)7r
+
4^o;
*
See
Figures
23 and 24 in Article 223.
t See A. G.
Greenhill, Messenger of
Math.,
vol. 8
(1877), p.
89,
and vol. 10
(1880), p.
83.
202
308 DISTORTION OF THE CROSS-SECTIONS
[CH.
XIV
(iii)
When the twisted
prism
is a hollow
shaft,
the inner and outer boundaries
being
circles which are not
concentric,
we
may
use the
conjugate
functions
|, rj
determined
by
the
equation
j;
+
,y
=
c
tan^d
+
i,);
and,
if
i)
=
a
represents
the outer
boundary,
and
;)=/3
the
inner,
we
may prove*
that
I ^9^/ ,,e""^coth/3sinhi(n-a)
+
e"""cothasinhre(/3-n)
^
=i
smhwO-a)
^
(iv)
When the boundaries are confocal
ellipses
and
hyperbolas
we
may
use the
conju-
gate
functions
|, ;;
determined
by
the
equation
In the case of a hollow
tube,
of which the section is bounded
by
two confocal
ellipses
^0
and
fi,
we
may prove
t that
223.
Graphic expression
of the results.
(a)
Distortion
of
t/ie cross-sections.
The curves
=
const, are the contour lines of the surface into which
any
cross-
section of the
prism
is distorted. These curves were traced
by
Saint-Venant for a number
of forms of the
boundary.
Two of the results are shown in
Fig.
22 and
Fig.
23. In both
Fig.
22.
cases the cross-section is divided into a number of
compartments,
4 in
Fig. 22,
6 in
Fig. 23,
and
changes sign
as we
pass
from
any compartment
to an
adjacent
com-
partment,
but the forms of the ciu'ves
(^=const.
are unaltered. If we think
[of
the axis
of the
prism
as
vertical,
then the curved surface into which
any
cross-section is strained
lies above its initial
position
in one
compartment
and below it in the
adjacent compart-
ments. Saint-Venant showed that the sections of a
square prism
are divided in this
way
into 8
compartments by
the
diagonals
and the lines drawn
parallel
to the sides
through
the centroid. When the
prism
is a
rectangle,
of which one
pair
of
opposite
sides is much
longer
than the other
pair,
there are
only
4
compartments separated by
the lines drawn
parallel
to the sides
through
the centroid. The
limiting
case between
rectangles
which
are divided into 4
compartments
and others which are divided into 8
compartments
*
H. M.
Macdonald, Cambridge
Phil. Soc.
Proc,
vol. 8
(1893J.
t
Of. A. G.
Greenhill, Quart.
J.
of Math.,
vol. 16
(1879).
Other
examples
of
elliptic
and
hyperbolic
boundaries are worked out
by Filon,
loc. cit.
p.
304.
222,
223]
OF A TWISTED PRISM
309
occurs when the ratio of
adjacent
sides is 1-4513. The
study
of the
figures
has
promoted
comprehension
of the result that the cross-sections of a twisted
prism,
of non-circular
section,
do not remain
plane.
y
Fig.
23.
(6)
Lines
of shearing
stress.
The distribution of
tangential
traction on the cross-sections of a twisted
prism
can be
represented graphically by
means of the lines of
shearing
stress. These lines are deter-
mined
by
the
equation
They
have the
property
that the
tangential
traction on the cross-section is directed at
any
point along
the
tangent
to that curve of the
family
which
passes through
the
point.
If
the curves are traced for
equidifFerent
values of
c,
the
tangential
traction at
any point
is
measured
by
the closeness of consecutive curves.
In the case of the
prism
of
elliptic
section
^-^{x^-f/)
=
-
{xW +fa^)/{a^
+
62),
and the lines of
shearing
stress are therefore concentric similar and
similarly
situated
ellipses.
In the case of the
equilateral triangle
V'
-
i (^^ +f)
=
-
J
a
-
1
[.r3
-
3x^^
+
3aa;2
-|-
3a
f],
and the lines of
shearing
stress are of the forms shown in
Fig.
24.
Fig.
24.
310 CALCULATION OF THE TORSIONAL
[CH.
XIV
224.
Analogy
to the form of a stretched membrane loaded
uniformly*.
Let a
homogeneous
membrane be stretched with uniform tension T and fixed at its
edge.
Let the
edge
be a
given
curve in the
plane
of
x, y.
When the membrane is
subjected
to
pressure,
of amount
p per
unit of
area,
it will
undergo
a small
displacement
z,
and is a function of x and
y
which vanishes at the
edge.
The
equation
of
equilibrium
of the membrane is
The function
^Tzjp
is determined
by
the same conditions as the function * of Article
219,
provided
that the
edge
of the membrane is the same as the
bounding
curve of the cross-
section of the twisted
prism.
It follows that the contour lines of the loaded membrane
are identical with the lines of
shearing
stress in the cross-section of the
prism.
Further the torsional
rigidity
of the
prism
can be
represented by
the volume contained
between the surface of the loaded membrane and the
plane
of its
edge.
We have seen
already
in Article 216 that the torsional
rigidity
is
given by
the
equation
^=.//((g-,)V(*..-)')**
or,
in terms of
1',
we have
-'//{(iy-()w_^
='^/*'^
^-/^
//* (3
+
?)
'^'^y
=
2^
I
{^dxdy,
since * vanishes at the
edge
and
^
+
^
+
2=0. It follows that the volume m
question
is
{pl4nT)C.
225.
Twisting couple.
The
couple
can be evaluated from
(6)
of Article 216 when the function
(j>
is known. We shall record the results in certain cases.
(a)
The circle.
If a is the radius of the circle the
twisting couple
is
^/xTira*
(15)
(b)
The
ellipse.
From the value of
(j>
in Article 221
(b)
we find that the
twisting couple
is
fiTTra'b'Ka'
+
b') (16)
(c)
The
rectangle.
From the result of Article 221
(c)
we find for the
twisting couple
the
formula
firiab (a=
+
b')
-
iir^ab {a'
-
b'-)
+
^/j-rb"-
l^-j
jji^x^-y^^
dxdy,
*
The
analogy
here described was
pointed
out
by
L.
Prandtl,
Phys.
Zeitschr.,
Bd. 4
(1903),
it affords a means of
exhibiting
to the
eye
the distribution of stress in a twisted
prism.
224,
225]
RIGIDITY OF A PRISM
311
where <I> stands for the series

(_)
Sinn
2fc
sin
2fe-
=o(2n
+
l) ,
(271+1)
Tra
Taking
one term of the
series,
we have a term of the
integral,
viz. :
(rzT 2L f fl -v.
(2w
+
l)7ra; (2
+
l) 7ry
(2/1
+
1)=
cosh
{(2n
+
l)7ra/26}
26
jjr^^"^
26
'^^^
W
"
,{2n
+
l)irx
.
(2n
+
l)7rw1
, ,
_
y
cosh
^

sin
^
26
I
^
Now
r ^Ginh
^^^+^^^'^
i^-
^^
r2acosh(2^+^i^
J.
a
26
'^'^
(2ft
+
1)
TT
L
26
2^
^2sinh(?!^
(2w+l)7r
cosh
^-"^^^^c^-.-^^V
2sinh(?^+l)^,
(
26
(2n+l)7r
26

(2n
+
l)7rM,
26
, ,,
<^^
26^^^
=
(2nnhr2(-ir,
*
.
(2rt
+
l)7r2/,
86^
/ ,^
26
J'
/
Hence the
twisting couple
is
equal
to
/4\*

1
/4\

1
^firab' +{-)
arab^ S
t^t ^,

ut6-'
(
-
)
S
T^i rx;
tanh
(2n
+
l)7ra
26
Since 2
{2n+ 1)-*
is
7rV96,
we
may
write down the value of the
twisting

=
o
couple
in the form
The series in
(17)
has been evaluated
by
Saint-Venant for numerous values of the
ratio a : b. When a > 36 it is
very nearly
constant,
and the value of the
twisting couple
is
nearly equal
to
fiTa63 (3'361)
. For a
square
the
couple
is
(4'4985) liza*.
The
twisting couple
was also calculated
by
Saint-Venant for a number of other forms
of section. He found that the resistance of a
prism
to torsion is often
very
well
expressed
by replacing
the section of the
prism by
an
ellipse
of the same area and the same moment
of inertia*. The formula for the
twisting couple
in the case of an
ellipse
of area A and
moment of inertia / is
iitA^JAtt^I.
*
Saint-Venant, Paris,
C.
li.,
t. 88
(1879).
312 TORSION OF A PRISM OF ^OLOTROPIC MATERIAL
[CH.
XIV
226. Torsion of
seolotropic prism.
The
theory
which has been
explained
in Article 216 can be extended to a
prism
of
seolotropic
material when the normal section is a
plane
of
symmetry
of structure.
Taking
the axis of z to be
parallel
to the
generators
of the
bounding surface,
we have the strain-
energy-function expressed
in the form
belonging
to
crystalline
materials that
correspond
with the
group C^ (Article 109).
The
displacement being expressed by
the formulae
(1),
the
stress-components
that do not vanish are
X^
and
Y^,
and these are
given by
the
equations
The
equations
of
equilibrium
are
equivalent
to the
equation
which must hold over the area of the cross-section
;
and the condition that the
bounding
surface
may
be free from traction is satisfied if the
equation
C66gf
cos(^,
>')+C44^
(cosy, v) +c^
|^
COS
(a',
i/)+g|
cos
(y,
p)j
=
C55ycos(A-, v)-C44.rcos(y, v)
-
c^, {x
cos
{x, v)- y cos(y, v)}
holds at all
points
of the
bounding
curve.
Exactly
in the same
way
as in the case
of
isotropy,
we
may prove
that the differential
equation
and the
boundary
condition
are
compatible,
and that the tractions across a normal section are
equivalent
to a
couple
of moment
rjj\citX^+c^f-2Ct^xy+Ct^x^-c^,y+Ci^(^x
g|-y
^j^dxdi/.
The
analysis
is
simplified considerably
in case
C45
=
0. If we
put
L for
C44
and M for
C55,
the differential
equation may
be written
and,
if
f{x, y)
=
is the
equation
of the
bounding curve,
the
boundary
condition
may
be written
dx dx
dy dy
"
dx
Zy'
We
change
the variables
by putting
/L+M
,
/L+M
,, ,
L+M
^=^\/^^' -^^-w^s^' '^^'^27{m-
Then <b' satisfies the
equation
The
equation /(a;, y)=0
becomes
P{x',y')=0,
where
Fix',i/)^f{x'
J^^, y'J^^,
226]
AND RECTANGULAR CROSS-SECTION 313
and the
boundary
condition is tranaformed into
da/ da/
*
dy' Sy'
~^
da/
di/'
which is
^
=/
cos
(.r', v)
-
x' cos
{y\ v),
if dv is the element of the normal to the transformed
boundary.
Thus
</>
can be found for
any boundary
if
<^'
can be found for an
orthographic projection
of that
boundary
;
and
the
problem
of
finding <^'
is the
simple
torsion
problem
which we considered before.
As an
example
we
may
take a
rectangular prism
with boundaries
given by
x=
a,
y=iVh.
We should find that the formula for
<^
is
(2ra
+
l)jrxv'.
IM^^'h^

(-)"
"
%hJM
sinh
i
I
cosh
-
26
V^
and that the
twisting couple
is
expressed by
the formula
1
3
a^L \irj =o(2?i+l)S
26
V-^ J
This formula has been used
by
W.
Voigt
in his researches on the elastic constants of
crystals. [See
Article
113.]
CHAPTER XV.
THE BENDING OF A BEAM BY TERMINAL TRANSVERSE LOAD.
227. Stress in a bent beam.
In Article 87 we described the state of stress in a
cylinder
or
prism
of
any
form of section held bent
by
terminal
couples.
The stress at a
point
consisted of
longitudinal
tension,
or
pressure, expressed by
the formula
tension
=

Mxjl,
where Mis the
bending
moment,
the
plane
of
{y, z)
contains the
central-line,
the axis of x is directed towards the centre of
curvature,
and / is the moment
of inertia of the cross-section about an axis
through
its centroid at
right
angles
to the
plane
of
bending.
In Article 95 we showed how an extension
of this
theory
could be made to the
problem
of the
bending
of a
rectangular
beam,
of small
breadth,
by
terminal transverse load. We found that the
requisite stress-system
involved
tangential
traction on the cross-sections as
well as
longitudinal
tensions and
pressures,
but that the
requisite
tension,
or
pressure,
was determined in terms of the
bending
moment
by
the same
formula as in the case of
bending by
terminal
couples.
This
theory
will now
be
generalized
for a beam of
any
form of section*.
Tangential
tractions on
the elements of the cross-sections
imply equal tangential
tractions,
acting
in
the direction of the
central-line,
on elements of
properly
chosen
longitudinal
sections,
the two
tangential
tractions at each
point constituting
a
shearing
stress. It is natural to
expect
that the
stress-system
which we seek to
determine consists of
longitudinal
tensions,
and
pressures,
determined as
above,
together
with
shearing
stress,
involving suitably
directed
tangential
tractions on the elements of the cross-sections. We shall
verify
this antici-
pation,
and shall show that there is
one,
and
only
one,
distribution of
shearing
stress
by
means of which the
problem
can be solved.
*
The
theory
is due to Saint-Venant. See
Introduction,
footnote
50,
and
p.
20.
227, 228]
SHEARING STRESS IN A BENT BEAM
515
228. Statement of the
problem.
To fix ideas we take the central-line of the beam to be
horizontal,
and
one end of it to be
fixed,
and we
suppose
that forces are
applied
to the cross-
section
through
this end so as to
keep
the beam in a
nearly
horizontal
position,
and that forces are
applied
to the cross-section
containing
the other end
in such a
way
as to be
statically equivalent
to a vertical load If
acting
in a line
through
the centroid of the section. We take the
origin
at the fixed
end,
and the axis of z
along
the
central-line,
and we draw the axis of x
vertically
downwards. Further we
suppose
that the axes of a; and
y
are
parallel
to the
principal
axes of inertia of the cross-sections at their centroids. We denote
the
length
of the beam
by
I,
and
suppose
the material to be
isotropic.
We
consider the case in which there are no
body
forces and no tractions on the
cylindrical bounding
surface.
-2
Mi
W
Fig.
25.
The
bending
moment at the cross-section distant z from the fixed end is
Wijj

z^
We assume that the tension on
any
element of this section is
given by
the
equation
Z,
=
-W{l-z)xlI (1)
where / stands for the
integral
I
Ix^dxdi/
taken over the area of the cross-
section. We assume that the stress consists of this tension
Z^
and
shearing
stress
having components
X^
and
Y^,
so that the
stress-components
Xx,
Yy,
Xy
vanish
;
and we seek to determine the
components
of
shearing
stress
X^
and
Fj.
Two of the
equations
of
equilibrium
become
dXJdz
=
0,
dY^jdz
=
0,
and it
follows that
Zj
and
F,
must be
independent
of z. The third of the
equations
of
equilibrium
becomes
BX,m_^]^^^
:
(2)
ox
dy
1
The condition that the
cylindrical bounding
surface is free from traction is
XiCos,{x, v)+ Y^cos{y, i/)
=
(3)
The
problem
before us is to determine
X^
and
Y^
as functions of x and
y
in accordance with the
following
conditions :

(i)
The differential
equation (2)
is satisfied at all
points
of the cross-
section of the beam.
316 DETERMINATION OF THE SHEARING STRESS
[CH.
XV
(ii)
The condition
(3)
is satisfied at all
points
of the
bounding
curve of
this section.
(iii)
The tractions on the elements of area of the terminal cross-section
{z
=
I)
are
statically equivalent
to a force
W,
directed
parallel
to the axis of
a;,
and
acting
at the centroid of the section.
(iv)
The
stress-system
in which
X^
=
Fj,
=
Xy
=
0, Z^
is
given by (1),
and
Xg, Fj
satisfy
the conditions
already
stated,
is such that the conditions of
compatibility
of
strain-components (Article 17)
are satisfied.
229.
Necessary type
of
shearing
stress.
The assumed
stress-system
satisfies the
equations
X,
=
Yy
=
Xy
=
0, Z,
=
-W{l-z)xlI,
^^J-^
=
o,
and
consequently
the
strain-components satisfy
the
equations
_
_
W{l-z)x
_ _ _
n
^^^*
-
^^v'
-
n
where E and a denote
Young's
modulus and Poisson's ratio for the material.
The
equations
of
compatibility
of the
type
yy
zz
^
^yz
dz^
dif dydz
are satisfied
identically,
as also is the
equation
2
^^^"
=

-I-
-
dxdy
dz
\
dx
dy
dz
The
remaining equations
of
compatibility
of this
type
become
/^fyf _
5ezx\
^ f^
_9 /3e^ _
^e
\
^
_
2a- W
dx\dx
dy)~
'
dy\dx dy
}
~
EI
'
From these
equations
we deduce the
equation
deyz
dezx
_
a
2o-Tr
'd^~JJ~'^''~
EI
^'
where 2t is a constant of
integration
;
and from this
equation
it follows that
e^z
and
Czx
can be
expressed
in the forms
eyz-rx^^,
ezx
=
-ry
+
^+^f,
(4)
where
^o
is a function of x and
y.
On
substituting
from these
equations
in the formulae
X^
=
/j^zx
and
Yz
=
/xgyj,
and
using
the relation
/x
=
^El{\
+
a),
we see that
equation (2)
takes the form
8'<^o y</,o 2(l
+
.r)TF
daf
"^
82/
^
El
*
-
"'
228,
229]
IN A BENT BEAM
317
and condition
(3)
takes the form
^
=
T{y
cos
(, v)-x
cos
(y, v)}
-
^^j
y^coa(a;, v).
These relations are
simplified by putting
W
<f>o
=
t4>
-
^
{X
+
i<Ta^
+
(I
+
i<T)xf\ (5)
Then
^
is the torsion function for the section
(Article 216),
and
;\;
is a
function which satisfies the
equation
S-lr--"
<'
at all
points
of a
cross-section,
and the condition
?J
=
-
{^o-*^
+
(1
-
i<7)2/}
cos
(x, v)-{2
+
a)xycos(y, v) (7)
at all
points
of the
bounding
curve. The
compatibility
of the differential
equation (6)
and the
boundary
condition
(7)
is shown
by observing
that,
since
the
integral
1
1
xdxdy
taken over the cross-section
vanishes,
the
integral
of the
right-hand
member of
(7)
taken round the
boundary
vanishes. The
problem
of
determining
the function
t^
from
equation (6)
and condition
(7) may
be
called the
"
flexure
problem
"
for the section.
When the functions
<^
and
;^
are known the
shearing
stresses
X^
and
Y^
are known in the forms
...(8)
The terms that contain t*
express
a
system
of tractions on the elements
of area of the
cross-section,
which are
statically equivalent
to a
couple
about
the axis z of moment
and the terms which contain W would
give
rise to a
couple
about the same
axis of moment
2(rT^//K^-4-^^^-^^^^-^'+*'^^^^}'^''^-
We
adjust
t so that the sum of these
couples
vanishes.
The tractions on the elements of area of a cross-section are
statically
equivalent
to a certain force at the centroid of the section and a certain
*
They
are of the same form as the tractions in the torsion
problem.
318 STRESS AND DISPLACEMENT
[CH.
XV
couple.
We show that the force is of
magnitude
W and is directed
parallel
to the axis of
x,
and that the
couple
is of moment
W{l

2)
and has its axis
parallel
to the axis of
y.
These statements are
equivalent
to the
equations
jjx^dxdy=-W,
{{\\dxdy
=
Q,
jjz,dxdy
=
0,
(9)
and
jjyZ.dxdy
=
0,
jj-
xZ.dxdy
=
W
(I
-
z), (((x Y,
-
yX^) dxdy
=
0. . .
.(10)
Now
by (2)
and
(3)
we
may
write down the
equations
=
TT
+
la;
{Xj
cos
{x, v)
+
Y^
cos
(y, v')]
ds
=
F.
In like
manner,
observing
that
jjxydxdy
vanishes,
we
may prove
the
second of
equations (9).
The third of these
equations
and the first two of
equations (10)
follow at once from the formula
(1)
for
Z^,
and the constant t
has
already
been
adjusted
so that the third of
equations (10)
shall be
satisfied.
The functions
<f>
and
%
are each
determinate,
except
for an additive
constant which does not affect the stress. We have therefore shown that the
problem
stated in Article 228 admits of
one,
and
only
one,
solution.
230. Ponnulse for the
displacement.
The
displacement
can be deduced from the strain without
determining
the forms of
(f>
and
)(.
The details of the work are as follows :

We have the
equation
dw__
W{l-z).v
'az~ EI
'
from which we deduce the
equation
Wl 1 W
>=
-;gFj^2
+
2 ^a:^2
+
<^', (11)
where
<^'
i.s a function of x and
y. Again,
we have the
equations
Wl W
El'"'' El'
du_Wl
1 W
2
aW
8(^0 d4>'
3 ~Ei^ 2E1^
'^^
EI^
+
a.r
"
Jx'
of which the second is obtained from
(11)
and the second of
(4).
These two
equations
are
compatible
if
Tx'=''-EI^-''El'''
229,
230]
Again,
we have the
equations
and these are
compatible
if
IN A BENT BEAM
dv Wl W
dy
EI EI
'
dz
dy dy
'
319
Further, by differentiating
the left-hand member of the
equation
^
+
^
=0 with
respect
to
z,
we obtain the
equation
dxdy
EI''
The three
equations
for
<^d
-
<^'
show that we must have
W
where
a, 0, y'
are constants. When we substitute for
<j>(,
from
(5)
we find the
following
expression
for <t>' :

W
The
displacement
w is now determined. When we substitute for
(j>'
in the
equations
for
dujcz
and
dv/dz,
we obtain the
equations
i=-^^^El^''-^''-
^-^ (-^^ -2'')J +^,
dv W
Tz^^^'-'^El'^-''-
From the
equations
for
dujdx
and
3m/0j
we obtain the
following
form for u :

W
u^-Tyz
+
-j^j[\l{z'^
+
^x^-\z,T{x''--f-)-lz^]+^z^F^{y),
where
/\ (?/)
is an unknown function of
y.
In like manner wo find the
following
form
for v.

W
v-rzx
+
^(T{l-z)xy-az
+
F.2{x),
where
F^ix)
is an unknown function of x. Since
duldy
+
dv/dx=0,
the functions
Fj, F.^
satisfy
the
equation
8^1 , dF^ ,
Tl'

-d^+^
+
'^El'^^'"'
and we must have
F,{y)=- io-
1^
?/-yi/
+a', F, (.r) ^yx+ff,
where
a, /3', y
are constants of
integration.
We have now found the
displacement
in the form
W
w
=
-T2/^+~[i(Z-^)cr(^-y=)
+
ii^'-K]-7y
+
^^
+
'.\
w ,
v
=
TZx+
r-ja{l-z)!}j
+
yx-az
+
/3 ,
till
^...(12)
w-
W
=
T<^
-
-^
[a; i}z
-
^z")
+
X
+
.*/]
-
^x
+
ay
+
7',
EI
320 SOLUTION OF THE PROBLEM OF FLEXURE
[CH.
XV
in which
a, jS,
7,
a, ^, 7'
are constants of
integration.
These
equations give
the most
general possible
form for the
displacement
(m,
v,
w)
when the stress
is determined
by
the conditions stated in Article 228.
The terms of
(12)
that contain
a, /3, 7,
a', ^', 7' represent
a
displacement
which would be
possible
in a
rigid
body,
and these constants are to be
determined
by imposing
some conditions of
fixity
at the
origin.
(Cf
Article
18.)
We have
supposed
that the
origin
is
fixed,
and we must therefore have
a!
=
0, /9'
=
0. We
shall,
in
general, suppose
that the additive constants
in
the
expressions
for
j>
and
;^
are determined so that these functions vanish at
the
origin.
Then we must also have
7'
=
0.
Besides
fixing
a
point,
we
may
fix a line
through
the
point.
We shall
suppose
that the linear element
which,
in the unstressed
state,
lies
along
the
axis of
y
retains its
primitive
direction. Then we must have a
=
0,
7
=
0.
Besides
fixing
a
point,
and a linear element
through
the
point,
we
may
fix a surface element
through
the line. The value of the constant
/3 depends
upon
the choice of this element. If we choose the element of the cross-
section,
we must have
dwjdx
=
at the
origin.
If we choose the element of
the neutral
plane
(i.e.
the
plane
x
=
0),
we must have
dujdz
=
at the
origin.
In the former case the central element of the cross-section at the fixed end
remains vertical
;
in the latter case the element of the central-line at the
fixed end remains horizontal. There is no reason for
assuming
that in all
practical
cases either of these conditions
holds;
most
probably
different
values of
^
fit the circumstances of different
particular
cases.
231. Solution of the
problem
of flexure for certain boundaries.
We shall now show how to find the function
x
^om the
equation
(6)
and
the condition
(7)
when the
boundary
of the section of the beam has one or
other of certain
special
forms. The constant which
may
be added to
%
will
generally
be chosen so that
x
vanishes at the
origin.
(a)
The circle.
The
equation
of the
bounding
curve is a;-
-t-
y^
=
a". In terms of
polar
coordinates
(r, 6)
the
boundary
condition at the curve r
=
a is
?2^
= -
a^ cos ^
jAo-
cos' ^
+
(1
-
lo-)
sin=
^1
-
a' sin 6
{(2
+
a)
sin 6 cos
6],
or
or
^
=
-(f-(-ia)a=cos^
+
ia=cos3^.
or
Since
X
is a
plane
harmonic function
within the circle r
=
a,
we must have
p^
=
-
(f
4-
1
C7-)
aV cos ^ 4-
i
r' cos
3^,
or
5^
=
-(f
+
i<7)aa;
+
i(a;'-3a;2/')
(13)
and denote
by
h. The value of / at a
point
on the
boundary
is
230, 231]
FOR CERTAIN FORMS OF SECTION
821
(6)
Concentric circles.
The beam has the form of a hollow tube. If
a^
ia the radius of the outer
circle,
and
Oj
that of the
inner,
we
may prove
that
x
is of the form
X=-ii
+
i<r)
|(aoHV)''
+
'
^[c<^+ircos3d+con8t.
(14)
In this case we cannot
adjust
the additive constant so as to make
x
vanish at the
origin,
but the
origin
is in the
cavity
of the tube.
(c)
The
ellipse.
The
equation
of the
bounding
curve ia
af/a'
+
y^/b'^
=
1. We introduce
conjugate
functions
^, tj by
means of the relation
x+
i.y
=
(a"
-
6')*
cosh
(f
+
iv),
d{x+ ly)
p/ab,
where
p
is the central
perpendicular
on the
tangent
at the
point.
The
boundary
condition
may
be written
or
^
=

6 cos
7; {i
a-a^ cos=
;
+
(1

| a-)
6^ sin''
'?)

sin
r)(2
+
<r)
ah sin
17
cos
rj ;
and this is the same as
If
=
-
[(i
+
f '^)
'^ +
(?
-
i-) ^']
'^o^
'?
+
[(5
+
i-)
'^
+
(i
-
i?) ^']
COS
37,.
Hence we must have
cosh c
X
=
-
[(i
+
i'^)
o?b +
(i
-
i<7) 6]
^^^
cos
n
+
H(i
+
i<^)
<^'h +
(i
-
la) 63]
^^1
cos
37;,
where
f
denotes the value of
^
at the
boundary,
so that
{a?
-
b^y>
cosh
f

=
a,
(a=
-
6=^)*
sinh
^,
=
b.
Now we have
{x
+
ij/y
=
(a=
-
60' i {cosh
3
(^
+
tT?)
+
3 cosh
(f
+
ii,)},
so that 4
^_^|
-
3
(^j:^
=
cosh
3?
cos
St,.
Also we have sinh
3^o
=
4 sinh"
f
+
3 sinh
Sa-
lience we find
X
=
-i{h
+
%<^)or
+
(\-ho-)b"-]x
, .. , X ., , ,,,,4(^-3^2/')-3:.(a-6')
L. E.
21
322 SOLUTION OF THE PROBLEM OF FLEXURE
[CH.
XV
In the above
analysis
we have
proceeded
as if a were
greater
than
b,
but
it is
easy
to
verify
that the final result holds also when b > a. In case b
=
a
it reduces to the result
already
found for the circle.
(rf) Confocal ellipses.
By
an
analysis
similar to the above the
problem might
be solved for a section bounded
by
two confocal
ellipses.
The result could not be
expressed rationally
in terms of x and
y.
Taking |(,
and
^j
to be the values of
^
which
correspond
with the outer and inner
boundaries,
and
writing
c for
(a^-b^)^,
we
may
show that
;^
=
cco8i7[(i- Jo-)ooshf-(}+|<7){cosh
|cosh licosh (lo+li)cosh f
-
sinh
lo
sinh
^j
sinh
(,$o
+
ii)
sinh
^}]
j-^^<.Q fj-o^^^'it /JLJ.1
N
sinh
^0
cosh 3
(^
-
g,)
-
sinh
g^
cosh 3
(^p
-
g
)"|
+
c3cos3,|^^cosh3|-(T^
+
Ja)
3sinh3(|-f,) J'
^^^^
(e)
The
rectangle.
The
equations
of the boundaries are x=
a,
y
=
b.
The
boundary
condition at x
=
a
is
||
=
-{i<7a=
+
(l-i<r)2/^}, (b>y>-b).
The
boundary
condition at
y
=
+ 6 is
^
=
+
(2
+
a)bx, {a>x>- a),
dy
We introduce a new function
)^ by
the
equation
x'
=
%-i(2
+
<r)(^-3a^') (17)
Then
')^
is a
plane
harmonic function within the
rectangle, dx/dy
vanishes at
y
=
b,
and the condition at a;
=
+ a becomes
^^'
=
-(l
+
cr)a^
+
<r/.
Now when
b>y
>

b the function
y'^
can be
expanded
in a Fourier's series
as follows :

2/
=
?
+

^
^-
cos
-ff-
.
Hence
^
^^^ ^^
expressed
in the form
. , mrx
-J
. smh

,
y;:^[-{l+a)a-'
+
lah']x
+
a~ I
^
Ji-
cos
-p,
...(18)
cosh-^
and,
by
means of this and
(17), x
can be written down.
(/)
Additional results.
The results for the circle and
ellipse
are included in the formula
X
=
Ax
+B
(3^-3x2/^);
231,
232]
FOK CERTAIN
FORMS OF
SECTION
323
the solution for the
ellipse
was first found
by adjusting
the constants A and B of this
formula,
and several other
examples
of the same method wore discussed
by
Saint-Venant.
Among
sections for which the
problem
is solved
by
this formula we
may
note the curve of
which the ordinate is
given by
the
equation
y
=

6
I
(1
-
.rVaT
I
. (a>x>
-
a).
'
The
corresponding
function
x
is
When
<r=i
the above
equation
becomes
a;2/a2+y</6<=l.
The curve is shown in
Fig.
2a
for the case where a
=
26.
Fig.
27.
As another
example
we
may
observe that the formula*
solves the
problem
for a section bounded
by
two arcs of the
hyperbola
x^
(1 -(-(7-)-yV=a*
and two
straight
lines
y=
a.
The section is shown in
Fig. 27,
o-
being
taken to be
i.
232.
Analysis
of the
displacement.
(a)
Curvature
of
the strained central-line.
The central-line of the beam is bent into a curve of which the curvatures
in the
planes {x, z)
and
{y, z)
are
expressed
with sufficient
approximation by
the values of
d'u/dz'
and
d^v/dz'
when x and
y
vanish. These
quantities
can
be calculated from the
expressions
for the
components
of strain
by
means of
the formulae
8^_92x_9ea
dPv
_deyi
de^
dz^
~
dz dx
'
dz''
~
dz
dy
'
or
they may
be calculated from
equations (12).
We find
d^^W(l-z)
d^v
EI dz'
=
0.
It follows that the
plane
of the curve into which the central-line is bent is
the
plane
of
(x, z),
and that its radius of curvature R at
any point
is
equal
to
EIIW{1

z).
The denominator of this
expression
is the
bending
moment.
Grashof,
Elasticitat und
Fettigkeit, p.
246.
212
324 ANALYSIS OF THE DISPLACEMENT
[CH.
XV
M
say
;
and therefore the curvature
l/R
of the central-line is connected with
the
bending
moment M
by
the
equation
M
=
EI/R
(19)
and the curvature at
any point
is the same as it would be if the beam were
bent
by
terminal
couples equal
to the value of ilf at the
point.
(b)
Neutral
plane.
The extension of
any longitudinal
filament is
given by
the
equation
e,,
=
-xlR (20)
It follows that filaments which lie in the
plane
a;
=
suffer no extension or
contraction
;
in other
words,
this
plane
is a
"
neutral
plane."
The
extension,
or
contraction,
of
any longitudinal
linear element is determined
by
its
distance from the neutral
plane
and the curvature of the
central-line,
by
exactly
the same rule as holds in the case of
bending by
terminal
couples.
(c) Obliquity of
the strained cross-sections.
The strained central-line is not at
right angles
to the strained cross-
sections,
but the cosine of the
angle
at which
they
cut is the
value,
at
any
point
of the
central-line,
of the
strain-component
e^x-
We shall denote it
by
So-
Then we have
shearing
stress at centroid
-^
.
"
rigidity
of material
'
and we
may
calculate
s
by
the formula
s,^-(W/EI){dxldxX
(22)
where the sufiix indicates that zero is to be substituted for x and
y
after
the dififerentiation has been
performed.
The
quantity
s^
is a small
constant,
so that all the strained cross-sections
cut the strained central-line at the same
angle ^tt
So-
The relative situation
of the strained central-line and an
initially
vertical filament is illustrated
by
Fig.
14 in Article 95.
If the element of the strained cross-section at the centroid of the fixed
end is
vertical,
the constant
/9
in the
displacement,
as
given by (12),
is
equal
to
So*.
When the
bounding
curve is the
ellipse x^/a^+i/^lb^
=
l,
we find
_
4W
2a'(l-t-<r)
+
5
If in
(21)
the
shearing
stress at the centroid were
replaced by
the
average shearing
stress
( Wjnab),
the estimated value of

would be too
small,
in a ratio
varying
from
J,
when a is
large compared
with
6,
to
|
when 6 is
large compared
with
at.
*
In Saint-Venant's memoir
/3
is identified with
Sj.
t
In
obtaining
these numbers a is
put equal
to
J.
232]
IN A BENT BEAM
When the
boundary
is a
rectangle
we find
3W{l
+
a)
iEab
A 00
(-1)
=1
, nna
n' cosh
J I
325
.(23)
The
expression
in
square
brackets was tabulated
by Saint-Venant,
o-
being
taken to be
^,
with the
following
results :

ajb
326 DISPLACEMENT AND SHEARING STRESS
[CH.
XV
cross-sections in their own
planes.
These
changes
are of the same kind as
those described in Article 88. It follows that the neutral
plane
is deformed
into an anticlastic surface. The strained central-line is one of the lines of
curvature of this surface
;
the
corresponding
centres of curvature are below
the neutral
plane,
and the
corresponding
radii of curvature are
expressed by
the formula
EIIW{1

z).
The other centre of curvature of the
surface,
at
any point
of the
central-line,
is above the neutral
plane;
and the corre-
sponding
radii of curvature are
expressed by
the formula
EIjcrW{l

z).
{g)
Distortion
of
the cross-sectioiis into curved
surfaces.
The
expression
for w
may
be written
W
w
=
r<l>
^Tf
X
(Iz

^z^)

I3a)
+
s^a;
_ j[
r
dx
Jo
+
xy'
]...
(26)
The term
T<f> corresponds
with the
twisting
of the beam
by
the
load,
and
we know that it
represents
a distortion of the cross-sections into curved
surfaces. The terms
x[W(lz

^z')/EI
+
0} represent
a
displacement by
which the cross-sections become at
right angles
to the strained central-line.
The term
s^x
represents
a
displacement by
which each cross-section is turned
back,
towards the
central-line,
through
an
angle
s,
as
explained
in
(c)
above.
The
remaining
terms in
W/EI represent
a distortion of the cross-sections
into curved
surfaces,
independent
of that which
depends upon T<f).
If we
construct the surface which is
given by
the
equation
+
Xf\
+
T<j},
.
.(27)
and
suppose
it to be
placed
so that its
tangent plane
at the
origin
coincides
with the
tangent plane
of a strained cross-section at its
centroid,
the strained
cross-section
will coincide with this surface.
Fig.
29.
232,
233]
IN A BENT BEAM
327
In the case of a circular
boundary
the value of the
right-hand
member of
(27)
is
and the contoiir lines of the strained cross-section are found
by equating
this
expression
to a constant. Some of these lines are traced in
Fig.
29.
233. Distribution of
shearing
stress.
The
importance
of the transverse
component
Y^
of the
tangential
traction
on the cross-sections
may
be seen in the case of the
elliptic boundary.
When
a is
large compared
with
b,
the maximum value of
Y^
is small
compared
with
that of
Xg ;
as the ratio of 6 to a
increases,
the ratio of the maximum of
Y^
to that of
X^
increases
; and,
when b is
large compared
with
a,
the maximum
of
F2
is
large compared
with that of
X^.
Thus the
importance
of
Y^
increases
as the
shape
of the beam
approaches
to that of a
plank.
We
may
illustrate
graphically
the distribution of
tangential
traction on
the cross-sections
by tracing
curves,
which are such that the
tangent
to
any
one of them at
any point
is in the direction of the line of action of the
tangential
traction at the
point.
As in Article
219,
these curves
may
be
called
"
lines of
shearing
stress." The differential
equation
of the
family
of
curves is
da;/X,
=
dy/Y (28)
+
{2
+
a)
^
dw
-
|-J
+
\,TX^
+
(1
-
|<r) 2/
j
dy
=
0.
11
Since
dXJdx
+
dY^jdy
is not
equal
to
zero,
the
magnitude
of the
shearing
stress is not measured
by
the closeness of
neighbouring
curves of the
family.
As an
example
we
may
consider the case of the
elliptic boundary.
The differential
equation
is
and this
may
be
expressed
in the form
2.r^{(H-<r)a2
+
o-62}--{2(l-)-o-)a2
+
62}4."'{2(l-ho-)a2
+
6-^}-y(l-2(r)a2=0.
ay y y
2(l+<r)o'+<)''
This
equation
has an
integrating
factor
y
(x+<r)a^+<rfi^
,
and the
complete primitive
may
be
expressed
in the form
where C is an
arbitrary
constant. Since
<r<\
all the curves of the
family
touch the
elliptic boundary
at the
highest
and lowest
points (
+
a, 0).
The case of a circular
328 THEORY OF COMBINED STRAIN
[CH.
XV
boundary
is
included,
and the lines of
shearing
stress are in this case
given by
the
equation 3+25
Some of these curves are traced in
Fig. 30,
o-
being
taken to be
J.
234. Generalizations of the
preceding theory.
(a) Asymmetric loading.
When the
load,
W
say,
is directed
parallel
to the axis of
y
instead of the axis of
x,
the
requisite stress-components are,
as
before, X,, Y,, Z^, given by
the
equations
z.=
W'{l-z)y
where I' denotes the
integral
|
iy'^dxdy
taken over the area of the
cross-section,
and
x'
i^
a
plane
harmonic function which satisfies the
boundary
condition
bL_
.(29)
Y=-{^
+
a)xy
cos
{x, v)
-
{Joy^
+
(
1
_
Jo-) a;2}
cos
(y, v).
The constant t is
adjusted,
as
before,
so that the tractions on a cross-section
may
not
yield
any couple
about the axis of z.
Apart
from a
displacement
which would be
possible
in a
rigid body,
the
displacement
is
given by
the
equations
W \
u=-Tyz
+
^j,cr{l-z)xy,
^
W
rzx
+
eT'
(4 ('
-
^)
<^
(.y'
-
^')
+
i^^'
-
*^}.
W
^=
r,l>-^,{y{lz-iz^)
+
x'+yx^}.
.(30)
233,
234]
IN A BENT AND
TWISTED BEAM
When the direction of the load is not that of one of the
principal
axes of the cross-
sections at their
centroids,
we
may
resolve tlie
load,
P
say,
into
corrnwnents
W and W
parallel
to the axes of x and
y.
The solution is to be obtained
by combining
the solutions
given
in Articles
229,
230 with that
given
here.
Omitting displacements
which would
be
possible
in a
rigid body
we deduce from the
expressions (12)
and
(30)
the
equations
of
the strained central-line in the form
W W
and this line is therefore a
plane
curve in the
plane
W'xir= WyjI.
The neutral
plane
is determined
by
the
equation e
=
0, and,
since
W W
this is the
plane
WxlI-\- W'y/r
=
0.
The neutral
plane
is therefore at
right angles
to the
plane
of
bending.
The load
plane
is
given by
the
equation y/x= W'j
W. Since / and /' are
resiiectively
the moments of inertia
of the cross-section about the axes of
y
and
x,
the result
may
be
expressed
in the form :

The traces of the load

plane
and the neutral
plane
on the cross-section are
conjugate
diameters of the
ellipse
of inertia of the cross-section at its centroid*.
(6)
Combined strain.
We
may
write down the solution of the
problem
of a beam held bent
by
terminal
couple
about
any
axis in the
plane
of its
cross-section, by
means of the results
given
in
Article 87
;
we have
merely
to combine the results for two
component couples
about the
principal
axes of the cross-section at its centroid.
By combining
the solution of the
problems
of extension
by
terminal tractive load
[Articles
69 and
70.(A)],
of torsion
(Chapter xiv.),
of
bending by couples,
and of
bending by
terminal transverse
load,
we
may
obtain the state of stress or strain in a beam deformed
by
forces
applied
at its ends
alone in such a
way
as to be
statically equivalent
to
any given
resultant and resultant
moment. In all these .solutions the
stress-components
denoted
by Xx, Yy, Xy
vanish.
As
regards
the
strength
of a beam to resist
bending
we
may
remark
that,
when the
linear dimensions of the cross-section are small
compared
with the
length,
the most
important
of the
-stress-components
is the
longitudinal tension,
and the most
important
of the
strain-components
is the
longitudinal extension,
and the
greatest
values are found
in each case in the sections at which the
bending
moment is
greatest,
and at the
points
of these sections which are furthest from the neutral
plane.
The condition of
safety
for
a bent beam can be
expressed
in the form:

The maximum
bending
moment must not
exceed a certain
limiting
value.
The condition of
safety
of a twisted
prism
was considered in Article 220. The
quantity
which must
not,
in this
case,
exceed a certain
limiting
value is the
shear;
and this is
generally greatest
at those
points
of the
boundary
which are nearest to the central-line.
When the beam is at the same time bent and
twisted,
the
components
of stress which are
different from zero are the
longitudinal
tension
Z^
due to
bending
and the
shearing
stres.ses
A'j
and
Y,.
If the
length
of the beam is
great compared
with the linear dimensions of the
cross-section the values of
Z^
near the section z
=
and the terms of
X,
and
Y^
that
depend upon twisting
can be
comparable
with each
other,
and
they
are
large compared
*
The result was
given by
Saint-Venant in the memoir on torsion of 1855.
330 THEORY OF COMBINED STRAIN
[CH.
XV
with the terms of
X,
and
Y^
that are due to
bending.
For the
purpose
of an estimate of
strength
we
might
omit the
shearing
stresses and
shearing
strains that are duo to
bending,
and take account of those
only
which are due to
twisting.
In
any
case in which the
stress-components X^, Y,, Z^
are different from zero and
Xj,
Yy, Xy vanish,
the
principal stress-components
can be found
by observing
that the
stress-quadric
is of the form

(2.Y,^-|-2I>-f-^2)
=
const.,
and therefore one
principal plane
of .stress at
any point
is the
plane
drawn
parallel
to the
central-line to contain the direction of the
resultant,
at the
point,
of the
tangential
tractions
on the cross-section. The normal traction on this
plane
vanishes,
and the values of the
two
principal
stresses which do not vanish are
i^;j[^,H4W+r/)]i (31)
In
any
such case the
strain-quadric
is of the form
^
[
-
o-.Z (a^H/
+
2^)
+
(1
+
(t)
2
(2X,.r
-(-
2
r.y
-I-
.Z^z)]
=
const.,
and the
principal
extensions are
equal
to
-#' ^4^w[^.^
+
4W+n^A (32)
the first of these
being
the extension of a line at
right angles
to that
principal plane
of
stress on which the normal traction vanishes.
(c) JEolotropic
material.
The
complexity
of the
problem
of Article 228 is not
essentially
increased if the material
of the beam is taken to be
seolotropic, provided
that the
pianos through any point,
which
are
parallel
to the
principal planes,
are
planes
of
symmetry
of structure. We
suppose
the
axes of
X,
y,
to be chosen in the same
way
as in Article
228,
and assume that the strain-
energy-function
has the form
\ {A , B, C, F, O, H) {e ,
e, e^f
+
J {LeJ
+
Me
J
+
Ne^^).
We denote the
Young's
modulus of the material for tension in the direction of the axis
of z
by J?,
and we denote the Poisson's ratios which
correspond respectively
with con-
tractions
parallel
to the axes of x and
y
and tension in the direction of the axis of
2, by o-j
and
(72.
We assume a
stress-system
restricted
by
the
equations
W
X^=Yy
=
Xy
=
Q, Z,=
-^{l-z)x
(33)
Then we
may
show that
X,
and
Y, necessarily
have the forms :
where
(^
and
x
^^^ solutions of the same
partial
differential
equation
.(34)
(
dx''
^if'jx
'
which
respectively satisfy
the
following boundary
conditions :

co8(.r, v)M~+coa{y,
v)L^
=
cos(.r, v)Mi/

coa(>/, v) Lx,
eoB(x,v)M^
+
cos
{y,v)L^
=
-
cos
{x, v)
M
Ua^x''
+
~^^^
'^^
y^
-cofi{y,v){E-M<ri)xy.
234,
235]
CRITICISMS of certain methods 881
Further we
may
show that the
displacement corresponding
with the
stress-system expressed
by (33)
and
(34) necessarily
has the form :
W
W
v=
TZX+^(l-z)ar^x1/
+
yX-aZ
+ff,
V
(35)
As in Article
230,
we
may
take
a'=j3'=y'
=
and a
=
y=0.
The constant of
integration
r
can be
adjusted
so that the traction at the loaded end
may
be
statically equivalent
to a
single force, W, acting
at the centroid of the terminal section in the direction of the axis
of X. The results
may
be
interpreted
in the same
way
as in Article 232.
235. Criticisms of certain methods.
(a)
In
many
treatises on
Applied
Mechanics* the
shearing
stress is calculated from
the'stress-equations
of
equilibrium,
without reference to the conditions of
comjiatibility
of
strain-components, by
the aid of certain
assumptions
as to the distribution of
tangential
traction on the cross-section. In
particular,
when the section is a
rectangle,
and the load
is a force W
parallel
to the axis of
x,
it is assumed
(i)
that
Y,
is
zero,
(ii)
that
X,
is
independent
of
y.
Conditions
(i)
and
(ii)
of Article
228,
combined with these
assumptions,
lead to the
following stress-system
:

X,=
r
=
Z=r,
=
0, 'V.=
^(3^-^),
Z,=
-j{l-z)x,
(36)
in which a is the area of the
cross-section,
and / is the moment of inertia
previously
so
denoted. The resultant traction / /
X^dxdy
is
equal
to W.
If this
stress-system
could be
correct,
there would exist functions
u, v,
w which would
be such that
0i)
_
o- IF . du
h/~~m^
''''
dz'
Now we have the identical
equation
dxdydz
~
dxdy \dy
'^
dz) 3^
\dz
^
dx)
"^
8^82
\dx
^
oy)
'
but this
equation
is not consistent with the above values for
dv/dy,...
; for,
when these
values are
substituted,
the left-hand member is
equal
to
2<rW/ET,
and the
right-hand
member is
equal
to zero. It follows that the
stress-system expressed by (36)
is not
jxjssible
in an
isotropic
solid
body.
We know
already
from Article 95 that the
stress-system (36) gives correctly
the
average
stress across the breadth of the
section,
and therefore
gives
a
good approximation
to the
actual stress when the breadth is small
compared
with the
depth.
The extent to which it
is
inadequate may
be estimated
by
means of the table in Article 232
(c) ;
for it would
give
for

the factor outside the

square
bracket in the
right-hand
member of
(23).
It fails also
to
give correctly
the direction of the
tangential
traction on the
cross-sections,
for it makes
this traction
everywhere
vertical,
whereas near the
top
and bottom
bounding
lines it is
nearly
horizontal.
*
See for
example
the treatises of Rankine and Graahof
(juoted
in the Introduction footnotes
94 and
95,
and those of
Ewing,
Bach and
Foppl quoted
in the footnote on
p.
110.
uu
,
dw W
f
I \ dw dv dv du
332 CRITICISMS OF CERTAIN METHODS
[CH.
XV
(6)
In the extension of this method to sections which are not
rectangular
it is
recognized*
that the
component Y^
of
shearing
stress must exist as well as
X,.
The case
selected for discussion is that in which the cross-section is
symmetrical
with
respect
to a
vertical axis. The
following assumptions
are made :

(i) A',
is
independent
of
y, (ii)
the resultants of
X,
and
F,
at all
points
/" which have a
given
x meet in
a
point
on the axis of x. To
satisfy
the
boundary
condition
(3)
this
point
must be that marked T in
Fig. 31,
viz. the
point
where the
tangent
at P to the
bounding
curve of the section meets this axis.
To
express
the
assumption (ii) analytically,
let
r)
be
the ordinate
{NP)
of P and
y
that of
P',
then
.(37)
Equation (2)
then becomes
dX, Idr,
Wx
and the solution which makes
X^
vanish at the
highest
point
Ix
a)
is
and it is
easy
to see that this solution also makes
A',
vanish at the lowest
point.
The
stress-system
obtained
by
these
assumptions
is
expres.sed by
the
equations
Wx{l-z)
^
A.=
y,=A>0,
A'.=
-J/^_^^rf., r^^-^^JLJ[^x,dx,
z,^
.(38)
it satisfies the
equations
of
equilibrium
and the
boundary condition,
and it
gives
the
right
value W for the resultant of the
tangential
tractions on the section.
But,
in
general,
it is
not a
possible stress-system,
for the same reason as in the case of the
rectangle,
viz. the
conditions of
compatibility
of
strain-components
cannot be satisfied.
(c)
These conditions
may
be shown
easily
to lead to the
following equation
:

i^^tlM
=
w^'
''''
which determines
>;
as a function of
x,
and therewith determines those forms of section for
which the
stress-system (38)
is a
possible
one. To
integrate (39)
we
put
/:
xr]
dx
=
^,
.
.(40)
and then
^
satisfies the
equation
. Th
I
2a'
^
=
c\(a'-x)''-*-'''(x
+
a
l+<r'
where
|', |"
mean
d^jdx, d^^/dx\
The
complete primitive
can be shown to be
2
]
H-cr
,a+a'y
See,
in
particular,
the treatiee of Orashof
already
cited.
235]
OF TREATING THE PROBLEM OF FLEXURE
333
where
C,
a and a' are
arbitrary
constants. On
eliminating ^ by
means of the relation
(40)
we see that the
equation
of the
bounding
curve must have the form
P /J V
2a' 2a - .
''=^rf^L^'*'-'')''^"'('^+")"^'''J
'

(*')
The constants a and a'
express
the
height
of the
highest point
of the
curve,
and the
depth
of its lowest
point,
measured from the centroid.
Unless the
bounding
curve of the section has one of the forms included in
equation (41)
the stress is not
correctly given by (38).
It
may
be observed
that,
if the section is
symmetrical
with
respect
to the axis of
y,
so that a'
=
a,
the
equation (41)
is of the
form
iti/b)^''+x^/a^=l.
We saw in Article 231
(/)
that the
problem
of flexure could be
solved for this
section,
and the curve was traced in
Fig.
26 for the case where
(r=^
and
a
=
26.
(d)
We
may
observe that in the case of the
elliptic (or circular) boundary
this
method would make the lines of
shearing
stress
ellipses, having
their axes in the same
direction as those of the
bounding
curve and
touching
this curve at the
highest
and lowest
points. Fig.
30 shows that the correct curves are flatter than these
ellipses
in the
neighbourhood
of these
points.
In
regard
to the
obliquity
of the strained
cross-sections,
the method would
give
for
Sq
the value 8
W{\+a-)/SEnab,
which is
nearly
correct when the
breadth is
small,
or b is small
compared
with
a,
but is too small
by
about 5
per
cent, in
the case of the
circle,
and
by nearly
20
per
cent, when b is
large compared
with a.
(e)
The existence of a term of the form
^z
in the
expression
for the deflexion
[Article
232
(rf)]
has been
recognized by
writers of technical treatises. The term was
named
by
Rankine
(loc. cit.)
"
the additional deflexion due to
shearing."
In view of the
discussion at the end of Article 230
concerning
the
meaning
of the constant
/3,
the name
seems not to be a
good
one.
(/)
The theorem of Article 120 is sometimes used to determine the additional
deflexion*. The theorem
yields
the
equation
J
/ /
{X,u+Y^v
+
Z,w),^idxdy-^
j
I
{X,u+ Y.v-\-Z^w\^^dxdy
=ijjf[{^^'+
i7+2,^-2^(,}\Z,
+
...)}/E+{A7+
7,^
+
Xy^)ly]dxdydz. ...(42)
When the tractions over the ends are
assigned
in a
special
manner in accordance with
the formulae
(1)
and
(8),
so that the
displacement
is
given by (12),
the first term of the
left-hand member of
(42)
becomes
\ WH^/EI+^
Wfil,
and the second term becomes
-
iW^l-i
jj
[{X,u+ r,v),^,-{Wlx/I){,r<t>- W(x
+
.vy^)/EI}]dxdy,
where the
expression
under the
sign
of
integration
is
indeiJendent
of
j3.
The
right-hand
member of
(42)
becomes
IWH^IEI+ilfi-^
I /
{X}-\-Y})dxdy,
which also is
independent
of
/3. Thus,
in this
case, equation (42)
fails to determine the additional deflexion. When
the tractions over the ends are not distributed
exactly
in accordance with
(1)
and
(8),
the
displacement
is
practically
of the form
given by (12)
in the
greater part
of the
beam,
but
must be
subject
to local
irregularity
near the ends. The left-hand member of
(42)
is
approximately equal
to
\
IKS,
where 8 is the deflexion at the loaded
end,
and the
right-
hand member is
approximately equal
to
\ WW^jEI ; but,
for a closer
approximation
we
should
require
a
knowledge
not
only
of
X,
and
T,
in the
greater part
of the
beam,
but also
of the terminal
irregularity.
*
See
e.g.
W. J. M.
Rankine,
loc.
cit.,
or J.
Perry, Applied
Mechanics
(London, 1899), p.
461.
CHAPTER XVI.
THE BENDING OF A BEAM LOADED UNIFORMLY ALONG ITS LENGTH.
236. In this
Chapter
we shall discuss some
problems
of the
equilibrium
of an
isotropic body
of
cylindrical
form,
by imposing particular
restrictions on
the character of the stress.
Measuring
the coordinate z
along
the
length
of the
cylinder,
we shall in the first
place suppose
that the stress is in-
dependent
of
z,
then that it is
expressed by
linear functions of
z,
and
finally
that it is
expressed by quadratic
functions of z. We shall find that the
first two restrictions lead to solutions which have been obtained in
previous
Chapters*,
but that the
assumption
of
quadratic
functions of z enables us to
solve the
problem
of the
bending
of a beam
by
a load distributed
uniformly
along
its
length.
237. Stress uniform
along
the beam.
We take the axis of z to be the central-line of the
beam,
and the axes of
X and
y
to be
parallel
to the
principal
axes of the cross-sections at their
centroids.
We
suppose
that there are no
body
forces,
and that the
cylindrical
bounding
surface is free from traction. We
investigate
those states of stress
in which the
stress-components
are
independent
of z.
The
equations
of
equilibrium
take the form
^x
^
dy
~
'
dx^
dy
~
'
dx
^
dy
~
'
^ '
and the conditions
which hold at the
cylindrical boundary
are
cos
(, v)
Xx
+
cos
{y, v) Xy
=
0,
cos
{x, v) Xy
+ cos
{y, v)
Yy
=
0,
cos
(x, v)
X;i
+
cos
{y, v)
F^
=
0. . .
.(2)
The conditions of
compatibility
of
strain-components
take the forms
-df^^'
d^-^'
dxdy^^'
^^^
-i
'^^w-S^r'

^'^
' *
Cf. W.
Voigt, Qnttingen Abhandlungen,
Bd. 34
(1887).
236,
237]
BEAM STRESSED UNIFORMLY ALONG ITS
LENGTH
335
The
equations (3)
show that
e^z
is a linear function of x and
y, say
e
=
e-
Kx-K'y
(6)
where
e, ,
k are constants. Whenever this is the case
equations (1)
and
conditions
(2)
lead to the conclusion that
X^.,
Yy, Xy
vanish. To
prove
this
we observe
that,
if
u,
v' are
any
functions of x and
y,
these
equations
and
conditions
require
that
the
integration being
taken over the cross-section
;
for the left-hand member
is at once transformable into
\{[Xx
cos
{x, v)
+
Xy
cos
(y, v)]
u
+
{Xy
cos
{x, v)
+
Yy
cos
{y, v)] v]
ds
where ds is an element of arc of the
bounding
curve of the cross-section.
Now in
equation (7) put
(i)
u
=x,
v'
=
0,
we find
|
/
X^dxdy
=
0,
(ii)
u'
=
x^,
v'
=
0,
we find 1 1
xXxdxdy
=
0,
(iii)
u'
=
xy,
v'
=

^af,
we find 1 1
yX^dxdy
=
0;
and in like manner we
may prove
that
1 1
Yydxdy
=
0,
\\
xYydxdy
=
0,
\\
yYydxdy
=
0.
It follows from these results and
(6)
that
1
1
XxBzzdxdy
=
0,
\\
YyBzzdxdy
=
0.
Again,
in
equation (7)
let
u',
v be the
components parallel
to the axes of
X and
y
of the
displacement
which
corresponds
with the stress
X^, ...,
then
this
equation
becomes
jj(Xxe^x+
Yyeyy
+
Xye^)dxdy
=
(8)
But we have
X^e^x
+
Yyeyy
=
-cr(Xx+ Yy)
e
+
^-
(1
+
<t) {(1
-
a) (Xj'
+
Yy')
-
2<rXxYy}
.
The
integral
of the term

a
(X^
+
Yy)
e^z vanishes,
and the
quadratic
form
(l-a-)(X/-\-
Yy^)-2a-XxYy
is definite and
positive,
since
<t<^;
also we
have
Xyexy
=
i^'^Xy".
Hence the
expression
X^exx
+
YyCyy
+
XyCxy
is
336 BEAM UNDER STRESS WHICH VARIES
[CH.
XVI
necessarily positive,
and
equation (8)
cannot be satisfied unless
Xx,
Yy, Xy
vanish
identically.
It follows that we must have
^zx
^
^^zz y
^yy
^
^^zz
^xy
^^^j W/
where
e
is
given by (6) ;
and then
equation (5)
is satisfied
identically.
The
remaining equations
and conditions are the third of the
equations
(1),
the third of the conditions
(2), equations (4),
and the relations
X^^fieix,
Yz
=
fjLeyz.
From these we
find,
as in Article
229,
that the mo.st
general
forms for
e^^;
^yz
^-re
"-
=
Ka"^~^)'
'''^
=
^(i
+
'^)
^^^^
where t is a constant of
integration,
and
<f)
is the torsion function for the
cross-section
(Article 216).
The strain is
expressed by equations (6), (9), (10),
and it follows that the
most
general
state of strain which is consistent with the conditions
(i)
that
the stress is uniform
along
the
beam,
(ii)
that no forces are
applied
to the
beam
except
at the
ends,
consists of the strain associated with
'simple
longitudinal
tension
(c
Article
69),
two
simple
flexures
involving
curvatures
K and K in the
planes
of
{x, z)
and
{y, z) [cf
Article
87],
and torsion t as in
Chapter
xiv.
The theorem
proved
in this Article for
isotropic solids, viz., that,
if
e
is linear in
X and
y,
and if there are no
body
forces and no surface tractions on the
cylindrical
boundary,
the
stress-components
X^,
Yy, Xy
must
vanish,
is true also for
seolotropic
materials, provided
that the
plane
of
{x,
y)
is a
plane
of
symmetry*.
238. Stress
varying uniformly along
the beam.
We take the axes of
x,
y,
z in the same
way
as
before,
and retain the
suppositions
that there are no
body
forces and that the
cylindrical bounding
surface of the beam is free from traction
;
and we
investigate
those states
of stress in which the
stress-components,
and
strain-components,
are linear
functions of z. We write the
stress-components
and
strain-components
in
such forms as
X^
=
X^i')^
+
Z^w, e:,^
=
e^^'2:
-i-
e^^<'
(11)
The
equations
of
equilibrium
take such forms as
-f-^-'-t^)-^-^-^-'"-'
<->
and the conditions at the
cylindrical boundary
take such forms as
z
{cos
{x,
v)
Z^w
\
cos
(2/, v) ZyW}
+ cos
(x, v)
Z^.^
+
cos
(y, v) Zj,('
=
0. .
..(13)
*
J.
Boussinesq,
J. de Math.
(LiouvilU), (Ski. 2),
t. 16
(1871).
237,
238]
UNIFORMLY ALONG ITS LENGTH 837
The conditions of
compatibility
of
strain-components
are
d'e^
da?
J^z""
ae,^"'
rt
9'e"' ^'"'
5<'^"'
n
aa? 9a;
9y
"
ay 9y
9a;
9y
9a;
9?/
9a;
ay
...(14)
with
9a;
V
9a; 9 v /
9a;
V
9a;
dy
J
9w 9a;
'
and
9yV
9a;
dy
J
dy\
dx
dy
J
dx
dy
'
r...(15)
\
da?
dy' dxdy
) da?
"^
dy^ dxdy
^ '
In all these
equations
the terms
containing
z and the terms
independent
of z
must vanish
separately.
The relations between
components
of stress and
components
of strain take such forms as
in which the terms that contain
z,
and those which are
independent
of
z,
on
the two sides of the
equations
must be
equated severally.
Selecting
first the terms in
z,
we observe that all the letters with
index
(1) satisfy
the same
equations
as are satisfied
by
the same letters
in Article
237,
and it follows that we
may put
Bzz

i l^iX Kj
y,
d<f>
-'' =
T,(^-2/
'[dx'^J'
^v'
~^'l9y
+
x),
(17)
in which
e^, Ku k/, Tj
are
constants,
and
^
is the torsion function for the
cross-section.
Again, selecting
the terms
independent
of
z,
we find from the first two
of
equations (12)
=
ly {cos {x, v)
Z^<>
+
cos
{y, v) Zj,"" j
-
x
{cos {x, v) Z"
+
cos
(y,
v) F^
""
}
ds,
which vanishes
by
the first two of
equations (13).
Also we have
by (17)
[j"{a;F,<"
-
2/X,w) dajdy
=
/.t,
jJ|a;
+
2/^
+
a;
9y
~
^
gj}
*^'^^'
22
338 BEAM UNDER STRESS WHICH VARIES
[CH.
XVI
where the
integral
on the
right
is the coefficient of
fi
in the
expression
for the
torsional
rigidity
of the beam. It follows that
Ti
must
vanish*,
and hence
that
X^>
and
F^"*
vanish.
By selecting
the terms
independent
of ^^ in the third of
equations (12)
and
conditions
(13)
we find the differential
equation
ox
oy
and the
boundary
condition
Z^i"'
cos
{x, v)
+
Fjf'
cos
iy, v)
=
0,
which are inconsistent unless
lb
Z^^HxAy
=
0.
Since
Z^'"
=
E
(e^

k^x

K^y),
this
equation requires
ei
to vanish.
We
may
now rewrite
equations (17)
in the form
e"' =-ic,x-
K,'y,
e^^f"
=
ey,/^
=
-
ae<",
e^^^m
=
e^^w
=
e^w
=
0. . .
.(18)
Since
X^'"
and
Fj*" vanish,
we
find,
by selecting
the terras
independent
of z in the first two of
equations (12)
and conditions
(13),
that
Xa,'"',
Fy'"', X<*
vanish and that
622""
is a linear function of x and
y.
We
may
therefore
put
e(>
=
6<,-a;-/c'2/,
e^.^""
=6^,!,""
=
-<re,,"'>,
ey>"=0,
(19)
where
e, k, Kq
are constants.
Equation (16)
is satisfied
identically.
Further,
by selecting
the terms
independent
of z in the third of
equations
(12),
and the third of conditions
(13),
and in
equations (15),
we
find,
as in
Articles 229 and 234
(a),
that
e^x^"^
and
e^i""
must have the forms
(20)
'"
=
^

-
y)
+
i
fl
+
h<r^
+
(1
-
i'^)
f\
+
/
{ll
+
(2
+
<-)
^y]
.
where
%
and
;y;'
are the flexure functions for the
cross-section,
corresponding
with
bending
in the
planes
of
{x, z)
and
{y, z),
and
To
is a constant.
We have shown
that,
in the
body
with a
cylindrical boundary,
the most
general
state
of stress consistent with the conditions that no forces are
applied except
at the
ends,
and
that the
stress-components
are linear functions of
z,
has the
properties (i)
that
X,
azid
Y,
are
independent
of
z, (ii)
that
X^, Yy, X^
vanish. Thus the
only stress-component
that
*
This conclusion is otherwise evident
;
for if
r,
did not vanish we should have twist of
variable amount
t-^z
maintained
by
tractions at the ends. The torsional
couples
at different
sections could not then balance.
238,
239]
UNIFORMLY ALONG ITS LENGTH
339
depends upon
z is
Z,
which is a linear function of z.
Conversely,
if there are no
body
forces and
X,.,
y, X^
all
vanish,
the
equations
of
equilibrium
become
dz
'
dz
'
dx'^
3y
'^
dz \
and it follows from these that
X,
and
V,
are
independent
of z and that
Z,
is a linear func-
tion of z. Thus the condition that the stress varies
uniformly along
the beam is the same
as the conditions that
X^,
}', Xy
vanish*.
The most
general
state of strain which is consistent w^ith the conditions
(i)
that the stress varies
uniformly along
the
beam,
(ii)
that no forces are
applied
to the beam
except
at the
ends,
consists of extension due to terminal
tractive
load,
bending by
transverse
forces,
and
by couples, applied
at the
terminal
sections,
and torsion
produced by couples applied
to the same
sections about axes
coinciding
with the central-line. The resultant force
at
any
section has
components parallel
to the axes of
x,
y,
z which are
equal
to
-EIk^, -EI'k;, Eo,
where
I=jjx-dxdy
and I'
=
jl
y^dwdy;
and the resultant
couple
at
any
section has
components
about axes
parallel
to the axes of
x,
y
which are
equal
to
-
EF
('
-I-
K^Z),
EI
(/Co
-f
K,Z),
and a
component
about the axis of z which is
equal
to
^T,\\{x^
+
f,-xf^-y^)dxdy
+
^,
jj
| ^
-
3/
^
-1-
(2
-h
io-) :^y
-
(1
-
i a)
y}
da^dy
+
^^^^j\^x^-X^-y^^-(2
+
\a)xf
+
{\-\a)f\^dxdy.
The solutions of the
problems
thus
presented
have been discussed in
previous
Chapters.
239.
Uniformly
loaded beam. Reduction of the
problem
to one
of
plane strainf
.
Taking
the axes in the same
way
as
before,
we .shall now
suppose
that all
the
components
of stress and strain are
expressed by quadratic
functions of z
so that for
example
X^
=
X^^^z'^
+
Z^i"^
-t-
Z^i>, e^^
=
e^^'%=
+
e^^<'^
-|-
e^^w (21 )
*
For the
importance
of these results in connexion with the historical
development
of the
theory,
see Introduction,
p.
21.
t
The
theory
is due to J. H.
Michell,
Quart.
J.
of
Math.,
vol. 32
(1901).
222
340 REDUCTION OF THE PROBLEM OF THE
[CH.
XVI
We shall
suppose
also that there is
body
force,
specified by components
X,
Y
parallel
to the axes oi
x,
y,
and surface traction on the
cylindrical boundary,
specified similarly by
Xy, Y,,
these
quantities being independent
of z. Then
in the
equations
of
equilibrium,
the
boundary
conditions,
the
equations
of
compatibility
of
strain-components,
and the stress-strain
relations,
the terms
of the
second,
first and zero
degrees
in z
may
be taken
separately.
Selecting
first the terms that contain
z-,
we
find,
exactly
as in Article
238,
that we
may put
fi
(2)
_
_
p
(2)
p
m
0
.(22)
where
e^, k^, k^, r^
are
constants,
and
</>
is the torsion function for the section.
Again, selecting
the terms that contain
z,
we
may
show that
t^
and
ej
must
vanish,
and that we
may put
p
m

p
(1)


rrp
W
P
<"
=
fd4> 9%
V
^!/
(23)
where
e,, k,, Ki, t,
are
constants,
and
p^
and
^
are the two flexure functions
for the section.
For the determination of
Xj;*"',
... we have the
equations
of
equilibrium
dx
dy
+
-?/' =0,
.(24)
dx
dy
and the
boundary
conditions
Z<'
cos
(x, v)
+
Zy<)
cos
{y, v)-X,^0,\
Z^'i
cos
{x, v)
+
Fy""
cos
(y, I/)
-
F,
=
0,
i
(25)
Z^'<"
cos
{x, v)
+
F^"
cos
{y,
v)
=
0.
j
The third of
equations (24)
and of conditions
(2.5)
are
incompatible
unless
,the
constant
ei
of
(23)
vanishes.
1
239]
UNIFORMLY LOADED BEAM
341
Further we have
e^Kj'"*,
... and
X^^'^
... connected
by
the
ordinary
stress-
strain
relations,
and we have the
equations
of
compatibility
of strain-
components
in the forms
with
dxdy
dec
dy
'
)x\
dx
dy
)
'
~
'
.(26)
9a;
\
dx
dy
de^
dy
\
dx
dy
20-Ac,
=
0,
and
S^e^'"'
^
SV*
_ 9'e^"
,.(27)
.(28)
'dy^
da?
dxdy
Equations (26)
give
us the form of
e^^"",
viz.:
ejz<'
=
eo
-
ic,x
-
K^y
+
1k^
{x
+
xy^)
-|-
2;
{x
+
<^y)
+
-^1^^

-(29)
and,
by
a similar
process
to that in Article
238,
we find
'
(30)
dx
*!yz
=
^(af+'')+'''{af
+
^^
+
'">^+^''{^'+*'^^'
+
(^-*'^>'^}',
wherein
e, , ', To
are
constants,
and
<^, x^ X
^r^ the functions
previously
named.
It remains to determine
Z^<',
F^"", Zy"
from the first two of
(24),
the
first two of
(25),
the
appropriate
stress-strain relations and the
equation (28).
This determination
requires
in effect the solution of a
problem
of
plane
strain. If we
put
Z,w
=
Xe,,i>
+
XJ,
Fy('
=
\e()
+
F,/, (31)
then the
equations
of the
problem
of
plane
strain are
dJi^dXjT ^
dx
dy
dXy^^dYy' _^
pZ
+
Z.o'+X
dx
dp
<'
(>y .
=
0,
=
0.
.(32)
dx
dy
together
with
equation (28),
the
equations
XJ
=
\eyy'<
+{\+
2/^)
e^,<',
Yy'
=
Xe,^i'
+
(X
+
2^) e,^i<", Z/
=
/ie^'',
(33)
342 NATURE OF THE SOLUTION OF THE PROBLEM
[CH.
XVI
and the
boundary
conditions
X;
cos
{x,
v)
+
Zj,<
cos
(y, v)
=
[Z,
-
Xe"
cos
{x,
v)\
\
Zi'
cos
{x, v)
+
Yy
cos
(y, V)
=
[
F,
-
\e<'
cos
(y, i/)].
J
The
expressions
in
square
brackets in
(32)
and
(34) may
be
regarded
as
known.
The
theory
here
explained
admits of extension to
any
case in which the forces
applied
to
the beam
along
its
length
have
longitudinal components
as well as transverse
components,
provided
that all these
components
are
independent
of z*. This restriction
may
be
removed,
and the
theory
extended further to
any
case in which all the forces
applied
to the
beam
along
its
length
are
represented by
rational
integral
functions of
z\.
240. The constants of the solution.
Let
W,
W denote the
components parallel
to the axes of x and
y
of the
uniform
load,
so that we have
W=(jpXdxdy
+
j
X,dS
with a similar formula for W. From
equations (32)
and
(34)
we find
F
=
-
[[
Z,<'>ddy,
W'
=
-j(Y,''^dwdy
(35)
Now we
may
write down the
equations
//z..^.rf,=//{A(.Z.)4^(.F.)-.g^4-fj}^.c^3/
=
lx{X,
cos
(x, v)
+
Y,
cos
{y, v)}
ds +
jjx
{^^<"
+
22^2<^'
} dxdy
=
-EI{ki
+
2zk^),
with similar
equations
for 1 1
Y^dxdy.
Hence we find
2EIk^=W, '2.EI'k^=W' (36)
Thus the constants
2, k^
are determined in terms of the load
per
unit of
length.
If the
body
forces and the surface tractions on the
cylindrical bounding
surface
give
rise to a
couple
about the axis of
z,
the moment of this
couple
is
11
p(xY-yX)dxdy+
\
(xY,
-
yX,) ds,
and from
equations (32)
and
(34)
we find that this
expression
is
equal
to
-JJ{^F,<"-2/Z,')da;dy.
*
J. H.
Michell,
loc. cit.
p.
339.
t
E.
Almansi, Rovie,
Ace. Lincei Rend,
(Ser. C),
t. 10
(1901).
239-241]
OF THE
UNIFORMLY LOADED BEAM 343
On
substituting /tej">
for
Z^'"
and
fiey/^'
for
F,<",
and
using
the
expressions
given
in
(23)
for
e^t'"
and
e^,"',
we have an
equation
to determine
t,.
When
no
twisting couple
is
applied along
the
length
of the
beam,
and the section is
symmetrical
with
respect
to the axes of x and
y,
Tj
vanishes.
The constants
k^, k.^', t,
depend,
therefore,
on the force- and
couple-
resultants of the load
per
unit of
length.
The terms of the solution which
contain the
remaining
constants
e, ac, k', k,, ,', To
are the same as the terms
of the
complete
solution of the
problem
of Article 238. These constants
depend
therefore on the force- and
couple-resultants
of the tractions
applied
to the terminal sections of the beam. Since the terms
containing
k^, /, t,
alone would involve the existence of tractions on the normal
sections,
the
force- and
couple-resultants
on a terminal section mu.st be
expressed by
adding
the contributions due to the terms in
k^, ./, t,
to the contributions
evaluated at the end of Article 238. The
remaining
constants
fo,
... are then
expressed
in terms of the load
per
unit of
length
and the terminal forces and
couples.
When the functions
<f>, %, x'
are known and the
problem
of
plane
strain is
solved,
we know the state of stress and strain in the beam bent
by
uniform
load,
distributed in
any assigned way,
and
by
terminal forces and
couples.
As in
Chapters
Xiv. and
XV.,
the terminal forces and
couples may
be of
any
assigned
amounts, but the tractions of which
they
are the statical
equivalents
must be distributed in certain definite
ways.
241. Strain and stress in the elements of the beam.
Three of the
components
of strain are determined without
solving
the
problem
of
plane
strain. These are
e^^, e^x,
Syz-
We have
ea
=
e,-{ic
+
KiZ+K^z')x- (;'
-I-
k^z
+
k^z^) y
-I-
2,,
(^
-I-
xy^)
\
-l-2;(x'-l-a;3/)
+
T,</),
:
=
(To
+
r;z)
(g
-
y)
+
{k,
+
2,^)
|?|
+
^
<7^
+
(1
-
J <r) 3/=
-I-
(/
-I-
2k^z)
\-^
+
{2
+
a)xy
y --(ST)
eyz
=
(To
+
T,\z)
(1^
+
^)
+
(^>
+
2=^)
{^
+
(2
+
<^)
^y}
.
+
(/
+
2k^z)
||^
+
(1
-
i<r)
x-' +
\ay-
The constant
eo
is the extension of the central-line.
We shall see
presently
that,
in
general,
it is not
proportional
to the resultant
longitudinal
tension.
The constants
To
and
Tj
are
interpreted by
the observation
that
To-I-Tj^
is the
twist of the beam.
To
interpret
the constants denoted
by
/c,, ...,
we observe that the
344 NATURE OF THE SOLUTION OF THE PROBLEM
[CH.
XVI
curvature of the central-line in the
plane
of
{x, z)
is the value of
d'^u/dz^
when
a;
=
y
=
0. Now we have
3%
_
dezx de^i
dz' dz dx
=
{ko+
KiZ
+
K^z-)

Tiy
+
Kia-(x-

7f)
+
2K2'a-xy, (38)
and
therefore the curvature in
question
is
ko
+
kiZ
+
k^z^.
In like manner we
should find that the curvature of the central-line in the
plane
of
(y,
z),
estimated as the value of
d-v/dz'
when x
=
y
=
0,
is
k^'
+
ic^'z + /CaV.
The
presence
of the terms
eo
+
22
(X
+
*7)
+
22'
(%'
+
a^y)
+
Ti<j>
in the
expression
for
e^
shows that the
simple
relation of the extension of the
longitudinal
filaments to the curvature of the
central-line,
which we noticed
in the case of
bending by
terminal forces
[Article
232
(b)],
does not hold
in the
present problem.
Of the
stress-components
two
only,
X^
and
Y^,
are determined without
solving
the
problem
of
plane
strain. The resultants of these for a cross-section
are
respectively EI(ki
+
2k2z)
and
Er{K-!
+
2KoZ).
The distribution
over the cross-section of the
tangential
tractions
X^
and
Yz
which are
statically equivalent
to these resultants is the same as in Saint-Venant's
solutions
(Chapter XV.).
When there is twist
To
-\-
TiZ,
the tractions
X^
and
F^
which
accompany
the twist are distributed over the cross-sections in the
same
way
as in the torsion
problem (Chapter xiv.).
The
stress-component
Z,,
is not
equal
to
Ee^^
because the stress-
components Xx,
Yy
are not
zero,
but the force- and
couple-resultants
of the tractions
Z^
on the elements of a cross-section can be
expressed
in
terms of the constants of the solution without
solving
the
problem
of
plane
strain. The resultant of the tractions
Z^
is the resultant
longitudinal
tension.
The moments of the tractions
.Z^
about axes drawn
through
the centroid of a
cross-section
parallel
to the axes of
y
and x are the
components
about these
axes of the
bending
moments at the section.
To
express
the resultant
longitudinal
tension we observe that
jjz,dxdy=jjz,^''>dxdy
=
jj[Ee^z^'+<T{Xx^'
+
Yy^o^)]dxdy.
Now we
may
write down the
equations
jfx,^^dxdy
=
//
{I
(.Z.<o.)
+
|(.Z,.o.)
-
.
(Sf!
-,
l^f)]
dxdy
=
{x
{Z^'
cos
{x, v)
+
Zj,'"'
cos
(y, v)]
ds +
jjx
(Z^
-f-
pX) dxdy.
241,
242]
OF THE UNIFORMLY LOADED BEAM 345
The
integral
jj
Yy'"* dxdy may
be transformed in the same
way,
and hence
we find the formula
^^
Z,
dxdy
=
^^
[EeJ"^
+
ax
(Z,<>>
+
pX)
+
ay ( F,'"
+
pT)] dxdy
+
0-J
{xX,
+
yY,)ds (39)
Since the resultant
longitudinal
tension is the same at all
sections,
and is
equal
to the
prescribed
terminal
tension,
this
equation
determines the
constant
e,,.
To
express
the
bending
moments,
let M be the
bending
moment in the
plane
of
{x, z).
Then
M=-
II
xZ.dxdy (40)
and therefore we have
m
dz
'
-jj
X
{Z,^'>
+
2zZ,^'>)
dxdy
=
EI
(,
+
^zic,).
This
equation
shows that M is
expressible
in the form
M= EI
{Ka
+
KiZ
+
K^Z-)
+ (iO\XSt
(41)
In like manner we
may
show that the
bending
moment in the
plane
of
(y,
z)
is
expressible
in the form
EF
('
+
K^z
+
K^'z")
+ const.
We shall show
immediately
how the constants
may
be determined.
242. Relation between the curvature and the
bending
moment.
We shall consider the case in which one end ^^
=
is held
fixed,
the other
end z
=
l is free from
traction,
and the load is
statically equivalent
to a
force W
per
unit of
length acting
at the centroid of the cross-section in the
direction of the axis of x*. The
bending
moment M is
given by
the
equation
M=^W{l-zy, (42)
and the
comparison
of this
equation
with
(41) gives
the
equations
K,=:-Wl/EI,
K,
=
^W/EI.
(43)
We observe
that,
if the constant added to the
right-hand
member of
(41)
were
zero,
the relation between the
bending
moment and the curvature
would be the same as in uniform
bending by
terminal
couples
and in
bending
*
The
important
case of a beam
supported
at tlie
ends,
and
carrying
a load IC
per
unit of
length,
can be treated
by compounding
the solution for a beam with one end
free,
bent
by
the
uniform
load,
with that for a beam bent
by
a terminal transverse load
equal
to
-
J
Wl.
346 NATURE OF THE SOLUTION OF THE PROBLEM
[CH.
XVI
by
terminal load. The constant in
question
does not in
general
vanish.
To determine it we observe that the value of ilf at 2;
=
is
-
jj
X
[^e^'-'
+
o-(Z^<'
+
F,/')] dxdy,
and therefore
M-EI{K,
+
K^z +
K^z'')=\^l^-x{E{eJ''^
+
K,x)
+
a{X^'-''^
+
Yy^''^)]dxdy.
(44)
Now we
may
write down the
equations
x{X,^'>^+Yy'^)da:dy
I!
9^(i(^-2/0^.''
+^2/^.'"'}
+|{i('=-2/")^.''+*yi';"'}
dxdy
=
j
IH'^- f)X.
+
^yY.]ds+
jj[i(af-f){pX
+
X,^^^)
+
xyipY+Y,^'^)]dxdy.
Hence we have the result
M
-
EI
(/Co +Kj2+
K^Z^)
=
-\\Ex (gjj*"'
+
Kf^x) dxdy
-
a
\
{^{x-
-
y^)
X, +
xyY,}
ds
J J J
-o-jj[i(x'-
f) (pX
+
X,> )
+
xy (p
Y
+
F,'
)] dxdy (45)
Since M is
given by (42)
this
equation
determines the constant
.
The
right-hand
member of
(45)
is the value of the added constant in the
right-
hand member of
(41).
The result that the
bending
moment is not
proportional
to the
curvature*,
when
load is
applied along
the
beam, may
be illustrated
by
reference to cases in which curvature
is
produced
without
any bending
moment. One such case is afforded
by
the results of
Article
87,
if we
simply interchange
the axes of
7/
and 2. It then
appears
that a stress-
system
in which all the
stress-components except Yy vanish,
while

has the form

Eax,
can be maintained
by
surface tractions of amount Eaxcoa
(j/, v) parallel
to the axis of
y.
These tractions are
self-equilibrating
on
every section,
and there is no
bending
moment.
The
corresponding displacement
is
given by
the
equations
M=
^a((T^2+y^

trz^),
V
=
axy,
w=
(toXZ,
80 that the central-line
(^=0, y
=
Q)
is bent to curvature o-a.
Another case is afforded
by
the state of stress
expressed by
the
equations
X^=Eax, Yy=Eax, Xy=-Eay, X,= Y,=Z.=0,
which can be maintained
by
surface tractions of amounts
&i{.cos(:K, v)

ycos{y, v)], Ea{x

con
(j/, v)ycos{x, v)}
parallel
to the axes of x and
y.
These tractions are
self-equilibrating
on
every section,
*
The result was obtained first
by
K. Pearson. See
Introductimi,
footnote 92. The formula
(45)
is due to J. H.
Michell,
he. cit.
p.
339. The amount of the extra curvature in some
special
cases is calculated in Article 244.
242-244]
OF THE UNIFORMLY LOADED BEAM 847
and there is no
bending
moment. The
corresponding displacement
is
given by
the
equations
u
=
a{^(l-cr)x^-^{S
+
a-)2/-
+
az^},
y
=
a(l o-)^,
w=

2aaxz,
and the curvature of the central-line is 2<ra.
If we consider a slice of the beam between two normal sections as made
up
of filaments
having
a direction transverse to that of the
beam,
and
regard
these filaments as bent
by
forces
applied
at their
ends,
it is clear that the central line of the beam must receive a
curvature, arising
from the contractions and extensions of the
longitudinal filaments,
in
exactly
the same
way
as transverse filaments of a beam bent
by
terminal load receive a
curvature. The
tendency
to antiolastic curvature which we remarked in the case of a
beam bent
by
terminal loads aifords an
explanation
of the
production, by
distributed
loads,
of some curvature over and above that which is related in the
ordinary way
to the
bending
moment. This
explanation suggests
that the effect here discussed is
likely
to be
most
important
in such structures as
suspension bridges,
where a load carried
along
the
middle of the
roadway
is
supported by
tensions in rods attached at the sides.
243. Extension of the central-line.
The fact that the central-line of a beam bent
by
transverse load
is,
in
general,
extended
or contracted was noted
long ago
as a result of
experiment*,
and it is not difficult to see
beforehand that such a result must be true.
Consider,
for
example,
the case of a beam of
rectangular
section loaded
along
the
top.
There must be
pressure
on
any
horizontal
section
increasing
from zero at the lower surface to a finite value at the
top.
With this
pressure
there must be associated a contraction of the vertical filaments and an extension
of the horizontal filaments. The value of the extension of the horizontal central-line is
determined
by
means of the formula
(39).
Since the stress is not
expressed completely by
the vertical
pressure,
this extension is not
expressed
so
simply
as the above
argument
might
lead us to infer.
The result that
fo?^*^ ^^^V
^ otherwise
expressed by saying
that the neutral
plane,
if
there is
one,
does not contain the central-line. In
general
the locus of the
points
at which
e vanishes,
or there is no
longitudinal
extension, might
be called the
"
neutral-surface."
If it is
plane
it is the neutral
plane.
244. Illustrations of the
theory.
(a)
Form
of
the solution
of
the
problem of plane
strain. When the
body
force is
the
weight
of the
beam,
and there are no surface
tractions,
we
may
make some
progress
with the solution of the
problem
of
plane
strain
(Article 239)
without
finding x-
In this
case, putting X=g,
F=0,
we see that the solution of the
stress-equations (32)
can be
expressed
in the form
X;
=
~"-XeJ)-5'p.r-2<2^[x
+
J<r.r3
+
(l-i<r)^/],
Yy'J^^-XeJ")
-2.,^[x
+
(l+|<r)a.y],
^
(46)
"
dxdy'
where Q must be
adjusted
so that the
equation
of
compatibility (28)
is satisfied. We
may
show that this
equation
leads to the
following equation
for Q :

Vi<Q
=
2^iK2(2-f (T).r
(47)
If we take the
particular
solution
^^;^.,(2
+
,r)_^
(48)

90
*
Fabr6,
Paris C.
R.,
t. 46
(1858).
348 SPECIAL SOLUTIONS OF THE PROBLEM
[CH.
XVI
we find for
Xj',
... a set of values
involving
surface
traction,
and an additional stress-
system
must be
superposed
so as to annul this surface traction without
involving any body
force
;
in other words a
complementary
solution of
Vi*Q
=
must be added to the value of
Q
given
in
(48),
and this solution must be
adjusted
so that the
boundary
conditions are
satisfied.
(ft)
Solution
of
the
problem of plane
strain
for
a beam
of
circular section bent
by
its
ovm
weight.
When the
boundary
is a circle
x^+y^^a^,
we have
x=-{i+y)a''^H{=^-^=^y^)\ (49)
and the surface values of the
stress-components given by (46),
when Q is
given by (48),
can
be
simplified by observing that,
in accordance with
(36), gp
=
fXK^cfi (1-1- o-).
It will be found
that these values are
given by
the
equations
^x'=/xK2^
(:i-3-|-3^2)_xe(0)- J^^,,(;,J-3^y),
I
ry'=IJiK^~,-i5x^
+
Sxy^)-\eJ0)
+
^^K^{l+ia)a'-.v
+
i,,<rK^(a;^-3xy^)\ (50)
24
2-
24
^/=-M<2^"^(y'
+
3y^-^).
The surface tractions
arising
from the terms in
^ikj
~^r^
'^*" ^ annulled
by superposing
the
stress-system*
,, ,
2
+(r
, ,
2
+
a-

2
+
it
, ,.,,
-U
=
--^
IJi.'^2"''x, Fj
= -
-g-
/iKjffl^^, Xy''
=
-^^
iiK^a^y
(ol)
The surface tractions
arising
from the term in
jiK^a'^x
can be annulled
by superposing
the
stress-system
x.'=o, r;=-M<2(i+|o-)A-, jr/=o (52)
The surface tractions
arising
from the terms in
(utk^ (afl-Sxy^)
can be annulled
by super-
posing
the
stress-system
X,'
=
/xo-<2^(J.272-|y-t-2'V'), r,'
=
^^aK^x{-.1-^:^
+
1f
+
^a^},
A'/
=
^a<0{-x^
+
^(2/''-a^)} (53)
The
stress-components J^x, Fy, X(")
are therefore
determined,
and thus the
problem
of
plane
strain is solved for a circular
boundary.
(c)
Correction
of
the curvature in this case. In the case of a beam of circular
section bent
by
its own
weight
we
may
show that
(o
=
0,
or the central-line is
unextended,
and that
2gp
P
(
7-H2o- +4(T^a-^'
"o-E
If the curvature were calculated from the
bending
moment
by
the
ordinary
rule the second
term in the bracket would be absent. Thus the correction to the curvature
ari.sing
from
the distribution of the load is small of the order
"linear dimension of cross-sectionl^
length
of beam
J
A consideration of the form of
(45)
would show that this result holds in
general
for a beam
bent
by
its own
weight
+.
P
*
Some of the solutions of the
problem
of
plane
strain in a circular
cylinder
which are
required
here were
given
in Article 186.
t
Solutions of the
problem
of the
bending
of a circular or
elliptic cylinder by
loads dis-
tributed in certain
special ways
have been
given by Pearson, Quart.
J.
of
Math.,
vol. 24
(1889),
and
by
Pearson and
Filon, Quart.
J.
of Math.,
vol. 31
(1899).
244]
OF THE UNIFORMLY LOADED BEAM 349
{d)
Narrmv
rectangular
heam loaded
along
the
top.
The
theory may
be ilhistrated further
by
the case of a beam of
rectangular
section
and small breadth loaded
uniformly along
its
upjier
surface. We shall treat the
problem
as one of
generalized plane stress*,
and we shall
neglect
the
weight
of the beam. Let 2a
be the
depth
of the
beam,
26 the
breadth,
and I the
length.
Take the axis of z
along
the
horizontal
central-line,
and the axis of oc
vertically
downwards at the fixed
end,
z=0. Let
Wdenote the load
per
unit of
length.
The
average stress-components X^, Z,, X,
can be
expressed
in the forms
X,=
-^~-^EK^{fl^x~l^\
Z,=
-EAx+
%EK.iX^~E(K^z^K.^z^x,
X.=
-
^E^a^-x^) (ki-)-2k22),
where,
in order to
satisfy equation (42),
we must have
ZW Z\n . ZWV-
/,
2a2
A
.(66)
0+wO
^'^)
4AV6' 8^a36
The curvature of the central-line can be shown to be
^-(2-|-<r) K.2a^-|-K,z-H(C22^,
which is
equal
to
The term
containing (|-f
o-)
a^
gives
the correction of the curvature that would be
calculated
by
the
ordinary
rule.
The extension of the central-line can be shown to be o- WI4bE
;
it is
just
half as
great
as the extension of the beam when free at the
ends, supported along
the
base,
and
carrying
the same load
along
the
top.
The neutral surface is
given by
the
equation
^[3^-^
+
3(f
+
<r)-(2
+
<r)^]
=
2,ra.
At a considerable distance from the free end the
depth
of this surface below the central-
line is
nearly equal
to
^aa^l{l

zy.
The result that the neutral surface is on the side of
the central-line towards the centres of curvature has been verified
experimentally
t.
() Doubt)/ stipported
beam. If we
superijose
on the
stress-system
found in
(55)
that due to a load

^
Wl at the end z
=
l,
we shall obtain the solution for a narrow rect-
angular
beam bent
by
uniform load W
per
unit of
length
and
supported
at both ends.
The additional
stress-system
is
given,
in accordance with the results of Article
95,
by
the
equations
and the
average
stress in the beam is
expressed by
the formulae
- "i w
*
The
problem
has been discussed
by
J. H.
Micliell, Quart.
J.
of Math.,
vol. 31
(1900),
and
also
by
L. N. G.
Filon,
Phil. Trans.
Hoy.
Soc.
(Ser. A),
vol. 201
(1903),
and Proc.
Roy. Soc,
vol. 72
(1904).
t
See a
paper by
E. G.
Coker, Edinburgh Boy.
Soc.
Trans.,
vol. 41
(1904), p.
229.
.(67)
CHAPTER XVII.
THE THEORY OF CONTINUOUS BEAMS.
245. Extension of the
theory
of the
bending
of beams.
In
previous Chapters
we have discussed certain exact solutions of the
problem
of the
bending
of beams
by
loads which are
applied
in
special
ways.
In the
problem
of the beam bent
by
a load concentrated at one
end
(Chapter
XV.)
we found that the
"
Bernoulli- Eulerian
"
theorem of the
proportionality
of the curvature to the
bending
moment is verified. In the
problem
of the beam bent
by
a load distributed
uniformly along
its
length
(Chapter XVI.)
we found that this theorem is not
verified,
but
that,
over and
above the curvature that would
present
itself if this theorem were
true,
there
is an additional constant
curvature,
the amount of which
depends upon
the
distribution over the cross-section of the forces
constituting
the load. We
appear
to be
justified
in
concluding
from these results
that,
in a beam
slightly
bent
by any
forces,
the law of
proportionality
of the
bending
moment
to the curvature is
sufficiently
exact at sections which are at a considerable
distance from
any place
of
loading
or of
support,
but
that,
in the
neighbour-
hood of such a
place,
there
may
be an additional local curvature. We
endeavoured to trace the circumstances in which the additional curvature
can become
very important,
and we solved some
problems
in which we found
it to be
unimportant.
From the results that we obtained we
appear
to
be
justified
in
concluding
that,
in most
practical problems relating
to
long
beams,
the additional curvature is not of
very
much
importance.
The state of stress and strain that is
produced
in the interior of a
beam,
slightly
bent
by any
forces,
may
be taken to be
given
with sufficient
approximation by
Saint-Venant's solution
(Chapter xv.)
at all
points
which
are at a considerable distance from
any place
of
loading
or of
support*
;
and
again,
at a
place
near the middle of a considerable
length
over which the
load is distributed
uniformly
or
nearly uniformly, they may
be taken to be
*
This view is confirmed
by
L. Pochhammer's
investigation
of the strain in a circular
cylinder
deformed
by given
forces. See his
Untersuchungen
Uber das
Gleichgewicht
det elastischen
Stabes,
Kiel,
1879.
245]
EFFECT OF SURFACE LOADING OF BEAMS
861
given
with sufficient
approximation by
Michell's solution
(Chapter xvi.).
But we have not so detailed information in
regard
to the state of stress or
strain near to a
place
of concentrated load or to a
place
of
support.
Near to
such a
place
the actual distribution of the forces
applied
to the beam must
be
very
influential.
Attempts
have been made to
study
the state of strain
at such
places experimentally.
In the research of Carus Wilson* a beam of
glass
of
rectangular
section,
supported symmetrically
on two rollers
B, C,
was
bent
by
means of a third roller A above its
middle,
and the state of strain in
the line AI)
(Fig.
32)
was examined
by
means of
polarized light
transmitted
352 EFFECT OF SURFACE LOADING OF BEAMS
[CH.
XVII
honzontal
tension, Z
=
_
(^^- -^j
+
_
(^-_ -jy,
rtical
pressure,
-
F,
=

{-
-
|)
,
vert
where b is the
depth
of the
beam,
and 2a is the
span
BC. The stress is
equivalent
to mean tension
unaccompanied by shearing
stress at those
points
at which
Xx=Yy.
In order that these
points may
be real we must
have
Ga/b
>
40/7r,
or
(span/depth)
> 4-25
nearly.
When this condition is
satisfied there are two such
points.
The
positions
of these
points
can be
determined
experimentally,
since
they
are characterized
by
the absence of
any doubly
refractive
property
of the
glass,
and the actual and calculated
positions
were found to
agree very closely.
A
general
theory
of two-dimensional
problems
of this character has been
given by
Filon*.
Among
the
problems
solved
by
him is included that of a
beam of infinite
length
to one side of which
pressure
is
applied
at one
point.
The
components
of
displacement
and of stress were
expressed by
means of
definite
integrals,
and the results are rather difficult to
interpret.
It is clear
that,
if the solution of this
special problem
could be obtained in a
manage-
able
form,
the solution of such
questions
as that discussed
by
Stokes could be
o.btained
by synthesis.
Filon concluded from his work that Stokes's value for
the horizontal tension
requires
correction,
more
especially
in the lower half
of the
beam,
but that his value for the vertical
pressure
is a
good approxi-
mation. As
regards
the
question
of the relation between the curvature and
the
bending
moment,
Filon concluded that the Bernoulli-Eulerian theorem is
approximate!}'
verified,
but
that,
in
applying
it to determine the deflexion
due to a concentrated
load,
account
ought
to be taken of a term of the same
kind as the so-called
"
additional deflexion due to
shearing
"
[Article
235
(e)].
Consider
for
example
a beam BG
supported
at both ends and
carrying
a
concentrated load W at the middle
point
A
(Fig. 38).
Either
part,
AC or
44w
iw.
Y
W
Fig.
33.
*
Phil. Tram.
Roy.
Soc.
(Ser. A),
vol. 201
(1903).
Reference
may
also be made to a thesis
by
C. Eibidre,
Sur divers cas de la
flexion
des
priemes rectangles, Bordeaux,
1888.
245,
246]
CONTINUOUS beams
858
AB,
of the beam
might
be treated as a
cantilever,
fixed at A and bent
by
terminal load
^Wacting upwards
at the other
end;
but Saint-Venant's solution
would not be
strictly applicable
to the
parts
AB or
AC,
for the cross-sections
are distorted into curved surfaces which would not fit
together
at A. In
Saint-Venant's solution of the cantilever
problem
the central
part
of the cross-
section at A is
vertical,
and the
tangent
to the central-line at A makes with
the horizontal a certain small
angle
s^.
[Article 232(c).]
Filon concluded
from his solution that the deflexion of the
centrally
loaded beam
may
be
determined
approximately by
the double cantilever
method,
provided
that the
<;entral-line at the
point
of
loading
A is taken to be bent
through
a small
angle,
so that AB and AG are inclined
upwards
at the same small
angle
to
the horizontal. He estimated this small
angle
as about
f-So-
The correction of the central deflexion which would be obtained in this
way
would be
equivalent,
in the case of a narrow
rectangular beam,
to
increasing
it
by
the fraction
45(]?^/16Z2
of
itself,
where I is the
length
of the
sjian,
and d is the
depth
of the beam. The
correction is therefore not
very important
in a
long
beam.
It must be understood that the
theory
here cited does not state that the central-line is
bent
through
a small
angle
at the
point immediately
under the concentrated load. The
exact
expression
for the
displacement
shows in fact that the direction is continuous at this
point.
What the
theory
states is that we
may
make a
good approximation
to the deflexion
by assuming
the Bernoulli-Eulerian curvature-theorem

which is not
exactly
true

and
At the same time
a.ssuming
a
discontinuity
of direction of the central-line

which does not

really
occur.
246. The
problem
of continuous beams*.
In what follows we shall
develope
the
consequences
of
assuming
the
Bernoulli-Eulerian curvature-theorem to hold in the case of a
long
beam,
of
small
depth
and
breadth,
resting
on two or more
supports
at the same
level,
and
bent
by
transverse loads distributed in various
ways.
We shall take the beam
to be
slightly
bent in a
principal plane.
We take an
origin anywhere
in the line
of the
supports,
and draw the axis of a:
horizontally
to the
right through
the
supports,
and the axis of
y vertically
downwards. The curvature is
expressed
with sufficient
approximation
by d^yjdx'.
The tractions exerted across a
normal section of the
beam,
by
the
parts
for which x is
greater
than it is at
the section
upon
the
parts
for which x is
less,
are
statically equivalent
to
a shearinof force
N,
directed
parallel
to the axis of
y,
and a
couple
G in the
plane
of
{x, y).
The conditions of
rigid-body equilibrium
of a short
length
Ax of the beam between two normal sections
yield
the
equation
s+^=^

*
The
theory
was initiated
by
Navier. See
Introduction, p.
22.
Special
oases have been
discussed
by many
writers, among
whom we
may
mention
Weyrauch, Aufgaben
zur Theorie
elaslischer
Ki'irper, Leipzig
1885.
L. E.
-23
354
CONTINUOUS BEAMS
[CH.
XVII
The
couple
G is taken to be
expressed by
the
equation
(2)
da?
where B is the
product
of
Young's
modulus for the material and the moment
of inertia of a normal section about an axis
through
its centroid at
right
angles
to the
plane
of
{x, y)*.
The senses of the force and
couple,
estimated
as
above,
are indicated in
Fig.
34.
Except
in
estimating
B no account is
taken of the breadth or
depth
of the beam.
N
Fig.
34.
In the
problems
that we shall consider the
points
of
support
will be
taken to be at the same level. At these
points
the condition
y
=
must be
satisfied.
At a free end of the beam the conditions iV"=
0,
G
=
must be
satisfied. At an end which rests
freely
on a
support (or
a
"
supported
"
end)
the conditions are
y
=
0,
G
=
0. At an end which is
"
built-in
"
{encastr6)
the direction of the central-line
may
be taken to be
prescribed f.
In the
problems
that we shall solve it will be taken to be hoiizontal. The
displace-
ment
y
is to be determined
by equating
the flexural
couple
G at
any
section,
of which the centroid is
P,
to the sum of the moments about P of all the
forces which act
upon any portion
of the
beam,
terminated towards the left
at the
section^.
This method
yields
a differential
equation
for
y,
and the
constants of
integration
are to be determined
by
the above
special
conditions.
The
expressions
for
y
as a function of x are not the same in the two
portions
of the beam
separated by
a
point
at which there is a concentrated
load,
or
by
a
point
of
support,
but these
expressions
must have the same value at the
point
;
in other
words,
the
displacement y
is continuous in
passing through
the
point.
We shall assume also that the direction of the
central-line,
or
dyjdx,
is continuous in
passing through
such a
point. Equations (1)
and
(2)
show that the
curvature,
estimated as
d-y/dx\
is continuous in
passing
through
the
point.
The difference of the
shearing
forces N calculated from
the
displacements
on the two sides of the
point
must balance the concen-
trated
load,
or the
pressure
of the
support
;
and thus the
shearing
force,
and therefore also
d^yjda?,
is discontinuous at such a
point.
*
B
is often called the
"
flexural
rigidity."
t Such an end is often described as
"clamped."
X
This
is,
of
course,
the same as the sum of the
moments,
with reversed
signs,
of all the
forces which act
upon any portion
of Uie beam terminated towards the
right
at the section.
246,
247]
EFFECT OF LOAD ON ONE SPAN
365
247.
Single span.
We consider first a number of cases in which there are two
points
of
support
situated at the ends of the beam. In all these cases we denote the
length
of the
span
between the
supports by
I.
(a)
Terminal
forces
and
couples.
M. M,
Fig.
35.
Let the beam be
subjected
to forces Y and
couples
M^
and
M^
at the ends
A and B. The forces Y must be
equal
and
opposite,
and,
when the senses
are those indicated in
Fig.
35,
they
must be
expres.sible
in terms of
Mi
and
ilfo
by
the
equation
IY=M-M,.
The
bending
moment at
any
section x is
{I

x)
Y
+
M^,
or
M(l-x)ll
+
M,x/l.
The
equation
of
equilibrium
is
accordingly
_
d\
,. I

-
+
^"I-
Integrating
this
equation,
and
determining
the constants of
integration
so that
y may
vanish at a;
=
and at a;
=
^,
we find that the deflexion is
given
by
the
following equation
:
By=-^l-'x{l-x){Mo{2l-x)
+
Mi(l
+
x)} (3)
The deflexion
given by
this
equation may
be described as
"
due to the
couples
at the ends of the
span."
(b) Uniform
load.
Supported
ends.
Aiwl
iwli
Z\
-2\
Fig.
36.
Taking
w to be the
weight per
unit of
length
of the
beam,
we observe
that the
pressures
on the
supports
are each of them
equal
to
^wl.
The
232
356 EFFECT OF LOAD
[CH.
XVII
moment about
any point
P of the
weight
of the
part
BP of the beam is
}^w{l

xf,
and therefore the
bending
moment at
P,
estimated in the sense
already explained,
is the sum of this moment and
^wl{l x),
or it is

^wxQ

x).
The
equation
of
equilibrium
is
accordingly
Integrating
this
equation,
and
determining
the constants of
integration
so that
y may
vanish at a;
=
and &% x
=
l,
we find the
equation
ay
=
i^wx{l-x)[r
+
x{l-x)\ (4)
If we refer to the middle
point
of the
span
as
origin, by putting
x
=
^l
+ x',
we find
By
=
q-,w{\i:^-x''){ll^-x'y
(c) Uniform
load. Built-in ends.
The solution is to be obtained
by adding
to the solution in case
(J)
a solution of case
(a)
adjusted
so that
dyjdx may
vanish at x
=
and x=l. It is clear from
symmetry
that
M^
=
Mf,
and Y=0. We have therefore
By
=
^iwx {l-x)(P+lx-T^)- ^Mx (I
-
x),
where Mis written for
J/q
or
M^.
The terminal conditions
give
M=^wl\
and the
equation
for the deflexion becomes
By=iiwx^{l-xf,
or,
referred to the middle
point
of the
span
as
origin
of
y,
it becomes
(rf)
Concentrated load.
Supported
ends.
2\ Ts
w
Fig.
37.
Let a load W be concentrated at a
point
Q
in
AB,
at which x
=
^.
We
shall write
^'
for
l-^,
so that
AQ=^
and
BQ
=
^'.
The
pressures
on the
supports
A and B are
equal
to
W^'/l
and
W^/l respectively.
The
bending
moment at
any point
in
AQ,
where
f
>a;>0,
is

W^'x/l;
and the
bending
moment at
any point
in
BQ,
where I > x >
^,
is

W^(l

x)/l.
2'*7] ON ONE SPAN
The
equations
of
equilibrium
are
accordingly
857
We
integrate
these in the forms
B
(y
-
a; tau
a)
=
-
J
i"'
Tf^V,
B
(y
-
(i
-
a,')
tau
/3j
= -
^
i-
Ff (^
-
a;)^
where tan a and tan
^
are the downward
slopes
of the central-line at the
points
A and B. The
conditions of
continuity
of
y
and
dy/dx
at
Q
are
B
^
tan a
-
^
Z-'
TTff
=
Bf
tan
/3
-
^
-> Tf
^f,
B tan a
-
i
^'
}Fff
= -
B tan
/3
+
^
^l >f
fp.
These
equations
give
B tan a
=
^
/->
W^^' (^
+
2r),
B tan
,S
=
^
Z->
TT^f (2f
+
f)-
Hence in
AQ,
where
^
> >
0,
we have
By
=
^l-^W^'{^(^+2^')w-a?l (5)
and in
BQ,
where
l>x>^,
we have
^y=^i-'W^{^'m
+
^){i-x)-{i-xy\ (t))
We observe that the deflexion at
any point
P when the load is at
Q
is
equal
to the
deflexion at
Q
when the same load is at P.
The central deflexion due to the
weight
of the
beam,
as determined
by
the sokition
of case
{b),
is the same as that due to
|
of the
weight
concentrated at the middle of
the
span.
(e)
Concentrated load. Built-in ends.
Fig.
38.
To the vahies of
By given
in
(5)
and
(6)
we have to add the value of
By given
in
(3),
and determine the constants
M^
and
i/, by
the conditions that
dyjdx
vanishes
at ^-
=
and at x
=
l. We find
from which
i/o= W^^-^jl^ i/i= W^^'jl^
358 THE THEOREM OF
[CH.
XVII
Hence in
AQ,
where
^>x>0,
we have
B3/=JZT'ir|V{3f(Z-^)-^'a:},
and in
BQ,
where
l>x>^,
we have
We notice that the deflexion at P when the load is at
Q
is the same as the deflexion
at

when the same load is at P.
The
points
of inflexion are
given by dh//da;^=0,
and we find that there is an inflexion
at
Pi
in
AQ
whore
APi=:AQ.AB/(ZAQ
+
BQ).
In like manner there is an inflexion at
P^
in
BQ
where
BP^=BQ.AB/{3BQ
+
AQ).
The
point
where the central-line is horizontal is
given by d7//d.v
=
0. If such a
point
is in
AQ
it must be at a distance from A
equal
to twice
AP^,
and for this to
happen
AQ
must be
>BQ. Conversely,
if
AQ<BQ,
the
point
is in
BQ
at a distance from B
equal
to twice
BP^.
The forces
Yq
and
Fj
at the
supports
are
given by
the
equations
248. The theorem of three moments*.
'^ Ai
Boa B.
n
Ma Mb
M^ M^
Fig.
39.
Let
A, B,
G he three consecutive
supports
of a continuous beam
resting
on
any
number of
supports
at the same
level,
and let
M^t, M, Mc
denote the
bending
moment.s at
A, B,
C. Denote the
shearing
forces on the two sides
of the
support
B
by
B^
and
B^,
with a similar notation for the others. The
pressure
on the
support
B is
B^
+
B^.
Now
Bo
is determined
by taking
moments about A for the
equilibrium
of the
span
AB,
and
JSj
is determined
by taking
moments about C for the
equilibrium
of the
span
BG. Hence the
pressure
B^
+
Bi
can be
expressed
in terms of the
bending
moments at
A, B,G
when the manner of
loading
of the
spans
is known.
Again,
the deflexion in the
span
AB
may
be obtained
by adding
the deflexion due to the load on this
span
when its ends are
supported
to that due to the
bending
moments at the
ends.
[Article
247
(a).]
The deflexion in the
span
BG
may
be determined
*
The theorem is due to
Clapeyron.
See
Introduction, p.
22. Generalizations have been
given by
various writers
among
whom
may
be mentioned M.
L^vy, Statiqiie graphique,
t.
2,
Paris
1886,
who treats the case where the
supports
are not all in the same level
;
R. R.
Webb,
Cambridge
Phil. Soc.
Proc,
vol. 6
(1886),
who treats the case of variable flexural
rigidity;
K.
Pearson, Messenger of
Math.,
vol. 19
(1890),
who treats the case in which the
supports
are
slightly compressible.
247,
248]
THREE MOMENTS 359
by
the same method. The condition of
continuity
of direction of the central-
line at B becomes then a relation
connecting
the
bending
moments
&tA,B,
G.
A similar relation holds for
any
three consecutive
supports.
This relation
is the theorem of three moments.
By
means of this
relation,
combined with
the
special
conditions which hold at the first and last
supports,
the
bending
moments at all the
supports
can be calculated.
To
express
this
theory analytically,
we take an
origin anywhere
in the
line of the
supports,
and draw the axis of x
horizontally
to the
right,
and the
axis of
y vertically
downwards. We take the
points
of
support
to be at
x
=
a, b, c,
... The
lengths
of the
spans,
b

a,
c

b,
... will be denoted
by
^AB' ^Bc>---
We
investigate
a series of cases.
(a) Uniform
load.
Let w be the load
per
unit of
length.
The deflexion in AB is
given,
in accordance with the results of Article 247
(o)
and
(6), by
the
equation
^1/
=
ij
w
{cc
-
a) {b
-
x) {{b

(if -\-{x-a){b- x)]

-^{x- a) {b
-
x) {Mb (6
+ a;
-
2a)
+
M^{2b-x-
a)}/(b
-
a).
A similar
equation may
be written down for the deflexion in BC. The
condition that the two values of
dyjdx
at a.'
=
6 are
equal
is
-4^w(b-ay
+
:^
i2M
+
M^) ib-a)
=
^V wic-bf-^ {2Mj,
+
Ma) (c
-
b),
and the
equation
of three moments is therefore
Iab (M^
+
^Mb)
+
V' {2Mj,
+
Mc)
=
\w
(IW
+
Ibc^ (7)
To determine the
pressure
on the
support
B we form the
equations
of moments for AB
about
A,
and for BC about C. We have
^0^-
-
*
"'l^-?
-
-V + M,,
=
0,
Bi
ho
-
\
wIbc^ -Ms+Ma=0.
These
equations give ^
and
i?i,
and the
pressure
on the
support
B is
Bo+B^.
In this
way
the
pressures
on all the
supports may
be calculated.
(6) Equal spans.
When the
spans
are
equal, equation (7) may
be written as a linear diflference
equation
of the second order in the form
and the solution is of the form
^fn
=
^wP
+
Aa''
+
B^,
where A and B are
constants,
and a and
^
are the roots of the
quadratic
,j:2^4^.^i
=
0,
or we have
a=-2 +
V3, ^=-2-V3.
The constants A and B are to be determined from the values of M at the first and last
supports.
(c) Uniform
load on each
span.
Let
w,fB
denote the load
per
unit of
length
on the
sjjan AB,
and
Wg^
that on BC. Then
we
find,
in the same
way
as in case
(a),
the
equation
of three moments in the form
l^B (-if.-,
+
^Mg)
+ he
(2.1f
+^c)
=
i
Wab Ia^ +
\
Vbc Inch
360 GKAPHIC METHOD OF SOLUTION OF THE
[CH.
XVII
(d)
Concentrated load on one
span.
Let a load W be concentrated at a
point
Q
in BG
given by

=
|.
The
deflexion in AB is
given,
in accordance with the results of Article 247
(a),
by
the
equation
By
=
-i(a;-a)(b-x) {M^ {2b
-
x
-
a)
+
Maib
+
x-
2a)}/(b
-
a),
and that in
BQ
is
given by
By=^iW[(^-b){c-^K2c-b-^)(x-b)-{c-^)(a>-bmc-h)
-^(x-b)ic-x){Ms{2c-x-b)
+
Mc{c
+
x-2b)]/{c-b).
The condition of
continuity
of
dyjdx
at
=
6 is
l{M^
+
2M^){b-a)
=
^W{^-b){c-^){2c-b-mc-b)-l{2M+Mc){c-b\
and the
equation
of three moments for
A, B,
is therefore
Iab {M^
+
23/^)
+
lj,c {2M
+
Mc)
=
Wl^^^l^c
0-
+
kSBc),
(8)
where
I^q
and
Iq^.
are the distances of
Q
from B and C. In like
manner,
if D is the next
support beyond
G,
the
equation
of three moments for
B, G,
D is
Ibc {Mb
+
2Mc)
+
IcD i^Mc
+
Md)
=
Wls^l^c
(1
+
IbqIU)
(9)
249.
Graphic
method of solution of the
problem
of continuous
beams*.
The
equation
of
equilibrium (2),
viz. B
-7^
=
G,
is of the same form as the
equation determining
the curve assumed
by
a loaded
string
or
chain,
when
the load
per
unit
length
of the horizontal
projection
is
proportional
to

G.
For,
if T denotes the tension of the
string,
m the load
per
unit
length
of the
horizontal
projection,
and ds the element of arc of the
catenary
curve,
the
equations
of
equilibrium,
referred to axes drawn in the same
way
as in
Article
246,
are
mdx ^
d
f ,d%i\
dx -
T
;^
=
const.
=
T
say,
:r{fii)
+
'^7r.
=
^'
ds
\ ds)
ds
and these
lead,
by
elimination of
T,
to the
equation
It follows that the form of the curve assumed
by
the central-line of the
beam in
any span
is the same as that of a
catenary
or funicular curve
determined
by
forces
proportional
to Q^x on
any length
hx of the
span,
provided
that the funicular is made to
pass through
the ends of the
span.
The forces Ohx are to be directed
upwards
or downwards
according
as G is
positive
or
negative.
*
The method is due to Mohr. See
Introduction,
footnote 99.
248,
249]
PROBLEM OF CONTINUOUS BEAMS 361
The
tangents
of such a funicular at the ends of a
span
can be determined
without
finding
the
funicular,
for
they depend only
on the statical resultant
and moment of the fictitious forces GBx. To see this we take the ends of the
span
to be a?
=
and x
=
l,
and
integrate
the
equation (2)
in the forms
dy
P
da;
^
Jo
B
and hence we obtain the
equation
,<iy.
PG_, fK, sG
dx,
(;-.)|
+
,
-/;<,-.,!
dx,
'i-ii-PHy-^
B
dx,
from which it follows that
dxJo Jo
tB
\dxJi Jo
xG
dx.
These values
depend only
on the resultant and resultant moment of the
forces
GSx,
and therefore the direction of the central-line of the beam at the
ends of the
span
would be determined
by drawing
the
funicular,
not foi the
forces
GSx,
but for a
statically equivalent system
of forces.
The flexural
couple
G at
any point
of a
span
AB
may
be found
by adding
the
couple
calculated from the
bending
moments at the
ends,
when there is
no load on the
span,
to the
couple
calculated from the load on the
span,
when the ends are
"
supported."
The
bending
moment due to the
couples
at the ends of the
span
is
represented graphically by
the ordinates of the
line A'B' in
Fig.
40,
where AA' and BB'
represent
on,
any
suitable
scale the
bending
moments at A and B. The
bending
moment due to
A'
Fig.
40.
Fig.
41.
uniform load on the
span
is
equal
to
-^7vx(l

x),
as in Article 247
(b),
and
it
may
be
represented by
the ordinates of a
parabola
as in
Fig.
41. The
bending
moment due to a concentrated load is
equal
to

Wx
(l

^)/l,
when
^
> a; >
0,
and to
-
W
(l
-
x) ^jl,
when Z > a; >
^,
as in Article 247
(d)
;
and it
may
be
represented by
the ordinate of a broken line as in
Fig.
42. The
362 GRAPHIC METHOD OF SOLUTION OF THE
[CH.
XVII
bending
moment due to the load on the
span may
be
represented
in a
general way by
the ordinate of the thick line in
Fig.
43.
A'
Fig.
42.
Fig.
43.
The fictitious forces QBx are
statically equivalent
to the
following
:

(i)
a force
<f> represented by
the area of the
triangle
AA'B,
acting upwards
through
that
point
of trisection
g
of AB which is nearer to
A, (ii)
a force
<f>',
represented by
the area of the
triangle
A'BB',
acting upwards through
the
other
point
of trisection
g'
of
AB,
(iii)
a force
F,
represented by
the area
contained between AB and the thick line in
Fig.
43,
acting
downwards
through
the centroid of this area. We take the line of action of F to meet AB in the
point
G. When the load on the
span
is
uniform,
F
=
^wP,
and G is at the
middle
point
of AB. When there is an isolated
load,
F-\W ^{l

^),
and
G is at a distance from A
equal
to
^ (Z
+
f).
The forces F and the
points
G are known for each
span,
and the
points
g, g'
are known also. The forces
0, <^'
are
unknown,
since
they
are
propor-
tional to the
bending
moments at the
supports,
but these forces are connected
by
certain relations. Let
A^^, Ai,
... denote the
supports
in
order,
let
^i, <pi,Fi
denote the
equivalent system
of forces for the first
span
A^Ai,
and so on.
Let
Mo, Mj, M,
... denote the
bending
moments at the
supports.
Then we
observe,
for
example,
that
<pi
:
<f)2
=
M^.A^A^
:
M^.A^A^,
and therefore the
ratio
(^i'
:
^j
is known.
Similarly
the ratio
(f)^
:
<f>3
is
known,
and so on.
If the force.s
(fy, <!>',
as well as
F,
were known for
any span,
we could
construct a funicular
polygon
for them of which the extreme sides could be
made to
pass through
the ends of the
span
Since the direction of the
central-line of the beam is continuous at the
points
of
support,
the extreme
sides of the funiculars which
pass through
the common
extremity
of two
consecutive
spans
are in the same
straight
line. The various funicular
polygons belonging
to the different
spans
form therefore a
single
funicular
polygon
for the
system
of forces
consisting
of all the forces
(/>, (/>',
F.
250.
Development
of the
graphic
method.
The above results enable us to construct the funicular
just
described,
and
to determine the forces
<^,
or the
bending
moments at the
supports,
when the
249,
250]
PROBLEM OF CONTINUOUS BEAMS
363
bending
moments at the first and last
supports
are
given.
We consider the
case where these two
bending
moments are
zero*,
or the ends of the beam
are
"supported."
We denote the sides of the funicular
by
1, 2, 3,
... so that
the sides
1, 3, 6,...
pass through
the
supports
A, Ai, A.,,
... .
Fig.
44.
We consider the
triangle
formed
by
the sides
2, 3,
4. Two of its vertices
lie on fixed
lines,
viz.: the verticals
through g^'
and
g^.
The third vertex
F,
also lies on a fixed line. For the side 3 could be
kept
in
equilibrium by
the
forces
(f>/
and
^2
and the tensions in the sides
2, 4,
and therefore
F,
is on the
line of action of the resultant of
(pi
and
(p^;
but this line is the vertical
through
the
point
a^,
where
a^g^
=
A^g^'
and
a^/
=
.4
1^2,
for
<f>i
:
(f)^
=
A^g^'
:
Ajg,.
Again,
the
point
Cg
where the side 2 meets the vertical
through
.4,,
is
determined
by
the condition that the
triangle
formed
by
the sides 1 and 2
and the line
A^C^
is a
triangle
of forces for the
point
of intersection of the
sides 1 and
2,
and
AqG
represents
the known force
Fi
on the scale on which
we
represent
forces
by
lines. Since the vertices of the
triangle
formed
by
the sides
2, 3,
4 lie on three fixed
parallel
lines,
and the sides 2 and 3
pass
through
the fixed
points
C, and
A^,
the side 4
passes through
a fixed
point C^,
which can be constructed
by drawing any
two
triangles
to
satisfy
the stated conditions.
In the above the
point
C2
may
be taken
arbitrarily,
but,
when it is
chosen,
AfiGi
represents
the constant horizontal
component
of the tension in the
sides of the funicular on the same scale as that on which
AoG^
represents
the force
F^.
We
may
show in the same
way
that the vertices of the
triangle
formed
by
the sides
5, 6,
7 lie on three fixed vertical
lines,
and that its sides
pass
through
three fixed
points.
The vertical on which the intersection F, of
*
The sketch of the
graphic
method
given
in the text is not intended to be
complete.
For
further details the reader is referred to M.
Levy,
loc. cit.
p.
358. A
paper by Perry
and
Ayrton
in Proc.
Roy.
Soc,
vol. 29
(1879), may
also be consulted. The memoir
by
Canevazzi cited in the
Introdtiction,
footnote
99,
contains a
very
luminous account of the
theory.
364 GRAPHIC METHOD FOR CONTINUOUS BEAMS
[CH.
XVII
the lines 5 and 7 lies
passes through
the
point
a^,
where
a^f/s
=
J.jg','
and
a^g^'
=
A^gs.
The fixed
point C^, through
which the side 5
passes,
is on the
vertical
through
C^,
and at such a distance from
Oj
that this vertical and the
sides 4 and 5 make
up
a
triangle
of forces for the
point
of intersection of the
sides 4 and 5. The line
Cfi^
then
represents
the force F^ on a certain
scale,
which is not the same as the scale on which
A^C^
represents
F^,
for the
horizontal
projection
of
G^Ct
represents
the constant horizontal
component
tension in the funicular on the scale on which
CjCj
represents
F^.
Since
Ct
is
known,
the ratio of scales in
question
is
determined,
and
C^
is therefore
determined. The side 6
passes through
the fixed
point
A^,
and the fixed
point
Cv.
through
which the side 7
passes
can be constructed in the same
way
as
G^
was constructed.
In this
way
we construct two series of
points
(7,, C^,
...
C^k-i,
... and
Gi, O7,
...
Osi+i,
.... We construct also the series of
points
aj, a,,,
...
a^,
...
,
where
a/^g,!
=
A^g^+i
and
a^g^^^
=
A^g^. By
aid of these series of
points
we
may
construct the
required
funicular.
Consider the case of n
spans,
the end
A,i,
as well as
A^,
being simply
supported.
The line
joining
C^n-i
to
An
is the last side
(Sn 1)
of the
funicular,
since the force
^',
like
<^],
is zero. The side
(.3n

2)
meets the
side
(3

1)
on the line of action of
F^,
and
passes through
the
point
Csn-o.
Let this side
(3n

2)
meet the vertical
through
a_i
in
F_i.
Then the line
Vn-\ C'3_4
is the side
(3n

4).
The side
(3m

3)
is determined
by joining
the
point
where the side
(3/i

2)
meets the vertical
through ^
to the
point
where the side
(3n

4)
meets the vertical
through
(7'_j
. This side
(3?i

3)
necessarily passes through
A^-i
in
consequence
of the mode of construction
of the
points
G.
Proceeding
in this
way
we can construct the funicular.
When the funicular is constructed we
may
determine the
bending
moments at the
supports by
measurement
upon
the
figure.
For
example,
let the side 4 meet the vertical
through
A^
in
S^.
Then
A^Si
and the sides 3
and 4 make
up
a
triangle
of forces for the
point
of intersection of 3 and 4.
The horizontal
projection
of either of the sides of this
triangle
which are not
vertical is
^A^A,,.
Hence
A^Si
represents
the force
<p
on the same scale as
^A-^A^ represents
the horizontal tension in the sides of the funicular. Thus
A-^SijA^A^ represents
the force
<^2
on a constant scale. But
(f>.2
represents
the
product
of
il/i
and
A^A2
also on a constant scale. Hence
A^S^IA-^A.? repre-
sents the
bending
moment at
A-^
on a constant scale. In like
manner,
if the
side .3^+1 meets the vertical
through
A^.
in the
point S^,
then
A^,S^IA,cAk+\
represents
the
bending
moment at
A^.
CHAPTER XVIII.
GENERAL THEORY OF THE BENDING AND TWISTING OF THIN RODS.
251. Besides the
problem
of continuous beams there are
many physical
and technical
problems
which can be treated as
problems concerning long
thin
rods, and,
on this
understanding,
are
capable
of
approximate
solution. In
this
Chapter
we shall consider the
general theory
of the behaviour of such
bodies,
reserving
the
applications
of the
theory
for
subsequent Chapters.
The
special
circumstance of which the
theorj'
must take account is the
possibility
that the relative
displacements
of the
parts
of a
long
thin rod
may
be
by
no
means
small,
and
yet
the strains which occur in
any part
of the rod
may
be
small
enough
to
satisfy
the
requirements
of the mathematical
theory.
This
possibility
renders
necessary
some
special
kinematical
investigations, subsidiary
to the
general analysis
of strain considered in
Chapter
I.
252. Kinematics of thin rods*.
In the unstressed state the rod is taken to be
cylindrical
or
prismatic,
so
that
homologous
lines iu different cross-sections are
parallel
to each other.
If the rod is
simply
twisted,
without
being
bent,
linear elements of different
cross-sections which are
parallel
in the unstressed state become inclined to
each other. We select one set of linear
elements,
which in the unstressed
state are
parallel
to each other and lie
along principal
axes of the cross-
sections at their centroids. Let
B/he
the
angle
in the strained state between
the directions of two such elements which lie in cross-sections at a distance
Ss
apart.
Then lim
Sf/Ss
measures the tndst.
When the rod is
bent,
the twist cannot be estimated
quite
so
simply.
We shall
suppose
that the central-line becomes a tortuous curve of curvature
1/p
and measure of
tortuosity 1/S.
We take a
system
of fixed axes of
x,
y,
z
of which the axis of z is
parallel
to the central-line in the unstressed
state,
and the axes of
x,
y
are
parallel
in the same state to
principal
axes of the
*
Cf. Kelvin and
Tait,
Nat.
Phil.,
Part
I, pp.
94 et
seq.,
and
Kirchhoff,
J.f.
Math.
(Crelle),
Bd. 56
(1859),
or Ges.
Abhandlungen (Leipzig 1882), p. 285,
or
Vorlesungen
iiber math.
Physik,
Mechanik, Vorlesung
28.
366 EXTENSION CURVATURE AND TWIST
[CH.
XVIII
cross-sections at their centroids. Let P be
any point
of the
central-line, and,
in the unstressed
state,
let three linear elements of the rod issue from P in
the directions of the axes of
x,
y,
z. When the rod is deformed these linear
elements do not in
general
continue to be at
right angles
to each
other,
but
by
means of them we can construct a
system
of
orthogonal
axes of
x, y,
z.
The
origin
of this
system
is the
displaced position
Pj
of
P,
the axis of z is the
tangent
at
Pj
to the strained
central-line,
and the
plane {x, z)
contains the
linear element
which,
in the unstressed
state,
issues from P in the direction
of the axis of x. The
plane
of
{x, z)
is a
"
principal plane
"
of the rod. The
sense of the axis of x is chosen
arbitrarily.
The sense of the axis of z is
chosen to be that in which the arc s of the
central-line,
measured from some
assigned point
of
it,
increases
;
and then the sense of the axis of
y
is deter-
mined
by
the condition that the axes of
x,
y,
z in this order are a
right-handed
system.
The
system
of axes constructed as above for
any point
on the strained
central-line will be called the
"
principal
torsion-flexure axes
"
of the rod at
the
point.
Let P' be a
point
of the central-line near to
P,
and let
P/
be the
displaced
position
of P'. The
length
&i
of the arc
PiP^'
of the strained central-line
may
differ
slightly
from the
length
& of PP'. If e is the extension of the
central-line at
Pj
we have
lim(8s,/&)
=
(l
+) (1)
s=o
The extension e
may
be zero. For
any application
of the mathematical
theory
of
Elasticity
to be
possible,
it must be a small
quantity
of the order of
the strains
contemplated
in the
theory.
Suppose
the
origin
of a frame of three
orthogonal
axes of
x,
y,
z to move
along
the strained central-line of the rod with unit
velocity,
and the three
axes to be directed
always along
the
principal
torsion-flexure axes of the rod
at the
origin
of the frame. We
may
resolve the
angular velocity
with which
the frame rotates into
components
directed
along
the instantaneous
positions
of the axes. We shall denote these
components by
k, k,
t. Then k and k'
are the
components of
curvature of the strained central-line at
Pj,
and t is the
twist
of the rod at
P,.
These statements
may
be
regarded
as definitions of the twist and com-
ponents
of curvature. It is clear that the new definition of the twist coincides
with that which was
given
above in the case of a rod which is not
bent,
and
that
K,
It are the
curvatures,
as defined
geometrically,
of the
projections
of
the strained central-line on the
planes
of
(y,
^)
and
{x, z),
and therefore the
resultant of k and k is a vector directed
along
the binormal of the strained
central-line
and
equal
to the curvature
\jp
of this curve.
253. Kinematical formulae.
We
investigate
in the first
place
the relation between the twist of the rod
252,
253]
OF A THIN BOD
367
and the measure of
tortuosity
of its strained
central-line. Let
I, m,
n denote
the direction cosines of the binormal of this curve at
Pi
referred to the
prin-
cipal
torsion-flexure axes at
P,,
and let
/', m',
n' denote the direction cosines
of the binormal at
P/
referred to the
principal
torsion-flexure axes at
P,'.
Then the limits such as lim
{l'-l)/Bs,
are denoted
by dl/ds.,
....
Agaiu
let
s,=o

l
+
Bl,...
denote the direction cosines of the binormal at
P/
referred to the
principal
torsion-flexure axes at
Pj.
We have the formulse*
lim
Sl/Ssi
=
dl/dSi

niT +
hk,
lim
Bm/Ssi
=
dm/dsi

n/c
-f
Ir,
\
s,=o
lim
Bn/Ssj
=
dn/dsi

Ik'
+
niK.
The measure of
tortuosity 1/2
of the strained central-line is
given by
the
formula
1/2^
=
lim
[(Blf
+
(Bmy
+
{Bny]/(Bs,y,
and the
sign
of S is determined
by choosing
the senses in which the
principal
normal,
binormal and
tangent
of the curve are drawn. We
suppose
the
prin-
Fig.
45.
cipal
normal
(marked
n in
Fig.
45)
to be drawn towards the centre of
curvature,
and the
tangent
to be drawn in the sense in which
s, increases,
and we choose the sense in which the binormal
(marked
b in the
figure)
is
drawn in such a
way
that the
principal
normal,
the binormal and the
tangent,
taken in this
order,
are
parallel
to the axes of a
right-handed system.
Now
we
may put
l
=
Kp
=

cos
f,
m
=
k'p
=
sin
f,

=
0,
where
p
is the radius of curvature
;
and then iir
/is
the
angle
between the
*
Cf. E. J.
Routh,
Dynamics of
a
syxtem of rigid
bodies
(London 1884),
Part
II,
Chapter
I.
368 KINKMATICAL FORMULA
[CH.
XVIII
principal plane {x, z)
of the rod and the
principal
normal of the strained
central-line. On
substituting
in the
expression
for
l/S'^,
and
making
use of
the above
convention,
we find the
equation
_dl
1
(2)
in which
ts,nf=

(K'/K) (3)
The
necessity
of
introducing
such an
angle as/
into the
theory
was noted
by
Saint-Venant*. The case in which
/
vanishes or is constant was the
only
one considered
by
the earlier writers on the
subject.
The linear elements of
the deformed rod which issue from the strained central-line in the direction
of the
principal
normals of this curve
are,
in the unstressed
state,
very
nearly
coincident with a
family
of lines at
right angles
to the central-line.
If
f
vanishes or is constant these lines are
parallel
in the unstressed state.
We
may
describe a state of the bent and twisted rod in which
f
vanishes
or is constant as such that the
rod,
if
simply
unbent,
would be
prismatic.
When
/
is variable the
rod,
if
simply
unbent,
would be a twisted
prism,
and the twist would be
df/dst.
With a view to the calculation of
k,
k
,
t we take the axes of
x,
y,
z at
P,
to be connected with
any system
of fixed axes of
x,
y,
z
by
the
orthogonal
scheme
(4)
253]
RELATING TO A THIN BOD 369
the
equations
are
multiplied by
1
+ e. If
k, k,
t are themselves
small,
and
quantities
of the order k are
neglected,
the factor 1
+
e
may
be
replaced by
unity.
If
K, K,
T are not
regarded
as small
quantities,
a first
approximation
to their values can be obtained
by replacing
1
+ e
by unity.
For the esti-
mation of
K, K,
T we
may
therefore
ignore
the distinction between
ds^
and ds
and write our formula3
dl.

=
?3 -T- +
rria

=
+ ri;
dm^ drhi
ds
dlj
ds
ds
dm..
ds
-=h'iF
+
^h^
+
n.'^. }
, dL dm.
ds
drii
.(6)
The direction cosines
li,...
can be
expressed
in terms of three
angles
6,
yjr, <f),
as is usual in the
theory
of the motion of a
rigid body.
Let be
the
angle
which the axis of z at
P,
makes with the fixed axis of
z, rjr
the
angle
which a
plane parallel
to these axes makes with the fixed
plane
of
(x, z),
(p
the
angle
which the
principal plane (x, z)
of the rod at
P,
makes with the
plane
zP^z.
Then the direction cosines in
question
are
expressed by
the
equations
l^= -sin-\//-sin (^
+
cos
r//'
cos
<^
cos
d, ij
=
cos\/^sin<^
+
sin\Jf cos<^cos
6, ,= -sinflcos<^,'\
;,=
-sini/<-cos<^-cosi|f sin0co.s5, i.2
=
cos\|^cos(/>- sinx^siiK^cos^, ^2=
sin5sin(/),y
l^=
sin 5 cos
xl',
=
8in
Bsm-^,
ra,=
COS 6.
(-!)
Fig.
46.
L. E.
24
370
EQUATIONS
OF
EQUILIBRIUM
[CH.
XVIII
The relations
connecting d0/ds, dy^jds, d^jds
with
k, k,
t are obtained at
once from
Fig.
46
by observing
that
k, k',
t are the
projections
on the
principal
torsion-flexure axes at
Pj
of a vector which is
equivalent
to vectors
dd/ds, difr/ds, d(j)/ds
localized in certain lines. The line
Pjf
in which
dd/ds
is localized is at
right angles
to the
plane
zP^z,
and
dyjr/ds
and
dcfijds
are
localized in the lines
PjZ
and
PiZ.
We have therefore the
equations
d0 . .
dyjr
.
^
.
,
dd
.
dxjr
. . .
,
K-T-
Sin
9

-J-
sin v cos
<p,
k
=
-^
cos
<p
+
-~-
sin a sin
^,

-i*-f--
(>
254.
Equations
of
equilibrium.
When the rod is deformed the action of the
part
of it that is on one side
of a cross-section
upon
the
part
on the other side is
expressed,
in the usual
way, by
means of tractions estimated
per
unit of area of the section. These
tractions are
statically equivalent
to a force
acting
at the centroid of the
section and a
couple.
The axis of z
being
directed
along
the
tangent
to the
central-line at this
centroid,
the tractions on the section are denoted
by
Xi, Yz, Zz.
The
components parallel
to the axes of
x,
y,
z of the force- and
couple-resultants
of these tractions are
N,
N'
,
T and
G, G', H,
where
N=llx,dxdij,
N'^jJY.dxdy,
T
=
jjz,dxdy,
)
i-(9)
G
=
IJyZ,dxdy,
G'
=
[(
-xZ.dxdy,
H=[j{xY,-yX,)dxdy,
the
integrations being
taken over the area of the section. The forces
N,
N'
are
"shearing-forces,"
the force T is the
"tension,"
the
couples
G,
G' are
"flexural
couples,"
the
couple
H is the "torsional
couple."
The forces
N, N',
T will be called the
stress-resultants,
and the
couples
G, G',
H the
stress-couples.
The forces
applied
to the rod are estimated
by
means of their force- and
couple-resultants per
unit of
length
of the
central-line, and,
in thus
estimating
them,
we
may disregard
the extension of this line. Let the forces
applied
to
the
portion
of the rod between the cross-sections drawn
through
Pj
and
P/
be reduced
statically
to a force at
Pj
and a
couple
;
and let the
components
of this force and
couple,
referred to the
principal
torsion-flexure axes at
Pj ,
be denoted
by [Z], [F], [Z],
and
[K\, [K'\ [6].
When
P/
is
brought
to
coincidence with
Pj
all these
quantities
vanish,
but the
quotients
such as
[XJ/Ss
can have finite limits. Let us write
lim
[Xyts
=
Z,
. . .
,
lim
[Z]/&
=
K,...;
=o as=o
then
X, Y,
Z are the
components
of the force-resultant at
Pi
per
unit of
253-254]
OF A THIN ROD 371
length
of the
central-line,
and
K, K',
are the
components
of the
couple-
resultant.
Now the forces
applied
to the
portion
of the rod contained between two
cross-sections balance the resultant and resultant moment of the tractions
across these sections. Let S denote the excess of the value of
any quantity
belonging
to the section
through
P/
above the value that
belongs
to the
section
through
Pj,
let
x,
y,
z denote the coordinates of
Pj
referred to fixed
axes, x',
y',
z' those of
any point
on the central-line between
P,
and
P/.
Using
the scheme
(4),
we can at once write down the
equations
of
equi-
librium of the
portion
in such forms as
h{kN
+
kN'
+
kT)
+
\
{l,X
+
kY
+
kZ)
ds
=
0,
J
8
and
B{kO
+
kG'+kH)
+
Bij {{n.M
+ n,N'
+
n,T)
+
8
(n,N
+n,N'
+
7hT)}
-
Sz
{{m,N
+
nhN' +
m^T)
+
S
{in,N
+
nhN'
+
m,T)}
fs+Ss
+
{(y'
-
y)
(^i^
+ 2
^+
'>hZ)
-
{z
-
z) {m^X
+
m.
Y+
m^Z)]
ds
J
s
re+is
+
(l,K
+ LK' +
k@)ds
=
0.
J s
We divide the left-hand members of these
equations by
Bs,
and
pass
to a
limit
by diminishing
Bs
indefinitely.
This
operation requires
the
performance
of certain differentiations. The results of
differentiating
li,
... are
expressed
by equations
(5),
since the extension of the central-line
may
be
disregarded.
We choose the fixed axes of
x,
y,
z to coincide with the
principal
torsion-
flexure axes of the rod at
Pj. Then,
after the differentiations are
performed,
we
may put
^i
=
1, wh
=
0,
and so on. The limits of
Bx/Bs, ByfBs, Bz/Ss
are
0, 0,
1. The limits of such
quantities
as
rs+Ss rs+Ss
(Bs)-' {l,X
+
LY+hZ)ds, (Bs)-' (lJ<:
+
l,K'
+
l,@)ds
are
X, Y,
Z -and
K, K',
@. The limits of such
quantities
as
(Bs)-'
r'
V
-
y) (i^
+ ^ Y+
n,Z)
ds
are zero. We
have, therefore,
the
following
forms for the
equations
of
equilibrium*:
as
y (10) ^-Tk
+ Nt+ 7
=
0,
ds
~-Nk'+N'k+Z=0,
ds
*
The
equations
were
given by Clebsch, Elasticitat, 50,
but
they
were
effectively
contained in
the work of
Kirchhoff,
loc. cit.
p.
3(55.
242
372 rORMULffi FOR THE STRESS-COUPLES
[CH.
XVIII
and
^-G't
+
Hk'-N'
+
K
=
0,
^
as
^-Hic+Gr
+ If+K'
=
0,
"j-
(11)
^-^-Gk'
+ G'k
+
@
=
0.
as
In addition to these
equations
there will in
general
be certain
special
conditions which hold at the ends of the rod. These
may
be conditions of
fixity,
or the forces and
couples applied
at the ends
may
be
given.
In the
latter case the terminal values of the stress-resultants and
stress-couples
are
prescribed.
These
special
conditions
may
be used to determine the constants
that are introduced in the
process
of
integrating
the
equations
of
equilibrium.
255. The
ordinary approximate theory.
The
equations
of
equilibrium
contain nine unknown
quantities
:
If, iV', T,
G, G', H, K, K,
T. It is clear
that,
if three additional
equations connecting
these
quantities
could be
found,
there would be suflScient
equations
to deter-
mine the curvature and twist of the rod and the stress-resultants and
stress-couples.
The
ordinary approximate theory

a
generalization
of the
"
BernouUi-Eulerian
"
theory

consists in
assuming
that the
stress-couples
are connected with the curvature and twist of the rod
by equations
of
the form
G
=
Ak,
G'
=
Bk, H=Ct,
(12)
where
A, B,
G are constants
depending
on the elastic
quality
of the material
and the
shape
of the cross-section. The nature of this
dependence
is known
from the results obtained in
comparatively simple
cases. For
isotropic
material
we should have
A
=
E(ok\
B
=
E(ok'\
where E is
Young's
modulus for the
material,
m is the area of the cross-
section,
and h and k' are the radii of
gyration
of the cross-section about the
axes of a; and
y,
which are
principal
axes at its centroid. In the same case G
would be the torsional
rigidity
considered in
Chapter
xiv. If the cross-
section of the rod has kinetic
symmetry,
so that A
=
B,
the flexural
couples
G, G',
as
expressed
in the formulse
(12),
are
equivalent
to a
single couple,
of
which the axis is the binormal of the strained
central-line,
and the
magnitude
is
B/p,
where
p
is the radius of curvature of this curve.
The
theory
is
obviously incomplete
until it is shown that the formulse
(12)
are,
at least
approximately,
correct. An
investigation
of this
question,
based
partly
on the work of KirchhofFand Clebsch* will now be
given.
*
See
Introduction, pp.
23 and 24.
254-256]
STRAIN IMPLIED IN BENDING AND TWISTING
373
256. Nature of the strain in a bent and twisted rod.
In Kirchhoff's
theory
of thin rods much
importance
attaches to certain
kinematical
equations.
These
equations
are not free
from
difificulty,
and the
following investigation,
which is direct if a little
tedious,
is offered as a sub-
stitute for the kinematical
part
of Kirchhoff's
theory.
We
suppose
that a thin
rod is
actually
bent,
so that the central-line has a certaia
curvature,
and
twisted,
so that the
"
twist
"
has a certain
value,
and we seek to ascertain the
restrictions,
if
any,
which are
thereby imposed upon
the strain in the rod.
For the sake of
greater generality
we shall
suppose
also that the central-line
undergoes
a certain small extension.
Now we can
certainly imagine
a state of the rod in which the cross-
sections remain
plane,
and at
right angles
to the
central-line,
and suffer no
strain in their
planes
;
and we
may suppose
that each such section is so
oriented in the normal
plane
of the strained central-line that the
twist,
as
already
defined,
has the
prescribed
value. To
express
this state of the rod
we denote
by
x,
y
the coordinates of
any point
Q,
lying
in the cross-section of
which the centroid is
P,
referred to the
principal
axes at P of this cross-
section. When the section is
displaced bodily,
as
explained
above,
the
point
P moves to
P,,
and the coordinates of
Pi,
referred to
any
fixed
axes,
may
be
taken to be
x,
y,
z. The
principal
axes at P of the cross-section
through
P
are moved into the
positions
of the axes of
x,
y
at
Pj
defined in Article 252.
The state of the rod described above is therefore such that the
coordinates,
referred to the fixed
axes,
of the
point Qi,
to which
Q
is
displaced,
are
x
-f-
Zi
4-
l^y, y
+
m^x
+
m^y,
z
-}-
n^x -f-
ri^y,
where
li,
... are the direction cosines defined
by
the scheme
(4).
Any
state of the
rod,
which involves the
right
extension and curvature
of the central-line and the
right
twist,
may
be derived from the state
just
described
by
a
displacement
which,
in the case of a thin
rod,
must be
small,
for one
point
in each cross-section and one
plane
element drawn
through
each
tangent
of the central-line are not
displaced.
Let
^, ;, f
be the
components
of this additional
displacement
for the
point Q,
referred to the axes of
x,
y,
z
at the
point
Pj.
The
coordinates,
referred to the fixed
axes,
of the final
position
of
Q
are
x
+
k{x+^)
+
L{y
+
v)
+
k^,
y
+
Wi(a;-f-^)-(-m2(y-t-i7)-f-TO3f,
z-l-ni(a;-)-^)
+
n2(2/
+
'7)
+
n3? (13)
To estimate the strain in the rod we take a
point
Q'
near to
Q.
In the
unstrained state
Q' will,
in
general,
be in a normal section different from that
drawn
through
P. We take it to be in the normal section drawn
through
P',
so that the arc PP'
=
hs. We take the coordinates of
Q'
referred to the
principal
axes at P' of the cross-section drawn
through
P' to be a; +
hx,
y
-(-
by.
374 NATURE OF THE STRAIN IN A
[CH.
XVIII
Then
Sx,
By,
Bs are the
projections
on the fixed axes of the linear element
QQ'.
We take r to be the
length
of this
element,
and write
Bx

Ir,
By
=
inr,
Bs
=
nr,
so that
I, m,
n are the direction
cosines,
referred to the fixed
axes,
of the line
QQ'.
We can write down
expressions
like those in
(13)
for the coordinates of
the final
position
of
Q',
and we can therefore
express
the
length
r^
of the line
joining
the final
positions
of
QQ'
in terms of r and
I, vi,
n. Since the direction
I, m,
n is
arbitrary,
the result
gives
us the six
components
of strain.
In
obtaining
the
length
r^
we must
express
all the
quantities
which
involve ?
correctly
to the first
order,
but
powers
of r above the first
may
be
neglected.
To obtain the
expressions
for the coordinates of the final
position
of
Q'
we note the
changes
that must be made in the several terms of
(13).
The
quantities
x,
y,
z, li,
... are functions of s
only,
but the
quantities f
, rj, ^
are functions oi
x,
y,
s. We must therefore in
(13)
replace
,
,
9x
1 3y
, 3z
X
by
X +
^
nr,
y by y
+
^
nr,
z
by
z
+
^
nr,
X
hy
X
+
Ir,
y hy y
+
mr,
^ky^
+
Sir
+
Smr
+
f^nr,....
Further the
quantities 9x/3s,
. . . are
given by
the
equations
j^
=
(l+e)k,
^
=
(l
+
e)m
^
=
(1
+
e)
3,
and the
quantities dlj/ds,...
are
given by
the
equations
|^
=
(l
+
e)(4T-U')>-
where the coeflScients of
(1
+
e)
are the
right-hand
members of
equations (5).
It follows that the difference of the x-coordinates of the final
positions
of
Q
and
Q'
is
(1
+
e)
;3
+
:
I
(l
+
^j
i
+
1^
m
+
^
nl
+
(1
+
e) (kr
-
k
')
n
{x
+
f)
256]
BENT AND TWISTED ROD
375
For the differences of the
y-
and z-coordinates we have similar
expressions
with
nil, m,, 73
and
n^, n^, n,
respectively
in
place
of
l^, l^, I3.
Since the
scheme
(4)
is
orthogonal,
the result of
squaring
and
adding
these ex-
pressions
is
r- l
+
^^]l
+
^Im
+
f^n
+
(l+e)n{K'^-r(y
+
r,)}
dx. ds
+
?=
+
r=
dx
\
dyj
dx
771 +
^
n
+
{l+)n [t {x
+
^)
-
K^}
^ +
^"^
+
s?
+
(i
+
^){i
+
(y
+
'?)-' (*
+
!)} .(14)
dx
dy
ds
and this is
Vi".
We have therefore
expressed
Vi'
in the form of a
homogeneous
quadratic
function of
I, m,
n.
Now,
the strains
being
small, r-^
is
nearly equal
to
r,
and we can write
n^
=
r^(l
+
2e),
where e is the extension in the direction
I, m,
n. Further we shall have
e
=
Bxxl'^
+
Byym^
+
e^n^
+
Cyzmn
+
Czxnl
+
Cxylm,
where the
quantities
Cxx,
are the six
components
of strain. The coefficient
of I in the first line of the
expression (14)
must be
nearly equal
to
unity,
and the coefficients of m and n in this line must be
nearly
zero. Similar
statements mutatis mutandis hold with
regard
to the coefficients of
I, m,
n in
the
remaining
lines. We therefore obtain the
following expressions
for the
components
of strain :
d^ . 3'?
and
-^ __. _?|
, ?5
.(15)
^=i+i+(i+^){'^'^-^(^+''^J'
?j.
.(16)
e,,
=
6 +
^+(l+6){(2/+7;)-'(a;
+
^)}.
In
obtaining
the formulae
(15)
and
(16)
we have not introduced
any
approximations except
such as arise from the consideration that the strains
are
"
small," and,
in
particular,
that
e,
being
the extension of the
central-line,
must be small. But we can
see,
without
introducing any
other
considerations,
that the terms of
(16),
as
they
stand,
are not all of the same order of
magni-
tude. In the first
place
it is clear that the terms

ry,
tx,
Ky,

k'x must be
small
;
in other
words,
the linear dimensions of the cross-section must be
small
compared
with the radius of curvature of the
central-line,
or with the
reciprocal
of the twist. Such terms as
k'^, tt;,
... are small also. We
may
.(17)
376 NATURE OF THE STRAIN IN A
[CH.
XVIII
therefore omit the
products
of e and these small
quantities,
and rewrite
equations (16)
in the forms
Now the
position
of the
origin
of
x,
y,
and that of the
principal plane
of
{x, z),
are unaffected
by
the
displacement (^, t), if),
and therefore this
displace-
ment is
subject
to the restrictions :
(i) ^' V> ^
vanish with x and
y
for all values of
s,
(ii) dr)/dx
vanishes with x and
y
for all values of s.
We conclude
that,
provided
that the strain in the rod is
everywhere
small,
the
necessary
forms of the
strain-components
are
given by equations
(15)
and
(17),
where the functions
^, -rj, f
are
subject
to the restrictions
(i)
and
(ii).
257.
Approximate
formulae for the strain.
We have now to introduce the
simplifications
which arise from the
consideration that the rod is
"
thin." The
quantities ^, rj, ^ may
be
expanded
as
power
series in x and
y,
the coefficients in the
expansions being
functions
of s
;
and the
expansions
must be valid for
sufficiently
small values of x and
y,
that is to
say
in a
portion
of the rod near to the central-line*. There are
no constant terms in these
expansions
because
^, rj, f
vanish with x and
y.
Further
d^/dx
and
d^/dy
must be small
quantities
of the order of admissible
strains,
and therefore the coefficients of those terms of
^
which are linear in
X and
y
must be small of this order. It follows that
^
itself must be small of
a
higher
order, viz.,
that of the
product
of the small
quantity d^/dx
and the
small coordinate x. Similar considerations
apply
to
rj
and
f.
As a first
step
in the
simplification
of
(17)
we
may
therefore omit such terms as

rr], k'^.
When this is done we have the
formulse-j*
and these with
(15)
are
approximate expressions
for the
strain-components.

The
expansions may
not be valid over the whole of a cross-section. The failure of
Cauchy's
theory
of the torsion of a
prism
of
rectangular
cross-section
(Introduction,
footnote
8.5) sufficiently
illustrates this
point.
But the
argument
in the text as to the relative order of
magnitude
of
such terms as
ry
and such terms as
tt;
could
hardly
be affected
by
the restricted
range
of
validity
of the
expansions.
t
It
may
be observed that Saint-Venant'a formulte for the torsion of a
prism
are included
256-258]
BENT AND TWISTED ROD 377
Again
we
may
observe that similar considerations to those
just
adduced
in the case of
f apply
also in the case of
d^/ds
;
this
quantity
must be of the
order of the
product
of the small
quantity d-^jdxds
and the small coordinate
X,
which is the same as the order of the
product
of the small
quantity d^jdx
and the small fraction
xjl,
where I is a.
length comparable
with
(or equal to)
the
length
of the rod.
Thus,
in
general, d^jds
is small
compared
with
3f/9x.
Similar considerations
apply
to
977/3,9
and
d^/ds*.
As a second
step
in the
simplification
of
(17)
we
may
omit
d^/ds, drjjds, d^/ds
and obtain the
formulaef
^'"^dx~'^^'
^"'^87"^'
e,,^
=
e-KX
+
Ky (19)
Again
we
may
observe that in Saint-Venant's solutions
already
cited e
vanishes,
and in some solutions obtained in
Chapter
xvi. e is small
compared
with KX. In
many important problems
e is small
compared
with such
quan-
tities as TX or KX. Whenever this is the case we
may
make a third
step
in
the
simplification
of the formulae
(17) by omitting
e.
They
would then read
^'^^fx~'^y' ^"^=97,
+
- e=-/c'a;
+
Ky (20)
With these we must associate the formulse
(15),
and in the set of formulae we
may suppose,
as has been
explained,
that
^, 77, f
are
approximately independent
of s.
It
appears
therefore that the most
important
strains in a bent and twisted
rod arc
(i)
extension of the
longitudinal
filaments related to the curvature of
the central-line in the manner noted in Article 232
(6), (ii) shearing
strains
of the same kind as those which occur in the torsion
problem
discussed in
Chapter
Xiv.,
(iii)
relative
displacement
of elements of
any
cross-section
parallel
to the
plane
of the section. The last of these strains is
approxi-
mately
the same for different cross-sections
provided
that
they
are near
together.
258. Discussion of the
ordinary approximate theory.
To determine the stress-resultants and
stress-couples
we
require
the
values of the
stress-components
X^, Y^, Z^.
Since
E
'(l-t-<7)(l-2<r)
in the formulse
(15)
and
(18) by putting |
=
'!
=
;
and his formuloB for
bending by
terminal
load are included
by putting
J= -(rKxy
+
iJK'(x--y-},
7;
=
itk'j;)/
+ A itk
(x^-y^).
In each case
f
must be determined
appropriately.
*
The
result,
so far as
d^jda
and
Otj/os
are
concerned,
is
exemplified by
Saint-Venant's formula
just
cited. In Saint-Venant's solutions
f
is
where
x
aid
x'
^.re the flexure
functions,
and
<i>
is the torsion function,
for the cross-section.
The functions
x
and
x'
are small of the order a-x,
where a is an
appropriate
Unear dimension
of the cross-section. In this case
f
is
actually independent
of s.
t
These are Kirchhoff's formulse.
^z
=
/-, , _N/i i^^^ l"" (^^
+
^yj)
+
(1
-
O") ^z:
378 APPROXIMATE FORMULA FOR THE
[CH.
XVIIl
where E is
Young's
modulus and a is Poisson's ratio for the
material,
the
expression
for this
stress-component
cannot be obtained without
finding
the
lateral extensions
exx,
eyy, given by
the formulae
(15),
as well as the
longi-
tudinal extension
e^, given by
the third of
(17), (18), (19)
or
(20).
To
express
the
stress-components completely
we
require
values for
^, rj, f,
and these
cannot be found
except by solving
the
equations
of
equilibrium subject
to
conditions which hold at the
cylindrical
or
prismatic bounding
surface of
any
small
portion
of the rod. If the rod is
vibrating,
the
equations
of small
motion
ought
to be solved. We
may,
however,
approximate
to the stress-
resultants and
stress-couples by retracing
the
steps
of the
argument
in the
last Article.
When there are no
body
forces or kinetic
reactions,
and the
initially
cylindrical bounding
surface of the rod is free from
traction,
the
portion
between
any
two
neighbouring
cross-sections is held in
equilibrium by
the
tractions on its ends.
According
to our final
approximation, expressed by
equations (15)
and
(20), ^, rj, f
are
independent
of
s, and,
in the
portion
of the
rod
considered, k,
k
,
r also
may
be
regarded
as
independent
of s. This
portion
of the rod
may
therefore be
regarded
as a
prism
held strained
by
tractions on
its ends in such a
way
that the
strain,
and therefore also the
stress,
are the
same at
corresponding points
in the intermediate cross-sections. The theorem
of Article 237 shows
that,
in such a
prism,
the
stress-components
Xx,
Yy, Xy
must
vanish, and,
since
e^
is
given by
the third of
(20),
we must have
a!=|-'("-")- |+S=
(^i)
Further the
stress-components
X., F.,
Z, must be
given by
Saint-Venant's
formulffi
^^
=
'"KS"4
^^
=
^^9^
+
4
^=
=
-^'('^--2/).-(22)
where
^
is the torsion function for the section
(Article
216).
The stress-
couples
are then
given by
the formulae
(12)
of Article 255. To this order of
approximation
the stress-resultants vanish.
When we retain
e,
as in the formulae
(19),
no modification is made in the
formulae for the
stress-couples,
and the
shearing
forces still vanish. To the
expression cr{K'x

Ky)
in the
right-hand
member of
(21)
we must add
the term

ae,
and the tension is
given by
the formula
T
=
Eo)e,
(23)
where m is the area of the cross-section.
When we abandon the
supposition
that
f, rj, ^
are
independent
of
s,
we
may
obtain a closer
approximation by assuming
that the
strains,
instead of
being
uniform
along
the
length
of a small
portion
of the
rod,
vary imiformly
along
this
length.
When there are no
body
forces,
and the
initially
258,
259]
TENSION and stress-couples 379
cylindrical
boundary
is free from
traction,
the theorem of Article 238 shows
that the
only possible
solutions are Saint-Venant's. The
stress-couples
and
the tension are
given by
the same
formulae as
before,
but the
shearing
forces
do not vanish.
In the
general
case,
in which forces are
applied
to
parts
of the rod other
than the
ends,
we
ought
to retain the formulie
(17)
for the
strains,
and the
formulaj
(21)
do not hold. We know from the
investigations
of
Chapter
XVI.
that the formula
(12)
and
(23)
are not
exact,
although they may
be
approxi-
mately
correct. The
corrections that
ought
to be made in them
depend upon
the distribution of the
applied
forces over the cross-sections.
From this discussion we
may
conclude that the formuloe
(12)
and
(23)
yield good approximations
to the values of the
stress-couples
and the tension
in
parts
of the rod which are at a distance from
any place
of
loading
or
support,
but
that,
in the
neighbourhood
of such
places, they
are of doubtful
validity.
Since the
equations (10)
and
(11)
combined with the formulas
(12)
determine all the
stress-resultants as well as the curvature and
twist,
the formula
(23)
determines the
extension f.
In
ordinary
circumstances e is small in
comparison
with such
quantities
as
kx,
which
represent
the extensions
produced
in non-central
longitudinal
filaments
by bending.
This
may
be seen as follows :

the order of
magnitude
of T
is,
in
general,
the same as that
of
^V,
or
If',
and this order
is, by equations (11),
that of
dO/ds.
Hence the order of e is
that of
{Ea>)~^ (dO/ds).
Now k is of the order
OjEao^,
where a is an
appropriate
linear
dimension of the
cross-sections,
and the order of kx is therefore that' of
{Ea>)~^{Oja).
Thus kx
is,
in
general,
a
very
much
larger quantity
than t.
In
any problem
in which
bending,
or
twisting,
is an
important
feature we
may,
for a
first
approximation, regard
the central-line as uuextended.
The
potential energy per
unit of
length
of the rod is
easily
found from
equations (21)
and
(22)
in the form
\{Ak'^
+
Bk''^ +
Ct'^) (24)
If there is no curvature or twist the
potential energy
is
\Ewi\
259. Rods
naturally
curved*.
The rod in the unstressed state
may possess
both curvature and
twist,
the
central-line
being
a tortuous
curve,
and the
principal
axes of the cross-sections
at their centroids
making
with the
principal
normals of this carve
angles
which
vary
from
point
to
point
of the curve. The
principal
axes of a cross-section
at its centroid and the
tangent
of the central-line at this
point
form a triad of
orthogonal
axes of
x^,
y^,
z^,
the axis of
3
being
directed
along
the
tangent.
We
suppose
the
origin
of this triad of axes to move
along
the curve with unit
*
The
theory
is
substantially
due to
Clebscli, Kliisticitiit,
55. It had been indicated in
outline
by Kirchhoff,
loc. cit.
p.
365,
380
EQUATIONS
OF
EQUILIBRIUM
OF A ROD
[CH.
XVIII
velocity.
The
components
of the
angular velocity
of the
moving
triad of
axes,
referred to the instantaneous
positions
of the
axes,
will be denoted
by
Ko, Ko, To.
Then
Ko, /c'
are the
components
of the initial
curvature,
and
To
is
the initial twist. If
l/2o
is the measure of
tortuosity
of the central-line at
any point,
and
^tt/,
is the
angle
which the
principal plane
of
(.Vo, Zo)
at the
point
makes with the
principal
normal of the
central-line,
we have the
formulae
tan/o
=
-
Ko'Iko, t
=
Ijlo
+
dfjds, (25)
which are
analogous
to
(2)
and
(3)
in Article 253.
When the rod is further bent and
twisted,
we
may
construct at each
point
on the strained central-line a
system
of
"
principal
torsion-flexure
axes,"
in the
same
way
as in Article
252,
so that the axis of z is the
tangent
of the strained
central-line at the
point,
and the
plane
of
(, z)
contains the linear element
which,
in the unstressed
state,
issues from the
point
and lies
along
the axis
of
Xf,.
By
means of this
system
of axes we
determine,
in the same
way
as
before,
the
components
of curvature of the strained central-line and the twist
of the rod. We shall denote the
components
of curvature
by
kj, /c/,
and the
twist
by
Tj.
The
equations
of
equilibrium
can be written
down,
by
the method of
Article
254,
in the forms
dN
ds
dN;
ds
dT
ds
-
N't,
+
Tk,'
+
Z
=
0,
-Nk^ +N'k,
+
Z=0,
.(26)
and
dG
\
^-G't,
+
Hk,'
-N'
+
K=0,
ds
dG'
-
Hk,
+
Gt,
+
N'
+
K'
=
0,
y
dH
ds
-GK^+G'K,+
(d
=
0.
.(27)
The rod could be held
straight
and
prismatic by
suitable
forces, and,
according
to the
ordinary approximation (Article 255),
the
stress-couples
at
any
cross-section would be

Ak^, Bk^,

Cto.
The
straight prismatic
rod
could be bent and twisted to the state
expressed by
i, k,', Tj
and
then,
according
to the same
approximation,
there would be additional
couples
Aki, Bki, Ct,.
The
stress-couples
in the rod when bent and twisted from
259]
WITH INITIAL CURVATURE AND TWIST 381
the state
expressed by
k^, k^, t^
to that
expressed by
k^, ,', Tj
would then be
given by
the formulae*
G
=
A{k,-k,), G'=(/-0. H=CiT,-r,) (28)
It is clear from the discussion in Article 258 that these formula) can be used with
greater certainty
if the rod is
subjected
to terminal forces and
couples only
than if forces
are
applied
to it
along
its
length.
It
may
be noted
that,
even when the cross-section of the rod has kinetic
symmetry
so that A
=
B,
the flexural
couples
are not
equivalent
to a
single couple
about the bi normal
of the strained central-line unless
ki'/k^'^ki/k^.
When this condition is satisfied the
flexural
couple
is of amount
iJ(l/pi

l/pu),
where
p,
and
p^
are the radii of curvature of
the central-line in the unstressed and stressed states.
The above method of
calculating
the
stress-couples requires
the ratios of the thickness
of the rod to the radius of curvature and to the
reciprocal
of the twist to be small of the
order of small strains
contemplated
in the mathematical
theory
of
Elasticity.
Unless this
condition is satisfied the rod cannot be held
straight
and untwisted without
producing
in
it strains which exceed this order. It
is, however,
not
necessary
to assume that this
condition is satisfied in order to obtain the formula!
(28)
as
approximately
correct formulse
for the
stress-couples.
We
may apply
to the
question
the method of Article
256,
and take
account of the initial curvature and twist
by
means of the
equations
lr=8x-
ijTfi Ss,
mr
=
Sy
+xtu Ss,
nr
=
8s{l
-
k^x
+
K(,y),
where
y
stands for
K^y-
k^x.
We should then find instead of
(14)
'.-[(-i)'4|-i^,|-.(^i-'|))+"r4"''-<"->
!
+<l'+|*('->+rf,{l*-(|-'|)}+^"i<-->^-<-;'->
-Ki'l+Ki";}

In
deducing approximate expressions
for the
strain-components
we denote
by [y] any
quantity
of the order of the ratio
(thickness)/(radius
of
curvature)
or
(thickness)/(reciprocttl
of
twist),
whether initial or
final,
and
by [e] any quantity
of the order of the strain.
Thus,
Toy
and
r^y
are of the order
[y]
; d^jdx
and
(kj
-
ico)y
are of the order
[e].
If,
in the above
expression
for
ri^,
we
reject
all terms of the order of the
product [y] [e]
as well as all terms
of the order
[e]-,
we find instead of
(19)
the formula;
ez^=S-{Ti-Tf,)y, ey,=
S +
{ri-Ta)x,
e
=
64-(ci-(c)y-(ci'- k,/)^.
From these we could deduce the formulas
(28)
in the same
way
as
(12)
are deduced
from
(19),
and
they
would be
subject
to the same limitations.
*
These
formulse,
due to
Clebsch,
were obtained
also, by
a
totally
different
process, by
A. B.
Basset,
Amer. J.
of Math.,
vol. 17
(1895).
CHAPTER XIX.
PROBLEMS CONCERNING THE
EQUILIBRIUM
OF THIN RODS.
260. Kirchhoff's kinetic
analogue.
We shall
begin
our
study
of the
applications
of the
theory
of the last
Chapter
with a
proof
of Kirchhoff's
theorem*,
according
to which the
equations
of
equilibrium
of a thin
rod,
straight
and
prismatic
when un-
stressed,
and held bent and twisted
by
forces and
couples applied
at its ends
alone,
can be identified with the
equations
of motion of a
heavy rigid body
turning
about a fixed
point.
No forces or
couples being applied
to the rod
except
at the
ends,
the
quantities
X, T,
Z and
K, K',
in
equations (10)
and
(11)
of Article 2-54
vanish.
Equations (10)
of that Article become
rIN dN' fJT
"^-N't+Tk'=0, ^-Tk
+
Nt=0,
^-N>c'
+
N'k=0,
...(1)
which
express
the
constancy,
as
regards magnitude
and
direction,
of the
resultant of
iV, N',
T
; and,
in
fact,
this resultant has the same
magnitude,
direction and sense as the force
applied
to that end of the rod towards which
s is measured. We denote this force
by
R.
Equations
(11)
of Article 254
become,
on substitution from
(12)
of
Article
255,
and omission of
K, K', 0,
a'^^-{B-C)k't
=
N',
B^-{C-A)tk
=
-N,
C~-{A-B)kic'
=
0.
(2)
The terms on the
right-hand
side are
equal
to the moments about the axes
of
X,
y,z
o{ a, force
equal
and
opposite
to R
applied
at the
point (0, 0, 1).
We
may
therefore
interjjret equations (2)
as the
equations
of motion of a
top,
that is to
say
of a
heavy rigid body turning
about a fixed
point.
In this
analogy
the line of action of the force R
(applied
at that end of the rod
towards which s is
measured) represents
the vertical drawn
upwards,
s
repre-
sents the
time,
the
magnitude
of R
represents
the
weight
of the
body.
A, B,
C
represent
the moments of inertia of the
body
about
principal
axes at the fixed
point, (,
k',
t) represents
the
angular velocity
of the
body
referred to the
*
G.
Kirchhofl,
toe. cit.
p.
3G5.
I
260,
261]
THEOREM OF THE KINETIC ANALOGUE 383
instantaneous
position
of this triad of axes. The centre of
gravity
of the
body
is on the (7-axis at unit distance from the fixed
point
;
and this
axis,
drawn from the fixed
point
to the centre of
gravity
at the instant
s,
is
identical,
in direction and
sense,
with the
tangent
of the central-line of the
rod,
drawn in the sense in which s
increases,
at that
point
Pj
of this line
which is at an arc-distance s from one end. The
body
moves so that its
principal
axes at the fixed
point
are
parallel
at the instant s to the
principal
torsion-flexure axes of the rod at
Pj.
On
eliminating
N and N' from the third of
equations (1) by
the aid of
equations
(2),
we find the
equation
-Y'
+
'A-K^
+ Bk
-r- +
{A-B)
TKK
=
0,
as as as
or,
by
the third of
(2),
jJT+^(AK^
+
BK'"-
+
Cr')]=0,
giving
the
equation
T + h
(Ak'
+
Bk"'
+
Gt'')
=
const
(3)
This
equation
is
equivalent
to the
energy-integral
of the
equations
of
motion of the kinetic
analogue.
261. Extension of the theorem of the kinetic
analogue
to rods
naturally
curved*.
The theorem
may
be extended to rods which in the unstressed state have curvature and
twist, provided
that the
components
of initial curvature
<, /'
and the initial twist
tq,
defined as in Article
259,
are constants. This is the case
if,
in the unstressed
state,
the
rod is
straight
but not
prismatic,
in such a
way
that
homologous
transverse lines in
different cross-sections lie on a
right
helicoid
;
or if the central-line is an arc of a
circle,
and the rod free from twist
;
or if the central-line is a
portion
of a
helix,
and the rod has
such an initial twist
that,
if
simply unbent,
it would be
prismatic.
When the rod is bent and twisted
by
forces and
couples applied
at its ends
only,
so that
the
components
of curvature and the
twist,
as defined in Article
259,
become
k,, icj', tj,
the
stress-resultants
N, N',
T
satisfy
the
equations
rlN rIN' rlT
^-iV'r,+
?'./=0,
'^-r<,-fiVVi
=
0,
^
-#..'
+
.
v.,
=
(4)
These
equations express
the result that
N, N',
T are the
components, parallel
to the
principal
torsion-flexure axes at
any section,
of a force which is constant in
magnitude
and
direction. We denote this
force,
as
before, by
R. Since the
stress-couples
at
any
section
are ^4
(k,
-
kq),
B
{k{
-
kq'),
C
(t,
-
t^
we have the
equations
\
ds
384 THE PROBLEM OF THE ELASTICA
[CH.
XIX
The kinetic
analogue
is a
rigid body turning
about a fixed
point
and
carrying
a
flywheel
or
gyrostat rotating
about an axis fixed in the
body.
The centre of
gravity
of the
flywheel
is at the fixed
point.
The direction cosines
I, m,
n of the axis of the
flywheel,
referred to
the
principal
axes of the
body
at the
point,
and the moment of momentum h of the
fly-
wheel about this
axis,
are
given by
the
equations

AK.ti
=
hl, BK^
=
hm,
-CTf^
=
hn.
.(6)
The
angular velocity
of the
rigid body
referred to
principal
axes at the fixed
point
is
(kj, K]', tj)
and the
interpretation
of the
remaining symbols
is the same as before.
262. The
problem
of the elastica*.
As a first
application
of the theorem of Article 260 we take the
problem
of
determining
the forms in which a thin
rod,
straight
and
prismatic
in the
unstressed
state,
can be held
by
forces and
couples applied
at its ends
only,
when the rod is bent in a
principal plane,
so that the central-line becomes a
plane
curve,
and there is no twist. The kinetic
analogue
is then a
rigid
pendulum
of
weight
R,
turning
about a fixed horizontal axis. The motion
of the
pendulum
is determined
completely by
the
energy-equation
and the
initial conditions. In like manner the
figure
of the central-line of the rod is
determined
completely by
the
appropriate
form of
equation (3)
and the
terminal conditions.
We take the
plane
of
bending
to be that for which the flexural
rigidity
is
B. Then k and t
vanish,
and the
stiess-couple
is a
flexural
couple
G', =Bk',
in the
plane
of
bending.
The stress-resultants are a tension T and a
shearing
force
N,
the latter directed towards the centre of
curvature. Let 6 be the
angle
which the
tangent
of the central-line at
any point,
drawn in the sense
in which s
increases,
makes with the line of action
of the force R
applied
at the end from which s is
measured
(see Fig. 47).
Then we have T=

Rcosd,
and K
=

dO/ds,
and the
equation (3)
becomes
-Rcose
+
^B (dd/dsy
=
const
(7)
In the kinetic
analogue
B is the moment of
inertia of the
pendulum
about the axis of sus-
pension,
and the centre of
gravity
is at unit
distance from this axis. The line drawn from
the centre of
suspension
to the centre of
gravity
at the instant s makes an
angle
6 with the vertical drawn downwards.
*
The
problem
of the elastica was first solved
by
Euler. See
Introduction, p.
3. The
gystematie application
of the theorem of the kinetic
analogue
to the
problem
was worked out
by
W.
Hess,
Math.
Ann.,
Bd. 25
(1885).
Numerous
special
cases were discussed
by
L. Saalschiitz,
Der belastete
Stab, Leipzig,
1880.
261-263]
FOKMS OF THE ELASTICA 385
Equation (7)
can be obtained
very simply by
means of the
equations
of
equilibrium.
These
equations
can be
expressed
in the forms
T=~Rcos0,
N=-Rsme,
^S+^=0,
as
from
which,
by putting
G'
=
-B
(dOjds),
we obtain the
equation
id'd/ds')
+ Bsme
=
0,
(8)
and
equation (7)
is the first
integral
of this
equation.
The
shape
of the
curve,
called the
elastica,
into which the central-line is
bent,
is to be determined
by
means of
equation (7).
The results take different
forms
according
as there
are,
or are
not,
inflexions. At an inflexion
dO/ds
vanishes,
and the flexural
couple
vanishes,
so that the rod can be held in the
form of an
inflexional
elastica
by
terminal force
alone,
without
couple.
The
end
points
are then
inflexions,
and it is clear that all the inflexions lie on the
line of action of the terminal force R

the line
of
thrust. The kinetic
analogue
of an inflexional elastica is an
oscillating pendulum.
Since the interval of
time between two instants when the
pendulum
is
momentarily
at rest is a
constant,
equal
to half the
period
of
oscillation,
the inflexions are
spaced
equally along
the central-line of the rod. To hold the rod with its central-
line in the form of a
non-inflexional
elastica terminal
couples
are
required
as
well as terminal forces. The kinetic
analogue
is a
revolving pendulum.
In
the
particular
case where there are no terminal forces the rod is bent into an
arc of a circle. The kinetic
analogue
in this case is a
rigid body revolving
about a* horizontal axis which
passes through
its centre of
gravity.
If the central-line of the
rod,
in the unstressed
state,
is a
circle,
and there is no initial
twist,
the kinetic
analogue (Article 261)
is a
pendulum
on the axis of which a
flywheel
is
symmetrically
mounted. The motion of the
pendulum
is
independent
of that of the
fly-
wheel,
and in like manner the
possible figures
of the central-line of the rod when further
bent
by
terminal forces and
couples
are the same as for a
naturally straight
rod. The
magnitude
of the terminal
couple
alone is altered
owing
to the initial curvature.
263. Classification of the forms of the elastica.
() Inflexional
elastica.
Let s be measured from an
inflexion,
and let a be the value of 6 at the inflexion
=
0.
We write
equation (7)
in the form
^5^^Y-f-72(cosa-cos(9)
=
(9)
To
integrate
it we introduce Jacobian
elliptic
functions of an
argument
u with a modulus
k which are
given by
the
equations
u=sJ{RIB),
k
=
8iuia
(10)
Then we have
^^
=
2kcn{u
+
K), sin|5
=
/tsn(M
+
A'), (11)
where A' is the real
quarter period
of the
elliptic
functions. To determine the
shape
of the
L. E.
25
386
FORMS OF THE ELASTICA
[CH.
XIX
curve,
let
x, y
be the coordinates of a
point
referred to fixed
axes,
of which the axis of x
coincides with the line of thrust. Then we have the
equations
dnjds

cos
6, dy/ds
=
sin
6,
and these
equations give
x=y/('|V-+2{^am(M+A')-^amA'}],
1
i
(12)
y=-2/:^Qcn(
+
Z),
J
where E&iau denotes the
elliptic integral
of the second kind
expressed by
the formula
/
Ea,mu=
I dn^udu,
Jo
and the constants of
integration
have been determined so that x and
y may
vanish with s.
The inflexions are
given by
cos 5=cos
a,
or
sn^{u+X)

l,
and therefore the arc between
two consecutive inflexions is
2^{S/R). K,
and the inflexions are
spaced equally aloug
the
axis of X at intervals
2^(BIR){'ZEa,mK-K).
The
points
at which the
tangents
are
parallel
to the line of thrust are
given by
sin 6
=
0,
or sn
{u
+
IC)dn(u
+
K)
=
0,so
that u is an imeven
multiple
of K. It follows that the curve
forms a series of
bays, separated by points
of inflexion and divided into
equal half-bays by
the
points
at which the
tangents
are
parallel
to the line of thrust.
The
change
of the form of the curve as the
angle
a increases is shown
by Figs.
48

55.
When a >
i^ ff,
X is
negative
for small values of
u,
and has its
numerically greatest negative
value when u has the smallest
positive
value which satisfies the
equation An^{u+K)=^.
Let
Ml
denote this value. The value of u for which x vanishes is
given by
the
cquiition
M
=
2
{^am {u
+
h")

A'am
K}.
When u exceeds this
value,
x is
positive,
and x has a maximum
value when u
=
2K

u^
.
Figs.
50

52 illustrate cases in which

x^
is
respectively greater than,
equal to,
and less than
| xj ]. Fig.
53 shows the case in which
xjj-=0
or 2'amA'
=
/r.
This
happens
when a
=
130
approximately.
In this case all the double
points
and inflexions
coincide at the
origin,
and the curve
may
consist of several
exactly equal
and similar
pieces lying
one over another.
Fig.
54 shows a case in which
2Ea,uiK<K,
or
x^<0;
the
curve
proceeds
in the
negative
direction of the axis of x. The
limiting
case of
this,
when
a
=
7r,
is shown in
Fig. 55,
in which the rod
(of
infinite
length)
forms a
single loop,
and the
pendulum
of the kinetic
analogue
starts close to the
position
of unstable
equilibrium
and
just
makes one
complete
revolution.
(6) Non-inflexional
elastica.
When there are no inflexions we write
equation (7)
in the form
*^(sy=^"=-'^+^(^+^v^)'
^^^'
where k is less than
unity,
and we introduce Jacobian
elliptic
functions of modulus k and
argument m,
where
u=k-HJ{RIB) (14)
We measure s from a
point
at which 6 vanishes. Then we have
:
=
^^f;gjdnM,
sin^e
=
snM, (15)
M 2
ds
263] FORMS OF THE
ELASTICA
387
Fig.
48.
Fig.
49.
Fig.
52.
Fig.
50.
Fig.
51.
Fig.
54.
Fig.
55.
252
388 BUCKLING OF A STRUT
and the coordinates x and
y
are
expressed
in terms of m
by
the
equations
[CH.
XIX
.(16)
^ >
Fig.
56.
in which the constants of
integration
are chosen so that x vanishes with
s,
and the axis of
X is
parallel
to the line of action of
R,
and at such a distance from it that the force R and
the
couple
-
B
{d6/dg)
which must be
applied
at the ends of the rod are
statically equivalent
to a force R
acting along
the axis of x. The ciirve consists of a series of
loops lying
altogether
on one side of this axis. The form of the curve is shown in
Fig.
56.
264.
Buckling
of
long
thin strut under thrust*.
The
limiting
form of the elastica when a is
very
small is obtained
by
writing
6 for sin 6 in
equation (8).
We have
then,
as first
approximations,
6*
=
a cos
{s^(RIB)],
x
=
s,
y
=
as/{B/R)
sin
{x^/iR/B)},
. .
.(17)
so that the curve is
approximately
a curve of sines of small
amplitude.
The
distance between two consecutive inflexions is
rr'J{BjR).
It
appears
therefore
that a
long straight
rod can be bent
by
forces
applied
at its ends in a direction
parallel
to that of the rod when
unstressed,
provided
that the
length
I and the
force R are connected
by
the
inequality
l'R>-7r"-B
(18)
If the direction of the rod at one end is constrained to be the same as that
of the
force,
the
length
is half that between consecutive
inflexions,
and the
inequality (18)
becomes
l'R>\7r'B (19)
The
theory
was initiated
by
Euler. See
Introduction, p.
3.
263-265]
UNDER THRUST
389
If the ends of the rod are constrained to remain in the same
straight
line,
the
length
is twice that between consecutive
inflexions,
and the
inequality (18)
becomes
l'R>i'7r'B
(20)
These three cases are illustrated in
Fig.
57.
Any
of these results can be obtained
very easily
without
having
recoui-se to the
general
theory
of the elastica. We take the second
case,
and
suppose
that a
long
thin rod is set
up
vertically
and loaded at the
top
with a
weight R,
while the lower end is constrained to
remain vertical*. Let the axes of x and
y
be the vertical line drawn
upwards through
the
lowest
point
and a horizontal line drawn
through
the same
point
in the
plane
of
bending,
as shown in
Fig.
57 b. If the rod is
very slightly bent,
the
equation
of
equilibrium
of the
portion
between
any
section and the loaded end
is,
with sufficient
approximation,
-5g+^(y.-y)=o,
where
y,
is the
displacement
of the loaded end. The solution of this
equation
which
satisfies the conditions that
y
vanishes with
x,
and that
y=y,
when x
=
l,
is
L
sin
""
sin
{I ^(R/B)} J'
and this solution makes
dy/dx
vanish with x if cos
{!.J{RjB)}
=
0. Hence the least value
of I
by
which the conditions can be satisfied is
^irJiBjR).
From the above we conclude
that,
in the case
represented by Fig.
57
h,
if
the
length
is
slightly greater
than
^ir'^{BjR),
or the load is
slightly greater
than
Itt^B/1^,
the rod bends under the
load,
so that the central-line assumes
the form of one
half-bay
of a curve of sines of small
amplitude.
If the
length
of the rod is less than the critical
length
it
simply
contracts under the load.
If the
length
is
greater
than the critical
length,
and the load is
truly
central
while the rod is
truly cylindrical,
the rod
may simply
contract
;
but the
equi-
librium of the rod thus contracted is unstable. To
verify
this it is
merely
necessary
to show that the
potential energy
of the
system
in the bent state is
less than that in the contracted state.
265.
Computation
of the
strain-energy
of the strut.
Let the
length
I be
slightly greater
than
\n ^/{BjR).
Let a> denote the area of the
cross-section of the
rod,
and E the
Young's
modulus of the material. If the rod
simply
contracts,
the amount of the contraction is
RjEco,
the loaded end descends
through
a
distance
Rl/Ea>,
and the loss of
potential energy
on this account is
R^l/Ea.
The
potential
energy
of contraction is
iR^l/Ea>.
The
potential energy
lost in the
passage
from the
unstressed state to the contracted state
is, therefore, iR^l/E<o.
If the rod bends into the form of one
half-bay
of an elastica of small
angle a,
the
potential energy
lost
through
the descent of the load is
R{1
j
cos 6
ds),
where the
integra-
tion extends over the
length
of the rod. The
potential energy
of
bending
ia
^BJ
{dd/dsyds,
*
We
neglect
the
weight
of the rod. The
problem
oi the
bending
of a vertical rod under its
own
weight
will be considered in Article 276.
890 BUCKLING OF A STRUT
[CH.
XIX
-or R
l(cos^-coso)<fe.
The tension T aX
any
section is
Eat,
where is the extension of
the
central-line,
and it is also

R cos
6,
and therefore the
potential energy
of contraction
is
\Eai
\
{p,
COB 6
jEafds
or
^{IP/E<o){l-
I sin^ 6
ds}.
Hence the loss of
potential energy
in the
passage
from the unstressed state to the bent state is
R{l{l+coaa)-2jcoaeds}-^{RyEw){l-J!im^6ds}.
The excess of the
potential energy
in the contracted state above that in the bent state
is therefore
R{l{l
+
cosa)-2
I
cos e
ds}
+
^ (R^/Ea)
j
Bin" 6
ds-lR^jEa (21)
Now we have
/
cos
dds=^{BIR) {2Ea.m K-K)=l{2K-^
Earn K~
1),
since
l=^KJ{BIR).
Also we have
A'=*,r(l+p2+^^+...), Ea.mK=in{l-ik^-ij(:*-...),
and therefore
2K--^Ea.mK-l
=
l-F--lk*-....
Hence the above
expression (21)
is
Rl
{ik*
-
RIEa>} +i {R^/Ea>)
jfiin^eds
(22)
If we denote the
length ^n ^(B/R) by l^,
we have
?o(l+p-H. ..)
=
?,
and therefore h'^
=
A{ljlf) 1) nearly,
so that
ik*-R/E<o=4{l/lo-l)^-RIE<o,
and this is
positive
if
l>lo{l+is/{IilEc.)},
or
l>l^+in^{B/Eu>).
The
quantity
B is the
product
of Ea and the
square
of the radius of
gyration
of the
cross-section about an axis drawn
through
its centroid at
right angles
to the
plane
of
bending. Denoting
this radius of
gyration by c,
we find that the
potential energy
in the
contracted state is
certainly greater
than that in the bent state if
l>lo
+
i^c (23)
The term
Jjrc
constitutes a correction of the formula for the critical
length
as
ordinarily
calculated
;
it
is,
of
course,
immaterial in
any
case to which the
theory
of thin rods could
be
applicable.
Another correction of the same order of
importance
would result from
taking
account of the
special
state of the
parts
of the rod that are near the ends. If the
forces
applied
to hold the lower end are so distributed that the
theory
of thin rods
gives
an
adequate
account of the strain near this
end,
then the terminal section is not fixed as a
whole,
and some work is done
by
the tractions at this end
[cf.
Article 235
(/)].
If,
on the
other
hand,
it is
kept fixed,
then there are "local
perturbations"
near the
end,
and the
additional
energy
that
depends upon
them has not been taken into account. There will
be similar local
perturbations
near the loaded end.
266. Resistance to
buckling.
The strains
developed
in the
rod,
whether it is short and
simply
contracts
or is
long
and
bends,
are
supposed
to be elastic
strains,
that is to
say
such as
265,
266]
UNDER THRUST
891
disappear
on the removal of the load. For Euler's
theory
of the
buckling
of
a
long
thin
strut,
explained
in Article
264,
to have
any practical bearing,
it is
of course
necessary
that the load
required,
in accordance with
inequalities
such as
(19),
to
produce bending
should be less than that which would
produce
set
by crushing.
This condition is not satisfied unless the
length
of the strut is
great compared
with the linear dimensions of the cross-section.
In view of the lack of
precise
information as to the conditions of
safety
in
general (Chapter iv.)
and of failure
by crushing (Article 189),
a
precise
estimate of the smallest ratio of
length
to diameter for which this condition
would be satisfied is not to be
expected.
The
practical question
of the conditions of failure
by buckling
of a rod or
strut under thrust involves some other considerations. When the thrust is
not
truly
central,
or its direction not
precisely
that of the
rod,
the
longitudinal
thrust is
accompanied by
a
bending couple
or a transverse load. The con-
traction
produced by
the thrust R is
RjEco.
When the thrust is not
truly
central,
the
bending
moment is of the order
Re,
where c is some linear
dimension of the
cross-section,
and the extension of a
longitudinal
filament
due to the
bending
moment is of the order
Rc'/B,
which
may easily
be two or
three times as
great, numerically,
as the contraction
R/Ea>.
The
bending
moment
may,
therefore,
produce
failure
by buckling
under a load less than
the
crushing
load.
Again,
when the line of thrust makes a small
angle
with the
central-line,
the transverse load R sin
yS
yields,
at a distance
compar-
able with the
length
I of the
rod,
a
bending
moment
comparable
with
IRsia^;
and the extension of a
longitudinal
filament due to this
bending
moment is
comparable
with IRc sin
^/B.
Thus even a
slight
deviation of the direction
of the load from the central-line
may produce
failure
by buckling
in a
fairly
long
strut. Such causes of failure as are here considered can best be discussed
by
means of Saint-Venant's
theory
of
bending (Chapter XV.) ; but,
for a
reason
already
mentioned,
a
precise
account of the conditions of failure
owing
to such causes is
hardly
to be
expected.
It is clear that such considerations as are here advanced will be
applicable
to other cases of
buckling
besides that of the
buckling
of a rod under thrust.
The
necessity
for them was
emphasized by
E. Lamarle*. His work has
been discussed
critically
and
appreciatively by
K.
Pearsoni".
In recent
years
the conditions of
buckling
have been the
subject
of considerable
discussion
:|:.
* '
Mem. sur la flexion du
boia,'
Ann. des travaux
publics
de
Belgique,
t. 4
(1846).
t
Todhunter and Pearson's
History,
vol.
1, pp.
678 et
seq.
*
Reference
may
be made to the
writings
of J.
Kiibler,
C. J.
Kriemler,
L. Prandtl in Zeitschr.
d. Deut&chen
Ingenieure,
Bd. 44
(1900),
of Kiibler and Kriemler in Zeitschr.
/.
Math. u.
Phys.
Bde. 45-47
(1900-1902),
and the dissertation
by Kriemler,
'
Labile u. stabile
Gleichgewiohts-
figuren...auf Biegung beanspruchter
Stiibe...
'
(Karlsruhe, 1902).
392
ELASTIC STABILITY
[CH.
XIX
267. Elastic
stability.
The
possibility
of a
straight
form and a bent form with the same terminal
load is not in conflict with the theorem of Article
118,
because the thin rod
can,
without
undergoing
strains
greater
than are
contemplated
in the mathe-
matical
theory
of
Elasticity,
be deformed in such a
way
that the relative
displacements
of its
parts
are not small*.
The
theory
of the
stability
of elastic
systems, exemplified
in the discussion
in Art.icles
264, 265,
may
be
brought
into connexion with Poincare's
theory
of
"equilibrium
of bifurcation
f."
The form of the rod is determined
by
the
extension e at the loaded end and the total curvature a
;
and these
quantities
depend upon
the load
R,
the
length
I and flexural
rigidity
B
being regarded
as constants. We
might represent
the state of the rod
by
a
point,
determined
by
the coordinates e and
a, and,
as R
varies,
the
point
would describe a curve.
When R is smaller than the critical
load,
a
vanishes,
and the
equilibrium
state,
defined
by
e as a function of
R,
is stable. When R exceeds the critical
value,
a
possible
state of
equilibrium
would still be
given by
a
=
;
but there
is another
possible
state of
equilibrium
in which a. does not
vanish,
and in
this state a and e are determinate functions of
R,
so that the
equilibrium
states for
varying
values of R are
represented by points
of a certain curve.
This curve issues from that
point
of the line a
=
which
represents
the
extension,
or rather
contraction,
under the critical load. Poincar(5 describes
such a
point
as a
"
point
of
bifurcation,"
and he shows
that,
in
general,
there
is an
"
exchange
of stabilities
"
at such a
point,
that is to
say,
in the
present
example,
the states
represented by points
on the line a
=
0,
at which e numeri-
cally
exceeds the extension under the critical
load,
are
unstable,
and the sta-
bility
is transferred to states
represented by points
on the curve in which
a=^0.
268.
Stability
of inflexional elastica.
When the lower end of the loaded rod is constrained to remain
vertical,
and the
length
I
slightly
exceeds
^inJiB/Ii),
a
possible
form of the central-line is a curve of sines of small
amplitude having
two
inflexions,
as in
Fig.
58
(6).
Another
possible
form is an elastica
illustrated in
Fig.
58
(c).
In
general,
if n is an
integer
such that
i{2n
+
l)n>ls/{R/B)>Js{2n-l)^, (24)
n forms besides the unstable
straight
form are
possible,
and
they
consist
respectively
of
1, 3,
... 2
-
1
half-bays
of different curves of the elastica
family.
The forms of these curves
are
given respectively by
the
equations
K=lJiR/B)xihh-,V(2n-l)] (25)
We shall show that all these forms
except
that with the
greatest K,
that is the smallest
number of
inflexions,
are
iinstablej;.
*
Cf. G. H.
Bryan, Cambridge
Phil. Soc.
Proc,
vol. 6
(1888).
t
Acta
Mathematica,
t. 7
(1885).
J
The result is
opposed
to that of L.
Saalschiitz,
Der belastete Stab
(Leipzig 1880),
but I do
not think that his
argument
is
quite convincing.
The result stated in the text
agrees
with that
obtained
by
a different method
by
J.
Larmor,
loc. cit.
p.
383.
267,
268]
STABILITY OF INFLEXIONAL ELASTICA 393
Omitting
the
practically unimportant potential energy
due to extension or contraction
of the
central-line,
we
may
estimate the loss of
potential energy
in
passing
from the
unstressed state to the bent state in which there are
r+1
inflexions,
in the same
way
as in
Article
265,
as
/J[^(l
+cos
and this is
a)-.j[
COS 6
ds], (26)
{2r+l)y/{BR)(iKr-4Er-^Krk,^), (27)
Fig.
58.
where
E^
is written for
EaiaKr,
and the suffix r indicates the number
{r
+
l)
of inflexions.
We
compare
the
potential energies
of the forms with
r+1
and
s+l
inflexions,
s
being
greater
than r. Since
{2r+l)K,={2s+l)K (28)
the
potential energy
in the form with s
+
l inflexions is the
greater
if
(2
+
l) (2E,
+
A\i.;^)>{2r
+
-i) (2^,+.ff,V).
Since
this condition is
But,
since
Ea.mK={l-k^)(K+k^\,
<>-('-|f)"'-'''('+l'S)
''
it follows that
{l-/c^)
{^
+
-f^ jl)
diminishes as k increases. Now when
s>r, K,<Kr,
and
k,<kr;
and therefore the
inequality (29)
is satisfied.
In the case illustrated in
Fig.
58 the three
possible
forms are
(a)
the unstable
straight
form, (6)
the
slightly
bent form with two
inflexions, (c)
the bent form with one inflexion.
The
angle
a for the form
(c)
is
given by A'=37r,
and it lies between 175 and 176.
394 ROD OF
EQUAL
FLEXIBILITY
[CH.
XIX
It
may
be obsei-ved that the conclusion that the stable form ia that with a
single
inflexion is not in conflict with Poincard's
theory
of the
exchange
of stabilities at a
point
of
bifurcation,
beoavise the loci in the domain of e and a which
represent
forms with two or
more inflexions do not issue from the locus which
represents
forms with one inflexion but
from the locus a
=
which
represents straight
forms.
The
instability
of forms of the elastioa with more than the smallest
possible
number of
inflexions between the ends is well known as an
experimental
fact.
Any particular
case
can be
investigated
in the same
way
as the
special
case discussed
above,
in which the
tangent
at one end
is, owing
to
constraint, parallel
to the line of thrust. An
investigation
of this kind
cannot, however,
decide the
question
whether
any particular
form is stable or
unstable for
displacements
in which the central-line is moved out of its
plane.
This
question
has not been solved
completely.
One
special
case of it will be considered in
Article
272(e).
269. Rod bent and twisted
by
terminal forces and
couples.
We resume now the
general problem
of Article
260,
and
express
the
directions of the
principal
torsion-flexure axes at
any point
Pi
on the strained
central-line
by
means of the
angles
0,
\fr, (f>
defined in Article 253. We
choose as the fixed direction
PjZ
in
Fig.
46 of that Article the direction of
the force
applied
to the rod at the end towards which s is measured. The
stress-resultants
N, N',
T are
equivalent
to a force R in this
direction,
and
therefore
{N,
N',
T)
=
R(-
sine cos
0,
sin sin
^,
cos
6') (30)
Equation (3)
of Article 260 becomes
i(AK^
+ BK'"-
+
CT-)
+ Rcos0
=
const
(31)
Since the forces
applied
at the ends of the rod have no moment about the
line
PjZ,
the sum of the
components
of the
stress-couples
about a line drawn
through
the centroid of
any
section
parallel
to this line is
equal
to the corre-
sponding
sum for that terminal section towards which s is measured. We
have therefore the
equation

Ak sin cos
<f>
+
Bk sin sin
<^
-f
Gt cos
=
const
(32)
The
analogue
of this
equation
in the
problem
of the
top expresses
the
constancy
of the moment of momentum of the
top
about a vertical axis
drawn
through
the fixed
point.
The
equations (31)
and
(32)
are two
integrals
of the
equations (2)
of
Article
260, and,
if a third
integral
could be
obtained,
d0/ds, dyjr/ds, dcfi/ds
would be
expressible
in terms of
0,
yjr, j>,
and the
possible
forms in which the
rod could be held
might
be found. In the
general
case no third
integral
is
known
; but,
when the two flexural
rigidities
A and B are
equal,
the third of
these
equations yields
at once the
integral
T
=
const
(33)
The
quantities
k, k,
t are
expressed
in terms of
0,
i/f, 0, d0/ds,
...
by
equations (8)
of Article
253,
and the
equations (31), (32), (33)
can be
268-270]
IN ALL PLANES DRAWN THROUGH ITS CENTRAL-LINE
395
integrated*
so as to
express
6,
yjr, (f>
as functions of
s,
and then the form
of the central-line is to be determined
by
means of the
equations
dx . .
, dy
.

.
,
dz .
T-
=
sm a cos
yjr,
-^
=
sin ff sm
y,
~
=
cos
6,
where
x,
y,
z are coordinates referred to fixed axes.
We shall not
proceed
with this
general theory,
but shall consider some
important special
cases.
270. Rod bent to helical
formf.
The
steady
motion of a
symmetrical top,
with its axis of
figure
inclined at
a constant
angle ^tt

a to the vertical drawn
upwards,
is the
analogue
of a
certain
configuration
of a bent and twisted rod for which A.
=
B.
Putting
e
=
^iT-a, ddjds
=
0,
we
have,
by (8)
of Article
253,
d'Jr
, , rfi/r
.
,
dd) . dylr
K
=
-J-
COS a cos
(p,
K.
=
-J-
cos a sin
<p,
r
=
-p
+
sin a
-~-
,
and,
by (31), (32), (33)
of Article
269,
T
=
const.,
-
4-
k'-
=
const.,
dyjr/ds
=
const.
The curvature of the central-line is constant and
equal
to cos a
(d-^/ds),
and
the binormal of this curve lies in the
plane
of
{x, y)
and makes an
angle <j)
with the axis of x reversed. It follows that
<^
is identical with the
angle
denoted
by f
in Article
253,
and that the measure of
tortuosity
of the curve
is sin a
(dyfr/ds).
Since the central-line is a curve of constant curvature and
tortuosity,
it is a helix traced on a
rigid
circular
cylinder.
The axis of the
helix is
parallel
to the line of action of
R,
and a is the
angle
which the
tangent
at
any point
of the helix makes with a
plane
at
right angles
to
this axis.
Let r be the radius of the
cylinder
on which the helix lies. Then the
curvature
'l//>
and the measure of
tortuosity 1/2
are
given by
the
equations
l/p
=
cos'a/r, 1/2
=
sin a cos
a/r, (34)
and we
may
write
K

cos(j)Cos-a./r,
'
=
sin^cos^a/r, d-\}r/ds
=
cos
a/r, d<f>jds
=
T

smacosa./r.
(35)
From
equations (2)
of Article 260 we find
(iV,
iV
')
=
(-
cos
(^,
sin
^) [Ot
cos^
a/r
-
sin a cos^
/?=],
and then from
equations (30)
we find
R= Ct cos
a/r
JSsin
acos^'a/j'^ (36)
*
See F. Klein u. A.
Sommerfeld,
Theorie des
Kreiseh,
Heft
2, Leipzig
1898,
or E. T.
Whittaker, Analytical Dynamics, Cambridge
1904.
t
Cf.
Kirchhoff, loc. cit.
p.
365.
[
396 ROD BENT TO HELICAL FORM
[CH.
XIX
The terminal force is of the nature of tension or
pressure according
as
the
right-hand
member of
(36)
is
positive
or
negative.
(See
.R
Fig. 59.)
For tlie force to be of the nature of
tension,
r
-^^
must exceed Bsin a cos
a/
Cr.
^

\-~~y^
The axis of the terminal
couple
lies in the
tangent plane
of
") the
cylinder
at the end of the
central-line,
and the
components
^ of this
couple
about the binormal and
tangent
of the helix at
"\ this
point
are B cos^
a/r
and Or. The
components
of the same
^ couple
about the
tangent
of the circular section and the
'>\ generator
of the
cylinder
at the same
point
are, therefore,
>i Rr and
K,
where K is
given by
the
equation
^^^^^^
K=Gt sin a -\- Bcos^
a/r (37)
Ip
It follows that the rod can be held so that it has a
given
FiK 59
twist,
and its central-line forms a
given
helix,
by
a wrench of
which the force R and the
couple
K are
given by equations (36)
and
(37),
and the axis of the wrench is the axis of the helix. The force and
couple
of the wrench are
applied
to
rigid pieces
to which the ends of the rod
are attached.
The helical form can be maintained
by
terminal force
alone,
without
any couple
;
and
then the force is of
magnitude
Bcos^
a/r^
sin
a,
and acts as thrust
along
the axis of the helix.
In this case there must be twist of amount
-
B cos'
a/Cr
sin a. The form can be maintained
also
by
tei-minal
couple alone,
without
any
force
;
and then the
couple
is of
magnitude
Bcos
a/r,
and its axis is
parallel
to the axis of the helix. In this case there must be twist
of amount Bsin a cos
a/Cr.
When the state of the rod is such
that,
if
simply unbent,
it would be
prismatic, d(f>lds
vanishes,
and the twist of the rod is
equal
to the measure of
tortuosity
of the central-line
(cf.
Article
253).
To hold the rod so that it has this
twist,
and the central-line is a
given
helix,
a wrench about the axis of the helix is
required ;
and the force R and
couple
K of the
wrench are
given by
the
equations
E=
-{B-C)amacos''a/r^,
K
=
{B
cos^ a
+
CHin^
a)
coa
air.
271.
Theory
of
spiral springs*.
When the sections of the rod have kinetic
symmetry,
so that
A=B,
and
the unstressed rod is helical with such initial twist
that,
if
simply
unbent,
it
would be
prismatic,
we
may express
the initial state
by
the formulae
o
=
0, '
=
cos^
a/r,
To
=
sin a cos
a/r
(38)
By
suitable terminal forces and
couples
the rod can be held in the state
expressed by
the formulae
i=0, /Ci'
=
cos^ai/rj,
Ti
=
sinaicosaj/r,,
(39)
*
Cf. Kelvin and
Tait,
Nat. Phil. Part n.
pp.
139 et
seq.
270,
271]
SPIRAL SPRINGS
397
where
i\, a^
are the radius and
angle
of a new helix. The
stress-couples
at
any
section are then
given by
the
equations
_
,
_
p
/cos^a,
cos'aN
_
^
/sin
!
cos
a,
sin a cos a\
~
'
~
V n
r
)'
~
\ r,
~
r
)'
and the stress-resultants are
given by
the
equations
N
=
0,
T
=
N't&na
^y,
_
^
cos"
ai
/sin
Kj
cos
a^
sin a cos a\ sin
a^
cos
a,
/cos^
a^
cos' a
All the
equations
of Article 259 are satisfied. The new
configuration
can
be maintained
by
a wrench of which the axis is the axis of the
helix,
and the
force R and
couple
K are
given by
the
equations
n

O
^^
"'
(
^^^
**'
^^^
"'
^^^ ^ ^
"^ n^^^
'*' /cos'
Hi
cos^ a\
~
n
\ r-i
r
J
n V n
r J
'
-.^
.
/sin
a,
cos
!
sin a cos a\

/cos"
a,
cos"a\
ir=(;sinai
^
+5cosai(
.
(40)
The
theory
of
spiral springs
is founded on this result. We take the
spring
in the unstressed state to be determined
by
the
equations (38),
so that the
central-line is a helix of
angle
a traced on a
cylinder
of radius
r,
and the
principal
normals and binormals in the various cross-sections are
homologous
lines of these sections. We take I to be the
length
of the
spring,
and h to be
the
length
of its
projection
on the axis of the
helix,
then the
cylindrical
coordinates
r, 6,
z of one end
being
r, 0, 0,
those of the other- end are
r,
'x^,
h,
where
T^
=
(i
cos
a)/r,
/i
=
Zsina
(41)
We
suppose
the
spring
to be deformed
by
a wrench about the axis of the
helix,
and take the force R and
couple
K of the wrench to be
given.
We
shall
suppose
that the central-line of the strained
spring
becomes a helix of
angle
ai
on a
cylinder
of radius
r^,
and that the
principal
normals and
binormals continue to be
homologous
lines in the cross-.sectious. Then R
and K are
expressed
in terms of
a^
and
r^
by
the
equations (40).
AVhen the
deformation is small we
may
write
r-\-hr
and a-l-Sa for
?,, ,,
and
suppose
that small
changes 3;^
and hh are made in
%
and h. We have
84
=
{I
cos
a)
6a,
8%
=

[{I
sin
a)/r]
Sa

[(I
cos
a)/r"] hr,
from which
ha.

{hh)/(l
cos
a), (Sr)/?'"
=
(sin
a. Sh
+
r cos a.
Sx)/lr
cos" a.
^r c
sin a cos a . Br cos 2a
t,
Hence o
=

sm a cos a

H da
r r r
8h .
Sy
=
cos a
^
+ sm a
-^
,
398 SPECIAL PROBLEMS CONCERNING
[CH.
XIX
,
5,008"
a . Bi- - . Ba
and S
=

cos'' a

2 sin a cos a

cos a. sin
a..
It follows that the force R and the
couple
K are
expressed
in terms of
I, r, a, Bh,
B^ by
the
equations
R
=
^[(C
cos= a
+
B sin=
a)Bh
+
(C- B)
sin a cos a .
rSxl
1
,
-(42)
K
=
j-[(G-B)
sin a cos a . Bh +
(Csin-
a
+
B cos-
a) rBx].
If the
spring
is deformed
by
axial force
alone*,
without
couple,
the axial
displacement
Sh and the
angular displacement 8x
are
given by
the
equations
Sk
=
lr^(^^
+
^^'^E,
8x
=
frsinac0sa(^-i)/?.
If the cross-section of the
spring
is a circle of radius
a, 1/C- IjB
is
icrjEna*,
where o-
is Poisson's ratio and E is
Young's
modulus for the material. Hence both bh and
hx
are
positive.
In the same case br is
negative,
.so that the
spring
is coiled more
closely
as it
stretches.
272. Additional results.
(a)
Rod
subjected
to terminal
cottples.
When a rod which is
straight
and
prismatic
in the unstressed state is held bent and
twisted
by
terminal
couples,
the kinetic
analogue
is a
rigid body moving
under no forces.
The
analogue
has been worked out in detail
by
W. Hesst. When the cross-section has
kinetic
symmetry
so that A
=
B,
the
equations
of
equilibrium
show that the twist t and
the curvature
(it^
+
k'-)
are
constants,
and
that,
if we
put
as in Article 253
tan/=-ic7K,
then
B{df/ds)={B-C)T.
It follows that the measure of
tortuosity
of the central-line is
Ct/B, and, therefore,
that
this line is a helix traced on a circular
cylinder.
If we use Euler's
angles 6, \|/-, (p
as in
Article
253,
and take the axis of the helix to be
parallel
to the axis of z in
Fig.
46 of that
Article,
6 is
constant,
and
^n

d is the
angle
a of the helix. The axis of the terminal
couple
is the axis of the
helix,
and the
magnitude
of the
couple
is B cos
a/r,
as we found
before,
r
being
the radius of the
cylinder
on which the helix lies.
*
The results for this case were found
by Saint-Venant,
Paris C.
R.,
t. 17
(1843).
A
number of
special
cases are worked out
by
Kelvin and
Tait,
loc,
cit.,
and also
by
J.
Perry,
Applied
Mechanics
(London, 1899).
The
theory
has been verified
experimentally by
J. W.
Miller,
Fhys.
Rev. vol. 14
(1902).
The vibrations of a
spiral spring supporting
a
weight
so
great
that
the inertia of the
spring may
be
neglected
have been worked out in accordance with the above
theory by
L. K.
Wilberforce,
Phil.
Mag. (Ser. 5),
vol. 38
(1894).
t
Math. Ann. Bd. 23
(1884).
271,
272]
THE
EQUILIBBIUM
OF THIN RODS 399
(b) Straight
rod mth initial tioist.
When the rod in the unstressed state has twist
tq
and no
curvature,
and the cross-
section has kinetic
symmetry
so that A
=
B,
the rod can be held bent so that its central-hne
has the form of a helix
(a, r),
and twisted so that the twist is
r,, by
a wrench about the axis
of the helix
;
and the force R and
couple
K of the wrench are found
by writing t,
-
t^
for t
in
equations (36)
and
(37)
of Article 270.
(c)
Rod bent into cimdar
hoop
and tioisted
uniformly.
When the rod in the unstressed state is
straight
and
prismatic,
and the cross-section
has kinetic
symmetry,
one of the forms in which it can be held
by
terminal forces and
couples
is that in which the central-line is a
circle,
and the twist is uniform
along
the
length.
The tension
vanishes,
and the
shearing
force at
any
section is directed towards
the centre of the
circle,
and its amount is
Crjr,
where r is the radius of the circle.
id) Stability of
rod
subjected
to
twisting couple
and thrust.
When the
rod, supposed
to be
straight
and
prismatic
in the unstressed
state,
is held
twisted,
but without
curvature, by
terminal
couples,
these
couples may
be of such an
amount as could hold the rod bent and twisted, li A=B the
central-line,
if it is
bent,
must be a helix. When the
couple
K is
just great enough
to hold the rod bent without dis-
placement
of the
ends,
the central-line
just
forms one
complete
turn of the
helix,
the radius
r of the helix is
very small,
and the
angle
a of the helix is
very nearly equal
to
Jtt.
We
have the
equations
A'=CV
=
/"' cos
a,
^cosa
=
27rr,
where t is the
twist,
and I the
length
of the rod. Hence this
configuration
can be
maintained if
2Tr/l=K/B.
We infer
that,
under a
twisting
couple
which exceeds
'UttBjI,
the
straight
twisted rod is
unstable.
This
question
of
stability may
be
investigated
in a more
general
manner
by supposing
that the rod is held
by
tenninal
thrust R and
twisting couple
K in a form in which the
central-line is
very nearly straight.
The kinetic
analogue
is
a
symmetrical top
which moves so that its axis remains
nearly upright.
The
problem
admits of a
simple
solution
by
the use of fixed axes of
x, y, z,
the axis of z
coinciding
with
the axes of the
applied couples
arid with the line of thiiist.
The central-line is near to this
axis,
and meets it at the
ends. The twist r is
constant,
and the torsional
couple
Cl-
ean be
equated
to K with sufficient
approximation.
The
flexural
couple
is of amount
BIp,
where
p
is the radius of
curvature of the
central-line,
and its axis is the binormal
of this curve. The direction cosines of this binormal can be
expressed
in such forms as
/dy
d-7.
\ds
ds'
dz
c?^y\
ds
d^J
'
and therefore the
components
of the flexural
couple
at
any
section about axes
parallel
to the axes of x and
y
can be
expressed
with sufficient
approximation
in the forms
Fig.
60.
-B
d^
ds"
dH
400 SPECIAL PROBLEMS CONCERNING
[CH.
XIX
For the
equilibrium
of the
part
of the rod contained between this section and one end
we take moments about axes drawn
through
the centroid of the section
parallel
to the axes
of X and
y,
and we thus obtain the
equations
(43)
ds^ as
The
complete primitives
are
x
=
Xi
sin
(5'i
+
f,) +Z2sin (ya^+^z).
y=ZiCos (ji+fi)
+
Z2
cos
(g'2
+
,),
where
Li, L^, i, (2
are
arbitrary
constants,
and
q^, q^
are the roots of the
equation
Bq^
+
Kq-R
=
0.
The terminal conditions are
(i)
that the coordinates x and
y
vanish at the ends s
=
and
s
=
l, (ii)
that the axis of the terminal
couple
coincides with the axis of z. The
equations
(43)
show that the second set of conditions are satisfied if the first set are satisfied. We
have therefore the
equations
i,
sin
1
+
L^
sin
t^
=
0, X,
cos f
i
+
Xj
cos f
2
=
0,
and
L^sm{q-J,
+
(^
+
L^8va.{q2l
+
e^Q, L^coa{q.J,
+
e-^
+
L^cos{q2l^(^
=
Q.
On
substituting
for
L^
cos
c^
and
L^
sin
fj
from the first
pair
in the second
pair,
we find the
equations
2;,{8in(5i?
+
fi)-sin(g'2?
+
fi)}
=
0, L^{co&{qil
+
fi)-cos{q.^l
+
fy)]
=
0,
from which it follows that
q^l
and
q^l
difffer
by
a
multiple
of Stt. The least
length
I
by
which the conditions can be satisfied is
given by
the
equation
2t/^=I?i-?2|.
or
n^_:^4.^
The rod
subjected
to thrust R and
twisting couple
K is therefore unstable if
P^4B^^ B
^ '
This condition* includes that obtained above for the case whore there is no
thrust,
and
also that obtained in
(18)
of Article 264 for the case where there is no
couple.
If the rod
is
subjected
to tension instead of
thrust,
R is
negative,
and thus a suflScient tension will
render the
straight
form stable in
spite
of a
large twisting coupla
(e) Stahility of flat
blade bent in its
planef.
Let the section of the rod be such that the flexural
rigidity
B,
for
bending
in one
principal plane,
is
large compared
with either the flexural
rigidity A,
for
bending
in the
perpendicular plane,
or with the torsional
rigidity
C. This would be the case
if,
for
example,
the cross-section were a
rectangle
of which one
pair
of sides is much
longer
than
the other
pair.
Let the
rod,
built in at one end so as to be
horizontal,
be bent
by
a vertical
transverse load R
applied
at the other end in the
plane
of
greatest
flexural
rigidity.
We
*
The result is due to A. G.
Greenhill,
Proc, Inst. Mech.
Engineers,
1883.
t
Of. A. G. M.
Michell.PMi. Mag. (Ser. 5),
vol. 48
(1899),
andL.
Prandtl,
'
Kipperscheinungen
'
{Diss.),
Nurnberg
1899.
272]
THE
EQUILIBRIUM
OF THIN RODS
^Ol
shall use the notation of Article
253,
and
suppose,
as in Article
270,
that the line of action
of the load R has the direction and sense of the line
P,z,
and we shall take the
plane
of
(z, x)
to be
parallel
to the vertical
plane containing
the central-line in the unstressed state.
If the
length I,
or the load
R,
is not too
great,
while the flexural
rigidity
B is
large,
the rod
will be
slightly
bent in this
piano,
in the manner discussed in
Chapter
XV.
But,
when the
length,
or
load,
exceed certain
limits,
the rod can be held
by
the terminal
force,
directed as
above
stated,
in a foi-m in which the central-line is bent out of the
plane (x, z),
and then
the rod will also be twisted. It will
appear
that the defect of torsional
rigidity
is
quite
as
influential as that of flexural
rigidity
in
rendering possible
this kind of
buckling.
Fig.
61.
Let s be measured from the fixed end of the
central-line,
and let
x,, yi, Zj
be the
coordinates of the loaded end of this line. Let
x, y,
z be the coordinates of
any point Pj
on the strained central-line. For the
equilibrium
of the
part
of the rod contained between
the section drawn
through Pj
and the loaded end we take moments about axes drawn
through Pj parallel
to the fixed axes.
Using
the direction cosines defined
by
the scheme
(4)
of Article
253,
we have the
equations
-(JKmi-l-5ic'm2-l-C7-m3)-(Xi-x)if
=
0,
I
(45)
AKn^
+ BK'n., +
Crn^
=-0.
j
When we substitute for
k, k',
t from
equations (8)
of Article
253,
and for
l^,
... from
equations (7)
of the same
Article,
we have
AkI^
+
Bk'I^
+
CtI,
=
[{A
sin^
<\)
+
Bcos^
<^)
sin
^
+
{AB)ii\n<^
cos cos
^
cos
^] j-
-t-
Ccos
^&md
-j^
-)-[ (^cos'^'c^-f ^sin2(^)cosi^sin5cos5-(-(.l

5)sin<^cos(^sini//-sin54-C'cos\/'sin^cosfl]-T-,
A
(c7i -)-^(c'to2
+
Crm^
=
[(A am^4)
+
Bcos'^(f))cos\ff
+
(A-B)ain<l)Coa(f>sia
yjfCoaffl-^
+
CBinOsmyfr
-^
^dyjf

[{A coa^(t)
+
Biiin^(j))am-^aindcoad
+
(A-B)ain<f)Coa<f>coa\lraind

CiiiQylramBcoa6]
-j-,
AKny>r
Bk
'2
+
CWij
=
-{A-B)am^cos<i>am6'^-\-Ccoa6'^-\-{Aain^ecoa'^^
+
Bam^ea\.v?<i,
+
Ccoa^6)^.
L. E. 26
402 ROD BENT BY FORCES
'
[CH.
XIX
In
equations (45)
we now
approximate by taking
A and C to be small
compared
with
B,
and 6 to be
nearly equal
to
^tt,
while
<^
and
-^
are
small,
and also
by taking Xj
to be
equal
to I and x to be
equal
to . We
reject
all the
obviously unimportant
terms in the
expressions
for
(4(c?i+ ...),
We thus find the
equations
Since
rfy/(pfo=?n3=sindsini/^=\^ nearly,
we deduce from the first and second
equations
of
this set the
equation
and from the second and third
equations
of the same set we deduce the
equation
and,
on
eliminating dyjrjdt
between the two
equations
last
written,
we find the
equation
^
J
+
5-(^-^)^<^=0
(46)
This
equation
can be transformed into Bessel's
equation by
the substitutions
S=i{l-s)^RIJ{Aa), </,
=
,(;-.)* (47)
It becomes
and the
primitive
is of the form
4>H^'Ji {l)+S'J.^
(I)] (l-s)i, (48)
where A' and B' are constants.
Now when a
=
l, d\j//ds vanishes,
and the
twisting couple
Ct
vanishes;
hence
d(j)/d
vanishes. This condition
requires
that A' should vanish.
Further, <^
vanishes when
s=0,
and thus the critical
length
is
given by
the
equation J_, (|)
=
at
^
=
^PR/^{AC),
or
2.6 ^C^'"^*'
''
2.4...
(2).
6.14...
(8-2)
^"C^""
The lowest root of this
equation
for R'
1*1
AC is 16
nearly,
and we infer that the rod bent
by
terminal transverse load in the
plane
of
greatest
flexural
rigidity
is unstable if
l>y(AC)'/R%
where
-y
is a number
very nearly equal
to 2.
The result has been verified
experimentally by
A. G. M. Michell and L. Prandtl. It
should be observed that the
rod,
if of such a
length
as that
found,
will be bent a
good
deal
by
the load
R,
unless B is
large compared
with A and
C,
and thus the above method is not
applicable
to the
general problem
of the
stability
of the elastica for
displacements
out of
its
plane.
273. Rod bent
by
forces
applied along
its
length.
When forces and
couples
are
applied
to the rod at other
points,
as well as
at the
ends,
and the
stress-couples
are assumed to be
given by
the
ordinary
approximations (Article 255),
forms are
possible
in which the rod could not
be held
by
terminal forces and
couples only.
When there are no
couples
except
at the
ends,
the third of
equations (11)
of Article 254 becomes
272-274]
APPLIED ALONG ITS LENGTH
403
and this
equation
shows that to hold the rod bent to a
given
curvature
without
applying couples along
its
length,
a certain rate of variation of
the twist
along
the
length
is
requisite.
In other words a certain
twist,
indeterminate to a constant
prh,
is
requisite.
When there are no
applied couples except
at the
ends,
and the curvature
is
given,
while the twist has the
required
rate of
variation,
N and N' are
given by
the first two of
equations (2)
of Article 260. The
requisite
forces
X, Y,
Z of Article 254 and the tension T are then connected
by
the three
equations (10)
of that Article. We
may
therefore
impose
one additional con-
dition
upon
these
quantities.
For
example,
we
may
take Z to be
zero,
and
then we learn that a
given
rod can be held with its central-line in the form of
a
given
curve
by
forces which at each
point
are directed
along
a normal to the
curve,
provided
that the rod has a suitable twist.
Similar statements are
applicable
to the case in which the
rod,
in the
unstressed
state,
has a
given
curvature and twist.
As an
example*
of the
application
of these remarks we
may
take the case of a rod
which in the unstressed state forms a circular
hoop
of radius
r^,
with one
principal
axis of
each cross-section inclined to the
plane
of the
hoop
at an
angle /,
the same for all cross-
sections. We denote
by
B the flexural
rigidity corresponding
with this axis. The initial
state is
expressed by
the
equations
'=o=-''o"'cos/(|, Ko'=''o~'sin/, 7-o
=
0.
Let the rod be bent into a circular
hoop
of radius
r^,
with one
principal
axis of each
cross-section inclined to the
plane
of the
hoop
at an
angle/;,
the same for all cro.ss-sections.
The state of the rod is then
expressed by
the
equations
(Ci=-ri-icos/i, Ki'
=
r,-isin/,, ri
=
0.
To hold the rod in this state forces must be
applied
to each section so as to be
equivalent
to a
couple
about the central-line
;
the amount of this
couple per
imit of
length
is
--
{A sin/i cos/
-
B
cos/, sin/)
-^{A-B)
sin/, cos/,.
274. Rod bent in one
plane by
uniform normal
pressure.
We consider next the
problem
of a rod held bent in a
principal plane
by
normal
pressure
which is uniform
along
its
length.
The
quantity
X
of Article 254
expresses
the
magnitude
of this
pressure per
unit of
length.
Let F denote the resultant of the
shearing
force N and the tension T at
any
cross-section, F,^,
Fy
its
components parallel
to fixed axes of x and
y
in the
plane
of the bent central-line. We
may
obtain two
equations
of
equilibrium by resolving
all the forces which act
upon any portion
of the rod
parallel
to the fixed axes. These
equations
are
#i^.
+
zf
=
0,
^F,-xf^^O.
as as as
'
as
*
Cf. Kelvin and
Tait,
Nat.
Phil.,
Part
ii., pp.
166 et
seq.
262
404 ROD BENT BY NORMAL FORCES
[CH.
XIX
It follows that the
origin
can be chosen so that we have
F,=
-yX, F,
=
xX;
and therefore the
magnitude
of F at
any point
P of the strained central-line
is
rX,
where r is the distance
OP,
and the direction of F is at
right angles
to
OP. This result can be
expressed
in the
following
form :

Let
P,
and P be
any
two
points
of the strained
central-line,
and let
Fi
and
F^
be the result-
ants of the
shearing
force and tension on the cross-sections
through
P,
and
Pj,
the senses of
F^
and
F^
being
such that these forces arise from the action
of the rest of the rod on the
portion
between
Pi
and
P^.
From
Pj, P^
draw lines
PiO, Pfi
at
right angles
to the directions of
F^, F^.
We
may
regard
the arc
P1P2
as the limit of a
polygon
of a
large
number of
sides,
and
this
polygon
as in
equilibrium
under the flexural
couples
at its
ends,
the
forces
Pi, F^,
and a force Xhs directed at
right angles
to
any
side of the
polygon
of which the
length
is hs. The forces are at
right angles
to the
sides of the
figure
formed
by
OPi, OP2
and this
polygon,
and are
propor-
tional to them
;
and the
lengths
of
OPi
and
OP2
are
F^jX
and
F^jX.
The
senses in which the lines must be drawn are indicated in
Fig.
62*.
Pi
Fig.
62,
Then
,dr
ds
The
stress-couple
0' satisfies the
equation
dO'
Let r denote the distance OP.
-
N=-F
^
=
-rX
dr
dr
ds ds'
Hence we have G'

|Xr^
+
const.
In the
particular
case where the central-line in the unstressed state is a
straight
line or a
circle,
the curvature
Ijp
of the curve into which it is bent is
given by
the
equation
P//3
=
iZr=
+
const
(49)
The
possible
forms of the central-line can be determined from this
equation f.
*
In the
right-hand figure OP^P.^
is shown as a
force-polygon.
The
theory
is due to M.
Ldvy,
J. de Math.
(Liouville), {Sii. 3),
t. 10
(1884).
t
The
complete integration
of
equation (49) by
means of
elliptic
functions was effected
by
G. H.
Halpben, Paris,
C.
B.,
t. 98
(1884).
See also his Traite des
fonctwns elliptiques,
Partie
2,
Ch. 5
(Paris, 1888).
The
subject
has been
investigated
further
by
A. G.
Greenhill,
Math.
Ann.,
Bd. 52
(1899).
274-276]
HEIGHT CONSISTENT WITH STABILITY 405
275.
Stability
of circular
ring
under normal
pressure.
When the central-line in the unstressed state is a circle of radius
a,
and the rod is
very
slightly bent, equation (49)
can be written in the
approximate
form
d^u
_
X
where
l/u
and 6 are the
polar
coordinates of a
point
on the central-line referred to as
origin,
and c is a constant. The value of u differs
very
little from
1/a,
and we
may
there-
fore
put
u
=
l/a
+
^,
where
^
is
small,
and obtain the
approximate equation
Hence
^
is of the form
^QCOH{nd+y),
where
^^
and
y
are
constants,
and n is
given by
the
equation
Now the function
$
must be
periodic
in d with
period Stt, for, otherwise,
the rod would
not continue to form a
complete ring.
Hence n must be an
integer.
If n were
1,
the circle
would be
displaced
without deformation. The least value of the
pressure
X
by
which
any
deformation of the circular form can be
produced
is obtained
by putting
n
=
2. We infer
that,
if
X<ZBIa\
the
ring simply
contracts under the
pressure,
but the
ring
tends to
collapse
if
X>ZBIaK
(50)*
276.
Height
consistent with
stabilityf.
As a further
example
of the
equilibrium
of a rod under forces
applied
along
its
length,
we consider the
problem
of a vertical
column,
of uniform material and
cross-section,
bent
by
its own
weight.
Let a
long
thin rod be set
up
in a vertical
plane
so that the
lower end is constrained to remain
vertical,
and
suppose
the
length
to be so
great
that the rod bends. Take the
origin
of
fixed axes of x and
y
at the lower
end,
draw the axis of x
vertically upwards
and the axis of
y horizontally
in the
plane
of
bending. (See
Fig.
63.)
For the
equilibrium
of the
portion
of the rod contained between
any
section and the free
end,
we
resolve
along
the normal to the
central-line,
and
then,
since
the central-line is
nearly
coincident with the axis of
x,
we
find the
equation
where Wis the
weight
of the rod. The
equation
of
equilibrium dGjds
+
N=0
can, therefore,
be
replaced by
the
approximate
equation
.1-
Fig.
63.
B^P.W- I
p
=
0, .(51)
where
p
is written for
dy/dx.
The terminal conditions are that
dpjdx
vanishes
at x
=
l,
and
y
and
p
vanish at x
=
0.
*
The result is due to M.
Levy,
he. cit.
t
The
theory
is due to A. G.
Greenhill, Cambridge
Phil. Soc.
Proe.,
vol. 4
(1881).
It has
been discussed
critically by
C.
Chree, Cambridge
Phil. Soe.
Proc.,
vol. 7
(1892).
406 HEIGHT CONSISTENT WITH STABILITY
[CH.
XIX
Equation (51)
can be transformed into Bessel's
equation by
the sub-
stitutions
^=i\/(5)^^-'^^''
-p='?(^-^)^ (52)
It becomes
and the
primitive
is of the form
p
=
[A'J^{i)
+
B'J_^m{l-x)\
(53)
where A' and B' are constants.
To make
dp/dx!
vanish at x
=
l we must have
A'=0,
and to make
p
vanish at x
=
we must have
J^_j(f)
=
at
^
=
^l(W/B)K
Hence the
critical
length
is
given by
the
equation
3.2 B
^-"^v )
3.6...(3n).2.5...(3n-l)
B"
The lowest root of this
equation
for
PWjB
is
(7
. 91
...),
and we infer that the
rod will be bent
by
its own
weight
if the
length
exceeds
(2.83 ...) V(5/W).
Greenhill
(loc.
cit.
p. 405)
has worked out a number of cases in which the
rod is of
varying
section,
and has
applied
his results to the
explanation
of the
forms and
growth
of trees.
CHAPTER XX.
VIBRATIONS OF RODS. PROBLEMS OF DYNAMICAL RESISTANCE.
277. The vibrations of thin rods or
bars,
straight
and
prismatic
when
unstressed,
fall
naturally
into three classes :
longitudinal,
torsional,
lateral.
The
"
longitudinal
"
vibrations are characterized
by
the
periodic
extension
and contraction of elements of the
central-line, and,
for this
reason,
they
will
sometimes be described as
"
extensional." The
"
lateral
"
vibrations are
characterized
by
the
periodic bending
and
straightening
of
portions
of the
central-line,
as
points
of this line move to and fro at
right angles
to its
unstrained direction
;
for this reason
they
will sometimes be described as
"flexural." In
Chapter
xii. we
investigated
certain modes of vibration
of a circular
cylinder.
Of these modes one class are of
strictly
torsional
type,
and other classes are
effectively
of extensional and flexural
types
when the
length
of the
cylinder
is
large compared
with the radius of its cross-section.
We have now to
explain
how the
theory
of such vibrations for a thin rod of
any
form of cross-section can be deduced from the
theory
of
Chapter
xviii.
In order to
apply
this
theory
it is
necessary
to assume that the
ordinary
approximations
described in Articles 2.55 and 258 hold when the rod is
vibrating.
This
assumption may
be
partially justified by
the observation
that the
equations
of motion are the same as
equations
of
equilibrium
under
certain
body
forces

the reversed kinetic reactions. It then amounts to

assuming
that the mode of distribution of these forces is not such as to
invalidate
seriously
the
approximate equations (21), (22), (23)
of Article
258. The
assumption may
be
put
in another form in the statement
that,
when the rod
vibrates,
the internal strain in the
portion
between two
neighbouring
cross-sections is the same as it would be if that
portion
were
in
equilibrium
under tractions on its
ends,
which
produce
in it the instan-
taneous
extension,
twist and curvature. No
complete justification
of this
assumption
has been
given,
but it is
supported by
the
results,
already
cited,
which are obtained in the case of a circular
cylinder.
It seems to be
legitimate
to state that the
assumption gives
a better
approximation
in the
case of the
graver
modes of
vibration,
which are the most
important,
than in
408
EXTENSIONAL,
TORSIONAL AND FLEXURAL MODES
[CH.
XX
the case of the modes of
greater frequency,
and
that,
for the
former,
the
approximation
is
quite
sufficient.
The various modes of vibration have been
investigated
so
fully by
Lord
Rayleigh*
that it will be
unnecessary
here to do more than obtain the
equations
of vibration. After
forming
these
equations
we shall
apply
them
to the discussion of some
problems
of
dynamical
resistance.
278. Extensional vibrations.
Let w be the
displacement, parallel
to the
central-line,
of the centroid of
that cross-section
which,
in the
equilibrium
state,
is at a distance s from
some chosen
point
of the line. Then the extension is
dw/ds,
and the tension
is E(o
{dw/ds),
where E is
Young's
modulus,
and tu the area of a cross-
section. The kinetic
reaction,
estimated
per
unit of
length
of the
rod,
is
pm (dhujdf),
where
p
is the
density
of the material. The
equation
of
motion,
formed in the same
way
as the
equations
of
equilibrium
in Article
254,
is
^^=^a>
^^>
The condition to be satisfied at a free end is
dw/ds
=
0;
at a fixed end w
vanishes.
If we form the
equation
of motion
by
the
energy-method (Article 115)
we
may
take
account of the inertia of the lateral motion t
by
which the cross-sections are extended
or contracted in their own
planes.
If x and
i/
are the coordinates of
any point
in a cross-
section,
referred to axes drawn
through
its
centroid,
the lateral
displacements
are
-
(TX
(dwlds),

(ry (dw/ds),
where a- is Poisson's ratio. Hence the kinetic
energy per
unit of
length
is
y-m-'Hm'}-
where A' is the radius of
gyration
of a cross-section about the central-line. The
potential
energy per
unit of
length
is
and, therefore,
the variational
equation
of motion is
where the
integration
with
respect
to is taken
along
the rod. In
forming
the variations
we use the identities
[dsdtdsdt ds^dfi
)~di\dsdt
dt dsdfi
)'^dt\dsdt
ds ds^dt
)'
*
Theory of Sound,
Chapters
vii. and viii.
t The lateral strain is
already
taken into account when the tension is
expressed
as the
product
of K and oi
(dwlds).
If the
longitudinal
strain alone were considered the constant that enters into
the
expression
for the tension would not be E but X-h
2yn.
'
277-280]
OF VIBRATION OF THIN RODS 409
and,
on
integrating by parts,
and
equating
to zero the coefficient of 8w under the
sign
of
double
integration,
we obtain the
equation
By retaining
the term
parK^dhD/ds^dfl
wo should obtain the correction of the
velocity
of
wave-proiwigation
which was found
by
Pochhamnier and Chree
(Article 201),
or the
correction of the
frequency
of free vibration which was calculated
by
Lord
Rayleigh*.
279. Torsional vibrations.
Let
yfr
denote the relative
angular displacement
of two
cross-sections,
so
that
d^lr/ds
is the twist of the rod. The centroids of the sections are not
di.splaced,
but the
component displacements
of a
point
in a cross-section
parallel
to axes of x and
y,
chosen a,s
before,
are

ylry
and
yjrx.
The torsional
couple
is C
(dyfr/ds),
vihere C is the torsional
rigidity.
The moment of the
kinetic reactions about the
central-line,
estimated
per
unit of
length
of the
rod,
is
pwK'-id-yjr/dt").
The
equation
of
motion,
formed in the same
way
as
the third of the
equations
of
equilibrium (11)
of Article
254,
is
'-'^'a?-3'
W
The condition to be satisfied at a free end is
dyfr/ds
=
;
at a fixed end
yjr
vanishes.
When we
apply
the
energy-method,
wo
may
take account of the inertia of the motion
by
which the cross-sections are deformed into curved surfaces. Lot
cj)
be the torsion
function for the section
(Article 216).
Then the
longitudinal displacement
is
(^(3\^/3),
and the kinetic
energy
of the rod
per
unit of
length
is
The
potential energy
is
|C(&^/3)',
and the
equation
of
vibration,
formed as
before,
is
,.K^^-,{f^^,.),%
=
C
3V
8

By inserting
in this
equation
the values of C and
j<f)^du>
that
liclong
to the
section,
wo
could obtain an
equation
of motion of the same form as
(2)
and could work out a
correction for the
velocity
of
wave-propagation
and the
frequency
of
any
mode of vibration.
In the case of a circular
cylinder
there is no correction and the
velocity
of
propagation
is
that found in Article 200.
280. Flexural vibrations.
Let the rod vibrate in a
principal plane,
which we take to be that of
(x, z)
as defined in Article 2.52. Let u denote the
displacement
of the
centroid of
any
section at
right angles
to the unstrained central-line. We
may
take the
angle
between this line and the
tangent
of the strained central-
line to be
8m/9s,
and the curvature to be
'bhij'bs^.
The flexunil
couple
Q' is
Bd'^ujds-,
where B
=
Eaik"',
k'
being
the radius of
gyration
of the cross-section
*
Theory of
Sowtd,
157.
410 FLEXURAL VIBRATIONS OF RODS
[CH.
XX
about an axis
through
its centroid at
right angles
to the
plane
of
bending.
The
magnitude
of the kinetic
reaction,
estimated
per
unit of
length,
is,
for a
first
approximation, pto {dhojdt^),
and its direction is that of the
displacement
u.
The
longitudinal displacement
of
any point
is

a;
(du/ds) ;
and therefore
the moment of the kinetic
reactions, estimated
per
unit of
length,
about an
axis
perpendicular
to the
plane
of
bending
is
pcok'^ (d^a/dsdt-).
The
equations
of vibration formed in the same
way
as the second
equation
of each of the
sets of
equations
of
equilibrium (10)
and
(11)
of Article 254 are
^
=
p-,
Ecok^-^~+N
=
p.k^^^,
(4)
and,
on
eliminating
N,
we have the
equation
of vibration
PKd^-^ d^t^)
=
-^^^
^^>
If
"
rotatory
inertia
"
is
neglected
we have the
approximate equation
pw
=
-^^
d^
^^)
and the
shearing
force N at
any
section is

Ecok'^d^u/ds^
At a free end
dHi/ds'
and
d^u/di^
vanish,
at a
clamped
end u and
du/ds
vanish,
at a
"
supported
"
end u and
d''u/ds^
vanish.
By retaining
the term
representing
the effect of
rotatory
inertia we could obtain
a correction of the
velocity
of
wave-propagation,
or of the
frequency
of
vibration,
of the
same kind as those
previously
mentioned*. Another
correction,
which
may
be of the
same
degree
of
importance
as this when the section of the rod does not
possess
kinetic
symmetry, may
be obtained
by
the
energy-method, by taking
account of the inertia of the
motion
by
which the cross-sections are distorted in their own
planes
t. The
components
of
displacement parallel
to axes of .v and
y
in the
plane
of the
cross-section,
the axis
of X
being
in the
plane
of
bending,
are
o^u
and the kinetic
energy per
unit of
length
is
expressed correctly
to terms of the fourth
order in the linear dimensions of the cross-section
by
the formula
where k is the radius of
gyration
of the cross-section about an axis
through
its centroid
drawn in the
plane
of
bending.
The term in o-
(k"-
-
X--') depends
on the inertia of the
motion
by
which the cross-sections are distorted in their
planes,
and the term in k'^
depends
on the
rotatory
inertia. The
potential energy
is
expressed by
the formula
i^^^-(sy

Cf. Lord
Bayleigh, Theory of Sound,
186.
t
The cross-sections are distorted into curved surfaces and inclined
obliquely
to the strained
central-line,
but the inertia of these motions would
give
a much smaller correction.
280,
281]
LONGITUDINAL IMPACT
411
The variational
equation
of motion is
^H4KST-<--'^lll,--(S)')--(2)l
In
forming
the variations we use the identities
a 8m a^M dtc d^8u
a a2 "a
^"^
a^
\ds^
ds ~d?
*";

OS
a*M a /. a^M
,
awazsMN
,
a /. a%
dsudhi\
^^"
aiw
=
^i
(,^"
a72a<
+
w 1^
j
+
ai
V*"
a^
"
aT
a^j
as well as identities of the
types
used in Article 278. The
resulting equation
of motion is
p[g-{^^(l-.)4-.M3||.]=-^^iS
(7)
Corrections of the
energy
such as that considered here
will,
of
course,
aflfect the
terminal conditions at a
free,
or
supported end,
as well as the differential
equation
of
vibration. Since
they
rest on the
assumption
that the internal strain in
any
small
portion
of the
vibrating
rod contained between
neighbouring
cross-sections is the same as
in a
prism
in which the
right extension,
or
twist,
or curvature is
produced by
forces
applied
at the ends and
holding
the
prism
in
equilibrium, they
cannot be
regarded
as
very
rigorously
established. Lord
Rayleigh {loc. cit.)
calls attention to the increase of im-
portance
of such corrections with the
frequency
of the vibration. We have
already
remarked that the
validity
of the fundamental
assumption
diminishes as the
frequency
281. Rod fixed at one end and struck
longitudinally
at the other*.
We shall illustrate the
application
of the
theory
of vibrations to
problems
of
dynamical
resistance
by solving
some
problems
in which a.
long
thin rod is
thrown into extensional vibration
by
shocks or
moving
loads.
We take first the
problem
of a rod fixed at one end and struck at the
other
by
a massive
body moving
in the direction of the
length
of the rod.
We measure t from the instant of
impact
and s from the fixed
end,
and
we denote
by
I the
length
of the
rod,
by
m the ratio of the mass of the
striking body
to that of the
rod,
by
V the
velocity
of the
body
at the instant
of
impact, by
w the
longitudinal displacement,
and
by
a the
velocity
of
propagation
of extensional waves in the rod.
The differential
equation
of extensional vibration is
W^""^^
^^)
The terminal condition at s
=
is w
=
0. The terminal condition at
s=l is the
equation
of motion of the
striking body,
or it is
,
d-w
,
dw , ,
"^^W^-'^'d^
^^^
*
Cf. J.
Boussinesq, Applications
des
potentiels..., pp.
508 et
seq.,
or Saint-Venant in the
'Annotated
Clebsch,'
Note
finale
du

60 and
Changements
et additions.
412
LONGITUDINAL IMPACT
[CH.
XX
since the
pressure
at the end
is,
in the notation of Article
278,

Em
(dw/ds),
and
Em/a^
is
equal
to the mass of the rod
per
unit of
length.
The initial
condition is
that,
when
t=0,
w
=
for all values of s between and
I,
but &t s
=
l
lim
(dwldt)
=
-V, (10)
<=+o
since the
velocity
of the struck end
becomes,
at the instant of
impact,
the same as that of the
striking body.
We have to determine w for
positive
values of
t,
and for all values of
s between and
I,
by
means of these
equations
and conditions. The first
step
is to
express
the solution of the differential
equation (8)
in the form
w=f(at-s)
+
Fiat
+
s), (11)
where
/
and F denote
arbitrary
functions.
The second
step
is to use the terminal condition at s
=
to eliminate one
of the
arbitrary
functions. This condition
gives
in fact
f{at)
+
F{at)
=
0,
and we
may,
therefore,
write the solution of
equation (8)
in the form
w=f(at-s)-f{at
+
s) (12)
The third
step
is to use the initial conditions to determine the function
/
in a certain interval. We think of
/
as a function of an
argument
^,
which
may
be
put equal
to at

s or at + s when
required.
Since
dw/ds
and
dw/dt
vanish with t for all values of s between and I we
have,
when
l>^>0,
-/'(-^)-/'(f)
=
0,
/'(-?)-/'(?)
=
0.
Hence it follows
that,
when
l>^>

I,
f {^)
vanishes and
/(f)
is a
constant which can be taken to be zero
;
or we have the result
when
1>^>-1,
/(0
=
(13)
The fourth
step
is to use the terminal condition
(9)
at s
=
l to form an
equation by
means of which the value of
/(?)
as a function of
f
can be
determined outside the interval I >
^>

I. The
required equation,
called
the
"continuing equation*,"
is
ml
[/" {at
-
I) -f" (at
+
I)] =/' (at
-
1) +/'
(at
+
I),
or,
as it
may
be
written,
f"(0
+
(Vrnl)f'(0=f"(^-2l)-(l/7nl)f'(^-2l) (14)
W^e
regard
this
equation
in the first instance as an
equation
to determine
/'(f).
The
right-hand
member is
known,
it has in fact been shown to be
zero,
in the interval 3l>
^>
I. We
may
therefore determine the form of
f'(0
in this interval
by integrating
the
equation (14).
The constant of
integration
is to be determined
by
means of the condition
(10).
The function

i^quation promotrice
of Saint-Venant.
281]
ROD STRUCK AT ONE END
418
f'{^)
will then be known in the interval Sl>
^>
I,
and therefore the
right-
hand member of
(14)
is known in the interval 5i >
f
> SI. We determine
the form of
/'(^)
in this interval
by integrating
the
equation (14),
and we
determine the constant of
integration by
the condition that there is no dis-
continuity
in the
velocity
at s
=
I after the initial instant. The function
/'(f)
will then be known in the interval ol>
^>
31.
By proceeding
in this
way
we can determine
/'(5')
for all values of
^
which exceed

I.
The
integral
of
(14)
is
always
of the form
/' (0
=
Ce-fw
+
e-f/^|ef'""
{/" (f- 21)
-
i^/'
(?
-
20) df,
. .
.(15)
where C is a constant of
integration.
When M>
^>
I the
expression
under
the
sign
of
integration
vanishes,
and
/' (f)
is of the form Ce~^"^. Now the
condition
(10) gives
a[f'(-l
+
o)-f'(l
+
0)]
=
-V,
OTf'{l
+
0)=V/a.
Hence Oe~""'=
Vja,
and we have the result
when.S?>r>?, /'(?)
=
^e-<f-''""
-.(16)
We observe that
/'(f)
is discontinuous at
f
=
i.
When 51 >
^>2l
we have
f" (^_ 21)
-
(1/mO/' (f
-
20
= -
2
(VImla)
e-if-^/m^,
and
equation (15)
can be written
/' (f)
=
Ce-f""'
-
2{Vlmla) (f
-
30
g-'f-'""^.
The condition of
continuity
of
velocity
at s
=
Z at the instant t
=
2l/a gives
/' {I
-
0) -/'
m
-
0) =/' (^
+
0) -/' (3^
+
0),
V V
or
-
e-'i""
=
- -
Ce-'"",
a
a
giving"
C=(F/a)(e>"
+
e=""').
Hence,
when 51 >
^>3l,
/'(0
=
Ze-<^-<./"^ +
^|l
-
J^(f-30}e-'<-'"""'
(17)
When 7l>
^
> 51 we have
/"
(?- 20
-
^/'
(?
-
20
=
-
^
[^-'^-^""^
+
2e-'f-=""']
4F
+;^(r-50e-<-"^
m,''I'a
and
equation
(15)
can be written
mla
f(Q
=
C'e-*'
-
-^(?-
50 [e-'f-'"'""
+
2e-'f-"""^]
+
^-
(?- 50'e-<f-'""^
414 LONGITUDINAL IMPACT
[CH.
XX
The condition of
continuity
of
velocity
at s
=
i at the instant t
=
U/a
gives
f'(M-0)-f'{5l-0)=f'(3l
+
0)-f'(ol
+
0),
V V 4F V
or

e-*
(e"*"*
+
e-^"")
+

g-^*"
=

(e-^""
+
1)
-
Ce-""",
a a ma a
giving
C
=

je'""
+
^1 ]
e*"* + e^'A .
Hence,
when 7? >
f
>
51,
/'(?)
=
^
e-'^-''""' +
^
{l
-
^
(^
-
^^)}
^~
""^"""
+
V
a
+Jj,(r-5o^ie-'f-""'
T
..(18)
The function
/(f)
can be determined
by integrating /'(?"),
and the
constant of
integration
is to be determined
by
the condition that there is
no sudden
change
in the
displacement
at s
=
l. This condition
gives, by
putting
t
=
0,
2l/a,
. . . such
equations
as
0=/(-Z
+
0)-/(Z
+
0),
f(l-0)-f(3l-0)=f{l
+
0)-f(Sl
+
0),
from
which,
since
f(

l
+
0)
and
f(l0)
vanish,
we find
/(Z+O)
=
0=/(i-O), f{3l
+
0)=f(Sl-0),
....
Hence there is no
discontinuity
in
/(f),
as is otherwise
evident,
since
/'(f) possesses only
finite discontinuities
separated by
intervals in which it
is continuous. We have therefore
merely
to
integrate /'(f)
in each of the
intervals '3l>
^>l,
5l>
^>3l,
... and determine the constants of
integration
so that
f(l)
=
and
/(f)
is continuous. We find the
following
results :
when 'M>
^>l,
fiO
=
(mlV/a){l
-
e-(f-"^)
;
when 5Z >
f
>
3Z,
when 7i >
f
>
ol,
y
/(0-^li--^
'|
+
!^(i+'
'-3l)lml
mlV
1+-
H'
(f-30}e--
(f-5me"'^~""^;
..(19)
The solution
expresses
the result
that,
at the instant of
impact,
a wave
of
compression
sets out from the struck
end,
and travels towards the fixed
end,
where it is reflected. The motion of the
striking body generates
a continuous series of such
waves,
which advance towards the fixed
end,
and are reflected there.
281,
282]
ROD STRUCK AT ONE KND
416
In the above solution we have
proceeded
as if the
striking body
became attached to the
rod,
so that the condition
(9)
holds for all
positive
values of t
; but,
if the bodies remain
detached,
the solution continues to hold so
long only
as there is
positive pressure
between
the rod and the
striking body. When,
in the above
solution,
the
pressure
at $=l becomes
negative,
the
impact
ceases. This
happens
when
/' {at-l}+f' (ai
+
l)
becomes
negative.
So
long
as 21 >at >0 this
expression
is
equal
to
( V/a)
e"""'"' which is
positive.
When
4l> at>
21,
it is
which vanishes when
2at/ml=4/m
+
2
+
e^^''",
and this
equation
can have a root in the
interval
4l>at>2l if 2
+
e~''"'<4jm.
Now the
equation
2
+
e-^"'

4/m
has a root
lying
between m
=
l and
m=2,
viz. : m= r73....
Hence,
if
to<1'73,
the
impact
ceases at
an instant in the interval
4lla>t>2lja,
and this instant is
given by
the
equation
t=-(2 +
m
+ime-^l'").
a
^
If 77J>1'73 we
may
in like manner determine whether or no the
impact
ceases at
an instant in the interval
6l/a>t>4l/a,
and so on. It
may
be shown also that the
greatest compression
of the rod occurs at the fixed
end,
and
that,
if to <
5,
its value
is
2(1 +6"-/) V/a, but,
if
to>5,
its value is
approximately equal
to
(l+>Jm)Vla.
If the
problem
were treated as a statical
problem by neglecting
the inertia of the
rod,
the
greatest compression
would be
Jm ( V/a).
For further details in
regard
to this
problem
reference
may
be made to the authorities cited on
p.
411.
282. Rod free at one end and struck
longitudinally
at the other*.
When the end
=
is
free, dwjds
vanishes at this end for all values of
t,
or we have
-f'{at)
+
F'{at)=0.
Hence we
may put /'(f)=/(f)
'lud write instead of
(12),
w=f{at-s)+f{at+s),
and,
as
before,
we find
that/(f)
vanishes in the interval
l>(>

I.
The
continuing equation
is now
/"(f)
+
(l/mO/'(0=-/"(f-20
+
(lM)/'(f-20
and the
discontinuity of/'(f)at f=Z
is determined
by
the
equation
ay'{-l+0)
+
f'{l
+
Q)\=-V,
or
/' (^+0)=
-
F/a.
Hence we find the results :
when
ZI>^>1, f(f)
=
-
-<.-(<:-')/",
when
bl>(>Zl,
/'(0=
_Z,-,C-.,..
+
Z{l_ ^^(f_30}.--)/nU
Now the extension at s
=
l is
/' {at
+
l)-f' {at I), and,
until i
=
2l/a,
this is
-(K/a) -''"",
which is
negative,
so that the
pressure
remains
positive
until the instant t
=
2l/a ; but,
immediately
after this
instant,
the extension becomes
( V/a) (2- e"-'"*),
which is
positive,
so
*
Cf. J.
BouBsinesq,
loc. cit.
p.
411.
416 DYNAMICAL RESISTANCE
[CH.
XX
that the
pressure vanishes,
and the
impact
ceases at the instant t
=
2l/a,
that is to
say
after the time taken
by
a wave of extension to travel over twice the
length
of the rod.
The wave
generated
at the struck end at the instant of
impact
is a wave of
compression ;
it is reflected at the free end as a wave of extension. The
impact
ceases when this
reflected wave reaches the end in contact with the
striking body.
The state of the rod
and the
velocity
of the
striking body
at this instant are determined
by
the above formulae.
The
body
moves with
velocity
Ve~'^l"' in the same direction as before the
impact ;
and the
rod moves in the same
direction,
the
velocity
of its centre of mass
being mV(l -g-^/).
The
velocity
at
any point
of the rod is 2 Fe"'/'" cosh
{s/ml),
and the extension at
any point
1
of it is 2
( V/a)
e
"
sinh
(s/nd),
so that the rod rebounds
vibrating.
283. Rod loaded
suddenly.
Let a massive
body
be
suddenly
attached without
velocity
to the lower
end of a
rod,
which is
hanging vertically
with its
upper
end fixed. With a
notation similar to that in Article
281,
we can write down the
equation
of
vibration in the form
d^w

dhu
,-
..
w-'^d?-^^'
(20>
and the value of w in the
equilibrium
state is
^gs (21

s)/a'.
Hence we
write
w
=
^gs(2l-s)/a'
+
w',
(21)
and then w' must be of the form
w'
=
<j>(at-s)-<f>(at
+
s),
(22)
'
and,
as
before,
we find
that,
in the interval I >
^>

I,
(j)(^)
vanishes.
The
equation
of motion of the attached mass is
fd'w'X
a"
fdw'\
which
gives
the
continuing equation
f'(D
+
;^,f(O
=
f'(r-20-^-,^'(r-20-f.,
(24)
and the constants of
integration
are to be determined so that there is no
discontinuity
of
velocity
or of
displacement.
We find the
following
results :

when
U>^>1,
,^'(^)
=
_iimi[l-e-<f-'""^}
w
^(r)=-|"^*{y-i+e-'^-'"""
.(25)
Further the
equations by
which
^'(f)
are determined in this
problem
can be identified with those
by
which
/(S')
was determined in Article 281
by
writing

gja
for V. The solution is not restricted to the
range
of values of
t within which the tension at the lower end remains
one-signed.
282,
283]
ROD LOADED SUDDENLY
417
The
expression
for the extension at
any point
is
g{l- s)lafi
-
<p' {at -g)-<f)' (at+t),
and,
at the fixed
end,
this is
equal
to
lg/a^-24,'{at),
or
lgja^
+
^{glaV)f{at),
where
/
is the function so denoted in Article 281. The maximum value occurs when
f{at)
=
Q.
Taking
wi
=
l,
so that the attached mass is
equal
to the mass of the
rod,
we find from
(16)
that
f{at)
does not vanish before t
=
Zlja,
but from
(17)
that it vanishes between
t
=
3lja
and t
=
5l/a
if the
equation
\+e'-{l-2{i-3l)ll}=0
haa a root in the interval
5l>(>Sl.
The root is
f
=
i{3+i (l
+
l/e^)},
or
f=;(3-668),
which is in this interval. The
greatest
extension at the fixed end is
^
{1
+
2e-2-58
[
-
1
+
e2
{1
+
2
(0-568)}]},
or
{Ig/a^) (1 +4e-o-^),
or
(327) Ig/aK
The statical strain at the fixed
end,
when the rod
supports
the attached mass in
equilibrium,
is
2lgla^,
and the ratio of the maximum
dynamical
strain to this is 1'63 : 1. This strain occurs at the instant t
=
(3'568)i/a.
Taking
m=2,
so that the attached mass is twice the mass of the
rod,
we find from
(16)
that
/' {at)
does not vanish before t
=
Zlja,
but from
(17)
that it vanishes between
t=Zlja
and
t=-bl/a
if the
equation
l+e{l-(f-30/^}
=
has a root in the interval
bl>(>3l.
The root is
f=Z(44-l/e),
or
f=?(4'368),
which is
in this interval. The
greatest
extension at the fixed end is
^4
{1
+4e-i(3-388)
[
_
1
+(1
+
1-368) e]},
or
lg/a^{l+Be~'''^
or
(5-04) Z^/o^.
The statical strain in this case is
Zlgja*,
and the
ratio of the maximum
dynamical
strain to the statical strain is 1'68 : 1. This strain
occurs at the instant <
=
(4-368) ^/a.
Taking
to
=
4,
so that the attached mass is four times the mass of the
rod,
we find
that
f{at)
does not vanish before t
=
5l/a,
but from
(18)
that it vanishes between
t=6l/a
and
t^llja
if the
equation
l-M(f-50/}*
+
[l-(f-50/^
+
i(f-5W^2]e=0
has a root in the interval
1l>^>bl.
The smaller root is
f=i (6-183),
which is in this
interval. The
greatest
extension at the fixed end is
where
f
is
given by
the above
equation.
The extension in
question
is therefore
%
[9
+8e-i(i-"3)
{2e-*- (1-183)}],
which is found to be
(9-18) {Igja-).
The statical strain in this case is 5
{Ig/a^),
and the
ratio of the maximum
dynamical
strain to the statical strain is 1-84
nearly.
This strain
occurs at the instant <
=
(6-183) l/a.
The
noteworthy
result is
that,
even when the attached mass is not a
large multiple
of
the mass of the
rod,
the
greatest
strain due to sudden
loading
does not fall far short
of the theoretical
limit,
viz. twice the statical strain.
(Cf.
Article
84.)
L. E. 27
418 DYNAMICAL RESISTANCE
[CH.
XX
284.
Longitudinal impact
of rods.
The
problem
of the
longitudinal impact
of two rods or bars has been
solved
by
means of
analysis
of the same kind as that in Article 281 *. It is
slightly
more
complicated,
because different undetermined functions are
required
to
express
the states of the two
bars;
but it is
simpler
because
these functions are themselves
simple.
The
problem
can be solved also
by
considering
the
propagation
of waves
along
the two rods+. The extension e
and
velocity
v at the front of an extensional wave
travelling along
a rod are
connected
by
the
equation
e
=

vja. (Of.
Article
205.)
The same relation
holds at
any point
of a wave of
compression travelling entirely
in one
direction,
as is obvious from the formula
w=f{at

s)
which characterizes
such a wave. When a wave of
compression travelling along
the rod reaches
a free
end,
it is reflected
;
and the nature of the motion and strain in the
reflected wave is most
simply investigated by regarding
the rod as
produced
indefinitely,
and
supposing
a wave to travel in the
opposite
direction
along
the continuation of the rod in such a
way
that,
when the two waves are
superposed,
there is no
compression
at the end section. It is clear that the
velocity propagated
with the
"image"
wave in the continuation of the rod
must be the same as that
propagated
with the
original
wave,
and that the
extension
propagated
with the
"
image
"
wave must be
equal numerically
to
the
compression
in the
original wavej.
Now let
I,
I' be the
lengths
of the
rods,
supposed
to be of the same
material and cross-section
,
and let
V,
V be their
velocities,
supposed
to be
in the same sense. We shall take I > I'. When the rods come into contact
the ends at the
junction
take a common
velocity,
which is determined
by
the
condition that the
system consisting
of two
very
small
contiguous portions
of
the
rods,
which have their motions
changed
in the same
very
short
time,
does
not,
in that
time,
lose or
gain
momentum. The common
velocity
must
therefore be
i{V+ V).
Waves set out from the
junction
and travel
along
both
rods,
and the
velocity
of each element of either
rod,
relative to the rod
as a
whole,
when the wave reaches
it,
is
J (
F~
V),
so that the waves are
waves of
compression,
and the
compression
is
^ (
F^
V')/a.
To trace the
subsequent
state of the shorter rod
I',
we think of this rod as
continued
indefinitely beyond
the free
end,
and we reduce it to rest
by
impressing
on the whole
system
a
velocity equal
and
opposite
to V. At the
instant of
impact
a
positive wave||
starts from the
junction
and travels
along
the rod
;
the
velocity
and
compression
in this wave are
^ (
V'v
V)
and

Saint-
Venant,
J. de math.
(Liouville), (S6r. 2),
t. 12
(1867).
t
Cf. Kelvin and
Tait,
Nat.
Phil.,
Part
I, pp. 280,
281.
t
Cf. Lord
Bayleigh, Theo7-y of
Sound,
vol.
2,
257.
Saint-Venant,
loc.
cit.,
discusses the case of different materials or sections as well.
II
An extensional wave is
"
positive
"
or
"
negative" according
as the
velocity
of the material
is in the same sense as the
velocity
of
propagation
or in the
oppo.site
sense.
284]
LONGITUDINAL IMPACT OF RODS 419
^ (
F~
V')ja.
At the same instant a
negative
"
image
"
wave starts from the
section distant 21' from the
junction
in the fictitious continuation of the
rod;
the
velocity
and extension in this
"image"
wave are
^(F-^/F')
and
^(F<*F')/a.
After a time
I'/a
from the instant of
impact
both these waves reach the free
end,
and
they
are then
superposed. Any part
of the actual rod in which
they
are
superposed
becomes unstrained and takes the
velocity
V/^V.
When the reflected wave reaches the
junction,
that is to
say
after a time
2l'la
from the instant of
impact,
the whole of the rod I' is
moving
with the
velocity
F"^
F',
and is unstrained.
Hence,
superposing
the
original velocity
F',
we have the result
that,
after the time taken
by
an extensional wave to
travel over twice the
length
of the shorter
rod,
this rod is unstrained and is
moving
with the
velocity
F
originally possessed by
the
longer
rod.
To trace the state of the
longer
rod I from the
beginning
of the
impact,
we think of this rod as continued
indefinitely beyond
its free
end,
and we
reduce it to rest
by impressing
on the whole
system
a
velocity equal
and
opposite
to F. At the instant of
impact
a
positive
wave starts from the
junction
and travels
along
the
rod;
the
velocity
and
compression
in this wave
are
^ (
F^
F')
and
\(V'^V')/a.
At the same instant a
negative "image"
wave starts from the section distant 21 from the
junction
in the fictitious
continuation of the
rod;
the
velocity
and extension in this
"image"
wave
are
J (F'^F')
and
^{V'^V')ja.
After a time 21'
/a
from the instant of
impact
the
junction
end becomes free from
pressure,
and a rear surface of the
actual wave is formed.
Hence,
the rod
being regarded
as continued in-
definitely,
the wave of
compression
and the
"image"
wave of extension are
both of
length
21'.
Immediately
after the instant
2l'/a
the
junction
end
becomes unstrained and takes zero
velocity.
Hence,
superposing
the
original
velocity
F,
we see that this end takes
actually
the
velocity
V,
so that the
junction
ends of the two rods remain in contact but without
pressure.
The state of the
longer
rod I between the instants
2l'/a
and
2l/a
is
determined
by superposing
the waves of
length
21',
which started out at the
instant of
impact
from the
junction
end and the section distant 21 from it in
the fictitious continuation of the rod. After a time
greater
than
l/a
these
waves are
superposed
over a finite
length
of the
rod,
terminated at the free
end,
and this
part
becomes unstrained and takes a
velocity
Vi^V,
the
velocity

F
being supposed,
as
before,
to be
impressed
on the
system.
The
state of the rod at the instant
2l/a
in the case where I > 21' is different from
the state at the same instant in the case where I < 21'. li l> 21' the wave
of
compression
has
passed
out of the
rod,
and the wave of extension
occupies
a
length
21' terminated at the
junction.
The strain in this
portion
is
extension
equal
to
^ (
F~
V')la
and the
velocity
in the
portion
is
|(
F~
F'),
the
velocity

F
being impressed
as before. The remainder of the rod is
unstrained and has the
velocity
zero.
Hence,
superposing
the
original
272
420 DYNAMICAL RESISTANCE
[CH.
XX
velocity
V,
we see that a
length
I

21' terminated at the free end has at this
instant the
velocity
V and no
strain,
and the remainder has the
velocity
|(F+ V)
and extension
| {Vo'V')/a.
The wave in the rod is now reflected
at the
junction,
so that it becomes a wave of
compression travelling away
from the
junction,
the
compression
is
^(V'V')/a
and the
velocity
of the
junction
end becomes V. The ends that came into contact have now
exchanged
velocities,
and the rods
separate.
If I < 21' the waves of
compression
and extension
are,
at the instant
2lja,
superposed
over a
length equal
to 21'

I terminated at the free
end,
and the
rest of the rod is
occupied by
the wave of extension. The
velocity
V
being
impressed
as
before,
the
portion
of
length
21'

I terminated at the free end
is unstrained and has the
velocity
Vf^V,
and the
remaining portion
has
extension
^(F~y')/a
and
velocity ^iV'^V').
Hence,
superposing
the
original velocity
F,
we see that a
length
21'

I terminated at the free end

has at the instant the
velocity
V and no
strain,
and the remainder has the
velocity \{V
+
V)
and the extension
\ (
F~
V')la.
The wave is reflected at
the
junction,
as in the other
case,
and the
junction
end takes the
velocity
F'.
In both cases the rods
separate
after an interval
equal
to the time taken
by
a wave of extension to travel over twice the
length
of the
longer
rod.
The shorter rod takes the
original velocity
of the
longer,
and rebounds
without strain
;
while the
longer
rebounds in a state of vibration. The
centres of mass of the two rods move after
impact
in the same
way
as if
there were a
"
coefficient of restitution
"
equal
to the ratio V : I.
285. Problems of
dynamical
resistance
involving
transverse
vibration.
The results obtained in Articles 281

284 illustrate the

general
character of
dynamical
re8istance.s. Similar methods to those used in these Articles cannot be
emplo3'ed
in
problems
that involve transverse vibration for lack of a
general
functional solution of the
equation (6)
of Article 280*. In such
problems
the best
procedure
seems to be to
express
the
displacement
as the sum of a series of normal
functions,
and to
adjust
the
constant coefficients of the terms of the series so as to
satisfy
the initial conditions. For
examples
of the
application
of this method reference
may
be made to Lord
Rayleight
and Saint-
Venantt.
A
simplified
method of
obtaining
an
approximate
solution can sometimes be
employed.
For
example, suppose
that the
problem
is that of a rod
"
supported
"
at both ends and
struck
by
a massive
body moving
with a
given velocity.
After the
impact
let the
striking
body
become attached to the rod. At
any
instant after the instant of
impact
we
may,
for an
approximation, regard
the rod as at rest and bent
by
a certain transverse load
applied
at the
point
of
impact.
It will
have,
at the
point,
a certain
deflexion,
which is
determined in terms of the load
by
the result of Article 247
(d).
The load is
equal
to the
pressure
between the rod and the
striking body,
and the deflexion of the rod at the
point
*
Fourier's solution
by
means of definite
integrals, given
in the Bulletin des Sciences & la
Societi
philomatique,
1818,
(cf.
Lord
Eayleigh, Theory of Sound,
vol.
1, 192),
is
applied
to
problems
of
dynamical
resistance
by
J.
Boussinesq, Applications
des
Potentiels, pp.
456 et
seq.
t
Theory of Sound,
vol.
1,
168.
t
See the 'Annotated
Clebsch,'
Note du

61.
284-286]
WHIRLING OF SHAFTS 421
of
impact
is
equal
to the
displacement
of the
striking body
from its
position
at the instant
of
impact.
The
equation
of motion of the
striking body, supix)sed subjected
to a force
equal
and
opi,)osite
to this transverse
load,
combined with the conditions
that,
at the
instant of
impact,
the
body
has the
prescribed velocity,
and is
instantaneously
at the
point
of
impact,
are sufficient conditions to determine the
displacement
of the
striking
body
and the
pressure
between it and the rod at
any subsequent
instant. In this
method,
sometimes described as Cox's
method*,
the deflexion of the rod
by
the
striking body
is
regarded
as a statical
efifect,
and thus this method is in a sen.se an
anticipation
of Hertz's
theory
of
impact (Article 139).
It has
already
been
pointed
out that a similar method
was used also
by
Willis and Stokes in their treatment of the
problem
of the
travelling
loadt.
A somewhat similar method has been
employed by
Lord
RayleighJ
for an
approximate
determination of the
frequency
of the
gravest
mode of transverse vibration of a rod. He
set out from a
general
theorem to the effect that the
frequency
of
any dynamical system,
that would be found
by assuming
the
displacement
to be of a
specified type,
cannot be le-ss
than the
frequency
of the
gravest
mode of vibration of the
system.
For a rod
clami)ed
at
one end and free at the
other,
he showed that a
good approximation
to the
frequency may
be made
by assuming
the
displacement
of the rod to be of the same
type
as if it were
deflected
statically by
a transverse
load,
concentrated at a distance from the free end
equal
to one
quarter
of the
length.
This method has
recently
been the
subject
of some
discussion
.
It has been shown to be
ajiplicable
to the determination of the
frequency
of
the
gravest
mode of transverse vibration of a rod of variable cross-section
||.
It has been
shown also that a method of successive
approximation
to the various normal fxmctions for
such a
rod,
and their
frequencies,
can be founded
upon
such solutions as Lord
Rayleigh's
when these solutions are
regarded
as first
approximationslT.
286. The
whirling
of shafts**.
A
Jong
shaft
rotating
between
bearings
remains
straight
at low
speeds,
but when the
speed
is
high enough
the shaft can rotate
steadily
in a form in
which the central-line is bent. The shaft is then said to
"
whirl." Let u be
the transverse
displacement
of a
point
on the
central-line,
il the
angular
velocity
with which the shaft rotates. The
equation
of
motion,
formed in
the same
way
as
equation (6)
in Article
280,
is
pg-,n..= -^^'^|^
(26)
and the solution of this
equation
must be
adjusted
to
satisfy appropriate
*
H.
Cox, Cambridge
Phil. Soc.
Trans.,
vol. 9
(1850).
Cf. Todhunter and Pearson's
Hittory,
vol.
I,
Article 1435.
t
See
Introduction, p.
26.
J Theory of Sound,
vol.
1,
182.

C. A. B.
Garrett,
Phil.
Mag. (Ser. 6),
vol. 8
(1904),
and C.
Chree,
Phil.
Mag. (Ser. 6),
vol. 9
(1905).
II
J.
Morrow,
Phil.
Mag. (Ser. 6),
vol. 10
(1905).
Some
special
cases of the vibrations of a
rod of variable
section,
in which the exact forms of the normal functions can be determined
in terms of Bessel's
functions,
were discussed
by Kirchhoff,
Berlin
Monatsberichte, 1879,
or Ge$.
Abhandlungen, p.
339.
IT A.
Davidoglou,
'
Sur
I'^quation
des vibrations transversales des
verges ^lastiques,'
Paris
{These),
1900.
**
Cf. A. G.
Greenhill,
Proc. Inst. Mech.
Engineers,
1883.
422
WHIRLING OF SHAFTS
[CH.
XX
conditions at the ends of the shaft. We shall consider the case in which the
ends s
=
and s
=
l are
"
supported."
If the rod rotates
steadily d''u/dt'
vanishes,
and the
equation
is the same as that for a rod
executing simple
harmonic vibrations of
period 27r/n.
In order that the
equation
ds*
.(27)
may
have a solution which makes u and
dhijds'^
vanish at s
=
and at s
=
l,
the
speed
of rotation H must be such that
n/27r
is
equal
to the
frequency
of
a normal mode of flexural vibration of the
doubly-supported
shaft. Thus the
lowest
speed
at which
whirling
takes
place
is such that
fl/2'7r
is
equal
to the
frequency
of the
gravest
mode of flexural vibration of such a shaft. If we
write
the
possible
values of m are
given by
the
equation
sin ml
=
0,
and the
smallest value of H for which
whirling
can take
place
is
The
whirling
of unloaded shafts
rotating
under various terminal conditions has been
dealt with
by
A. G. Greenhill
(loc.
cit).
The
important
technical
problem
of a shaft
carrying loads, pulleys
for
example,
has been discussed
theoretically
and
experimentally by
S.
Dunkerley*.
He found that the direct
application
of the method of normal
functions,
illustrated
above,
led to
very complicated results,
and
proposed
to fall back on an
empirical assumption.
The
subject
has been discussed further
by
C.
Chreet, by
the aid
of Lord
Rayleigh's
statical method of
determining frequencies (Article 285).
Phil. Tram. R. Soc.
(Ser. A),
vol. 185
(1894).
+
Phil.
Mag. (Ser. 6),
vol. 7
(1904).
CHAPTER XXI.
SMALL DEFORMATION OF NATURALLY CURVED RODS.
287. In the
investigations
of
Chapters
xviii. and xix. we have
given
prominence
to the consideration of modes of deformation of a thin rod which
involve
large displacements
of the central-line and twist that is not
small,
and we have
regarded
cases in which the
displacement
of the central-line and
the twist are small as
linaiting
cases. This was the method
followed,
for
example,
in the
theory
of
spiral springs (Article 271).
In such cases the
formulae for the
components
of curvature and twist
may
be
calculated,
as has
been
explained, by treating
the central-line as unextended. We can
give
a
systematic
account of such modes of deformation as involve small
displace-
ments
only by introducing quantities
to denote the
components
of the dis-
placement
of
points
on the
central-line,
and
subjecting
these
quantities
to
a condition which
expresses
that the central-line is not extended*.
288.
Specification
of the
displacement.
The small deformation of
naturally straight
rods has been
sufficiently
investigated already,
and we shall therefore
suppose
that,
in the unstressed
state,
the rod lias curvature and twist. As in Article
2.59,
we shall use
a
system
of axes of
a;,
y^,
z,,,
the
origin
of which moves
along
the unstrained
central-line with unit
velocity,
the axis of
z
being always
directed
along
the
tangent
to this
line,
and the axes of
Xf,
and
y^ being
directed
along
the
principal
axes of the cross-sections at their centroids. We have denoted
by
^
TT

/o
the
angle
which the axis of
a;
at
any point
makes with the
principal
normal of the unstrained central-line at the
point,
and
by
k^, Kq, Tq
the com-
ponents
of initial curvature and the initial twist. We have the formula
'^0 /'^o
=

tanyo-
*
The
theory
was
partially
worked out
by
Saint-Venant in a series of
papers
in Paris C.
R.,
t. 17
(1843),
and more
fully by
J. H.
Michell, Messenger of Math.,
vol. 19
(1890).
The latter has
also obtained some exact solutions of the
equations
of
equilibrium
of an elastic solid
body
bounded
by
an
incomplete tore,
and these solutions are
confirmatory
of the
theory
when the tore is thin.
See London Math. Soc.
Proc,
vol. 31
(1900), p.
130.
424
KINEMATICS OF SMALL DEFORMATION
[CH.
XXI
The curvature
1/po
and the
tortuosity 1/So
of the central-line are
given by
the formulae
(i/p)^
=
Af +
0, i/s
=
T-d/;/ds,
in which s denotes the arc of the central-line measured from some chosen
point
of it.
When the rod is
slightly
deformed,
any particle
of the central-line under-
goes
a small
displacement,
the
components
of
which,
referred to the axes of
o.
Vo,
2'o,
with
origin
at the unstrained
position
P of the
particle,
will be
denoted
by
u, v,
w. The rod will receive a new curvature and
twist, defined,
as in Articles 252 and
259,
by
means of a
moving system
of
"principal
torsion-flexure axes." We recall the conventions that the axis of z in this
system
is directed
along
the
tangent
of the strained central-line at the
point
Pi
to which P is
displaced,
and that the
plane
of
(x, z)
is the
tangent plane
at
Pi
of the surface made
up
of the
aggregate
of
particles
which,
in the
unstressed
state,
lie in the
plane
of
{x^, Zo)
at P. We have denoted the
components
of curvature and the twist of the strained central-line at
Pi
by
i, i'. Tj.
When the
displacement
(m,
v,
w)
of
any point
of the central-line is
known,
the
tangent
of the strained central-line at
any point
is
known,
and it
is clear that one additional
quantity
will suffice to determine the orientation
of the axes of
{x, y, z)
at
Pi
relative to the axes of
(a;, y,
z^)
at P. We shall
take this
quantity
to be the small
angle
/3
which the
plane
of
{x, z)
makes
with the
plane
of
(, Za)-
The relative orientation of the two sets of axes
may
be determined
by
the
orthogonal
scheme of transformation
288-290]
OF A THIN CURVED ROD
425
lira
(I
-
u)/Bs
are the direction cosines
L,
Let
(', v',
w')
be the dis-
4s
=
placement
of F referred to the axes of
x^.y^,
z
at
P',
and
(f/', V, W)
the
same
displacement
referred to the axes of
x^^,
y,
z^
at P. Then
il V. D
=
(8^0
+
U',
Sy
+
V, Bz,
+
W).
The limits of
Bx^lBs, Si/o/Bs, Bzo/Ss
are
0, 0,
1. The limits of
(!*'-)/&,
...
are
dujds,
... and we have the usual formulae connected with
moving
axes in
such forms as
.. U'

u du
,
urn
s
=
T VTo
+
WK^
.
s=o
OS ds
Hence we obtain the
equations
T
du
,
,
tr
dv
^T ^
dw
, ..
L3=-^-vro-\-'WKD,
-'3=j

WKo+ur^, iVs=l
+
-j uk^ +vko.
...{2)
The
equation
L^"
+
M^
+
N^
=
1
leads,
when we
neglect squares
and
products
of
u, V, w,
to the
equation
dw
,
-
,.
T UK +VKo=0,
(3)
which
expresses
the condition that the central-line is unextended. In conse-
quence
of this
equation
we have
^3=
1.
The direction cosines of the axes of
x,
jr
at
P,,
referred to the axes of
^0,
2/0.
Zo
at
P,
are determined
by
the conditions that the
plane
of
(x.z)
makes
a small
angle
/3
with the
plane
of
(a;, z^),
and that the scheme of transforma-
tion
(1)
is
orthogonal
and its determinant is 1. These conditions
give
us
, (4)
These
equations might
be found otherwise from the formulae
(7)
of Article
253
by. writing
Zj,
... instead of
k,
...
,
taking
6 to be
small,
and
putting
y8
for
<p
+
-\{r. They
are,
of
course,
correct to the first order in the small
quantities v, v, w, /8.
290. Curvature and twist.
For the calculation of the
components
of curvature and the twist we have
the formulae
(6)
of Article
253,
in which
k^,...
are written for
/c,
.... In
those formulae
l^,
... denoted direction cosines of the axes oi
x,
y,
z referred to
fixed axes. Here we have taken
L^,
... to denote the direction cosines of the
axes of
X,
y,
z at
Pi
referred to the axes of
x^,
y^,
Zo
at P. If P' is a
point
near to
P,
so that the arc PP'
=
Bs,
and
P/
is the
displaced position
of
P',
we
may
denote
by
Z,',
. . . the direction cosines of the axes of
x,
y,
z at
P/
referred
to the axes of
x^, , z^
at
P',
and then the limits such as lim
(Z//

Zi)/Ss
are
4=0
426 CURVED ROD SLIGHTLY DEFORMED
[CH.
XXI
the differential coefficients such as
dLj/ds.
Let the fixed axes of reference
for
li,
... be the axes of
w^,
y,
z
at
P,
and let
l^
+
U-i,
... denote the direction
cosines of the axes of
x,
y,
z at
P,'
referred to these fixed axes. Then the
limits such as lim
hl^lis
are the differential coeflScients such as
dl^jds.
It is
=o
clear
that,
at
P, li
=
Li,
... but that
dli/ds 4 dLJds,
.... We have in fact the
usual formulae connected with
moving
axes,
viz. :

dli/ds
=
dL^/ds

M^To
+
N^Ko',
dmjds
=
dMi/ds

N^k^
+
L^t^,
duj/ds

dNJds

L^k^
+
M^k^,
with similar formulae for
dljds,
... and
dl^jds,
....
In the formulae
(6)
of Article 253 we write
k^,
... for
k, ...,
put
W3
=
iV^3=l,
replace
^,
...
by
the values found for
L^,
... in
(2)
and
(4),
and substitute the
values
just
found for
dl^jds, Rejecting
terms of the second order in the
small
quantities
u, v, w, /3,
we obtain the
equations
dM,
fCi

fCfj
-T
pKfj ToZ-,,
Ti
=
To
+
T-
+
/C0X/3
+
K0M3,
in which
L,
and
M^
are
given by
the first two of
equations (2).
.(5)
291.
Simplified
formulae.
The formulae are
simplified
in the case where
f^
=
-J-tt.
In this case the
axis of
,
which is a
principal
axis of a cross-section at a
point
of the
unstrained
central-line,
coincides with the
principal
normal of this curve at
the
point.
When this is the case we have
Ko
=
0, K^
=
1/po,
To
=
l/2o.
r
du V W
i
i
ds
Po
1
2,
d
ds
d
P
dv u
ds
2c
T,r
dv
N.= l,
_. /du
Po
ds
\ds
V w
2
p,
dv
1
fdu
V w
2o \ds 2o
Po.
dv u
"^
2
r
=1
^
liiii
'
2o
ds
po\ds
2,
J 2ol
2J-
The condition that the central-line is unextended is
dw
ds
u
Po'
.(6)
(7)
290-292]
BENDING OF CIRCULAR RING 427
The measures of curvature and
tortuosity
and the direction cosines of the
principal
normal and binormal can be calculated from these formulae or from
the more
general
formulae of Article 290.
292. Problems of
equilibrium.
The
theory
is
applicable
to such
problems
as the deformation of the links
of chains*
by
the
pressure
of
adjacent
links,
and it
may
be used also to
give
an account of the behaviour of arches
f,
the link or the arch
being
treated
as a thin curved rod. The
equations
of
equilibrium
have been
given
in
Article
259,
and we have found in
preceding
Articles of this
chapter expres-
sions for all the
quantities
that occur in terms of the
displacement (m,
v,
w)
and the
angular displacement
/S,
the
quantities
u, v,
w
being
themselves con-
nected
by
an
equation (3)
or
(7). Naturally any special problem,
such as
those
mentioned,
is of a
very
technical
character,
and we shall content our-
selves here with a
slight study
of some cases of the
bending
of a rod in the
form of an
incomplete
circular
ring.
(a) Incomplete
circular
ring
bent in its
plane.
9
Let the unstrained central-line be a circle of radius
a,
and let be the
angle
between the radius drawn from the centre of the circle to
any point
on
it and a chosen
radius,
then
po

ds/dO
=
a.
The
displacement
u is directed
along
the radius drawn
inwards,
and the
displacement
w is directed
along
the
tangent
of the circle in the sense in
which increases. We shall
suppose
that the
plane
of the circle is a
principal plane
of the rod at
any point,
and that the flexural
rigidity
for
bending
in this
plane
is B. Then
v, /S
and
l/S^
vanish,
and the condition
that the central-line is unextended is
dw ..
d^="
^>
The flexural
couple
G' in the
plane
of the circle is
n.^Bfdhvdw\_
^
aAdO^ ^dO)'
^^^
the other flexural
couple
and the torsional
couple
vanish.
Let the rod be bent
by
forces
having components
X,
Z
per
unit of
length
*
E.
Winkler,
Der
Civilingenieur,
Bd. 4
(1858).
Winkler's memoir is described at
length
and corrected in detail in Todhunter and Pearson's
History,
vol.
2,
pp.
422 et
seq.
+
M.
Bresse,
Recherches
analytiques
sur la
flexion
et la resistance des
pieces
courbes,
Paris 1854.
An account of this treatise also is
given
in Todhunter and Pearson's
History,
vol.
2, pp.
352
et
seq.
H. T.
Eddy,
Amer. J.
of Math.,
vol. 1
(1878),
has
proposed
a
graphical
method of treat-
ment of the
problem
of arches.
428 BENDING OF INCOMPLETE
[CH.
XXI
directed
along
the radius and
tangent
at
any point.
The
equations
of
equilibrium
obtained from
(26)
and
(27)
of Article 259 are
di\r
^
dr --
,^
dG'
^
: + r+Ztt
=
0, -T?i-N
+Za
=
0, -y.+Na
=
0.
.(10)
AT
-
_
( )
a'
[dd*

d(fi)
d'w d*w d'^w
=
(S-)-
.(11)
.(12)
Fig.
64.
de
' '
'
dO
'
'
d0
Hence we find that the
shearing
force JV" and the tension T are
expressed
in terms of w
by
the
equations
and that w satisfies the
equation*
-(
^^
a
U^'
de*
We note the
following
results :

(i)
When the rod is
slightly
bent
by couples equal
to K
applied
at its ends in its
plane,
the central-line remains
circular,
but its radius is
reduced
by
the fraction
Ka/B
of itself.
(ii)
When the ends of the rod are
given by
5=
a,
so
that the line
joining
them subtends an
angle
2a at the
centfe,
and the rod is
slightly
bent
by
forces
equal
to R
acting
as tension
along
this line as in
Fig. 64,
the
displace-
ment is
given by
the
equations
w=
-
{ofllilB)
6
(cos a+i
cos
6), u=dw/de.
(ill)
When the rod is
slightly
bent
by
forces
equal
to
S, applied
as shown in
Fig.
65
to
rigid pieces
attached to its ends and
extending
across
the chord of the
incomplete ring,
the
displacement
is
given by
the
equations
w
=
-
J (a^S/B)
e sin
6,
u
=
dw/dd.
(iv)
When the rod forms a
complete
circular
ring,
and
is
slightly
bent
by
normal
pressures equal
to
Xi applied
at the
opposite
ends of a
diameter,
we measure 6 from
this diameter as shown in
Fig. 66,
and find for the dis-
placement
w a,t a.
point
on that side of this diameter in which 7r>d>0
w
=
-
Xj (a3/fi) [e/n-i(l-coa 6-^0 81116)1
u
=
dw/d6.
The
displacements
are
clearly
the same at
any
two
points symmetrically
situated on
opposite
sides of this
diameter.
We
may
deduce the value of u at
any point,
and we
may prove
that the diameter which coincides with the
line of thrust is shortened
by {{7r^

8)/4TT}{A\a^/B),
while the
perpendicular
diameter is
lengthened by
{(4-,r)/2-}(Xia3/5)t.
(v)
When the rod forms a
complete
circular
ring
of
weight W,
which is
suspended
from a
point
in its cir-
cumference,
we measure 6 from the
highest point,
and
Fig.
65.
Fig.
66.

Cf. H.
Lamb,
London Math. Sac.
Proc,
vol. 19
(1888), p.
365.
text under the numbers
(i)

(v)
are taken from this
paper.
+
These results are due to Saint-
Venant,
Paris C.
R.,
t. 17
(1843).
The results
given
in the
292]
CIRCULAR RING
429
find for the
displacement
w a.t a,
point
for wliich
n>6>0 the value
w=-
W{a^lB) (Stt)-' {((9
-
ff
)2
sin 5
-
4
(5
-
ff)(l
-
cos
6)
-
n^ sin
6}
;
the
displacement
is the same at the
corresponding point
in the other half of the
ring.
In this case we
may prove
that the amounts
by
which the vertical diameter is
lengthened
and the horizontal diameter shortened are the halves of what
they
would be if
the
weight
Wwere concentrated at the lowest
point.
(vi)
When the rod forms a
complete
circular
ring
which rotates with
angular velocity
a about one
diameter*,
taken as axis of
y,
its central-line describes a surface of revolution
of which the meridian curve is
given by
the
equations
y
=
o cos fl
+
^j (mw^a^/B) (1
-
cos'^),
where m denotes the mass of the
ring per
unit of
length,
and 6 is measured from the
diameter about which the
ring
rotates. This diameter is shortened and the
perpendicular
diameter
lengthened by
the same amount
J {ma>^a^/B).
(b) Incomplete
circular
ring
bent out
of
its
plane.
As before we take a for the radius of the
circle,
and
specify
a
point
on it
by
an
angle
;
and we take the
plane ^ ^ .^
..
of the circle to be that
principal plane
/
'J^-'e
/'
yi'
of the rod for which the flexural
rigidity
,'^''
V''
^r
is B. We consider the case where the
/' \
^>^
rod is bent
by
a load
W,
applied
at the
/'
\_^^
end
=
a in a direction at
right angles
/
^^^
to this
planef,
and is fixed at the end /
^^_^,,,,^'''^^
=
0,
so that the
tangent
at this
point
f
is fixed in
direction,
and the transverse
linear element
which,
in the unstressed
^
state,
is directed towards the centre of
the circle is also fixed in direction. Then
m, v, w, /S,
dujdd, dvjdO
vanish
with 6.
The stress-resultants
N, N',
T at
any
section are
statically equivalent
to
the force
W,
of which the direction is
parallel
to that of the axis of
y^
at
any
section,
and we
have, therefore,
N
=
^W, N'=W,
T=:iW/a).{dv/d9) (13)
The
equations
of moments
are, therefore,
+^='^^' f
=-^^'
S-^=
(1^)
From the first and third of
these,
combined with the conditions that
G and H vanish when 6
=
a,
we find
G==-aW
sm{a-0),
H
=
aW
{1-
cos
(a-
0}] (15)
*
G. A. V.
Peschka,
Zeitschr.
f.
Math. u.
Phys. (SchlSmilch),
Bd. 13
(1868).
t
The
problem
has been discussed
by
Saint-
Venant,
Parit C.
R.,
t. 17
(1843),
and
by
H.
Eesal,
J. de Math.
{LiouvilU), (S6r. 3),
t. 3
(1877).
430 VIBRATING CIRCULAR RING
[CH.
XXI
Now we have
G
=
-
A
/d^
a'
W ^)'
^=a^<f^(^
+
^)' .(16)
.(17)
and from these
equations
and the terminal conditions at ^
=
we can obtain
the
equations
Wa?
v
+
aff=
-jr
{^

sin a
+ sin
(a

6)\
,
V
=
f^ {(Q

sin
Q)

sin o
(1

cos
&)}
+
^Wa^
('^+ ^
{(?cos (a
-
6')
-
sin e cos
a}.
We
may prove
also that w and w are small of the order v*.
293. Vibrations of a circular
ring.
We shall illustrate the
application
of the
theory
to vibrations
by
con-
sidering
the free vibrations of a rod
which,
in the unstressed
state,
forms
a circular
ring
or a
portion
of such a
ring,
and we shall restrict our work to
the case where the cross-section of the
ring
also is circular. We denote the
radius of the cross-section
by
c,
and that of the circle formed
by
the central-
line
by
a,
and we take the
displacement
u to be directed
along
the radius
drawn towards the centre of the latter circle. The
equations
of
motion,
formed as in Articles 278

280,
are
.(18)
and
^^
, IT AT' 12
^^
dH
dd
-G:
1 2
S^/S
hernia
^^,
.(19)
c* /dv
in which m is the mass of the
ring per
unit of
length,
and
(20)
E
being
the
Young's
modulus and
/i.
the
rigidity
of the material of the
ring.
The above
equations
with the condition
dw
dd
=
u
.(8 bis)
yield
the
equations
of motion.
292,
293]
VIBRATING CIRCULAR RING 431
It is clear that the above
system
of
equations
falls into two sets. In the
first set V and
/8
vanish,
and the motion is
specified by
the
displacement
u or
w,
these variables
being
connected
by equation (8);
in this case we have
flexural vibrations of the
ring
in its
plane.
In the second set u and w
vanish,
and the motion is
specified by
v or
/8,
so that we have flexural
vibrations
involving
both
displacement
at
right angles
to the
plane
of the
ring
and twist.
It
may
be shown in the same
way
that the vibrations of a
curved rod fall
into two such classes whenever the central-line of the
unstressed rod is a
plane
curve,
and its
plane
is a
principal plane
of the rod at each
point.
In
case the central-line is a curve of double curvature there is no such
separa-
tion of the modes of vibration into two
classes,
and the
problem
becomes
extremely complicated*.
(a)
Fleocural vibrations in the
plane of
the
ring.
We shall
simplify
the
question by neglecting
the
"rotatory
inertia."
This amounts to
omitting
the
right-hand
member of the second of
equations
(19).
We have then
J
Eire*
(d'w
d*w d-w\ d'
f dhv\
The normal functions for free vibration are determined
by taking
w to be
of the form W cos
{pt
+
e),
where W is a function of ff. We- then have the
equation
d^W d*W d'W /
4'may
\
A,may
dd"
^
W
"^
d0-'
V
Em* )
"^
'Em'~
^
~
"
The
complete primitive
is of the form
TF
=
2
(A^
cos
n^O
+
B^
sin
n^9),
where
n^, n^, n,
are the roots of the
equation
n^
(n -\f
=
{n^
+
\) (4>may/E-rrC*).
If the
ring
is
complete
n must be an
integer,
and there are
vibrations
with n
wave-lengths
to the
circumference,
n
being any integer greater
than
unity.
The
frequency
is then
given by
the
equation f
^ _
Eire'
nHn-'-iy
^
~
ima* n^^"
^ ^^
*
The vibrations of a rod of which the natural form is helical have been
investigated by
J. H.
Michell,
loc. cit.
p. 423,
and also
by
the
present writer, Cambridge
Phil. Soc.
Traits.,
vol. 18
(1899).
t The result is due to E.
Hoppe,
J.
f.
Math.
(Crelle),
Bd. 73
(1871).
432
VIBRATING CIRCULAR RING
[CH.
XXI
When the
ring
is
incomplete
the
frequency equation
is to be obtained
by
forming
the conditions that
N, T,
0' vanish at the ends. The result is diffi-
cult to
interpret except
in the case where the initial curvature is
very slight,
or the radius of the central-line is
large compared
with its
length.
The
pitch
is then
slightly
lower than for a
straight
bar of the same
length,
material
and cross-section*.
(6)
Flexural vibrations at
right angles
to the
plane of
the
ring.
We shall
simplify
the
problem by neglecting
the
"
rotatory
inertia,"
that
is to
say
we shall omit the
right-hand
members of the first and third of
equations (19);
we shall also
suppose
that the
ring
is
complete.
We
may
then write
i>
=
Fcos
{nO
-f-
a)
cos
{pt
+
e), /S
=
B' cos
{n6
+
a)
cos
{jpt
+
e),
where
F, B', a,
e are
constants,
and n is an
integer.
From the first and third
of
equations
(19)
and the second of
equations (18)
we find the
equations
n'
(aB'
+
n'V)
+
^
n'
(aB' +V)=
^^^
V,
^nUaB'+V)
+
(aB'
+
n'V)
=
0,
from which we obtain the
frequency equation f
ETTC^nHn^-iy
^
4:m.a* iv'
+
1
+
a
'
^^
where a is Poisson's ratio for the
material,
and we have used the relation
E=2jj,{l+ a-).
It is
noteworthy
that,
even in the
gravest
mode
(n
=
2),
the
frequency
differs
extremely
little from that
given by equation (21)
for the
corresponding
mode
involving
flexure in the
plane
of the
ring.
(c)
Torsional and extensional vibrations.
A curved rod
possesses
also modes of free vibration
analogous
to the torsional and
extensional vibrations of a
straight
rod. For the torsional vibrations of a circular
ring
we
take u and w to
vanish,
and
suppose
that v is small in
comparison
with
aj3,
then the
second of
equations (18)
and the first of
equations (19)
are satisfied
approximately,
and the
third of
equations (19)
becomes
approximately
For a
complete
circular
ring
there are vibrations of this
type
with n
wave-lengths
to the
circumference,
and the
frequency p/2w
is
given by
the
equation
jo2='"''^(l-ho-^2)
(23) ^
ma'
^
When n
=
0,
the
equations
of motion can be satisfied
exactly by putting
v=0 and
taking
3
to be
independent
of 6. The characteristic feature of this mode of vibration is that each

The
question
has been discussed
very fully by
H.
Lamb,
loc. cit.
p.
428.
f
The result is due to J. H.
Michell,
loc. cit.
p.
423.
1
293]
VIBRATING CIRCULAR RING 433
circular cross-section of the circular
ring
is turned in its own
plane through
the same
small
angle
fi
about the
central-line,
while this line is not
displaced*.
For the extensional modes of vibration of a circular
ring
we take v and
/3
to
vanish,
and
suppose
that
equation (8)
does not hold. Then the extension of the central-line is
a~^
(dw/dd-u),
and the tension T is
E7Tc''a~^(dwld6-u).
The
couples O,
If and the
shearing
force N' vanish. The
expressions
for the
couple
G' and the
shearing
force N
contain c* as a
factor,
while the
expression
for T contains c^ as a factor. We
may,
there-
fore,
for an
approximation,
omit O' and
N,
and
neglect
the
rotatory
inertia which
gives
rise to the
right-hand
member of the second of
equations (19).
The
equations
to be
satisfied
by
u and w are then the first and third of
equations (18),
viz. :
The
displacement
in free vibrations of
frequency p/Zn
is
given by equations
of the form
u
=
{A
sin n6 +
B COS
nd)
COS
(pt
+
t),
w
=
n(A
cos nd

Bam
n6)
cos
(pt+t),
where
iB
=
^(l-fn2)
(24)
When n
=
0,
w vanishes and u is
independent
of
6,
and the
equations
of motion are
satisfied
exactly.
The
ring
vibrates
radially,
so that the central-line forms a circle of
periodically
variable
radius,
and the cross-sections move without rotation.
The modes of vibration considered in
(c)
of this Article are of much
higher pitch
than
those considered in
(a)
and
(6),
and
they
would
probably
be difficult to excite.
*
The result that the modes of vibration
involving displacements
v and
j3
are of two
types
was
recognized by
A. B.
Basset,
London Math. Soc.
Proc,
vol. 23
(1892),
and
the frequency
of the
torsional vibrations was found
by
him.
L. E.
28
CHAPTER XXII.
THE STRETCHING AND BENDING OF PLATES.
294.
Specification
of stress in a
plate.
The internal actions between the
parts
of a thin
plate
are most
appro-
priately expressed
in terms of stress-resultants and
stress-couples
reckoned
across the whole thickness. We take the
plate
to be of thickness
2h,
and on
the
plane midway
between the
faces,
called the
"
middle
plane,"
we choose an
origin
and
rectangular
axes of x and
y,
and we draw the axis of z at
right
angles
to this
plane
so that the axes of
x,
y,
z are a
right-handed system.
We draw
any cylindrical
surface G to cut the middle
plane
in a curve s.
The
edge
of the
plate
is such a surface as
G,
and the
corresponding
curve is
the
"
edge-line."
We draw the normal v to s in a chosen
sense,
and choose
the sense of .s so that
v, s,
z are
parallel
to the directions of a
right-handed
system
of axes. We consider the action exerted
by
the
part
of the
plate lying
on that side of G towards which v is drawn
upon
the
part lying
on the other
side. Let & be a short
length
of the curve
s,
and let two
generating
lines of
G be drawn
through
the extremities of hs to mark out on G an area A. The
tractions on the area A are
statically equivalent
to a force at the centroid of A
and a
couple.
We
resolve
this force and
couple
into
components
directed
along
1/, s,
z. Let
\T\ \S\ [iV]
denote the
components
of the
force,
\R\ \G\ {K\
those of the
couple.
When 8s is diminished
indefinitely
these
quantities
have zero
limits,
and the limit of
[^]/Ss
also is
zero,
but
[^J/Ss,
...
[G]/Ss,
may
be finite. We denote the limits of
[TJ/Ss,
...
by
7,
.... Then
T, S,
N
are the
components
of the stress-resultant
belonging
to the line
s,
and
H,
G
are the
components
of the
stress-couple belonging
to the same line. T is a
tension,
S and N are
shearing
forces
tangential
and normal to the middle
plane,
G is a flexural
couple,
and H a torsional
couple.
When the normal v
to s is
parallel
to the axis of
x,
s is
parallel
to the axis of
y.
In this case we
give
a suffix 1 to
T,
When the normal v is
parallel
to the axis of
y,
s is
parallel
to the
negative
direction of the axis of x. In this case we
give
a
suffix 2 to
T,
The conventions in
regard
to the senses of these forces
and
couples
are illustrated in
Fig.
68.
294,
295]
STRESS-RESULTANTS AND
STREBS-COUPLES
435
are
For the
expression
of
T,
... we take
temporary
axes of
x',
y',
z which
parallel
to the
directions of
v, s, z,
and
denote
by
X'^,
... the
stress-com
ponents
referred to these axes. Then we have the
formuls*
1.1
*T,
T^S,
O.
T,
Fig.
68.
T=f
X'^dz,
S=r
X'y.dz,
N^r
X\dz,
!
-h J -h J
-h
H=r
-zX'ydz,
0=r
zX'^dz;
!
-h J
-h
and,
in the two
particular
cases in which v is
parallel respectively
to the axes
of X and
y,
these formulae become
r,=
f*
X^dz, S,=
r
Xydz,
N,=
r
X,dz^
Hi=
I

zXy
dz, Oi
=
I
zXx dz,
and
S,=
r
-Xydz,
T,=
r
Yydz, N.^!"
Y,dz]
J
-h J -h J
-h
\
rh rh
0^=1
zYydz,
H:i=
I
zXydz.
J
-h J -h
We observe that in accordance with these formulae
82= 8i, Hi
=

Hj
(2)
.(3)
295. Transformation of stress-resultants and
stress-couples.
When the normal v to the curve s makes
angles
6 and
^tt

O with the
axes of X and
y,
T, S,
... are to be calculated from such formulae as
T=r
x'^dz,
J -h
in which the
stress-components
X'-^,
... are to be found from the formulae
(9)
of Article 49
by putting
li
=
cos6, nil
=
sin
0, 4= sinfl, m5
=
cos^, Mi
=
-j
=
Z3
=
7%
=
0, n,
=
\.
It is assumed that the
plate
is but
slightly
bent. Cf. Article 328 in
Chapter
xxrv.
282
436
EQUATIONS
OF
EQUILIBRIUM
[CH.
XXII
...(6)
We find r
=
r,
cos"
d+T^
sin 6
+
S^
sin
26,
\
S
=
\{T^
-
T;)
sm '26 -ir
8^ co%2e,
N
=
N,cos6
+
N^sin6,
I
(4)
G
=
G,
cos' 6 +
G^
sin=
6-H^
sin
26,
H
=
^{G,-G^)
sin 20 +
H,
cos 20.
,
Instead of
resolving
the stress-resultants and
stress-couples belonging
to
the line s in the directions
v, s,
z we
might
resolve them in the directions
X,
y,
z. The
components
of the stress-resultant would be :
parallel
to
,
Tcos^
Ssin0,
or
Tj
cos -1-
(Sj
sin
0,1
parallel
to
2/,
T sin -t- <S cos
0,
or
Tj
sin
+
/Sj
cos
0,
1
(5)
parallel
to
z, N^
cos -f
iV,
sin
;
J
and those of the
couple
would be :
about an axis
parallel
to
x,
Hcos

G sin
0,
or
H^
cos 6

G^
sin
0,'
about an axis
parallel
to
y,
H sind
+
G cos
0,
or
Gi
cos 6
Hi
sin 0.
296.
Equations
of
equilibrium.
Let C
denote,
as
before,
a
cylindrical
surface
cutting
the middle
plane
at
right angles
in a curve
s,
which we take to be a
simple
closed contour. The
external forces
applied
to the
portion
of the
plate
within C
may
consist of
body
forces and of surface tractions on the faces
(z
=
h and ^^
=

^)
of the
plate.
These external forces are
statically equivalent
to a
single
force,
acting
at the centroid P of the volume within
C,
and a
couple.
Let
[X'], [Y'], [Z']
denote the
components
of the force
parallel
to the axes of
x,
y,
z,
and
[L], [M'], [N']
the
components
of the
couple
about the same axes. When
the area to within the curve s is diminished
indefinitely by contracting
s
towards
P,
the limits of
[X'],
...
[L'],
... are zero and the limit of
[iV'J/ct)
also
is
zero,
but the limits of
[X']/(o,
...
may
be finite. We denote the limits of
[X']/o},
...
by
X',
Then
X', Y',
Z' are the
components
of the force-
resultant of the external forces estimated
per
unit of area of the middle
plane,
and
L',
M' are the
components
of the
couple-resultant
of the same
forces estimated in the same
way.
The
body
force
per
unit of mass is
denoted,
as
usual,
by {X,
Y,
Z),
and the
density
of the material
by p.
The definitions of
X', Y', Z', L',
M' are
expressed
analytically by
the formulae
\
Z'=f*
pXdz
+
{X,),^j,-{X,\^
J -h
Y'^r
pFd^
+
(F,),_A-(F,),=_,,
J -h
Z'
=1"
pZdz
+
(Z,%.^
-
{Z,%._H,
(7)
295-297]
and
OF A PLATE
437
.(8)
.(9)
/.'
=
f
-zp
Ydz
-
h
{(
F,),_,
+
(
F,),._,);
M'
=
r
zpXdz
+ h
[{X,\.
+
(Z,),._,).
J h
'
We
equate
to zero the force- and
couple-resultants
of all the forces
acting
on the
portion
of the
plate
within the
cylindrical
surface C. From the
formulae
(5)
we have the
equations
j(Ti
cos ^
+
,Sf,
sin
6)
ds +
Ux'dxdy
=
0,
1{T^
sin
e+S,
cos
6)
ds
+
jJY'dxdy
=
0.
|(iVi
cos e
+
N^
sin
6)
ds
+
(jz'dxdy
=
0,
where the
surface-integrals
are taken over the area within
s,
and the line-
integrals
are taken round this curve. From the formulse
(5)
and
(6)
we have
the
equations
j{{H,
cos
e-G^
sin
e)
+
y (N,
cos
O+N^
sin
0)}
ds
+
jj
(L'
+
yZ') dxdy
=
0,
1
[{G,
cos
e-H^
sin
e)-x {N^
cos O +
N^
sin
6)]
ds
+
jj
(M'
-
xZ') dxdy
=
0,
Ux
(T^
sin e
+
S,
cos
0)-y{T^cos0
+
8^
sin
61)}
ds +
(j{x7'
-
yX') dxdy
=
0.
(10)
Since cos and sin are the direction-cosines of the normal to s referred
to the axes of x and
y,
we
may
transform the
line-integrals
into surface-
integrals.
We thus find from
(9)
three
equations
which hold at
every point
of the middle
plane,
viz.
-
^^'+?^+z'=o
?^'
+ ^^+r
=
o
^'
+
^^
+ z'
=
o.
...(11)
9a;
dy
'
dx
dy
'
dx
dy
We transform the
equations (10)
in the same
way
and
simplify
the results
by using equations (11).
The third
equation
is
identically
satisfied. We
thus find two
equations
which hold at
every point
of the middle
plane,
viz.
dH,_dG^^
,
^Gi_dH,_^ ,^^
dx
dy
dx
dy
Equations (11)
and
(12)
are the
equations
of
equilibrium
of the
plate.
297.
Boundary
conditions.
In a thick
plate subjected
to
given
forces the tractions
specified by
Xy, F, Zy,
where v denotes the normal to the
edge,
have
prescribed
values at
every point
of the
edge.
When the
plate
is
thin,
the actual distribution of
438 BOUNDARY CONDITIONS
[CH.
XXII
the tractions
applied
to the
edge, regarded
as a
cylindrical
surface,
is of no
practical importance.
We
represent
therefore the tractions
applied
to the
edge by
their force- and
couple-resultants,
estimated
per
unit of
length
of the
edge-line,
i.e. the curve in which the
edge
cuts the middle surface. It follows
from Saint-Venant's
principle (Article 89)
that the effects
produced
at a
distance from the
edge by
two
systems
of tractions which
give
rise to the
same force- and
couple-resultants,
estimated as
above,
are
practically
the
same. Let these resultants be
specified by components
T, S,
N and
H,
G in
the senses
previously assigned
for
T, S,
N and
H, G,
the normal to the
edge-
line
being
drawn outwards. Let the sti-ess-resultants and
stress-couples
belonging
to a curve
parallel
to the
edge-line,
and not
very
near to
it,
be
calculated in accordance with the
previously
stated
conventions,
the normal
to this curve
being
drawn towards the
edge-line
;
and let
limiting
values of
these
quantities
be found
by bringing
the
parallel
curve to coincidence with
the
edge-line.
Let these
limiting
values be denoted
by
f, S,
N and
H,
U. It
is most
necessary
to observe that the statical
equivalence
of the
applied
tractions and the stress-resultants and
stress-couples
at the
edge
does not
require
the satisfaction of all the
equations
f=T, ^=S, F=N, H=B.,
=
G:
These five
equations
are
equivalent
to the
boundary
conditions
adopted by
Poisson*. A
system
of four
boundary
conditions was afterwards obtained
by
Kirchhoff
f,
who set out from a
special assumption
as to the nature of the
strain within the
plate,
and
proceeded by
the method of variation of the
energy-function.
The
meaning
of the reduction of the number of conditions
from five to four was first
pointed
out
by
Kelvin and
Tait;]:.
It lies in the
circumstance that the actual distribution of tractions on the
edge
which
give
rise to the torsional
couple
is immaterial. The
couple
on
any
finite
length
might
be
applied by
means of tractions directed at
right angles
to the middle
plane,
and
these,
when reduced to force- and
couple-resultants,
estimated
per
unit of
length
of the
edge-line,
would be
equivalent
to a distribution of
shearing
force of the
type
N instead of torsional
couple
of the
type
H. The
required shearing
force is
easily
found to be

3H/9s.
This result is obtained
by
means of the
following
theorem of Statics : A line-distribution of
couple
of
amount H
per
unit of
length
of a
plane
closed curve
s,
the axis of the
couple
at
any
point being
normal to the
curve,
is
statically equivalent
to a line-
distribution of force of amount

dH/ds,
the direction of the force at
any point
being
at
right angles
to the
plane
of the curve.
*
See
Introduction,
footnote 36. Poissou's solutions of
special problems
are not
invalidated,
because in all of them H vanishes.
t See
Introduction,
footnote 125.
X
Nat. Phil, first
edition,
1867. The same
explanation
was
given by
J.
Boussinesq
in 1871.
See
Introduction,
footnote 128.
297]
AT THE EDGE OF A PLATE
439
H-oH
A
H
A
H+SH
The theorem is
proved
at once
by forming
the force- and
couple-resultants
of the line-
distribution of force
-dfflds.
The axis of z
being
at
right angles
to the
plane
of the
curve,
the force at
any point
is directed
parallel
to the axis of
z,
and the force-resultant is
expressed by
the
integral
-"j-ds
taken round the closed curve. This
integral
vanishes.
The
components
of the
couple-resultant
about the axes of x and
y
are
expressed by
the
integrals
\
-y
-^^ds
and \x
-^
ds taken round the curve. If v denotes the direction of
the normal to tlie
curve,
we have
\
-y
-K-ds=
\h Jds=
\Hcos{x, v)d,
and
U~ds==
I
-ff^^ds=
I
H
coa(y, ,-)d3,
the
integrations
being
taken round the curve. The
expressions
/ II cos
(x, v)
dt and
j
II
cos(j/, v)ds
are the values of the
components
of the
couple-resultant
of the line-
distribution of
couple
H.
The theorem
may
be illustrated
by
a
figure.
We
may
think of the cvirve as a
polygon
of a
large
number of sides. The
couple Hbs,
belonging
to
any
side of
length hs,
is
statically
equivalent
to two forces each of
magnitude H,
directed at
right angles
to the
plane
of the curve
in
opposite senses,
and
acting
at the ends of the
side. The
couples belonging
to the
adjacent
sides
may similarly
be
replaced by pairs
of forces of
magnitude
11+ SH or II- dH as shown in
Fig. 69,
where SH means
{dII/ds)Ss.
In the end we are
left with a force

dll at one end of
any
side of
length Ss, or,
in the
limit,
with a line-distribution of force
-dlljdi.
From this theorem it follows
that,
for the
purpose
of
forming
the
equations
of
equilibrium
of
any portion
of the
plate
contained within a
cylindrical
surface
C,
which cuts the middle surface at
right angles
in a curve
s,
the
torsional
couple
if
maybe
omitted,
provided
that the
shearing
stress-resultant
iV is
replaced by
N
-dH/ds*.
Now the
boundary
conditions are
limiting
forms of the
equations
of
equilibrium
for certain short narrow
strips
of the
plate
;
the contour in which the
boundary
of
any
one of these
strips
cuts the
middle
plane
consists of a short arc of the
edge-line,
the two normals to this
curve at the ends of the
arc,
and the arc of a curve
parallel
to the
edge-line
intercepted
between these normals. The limit is taken
by
first
bringing
the
parallel
curve to coincidence with the
edge-line,
and then
diminishing
the
length
of the arc of the
edge-line indefinitely.
In accordance with the above
*
This result
might
be used in
forming
the
equations
of
equilibrium (11)
and
(12).
The line-
integrals
in the third of
equations (9)
and the first two of
equations (10)
would be written
/(.-f).. /{-si..,(.-f)[..
/jocos.-.(.-f)f<..
and these can be transformed
easily
into the forms
given
in
(9)
and
(10).
H-SH H H-t-SH
Fig.
69.
440 BOUNDARY CONDITIONS
[CH.
XXII
theorem we are to form these
equations by omitting
H and
H,
and
replacing
N and N
hy NdH/ds
and N
8H/9s.
The
boundary
conditions are thus
found to be
f
=
T, 5=S,
N-dHlds
=
'N-dU/ds,
G
=
G.
These four
equations
are
equivalent
to the
boundary
conditions
adopted
by
Kirchhoff.
In
investigating
the
boundary
conditions
by
the
process just
sketched we observe that
the terms contributed to the
equations
of
equihbrium by
the
body
forces and the tractions
on the faces of the
plate
do not
merely
vanish in the
limit,
but the
quotients
of them
by
the
length
of the short arc of the
edge-line
which is
part
of the contour of the
strip
also
vanish in the limit when this
length
is diminished
indefinitely.
If this arc is denoted
by
8i we have such
equations
as
lim
(Ss)-^
I
IX'dxdy^O,
lim
{h$)-'^ \ \{L' +yZ')dxdy
=
0,
6*=0
J J
6s
=
J J
the
integration being
taken over the area within the contour of the
strip.
The
equations
of
equilibrium
of the
strip
lead therefore to the
equations
lim
_
{hs)-^ UTcos.e-S^m6)ds
=
Q,
lim
_
{bs)-^
UTsm e +
Sco6)ds
=
0,\
in which the
integrations
are taken all round the contour of the
strip,
and
T,
... denote
the force- and
couple-reultants
of the tractions on the
edges
of the
strip,
estimated in
accordance with the conventions laid down in Article 294. We evaluate the contributions
made to the various
line-integrals by
the four lines in which the
edges
of the
strip
cut
the middle
plane.
Since the
parallel
curve is
brought
to coincidence with the
edge-line,
the contributions of the short
lengths
of the two normals to this curve have zero limits
;
and we have to evaluate the contributions of the arcs of the
edge-line
and of the
parallel
curve. Let
v^
denote the direction of the normal to the
edge-line
drawn outwards. The
contributions of this arc
may
be estimated as
(13)
and
{Tcos(^, v)-Scos(y, i'o)}S, {Tcos(^, i/^)
+
S cos
(.t;, ^o)} fi,
JN-
I
-
G cos
O, v)
+
yhii-
g^
j|
8,
|g
cos
{x, v^)
-xhi-
-^-J^
I
In
evaluating
the contributions of the arc of the
parallel curve,
we observe that the con-
ventions,
in accordance with which the
T,
...
belonging
to this curve are
estimated,
require
the normal to the curve to be drawn in the
opposite
sense to
vo,
and the curve to
be described in the
opposite
sense to the
edge-line,
but the arc of the curve over which
we
integrate
has the same
length
Ss as the arc of the
edge-line.
In the limit when the
parallel
curve is
brought
to coincidence with the
edge-line
we
have,
in accordance with
these
conventions,
T= f
,
,S'=
^, N=-N, 0=0,
H=
H,
dff/d.i
=
-
dff/ds,
and cos 5= -cos
(x, >/(,),
sin 6
=

cos
(i/, v,,).
297]
AT THE EDGE OF A PLATE 441
Hence the contributions of the arc of the
parallel
curve
may
be estimated as
{-fcos(^, vo)
+
^cos(y, vo)}Ss, {-fcos(y, vo)-<?cos(^, vo)]8i,
i-N+j-\ss,
and
jtfcosCy,
..)
+
.y(^-iV^+?g^^|as,
i
-a cos
(a;,
yf,)-xf-If+^^A\
lit.
On
adding
the contributions of the two
arcs, dividing by Ss,
and
equating
the
resulting
expressions
to
zero,
we have the
boundary
conditions in the forms
previously
stated.
In
general
we shall omit the bars over the letters
T,...,
and write the
boundary
conditions at an
edge
to which
given
forces are
applied
in the form
r=T,
S
=
S,
iV-^=N-^,
G
=
G
(14)
At a free
edge
T, S,
NdHjds,
G vanish. At a
"supported" edge
the
displacement
w of a
point
on the middle
plane
at
right angles
to this
plane
vanishes,
and
T, S,
also vanish. At a
clamped edge,
where the inclination
of the middle
plane
is not
permitted
to
vary,
the
displacement (u,
v, w)
of a
point
on the middle
plane
vanishes,
and
dwjdv
also
vanishes,
v
denoting
the
direction of the normal to the
edge-line.
The efiect of the mode of
application
of the torsional
couple may
be illustrated
further
by
an exact solution of the
equations
of
equilibrium
of
isotropic
solids*. Let
the
edge-line
be the
rectangle given by x=+a, y=+h.
The
plate
is then an extreme
example
of a flat
rectangular
bar. When such a bar is twisted
by opposing couples
about
the axis of
x,
so that the twist
produced
is
r,
we know from Article 221
(c)
that the dis-
placement
is
given by
""^^
2A
provided
that the tractions
by
which the torsional
couple
is
produced
are
expressed by
the formulae
(2
+
ll^
(2
+
l)^,r^
2<A
-
(-)-
""^^
2A
'^'
2A
2A
.
, (2714-1) Try (2M+l)ir
/ s
smh^^

,
--
cos^

~r^

2A
"
(-)"
2A 2A
cosn
2^
^V^A^6-F^*^g)^J(2;^tanh'
There are no tractions on the faces
z=h
or on the
edges y=b.
The total torsional
couple
on the
edge
a;
=
a is
2A
and of this one-half is contributed
by
the tractions
Xy
directed
parallel
to the middle
plane,
and the other half
by
the tractions
.1'^
directed at
right angles
to the middle
plane.
Kelvin and
Tait,
Nat.
Phil.,
Part 2,
pp.
267 et
seq.
442 BOUNDARY CONDITIONS
[CH.
XXII
When the
plate
is
very
thin the total torsional
couple
is
approximately equal
to
^firh^b,
so that the
average
torsional
couple per
unit of
length
of the
edge-lines
.r= +a
is
approximately equal
to
/xrA^.
At
any point
which is not near an
edge y=h,
the state
of the
plate
is
expressed approximately by
the
equations
u=
ryz,
v=
Tzx, w=Txy.
The traction
X^
is
nearly equal
to

iiirz
at all
points
which are not
very
near to the
edges y= b,
and the traction
X,
is
very
small at all such
points.
The distribution of
traction on the
edge
x=a is
very nearly equivalent
to a constant torsional
couple
such as
would be denoted
by -ffj,
of amount
j^firh?,
combined with
shearing
stress-resultants such
as would be denoted
by N'l, having
values whicli differ
appreciably
from zero
only
near
the corners
{x
=
a,
y=
b),
and
equivalent
to forces at the corners of amount
^fir/i^.
At
a distance from the free
edges y=+b
which exceeds three or four times the
thickness,
the stress is
practically expressed by giving
the value
-
2jitz
to the
stress-component X^
and zero values to the
remaining stress-components.
The
greater part
of the
plate
is in
practically
the same state as it would be if there were torsional
couples, siJecified by
Ili
=
fjfjiTh^
at all
points
of the
edges
x=
a,
and
H2=
^firh^
at all
points
of the
edges
y=
b.
Thus the forces at the corners
may
be
replaced by
a
statically equivalent
distri-
bution of torsional
couple
on the free
edges,
without
sensibly altering
the state of the
plate,
except
in a narrow
region
near these
edges.
Within this
region
the value of the torsional
couple ff2, belonging
to
any
line
y
=
const.,
which would be calculated from the exact
solution,
diminishes
rapidly,
from

^/xrA'to
zero,
as the
edge
is
approached.
The
rapid
diminution of
H2
is
accompanied,
as we should
expect
from the second of
equations (12), by large
values of
^1.
If we
integrate Ni
across the
region,
that is to
say,
if we form the
integral
I
N^dy,
taken over a
length, equal
to three or four times the
thickness, along any
line drawn at
right angles
to an
edge
y
=
b or
y=
-b and terminated at that
edge,
we find the value of the
integral
to be
very
nearly equal
to
^/iTA'.
This remark enables us to understand
why,
in the
investigation
of
equations (14),
the third of
equations (13),
viz. lim
(8)-'
I ( iV'--a \d$
=
0,
where the
integration
is
taken round the contour of a
"strip,"
as was
explained,
should not be
replaced by
the
equation
lim
{hs)~^\
Nds
=
0,
and also
why
the latter
equation
does not lead to the
result N='S. When
N,
Hare calculated from the state of strain which holds at a distance
from the
edge,
and
equations (14)
are established
by
the method
employed above,
it is
implied
that no substantial difference will be made in the results if the linear dimensions
of the
strip,
instead of
being
diminished
indefinitely,
are not reduced below
lengths equal
to three or four times the thickness. When the dimensions of the
strip
are of this
order,
the contributions made to the
integral
( Nds
by
those
parts
of the contour which are
normal to the
edge-line may
not
always
be
negligible
; but,
if
not, they
will be
practically
balanced
by
the contributions made to
j
-
(dff/ds)
ds
by
the same
parts
of the contour*.
298. Relation between the flexural
couples
and the curvature.
In Article 90 we found a
particular
solution of the
equations
of
equili-
brium of an
isotropic
elastic solid
body,
which
represents
the deformation of
*
Cf. H.
Lamb,
London Math. Soc.
Proc,
vol. 21
(1891), p.
70.
297,
298]
CURVATURE PRODUCED BY COUPLES 443
a
plate slightly
bent
by couples applied
at its
edges.
To
express
the result
which we then found in the notation of Article 294 we
proceed
as follows :

On the surface into which the middle

plane
is bent we draw the
principal
tangents
at
any point.
We denote
by
,, s,
the directions of these lines on
the unstrained middle
plane, hy
R^, JR,
the radii of curvature of the normal
sections of the surface drawn
through
them
respectively, by
Gi, 0,
the
flexural
couples belonging
to
plane
sections of the
plate
which are normal to
the middle surface and to the lines
s,, Sj
respectively.
We determine the
senses of these
couples by
the conventions stated in Article 294 in the same
way
as if
s, , s^,
z were
parallel
to the axes of a
right-handed system.
Then,
according
to Article
90,
when the
plate
is bent so that
i?i, R^
are
constants,
and the directions
Sj, s,
are
fixed,
the stress-resultants and the torsional
couples belonging
to the
principal planes
of section
vanish,
and the flexural
couples
Gi, Gi
belonging
to these
planes
are
given by
the
equations
Gi'
=
-B{l/Ri
+
ajR,), G,'=-Dil/R,
+
a/Ri), (15)
where,
with the usual notation for elastic
constants,
D
=
lEh'jil
-
<7')
=
f/i/i' (X,
4-
/ji)/{\
+
2/Li) (16)
The constant D will be called the
"
flexural
rigidity
"
of the
plate.
Now let the direction
Sj
make
angles <j>
and
^Tr

<f>
with the axes of x
and
y.
Then,
according
to
(4),
Gi, G^, Hi
are
given by
the
equations
Gi
cos^
<^
-f-
Gj
sin=
<})-HiSm2<j)
=
-B
(1/ie,
+
a-/R,),
Gi
sin'
<f>
+
G,
cos'
(/>
+
Zf,
sin
20
=
-
D
(l/iJ,
-|-
o-/^),
i {Gi
-
G,)
sin
2<f>
+
Hi
cos
2(^
=
0,
from which we find
.os^
(f> tim-<j> /sin'(^
cos'<^\"|
-RT^^.
^"vW'^Rr)]'
Gi
=
-I>
G,
=
-D
sin' cos'
^
/cos'
^
sin'
Ri R2 \ Ri
n
ir.
=
ii)(l-<r)sin2<^(-i--^
Again,
let w be the
displacement
of a
point
on the middle
plane
in the
direction of the normal to this
plane,
and write
_8'w _9'w _
3^ ,.,-,
""S^'
"^"92/'' '^'dxdy
^^
Then the indicatrix of the surface into which the middle
plane
is bent is
given,
with sufficient
approximation, by
the
equation
!'
+
K,y^
+
2Txy

const.
;
and,
when the form on the left is transformed to coordinates
|, rj,
of which
the axes coincide in direction with the lines
Sj, s^,
it becomes
444 CURVATURE PRODUCED BY COUPLES
[CH.
XXII
Hence we have the
equations
cos'd) sin=<f) sin' A cos'rf) _ .
^,/l
1\
*'=-fir'^^r' "'^^r^-w ^"^'^'^^'^U-i-J'
and the formulae for
Gi, 0,, Hi
become
Gi
=
-D{Ki
+
aK,), 0,
=
-D{k,
+
<tk,), H,
=
D{l-a)r. ...(18)
We shall show that the formulae
(18),
in which
ki, k^,
t are
given by (17),
and
D
by (16),
are either correct or
approximately
correct values of the stress-
couples
in a
very
wide class of
problems.
We observe here that
they
are
equivalent
to the statements that the flexural
couples belonging
to the two
principal planes
of section at
any point
are
given,
in terms of the
principal
radii of curvature at the
point, by
the formula;
(15),
and that the torsional
couples belonging
to these two
principal planes
vanish.
If s denotes the direction of the
tangent
to
any
curve drawn on the middle
plane,
and
V the direction of the normal to this
curve,
and if 6 denotes the
angle
between the directions
V, X,
we
find, by substituting
from
(17)
and
(18)
in
(4),
the
equations
.=
-.)[cos^.(S..5).sin^.(|^..).(l-.)sin..U],
^=Z)(l-.)[sin^cosfl(|^-g)
+
(cos^^-sin^^)|j].
We
may
transform these
equations,
so as to avoid the reference to fi.\ed axes of x and
y*,
by
means of the formulae
3
8-/,3
8
.,9-^3
35

35 1
/-.m
^
=
008^5;

smdi^ri ;^
=
co8e=--)-sinfl5-, ^
=
0, 5-=-i (1*)
eg
oy
ex ov ox
oy
ov as
p
where
p'
is the radius of curvature of the curve in
question.
We find
T^ f32w
/a^w 1
aw\i

,3
/3w\
,.,
These
equations
hold whenever the
stress-couples
are
expressed by
the formulae
(18).
In the
problem
of Article 90 we found for the
potential energy
of the
plate,
estimated
per
unit of area of the middle
plane,
the formula
or,
in our
present
notation,
iZ)[(/c,
+
,)^-2(l-o-)(Yi,-T')]
(21)
We shall find that this formula also is
correct,
or
approximately
correct,
in a
wide class of
problems.
299. Method of
determining
the stress in a
platef.
We
proceed
to consider some
particular
solutions of the
equations
of
equilibrium
of an
isotropic
elastic solid
body, subjected
to surface tractions
only,
which are
applicable
to the
problem
of a
plate
deformed
by given
forces.
*
Of. Lord
Bayleigh, Theory of Sound,
vol.
1,
216.
t
The method was worked out
briefly,
and in a much more
general fashion, by
J. H.
Michell,
London Math. Soc.
Proc,
vol. 31
(1900), p.
100.
298,
299]
METHOD OF DETERMINING THE STRESS 445
These solutions will be obtained
by
means of the
system
of
equations
for the
determination of the
stress-components
which were
given
in
(iv)
of Article 92.
It was there shown
that,
besides the
equations
^l^
+
^Iv
+ ^Ji^O
^Jy^^Jy + ^Ji^Q
9^i
+
^
+
^^
=
...(22)
dx
dy
dz
'
dx
dy
dz
'
dx
dy
dz

\ >
we have the two sets of
equations
1 3' 1 rl'0 1 9*0
^
^'~
l+adaf'
^
^''~
l+<7dy"
^
^'
1
+
a dz'
'
' '
'^'^'^^
and
1
+
cr
oydz
I + a dzdx
*
1
+a
oxay
where B
=
X^
+
Yy
+
Z, (2.5)
It was shown also that the function is
harmonic,
so that V=0
=
0,
and that
each of the
stress-components
satisfies the
equation V*/
=
0.
We shall
suppose
in the first
place
that the
plate
is held
by
forces
applied
at its
edge only.
Then the faces z
=
h
are free from
traction,
or we have
Xi
=
Fi
=
^2
=
when z
=
h.
It follows from the third of
equations (22)
that
dZJdz
vanishes a,t z
=
h and at ^
=

A. Hence
Z^
satisfies the
equation
V*^^
=
and the conditions
Z^

0,
dZ^/dz
=
at z
=

h. If the
plate
had
no boundaries besides the
planes
z
=
h,
the
only possible
value for
Z^
would
be zero. We shall take
Zz
to vanish*. It then follows from the
equations
V2
=
0, V%
=
-(1
+
o)-^9^0/^2^
that is of the form
0,
-f
^,,
where
0
and
0,
are
plane
harmonic functions of x and
y
which are
independent
of z.
For the determination of
X^, Y^
we have the
equations
dx^dy'^'
'
l+<Tdx'
^'
l
+
ady'
and the conditions that
X^
=
Fj
=
at 2
=
+ /i. A
particular
solution is
given by
the
equations
^.=lr^,<')t. >'.=-irf.<"
>f
w
We shall take
X^
and
Y^
to have these forms. When
X^, Y^, Z^
are known
general
formulfe can be obtained for
Xx,
Yy, Xy.
If
1
is a
constant, Xz
and
Fj
vanish as well as
Z^,
and the
plate
is then
in a state of
"plane
stress." If
,
depends upon
x and
y
the
plate
is in a
state of
"
generalized plane
stress
"
(Article 94).
We shall examine
separately
these two cases.
In like
manner,
when the
plate
is bent
by pressure applied
to its
faces,
we find a
particular
solution of the
equation
V'Z^
=
which
yields
the
prescribed
values of
Z^
at z
=
h and z
=
h,
and we deduce the most
general
form of which is consistent with this solution. We
proceed
to
*
J. H.
Michell,
loc.
cit.,
calls attention to the
analogy
of this
procednre
to the
customary
treatment of the condenser
problem
in Electrostatics.
446
PLATE IN A STATE OF PLANE STRESS
[CH.
XXII
find
particular
solutions of the
equations
satisfied
by
Xi
and
Y,
and to deduce
general
formulae for
Xx,
Yy,Xy.
300. Plane stress.
When
Xi, Yz, Zi
vanish
throughout
the
plate
there is a state of
plane
stress. We have
already
determined in Article 145 the most
general
forms
for the
remaining stress-components
and the
corresponding displacements.
We found for @ the
expression
=
0o
+
y9^, (27)
where

is a
plane
harmonic function of x and
y,
and
/3
is a constant. The
stress-components
X^,
Yy, Xy
are derived from a stress-function
p(; by
the
formulae
x,=
and
X
ba8 the form
df dxdy
'
X
=
Xo
+
^Xi-2r^
^'00,
.(28)
.(29)
where
V,Xo
=
o,
V,^Xi
=
/8
(30)
If we introduce a
pair
of
conjugate
functions
^, t]
oi x and
y
which are
such that
rA
=
12
=
@
dx
dy
"'
dy
dx'
(31)
.(33)
the most
general
forms for
^a
and
^.i
can be written
Xo
=
i^^+/. x^
=
\^{x-
+
f)+F, (32)
where
/
and F are
plane
harmonic functions. The
displacement (w,
v,
w)
is
then
expressed by
the formulae
The solution
represents
two
superposed stress-systems,
one
depending
on
o.
Xo'
^^^ *^ other on
^,
X\-
These two
systems
are
independent
of each other.
301. Plate stretched
by
forces in its
plane.
Taking
the
(o,
;^o) system,
we have the
displacement given by
the
equations
"
pH^*'^^
^y)
E
dy'
E
w=-^z@,
.(34)
299-301]
PLATE STRKTCHED IN ITS PLANE 447
where
Xo
is of the form
^x^ +/,
0,,
and
/
are
plane
harmonic
functions,
and
f, 7j
are determined
by (31).
The normal
displacement
of the middle
plane
vanishes,
or the
plate
is not bent. The stress is
expressed by
the formulae
V
X,,

dxdy
V^'
Z'^0
(35)
21
+
0-
The stress-resultants
T^, T^, (S,
are
expressed by
the
equations
"^'^U^'^^-lrr.
^
=
-a-4K-5rf-^'^^^")'
>
.(36)
The stress-resultants
N^, N,,
and the
stress-couples
G,, (?,, ^i
vanish. The
equations (11)
and
(12),
in which
X', V, Z', L', M', vanish,
are
obviously
satisfied
by
these forms.
When we transform the
expressions
for
Tj, T^, *S,
by
means of the
equa-
tions
(4),
we find
that,
at a
point
of the
edge-line
where the normal makes
an
angle
6 with the axis of
x,
the tension and
shearing-force
T,
S are
given
by
the
equations
T
=
fees''
(9 .- + sin' 6*
1^
-
2 sin 6 cos 6
5-.
V
dy^
ex'
dxdy.
^=U^cos^g-|,)-cos2^^-|^J(2Ax.-l-
)K-|rf^/^'e.),
+
0-
+ <T
When these
equations
are transformed
by
means of the formulae
(19)
so as to
eliminate the reference to fixed axes of x and
y,
they
become
Hi
1 d
s^
p'
dv.
^
=
-a.|a.r^-3-f
These
expressions
are
sufficiently general
to
represent
the effects of
any
forces
applied
to the
edge
in the
plane
of the
plate*.
If the forces are
applied by
means of tractions
specified
in accordance with
equations (35),
the solution
expressed by equations (34)
is exact
; but,
if the
applied
tractions at the
edge
are distributed in
any
other
way,
without
ceasing
to be
equivalent
to
*
The case of a circular
plate
was worked out in detail
by Clebsch, Elasticitdt,
42.
448 PLATE STRETCHED IN ITS PLANE
[CH.
XXII
resultants of the
types
T, S,
the solution
represents
the state of the
plate
with sufficient
approximation
at all
points
which are not close to the
edge.
It
may
be observed that the stress-resultants and the
potential energy per
unit of area
can be
expressed
in terms of the extension and
shearing
strain of the middle
plane.
If we
write
u, v,
for the values of u and v when
z=0,
and
put
3u 0v 3u 3v
wefind
'^
=
E\d^~''d^)' ''-\dx^~''~df)'
^i
+
^2=^^Q<"
and then we have
""' '
B
dxdy'
Ti
=
j-3^2
(f1
+ <^f
2)
-
4
^^7^
gp
(f
I
+
f
2),
The
potential energy per
unit of area can be shown to be
<t2 Eh^
rf
3^(^i
+
^2)P
, pfa
+
^a))" ,

f3^(^i
+
^2)-|n
+^rr;r3^2Lt
3^'
J I if~] I 3-^3y J J'
Some
special examples
of the
general theory
will be useful to us
presently.
(i)
If we
put
00
=
0, xo
is a
plane
harmonic
function,
and the state of the
plate
is one
of
plane
strain
involving
no dilatation or rotation
[cf
Article 14
{d)\.
We have
E dx' E
dy'
'
and
r.=
_r,=2A^,
.S,=
-2Ag5^.
(ii)
If
00
is constant we have
|
=
eo.i;, rj

Qt)y,
and we
may put x<s

\Qo{x^+y^).
and
then we have
and
Ti
=
T2=eoh, 8^
=
0.
This is the solution for uniform tension
GqA
all round the
edge.
(iii)
If
&n=ax,
where a is
constant,
we have
i
=
ia{x^-y^), r)
=
axy,
and we
may put
Xn^^aafi,
and then we have
and
7'i
=
0, ^2
=
2*0^, 'S'i=0.
A more
general
solution can be obtained
by adding
the
displacement given
in
(i).
.(38)
301,
302]
PLATE BENT TO A STATE OF PLANE STRESS 449
(iv) By taking
the function
xo
in
(i)
to be of the second
degree
in x and
y,
we
may
obtain the most
general
sohition in which the
stress-components
are
independent
of x and
y,
or the
plate
is stretched
uniformly.
The results
may
be
expressed
in terms of the
quantities
fi, e2i
^ that define the
stretching
of the middle
plane.
We should find for the stress-
components
that do not vanish the
expressions
A'^
=
E{,,
+
af;)Kl-a% r
=
.E(f2
+
,)/(l-<T''), X,
=
i'ar/(l+<r),
and for the
displacement
the
expressions
U
=
(iX +
^^1/,
V
=
fiy
+
^mX,
W= -0-2
(fi
+
f2)/(l-o).
302. Plate bent to a state of
plane
stress.
Omitting
in
equations (33)
the terms that
depend
on
&,
Xm
we have the
displacement given by
the
equations
"^
"
"
21:
^^'
^
^'
"
'"'"'^
^
^^^'
where
Xi
bas the form
Xi
=
i/8 (ar'
+
3/)
+
F,
and i' is a
plane
harmonic function.
The stress is
expressed by
the
equations
The stress-resultants
vanish,
and the
stress-couples
are
given by
the
equations
The
equations (11)
and
(12),
in which
X', Y', Z', 11,
W
vanish,
are
obviously
satisfied
by
these forms.
The normal
displacement
w of the middle
plane
is
given by
the
equation
w
=
-4('^
+
2/^)+T^-
^*^^
80 that the curvature is
expressed by
the
equations
"'"
E^ E dx''
"'"
E'^ E
dif
E
dxdy'
From these
equations
and the
equation V,';)^i
=
/8,
we find
SO that the formulae
(18)
hold.
The
stress-couples
at the
edge
are
expressible
in the forms
<^-H'&'-?^)-
''-i'-rM)
<">
L. E.
29
450 PLATE BENT TO A STATE OF PLANE STRESS
[CH.
XXII
and,
if the
edge
is
subjected
to
given
forces,
and
dHlds
must have
pre-
scribed values at the
edge.
Since
%i
satisfies the
equation ^i^Xi

^>
the
formulae
(41)
for and
dH/ds
are not
sufficiently general
to
permit
of the
satisfaction of such conditions. It follows that a
plate
free from
any
forces,
except
such as are
applied
at the
edge
and are
statically equivalent
to
couples,
will not be in a state of
plane
stress unless the
couples
can be
expressed by
the formulae
(41).
Some
particular
results are
appended.
(i)
When the
plate
is bent to a state of
plane
stress the sum of the
principal
curvatures
of the surface into which the middle
plane
is bent is constant.
(ii)
In the same case the
potential energy per
unit of area of the middle
plane
is
given
exactly by
the formula
(21).
(iii)
A
particular
case will be found
by taking
the function F introduced in
equations
(32)
to be of the second
degree
in x and
y.
Then
xi
^.Iso is of the second
degree
in x
andy,
and we
may
take it to be
homogeneous
of this
degree
without
altering
the
expressions
for
the
stress-components.
In this case w also is
homogeneous
of the second
degree
in x and
_y,
and
Ki, K2)
I" ^re constants. The value of
xi
is
and the
stress-components
which do not vanish are
given by
the
equations
E E E
1

(T^ 1

<r
i--r(T
(iv)
This case includes that discussed in Article
90,
and
becomes,
in
fact,
identical
with it when the axes of x and
y
are chosen so that t
vanishes,
that is to
say
so as to be
parallel
to the lines which become lines of curvature of the surface into which the middle
plane
is bent. Another
special
sub-case would be found
by taking
the
plate
to be
rectangular,
and the axes of x and
y parallel
to its
edges,
and
supposing
that
kj
and
k^ vanish,
while t
is constant. We should then find
u=Tyi,
V
-Tza,
w
=
Txy.
The stress-resultants and the flexural
couples
Cj , O-i vanish,
and the torsional
couples 5]
and
Hi
are
equal
to + i)
(1

o-)
t. The result is that a
rectangular plate
can be held in the
form of an anticlastic surface
w=Txy by
torsional
couples
of amount
D{\a-)T per
unit
of
length applied
to its
edges
in
proper senses,
or
by
two
pairs
of forces directed
normally
to the
plate
and
applied
at its corners*. The two forces of a
pair
are
applied
in like senses
at the ends of a
diagonal,
and those
applied
at the ends of the two
diagonals
have
opposite
senses. The
magnitude
of each force is 2Z)
(1
-
o-)
r.
303. Generalized
plane
stress.
When
Z^
vanishes
everywhere,
and
X^, F^
vanish at ^^
=
+
A,
we take the
values of
X^
and
Y^
to be
given by equations (26)
of Article 299. To
deteimine
Xx,
Yy, Xy
we have the first two of
equations (22)
and
(23),
the
third of
equations (24),
and
equation (25),
in which
Zz vanishes, X^
and
Yi
are taken to be
given by (26),
and @ has the form
0, -|-0i,
the functions

H.
Lamb,
London Math. Soc.
Proe.,
vol. 21
(1891), p.
70.
302,
303]
OR OF GENERALIZED PLANE STRESS 461
o
and
1
being plane
harmonic functions of
x,
y
and
independent
of z. The
stress
depending upon

has been determined in Article
300,
and we shall
omit
.
We have therefore the
equations
dx
dy
1
4-
o" 9a;
'
dx
dy
1
+ a
dy
'
^^^^
+
r!" II
=
0.
v=F,
+
/-'^
=
0,
v=z,
+
,
A-
.^
=
0.
1
+ o- oa;' 1
+
a
dy"
"
1
+
a
oxoy
(42)
From the first two of these
equations
we find
where
%'
is a function of
x,y,z\
and the last
equation
of
(42) gives
The
remaining equations
of
(42)
can now be transformed into the forms
dxdyXdz''
1
+
0- V
These
equations
show that the
expression
^
-
^^
^^^
i^ ^ linear function
of X and
y,
and we
may
take it to be zero without
altering
the values of
Xx,
Yy, Xy.
We therefore write
^=^^^^ir^./^-
(^^>
where
V,V=-[^@i
(45)
If we introduce two
conjugate
functions
fi, i;,
of
x,
y
which are such that
dx I
we
may express p^/
in the form
9|i^9^i^Q
8|i^_3^
(4gN
dx
dy
' '
dy
dx
^^=-w^f^^^^^
^''^
where
J*"]
is a
plane
harmonic function. Thus the form of
y^,
and therewith
also that of
X^,
Yy, Xy,
is
completely
determined.
The
displacement
is determined
by
the
equations
of the
types
8w 1
/
y ^r ys
dw dv 2
(1
+
0-)
y
292
452
PLATE BENT TO A STATE OF
[CH.
XXII
in which
Z^ vanishes, X^
and
Fj
are
given by (26),
and
X^,
Yy, Xy by (43).
The
resulting
forms for
u, v,
w are
V
=

E
1
E
(i+.),9g:+^(.2_
I/O \ ^
^i'
\
(I
+
.).^^'.H2-.)^fj.
.(48)
w=
^,[(l
+
<r)x/4(/<^-i7^0.]-
I
304. Plate bent to a state of
generalized plane
stress.
The normal
displacement
w of the middle
plane
is
given by
the
equation
w
=
i{A0,+(l
+
r)x/!, .(49)
and,
since
Vj^,
=
0,
we have
by (45)
Vi*w
=
0,
(50)
where
Vi^
denotes the
operator d*/dx*
+
d^ldij*
+
2d*/da^dy^,
and then
The
stress-components
are
given
by
the
equations
-^
zw +
j^^
[h'z
-
K2
-
<t)z'}
V,^
)
1
+
0- 1

o-^
'
z.=-
303,
304]
GENERALIZED PLANE STRESS
453
The
stress-resultants and
stress-couples
belonging
to
any
curve of which
the normal is v can be
expressed
in the forms
vv
y...(o4)
H
=
where
p
denotes the radius of curvature of the curve. At a
boundary
to
which
given
forces and
couples
are
applied
G and
N-dH/ds
have
given
values. The solution is
.sufficiently general
to admit of the
satisfaction of
such
boundary
conditions. The solution
expressed
by (48)
is exact if the
applied
tractions at the
edges
are distributed in
accordance with
(52),
in
which w
satisfies
(50) ; but,
if
they
are
distributed
otherwise,
without
ceasing
to be
equivalent
to resultants of the
types
N, 0, H,
the solution
represents
the state of the
plate
with sufficient
approximation
at all
points
which are
not close to the
edge.
The
potential energy per
unit of area can be shown to be
j.[,.,--.(i-,){--(||)}]
-5,'M.['
272+640-H-50-'
rasviVa^riV
_ /a27j2w\n
420(1-0-) \_ dx^^^dy"^ \^xdy)\
^^^^
The results here obtained include those found in Article 302
by
putting
1
=
y9.
Equations (53)
show that the
stress-couples
are not
expressed by
the formuljB
(18)
unless the sum of the
principal
curvatures is a constant or
a linear function of x and
y.
In like manner the formula
(21)
is not verified
tlnless the sum of the
principal
curvatures is constant
;
but these formulae
yield approximate expressions
for the
stress-couples
and the
potential
energy
when h is small.
The
theory
which has been
given
in Art. 301 and in this Article consists
rather in the
specification
of forms of exact solutions of the
equations
of
equilibrium
than in the determination of
complete
solutions of these
equations.
The forms contain a number of unknown
functions,
and the
complete
solutions
are to be obtained
by adjusting
these functions so as to
satisfy
certain
differential
equations
such as
(50)
and certain
boundary
conditions. These
forms can
represent
the state of strain that would be
produced
in a
plate
of
any shape by any
forces
applied
to the
edge,
in so far as these forces are
expressed adequately by
a line-distribution of
force,
specified by components,
T, S,
N

3H/9s,
and a line-distribution of flexural
couple
G.
454 CIRCULAR PLATE CENTRALLY LOADED
[CH.
XXII
305. Circular
plate
loaded at its centre*.
The
problem
of the circular
plate supported
or
clamped
at the
edge
and loaded at the
centre
may
serve as an
example
of the
theory just given.
If a is the radius of the
plate,
and r denotes the distance of
any point
from the
centre,
we
may
take w to be a function
of r
only,
and to be
given by
the
equation
where
W,
A are
constants,
and then we have on
any
circle of radius r
W
and the resultant
shearing
force on the
part
of the
plate
within the circle is W. Hence W
is the load at the centre of the
plate.
The
complete primitive
of
(56)
is
-^(r^log^
+
r'^^+iAr'
+
B+Clogr,
where B and C are constants of
integration.
If the
plate
is
complete up
to the
centre,
C must
vanish,
and we take therefore the solution
w
=
^(r^\og^+A+iAr^
+
B.
The flexural
couple
O at
any
circle r=a is
given by
the
equation
We
may
now determine the constants A and B. It the
plate
is
supported
at the
edge,
so that w and O vanish at r
=
a,
we find
W
^'-2nl)\
and the central
deflexion,
which is the value of

w at
r=0,
is
If the
plate
is
clamped
at the
edge,
so that w and
dw/dr
vanish a.t r
=
a,
we have
w
=
g^[rMog^-|(a.'-r2)],
(58)
and the central deflexion is
Wa'^/lQirD.
If the
plate
is
very
thin the central deflexion is
gi-eater
when it is
supported
at the
edge
than when it is
clamped
at the
edge
in the ratio
(3
+
(r)
:
(1+0-),
which is 13 : 5 when o-
=
J,
306. Plate in a state of stress which is
uniform,
or varies
uniformly,
over its
plane.
When the stress in a
plate
is the same at all
points
of
any plane parallel
to the faces
of the
plate
the
stress-components
are
independent
of x and
y,
and the
stress-equations
of
equilibrium
become
?^'=0 ^'=0 ^-^=0
dz
"'
dz
"'
dz
"
If the faces of the
plate
are free from traction it follows that
X^, Y Z^ vanish,
or the
plate
is in a state of
plane
stress. The most
general
state of
stress, independent
of x and
y,
which can be maintained in a
cylindrical
or
prismatic body by
tractions over its curved
surface can be obtained
by adding
the solutions
given
in
(iv)
of Article 301 and
(iii)
of
*
EesnltB
equivalent
to those obtained here were
given by
Saint-Venant in the
'
Annotated
Clebsch,'
Note du

45.
305,
306]
UNIFORMLY VARYING STRESS
455
Article 302. In these cases the stress is uniform over the cross-sections of the
oyUnder
or
prism.
When the
stress-components
are linear functions of x and
y
the stress varies
uniformly
over the cross-sections of the
cylinder
or
prism.
We
may
determine the most
general
possible
states of stress in a
prism
when the ends are free from
traction,
there are no
body forces,
and the
stress-components
are linear functions of x and
y.
For this
purpose
we should
express
all the
stress-components
in such forms as
X^==Xx^Xi'y\Xi\
where
AV, X^", Ay)
are functions of z. When we introduce these forms into the various
equations
which the
stress-components
have to
satisfy,
the terms of these
equations
which
contain
x,
or
y,
and the terms which are
independent
of x and
y
must
separately satisfy
the
equations.
We take fii-st the
stress-equations
of
equilibrium.
The
equation
'bx
dy
dz
'
combined with the conditions that
X,
vanishes at 2
=
A, gives
us the
equations
3 V
()
X.'=0, X."
=
0,
^4..V^'-|-AV'
=
0,
and in like manner we have the
equations
arc)
It follows that
A'j
and
F,
are
independent
of x and
y.
The third of the
stress-equations
becomes therefore
dZJdz
=
0, and,
since
Z,
vanishes at the faces of the
plate {z= h),
it
vanishes
everywhere.
Again
e is of the form
xe'+yQ"
+e^''),
where
0', e",
eC* are functions of
z, and,
since e is an harmonic
function, they
must be linear functions of z. The
equation
1 3^6
V^Xt=
-
.^-^
takes the form
B^A'/Bz^s constant,
so that
d^X,/S!?=0.
Since
AT,
satisfies
i+a-oxoz
this
equation
and vanishes at
z=h,
it must contain i^-K^ as a
factor,
and since it is
independent
of x and
y
it must be of the form A
(z^

A^),
where A is constant. Like
statements hold
concerning Y,.
It follows
that,
if a
cylindrical body
with its
generators parallel
to the axis of z is free
from
body
forces and from traction on the
plane ends,
the most
general type
of stress which
satisfies the condition that the
stress-components
are linear functions of x and
y
is included
under the
generalized plane
stress discussed in Article 303
by taking 60
and
e,
to be huear
functions of x and
y
and
restricting
the
auxiliary plane
harmonic functions
/
and
Fi
introduced in
equations (32)
and
(47)
to be of
degree
not
higher
than the third.
It
may
be shown
that,
in all the states of stress in a
plate
which are included in this
category,
the
stress-components
are
expressible
in terms of the
quantities fj, 21 ^>
which
define the
stretching
of the middle
plane,
and
k,, kj, t,
which define the curvature of the
surface into which this
plane
is
bent, by
the formulfe
A;x=
j-~2
ll
+
0'f2-(Kl+0':2}2])
E
1
ry=
:p-2
{f2-|-0-l
-
(K2-f- (7-Ki)2},
E
^v=i:f:^{i^-'"^}' \
(59)
E{k^-z^)d
^='-i
i_o.2
^('"
+
''^)'
^.
=
0.
456 PLATE BENT BY PRESSURE
[CH.
XXII
The stress-resultants and
stress-couples
are
expressed by
the formulse
- iEh
, ^
.
-,
^Eh
, ^
.
^
Eh. .
A',=
-i)^(M
+
2),
^2=-i)|(<,
+
.2),
^
^^^
and the
potential energy per
unit of area is
-iS[p^H-pm
<'>
307. Plate bent
by pressure
uniform over a face.
When the face z=h is
subjected
to uniform
pressure p,
we have
V*Z^
=
everywhere, dZ^jdz
=
at z
=
h and z
=

h, Z^^

p
at z
=
h, Z^
=
at
z
=
h. A
particular
solution is
Z,
=
ih-'p (2
+
hy(z- 2/i)
=
i /t-> (^
-
Wz
-
2h') (62)
and we take this to be the value of
Z^.
To determine we have the
equations
V^@
=
0, d'@ldz''
=
-
I (1
-f
<7) h-'i7z,
of which the most
general
solution has the form
@
=
-l(l+a) h->pz>
+
f (1
+
0-) h-'pz
(x^
+
f)
+
z,
+
0,
where
0i
and

^re
plane
harmonic functions. We
may
omit the terms
z@i
and
0
because the
stress-systems
that would be calculated from them
have been found
already.
We take therefore for @ the form

=
-{{l
+
a)h-'p2^+^{l
+
<r)h-^pz{x^
+
f)
(63)
To determine
X^
and
Fz
we have the
equations
and the conditions that
X^
and
Yg
vanish &t z
=
h and at ^
=

h. A
particular
solution is
X^
=
lh-'p{h''-z^)x,
Y,
=
pi-'p(h'-z')y,
(64)
and,
as in Article
299,
we take
X^
and
Y^
to have these values.
To determine
X^,
Yy, Xy
we have the
equations
dx
df
4A
'
dx
dy'
4,h'
'
V>X
=
VF
=
-fA->, V^Xy=Q,
X^ +
Yy
=
lh-p [- (2
-I-
ff)
^^
+
3^
{Kl
+
0-)
(^=
+
y')
+
^
+
2A].
)
I...
(65)
306,
307]
APPLIED TO ONE FACE 457
To
satisfy
the first two of these
equations
we take
JSl",,
Yy, Xy
to have
the forms
where
x
must
satisfy
the
equation
and then the
remaining equations
of
(65)
show that
al
+
^>'x+lg^(^+2/^)
must be a linear function of x and
y.
As in
previous
Articles,
this function
may
be taken to be zero without
altering
X^,
Yy
or
Xy,
and therefore
;^
must
have the form
where
;^i"
and
x"
^re functions of x and
y
which
satisfy
the
equations
'^.V=-i(i-<^)|,(^+2/0+ff.
ViV'=ip;
(66)
and we
may
take for
X\\ Xo"
^^^
particular
solutions
X"
=
-
I
(1
-
<r)
g
(x>
+
ff
+
T^f
(^
+
f),
\
.(67)
Xo'^hPi'^'
+
f)-
More
general integrals
of the
equations (66)
need not be taken because the
arbitrary plane
harmonic functions that
might
be added to the solutions
(67)
give
rise to
stress-systems
of the
types already
discussed.
The
expressions
which we have now found for
Xx,
Yy, Xy
are
z
Yy
=
iP
+
iplA^'
+
f
+
h^)-^M^-<^)p|ii^^'
+
f)-^^pi'
(68)
The
stress-components being given by (62), (64)
and
(68),
the
corresponding
displacement
is
given by
the formulae
_
l+a-
px
p^y
E 8A
V
=
-
^^^ H
[(2
-
a)
^
-
2h'z
-2h'-U^-<^)^(^
+
f)l
u
=
--^g-^[(2-a)2'-Sh'z-2h'-i{l-<7)z{x'
+
f)]
w
=
l+cr
p
[(l-h<7)^^-6/iV-8fe-t-3(A^-<ra=)(*'H2/^)-|(l-o-)(^+y')]-
(69)
458 PLATE BENT BY PRESSURE
[CH.
XXII
It is
noteworthy
that when the
displacement
is
expressed by
these formulae
the middle
plane
is
slightly
stretched. We
have,
in
fact,
when
=
The stress-resultants and
stress-couples
are
given by
the formulae
16
H,
=
-l{l-a-)pxy.
These forms
obviously satisfy equations (11)
and
(12)
in which
X', Y', L',
M'
vanish and Z' is
replaced by p.
The middle
plane
is bent into the surface
expressed by
the
equation
20
P'''
307,
308]
APPLIED TO ONE PACE
and thence we
obtain,
in the same
way
as
before,
the formulae
^*
=
l
tl2/*=
-
(6
+
'T)
^=
+ i
(5
+
<r)
^
+
f (1
+
.7) y'],
459
F
=
X
=
[Sh?-{2
+
da)z'+[ii{l
+
oa)a?
+
%{l+<T)f]zl ...(73)
The
displacement
is then
given by
the formulae
1
+
q-
jOp
+
^
^
(5*'^
+
f) ('
+
f)
+
^zx'h'
-
^
(3^
+
1/^)^1
,
1+q-
po
E W
E 4-h'
U?xy
+
-g-^
zxy {x"
+
y^)
^
xy^
,
h
(74)
-
^-^
X
(x^
4-
y-J
+
H*'"
+
3a;!/^)
A''
^^H;
The middle
plane
is
slightly
stretched in a direction at
right angles
to
that
along
which the
pressure
varies. We
have,
in
fact,
when z
=
du dv ..
poX
dv du
-
=
0,
-=^1
+
.)-,
g^
+
g^
=
0.
The stress-resultants and
stress-couples
are
given by
the formulae
N,
=
ip,(Sx'-hy')
+
j%Poh',
N,
=
\p,xy,
G'i
=
iVPo[H5
+
o-)^
+
(1+<t)/^
+
|(14-<7)/iH
'
(7-5)
(?=
=
tV Po [i (1
+
So-)
a;
+
(
1 +
<r) a;y
-
1 (2
+
3<7) fe],
^i
=
T'ffi>o [- Ml
-
<^)
(3^2/
+
2/)
+
1(2
-
<^)
%]
These forms
obviously satisfy equations (11)
and
(12)
in which
X', Y',
L'
,
M'
are
put equal
to zero and Z' is
put equal
to

pt,x.
The middle
plane
is bent into the surface
expressed by
the
equation
^^(^a?
+
y^y-^^J^^{^
+
Zf)
, (76)
24
-
00-
+
0-2
1 Po
and we find
G^i
=
-
jD
(,
+
aK^)
-I-
^'i
G^
=
-D{k.
+
a-K^)
+
jV
^'
1-cr
8
-I- 9o- -1- 3ct=
1-0-
Poh'x,
Poh'^x,
^,
=
i)(l-<7)T-^My
The formulae
(18)
are
approximately
correct when h is small.
460 PLATE BENT BY PRESSURE
[CH.
XXII
309. Circiilar
plate
bent
by
uniform
pressure
and
supported
at
the
edge.
When a
plate
whose
edge-line
is a
given
curve is
slightly
bent
by pressure,
which is
uniform,
or varies
uniformly,
over one
face,
the
stress-system
is to be obtained
by
com-
pounding
with the solution obtained in Article 307 or 308 solutions of the
types
discussed
in Articles 301 and 302 or
303,
and
adjusting
the latter so that the
boundary
conditions
may
be satisfied. We shall discuss the case of a
clamped edge presently.
When the
edge
is
supported,
the
boundary
conditions which hold at the
edge-line
are
w
=
0, 0=0,
T=S=0
(77)
Let the
plate
be
subjected
to uniform normal
pressure p
and
supported
at the
edge,
and let the
edge-line
be a circle r=a. The solution
given
in
(71) yields
the
following
values for
vi, O, T,
S a.t r
=
a :

The solution
given
in
(ii)
of Article 301
yields
the values
w
=
0,
=
0, T=^ph,
S=0
when
Gj
is
put equal
to
^p.
The solution
given
in Article 302
yields
zero values for T
and
S,
and it
may
be
adjusted
to
yield
constant values for w and O at r
=
a
by putting
Xi=i/3 {^^+2/'')
+
y,
where
y
is a constant. These values are
If we
put
Sp/S
+
o; 2,3-0-
\
_3(l-<r)pc^f5
+
(ra^ 8
+
(r
+
<r^ h^\
the values of w and (? at
r=a,
as
given by
the .solutions in Article 302 and in Article
307,
become identical.
We
may
now combine the three solutions so as to
satisfy
the conditions
(77)
at r=a.
We find the
following expressions
for the
components
of
displacement
'
E
'=^
+
f
[-H|f^{(3
+
<r)a2-2(l
+
a)r2}-A|(5-f2<r
+
,r2)-fJsJ(l-^,r)2],
(78)
where w
=
-
J

(a^
-
^)
{j g.+-%=
-
r^)
+
i
i^+
"^^^j
(79)
The stress-resultants and
stress-couples
at the
edge
vanish with the
exception
of
N,
which
is
equal
to
^pa.
The middle
plane
is bent into the surface
expressed by
the
equation (79),
and the
right-hand
member of this
equation
with its
sign changed
is the deflexion at
any point.
The
comparison
of this result with
(57)
of Article 305 shows
that,
when the
plate
is
thin,
the central deflexion due to
uniformly
distributed load is the same as for a load concen-
trated at the centre and
equal
to
|^(5-|-(r)/(3-(-o-)
of the total distributed load.
The middle
plane
is stretched
uniformly,
and the amount of the extension of
any
linear
element of it is
^(rpjE.
This is half the amount
by
which the middle
plane
would be
309,
310]
APPLIED TO ONE FACE 461
stretched if one face of the
plate
were
supported
on a smooth
rigid plane
and the other
were
subjected
to the
pressure p.
Linear filaments of the
plate
which are at
right angles
to its faces in the unstressed
state do not remain
straight
or normal to the middle
plane.
The curved lines into which
they
are deformed are of the
type expressed by
the
equation
where U is the radial
displacement,
and
Coi ^'ii ^3
are
given by
the formulae
^O
=
*0"^l
Tj
_
_ w_ 2+^:1^
^
Eh? 8
These lines are of the same form as those found in Article 95 for the deformed
shapes
of
the
initially
vertical filaments of a narrow
rectangular
beam bent
by
a vertical load. The
tangents
to these lines cut the surface into which the middle
plane
is bent at an
angle
J,r-|(l+<r)/>r/j?A.
310. Plate bent
by
uniform
pressure
and
clamped
at the
edge.
Let
(u,
V, w)
be the
displacement
of
any point
of the middle
plane.
When the
plate
is
clamped
at the
edge
the conditions which must be satisfied
at the
edge-line
are
u
=
0,
v
=
0,
w
=
0,
5w/Si/
=
0,
(80)
V
denoting
the direction of the normal to the
edge-line.
We seek to
satisfy
these conditions
by
a
synthesis
of the solutions in Articles
301,
303 and 307.
We have
"
=
J
1
^-(n-<7)^Vi(i
+
<^)py]
In these
expressions f
and
t)
are
conjugate
functions of x and
y
which are
related to a
plane
harmonic function
0,,
by
the
equations
dx
dy
"
dy
dx'
and
Xo
is of the form
^x^ +/,
where
/
is a
plane
harmonic function. The
functions

and
/
must be
adjusted
so that u and v vanish at the
edge-
line. One
way
of
satisfying
these conditions is to take

to be constant.
If we
put
we shall have
Xo- ~i
Y^r~
P (^
+
2/")
'
and then u and v vanish for all values of x and
y.
462 PLATE BENT BY PRESSURE
[CH.
XXII
We
may
show that this is the
only way
of
satisfying
the conditions. For this
purpose
we
put
and then we have to show that there is
only
one
way
of
choosing 60, i, r), xo
which will
make U and V take
given
values at a
given boundary.
This is the same
thing
as
showing
that if U and F vanish at the
boundary they
vanish
everywhere.
Since
Vj2;(o
=
6oi
we have
dx
dy
1

o-
\dx 3y
/
'
a|^_a,^_ /0F'_ac[\
dy
dx
^ydx (ly )'
_J_d_/dU 3F\_ il^/3F_3I^\
Q
la-dx\dx dy J
^
dy \dx dy J
'
l

a-dy\dx dy J -8.2;\8ar dy J
jjl^ {r^a-
di\d^'^d^)~^d^\d^-~ ^)]
,

f
1 d
(dU dV\ ,
d /dV
dU\)-\j
,

the
integration being
extended over
any part
of the middle
plane.
When it is extended
over the area within the
edge-line,
and U and V vanish at the
edge-line,
the
integral
can
be transformed into
and we have also
Since
Vi2|=0,
we have
and we have also
It follows that
and this cannot vanish unless
dUdV
^
,dVdU.
5-
-I-
3- =0,
and -=
^
=
0-
ox
cy
ox
oy
It follows that F and U would be
conjugate
functions of x and
y
which vanish at the
edge-line, they
would therefore vanish
everywhere.
The form of w is
given by
the
equation
w
=
>
[(1
+
cr) x/
+
/*=e,
-
I^f^
{00^
+
yj +3^l^p{a?
+
f)\
,
(81)
where
Hj
is a
plane
harmonic function and
^^x^

~
i
i-
-^^J
solution
of the
equation
DV*vi
=

p
can be thrown into this form. To determine w
we have the
equation
and the
boundary
conditions,
viz. :
w
=
and
dwjdv
=
at the
edge-line.
There is
only
one value of w which satisfies these con-
ditions. When w is known
@i
is
given by
the
equation
V.=w=i
.(l-a)@,-l^-j^p(x^
+
f)
+
i^-+''l> ...(82)
and
Xi'
is
given by (81).
310,
311]
APPLIED TO ONE FACE
463
As an
example
we
may
take the case of a circular
plate
of radius a. The deflexion w
is
given by
the
equation
,
'w=
-
A|('-'-^)^
(83)
where r denotes distance from the centre. The central deflexion is one
quarter
of that
which would be
produced by
the same total load concentrated at the centre
(Article 306).
Another
example
is afforded
by
an
elliptic plate*
of which the
boundary
is
given by
the
equation x^la^+y^jb-^l.
It
may
be shown
easily
that
ts(>-^g/(M.-i.)
<)
In the case of the circular
plate equations (82)
and
(83)
show that
61
is
constant,
and
it is therefore convenient to use the solution in the form
given
in Article 302 instead of
Article 303. "We have
w=-i|r2
+
^^X.
+
A(l+<^)/^3{'^'^-Hl-T)r^},
where
Vi'^xi
=
^-
Of
comparing
this form with
(82)
we see that
,.
=
^^-J(l-.)^, ^=-A|(l
+
.)[a-i^].
The
complete expressions
for the
components
of
displacement
are then
given by
the
equations
ozf,
1
+ 0-
, , ,
o-
5
, <r , zh^ , l-2cr
,,~|
where w is
given by (83).
In this case the middle
plane
is bent without extension. Linear
elements of the
plate which,
in the unstressed
state,
are normal to the middle
plane
do
not remain
straight,
nor do
they
out at
right angles
the surface into which the middle
plane
is bent.
311. Plate bent
by uniformly varying pressure
and
clamped
at
the
edge.
We seek to
satisfy
the conditions
(80)
at the
edge-line by
a
synthesis
of
the solutions in Articles
301,
303 and 308. For u and v we have the forms
u
=
^[|-(l
+
^)^g|-i(l
+
<r)i,o2/^],
in which the unknown functions must be chosen so that u and v vanish at
the
edge-line.
We
may
show in the same
way
as in Article 310 that these
conditions cannot be satisfied in more than one
way.
The unknown functions
depend upon
the
shape
of the
edge-line.
*
The result was communicated to the Author
by
Prof. G. H.
Bryan.
1
464 PLATE BENT BY TRANSVERSE FORCES
[CH.
XXII
When this line is a circle or an
ellipse
the conditions
may
be satisfied
by assuming
for
i> ?. 6o> Xo
t^ iorras
where
aj, /3i, yi
are constants. For a circle of radius o we should find
3jPo(l
+
o-) yo(3
+
5cr)
__
aV(l
+
o-) y(,
"~
6-2o-
' "
4(6-2o-)'
^'~
6-2i7
'
and thence u=
-""i^^ tt" ('-'').
v=0.
For an
ellipse given by
the
equation x^la'^+y^Jb^=l
we should find
{
l
+
a){a^
+
mPo
ft ^
K(l
+
3o-)
+
26^(l
+
o-)}po
_ a{l+^)pa^b^
"'"
2a^(l-<r)
+
462
' **'
4
{2a^(l -<7)
+
46'''}
'
^^
2a'
(1- a-)
+
46^'
andthence
^=:^
^#(1 -.)+462 (o-^
+
p-
V'
"=-
In these cases the middle
plane
is
slightly
extended.
Again
the form of w is
given by
the
equation
KV(^
+
2/O^-if^(^
+
32/0],
...(86)
vv
=
^[(l
+
<r)x/
+
/^'e).]-i^
so that w satisfies the
equation
DVi^w
=
poX
and the conditions
w
=
0,
dwjdv
=
0,
at the
edge-line.
These conditions determine w. When w is
known, ,
is
given by
the
equation
'
E
'
D
U^^
+
2/0-^l^]>
(87)
Po^
I
and
;^'
is
given by (86).
For the circle we have
w=
-Tk^('-'-V;
(88)
and for the
ellipse*
we have
"-A('-^-,!;)7cv^i)
<)
312. Plate bent
by
its own
weight.
When the
plane
of the
plate
is
horizontal,
and the
plate
is bent
by
its own
weight,
the solution is to be obtained
by superposing
two
stress-systems.
In
one of these
stress-systems
all the
stress-components except
Z^ vanish,
and
Z^
*
The result was communicated to the Author
by
Prof. G. H.
Bryan.
311-313]
APPROXIMATE THEORY
465
is
gp
(z
+
h),
the axis of z
being
drawn
vertically upwards.
The
corresponding
displacement
is
given by
the
equations
u
=
-a-gp{z+h)x/E,
v
=
-
affp{z
+
h)y/E,
w
=
yp{z''
+ 2hz
+
aix'
+
y')}/E.
(90)
In the second
stress-system
there is
pressure
2gph
on the face z
=
h of the
plate,
and the solution is to be obtained from that in Article 307
by writing
2gph
for
p.
The surface into which the middle
plane
is bent is
expressed by
the
equation
w=^^(.^
+
yO-.V^J-(-^
+
^')(-'
+
2/=-i^y,
(91)
and the
stress-couples
are
given by
the
equations
r m ^
- 24
+ 230- -I- 3<T
,,
G^i
=
-
Z
(,
+
<TK^)
+
gQ
^
_
.
gph",
n n/ X
24
-(- 23<7
-f-
3o-
,,
(x^
=
-D{k,
+
<jk{}
+
g^
,
J
_-
-
gph\
H,=
D(l-a)
T.
The formulat!
(18)
are
approximately
correct when h is small.
To
satisfy
the
boundary
conditions in a
plate
of
any assigned shape,
supported
in
any specified way,
we must
compound
with the solution here
indicated solutions of the
types
discussed in Articles 301 and
303,
and
adjust
the latter solutions so as to
satisfy
these conditions.
313.
Approximate theory
of the
bending
of a
plate by
transverse
forces*.
In all the solutions which we have found the formulae
(18)
of Article 298
are either correct or
approximately
correct. We seem to be
justified
in con-
cluding
that,
in a
plate slightly
bent
by
transverse
forces,
these formulas
may
be taken to
give
a sufficient
approximation
to the
stress-couples.
In a
plate
80 bent the
appropriate equations
of
equilibrium
are
ox
dy
ox
dy
ox
ay
By eliminating
N-^
and
N^
from these we obtain the
equation
dx^
dy'' dxdy
and
by substituting
from
(17)
and
(18)
in this
equation
we find the
equation
DV,*w
=
Z'
(92)
*
For authorities in
regard
to the
approximate theory,
see
Introduction,
pp.
27

29. A
general
justification
on the same lines as that of the
corresponding theory
for rods
(Article 258)
will be
found in Article 329 of
Chapter
xxiv. A
very
elaborate
investigation
of exact solutions for
various distributions of load has been
given by
J.
Dougall, Edinburgh
R. Soc.
Trans.,
vol. 41
(1904).
In this
investigation
the correctness of the
approximate theory
is verified for all cases
of
practical importance.
L. E.
30
466 APPROXIMATE THEORY
[CH.
XXII
The
stress-couples
0,
H at the
edge
are
given
in accordance with
(17)
and
(18) by
the formulae
To find an
expression
for the
shearing
force JV in the direction of the normal
to the
plane
of the
plate
we observe that
and then on
substituting
from
(17)
and
(18)
we find the formula
N=-B^Vi'v, (93)
ov
To determine the normal
displacement
w of the middle
plane
we have
the differential
equation
(92)
and the
boundary
conditions which hold at the
edge
of the
plate.
At a
clamped edge
w and
dw/dv
vanish,
at a
supported
edge
w and G
vanish,
at an
edge
to which
given
forces are
applied
NdHjds
and G have
given
values.
The same differential
equiation
and the same
boundary
conditions would
be obtained
by
the
energy
method
by assuming
the formula
(21)
for the
potential energy
estimated
per
unit of area of the middle
plane*.
In all the solutions which we have found the differential
equation (92)
is
correct whether the formulae
(18)
and
(21)
are
exactly
or
only approximately
correctf.
The solutions that would be obtained
by
the
approximate
method
described in this Article differ from the exact solutions that would be
obtained
by
the methods described in
previous
Articles
only by very
small
amounts
depending
on the small corrections that
ought
to be made in the
formulae
(18)
for the
stress-couples.
In
general
the form of the bent
plate
is
determined with sufficient
approximation by
the method of this Article.
314. Illustrations of the
approximate theory.
(a)
Circular
plate
loaded
symmetrically %
When a circular
plate
of radius a
supports
a load Z'
per
unit of area which is a function
of the distance r from the centre of the
circle, equation (92)
becomes
13
r dr
HHrny^''"-
the direction of the
displacement
w
being
the same as that of the load Z'. We shall
record the results in a series of cases.
*
The
process
of variation is worked out
by
Lord
Rayleigh, Theory of Sound,
215.
+
A more
general
form which includes
(92)
in the
special
cases
previously
discussed is
given
by
J. H.
Michell,
loc. cit.
p.
444.
t
The
general
form of the solution and the
special
solutions
(i)

(iv)
were
given by
Poisson
in his memoir of 1828. See
Introduction,
footnote 36. Solutions
equivalent
to those in
(v)
and
(vi)
were
given by
Saint-Venant in the 'Annotated
Clebsch,'
Note du

45.
313,
314]
OF THE BENDING OF PLATES
467
(i)
When the total load W is distributed
uniformly
and the
plate
is
supported
at
the
edge
(ii)
When the total load W is distributed
uniformly
and the
plate
is
clamped
at
the
edge
(iii)
When the load W is concentrated at the centre and the
plate
is
supported
at
the
edge
W
''8irn\
w
=
.^[-rMog^+i^(a-r2)].
(iv)
When the load W is concentrated at the centre and the
plate
is
clamped
at
the
edge
(v)
When the total load W is distributed
uniformly
round a circle of radius h and
the
plate
is
supported
at the
edge,
w takes different forms
according
as r
>
or < 6.
We find
(vi)
When the total load W is distributed
uniformly
round a circle of radius 6 and
the
plate
is
clamped
at the
edge,
we find
(ft) Application of
the method
of
inversion*.
The solutions
given
in
(iii)
and
(iv)
of
(a),
or in Article
305,
show
that,
in the
neighbourhood
of a
point
where
pressure
P is
applied,
the
displacement
w in the direction
of the
pressure
is of the form
{Pj8nI>)r^logr+i,
where
f
is an
analytic
function of
X and
7/
which has no
singularities
at or near the
point,
and r denotes distance from the
point.
Since w satisfies the
equation Vi''w
=
at all
points
at which there is no load we
may
apply
the method of inversion
explained
in Article 154. Let & be
any point
in the
plane
of the
plate,
P
any point
of the
plate,
P' the
point
inverse to P when (/ is the centre of
inversion, k/, y'
the coordinates of
i",
R' the distance of P' from
0\
w' the function
of
y, rf
into which w is transformed
by
the inversion. Then A'^w' satisfies the
equation Vj'*
(/J"''w')
=
0,
where V,'* denotes the
operator ^-^
+
.s-r;
+
2 -
,- ,
.
ox*
oy* ox^oy'
It is clear
that,
if w and
3w/3i/
vanish at
any bounding curve,
/i'^w' and d
(R'^vf')/dii'
vanish at the transformed
boundary,
v'
denoting
the direction of the normal to this
boundary.

J. H.
Michell,
London Math. Soc.
Proc,
vol. 34
(1902), p.
223.
302
468
APPROXIMATE THEORY
[CH.
XXII
We
apply
this method to the
problem
of a circular
plate clamped
at the
edge
and
loaded at one
point
0. Let 0' be the
inverse
point
of with
respect
to the
circle,
C the centre of the
circle,
and a
its
radius,
also let c be the distance of
from C. The solution for the
plate
O'-^--

[ l____i^ I
clamped
at the
edge
and
supporting
a
load W at C is
W
where r denotes the distance of
any
point
P from C. Now invert from (/
*''8-
70.
with constant of inversion
equal
to
o*/<!*-o'.
The circle inverts into
itself,
C inverts into
0,
P inverts into P' so
that,
if
OP'
=
R and 0'P'
=
R',
we have
r
a^jc
'
^[-r^log"
+
i(a^-r^)],
Hence R'W is
W

r aiP
,
cR' ,
I
.
oiS2\"l
It follows that the
displacement
w of a circular
plate
of radius a
clamped
at the
edge
and
supporting
a load W at a
point
distant c from the centre is
given by
the
equation
^[-iJMog^4'..i(^,ie'-7e^)],
(94)
where R denotes the distance of
any point
of the
plate
from
0,
and R denotes the distance
of the same
point
from the
point 0',
inverse to with
respect
to the circle.
We
may pass
to a limit
by increasing
a
indefinitely.
Then the
plate
is
clamped along
a
straight edge
and is loaded at a
point
0. If 0' is the
optical image
of in the
straight
edge,
the
displacement
in the direction of the load is
given by
the
equation
w=^[-iPlog|'
+
J(iJ'2-/J2)],
(95)
where
/?,
R' denote the distances of
any point
of the
plate
from the
points
and 0'.
The contour lines in these two cases are drawn
by
Michell
iJ,oc. cit.).
(c) Rectangvlar plaie supported
at a
pair of opposite edges.
Let the
origin
be taken at one corner and the axes of x and
y along
two
edges,
let the
other
edges
be
given by .r=2a, y
=
'2.b. We
expand
Z' in the form
_,
_ . rmrx . nirv
^=52Zsm
sm-g/,
where m and n are
integers.
Then a
particular
solution of
equation (92)
is
_
1 QZ'

sin
{mrnel^a)
sin
(nny/ib)
If the
edges
x=0 and x
=
2a are
supported
this solution satisfies the
boundary
con-
ditions at these
edges.
If all four
edges
are
supported
the solution satisfies all the
conditions,
but if the
remaining edges
are not
simply supported
we have to find a solution
314]
VIBRATIONS OF PLATES
469
W2
of tho
equation Vi''w.j
=
so that the sum
W1 + W2
may satisfy
the conditions at
y=0
and
y
=
2b. We assume for
W2
the form*
^r
. TilTTX
W2
=
2lsm-2^,
where
r,
is a function of
y
but not of x. Then
F
satisfies the
equation
and the
complete primitive
is of the form
r=J,cosh
1^+
B
sinh
'^^^
+
y
Mm'
cosh
'^
+
'B'
sinh
^^J
,
where
^^i B^, -Am, B^
arc undetermined constants. These constants can be
adjusted
so
as to
satisfy
the
boundary
conditions at
y=0
and
y
=
2bi.
(d)
Transverse vibrations
of plates.
The
equation
of vibration is obtained at once from
(92) by substituting
for Z' the
expression
-
^ph
-^
. We have
e<w yw d*w
_ 2ph
3% ,
dx*'^
dy*^ dMy^~
D S<2
^ '
When the
plate
vibrates in a normal mode w is of the form Wcos
{pt
+
),
where W is
a function of x and
y
which satisfies the
equation
8*W 8*W

BW
_3p(1-(t2)p'
and the
possible
values of
p
are to be determined
by adapting
the solution of this
equation
to
satisfy
the
boundary
conditions. From the form of the coefficient of W in
the
right-hand
member of this
equation
it
appears
that the
frequencies
are
proportional
to the
thickness,
and
inversely proportional
to the
square
of the linear dimension of the
area within the
edge-line.
The
theory
of those modes of transverse vibration of a circular
plate
in which the
displacement
IT is a function of distance from the centre was made out
by
PoissonJ,
and
the numerical determination of the
frequencies
of the
graver
modes of vibration was
effected
by
him. In this case the
boundary
conditions which he
adopted
become identical
with Kirchhoff's
boundary
conditions because the torsional
couple
H
belonging
to
any
circle concentric with the
edge-line
vanishes. The
general theory
of the transverse
vibrations of a circular
plate
was obtained
subsequently by Kirchhoff,
who
gave
a full
numerical discussion of the results. The
problem
has also been discussed
very fully by
Lord
Rayleighy.
The free vibrations of a
square
or
rectangular plate
have
not,
so
far,

This
step
was
suggested by
M.
L6vy,
Paris C.
P.,
t. 129
(1899).
t
The ease of four
supported edges
is discussed
by
Saint-Venant in the
'
Annotated
Clebsoh,'
Note du

73. A number of cases are worked out
by
E.
Estanave,
'Contribution k I'etade de
r^quilibre ilastique
d'une
plaque...' (These), Paris,
1900. Elastic constants are sometimes
measured
by observing
the central deflexion of a
rectangular plate supported
at two
opposite
edges
and loaded at the centre
;
see A. E. H.
Tutton,
Phil. Trans. R. Soc.
(Ser. A),
vol. 202
(1903).
t
In the memoir of 1828 cited in the
Introduction,
footnote 3G.
J.f.
Math.
(Crelle),
Bd. 40
(1850),
or Ges.
Abhandlungen, p. 237,
or
Vorlesungen
iiber math.
Physik,
Meclianik, Vorlesung
30.
II Theory of
Sound,
vol.
1, Chapter
x.
470 VIBRATIONS OF PLATES
[CH.
XXII
been determined
theoretically.
The case of
elliptic plates
has been considered
by
E. Mathieu* and M.
BartheWmyt.
(e)
Extensional vibrations
of plates.
We
may
in like manner
investigate
those vibrations of a
plate
which involve no
transverse
displacement
of
points
of the middle
plane, by taking
the stress-resultants
Tj, Tj) ^i
to be
given by
the
approximate formulae, [cf. (iv)
of Article
301],
_,
2Eh /9u
,
av\
^
%Eh /3v
,
8u\

Ek /Su dv\
^'=r^^l,si+'^^;' ^^=r3^Va^+'^3SJ' '^'=n:^(,^+aij'
or the
potential energy per
unit of area of the middle
plane
to be
given by
the formula
Eh r/au
8vY
,fau8v ./au
avyn
r^TT^
LV9^
+
^;
~
^
^^
~
""^
ja^ a^-
*
Va^
+
as; IJ

The
equations
of motion are
37,
a5i_
a^u 35.
o7'2_
av
32u^,,, v8'u^,,_
, av
p(l-<r'^)
32u
^
.(97)
i(^-'^)a-^+a7^+i(^+'^)3^=-V-3^-
At a free
edge
the stress-resultants denoted
by T,
S vanish. The form of the
equations
shows that there is a
complete separation
of modes of vibration
involving
transverse
displacement,
or
flexure,
from those
involving displacement
in the
plane
of the
plate,
or
extension,
and that the
frequencies
of the latter modes are
independent
of the
thickness,
while those of the former are
proportional
to the thickness.
The
theory
of the vibrations of
plates
has here been treated in a
provisional
manner.
Detailed discussion of the modes and
frequencies
of transverse vibration
appears
to be
unnecessary,
since
thej'
have been
investigated minutely by
the writers
already
cited.
Some
special
results in
regard
to extensional vibrations will be found in a Note at the end
of this book. A more detailed
investigation
of the
theory
on which the
equations
of
vibration are founded will be
given
in
Chapter
xxiv. See
especially
Article 333.

J. de Math.
(LiouvilU), (S6r. 2),
t. 14
(1869).
+ Mim. de I'Acad. de
Toulouse,
t. 9
(1877).
CHAPTER XXIII.
INEXTENSIONAL DEFORMATION OF CURVED PLATES OR SHELLS.
315. A CURVED
plate
or shell
may
be described
geometrically by
means
of its middle
surface,
its
edge-line,
and its thickness. We shall take the
thickness to be constant and denote it
by
2h,
so that
any
normal to the
middle surface is cut
by
the faces in two
points
distant h from the middle
surface on
opposite
sides of it. We shall
suppose
that the
edge
of the
plate
cuts the middle surface at
right angles;
the curve of intersection is the
edge-line.
The case in which the
plate
or shell is
open,
so that there is an
edge,
is much more
important
than the case of a closed
shell,
because an
open
shell,
or a
plane plate
with an
edge,
can be bent into an
appreciably
different
shape
without
producing
in it strains which are too
large
to be
dealt with
by
the mathematical
theory
of
Elasticity.
The like
possibility
of
large changes
of
shape accompanied by very
small
strains was
recognised
in
Chapter
xviii. as an essential feature of the
behaviour of a thin rod
;
but there is an
important
difference between the
theory
of rods and that of
plates arising
from a certain
geometrical
restriction.
The extension of
any
linear element of the middle surface of a strained
plate
or
shell,
like the extension of the central-line of a strained
rod,
must be small.
In the case of a rod this condition does not restrict in
any way
the
shape
of the
strained central-line
;
and this
shape may
be
determined,
as in
Chapters
XIX.
and
XXI.,
by taking
the central-line to be unextended.
But,
in the case of the
shell,
the condition that no line on the middle surface is altered in
length
restricts the strained middle surface to a certain
family
of
surfaces,
viz. those
which are
applicable upon
the unstrained middle surface*. In the
particular
case of a
plane plate,
the strained middle surface
must,
if the
displacement
is
inextensional,
be a
developable
surface. Since the middle surface can
undergo
but a
slight
extension,
the strained middle surface can differ but
slightly
from one of the surfaces
applicable upon
the unstrained middle
surface
;
in other
words,
it must be derivable from such a surface
by
a
displacement
which is
everywhere
small.
*
For the literature of the
theory
of surfaces
applicable
one on another we
may
refer to the
Article
by
A.
Voss,
'
Abbildung
und
Abwickelung
zweier Flachen auf einander
'
in
Ency.
d. viath.
Wiss.,
III. D 6a.
472 CURVATURE OF A THIN SHELL
[CH.
XXIII
316.
Change
of curvature in inextensional deformation.
We
begin
with the case in which the middle surface is deformed without
extension
by
a
displacement
which is
everywhere
small. Let the
equations
of the lines of curvature of the unstrained surface be
expressed
in the forms
a
=
const, and
/8
=
const.,
where a and
/9
are functions of
position
on the
surface,
and let
R^, R^
denote the
principal
radii of curvature of the surface
at a
point,
ii,
being
the radius of curvature of that section drawn
through
the
normal at the
point
which contains the
tangent
at the
point
to a curve
of the
family
/3
(along
which a is
variable).
When the shell is strained
without extension of the middle
surface,
the curves a
=
const, and
/3
=
const,
become two families of curves drawn on the strained middle
surface,
which
cut at
right angles,
but are not in
general
lines of curvature of the deformed
surface. The curvature of this surface can be determined
by
its
principal
radii of
curvature,
and
by
the
angles
at which its lines of curvature cut the
curves a and
^8.
Let
-jt
+
^^d
^^^
t5"
+
^
"5"
t>e the new
principal
curvatures
til -f^l -tii ^2
at
any point.
Since the surface is bent without
stretching,
the measure of
curvature is
unaltered*,
or we have
[Ri'^^RJ KR.'^^RJ^RJi,'
or,
correctly
to the first order in 8
-0-
and S
-j,-
,
ill R2
i'hi'k='
('>
Again
let
yjr
be the
angle
at which the line of curvature associated with
the
principal
curvature
-5-
+
8
-5-
cuts the curve
/3
=
const, on the deformed
Ml Hi
surface,
and let
ii,', R^
be the radii of curvature of normal sections of this
surface drawn
through
the
tangents
to the curves
/9
=
const, and a
=
const.
In
general y^
must be
small,
and
R^, iJ/
cau differ but little from
R^, R^.
The indicatrix of the
surface,
referred to axes of x and
y
which coincide with
these
tangents,
is
given by
the
equation
a? xP
/
1 1 \
^/
+
-^
+
a;y
tan
2i/r
(
-^,
-
^\
=
const.
Referred to axes of
f
and
t)
which coincide with the
tangents
to the lines
of
curvature,
the
equation
of the indicatrix is
^'(i
+
^i)
+
''^(4
+
4)
=
--*'
*
The theorem is due to
Gauss, 'Disquisitiones generales
circa
superficies
curvas,' Gottingen
Comm.
Bee,
t. 6
(1828),
or
Werke,
Bd.
4, p.
217. Cf.
Salmon,
Geovietry of
three
dimensiotis,
4th
edition,
p.
355.
316]
SLIGHTLY BENT WITHOUT EXTENSION 473
and therefore we have
T?
' *
T>
'
I?
'*'
D '7> 'I?
*
til li^ Jtti li^ -^1 -^2
h-(2)
The
bending
of the surface is determined
by
the three
quantities /c,, k
t
defined
by
the
equations
The curvature
1/-R'
of the normal section drawn
through
that
tangent
line of the strained middle surface which makes an
angle
w with the curve
y8
=
const, is
given by
the
equation
1 cos* ft) sin* ft)
.
-n>
=
p/
H

d7-
+
''r sin ft) cos
ft),
and the curvature
1/R
of the
corresponding
normal section of the unstrained
middle surface is
given by
the
equation
1
_
cos* ft) sin* ft)
R
Ri -Rj
so that the
change
of curvature in this normal section is
given by
the
equation
^i

-g
=
i
cos* ft)
+
Ka
sin* ft)
+
2t sin ft) cos ft)
(4)
We shall refer to
k^, k^,
t as the
changes of
curvature.
In
general,
if
R^ ^ R.^,
equations (2) give, correctly
to the first
order,
1^
1,
O
p-

/Cji
P
""
I?
'
jfl Xtj JtVj Jti
For
example,
in the case of a
cylinder,
or
any developable
surface,
if the
lines
/8
=
const, are the
generators,
Ki vanishes,
and tan
2-<fr

2x723.
The case of a
sphere
is somewhat
exceptional
because of the indeterminate-
ness of the lines of curvature. In this
case,
putting
i2i
=
i?2,
we find from
(1)
and then we
have, correctly
to the first
order,
/Cj
+
/tj
=
0,
tan
21/r
=
2t/(/c,

/Co)
=
t/zci
,
and,
correctly
to the second
order,
8
pj
=
-
iK.2
+
T*
=
Ki*
+
T*,
474 STRAIN IN A THIN SHELL
[CH.
XXIII
but
,
and
k^
are not
equal
to 5
^j-
and S
^j-
unless t
=
0,
and -vir is not small
unless T is small
compared
with
k^.
The result
that,
in the case of a
cylinder slightly
deformed without
extension, ki=0,
or there is no
change
of curvature in normal sections
containing
the
generators,
has been
noted
by
Lord
Rayleigh
as "the
principle upon
which metal is
corrugated."
He has also
applied
the result
expressed
here as
k^IR.^
+
k^jRi
=
to the
explanation
of the behaviour of
Bourdon's
gauge*.
317.
Typical
flexural strain.
We
imagine
a state of strain in the shell which is such
that,
while no line
on the middle surface is altered in
length,
the linear elements
initially
normal
to the unstrained middle surface remain
straight,
become normal to the
strained middle
surface,
and suffer no extension or contraction. We
express
the
components
of strain in this state with reference to axes of
x,
y,
z,
which
are directed
along
the
tangents
to the curves
/3
and a at a
point
Pi
on the
strained middle surface and the normal to this surface at
Pi.
Let P be the
point
of the unstrained middle surface of which
Pj
is the
displaced position,
and let hs be an element of arc of a curve
s,
drawn on the unstrained
surface,
and
issuing
from
P;
also let R be the radius of curvature of the normal
section of this surface drawn
through
the
tangent
to s at P. The normals
to the middle surface at
points
of s meet a surface
parallel
to the middle
surface,
and at a small distance z from
it,
in a
corresponding
curve,
and the
length
of the
corresponding
element of arc of this curve is
approximately
equal
to
{{R

z)jR}hs-f.
When the surface is bent so that R is
changed
into
R',
and z and Ss are
unaltered,
this
length
becomes
[{R'

z)jR']
Bs
approximately.
Hence the
extension^
of the element in
question
is
/R'-z
R-z\/R-z
.
^ , /I 1\
[~R'
W)/-R-'
^''
^PPro^i'^tely,
-z{^-^,-^).
Let the
tangent
to s at P cut the curve
yS
at P at an
angle
w. The
direction of the
corresponding
curve on the
parallel
surface is
nearly
the
same
;
and the extension of the element of arc of this curve can be ex-
pressed
as
Bxx
cos' w +
Cyy
sin'' (o
+
exy
sin eu cos w.
Equating
the two
expressions
for this
extension,
and
using (4),
we find
exx
cos^ Q)
+
Byy
sin' m
+
e^y
sin to cos to
=

^
(ki
cos^ (o +
k^
sin' <b
+ 2t sin eo cos
w),
and therefore
e^x
=

zki,
eyy
=

zk^,
exy
=

2zt.
*
Proc. R.
Soc,
vol. 45
(1889), p. 105,
or
Scientific Papers,
vol.
3, p.
217.
+ Near a
point
on the middle surface the
equation
of this surface can be taken to be
2l'=^IM^+ifjIi2,
and the coordinates of the
point
in which the normal at
(|', r/')
meets the
parallel
surface can be
shown, by forming
the
equations
of the
normal,
to be
approximately
^ (1 -2/7Ji)
and
7)'(1 -/iJ). Putting |'
=
S.
cosw, n'
=
Ss . sin
u,
and
neglecting z'^jB^'
and
z'^/R/,
we obtain the result stated in the text.
J
Cf. Lord
Kayleigh, Theory of Sound,
2nd
edition, p.
411.
316,
317]
SLIGHTLY BENT WITHOUT EXTENSION 475
In the
imagined
state of strain
e^x,
Syz,
e
vanish. With this strain we
may compound any
strain
by
which the linear elements
initially
normal
to the unstrained middle surface become
extended,
or
curved,
or inclined
to the strained middle surface. The most
important
Case is that in which
there is no traction on
any
surface
parallel
to the middle surface. In this
case the
stress-components
denoted
by
X^, Yz, Z^ vanish,
and the strain-
components
e^,
Syz,
ezz
are
given by
the
equations
Bzx
=
0,
ej,z
=
0, gj^
=
-
(o-/(l
-
g)\ (e
+
Byy),
where cr is Poisson's ratio for the
material,
supposed isotropic.
In this state
of strain the linear elements
initially
normal to the unstrained middle surface
remain
straight,
become normal to the strained middle
surface,
and suffer
a certain extension
specified by
the value of
e^,
written above. It is clear
that this extension can have
very
little effect* in
modifying
the
expressions
for
Bxx,
eyy, Cxy,
and we
may
therefore take as
approximate expressions
for the
strain-components
exx
=
-ZKi,
eyy
=
ZK2, e
=
Y-
0(;i-|-/co), ea:y
=
-2TZ, e2t
=
ej,j
=
0.
...(5)
This state of strain
may
be described as the
typical flexural
strain.
The
corresponding stress-components
are
E E
Xx
= -
,
2
Z
{Ki
+
<TKi),
Yy
= -
z
(Kj
+
0-/Ci),
1

(7 1

O"
Xy=^~^zT,
X,= Y,
=
Z,
=
0,
where E is
Young's
modulus for the material. The
strain-energy-function
takes the form
i
fl^.
K*'
+
^^)'
-
2
(1
-
<7) (,,
-
T)].

The
potential energy of bending,
estimated
per
unit of area of the middle
surface,
is obtained
by integrating
this
expression
with
respect
to z between
the limits
-
h and
/;,
the thickness of the shell
being
2h. The result can be
written
^D[(k,
+
K,y-2il
-
<t)(>c,>c,-t%
(6)
where D is the "flexural
rigidity" f^VCl
-o-=).
In the case of a
cylinder,
or
any developable
surface,
this
expression
becomes
^
D
{k2^
+
2(1-<t)t^}.
In the case of a
sphere
it becomes
| fih" (k,'
+
r),
or
f/x/t"
[S
-^j
,
where
fj,
is
the
rigidity
of the material
f.
*
It will be seen iu the more
complete
investigation
of Article 327 below that such effects are
not
entirely negligible.
t
These are the
expressions
used
by
Lord
Rayleigh, Theory of
Sottnd,
2nd edition, Chapter
x a.
476 CALCULATION OF THE CHANGE OF CURVATURE
[CH.
XXIII
318. Method of
calculating
the
changes
of curvature.
The conditions which must be satisfied
by
the
displacement
in order
that the middle surface
may
suffer no extension
may
be fouud
by
a
straight-
forward method. Let ABa be the element of arc of a curve
/3
=
const, between
two curves a and a +
Ba, BB^
the element of arc of a curve a
=
const, between
two curves and
/3
+
8/8 ;
also let
x',
y',
z' be the coordinates of a
point
on
the strained middle surface referred to
any
suitable axes. We form
expres-
sions for x
,
y',
z in terms of the coordinates of the
point
before strain and
of
any
suitable
components
of
displacement.
Since curves on the middle
surface retain their
lengths,
and cut at the same
angles
after strain as before
strain,
we must have
^[(ST-(IT-T-. I[(l)"-(l) m
=
1,
dx dx'
dy' dtf
dz dz'
da
d/S
da
d/S
da
9yS
These
equations give
us three
partial
differential
equations connecting
the
components
of
displacement.
The
changes
of curvature also
may
be calculated
by
a
fairly straightforward
method. The direction cosines
I, m,
n of the normal drawn in a
specified
sense to the strained middle surface can be
expressed
in such forms as
-AB\dad^ dad^J'
and the
ambiguous sign
can
always
be determined. The
equations
of the
normal are
x

x'
y

y'
z

z'
,
L m n
'
and,
if
{x,
y, z)
is a centre of
principal
curvature,
we have
x
=
x'
+
Ip
,
y
=
y'-\- mp',
z
=
z'
+
np,
where
p
is the
corresponding principal
radius of curvature
;
p
is estimated as
positive
when the normal
{I,
m, n)
is drawn from
(x', y', z')
towards
(x, y,
z).
If
(a
+
Ba, ^
+
B^)
is a
point
on the surface near to
{x, y',
z')
on that line of
curvature
through (x', y', z')
for which the radius of curvature is
p',
the
quantities
x,
y,
z,
p
are
unaltered,
to the first order in
Ba, S/3,
by changing
a
into a -f- 8a and
/3
into
/3
-I-
Sy9.
The
quantity
we have
already
called tan
i/r
is one of the two values of the ratio
BB^/ABa.
Hence tan
yjr
and
p'
are
determined
by
the
equations
|8..|a..,-g8.+|s/,)
=
o,
^-^^p^i'-^^)-"-
I
318,
319]
THIN CYLINDRICAL SHELL
477
These three
equations
are
really equivalent
to
only
two,
for it follows
from the mode of formation of the
expressions
for
I, m,
n,
and from the
equation
I- +
in'
+
ri^
=
I
, that,
when we
multiply
the left-hand members
by
I, m,
n and add the
results,
the sum vanishes
identically. By
elimin-
ating
the ratio
Sa/^/3
from two of these
equations
we form an
equation
for
p,
and the values of
Ijp'
are
^
+
B^
and
o+^-p
I
^y eliminating p'
from two of the
equations
we form an
equation
for
8/9/Sa,
which determines
tan
i|r.
We shall
exemplify
these methods in the cases of
cylindrical
and
spherical
shells. In
more difficult
cases,
or when there is extension as well as
change
of
curvature,
it is advisable
to use a more
powerful
method. One such method will be
given
later
;
others have been
given by
H. Lamb* and Lord
Rayleight.
The results for
cylindrical
and
spherical
shells
may,
of
course,
be obtained
by
the
general
methods
;
but these cases are so
important
that
it seems to be worth while to show how
they may
be
investigated by
an
analysis
which
presents
no difficulties
beyond
the
manipulation
of some rather
long expressions.
The
results in these cases were obtained
by
Lord
RayleighJ.
319. Inextensional deformation of a
cylindrical
shell.
(a)
Formuhe
for
the
displacement.
When the middle surface is a circular
cylinder
of radius
a,
we take the
quantities
a and
yS
at
any point
to be
respectively
the distance
along
the
generator
drawn
through
the
point,
measured from a fixed circular
section,
and the
angle
between the axial
plane
containing
the
point
and a fixed axial
plane
;
and we write x and
<^
in
place
of
a and
/8.
We resolve the
displacement
of the
point
into
components
: u
along
the
generator,
v
along
the
tangent
to the
circular
section,
w
along
the normal to the
surface drawn inwards. The coordinates
x',
y',
z of the
corresponding point
on the
strained middle surface are
given by
the
equations
x'
=
x-^ u,
y'=(a

w)
cos
(^

t) sin
^,
z'
=
{a

w)
sin
<^
+ v cos
<p.
The conditions that the
displacement may
be inextensional are
x,x
Fig.
71.
m^S'*
fix)
=
1, =1,
dx
9^
dx
d<j>
dx
dtf)
*
London Math. Soc.
Proc,
vol. 21
(1891), p.
119.
t
Theory of
Soimd,
2nd
edition,
vol.
1, Chapter
xa.
t
London Math. Soc.
Proc,
vol. 13
(1882),
or
Scientific Papers,
vol.
1, p.
551,
and the
paper
cited on
p.
474
supra.
See also
Theory of Sound,
2nd
edition,
vol.
1, Chapter
x a.
478
INEXTENSIONAL DEFORMATION OF A
[CH.
XXIII
On
writing
down the
equations
dx'
,
du
dy'
dw . dv .
,
dz' dw .
,
dv
,
:
=
1
+

, ^
=
- -
cos A
-
^
sin
d),
^
=
-
^
sm
^
+
^
cos
^,
dx dx dx dx
^
dx
^
dx dx
^
dx
^
dx' du
dif fdw
\
,
( , ^v\
.
we see that these conditions
are,
to the first order in
u, v, w,
9it _ dv dv Idu
,,
dx
d(f)
dx
ad(f>
These
equations
show that u is
independent
of
x,
and v and w are linear
functions of x.
If the
edge-line
consists of two circles x
=
const., u, v,
w must be
periodic
in
(^
with
period
2-77,
and the most
general possible
forms are
M
=
-
2
-
S
sin
(n</)
+
/9),
t)
=
2
[
J

cos
{nj>
+
a)
+
Bx
cos
{n<f>
+
/3)],
w
=

2 n
[An
sin
(n^
+
a)
+ BnX
sin
(n<f)
+
/3)],
(8)
where
A^, , , /3
are
constants,
and the summations refer to different
integral
values of n.
(b) Changes of
curvature.
The direction cosines
I, m,
n of the normal to the strained middle surface
drawn inwards are
^
^
1
/sy
a/
_
a^'
ay\
a
\dx
d<fi
dx
d<j}J
'
We write down the values of
dx'jdx,
...
simplified by using (7),
in the forms
dx'
^ dy'
1 9 .
,
dw
,
dz' \du
,
dw .
,
^-
=
1,
^
=
-
.r-
sin (i
-
^-
cos
<f),
5-
=
^
cos d)
-
^
sin
<f>,
dx dx
adcj)
^
dx
^
dx
adq>
^
ox
dx' du
dy'

^ / ,
SwN
.
^
dz'
. / , 9wV
.
,
and we
find,
to the first order in
u, v, w,
,
dw
l
=
-T-,
m
dx
1
I
/
dw\ .
,
.
,
1 /
dw\
.
=
-oos<f>
+
-{v
+
^)sm<l>,
n
=
-sin<j>--[v
+
^jcos<f,.
The
principal
radii of curvature and the directions of the lines of curva-
ture are
given by
the
equations
j^
/a*'
8y'
_ a^' 3'\
i/a'am
aza^'_amaa;'_ay'az\
p'
\dx
d(f)
dx
d<^J p
\dx
d<l>
dx
dtf)
dx
d^
dx
d^J
dl dm dm dl
_
^
1
dx
d<f>
dx
d<l>
319,
320]
and
THIN CYLINDRICAL SHELL
^
\dx
ox ox ox!
^
\d<j> d(f) o<f> d<p
5j ^
fdx' dm dm dx'
_
dy'
dl dl
dy'\
_
^
\dx
dd) dx dd) dx
d<f>
dx
dd>J
479
9<^
dx
d<p
dx
d<f>
dx
d<f>)
For the
purpose
of
calculating
the coefficients in these
equations
we write
down the values of
dl/dx,
...,
simplifying
them
slightly by
means of
(7)
and
by
the observation that v and w are linear functions of x. We have
dl
=
0,
dw\
a
dx\
d^J
'
dm sin
^
d
dx
d*w dm .
.
f,
1
fd^w \) .
cos (6/
dw\
dx
d(}> dxd<j)
'
d<f>
We know beforehand
that,
when terms of the second order in
u, v,
w are
neglected,
one value of
1/p'
is zero and the other is
Ija
+
K^;
also the value
of
aB(f>ISx
is tan
yfr,
and tan
2fjr
=

2aT. We can now write down the above
equations
for
p'
and
BxlB<f>
in the forms
(correct
to the first order in
u, v, w)
(l
+
,jj^_sin<^-(
+
^-^)cos<^^
"
.
,
1
/
dw\ . 1 /0%
\
.
."1
+
[^sin<^
+
-(.
+
g^jcos,^
+
-(^^
+
wjsm<^J
=
0,
,19/
dw\ .
,
1
,
, /9w ^^\ J.
+
-
tan <r
a
.
,
1
/
dw\ . 1
(d-w
\ .
'
d^J
^
a
V9^'
The former of these
gives,
to the first order in
u, v, w,
_
1 /d^w
"'"a'Kd
and the latter
gives,
to the same
order.
W^"")'
=
0.
.(9)
tan2V.=
-2i(.
+
|^),
or
1 9 /
a 9a; V
dw
'd~4
.(10)
With the values of
, v,
w
given
in
(8)
these results become
.(11)
2
'^^'~ [An
sin
{n<p
+
)
+
BnX
sin
{n(f>
+
^)],
rt-
^
T
=
-
S Bn
cos
(n</>
+
^).
a
320. Inextensional
deformation
of a
spherical
shell.
(a)
Formulce
for
the
displacement.
When the middle surface is a
sphere
of radius a we take the coordinates
480 INEXTENSIONAL DEFORMATION
[CH.
XXIII
a and
yS
to be
ordinary spherical polar
coordinates,
and write
6,
<^
for
a, /9.
The
displacement
is
specified by
components
u
along
the
tangent
to
the meridian in the direction of
increase of
6,
v
along
the
tangent
to the
parallel
in the direction of
increase of
<^,
w
along
the normal
to the surface drawn inwards. The
Cartesian coordinates of a
point
on
the strained middle surface are
given
by
the
equations
x
={a

w)
sin 6 cos
<^
+
u, cos 6 cos
(^

v sin
^,
y'
=
{a

w)
sin 9 sin
<^
+
u cos sin
(f)
+ V cos
<f>,
/
=
(a

w)
cos

u sin 0.
The conditions that the
displacement may
be inextensional are
a
l\d0)
"*"
\d0)
^
\d0j
J
'
a sin
l\d(tij
^
{d4>)
"^
\d<t>J ]
~
'
dx' dx'
"dy dy
dz' dz'
_
.
d0df'^d0d^'^d0d^~
We write down the
equations
Fig.
72.
fa
-
w +
^j
cos ^
-
(^
+
uj
sin cos
<^
-
^^
sin
</>,
(a-w +
^j cos0-(^
+
M
j
sin ^ sin
^
+
^^
cos
^,
d0~
d0
dz_
d0
and
9^'
ay
d<}>
dz' du .
a
dw
n
The conditions that the
displacement may
be inextensional
are,
to the first
order in
u, v, w,
du
.
a /J ,
S^
m;
=
^
,
w sm ^
=
M cos t/
+
g^
,
w)
sin ^ +
M cos ^
+
^
sin
</>
9i>'
-
=
(a

w)
sin ^
+
M cos ^ +
^-r
cos
+
^^
COS ^
-

^-r
sm ^
~du

dw . S
^-7 cos ^

i;

^-r
sm d'
cos<^,
sin^.
,dv
sin ^
^
+
cos (/
[ 5^
"(r-'r'-")^""'^-*""""'-?
"'")''
320]
OF A THIN SPHERICAL SHELL
481
or,
as
they may
be written,
9m
w- ^'"^
de sine
~d<f>
sine
'
d<j>^^'^^^'^^de8^
0.
...(12)
u
=
sin 2
i;
=
sin
S
w
=
2
.(13)
The last two of these
equations
show that
w/sin
and
?;/8in
d are
conjugate
functions of
log (tan ^0)
and
<j).
If the
edge-line
consists of two circles of
latitude, u, v,
w must be
periodic
in
<f)
with
period
27r,
and the most
general possible
forms for them are
e 6
An
tan"
-
cos
(nc/)
+
a)
+

cot"
-
cos
(n<^
+
/3)
6 e
An
tan"
-
sin
{nj>
+
a)
-
5
cot"
^
sin
(n<^
+
/9)
a
(/I
+
cos
6)
An
tan"
-
cos
(n<^
+
a)
it
6 "I

{n

cos
6)
Bn
cot"
^
cos
(n<^
+
;8)
,
where
An, Bn, . ^n
are
constants,
and the summations refer to different
integral
values of n.
If in the formulae
(12)
we
put
n=0 we find
displacements
of the
type
M=
^osindcosa, D=Josinflsina, w=A,^co&6cosa,
the terms in B
being
of the same
type.
The
components
of this
displacement
in the
directions of
n/,
y',
/ are
-
.^0
sin a sin 5 sin
<^,
^o
sin a sin 5 cos
<^, AqCOSo,
and this
displacement
is
compounded
of a translation

JqCoso
in the direction of the
axis of z' and a rotation
Af^a'^
sin a about this axis.
If in the formulae
(12)
we
put
n
=
1,
we find
displacements
of the
types
u
=
Ai{\-coa6)<X)a{<^
+
a), =J,(1

cos5)8in(0-l-a), w=^i8inflcos((^-|-a),
and
M
=
,(l+cos(9)cos(<^+/3), >=-5i(l+cosd)sin(<^+^),
w=
-Z!isinflcos(<^+/3).
The former is
equivalent
to a translation
(

^jcosa, ilisina, 0)
and a rotation
^,a~'(sina, cosa, 0);
and the latter is
equivalent
to a translation
(^icosft -^isinftO)
and a rotation
Bia~'^{am^, cosft 0).
It
appears
from what has
just
been said that all the
displacements
obtained from
(13) by putting
w
=
or 1 are
possible
in a
rigid body,
and the
terms for which n has these values
may
be omitted from the summations.
Similar results can be
proved
in the case of
cylindrical
shells.
If the
edge-line
consists of one circle of
latitude,
and the
pole
^
=
is
included,
we must omit from
(13)
the terms in
cot"^^, (
>
1),
for these
terms become infinite at the
pole.
If the
sphere
is
complete
the terms in
tan"^^, (n>l),
must be omitted
also;
that is to
say
no inextensional
L. E.
31
482 INEXTENSIONAL DEFORMATION
[CH.
XXIII
displacements
are
possible
in a
complete spherical
shell
except
such as are
possible
in a
rigid body*.
(b) Changes of
curvature.
We form next
expressions
for the direction cosines
I, m,
n of the normal
to the deformed
surface,
by
means of such formulae as
~
a^sme\d<bde
d<i> dd
and for this
purpose
we first write down the
expressions
for
dx'jdd,
. . .
simplified
by
means of
equations (12).
We have
dx' -
(dw , \

-J
.
3
.
^
=
a cos cos

I
^n
+
w
)
sin cos
9

^
sm
<p.
dy n , /Sw \ n ,
Sw
,
^
=
acos
t/sin^
I
^
+
Mj
sin ^sin
9
+
^cos^,
and
^
=

a sin t' sin + I
5-7
cos y

w

^
sin el cos
9,
^
=
a sin 6 cos
<^
+
( ^
cos 6

v

i^&in6\
sin
^,
dz' du .
n
dw
n
Hence we
have,
to the first order in
u, v, w,

n ,
1
/aw
\
n .
1
/
1 awN .
,
{
=

sin 6/ COS d>
hwf
+
w cos d cos
9
+
-
w
+
-.
^
;
-
sin
9,
^
a
\30 J
^
a\
sm.6
d<fij
^
.... 1 /aw
,
\
/]

.
1
/
,
1 awN
,
??i
=
sin a sin
9
U-n
+
m cos t/ sin

[v
+
-.

^
;r- cos
rf),
^
a
Vaa /
^
a
\
sin d'
d<f>J
^
n
=

cosO+-(^
+ u]smd.
Exactly
as in the ca.se of the
cylinder,
the
principal
curvatures and the
directions of the lines of curvature are determined
by
the
compatible
equations
*
The result is in accordance with the theorem that a closed surface cannot be bent without
stretching.
This theorem is due to J. H.
Jellett,
Dublin Trans. li. Irish
Acad.,
vol. 22
(1855).
320]
OF A THIN SPHERICAL SHELL
483
and we therefore write down the
following equations,
in which we
put
for
shortness X
=
-
(x^ +m),
Y=-(v + -.

7,^^-7],
a
\dO J a\
sin
d6j
and
dl
dO'
dm
W'
dn
dd''
m
d<j>"
dm
dx
dY .
I
+
-^
1 cos ^ cos
(^
+
X sin ^ cos
<^
+
j^
sin
<f),
dd
1
+
x^j
COS ^ sin
^
+ Z sin ^ sin
^

^-
cos
<f>,
1
+
^j
sin ^ + Z cos
0,
sin ^
+
2'' cos ^
+ ;-7-
)
sin 6
dX
d4>
cos 6
Y]
cos
<^,
dY
sin ^
+
A'' cos ^
+
^^ j
cos
-
(
~
cos
-
F
j
sin
^,
(dX
dn . .dX
a^
=
^^'^%-
Our
procedure
in this case must be a little different from that
adopted
in the case of the
cylinder
because,
to the first
order,
the sum and
product
of the
principal
curvatures are unaltered
by
the strain. We therefore
begin
by finding
the
equation
for
tani^,
or sin
OB^/Bd.
This
equation may
be
written
/dx' tan
ilr
da/\
fdm.
tan
i/r
dm\
_
idy'
tan
-^ dy'\
/dl
tan
^Jr
dl
[W
"*
"sST^
d^J
[W
*"
"smT
d^)~[de'^
"sm"^
3^7
[w
^inJ
3^,
)
and,
by
direct substitution of the values written above for
dx'jdd,
...,
it is
found to be
( -.^

a
TT^ I
cos ^
H

-.

^
a
(
sin ^
"^

X cos ^

^)
cos 6
\d6
do J sin p
V
,az
dd
dY
dcj>
-SI{''(lf''-^)-(r-"1"^)}''*=-
Now we have
dv_ dY__d
/ 1 dw
de "'de~ dd Isin^
d4>.
)
a
[-^
cos 6
')-
^^cos^-.-^g^sin^)
sin cos 6
dd
Vsini
dw\
e
d^)
'
/.
^dX

n
dY\
. ./d'w du\
aC^ ^ \
^" 1 ^'^
d^w \ /
I
=
sin 6
(d^w
\ 9-m;
,
.
/,9w ,
]
where,
in the last
line,
use has been made of the
equations (12).
But,
since
312
484 INEXTENSIONAL
DEFORMATION OF A SPHERICAL SHELL
[CH.
XXIII
v)
=
dujdd,
and u satisfies the
equation
obtained
by eliminating
v fiom the
second and third of
(12),
viz.
1 f^w
^dw
Id/,
.^d^u ,
.
^ ^du \

it follows that
1 3'K
o'w
,8w
+
w
-
w.
Hence the
equation
for tan
i|r
becomes
8
/
1
^^/;^
tan
2y
=
;
dO
Vsin
6>
8^
One of the
equations
for
determining p
is
re'^-^'^-p'il'e''-^'^)-''
or
f
1
+
^Vin
^
+
^ cos ^ +
If
tan
^r
a \
off
J 00
^'
sin^+Zcos^ +
i(|^
+
cot0~]tani/r
=
l+9:J
+
t^H-lf|'+cot^
a^
,3w\
sin 6^
(9<^
a
\d(j> d<pJ.
I
)
sec
2^^.
But,
using
the notation of Article
316,
we have
-.

=
,
cos' ylr
+ K.i
sin' -Jr
+
t sin
2>^
p
a
It follows that
=
i (cos 2yjf
+
sin
2-v^
tan
2y}f)
=
/fj
sec
2-\^.
_ 1^
/aw
"'
~
tt'
la^
+
tv
iif-
1 3w'
"
a' S^ Vsin
a<^
With the values of
u, v,
w
given
in
(13)
we now find
e
2
Ki


fC^

2^
=
-S
n^

n
320,
321]
INEXTENSIONAL
VIBRATIONS OF A THIN SHELL 485
321.
Inextensional vibrations.
If we assume that the state of strain in a
vibrating
shell is that which
has been described in Article 317 as the
typical
flexnral
strain,
we
may
calculate the
frequency
of vibration
by
forming expressions
for the kinetic
and
potential
energies*.
We illustrate this method in the cases of
cylindrical
and
spherical
shells.
(i) Cylindrical
shell.
The kinetic
energy,
estimated
per
unit of area of the middle
surface,
is
where
p
is the
density
of the
material,
and
u, v,
w are
given by (8),
in which
the coefficients
An, B^
are to be
regarded
as functions of t. The kinetic
energy
T of the
vibrating
shell is obtained
by
integrating
this
expression
over the area of the middle surface. If the ends of the shell are
given by
x=
l,^e
find
T^2Trpalh'Z
<'-")('i=)'M^--'>"i(t)"
...(16)
The
potential energy
of
bending,
estimated
per
unit of area of the middle
surface,
is
Ji)[,^+2(l-o-)T=],
where
k^
and t are
given by (11).
The
potential energy
V of the
vibrating
shell is obtained
by integrating
this
expression
over the area of the middle
surface. We find
F=
DttZS
iHLSlll
[n'An"
+
{inH^+
2
(1
-
<r) a') B^^] (17)
The coefficients
An, Bn
in the
expressions (8)
for the
displacement may
be
regarded
as
generalized
coordinates,
and the
expressions
for T and V show
that
they
are
''
principal
coordinates,"
so that the various modes of vibration
specified by
different A's or B's are executed
independently
of each other.
The vibrations in which all the B's and all but one of the A's vanish are two-
dimensional and take
place
in
planes
at
right angles
to the axis of the
cylinder.
The
type
is
expressed by
the
equations
-M
=
0,
V
=
An
cos
nj),
w
=

nAn
sin
n<l>,
in which
A
is
proportional
to a
simple
harmonic function of the time with
a
period 2Tr/p,
and
p
is
given by
the
equation
P
~2pha*
if+l
~~'ip(\-a^)a*
n^
+
1
^ '
*
The
theory
of inextensional vibrations is due to Lord
Bayleigh,
London Math. Soc.
Proc.,
vol. 13
(1881),
or
Scientific Papers,
vol.
1, p. 551,
and Proc. R.
Soc,
vol. 45
(1889),
p. 105,
or
Scientific Papers,
vol.
3, p.
217. See also
Theory of
Sound,
second edition. Chapter
x a. A
discussion of the conditions for the existence of
practically
inextensional modes of vibration will
be
given
in
Chapter
xxiv.
infra.
486
IN EXTENSIONAL VIBRATIONS
[CH.
XXIII
The vibrations in which all the A's and all but one of the B's vanish are
three-dimensional. The
type
is
expressed by
the
equations
u
=

sin
nd),
v
=
xBn
cos
n<h,
w=

nxB^
sin
nd),
n
and the
frequency pl^ir
is
given by
the
equation
s _
Eh'
vr(ji^j-\
y
1 4- 6
(1
-
(t) aVn'P
, ,
^
"3/3(1
-o-'')a*
n^+1
\
+
Za'ln^{v?+\)L^
^ '
If either n or
Ija
is at all
large
the two values of
p belonging
to the same
value of n are
nearly equal.
(ii) Spherical
shell.
We shall
suppose
the middle surface to be bounded
by
a circle of latitude
d
=
a,
and that the
pole
^
=
is included. Then in
(13)
and
(15)
the co-
efficients
Bn
vanish. The kinetic
energy
T is
given by
the
equation
T
=
irpa'h-Z
(^jr
sin ^
(2
sin^
+
(cos
+
nf}
tan^"
^d0^.
. .
.(20)
The
potential energy
of
bending,
estimated
per
unit of area of the middle
surface,
is
^ fih^ (k^^
+
t'),
where
:,
and t are
given by (15)
with the B's
omitted. Hence the
potential energy
V of the
vibrating
shell is
given by
the
equation
_n(n^-l)^4^/^tan-2 3i^^J
(21)
The coefficients
A^
in the
expressions
for the
components
of
displacement
can be
regarded
as
"
principal
coordinates
"
*
and the
frequency
can be
written down.
In a
principal
mode the
type
of vibration is
expressed by
the
equations
a
a
u
=
An
sin tan"
-
cos
n^,
v
=
An
sin tan"
-
sin
n<f>,
w
=
An
{n
+
cos
0)
tan"
^
cos
n^,
in which
An
is
proportional
to a
simple
harmonic function of the time. The
frequency jpn/27r
is
given by
the
equation
j,
=
I
^
^]n''
(,
-
1)
( I"

tan'"
I -^y(
Tsin ^
{2
sin" ^
+(cos
(?
+
n)'l
tan="
I
d6'V
In this
expression
n
may
be
any integer greater
than
unity.
*
When the
edge-line
consists of two circles of
latitude,
so that the coefficients II occur as well
as the coefficients
A,
the A'b and -B's are not
principal coordinates,
for terms
containing
such
products
as
(dAJdt). (dBJdt)
occur in the
expression
for T. See Lord
Bayleigh, Theory of
Sound,
second
edition. Chapter
x a.
v=%.,l^
321]
OF A THIN SHELL
487
The
integrations
can
always
be
performed.
"We have
fsinfl{2sin25
+
(co8d
+
)2}tan2sc^(9= [^ ^^~f''' \(n-\f+2ln
+
\)x-x^dx,
Jo
'2
jl+cosa
a:"
"' ' ^ ' '
and the second of these can be evaluated for
any integral
value of n. In the case of a
hemisphere {a
=
^n)
Lord
Eayleigh {loc. ck.)
finds the
frequencies p^y jOj, pi
for n
=
% 3,
4
to be
given by
^-WSS^^-^^^^' ^-^V(i?)^^^-^^)' ^-JV(IJ)^^-^^^)-
In the case of a saucer of 120
(a
=
Jtt)
he finds
In the case of a
very
small
aperture
in a
nearly complete sphere {a

v
nearly)
the
frequency
calculated from the above formula* is
given approximately by
"P"
"a^Zp {n-af
*
Cf. H.
Lamb,
loc. cit.
p.
477.
CHAPTER XXIV.
GENERAL THEORY OF THIN PLATES AND SHELLS.
322. Pormulse
relating
to the curvature of surfaces.
For the
investigations
in the last
Chapter
the elements of the
theory
of the curvature of surfaces are
adequate.
For the
purpose
of
developing
a more
general
method of treatment of the
problem
of curved
plates
or shells
we shall
require
some further results of this
theory.
It seems best to
begin
by obtaining
these results.
Let
a, /S
denote
any
two
parameters by
means of which the
position
of
a
point
on a surface can be
expressed,
so that the
equations
a
=
const.,
;8
=
const,
represent
families of curves traced on the surface. Let
%
be the
angle
between the
tangents
of these curves at
any point
;
^
is in
general
a function of a and
yS.
The linear element ds of
any
curve traced on the
surface is
given by
the formula
{dsy
=
A'
(day
+
^{(1^)^+
2ABcos
xdctd^,
(1)
where A and B
are,
in
general,
functions of
a, /9.
Let a
right-handed system
of
moving
axes of
x,
y,
z be constructed so that the
origin
is at a
point (a, /3)
of the
surface,
the axis of z is the normal to the surface at the
origin,
drawn
in a chosen
sense,
the axis of x is the
tangent
to the curve
/3
=
const, which
passes through
the
origin,
drawn in the sense of increase of
a,
and the axis of
y
is
tangential
to the
surface,
and at
right angles
to the axis of x*. When
the
origin
of this triad of axes moves over the surface the directions of the
axes
change.
If t
represents
the
time,
the
components
of
velocity
of the
origin
are
. da.
, j>dy8 D<^/3
. -
parallel
to the instantaneous
positions
of the axes of
x,
y,
z. The
components

When the curves a

=
const. and
/3=oonst.
cut at
right angles
we
suppose
that the
parameters
a and
^,
and the
positive
sense of the normal to the
surface,
are so chosen that the directions in
which a and
/3
increase and this normal are the directions of a
right-handed system
of axes.
322]
CURVATURE OF SURFACES 489
of the
angular velocity
of the
system
of
axes,
referred to tliese same
directions,
can be
expressed
in the forms
da
d/3
da
dl3
P^di+P^dt' 'i^dt^'i^dt
da
dy9
''
dt '^'''df
in which the
quantities jo,
,
... are functions of a and
/3.
The
quantities p,,
... are connected with each other and with
A, 2?,
^ by
the
systems
of
equations (2)
and
(3)
below. These results
may
be obtained
as follows:

Let
X, y,
z denote the coordinates of a fixed
point
referred to the
moving
axes. Then
X, y,
z are functions of a and
ft
and the conditions that the
point
remains fixed while the
axes move are the three
equations
a^ da dx
d^ /'
da
,
dff\
/ da
,
dff\
,
.da
, d0
daTt-^mdr^{:^dt+'--'dt)+'['i^dt+^-^di)-^'^di+^di'''^=^'
dy
da
dy dff ( da, fl?/3\
/ da
dff\ _rf/3
.
dz da dz da
(
da
,
dff\
(
da
, rf/3\
.
dadi-^radt-''[i^dt+^^ di)
+"
[p^
dt+P^
di)
=^-
Since these hold for all values of
da/dt
and
d^/dt,
we have the six
equations
dx
da
=
-A+riy-qiZ,
dy
=Pi^-r,x,
dz
dx

du

.
,
dz
The conditions of
compatibility
of these
equations
are three
equations
of the form

(J~\
=
^ (^\ ; and,
in
forming
the differential
coefficients,
we
may
use the above
op \OaJ
da
\Oji/
expressions
for
dx/da,
.... The results must hold for all values of
x, y,
z.
The
process just
sketched leads to the
equations'
da'
(2)
and
OX l_^
(dA
da B sin
x
\9/3
1
dB
dA
r,=
(dB_
|
+
2sinx
=
|cosx.
.(3)
*
The sets of
equations (2)
and
(3)
were obtained
by
D.
Codazzi,
Paris iUm.
par
divers
savants,
t. 27
(1882).
490
CURVATURE OF SURFACES
[CH.
XXIV
To
express
the curvature of the surface we form the
equations
of the
normal at
(a
+
Sa, /3
+
S/3)
referred to the axes of
x,
y,
z at
(a, /S).
The
direction cosines of the normal
are,
with sufficient
approximation, {q^o.
+
g^SyS),

(jJiSa+jJaS/S),
1,
and the
equations
are
a;

(J.Sa
+
JS8/3cosT^)
_
y

BB0
sin
x
_
It follows that the lines of curvature are
given by
the differential
equation
Ap^ (daf
+
B
(p.,
cos
x
+
qi
sin
x) (d^T
+
{Ap-2
+
^
{p^
cos
x
+
9^
sin
x)]
dad^=0,
(4)
and that the
principal
radii of curvature are the roots of the
equation
R"
(Pi?2
-
p^qi)
-
R
{Ap2
-
B
{p,
cos
x
+
qi
sin
x)}
+
AB sin
^
=
0. . .
.(5)
From these results the
equation
of the indicatrix of the surface is
easily
found to be
-|-^+Unx-^*^)^^
+
'5-^
=
'^^^*
^'^
The measure of curvature is
given by (5)
and the third of
(2)
in the form
1
/dvi
_ drj\
AB sin
X
W
3a/
323.
Simplified
formulse
relating
to the curvature of surfaces.
When the curves a
=
const, and
y8
=
const, are lines of curvature on the
surface the formulae are
simplified very
much. In this case the axes of x and
y
are the
principal tangents
at a
point,
the axis of z
being
the normal at the
point.
We have
X
=
l'^. i'.
=
0>
?=
=
0, (7)
and the roots of
equation (5)
are
-
A/q^
and
B/p^.
We shall write
J___2l J__^
(8\
R,~
A'
R,
B
^ ^
so that
Rj, R.2
are the radii of curvature of normal sections of the surface
drawn
through
those
tangent
lines which are axes of
x,
y
at
any point.
We
have also
IdA
_18:?
*"'
Bd^'
^'
Ada'
AB
R.
(9)
?A da\Ada) d^\Bd^J'
d
(B\
IdB d
(A\_]^dA
,
Q.
322-324]
DEFORMATION OF A THIN SHELL 491
324. Extension and curvature of the middle surface of a
plate
or shell.
In
general
we shall
regard
the middle surface in the unstressed state
as a curved
surface,
and take the curves a
=
const, and
/3
=
const, to be the
lines of curvature. In the case of a
plane plate
a and
/8
may
be
ordinary
Cartesian
coordinates,
or
they may
be curvilinear
orthogonal
coordinates. In
the case of a
sphere
a and
/8
could be taken to be
ordinary spherical polar
coordinates.
Equations (7)

(10)
hold in the unstressed state. When
the
plate
is deformed the curves that were lines of curvature become two
families of curves traced on the strained middle
surface,
which cut each other
at an
angle
that
may
differ
slightly
from a
right angle.
We denote this
angle by %
and its cosine
by
sr,
and we denote
by
e,
and
e^
the extensions
of linear elements
which,
in the unstressed
state,
lie
along
the curves
/3
=
const, and a
=
const. The
quantities
a and
yS
may
be
regarded
as
parameters
which determine a
point
of the strained middle
surface,
and
the formula for the linear element is
{dsY
=
4'^
(1
+
e,y {daf +B'(l+
,y
(d^f
+
2AB{1
+
e,)
(1
+
e,)
tr
rfad/3.
As in Article
322,
we
may
construct a
system
of
moving orthogonal
axes of
x,
y,
z with the
origin
on the strained middle
surface,
the axis of z
along
the
normal at the
origin
to this
surface,
and the axis of x
along
the
tangent
at the
origin
to a curve
/3
=
const. The
components
of
velocity
of the
origin
parallel
to the instantaneous
positions
of the axes of x and
y
are
^(l+,0g
+
5(l
+
e.)^f, 5(l+e.)sin/|."
The
components
of
angular velocity
of the triad of axes referred to these
same directions will be denoted
by
.da.
,dQ ,da ,d^
,da
,
,d/3
P^dt+P^dt'
'i'di
+
'i^dt' ''^dt^''^dt-
Then in
equations
(2)
and
(3)
we must
replace Ahy A{1
+
e^), BhyB{l+
ej),
Pi, ^2-

i\
by Pi'> Pi,

^2'-
The directions of the lines of curvature of the
strained middle
surface,
the values of the sum and
product
of the
principal
curvatures,
and the
equation
of the indicatrix are found
by making
similar
changes
in the formulae
(4)

(6).
If we retain first
powers
only
of
e,, e^, ct,
equations
(3) give
,_
IdA
3ot
^9^,i2M_:4^'^
1 dB
_'^^_li.^
+
^^
Adi
~Ad^
Ada'^Ada'
r,= ,^ -^^-^T-+^^%y
(11)
9'\P' -^^'4-6^1+6
2l
B^J-'^A^^'A^'^B'
492 EXTENSION AND BENDING
[CH.
XXIV
The indicatrix of the strained middle surface is
given,
to the same order
of
approximation, by
the formula
-
1'
(1
-
6,)
a^
+
!'
(1
-
e,)
-
5-'
r^rj
y
+
2
2^
(1
-
e.) ^i/
=
const.
If
jRi', iij'
denote the radii of curvature of normal sections of the strained
middle surface drawn
through
the axes of x and
y
at
any point,
and
\}r
the
angle
which one of the Hnes of curvature of this surface drawn
through
the
point
makes with the axis of x at the
point,
we
have,
to the same
order.
tan2^
=
-^(l-e,)y
...(12)
It is clear from these formulae
that,
when the extension is
known,
the state
of the strained middle surface as
regards
curvature is defined
by
the
quantities
-q^lA, p,'/B, p,'/A.
We shall write

^

^
=
i, ^ 7r
=
'^2i a=''' (13)
A
Ml
n
ji^
A.
and shall refer to
Ki, k^,
t as the
"changes
of curvature." In the
particular
cases of a
plane plate
which becomes
slightly
bent,
and a shell which under-
goes
a small inextensional
displacement,
these
quantities
become identical
with those which were denoted
by
the same letters in
Chapters
xxii. and
XXIII.
The measure of curvature is
given by
the formula
1

e,

2
(dvi
dr^\ /8rV
AB
V5/3
doL
where
r/, r^
are
given by
the first two of
(11).
When there is no extension
the values of
r/,
r,' for the deformed surface are identical with those of
r^, r^
for the unstrained
surface,
and the measure of curvature is unaltered
by
the
strain
(Gauss's theorem).
The sum of the
principal
curvatures,
being equal
to
1/Ri
+I/R2',
can be found from the formulae
(12).
325. Method of
calculating
the extension and the
changes
of
curvature.
To calculate
ei,
...
p/,
... in terms of the coordinates of a
point
on the
strained middle
surface,
or of the
displacement
of a
point
on the unstrained
middle
surface,
we introduce a scheme of nine direction cosines
expressing
the directions of the
moving
axes of
x,
y,
z at
any point
relative to fixed axes
of
X,
y,
z. Let the scheme be
324,
325]
OF A THIN SHELL
493
494 EXTENSION AND BENDING OF A
[CH.
XXIV
The forraulse
(18)
enable us to calculate
ei, e^, ct,
and the formulae
(19)
give
us the means of
calculating p/,

326. Formulse
relating
to small
displacements.
Let
u, V,
w denote the
components
of
displacement
of
any point
on the
unstrained middle surface referred to the
tangents
at the
point
to the curves
^
=
const, and a
=
const, and the normal at the
point
to the surface. We
wish to calculate the extension and the
changes
of curvature in terms of
u, V,
w and their differential coefficients with
respect
to a and
/3.
(a)
The extension.
According
to the formulse
(18)
we
require expressions
for
9x/3a,
... where
X,
y,
z are the coordinates of a
point
on the strained middle surface referred
to fixed axes. We shall choose as these fixed axes the lines of reference for
u, V,
w at a
particular point
on the unstrained middle
surface,
and obtain the
required expressions by
an
application
of the method of
moving
axes.
Let
P(n, 3)
be the chosen
point
on the unstrained middle
surface, P'(a
+
da, ^
+
8^)
a
neighbouring point
on this surface. The lines of reference for
u, v,
w are a triad of
moving axes,
and the
position
of these axes when the
origin
is at P' is to be obtained
from the
position
when the
origin
is at P
by
a small translation and a small rotation.
The
components
of the
translation,
referred to the axes at
P,
are
Aha^ -SS/S,
0. The
components
of the
rotation,
referred to the same
axes,
are
given by
the results in
Article 323 in the forms
Jit' El' d^
H'^ da A'
When P is
displaced
to
Pi
and P' to
Pi',
the
x,
y,
z of
Pi
are the same as the
u, v,
w
of P
;
the
X, y,
z of
Pi
are
^
X
+
(dx/da)
8a +
(ax/8/3) 8ft
...
,
and the
u, v,
w of P' are
u
+
(du/da)
8a +
(3m/3/3) 8ft
... .
These
quantities
are connected
by
the
ordinary
formulse
relating
to
moving axes,
viz. :
^8a
+
|83
=
^8a
+
f^-"
da
6/3 \d(
lSo\ f dASadB8^\^ [ Aha\
da^''
+
d-^^^'^^^+[fa^''
+
r^^^)-'"li;+''[-WB'^TaA)'
g-8a
+
g^8^= (^_8
+
^80J_(^__j+._;
and in these formulse we
may equate
coefficients of 8a and
8ft
The above
process
leads to the
following expressions
for
9x/9a,
.
ax
aa"
325, 326]
THIN SHBXL SLIGHTLY DEFORMED
493
When
products
of
u, v,
w and their
differential coefficients are
neglected
the
formulae
(18)
and
(20) give
_
1 9m V dA w
^'~Ad'a'^ABd^~}t,'
1 dv u dB
'
BdlS^
AB da'
w
\
.(21)
_\
dv \ du u dA V dB
'^~Ada'^Bd^~ABd^~Alida.'
These formulae determine the extension.
When the
displacement
is inextensional
u, v,
w
satisfy
the
system
of
partial
differential
equations
obtained from
(21) by
equating
the
right-hand
members to zero. As we saw in
particular cases,
in Articles 319 and
320,
the
assumption
that the
displacement
is inextensional is almost
enough
to
determine the forms of
, v,
w as functions of a and
^.
(b)
The
changes of
curvature.
According
to the formulae
(19)
we
require expressions
for the direction
cosines
l^,
... of the
moving
axes referred to the fixed
axes;
we
require
also
expressions
for
dl,/da.,
We shall choose our fixed axes as before to be
the lines of reference for
u, v,
w at one
point
P of the unstrained middle
surface.
By (15), (16), (17), (20), (21)
we can write down
expressions
for the
values of
Ij,
... at the
corresponding point
Pj
of the strained middle surface
in the forms
.(22)
k=l
[CH.
XXIV
.(23)
496 EXTENSION AND BENDING OF A
Then we have the values
,_
^
_
1 dv
_
u dA
iv_lSw
u
J
_
\dv u dA
M ^ at

^ ^^*' ^
^'~~Ad~a'^ABd^'
'~ '
^'~Bd^^R,'
and these hold for all
points.
We
apply
the method of
moving
axes to
deduce
expressions
for
dli/da,
...;
and then we form the
expressions
for
p^,
...
in accordance with
(19).
The direction cosines of the axes of
x, y,
z at a
neighbouring point Pi,
referred to the
lines of reference for
u, v,
w at
F',
would be denoted
by Li
+
(dLi/da)
8a
+
(3ii/9y3) 8^,
...
;
the direction cosines of the axes of
x, y,
z at
P^',
referred to the fixed
axes,
which are the
lines of reference for
m, v,
w at
P,
would be denoted
by
li
+
(dli/da)
8a
+
(?^i/3/3) 6^,
. . . . Since
the
components
of the rotation of the lines of reference for
u, v,
w are
gS/3
Ma
'
Ri
'
dAha dB
'd^
B'^ da A'
we have the
ordinary
formulee connected with
moving
axes in the forms
with similar formulae in which the suffix 1 attached to
I, m,
n and
L, M,
N is
replaced
successively by
2 and 3. On
substituting
for
Z],
... the values
given
in
(23),
we find
da
"
AB
S/3 \da
B
dfi) R^ \da
'^
rJ
'
8/3
~
A^
da\da BdfiJ'
dnii
d /I dv u dA\
}_dA
Bd^'
3nii
dill
drii
_d
/Idv u dA\ B
(\
dw u\ \ dB
~d^\Ada
AB
dp) RiKA
da'^Rj'^A
da'
_3/l3w
^\ ,
-^
~di\Ad^'^Ri)'^Ri'
~d^\A da'^Ri)'^Ri\Ada
AB
dfij'
326]
and
THIN SHELL SLIGHTLY
DEKORMED
497
3/3
_
3 / 1 dv u dA\ 1 dB
3/3 V
Ada AB
d^J
A da
'
8i2_
1 dA /dv u dA\
^~AB
d^\S^~Bd^)'
8/8 E2 W aj
A^ da
\da
B
d^J
'
^
_^f]^'^iB
v\
_
I /dv u dA\
da
~da\'SdB'^Jt'J~Ri\da~Bd^)'
dui
_
d /I dw v\ B
.12/ iV2
In
calculating p^,
... from the formulae
(19),
we write for
li,
... the values
given
in
(22),
and for
dli/da,...
the values
just found,
and we observe
that,
since the scheme
(14)
is
orthogonal,
two of the formulse
(19)
can be written
8?i 6;, , /, 3^1 .
dm,
.
3rei\
, /, -, ,
The
process just
described leads to the formulae
,
_
d /I dw v\ 1 dA /I dw u\ 1 /dv
=33"""''^"""'3^y
-+n3
dn{\
u dA
d
/I
dw
*'
R]
doL
U
da
"^
R
\ IdA
f\dw
B
d/3 \B
da
^'
~
Bdl:i'^da\Ada
AB
d^
J
'^
RAB d$
'^
BJ
'
i
P'~R,'^d^\Bd^'^ R,)'^ Adi\Ada'^ RJ
'
y...(24)
,
d
fl
dw u
^'^~d^\Adi'^R,
'^
AdaKBd/B'^ R,
v_\_
B
RJ AR,
dv
da'
u dA
Bd^
V...(25)
^''~AdcL'^d0 [a
da
u dA
ABdB
B
R2
1 dw u
Ada^%
We can now write down the formulae for the
changes of
curvature in the
forms
u\ .
\ dA
_
1
a^
/\dw
"'"AdaKAda^R,
+
1 d /I dw V
"'^Bd^KBdB R:
_'^d_/^dw
''~
Ada\B'd$
+
ABd^
ABdi
v\
\ dAdw
rJ
~
A'B
W
da
1 dw V
d$^Rj'
(s
Ada^
rA^
1
d_v
AR,
da
L. E.
(26)
32
498 STRAIN IN A BENT
[CH.
XXIV
The above formulee admit of various verifications :
(i)
In the case of a
plane plate,
when a and
ff
are Cartesian
coordinates,
we have
d^w
_dhv
dhv
""Sii^'
"''W '"~3^3|3"
These results
agree
with the formulee in Article 298.
(ii)
In the cases of
cylindrical
and
spherical shells,
the conditions that the
displace-
ment
may
be inextensional can be found as
particular
cases of the formulse
(21),
and the
expressions
for the
changes
of
curvature,
found
by simplifying (26)
in accordance with these
conditions, agree
with those obtained iu Articles 319 and 320.
(iii)
Let a
sphere
be
slightly
deformed
by purely
normal
displacement,
in such a
way
that the radius becomes a
+bPn (cos 6),
where 6 is
small, P^
denotes
Legendre's
mth
coefficient,
and d is the co-latitude. The sum and
product
of the
principal
curvatures
of the deformed siu^iace can be
shown, by
means of the foimulae of this Article and those
of Article
324,
to be
l+X{n-l){n
+
2)P{coae)
and
i
-t-
i(i-l)(
+
2)P,(cos e),
Cb U KM, (Ji
correctly
to the first order in 6. These are known results.
(iv)
For
any surface,
when
fj, 2)
'>^ are
given by (21),
and
pi,
... are
given by (24)
and
(25), equations (11)
are satisfied
identically, squares
and
products
of
m, v,
w and their
differential coefBcients
being,
of
course,
omitted.
327. Nature of the strain in a bent
plate
or shell.
To
investigate
the state of strain in a bent
plate
or shell we
suppose
that
the middle surface is
actually
deformed,
with but
slight
extension of
any
linear
element,
so that it becomes a surface
differing
but
slightly
from some
one or other of the surfaces which are
applicable upon
the unstrained middle
surface. We
regard
the strained middle surface as
given
;
and we
imagine
a
state of the
plate
in which the linear elements that are
initially
normal to
the unstrained middle surface remain
straight,
become normal to the strained
middle
surface,
and suffer no extension. Let P be
any point
on the unstrained
middle
surface,
and let P be
displaced
to
Pj
on the strained middle surface.
Let
X,
y,
z be the coordinates of
Pj
referred to the fixed axes. The
points
P
and
P,
have the same a and
/3.
Let
Q
be
any point
on the normal at P to
the unstrained middle
surface,
and let z be the distance of
Q
from
P,
reckoned
as
positive
in the sense
already
chosen for the normal to the surface. When
the
plate
is
displaced
as described
above, Q
comes to the
point
Qi
of which
the coordinates are
where,
as in Article
32.5, l^, m,, n^
are the direction cosines of the normal to
the strained middle surface.
The actual state of the
plate,
when it is deformed so that the middle
surface has the
assigned
form,
can be obtained from this
imagined
state
by
imposing
an additional
displacement upon
the
points Q^.
Let
^, rj, 5'
denote
the
components
of this additional
displacement,
referred to axes of
x,
y,
z
326, 327]
PLATE OR SHELL
499
with
origin
at
P,
which are drawn as
specified
in Article 324. Then the
coordinates of the final
position
of
Q
are
z
+
,f +n.i7+n,(
+
f). ...(27)
In these
expressions
Z,,
... are the direction cosine.s so denoted in Article
325,
X,
y,
z, /,,
...
3
are functions of a and
/8,
and
f, /,
are functions of
o, /9,
z.
We consider the
changes
which must be made in these
expressions
when,
instead of the
points
P, Q,
we take
neighbouring points
P', Q',
so that
Q*
is
on the normal to the unstrained middle suiface at
P',
and the distance
P'Q*
is ^
+ Bz,
where Bz is small. Let P be
(a, /3)
and P'
(a
+
Sa, /9 +
S/S),
where
8a and
S/3
are small
;
and let r denote the distance
QQ',
and
I, m,
n the
direction cosines of the line
QQ',
referred to the
tangents
at P to the curves
yS
=
const, and o
=
const, which
pass through
P and the normal to the un-
strained middle surface at P. The
quantities
a, ^,
z
may
be
regarded
as the
parameters
of a
triply orthogonal family
of surfaces. The surfaces z
=
const,
are
parallel
to the middle surface
;
and the surfaces a
=
const, and
/9
=
const,
are
developable
surfaces,
the
generators
of which are the normals to the
unstrained middle surface drawn at
points
on its several lines of curvature.
The linear element
QQ'
or r is
expressed
in terms of these
parameters by
the
formula
and the
projections
of this eleaient on the
tangents
to the curves
/3
=
const,
o
=
const.,
drawn on the middle
surface,
and on the normal to this surface
are
h, wx,
nr. Hence we have the formulae
^=za"^)'
^^
=
^(i-V)
'
^'^"''
^''^
In
calculating
the coordinates of the final
position
of
Q'
we have in
(27)
to
replace
k
by
/i
+
i,(r,'8a
+
r,'S/3)-/3(?i'Sa
+
^;a/3),
h
by
k
+
k{p:^OL+ p;B^)-k{r-;Ba
+
r^m.
k
by
k
+
h{q,'Ba+q,'B^)-hiPi^a+Pi'8^),
I'^y
^*fj'*?,'f4''
z
by
z + Sz.
We use also the formulas
(15)
and
(16)
for
dx/da,
... and the formulas
(28)
for
ia, g/3,
Bz.
32i
500
STRAIN IN A BENT
[CH.
XXIV
Let
ti
denote the distance between the final
positions
of
Q
and
Q'.
We
express
r,
as a
homogeneous quadratic
function of
I, m, n,
and deduce ex-
pressions
for the
components
of strain
by
means of the formula
r,
=
r
[(P
+
mi" +
m*)
+
2
{e^xf
+
eyytn'
+
e^Ji"
+
ey^mn
+
ez^nl
+
e^ylm)].
Now the difference
of the x-coordinates of the final
positions
of
Q
and
Q'
is
( It jut ^
l/I / 7 /X
^I"
. /I / 7 'X
Wir
)
+
(.
+
?)
|(Z,
^/
-
4K)
2-(T^7)
+
^^^*'
-
^=P''>
i/(i-'./i?o.
+
h
af
ir
Sf
7r
+
^^nr
8a ^
(1
-
^/ie.)
^
8;8

(1
-
z/R,)
^
dz
+
I2
dt)
It
dr)
mi
dr]
3a ^
(1
-
z/R,) ^d^{l- z/R,)
^
dz
+
1.
af
Ir
+
ar
mr /
dj\
1
''
Wa
A{1- z/R,)
^
3^
5
(1
-
z/R,j
'^
\
^
aW^'j
'
The dififerences of the
y-
and z-coordinates can be written down
by
sub-
stituting
nil, f>h, 1^3
and
,, Wa, n^
successively
for
^1, 4, h-
Since the scheme
(14)
is
orthogonal,
we find the value <>f
r,'-'
in the form
'-"['"r-U{i/"-3'-^^<'+4i}
m
-t-r"
I
l-zjR,
pi
B v+%U
+
t}^
1
a|
+
9fT
-i-rr
'
\-zlR,
ll -zIR.
i^^A^^^aaPi-^/ie, t
-6^^^
5a^j
(29)
+
n 1
+
a?'
a^.
In
deducing expressions
for the
components
of strain we observe
that,
in
order that the strains
may
be
small,
it is
clearly necessary
that the
quantities
l-z/i2.U
i2i/'
l-^/iJA

-B2/'
l-2/R,B
1-
327,
328]
PLATE OR SHELL
601
should be small. The third of
equations (11)
in Article 324 shows
that
p^'jA
+
q^jB
is a small
quantity,
and we see
therefore
that,
in the
notation
of
(13)
in Article
324,
the
quantities zk zk.^,
zr must be
small.
The
expressions
for the
components
of strain which we obtain
from
(29)
are
^(EX
yy
1
-
zjR^
1
ZK^
1
/sr
A
\d
l-r.'n^qn)].
^yy
^XU
"~
1
-
zjR^
ST
l-z/Ri
^'c-2+l.it~-p,'^+r,'^
B'd^
TZ
+
l-2/R,l-z/R.
+
-r--Vi(a^^.-f-.-f)-:4^^(|
l-z/R,
]R,
CM)
r^V
+
q^tj,r
...(30)
'dz'
1
(dK
^'"''dz^l-z/R.AKda
dz'^l-
z/R^

\d0
=
^
+
-qi'^+PiVJ
In these
expressions ^,
r], f
are functions of
a, 0,
z which vanish with z for all
values of
a, /8.
We observe that the values found in Article 317 for
exx<
e^y, e^y
would be
obtained from the above
by omitting
ej, e^,
et and
f, rj, f,
and
replacing
1

z/Ri
and 1

z/R.^.
by unity.
328.
Specification
of stress in a bent
plate
or shell.
The stress-resultants and
stress-couples
in a curved
plate
or
shell,
or in a
plane plate
which is
appreciably
bent,
may
be defined in a similar
way
to
that
adopted
in Article 294 for a
plane plate slightly
deformed. Let s denote
any
curve drawn on the strained middle
surface,
v the normal to this curve
drawn in a chosen sense on the
tangent plane
of the surface at a
point
Pi,
and let the sense of
description
of s be such that the directions of the normal
V,
the
tangent
to
s,
and the normal to the surface at
P,,
in the sense
already
chosen as
positive,
are
parallel
to the axes of a
right-handed system.
We
draw a normal section of the strained middle surface
through
the
tangent
to
s at
P],
and mark out on it a small area
by
the normal to the surface at
P,
and the normal to the
(plane)
curve of section at a
neighbouring point
P,'.
The tractions exerted across this
area,
by
the
portion
of the
plate
on that
side of s towards which v is
drawn,
upon
the
remaining portion,
are reduced
to a force at
P,
and a
couple.
The
average components
of this force and
couple per
unit of
length
of
PiPi
are found
by dividing
the measures of the
502
STRESS AND STRAIN IN A
[CH.
XXIV
components by
the measure of this
length.
The limits of these
averages
are
the stress-resultants and
stress-couples belonging
to the curve s at the
point
Pj.
We denote
them,
as in Article
294,
by
T, S, N, H,
G. For the
expres-
sion of them we take
temporary
axes of
x,
y,
z
along
the normal
v,
the
tangent
to
s,
and the normal to the strained middle surface at
Pj,
and denote
by
Z'x-.
the
stress-components
referred to these axes.
Then,
taking
B! to
be the radius of curvature of the normal section of the surface drawn
through
the
tangent
to s at
Pj,
we have the formulae
^=/>'K^-i')''^'
^=/>v(^-l^)^^' ^=/>'^(i-i-')'^^'
When we refer to the axes of
x,
y,
z
specified
in Article
324,
and denote
the stress-resultants and
stress-couples belonging
to curves which are normal
to the axes of x and
y respectively by attaching
a suffix 1 or 2 to
T, ...,
we
obtain the formulae
"^^^WM'-iM"' '^-WM'-iM''
'-f-M'-i^'']
H.
=
\\-zX,[x-^)dz.
G.=j\zX.{l--^^dz,
and
^-/>''(^-'i')'^' '^-L-^'-i)'-- ^-/>^(^-i>
H,
=
f_
ZXy
{\
-
-^)
dZ, (?.
=/*^
^Yy
(l
-
4-)
dZ,
(31)
(32)
in which
7?/
and
R^' denote,
as in Article
324,
the radii of curvature of
normal sections of the strained middle surface drawn
through
the axes of x
and
y.
We observe that the relations
S^
+
S.2
=
and
H^
+
Hi
=
0,
which hold in
the case of a
plane plate slightly
deformed,
do not hold when the strained
middle surface is
appreciably
curved. Tlie relations between the
T, S, N, G,
H
for an
assigned
direction of v and those for the two
special
directions x and
y,
which we found in Article 295 for a
plane plate slightly
deformed,
are also
disturbed
by
the
presence
of an
appreciable
curvature.
329.
Approximate
formulae for the
strain,
the stress-resultants
and the
stress-couples.
We can deduce from
(30)
of Article 327
approximate expressions
for the
components
of strain
by arguments precisely
similar to those
employed
in
Articles 2.57 and 259. Since
f, 77, f
vanish with z for all values of a and
/3,
and
d^jdz,
... must be small
quantities
of the order of admissible
strains,
328,
329]
^
BENT PLATE OR SHELL
503
f, 7], f
and their diflferential coefficients with
respect
to a and
yS
may,
for a
first
approximation,
be omitted.
Further,
for a first
approximation,
we
may
omit the
products
of
zjRi
or
zjR^
and
any component
of strain. In
particular,
since
q.^B+p^lA
is of the order
e,/jBi,
we omit the
product
of this
quantity
and z
; and,
for the same
reason,
we
replace
such terms as
^ttt
and
1

zjRi
zk

\--n
by
ei
and
zk^.
By
these
processes
we obtain the
approximate
X

Z\
SX\
formulae
*
Cxz
^
f
1 ZKx,
Cyy
=
63 ZKi,
gj-y
=
or
ZZTi e^a;
~
jj^
)
^v
~
57
>
^^^-
.(00)
In these
f, 77, i^roay,
for a first
approximation,
be
regarded
as
independent
of
a and
/3.
In case the middle surface is
unextended,
or the extensional
strains
e,, ej,
w are small
compared
with the Hexural strains
^,, 2, zt,
these
expressions may
be
simplified
further
by
the omission of
ei, e,,
ct.
The
approximate
formulae
(33)
for the
strain-components,
as well as the
more exact formulte
(30),
contain the unknown
displacements f, ?;, f,
and it is
necessary
to obtain values for these
quantities,
or at
any
rate for their
differential coefficients with
respect
to
z,
which shall be at least
approxi-
mately
correct.
We
begin
with the case of a
plane plate,
and take
a, ^
to be Cartesian
rectangular
coordinates,
so that A and B are
equal
to
unity,
and
1/iii
and
l/iij
vanish. In the forrauliB
(33) ^,
1], f
are
approximately independent
of
a, /8.
We consider a slender
cylindrical
or
prismatic portion
of the
plate
such as would fit into a fine hole drilled
transversely through
it. We
may
take the cross-section of this
prism
to be so small that within it
ei, ej,
a and
Ki, K2,r
may
be treated as constants. Then the
strain-components,
as ex-
pressed by (33),
are the same at all
points
in a cross-section of the slender
prism.
If there are no
body
forces and no tractions on the faces of the
plate,
we know from Article 306 that the stress in the slender
prism,
in which the
strains are uniform over
any
cross-section,
is
plane
stress.
Hence,
to this
order of
approximation
X^, Y^, Z^ vanish,
and we have
|l
=
0,
^-5
=
0,
f
=
--^[e,
+
e,-z{K,
+
K,)] (34)
OZ OZ OZ 1

(7
'
I
The
remaining stress-components
are then
given by
the
equations
E
E F E
Xx
=
1 , {ei
+
(re,-z (k,
+
(tk,)], Yy
=
,
[e
-h <Te,
-
^
(/c,
+
<t/c,)},
^=2(r-r^)(--2^^)--(''>
*
Equivalent
formulffi in the case of a
plane plate
were
given by
Kirchhoff, Vorlesungen
iiber
math.
Phyxik,
Mechanik, Vorlesung
30.
504
STRESS AND STRAIN IN A
[CH.
XXIV
From these results we
may
deduce
approximate
formulae for the stress-
resultants and
stress-couples.
For this
purpose
we omit from the formulae
(31)
and
(32)
the factors
(1
-
zjR^)
and
(1
-
zjRi).
We should obtain zero
values for
N^, N^,
while
T^,
... and
G^,
... would be
given by
the formulae
2Eh
,
, 2A'/i , .

y
Eh ,
T,
=
j^
(ei
+
o-e,),
T,
=
~^-^
(62
+
o-eO,
-
S,
=
^
=
^^
is,
. .
.(06)
and
G',
=
-Z)(^,
+
c7,), G,
=
-D{K,-\-aK,\
-H,
=
H,
=
D{\-o)t. ...(37)
To the same order of
approximation
the
strain-energy per
unit of area is
given by
the formula
[Ehl{\
-
a')] [(6,
+
,Y
-
2
(1
-
<7) {e,e,-i^^)]
+
isD [(,
+
K.,y
-
2
(1
-
<r) {k,k,
-
r%
. .
.(38)
To
get
a closer
approximation
in the case of a
plane plate
we
may regard
the strain in the slender
prism
as
varying uniformly
over the cross-sections.
Then we know from Article 306 that
X^
and
Fj
do not
vanish,
but the third
of
(34)
and the formula?
(35)
still
hold,
and therefore also
(36)
and
(37)
are
still
approximately
correct,
while
Ni
and
N^
are
given according
to the result
of Article 306
by
the formula?
These values for
N^, N^
could be found also from
(12)
of Article 296
by
omitting
the
couples
L',
M' and
substituting
for
d, 0^, H^
from
(37).
From this discussion of the case of a
plane plate
we
may
conclude that
the
approximate expressions (33)
and
(34)
for the
components
of strain are
adequate
for the
purpose
of
determining
the
stress-couples
; but,
except
in
cases where the extension of the middle
plane
is an
important
feature of the
deformation,
they
are
inadequate
for
determining
the stress-resultants. The
formula?
(37)
for the
stress-couples
are the same as those which we used in
Articles
313,
314. The results obtained in Articles
307, 308,
312 seem to
warrant the conclusion that the
expressions (37)
for the
stress-couples
are
sufficient
approximations
in
practically important
cases whether the
plate
is
free from the action of
body
forces and of tractions on its faces or
not.
In the case of a curved
plate
or shell we
may,
for a first
approximation,
use the formulae
(33)
and the theorem of Article 306 in the same
way
as for
a
plane plate.
Thus
equations (34)
and
(35)
are still
approximately
correct.
We
may
obtain from them the terms of lowest order in the
expressions
for
the stress-resultants of the
type
T,
S and the
stress-couples.
On
substituting
in the formulae
(31)
and
(32),
we
find,
to the first order in
h,
T,
=
:j -,
(1
+
ae,), T.,
=
(e,
+
<r,), -8,
=
S,=
-
w,
. .
.(36 his)
(39)
329]
BENT PLATE OR
SHELL
505
and,
to the third order in
h,
{?,
=
-
i)
|/ci
+
aK,
+
^,
(e,
+
o-6,)|
, G,
= -
i)
L, +
ffK,
+
i-, (f,
+
ffe,)|
/
ff,.7,o-,)(.+>|.,), /f...j,a_,)(,
+
i^,).'
This first
approximation
includes two extreme cases. In the first the
extensional strains
e^, e,,
m are small
compared
with the flexural strains
ZK-i, ZK.,,
zr. The
stress-couples
are then
given by
tlie formula?
G,
=
-D{K,
+
aKX
G.
=
-D(K,
+
a-K,), -H^
=
H,=D{\-<T)T,...{yiUs)
and the
strain-energy per
unit of area is
given by
the formula which we
found
by
irieans of a certain
assumption
in Article
317,
viz. :
\B [{k,
+
K,y
-
2
(1
-
<7) {k,k,
-
T^)],
but the stress-resultants are not
suSiciently
determined.
In the second extreme case the flexural strains
zk^, zk^,
zt are small
compared
with the extensional strains
e^, e^,
tn-. Then the
stress-resultants
of
type
T,
S are
given by
the formute
(36),
and the stress-resultants of
type
iV and the
stress-couples
are
unimportant.
The
strain-energy per
unit of area
is
given by
the formula
[Eb/{1
-
a^)] [(6,
+
6.,)^
-
2
(1
-
<7)
(6,6.,
-
i^')] (40)
When the extensional strains are
comparable
with the flexural
strains,
so
that,
for
example,
or is of the order
At,
the stress-resultants of
type
T,
S are
given
with sufficient
approximation by (36),
and the
stress-couples
are
given
with sufficient
approximation by (37),
while the
strain-energy per
unit of
area is
given by (38).
From this
analysis
of the various
possible
cases it
appears
that,
whenever
the
stress-couples
Gj, G^, H^, H2
need be calculated at
all,
they may
be
calculated from the formulas
(37)
instead of
(39).
When the extensional strains are
large compared
with the flexural
strains, approximate
equations
of
equihbriuni
can be formed
by
the method of variation described in Article
115, by taking
the
strain-energy per
unit of area to be
given by
the formula
(40).
In
the same case
approximate equations
of vibration can be formed
by using
this
expression
(40)
for the
strain-energy
and the
expression ph
I -*
-
j
+
(
^7
)
+
(
5;'
)
*"" ^^^ kinetic
energy per
unit of area.
The
strain-energy per
unit of area is
not,
in
general, expressed correctly
to the third
order in h
by (38).
The
complete expression
would contain additional terms. In
general
the
complete expression
for the
strain-energy
must be formed before
equations
of
equilibrium
and vibration can be obtained
by
the variational method*. We shall use
a difi'erent method of
forming
the
equations.
The
approximate expression (38)
for the
strain-energy suggests,
as the correct
form,
a function
expansible
in
rising powers
of
h,
and
having
for coefficients of the various
powers
of h
expressions
determined
by
the
displacement
of the middle surface
only.
*
A. B.
Basset,
Phil. Trans. B. Soc.
(Ser. A),
vol. 181
(1890).
506
STRESS AND STRAIN IN A
[CH.
XXIV
Lord
Rayleigh*
has called attention to the fact
that,
when there are tractions on the
faces of the
shell,
no such form is
possible,
and has illustrated the matter
by
the two-
dimensional
displacement
of a
cylindrical
tube
subjected
to surface
pressure.
In this
problem
the first
approximation, given by (40),
is undisturbed
by
the surface
pressures.
330. Second
approximation
in the case of a curved
plate
or shell.
In the case of an
appreciably
curved middle surface we can make some
progress
with
a second
approximation provided
that the
displacement
is small. Such an
approximation
is
unnecessary
unless the extensional strains
fi, f2,
^ are small
compared
with the flexural
strains
zki, zk^,
zt. We shall
suppose
that this is the case. In
calculating
the strains
fijcc,...
from
(30)
instead of
(33)
we observe that the term
f,(l -z/Rj)'"^ may
still be
replaced
by ti,
and that the term
-zki (1 -z/^i)"' may
be
replaced by zki-z^ki/Ri.
The values
of
f, ij, f
which were
given by
the first
approximation
are
1
=
0, ,,
=
0,
f=-j^{(.,
+
,,)0-i(<,
+
2)22},
and these values
may
be substituted in the first three of
(30). Further,
in the terms of
(30)
that contain
|, 17, f
we
may replace pi,
...
by
the
corresponding quantities relating
to the unstrained
shell,
that is to
say
we
may put pi'
=
q.J
=
0,
p.2'/Bl/li2, -qi'/A
=
llRi.
We
reject
all terms of the
types (\zjRi, fi:i, ki^z^.
We thiis obtain the
equations
2
*'
1 g
2
"i
+
^g
e-fi ZKi
^
^
2 1-<r^
A
'
.,
K2
Iff
,<\
+
K--z r (41)
-TZ^il/Ri
+
l/R^).
From the formula for
e^,,
we can calculate
Si
and
152 by
means of
(31)
and
(32)
of
Article
328,
and in this calculation we
may replace l/Ri
and
I/R2' by 1/Ri
and
1/^2.
We find
S,^
_^^+Z)(l-.)-^-ii>(l-.).(i-4-i),
^^w-i>(I-.)-^+ii)(l-.).(i-
+
i-j.
.(42)
S,=
-
In
calculating
a second
approximation
to
7*1
and
T2
we
may
not assume that
Z,
vanishes. As in the case of the
plane plate,
we take the shell to be free from the action
of
body
forces
and
of tractions on its faces. We observe that the axes of
x, y,
z
specified
in Article 323 are
parallel
to the normals to three surfaces of a
triply orthogonal family.
This is the
family
considered in Article
327,
and the
parameters
of the surfaces are
a, j3,
z.
We write
temporarily y
in
place
of
z,
and use the notation of Articles 19 and 58. The
values of
hi, h^, A3
are
given by
the
equations
k-' (-i). rM'-)- k-'-
We write down an
equation
of the
type
of
(19)
in Article 58
by resolving along
the normal
to the surface
y.
This
equation
is
-^(-(-i)(-i)"}]
-f(-r^(-(-i.))-f('-i)""^{K-0-
*
LoncUm Math. Soc.
Proc,
vol. 20
(1889), p. 372,
or
Scientific Faperi,
vol.
3, p.
280.
329-331]
BENT PLATE OR SHELL
507
Returning
to our
previous notation,
we write this
equation
s(-(-7a-v.}nM-(-i)>-M(-(.-A)(.-A)4
-ll (-4) '-f('-i)
'-.-
To obtain an
approximation
to
Z^,
we substitute in this
equation
for
A'^.,
... the vahies
given by
the first
approximation,
and
integrate
with
res{)ect
to z. We determine the
constant of
integration
so that
Z,
may
vanish at z
=
h and z= -A. We must omit the
terms
containing X^
and
}',
and use the
approximate
values
given
in
(35)
for
X^
and
}',.
Further we
may
omit the factors 1
-
zjRi
and 1
-
zjE-^
and such terms as f
i z//fi
. We thus
find the formula
'--li^^'-^tT^^'T^
^^'^
ET
rr P*
Now we have
and
hence, by
means of the formula for
e^^^i yy, ^t-,
we calculate
approximate
values for
r,,
To in the forms*
^.=^(,,.,),Z>[.(i--i-)-i^-^(..
+
.,(i-
+
j)
l-<
.(44)
\-<t\ RZ~^ Ri ;J'
1-o-V A R-i JI
J
The formute for the
stress-couples
are not affected
by
the second
approximation,
so far
at
any
rate as terms of the order
Dki
are concerned.
331.
Equations
of
equilibrium.
The
equations
of
equilibrium
are formed
by equating
to zero the resultant
and resultant moment of all the forces
applied
to a
portion
of the
plate
or shell. We consider a
portion
bounded
by
the faces and
by
the surfaces
formed
by
the
aggregates
of the normals drawn to the strained middle surface
at
points
of a curvilinear
quadrilateral,
which is made
up
of two
neighbour
-
ing
arcs of each of the families of curves a and
^.
Since the extension of
the middle surface is
small,
we
may neglect
the extensions of the sides of the
quadrilateral,
and we
may regard
it as a curvilinear
rectangle.
We denote
the
bounding
curves of the curvilinear
rectangle by
a,
a +
Bot, 0, ^
+
8/3,
and
resolve the stress-resultants on the sides in the directions of fixed axes of
*
The
approximate
forms of
S,
,
S.^, T^, T,
obtained in this Article
agree substantially
with
those found
by
a different
process by
A. B.
Basset,
loc. cil.
p. 50.5,
in the cases of
cylindrical
and
spherical
shells to which he restricts his discussion. His forms contain some additional terms
which are of the order here
neglected.
508 EQUATIONS
OF
EQUILIBRIUM
[CH.
XXIV
X,
y,
z which coincide with the
tangents
to
;S
and a at their
point
of intersection and the normal to
the strained middle surface at this
point (Fig. 73).
Fig.
74 shows the directions
and senses of the stress-resultants
on the
edges
of the curvilinear
rectangle,
those across the
edges
o + Sa and
/3
+
S/S being
distin-
guished by
accents. The axes of
the
stress-couples
H^, Gi
have the
^'
same directions as
jT,, S^;
those of
H^, Gi
have the same directions as
T'i, Si.
^
+
S/3
Fig.
73.
Fig.
74.
The stress-resultants on the side a of the
rectangle yield
a force
having
components
-T,BS^, -S,BS^, -N,BB^
parallel
to the axes of
x,
y,
z. The
corresponding component
forces for the
side a
+ Ba are to be obtained
by applying
the usual formulae
relating
to
moving
axes;
for the
quantities
T^, Si, iV,
are the
components
of a vector
referred to
moving
axes of
x,
y,
z,
which are defined
by
the
tangent
to the
curve
yS
=
const, which
passes through any point
and the normal to the
strained middle surface at the
point.
In
resolving
the forces
acting
across
331]
OF A THIN SHELL
509
the side a + Ba
parallel
to the fixed
axes,
we have to allow for a
change
of o into a
+
Sa,
and for the small rotation
(pi'Ba, qi'Ba, r,'8a).
Hence the
components parallel
to the axes of
x,
y,
z of the force
acting
across the side
a
+ Ba are
respectively
T,BB^
+ Sa
^
(T,BB^)
-
S,BB^
.
r.'Ba
+
N, BBff
.
q.'Ba,
S^BB0
+ Ba
^^
(S.BB^)
-
N,BB0.p,'Ba
+
T.BBfi
.
n'Ba,
N,B80
+
&a^
{N,BB0)
-
T.BSff
.
q/Ba
+
S,BB^.p,'Ba.
In like manner we write down the forces
acting
across the sides
/8
and
;8
+
S/3.
For
(3
we have
S,ABa, -T^ABa,
-N^ABa;
and for
/S
+
S/3
we have
-
S,ABa
-
S;8

(S^ABa)
-
T,ABa
.
r,'S/3
+
N,ABa
.
q,'Bl3,
T,ABa
+
8^
g|
(T.ABa)
-
N.ABa
.
p,'Bl3
-
S,ABa
.
n'B^,
N.ABa
+
8/3
~
{N.ABa)
+
S,ABx
.
q.'Bff
+
T,ABa.p^B0.
Let
X', Y',
Z' and
L', M', denote,
as in Article
296,
the
components,
parallel
to the axes of
x,
y,
z,
of the force- and
couple-resultant
of the
externally
applied
forces estimated
per
unit of area of the middle surface. Since the
area within the
rectangle
can be taken to be
AB8aB0,
we can write down
three of the
equations
of
equilibrium
in the forms
djT^
_
d{S,A)
_
^^,^^^B^ ^,rj,^^^
^
^^,^^^
^
q:N,A)
+ ABX'
=
0.
'
^-^^
+
^^-^
-
{p^'N,B +p,'N,A)
+
ir,%B
-
r,'S,A)
+ ABY'
=
0,\
(45)
?^^
+
^-^^
-
{q^'T.B- q:S,A)
+
{p:S,B +p,'T,A )
+ ABZ'
=
0.
Again
the moments
of the forces and the
couples acting
across the sides
of the
rectangle
can be written down. For the side a we have the
component
couples
-H,BB/3, -G,BB0, 0,
and for the side a +
Sa we have the
component couples
H,BB0
-f- Sa
^
{H,BB0)
-
G^BB^
.
r/Sa,
G, B8/3
-t-
8a
^
((?,58/3)
+
H,BB0
.
r/Sa,
-
H,BB0
.
q^'Ba
+
G, BB0
.
p,'Ba
;
510 BOUNDARY CONDITIONS
[CH.
XXIV
for the side
/9
we have the
component couples
G.ASa, -H.AZa, 0,
and for the side
/3
+
S/9
we have the
component couples
-
G,AU
-
ByS
g^
{G^Aha)-
H.AZol
.
r/8y8,
H,A
Sa
+
8/9
^
(H^ABa)
-
G,
A Ba .
n'B0,
GaABa.q,'B^
+
H^ABa.
p.'Bfi.
Further the moments about the axes of the forces
acting
across the sides
a
+ Ba and
fi
+
B^
can be taken to be
BB^.N^ABa, -ABol.N.BB^,
ABa .
S,BB^
+
BB^
.
S^ABa.
The
equations
of moments can therefore be written in the forms
d(H,B) d{G,A)
da
d0
-
(G,Bn'
+
H,Ar^')
+
{N,
+
L')
AB
=
0,
'S^^'J^HH.Br.'-G.Ar.')-(W.-M')AB
=
0,\-^''^
oa
()p
G,Bp/
+
G,Aq2
-
(H.Bq,'
-
H,Ap,')
+
{S,
+
S,)
AB
=
0.
Equations (45)
and
(46)
are the
equations
of
equilibrium.
332.
Boundary
conditions.
The
system
of stress-resultants and
stress-couples belonging
to a curve
s drawn on the middle surface can be modified after the fashion
explained
in Article
296,
but account must be taken of the curvature of the surface.
Regarding
the curve s as a
polygon
of a
large
number of
sides,
we
replace
the
couple
HBs
acting
on the side Bs
by
two
forces,
each of amount
H,
acting
at
the ends of this side in
opposite
senses in lines
parallel
to the normal to the
surface at one
extremity
of Bs
;
and we do the like with the
couples acting
on
the
contiguous
sides. If P'PP" is a short arc of
s,
and the arcs P'P and PP"
are each
equal
to
Bs,
these
operations
leave us with a force of a certain
magnitude
direction and sense at the
typical point
P. The forces at P and
P",
arising
from the
couple
on the arc
PP",
are each
equal
to
H,
and their
lines of action are
parallel
to the normal at
P,
the force at P
being
in the
negative
sense of this normal. The forces at P' and P
arising
from the
couple
on the arc P'P are each
equal
to
HBH,
and their lines of action are
parallel
to the normal at
P',
the force at P
being
in the
positive
sense of this normal.
Now let
i^i'", /i2*"
he the
principal
radii of curvature of the strained middle
surface at
P,
so that the
equation
of this surface referred to axes of
f, r),
z
which coincide with the
principal tangents
at P and the normal is
approxi-
mately
331-333]
AT THE EDGE OF A THIN SHELL
511
Also let
</)
be the
angle
which the
tangent
at P to P'PF' makes with the
axis of
^.
The
point
P" has coordinates

hs cos
^,

Bs sin
<j>,
0,
and the
direction cosines of the normal at F'
are,
with sufficient
approximation,
&cos<^/iJi<, Bssm<f>/R./'>,
1. The force at F
arising
from the
couple
on
F'F has
components i/&cos<^/iJi'",
HBs am
4>/R,^'\
H-SH
parallel
to the
axes of
f, r),
z. Hence the force at F
arising
from the
couples
on F'F and
FF" lias
components parallel
to the normal to s drawn on the
surface,
the
tangent
to s and the normal to the
surface,
which are
^Sssin<^cos^(l/ii,)-l/if,(')), Els\R',
-
SH,
where
R',
=
[cos-(^/i2,i"
+
sitfc^/i^,"*]-',
is the radius of curvature of the
normal section
having
the same
tangent
line as the curve s. Hence the
stress-resultants
T, S,
N and
stress-couples
H,
G can be
replaced by
stress-
resultants
r-l-i^sin20[l/i2,(')-l/ie,)),
S
+
H/R', N-dHjds, (47)
and a flexural
couple
G.
The
boundary
conditions at an
edge
to which forces are
applied,
or at a
free
edge,
can now be written down in the manner
explained
in Article 296.
The formulfe
(47)
are
simplified
in case the
plate
or shell is but little
bent,
for then the radii of curvature and the
position
of the
edge-line
relative to
the lines of curvature
may
be determined from the
unstrained,
instead of the
strained,
middle surface.
They
are
simplified
still more in case the
edge
is a
line of
curvature*,
for then H does not contribute to T.
333.
Theory
of the vibrations of thin shells.
The
equations
of vibration are to be formed
by substituting
for the*
external forces and
couples
X', Y',
Z' and
L',
M' which occur in
equations
(4.5)
and
(46)
of Article 331 the
expressions
for the reversed kinetic reactions
and their moments. If we
neglect "rotatory
inertia" the values to be
substituted for
L\
M' are zero. When we use the
components
u, v,
w of
displacement
defined in Article
326,
the
expressions
to be substituted for
{X', Y',
Z')
are
-
2ph (d^uI'dP,
dH)/df-, dhv/dt').
In
forming
the
equations
we omit all
products
of
u, v,
w and their
differential
coefficients; and,
since the stress-resultants
and
stress-couples
are linear functions of these
quantities,
we
may simplify
the
equations by
replacing p/,
...
by
their values in the unstrained
state,
that is to
say, by
the
values
given for^i,
... iu Article 323.
*
Tlie result
that,
in this
case,
H contributes to S as well as to N was noted
by
A. B.
Basset,
loc. cit.
p.
505. See also the
paper by
H. Lamb cited on
p.
477.
512
THEORY OF THE VIBRATIONS
[CH.
XXIV
The
equations (46)
of Article 331 become
1
\
d{H,B) d(G,A)
dA
rrdB)
1
\d(0,B) d{H,A)
dA
dB\
AB\
da
^
d^
~^'d0~^'da]~^'-^'
.(48)
and the
equations (45)
become
+
-j^
+
S, +
S,
=
0;
1
\d(3\B)
d(S,A)
dA dB)
N,
_.<
AB{
da
d/rf
'
'
8^ ^aaj R,
^
df'
1
\
d{S,B)
.
d{T,A )
dA
dB]
A\_^.d^v
AB\
da
^~W' ~-^'di:i~^'da\~R,~'^P''d'
}
-(49)
1
(8
{N,B)
d
(N,A)l
T,T,_
d^
AB\
da
^
d0~]
^Rr Rl
~
^
d
d'^w
df
'
The
equations (49),
some of the
quantities
in which are connected
by
the
relations
(48),
are the
equations
of vibration.
These
equations
are to be transformed into a
system
of
partial
differential
equations
for the determination of
u, v, w,
by expressing
the various
quantities
involved in them in terms of
u, v,
w and their differential
coefficients. This
transformation
may
be effected
by
means of the
theory given
in
preceding
Articles of this
Chapter. Equations
(37)
of Article 329
express G^, G^, H^, H^
in
terms of
k^, k.^, t,
and
equations (26)
of Article 326
express
/Cj, k^,
t in terms of
u, V,
w.
By
the first two of
equations (48)
therefoi-e we have
iV],
N.,
expressed
in terms of
u, v,
w.
Equations (36)
of Article 329
give
a first
approximation
to
Si, S^, 7\, T^
in terms of
e^, e^, sr,
and
equations (21)
of Article 326
express
,
ei, Ca,
TO- in terms of
u, v,
w. A closer
approximation
to
S^, S^, T^, T^
is
given
in
equations. (42)
and
(44)
of Article
330;
and
they
are there
expressed
in
terms of
K], ,
t as well as
ii,i^,tB\
so that
they
can still be
expressed
in
terms of
u, v,
w. When these
approximate
values are substituted in the
third of
equations (48)
it becomes an
identity.
When
iV"i, A^j, /S,, S^, T^, T^
are
expressed
in terms of
u, v, w,
the desired transformation is effected.
The
theo7-y of
the vibrations
of
a
plane plate, already
treated
provisionally
in Article 314
(d)
and
(e),
is included in this
theory.
In all the
equations
we have to take
1/jB,
and
l/Ri
to be zero. The
equations (48)
and
(49)
fall
into two sets. One set contains
d'^ujdP, d'^vjdP
and the stress-resultants of the
type
T,
S
;
the other set contains
d'^w/dt^,
the stress-resultants of
type
N,
and
the
stress-couples.
Now,
in this
case,
the stress-resultants of
type
T,
S are
expressible
in terms of
e,, 62,
or
by
the formulas
(36)
of Article
329,
and
,, 62,
w are
expressible
in terms of
u,
v
by
the formulae
_du _dv _dv
da
^'~da'
''~d(i' '^~da'^d^'
333]
OF PLATES AND SHELLS 513
a and
/3
being ordinary
Cartesian coordinates. Hence one of the two sets of
equations
into which
(-tS)
and
(49)
fall becomes identical with the
equations
of extensioiial vibration
given
in Article
314(e). Further,
the
stress-couples
are
expressible
in terms of
/C], k^,
t
by
the formula;
(37)
of Article
329,
and
Ki, K.,,
T are
expre.ssil)le
in terms of w
by
the formulae
_d'w _d'w _
d'w
"''d^'
"''d^'' ^~Sa8/i'
while
iV,
and
N2
are
expressible
in terms of the
stress-couples by
the
equations
_dG,dH,
^
_dG,
dH,
The second of the two sets of
equations
into which
(48)
and
(49)
fall is
equi-
valent to the
equation
of transverse vibration
given
in Article 314
(d).
In
applying
the results of Articles 329 and 330 to vibrations we make
a certain
assumption.
A similar
assumption
is,
as we noted in Article
277,
made
habitually
in the
theory
of the vibrations of thin rods. We assume in
fact that the state of strain within a thin
plate
or
shell,
when
vibrating,
is
of a
type
which has been determined
by using
the
equations
of
equilibrium.
For
example,
in the case of a
plane plate vibrating transversely,
we assume
that the internal strain in a small
portion
of the
plate
is
very nearly
the
same as that which would be
produced
in the
portion
if it were held in
equilibrium,
with the middle
plane
bent to the same curvature. Consider a
little more
closely
the state of a
cylindrical
or
prismatic portion
of a
plane
plate,
such as would fit into a fine hole drilled
transversely through
it. We
are
assuming
that,
when the
plate
vibrates,
any
such
prismatic portion
is
practically adjusted
to
equilibrium
at each instant
during
a
period.
This
being
.so,
the most
important components
of strain in the
portion,
when the
plate
vibrates
transversely,
are
given by
<ixx
=
-zic-^,
eyy
=
-ZKn,
exy
=
-'^ZT,
ea=[(Tl{].- a)]
z{Ki
+
k^),
and,
when it vibrates in its
plane, they
are
given by
exx
=
e-i,,
eyy
=
e2,
exy
=
'BT,
=

{o-li^ -cr)\ {e^+e^;
in both cases
e^^
is
adjusted
so that the
stress-component
Z^
vanishes. It is
clear that the
assumption
is
justified
if the
periods
of vibration of the
plate
are
long compared
with the
periods
of those modes of free vibration of the
prismatic
portion
which would involve strains of such
types
as are assumed. Now the
period
of
any
mode of transverse vibration of- the
plate
is
directly proportional
to the
square
of some linear dimension of the area contained within the
edge-line
and
inversely proportional
to the
thickness,
and the
period
of
any
mode of extensional vibration is
directly proportional
to some linear dimen-
sion of the area contained within the
edge-line
and
independent
of the
thickness,
while the
period
of
any
mode of free vibration of the
prismatic
portion, involving
strains of such
types
as those
assumed,
is
proportional
L. E.
33
514 THEORY OF THE VIBRATIONS
[CH.
XXIV
to the linear dimensions of the
portion,
or,
at an outside
estimate,
to the
thickness of the
plate.
There is
nothing
in this
argument peculiar
to a
plane plate
;
and we
may
conclude that it is
legitimate
to assume
that,
when a
plate
or shell is
vibrating,
the state of strain in
any
small
portion
is
practically
the
same,
at
any
instant,
as it would be if the
plate
or shell
were held in
equilibrium,
with its middle surface stretched and bent as it
is at the instant. We see also that we
ought
to make the reservation
that the
argument by
which the
assumption
is
justified
diminishes in
cogency
as the
frequency
of the mode of vibration increases*.
The most
important
result obtained
by
means of this
assumption
is the
approxi-
mate determination of the
stress-component Z,.
When there is
equilibrium
and the
plate
is
plane, i!j=0
to a second
approximation ;
when there is
equilibrium
and the middle surface
is
curved, Z,
vanishes to a first
approximation,
and
by
the second
approximation
we
express
it as
proportional
to
{Ifi

z^)
and to a function which is linear in the
principal
curvatures
and the
changes
of curvature. The results in
regard
to
Z,
as a function of h and z can be
illustrated
by
a
discussion,
based on the
general equations
of vibration of elastic solid
bodies,
of the vibrations of an infinite
plate
of finite thickness. Such a discussion has been
given
by
Lord
Rayleight ;
and from his results it can be shown
that,
in this
case,
there are classes
of vibrations in which
Z^
vanishes
throughout
the
plate,
and
that,
in the
remaining classes,
the
expression
for
Z,
can be
expanded
in
rising powers
of A and
z,
and the
expansion
contains no terms of
degree
lower than the fourth.
When the middle surface is curved the
components
of
displacement
u, V,
w must
satisfy
the differential
equations (49)
transformed as
explained
above,
and
they
must also
satisfy
the
boundary
conditions at the
edge
of the shell. At a free
edge
the flexural
couple
and the three linear com-
binations of the stress-res-ultants and the torsional
couple expressed
in
(47)
of Article 332 must vanish. The order of the
system
of
equations
is,
in
general, sufficiently high
to admit of the satisfaction of such
conditions;
but the actual solution has not been effected in
any particular
case.
A method of
approximate
treatment of the
problem depends upon
the
observation that the
expressions
for the
stress-couples,
and therefore also for
J^i, Ni,
contain as a factor J) or
^Eh^/{1

&"')
while the
expressions
for the
remaining
stress-resultants contain two
terras,
one
proportional
to
h,
and the
other to /t'. Both members of each of the
equations (49)
can be divided
by
h
;
and then those terms of them which
depend upon
e,, e^,
m are
independent
of
/(,
and the
remaining
terms contain ]\? as a factor. We should
expect
to
get
an
approximately
correct solution
by omitting
the terms in h^. When
this is done two of the
boundary
conditions at a free
edge,
viz.: those of the
type
G
=
0,
N

dHjds
=
0,
disappear
;
and the
system
of
equations
is of a
sufficiently high
order to admit of the satisfaction of the
remaining boundary
conditions. Since h has
disappeared
from the
equations
and
conditions,
the
*
The
argument
is
clearly applicable
with some modifications of detail to the
theory
of the
vibrations of thin rods.
t
London Math. Soc.
Proc,
vol. 20
(1889), p.
225,
or
Scientific Papers,
vol.
3, p.
249.
333]
OF PLATES AND SHELLS 515
frequency
is
independent
of the thickness. The extension of the middle
surface is the most
important
feature of the
deformation,
but it is
necessarily
accompanied by bending.
The
theory
of such extensional vibrations
may
be
obtained
very simply by
the
energy
method,
as was noted in Article 329.
The extensional modes of vibration of a thin shell are
analogous
to the
extensional vibrations of a thin
plane plate,
to which reference has
already
been
made in this Article and in
(e)
of Article 314. The consideration of the case
of a
slightly
curved middle surface shows at once that an
open
shell must
also
possess
modes of vibration
analogous
to the transverse vibrations of a
plane plate,
and
having frequencies
which aie much less than those of the
extensional vibrations. The existence of such mode-s of vibration
may
be
established
by
the
following argument
:

A
superior
limit for the
frequency
of the
gravest
tone can be found
by
assuming any
convenient
type
of vibration
; for,
in
any vibrating system,
the
frequency
obtained
by assuming
the
type
cannot be less than the least
frequency
of natural vibration*. If we assume as the
type
of vibration one
in which no line on the middle surface is altered in
length,
we
may
calculate
the
frequency by
means of the formulae for the kinetic
energy
and the
potential
energy
of
bending,
as in Article 321. Since the kinetic
energy
contains h as
a
factor,
and the
potential energy
li?,
the
frequency
is
proportional
to h. The
frequency
of such inextensional vibrations of a shell of
given
form can be
lowered
indefinitely
in
comparison
with that of
any
mode of extensional
vibration
by diminishing
/;. It follows that the
gravest
mode,
of vibration
cannot,
in
general,
be of extensional
typef.
If we assume that the vibration is of
strictly
inextensional
type
the forms
of the
components
of
displacement
as functions of
a,
/9
are,
as we saw in
Articles
319, 320,
and
326,
very narrowly
restricted. If
displacements
which
satisfy
the conditions of no extension are substituted in the
expressions
for
the stress-resultants and
stress-couples,
the
equations
of motion and the
boundary
conditions
cannot,
in
general,
be
satisfiedj.
It is
clear, therefore,
that the vibrations mu.'^t involve some extension. To constrain the shell to
vibrate in an inextensional mode forces would have to be
applied
at its
edges
and over its faces. When these forces are not
applied,
the
displacement
must differ from
any
which satisfies the conditions of no extension.
But,
in
any
of the
graver
modes of
vibration,
the difference must be
slight;
for,
otherwise,
the mode of vibration would be
practically
an extensional
one,
and
*
Lord
Eayleigh, Theory of
Sound,
vol.
1,
89.
t The case of a closed
sheet,
such as a thin
splierical
shell,
is an obvious
exception,
for there
can be no inextensional
displacement.
A shell of
given
small
thickness, completely
closed
except
for a small
aperture,
is also
exceptional
when the
aperture
is small
enough.
X
In the
particular
cases of
spherical
and
cylindrical
shells the failure of the inextensional
displacement
to
satisfy
the
equations
of motion and the
boundary
conditions can be
definitely
proved.
The case of
cylindrical
shells is dealt with in Article 334
(rf).
332
I
516 VIBRATIONS OF A THIN
[CH.
XXIV
the
frequency
could not be
nearly
small
enough.
From the form of the
equations
of vibration we
may
conclude that the
requisite
extension must
be
very
small over the
greater part
of the surface
;
but near the
edge
it
must be of sufficient
importance
to secure the satisfaction of the
boundary
conditions*.
334. Vibrations of a thin
cylindrical
shell.
It is convenient to illustrate the
theory by discussing
in some detail the
vibrations of a
cylindrical
shell. As in Article 319 we shall take a to be the
radius of the
shell,
and write x for a and
<^
for
/3,
and we shall
suppose
the
edge-line
to consist of two circles x
=
I and x
=

I.
According
to the
results of Article
326,
the extension and the
changes
of curvature are
given
by
the
equations
du
_
'
/
^^
A _dv
\ du
The
displacement being periodic
in
^
with
period
27r,
and the shell
being
supposed
to vibrate in a normal mode with
frequency p/Stt,
we shall take
u, V,
w to be
proportional
to
sines,
or
cosines,
of
multiples
of
<^,
and to a
simple
harmonic function of t with
period lirjp.
The
equations
of vibration then
become a
system
of linear
equations
with constant coefficients for the
determination of
u, v,
w as functions of x. We shall
presently
form these
equations;
but,
before
doing
so,
we consider the order of the
system.
The
expressions
for
e^, e^,
isr contain first differential coefficients
only;
that for
k^
contains a second dififerential coefficient. Hence
Gi
and
G^
contain second
differential
coefficients,
and
iV,
contains a third differential coefficient. The
third
equation
of
(49)
contains
d^w/daf
in a term which is omitted when we
form the
equations
of extensional vibration. Thus the
complete equations
of vibration will be of a much
higher
order than the
equations
of extensional
vibration. It will be seen
presently
that the former are a
system
of the 8th
order,
and the latter a
system
of the 4th order. The reduction of the order
of the
system
which occurs when the
equations
of extensional vibration axe
taken instead of the
complete equations
is of fundamental
importance.
It
does not
depend
at all on the
cylindrical
form of the middle surface.
*
The
difficulty arising
from the fact that inextensional
displacements
do not admit of the
satisfaction of the
boundary
conditions is that to which I called attention in
my paper
of 1888
(see
Introduction,
footnote
133).
The
explanation
that the
extension, proved
to be
necessary,
may
be
practically
confined to a narrow
region
near the
edge,
and
yet may
be
sufficiently
important
at the
edge
to secure the satisfaction of the
boundary conditions,
was
given
simul-
taneously by
A. B. Basset and H. Lamb in the
papers
cited on
pp.
50-5 and 477. These authors
illustrated the
possibility
of this
explanation by
means of the solution of certain statical
problems.
333,
334]
CYLINDRICAL SHELL
517
(a)
General
equations.
In accordance with what has been said
above,
we take
u=[/sinncj>coa{pt
+
e),
v=Vcosn<f)coa{pt
+
(),
w= Warn
ncj)
con
(pt
+
t), ...(50)
where
U, V,
Ware functions of x. Then we have
dU .
W+nV
'i
=
-^^^^^'P'>^iP^
+
f\
2=
8inn<^co8(pt
+
),
*
'
~
li^
am
ntp
con
[pt
+
t),
k^=
^
sm
m/>
cos
{pt
+
c),
T
=
-
T-
(
r+
M
W)
cos
n<^
cos
(jo<
+
f
).
Also
G^i
=
-/) sin
7M^
cos
{pt
+
()
(
-j-j
-
<r
^
J
,
f n ^ , ,, ^(
<^^
nV+n^W\
G^=
-Bsmn<t>coa(pt+()[a-
-^
-^
j,
ff,=Dcosn<f,co.(pt+.)'^-^ [n ^
+^j=-JT,.
The first two of
equations (48)
become
and we have
We have also
j^_dG,
IdH,
y_\^G, djh
*
dx "^a
d<f>
'
^~a
d<p~
dx
'
.r,=
-
2)cos<^cos(;,*
+
{-
^
-
-3
W+--
^
-
^^j
.
7T nf^ / _L N
'^
+
2o-^
Ki
2 + (r
K2~|
I',
=
i)[^^,(.,
+
,.)-2^-^ --2713^) -J,
.S,
=
4Z)(l-<r)[|,n7+r],
^^
=
iZ)(l-<r)[-|,w+^],
where
cj, k,,
... have the values
given
above. The
equations
of vibration are
or,
in terms of
U, V, W,
snrd tdU
W-\-nV\
l-tm/dV
^ nU\l^.

^
Idv [dx
-
"
^-
j
-
-T- a
W
+
VJJ
+
^''^^
+
AL2Tr=^^)^^+2(rr.)a3rfx('+'^")J
=
'^
("^
518 VIBRATIONS OF A THIN
[CH.
XXIV
m'y-^.E^rj^.,f^w
n (PV n^
,
^a'^djfi a**^^ a^ dx^
a^^J"'
^^^'
Ifi
\_a
dx
Dfd^W
J.n^cPW
ti^ wt-
h\_d.r*
o* cfo^
""
a<
^ "'
a^dx^^ a-
+
27rr^a^
W
-2(l^)a*('^^+'^)J=^
^^^^
The
boundary
conditions at :r=Z and x= -I are
and all the left-hand members can be
expressed
as linear functions of
U, V,
Wand their
differential coefficients with
resi^ct
to x.
The
system
of
equations
for the determination of
u, v,
w as functions of
x,
has now been
expressed
as a linear
system
of the 8th order with constant
coefficients. These coefficients contain the unknown constant
p^
as well as
the known constants h and n
;
and
>i,
being
the number of
wave-lengths
to
the
circumference,
can be chosen at
pleasure.
If we
disregard
the fact that
h is small
compared
with a or
I,
we can solve the
equations by assuming
that,
apart
from the
simple
harmonic factors
depending upon ^
and
t,
the
quantities
u, V,
w are of the form
^e'"*, i;e"^, fe'"*,
where
f, r), ^,
m are constants. The
constant m is a root of a determinantal
equation
of the 8th
degree,
which is
really
of the 4th
degree
in
m?,
for it contains no terms of
any
uneven
degree.
The coefficients in this
equation depend upon p^.
When m satisfies this
equation
the ratios
^
:
n
'
^
are
determined,
in terms of m and
p', by any
two of the three
equations
of motion.
Thus,
apart
from
<f>
and t
factors,
the
solution is of the form
r=\ r
=
\ r=\
in which the constants
^r, f/
are
arbitrary,
but the constants
//.
are
expressed
as
multiples
of them. The
boundary
conditions at x
=
l and
x
=

l
give eight homogeneous
linear
equations connecting
the
f, ^' ;
and
the elimination of the
^, f
from these
equations
leads to an
equation
to
determine
p'.
This is the
frequency equation.
(6)
Extensional vibrations.
The
equations
of extensional vibration are obtained
by omitting
the terms in
equations
(51)

(53)
which have the coefficient
DjL
The determinantal
equation
for m^ becomes a
quadratic.
The
boundary
conditions a,t x=
1
become
^1
=
0, <S'i
=
0,
or
dU W+nV

dV nU
^
o-
=0, ~j

I =0.
dx a dx a
334]
CYLINDRICAL SHELL
519
Since h does not occur in the differential
equations
or the
boiuidary conditions,
the
frequencies
are
independent
of h.
In the case of
symmetrical vibrations,
in which
u, v,
w are
independent
of
</>,
we take
U=UcO{pt-\-(), V=VcOS{pt
+
(),
W=WQ,Olii^pt-\-(),
and we find the
equations
The
boundary
conditions sA
x=l
are
dU W dV

-T
<r

=0,
-r-=0.
ax a ax
There are two classes of
symmetrical
vibrations. In the first class U and H'
vanish,
so that the
displacement
is
tangential
to the circular sections of the
cylinder.
In this
class of vibrations we have
,r
nnx E n^n^
V=,cos

, ^,^
=
2^-^^^-^,
where n is an
integer.
The.se vibrations are
analogous
to the torsional vibrations of a
solid
cylinder
considered in Article 200. In the .second class V
vanishes,
so that the
displacement
takes
place
in
planes through
the
axis,
and we find
,, ,
nnx .

. nnx
t/=fcos
T
,
W=Q8vaj,
where
|
and
f
are connected
by
the
equations
E<T nn
^^)
la
The
equation
for
p'^
is
,_ 2 E_ (\
n'^^V,
EH^^^
If the
length
is
great compared
with the
diameter,
so that
ajl
is
small,
the two
types
of
vibration are
(i)
almost
purely radial,
with a
frequency {Ejp (1

a'^)}'/2n-a,
and
(ii)
almost
purely longitudinal,
with a
frequency
n
{E/p)^/2l.
The latter are of the same kind as the
extensional vibrations of a thin rod
(Article 278).
A more detailed
investigation
of the extensional vibrations of
cylindrical
shells with
edges
will be found in
my paper
cited in the
Introdiiction,
footnote 133. For a shell of
infinite
length
the radial vibrations have been discussed
by
A. B.
Basset,
London Math.
Soc.
Proc,
vol. 21
(1891), p. 53,
and the various modes of vibration have been
investigated
very fully by
Lord
Rayleigh,
Proc. B.
Soc,
vol. 45
(1889), p. 443,
or
Scientific Papers,
vol.
3,
p.
244. See also
Theory of
Sound,
2nd
edition,
vol.
1, Chapter
x a.
(c)
Inextensional vibrations*.
The
displacement
in a
principal
mode of vibration is either two-dimensional and
given
by
the formulas
u
=
0,
v
=
Acoa{pt
+
e)coa{n(l>
+
a),
w=
-nAnCoa{pJ
+
(n)sm{m}>+a,),
*
See
Chapter xxiii.,
Articles 319 and 321.
520
VIBRATIONS OF A THIN
[CH.
XXIV
or else the
displacement
is three-dimensional and
given by
the formiiloB
u=
--j5cos(jB,'<
+
f')i"("^+/3). t'=^-Scos(jo7
+
f')cos(M(/>-|-/SJ,
w=-
nxB
cos
(pn'i+fn)
sin
((|) +0),
""^
^"
~2pAa4
m2
+
1
"r+3a2/M2(2+iy2-
All the values of
^
and
p'
are
proportional
to h.
(d)
Inexactness
of
the inexiensional
displacement.
To
verify
the
failure
of the assumed inextensional
displacement
to
satisfy
the
equations
of
motion,
it is sufficient to calculate
T2
from the
equations
of
motion,
and
compare
the
result virith the second of the formulae
(44). Taking
the two-dimensional vibration
specified
by
^,
we have the
equation
a- dx
ad<j>
^''"P'^'"
a^
^(l-^^2:j:ij^n(W>+)
cos
(p -)-)
2i)3(2-l)
Dn^{n^-\)
J

sin
(m(/)
-f-
a)
cos
{pt
+
f
) ;
{n^+\)a*
but we have also
^~
2(l-(r)
a
2
+a-
Dn{n^-1) .-,... ,
.
, ^
~2li^^)
fflS
^'^sm(n(^-fa)co8(^*-fO-
The two values of
T2
are
difierent,
and the
equations
of motion are not satisfied
by
the
assumed
displacement.
It is clear that a correction of the
displacement involving
but
slight
extension would enable us to
satisfy
the differential
equations.
Two of the
boundary
conditions are
^1=0, Ni-a-^dffild<t>
=
0. When the vibration
is
two-dimensional, Oi
is
independent
of
x,
and cannot vanish at
any particular
value of
X unless
A
=
0. When the vibration is
three-dimensional, Ni
and
Hi
are
independent
of
X,
and
^i

a~^dlfijd<f>
cannot vanish at
any particular
value of x unless
-fl
=
0. Thus the
boundary
conditions cannot be satisfied
by
the assumed
displacement.
The correction of
the
displacement required
to
satisfy
the
boundary
conditions would
appear
to be more
important
than that
required
to
satisfy
the differential
equations.
(e)
Nature
of
the correction to he
applied
to the inextensional
displace-
ment.
It is clear that the existence of
practically
inextensional vibrations is
connected with the fact
that,
when the vibrations are taken to be
exten.sional,
the order of the
system
of
equations
of vibration is reduced from
eight
to
four. In the determinantal
equation
indicated in
(a)
of this Article the
terms which contain w* and m" have h'^ as a
factor,
and thus two of the values
of rn' are
large
of the order
l/h.
The
way
in which the solutions which
depend
on the
large
values of m would enable us to
satisfy
the
boundary
334]
CYLINDRICAL SHELL
521
conditions
may
be illustrated
by
the solution of the
following
statical
problem*
:

A
portion
of a circular
cylinder
bounded
by
two
generators
and two
circular sections is held bent into a surfece of revolution
by
forces
applied
along
the
bounding generators,
the circular
edges being
free,
in such
a
way
that the
displacement
v
tangential
to the circular sections is
proportional
to the
angular
coordinate
(^
;
it is
required
to find the
displacement.
We are to have v
=
c4),
where c is
constant,
while u and w are
indepeudent
of
0.
Hence
du c

w .
d'lv c
''^dx' '^=~^
'
^"
=
0' '^'
=
3^' "^^a^^
^
=
0-
The stress-resultants
)S,,
& and the
stress-couples
H^, H^ vanish,
and we
have
The
equations
of
equilibrium
are
dx
'
d<f>
'
dx^^ a

and the
boundary
conditions at a;
=
+ Z are
T,
=
0, N,=0, G,=0.
We seek to
satisfy
these
equations
and conditions
approximately by
the
assumption
that the extensional strains
e,, e^
are of the same order as the
flexural strains
hxi, Iik^.
When this is the case
Ti
and
T^
are
given
with
sufficient
approximation by
the formulae
T,
=
(SD/h')
(,
+
ae,), T,
=
(3D/h') (e,
+
<re,).
To
satisfy
the
equation dTJdx
=
and the condition
T,
=
at x=
1
we
must
put
T]
=
0,
or
e,
=

ae^,
and then we have
Tj
=
3Z)(1

a^) ejh^.
The
equations
of
equilibrium
are now reduced to the
equation
while the
boundary
conditions At x=
1
become
d^w o'C
_
?^w
_
_
If we take c-w to be a sum of terms of the form
fe"",
then m' is
large
of
the order
Ijh
;
and the solution is found to be
w
=
c+Ci
cosh
{qxja)
cos
{qxja)
-t-
Cj
sinh
{qxja)
sin
{qxja),
*
This is the
problem
solved for this
purpose by
H.
Lamb,
loc. cit.
p.
477. The same
point
in the
theory
was illustrated
by
A. B.
Basset,
loc. cit.
p.
505, by
means of a different statical
problem.
522
VIBRATIONS OP A THIN
[CH.
XXIV
where
and
q'
=
(a|2h)^/{Sil-<7%
_
o-c sinh
(qlla)
cos
(glla)

cosh
(ql/a)
sin
(qlla)
'
~
~
^ 6inh(2qlla)+^m (2qlla)
^
_
trc sinh
{qlla)
cos
{qlla)
+
cosh
(yVa)
sin
{qlja)
''~~'f
sinh
(2gi/a)
+
sin
(2gi/tt)
The form of the solution shows that near the boundaries
61, 2, hx^,
hK are all
of the same order of
magnitude,
but
that,
at a distance from the boundaries
which is at all
large compared
with
{ah)i,
e^
and
e^
become small in com-
parison
with
hK^.
It
may
be shown
that,
in this statical
problem,
the
potential energy
due
to extension is
actually
of the order
V(^/a)
of the
potential energy
due to
bending*.
In the case of vibrations we
may
infer that the extensional
strain,
which is
necessary
in order to secure the satisfaction of the
boundary
con-
ditions,
is
practically
confined to so narrow a
region
near the
edge
that its
effect in
altering
the total amount of the
potential energy,
and therefore the
periods
of
vibration,
is
negligible.
335. Vibrations of a thin
spherical
shell.
The case in which the middle surface is a
complete spherical
surface,
and
the shell is
thin,
has been
investigated by
H.
Lambf by
means of the
general
equations
of vibration of elastic solids. All the modes of vibration are
extensional,
and
they
fall into two
classes,
analogous
to those of a solid
sphere investigated
in Article
194,
and characterized
respectively by
the
absence of a radial
component
of the
displacement
and
by
the absence of a
radial
component
of the rotation. In
any
mode of either class the
displace-
ment is
expressible
in
terms
of
spherical
surface harmonics of a
single integral
degree.
In the case of vibrations of the first class the
frequency |)/27r
is
connected with the
degree
n of the harmonics
by
the
equation
fa'plfi
=
{n- l){n
+
2), (54)
where a is the radius of the
sphere.
In the case of vibrations of the second
class the
frequency
is connected with the
degree
of the harmonics
by
the
equation
(55)
If n exceeds
unity
there are two modes of vibration of the second
class,
*
For farther details in
regard
to this
problem
the reader is referred to the
paper by
H. Lamb
already
cited.
t
iMndon Math. Soc.
Proc.,
vol. U
(1883), p.
50.
334,
335]
SPHERICAL SHELL
523
and the
gravest
tone
belongs
to the slower of those two modes of
vibration
of this class for which n
=
2. Its
frequency jp/27r
is
given by
i)
=
V(W/')a-'(l-176),
if Poisson's ratio for the material is taken to be
|.
The
frequencies
of all
these modes are
independent
of the thickness.
In the
limiting
case of a
plane plate
the modes of vibration fall into two
main
classes,
one
inextensional,
with
displacement
normal to tlie
plane
of the
plate,
and the other
extensional,
with
displacement parallel
to the
plane
of
the
plate. [See
Articles 314
(d)
and
(e)
and 333 and Note F at the end of
the
book.]
The case of an infinite
plate
of finite thickness has been discussed
by
Lord
Rayleigh
*,
starting
from the
general equations
of vibration of ela.stic
solids,
and
using
methods akin to those described in Article 214
supra.
There
is a class of extensional vibrations
involving displacement parallel
to the
plane
of the
plate ;
and the modes of this class fall into two
sub-classes,
in one of which there is no
displacement
of the middle
plane.
The other
of these two sub-classes
appears
to be the
analogue
of the
tangential
vibra-
tions of a
complete
thin
spherical
shell. There is a second class of extensional
vibrations
involving
a
component
of
displacement
normal to the
plane
of the
plate
as well as a
tangential component,
and,
when the
plate
is
thin,
the
normal
component
is small
compared
with the
tangential component.
The
normal
component
of
displacement
vanishes at the middle
plane,
and the
normal
component
of the rotation vanishes
everywhere;
so that the vibrations
of this class are
analogous
to the vibrations of the second class of a
complete
thin
spherical
shell. There is also a class of flexural vibrations
involving
a
displacement
normal to the
plane
of the
plate,
and a
tangential component
of
displacement
which is small
compared
with the normal
component
when
the
plate
is thin. The
tangential component
vanishes at the middle
plane,
so that the
displacement
is
approximately
inextensional. In these vibrations
the linear elements which are
initially
normal to the middle
plane
remain
straight
and normal to the middle
plane throughout
the
motion,
and the
frequency
is
approximately proportional
to the thickness. There are no
inextensional vibrations of a
complete
thin
spherical
shell.
The case of an
open spherical
shell or bowl stands between these extreme
cases. When the
aperture
is
very
small,
or the
spherical
surface is
nearly
complete,
the vibrations must
approximate
to those of a
complete spherical
shell. When the
angular
radius of the
apertiu-e,
measured from the included
pole,
is
small,
and the radius of the
sphere
is
large,
the vibrations must
approximate
to those of a
plane plate.
In intermediate cases there must be
vibrations of
practically
inextensional
type
and also vibrations of extensional
type.
*
London Math. Soc.
Proc,
vol. 20
(1889), p. 225,
or
Scientific Papers,
vol.
3, p.
249.
524 VIBRATIONS OF A THIN
[CH.
XXIV
Purely
inextensional vibrations of a thin
spherical
shell,
of which the
edge-line
is a
cii'cle,
have been discussed in detail
by
Lord
Rayleigh* by
the
methods described in Article 321
supra.
In the case of a
hemispherical
shell the
frequency p/iir
of the
gravest
tone is
given by
P
=
V(Wp)(AK) (4-279).
When the
angular
radius a of the
aperture
is
nearly equal
to
tt,
or the
spherical
surface is
nearly complete,
the
frequency ^/27r
of the
gravest
mode of
inextensional vibration is
given by p
=
':/(fi/p){h/a'{-n-

ay}(5'657). By
supposing
Tr

a to diminish
sufficiently,
while /) remains
constant,
we can
make the
frequency
of the
gravest
inextensional mode as
great
as we
please
in
comparison
with the
frequency
of the
gravest (extensional)
mode of vibration
of the
complete spherical
shell. Thus the
general argument by
which we
establish the existence of
practically
inextensional modes breaks down in the
ca.se of a
nearly complete spherical
shell with a small
aperture.
When the
general equations
of vibration are formed
by
the method
illu.strated above in the case of the
cylindrical
shell,
the
components
of
displacement being
taken to be
proportional
to sines or cosines of
multiples
of the
longitude <f),
and also to a
simple
harmonic function of
t,
they
are a
system
of linear
equations
of the 8th order for the determination of the com-
ponents
of
displacement
as functions of the co-latitude 6. The
boundary
conditions at the free
edge require
the
vanishing,
at a
particular
value of
6,
of four linear combinations of the
components
of
displacement
and certain of
their differential coefficients with
respect
to 0. The order of the
system
of
equations
is
high enough
to admit of the satisfaction of such conditions
;
and
the solution of the
system
of
equations, subject
to these
conditions,
would
lead,
if it could be
effected,
to the determination of the
types
of vibration and
the
frequencies.
The extensional vibrations can be
investigated by
the method illustrated
above in the case of the
cylindrical
shell. The
system
of
equations
is of the
fourth
order,
and there are two
boundary conditionsf.
In
any
mode of
vibration the motion is
compounded
of two
motions,
one
involving
no radial
component
of
displacement,
and the other no radial
component
of rotation.
Each motion is
expressible
in terms of a
single spherical
surface
harmonic,
but the
degrees
of the harmonics are not in
general integers.
The
degree
a
of the harmonic
by
which the motion with no radial
component
of
displace-
ment is
specified
is connected with the
frequency by equation (54),
in which
a is written for n
;
and the
degree
y8
of the harmonic
by
which the motion
*
London Math. Soc.
Proc,
vol. 13
(1881),
or
Scientific Papers,
vol.
1,
p.
551. See also
Theory of
Sound,
2nd
edition,
vol.
1,
Chapter
x a.
+ The
equations
were formed and solved
by
E.
Mathien,
J. de I'Ecole
potytechnique,
t. 51
(1883).
The extensional vibrations of
spherical
shells are also discussed in the
paper by
the
present
writer cited in the
Introduction,
footnote 133.
335,
336]
SPHERICAL SHELL
525
with no radial
component
of rotation is
specified
is connected with the
frequency by equation (55),
in which is
written for n. The two
degrees
a and
/3
are connected
by
a
transcendental
equation,
which is the
frequency
equation.
The vibrations do not
generally
fall into classes in the same
way
as those of a
complete
shell
; but,
as the
open
shell
approaches completeness,
its modes of extensional vibration tend to
pass
over into those of the com-
plete
shell.
The existence of modes of vibration which are
practically
inextensional is
clearly
bound
up
with the fact
that,
when the vibrations are assumed to be
extensional,
the order of the
system
of differential
equations
of vibration is
reduced from 8 to 4. As in the case of the
cylindrical
shell,
it
may
be shown
that the vibrations cannot be
strictly inextensional,
and that the correction
of the
displacement required
to
satisfy
the
boundary
conditions is more
important
than that
required
to
satisfy
the differential
equations.
We
may
conclude
that,
near the free
edge,
the extensional strains are
comparable
with
the flexural
strains,
but that the extension is
practically
confined to a narrow
region
near the
edge.
If we trace in
imagination
the
gradual changes
in the
system
of vibrations
as the surface becomes more and more
curved*,
beginning
with the case of a
plane plate,
and
ending
with that of a
complete spherical
shell,
one class of
vibrations,
the
practically
inextensional
class,
appears
to be
totally
lost. The
reason of this would seem to lie in the
rapid
rise of
frequency
of all the
modes of this class when the
aperture
in the surface is much diminished.
The theoretical
problem
of the vibrations of a
spherical
shell
acquires
great practical
interest from the fact that an
open spherical
shell is the best
representative
of a bell which admits of
analytical
treatment. It
may
be
taken as established that the vibrations of
practical importance
are inex-
tensional,
and the essential features of the
theory
of them
have,
as we have
seen,
been made out. The tones and modes of vibration of bells have been
investigated experimentally by
Lord
Rayleigh-f-.
He found that the nominal
pitch
of a
bell,
as
specified by English
founders,
is not that of its
gravest
tone,
but that of the tone which stands fifth in order of
increasing frequency ;
in this mode of vibration there are
eight
nodal meridians.
336. Problems of
equilibrium.
When a thin
plate
or shell is held deformed
by externally applied
forces,
the strained middle surface
must,
as we observed in Article
315,
coincide
very nearly
with one of the surfaces
applicable upon
the unstrained middle
surface. We
may
divide the
problem
into two
parts:
(i)
that of
determining
*
The
process
is
suggested by
H. Lamb iu the
paper
cited on
p.
477.
t
Phil.
Mag. (Ser. 5),
vol. 29
(1890), p.
1,
or
Scientific Papers,
voL
3, p. 318,
or
Theory of
Sound,
2nd
edition,
vol.
1, Chapter
X.
526
EQUILIBRIUM
CONFIGURATIONS
[CH.
XXIV
this
applicable
surface,
(ii)
that of
determining
the small
displacement by
which the strained middle surface is derived from this
applicable
surface.
This is the
procedure adopted by
Clebsch* in his treatment of the
problem
of finite deformation of
plane plates.
It
appears
that there is some
degree
of
indefiniteness
attaching
to this division of the
problem,
since
any
one of the
surfaces
applicable upon
the unstrained iniddle
surface,
and derivable one
from another
by displacements
of the order which we
regard
as
small,
would
serve
equally
well as a solution of the first
part
of the
problem.
Greater
precision may
be
imparted
to the
procedure
if we
regard
the two
steps
as
(i)
the determination of an inextensional
displacement,
which need not be
small,
(ii)
the determination of an additional
displacement involving
ex-
tensional strains at least of the same order of
magnitude
as the additional
flexural
strains,
and
possibly large
in
comparison
with them.
The first
step
is
analogous
to the determination of
equilibrium
configura-
tions of a thin
rod,
discussed in
Chapters
XIX. and xxi.
; but,
unless the
displacement
is
small,
little
progress
can be made. When the
displacement
is
small,
it
is,
as we
know,
very narrowly
restricted as
regards
its functional
character. This restriction carries with it a notable difference in
procedure
between the
problems
of rods and of
plates
or
shells,
and it also increases the
theoretical,
though
not the
practical, importance
of the second
step
in the
solution of the
problem.
These
points may
be illustrated
by
a
particular problem
:

Let a
hemispherical
shell be
deformed
by
a
string
stretched
tightly
with tension Fbetween two
oppo.site jwints
on its
edge.
In the notation of Article 320 we take these
points
to be 6
=
^t!-,
=
and
(f>
=
ir,
and we
suppose
that the
pole
6
=
ia included. The
type
of small inextensional
displace-
ment is
given by
the
equations
u

in0 2
Ant&n"-
cos
n(f>,
v=sm6 2
Jtan"- sinjid),
71
=
1!
^
n=2
"
!f= 2
(7i-{-cosd)Aia,n'^gCX)and).
-2
-i
The
potential energy
of
bending
V is
given according
to Article 321
(6) by
the
equation
r=|.,g2[^-i)M^i^J
=
r/i
~
2m
(m2
-
1) (2?i2
-
1) ^2.
The work done
by
the tension of the
string
in a small
displacement
is
i^2(l+cosnjr)8>^l,
and the increment of the
potential energy
of
bending
is
|,r,*5^2(2-l)(2n2-l)J8^.
*
Elasticitat,
70.
336]
OF A THIN SHELL
527
Hence we have
.
,
Fa^ 1+C08nr
jr^AS(2-l)(2n-l)'
80 that
A
vanishes when n is
odd, and,
when n is
even,
The inextensional
displacement
is now determined*.
In this solution the
necessity
of
satisfying boundary
conditions at the
edge
is left out
of
account,
and it is on account of these conditions that the second
step
in the
complete
solution,
viz. : that of
determining
a
subsidiary
extensional
displacement, acquires
so much
theoretical
importance.
From the first
part
of the solution we could calculate the flexural
couple
and the radial stress-resultant at the
edge.
In the case of small
displacements
the
equations
of
equilibrium
under no forces
except
at the
edge
are formed
by omitting
the
kinetic reactions in the
equations
of vibration. We thus have the forms of those
equations,
and we know that
they
are of a
sufficiently high
order to admit of the satisfaction of the
conditions
(i)
the tension and
shearing
force at the
edge vanish, (ii)
the flexural
couple
and
the radial stress-resultant at the
edge
have
given
values. If we take these
given
values to
be those calculated from the first
part
of the solution with revei-sed
signs,
the
displacement
which satisfies the
equations
of
equilibrium
and the
boundary
conditions is the
required
subsidiary displacement.
As in the case of vibrations the
subsidiary displacement
diminishes
rapidly
as the distance from the
edge increases,
and becomes
very
small as
soon as the distance from the
edge
is a considerable
multiple
of the mean
proportional
between the radius and the thickness. The method of
determining
the
subsidiary
displacement
in the case of a
cylindrical
shell was illustrated in Article 334
(e),
where the
inextensional
displacement
was u
=
0,
v
=
c(l),
w=c.
There are cases in which the first
part
of the solution
may
be omitted.
For
example,
no inextensional
displacement
can be
produced
in a
spherical
bowl
by
forces which are
symmetrically
distributed round the axis. The
bowl is
very
stiff,
but
not,
of
course,
infinitely
stiff,
to resist such forces. The
method of solution in such cases
may
be illustrated
by
the
problem
of a
hemispherical
bowl
resting
with its
edge
on a smooth horizontal
plane
and
deformed
by
its own
weight.
Specifying
the
displacement by components
u, v,
w as in Article
320,
and
using
the
results of Article
326,
we find
.^^^If?"
,,\ .,=
-l-J^Uuco,6-w.ine),
^
=
lg
+
JL^-.cot<)V ^
a\de /'
a am 6
\dct>
) a\o6 8mtfe<^ J
The stress-resultants are
given by
the formulae
(36)
of Article
329,
viz. :
The
equations
of
equihbrium
become
g>-K^,-,S,)
cot 5 +
4-^11
=
0, T,
+
T,
+
2ffpha
cos 6
=
0.
*
The method and this
example
of its
application
arc due to Lord
Kayleigh,
London ilath.
Sqc.
Proc,
vol. 13
(1881),
or
Scientific Papers,
vol.
1,
p.
651,
or
Theory of Sound, Chapter
xa.
528 STABILITY OF A THIN PLATE OR SHELL
[CH.
XXIV
Now,
the forces
being independent
of
<^,
the
displacements
are also
independent
of
<j),
and thus these
equations
become
.||^(^g_^)
+
,(cot5--)]
+
(l-<r)cotdg-cot5)+^^^!ii=^sind
=
0,
(^-v
cot 5
j
+
2 cot 6
(J^
-
V cot
e]=0,
.
d_
/8c
^+,.cotg-2..+
^^"'^J,-"
)cos^
=
0.
The
boundary
conditions at 5
=
i7r
are u
=
0,
v=0. The solution is
tt=2_ sin^-tan
-
+
sin51og(l+cos^) ,
v
=
0.
v=^^\ -^cosd-l+cos51og(l+cosd)
.
//i |_i
+(r
J
337. Problems of
stability.
In accordance with the
general principles explained
in Article 267 we see
that an extensional
configuration
of
equilibrium
of a thin
plate
or shell can
be unstable if there can exist both an extensional and an inextensional con-
figuration
of
equilibrium
with the same
externall}' applied
forces. In such
cases interest is centred in the determination of critical values for the
external
forces,
or for the linear dimensions of the
plate
or
shell,
which must
not be exceeded if the
system
is to be stable. We illustrate some methods
appropriate
to such
questions by
means of two
problems.
(a) Buckling of
a
rectangular plate
under thrust in
its
plane.
When the
length
and breadth of the
plate,
or the thrust at the
edge,
are not too
great,
the
plate simply
contracts in its
plane,
in the manner indicated in Article 301
;
but when
the linear
dimensions,
or the
thrusts,
are
great enough
it bends. We shall
suppose
that
it is
very slightly
bent.
We take the centre of the
rectangle
as
origin,
and lines
parallel
to the
edges
as axes of
X and
y,
and use x and
y
instead of a and
j3
in the formulae of Article
326,
in which we
put
^
=
JS
=
1 and
V/fi
=
1
IRi
=
0. We find
Also we have
P'-~^'-d^y'
?'=-a^'
V'
=
df'
''=^2.
'2-g^-
Omitting products
of diflFerential coefficients of
u, v, w,
wo find from
equations (46)
of
Article 331
'
ex
\0x' Cy'J'
'
cy \da^ dy^j
The first two of
equations (45)
of the same Article are satisfied
approximately,
when
X', Y',
Z'
vanish, by putting Ti
and
T^ equal
to constants and
-S,
and
S'2 equal
to zero.
We take
T\=

Pi, T'i^
-
P2,
336,
337]
COLLAPSE of boiler flues
529
where
/"j
and
P^
are the thrusts at the
edges
a;
=
con8t. and
y=cou8t.,
each estimated
per
unit of
length
of the
corresponding edge.
The third of
equations (46)
becomes
If the
plate
is
"supiwrted"
at the
edges
x=
a
and
y= +6,
we must have w=0 and
6*1=0
at x=
a,
and w=0 and
6*2
=
at
y=
b.
We have a solution of the form
wWsm \r ism
~-
-,
2a 26
where m and n are
integers
and W is a
constant, provided
that
This
equation gives
the critical thrusts. For
example,
if
I'i
=
l'i,
the critical value of
Pi
and
Pi
is
iZ)7r2(l/a2+l/6'^)*.
(6) Collapse of
a tube under external
pressure.
When a
cylindrical
shell of circular section is
subjected
to external
pressure p,
which
is not too
great,
it suflers a
purely
radial
displacement,
the amount of which can be
calculated
by
the method of Article 100
; but,
when
p
is too
great,
the shell bends under
the
pressure.
In the case of a
long cylinder, supposed
to bend
slightly
in two dimensions
without
stretching,
the
displacement
is
given, according
to Article
319, by
the
formulae,
M=0,
=2^cosn(^,
w=
-'SnA^amrKl),
and therefore we have
Ki=0, T=0, ic2
=
2{i(m2 l)/a'}.48in?^.
According
to the formulie
(24)
and
(25)
of Article
326,
all the
quantities p\,...
vanish
except p-i,
which is
1+0x2-
We shall write
p~!=:alR,
/?
=

a2)C2
+
...,
where R is the radius of curvature of the deformed cross-section of the middle surface.
The
ordinary approximation
to the
stress-couples
t
gives
6]=-Z)o-K2, (?2=-2>K2, Hi=-Hi=0;
and the first two of
equations (46) give
.v,_o, ly,-
-^.
The second and third of
equations (45) give
iar2_A'2_ 13^2, 2:2, o
a
dcj,
R
'
a
d4>
"^
R'^P"^-
Eliminating T2
from
these,
we find
1
S_fRd^,\
N,
pdR
ad(l>\a d(t>J'^ R'^ad(t>
'
*
The
problem
is
strictly analogous
to tliat of the
doubly pivoted
strut considered in Article
264. The above solution is due to G. H.
Bryan,
Loudon Math. Soc.
Proc,
vol. 22
(1891), p. 54,
who discusses a number of
special
cases.
t
It is assumed that the existence of
pressure
on the outer surface of the tube does not
seriously
affect the first
approximation
to the strain. The second
approximation
is not
required
for the calculation of the
stress-couples.
L. E. 34
530
COLLAPSE OF BOILER FLUES
[CH.
XXIV
or, neglecting
the
square
of
axo,
There can be a sohition in which
k-i
is
proportional
to sin
n<f)
if
^=^3(-l).
Hence the least value of
p
for which there can be a form other than circular is
SDja^.
We infer that the circular
cylinder
is unstable if the external
pressure
exceeds
3Z)/a'*.
The result
just
obtained admits of
application
to the
problem
of the
collapse of
boiler
flues.
The
pressure
of steam in a boiler is much in excess of the
pressure
of the air
in the
flues,
and it is found that
long
flues tend to
collapse
under the
pressure.
To obviate
this weakness it is usual to construct the flues in several detached
pieces
with massive
flanged joints,
thus
shortening
the effective
length
of the flue to the distance between
consecutive
joints.
Our result is that a flue of infinite
length
will not
collapse
unless the
pressure
exceeds
[2.7(1
-
cr^)] {hjaY,
where E and o- denote
Young's
modulus and Poisson's
ratio for the
material,
and
hja
is the ratio of the thickness to the diameter. The
portion
of the flue between two
joints
is
effectively
a thin
cylindrical
shell with fixed
ends,
and the
fixity
of the ends has the
consequence
that the middle surface cannot be bent without
stretching. If, howe%-er,
the
pressing
exceeds the critical
pressure,
and the
length
is
sufficiently great,
the extension
may
be
practically
confined to a narrow
region
near the
ends,
while the
greater part
of the surface bends almost without extension.
The most
interesting question
to be settled concerns the critical
length,
or the least
length
for which
collapse
is
possible
under the critical
pressure.
An exact numerical
value cannot be
obtained,
but an indication of the relations between the various dimensions
of the flue can be
gathered
from the
principles explained
in Article 334
(e).
For
collapse
to be
possible,
the effective
length,
or the distance between the
joints,
must be
great enough
for the inextensional
configuration
to be established over the
greater part
of the
length,
in
other
words,
it must be
great enough
to secure that the
subsidiary
extensional
displacement
required
to
satisfy
the terminal conditions shall diminish to a
negligible quantity
between
an end and the middle of the flue. From the method of solution
adopted
in Article 334
(e)
we can see at once that the distance
required
must be a
large multiple
of the mean
proportional
between the thickness and the diameter. It would
appear
therefore
that,
in
flues of different
sizes,
the rule for
spacing
the
joints, by
whicli the flues are
protected
against collapse, ought
to be : The distance between the
joints
should be
proportional
to
the
geometric
mean between the thickness and the diameter.

The result is due to G. H.

Bryan, Cambridge
Phil. Soc.
I'roc,
vol. 6
(1888), p.
287. The
analogous
result for a
ring
is
given
in Article 275
supra.
NOTES.
NOTE A.
Terminology
and notation.
Questions
of
notation,
and of the most
apjiropriate nomencktm-e,
for
elasticity
have
been much discussed. Keference
may
be made to the
writings
of W. J. M.
Rankine',
to Lord Kelvin's account of Rankine's
nomenclature^,
to K. Pearson's^ efforts after
consistency
and
uniformity,
to
pronouncements
on the
subject by
H. Lamb* and
W.
Voigf^.
The
following
tables show some of the more
important
notations for strain-
components
and
stress-components.
Strain-components.
Text"
532 NOTE A. TERMINOLOGY AND NOTATION
Kelvin and Tait's notation for
strain-components
and
stress-components
has been
adopted by
Lord
Rayleigh
and J. H.
Michell, among others,
and it was used in the first
edition of this book. KirchhofiTs notation for
stress-components
has met with
very general
acceptance,
but there seems to be no
equally suggestive
and convenient notation for
strain-components.
The notation
A',,, Yy, Z,
for the
components
of traction across a
plane,
the normal to which is in the direction
k,
is
supported by Voigt'.
The word "shear" has been used in the sense attached to it in the te.xt
by
Kelvin and
Tait. Rankine'2
proposed
to use it for what has here been called
"tangential
traction."
The word "traction" has been used in the sense attached to it in the te.\t
by
Kelvin and
Tait. Pearson^ uses "traction" in the sense here attached to "tension." The strains
which have here been called "extension" and
"shearing
strain" have been called
by
him
"stretch" and "slide." It
appears
to be desirable to maintain a distinction between
"
simple
shear,"
or
"pure shear,"
and
"shearing strain,"
and also between
"tangential
traction" and
"shearing
stress."
The "stress
equations"
of
equilibrium
or motion
(Article 54)
are called
by
Pearson^
"body-stress-equations,"
and the
equations
of
equilibrium
or motion in terms of
displace-
ments
(Article 91)
are called
by
him
"body-shift-equations."
The terms
"Young's
modulus," "rigidity,"
"modulus of
compression" (Articles 69, 73)
are
adopted
from Kelvin
and
Tait^;
these
quantities
are called
by
Pearson^ the
"stretch-modulus,"
the "slide-
modulus,"
and the "dilatation-modulus." The number here called "Poisson's ratio" is
called
by
Pearson^ the
"stretch-squeeze
ratio."
For
isotropic
solids Lamd" introduced the two constants X and
/j
of Article
69;
^
is
the
rigidity
and X
-|-|n
is the modulus of
compression.
Kelvin and Tait and Lord
Rayleigh
have used the letter n to denote the
rigidity.
Saint-Venant^ used the letter O.
Many
writers,
including
Clebsch and Kelvin and
Tait,
have used the letter
^,
as it is used in this
book,
to denote
Young's
modulus
;
in Lord
Eayleigh's Theory of
Sound the letter
q
is used.
Poisson's
ratio,
here denoted
by a,
has been denoted so
by
Kelvin and
Tait,
Clebsch and
Lord
Rayleigh
have denoted it
by /j.
Saint-Venant and Pearson
by i;.
In
many
of the
writings
of Italian elasticians the constants
(X-^2/i)/p
and
fujp
are
used,
and denoted
by
Q.'^ and w^
;
Q and a> are the velocities of irrotational and
equivoluminal
waves. Kirchhoff*
used two constants which he denoted
by
K and
6;
K \a the
rigidity,
and 6 is the number
o-/(l
-
2(r),
where a- is Poisson's ratio. Kelvin and Tait
''
used two constants
to,
n connected
with Lamd's X and
/i by
the
equations
m
=
X-(-^, n=fj..
In the case of
eeolotropic
solids there are
comparatively
few
competing
notations.
Pearson' has
suggested
the
following
notation for the elastic constants which we have
denoted after
Voigt^ by cn,...:

c=
I
xxxx
I , c,2=
I xxyy | ,
...
C44 | yzyz \ ,
....
The rule is that
any
suffix
1,
2 or 3 is to be
replaced by xx, yy
or
zz,
and
any
suffix
4,
5 or 6 is to be
rejJaced by yz,
zx or
xy.
The two first letters in
any symbol
refer to a
component
of
.stress,
as
X^,
and the two last letters to a
component
of strain as
e^^.
The
letters in either of these
pairs
can be
interchanged
without
altering
the
meaning
of the
symbol.
The conditions
(c,r=c), expressing
that there is a
strain-energy-function,
are
represented by
the statement that the two
pairs
of letters in a
symbol
are
interchangeable.
Cauchy's
relations
(Article 66)
amount to the statement that the order of the letters is
indifferent.
The constants
by
which the strain is
expressed
in terms of the
stress,
denoted in
Articles 72 and 73
by Cn/n, ...,
are denoted
by Voigt* hy ii, ...,
and this
usage
has been
"
Applied
Mechanics.
NOTE B. STRESS
588
followed
by
Liobisch".
Voigt'
has
proposed
the name "moduhis" for these
coefficients,
but this
proposal
seems to ruu counter to the
usage implied
in such
phrases
as
"Young's
modulus." Names for the coefficients
Cn,
... and
Cn/n,
... were
proposed by Rankine',
ond
accounts of his
terminology
will be found in Lord Kelvin's Baltimore Lecture* and in
Todhunter and Pearson's
History,
vol. 2.
NOTE B.
Tlie notion
of
stress.
One
way
of
introducing
the notion of stress into an abstract
conceptual
scheme of
Rational Mechanics is to
accept
it as a fundamental notion derived from
experience.
The
notion is
simply
that of mutual action between two bodies in
contact,
or between two
parts
of the same
body separated by
an
imagined
surface
;
and the
physical reality
of such modes
of action
is,
in this
view,
admitted as
part
of the
conceptual
scheme. It is
perhaps
in this
meaning
that we are to understand the dictum of Kelvin and Tait" that "force is a direct
object
of sense." This was the method followed
by
Euler'" in his formulation of the
principles
of
Hydrostatics
and
Hydrodynamics,
and
by Cauchy'"
in his earliest
writings
on
Elasticity.
When this method is
followed,
a distinction is established between the two
types
of forces which we have called
"body
forces" and "surface
traction.s,"
the former
being
conceived as due to direct action at a
distance,
and the latter to contact action.
Natural
Philosophers
have
not,
as a
rule,
been
willing
to
accept
distance actions and
contact actions as
equally
fundamental. It has been held
generally
that a more
complete
analysis
would reveal an
luiderlying identity
between the two modes of action. Sometimes
it has been
sought
to
replace
action at a distance
by
stress in a medium
;
at other times to
represent
actions
generally recognized
as contact actions
by
means of central forces
acting
directly
at a distance '^. As an
e.xample
of the former
procedure,
we
may
cite Maxwell's
stress-system equivalent
to electrostatic attractions and
repulsions'*.
The alternative
procedure
is
exemplified
in
many
of the
early
discussions of
Elasticity,
and an account
will be
given presently
of
Cauchy's
use of it to determine the stress-strain relations in a
crystalline
material'''.
Any
siich reduction of contact actions to distance actions tends
to obliterate the distinction between surface tractions and
body
forces,
and it has been
customary
to maintain the distinction
by
means of an
hypothesis concerning
the molecular
structure of bodies. In such theories as
Cauchy's
the
apparent
contact actions are traced
to distance actions between
"molecules,"
and these actions are
supposed
not to extend
beyond
a certain
region surrounding
a
"molecule,"
known as the
"region
of molecular
activity."
The
body forces,
on the other
hand,
are traced to distance actions which are
sensible at sensible distances. Thus a second
way
of
introducing
the notion of stress is to
base it
upon
an
hypothesis concerning
intermolecular forces.
"
Pliysikalische Krystallographie, Leipzig,
1891.
iJ
Nat.
Phil.,
Part
1, p.
220.
"
Berlin Hist, de
I'Acad.,
t. 11
(1755).
'"
Exercices de
mathematiques,
t. 2
(1827), p.
42.
Cauchy's
work dates from
1822,
see
Introduction,
footnote 32.
"
The fluctuation of scientific
opinion
in this matter has been sketched
by
Maxwell in a
lecture on
'
Action at a
distance,' Scientific Papers,
vol.
2, p.
311.
"
Electricity
and
Magnetism,
2nd edition
(Oxford 1881),
vol.
1,
Part
1, Chapter
v. Of.
Article 53
(vi) supra.
'" '
De la
pression
ou tension dans un
syst^me
de
points
materiels,'
Exercices de
mathimatiquei,
t. 3
(1828), p.
213.
534 NOTE B
A third
way
is found in an
application
of the
theory
of
energy.
'
Let us
suppose
that a
strain-energy-function exists,
and that the
equations
of
equilibrium
or vibration of a solid
body
are
investigated by
the method of Article
115,
and let the
energy
of that
portion
of the
body
which is contained within
any
closed surface S be increased
by increasing
the
displacement.
Part of the increment of this
energy
is
expressed
as a surface
integral
of the form
f fr(dW
dW dW
) "1
Now in the formulation of Mechanics
by
means of the
theory
of
energy,
"forces" intervene
as the coefficients of increments of the
displacement
in the
expression
for the increment
of the
energy.
The above
expression
at once
suggests
the existence of forces which act at
the surface
bounding any portion
of the
body,
and are to be estimated as so much
per
unit
of area of the surface. In this view the notion of stress becomes a
secondary
or derived
notion,
the fundamental notions
being energy,
the distinction of various kinds of
energy,
and the localization of
energy
in the medium. This method
appears
to be restricted
at
present
to cases in which a
strain-energy-function
exists.
Cauchy's investigation
of stress-strain relations in a
crystalline body.
The
body
is
supposed
to be made
up
of a
large
number of material
points,
or
particles,
which act
upon
one another at a distance
by
means of forces directed
along
the lines
joining
them in
pairs.
The force between two
particles
of masses
m,
m' at a distance r
apart
is taken to be an attraction of amount
mm'x (>),
and the function
x ('")
's
supposed
to vanish when r exceeds a certain finite value
R,
caUed
by Cauchy
the "radius of the
sphere
of molecular
activity."
The
particles
are
supposed
to
form,
when in
equilibrium
under no external
forces,
a
"homogeneous assemblage." By
this it is meant that all of
them have the same
mass,
and
that,
if three of them are situated at
points P, P', ,
and a
line
QQ'
is drawn from
Q, equal
and
parallel
to PP' and in the sense from P to
P',
there is
a
particle
at
Q'.
Let
X, y,
z be the
coordinates,
and M the
mass,
of
any particle
P. We draw a closed
curve round P in the
plane {jp)
which
passes through
P and is
parallel
to the
plane
of
(y, i),
so that all the radii vectores drawn from P to s exceed R. Let S be the area within
this curve. We shall
suppose
that all the linear dimensions of S are small
compared
with
ordinary
standards. The statical resultant of all the forces whose lines of action cross
p
within is a
force,
of which the
components parallel
to the axes are denoted
by
^xS, T^S, ZxS,
where
X^, Y^, Z^
are the
components
of the traction across the
plane
at P. But these
components
are also the sums of such
expressions
as
minii' x{ni)\i, inimix{rij)iiij, mim/ x(.rij)vij,
where
mf
denotes the mass of a
particle
situated on that side of the
plane
for which x is
greater
than the x of
P, m/
denotes the mass of a
particle
situated on the other side of the
plane, rtj
denotes the distance between these
particles, Xy, /in, vij
denote the direction
cosines of the line drawn from
m/
to
Wj.
The summation extends to all
pairs
so situated
that the line
joining
them crosses
p
within s and the distance
r,-,-
does not exceed R.
From the assumed
homogeneity
of the
assemblage
it follows that there is a
particle
Q
of mass m
(equal
to M or
mj
or
m/),
so situated that the line
PQ
is of
length
r
equal
to
Tij,
and is
parallel
to the direction
(Xij, in,, Vij).
Thus the terms of the above sums
may
be
replaced by
Mmxir)^, Mmx{r)fi, Mm.x{i')v,
STRESS
586
where r is the distance of a
particle
m from
M,
and
X,
fi,
v arc the direction cosines of the
line drawn from M to m. The summation
may
! effected
by
first
summing
for all the
pairs
of
particles (mj, m-)
which have the same
r, X, fi,
v and are so situated that the line
joining
them
crosses/)
within
,
then
summing
for all the directions
(X, /i, v)
on which such
pairs
of
particles occur,
and
lastly summing
for all the
pairs
of
particles
on
any
such line
whose distances do not exceed R. The first summation is efiected
by multiplying
the
expressions
such as
Mmx^
(r)
X
by
the number of
particles
contained in a
cylinder
of base
S and
height
rX. This number is
pSrKjM,
where
p
is the
density,
or mass
i)er
unit of
volume,
of the
system
of
particles.
Thus we
require
the sums of such
quantities
as
pmrX')( (r), pmrXn)^ (r), pmr\v}( (r).
If the sunmiation is extended to all directions
(X, fi, v)
round P in which
particles occur,
any
term will be counted
twice,
and therefore the
required expressions
for the
component
tractions
JT^,
... are
X^=ySmr\^x(''): l^x=ip2TOrX,ix(''). ^x^ip^nirXvxir),
in which the summations extend to all
particles
whose distance from P does not exceed R.
If there is no initial stress the six sums of these
types vanish,
or we have
SotcX^
x{r)
=
0, ..., 2mrX/ X W
=
0,
. . .
;
but,
when there is initial
stress,
the values of the six
components
of it at
any point
are
X^m,...,
where
XJ0)
=
ip27nr\^x(.r),
..., X/)
=
ip2rX^xM,

The stress-strain relations are obtained
by investigating
the small
changes
which are
made in the above
expressions
when the
system undergoes
a small relative
displacement.
As in Article
7,
we
may
take the unstrained
position
of if to be
given by
coordinates
X, y, z,
and its strained
position by
coordinates .r
+
w, y
+
c,
z
+
w. At the same time m is
displaced
from
{x+x,i/
+
y,
z
+
z)
to
(;r
+
x
+
M
+
u, ...),
where
u,
... are
given
with sufficient
approximation by
such formulie as
du du du
so that rX becomes ?'X
+
8
(rX),
where
and we have similar formidse for 8
(rp.),
8
{rv).
Also r becomes r
(1 +e),
where
e
=
e\-+ep.^
+ e,,v^
+
e,pv+e,xv\+e^\p,
and
p
becomes
p',
where
,
f
du dv dw\
The effect of these
changes
is to
give
us for
X^,
... such
expressions
as
X,
=
i
p'2
[m
^
~^^
{x (')
+
erx: ir)} {r\
+
8
(/X)}^]
,
Xy
=
i p'2
[m -^l^^^
{x
{r)
+
erx: (')} {^^
+

W){'/'
+
(''/')}]

When there is no initial

stress,
these
equations give
us the stress-strain relations in such
forms as
X^
=
y^ \mr{rx;
(r)
-
x (')}
X^
{e^^\'^
+
eyyp.'^->rev^
+
ey,p.v
+
e.^v\
+
e,:,\p)'\,
Xy
=
\pS, \mr{rx; (r)- x (?)} X^ {e^>?
+
eyyp?
+
ev'^
+
e,.pv-\-<^.^v\
+
e\p]'\
;
536 NOTE B. STRESS
and the elastic constants
Cn,
... are
expressed by
sums of the
tyjies
Cii
=
ip2 [mr {rx' (>)- x (r)} \*], Ci2
=
Cco
=
ip2 [mr {r^' (r)-x W} XV^,
Ci4=C66=ip2 [mr {rx' (r)
-
x W} XV"], <;i6
=
*p2 [wr {r/ (r)
-
x (r)} XV]-
There are 15 of
these,
Green's 21 coefficients
being
connected
by
the 6 relations which
have been called
Cauchy's
relations
(Article 66).
When there is initial stress we have to add to the above
expressions
for
X^
and
X^
the terms
'.'-(*aM-)*"- oy
dz
and
^^'1^^''"'S+^-''S+^^'"'^ Z,.(l-J),
with similar additions to the
remaining stress-components (Article 75).
The above
investigation
is
given
as an
example
of the kind of methods
by
which the
elements of the
theory
of
Elasticity
were
originally investigated.
A modification
by
which
the results
may
be made to accord with
experiment,
at
any
rate for
isotropic
solids,
has
been
proposed by
W.
Voigt,
Ann.
Phi/s. (Ser. 4),
Bd. 4
(1901).
NOTE C.
Applicatimis of
the metlwd
of moving
axes.
The
theory
of
moving
axes
may
be based on the result obtained in Article 35. Let a
figure
of invariable form rotate about an axis of which the direction
cosines,
referred to
fixed
axes,
are
I, m, n,
and let it turn
through
an
angle
bd in time &t. At the
beginning
of this interval of time let
any point belonging
to the
figure
be at the
point
of which the
coordinates,
referred to the fixed
axes,
are
x, y,
z
;
then at the end of the interval the same
point
of the
figure
will have moved to the
point
of which the coordinates are
x-\-{jn,z

ny)&Ya.h6

{x-l{lx->rmy-'rnz)\{\.
-cos
85),
Hence the
components
of
velocity
of the
moving point
at the instant when it
passes through
the
point (a?, y, z)
are
dd dd
,dd
d6
-y^'dt^'^'dr -'^dt^''''dt^
de
, ,d6
-""""di+^^df
We
may
localize a vector of
magnitude d6/dt
in the axis
(l, m, n),
and
specify
it
b}'
components a>^,
ay, a^,
so that
a^^ldd/dt,
.... This vector is the
angular velocity
of the
figure.
The
components
of the
velocity
of the
moving point
which is
passing through
the
point {x, y, z)
at the instant t are then
-ya^+zaiy,
-2o)x
+
.^<Bj, -xaiy+ya>x.
Let a triad of
orthogonal
axes of
{x, y', z'), having
its
origin
at the
origin
of the fixed
axes of
{x, y, z),
and such that
they
can bo derived from the axes of
(.r, y, z) by
a
rotation,
rotate with the
figure;
and let the directions of the
moving
axes at the instant t be
specified by
the scheme of nine direction cosines.
NOTE C. MOVING AXES
537
Let
^1,^2, ^3
denote the
comi^nents
of the
angular
velocity
of the
rotating figure parallel
to the axes of
*', y[, z',
so that
and let a
point (.(;', y', z')
move so as to be
invariably
connected with the
figure.
The
coordinates of this
point
referred to the fixed axes
are,
at the instant
t, It-v' +
l-iy'
+
ls/, ...,
and we
may equate
two
expressions
for the
components
of
velocity
of the
iwint.
We thus
obtain three
equations
of the
type
j^ili-v"
+
h!/'
+13!^)= -(mix;'
+m2!f' +m3z')(nidi+n202
+
n3&3)
Since the axes of
(y, /, 2')
can be derived from those of
(.r, y, 2) by
a
rotation,
we have
such
equations
as
The above
equations
hold for all values of
.r',
y', 2',
and
therefore, a/, y,
/
being
independent
of the
time,
we have the nine
equations
-^=1^63-
1^6,, ~r7
=
h6i~h6i-, -TT
=
k^2~h^\t
^
dmj
. . dm-,

dm-,
-^=m.ie3-m36.,, -^^
=711361-111163,
-^=mi62-m.26i,
V
(1)
dih
A A
dn., . .
dn3
-^^-
=
203-3*j>
-^j-
=
%^i-i(9.i,
-^=ni6.2-n.Ji.
j
Now let
w, V,
w be the
projections
on the fixed axes of
any vector, u', w',
w' the
projections
of the same vector on the
moving
axes at time t. We have such
equations
as
=
^1
f-J-
-
v'63
+
v/6A
+
h
(rff
-w'6i
+
M'fls
j
+
^3
(^
-
We^
+ v'6A
(2)
Hence the
projections
on the
moving
axes of that vector whose
projections
on the fixed
axes are
du dv dv>
'dt' dt' It
du'
,. ,.
di''
,. ,
dw'
,. ,.
are
-^-v'63+ti!6-2, -j^-w
61
+
11!
63,
-^^
-u'6.,+vd, (3)
We
may
abandon the condition that the
origin
of the
moving
axes coincides with that
of the fixed axes. The formulae
(1)
are
unaltered,
and the formuIiB
(2)
also are unaltered
unless
, V,
w are the coordinates of a
point.
Let
x,,, yo,
^o
be the coordinates of the
origin
of the
moving
axes referred to the fixed
axes, x, y,
z and
x', y',
z' those of
any moving point
referred
respectively
to the fixed axes and the
moving
axes. We have such formulte as
X
=
X(,-\-l^.lf +
1^1/
+
I3Z',
and therefore
dx
~dt'
Let
?<') V> ^''o'
be the
projections
of the
velocity
of the
origin
of
(of,
y", i!)
on the
instantaneous
positions
of the
moving axes,
then we have
-

+
^'
(S-^'^^+^'^^)
-
<S'
-^'^'
-^^=) +<S-^'^^-^'^)

'^-hu^
+h Vo
+
hWo'-
538
NOTE C
Hence the
projections
of the
velocity
of
any moving point upon
the instantaneous
positions
of the
moving
axes are
<
+
^'-y'3+'^2, f'+^^-2'^i
+
-^'3, Wo' +
^-^^2+y^i
(4)
These formulae can be utilized for the calculation of differential coefficients. Let
a, ft
y,
... be
any parameters,
and let a triad of
orthogonal
axes of
x', y\
z' be associated
with
any system
of values of the
parameters,
so
that,
when the
parameters
are
given,
the
position
of the
origin
of this triad and the directions of the axes are known. Let the
position
of a
point
relative to the variable axes be
supposed
to be known
;
the coordinates
x', y\
2' of the
point
are then known functions o{
a, ^,
y,
Let
x, y,
z be the coordinates
of the
point
referred to fixed axes. Then
x, y,
z also are functions of
a, /3,
y, ...,
and we
wish to calculate the values of
8.r/3a,
.... When
o, ^,
y,
... are altered the
origin
of the
variable axes
undergoes
a
displacement
and the axes
undergo
a
rotation,
and we
may
regard
this
displacement
and rotation as
being
effected
continuously
with certain velocities.
Thus we have a
velocity
of the
origin
and an
angular velocity
of the triad of axes. This
velocity
and
angular velocity being denoted,
as
before, by
their
components i<o', vd, iV
and
^1) ^2) 63,
referred to the instantaneous
positions
of the variable
axes,
the
quantities
Mo'j , ^1)
are linear functions of
dajdt, dfi/dt, ...,
and the coefficients of
da/dt,
... in
these functions are known functions of
a, ^,
y,
.... Thus we have such
equations
as
ox da dx
dff
d^di'^d^
di'^'
='.(<HS-lf+-)---)
^'.{.+(l-|S--)-''''-4
We
may equate
the coefficients of
da/dt, d^jdt,
... on the two sides of these
equations,
the
quantities o' 61,
...
being expressed
as linear functions of
dajdt,
....
In like
manner,
if
u, v,
w and
u', v',
w' denote the
projections
of
any
vector on the fixed
and variable
axes, equations (2) give
us formulso for
calculating du/da,
In
applications
of the method it is
generally
most convenient to take the fixed axes to coincide with the
positions
of the variable axes that are determined
by particular
values
a, j3,
y,
... of the
parameters,
then in
equations (2)
we
may put ?i
=
i2
=
M3=l
and
?2=...=0.
When this
is done the values of
du/da,
...
belonging
to these
particular
values of
a,
... are
given by
formulae of the
type
duda duds
dudy
/du' da du'
dff
du'
dy
\
, , , ,->
Tadt+r^dt^d-yI^--'={Tadt^d^dt^h^dt^--T''^'^''^''
^^^
The above
process
has been used
repeatedly
in
Chapters xviii., xxi.,
xxiv. As a
further illustration we take some
questions concerning
curvilinear
orthogonal
coordinates.
The coordinates
being a, /3,
y,
the
expression
for the linear element
being
{{da/h,f
+
{d^jhf+ {dylh^f) i,
and the variable axes
being
the normals to the
surfaces,
we have
,
1 da
MOVING AXES
539
curvature of the Kurface
/3.
In
Fig.
75 the
point
P is
(a, /3,
y), P,
is
(a
+
Sa, /3,
y), /
is
(a, ^
+
8/3,
y),

is
(a
+
Sa, ^ +8/i,
y).
The
length
of the arc
PP,
can be taken to be
a//
3
Fig.
75.
and the excess of the
length P.iQ
above
PPj is,
to the second
order, 8|8
5^ ( j^ )
. We
may
regard
the
tangents
to
PP^
at /" and
PiQ
at
P,
as
intersecting
in
T,
and take the
length
of
PP2
to be
Sff/Ai.
Then the
angle PTP^
is
-h2^fj\
8a. Hence the coefficient of
da/dt
in
63
is
-
ft2 KA
[y)
I" '''^^ manner the coefficient of
dSjdt
in
^3
is
A,
^ (
7-
).
We can now write down the formuto
^'='*'a^Wrf''^^^U)
.(6)

~J~di~""'d^\hJ
~dt'
The above
argument
shows that the
principal
curvatures of the surface
y, belonging
to
its lines of intersection with .surfaces a and
/3,
are
respectively
We have similar formula) for the
principal
curvatures of the surfaces a and
/3.
Let
L, M,
JV bo the direction cosines of a fixed line referred to the normals to the
surfaces at a
particular point (a, /3,
y),
and let
Z', M',
N' be the direction cosines of the
same line referred to the variable axes at
any point.
Then
L', M',
N' are functions of
a, j3,
y,
but
L, M,
Nare
indei)endent
of
a, ^,
y.
We
may
use the formulee
(5),
and in them
we
may replace u, v,
w
by L, M,
N and
', v',
v/
by L', M',
N'. We find
dN'
,,,
8/l\ SiV
-.,,
8/l\
8.V'_
8/l\ 8/n
These formulae were used in Article 68.
540 NOTE C. MOVING AXES
To
investigate expressions
for the
components of
strain and rotation^ we take
(', v', w')
to be the
displacement (t<, M/s, Uy),
and
{u, v, w)
to be the
displacement
referred to fixed
axes of
X, y,
z which coincide with the normals to the surfaces
a, /3,
y
at the
point (a, /3,
y).
Then we
have,
for
example,
at
(a, ^,
y)
Now
using
the formuliB
(5)
and
(6)
we have
duda
dud^
dudy
_
dua
da
dua
dfi
dua
dy
d^did^di'^dydi~^dtd^dt'^d^di
"''{^4(i)S-^^|S}+'^{^3i;(i;)S-^'4S}'
and from these we have
3^='^'
{
3f
+'^-^''^
d^iiy^'-^'y d-y {Q}
'
3m
f3M<.
,
3/l\l 3m
,
(3jf.
,
3
/IM
3y='^i8^-'^'''^3^UJ}' 3^=^M-3^-^"'^3^U)r
The formulee
(36)
of Article 20 and
(38)
of Article 21 can now be written down.
To
investigate
the
stress-equations^"
we take the same
system
of fixed
axes,
and consider
the resultants of the tractions on the faces of a curvilinear
parallelepiped
bounded
by
surfaces
a,
a
+8a, /3, /3 +8;3,
y, y
+
8y. (Cf. Fig.
3 in Article
21.)
We
may
take the areas
of the faces
a, A
y
to be
Ai, A2, A3,
where
Ai
=
8^8y//i2A3, A2=8y8a/A3^i, A3
=
8a
8/3/^1^2-
The tractions
per
unit of area across the surface a can be
expressed by Xa, Ya,
Z,
or
by aa^ ap, ya,
and the resultant tractions across the face
Aj
can be
expressed
as
XiAi, FaA], ZaAi
or as
aaAi, a/SAj, yaA,.
In the formulse
(5) X^Ai, FaAj, ZaAy
can
take the
places
of
m, v, w,
and
aaAi, a^Ai,
yaA,,
the
places
of
m', v',
vJ.
Similarly
XpAj, }0A2, .ZJ3A2
can take the
places
of
m, r, w,
and
a/3A2, ^^^2,
/37A2
those of
m', t>', i(/,
and so on. Now the
equations
of motion can be
expressed
in such forms as
a3^(X.A0
+
8^|(Z,A2)
+
8y^^(X,A3)4-.^.|ff=./4|g,
where the notation is the same as in Article 58. We have the
equations
3^(^-4;+3i(^-')f+^(^--')S
=
3ii
(""''')
5^
+
3^
(''^')
J
+
3^
('""^')
i
-
''^^'^='+>^'^^'
=
3^
(^^^)
S
+
sl
("'^^^)
f
+
3y
(^^^)
S-
^^A2^3+^;a.^.,
3^(^v-4:
+
3^(^vA3)f
+
3^(X,A3)J
=
3^
(^^^)
57
+
3^
^^^=)
f
+
1;
(^^^)
S-
^^=^'+>>^=<'2'
where
6^, 63
are
given
by (6). Equation (19)
in Article 58 can be written down at once.

Cf. E. B.
Webb, Messenger of Math.,
vol. 11
(1882), p.
146.
NOTES
D, E,
F
541
NOTE D.
Dr A.
Timpe
has called
my
attention to the fact that the form of the lines of stress in
Fig. 15, p. 195, api>ears
to indicate the existence of
points
at which more than two linos of
stress,
in the
plane
of the
figure,
meet. If there are such
j)oints
the stress at either of them
must be a
simple
tension or
pressure
at
right angles
to the
plane
of the
figure,
and two
principal
stresses vanish. The existence of such
points
has not been
proved
;
for the
positions
of the lines were
computed by Hertz,
loc. cit.
p. 195,
for the
jmrts
of the
figure
near to A'OA and to the lino drawn
through
at
right angles
to
A'OA,
and the rest of the
figure
was drawn
conjecturally.
NOTE E.
Stress in a beam loaded
uniformly.
Using
the notation of Article 244
(a), (6), (c)
I find the
following expressions
for the
stress-components
in a circular
cylinder
bent
by
its own
weight
:

^.=^[(5
+
2<r)(a2-^-^)-3(l-2a)/],
r=^[3(l
+2a)(a2-/)- (1
-
2<r)x2],
^=
"if
[(1
-
2<r)(a2-y^)
-
3(1 +2<T)a,-2],
Z,=/.(<,
+
2K22)[-(|
+
io-)(a2-a;2)
+
(J-i(r)/], Y,=),{K^-\-=l^.,z){^
+
,T)xy,
Z,= -'(Ko+<i'-
+
'<22')-'!?-M2^[i(9
+
13<r+40a2-(l-l-i<r)(a;2+y)].
The constant
k^
is
given by
the
equation
When the
beam,
of
length I,
is fixed
horizoritaHy
at 2
=
0,
and the end
=
^8
unloaded,
When the
beam,
of
length 2?,
is
supported
at the ends z
=
l and z=
-I,
these ends
being
at
the same
level,
Vn, 2
7 + 12(r +
4<rn
M
=
0, .,=
-.,\P^ra'^ "6(l+^rJ-
An
independent
calculation of the
displacement kindly
sent to me
by
Mr G. C.
Calliphronas
confirms these results.
NOTE F.
Extensional mbrations
of
a
plane plate".
The
equations
of vibration are
equations (97)
of Article 314
(e). They may
be
expressed
very simply
in terms of the areal dilatation A' and the rotation
w,
these
quantities being
defined
analytically by
the
equations
,
3u 9v

3v 3u
,,,
^=s+?3;' 2=^=8^-^
^^^
The
equations
take the forms
aA'
,,
,anr
p(l-(r2)82u
8A'
8w_
p (1
-(t^)
S^v ,,
-g^-Ci---)^
li-W
l^^^^-^'Tx-
E'dt^
^^^
In this form
they
are
readily
transformed to
any
suitable cm-vilinear coordinates.
2'
Equations
equivalent
to
(97)
of Article 314
(e)
were obtained
by
Poisson and
Cauchy,
see
Introduction,
footnotes
36 and 124. Poisson
investigated
also the
symmetrical
radial vibrations
of a circular
plate,
obtaining
a
frequency equation equivalent
to
(10)
of this
Note,
and
evaluating
the
frequencies
of the
graver
modes of this
type.
542 NOTE F. EXTENSIONAL VIBRATIONS OF PLATES
Consider more
particularly
the case of a
plate
with a circular
edge-line.
It is
appropriate
to use
plane polar
coordinates
r,
6 with
origin
at the centre of the circle.
Let
U,
V be the
projections
of the
displacement
of a
point
on the middle
plane upon
the
radius vector and a line at
right angles
to the radius vector. Then we have
\x.=
Ucoa6-Vs\n6,
v=U
ii\ne+
VcosB
(3)
\dU
and
3^
18F
2i<r
=
-=-
+
--
cr r r dd'
(4)
-^>
^1^^
=
2p(H-<r)i
dfi
,.(5)
.(6)
(7)
and the stress-resultants
belonging
to
any
circle r=const, are
T, S,
where
2M^p_ f.l^_I\] v-J^r^'
VI din
l-a^\_dr^'^\r^r d6jj'
l
+
a\_dr
r
'^
r dd
J
The
equations
of vibration
give
We
put
U=
Un
cos ?j5 cos
pt, V=Vnamnd
cos
pt,
where
U
and
F
are functions of
r,
and we write
K^
=
pil-(r')pyE, K'-^^2p{l+,r)p^lE. (8)
Then A' is of the form
A'J (kv)
cos n6 cos
pt,
and or is of the form B'
J (k't)
sin n6 cob
pt,
where A' and B' are
constants,
and
J
denotes Bessel's function of order n. The forms of
(I and V are
given by
the
equations
dr
and with these forms wo have
A'=

AK^J,^{icr)
cos n0 cos
pt,
2m
=
BK'^J{K'r)
sin n6
cospt.
We can have free vibrations in which V vanishes and U is
independent
of d
;
the
frequency equation
is
cLli (ko)
U^^A^^^KuBl-iiriyosnecospt,
1
=
-[r.^
^)
+
B^j;-^)]sin^cos^^
(9)
'
+
-JiiKa)
=
0, .(10)
a
being
the radius of the
edge-line.
We can also have free vibrations in which U vanishes
and V is
independent
of 6
;
the
frequency equation
is
dJi{K'a)
^
Jii KO)
da a
^ '
These two modes of
symmetrical
vibration
appear
to be the
homologues
of certain
modes of vibration of a
complete
thin
spherical
shell
(cf.
Article
335).
The mode in which
U vanishes and V is
independent
of 8 is the
homologue
of the modes in which there is no
displacement parallel
to the radius of the
sphere.
The mode in which V vanishes and V is
independent
of 6 seems to be the
homologue
of the
quicker
modes of
symmetrical
vibration
of a
sphere
in which there is no rotation about the radius of the
sphere.
In the
remaining
modes of extensional vibration of the
plate
the motion is
compounded
of two : one characterized
by
the absence of areal
dilatation,
and the other
by
the absence
of rotation about the normal to the
plane
of the
plate.
The
frequency equation
is to be
formed
by eliminating
the ratio A : B between the
equations
-''[^""r+(--^'-)-'-'""']*"""->E'"f'-5''-"'"]=->'1
(12)
These modes of vibration seem not to be of sufficient
physical importance
to make it worth
while to
attempt
to calculate the roots
numerically.
INDEX.
AUTHOES CITED.
I'l'lie
numbers
refer
to
pages.]
Abraham, M.,
44.
Airy,
G.
B., 17,
86.
Alibrandi, P.,
107.
Almansi, E., 22, 107, 239,
342.
Amagat,
E.
H.,
103.
Aron, H., 29.
Bach, C, 110,
331.
Barth(516my, A.,
470.
Basset,
A.
B., 24, 381, 433, 505, 507, 511,
516, 519,
521.
Bauschinger, J., 110, 111, 112, 113, 114,
115.
Beltrami, E., 49, 86, 132,
290.
Bernoulli,
Daniel, 3, 4,
5.
Bernoulli, James,
3.
Bernoulli,
James
(the Younger),
5.
Betti, E,, 16, 44, 47, 170,
171,
220.
Binet, J.,
23.
Blanchet,
P.
H.,
18.
Boltzmann, L., 116,
143.
BorcharcU,
C.
W., 53, 106, 235,
259.
Boscovich,
E.
J.,
6.
Boussinesq, J., 16, 20, 24, 26, 29, 87, 182,
185, 187, 190, 208, 228, 230, 302, 304,
336, 411, 415, 420,
438.
Braun, F.,
116.
Bresse, M., 24,
427.
Brewster,
D.,
87.
Bromwich,
T. J. I'
A., 274, 289,
297.
Bryan,
G.
H., 30, 392, 463, 464, 529,
530.
Burkhardt, H., 7,
15.
Butcher,
J.
G.,
116.
Calliphronas,
G.
C,
541.
Canevazzi, S., 23,
363.
Cardani, P.,
102.
Cauchy,
A. L., 8, 9, 10, 11, 12, 14, 18, 19,
27, 35, 57, 72, 79,
81, 105, 108, 376, 533,
541.
Cerruti, V., 16, 225, 228, 230, 235,
292.
Cesiro, E.,
55.
Chladni,
E. F.
F.,
5.
Chree, C, 120, 125, 143, 172, 239, 240,
247,
249,
253, 255, 259, 263, 266, 274, 275,
277, 405, 409, 421,
422.
Christoffel,
E.
B., 18,
287.
Cilley,
F.
H.,
108.
Clapeyron,
B. P.
E., 22,
358.
Clausius, E., 9,
10.
Clebsch, A., 14, 17, 21, 24, 27, 28, 29, 35,
176, 177, 204, 257, 371, 372, 379, 381, 447,
526.
Codazzi, D.,
489.
Coker,
E.
G.,
349.
Cornu,
M.
A., 102,
128.
Cosserat,
E. and
F., 125, 239,
240.
Coulomb,
C.
A., 3, 4,
118.
Coulon,
J.,
292.
Cox, H., 421.
Culman, K.,
23.
Darwin,
G.
H., 16, 118, 253, 256, 257.
Davidoglou, A.,
421.
Dougall, J., 184, 222,
225, 230, 233,
465.
Duhamel,
J. M.
C,
106.
Duhem, P.,
47.
Dunkerley, S.,
422.
Eddy,
H.
T.,
427.
Edwardes, D., 240,
260.
Estanave, E.,
469.
Euler, L., 3, 4, 5, 384, 388,
533.
Everett,
J.
D.,
103.
Ewing,
J.
A., 32, 84, 110, 111, 112,
113, 331.
Ewing
and
Eosenhain,
110.
Fabr(5,
347.
Filon,
L. N.
G., 135, 211, 263, 304, 348,
319,
352.
Flamant, 110,
208.
Fiippl,
A., 110, 117, 119,
331.
Fourier,
J. B.
J.,
420.
Fresnel, A.,
7.
Fuss,
P.
H.,
3.
544 INDEX
Galileo,
2.
Garrett,
C. A.
B.,
421.
Gauss,
C.
F.,
472.
Gehring, F.,
27.
Germain, Sophie,
5.
Goldschmidt, V.,
154.
Grashof, F., 22, 323, 331,
332.
Green, G., 11, 15, 18, 57, 83, 109,
288.
Greenhill,
A.
G., 143, 302, 307, 308, 400,
404, 405, 406, 421,
422.
Guest,
J.
J.,
119.
Hadamard, J., 29, 57,
292.
Halphen, G.-H.,
404.
Hamburger, M.,
197.
Harnack, A.,
63.
Hausmaninger, V.,
25.
Heppel,
J.
M.,
22.
Hertz, H., 16, 26, 190, 195, 213, 295,
541.
Hess, W., 24, 384,
398.
Hicks,
W.
M.,
220.
Hilbert, D.,
169.
Hilton, H.,
147.
Hooke, B.,
2.
Hopkins, W.,
45.
Hopkinson, J., 25, 107,
143.
Hoppe, E., 25,
431.
Ibbetson,
W.
J., 17,
133.
Jaerisoh, P., 18, 265,
272.
Jeans,
J.
H., 253,
274.
Jellett,
J.
H.,
482.
Jouravski,
22.
Kelvin, Lord, 12, 16, 40, 54, 55, 59, 77, 93,
97, 106, 107, 109, 115, 116, 117, 169, 180,
236, 245, 256, 257, 282, 287, 295,
531.
Kelvin and
Tait,
14, 20, 24, 29, 35, 57, 67,
69,
118, 129, 187, 248, 255, 257, 302, 365,
396, 403,
418, 438, 441, 531, 532,
533.
Kerr, J.,
87.
Kirchlioflf, G., 14, 23, 27, 28, 30, 49, 50,
97, 163, 167, 290, 365, 371, 372, 377, 379,
382, 395, 421, 438,
469, 503,
531.
Klein and
Sommerfeld,
395.
Konig, W.,
295.
Kohlrausch, B.,
107.
Kriemler,
0.
J.,
391.
Kiibler, J.,
391.
Lagerhjelm, P.,
97.
Lagrange,
J.
L.,
3.
Lamarle,
E.,
391.
Lamb, H., 18, 29, 55, 79, 170, 2.S0, 234,
265, 272, 274,
297, 428, 432, 442, 450,
477, 487, 511, 516,
521, 522, 525,
531.
Lam^, G., 13, 15, 17, 51, 55, 81, 87,
89,
117, 139, 142, 259,
531.
Lam6 and
Clapeyron, 11,
15.
Laplace,
P.
S.,
6.
Larmor, J., 164, 166, 249,
257, 291, 304,
383,
392.
Lauricella, G., 230, 233, 235, 239,
292.
U\y, M., 23, 358, 363, 404,
405,
469.
Lewis,
W.
J.,
154.
Liebisch, Th., 146, 159,
533.
Liouville, J.,
289.
Lipscbitz, B.,
97.
Lorenz, L.,
293.
MacCuUagb, J., 166,
288.
Macdonald,
H.
M., 289,
308.
Mallock,
H. E.
A., 102, 128,
142.
Marcolongo, E., 230,
236.
Mariotte, E.,
2.
Mascart,
M.
E., 87,
146.
Mathieu, E., 29, 470,
524.
Maxwell,
J.
C,
17, 82, 86, 87, 106, 116,
143, 220,
533.
Mesnager, A.,
208.
Meyer,
O.
E.,
116.
Michell,
A. G.
M., 400,
402.
Miehell,
J.
H., 17,
22,
24, 29, 86, 89, 133,
134, 139, 198, 200, 207, 208, 209, 212,
213, 215, 230, 259, 262, 339, 342, 346,
349, 423, 431, 432, 444, 445, 466,
467.
Miers,
H.
A.,
154.
Miller, J.
W.,
398.
Minohin,
G.
M., 67,
133.
Mohr, 0., 23, 118, 119,
360.
Morera, G.,
86.
Morrow, J., 102,
421.
MuUer-Breslau,
H. F.
B.,
23.
Navier, 7, 22, 25, .353.
Neumann, C, 169,
235.
Neumann,
F., 14, 87, 106, 152,
173.
Newton, I., 6,
195.
Ostrogradsky, M.,
18.
Pearson, K., 13, 22, 117, 346, 348, 358,
581.
See also Todhunter and Pearson.
Perry, J., 333,
398.
Perry
and
Ayrton,
363.
Pesehka,
G. A.
V.,
429.
Phillips, E.,
27.
Pochhammer, L., 16, 18, 25, 259, 275, 277,
350,
409.
Poincar^, H., 220, 235,
392.
Poisson, S.
D., 6, 8, 10, 11, 12, 17,
18, 25,
27, 274, 282, 289, 438, 466, 469,
541.
Poncelet,
J.
V., 12,
117,
120.
INDEX
545
Poynting
and
Thomson,
142.
Prandtl, L., 310, 391, 400,
402.
Purser, F., 144,
306.
Eankine,
W. J.
M., 22, 110, 288, 331, 333,
531,
532.
KayleiRh, Lord, 19, 26, 29, 96,
107, 170, 176,
276, 277, 280, 288, 289, 293, 295, 408, 409,
410, 411, 418, 420, 421, 444, 466, 469,
474, 475, 477, 485, 486, 506, 514, 515,
519, 523, 525,
527.
Eesal, H.,
429.
Bibifere, C,
352.
Bitter, A.,
23.
Bouth,
E.
J.,
867.
Saalschiitz, L., 384,
392.
Saint-Venant, B., 13, 14, 17, 19, 20, 21, 22,
23, 25, 26, 27, 29, 49, 57, 109, 117, 120,
126, 129, 158, 161, 299, .305, 307, 311,
314, 324, 325, 329, 368, 376, 377, 398,
411, 418, 420, 423, 428, 429, 454, 466,
469,
531.
Salmon, G., 43, 52,
472.
Schneebeli, H., 196,
197.
Schoenflies, A., 147,
154.
S^bert and
Hugoniot,
26.
Somigliana, C, 230, 232, 233,
235.
Stokes,
G.
G., 10, 11, 12,
18, 26, 38, 47, 96,
102, 290, 292, 294, 305, 351,
421.
Tait,
P.
G.,
196.
Tedone, 0., 230, 260,
292.
Thomson,
J.
J.,
55.
Thomson,
Sir
W.,
see
Kelvin,
Lord.
Timpe, A.,
541.
Tissot,
M.
A.,
63.
Todhunter and
Pearson, 7, 27, 67, 97,
109,
117, 118, 120, 391, 421,
427.
Tresca, H.,
114.
Tutton,
A. E.
H.,
469.
Unwin,
W.
C, 78, 110, 111, 115, 116,
117.
Verdet, E.,
7.
Vicat,
L.
J.,
114.
VoiRt, W., 14, 21, 25, 44, 79, 97, 116, 118,
154, 156, 158, 160, 313, 334, 531, 532,
536.
Volterra, V., 233,
292.
Voss, A.,
471.
Wangerin, A.,
260.
Warburg,
E.
G.,
116.
Webb,
E.
E., 69, 260, 358,
540.
Weber, H.,
230.
Weber, W.,
114.
Wehage,
H.,
119.
Weierstrass, K.,
169.
Weingarten,
J.,
87.
Wertheim, G., 13, 97,
107.
Weyrauch,
J.
J., 23,
353.
Whittaker,
E.
T.,
395.
Wilberforce,
L.
E.,
398.
Willis, E., 26,
421.
Wilson, Carus, 208,
351.
Winkler, E.,
427.
Wohler, A.,
117.
Young, T., 4,
7.
L. E.
35
INDEX.
MATTERS TREATED.
[The
numbers
refer
to
pages."]
Additional
deflexion of beams,
333.
^olotropy, 103, 146;
of
inertia, 288;
Curvi-
linear,
161
; produced by permanent set,
116.
Molotropic solid,
Constants and Moduluses
of,
103

105,
158
; propagation
of waves
in, 18,
286289.
Afterstrain,
114.
Angle,
Transmission of stress from
an,
209.
Anorthic
crystals,
155.
Anticlastic
cun'ature, 21, 128, 129, 325,
347.
Applicable surfaces,
471.
Applied Mechanics,
Treatises
on,
110
;
Criti-
cism of some methods used
in,
331

333.
Arches,
427.
Average
strains.
Determination
of,
171.
Axes, Moving,
536

540;
Principal,
of
strain,
37,
60
;
of
symmetry,
147
; Principal
torsion-
flexure,
of a
rod,
366.
Bars : see
Beams, Rods, Torsion,
&o,
Barytes,
Constants
for,
161.
Beams,
^olotropic,
159,
330.
Beams, Bent,
Curvature
of, 323, 345, 348,
349
;
Deflexion
of, 325, 333,
355358
;
Twist
pro-
duced
in, by
transverse
load, 325,
344
;
Stress
produced in, by
transverse
loads, 136,
314316, 331, 332, 336345,
541
; Shearing
stress
in, 136, 316318, 327, 331,
332
;
Dis-
placement in, 318320,
323327
;
Obliquity
of cross-sections
of, 137, 324, 325,
333
;
Distortion of cross-sections
of,
326
; Strength
of,
329
;
Extension
produced
in
by
trans-
verse
load, 347,
349. See also
Bending of
Beams, Bending moment,
Bernoulli-Eulerian
theory. Loading,
Neutral
plane,
&c.
Beams, Continuous,
22,
353
; Single span,
355
358
;
Several
spans,
358364.
BeUs,
Vibrations
of, 5, 29,
525.
Bending of Beams,
History
of
theory
of,
2, 3,
20,
21
; by couples, 126129,
159
; by
terminal
load, 136138,
314333
; by
uni-
form
load,
339349
; by
distributed
load.
350

353
;
of
particular
forms of section
(narrow rectangular, 136, 331,
349
; circular,
320, 327, 328, 348, 541; elliptic, 321, 324,
333
; rectangular, 322,
325
;
other
special
forms, 323,
332).
Bending moment, 314,
344
;
Relation
of,
to
curvature, 127, 323, 345,
350.
BernouUi-Eulerian
theory,
19, 350,
353.
Beryl,
Constants
for,
161.
Blade : see
Stability.
Body forces,
73
;
Particular
integrals for, 180,
217, 232, 246,
294.
Boiler-flues, Collapse
of,
530.
Boundary
conditions, 98, 132,
164
;
in torsion
problem,
299
;
in flexure
problem, 315, 317,
328
;
in
plates, 27,
437442
;
in
shells, 30,
510,
515
;
in
gravitating sphere,
250
;
in
vibrating sphere,
268
;
in
vibrating cylinder,
278,
280.
Bourdon's
gauge,
474.
Brass,
Constants
for, 13,
103.
Breaking
stress,
112.
Bridges, Travelling
load
on, 26,
421
; Suspen-
sion,
347.
Buckling,
Resistance
to,
390. See also
Stability.
Cannon : see Gun construction.
Cantilever, 21,
353.
Capillarity,
6.
Cartography,
63.
Cast
iron, 107, 112,
113.
Cast
metals,
95.
Cauchy's
relations, 14,
98.
Central-line,
of
prismatic body
or curved
rod,
127,
379.
Cliairn,
Links
of,
427.
Circular arc : see Arches and Circular
ring.
Circular
cylinder, Equilibrium
of,
under
any
forces, 257,
259
;
bent
by
its own
gravity,
348,
541
(see
also
Beams) ;
strained
sym-
metrically,
263
;
Vibrations
of,
18, 275280,
405.
INDEX
647
Circular
disk, Rotating,
144
; Eciuilibrium of,
under forces in its
plane,
213

215.
Circular
ring. Equilibrium
of, 385, 403, 427
430; Stability of, 23, 405;
Vibrations
of,
430433.
Clamped end,
of a
rod,
354.
Collapse
: see
Stability.
Collision : see
Impact.
Colour
fringes
: see
Light,
Polarized.
Combined
strain,
329.
Compatibility,
Conditions
of,
49.
Complex
variable : see
Conjugate functions.
Compression,
Modulus
of, 12, 101,
104
;
of a
body
under
pressure, 102, 105, 139, 141, 161,
172,
173
;
of a
sphere by
its own
gravitation,
140
;
of a
body
between
parallel planes,
172,
263
;
Centre
of, 184,
294.
Cones, Equilibrium
of,
200.
Conformal transformation
: see
Conjugate func-
tions.
Conjugate functions,
46, 211, 216, 259, 260,
301, 307,
321.
Conjugate property,
of normal
functions,
176
;
of harmonic
functions, 218,
240.
Constants, Elastic, Controversy concerning,
13
15
;
Definition
of,
97
; Magnitude of,
103
;
Thermal variations
of,
107
;
of
isotropic
bodies,
100
;
of
crystals,
160
; Experimental
determination
of, 22, 102, 142, 160,
469.
Continuing equation,
412.
Coordinates,
Curvilinear
orthogonal,
52
;
Strain
in terms
of, 54,
540
; Stress-equations
in
terms
of, 87, 164,
540
;
General
equations
in
terms
of, 138,
164. See also
Cylindrical
coordinates.
Polar
coordinates, Conjugate
functions.
Copper,
Constants
for,
103.
Crushing,
of
metals, 114;
of
cylindrical
test
pieces,
263.
Crystalline
medium : see
Aiolotropic
solid,
and
Crystals.
Crystals, Symmetry
of,
152
;
Classification
of,
154; Elasticity
of, 14, 156;
Elastic con-
stants
of,
160
;
Neumann's law of
physical
behaviour
of, 14,
152.
Cubic
capacity,
of a
vessel,
121.
Cubic
ciystals,
155,
159.
Curl,
of a
vector,
46.
Curvature : see
Beams, Rods, Plates, Shells,
Surfaces.
Cylindrical body of any form of section. Equili-
brium
of,
under
tension,
102
;
under
gravity,
124,
172
;
under fluid
pressure,
125
;
in a
state of
plane
strain,
259
;
in a state of stress
uniform
along
its
length,
334
;
in a state of
stress
varying uniformly along
its
length,
336
;
in a state of stress
uniform,
or
varying
uniformly,
over its
oross-sections,
454. See
also
Beams,
Plane
strain,
Pltine
treit,
Hods.
Cylindrical coordinates, 56, 89, 140, 161, 269,
260, 275.
Cylindricalflaw,
in twisted
shaft, 121,
304.
Cylindrical shells.
Equilibrium
of,
under
pres-
sure,
141.
Cylindrical shells. Thin,
Ineztensional deforma-
tion
of,
477
;
Vibrations of
(Inextensional,
488,
519
; general theory,
516
; extensional,
518)
;
Stability of,
under
pressure,
529.
Deflexion
: see Beams and Plates.
Density,
Table
of,
103.
Diagrams,
of
plane stress,
86
; stress-strain,
111.
Dilatation, Cubical, 41,
69
; Uniform,
44
;
Average
value
of,
172
;
in curvilinear co-
ordinates,
54
;
Waves
of, 18, 282, 286,
288
;
Centre
of,
184
;
Determination of, 221.
Discontinuity,
Motion of a surface
of,
283

286.
Displacement,
35.
Distortion,
Waves
of, 18, 282, 286, 288;
of
cross-sections of twisted
prism,
308
;
of
cross-sections of bent
beam,
326.
Divergence,
of a
vector,
46.
Double
force, 184, 234,
294.
Dynamical resistances, 26, 120,
411417,
420
422.
Earth,
In a state of initial stress, 107, 140;
Strained
by
its own
gravitation,
140,
248
;
Stress
produced in, by weight
of
continents,
253; EUiptioity
of
figure
of, 254;
Tidal
effective
rigidity of,
255257
;
Period of
spheroidal
vibrations
of,
273.
Earthquakes,
297.
Ease,
State
of.
111.
Elastica, 3, 24, 384388,
392.
Elastic
after-working,
114.
Elasticity,
defined,
90
;
Limits
of,
113.
Ellipsoid,
Solutions of the
equations
of
equi-
librium
in, 239, 240,
260.
Elliptic cylinders.
Solutions of the
equations
of
equilibrium in,
259,
260
;
Solution of the
torsion
problem
in,
305
;
Solution of the
flexure
problem
in,
321
; Confocal, 308,
322.
Energy,
Intrinsic,
91. See also Potential
Energy
and
Strain-energy-function.
Equilibrium,
General
equations of,
7

12, 82,
98, 130, 132,
138, 166,
217
;
of bifurcation,
392.
Equipollent
loads.
Elastic
equivalence
of,
129.
Existence
theorems, 169,
220.
Experimental
results.
Indirectness
of,
94.
Extension, 32, 40, 44,
59
;
of beam bent
by
548 INDEX
distributed
load, 347
;
of
plate
bent
by pres-
sure,
458464.
Extensional
vibrations,
of
rods, 408,
411

420
;
of a circular
ring,
433
;
of
plates, 470, 523,
541
;
of
shells,
515
;
of
cylindrical shells,
518
;
of
spherical shells,
524.
Extensions, Principal, 42,
60.
Extensometer, 94,
111.
Factors
of safety, 118,
120.
Failure : see
Rupture.
Fatigue,
117.
Flaws, 120,
304.
Flexure,
see Beams.
Flexure
functions, 317, 328,
338.
Flexure
problem,
317.
Flexural
rigidity,
of
beam,
354
;
of
rod,
372
;
of
plate,
443.
Flow : see
Plasticity.
Flue : see
Boiler-flues.
Fluid,
115.
Fluor
spar.
Constants
for, 160.
Flux
of energy,
in
vibratory motion,
174.
Fourier's
series,
306.
Fracture : see
Rupture.
Frameworks,
23.
Frequency equation,
176.
FreaneVs wave
surface,
288.
Funicular
polygon,
360.
Galileo's
problem,
2.
Girders : see Beams.
Glass,
Constants
of, 13,
103.
Graphic representation,
of
stress, 86;
of solu-
tion of the
problem
of continuous
beams,
360364;
of the
theory
of
torsion, 308;
and
flexure, 326,
327.
Gravitation : see
Compression
and Earth.
Gravity,
Effect
of,
on vibration of
sphere,
274
;
on surface
waves,
297.
Green's
functions,
219.
Green's
transformation,
83.
Groups,
of
transformations, 69, 148,
153.
Gun
construction,
143.
Hamiltonian
principle,
163.
Hardening, by overstrain,
113.
Hardness,
16.
Harmonic
function,
218.
Harmonics, spherical, 218, 236,
266.
Heat : see Tliermal
effects.
Height,
consistent with
stability,
405.
Helix : see
Rods,
Thin and
Springs.
Hertz's
oscillator, Type
of waves due
to,
295.
Hexagonal crystals,
155.
Hooke's
law, 2, 9,
12
; generalised,
95
;
ex-
ceptions to, 95,
110.
Hydrodynamic analogies,
to torsion
problem,
302.
Hysteresis,
116.
Identical
relations,
between
strain-components
:
see
Compatibility
,
conditions of.
Impact, 16,
195

197
;
of
spheres,
197
; Longi-
tudinal,
of
bars, 25,
418420
; Transverse,
of
bars,
420.
Incompressible solid, equilibrium of,
248
;
vibra-
tions
of, 271, 273
;
waves on surface
of,
296.
Inextensional
displacement,
in thin
rod,
425
;
in thin
shell, 476,
495.
Initial
stress, 107,
249.
Integration,
methods
of,
for
equilibrium, 15,
218, 220, 234, 239;
for
vibratory motion,
176,
289.
Intermolecular
action, 6, 7, 9, 10, 533,
534.
Invariants,
of
strain, 43, 60,
100
;
of
stress,
81.
Inversion,
of
plane strain,
212
; applied
to
plate,
467.
Iron,
Constants
of,
103
;
Elastic limits
of,
114
;
Yield
point of,
114. See also Cast iron.
Isostntic
surfaces, 87,
89.
Isotropic solids,
100.
Isotropy, Transverse,
157.
Kinematics,
of thin
rods, 365370,
423426
;
of thin
shells,
492498.
Kinetic
analogue,
for
naturally straight rod,
382
;
for
naturally
curved
rod,
383
;
for
elastica,
385
;
for rod bent and twisted
by
terminal forces and
couples,
394
;
for rod
subjected
to terminal
couples,
398.
Kinetic
moduluses, distinguished
from
static,
96,
97.
Lamina,
4.
Lead,
Constants
of,
103.
Length,
standards
of,
affected
by atmospheric
pressure, &c.,
120.
Light, Polarised,
Examination of
stress-systems
by
means
of, 87,
351.
Limitations,
of the mathematical
theory,
110.
Limits : see
Elasticity,
Limits of.
Lines
of shearing stress,
in
torsion, 303,
309
;
in
flexure,
327.
Lines
of stress,
87
;
for two bodies in
contact,
195,
541
;
for force at a
point,
199.
Load,
95
;
Sudden
application
or reversal
of,
120,
178
; travelling,
26,
421.
Loading,
Effect of
repeated,
116
; Asymmetric,
of
beams,
328
; Surface,
of
beams,
351.
Longitudinal vibrations,
of rods : see Exten-
sional
vibrations,
of rods.
Longitudinal
waves, 8,
11.
INDEX
549
Magnetometer, deflexion-bars,
121.
Matter, Constitution
of,
6.
Maxwell's
streng-sygtem, 82,
85, 133,
533.
Membrane,
Analogy of,
to twisted
shaft,
310.
Middle
third.
Rule
of, 84,
209.
Modulus, 104,
533. See also
Compression,
modulus
of,
Rigidity, Young's
modulus.
Molecular
hypothesis,
6

10.
Moments,
tlieorem of
three, 22,
358

360.
MonocUnic
crystals,
155.
Multicomtant
theory,
13.
Neutral axis : see Galileo's
problem.
Neutral
plane,
324.
Neutral
surface, 347,
349.
Nodal
surfaces,
of
vibrating sphere,
272.
Nonnal
forces,
rotl bent
by,
403.
Normal
functions,
of
vibrating system,
170.
Notation,
531.
Nuclei,
of
strain,
183, 198, 205,
294.
Oblique crystals,
155.
Optics,
influence of theories of in
stimulating
research in
Elasticity, 7, 8, 11,
30.
Orthogonal
surfaces,
52.
Perturbations, Local,
186.
Pendulum,
Analogy
to
elastica,
385.
Photo-elasticity
: see
Light,
Polarized.
Piezo-eleclricity,
146.
Piezometer, 94, 142,
162.
Plane,
Problem of
the, 15, 188190, 200,225
232.
Planes,
Principal,
of
stress,
79.
Plane
strain, 45,
134
; Displacement accompany-
ing,
201
;
Transformation
of,
211.
Plane
stress, 81, 135; Displacement accompany-
ing,
203
;
in
plate,
446.
Plane
stress. Generalised, 135,
205
;
in bent
beam, 136,
349
;
in
plate,
450.
Plasticity,
114.
Plate, elliptic. Bending of, by pressure, 463,
464;
Vibrations
of,
470.
Plate, Thick,
stretching of,
446

449
; bending
of, 129,
449465.
Plate,
thick
circular. Bending of,
by
central
load,
454; by
uniform
pressure, 460, 463;
by
variable
pressure,
464.
Plate, Thin,
bounded
by straight edge,
468.
Plate,
thin
circular,
406

468.
Plate, thin. Subjected
to forces in its
plane,
205216
; Boundary
conditions
for, 27,
437442; Bending of, 465;
Vibrations
of,
469, 512, 522,
541
;
General
theory
of,
503
;
Stability of,
528.
Plates, History
of
theory
of, 56,
2729.
Poisson's
ratio, 13, 101,
105.
Polar
coordinaUt, 66, 89, 188, 161, 198,
259,
274,
479.
Potassium
Chloride, ConstantB
for,
160.
Potential, Newtonian,
Theory of, 169,
218.
Potentiiil
energy,
of strained
body, Minimam,
168
;
Theorem
concerning,
170.
Potentials, Direct, Inverse, Logarithmic, 189,
190.
Pressure,
Hydrostatic, 79,
see also
Compretiion ;
docs not
produce fracture,
119
;
between
bodies in
contact,
190195.
Prism,
Torsion of : see Torsion.
Prismatic
crystals,
155.
Punching,
of
metals,
114.
Pyrites,
Constants
for,
160.
Quadric surfaces, representing
distribution of
strain, 37, 41, 60, 62, 65;
and of
stresB,
79,
82.
Quartz,
Constants
for,
160.
Radial
displacement, 139, 141,
161.
Radial
vibrations,
of
sphere,
273
;
of
spherical
shell, 274
;
of
cylindrical shell,
519.
Rariconstant
theory,
13.
Rays, equivalent, 146,
153.
Reality,
of the roots of the
frequency equation,
177.
Reciprocal strain-ellipsoid, 37,
60.
Reciprocal
theorem
(Betti's), 16, 170, 221, 223,
232.
Refraction, Double,
due to
stress,
87
;
to un-
equal heating,
106.
Resilience,
170.
Rhombic
crystals,
155.
Rhombohedral
crystals,
155.
Rigid body displacement, superposable upon
displacement
determined
by strain,
50
;
or
stress,
166.
Rigidity, 101,
105. See also Flexural
rigidity.
Torsional
rigidity.
Tidal
effective rigidity.
Ring
: see Circular
ring.
Rocksalt,
Constants
for,
160.
Rods, Naturally
curved. Approximate theory
of,
379, 423;
Problems
concerning, 383, 403,
405,
427430
;
Vibrations
of,
430433.
Rods, Thin,
Kinematics
of, 365; Equations
of
equilibrium of,
370
;
Strain
in,
373
; Ap-
proximate theory of, 372,
377
;
Problems of
equilibrium
concerning,
382

406
;
Vibrations
of,
407411
;
of variable section, 406,
421
;
Problems of
dynamical
resistance
concerning,
411421.
Rotation,
of a
figure,
67
;
Strain
produced
in a
cylinder by,
143
;
Centre
of, 185,
294.
Rotation, Components of, 39, 55,
71
;
Deter-
mination
of, 224,
231.
550
INDEX
Rotationally
elastic
medium,
166.
Rupture, Hypotheses concerning
the conditions
of,
117.
Safety
: see Factors
of Safety
and
Rupture.
Saint-Venant's
principle,
129.
Scope,
of the Mathematical
theory,
119.
Screw-propeller shafts,
119.
Semi-inverse
method,
19.
Set,
111.
Shafts
: see
Rotation, Torsioji, Whirling.
Shear, Pure,
33
; Simple,
33, 68,
69
;
Used
by
Kelvin and Tait to denote a
strain, 532; by
Bankine to denote a
stress,
532.
Shearing strain,
45.
Shearing stress,
80
;
cone
of,
82. See also
Beams and Line.i
of shearing
stress.
Shells, Thin,
Inextensional
displacement
of,
471487
;
General
theory of, 29,
488316.
Simple solutions,
182,
187. See also
Nuclei,
of
strain.
Sound
waves,
96.
Sphere,
Problem of
the,
15
; Equilibrium of,
140,
236

257
;
with
given
surface
displace-
ments,
236
;
with
given
surface
tractions,
240
;
under
body forces,
246
;
Vibrations of,
268274.
Spherical cavity,
in infinite
solid,
245.
Spherical shell. Equilibrium
of,
under
pressure,
139,
161
;
under
any
surface
tractions, 245,
259
;
Vibrations
of,
274.
Spherical shell. Thin,
Inextensional deforma-
tion
of,
477
;
Inextensional vibrations
of,
486; Extensional and other vibrations
of,
522
; Equilibrium of, 526,
527.
Spheroidal vibrations, 273,
274.
Springs, Spiral, 23,
396.
Stability,
General criteria
for, 30, 97, 392;
Strength dependent on, 120;
of
strut, 3,
388391
;
of
elastica,
392
;
of rod
subjected
to
twisting couple
and
thrust,
399
;
of flat
blade bent in its
plane,
400
;
of
ring
under
pressure, 24,
405
; Height
consistent
with,
405
;
of
rotating shaft,
421
;
of
plate
under
thrust,
528
;
of tube under
pressure,
529.
Statical
metliod,
of
determining frequencies
of
vibration, 421,
422.
Steel,
Constants
for,
103
;
Elastic
limit,
and
Yield-point of,
114.
Strain, Cauchy's theory of,
8
; Examples of,
32,
33
;
Homogeneous, 30,
0471
; Pure, 39,
65
;
Determined
by displacement,
39
;
Com-
ponents of, 40,
59
;
Transformation
of,
42
;
Invariants
of, 43, 44, 60; Types
of,
44,
45
;
Kesolution
of,
into irrotational dilatation
and
equivoluminal distortion, 47
;
Identical
relations between
components of,
49
;
Dis-
placement effectively
determined
by,
50
;
referred to curvilinear
coordinates, 54,
540
;
General
theory of,
57

64; Composition
of
finite
homogeneous,
69
; Appropriated by
Rankine to denote relative
displacement,
531.
Strain-ellipsoid,
37,
61.
Strain-energy-furwtioii,
Introduction
of,
11
;
Existence
of, 12, 92;
Form
of, 96;
in
isotropic solids, 99,
152
;
in
saolotropic
solids,
98
;
in
crystals,
156
;
in solids ex-
hibiting
various
types
of
seolotropy,
157 ;
Generalisation
of,
97.
Strength, Ultimate,
112.
Stress, Cauchy's theory
of, 8;
Notion
of, 72,
533
;
Specification of,
75
; Components
of,
77
;
Measure
of,
77
;
Transformation of,
78
;
Types of, 79;
Eesolution
of,
into mean
tension and
shearing stress,
81
;
Uniform
and
uniformly varying,
84, 99,
101,
123
;
Graphic representation
of,
86
;
Lines
of,
87
;
Methods of
determining,
98
;
Direct deter-
mination
of, 17, 132, 133,
445
; Appropriated
by
Kankiue to denote internal
action,
531.
Stress-difference
: see
Rupture.
Stress-equations,
82
;
referred to curvilinear
coordinates, 87, 164,
540.
Stress-functions, 17, 85, 133, 201204,
260
263.
Stress-resultants and
stress-couples,
in
rod, 370;
in
plate, 28,
434
;
in
shell,
502.
Stress-strain
relations, 94, 95, 97, 100,
101.
Strut : see
Stability.
Supported etui,
of a
rod,
354
; edge
of a
plate,
441.
Surface of
revolution. Equilibrium
of solid
bounded
by,
260.
Surface
tractions,
73.
Surfaces,
Curvature
of,
488.
Suspension bridges,
347.
Symbolic notations, 287,
532.
Symmetry,
Geometrical,
147
; Alternating,
147;
Elastic,
148
;
of
crystals,
152
; Types
of,
157.
Tangential Traction, 77,
82.
Tenacity,
112.
Tension,
73
; Mean,
81.
Terminology,
531.
Testing machines.
111.
Tetragonal crystals,
155.
Thermal
effects, 93,
106.
Thermodynamics, Application of, 91,
93.
Thermo-elastic
equations,
106.
'Tidal
deformation,
of solid
sphere,
255.
Tidal
effective rigidity,
of the
Earth,
256.
Time-effects,
114.
Topaz,
Constants
for,
161.
INDEX
561
Tore,
Incomplete,
torsion and flexure
of,
423.
Torsion,
History
of
theory of, 4,
19

21
;
of a
bar of circular
section,
126
;
of
prisms
of
isotropic
material,
298

311
;
of
prisms
of
SBolotropic
material, 312;
of
prisms
of
special
forms of
section,
305

308, 313;
Stress and
strain,
that
accompany, 44, 298,
303
; strength
to
resist,
804.
Torsion
function,
300.
Torsion
problem,
300.
Torsional
couple,
in
rod,
376
;
in
plate,
434
;
in
shell,
,502.
Torsional
rvjidity,
300
;
Calculation
of,
310.
Torsional
vibrations,
of
cylinder, 276;
of
rod,
409
;
of circular
ring,
432.
Tortuosity,
Measure
of,
of central-line of
rod,
367, 368, 380, 395, 424,
427.
Traction,
72
;
used
by
Pearson in sense of
Tension,
532.
Transmission
offorce,
from
point
of
application,
180,
206. See also
Plane,
Problem of the.
Trees,
406.
Triclinic
crystals,
155.
Tubes : see
Cylindrical
shells.
Twinning,
of
crystals,
160.
Twist,
of a
rod, 365,
366.
Typical flexural $train,
476.
Uniqueness of solution,
of the
equations
of
equilibrium, 167;
of the
equations
of vibra-
tory motion, 173; Exceptions
to, 80,
892.
Variational
equation, 163; DifiSonlty
of
forming,
in case of thin
shells,
505.
Variations,
Calculus
of,
169.
Vibrations,
General
theory of,
176

178.
Viscosity,
115.
Watch
spring,
116.
}Vaves, Propagation
of,
in
isotropic media, 11,
18, 281286,
289295
;
in
aeolotropic media,
18, 286289;
in infinite
cylinder, 276280;
over surface of
solid,
295

297.
Wave
surfaces,
287.
Whirling,
of
shafts,
421.
Wires : see
Rods,
Thin.
Work,
done
by
external
forces,
91.
Yield
point,
112.
Young's
modulus, 4;
in
isotropic solid, 101;
in
aeolotropic solid,
105, 158; Quartic
surface
for, 105,
159.
CAMBRIDGE : PRINTED BY JOHN
CLAY,
M.A. AT THE UNIVERSITY PRESS.
-?
PLEASE
DO
NOT REMOVE
CARDS
OR SLIPS
FROM
THIS
POCKET
UNIVERSITY
OF TORONTO
LIBRARY
CiA Love,
Augustus
Edward
Hough
931
A treatise
oa the
L6
mathematical theory
of
1906
elasticity
P&ASci.