1 

Lecture 6

To Find Radius of Convergence From Ratio of Consecutive


Terms an / an 1 of the Power Series  a n ( z  z0 ) 
n , an  0
n 0
for all n .

The formula for radius of convergence in terms of an 1 / an
does not work if an = 0 for infinitely many n’s.


The Series  n
a ( z  z 0 ) n
with an  0 for all n is called a
n 0
Power Series with Gaps, if n  n .

Theorem 3.
1/( n n 1 )1 1/(   )
If lim an / an 1  lim an / an 1 n n 1 ,
exists, then
n  R n
where R is the radius of convergence of the power series

 n
a ( z  z 0 ) n
, an  0 for all n and {n } is any increasing
n 0
sequence of non‐negative integers such that
n   as n   .

2 
 

sup 1/( n n 1 ) B*

Proof. Let lim an / an 1  . Then,
n  inf A*

( A*   )n n1  | an / an 1 |  ( B*   )n n1 for all n  n0 .

n n0 ( n n 1 ) ....... ( n0 1 n0 )
 an0 ( A*   )  an  an0 ( B*   )
an0 1 an
(since an  an0 .... )
an0 an 1
1/ n 1/ n
 lim inf an  A* and limsup an  B* .
n  n 

1/(   )
Since, lim an / an 1 n n 1 exists, A*  B* .
n 

1/ n 1 1/  1/(   )
 lim an exists and  lim an n  lim an / an 1 n n1
n  R n n 

3 
 


1
Example.  n ( z  z0 ) .
n2

n 0 2

The radius of convergence R of the above power series is
given by
2 2
n 1 1/( n  ( n 1) )
1 2
 lim  n 
R n  2 

1/(2 n 1)
1
 lim   1
n   2 

4 
 

Radius of Convergence of Product of Power Series

Let

 n
a ( z  z n
0 has radius of convergence R1
)
n 0
and

 bn ( z  z0 )n has radius of convergence R2
n 0


Hadamard Product:  an bn  z  z0  is called the Hadamard
n

n 0
Product of the above two power series.

Let R* be its radius of convergence.

1/ n 1/ n 1/ n
Since, limsup an bn  limsup an limsup bn (prove!)
n  n  n 
1 1
 *
  R *
 R R
1 2  [min( R ,
1 2R )]2

R R R
1 2
5 
 

*Cauchy Product:

The power series  ( a0bn  ...  an b0 ) ( z  z0 )n (*)
n 0
is called the Cauchy Product of the above two Power series.
Let R be its radius of convergence.

Proposition. If the radius of convergence of

f ( z )   an  z  z0  is R1 and radius of convergence of
n

n 0

g ( z )   bn  z  z0  is R2 , then the radius of convergence of
n

n 0
their Cauchy product is R  min( R1 , R2 ) .

Proof. WLOG assume that z0 = 0.
n
Let Sn ( z ) be n th
partial sum  ak z k (1)
k 0
n
Tn ( z ) be n th partial sum k
b z k
(2)
k 0
n
Pn ( z ) be n th partial sum  (a0bk  ...  ak b0 ) z k .
k 0
To show: If f(z) is limit of (1) as n  and g(z) is limit of (2)
as n  , then (*) has the sum f(z).g(z) in
z  z0  R  min( R1 , R2 ) , where R1 is the radius of
convergence for power series of f(z) and R2 is the radius of
convergence for power series of g(z).
6 
 

Now,
Pn ( z )  a0b0  ( a0b1  a1b0 ) z  ...  ( a0bn  ...  an b0 ) z n
 a0Tn ( z )  a1Tn 1 ( z ) z  ...  anT0 ( z ) z n

 a0 ( n ( z )  g ( z ))  a1 ( n 1 ( z )  g ( z )) z  ...  an ( 0 ( z )  g ( z )) z n ,
where  n ( z )  Tn ( z )  g ( z )  Tn ( z )   n ( z )  g ( z )

 [ Sn ( z ) g ( z )]  [a0 n ( z )  a1 n 1 ( z ) z  ...  an 0 ( z ) z n ] .

Since,  n ( z ) 0 as n   in z  R2 ,  n ( z )   n  N and

further  ( z )   an z   in z  R1 ,
n

n 0
 n ( z )  a0 n ( z )  a1 z n 1 ( z )  ...  an z n 0 ( z )

 an z n 0 ( z )  ...  an  N z n  N  N ( z )

 an ( N 1) z n ( N 1) N 1 ( z )  ...  a0 n ( z )

 an z n 0 ( z )  ...  an  N z n  N  N ( z )
  ( z ) in z  min( R1 , R2 )
 lim  n ( z )   ( z ) ( since an zn  0 as n  )
n 
 lim  n ( z )  0, ( since  is arb.)
n 
 lim Pn ( z )  lim Sn ( z ) g ( z )  f ( z ). g ( z ) .
n  n 
 
 
7 
 

Complex Integration

Let C : z (t ), a  t  b , be a  zn  
continuously differentiable
curve, i.e. z(t) is a continuously  zn 1  

differentiable function in [a, b] .

For any partition  z2   
C : z (t )  
 z1  
{a  t0 , t1 ,..., tm 1 , tm  b}of [a , b] ,
 z0  

let z (t j )  z j , j  0,1,..., m
a  t0 t1    t2    tn1    tn   

Let the function f : C  C be continuous on the curve C .
n
Consider the sum Sn   f ( m )( zm  zm 1 )
m 1
where,  m is any point on the curve lying between zm 1 and
zm .

Definition. Complex Integration of f on C is defined as

 f ( z ) dz  lim Sn , provided max zm  0 as n  
C n  1 m  n



8 
 

Equivalently, using the definition of integral of real functions,
b
 f ( z ) dz   f ( z (t )) z (t ) dt . (*)
C a

Note. The definition (*) is independent of the parametric
representation of C. For if w(t) , c  t  d , is another
parametric representation of C . Let  :  c, d    a, b be one‐
one onto differentiable function such that  ( c )  a ,  ( d )  b
and w(t )  z ( (t )) . Then,
d d
 (t ) dt   f ( z ( (t ))) z( (t )) (t ) dt
 f ( w(t )) w
c c

 (d ) b
  f ( z ( x )) z ( x ) dx   f ( z ( x )) z( x ) dx
 (c) a


z (t )  
w( t )  
C 
a  b  c  d 


 (t )  



9 
 

For studying the properties of integration of a function

f : C  C, we need the definition and properties of a function
F : [a, b]  C.

Integration of Functions F : [a, b]  C

Let F (t )  u (t )  i v (t ), a  t  b . Define,

b b b
 F (t ) dt   u (t ) dt  i  v (t ) dt
a a a

Properties.

b b
(i) Re  F (t ) dt   Re F (t ) dt
a a
b b
(ii)   F (t ) dt    F (t ) dt ,   a complex constant
a a
b b
(iii) |  F (t ) dt |   | F (t ) | dt
a a

Proof: (i) and (ii) immediately follow from the definition.






10 
 

b
i0
(iii) Let r0 e   F (t ) dt . Then,
a
b
r0   e i0 F (t ) dt
a

 
b
  Re e i0 F (t ) dt (using (i ))
a
b b
 i0
  e F (t ) dt   F (t ) dt (using property of real integral)
a a

11 
 

Properties of Complex Integration  f ( z ) dz

C

(1)  f ( z ) dz    f ( z ) dz (use that if parametric rep. of C is
C C
z (t ), a  t  b , then parametric rep. of –C is z ( t ) : b  t   a
or, alternatively, z (b  ( a  t )) : a  t  b )


(2)   f ( z ) dz    f ( z ) dz ( easily follows from definition )
C C

(3)
 ( f  g ) dz   f dz   g dz ( follows easily from definition )
C C C
12 
 

(4) If C1 is continuous curve from 1 to 1 , C2 is a

continuous curve  2 to  2 and 1   2 and the curve C is
union of curves C1 and C2 , then

 f ( z ) dz   f dz   f dz
C C1 C2

(5)  f ( z ) dz  ML, where L is the length of C and
C
f ( z )  M for z  C .

(The property (5) is called ML‐Estimate of the integral).

13 
 

Proof (4):

Let C1 : z1 (t ), 0  t  1 and C2 : z2 (t ), 0  t  1 .

 z (2t ), 0  t  1 / 2
Then, C = z(t), where z (t )   1
 z2 (2t  1), 1 / 2  t  1

Therefore,

1
 f ( z ) dz   f ( z (t )) z (t ) dt
C 0

1/2 1
  f ( z (t )) z(t ) dt   f ( z (t )) z (t ) dt
0 1/2

1/2 1
  f ( z1 (2t ))( z1 (2t ).2) dt   f ( z2 (2t  1)) ( z2 (2t  1).2) dt
0 1/2
1 1
  f ( z1 (t ) ( z1 (t ).) dt   f ( z2 (t ) z2 (t ) dt
0 0
  f ( z ) dz   f ( z ) dz
C1 C2

14 
 

Proof (5):
b
 f ( z ) dz   f ( z (t ) z (t ) dt
C a

b
 M  z (t ) dt  ML
a

15 
 

Example 1: Show that

dz
 4  4 2, where C : The line segment joining i and 1.
Cz
        i   
Solution:
y  1 x
L = length of C = 2

1 
On C, z  x  y  x  (1  x )
2 2 2 2

4 1 1 1
 z  ( x 2  (1  x )2 )2  (2 x 2  2 x  1)2  [2( x  )2  ]2 
2 2 4

Using Property 5, we now get the required estimate of the
integral in Example 1.