/// 5/// SEDIMENTATION

Sedimentation is the movement of particles or macromolecules

in an inertial field. Its applications in separation technology are extremely widespread.
Extremes of applications range from the settling due to gravity of tons of solid waste
and bacteria in wastewater treatment plants to the centrifugation of a few microliters of
blood to determine packed blood cell volume (“hematocrit”) in the clinical laboratory.
Accelerations range from 1 × g in flocculation tanks to 100,000 × g in ultracentrifuges
for measuring the sedimentation rates of macromolecules. In bioprocessing, the most fre-
quent applications of sedimentation include the clarification of broths and lysates, the col-
lection of cells and inclusion bodies, and the separation of fluids having different densities.
Unit operations in sedimentation include settling tanks and tubular centrifuges
for batch processing, continuous centrifuges such as disk centrifuges, and less fre-
quently used unit operations such as field-flow fractionators and inclined settlers.
Bench scale centrifuges that accommodate small samples can be found in most
research laboratories and are frequently applied to the processing of bench scale cell
cultures and enzyme preparations. Certain high-speed ultracentrifuges are used as
analytical tools for the estimation of molecular weights and diffusion coefficients.
The chapter begins with a description of the basic principles of sedimentation,
followed by methods of characterizing laboratory and larger-scale centrifuges. Two
important production centrifuges, the tubular bowl centrifuge and the disk-stack cen-
trifuge, are analyzed in detail to give the basis for scale-up. Ultracentrifuges, impor-
tant for analytical and preparative work, are then analyzed. The effect of flocculation
of particles on sedimentation is presented, and sedimentation of particles at low
accelerations is discussed. The chapter concludes with a description of centrifugal
elutriation.

5.1 INSTRUCTIONAL OBJECTIVES

After completing this chapter, the reader should be able to do the following:

r Determine the velocity of a sedimenting particle and calculate sedimentation

times, equivalent times, and sedimentation coefficients in gravitational and
centrifugal fields.
185
186 // BIOS E PA R AT IONS SC IENCE AND E N GINEE RIN G

r Perform engineering analyses and scaling calculations on tubular bowl and

disk-stack centrifuges.
r Choose the appropriate centrifuge for particular liquid-solid separations.
r Calculate molecular weight from ultracentrifuge data.
r Explain the behavior of sedimenting flocs.
r Discern the relative importance of diffusion in a sedimentation operation.
r Explain how inclined sedimentation, field-flow fractionation, and centrifugal
elutriation work.

5.2 SEDIMENTATION PRINCIPLES

The most frequently encountered inertial fields are gravitational acceleration,
g = 9.8 m/s2, or centrifugal acceleration, 2R, where R is the distance of the particle
from the center of rotation, and is angular velocity (rad/s). The same theory can
be used to describe sedimentation in both types of inertial fields.

5.2.1 Equation of Motion

Analysis begins with the equation of motion of a particle of radius a and density
having mass m = (4/ 3)πa3 in an inertial field moving radially at a distance R from
the center of rotation (Figure 5.1)
Assuming the particle is spherical,

dv ⎛ 4⎞ ⎛ 4⎞
(5.2.1)  m = FiR + FbR + FdR = ⎜ ⎟ π a3 ρω 2 R − ⎜ ⎟ π a3 ρ0ω 2 R − 6π µ a υ
dt ⎝ 3⎠ ⎝ 3⎠

Direction of movement

Particle of radius a

Center of rotation
FIGURE 5.1 Particle moving radially under the influence of an inertial field at distance R from the
center of rotation.
Sedimentation // 187

where subscripts designate forces in the R direction, due to i, inertial acceleration, b,

buoyancy due to the density of the medium 0 through which the particle sediments,
and d, the Stokes drag force, which under conditions of creeping flow is proportional
to the velocity and viscosity µ [1]. Solving Equation (5.2.1) for velocity in a cen-
trifugal field at steady state (all forces balanced, so dv/dt = 0) gives

2 a 2 (ρ − ρ0 )ω 2 R
(5.2.2) υ =
9µ

commonly called the “centrifuge equation.” If particles are allowed to sediment only
in the presence of gravity, then the inertial acceleration is g = 9.8 m/s2, and 2R in
Equation (5.2.2) is replaced by g. If, however, the particle is moving outwardly from
the center of rotation in a centrifuge, R is not constant and is related to the velocity
by v = dR/dt, which gives from Equation (5.2.2) after rearrangement,

dR ⎡⎣2 a (ρ − ρ0 )ω ⎤⎦ dt
2 2
(5.2.3)  =
R 9µ

with the following initial condition:

(5.2.4)  at t = 0, R = R0

This equation can be integrated to give

⎛ R ⎞ 2 a 2 (ρ − ρ0 )ω 2 t
(5.2.5)  ln ⎜ ⎟= 9µ
⎝ R0 ⎠

This is a useful equation that relates time to the distance traveled by the particle.

5.2.2 Sensitivities
Creeping flow conditions are usually satisfied in sedimentation. Calculating the
Reynolds number for spherical particles

(5.2.6) Re = 2a υρ
µ

Creeping flow occurs at Reynolds numbers less than about 0.1 [1]. Table 5.1 shows
the sedimentation velocities for important bioparticles and biomolecules calculated
from Equation (5.2.2) with 0 = 1.0 g/cm3, µ = 0.01 g cm−1 s−1 (poise), and repre-
sentative values of and a. The sedimentation velocity and Reynolds number results
shown in Table 5.1 for yeast cells and bacterial cells at gravitation acceleration can
be multiplied by a centrifuge’s centrifugal acceleration to give the corresponding
188 // BIOS E PA R AT IONS SC IENCE AND E N GINEE RIN G

TABLE 5.1 Calculated Settling Velocities and Reynolds Number for Example Bioproducts (Assumes
0 = 1.0 g/cm3 and µ = 1.0 cp)

Dimensionless Sedimentation
Bioparticle or Sedimentation Density, acceleration velocity, Reynolds
biomolecule Radius, a (µ m) (g/cm3) (G =  2R/g) v (cm/h) number, Re
Yeast cell 2.5 1.1 1 0.5 7 × 10−6
Bacteria cell 0.5 1.1 1 0.02 6 × 10−8
Protein 0.005 1.3 104 0.06 2 × 10−9

values for operation in the centrifuge; for example, at a dimensionless acceleration

of 10,000, the Reynolds number for yeast cells is 0.07, which means that the flow is
still creeping.
It is clear from Table 5.1 that gravitational sedimentation is too slow to be practi-
cal for bacteria, and conventional centrifugation is too slow for protein macromol-
ecules. In the case of true particles, flocculation (see Section 5.6) is often used to
increase the Stokes radius a, while ultracentrifugation (see Section 5.5) is used in
macromolecular separations.
When particle density and solvent density are equal, the sedimentation velocity v
is zero, and the process is called isopycnic or equilibrium sedimentation. This fact is
exploited in the determination of molecular densities and in the separation of living
cells. A density gradient or a density shelf is employed in such cases. Densities of
representative cells, organelles, and biomolecules measured by this method are given
in Table 5.2 [2–5]. The density of Amoeba proteus cells is low because these cells
contain fat vacuoles of low density and have no cell wall.
An example of a density shelf used for the separation of cells is the preparation of
lymphocytes by sedimentation. The goal of this separation is to remove erythrocytes
from a leukocyte population on the basis of a density shelf. By combining Ficoll,
a high molecular weight polymer, and Hypaque, a heavily iodinated benzoic acid
derivative, in appropriate proportions in aqueous buffers, it is possible to achieve a
density around 1.07 g/cm3 in isotonic solutions. At this density most white blood cell
subpopulations will float and nearly all red blood cells will sediment.
When the concentration of sedimenting particles increases, the sedimentation
velocity has been found to decrease, a phenomenon known as “hindered settling.” This
effect has been quantified by the following expression for particles of any shape [6]:

(5.2.7) υc = υ (1 − ϕ )n

where vc is the sedimentation velocity of particles in a concentrated suspension, v is

the velocity of individual particles [Equation (5.2.2)], is the volume fraction of the
particles, and n is a function only of the shape of the particle and of the Reynolds
number. For spherical particles with Re < 0.2 (usually satisfied during sedimenta-
tion), the exponent n has been found to be 4.65. Equation (5.2.7) may also be applied
Sedimentation // 189

TABLE 5.2 Measured Values of the Density of Representative Cells,

Organelles, and Biomolecules

Cell, organelle, or biomolecule Density, (g/cm3) Ref.

Escherichia coli 1.09a 2
Bacillus subtilis 1.12 3
Arthrobacter sp. 1.17 4
Saccharomyces pombe 1.09 2
a
Saccharomyces cerevisiae 1.11 2
Amoeba proteus 1.02 2
a
Murine B cells 1.06 2
Chinese hamster ovary (CHO) cells 1.06 2
Peroxisomes 1.26a 5
a
Mitochondria 1.20 5
Plasma membranes 1.15a 5
a
Proteins 1.30 5
Ribosomes 1.57a 5
a
DNA 1.68 5
RNA 2.00a 5

a
Average value.

TABLE 5.3 Effect of Particle Volume

Fraction on the Particle Sedimentation
Velocity for Spherical Particles

vc / v
0.01 0.95
0.05 0.79
0.10 0.61
0.20 0.35

to particles of any size in a polydisperse system, using the volume fraction for all the
particles in the calculation [7].
The magnitude of the hindered settling effect for spherical particles as a function
of the particle volume fraction can be seen in Table 5.3. Note that hindered settling
can be significant for particle concentrations of a few percent or greater.

5.3 METHODS FOR ANALYSIS OF SEDIMENTATION

While the fundamental principles of sedimentation are important to a basic understanding
of this subject, other less theoretical methods have been developed for the actual practice
of sedimentation. Equilibrium sedimentation, the sedimentation coefficient, equivalent
sedimentation time, and sigma analysis are some of the more important of these methods.
190 // BIOS E PA R AT IONS SC IENCE AND E N GINEE RIN G

5.3.1 Equilibrium Sedimentation

Some products can be isolated on the basis of their density. In ultracentrifugation
this is true of nucleic acids, and isopycnic (same-density) sedimentation was the
method originally used to demonstrate the semiconservative replication of DNA [8].
When  =  0 in Equation (5.2.2), then v = 0 and all inertial motion stops; therefore, if
the density of the solution is known at the location where motion stops, the density
of the solute or suspended particle is known. In most applications inertial motion
is arrested in a density gradient, so that the density of the medium increases below
the arrest point, and the buoyant force [Equation (5.2.1)] is greater than the inertial
force, which causes the particle to return to its isopycnic level. Centrifugation can
therefore be used analytically to determine particle or macromolecule density, as
discussed earlier.
In practice, there are at least three routes to the establishment of conditions for
isopycnic sedimentation—the creation of a region in the sedimentation vessel where
0 ≥ . One method is to layer solutions of decreasing density, starting at the bot-
tom of the vessel and proceeding until the vessel is filled. The resulting gradient is
like a staircase until diffusion smooths it out. Another is to centrifuge at extremely
high speed, resulting in isothermal stratification of a density-forming solute, such
as CsCl. Such a gradient is not necessarily linear. The most widely used method of
forming a density gradient is the gradient mixing method, in which two cylindrical
containers, one containing a concentrated solution and the other containing a dilute
solution and a stirring apparatus, are linked as in Figure 5.2 to produce an outflow
with a linear salt gradient. For these gradient mixers, the time-dependent solute con-
centration is as follows [9]:

c1,0 + Q(c2 − c1,0 )

(5.3.1)  c(t ) = t
2V0

where Q is the outflow rate due to pumping or gravity feed, V0 is the initial volume in
each cylinder, c1,0 is the initial concentration of solute in the mixed chamber, and c2

Paddle
Salt stirrer
solution Salt
solution

Magnetic stirrer
FIGURE 5.2 Two types of a linear gradient mixer.
Sedimentation // 191

is the concentration of solute in the nonmixed chamber (constant). By programming

Q, one can set up a variety of salt gradients.

5.3.2 Sedimentation Coefficient

When a body force is applied, velocity through a viscous medium is usually propor-
tional to the accelerating field (examples are electric, magnetic, and inertial). In the
case of sedimentation, the resulting constant, a property of both the particle and the
medium, is the sedimentation coefficient, which is defined as

υ
(5.3.2) s ≡
ω2 R

Comparing this equation with Equation (5.2.2), we see that

2 a 2 ( ρ − ρ0 )
(5.3.3)  s =
9µ

which defines s in terms of only properties of the particle and the medium. This coef-
ficient is usually expressed at 20°C and under conditions (viscosity and density) of
pure water as

s20,w (s )

The sedimentation coefficient is often expressed in svedberg units, where 10−13

s  =  1 svedberg unit (S), named after the inventor of the ultracentrifuge, Theodor
Svedberg.

EXAMPLE 5.1
Application of the Sedimentation Coefficient In 1974 D. E. Koppel measured the
sedimentation coefficient (s20,w) for the smaller ribosomes from Escherichia coli at
70 S (Koppel, D. E., Biochemistry, 13, 2712, 1974). Estimate how long it would take
to completely clarify a suspension of these ribosomes in a high-speed centrifuge
operating at 10,000 rpm with a tube containing the ribosome suspension in which the
maximum distance of travel of particles radially outward is 1 cm and the initial dis-
tance from the center of rotation to the particles nearest the center of rotation is 4 cm.

Solution
We can write Equation (5.3.2) as

dR 1
s=
dt ω 2 R
192 // BIOS E PA R AT IONS SC IENCE AND E N GINEE RIN G

or
dR
ω 2 s dt =
R

We integrate this equation with the following initial condition:

at t = 0, R = R0 (distance from center of rotation to the particles nearest the center
of rotation)
to give
R
ω 2 st = ln
R0

To determine the maximum time required, we evaluate R at the maximum travel

of the cells measured from the center of rotation (5 cm):

⎛ R⎞ 1h
ln ⎜ ⎟ ln(5 / 4)
⎝ R0 ⎠ 3600 s
t= = = 8.1 h
ω 2s ⎛ rev 2π rad 1 min ⎞
2
−13
⎜⎝ 10, 000 min × rev × 60 s ⎟⎠ (70 × 10 s)

This should not be an unreasonable amount of time to centrifuge the ribosomes.

However, since the time varies inversely with the square of the rotation speed, the
time can be reduced to 2 h by doubling the speed.

5.3.3 Equivalent Time

To assess the approximate properties of a particle type to be separated, it is sometimes
convenient to calculate an “equivalent time.” To do this, we first define a dimension-
less acceleration, G, the ratio of the centrifugal to gravitational acceleration for a
particular centrifuge:

ω2R
(5.3.4)  G ≡
g

where R is usually defined as the radius of the centrifuge bowl. Thus, this dimension-
less unit is measured in “g’s”—multiples of the earth’s gravitational acceleration.
A rough approximation of the difficulty of a given separation by centrifugation is the
product of the dimensionless acceleration and the time required for the separation.
This product is called the equivalent time for the separation, and is written as

ω2R
(5.3.5)  Equivalent time ≡ Gt = t
g
Sedimentation // 193

Typical values of equivalent time are as follows: 0.3 × 106 s for eukaryotic cells,
9 × 106 s for protein precipitates, 18 × 106 s for bacteria, and 1100 × 106 s for
ribosomes [10].
The equivalent time for the centrifugation of cells or biological particles of
unknown sedimentation properties may be estimated in a laboratory centrifuge.
Samples are centrifuged for various times until a constant volume of packed cells is
reached. The equivalent time Gt is calculated as the product of the G for the particu-
lar centrifuge and the time required to reach constant packed-cell volume. A centri-
fuge that has commonly been used for this determination is the Gyro-Tester (Alfa
Laval, Inc.).
One approach to scale-up of a centrifugal operation is to assume constant equiva-
lent time:

(5.3.6)  (Gt )1 = (Gt )2

where the subscripts refer to centrifuges 1 and 2, respectively.

EXAMPLE 5.2
Scale-up Based on Equivalent Time If bacterial cell debris has Gt = 54 × 106 s [10],
how large must the centrifuge bowl be, and what centrifuge speed is needed to effect
a full sedimentation in a reasonable amount of time?

SOLUTION
Assume that a reasonable amount of time is about 2 h. From Equation (5.3.5) for Gt,
we can estimate the centrifuge speed if we know the centrifuge bowl size and the
time of centrifuging. It is reasonable to have a centrifuge that is 10 cm in diameter.
Solving Equation (5.3.5) for using these values gives
1/ 2
⎛ m⎞
1/ 2 54 × 106 s × 9.81 2
⎛ Gtg ⎞ ⎜ s ⎟ rad 1 rev 60 s
ω =⎜ =⎜ = 1213 × × = 11, 590 rpm
⎝ Rt ⎟⎠ ⎟ 2π rad min
⎜ 0.05 m × 2(3600) s ⎟ s
⎝ ⎠

This speed can be achieved in a production tubular bowl centrifuge (see later:
Table 5.5).

5.3.4 Sigma Analysis

The more commonly used analysis in industry is “sigma analysis,” which uses the
operation constant to characterize a centrifuge into which feed flows at volumetric
194 // BIOS E PA R AT IONS SC IENCE AND E N GINEE RIN G

flow rate Q. Since the engineer often needs to determine Q on scale-up, a convenient
relationship is

(5.3.7) Q = {υ g }[ ∑ ]

where vg is the sedimentation velocity at 1 × g, namely,

(ρ − ρ0 )g
(5.3.8) υ = 2 a 2
g
9µ

and represents the geometry and speed of the centrifuge, as derived in the dis-
cussion of individual centrifuges in the next section; can also be thought of as
the cross-sectional area equivalent of the centrifuge, with units of length squared.
Therefore, in Equation (5.3.7) the accolades {} indicate properties of the particle to
be separated and of the fluid in which separation is occurring, and the square brack-
ets [] indicate properties of the centrifuge.
The sedimentation velocity at 1 × g can be estimated directly by using Equation
(5.3.8) if relevant properties of the system are known, or it can be measured in the
laboratory. Combining Equations (5.3.8) and (5.2.5), we obtain

⎛ R⎞
g ln ⎜ ⎟
⎝ R0 ⎠
(5.3.9) υ =
g
ω2t
This is a useful equation for determining vg in the laboratory. Parameters in Equation
(5.3.9) can be measured as follows. The minimum time t to clarify the sample in a
laboratory centrifuge at speed is determined; R and R0 are the distances from the
center of rotation to the top of the packed solids and to the top of the liquid in the
centrifuge tube, respectively. Since the sample cannot be observed directly when the
centrifuge is being operated, a series of experiments would be required to determine
the time for centrifuging at which R is constant.

5.4 PRODUCTION CENTRIFUGES: COMPARISON

AND ENGINEERING ANALYSIS
The common types of production centrifuges are illustrated in Figure 5.3, and a
comparison of the advantages and disadvantages of the different centrifuge designs
is given in Table 5.4 [11]. In the tubular centrifuge, solids deposit on the wall of the
bowl, and feed continues until the bowl is almost full, at which time the operation
is stopped and the solids removed. This type of centrifuge works well for particles
of relatively low sedimentation coefficient that must be recovered, such as pro-
tein precipitates. The disk centrifuges have a relatively high sedimentation area
for their volume and allow for continuous or intermittent solids discharge; they
Sedimentation // 195

Solids

(a) (b) (c)

Solids

(d ) (e) (f )
FIGURE  5.3 Common types of production centrifuges:  (a) tubular bowl, (b) multichamber, (c)
disk, nozzle, (d) disk, intermittent discharge, (e) scroll, and (f) basket. Arrows indicate the path of
the liquid phase; dashed lines show where the solids accumulate.

have been successfully used for the centrifugation of cells and cell lysates, where
the entire process often must be contained to avoid the escape of aerosols. The
scroll (or decanter) and basket centrifuges are typically used for particles that sedi-
ment relatively rapidly and can be washed well as packed solids, such as antibiotic
crystals.
Of all the centrifuges, the tubular and the disk types are probably the most
likely to be found in a bioseparation process involving the recovery of a protein
produced by cells. The capabilities of tubular and disk centrifuges are given in
Table 5.5 [12]. Note that there is generally a reduction in the maximum g-force
(dimensionless acceleration) as the diameter of the bowl increases. The tubular
bowl and disk centrifuges are analyzed to develop the value that can be used in
scaleup.

5.4.1 Tubular Bowl Centrifuge

The tubular bowl centrifuge allows fluid to enter at one end of a rotating cylinder
and exit at the opposite end, while particles move toward the wall of the cylinder
and captured both at the wall and by a weir at the exit, as indicated in Figure 5.4.
The liquid enters the bowl through an opening in the center of the lower bowl
head. The liquid is pushed by centrifugal force toward the periphery of the rotat-
ing bowl. Clarified liquid overflows a ring weir in the upper bowl head, the radius
TABLE 5.4 Comparison of Production Centrifugesa

System Advantages Disadvantages

Tubular bowl (a) High centrifugal force (a) Limited solids capacity
(b) Good dewatering (b) Foaming unless special skimming
(c) Easy to clean or centripetal pump used
(d) Simple dismantling of bowl (c) Recovery of solids difficult
Chamber bowl (a) Clarification efficiency remains (a) No solids discharge
constant until sludge space full
(b) Large solids holding capacity (b) Cleaning more difficult than
tubular bowl
(c) Good dewatering (c) Solids recovery difficult
(d) Bowl cooling possible
Disk centrifuge (a) Solids discharge possible (a) Poor dewatering
(b) Liquid discharge under pressure (b) Difficult to clean
eliminates foaming
(c) Bowl cooling possible
Scroll or decanter (a) Continuous solids discharge (a) Low centrifugal force
centrifuge (b) High feed solids concentration (b) Turbulence created by scroll
Basket centrifuge (a) Solids can be washed well (a) Not suitable for soft biological
solids
(b) Good dewatering (b) No solids discharge
(c) Large solids holding capacity (c) Recovery of solids difficult

a
See reference [11].

TABLE 5.5 Capabilities of Tubular and Disk Centrifugesa

Bowl diameter Speed Maximum dimensionless Throughput

Type (mm) (rpm) acceleration (G), 2R/g (liters/min)
Tubular bowl 44 50,000 61,400 0.2–1.0
105 15,000 13,200 0.4–38
127 15,000 16,000 0.8–75
Disk with nozzle 254 10,000 14,200 40–150
discharge 406 6,250 8,850 100–570
686 4,200 6,760 150–1500
762 3,300 4,630 150–1500

a
See reference [12].
Sedimentation // 197

Exit
Ring weir

R0

R1
Liquid
L
interface

R
z

Feed
FIGURE 5.4 Cross section of a tubular centrifuge in operation.

of which establishes the depth of the pool of liquid around the periphery of the
rotating bowl.
As in most engineering calculations, we wish to determine the flow rate Q. The
equations of motion that give the trajectory of sedimented particles are, first, in the
radial direction from Equation (5.2.2)

dR 2 a 2 (ρ − ρ0 )ω 2 R
(5.4.1)  =
dt 9µ

then in the axial direction, due only to pumped flow, Q

dz Q Q
(5.4.2)  = =
dt A π ( R0 − R12 )
2

where A is the cross-sectional area for liquid flow in the centrifuge. These equations
of motion are combined to give the trajectory equation

dR
dR
(5.4.3)  dt =
dz dz
dt

Substituting Equations (5.4.1) and (5.4.2) into this ratio, integrating dR between R0
and R1, and integrating dz between 0 and L and solving for Q gives
198 // BIOS E PA R AT IONS SC IENCE AND E N GINEE RIN G

⎡ ⎤
⎢ ⎥
⎪⎧ 2 a (ρ − ρ0 ) ⎪⎫ ⎢ π L ( R0 − R1 )ω ⎥
2 2 2 2
(5.4.4)  Q = ⎨ ⎬⎢ ⎥
⎩⎪ 9µ ⎭⎪ ⎢ ⎛R ⎞
ln ⎜ 0 ⎟ ⎥
⎢⎣ ⎝ R1 ⎠ ⎥⎦

In view of Equations (5.3.7) and (5.3.8), the first factor in Equation (5.4.4) can be multi-
plied by g while the second is divided by g to give, again, Equation (5.3.7) for analysis:

(5.4.5)  Q = {υ g }[ ∑ ]

where, for a tubular bowl centrifuge,

π L ( R02 − R12 )ω 2
(5.4.6)  ∑ =
⎛R ⎞
g ln ⎜ 0 ⎟
⎝ R1 ⎠

In practice, one uses a benchtop centrifuge to determine vg and chooses a cen-

trifuge with throughput in the desired range. The centrifuge manufacturer supplies
for the selected centrifuge, and the flow rate is determined from Equation (5.4.5).
Alternatively, if the flow rate is constrained by some other factor, a custom centrifuge
with a specific can be constructed. This is also true for the continuous disk-stack
centrifuge, which is the subject of the next section.
A tubular bowl centrifuge operates in the batch mode. The bowl begins spinning,
either dry or full of water, which will be displaced by the solids suspension. The sol-
ids suspension is fed from the bottom of the bowl, as previously described, typically
through a jet that sprays the suspension onto the wall of the bowl. If the bowl begins
empty, the suspension distributes rapidly up the length of the bowl in a thin film. As
the suspension is fed, the liquid layer grows from the wall toward the center of the
bowl, until the thickness of the liquid layer is equal to the weir at the top of the bowl.
At this point, clarified liquid begins to exit the centrifuge at the rate of the feed. The
residence time of fluid in the bowl is

π L ( R02 − R12 )
(5.4.7)  τ =
Q

As solids build up on the wall of the bowl, the inner radius of the bowl is effectively
decreased, reducing the residence time in the bowl, but also reducing the sedimen-
tation distance and the maximum sedimentation velocity [see Equation (5.2.2)].
Referring to Equation (5.4.6), it can be seen that decreases with time during tubu-
lar bowl operation, as R0 approaches R1. Therefore, in order to continue to capture
particles in the tubular bowl, the flow rate has to be decreased as operation proceeds.
Sedimentation // 199

Practically, this is not done—the centrifuge is fed until particles break through, and
then the operation is discontinued and the bowl emptied of solids. Typically, the sol-
ids occupy no more than 80% of the bowl volume.
There are now commercial tubular bowl centrifuges which have mechanical
means to discharge solids. This feature makes operation of the tubular bowl centrifuge
more convenient and likely reduces downtime. However, the centrifuge still has to be
stopped, solids discharged and then restarted, which reduces the efficiency overall.
Typically, tubular bowl centrifuges can concentrate solids to essentially 100%
wet weight, or the weight you would measure by centrifuging a sample in the labo-
ratory and decanting the supernatant. The solids are not required to flow in order to
operate the equipment; in fact, a “tight pellet” is preferred. This is an advantage of a
tubular bowl centrifuge over a disk centrifuge.

EXAMPLE 5.3
Complete Recovery of Bacterial Cells in a Tubular Bowl Centrifuge It is desired
to achieve complete recovery of bacterial cells from a fermentation broth with a pilot
plant scale tubular centrifuge. It has been determined that the cells are approximately
spherical with a radius of 0.5 µm and have a density of 1.10 g/cm3. The speed of the
centrifuge is 5000 rpm, the bowl diameter is 10 cm, the bowl length is 100 cm, and
the outlet opening of the bowl has a diameter of 4 cm. Estimate the maximum flow
rate of the fermentation broth that can be attained.

SOLUTION
The flow rate can be estimated from Equation (5.4.5) by determining the settling
velocity under gravity (vg) and the factor for the centrifuge. We can estimate vg
from Equation (5.3.8) and assuming the viscosity µ is the same as for water (1.0 cp):

g m 106 cm
m3
2 a ( ρ − ρ0 ) g
2 2(0.5 × 10 −6 m )2 × (1.10 − 1.00) × 9 .81 ×
υg = = cm 3 s2 m3
9µ ⎛ g ⎞
9 ⎜ 0.01
⎝ cm s ⎟⎠
= 5.45 × 10 −6 cm/s

For complete recovery of the cells, we can use Equation (5.4.6) to estimate :

2
⎛ rev 2π rad ⎞
Σ=
πL ( R02 − R12 )ω 2
=
π (100 cm ) × (52 − 22 ) cm 2 × ⎜ 5000
⎝ min
×
rev ⎟⎠
⎛R ⎞ m 100 cm ⎛ 60 s ⎞
2
g ln ⎜ 0 ⎟ 9.81 × ln (5 / 2) × ×⎜
⎝R ⎠ 1 s 2
m ⎝ min ⎟⎠
= 2.01 × 106 cm 2
200 // BIOS E PA R AT IONS SC IENCE AND E N GINEE RIN G

From Equation (5.4.5):

liter 60 s liter
Q = υ g ∑ = (5.45 × 10 −6 cm/s ) × (2.01 × 106 cm 2 ) × × = 0.66
103 cm3 min min

5.4.2 Disk Centrifuge

A disk centrifuge is a system of rapidly rotating concentric inverted cones placed
close together to minimize the time to capture dense particles or liquids at rela-
tively high dimensionless accelerations. In this configuration (Figure 5.5), the
feed suspension enters on the axis of rotation and is forced to the bottom of the
rotating bowl. Pressure forces the suspension upward. The heavier fluid is forced
through holes at the end of each disk channel until it reaches the outer periphery
of the bowl. The lighter fluid flows up the disk channels and out of the centri-
fuge. The heavier sediment flows out through a nozzle if it is open; otherwise it
collects on the outer wall of the bowl. Depending on whether the nozzle is open
or closed, this centrifuge is operating in continuous or batch mode, respectively.
The bowl can be designed to open intermittently at the periphery in the semicon-
tinuous mode.
To determine the maximum feed rate Q, a simplified diagram of a single zone
between two disks is used, as shown in Figure 5.6. For convenience, x-y coordinates
are placed in a vertical plane that intersects the rotor axis with the x axis parallel to
the disk surface. Equations of motion of a suspended particle are then determined
with the constraint that every particle must sediment from the lower to the upper wall
of a pair of disks. Thus, a suitable relationship must be obtained between 0, the flow
Feed

Liquid
discharge

Conical disk
channels

Light
liquid

Heavier
sediment

Bowl

Disk channels
entrance
FIGURE 5.5 Cross-sectional diagram of a disk centrifuge, showing the path of the liquid flow and
the collection of solids at the periphery.
Sedimentation // 201

θ
x
y νω

R1 l
νω
R
ν0
R0

FIGURE  5.6 Diagram of the zone between two disks and the definition of variables for a disk
centrifuge.

velocity, and , the sedimentation velocity of the particles. The following assump-
tions simplify the analysis:

1. Typically, 0, the flow velocity, is much greater than .

2. v0 = Q/A, where A decreases as particles move toward the center.
3. The fluid velocity 0 is a function of y and goes to zero at the surface of the
disks.

The equation of motion in the x direction is

dx
(5.4.8)  = υ0 − υω sin θ
dt

This is simplified by the assumption that 0 is much greater than sin . The average
value of 0, denoted < 0>, is given by the flow rate Q divided by the cross-sectional
area A perpendicular to the flow for n disks:

(5.4.9)  υ = Q = Q
0
A n( 2 π R l )

where the R is the radial distance from the center of rotation, which is varying, and
l is the spacing between disks, which is fixed. The local value of 0 can be found by
multiplying < 0> by a function f (y) that gives the velocity variation between the
disks:

(5.4.10) υ = Q
0 f ( y)
n(2 π R l )
202 // BIOS E PA R AT IONS SC IENCE AND E N GINEE RIN G

Integrating 0 across the space between the disks gives the average value of 0:

l
∫ υ0 ( y)dy =
(5.4.11)  0
Q
l n (2π Rl )

From Equations (5.4.10) and (5.4.11), it is easily deduced that

l
∫ f ( y)dy = 1
(5.4.12)  0
l

From Equations (5.4.8) and (5.4.10), neglecting sin in comparison to 0, we

have

dx Q
(5.4.13)  = f ( y)
dt n (2π Rl )

The equation of motion in the y direction is centrifugal particle motion:

dy 2 a 2 (ρ − ρ0 )ω 2 R
(5.4.14)  = υω cos θ = cos θ
dt 9µ

where the centrifugal velocity is obtained from Equation (5.2.2). The slope of the
trajectory of particles moving between any pair of disks is determined by combining
the equations of motion in the x and y directions:

dy
dy 4 nπ a 2 (ρ − ρ0 )ω 2 R 2 l cos θ
(5.4.15)  = dt =
dx dx 9µQf ( y)
dt
Since dx = −dR/sin , Equation (5.4.15) can be rearranged to give

4 nπ a 2 (ρ − ρ0 )ω 2 R 2 cot θ
(5.4.16) Q f ( y)dy = − dR
l 9µ

For the boundary conditions of integration, we focus on the particles that are the
most difficult to capture: such a particle entering at y = 0 and R = R0 would exit at
y = l and R = R1 (see Figure 5.6). After integration of Equation (5.4.16) and using
the result of Equation (5.4.12), the result can be rearranged after multiplying and
dividing by g to yield

⎪⎧ 2 a (ρ − ρ0 )g ⎪⎫ ⎡ 2 nπ ω ( R0 − R1 )cot θ ⎤
2 2 3 3
(5.4.17) Q = ⎨ ⎬⎢ ⎥ = {υ g }[ ∑ ]
⎪⎩ 9µ ⎪⎭ ⎢⎣ 3g ⎥⎦
Sedimentation // 203

Therefore, in a sensitivity analysis it is seen that the factor depends on the cube
of the bowl radius, the cotangent of the disk acute angle, the number of disks in the
stack, and, as in the tubular centrifuge, the square of the rotor speed. The disk acute
angle made by the conical disks is typically between 35˚ and 50˚ [13].
There are three types of disk centrifuges, depending on the mode of discharg-
ing the solids. In the batch mode, solids accumulate at the periphery of the bowl
until the bowl is nearly full, which is evidenced by turbidity of the liquid flowing
out of the centrifuge. At this point, the centrifuge needs to be shut off and the sol-
ids removed manually. The solids content of the feed needs to be low (0 to 1 vol.
%) in order that the downtime for solids removal does not become excessive. In
the semicontinuous mode, the bowl is opened intermittently to allow solids to dis-
charge through ports at the periphery. The solids content of the feed can be in the
range of 0 to 10 vol. % for the intermittent discharge, and the solids exiting can
be nonflowable. For continuous operation, nozzles are placed at the periphery of
the centrifuge, and the solids content of the feed is in the range of 6 to 25% vol.
%. The solids exiting from the nozzles are flowable. Nozzles are distributed around
the periphery of the centrifuge and typically range from 12 to 24, depend on the
size of the centrifuge [13].
The time between discharges in the semi-continuous mode of operation of a disk
centrifuge with volumetric feed rate Q can be estimated from [13]:

Vsϕ e
(5.4.18)  t d =
Qϕ f

where Vs is the solids holdup volume of the bowl (which typically is 40 to 50% of
the entire bowl volume) and e and f are the volume fractions of solids in the exiting
sediment and feed, respectively.

5.5 ULTRACENTRIFUGATION
Ultracentrifuges operate over a range of inertial accelerations of 50,000 to
100,000 × g. These accelerations are so great that it is possible to sediment very
small particles and even macromolecules in solution by an ultracentrifuge. In ana-
lytical ultracentrifugation [14], a sample volume of less than 1.0 ml is centrifuged
in an optical cell while the concentration is monitored optically as a function of
distance from the center of rotation. In preparative centrifugation, samples of up
to 50 ml are centrifuged in a batch operation and collected, usually as a function of
final distance from the center of rotation. This collection is usually accomplished
by carefully puncturing the bottom of the centrifuge tube and collecting the out-
flow in a series of tubes.
204 // BIOS E PA R AT IONS SC IENCE AND E N GINEE RIN G

5.5.1 Determination of Molecular Weight

One of the most important uses of ultracentrifugation has been the determination of
molecular weights by the combined measurement of sedimentation and diffusion coef-
ficients—the “sedimentation-diffusion molecular weight.” Proteins are the most com-
mon macromolecular compounds that are homogeneous with respect to molecular
weight, and it is for these compounds that the sedimentation-diffusion method has
been mainly used. The molecular weights of proteins, as well as for viruses, obtained
by this method have been found to be in excellent agreement with those obtained by
other methods [15]. Today, protein and nucleic acid molecular weights are commonly
determined by mass spectroscopy, electrophoresis, or chromatography (see Chapter 2).
The development of the dependence of molecular weight on the sedimentation
and diffusion coefficients starts with the equation of motion for a sedimenting mol-
ecule at steady state [Equation (5.2.1)], which can be written as follows:

(5.5.1)  V (ρ − ρ0 )ω R − 6πµ aυ = 0
2

where V is the volume of a single molecule. If we let V = mV = m ρ, where V is the

molecule’s specific volume and m is the mass of a single molecule, then Equation
(5.5.1) can be rearranged to give

6π µ aυ
(5.5.2) m =
(1 − V ρ0 )ω 2 R

Viscosity can be related to the diffusion coefficient by the Stokes-Einstein equation:

µ 1
(5.5.3) D =
kT 6 π a

where k is Boltzmann’s constant. Substituting the Stokes-Einstein equation into

Equation (5.5.2) and making use of the definition of the sedimentation coefficient s
[Equation (5.3.2)] and of Boltzmann’s constant (= R/N, where R is the universal gas
constant and N is Avogadro’s number) leads to
sRT
(5.5.4) M =
D (1 − V ρ0 )
where M is the molecular weight and here R is the gas constant. The partial specific
volume V can be easily determined as the slope of a plot of the specific volume of
the solution versus the weight fraction of the substance (see, e.g., the determination
of M of chymotrypsinogen by Schwert [16]). For accurate determinations of M using
this method, it is necessary that s and D be extrapolated to zero concentration. The
diffusion coefficient of the macromolecule can be measured from concentration pro-
files at equilibrium during sedimentation with a density gradient, or more commonly,
in a separate procedure such as by the free-diffusion method [15].
Sedimentation // 205

5.6 FLOCCULATION AND SEDIMENTATION

After cells have been lysed or bioparticles have been dispersed, it is often useful to
hasten the subsequent sedimentation or filtration step by reversibly increasing the
size of the particles to be separated. To this end, flocculation is used, and it occurs as
the result of adding a suitable chemical called a “flocculant” or by the selection of
naturally flocculating cells for fermentation, as in the case of lager yeast. Flocculants
can act by forming interparticle molecular “bridges” between particles, in which
case the flocculants are usually polymers or oligomers; they can also act by reducing
the repulsive forces between cells, usually by reducing the strength of the electro-
static field (for details see Chapter 3, Section 3.4).
Centrifugation is frequently used after the addition of a flocculant to uncleared
broth or cell lysates. While the resulting aggregates are often treated as Stokes spheres,
they are actually open structures in which internal convection may occur. Various
theories of sedimentation of flocs have therefore been proposed, in view of the impor-
tance of this motion in the water-processing (sewage treatment) industry [17].
Chapter 3 introduces the subject of flocculation but not the collection of the
flocs that are produced. In process engineering, the sedimentation of flocs is
usually treated empirically. Nevertheless, a number of formal treatments of this
problem have been carried out. In the standard Equation (5.2.2) for the sedimen-
tation of spheres, empirical adjustments for sedimentation velocity of flocs can
be made on the basis of the reduction of density by the void volume fraction
of the floc and the reduction of the drag force by the drag reduction factor Ω,
which is also a function of the void volume fraction and the floc radius a. The
resulting Stokes sedimentation velocity of spherical flocs composed of particles
of density is then

2 a 2 (1 − ε )(ρ − ρ0 )ω 2 R
(5.6.1)  υ =
9µΩ(ε , a )

To estimate the drag force reduction due to liquid flow through the void volume
of the floc, a dimensionless, normalized diameter β ≡ a / k f is defined, in which
1/ 2

kf is the permeability of the floc for the fluid in which it is settling. The drag force
reduction factor has been related to the void volume fraction and the floc radius by a
variety of closely related functions, the most widely used of which seems to be [18]:

⎛ tanh β ⎞
2β 2 ⎜ 1 −
⎝ β ⎟⎠
(5.6.2) Ω(ε, a ) =
⎛ tanh β ⎞
2β 2 + 3 ⎜ 1 −
⎝ β ⎟⎠
206 // BIOS E PA R AT IONS SC IENCE AND E N GINEE RIN G

The determination of requires knowing the floc permeability kf, which can be cal-
culated by using the following form of the Carman-Kozeny equation:

ε2
(5.6.3) k f =
KS 2 (1 − ε )2

where Kozeny’s constant K  =  4.8 [19], and S is Carman’s specific surface area,
defined as the area of nonporous materials exposed to the liquid per unit volume,
as determined from the geometry of the individual particles that form the aggregate.
In addition to the collective behavior of particles in flocs, particles do not behave
totally independently when they are suspended at high concentration, such as at the
outer rim of a centrifuge or the bottom of a settling tank.

5.7 SEDIMENTATION AT LOW ACCELERATIONS

At low g forces, the sedimentation rate slows down and can sometimes be similar to
the rate of transport by diffusion, which is not affected by inertial forces. Here, we
examine ways to compare sedimentation and diffusion rates. The case of isothermal
settling is evaluated, which can lead to exponential distributions of concentration as
a function of height for small particles. Inclined sedimentation and field-flow frac-
tionation, two operations that separate suspended particles on the basis of inertial
motion at 1 × g, are also described.

5.7.1 Diffusion, Brownian Motion

Einstein described diffusion as the consequence of a “random walk” by particles
due to their thermal energy kT (k = Boltzmann’s constant). The surprisingly simple
result was

2
(5.7.1)  x = 2D t

where <x2> is the mean square distance traveled by a particle having diffusion coef-
ficient D in time t. Using the Stokes-Einstein equation [Equation (5.5.3)] for spheri-
cal particles of radius a undergoing Brownian movement in a fluid of viscosity µ, we
can relate the diffusion coefficient to the thermal energy kT as follows:

kT
(5.7.2) D =
6π µ a

In a concentration gradient, the net unidirectional flux of particles is proportional to

D and the gradient, dc/dx, of the particle concentration c. Diffusion is not affected
by gravity. Diffusion and sedimentation velocities, however, are sometimes similar,
and their sum results in gradual settling.
Sedimentation // 207

5.7.2 Isothermal Settling

If the temperature T does not change over the height h of an ensemble of particles,
then the mean kinetic energy, which is proportional to k T, of all particles is the same
at all heights. The potential energy of a particle of mass m is usually expressed as
mgh; but if the particles are subject to buoyant forces in the fluid, the potential energy
becomes V ( − 0)gh for particle volume V. From the Boltzmann distribution rule,
the concentration of particles at height h at equilibrium is therefore

⎡ −V (ρ − ρ0 )gh ⎤
(5.7.3) c(h) = c(0)exp ⎢ ⎥
⎣ kT ⎦

This means that concentration is an exponential function of height under isothermal

conditions and that large, dense particles with potential energy much greater than k T
(from mammalian cells to marbles) will be concentrated at h = 0 and that small par-
ticles (mainly molecules) will have c(h) ≈ constant. However, submicrometer organic
particles and certain macromolecules have values of V and that lead to measurable
exponential distributions of c(h).

5.7.3 Convective Motion and Péclet Analysis

One way to determine whether c(h) will be distributed as in Equation (5.7.3) is to
estimate the value of the Péclet number, Pe, commonly used in the dimensionless
analysis of fluid motion relative to diffusive transport. The Péclet number is a ratio
of the sedimentation velocity to the characteristic rate of diffusive transport over
distance L:
υ
Pe =
(5.7.4)  D L

If Pe < 0.1, diffusion is dominant; and c(h) is approximately constant or distributed.

In this case, Equation (5.7.3) should be used. If Pe > 10, sedimentation dominates,
and the concentration will eventually be high at h = 0 and 0 elsewhere.

5.7.4 Inclined Sedimentation

Rapid removal of high-density solids can be achieved at 1  × g by using inclined
sedimentation. An inclined settler is shown in Figure 5.7; the dimensions and the
flow rates are labeled. In a typical application, feed containing suspended particles is
pumped into the settler near or at its lower end at flow rate Qf , particle-free overflow
exits the upper end at flow rate Qo , and particle-rich suspension leaves in the under-
flow at rate Qu. This is an excellent method for harvesting supernatants continuously
or batchwise from particle (cell)-laden broth. The material balance relationships are

(5.7.5)  Q f = Qu + Qo
208 // BIOS E PA R AT IONS SC IENCE AND E N GINEE RIN G

(5.7.6)  c f Q f = cuQu + coQo

where the c’s designate particle concentrations. It is often desirable that co = 0. To
achieve this, Qo must equal the volumetric clearing rate S( ). This volumetric clear-
ing rate S is equal to the vertical settling velocity g of the cells multiplied by the
horizontal projected area of the upward-facing surfaces of the channel onto which
the cells may settle, which is given by the following equation [20]:

(5.7.7) S = υ g w ( L sin θ + b cos θ)

where is the angle of inclination of the plates from the vertical, and b, w, and L
are the height, width, and length of the rectangular settler, respectively (Figure 5.7).
Inclined settlers are designed so that the path to the completion of sedimenta-
tion of a particle is extremely short, only a few millimeters, before the sedimented
particles begin to be convected toward the underflow. If the particulate fraction is
desired, it can be batch concentrated by continuous recycle of the underflow back
to the tank while the overflow bleeds off the supernatant. By the use of appropriate
settings, governed and predicted by the foregoing equations, it is also possible to
remove small particles in the overflow while retaining larger ones in the underflow,
thereby effecting a binary particle classification by size [20].
An important application of inclined settlers is in the removal of unproductive or
parasitic cells from bioreactors or cell culture systems; this method was used to remove
nonviable hybridoma cells from a cell culture, which resulted in high viable cell concen-
trations and high monoclonal antibody productivity over a 2-week culture period [20].

Pump
b
Overflow, Qo

Sampling
θ L
Sampling Inclined settler

Feed
Qf Underflow,Qu

FIGURE 5.7 Diagram of an inclined settler system indicating variables used in the mass balance
[Equation (5.7.5)] and volumetric clearing rate [Equation (5.7.7)].
Sedimentation // 209

Since inclined settlers can be scaled up directly by increasing the area for set-
tling, they potentially could be used at larger than laboratory scale. These settlers
operate at much higher capacities than vertical settlers because cells need to settle
only a distance of order b (see Figure 5.7) in an inclined settler, compared with a
distance of order L in a vertical settler.

5.7.5 Field-Flow Fractionation

Field-flow fractionation (FFF) is designed to separate particles of different sizes on
the basis of the hydrodynamics of a very thin, flat, horizontal channel through which
a sample suspension is pumped and subjected to the laminar flow velocity gradient
in the channel as shown in Figure 5.8. In the case of sedimentation, the driving force
toward the lower channel wall is gravity or centrifugal sedimentation. In the latter
case, the channel is “wrapped” around the perimeter of a rotating centrifuge. Steep
velocity gradients occur at the upper and lower walls of the channel, and these result
in a distribution of larger particles toward the center and smaller particles toward the
lower wall. The separation of particles by this phenomenon is called “steric” FFF. As
particle transport proceeds along the lower wall, the particles bunch up according to
their velocity; thus the lower wall is called an “accumulation wall.” For continuous
operation, a horizontal splitter at the outlet permits the collection of large particles
in an upper outlet and small particles in a lower outlet. The governing equation for
field-flow fractionation is

(5.7.8) R = 6 γ a
b

where R is the retention ratio, defined as the ratio of the particle velocity to the mean
fluid velocity, a is the particle radius, b is the channel height, and is the “steric fac-
tor” that determines the particle net velocity and is approximately equal to (1 + /a),
where is the distance between the particle and the accumulation wall, as noted in

Sample injection
(inflow)
Field
Fraction collection
(outflow) Field
Flow vectors

FFF Parabolic flow profile

Acc cha
umu nnel Flow
3
latio 2
nw 1
all

δ1 δ2 δ3
(a) (b)

FIGURE 5.8 (a) Exploded view of a field-flow fractionation (FFF) channel. The “field” is the driving
force for separation, expected to act differentially on particles of different types. The field could
be gravitational, electrical, thermal, adsorptive, or steric, to name a few. (b) The principle of steric
FFF. Larger particles protrude into the higher-velocity region of laminar flow, hence are carried
farther in a specific amount of time than their smaller counterparts.
210 // BIOS E PA R AT IONS SC IENCE AND E N GINEE RIN G

Figure 5.8. This technique has been shown to separate two cell-sized latex popula-
tions having sphere diameters of 10 and 15 µm, and also to separate white and red
blood cells [21].

5.8 CENTRIFUGAL ELUTRIATION

Centrifugal elutriation is similar to inclined sedimentation and field-flow fraction-
ation in that sedimentation takes place in the presence of fluid flow. In centrifu-
gal elutriation, also called counterstreaming centrifugation, the effective length of a
sedimentation path is greatly extended by continuously pumping a counterstreaming
fluid in the opposite direction to that of sedimentation. To accomplish this, “elutria-
tion rotors” have been designed. Such rotors perform functions similar to those of
the tubular bowl and disk-stack centrifuges. The sample suspension is held in the
rotor chamber, and the particles remain there as long as the two opposing forces are
in balance. By incremental increases in the flow rate of the fluid or by decreases in
the centrifugal force, distinct populations of particles (including cells) with relatively
homogeneous sizes can be eluted out of the rotor sequentially. The objective of elu-
triation is thus to collect particles of a specified volume by modifying the velocity of
the eluent fluid or the angular velocity of the rotor [22]. Both these variables must be
controlled with extreme care if precision of separation is to be achieved. The volume
and radius of the largest particle that can escape against the counterflow and be col-
lected in the effluent can be determined from the equation of motion for a spherical
particle in an inertial field [Equation (5.2.1)], using the velocity v0 for the eluting
fluid [22]:
3/ 2
⎡ υ0 µ ⎤
9π 2 ⎢ ⎥
⎣ R ( ρ − ρ0 ) ⎦
(5.8.1)  V =
ω3
1/ 2
⎡ υ0 µ ⎤
3⎢ ⎥
⎣ 2 R ( ρ − ρ0 ) ⎦
(5.8.2)  a =
ω
where V and a are the volume and radius of the particle, respectively. The capacity
of centrifugal elutriation in the batch mode is about 108 particles of about 10 µm
diameter.

5.9 SUMMARY
Sedimentation is the movement of particles or macromolecules in an inertial
field. Inertial accelerations vary from 1 × g in flocculation tanks to 100,000 × g in
ultracentrifuges.
Sedimentation // 211

r Sedimentation, like filtration, is used in early stages in downstream bioprocess-

ing mainly for liquid–solid separations. Particles can be separated at large scale
in continuous centrifuges, and macromolecules can be separated, either for
analysis or collection, at small scale by ultracentrifugation at very high speeds.
r The sedimentation velocity of a particle depends on the square of its radius a
and linearly on the difference between its density and that of the suspending
solvent 0:
2 a 2 (ρ − ρ0 )ω 2 R
υ=
9µ

where is angular velocity (rad/s), R is the distance of the particle from the
center of rotation, and µ is fluid viscosity. The term 2 R is the centrifugal
acceleration.
r Corrections related to sedimentation calculations are required when particle
concentrations are high enough to hinder settling. Functions of concentration
are available for this correction.
r The sedimentation coefficient s, a property of both the particle and the medium,
is defined as

υ
s≡
ω2R

which leads to
2 a 2 ( ρ − ρ0 )
s=
9µ

r Equivalent time is the product Gt, where G is defined as

ω2R
G≡
g

One approach to scale-up of centrifugation is to assume constant equivalent

time.
r Engineering analyses and scaling calculations on large-scale centrifuges are
often performed by means of “sigma analysis,” which uses the operation con-
stant to characterize a centrifuge into which feed flows at volumetric flow rate
Q. Since the engineer often needs to estimate Q, a convenient relationship is

Q = {υ g }[ ∑ ]
212 // BIOS E PA R AT IONS SC IENCE AND E N GINEE RIN G

where g is the sedimentation velocity at 1 × g,

2 a 2 ( ρ − ρ0 ) g
υg =
9µ

and represents the geometry and speed of the centrifuge.

r The appropriate centrifuge for a particular liquid-solid separation is chosen
on the basis of the objectives of the operation, the solids content and flow rate
of the feed stream, and the maximum acceleration at which the device can
operate.
r The molecular weight of a macromolecule such as a protein can be determined
from ultracentrifuge data, which measure the sedimentation and diffusion
coefficients.
r After flocculation, loosely structured, porous particles form, and these sediment
more rapidly than their solid counterparts having the same size. Correction fac-
tors allow the calculation of the velocity of flocculated particles.
r Diffusion occurs in sedimentation operations, and its significance, specifically
at low accelerations, can be ascertained by calculating the Péclet number.
r Two useful sedimentation operations at 1 × g are inclined sedimentation and
field-flow fractionation.
r In centrifugal elutriation, a counterstreaming fluid is continuously pumped in
the opposite direction to that of sedimentation. This greatly extends the sedi-
mentation path and allows distinct populations of particles (including cells)
with relatively homogeneous size to be eluted out of the centrifuge rotor
sequentially.

NOMENCLATURE
a radius of particle (μm)
A cross-sectional area (cm2)
b height (cm)
c concentration (M, or g liter−1)
D diffusion coefficient (cm2 s−1)
g gravitational acceleration (9.8066 m s−2)
G multiple of gravitational acceleration (= 2R/g) (dimensionless)
h height (cm)
Fx force in x direction (N)
k Boltzmann’s constant (1.3807 × 10−23 J K−1)
kf floc permeability (cm2)
K Kozeny’s constant (= 4.8) (dimensionless)
m mass of particle or molecule (g)
M molecular weight (Daltons)
Sedimentation // 213

n exponent in Equation (5.2.7) (dimensionless)

n number of disks in a disk centrifige (dimensionless)
N Avogadro’s number (6.0221 × 1023 molecules mol−1)
l spacing between disks in a centrifuge (cm)
L distance (cm)
L length of an inclined settler (cm)
Pe Péclet number (= vL /D ) (dimensionless)
Q flow rate (cm3 s−1)
R distance from center of rotation (cm)
R retention ratio in field-flow fractionation (= ratio of particle velocity to the
mean fluid velocity) (dimensionless)
R gas law constant (8.3145 J mol−1 K−1)
Re Reynolds number (2av /μ) (dimensionless)
s sedimentation coefficient [Equation (5.3.2)] (s)
S volumetric clearing rate in an inclined settler (cm3 s−1)
t time (s)
td time between discharges of a solids-ejecting disk centrifuge (s)
T temperature (K)
sedimentation velocity (cm s−1)
c sedimentation velocity in a concentrated suspension (cm s−1)
g sedimentation velocity at 1 × g (cm s−1)
sedimentation velocity at angular velocity (cm s−1)
0 fluid velocity (cm s−1)
V volume of a molecule or particle (cm3)
V0 volume of cylinder in a gradient mixer [Equation (5.3.1)] (cm3)
Vs volume of solid holdup in the bowl of a disk centrifuge (cm3)
V specific volume of a molecule (cm3 g−1)
w width of an inclined settler (cm)
Greek Letters
1/ 2
normalized diameter of a floc (= a / k f ) (dimensionless)
steric factor in field-flow fractionation (≅ 1 + δ / a ) (dimensionless)
distance between particle and accumulation wall in field-flow fractionation (cm)
void fraction of particles in a floc (dimensionless)
centrifuge disk acute angle (Figure 5.6) (degrees)
angle of inclination of an inclined settler from vertical (degrees)
μ viscosity of fluid (g cm−1 s−1 = poise)
density of particle or molecule (g cm−3)
0 density of medium (g cm−3)
operation constant of a centrifuge (cm2)
residence time of fluid in the centrifuge bowl (s)
volume fraction of particles (dimensionless)
214 // BIOS E PA R AT IONS SC IENCE AND E N GINEE RIN G

e volume fraction of solids in the exiting sediment (dimensionless)

f volume fraction of solids in the feed (dimensionless)
angular velocity (rad s−1)
drag reduction factor [Equation (5.6.2)] (dimensionless)

PROBLEMS

5.1 Sedimentation versus Filtration Four particulate materials, A, B, C, and D, are sus-
pended in water (density = 1.00 g/cm3) and have the properties given in Table P5.1.
Choose between sedimentation and filtration for the separation of the following pairs
of particles from one another in mixed suspensions in water. Explain your answer in
each case.
(a) A from B
(b) B from C
(c) C from D

TABLE P5.1

Particle name Density (g/cm3) Radius (µ m)

A 1.05 10
B 1.05 12
C 1.01 50
D 1.04 25

5.2 Strategies for Product Separation Yeast cells (a = 3 µm) in a fermentor secrete a
low molecular weight product at a concentration that produces uniform rod-shaped
crystals 2  × 6 µm at about 20 times the number concentration (particles/ml) as the
cells. Using concise statements, design two possible strategies that take advantage of
the particulate nature of the product to separate the product from the broth and from the
cells. What additional information about the product crystals would be useful?
5.3 Isopycnic Sedimentation You wish to capture 3 µm particles in a linear density gra-
dient having a density of 1.12 g/cm3 at the bottom and 1.00 at the top. You layer a thin
particle suspension on the top of the 6 cm column of fluid with a viscosity of 1.0 cp and
allow particles to settle at 1 g.
(a) How long must you wait for the particles you want (density = 1.07 g/cm3) to sedi-
ment to within 0.1 cm of their isopycnic level? Is it possible to determine the time
required for particles to sediment to exactly their isopycnic level?
(b) If instead of 1 g you use a centrifuge running at 800 rpm, and the top of the fluid
is 5 cm from the center of rotation, how long must you centrifuge for the particles
to move to within 0.1 cm of their isopycnic level?
5.4 Time Required for Sedimentation by Gravity A certain reagent is added to a sus-
pension of cells 4 µm in diameter. These cells have a density of 1.08 g/cm3, and they are
suspended in liquid with a density of 1.00 g/cm3 and viscosity of 1.0 cp. This reagent
causes about half of the cells to form fairly solid aggregates, all of which are 90 µm
Sedimentation // 215

in diameter and have density midway between that of the liquid and the cells. How
much time is required for all the aggregates to sediment to within 1 cm of the bottom
of a vessel filled with suspension that is 0.5 m high? Approximately what fraction of
the single cells would have sedimented to this depth in the same amount of time? How
much time is required for all the single cells to sediment to within 1 cm of the bottom
of the vessel?
5.5 Time Required for Sedimentation in a Centrifuge Using the results of Problem
5.4, determine the diameter and speed of a centrifuge required to reduce the total sedi-
mentation time for the aggregates by a factor of ten, assuming you will use containers
that are 20 cm high in the centrifuge. Also assume that the center of rotation is 3 cm
from the tops of the containers. How much time must the same centrifuge be operated
to also sediment all the single cells? For simplicity, assume a swinging-bucket type of
centrifuge, in which the axis of the cylindrical vessel is horizontal, hence parallel to the
direction of sedimentation.
5.6 Determination of Sigma for a New Pilot Scale Centrifuge You test a new pilot
scale centrifuge by doing a breakthrough experiment using yeast as test particles. The
yeast were previously found to sediment at 100 µm/s in a laboratory centrifuge oper-
ated at an acceleration of 500 × g. The breakthrough flow rate is found to be 10 liters/
min. What is the sigma ( ) of this new centrifuge?
5.7 Bench Scale Tests for a Tubular Bowl Centrifuge You can bench-test a tubu-
lar bowl separation by first characterizing the product in a test-tube centrifugation.
Without actually knowing the size and density of the particles in the suspension, derive
an expression for the angular velocity required to capture the solids at a given volu-
metric flow rate Q in terms of the geometry of the tubular bowl and the quantities you
would measure in the test tube centrifugation.
5.8 Recovery of E. coli in a Tubular Bowl Centrifuge You are using a tubular bowl
centrifuge to recover E. coli cells containing an important bioproduct from a fermenta-
tion broth. In a preliminary run you find that 50% of the cells are recovered at a flow
rate of 5 liters/min and rotation speed of 6000 rpm.
(a) To increase the yield to 95% using the same centrifuge, what must the flow rate be?
(b) How much does the sedimentation velocity change if you double the rotation
speed?
5.9 Scale-up of a Disk Centrifuge Based on Laboratory Data Determine the maxi-
mum flow rate for the clarification of a suspension of lysed Escherichia coli cells by
a plant scale disk centrifuge based on laboratory data. The plant centrifuge has a bowl
diameter of 25.4 cm and capabilities shown in Table 5.5. For this centrifuge, you also
know that  = 42°, R1 = 8 cm, R0 = 20 cm, and number of disks = 100 (see Figure 5.6).
In a laboratory centrifuge, you determined that it took a minimum of 17 min to
clarify the cell lysate at 12,000 rpm. The top of the culture being centrifuged was
32 mm from the center of rotation, and the top of the packed solids was 79 mm from
the center of rotation after 17 min.
5.10 Determination of Molecular Weight by Ultracentrifugation A  new biopharma-
ceutical “X” has been discovered. Only crude extracts are available, and the material
is known to be a macromolecule. You are given a preparative ultracentrifuge and asked
to estimate the molecular weight of the macromolecule. You then do two experiments.
216 // BIOS E PA R AT IONS SC IENCE AND E N GINEE RIN G

In the first one, you set up a linear sucrose density gradient and layer a crude sample of
X on top of it, using a 5 cm long centrifuge tube (completely filled) and centrifuge to
equilibrium for 3 days (72 h) at 25,000 rpm. In the second experiment, you place the
sample in its dilute buffer (viscosity the same as water) directly into two of the same
plastic ultracentrifuge tubes and run the centrifuge for 24 h at 25,000 rpm, at which
time you stop and remove fractions from one of the two tubes. After an additional 72 h
at 25,000 rpm, you stop the centrifuge and remove fractions from the remaining tube.
In all three cases, you collect 20 fractions, and each fraction corresponds to a 2.5 mm
layer, so fraction 1 came from the bottom of the tube and fraction 20 came from the top.
Assume that the top of each centrifuge tube is 3 cm from the center of rotation.
The company biologist then takes the three sets of 20 fractions and tests them on cell
cultures. The biologist returns data to you in terms of biological activity units in each
fraction, on a scale that is known to be linear with product mass. You then plot the data
from the three centrifuge tubes (Figure P5.10).
Units of biological activity per fraction

Units of biological activity per fraction

40 1.2 40 24 + 72 h
Density (g/cm3)

24 h
1.1

0 1.0 0
0 4 8 12 16 20 0 4 8 12 16 20
Fraction number Fraction number
(a) (b)

FIGURE P5.10 Ultracentrifugation results. (a) First experiment: crude extract layered on the top of
a tube with a linear sucrose density gradient and centrifuged to equilibrium for 72 h. (b) Second
experiment: crude extract in its dilute buffer in each of two of the same plastic tubes with one tube
centrifuged for 24 h and the other tube for 72 h. Centrifuge speed 25,000 rpm. Fractions, each cor-
responding to a 2.5 mm layer in the tube, are numbered from the bottom to the top of each tube.
(a) From the appropriate equations for ultracentrifugation, use the data to estimate
the molecular weight of product X.  (Hint: Use widths of profiles to estimate
diffusivity.)
(b) Make a sketch of the method for collecting fractions.
(c) After you completed your material balance calculations on the tubes, which were
8 mm in diameter, the biologist told you that each tube originally contained 20 units
of biological activity. Speculate about why the material balance did not close.
5.11 Isothermal Settling Based on the data in Table 5.1, estimate the reduced con-
centration profile, c(h)/c(0), for the isothermal settling at 1 g of a protein with a
sedimentation radius of 0.005 µm at ambient temperature up to a height of 100 cm.
Recalculate the profile for a protein with a sedimentation radius of 0.002 µm. Also
calculate the molecular weight of each protein. Explain the meaning of the concen-
tration profiles.
Sedimentation // 217

5.12 Cost of Centrifugation In a bioprocess for the production of foreveron (a hypotheti-

cal youth-giving protein made by Kindergen, Inc.), a centrifuge is used to remove the
cells from the culture broth. The centrifuge operates continuously at 20 kW and pro-
cesses 20 liters of feed culture broth per hour, discharging a liquid supernatant phase
containing 92 vol% of the feed liquid and no cells. The cell paste output has a density
of 1.08 g/cm3. The feed is processed batchwise in 40 liters of feed per batch. It takes a
busy technician ($20/h) 15 min to start the feed flow and 15 min to collect each batch
and deliver the supernatant and sediment to the next steps. Kindergen paid $50,000 for
the centrifuge, which has a useful life of 10 years, buys a service contract for $5000/
year, and replaces the rotor every year for $10,000. The centrifuge is used only for
foreveron processing at two batches per day, 300 days a year.
Calculate the cost of the centrifugal processing per kilogram of cell paste produced.
What contributes the most to the cost of centrifuging?
5.13 Estimation of Flow Rate for Centrifugation of Yeast Cells A maximum flow rate
of 50 liters/min was achieved for the centrifugation of bacteria cells in a tubular centri-
fuge. The cells were 2.0 μm in diameter and had a density of 1.08 g/cm3. The medium
had a density of 1.01 g/cm3 and viscosity of 1.2 cp.
It is desired to centrifuge yeast cells in this same centrifuge. The yeast cells have a
diameter of 5.0 μm and a density of 1.10 g/cm3. The medium has a density of 1.02
g/cm3 and a viscosity of 1.3 cp. Estimate the maximum flow rate that can be used to
centrifuge the yeast cells.
5.14 Determination of Sigma from Laboratory Data In a laboratory centrifuge, it was
found that it took 300 s to centrifuge cells at 2000 rpm to a constant height of packed
cells. The top of the cell suspension and the top of the packed cells were 5 cm and 9 cm,
respectively, from the center of rotation of the centrifuge. For processing these cells in
a tubular bowl centrifuge at a flow rate of 20 liters/min, what should the value of be?

REFERENCES
1. Bird, R. B., Stewart, W. E., and Lightfoot, E. N. (2002). Transport Phenomena, 2nd ed., Wiley,
New York.
2. Kubitschek, H. E. (1987). Buoyant density variation during the cell cycle in microorganisms.
Crit. Rev. Microbiol., vol. 14, p. 73.
3. Hart, A., and Edwards, C. (1987). Buoyant density fluctuations during the cell cycle of Bacillus
subtilis. Arch. Microbiol., vol. 147, p. 68.
4. Illmer, P. (2001). How big is a bacterium? Approximation to a correct total cell density of
Arthrobacter sp. FEMS Microbiol. Lett., vol. 196, p. 85.
5. Sheeler, P. (1981). Centrifugation in Biology and Medicine, Wiley-Interscience, New York.
6. Richardson, J. F., and Zaki, W.  N. (1954). Sedimentation and fluidisation:  Part I. Trans. Inst.
Chem. Eng., vol. 32, p. 36.
7. Richardson, J. F., and Shabi, F. A. (1960). The determination of concentration distribution in a
sedimenting suspension using radioactive solids. Trans. Inst. Chem. Eng., vol. 38, p. 33.
8. Meselson, M., and Stahl, F. W. (1957). The replication of DNA in E. coli. Proc. Natl. Acad. Sci.
USA, vol. 44, p. 671.
9. Ladisch, M. R. (2001). Bioseparations Engineering, Wiley, New York.
10. Belter, P. A., Cussler, E. L., and Hu, W.-S. (1988). Bioseparations, Wiley, New York.
218 // BIOS E PA R AT IONS SC IENCE AND E N GINEE RIN G

11. Bell, D. J., Hoare, M., and Dunnill, P. (1983). The formation of protein precipitates and their cen-
trifugal recovery. In Advances in Biochemical Engineering/Biotechnology, vol. 26, A. Fiechter,
ed., Springer-Verlag, Berlin, p. 1.
12. Perry, R. H., Green, D. W., and Maloney, J. D., eds. (1997). Liquid-solids operations and equip-
ment (Section 18). Perry’s Chemical Engineers’Handbook, McGraw-Hill, New York.
13. Leung, W. W. (2007). Centrifugal Separations in Biotechnology, Academic Press, Amsterdam.
14. Schachman, H. K. (1959). Ultracentrifucation in Biochemistry, Academic Press, New York.
15. Tanford, C. (1961). Physical Chemistry of Macromolecules, Wiley, New York.
16. Schwert, G. W. (1951). The molecular size and shape of the pancreatic proteases. II. Chymotry-
psinogen. J. Biol. Chem., vol. 190, p. 799.
17. Li, D., and Ganczarczyk, J. (1992). Advective transport in activated sludge flocs. Water Environ.
Res., vol. 64, p. 236.
18. Neale, G., Epstein, N., and Nader, W. (1973). Creeping flow relative to permeable spheres.
Chem. Eng. Sci., vol. 28, p. 1865.
19. Adler, P. M. (1986). Transport processes in fractals. VI. Stokes flow through Sierpinski carpets.
Phys. Fluids, vol. 29, p. 15.
20. Davis, R. H., Lee, C.-Y., Batt, B. C., and Kompala, D. S. (1991). Cell separations using differen-
tial sedimentation in inclined settlers. In Cell Separation Science and Technology, D. S. Kompala
and P. Todd, eds., ACS Symposium Series 464, American Chemical Society, Washington, DC,
p. 113.
21. Giddings, J. C., Barman, B. N., and Liu, M.-K. (1991). Separation of cells by field-flow frac-
tionation. In Cell Separation Science and Technology, D. S. Kompala and P. Todd, eds., ACS
Symposium Series 464, American Chemical Society, Washington, DC, p. 128.
22. Keng, P. C. (1991). High-capacity separation of homogeneous cell subpopulations by centrifugal
elutriation. In Cell Separation Science and Technology, D. S. Kompala and P. Todd, eds., ACS
Symposium Series 464, American Chemical Society, Washington, DC, p. 103.