You are on page 1of 15

Chemical Engineering Science 61 (2006) 332 346

www.elsevier.com/locate/ces

A generalized population balance model for the prediction of particle size


distribution in suspension polymerization reactors
Costas Kotoulas, Costas Kiparissides
Department of Chemical Engineering, Aristotle University of Thessaloniki and Chemical Process Engineering Research Institute,
P.O. Box 472 541 24 Thessaloniki, Greece
Received 21 February 2005; received in revised form 1 July 2005; accepted 3 July 2005
Available online 22 August 2005

Abstract
In the present study, a comprehensive population balance model is developed to predict the dynamic evolution of the particle size
distribution in high hold-up (e.g., 40%) non-reactive liquidliquid dispersions and reactive liquid(solid)liquid suspension polymerization
systems. Semiempirical and phenomenological expressions are employed to describe the breakage and coalescence rates of dispersed
monomer droplets in terms of the type and concentration of suspending agent, quality of agitation, and evolution of the physical,
thermodynamic and transport properties of the polymerization system. The xed pivot (FPT) numerical method is applied for solving the
population balance equation. The predictive capabilities of the present model are demonstrated by a direct comparison of model predictions
with experimental data on average mean diameter and droplet/particle size distributions for both non-reactive liquidliquid dispersions
and the free-radical suspension polymerization of styrene and VCM monomers.
2005 Elsevier Ltd. All rights reserved.
Keywords: Population balance model; Suspension polymerization; PVC; Polystyrene

1. Introduction
Suspension polymerization is commonly used for producing a wide variety of commercially important polymers
(i.e., polystyrene and its copolymers, poly(vinyl chloride), poly(methyl methacrylate), poly(vinyl acetate)). In
suspension polymerization, the monomer is initially dispersed in the continuous aqueous phase by the combined
action of surface-active agents (i.e., inorganic or/and watersoluble polymers) and agitation. All the reactants (i.e.,
monomer, initiator(s), etc.) reside in the organic or oil
phase. The polymerization occurs in the monomer droplets
that are progressively transformed into sticky, viscous
monomerpolymer particles and nally into rigid, spherical
polymer particles of size 50500 m (Kiparissides, 1996).
Corresponding author. Tel.: +30 2310 99 6211;
fax: +310 2310 99 6198.
E-mail addresses: cypress@cperi.certh.gr,
cypress@alexandros.cperi.certh.gr (C. Kiparissides).

0009-2509/$ - see front matter 2005 Elsevier Ltd. All rights reserved.
doi:10.1016/j.ces.2005.07.013

The polymer solid content in the fully converted suspension


is typically 3050% w/w.
The suspension polymerization process can in general be
distinguished into two types (Kalfas, 1992): the bead polymerization, where the polymer is soluble in its monomer
and smooth spherical particles are produced; and the powder polymerization, where the polymer is insoluble in its
monomer and, thus, precipitates out leading to the formation of irregular grains or particles. The most important
thermoplastic produced by bead suspension polymerization
is polystyrene (PS). In the presence of volatile hydrocarbons (C4C6), foamable beads (the so-called expandable
polystyrene, EPS) can be produced. On the other hand,
poly(vinyl chloride) (PVC), which is the second largest thermoplastic manufactured in the world, is an example of powder polymerization.
One of the most important issues in suspension polymerization process is the control of the nal particle size
distribution (PSD) (Yuan et al., 1991). The initial monomer
droplet size distribution (DSD) as well as the nal polymer

C. Kotoulas, C. Kiparissides / Chemical Engineering Science 61 (2006) 332 346

Fig. 1. Schematic representation of drop breakage and coalescence mechanism.

PSD in general depend on the type and concentration of


the surface-active agents, the quality of agitation and the
physical properties (e.g., density, viscosity and interfacial
tension) of the continuous and dispersed phases. The transient droplet/particle size distribution is controlled by two
dynamic processes, namely, the drop/particle breakage and
coalescence rates. The former mainly occurs in regions of
high-shear stress (i.e., near the agitator blades) or as a result
of turbulent velocity and pressure uctuations along the surface of a drop. The latter is either increased or decreased by
the turbulent ow eld and can be assumed to be negligible
for very dilute dispersions at sufciently high concentrations
of surface-active agents (Chatzi and Kiparissides, 1992).
When drop breakage occurs by viscous shear forces, the
monomer droplet is rst elongated into two uid lumps separated by a liquid thread (see Fig. 1a). Subsequently, the
deformed monomer droplet breaks into two almost equalsize drops, corresponding to the uid lumps, and a series of
smaller droplets corresponding to the liquid thread. This is
known as Thorough breakage. On the other hand, a droplet
suspended in a turbulent ow eld is exposed to local pressure and relative velocity uctuations. For nearly equal densities and viscosities of the two liquid phases, the droplet
surface can start oscillating. When the relative velocity is
close to that required to make a drop marginally unstable,
a number of small droplets are stripped out from the initial
one (see Fig. 1b). This situation of breakage is referred to as
erosive one. Erosive breakage is considered to be the dominant mechanism for low-coalescence systems that exhibit
a characteristic bimodality in the PSD (Chatzi and Kiparissides, 1992; Ward and Knudsen, 1967).
Two different mechanisms have been proposed in the literature to describe the coalescence of two drops in a turbulent ow eld. The rst one (Shinnar and Church, 1960)
assumes that after the initial collision of two drops, a liquid
lm of the continuous phase is being trapped between the
drops that prevents drop coalescence (see Fig. 1c). However,

333

due to the presence of attractive forces, draining of the liquid lm can occur leading to drop coalescence. On the other
hand, if the kinetic energy of the induced drop oscillations
is larger than the energy of adhesion between the drops,
the drop contact is broken before the complete drainage of
the liquid lm. The second drop coalescence mechanism
(Howarth, 1964) assumes that immediate coalescence occurs when the approach velocity of the colliding drops at
the collision instant exceeds a critical value. In other words,
if the turbulent energy of collision is greater than the total
drop surface energy, the drops will coalesce (see Fig. 1d).
Surface-active agents play a very important role in the
stabilization of liquidliquid dispersions. One of the most
commonly used suspension stabilizers is poly(vinyl acetate)
that has been partially hydrolyzed to poly(vinyl alcohol)
(PVA). By varying the acetate content (i.e., degree of hydrolysis), one can alter the hydrophobicity of the PVA and,
thus, the conformation and surface activity of the polymer
chains at the monomer/water interface (Chatzi and Kiparissides, 1994). The solubility of the PVA in water depends on
the overall degree of polymerization (i.e., molecular weight),
the sequence chain length distribution of the vinyl alcohol and vinyl acetate in the copolymer, the degree of hydrolysis and temperature. Depending on the agitation rate,
the concentration and type of surface-active agent, the average droplet size can exhibit a U-shape variation with respect to the impeller speed or the degree of hydrolysis of
PVA. This U-type behaviour has been conrmed both experimentally and theoretically and has been attributed to
the balance of breakage and coalescence rates of monomer
drops.
In regard with the droplet/particle breakage and coalescence phenomena, the suspension polymerization process
can be divided into three stages (Hamielec and Tobita, 1992;
Maggioris et al., 2000). During the initial low-conversion
(i.e., low-viscosity) stage, drop breakage is the dominant
mechanism. As a result the initial DSD shifts to smaller
sizes. During the second sticky-stage of polymerization, the
drop breakage rate decreases while the drop/particle coalescence becomes the dominant mechanism. Thus, the average
particle size starts increasing. In the third stage, the PSD
reaches its identication point while the polymer particle
size slightly decreases due to shrinkage (i.e., the polymer
density is greater than the monomer one).
For the PS process, Villalobos et al. (1993) reported
that the end of the rst stage occurs at approximately 30%
monomer conversion, corresponding to a critical viscosity
of about 0.1 Pa s, while the second stage extends up to 70%
monomer conversion. In the VCM powder polymerization,
at monomer conversions around 1030%, a continuous
polymer network is commonly formed inside the polymerizing monomer droplets that signicantly reduces the
drop/particle coalescence rate (Kiparissides et al., 1994).
Cebollada et al. (1989) reported that the PSD is essentially
established up to monomer conversions of about 3540%
(i.e., end of the second stage).

334

C. Kotoulas, C. Kiparissides / Chemical Engineering Science 61 (2006) 332 346

The paper is organized as follows. In Section 2, a generalized population balance model is developed to describe
the dynamic evolution of PSD in batch suspension polymerization reactors. The model takes into account the dynamic
evolution of the physical properties of the continuous and
dispersed phases, in terms of the variation of monomer conversion and the turbulent intensity characteristics of the ow
eld, as well as their relative effects on breakage and coalescence mechanisms. In Section 3 of the paper, the xed pivot
technique (FPT) (Kumar and Ramkrishna, 1996) is applied
for solving the general population balance equation, governing the PSD developments, in terms of the polymerization
conditions (e.g., monomer to water volume ratio, temperature, type and concentration of stabilizer, impeller energy
input, etc.) and the polymerization kinetic model. An extensive analysis on the robustness of the numerical method
is carried out in regard with the convergence of the solution and the conservation of the total mass in the system.
Finally, in Section 4 of the paper, the capabilities of the
present model are demonstrated by a direct comparison of
model predictions with experimental data on average mean
diameter and droplet/particle size distributions for both nonreactive liquidliquid dispersions and the free-radical suspension polymerization of styrene and VCM monomers.
2. Model developments
To follow the dynamic evolution of PSD in a particulate
process, a population balance approach is commonly employed. The distribution of the droplets/particles is considered to be continuous in the volume domain and is usually described by a number density function, n(V , t). Thus,
n(V , t) dV represents the number of particles per unit volume in the differential volume size range (V , V + dV ). For
a dynamic particulate system undergoing simultaneous particle breakage and coalescence, the rate of change of the
number density function with respect to time and volume is
given by the following non-linear integro-differential population balance equation (Kiparissides et al., 2004):
 Vmax
j[n(V , t)]
=
(U, V )u(U )g(U )n(U, t) dU
jt
V
 V /2
+
k(V U, U )
Vmin

n(V U, t)n(U, t) dU n(V , t)g(V )


 Vmax
n(V , t)
k(V , U )n(U, t) dU .
(1)
Vmin

The rst term on the right-hand side (r.h.s.) of Eq. (1) represents the generation of droplets in the size range (V , V +dV )
due to drop breakage. (U, V ) is a daughter drop breakage
function, accounting for the probability that a drop of volume V is formed via the breakage of a drop of volume U .
The function u(U ) denotes the number of droplets formed
by the breakage of a drop of volume U and g(U ) is the

breakage rate of drops of volume U . The second term on the


r.h.s. of Eq. (1) represents the rate of generation of drops in
the size range (V , V +dV ) due to coalescence of two smaller
drops. k(V , U ) is the coalescence rate between two drops of
volume V and U . Finally, the third and fourth terms represent the drop disappearance rates due to drop breakage and
coalescence, respectively. Eq. (1) will satisfy the following
initial condition at t = 0:
n(V , 0) = n0 (V ),

(2)

where n0 (V ) is the initial drop size distribution of the dispersed phase. In the present study, the initial monomer DSD
was assumed to follow a normal distribution.
2.1. Breakage and coalescence rates
The solution of the population balance equation (Eq. (1))
presupposes the knowledge of the breakage and coalescence
rate functions. In the open literature, several forms of g(V )
and k(V , U ) have been proposed to describe the drop breakage and coalescence rate functions in liquidliquid dispersions (Coulaloglou and Tavlarides, 1977; Narsimhan et al.,
1979; Sovova, 1981; Chatzi et al., 1989). According to the
original work of Alvarez et al. (1994) and the proposed
modications of Maggioris et al. (2000), the breakage and
coalescence rates can be expressed in terms of the breakage, b , and collision, c , frequencies and the, respective,
Maxwellian efciencies, b and c :
g(V ) = b (V )eb (V ) ,
k(V , U ) = c (V , U )e

c (V ,U )

(3)
,

(4)

b and c denote the corresponding ratios of required to


available energy for an event to occur.
In the present study, the breakage of a drop exposed to a
turbulent ow eld was supposed to occur as result of energy transfer from an eddy to a drop having a diameter equal
to the eddy wave length, Dv . Eddy uctuations with wavelengths smaller (larger) than the drop diameter Dv produce
an oscillatory (rigid body) motion of the drop that do not
lead to breakage (Alvarez et al., 1994). The frequency term,
b (V ), was assumed to be equal to the inverse uctuation
time period, corresponding to the time required for a drop
to reach its mean drop displacement:
b (V ) = u(Dv )/Dv ,

(5)

where u(Dv )2 is the mean square of the relative velocity


between two points separated by a distance Dv , or the mean
square uctuation velocity of drops of diameter Dv .
For drops in the inertial subrange of turbulence (i.e.,
 < Dv  L), the energy spectrum will be independent of
the kinematic viscosity, c , and the mean uctuation velocity is solely determined by the rate of energy dissipation
(Hinze, 1959):
u(Dv )2 = kb (s Dv )2/3 ,

(6)

C. Kotoulas, C. Kiparissides / Chemical Engineering Science 61 (2006) 332 346

where kb is a model parameter and s is the average energy


dissipation rate for the dispersion.
For droplets in the viscous dissipation range (i.e., Dv < ),
the inertial forces are of the same order of magnitude as
the viscous shear forces and the mean square of the relative
velocity between two points separated by a distance Dv will
be given by (Shinnar, 1961)
u(Dv )2 = kb Dv2 (s /c ).

and inelastic dispersions (Yo ), Eq. (11) reduces to






1
12
Ve =
exp
1 .
(13)
12
Re
The scalar quantity Cds in Eq. (10) can be expressed in
terms of the numbers and volumes of daughter and satellite
drops (Chatzi and Kiparissides, 1992):

(7)

For high values of the dispersed phase volume fraction, the


damping effect of the dispersed phase on the local turbulent intensity needs to be taken into account. Doulah (1975)
proposed the following cubic equation for the calculation of
the average energy dissipation rate of the liquidliquid dispersion, s , in terms of the average energy dissipation rate
of the continuous phase, c , and the kinematic viscosities of
the continuous, c , and liquidliquid dispersion, s :
s /c = (c /s )3 .

(8)

Thus, in the presence of a high-volume dispersed phase, the


overall viscosity of the system increases and, therefore, the
energy dissipation rate for the system decreases.
According to Alvarez et al. (1994), for an effective drop
breakage to occur, the drop surface energy and drop viscoelastic resistance must be overcome. Considering that the
ow within a drop can be described as one-dimensional
simple-shear ow of a Maxwell uid, the breakage efciency
can be expressed as follows:
b = ab (Dv ),

(9)

Cds =

C
2
+ ds ,
Re(1 + Re Ve)
We

(10)

where Re and We denote the drop Reynolds and Weber


numbers, respectively. The dimensionless quantity Ve accounts for drop viscoelasticity and is a function of the drop
Reynolds number and its physical properties:


Yo
1a
Ve =
exp
a
2Re Yo



1+a
1
1a
a

exp
(11)
1a
1+a
ReYo
12
where
Yo =

2d
,
d Ed Dv2

1 48Yo .

(12)

In the present study, the dispersed-phase elasticity modulus, Ed , was approximated by the product of the polymer
elasticity modulus, Ep , and the fractional monomer conversion, (i.e., Ed = Ep ). For highly viscous (Re < 1)

Nda r 2/3 + Nda


(Nda r + Nsa )2/3

1,

r = Vda /Vsa ,

(14)

where Nda is the number of daughter drops of volume Vda


and Nsa is the number of satellite drops of volume Vsa . In
the present study, the number of daughter drops was set
equal to 2, the volume ratio of daughter to satellite drops, r,
was considered to be constant, while the number of satellite
drops was calculated as a function of the parent drop size
(Chatzi and Kiparissides, 1992):
1/3

Nsa = integer(Snsa Du ),

(15)

where Snsa is a model parameter estimated from experimental measurements on DSD or PSD.
Assuming that the daughter and satellite drops are normally distributed about their respective mean values with
standard deviations of da and sa , one can derive the following expression for the distribution of drops of volume
V , formed via the breakage of a drop of volume U:
u(U )(U, V )


(V Vda )2
= Nda
exp
2 2da
da 2



1
(V Vsa )2
+ Nsa
.
exp

2 2sa
sa 2

where ab is a model parameter and (Dv ) is the ratio of


required to available energy for a drop of diameter Dv to
break:
(Dv ) =

335

(16)

It should be noted that the daughter drop number density


function, u(U )(U, V ), should satisfy the following number
and volume conservation equations:
 U
u(U )(U, V ) dV = u(U ),
0
 U
V u(U )(U, V ) dV = U .
(17)
0

Accordingly, one can calculate the volumes of daughter and


satellite drops, formed by the breakage of a drop of volume
U, in terms of r, Nda and Nsa (Chatzi and Kiparissides,
1992):
Vda =

U
,
Nda + Nsa /r

Vsa =

U
.
Nda r + Nsa

(18)

Assuming that the drop collision mechanism in a locally isotropic ow eld is analogous to collisions between
molecules as in the kinetic theory of gases, the collision frequency between two drops with volumes V and U can be

336

C. Kotoulas, C. Kiparissides / Chemical Engineering Science 61 (2006) 332 346

expressed as (Maggioris et al., 2000)


c (V , U ) = kc (Dv2 + Du2 )(u(Dv )2 + u(Du )2 )1/2 .

(19)

For deformable drops, that is generally the case for low


interfacial tension dispersions or large-size drops, the drop
coalescence efciency can be expressed as (Coulaloglou and
Tavlarides, 1977)


c c s Dv Du 4
(
)
c (V , U ) = ac
,
(20)
Dv + D u
2
where ac is a model parameter. c , c , and  are the viscosity and density of the continuous phase, the interfacial
tension between the dispersed and aqueous phases and the
dispersed phase volume fraction, respectively.
At high monomer conversions, when the polymerizing
monomerpolymer particles behave like rigid spheres, the
coalescence efciency can be expressed as (Coulaloglou and
Tavlarides, 1977)

(b)
c (V , U ) = ac

c
1/3
c s (Dv

+ Du )4/3

(21)

In general, the monomer drops will behave like deformable drops at the beginning of polymerization while,
at high monomer conversions, they will behave like rigid
polymer particles. Thus, the coalescence efciency over the
whole monomer conversion range can be written as
exp{c (V , U )} = (1 ) exp{(a)
c (V , U )}
+ exp{(b)
c (V , U )},

(22)

where is the fractional monomer conversion.


2.2. Evaluation of the physical properties
The density of the suspension system, s , can be calculated as a weighted average of the densities of the dispersed
( d ) and continuous ( c ) phases (Bouyatiotis and Thornton,
1967):
s = d  + c (1 ).

(23)

The density of the dispersed phase will in turn be a function of the corresponding densities of the polymer ( p ) and
monomer ( m ) and the extent of monomer conversion, :
1


1
d =
+
.
(24)
p
m
The viscosity of the liquid(solid)liquid dispersion
was calculated by the following semi-empirical equation
(Vermeulen et al., 1955):


1.5d 
c
s =
,
(25)
1+
1
d + c
where d and c are the viscosities of the dispersed and
continuous phases, respectively.

For the heterophase suspension polymerization of VCM,


the viscosity of the polymerizing monomer droplets, d , can
be calculated by using the Eulers equation (Krieger, 1972):

d = m 1 +

0.5[n]p
1 p /cr

2
,

(26)

where p is the volume fraction of the polymer in the dispersed phase, given by p = ( d / p ). cr is the polymer
volume fraction corresponding to the critical monomer conversion c , at which a 3-D polymer skeleton is formed inside the polymerizing monomer drops. When p cr , the
dispersed-phase viscosity approaches innity, indicating the
formation of a rigid structure. Thus, for values of p larger
than the critical value cr , the dispersed phase viscosity was
assumed to remain constant.
For the VCM suspension polymerization, the value of
cr was taken to be equal to 0.3, which corresponds to
the monomer conversion at which a continuous polymer
network is formed inside the polymerizing VCM droplets
(Kiparissides et al., 1994). For the suspension polymerization of styrene, the value of cr was set equal to the 0.7
which corresponds to the monomer conversion at which particle coalescence stops. Finally, the intrinsic viscosity of the
polymer solution, [n], was calculated by the well-known
MarkHouwinkSakurada (MHS) equation as a function of
the weight average molecular weight of the polymer, Mw :
[] = kM aw .

(27)

The viscosity of the continuous phase depends on the


concentration and type of stabilizer that, in turn, affects the
PSD (Cebollada et al., 1989). Okaya (1992) employed the
SchulzBlaschke equation to calculate of the viscosity of
aqueous PVA solutions:

c = w

[PVA ]CPVA
1+
1 0.45[PVA ]CPVA


,

(28)

where c , w , [nPVA ] and CPVA are the viscosities of the


aqueous PVA solution and pure water, the intrinsic viscosity
and the stabilizer concentration, respectively.
In the open literature, a great number of papers have been
published, dealing with the behaviour of polymer molecules
at interfaces. Prigogine and his collaborators (Prigogine and
Marechal, 1952; Defay et al., 1966) presented a remarkably
simple theory on the surface tension of polymer solutions.
Although the Prigogine theory refers specically to the surface tension of polymer solutions, it is equally applicable to
the prediction of interfacial tension between a polymer solution and an immiscible liquid or a solid (Siow and Patterson,
1973). In the present study, the model of Siow and Patterson
(1973) was employed for the calculation of the interfacial
tension between the aqueous and the dispersed phase, . The
change in the interfacial tension with monomer conversion

C. Kotoulas, C. Kiparissides / Chemical Engineering Science 61 (2006) 332 346

was taken into account as in the original work of Maggioris


et al. (2000).

k=1

The numerical solution of the PBE commonly requires


the discretization of the particle volume domain into a number of discrete elements. Accordingly, the unknown number density function is approximated at a selected number
of discrete points, resulting in a system of stiff, non-linear
differential equations that is subsequently integrated numerically. Several numerical methods have been proposed in the
literature for the solution of the general PBE (Eq. (1)) in
the continuous or its equivalent discrete form (Kiparissides
et al., 2004). In the present study, the FTP of Kumar and
Ramkrishna (1996) was employed for solving the continuous PBE (Eq. (1)).
Assuming that the number density function remains constant in the discrete volume interval (Vi to Vi+1 ), one can
dene a particle number distribution, Ni (t), corresponding
to the i element:
 Vi+1
Ni (t) =
n(V , t) dV = ni (V , t)(Vi+1 Vi ).
(29)
Vi

Following the original developments of Kumar and


Ramkrishna (1996), the total volume domain (Vmin to
Vmax ) is rst divided into a number of elements. The
drop/particle population, Ni (t), corresponding to the size
range (Vi , Vi+1 ), is then assigned to a characteristic size xi
(also called grid point) as shown in Fig. 2. To account for
the formation of new particles of volume V that does not
correspond to the characteristic grid points (xi , xi+1 ), the
following approximation is applied. The particle number
fractions a(V , xi ) and b(V , xi+1 ) are assigned to the particle populations at the grid points xi and xi+1 , respectively,
so that two desired population properties of interest (e.g.,
total number and volume) are exactly preserved:
a(V , xi )f1 (xi ) + b(V , xi+1 )f1 (xi+1 ) = f1 (V ),

(30)

a(V , xi )f2 (xi ) + b(V , xi+1 )f2 (xi+1 ) = f2 (V ).

(31)

By integrating Eq. (1) over the discrete size interval


(Vi , Vi+1 ) and properly accounting for the respective drop
breakage and coalescence terms, the following set of

Vi-2

xi-1

Vi-1

xi

Vi

xi+1

Vi+1

Fig. 2. Discretization of the particle volume domain.

discretized equations can be derived:


M

dNi
=
nb (xi , xk )g(xk )Nk (t) +
dt

3. Numerical solution of the population balance


equation

xi-2

337

Vi+2

j
k
j,k
xi1  xj +xk  xi+1



1
1 j,k nc (xi , V )k(xj , xk )Nj (t)Nk (t)
2
M

g(xi )Ni (t) Ni (t)
k(xi , xk )Nk (t),
(32)
k=1

where nb (xi , xk ) denotes the fraction of drops/particles of


size xi resulting from the breakage of a drop/particle of size
xk . To preserve the number and volume of particles in the
size range (Vi , Vi+1 ), nb (xi , xk ) must satisfy the following
equation:

nb (xi , xk ) =

xi+1

xi+1 V
(xk , V )u(xk ) dV
x
i+1 xi
xi
 xi
V xi1
(xk , V )u(xk ) dV .
+
x
xi1 i xi1

(33)

The respective expression for nc (xi , V ), accounting for the


fraction of drops/particles of size V (=xj + xk ) formed via
the coalescence of two drops/particles of volumes xj and
xk , will be given by
x
i+1 V

xi+1 xi
nc (xi , V ) = V x
i1

xi xi1

xi V xi+1 ,
(34)
xi1 V xi .

Accordingly, from the calculated values of Ni (t), one can


easily obtain the values of the average number density function, n i (V , t):
ni (V , t) =

Ni (t)
.
(Vi+1 Vi )

(35)

It is often desirable to know the number diameter density


function, n(D, t). By noting that n(V , t) dV = n(D, t) dD,
one can easily calculate the average number diameter density
function, ni (D, t), in terms of Ni (t):

Ni (t) =

Di+1
Di

n(D, t) dD = n i (Di , t)(Di+1 Di ).

(36)

In many cases, experimental measurements on PSD are


given in terms of number or volume fractions, from which
it is not easy to derive the actual particle number distribution, Ni (t), or the number volume density function,
n(V , t). Thus, it is better to compare directly the available
experimental measurements on number and volume fraction
distributions with respective simulation results obtained
from the numerical solution of the PBE. Accordingly, one
can dene the following number, A(D, t), and volume,

338

C. Kotoulas, C. Kiparissides / Chemical Engineering Science 61 (2006) 332 346

Table 1
Physical/transport properties and model parameters for VCM/PVC system

m = 947 1.746(T 273.15) (Kg/m3 )


p = 103 exp(0.4296 3.274 104 T ) (Kg/m3 )
m = 2.1 104 106 (T 273.15) (Kg/m s)
c = 1011 0.4484(T 273.15) (Kg/m3 )
c = 0.08 exp(1.5366 102 T ) (Kg/m s)

(Kiparissides et al., 1997)


(Kiparissides et al., 1997)
(Kiparissides et al., 1997)

0.851 (m 3 /Kg)
[n] = 1.087 105 1.67 108 (T 273.15)Mw
3
5
[s ] = 9.13 10 + 4.317 10 DP
Ep = 2.4 109 (Kg/m s2 )
r = 35, SNsa = 110, kb = 324, ab = 33, kc = 3 107 , ac = 2 109 , ac = 1 102

(Polymer Handbook)
(Okaya, 1992)
(Polymer Handbook)
(This study)

Table 2
Physical, transport properties and model parameters for styrene/PS system

m = 923.6 0.887(T 273.15) (Kg/m3 )


p = 1050 0.602(T 273.15) (Kg/m3 )

(Achilias and Kiparissides, 1992)


(Achilias and Kiparissides, 1992)

(Kg/m s)
0.722 (m 3 /Kg)
[n] = 1.38 105 Mw
Ep = 3.38 109 (Kg/m s2 )
r = 35, SNsa = 50, kb = 400, ab = 33, kc = 4 107 , ac = 5 109 , ac = 3 103

(Achilias and Kiparissides, 1992)


(Achilias and Kiparissides, 1992)
(Polymer Handbook)
(This study)

m = 10528.64(1/T 1/276.71)

AV (D, t), probability density functions:

4. Results and discussion

fNi
n(D, t)

,
Nt (t)
Di+1 Di
( D 3 /6)n(D, t)
fVi
.

AV (D, t) =
Vt (t)
Di+1 Di

The predictive capabilities of the proposed model were


demonstrated by a direct comparison of model predictions with experimental data on average mean diameter
and droplet/particle size distribution of both non-reactive
liquidliquid dispersions of styrene and VCM in water,
and free-radical suspension polymerization of styrene and
VCM. For polymerization systems, the general population balance model (see Eq. (1)) was solved together with
the pertinent molecular species differential equations (see
Appendix A) describing the molecular weight developments (e.g., number and weight average molecular weights)
and the polymerization rate in the heterophase suspension
system. In addition to the above dynamic equations, the
dynamic model included all the necessary algebraic equations, describing the variation of the kinetic rate constants,
and the physical, transport and thermodynamic properties of the multi-phase system with respect to the reactor
operating conditions (e.g., temperature, monomer mass,
etc.) and the fractional monomer conversion. Additional
details, regarding the kinetic mechanism (e.g., gel- and
glass-effect), phase equilibrium calculations (e.g., monomer
and initiator partitioning, number of phases in the system,
etc.) can be found in the publications of Kiparissides et al.
(1997), Kotoulas et al. (2003) and Krallis et al. (2004). In
Tables 1 and 2, the physical, transport and model parameters for the VCM/PVC and styrene/PS systems are reported.
It should be noted that the numerical values of the key
model parameters (i.e., kb , ab , kc , ac ) were estimated by
tting the model predictions to experimental data on DSD
of liquidliquid aqueous dispersions of styrene and VCM.
The system of non-linear ordinary differential equations

A(D, t) =

(37)

A(D, t) dD and AV (D, t) dD represent the number (fNi )


and volume (fV i ) fractions of particles in the size range
(D, D + dD), respectively. Nt (t) and Vt (t) denote the respective total number and volume of particles per unit volume of the reaction medium. It is apparent that the number and volume probability density functions will satisfy the
following normalization conditions:


Dmax

Dmin


A(D, t) dD = 1,

Dmax
Dmin

AV (D, t) dD = 1.

(38)

Very often experimental measurements on some average


particle diameter are only available. In general, the average
particle diameter, Dqp , can easily be calculated, in terms of
the number probability density function, using the following
equation (Chatzi and Kiparissides, 1992):
(Dqp )qp

 Dmax
q
=
D A(D, t) dD
Dmin

Dmax

Dmin

D p A(D, t) dD,

(39)

where q and p are characteristic exponents that dene the


desired average particle diameter (e.g., mean Sauter diameter, D32 , average number diameter, D10 , average volume
diameter, D30 , etc.).

10

-2

10

-3

10

-4

10

-5

10

-6

10

-7

10

-8

10

-9

10

-10

10

-11

10

-12

10

-13

10

-14

339

0.009

NE
NE
NE
NE

200

=
=
=
=

30
50
80
100

400

600

800

1000

1200

1400

1600

Volume Probability Density Function (m-1 )

-1

Volume Probability Density Function (m )

C. Kotoulas, C. Kiparissides / Chemical Engineering Science 61 (2006) 332 346

Final Distributions
D = 200 m
D = 700 m
D = 1000 m

0.006
0.005
0.004
0.003
0.002
0.001

200

400

600
800
1000
Particle Diameter (m)

1200

1400

Fig. 4. Effect of the initial DSD on the calculated volume probability


density function of styrene droplets in water (non-reactive case).
600

Sauter Mean Diameter, D32 (m)

(Eqs. (32)(34)) together with the necessary kinetic equations (see Appendix A) were numerically integrated using
the Gear predictorcorrector DE solver.
Validation of the numerical method: The accuracy and
convergence characteristics of the numerical method (FTP)
were rst assessed by varying the total number of discretization points, the size of the total volume domain and the initial
DSD. Fig. 3 shows the effect of the number of equal-size discrete elements (i.e., 30, 50, 80 and 100) on the volume probability density function for the styrene suspension polymerization. The diameter domain extended from 1 to 2000 m
while the initial DSD followed a Gaussian distribution with
a mean value of D0 = 1000 m and a standard deviation of
D = 100 m. As can be seen, the volume probability density function converges to the same distribution for values
of the number of elements NE80. In the present study,
it was assumed that the numerically calculated distribution
converged to the correct one when the total mass of the dispersed phase (i.e., monomer plus polymer), given
 by the rst
moment of particle number distribution, ( d N
k=1 Vi Ni (t)),
differed from the initial monomer mass by less than 23%.
When the upper limit of the total diameter domain, Dmax ,
was reduced from 2000 to 1200 m, it was found that the
number of discrete elements, required for the satisfaction
of above mass conservation criterion, was NE 50. Thus, it
was concluded that the numerical solution converged to the
correct distribution when the size of the discrete elements
(i.e., the ratio of the total diameter domain over the number of elements) was smaller than 25 m. A similar rule was
found to be applicable to the VCM/PVC suspension polymerization system.
In liquidliquid dispersions the nal DSD is controlled by
the dynamic equilibrium between drop breakage and coalescence rates. Thus, for the same operating conditions (e.g.,
input power, dispersed phase volume fraction, temperature,
etc.) the nal DSD should be independent of the initial DSD.
Fig. 4 illustrates the effect of the initial DSD on the nal

0.007

0.000

Particle Diameter (m)

Fig. 3. Effect of the number of volume elements on the calculated volume


probability density function (free-radical suspension polymerization of
styrene).

Initial Distributions
D = 200 m = 50 m
D = 700 m =100 m
D = 1000 m =100 m

0.008

D = 200 m = 50 m
D = 700 m = 100 m
D =1000 m = 100 m

550
500
450
400
350
300
250
200
0

50

100

150

Time (min)

Fig. 5. Effect of the initial DSD on the dynamic evolution of the Sauter
mean diameter of styrene droplets in water (non-reactive case).

DSD at dynamic equilibrium for the styrenewater dispersion system. As can be seen, the calculated nal DSD is not
affected by the initial condition. On the other hand, the time
required for the system to attain its nal DSD is affected by
the initial DSD condition. Fig. 5, clearly depicts the variation of the Sauter mean droplet diameter with respect to
time. In all cases, the drop breakage and coalescence rate
functions were the same. It is apparent that the time required
for the liquidliquid dispersion to reach its dynamic equilibrium distribution is larger when the initial DSD had a mean
value of D0 = 200 m. On the other hand, no signicant
differences in the required times for the system to reach its
dynamic equilibrium were observed when the mean value
of the initial DSD changed from 1000 to 700 m. Notice
that in the former case (i.e., D0 = 200 m and D = 50 m)
the drop coalescence mechanism controls the dynamic evolution of DSD, while in the later case (i.e., D0 = 1000 m
and D = 100 m) the DSD evolution is mainly controlled
by the drop breakage mechanism.

340

C. Kotoulas, C. Kiparissides / Chemical Engineering Science 61 (2006) 332 346


60

Initial DSD

D = 80 m

D0 = 700 m

D 10 (
D 32 (

D0 =1000 m

D = 100 m

D =100 m

Mean Droplet Diameter (m)

D0 = 200 m

-1

Volume Probability Density Function (m )

0.010

0.008
Final PSD

0.006

(D =1000 m)
(D = 700 m)
(D = 200 m)

0.004

0.002

sim)
sim)

50

40

30

20
0

0.000
0

200

400

600

800

1000

20

Fig. 6. Effect of the initial DSD on the calculated polystyrene particle


size distribution (suspension polymerization of styrene).

60

80

100

Time (min)

1200

Particle Diameter (m)

40

Fig. 8. Dynamic evolution of the calculated and measured mean


diameter of VCM droplets (non-reactive case: monomer volume fraction = 0.1; CPVA (72.5% degree of hydrolysis) = 0.02%;
temperature = 55 C; N = 500 rpm).

550
5 min
10 min
30 min
120 min
Experimental

-1

Volume Probability Density Function (m )

Sauter Mean Diameter, D32 (m)

0.06
500
450
400
350
300
D = 200 m = 80 m
D = 700 m = 100 m
D =1000 m = 100 m

250
0

50

100

150

200

250

300

350

Time (min)

0.05
0.04
0.03
0.02
0.01
0.00
0

30

60

90

120

150

Particle Diameter (m)


Fig. 7. Effect of the initial DSD on the dynamic evolution of the Sauter
mean diameter of styrene droplets in water (suspension polymerization
of styrene).

It should be noted that when the polymerization in the


monomer droplets starts before the system has reached its
liquidliquid equilibrium distribution, the nal PSD in the
suspension system will not be independent of the initial DSD
condition. In Figs. 6 and 7, the effect of the initial DSD
on the nal PSD is depicted for the free-radical suspension
polymerization of styrene, assuming that the polymerization in the monomer droplets starts at time zero (i.e., before the liquidliquid dispersion reaches its dynamic equilibrium point). As can be seen, as the average size of the
initial monomer DSD increases, the nal PSD is shifted to
larger sizes. The reason is that drop breakage ceases before
the liquidliquid dispersion has reached its nal equilibrium
distribution.
Vinyl Chloride suspension polymerization. The dispersion
of vinyl-chloride monomer (VCM) in aqueous PVA solutions has been studied experimentally by Zerfa and Brooks

Fig. 9. Dynamic evolution of the calculated distribution of VCM droplets


(non-reactive case: experimental conditions as in Fig. 8; discrete points
represent experimental measurements).

(1996a,b) under different conditions (e.g., monomer holdup, agitation speed and type and concentration of stabilizers).
Fig. 8 illustrates the dynamic evolution of the number mean
diameter, D10 , and the Sauter mean diameter, D32 , of VC
monomer droplets in the dispersion. The monomer volume
fraction in the dispersion was 0.1, the temperature was kept
constant at 55 C, the agitation speed was set at 500 rpm,
while 200 ppm of PVA with a degree of hydrolysis equal to
72.5% were added to the aqueous phase for the stabilization of the VCM droplets (Zerfa and Brooks, 1996b). The
continuous lines represent simulation results while the discrete points the experimental measurements. As can be seen,
the droplet size initially reduces (i.e., due to the dominant
drop breakage mechanism) and reaches its nal dynamic
equilibrium value, at approximately 30 min. The evolution
of DSD is shown in Fig. 9. Initially, the volume probability

341

200

0.07
250 rpm
350 rpm
500 rpm
650 rpm

0.06

(
(
(
(

sim)
sim)
sim)
sim)

Volume Mean Diameter, D30 (m)

Volume Probability Density Function (m-1)

C. Kotoulas, C. Kiparissides / Chemical Engineering Science 61 (2006) 332 346

0.05
0.04
0.03
0.02
0.01
0.00
0

30

60

90

120

150

Experimental
Simulation

180
160
140
120
100
80

180

Particle Diameter (m)

60
0

Fig. 10. Calculated and experimentally measured distributions of


VCM droplets in water at different agitation rates (monomer
volume fraction = 0.1; CPVA (72.5% degree of hydrolysis) = 0.03%;
temperature = 55 C).

density function of VCM is broad. However, as the agitation


continues, it becomes narrower and shifts to smaller sizes.
The predicted steady-state DSD (at 120 min) is in excellent
agreement with the experimentally measured one (black discrete points). It is clear that the proposed model is capable
of predicting satisfactorily the dynamic evolution of VCM
distribution as well as its mean droplet value.
Fig. 10 illustrates the effect of the agitation rate on the
steady-state volume probability density function of VCM
droplets in the aqueous dispersion. All other experimental
conditions were similar to those of Fig. 8, except the concentration of the PVA stabilizer, which was 300 ppm (Zerfa
and Brooks, 1996a). The discrete points represent the experimental measurements while the continuous lines the model
predictions. As can be seen, as the agitation rate increases
the DSD shifts to smaller sizes and becomes narrower due
to the increased drop breakage rate. In all cases, the model
results are in very close agreement with the experimental
data. An additional comparison study was carried out for a
VCM dispersed volume fraction, of 0.2. It was found that
the VCM droplet distribution shifted to larger sizes as the
monomer hold-up increased. Again, simulation results were
in excellent agreement with experimental measurements on
DSD.
Subsequently, experimental measurements on the average
particle size and PSD were compared with model predictions
for the free-radical suspension polymerization of VCM. The
experimental data for the PVC system were provided by
ATOFINA. The experiments were carried out in a 30 L batch
reactor, using 40% v/v VCM in water. The polymerization
temperature was set at 56.5 C while the agitation speed
remained constant at 330 rpm.
Fig. 11 depicts the variation of the volume mean diameter with respect to polymerization time. The continuous
line represents the simulation results and the discrete points
the experimental measurements. Initially, the mean diameter

50

100

150

200

250

300

Time (min)

Fig. 11. Dynamic evolution of calculated and experimentally measured volume mean diameter of PVC particles (reactive case: temperature=56.5 C;
dispersed phase volume fraction = 0.4; agitation rate = 330 rpm).

shifts to smaller values due to the dominant drop breakage


mechanism. Subsequently, the drop breakage rate is reduced
while the drop coalescence rate increases because of the increased viscosity of the dispersed phase. Thus, the mean
particle diameter increases until a monomer conversion of
about 75%. After this point, the drop coalescence rate ceases
and the PSD remains almost constant. It is apparent that the
present model predicts very well the dynamic behaviour of
the PSD for the free-radical suspension polymerization of
VCM. In Fig. 12, experimental measurements (dash lines)
and simulation results (continuous lines) on PSD are plotted at four different conversion levels (i.e., 55%, 65%, 75%
and 83%). As can been seen, the simulation results are in
very good agreement with the experimental measurements.
It should be pointed out that all the simulation results on
VCM suspension polymerization (i.e., for both reactive and
non-reactive cases) were obtained using the same values of
the model parameters (see Table 1).
Styrene suspension polymerization. The dynamic evolution of styrene DSD in aqueous dispersions was experimentally studied by Yang et al. (2000). Fig. 13 illustrates the
dynamic evolution of the Sauter mean diameter of styrene
droplets for two different monomer volume fractions. The
temperature was kept constant at 25 C, the agitation speed
was 350 rpm, while 500 ppm PVA were added to the aqueous phase for the stabilization of the dispersion. The PVA
had a degree of hydrolysis equal to 88%, while its molecular weight varied between 30,000 and 50,000 g/mol. The nal DSD at dynamic equilibrium was attained after 150 min
of stirring. Apparently, the model predicts fairly well the
dynamic evolution of the Sauter mean diameter of styrene
droplets, as well as the effect of monomer volume fraction.
In Fig. 14, the time evolution of DSD is depicted for the
case of styrene volume fraction of 0.1. It is evident that the
model predictions on DSD are in very good agreement with

342

C. Kotoulas, C. Kiparissides / Chemical Engineering Science 61 (2006) 332 346

0.014
Conversion 55%

0.012

Experimental
Simulation

0.010
-1

Volume Probability Density Function (m )

Conversion 65%
Experimental
Simulation

0.008
0.006
0.004
0.002
0.000
0.014
Conversion 75 %

0.012

Conversion 83%

Experimental
Simulation

0.010

Experimental
Simulation

0.008
0.006
0.004
0.002
0.000
0

100

200

300

400 0

100

200

300

400

Particle Diameter (m)


Fig. 12. Predicted and experimentally measured distributions of PVC particles at four different conversion levels: 55%, 65%, 75% and 83% (experimental
conditions as Fig. 11).

Sauter Mean Diameter, D32 (m)

160

sim)
sim)

140

120

100

80
0

50

100

150

200

0.018

-1

Volume Probability Density Function (m )

= 0.05 (
= 0.10 (

250

Time (min)

Fig. 13. Dynamic evolution of calculated and experimentally measured


Sauter mean diameter of styrene droplets at two different monomer volume
fractions (non-reactive case: temperature=25 C; agitation rate=350 rpm;
CPVA (88% degree of hydrolysis) = 0.05%).

experimental measurements (discrete points). Fig. 15 illustrates the effect of the agitation rate on the steady-state DSD
of the styrene droplets. The operating conditions were as in
Fig. 13, while the monomer volume fraction was 0.1. As in
the case of the VCM dispersion, the mean size of styrene
droplets decreases with the agitation rate while the DSD be-

5 min (
30 min (
120 min (

0.015

sim)
sim)
sim)

0.012
0.009
0.006
0.003
0.000
50

100

150

200

250

300

Particle Diameter (m)

Fig. 14. Dynamic evolution of calculated and experimentally measured


distributions of styrene droplets in water (non-reactive case: experimental
conditions as Fig. 13; monomer volume fraction = 0.1).

comes narrower. In all cases, the simulation results are in


very good agreement with the experimental data that clearly
underlines the predictive capabilities of the present comprehensive population balance model.
Finally, the present model was employed to predict the
dynamic evolution of PSD in the free-radical suspension
polymerization of styrene. More specically, the effects of

-1

Volume Probability Density Function (m )

C. Kotoulas, C. Kiparissides / Chemical Engineering Science 61 (2006) 332 346

to the system at three different conversion levels (i.e., 0, 50%


and 100%). In Fig. 16a, model results (continuous lines)
are compared with experimental data (dash lines) on PSD
for the case of n-pentane addition at zero monomer conversion. In Fig. 16b, the corresponding distributions are illustrated for the case of n-pentane addition at 50% monomer
conversion. The last case (see Fig. 16c) corresponds to the
addition of n-pentane at the end of polymerization. As can
been seen, for all cases, there is a close agreement between
experimental and simulation results on PSD, indicating the
predictive capabilities of the model for the free-radical suspension polymerization of styrene. It should be noted that,
in the presence of n-pentane, the EPS particles are more
uniform while the PSD becomes narrower.
Villalobos et al. (1993) also investigated experimentally
the effect of the stabilizer concentration on PSD. More
specically, experiments were carried out at different TCP
concentrations, in the presence of 7.5% w/w n-pentane,
added at 50% monomer conversion. In Fig. 17 the predicted
and experimental PSDs of the EPS particles are shown for
the three different concentrations of the surface-active agent.
As can be seen, as concentration of the surface-active agent
decreases (i.e., the interfacial tension increases) the PSD becomes broader and shifts to larger sizes. Apparently, there is
a very good agreement between calculated and experimental
measurements on the volume probability density function.

0.040
250 rpm (
450 rpm (
650 rpm (

0.035

sim)
sim)
sim)

0.030
0.025
0.020
0.015
0.010
0.005
0.000
0

50

100

150

200

250

300

350

Particle Diameter (m)

Fig. 15. Effect of agitation rate on the calculated and experimentally measured distributions of styrene droplets in water
(non-reactive case: temperature = 25 C; agitation rate = 350 rpm; CPVA
(88% degree of hydrolysis) = 0.05%; monomer volume fraction = 0.1).

-1

Volume Probability Density Function (m )

n-pentane and concentration of stabilizer on the PSD were


investigated for the expandable PS suspension polymerization process. Experimental measurements on PSD were
taken from the work of Villalobos et al. (1993). The freeradical styrene suspension polymerization was carried out
in a 1-gal reactor vessel. The dispersed monomer volume
fraction was 0.4. The polymerization took place at 105 C in
the presence of 1,4-bis(terbutyl peroxycarbo) cyclohexane
(TBPCC) bifunctional initiator. The initiator concentration
was 0.01 mol/L-styrene in all the experimental cases, while
tricalcium phosphate (TCP) was used as surface-active agent
at three different concentrations (i.e., 7.5, 5.0 and 3.5 gr/L).
In Fig. 16, experimental measurements and simulation results on PSD are shown for three different addition policies
of n-pentane into the reactor. More specically, 7.5% w/w
of n-pentane with respect to the styrene mass was added
0.0030

(a)

Experimental
Simulation

0.0025

343

5. Conclusions
A comprehensive population balance model coupled with
a system of differential equations governing the conservation of the various molecular species present in the system
has been developed to describe the dynamic evolution of
the DSD/PSD in free-radical suspension polymerization reactors. The xed pivot technique (FPT) was employed for
solving the PBE. The robustness of the numerical method

(b)

Experimental
Simulation

(c)

Experimental
Simulation

0.0020
0.0015
0.0010
0.0005
0.0000
0

200

400 600 800 1000 1200


Particle Diameter (m)

200

400

600

800 1000 1200

Particle Diameter (m)

200

400

600

800 1000 1200 1400

Particle Diameter (m)

Fig. 16. Predicted and experimentally measured distributions of EPS particles for different n-pentane addition policies. (a) 7.5% w/w n-pentane (wrt
styrene) at = 0%; (b) 7.5% w/w n-pentane (wrt styrene) at = 50% ; (c) in the absence of n-pentane (temperature = 105 C; dispersed phase volume
fraction = 0.4, [Io ] = 0.01 mol TBPCC/L-styrene; [TCP] = 7.5 g/L).

344

C. Kotoulas, C. Kiparissides / Chemical Engineering Science 61 (2006) 332 346

Volume Probability Density Function (m)

0.0035
7.5 % w/w (
5.0 % w/w (
3.5 % w/w (

0.0030

Re
Snsa
t
u(V )

sim)
sim)
sim)

0.0025

u(Dv )2

0.0020
0.0015

Vda , Vsa

0.0010

V , U, x
We

0.0005
0.0000
0

500

1000

1500

2000

Particle Diameter (m)

Fig. 17. Effect of surface-active concentration on the calculated and


experimentally measured distributions of EPS particles at three different
quantities of surface-active agent (TCP) (experimental conditions as in
Fig. 16).

Greek letters
ab , ac

model parameters

(U, V )

daughter droplets probability function,


1/m3
average energy dissipation rate per unit
mass, m2 /s3
microscale of turbulence, m
breakage and coalescence efciencies
viscosity, Kg/m s
kinematic viscosity, m2 /s
density, Kg/m3
interfacial tension, Kg/s2
standard deviation of the distribution
for daughter and satellite drops
dispersed phase volume fraction
volume fraction of the polymer in the
dispersed phase
monomer conversion
breakage and coalescence frequencies,
1/s

was examined in regard with its convergence characteristics and accuracy in terms of the mass conservation of
the monomer, initially loaded into the reactor. The predictive capabilities of the model were demonstrated via the
successful simulation of experimental measurements on
DSD/PSD and the average droplet/particle diameter for both
non-reactive liquidliquid dispersions and the free-radical
suspension polymerization of styrene and VCM.


b , c




da , sa

Notation


p

A(D, t), Av (D, t) number and volume probability density


functions, 1/m
CPVA
concentration of surface-active agent,
Kg/m3
D
diameter, m
DP
degree of polymerization of the PVA
stabilizer
E
elasticity modulus, Kg/m s2
g(V )
breakage rate, 1/s
k(V , U )
coalescence rate, m3 /s
kb , kc
model parameters
L
macroscale of turbulence, m
Mw
weight average molecular weight,
Kg/kmol
n(V , t)
number density function, 1/m6
[n]
intrinsic viscosity, m3 /Kg
NE
number of discrete elements
Ni
number of particles having volume
equal to xi per reactor unit volume,
1/m3
Nda , Nsa
number of daughter and satellite
droplets per breakage events
r
volume ratio of daughter over the satellite drops

Reynolds number
model parameter
time, s
number of droplets formed by a breakage of a droplet of volume V
mean square of the relative velocity between two points separated by a distance D, m/s
volumes of daughter and satellite
drops, m3
volumes, m3
Weber number


b , c
Subscripts
c

continuous phase

d
m
p
s
w

dispersed phase
monomer
polymer
suspension system
water

Acknowledgement
The authors gratefully acknowledge ARCHEMA (exATOFINA Chemicals) for providing the experimental data
for PVC suspension polymerization.

Appendix A
The free-radical polymerization of vinyl monomers in
general includes the following chain initiation, propagation,

C. Kotoulas, C. Kiparissides / Chemical Engineering Science 61 (2006) 332 346

chain transfer to monomer and bimolecular termination reactions (Kiparissides et al., 1997):
decomposition of initiators
kd,i

Ii 2R ,

Live polymer moment rate equations



 k  
Nm

k
r k
2fk kdk Ik + kp M
r k =
r
k=1

i = 1, 2, . . . , Nm ,

k  
1 j k
r kr
rk = kfm Mk + ktc
r
2

chain propagation

r=0

kp

+ ktd k 0 + kz Zk .

Pn + M Pn+1 ,
kf m

Pn + M Dn + P1 ,
termination by combination

Mn =

ktc

Pn + Pm Dn+m ,
ktd

Pn + Pm Dn + Dm ,
inhibition of live radical chains

M
d
0 .
= kp
dt
M0

kz

Pn + Z Dn + Z ,
where Ii , R , M and Z denote the initiator, primary radicals,
monomer and inhibitor molecules, respectively, and Pn and
Dn , the corresponding live and dead polymer chains,
having a degree of polymerization n.
In the free-radical polymerization of VCM, the polymer
is insoluble in its monomer, thus, precipitates out to form
a separate phase (i.e., the polymer-rich phase). Thus, the
elementary reactions presented above take place in both the
monomer-rich and polymer-rich phases (Kiparissides et al.,
1997). Additional details, regarding the kinetic modeling
of free-radical polymerization of styrene and VCM (e.g.,
gel- and glass-effect), phase equilibrium calculations (e.g.,
monomer and initiator partitioning, number of phases in the
system, etc.), can be found in the publications of Kiparissides
et al. (1997, 2004) and Kotoulas et al. (2003).
The method of moments is invoked in order to reduce the
innite system of molar balance equations, required to describe the molecular weight distribution developments. Accordingly, the average molecular properties of the polymer
(i.e., Mn , Mw ) are expressed in terms of the leading moments of the dead polymer molecular weight distribution.
The moments of the total number chain length (TNCL) distributions of live radical and dead polymer chains can be
dened as (Krallis et al., 2004)

n=i

nk D n .

(1 + 1 )
MWm ,
(0 + 0 )

Mw =

(2 + 2 )
MWm .
(1 + 1 )

(A.4)

Finally, the total monomer conversion can be calculated by


the following expression, assuming that the long chain hypothesis holds true (i.e., the monomer is mainly consumed
via the propagation reaction):

termination by disproportionation

k =

(A.3)

The number- and weight-average molecular weights


can be expressed in terms of the molecular weight of the
monomer, MWm , and the moments of the TNCLDs of live
and dead polymer chains:

chain transfer to monomer

nk Pn ,

(A.2)

Dead polymer moment rate equations

kp

R + M P1 ,

r=0

+kfm M(0 k )(ktc +ktd )k 0 kz Zk .

chain initiation

k =

345

(A.1)

n=i

Accordingly, one can easily derive the corresponding moment rate functions:

(A.5)

References
Achilias, D.S., Kiparissides, C., 1992. Development of a general
mathematical framework for modeling diffusion-controlled free-radical
polymerization reactions. Macromolecules 25 (14), 37393750.
Alvarez, J., Alvarez, J., Hernandez, M., 1994. A population balance
approach for the description of particle size distribution in suspension
polymerization reactors. Chemical Engineering Science 49, 99113.
Bouyatiotis, B.A., Thornton, J.D., 1967. Liquidliquid extraction studies
in stirred tanks. Part I. Droplet size and hold-up measurements in a
seven-inch diameter bafed vessel. Institution of Chemical Engineers
(London) Symposium Series 26, 4350.
Cebollada, A.F., Schmidt, M.J., Farber, J.N., Cariati, N.J., Valles, E.M.,
1989. Suspension polymerization of vinyl chloride. I. Inuence of
viscosity of suspension medium on resin properties. Journal of Applied
Polymer Science 37, 145166.
Chatzi, E.G., Kiparissides, C., 1992. Dynamic simulation of bimodal
drop size distributions in low-coalescence batch dispersion systems.
Chemical Engineering Science 47, 445456.
Chatzi, E.G., Kiparissides, C., 1994. Drop size distributions in high
holdup fraction suspension polymerization reactors: effect of the degree
of hydrolysis of PVA stabilizer. Chemical Engineering Science 49,
50395052.
Chatzi, E.G., Gavrielides, A.D., Kiparissides, C., 1989. Generalized model
for prediction of the steady-state drop size distribution in batch stirred
vessel. Industrial Engineering Chemistry Research 28, 17041711.
Coulaloglou, C.A., Tavlarides, L.L., 1977. Description of interaction
processes in agitated liquidliquid dispersions. Chemical Engineering
Science 32, 12891297.
Defay, R., Prigogine, I., Bellemans, A., Everett, D.H., 1966. Surface
Tension and Adsorption. Wiley, New York.
Doulah, M.S., 1975. On the effect of holdup on drop sizes in liquidliquid
dispersions. Industrial Engineering Chemistry Fundamentals 14,
137138.

346

C. Kotoulas, C. Kiparissides / Chemical Engineering Science 61 (2006) 332 346

Hamielec, A.E., Tobita, H., 1992. Polymerization processes. Ullmanns


Encyclopedia of Industrial Chemistry, vol. A21. VCH Publishers,
New York, pp. 305428.
Hinze, J.O., 1959. Turbulence. McGraw-Hill, New York.
Howarth, W.J., 1964. Coalescence of drops in a turbulent ow eld.
Chemical Engineering Science 19, 3338.
Kalfas, G.A., 1992. Experimental studies and mathematical modeling of
aqueous suspension polymerization reactors. Ph.D. Thesis, University
of WisconsinMadison, USA.
Kiparissides, C., 1996. Polymerization reactor modeling: a review of recent
developments and future directions. Chemical Engineering Science 51,
16371659.
Kiparissides, C., Achilias, D.S., Chatzi, E., 1994. Dynamic simulation
of primary particle-size distribution in vinyl chloride polymerization.
Journal of Applied Polymer Science 54, 14231438.
Kiparissides, C., Daskalakis, G., Achilias, D.S., Sidiropoulou, E., 1997.
Dynamic simulation of industrial poly(vinyl chloride) batch suspension
polymerization reactors. Industrial Engineering Chemistry Research
36, 12531267.
Kiparissides, C., Alexopoulos, A., Roussos, A., Dompazis, G., Kotoulas,
C., 2004. Population balance modelling of particulate polymerization
processes. Industrial Engineering Chemistry Research 43, 72907302.
Kotoulas, C., Krallis, A., Pladis, P., Kiparissides, C., 2003. A
comprehensive kinetic model for the combined chemical and thermal
polymerization of styrene up to high conversions. Macromolecular
Chemistry Physics 204, 13061314.
Krallis, A., Kotoulas, C., Papadopoulos, S., Kiparissides, C., Bousquet, J.,
Bonardi, C., 2004. A comprehensive kinetic model for the free-radical
polymerization of vinyl chloride in the presence of monofunctional
and bifunctional initiators. Industrial Engineering Chemistry Research
43, 63826399.
Krieger, I.M., 1972. Rheology of monodispersed lattices. Advances in
Colloid and Interface Science 3, 111127.
Kumar, S., Ramkrishna, D., 1996. On the solution of population balance
equations by discretizationI. A xed pivot technique. Chemical
Engineering Science 51, 13111332.
Maggioris, D., Goulas, A., Alexopoulos, A.H., Chatzi, E.G., Kiparissides,
C., 2000. Prediction of particle size distribution in suspension
polymerization reactors: effect of turbulence nonhomogeneity.
Chemical Engineering Science 55, 46114627.

Narsimhan, G., Gupta, G., Ramkrishna, D., 1979. A model for translational
breakage probability of droplets in agitated lean liquidliquid
dispersions. Chemical Engineering Science 34, 257265.
Okaya, T., 1992. General properties of polyvinyl alcohol in relation to its
applications. In: Finch, C.A. (Ed.), Polyvinyl Alcohol Developments.
Wiley, New York, pp. 130.
Prigogine, I., Marechal, J., 1952. The inuence of differences in molecular
size on the surface tension of solutions. Journal of Colloid Science 7,
122127.
Shinnar, R., 1961. On the behavior of liquid dispersions in mixing vessels.
Journal of Fluid Mechanics 10, 259277.
Shinnar, R., Church, J.M., 1960. Predicting particle size in agitated
dispersions. Industrial Engineering Chemistry Research 35, 253256.
Siow, K.S., Patterson, D., 1973. Surface thermodynamics of polymer
solutions. Journal of Physical Chemistry 77 (3), 356368.
Sovova, H., 1981. Breakage and coalescence of drops in a batch
stirred vessel. II. Comparison of model and experiments. Chemical
Engineering Science 36, 15671573.
Vermeulen, T., Williams, G.M., Langlois, G.E., 1955. Interfacial area in
liquidliquid and gasliquid agitation. Chemical Engineering Progress
51, 85F95F.
Villalobos, M.A., Hamielec, A.E., Wood, P.E., 1993. Bulk and suspension
polymerization of styrene in the presence of n-pentane. An evaluation of
monofunctional and bifunctional initiation. Journal of Applied Polymer
Science 50, 327343.
Ward, J.P., Knudsen, J.G., 1967. Turbulent ow of unstable liquidliquid
dispersions: drop sizes velocity distributions. A.I.Ch.E. Journal 13,
356371.
Yang, B., Takahashi, K., Takeishi, M., 2000. Styrene drop size and size
distribution in an aqueous solution of poly(vinyl alcohol). Industrial
Engineering Chemistry Research 39, 20852090.
Yuan, H.G., Kalfas, G., Ray, W.H., 1991. Suspension polymerization.
JMSReviews in Macromolecular Chemical Physics C31, 215299.
Zerfa, M., Brooks, B.W., 1996a. Prediction of vinyl chloride drop sizes
in stabilized liquidliquid agitated dispersion. Chemical Engineering
Science 51 (12), 32233233.
Zerfa, M., Brooks, B.W., 1996b. Vinyl chloride dispersion with relation
to suspension polymerization. Chemical Engineering Science 51 (14),
35913611.

You might also like