Chemical Engineering Science 60 (2005) 5574 – 5589

www.elsevier.com/locate/ces

Characteristic intervals in suspension polymerisation reactors: An

experimental and modelling study
Fatemeh Jahanzada , Shahriar Sajjadib , Brian W. Brooksa,∗
a Department of Chemical Engineering, Loughborough University, Loughborough, Leicestershire, LE11 3TU, UK
b Division of Engineering, King’s College London, London, WC2R 2LS, UK

Received 30 November 2004; received in revised form 15 April 2005; accepted 15 April 2005
Available online 29 June 2005

Abstract
Suspension polymerisation of methyl methacrylate was carried out as a model to elaborate on the evolution of particle size average and
distribution in the course of polymerisation. Four characteristic intervals in the evolution of particle size were identified as: transition,
quasi-steady-state, growth, and identification stages. The effects of stabiliser and initiator concentrations, monomer hold up, reaction
temperature, and agitation speed on the characteristic intervals, as well as the kinetics of polymerisation, were examined. The transition
stage, which has been totally ignored in the literature, was found to have significant effect on the evolution of particle size. The transition
stage is shortened by increasing the rate of polymerisation in the drops (either by increasing initiator concentration or using a higher
reaction temperature). Increasing the impeller speed and stabiliser concentration will also lead to a shorter transition period. However,
the delayed adsorption of the stabiliser on the surface of drops will prolong the transition stage. It is shown that the occurrence of the
quasi-steady state depends on the polymerisation conditions. The quasi-steady state occurs only if the balance between drop break up
and coalescence can be maintained. This requires a high rate of drop break up within a period of time during polymerisation (i.e., a low
rate of polymerisation in the drops by using a low initiator concentration and reaction temperature, a high agitation speed and a high
stabiliser concentration). The mechanisms underlying the growth stage are explained in terms of the overall rates of drop break up and
coalescence in the course of polymerisation reactions. It is also shown that the onset of growth stage cannot be defined in terms of a
critical conversation or viscosity, and it depends on the polymerisation conditions including mixing. The growth stage occurs if drops are
not sufficiently stable against both break up and coalescence. The onset of the growth stage is advanced with a decrease in the rate of drop
break up (e.g., decreasing agitation speed and stabiliser concentration). The growth stage can be totally eliminated from a polymerisation
process if dispersions with a static steady state can be formed. That requires a high concentration of stabiliser, or a low concentration of
monomer, to be used. A population balance model, which included the transition stage and the delayed adsorption of the stabiliser, was
developed that is capable of predicting the evolution of drop size in the suspension polymerisation.
䉷 2005 Elsevier Ltd. All rights reserved.

Keywords: Suspension polymerisation; Characteristic intervals; Break up; Coalescence; Particle size; Population balance modelling; Methylmethacrylate

1. Introduction distribution are two of key factors to guarantee the quality

Suspension polymerisation is widely used in industry to
produce high value-added particulate products such as chro-
matographic separation media, ion exchange resins and sup-
ports for enzyme immobilisation. For all these products the
average particle size and the breadth of the particle size
∗ Corresponding author. Tel.: +44 1509 222 510; fax: +44 1509 223 923.
E-mail address: b.w.brooks@lboro.ac.uk (B.W. Brooks).
0009-2509/$ - see front matter 䉷 2005 Elsevier Ltd. All rights reserved.
doi:10.1016/j.ces.2005.04.063
of the product. In comparison with the number of studies
appearing in the literature on the kinetics of polymerisa-
tion and/or properties of final polymer particles, less atten-
tion has been paid to the study of evolution of particle size
and particle size distribution. The main reason for this is
perhaps the complexity of the phenomena that determines
particle size.
In principle, a balance between the rate of drop break up
and coalescence determines the size of drops in suspension
 F. Jahanzad et al. / Chemical Engineering Science 60 (2005) 5574 – 5589
polymerisation. Therefore, the drop size is a strong func-
tion of several parameters such as densities and viscosities
of continuous and dispersed phases, interfacial tension, type
and concentration of suspending agent, dispersed-phase hold
up, type of impeller and stirring speed, as well as the kinetics
of polymerisation. Suspension polymerisation reactions gen-
erally produce particles with a broad or bimodal size distri-
bution. Bead particles, which are the intended product from
a suspension polymerisation, with diameters in the range of
10
m to 5 mm, are usually accompanied with unintended
smaller particles. Many aspects of suspension polymerisa-
tions have been reviewed in the literature (Brooks, 1990;
Dowding and Vincent, 2000; Vivaldo-Lima et al., 1997).
When two immiscible phases are brought into contact by
agitation, drops are formed and their size depends on many
variables including the agitation speed. An increase in the
agitation speed enhances the rate of drop break up and thus
favours the formation of smaller drops. However, at very
high agitation speeds the drop size may increase with ag-
itation speed due to an increase in the rate of coalescence
because of the very large surface area of drops and reduced
effectiveness of the suspending agent molecules on the in-
terface. A U-shape dependence of the mean drop size on
the agitation speed has been reported by several investiga-
tors (Chatzi and Kiparissides, 1994; Johnson, 1980; Zhou
and Kresta, 1998). Yang et al. (2001) applied a variable
agitation-speed method during the suspension polymerisa-
tion of styrene to limit the drop size enlargement during the
growth stage and to control the final polymer particle size.
One of the crucial factors in the control of particle size in
suspension polymerisation reactors is the type and the con-
centration of the stabiliser(s) used. The majority of stabilis-
ers used in this process are either water-soluble polymeric
materials or inorganic particles. Suspending agents adsorb
at the monomer/water interface and, thus, enhance the sta-
bility of drops against coalescence. Water-soluble stabilisers
also facilitate the drop break up by reducing the interfacial
tension. Partially hydrolysed poly vinyl acetate or poly vinyl
alcohol (PVA) is one of the polymeric water-soluble stabilis-
ers used in suspension and emulsion polymerisation pro-
cesses. A number of studies have been reported on the sus-
pension polymerisation with PVA as stabiliser (e.g., He et al.,
2002; Konno et al., 1982; Mendizabal et al., 1992; Zerfa
and Brooks, 1998). It has been shown that the best PVA
grade for using as a stabilising agent in suspension poly-
merisation is the one with a degree of hydrolysis of 80–90%
and molecular weight of above 70,000 (Castellanos et al.,
1991; Mendizabal et al., 1992). This grade of PVA forms a
thicker and stronger layer on the water/monomer interface
and has less tendency to be desorbed. Using PVAs with a
very low degree of hydrolysis (< 80%) or low molecular
weight results in coagulation and agglomeration of poly-
mer particles and formation of polymer lumps, while using
PVAs with a very high degree of hydrolysis (> 90%) re-
sults in unstable dispersions leading to formation of shape-
less polymer bulk (Mendizabal et al., 1992). Increasing the
5575
concentration of stabiliser in general, and PVA in partic-
ular, has been found to increase its ability to stabilise the
dispersion due to steric and Marangoni effects. In both sim-
ple liquid–liquid dispersion and suspension polymerisation,
a smaller mean drop/particle size and sharper drop/particle
size distribution have been obtained with increasing PVA
concentration (Chatzi and Kiparissides, 1994; Konno et al.,
1982; Lazrak et al., 1998).
It is always desired in industry to increase the output of
manufacturing units at minimum cost. In suspension poly-
merisation processes this target is usually met by increasing
the monomer hold up. However, the incorporation of more
monomer destabilises the particles and also increases the
possibility of thermal runaway. Therefore, a careful exam-
ination of interaction of parameters involved in suspension
polymerisation reactors is required. Increase in monomer
hold up enhances the dispersion viscosity and thus damps
the turbulence of the system, and also increases the colli-
sion frequency of dispersed drops. All these result in larger
polymer particles (Kalfas et al., 1993; Konno et al., 1982;
Lazrak et al., 1998).
One important aspect of monomer-hold up can be seen in
association with monomer solubility in the water phase. Wa-
ter solubility of monomer is one of the factors that can affect
the kinetics of the reaction and the quality of the product.
When the solubility of the monomer in the aqueous phase
is negligible (as it is for styrene, for example) the kinetics
of suspension polymerisation can be described as similar
to that of the corresponding homogeneous bulk polymeri-
sation. However, in the case of monomers with a moderate
to high water solubility, some of the monomer resides in
the water phase and does not get involved in the polymeri-
sation reaction. As polymerisation proceeds in the drops,
monomer transfers from the aqueous phase into the drops
to replace the monomer that has polymerised. This effect
causes some deviations from the kinetics of bulk polymeri-
sation. Kalfas et al. (1993) carried out a series of suspen-
sion homo- and copolymerisations of monomers with differ-
ent water solubility. Their results showed that mass transfer
limitations in suspension polymerisation of partially water-
soluble monomers (methyl methacrylate, vinyl acetate and
acrylonitrile) at low monomer-to-water ratios is consider-
able and leads to a lower final conversion in comparison
with higher monomer-to-water ratios. Water solubility of
monomer also causes the formation of emulsion particles in
the aqueous phase. These particles are the result of the ho-
mogenous nucleation occurring in the water phase and be-
cause of segregation of radicals in these particles, the poly-
mer formed usually has a higher molecular weight than the
polymer formed in larger particles of the suspension poly-
merisation (Cunningham, 1999).
The variation in reaction temperature will affect the prop-
erties of both monomer and oil phases in a suspension poly-
merisation. A 10◦ C temperature increase results in a two-
to threefold increase in the rate of polymerisation (Odian,
1991). The effects of temperature increase on the properties
5576 F. Jahanzad et al. / Chemical Engineering Science 60 (2005) 5574 – 5589
of the water phase containing a stabiliser are complex. These
mainly depend on how the stabiliser distributes between the
two phases at higher temperatures. It seems that for PVA
there is an overall trend of decreasing drop stability with
increasing temperature, though the magnitude of reduction
is quite different for different types of PVA (Lazrak et al.,
1998). It has been shown that this effect is not due to a de-
crease in viscosity of the continuous phase with increasing
temperature (He et al., 2002).
In this paper the characteristic intervals of a typical sus-
pension polymerisation, in terms of average drop size, are
clarified. The effects of variations in process and formula-
tion variables on these intervals will be discussed. Methyl
methacrylate (MMA) has been selected as a “model”
monomer. A mathematical model was developed to elab-
orate on the kinetics of polymerisation as well as the
evolution of drop size average and distribution.
2. Experimental
2.1. Materials and experimental set up
MMA (analytical grade from Aldrich) was distilled at re-
duced pressure to remove inhibitors. PVA (Mw = 85, 000 −
146, 000, degree of hydrolysis = 87–89%, from Aldrich)
and lauroyl peroxide (LPO) (97%, from Aldrich) were used
as stabiliser and initiator, respectively, without any further
purification. Distilled water was used as the continuous
phase. The experiments were carried out using a 1-l jack-
eted glass reactor (internal diameter 10 cm) equipped with
four 90◦ baffles. A four-bladed flat turbine impeller with a
diameter of 5 cm was used for agitation. The temperature
of the vessel content was controlled within ± 0.5◦ C of the
desired reaction temperature by pumping water with ap-
propraite temperature through the jacket. Nitrogen purging
was carried out for 30 min before the monomer phase was
added to the aqueous phase. Samples were withdrawn, at
the desired time intervals, from the reaction vessel using
a hypodermic syringe to measure the monomer conversion
and drop/particle size distribution. In order to avoid irrepro-
ducibility, the sampling point was fixed at 2 cm away from
the impeller shaft and 3 cm below the suspension level.
Experiments (unless otherwise stated) were carried out at
70◦ C, with an agitation speed of NI =500 rpm, the monomer
volume fraction was
d = 0.20 and PVA and LPO concen-
trations were 1.0 g/l (based on water phase) and 1.0 wt%
(based on monomer phase), respectively.
2.2. Measurements
2.2.1. Monomer conversion
Monomer conversion was measured gravimetrically.
Small quantities of the dispersion were withdrawn from the
reactor and transferred into the small weighed aluminium
foil dishes. The samples, contained in the dishes, were
weighed and then quenched by adding cold acetone. The
dishes were kept in the vacuum oven at 50◦ C for at least
18 h. Monomer, acetone, and water were driven off by dry-
ing to constant weight. Then the dishes were weighed again
to calculate the monomer conversion using a simple mass
balance.
2.2.2. Interfacial tension
Interfacial tensions between MMA and aqueous solutions
of PVA were measured using a Du Nouy ring tensiometer.
Measurement was repeated at 25◦ C three times for each sam-
ple. The average value was taken as the interfacial tension.
2.2.3. Monomer concentration in the aqueous phase
A gas–liquid chromatograph (Pye Unicam 304-FID GLC)
was used to determine the equilibrium MMA concentration
in the aqueous phase. A general-purpose capillary column,
wall coated-fused silica-CP-SIL5, was used together with a
flame ionisation detector (FID). The carrier gas was helium.
Methanol was used as the internal standard and the corre-
sponding calibration curve was made using standard meth-
ods. Samples taken from the reactor at appropriate intervals
were kept in capped test tubes and centrifuged for phase sep-
aration. Samples from the aqueous phase were withdrawn
with a syringe and weighed. A known amount of a Methanol
solution was added to each sample before injecting to the
GLC. Three measurements were carried out for each sample
to assure about the reproducibility of the measurement.
2.2.4. Drop/particle size distribution (DSD/PSD)
A particle sizer (Coulter LS130), based on Fraunhofer
diffraction, was used for DSD/PSD measurements. Prior to
any measurement, the sample cell of the instrument was
thoroughly cleaned and filled with an aqueous solution of
sodium lauryl sulphate (0.2 g/l) to prevent coalescence of
the droplets during measurement. At desired time intervals,
a drop of the reactor content was immediately transferred
to the sample cell. The cell content was mixed gently by a
magnetic stirrer, provided in the cell, to prevent coalescence
of droplets during the measurement. Sampling and size mea-
surements were continuously carried out in the course of the
process.
Drop size measurements required the dispersion to be
highly diluted in water whereupon the monomer in the
monomer drops or unreacted monomer inside the particles
may diffuse out of the particles. Water-solubility of MMA
at 25◦ C is around 15 g/l. A simple mass balance shows that
dilution of samples with 102 times the volume of water
would be enough to transfer almost all MMA monomer out
of the drops, leading to their shrinkage. To avoid underesti-
mation of drop size, the distilled water used for the dilution
of the samples, was saturated with the monomer prior to
preparation of the surfactant solution.
The particle sizer Coulter LS130 has 84 channels (bins)
which accommodate size ranges that change logarithmically.
 F. Jahanzad et al. / Chemical Engineering Science 60 (2005) 5574 – 5589 5577

These channels cover the size range of 0.43–822
m. The 100
output of the particle sizer is in terms of volume of drops
in each bin size, which can be easily transformed to volume
80
frequency (f ) defined as quasi identification
v(d) teady state
f (d) =
d , (1) 60
v(d)

D 32 (µm)
max
dmin

where v(d) is the volume of drops with a diameter between

40
ln(d) and ln(d)+  ln(d). dmin and dmax are the diameters of
the smallest and the largest drops measured by the particle
sizer, respectively. The normalised distribution in terms of 20
diameter can be calculated using the following equation: transition growth

f (d) 0
F (d) = , (2)
d(d) 0 30 60 90 120
(a) time (min)
d(d) is the variable bin size between ln(d) and ln(d) +
 ln(d). It should be noted that the bin size is constant only
if a logarithmic scale is considered. F (d) has the dimension 1.E+07
of inverse length unit (
m−1 ) and satisfies the normalisation
Rc
condition:
1.E+04
d
Rate (1/sec.cm3)
max
F (d) · d(d) = 1. (3) Rb
dmin
1.E+01
The Sauter mean diameter (D32 ) was calculated using the
following equation:

dmax 1.E-02
d f (d)
D32 =
d min . (4)
dmin (f (d)/d)
max

1.E-05
0 30 60 90 120
3. Results and discussion (b) time (min)

3.1. Typical behaviour Fig. 1. Characteristic intervals in the evolution of mean drop size (a) and
theoretical predictions of overall rates of break up and coalescence (b) in
MMA suspension polymerisation at the basic conditions.
Before discussing the results in detail, it is helpful to con-
sider the characteristic intervals of a typical suspension poly-
Table 1
merisation, which were identified in this research (Jahanzad, Characteristics of different stages in a suspension polymerisation
2004) as the transition, quasi-steady-state, growth and iden-
tification stages. The size of drops in suspension polymeri- Stage Feature Result
sation, in principle, is determined by a balance between the Transition Rbreak-up  Rcoalescene Decrease in drop size
overall rates of drop break up (Rb ) and coalescence (Rc ). Initial drops formed
Depending on the magnitudes of these rates, different stages Quasi-steady Rbreak-up ≈ Rcoalescene Almost constant drop
can be observed in the course of polymerisation. Fig. 1a il- state size
Growth Rbreak-up
Rcoalescene Increase in drop size
lustrates typical behaviour of time evolution of drop size in Identification Rbreak-up = Rcoalescene = 0 Final particle size
suspension polymerisation. Fig. 1b shows the correspond- achieved
ing theoretical predictions of the overall rates of drop break
up and coalescence (per cm3 of the reactor volume) during
different intervals. The expressions used to calculate Rb and in size until they reach an almost constant average size. This
Rc have been given in Section 4 (Model Development). The stage has not received any detailed attention in the literature.
characteristics of each interval (or stage) are summarised in Quasi-steady-state stage: during this stage the rates of
Table 1. The significance of each stage is as follows. drop break up and drop coalescence are almost balanced
Transition stage: during this stage the drop size decreases leading to a quasi-steady-state drop size and distribution. At
exponentially and drop size distribution narrows dramati- the quasi-steady state, coalescence removes as many drops
cally due to a higher rate of drop break up in comparison from the dispersion as are made via the break-up mechanism.
with drop coalescence. During this stage drops are reduced Therefore, this steady state is often of a dynamic nature. The
5578 F. Jahanzad et al. / Chemical Engineering Science 60 (2005) 5574 – 5589
existence of this stage in suspension polymerisation has been
questioned by some investigators (Machado et al., 2000).
However, as we will show later, the occurrence of this stage
is not always possible and depends on the polymerisation
conditions.
Growth stage: during this stage the rate of drop break
up falls considerably below the rate of drop coalescence
(Rc  Rb ) due to the viscosity build up in drops, leading to
the drop enlargement and size distribution broadening. This
stage has been traditionally called a “sticky stage” or “tacky
stage” in the literature perhaps because of the impression
that drop coalescence is boosted with the drop tackiness as
a result of increasing drop viscosity. However, the present
authors believe that this may be a misleading title for this
stage. This can be better explained if we look at the mecha-
nism of drop coalescence. Before two drops can participate
in a coalescence event, they have to approach each other,
and must remain together long enough for the intervening
liquid film to drain to its critical thickness, whereupon film
rupture occurs. In fact, the rate of drop coalescence is con-
trolled by liquid drainage in the film between approaching
droplets, and more significantly, by the rigidity of the two
corresponding drop/water interfaces. Increasing the viscos-
ity of dispersed phase in mobile-interface systems leads to
partial immobility of the interface and so decreases the prob-
ability of coalescence (Liu and Li, 1999). This implies that,
with increasing viscosity, particles are more stable against
coalescence than in the earlier stages. Viscosity build up in
the polymerising drops diminishes both the rates of drop
break up and coalescence. With increasing drop viscosity,
a point is reached where drops cannot be easily broken up
but they can still undergo coalescence. As a result, drops
start growing by coalescence. The occurrence of the growth
stage depends on the drop viscosity, drop stability and mix-
ing conditions as will be discussed later.
Identification stage: this stage starts when the liquid–liquid
dispersion fully transforms to a solid–liquid suspension
with constant particle size average and distribution. At this
stage, drops have a very high viscosity and behave like solid
particles so they cannot coalesce with each other and keep
their identity for the remainder of the process. The onset of
this stage has been called the “Identification point”. This
point is mainly determined by the glass transition tempera-
ture of the reaction mixture, which is a strong function of
polymer phase fraction in the particles, and is independent
of the mixing conditions.
Having clarified the characteristic intervals, in the follow-
ing section we evaluate the effects of different process and
formulation variables on each stage.
3.2. Initiator concentration
The kinetics of polymerisation can play an important role
in the evolution of particle size and size distribution in
suspension polymerisation. The viscosity of drops and its
1
4.0 wt%
1.0 wt%
0.8
0.25 wt% LPO
0.6
conversion
0.4
0.2
0
0 50 100 150 200
time (min)
Fig. 2. Conversion-time variations for MMA suspension polymerisation
with different LPO concentrations.
variation with time is, in fact, a reflection of the rate of poly-
merisation inside the drops. Fig. 2 shows the conversion-
time variations for the runs with different LPO concentra-
tions. The overall rate of polymerisation increased with LPO
concentration. Generally the onset of the gel effect, when
the rate of polymerisation increases significantly, depends
on two factors; the volume fraction of polymer in the mix-
ture (or conversion) and the molecular weight of the poly-
mer produced (Marten and Hamielec, 1979). Fig. 2 indicates
that for the LPO concentrations of 0.25, 1.0, and 4.0 wt% the
onset of the gel effect happened at the conversions of 0.35,
0.40, and 0.45, respectively. By increasing the LPO concen-
tration shorter polymer chains were produced and the gel
effect was postponed to higher conversions. The rate of the
polymerisation reaction during the gel effect increased sig-
nificantly with LPO concentration, as was expected. Gener-
ally, a higher final conversion was achieved with increasing
LPO concentration.
Fig. 3a illustrates the variations of drop Sauter mean di-
ameter (D32 ) with reaction time (t). This figure shows that
the largest decrease in the average size of drops during the
transition stage was obtained for the lowest LPO concentra-
tion used. This is due to a lower rate of viscosity build up in
the drops containing a lower concentration of LPO. A high
viscosity of drops, associated with a higher rate of polymeri-
sation at high LPO concentration, damps the rate of drop
break up so that the rates of drop break up and drop coales-
cence are balanced earlier. This also results in a larger drop
size in the beginning of the quasi-steady-state stage. During
this stage, drops underwent a slight size increase with time.
Obviously, the quasi-steady state lasted longer for the lower
LPO concentration.
Note that time is not a good representative of the sta-
tus of the polymerising drops when they contain different
concentrations of the initiator. The evolution of D32 can
 F. Jahanzad et al. / Chemical Engineering Science 60 (2005) 5574 – 5589 5579

120 1

1.0 wt%
0.8
90 4.0 wt%

0.6

conversion
PVA (g/l)
D 32 (µm)

60 0.1

0.25 wt% LPO 0.4 0.2

0.5
30
0.2 1.0

4.0
0 0
0 50 100 150 200 0 30 60 90 120 150
(a) time (min)
time (min)
120
Fig. 4. Conversion-time variations for MMA suspension polymerisation
0.25 wt% LPO with different PVA concentrations.

90
400
0.1 g/ l
1.0 wt% 0.2 g/ l
D 32 (µm)

60
300

4.0 wt%
D 32 (µm)

30 200 0.5 g/ l

1 g/l
0 100
0 0.2 0.4 0.6 0.8 1 4 g/ l
(b) conversion

Fig. 3. Variations of Sauter mean diameter with time (a) and conversion 10 g/ l PVA
(b) for different LPO concentrations. 0
0 30 60 90 120 150
time (min)
be better explained if the comparison is made in terms of
monomer conversion as shown in Fig. 3b. This figure reveals Fig. 5. Variations of Sauter mean diameter with time for different PVA
that growth stage is delayed with increasing LPO concentra- concentrations.
tion. This is because low molecular weight polymer chains
are produced with increasing initiator concentration. As a
result of a low viscosity, the balance between drop break 3.3. Stabiliser concentration
up and drop coalescence can be maintained up to a higher
conversion. Fig. 4 shows the conversion-time variation for runs with
At low initiator concentrations, the rate of polymerisation different PVA concentrations. All the data points fall on the
is slow and polymerising drops are subject to more collision same curve confirming that PVA concentration does not af-
before the identification point can be achieved. As a result, fect the rate of polymerisation. This also indicates that pos-
the final size of particles obtained from the polymerisations sible grafting to PVA molecules does not have a significant
increased with decreasing LPO concentration. A recent work effect on the kinetics of polymerisation (Okaya et al., 1999).
on the effect of initiator concentration in MMA suspension Fig. 5 illustrates the variations of D32 with time for dif-
polymerisation with benzoyl peroxide as initiator showed ferent PVA concentrations. The rate of drop break up is a
similar results (Lazrak et al., 1998). strong function of interfacial tension, particularly during the
5580 F. Jahanzad et al. / Chemical Engineering Science 60 (2005) 5574 – 5589
16
12
interfacial tension (dyn/cm)
8 70 min
0
0 2 4 6
PVA concentration (g/ l)
Fig. 6. Variations of interfacial tension between MMA and aqueous PVA
solution with PVA concentration at 25◦ C.
transition stage when the viscosity of drops is still very low.
As a result, a shorter transition stage and a lower steady-state
D32 were achieved with increasing PVA concentration. The
difference, however, became less important with increasing
PVA concentration. This is confirmed by a diminishing dif-
ference in the interfacial tension, with PVA concentration,
as shown in Fig. 6. Values of interfacial tension in the reac-
tor (at 70◦ C) will be slightly different from those shown in
Fig. 6.
The occurrence of a steady state in a suspension polymeri-
sation is not always certain, and its existence depends on
the polymerisation kinetics, mixing condition and the type
and concentration of the stabiliser used. With the PVA con-
centration of 0.1 g/l, the steady state was never reached and
drops started growing, soon after the minimum drop size was
reached, until the mass coagulation of particles occurred af-
ter 20 min. For the PVA concentrations of 0.2 and 0.5 g/l, the
quasi-steady state lasted until about 30 and 50 min, respec-
tively, followed by the growth stage. For the PVA concen-
trations of 1.0 g/l and more, the polymerisations produced
almost the same steady-state drop size until about 60 min.
Note that if drops are reduced to a size that cannot be fur-
ther broken by agitation and if they are also sufficiently pro-
tected against drop coalescence, then drops do not undergo
any transformation during mixing and a static steady state is
established. This implies that with increasing the stabiliser
concentration, the nature of the steady-state stage gradually
shifts from a dynamic one towards a static one (Jahanzad et
al., 2004). Fig. 5 shows that the increase in drop size in the
growth stage is very small for the run with the PVA concen-
tration of 10.0 g/l.
The onset of the growth stage was found to be influenced
by the concentration of PVA. At a low concentration of PVA,
the interfacial tension is sufficiently large (see Fig. 6) to
0.03
1 min
20 min
40 min
0.02
F (d) (1/µm)
65 min
0.01
0
0 50 100 150 200
8 10
diameter (µm)
Fig. 7. Time evolution of particle size distribution in MMA suspension
polymerisation using 1.0 g/l PVA.
allow for a sharp reduction in the rate of drop break up with
increasing viscosity. Note that the poor coverage of drops
with PVA, as well as delayed adsorbtion (discussed later in
Section 4.1), enhance the drop instability. These cause the
drops to start growing at a low conversion. The higher the
PVA concentration, the less was the particle enlargement
during the growth stage. Drops with higher stabiliser cover-
age are less susceptible to coalescence after a collision. As
a result, a sharper rise in drop sizes is observed during the
growth stage at a lower PVA concentration. The results show
that no critical conversion or critical viscosity can be defined
to mark the onset of a growth stage. The critical conversion
or viscosity defined in the literature (Villalobos et al., 1993)
for the onset of this stage, thus, may be only applicable for
fixed polymerisation conditions and cannot be generalised.
It is also clear from the results that the gel effect cannot be
considered as the sole cause of the growth stage.
The onset of the identification stage was not affected by
PVA concentration implying that it is only determined by the
glass temperature of the reaction mixture. However, some
limited flocculation was observed at this point at lower PVA
concentrations.
Fig. 7 shows the time evolution of particle size distribu-
tion for the run with PVA concentration of 1.0 g/l. A small
peak appeared in each distribution at the lower end of the
size range, with volume fraction of about 0.05, but those
peaks have not been shown in Fig. 7 for more clarity. Their
significance has been discussed elsewhere (Jahanzad et al.,
2004). Fig. 7 shows that the drop size distribution initially
narrowed and that drop sizes changed to lower values with
time (transition stage). At later times, drop sizes showed a
quasi-steady state and then shifted to a higher values dur-
ing the growth stage. The PSD remained unchanged after
70 min when the identification point was reached.
 F. Jahanzad et al. / Chemical Engineering Science 60 (2005) 5574 – 5589
200
220 rpm
150 300 rpm
D 32 (µm)
100
500 rpm
50
0
0 50 100 150 200
time (min)
Fig. 8. Variations of Sauter mean diameter with time in MMA suspension
polymerisation at different agitation speeds.
3.4. Agitation speed
Three different agitation speeds were used. The results
indicate that agitation speed does not have a significant ef-
fect on the rate of polymerisation, as was expected (not
shown). However, the size of drops was significantly af-
fected by the agitation speed. Fig. 8 shows the variations of
D32 with time. Generally, the approach to the steady state
was shortened with increasing agitation speed as a result of
a more extensive drop break up. Similarly, the duration of
the steady-state stage was prolonged at high impeller speed.
For NI =220 rpm, the steady-state stage did not exist and the
particle growth (growth stage) started just after a minimum
drop size was reached. The onset of the growth stage oc-
curred approximately at the conversions of about 0.20, 0.40
and 0.60, for NI = 220, 300, and 500 rpm, respectively. It
can be inferred from the data depicted in Fig. 8 that the con-
version, as well as the viscosity, at which the particle growth
started, are not constant and could vary with the impeller
speed.
Note that the rate of drop break up is a strong function of
the agitation speed. At a low agitation speed, an infinitesimal
increase in the drop viscosity will lead to an imminent reduc-
tion in the rate of drop break up so that the balance between
drop break up and coalescence could not be maintained fur-
ther. This will result in a gradual enlargement of drops with
time. The balance could be better maintained at higher im-
peller speed, due to a more extensive break up, which allows
a quasi-steady state to form. It should be noted that the rate
of drop coalescence increases with agitation speed to main-
tain the balance as more drops are formed. The drop–drop
collision frequency increases with the agitation speed, but
the efficiency of coalescence decreases due to decline in the
5581
1.0
φd
0.025 0.05
0.8
0.10 0.20
0.30 0.40
0.6
conversion
0.4
initial final
stage stage
inter-
0.2
mediate
stage
0.0
0 30 60 90 120
time (min)
Fig. 9. Conversion-time variations for MMA suspension polymerisation
with different monomer volume fractions.
contact time. The net result, however, is an increase in the
rate of drop coalescence with agitation speed.
3.5. Monomer hold up
Suspension polymerisations of MMA were carried out
with the agitation speed of 300 rpm using different monomer
volume fractions (
d ). Fig. 9 shows the conversion-time
variation for this series. The rate of polymerisation can
be better described if the following conversion regions are
considered:
Initial stage: the initial rate of polymerisation shows an
increase with increasing monomer volume fraction. MMA
has a water solubility of 1.6 wt% at 70◦ C. The monomer in
the dispersed phase saturates the water continuous phase be-
fore much polymerisation can take place. This reduces the
amount of monomer available in the monomer drops. Know-
ing that the locus of polymerisation in suspension polymeri-
sation is, in fact, the monomer drops and assuming that the
polymerisation of MMA in the water phase is negligible, the
overall rate of polymerisation is expected to decrease with
decreasing
d . For
d = 0.025, 60% of the initial monomer
is dissolved in the water phase and is absent from the locus
of polymerisation. Obviously, the rate of polymerisation was
mostly affected for this case. Even with
d =0.05, polymeri-
sation started with 30% of the initial monomer charge in the
water phase, and so the rate of polymerisation was conse-
quently affected. With
d values higher than 0.05, the initial
rate of polymerisation became less affected by
d . It is worth
noting that the concentration of initiator in the monomer
drops maybe greater than that indicated by the overall for-
mulation because of partial solubility of the monomer in the
water phase. This may slightly compensate the fall in the
5582 F. Jahanzad et al. / Chemical Engineering Science 60 (2005) 5574 – 5589
0.03
φd
0.20
0.30
0.02
w MMA,aq
0.01
0.00
0 30 60 90 120
time (min)
Fig. 10. Variations of MMA concentration in the aqueous phase with time
in suspension polymerisations with two different MMA volume fractions.
rate of polymerisation due to presence of monomer in the
water phase. However, the rate of polymerisation is always
a decreasing function of
d .
Intermediate stage: this stage, which is manifested by the
gel effect, appears at different conversions depending on
the
d values used. For the higher values of
d , the gel
effect started at a conversion of 0.40, which corresponds to
the onset of gel effect for the MMA polymerisation in bulk
at the same conditions. With decreasing
d , the onset of
gel effect is shifted toward a lower conversion. For
d =
0.025, for example, the gel effect started at the conversion
of 0.20 because a substantial fraction of the monomer was
dissolved in the aqueous phase. It should be noted that the
onset of the gel effect depends on the monomer conversion
inside the drops and not on the overall monomer conversion
in the reactor. This means that for the MMA suspension
polymerisations, it is expected that the onset of the gel effect
appears at the particle conversion of 0.40, independent of
the values for the corresponding overall conversions.
Final stage: generally a lower final conversion was
achieved for the lower
d . This is in agreement with the
results reported by Kalfas et al. (1993). This effect is more
significant as
d decreases to very low values. The limited
conversion is a result of monomer partitioning between
monomer drops and water phase. Though at high conver-
sions the monomer dissolved in the water phase diffuses
into the polymer particles and reacts there, but the diffusion
process is extremely slow.
The monomer concentration in the water phase shows
simultaneous variation with that in the polymer particles.
Fig. 10 shows the time variations in the MMA weight
fraction in the aqueous phase (wMMA,aq ) during the poly-
merisations with
d = 0.20 and 0.30. Both systems behaved
300
φd
0.025 0.05
0.10 0.20
200 0.30 0.40
D 32 (µm)
100
0
0 20 40 60 80 100
time (min)
Fig. 11. Variations of Sauter mean diameter with time in suspension
polymerisations with different MMA volume fractions.
similarly. For both monomer hold ups, MMA concentration
in water gradually decreased from about 15 g/l, the equilib-
rium solubility of MMA in water, by diffusion of monomer
molecules from the water phase into the polymerising drops.
The monomer transport occurs to maintain the thermody-
namic equilibrium between phases as a result of conversion
of monomer to polymer in the polymerising drops. During
the gel effect, when the rate of monomer conversion in-
creased significantly (after about 55 min corresponding to
the conversion of 0.40), a sudden drop in the MMA con-
centration in the water phase is observed. This is a result
of rapid monomer molecules emigration from the water
phase to the polymer/monomer drops in order to maintain
the equilibrium. As the conversion reached higher values,
the MMA concentration in the water phase dropped to a
value as low as 0.2 wt%. Monomer transport at this stage
was quite slow due to solidification of polymer particles.
Fig. 11 compares the variation in D32 with time for dif-
ferent values of
d . Generally, the number of drop encoun-
ters increases with monomer hold up. This, together with a
lower surface coverage by the stabiliser, at a constant sta-
biliser concentration, will boost the rate of drop coalescence
and result in larger drops during the transition stage and the
quasi-steady-state stage. It also gives a steeper rise in the
drop size during the growth stage. The final size of particles
increased with
d , similar to what was shown by other in-
vestigators (Kalfas et al., 1993; Konno et al., 1982; Lazrak
et al., 1998).
According to Fig. 11, the drop size showed only a small
increase with time with lower values of
d , but showed an
appreciable increase during the growth stage for the runs
with higher values of
d . For the lowest values of
d used,
0.025 and 0.05, the conversion at which drop enlargement
 F. Jahanzad et al. / Chemical Engineering Science 60 (2005) 5574 – 5589
1 180
80 C
0.8
70 C
60 C 50 C
0.6
conversion
0.4
0.2
0
0 50 100 150 200 250
time (min)
Fig. 12. Conversion-time variations for MMA suspension polymerisation
at different temperatures.
occurred was around the value of 0.20, which corresponds
to the onset of gel effect for these runs as mentioned earlier.
Except for the two lowest
d used, the conversion range of
the quasi-steady-state stage seems to narrow with increasing

d . This implies that with an enhancement of the rate of
drop coalescence with
d , the balance between drop break
up and coalescence is perturbed and drops enter the growth
phase earlier. The results suggest that the onset of the growth
stage is advanced with increasing
d . For the highest
d
used, 0.40, particle coalescence occurred extensively at x =
0.45 leading to the formation of some polymer lumps. The
measurement of conversion for
d = 0.40 was hampered by
gross coagulation of particles.
3.6. Reaction temperature
Experiments were carried out at different temperatures
with the agitation speed of 300 rpm. The conversion-time
variations for the runs are depicted in Fig. 12. It is evident
from this figure that the rate of polymerisation increases
with temperature. The onset of the gel effect was recorded at
210, 100, 55, and 32 min (corresponding to the conversions
of 0.30, 0.35, 0.40, and 0.45, respectively) for the polymeri-
sations carried out at 50, 60, 70, and 80◦ C, respectively. The
onset of the gel effect occurred at lower conversions with de-
creasing temperature because of the high-molecular weight
polymer formed. The duration of the gel effect, however,
was shortened with increasing temperature. This implies that
the viscosity increase during the gel effect will occur in a
short period of time at higher temperatures (about 5 min for
80◦ C, for example).
Fig. 13 shows the variations of D32 with time for the runs
at different temperatures. Generally the final particle size
increased with the reaction temperature. Similar results have
5583
80 C
70 C
150
60 C
120
D 32 (µm)
90 50 C
60
30
0
0 50 100 150 200 250 300 350
300 350
time (min)
Fig. 13. Variations of Sauter mean diameter with time at different tem-
peratures.
been reported (Lazrak et al., 1998). According to this figure
the reduction in D32 during the transition stage appears to be
more appreciable at the lower temperatures. This is because
of lower rate of polymerisation, which allows drop break up
to be operative for longer times. For the higher temperatures,
such as 80◦ C, the rise in drop viscosity depressed the rate
of drop break up and caused the transition stage to become
shorter by early balancing of the rates of drop break up and
drop coalescence. As a result of a depressed rate of break
up, a higher quasi-steady-state drop size was achieved with
a temperature increase.
Fig. 13 shows that D32 increases slightly during the so
called quasi-steady-state stage for the lower temperatures
(50◦ C and 60◦ C) after a few percent conversion. However,
the quasi-steady state continued for a longer period of time.
At these temperatures, the rate of viscosity build up in drops
with conversion is so high (because of high molecular weight
polymer) that the rate of drop break up is extensively im-
paired as reaction proceeds. As a result, the rate of drop break
up falls behind that of coalescence. This leads to a constant,
though small, rise in the size of drops. At higher reaction
temperatures the quasi-steady state (constant drop size) was
more observable in terms of conversion, though it was very
short in terms of time. The molecular weight of polymer,
and the polymer viscosity, is significantly reduced at high
temperatures. Note that this is a double effect because be-
sides formation of low molecular weight polymers with low
viscosity, polymers are less viscous at higher temperatures.
But the higher rate of monomer conversion at higher tem-
peratures disturbs the balance of break up and coalescence
and shortens the quasi-steady-state stage.
For all runs, significant rise in the drop size was observed
during the growth stage, in association with the gel effect,
where a massive increase in the drop viscosity occurred
5584 F. Jahanzad et al. / Chemical Engineering Science 60 (2005) 5574 – 5589

within a short period of time. The growth was less signif-
icant for the lower temperatures and became steeper with
a temperature increase. If the stabiliser performance does
not vary with temperature, then a greater drop growth is ex-
pected at a lower temperature because of a longer polymeri-
sation time that allows more coalescence. This is in contrast
to the results obtained. Therefore, it seems that the growth of
the drops at higher temperatures could be due to loss of sta-
bility of the polymerising drops. This implies that PVA be-
comes less effective in protecting polymerising drops against
coalescence with increasing temperature, as has been sug-
gested before (Lazrak et al., 1998). The reasons for the loss
of particle stability at higher temperatures are not clear. In-
creased rate of chain transfer and grafting to PVA (Okaya
et al., 1999) and dissolution of PVA in the monomer phase
at higher temperatures could be possible reasons.
distribution, which is considered a normal distribution. n(u)
is the number of drops formed by breaking a drop with
volume u, which is assumed to be 2 (binary break up). The
calculated n(dv , t), is then transformed to volume probable
density function, F (dv , t), using Eq. (2).
The rates of drop break up and coalescence were cal-
culated using the correlations introduced by Alvarez et al.
(1994) for suspension polymerisation of styrene (Eqs. (6)
and (7)). They included the effects of viscoelastic proper-
ties of polymerising drops on the rates of drop break up and
coalescence:
Rate of drop break up:
b(v) = b (v) exp[−b (v)]. (6.1)
kb 1/3
b (v) = 2/3
, (6.2)
(1 +
d )dv

4. Model development b (v) = ab (dv ), (6.3)

6 1
To date, a number of mathematical models have been re- (dv ) = + , (6.4)
Re(dv )[1 + Re(dv )Ve (dv )] W e(dv )
ported for suspension polymerisation systems. These models
 
consider the effects of viscoelastic properties of polymeris- 1− 1
Ve (dv ) =  exp − − ,
ing drops on the rates of drop break up and coalescence 2Y0 Re(dv ) 12
(Alvarez et al., 1991,1994; Chen et al., 1999; Machado et al.,
2d
2000) as well as the nonhomogeneity of the turbulent flow Y0 = , (6.5)
(Maggioris et al., 1998, 2000; Vivaldo-Lima et al., 1998). d Ed dv2
A mathematical model was developed in this research us- 
= 1 − 48Y0 ,
ing the population balance modelling. The tank Reynolds
Y0
numbers for all conditions used were always above 15,000 =
(1 − )
which implies that a fully developed turbulent flow was dom-  

inant in the system. A homogeneous turbulent flow during (1 − )2 1−

× 1+ − exp − . (6.6)
the suspension polymerisation was assumed. A brief intro- 1+ 2Y0 Re(dv )
duction to the model is presented here. For details, readers
are referred to Jahanzad (2004). In the developed model the Rate of drop coalescence:
evaluation of drop volume distribution has been carried out
by numerical solution of the population balance equation c(v, u) = c (v, u) exp[−c (v, u)], (7.1)
(PBE) for dispersed drops, which is an integro-differential 
1/3 2/3 2/3
equation (Eq. (5)), to predict the evolution of drop size dur- c (dv , du ) = kc (dv2 + du2 ) (dv + du ), (7.2)
ing the polymerisation. 1 +
d
 dv max c (v, u) = ac (dvu ) − bc /W e(df ),
dn(dv , t)
= (/2)(6)2/3 v 2/3 b(u)(u, v)n(u) −1
dvu (v, u) = (dv−1 + du−1 ), df = (dv3 + du3 )/dvu
2
. (7.3)
dt dv
× n(du , t) d(du ) − b(v)n(dv , t) + v 2/3 where  and  are the frequency and the efficiency of
 dv/2
the break up (with subscript b) or coalescence (with sub-
× c(v − u, u)n(dv−u , t)n(du , t) d(du )
dv min script c).  is the average energy dissipation rate of the sys-
 dvmax−v tem. Re(d) and W e(d) are the Reynolds and Weber num-
− n(dv , t) c(v, u)n(du , t) d(du ), (5) bers of a drop with diameter d, respectively. Ed ,
d and
dv min
d are Young’s elasticity modulus, viscosity and density of
where n(dv , t) is the number of drops with diameter dv the dispersed phase, respectively. Ed is considered as a lin-
(and volume v) at time t, dv min and dv max are the minimum ear function of polymer–monomer conversion, Ed = xE p ,
and maximum drop diameters, respectively, which can exist where Ep is Young’s elasticity modulus of pure polymer
in the dispersion. b and c are the rates of drop break up (=3.2 × 1010 g/cm s2 , Ferry, 1980). kb , ab , kc , ac and bc are
and coalescence, respectively. (u, v) is the probability of adjustable parameters.
forming a drop with volume v from break up of a drop The method of moments was used for prediction of the
with volume u or, in other words, the daughter drop volume kinetics of MMA bulk polymerisation (Chiu et al., 1983)
 F. Jahanzad et al. / Chemical Engineering Science 60 (2005) 5574 – 5589 5585

and gel and glass effects were predicted by the correlations Table 2
proposed by Baillagou and Soong (1985) for bulk polymeri- Fitted values for the adjustable parameters of the model
sation of MMA. Monomer and stabiliser partitioning were Adjustable parameter Fitted value
modelled using the modified Flory–Huggins theory for poly-
kb 1.5
mer solutions (Kalfas and Ray, 1993) and the Langmuir-
ab 0.03
type interfacial adsorption (Adamson, 1976), respectively. kc (1/cm3 ) 3 × 10−6
Variations of drop viscosity with monomer conversion was ac 1 × 10−5
estimated using the correlation presented by Baillagou and bc 3.0
Soong (1985).
The set of differential equations for the moments of poly-
mer molecular weight have been solved by a Runge–Kutta 100
4th order method. The PBE (Eq. (5)) was solved numerically
1
for n(dv , t). The integro-differential PBE was transformed
into a system of differential equations by discretising the 0.8
80

θ/θeq
range of drop diameters according to the method of classes 0.6
(Chatzi and Kiparissides, 1992). The size range was divided 0.4
into a number of equally spaced intervals. Up to 100 points 60 0.2

were used for the numerical integration. A Runge–Kutta 4th
order method was also used for solving the differential equa-
tions over the time. The integral terms in the PBE were ap-
proximated using a composite Simpson’s rule. For an even
number of intervals Simpson’s 1/3 rule was applied, while
for an odd number of intervals Simpson’s 3/8 rule was em-
ployed. The initial condition, for solving the population bal-
ance equation, was assumed to be a normal distribution with
a D32 equal to the first experimental point which was gener-
ally available within 30 seconds after the start of the mixing.
All the physical constants and parameters used in the ki-
netic model (method of moments), for MMA free radical
polymerisation, were extracted from the literature, mostly
after Baillagou and Soong (1985). The adjustable parame-
ters (kb , ab , kc , ac and bc ), used for calculating the rates of
drop break up and coalescence, were obtained for the best
fit to the experimental D32 -time data. The parameter kb in
the break up frequency model (Eq (6.2)) determines the rate
of eddy-drop collisions leading to the drop break up. With
increasing kb the frequency of break up increases. The pa-
rameter ab in Eq. (6.3) indicates the efficiency of drop break
up. The higher the values of ab , the lower the required en-
ergy for drop deformation leading to a higher probability for
drop break up. The parameter kc in Eq. (7.2) determines the
frequency of drop–drop collision. The collision frequency
increases with kc . The parameters ac and bc in Eq. (7.3)
indicate the importance of coalescence efficiency. A higher
ac and lower bc result in a higher coalescence efficiency
due to lower required energy for drop deformation and film
drainage processes.
In order to estimate the values of the adjustable param-
eters for suspension polymerisation of MMA, a two-stage
approach was used. We first applied the model to the D32 -
time experimental data of the non-reacting low-hold up
MMA/water dispersion (
d = 0.05) while ignoring coa-
lescence. In this way, the estimated values for break-up
adjustable parameters to give the best fit to the experi-
mental data were obtained. Then using these values for a
higher dispersed phase hold up (
d = 0.20), the coalescence
D 32 (µm)
0
0 0.2 0.4 0.6 0.8 1
40 t/teq
20
0
0 30 60 90 120
time (min)
Fig. 14. Model prediction of drop size evolution in a simple MMA–water
dispersion; solid and dashed lines are model predictions with and without
considering Eq. (8), respectively, at the basic conditions. The small graph
illustrates Eq. (8).
adjustable values were estimated. The fitted values for the
adjustable parameters are shown in Table 2.
4.1. Delayed stabiliser adsorption
The model was initially checked for MMA–water disper-
sions in the absence of initiator. Experimental data for non-
reactive MMA-water dispersions have been reported else-
where (Jahanzad et al., 2005). The model predicted a very
quick transition stage in comparison with the experimental
data as shown in Fig. 14 (dashed line). It is known that the
steady-state drop size is independent of the magnitudes of
the rate of drop break up and coalescence and only depends
on their relative value (Rb /Rc ) (Sovova, 1981). However,
the approach to steady state depends on the magnitudes of
Rb and Rc . Mathematically, it is possible to slow down the
approach to the steady state by decreasing the ratio of the
rate of drop break up over that of drop coalescence without
affecting the steady-state value of drop size. However, this
approach is not justified because while it produces a good
fit for the transition stage, it cannot give a satisfactory fit
to D32 -time data during the growth stage because of under-
estimation of the rate of drop break up. There are reports
5586 F. Jahanzad et al. / Chemical Engineering Science 60 (2005) 5574 – 5589

in the literature that the adsorption of PVA molecules on 1

the surface of drops in liquid–liquid dispersions is a slow 4.0 wt%
process (Lazrak et al., 1998; Mendizabel et al., 1992; Zerfa
and Brooks, 1998). Zerfa and Brooks (1998) and Mendiz- 0.8
1.0 wt%
abel et al. (1992) claimed that while the adsorption of PVA
might not be slow, the reconfiguration of PVA molecules
0.25 wt% LPO
on the surface of drops, required for a better protection of 0.6

conversion
drops, will take time. A simple empirical correlation (Eq.
(8)), based on the data of Lazrak et al. (1998), was devel-
oped to predict the evolution of instantaneous surface cov- 0.4
erage () with time. The equilibrium time (teq ) within which
the equilibrium value (eq ) is reached was taken as 30 min
(Lazrak et al., 1998). The Langmuir-type interfacial adsorp- 0.2
tion (Eq. (9)) was considered.
 = eq (1 + 0.14Ln(t/teq )) for t
teq ,
(8) 0
 = eq for t > teq ,
0 50 100 150 200
ka csa time (min)
eq = , (9)
1 + ka csa
where ka and csa are the partitioning coefficient and aqueous
phase concentration of the suspending agent, respectively.
In Fig. 14 the solid line shows the model prediction with the
Fig. 15. Experimental (symbol) and theoretical (line) conversion-time for
suspension polymerisation of MMA with different LPO concentrations.
integration of this term gives the total rate of coalescence:
incorporation of delayed PVA adsorption. The small graph in  dy max   dv/2
Fig. 14 illustrates the variations of /eq with dimensionless Rc (t) = v 2/3 c(v − u, u)n(dv−u , t)
time (t/teq ) according to Eq. (8). dy min dv

min
The partitioning of PVA molecules and its effect on in-
terfacial tension (
) and drop surface coverage () can be × n(du , t) d(du ) d(dy ). (13)
calculated by the following equations along with Eq. (9):
cs = csa + a c∞ , (10) The above expressions were used to produce the predictions
shown in Fig. 1b.

=
0 − k
, (11)
4.3. Model predictions
where a is the interfacial area per water volume unit, cs is the
total concentration of the stabiliser, c∞ is the concentration
Here a few examples of model predictions in compari-
of the stabiliser (mass/area) at  = 1,
0 is the interfacial
tension in the absence of the stabiliser and k
is constant. We
used the same values for ka and k
as obtained by Lazrak
et al. (1998), using their experimental data for interfacial
tension of the same system at 70◦ C (10.56 l/g and 10.40 for
ka and k
, respectively). The value of c∞ (=1×10−6 g/cm2 )
was estimated from the data reported by Van Den Boomgaard
et al. (1978).
4.2. Estimation of Rb and Rc
The overall rate of drop break up at time t, Rb (t), can be
calculated by integrating the rate of break up for drops with
diameter dv at time t (the second term in the right hand side
of Eq. (5)) over the range of drop size of dv min to dv max :
 dv max
Rb (t) = [b(v)n(dv , t)] d(dv ). (12)
dv min
The third term of the right hand side of Eq. (5) gives the
number of coalescence events at time t which form drops
with diameter dv from coalescing smaller drops. Therefore
son with experimental data are presented. Figs. 15 and 16
compare the experimental and theoretical conversion-time
for MMA suspension polymerisation with different LPO
concentrations and monomer volume fractions, respectively.
These figures show a good agreement between the model
predictions and the experimental kinetics data. Fig. 16 shows
the capability of the model to predict the kinetics of suspen-
sion polymerisation for monomers with partial water solu-
bility. The small graph in Fig. 16 compares the experimental
and model results for weight fraction of MMA in aqueous
phase during the suspension polymerisation with monomer
volume fraction of 0.2.
Figs. 17 and 18 illustrate the experimental data and model
predictions of drop size evolution in suspension polymerisa-
tion of MMA with different PVA and LPO concentrations,
respectively. As it is seen, the model can predict all four char-
acteristic intervals and the effects of stabiliser and initiator
concentrations on these intervals. In Fig. 19 the experimen-
tal and model results for the evolution of PSD are compared
for one of the experiments. Again the agreement between
the theoretical and experimental results is acceptable.
 F. Jahanzad et al. / Chemical Engineering Science 60 (2005) 5574 – 5589 5587

1.0 150
φd
0.025
0.8 0.20 0.25 wt% LPO
model
1.0 wt%
100
0.6
4.0 wt%
conversion

D 32 (µm)
0.4
0.03
WMMA,aq

50
0.2
0
0 120
time (min)
0.0
0 30 60 90 120
time (min) 0
0 50 100 150 200
Fig. 16. Experimental (symbol) and theoretical (line) conversion-time for time (min)
suspension polymerisations with different MMA hold ups. The small
graph shows the monomer weight fraction in water phase during the Fig. 18. Experimental (symbol) and theoretical (line) D32 for suspension
polymerisation. polymerisations with different LPO concentrations.

200 0.05

0.5 g /l PVA 40 min

0.04
150

0.03 0.5 min

F(d) (1/µm)
D 32 (µm)

100 final PSD

1 g/l
0.02

50 10 g/l
0.01

0 0.00
0 50 100 150 200 0 50 100 150 200
time (min) diameter (µm)

Fig. 17. Experimental (symbol) and theoretical (line) D32 for suspension Fig. 19. Experimental (symbol) and theoretical (line) evolution of particle
polymerisations with different PVA concentrations. size distribution for MMA suspension polymerisation with 1.0 g/l PVA.
The curves for 0.5 and 40 min and Final PSD represent approximately the
transition stage, quasi-steady-state, and identification stage, respectively.

5. Conclusions
Suspension polymerisations can be characterised by up
to four intervals. They are: transition stage during which
drops are reduced in size, quasi-steady-state stage with a
rather constant drop size, growth stage with a sharp increase
in the size of drops, and identification stage where the full
transformation of drops to particles occurs. The existence,
duration, and the boundaries of these intervals are subject to
massive change with polymerisation conditions. The results
showed that:
• The transition stage is shortened by increasing the rate
of polymerisation in the drops. This could be done by
increasing the initiator concentration or using a higher
reaction temperature. Increasing the impeller speed and
PVA concentration will also lead to a shorter transition
period. However, the delayed adsorption of the stabiliser
on the surface of drops will prolong the transition stage.
5588 F. Jahanzad et al. / Chemical Engineering Science 60 (2005) 5574 – 5589

• The quasi-steady state occurs only if the balance between Ed Young’s elasticity modulus of
drop break up and coalescence can be maintained. This re- monomer–polymer mixture
quires that a high rate of drop break up should be operative Ep Young’s elasticity modulus of pure polymer
in the course of a polymerisation. This can be assured via f volume frequency, Eq. (1)
a low rate of polymerisation in the drops, i.e., low initiator F normalised distribution, cm−1 , Eq. (2)
concentration and reaction temperature, which facilitates ka partitioning coefficient, cm3 g−1 , Eq. (9)
drop break up. Alternatively, a high rate of drop break up kb , kc adjustable parameters of break up and
can be maintained by using a high agitation speed and collision frequencies, respectively
high stabiliser concentration. Low impeller speed and k
constant in Eq. (11), dyn cm−1
low PVA concentration, for example, may remove the n average number of daughter drops formed by
quasi-steady-state stage completely and drops start grow- breaking of a mother drop
ing right after the transition stage. At a high stabiliser NI agitation speed, rpm
concentration, all stages overlapped so that the drops Rb , Rc overall rates of drop break up and
formed after the transition stage were maintained in the coalescence, respectively, s−1 cm−3
reaction medium and were gradually converted to poly- Re Reynolds number
mer beads with the least interaction with other particles t time, min
(static steady state). v drop/particle volume, cm3
• A growth stage occurs if drops are not sufficiently sta- Ve elasticity parameter, defined by Eq. (6.5)
ble against both break up and coalescence. The onset of We Weber number
the growth stage is advanced with a decrease in the rate wMMA,aq MMA weight fraction in the aqueous phase
of drop break up, i.e., decreasing stirring speed and PVA x conversion of monomer to polymer
concentration. The onset of the growth stage was delayed Y0 elasticity parameter, defined by Eq. (6.5)
with decreasing initiator concentration and temperature.
The growth stage can be totally eliminated from a poly-
merisation process if dispersions with a static steady state Greek letters
can be formed. That requires a high concentration of sta-
biliser, or a low concentration of monomer to be used. elasticity parameter, defined by Eq. (6.6)
• The onset of the identification stage depends only on  daughter drop volume distribution, Eq. (5)
the variables which alter the rate of polymerisation such  average energy dissipation rate, cm2 s−3
as temperature and initiator concentration. Mixing condi-  instantaneous surface coverage of
tions do not play an important part here. drops, Eq. (8)
eq equilibrium surface coverage of drops,
A population balance modelling approach was used to model Eq. (9)
the suspension polymerisation of MMA. Predicted kinetics  elasticity parameter, defined by Eq. (6.6)
features and drop size behaviour were compared with ex- b , c efficiencies of drop break up and
perimental data. coalescence, respectively, Eqs. (6.3)
and(7.3)

d viscosity of dispersed phase, g cm−1 s−1
d density of dispersed phase, g cm−3
Notation

0 interfacial tension in the absence of the
stabiliser, dyn cm−1
ab , ac adjustable parameters of break up (Eq. (6.3))
d volume fraction of dispersed phase
and coalescence (Eq. (7.3)) efficiencies, b , c drop break up and collision frequency,
respectively respectively, s−1 , Eqs. (6.2) and (7.2)
b rate of drop break up, s−1 , Eq. (5)  efficiency parameter, defined by Eq. (6.4)
bc adjustable parameter of coalescence
efficiency, Eq. (7.3)
c rate of drop coalescence, s−1 , Eq. (5) References
csa stabiliser concentration in the aqueous
phase, g cm−3 Adamson, A.W., 1976. Physical Chemistry of Surfaces. Wiley, New York.
c∞ stabiliser concentration (mass/area) at Alvarez, J., Alvarez, J.J., Martinez, R.E., 1991. Conformation of the
 = 1, g cm−2 particle size distribution in suspension polymerisation. The role of
kinetics, polymer viscosity and suspension agent. Journal of Applied
d drop/particle diameter, cm
Polymer Science 49, 209–221 (Applied Polymer Symposium).
dv max maximum drop size, cm Alvarez, J., Alvarez, J.J., Hernandez, M., 1994. A population balance
dv min minimum drop size, cm approach for the description of particle size distribution in suspension
D32 Sauter mean diameter, cm, Eq. (4) polymerisation reactors. Chemical Engineering Science 49 (1), 99–113.
 F. Jahanzad et al. / Chemical Engineering Science 60 (2005) 5574 – 5589
Baillagou, P.E., Soong, D.S., 1985. Free-radical polymerisation of methyl-
methacrylate in tubular reactors. Polymer Engineering and Science 25
(4), 212–231.
Brooks, B.W., 1990. Basic aspects and recent developments in suspension
polymerisation. Makromolecular Chemistry, Makromolecular
Symposium 35,36, 121–140.
Castellanos, J.R., Mendizabal, E., Puig, J.E., 1991. A quick method for
choosing a protecting colloid for suspension polymerisation. Journal of
Applied Polymer Science 49, 91–101 (Applied Polymer Symposium).
Chatzi, E.G., Kiparissides, C., 1992. Dynamic simulation of bimodal
drop size distributions in low-coalescence batch dispersion systems.
Chemical Engineering Science 47 (2), 445–456.
Chatzi, E.G., Kiparissides, C., 1994. Drop size distributions in high holdup
fraction dispersion systems: effect of the degree of hydrolysis of PVA
stabiliser. Chemical Engineering Science 49 (24B), 5039–5052.
Chen, Z., Pauer, W., Moritz, H.U., Pruss, J., Warnecke, H., 1999. Modeling
of the suspension polymerisation process using a particle population
balance. Chemical Engineering and Technology 22, 609–616.
Chiu, W.Y., Carratt, G.M., Soong, D.S., 1983. A computer model for the
gel effect in free-radical polymerisation. Macromolecules 16, 348–357.
Cunningham, M.F., 1999. Microsuspension polymerisation of methyl
methacrylate. Polymer Reaction Engineering 7 (2), 231–257.
Dowding, P.J., Vincent, B., 2000. Suspension polymerisation to form
polymer beads. Colloids and Surfaces A: Physicochemical and
Engineering Aspects 161, 259–269.
Ferry, J.D., 1980. Viscoelastic Properties of Polymers. 3rd ed. Wiley,
New York.
He, Y., Howes, T., Litster, J.D., Ko, G.H., 2002. Experimental study
of drop-interface coalescence in the presence of polymer stabilisers.
Colloids and Surfaces A: Physicochemical and Engineering Aspects
207, 89–104.
Jahanzad, F., 2004. Evolution of particle size distribution in suspension
polymerisation reactions. Ph.D. Thesis, Loughborough University,
Loughborough, UK.
Jahanzad, F., Sajjadi, S., Brooks, B.W., 2004. On the evolution of
particle size average and size distribution in suspension polymerisation
processes. Macromolecular Symposia 206, 255–262.
Jahanzad, F., Sajjadi, S., Brooks, B.W., 2005. Comparative study of
particle size in suspension polymerization and corresponding monomer-
water dispersion. Industrial and Engineering Chemistry Research 44
(11), 4112–4119.
Johnson, G.R., 1980. Effects of agitation during VCM suspension
polymerisation. Journal of Vinyl Technology 2, 138–143.
Kalfas, G., Ray, W.H., 1993. Modeling and experimental studies
of aqueous suspension polymerisation processes. I. Modeling and
simulation. Industrial and Engineering Chemistry Research 32,
1822–1830.
Kalfas, G., Yuan, H., Ray, W.H., 1993. Modeling and experimental studies
of aqueous suspension polymerisation processes. II. Experiments in
batch reactors. Industrial and Engineering Chemistry Research 32,
1831–1838.
Konno, M., Arai, K., Saito, S., 1982. The effect of stabiliser on coalescence
of dispersed drops in suspension polymerisation of styrene. Journal of
Chemical Engineering of Japan 15 (2), 131–135.
Lazrak, N., Le Bolay, N., Ricard, A., 1998. Droplet stabilization in high
holdup fraction suspension polymerisation reactors. European Polymer
Journal 34, 1637–1647.
5589
Liu, S., Li, D., 1999. Drop coalescence in turbulent dispersions. Chemical
Engineering Science 54, 5667–5675.
Machado, R.A.F., Pinto, J.C., Araujo, P.H.H., Bolzan, A., 2000.
Mathematical modeling of polystyrene particle size distribution
produced by suspension polymerisation. Brazilian Journal of Chemical
Engineering 17, 395–405.
Maggioris, D., Goulas, A., Alexopoulos, A.H., Chatzi, E.G., Kiparissides,
C., 1998. Use of CFD in prediction of particle size distribution in
suspension polymer reactors. Computers and Chemical Engineering 22
(Suppl. S), S315–S322.
Maggioris, D., Goulas, A., Alexopoulos, A.H., Chatzi, E.G., Kiparissides,
C., 2000. Prediction of particle size distribution in suspension
polymerisation reactors: effect of turbulance nonhomogeneity.
Chemical Engineering Science 55, 4611–4627.
Marten, F.L., Hamielec, A.E., 1979. High-conversion diffusion-controlled
polymerisation of methyl methacrylate. American Chemical Society,
Symposium Series 104, 43.
Mendizabal, E., Castellanos-Ortega, J.R., Puig, J.E., 1992. A method for
selecting a polyvinyl alcohol as stabiliser in suspension polymerisation.
Colloids and Surfaces 63, 209–217.
Odian, G., 1991. Principles of Polymerisation. 3rd ed. Wiley, New York.
Okaya, T., Suzuki, A., Kikuchi, K., 1999. Importance of grafting in the
emulsion polymerisation of MMA using PVA as a protective colloid.
Effect of initiators. Colloids and Surfaces A: Physicochemical and
Engineering Aspects 153, 123–125.
Sovova, H., 1981. Breakage and coalescence of drops in a batch stirred
vessel—I. Comparison of continuous and discrete models. Chemical
Engineering Science 36 (1), 163–171.
Van Den Boomgaard, T., King, T.A., Tadros, T.F., Tang, H., Vincent, B.,
1978. The influence of temperature on the adsorption and adsorbed
layer thickness of various molecular weight fractions of poly(vinyl
alcohol) on polystyrene latex particles. Journal of Colloid and Interface
Science 66 (1), 68–76.
Villalobos, M.A., Hamielec, A.E., Wood, P.E., 1993. Bulk and suspension
polymerisation of styrene in the presence of n-pentane. An evaluation of
monofunctional and bifunctional initiation. Journal of Applied Polymer
Science 50, 327–343.
Vivaldo-Lima, E., Wood, P.E., Hamielec, A.E., Penlidis, A., 1997. An
update review on suspension polymerisation. Industrial and Engineering
Chemistry Research 36, 939–965.
Vivaldo-Lima, E., Wood, P.E., Hamielec, A.E., Penlidis, A., 1998.
Calculation of the particle size distribution in suspension
polymerisation using a compartment-mixing model. The Canadian
Journal of Chemical Engineering 76, 495–505.
Yang, B., Takahashi, K., Takeishi, M., 2001. Unsteady stirring method used
in suspension polymerisation of styrene. Journal of Applied Polymer
Science 82, 1873–1881.
Zerfa, M., Brooks, B.W., 1998. Experimental investigation of PVA
adsorption at the vinyl chloride water interface in monomer suspension.
Colloids and Surfaces A: Physicochemical and Engineering Aspects
132 (2–3), 267–273.
Zhou, G., Kresta, S.M., 1998. Correlation of mean drop size and minimum
drop size with the turbulence energy dissipation and the flow in an
agitated tank. Chemical Engineering Science 53 (11), 2063–2079.