Chemical Engilwering Science, Vol. 40, No. I, pp. 75-86, 1985. 0009 -2509/85 :!i3.

oo + o_oo
Printed in Great Britain. Pergamon Press Ltd.

MAJOR FACTORS CONTRIBUTING TO THE

NONLINEAR KINETICS OF FREE-RADICAL
POLYMERIZATION

P. E. BAILLAGOU and D. S. SOONG t

Department of Chemical Engineering, University of California, Berkeley, CA 94720, U.S.A.

(Received 16 January 1984)

A�tract-Predictions based on a recently developed model for nonisothermal free-radical polymerization

with ge� e ��t are made for batch and plug-flow systems. The often observed nonlinear kinetic phenomena,
such as 1gmt1on, thermal runaway, dead ending and autoacceleration are clearly separated into parametrically
dependent and gel effect induced categories. This analysis identifies major causes for the difficulties
encountered in polymerization reaction engineering and serves as the basis for future modeling of realistic
processes, where the fluid mechanics also plays an important role in governing reactor performance and
product quality.

INTRODUCTION weight distribution (Martin et al., 1972; Nunes et al.,

The conversion of monomer into polymer reduces the 1982). In general, mechanical properties are improved
free volume of the reaction mixture. This results in a by low polydispersity and high molecular weights. The
great increase of the viscosity of the polymerizing typical goal is to maintain polydispersity to as low as
medium; mixing and pumping problems can occur possible arrd to achieve a minimum molecular weight

even at low conversions ( - 30 %). When the reactor is
operated below the glass transition temperature of the
polymer, total conversion cannot be achieved because
monomer diffusion to reacting radicals is seriously
hampered by the high viscosity of the reaction mixture.
A related consequence of the decrease in the free
in the neighbourhood of 105• Since the product quality
is determined by the processing history experienced by
the reaction fluid, gel effect and thermal run-away must
generally be avoided. Sacks et al. ( 1973) have de­
termined the optimal temperature history in a batch
reactor and shown that product quality is indeed

volume is the gel effect (Trommsdorff et al., 1947),
which causes autoacceleration of the overall rate of
polymerization, observed even under isothermal con­
ditions. This phenomenon, particularly strong in the
case of MMA polymerization, will be a focus of this
intimately related to the control of the reactor.
The use of tubular reactors is inherently an attract­
ive alternative for the mass production of high quality
polymers. Continuous operation allows high through­
puts at a low cost. This reactor is distributed; the

work. governing space variable can be used for process

Polymerization reactions are highly
(releasing 10-20 kcal/g mol). Unfortunately, mono­
mer-polymer solutions have small thermal diffusivities
and low heat capacities. Convective heat transfer is
retarded by the high viscosity of the fluid, which
decreases mixing efficiency. This unfavourable combi­
nation of the thermodynamics of the reaction and the
physical properties of the reactive mixture makes
reaction temperature very difficult to control. Thermal
run-aways often ensue. We will thus devote much
attention to this phenomenon in the present study as
well.
Despite the difficulties confronting polymerization
reaction engineers, the use of polymers as structural
materials often generates stringent performance re­
quirements of a polymerization process, e.g. steady
production rates and reproducible mechanical pro­
perties of the product. Steady throughputs are realized
more easily and at lower costs b y continuous processes.
The important mechanical properties of a polymer
have been shown to depend strongly on the molecular
t Author to whom correspondence should be addressed.
exothermic optimization, similar to the use of temperature history
in a batch reactor to generate desirable polymers.
Hence, in principle a tubular reactor would combine
the advantages of both a CSTR (continuous) and a
batch reactor (optimizable). However, in practice
stirred tanks are easier to control and have better
mixing characteristics than tubular reactors. Mixing
and heat transfer efficiencies in tubular reactors are
generally not sufficient to overcome process difficulties
accompanying polymerization. Furthermore, the vel­
ocity profile in a real tubular reactor is far from that of
a plug-flow reactor. A longer residence time and
therefore higher conversion is to be expected near the
wall. Consequently, a very high viscosity gradient
develops along the radius of the tube, which in turn
aggravates the already nonlinear velocity profile. This
distorted velocity profile eventually leads to the forma­
tion of a thick stagnant layer of polymer, which slowly
coats the wall and reduces the capacity of the reactor
(Lynn and Huff, 1971).
Most publications in the polymerization reaction
engineering literature address only parts of the above
problems. The gel effect was experimentally studied
and the results modeled, but only for isothermal

75
76 P. E.
conditions. On the other hand, non-isothermal be­
haviour was investigated, but the gel effect was neglec­
ted in the analysis. The kinetic models developed to
date have been extensively used to predict molecular
weight averages. However, attempts to correlate the
entire molecular weight distribution with reaction
conditions have seldom been made. Incomplete mathe­
matical models are also often found in tubular reactor
literature. Some papers emphasi.z.e the kinetics but
ignore the transport problem by assuming plug-ft.ow.
The opposite has also been reported, i.e. extensive
study of the velocity profile with a poor kinetic model.
Most previous works on tubular polymerizers focus on
the polymerization of styrene. This system is not
representative of all the polymerization difficulties,
because it exhibits only a very weak gel effect. In
contrast, our project concentrates on the polymeriz­
ation of methyl methacrylate, whose gel effect is much
more severe. The polymerization of methyl meth­
acrylate poses particularly challenging problems in our
studies.
The present program is divided into several phases.
Table I. Proposed free-radical polymerization mechanism
kd R"
I 2
k.
1.
R + M P
i
kp
p.; + M p� +
kt:m
p� + M Dn + Pi
kts
P,; + s
Dn+ Pi
kt:d
p,; + Pm Dn+Dm
ktc
p� + Pm 0n+m
where
I Initiator
R Primary radical
M Monomer molecule
S Solvent molecule
P� Macroradical of length n
00 Dead polymer of length n
BAILLAGOU and D. s. SoONG
First, a r�listic gel effect model has been developed to
describe the polymerization of methylmethacrylate
over the entire course of reaction (Chiu et al., 1983).
This model is here used for the simulation of non­
isothermal reactions in plug-ft.ow reactors. At this level
of approximation, the rheology is decoupled from
kinetic considerations. Autoacceleration, ignition and
thermal run-away phenomena can be traced distinctly
to separate causes by simple kinetics analyses. Next,
the complete kinetic model is employed together with a
realistic viscosity constitutive relation for accurate
descriptions of the velocity, conversion and tempera­
ture profiles in tubular reactors. The predictions of this
model are compared with the plug-flow simulations.
Experiments have been conducted to verify model
predictions and to pin-point the causes of problems
often encountered in tubular polymerization pro­
cesses. Finally, a new operating strategy, polymeriz­
ation at elevated pressures and above Tg temperatures,
is explored. The results of such experiments will be
disclosed. The above segments of this ongoing research
program will be published in a series of papers. This
Initiation
1
Propagation
Chain transfer to monomer
Chai.n transfer to solvent
Termination by disproportionation
Termination by combination
 Free-radiail polymerization 77

paper only addresses issues concerning the major non­
isothermal and nonlinear features of polymerization
reaction with gel effect.
LITERATURE REVIEW ON NON-ISOTHERMAL
POLYMERIZATION REACTIONS
Exothermic polymerization reactions occurring in a
low thermal conductivity medium create a difficult
heat transfer problem which seriously limits achieve­
ment of isothermal conditions in tubular reactors. All
the rate constants of the elementary steps in a free­
radical mechanism are assumed to have an Arrhenius
temperature dependence. However, the initiation pro­
cess requires a much higher activation energy than
termination. An increase in temperature consequently
affects kd much more than kt. Therefore, this strong
positive feedback further accelerates the reactor tern-
perature rise, making run-away a likely event during
polymerization. Biesenberger and Capinpin exten­
sively studied the behaviour of a non-isothermal
polymerization reactor, its sensitivity to the operating
conditions and the consequences of temperature vari­
ations on the conversion and molecular weight distri­
bution. They classified the principal non-isothermal
phenomena into run-away, ignition (parametrically
sensitive run-away) and dead ending (incomplete con­
version of the monomer due to initiator burning out
prematurely at high temperatures) (Biesenberger and
Capinpin, 1974). A dimensional analysis inspired by
explosion and combustion theories was also carried
out by these authors. Using dimensionless groups, they
derived criteria to predict run-away (RA), ignition (IG)
and dead ending (DE) in non-isothermal polymeriz­
ation. These predictions were compared with com­
puter simulations (Biesenberger et al., 1976) and with
experiments (Sebastian and Biesenberger, 1976).

Table 2. Kinetic equations for the proposed free-radical mechanism

r:t a -kd I I l (1)

rR - 2fkd(IJ-k i[R0)(H] ( 2)

rM - -k1[R"] [MJ-(kp+kt };\0(M) ( 3)

m

k p [M] ( [ P,'.,- 1-(P.;_ J )-(kt

1 m
(M)+kt
s
I SJ} Ip�]-( p� IL
m�
k t IP.;, l (5)

( 6}

n-1

kt [P,'.,J [H]+kr [P.;_J [SJ+kt IP,;J:i..0+0.Skt L (P� I [P�ml (7)

m s d c m-1

with the symbols used in Table 1 and the following

no tations for the rate constants:

kd - initiator decomposition

k i - initiation

kp - propagation

ktm- transfer to monomer

kts - transfer to solvent

kc - termination by dis proportionation

d
k tc 'O
termination by combination
78 P. E. BAJLLAGOU and D. S. SooNG

The above modeling work was somewhat refined by
Brooks (1983) who showed that the boundary between
ignition and extinction can be more accurately located,
if changes of densities and heat capacities with tem­
perature are taken into account in the analysis. His
conclusions are consistent with the important ideas
already formulated by Biesenberger and co-workers.
Work done in the U.S.S,R. about non-isothermal
polymerizations was reviewed by Stolin et al. (1979).
Again, the analogy with combustion theories allows
the utilization of Semenov and Frank Kamenetsky's
work in the treatment of thermal effects on the free­
radical polymerization kinetics.
Another cause of non-linearity in polymerization
kinetics is due to the reduction of the free volume of the
mixture with conversion. This phenomenon, called gel
effect, occurs even under isothermal conditions and is
entirely different in nature from the previously dis­
cussed thermal autoaccelerations. The origin of the gel
effect was discovered by Trommsdorff (1947). The rate
of polymerization is not accelerated by an increase of
the rate of initiation, but by a decrease of the rate of
termination. The termination steps are more severely
diffusion controlled than the propagation reaction,
because they involve reactions between macroradicals
whose mobility is seriously reduced by the diminishing
free volume of the polymerizing mixture. When the
conversion in the reactive fluid approaches its glass

Table 3. Mass and energy balance equations for a plug-flow reactor at steady-state

dx
( l )
dz uz

d I Ij l duz
- --( r -(I)
1
-- ) ( 2)
dz uz dz

d l duz
---
ll.o
- --( r - ll.o -- ) ( 3)
dz ll.o dz
Uz

dT l
rT
( 4)
- -- ( -- )
dz Uz pcv

with:

rx � (kp+kt )(l-x)ll.0 ( 5)
m
r
I
- -kd [I} ( 6}
')
( 7}
-
rll.0 - 2fkd[ I )-kcll.0

rT - - AHpkp [M]ll.0+ (8)

where:

x conversion

(I] init:iator concentrat:ion

[M J .. monomer concentration
00

total_ radical concentration - L ( P�)

n-l.
T temperature of the reaction mixture
Tbath' temperature of the bat:h

axial_ ve1-ocity
uz
p density of t:he mixt:ure

Cv hea t capacity of the mixture

n� heat of pol_ymerization

U overall heat transfer coefficient bct�1cen the reaction riix t ur e

and the bath f 1-uid

H rhJrrtcteristic linear dir."':.cnsion .. volunc over tot:al heat t:rn.nsfer

area
 Free-radical polymerization 79

transition at the reaction temperature, movement of
the monomer molecules is also impeded. The rate of
propagation thus decreases and the reaction stops
before complete depletion of monomer or initiator.
Accurate constitutive equations must be derived to
describe the dependence of the apparent propagation
and termination rate constants on conversion, tem­
perature and molecular weight. Several models have
been proposed in the literature. These are reviewed by
Carratt (1983). All models agree on the principle of
relating the termination rate to macroradical diffusion.
Differences exist in the choice of the theory to describe
the diffusion processes.
MODEL DERIVATION
This model describes the non-isothermal bulk or
solution polymerization of methyl methacrylate in a
plug-flow reactor at steady state. Axial diffusive mass
and heat transfer are neglected. Since the quasi-steady
state approximation is not used, three differential mass
balance equations are required (conservation of
monomer, initiator and total radical concentrations).
Non-isothermal effects are accommodated by coup­
ling the mass balances to an energy balance equation
through the Arrhenius expressions for the rate con-
stants. Conservation of energy accounts for the heat
generated by the propagation reaction and heat re­
moval through the wall of the reactor and by convec­
tive fluid flow. The gel effect is modeled by the
constitutive equations derived by Chiu et al. (1983).
Volume contraction due to reaction as well as the
temperature dependence of the densities of monomer
and solvent are allowed for in the model. The density
of the mixture is a function of the conversion, the
amount of solvent introduced in the feed and the
temperature. Heat capacities of the monomer, solvent
and polymer are taken to be constant, independent of
temperature. Monomer and polymer are assumed to
have the same heat capacity on a unit mass basis.
The complete mechanism of MMA polymerization
is summarized in Table 1. Application of simple kinetic
laws to the polymerization mechanism allows the
derivation of the rate of accumulation for the species
(Table 2). The rate of accumulation of the total
concentration of radicals l::'=i [P.] = A.0 [eq. (7),
Table 2] was obtained by adding all the free radical
kinetic equations.
The set of coupled first order differential equations
is listed in Table 3. The Arrhenius temperature depen­
dences of the rate constants are shown in Table 4.
Table 5 gives the physical properties of the chemical

Table 4. Kinetic rate constants used in the models

f - 0.58

kd(min-1) • 6.32•1016 exp [-30.66 (kcal/mole}/RT(°K})

(Tobolsky and Baysal, 1953}

k p(l./min.mol) • 2.95•107 exp (-4.35 (kcal/mole)/RT(°K))

(Mahabacli and 01Drlscol 1, 1977)

k
t
m - 9.48•103 exp (-13.38 (kcal/mole)/RT(°K))
k
p
(Stickler and Meyerhoff, 1978).

k
t
8 1.01•10 3 exp [-11.40 (kcal/mole)/RT(°K))
k
•

p
(Gopalan and S anthappa, 1957).

k (I/min.mol) 9 exp (-0.701 (kcal/mole)/RT(°K))

t • 5.88•10

(Hahabadi and 01Driscoll, 1977).

k
t
c
- 3 .956•10-4 exp (4.09 (kcal/mo1e)/RT(°K))
�
(Be�ington, et al., 1954).
80 P. E. BAILLAGOU and D. s. SooNG

Table 5. Physical properties of the polymerizing mixture

MMA)
3
pH(g/cm ) 0.9665 0.0011 T(°C) (our data for

p8(g/cm )
3
0.883 0.0009 T(°C) (our data for to1uene)

-
C - - 0.1946 - 0.916•10 3 T(°C) (our data. assum:Lng

C (ca1/g• ° C) 0.4 (Ja�sLnghan� and Ray, 1977)

p
H

C (ca1/g•°C) 0.535 {Perry and ChL1ton, 1973)

p
s
WH(g/mo1) 100.13 (Weast, 1982)

w8(g/m.o1) 92.15 {Weast. 1982)

Table 6. Gel effect constitutive equations

w�th:

4
3. 46 • 10
4.4533 • 1018 [I]F exp(­ )
R{T+273 .15)

15 2. 8 • 10 4
2 • 5 29 2•10 exp ( - )
R(T+273.15)

(1 - 91p)
D - exp (2.303 ������ )
A + B(l )
-91p

A 0 .168

B 0.03

where:

TgpLB the g1ass transLtLon temperature of PMHA {ll4°C )

91 p �a the vo1ume fractLon of po1ym.er

 Free-radical polymerization 81

species involved in the process. The gel effect constitut­ Residence time (min)

ive equations are summarized in Table 6. 0 10 20 :30

This extremely stiff non-linear initial value problem (b)
180
is numerically integrated by Gear's method. The
RA
---
results of the calculation are the conversion, tempera­ 2
ture, initiator and radical concentration profiles. 3

160

4
RESULTS AND DISCUSSIONS
IG
Most of the previous modeling and experimental
efforts about polymerization kinetics were devoted to � 140 GE l

batch systems. Since plug-flow and batch reactor

models are mathematically equivalent, our results can
be easily reduced to the batch reactor frame in order to
check their agreement with previous studies. After
suitable transformation (multiplication of the right­
hand side of the plug-flow equations by Q/nR 2, Q
being the volumeti"ic flow rate and R the inner radius of
the reactor), this model gives predictions in complete
agreement with the work by Chiu et al. (1983) on
isothermal batch polymerizations. This confirms the
Axial distance (cm)
consistency between the numerical techniques and
basic equations of the two works. The non-isothermal
Residence lime (min)
simulations of this model are shown in Figs 1-5. This o 1 o 2or=------ 3 0
set of figures examines parameter sensitivity of the
1 0-<1 r______r_____� �,=-,
process in detail, including the effects of varying an Cc)
initial condition (feed temperature, concentration of
initiator or fraction of solvent in the feed) or a system
parameter (overall heat transfer coefficient through the
'
wall, or bath temperature). The horizontal axis can be ... 10
0
scaled either in residence time or distance. The distance .§.
scale chosen here corresponds to our experimental c:
.2
conditions, where the feed flow rate is 1 cm3 s -1 into a �
1 cm radius tube. The residence time scale allows ready c
Q)
u
use of the predictions for other tubular or batch c
8
reactors. 0
u
The results show the details of the transition from 'O
quasi-isothermal to run-away behaviour. With gentle e
Q)

feed conditions, the temperature profiles reach a low "'

plateau after a smooth maximum. As the initial

E 107
=>

E
(.)
=>

Residence time(minl
10>!-_____,.2.,.,_______�
. 0�______,
10
-8
10
<a>
0 200 400 600
0.8 Axial dis1ance (cm)

Fig. L Non-isothermal polymerization of MMA. Influence

of AIBN concentration in the feed on the conversion,
06 temperature and cumulative radical concentration profiles
0
c:
(or histories) simulated by a plug ftow (or batch reactor)
� model. U = 15 Btu/h ft2°F, s = 0, TF = 80°C, Tbath = 80°C.
...
> [AIBNJF (moljl): (1) 0.0320, (2) 0.0258, (3) 0.0226, (4) 0.0193,
(.) 04
c:
0 (5) 0.0177, (6) 0.0161, (7) 0.0129.

conditions become more severe, two temperature

0.2
maxima can be observed. The first peak is a de­
generated run-away, showing that the heat transfer is
barely sufficient to keep the temperature under con­
o �:.____...____,____,__---'=.h---1..----'
0 200 400 trol. The second maximum is induced by the gel effect.
E;iOC)
Axial distance fem] With increasing severity of the initial condition, the

CF.S 40:1-F
82 P. E. BAILLAGOU and D. s. SooNG

Residence time(minl Residence lime (min)

1.o o
;::-------''"'o'-------=2=,o;=:-______........,. 10-40r------•�o"-------2=ro_____�3�0
�
Ca> Cc)

c: c:
0 0
·;:
� �
Q)
> c
c: "'
0 ...,
(_) c::

8
0
'6
...,

e
"'

� 167
:::>

E
600
:::>

200 400 (_)

Axial distance (cm)

-a
10
Residence lime(minl
0
iscP.������'0��2�0'--���---"3�0,__. 200 400
Axial distance (cm I
600

(b)
Fig. 2. Non-isothermal polymerization of MMA. lnftuence
of the feed temperature on the conversion, temperature and
I
160 cumulative radical concentration profiles (or histories) simu­
IG 2 3
GEI 4 lated by a plug ftow (or batch reactor) model
GEi U = 15 Btu/h ft2 °F, [AIBN]F = 0.0258 mol/1, s = 0, Tbath
GEi
= 70°C. TF (0C): {1) 100, (2) 96, (3) 95, (4) 90, (5) 80.
140

and the hot spot is shifted towards the inlet of the

reactor until early run-aways are generated (Figs 4
and 5).
The conversion plots corroborate essentially the
same points observed in the temperature profiles.
Under gentle feed conditions, the conversion increases
linearly with the axial distance. If the conditions
become more severe, the conversion profiles show an
acceleration of the reaction near the outlet of the
0 400 600
reactor. This onset of non-linearity is shifted towards
Axial distance (cm)
the inlet of the reactor as the severity of the initial
conditions increases. The slope of the conversion
profiles suddenly becomes almost vertical until the
temperature profile suddenly becomes concave up­ conversion abruptly levels off, indicating initiator
wards and the temperature reaches a high, sharp burn-out and incomplete reaction.
maximum at the early stage of the reaction. The The radical concentration plots exhibit charac­
mixture then cools pff rapidly to the surrounding teristics consistent with the hot spots and non­
temperature, indicating that the reaction has stopped. linearities observed in the temperature and conversion
Further changes in the initial condition do not affect profiles. The early run-aways coincide with very rapid
the temperature profiles significantly (Figs 1-3). Under increases in the radical population followed almost
good heat transfer conditions, the temperature profile immediately by its total disappearance. A rapid growth
becomes ftat without exhibiting any initial maximum. of the radical population is also observed in the case of
An increase of the surrounding temperature or a hot spots generated at the end of the reaction.
decrease of the heat transfer coefficient causes the However, radical concentration does not entirely
temperature to stabilize at a higher level. Additional vanish, but levels off or even increases after the sharp
changes in the same direction cause the formation of a maximum. An interpretation of these observations will
hot spot close to the outlet of the reactor. Eventually, be proposed below based on analyses of polymeriz­
the breadth of the temperature plateau is shortened ation reactions previously reported in the literature.
 Free-radical polymerization 83

Residence Residence tm
i elminl

1.00 10 10-4 �0�����--'-;10"--=2�0'--��-3�0"'""'

(a) (C)

0.8
�
"' 10
Ci
.s
c
06
. c:
0 0
+=
� �
"'
> c
c Q)
0 04 u
u c
0
u
Ci
u
'O
0.2 e
...
-�
!2
:>

E
0 :>
0 200 400 600 u
Axial distance !cm)

-6
10

Residence timeCminl 0 200 400 600

d 10 20 30 Axial distance Ccml

(b) Fig. 3. Non-isothermal polymerization of MMA. Influence

160 of the weight fraction of solvent (toluene) in the feed on
the conversion, temperature and cumulative radical con­
centration profiles (or histories) simulated by a plug flow (or
batch reactor) model. U = 15 Btu/h ft2°F, [AIBN]F
160 = 0.0258 moljl, TF = 80°C, Tbat
h = 80°C. s; (1) 0.00, (2) 0.10,
(3) 0.15, (4) 0.20.

.t 140

�
:::> by Sebastian and Biesenberger (1978) who experimen­
0
tally observed the transition from controlled to un­
� �20
controlled operation caused by small variations of the
E
...
I- initiator concentration in the feed. Their experimental
temperature histories are quite similar to our simu­
100 lation results for the same conditions (Fig. 3). As the
initiator concentration in the feed increases, the tem­
perature maximum at the begining of the reactor
80 becomes more pronounced, and the peak suddenly
0 becomes much sharper. This first run-away is para­
AKiol distance (cm)
metrically sensitive and is therefore an ignition, ac­
cording to Biesenberger's terminology. The high acti­
vation energy of the initiator decomposition, re­
The control of reactor temperature can be easily lost sponsible for the strong temperature dependence of
during two delicate phases of the polymerization the rate of initiation, is the source of this parametric
process. At the beginning of the reaction, the radical sensitivity. However, further increases in the initiator
population builds up very rapidly, until a steady state is concentration do not alter the temperature maximum,
reached, at which point the loss of radical popuJation and change the location of the peak only slightly. The
by termination balances the rate of production by parametric sensitivity disappears and the hot spot
initiation [eq. (7), Table 3]. This rapid establishment of following ignition is simple run-away. A similar pat­
the radical population releases much heat in a narrow tern describes the influence of all other parameters of
region of the reactor. This is the major cause of the the process.
early run-aways described by Biesenberger. The tem­ The high temperatures generated during a run-away
perature plots (Figs 1-5) show that the transition to or an ignition considerably enhance the rate of in­
run-away behaviour can be induced by very small itiator depletion. The initiator concentration vanishes
changes in any of the parameters or operating con­ rapidly and causes depletion of the radical population.
ditions investigated here. This behaviour was reported The reaction stops by exhaustion of initiator before
84 P. E. BAILLAGOU and D. S. SooNG

Residence Residence time (min)

1.00 10 10-4 or-------'1,,o'---'2"'or-=3<ro�

Ca> Cc)
5
0.8
� 10
0
QJ

.§.
c:
06
0

·�
Cl>
>
c::
0
u
04

0.2

0
0 2'JO 400 600
A"-•al distance lcml

-6
10

Residence time{minl
0 200 400 600
10 20 30 Axial distance (cm)
0
RA
(b) Fig. 4. Non-isothermal polymerization of MMA. Influence
180 of the overall heat transfer coefficient on the conversion,
temperature and cumulative radical concentration profiles (or
2
IG histories) simulated by a plug flow (or batch reactor) model.
(AIBNJF = 0.0258 mol/J, s = 0, TF = 80°C, Tbath = 80°C.
160 U (Btu/h ft2 °F) (1) O, (2) 15, (3) 20, (4) 30, (5) 60, (6) oo .
5
4 GE!
3 GE!
GE!
� 140
� effect occurs at an advanced stage of the reaction when
:>

0 the concentration of reactants is low, the temperature

:g_ 120 rise and its consequences are less deleterious than the
E
"'
I- early run-aways or ignitions discussed above. The
radical population does not entirely disappear after a
gel effect induced hot spot, because the diminution of
k1 caused by the gel effect limits the rate of radical
termination [eq. (7), Table 3] . Dead ending is therefore
less severe, since part of the radical population survives
200 400 600 the hot spot, allowing the conversion to proceed
Axial rlistance (cm)
(although at a slower rate) beyond the gel effect
induced temperature rise.

total conversion of monomer is reached. This phenom­ CONCLUSIONS

enon of dead ending can be observed in the conversion
curves (Figs 1-5) corresponding to uncontrolled tem­
perature. It is particularly severe in cases of early run­
aways. When the radical population grows under
control, the temperature profile levels off to approach a
plateau, often after a broad, yet small maximum. The
conversion increases gradually with distance when the
temperature is steady.
The other difficult phase of the reaction occurs at the
onset of gel effect. The heat generated by the autoac­
celeration of the· reaction causes the temperature to
rise. Gel effect induced (labeled GEI on the plots) run­
aways are not parametrically sensitive. Since the gel
Two entirely different mechanisms are found to be
responsible for the non-linearities exhibited by poly­
merization reactions. The hot spots observed in the
simulations are attributed to either the rapid forma­
tion of the radical population at the beginning of the
reaction or to the gel effect. The consequences of both
phenomena are amplified by the positive feedback
inherent in the kinetics. The existence of these two
separate mechanisms inducing autoacceleration has
not been clearly identified in the literature. Pre­
vious studies about non-isothermal polymeriz­
ation (Biesenberger and Capinpin, 1974; Biesenberger
et al., 1976; Brooks, 1983) ignore or neglect the gel
 Free-radical polymerization 85

Residence time (mini Residence time (mini

LOO
10 20
-4
10 or=-
---''"'o"---=2.,,0'---'3::;;0�
Co) Cc)

08
4
�
"' 10
0
0.6 s
0
c:
<=
0
�
� �
Q,)
> c:
"'
c:
0 04 '-'
u <=
0
'-'
0

-c
v

0.2 e
.,
>
� -7
S! 10
:::>

E
0 :::>
0 200 400 600 u
Axial distance Ccml

Residence time (min l 0 200 400 600

� 0'----
1800 ------'1
�
- ----'2
�o;:o...____
. --=3�0'""' Axial distance (cm)

Cb) Fig. 5. Non-isothermal polymerization of MMA. Influence

RA 5 of the bath temperature on the conversion, temperature and
EI cumulative radical concentration profiles (or histories)
160
simulated by a plug flow (or batch reactor) model.
U = 15 Btu/h ft2 °F, [AIBN]F = 0.0258 molfl, s = 0, TF
= 70°C. Tbath (0C): (1) 80, (2) 77, (3) 76, (4) 75, (5) 72.5, (6) 70.

pattern in a tubular reactor. Experimental results will

be compared with predictions of the extended models
in future publications.

Acknowledgement-This work was supported by the

National Science Foundation via CPE 8103632.

600
Axial distance (cm)
effect, because the polymerization is dead ended before
:reaching the onset of gel effect. Since the gel effect is
not yet very well understood, most analyses are still
made at isothermal conditions to isolate the influence
of free volume diminution and chain entanglements
from the thermal effects. The predictions of this model
in limiting cases (isothermal with gel effect and dead­
ended run-away at low conversion) are consistent with
the studies reported in the literature. This model can
therefore be expected to describe properly the import­
ant phenomena governing the kinetics of
isothermal polymerization reactions. These equations
will constitute the backbone of several extensions, e.g.
simulation of the molecular weight distribution and
detailed modeling of mass transfer, reaction and flow
REFERENCES
Bevington J. C., Melville H. W. and Taylor R. P., 1954, The
termination reaction in radical polymerizations--11.
Polymerizations of styrene at 60°C and of methyl meth­
acrylate at 0 and 60°C, and the copolymerization of these
monomers at 60°C. J. polym. Sci. 14 463.
Biesenberger J. A. and Capinpin R., 1974, A study of chain
addition polymerization with temperature variations-- 1.
Thermal drift and its effects on polymer properties. Polym.
Engng Sci. 14 737.
Biesenberger J. A., Capinpin R. and Yang J.C., 1976, A study
of chain addition polymerization with temperature vari­
ations--11. Thermal runaway and instability-- computer
study. Polym. Engng Sci. 16 101.
Brooks B. W., 1983, Thermal stability of polymerization
reactors. J. appl. Polym. Sci. 28 619.
non­ Carratt G. M., 1983, Free radical polymerization of methyl
metbacrylate with gel effect. M.S. Thesis, University of
California, Berkeley.
Chiu W. Y., Carratt G. M. and Soong D. S., 1983, A computer
model for the gel effect in free-radical polymerization.
Macromolecules 16 348.
86 P. E. BAILLAOOU and D. S.
Gopa]an M. R. and Santhappa M., 1957. Methyl ethyl ketone
peroxide as an initiator in the polymerization of vinyl
monomen. J. polym. Sci. 25 333.
Jaisinghani R. and Ray W. H., 1977, On the dynamic behavior
of a class of homogeneous continuous stirred tank pol­
ymerization reactors. Chem. Engng Sci. 32 811.
Lynn S. and Huff J. E., 1971, Polymerization in a tubular
reactor. A.l.Ch.E. J. 17 475.
Mahabadi H. K. and O'Driscoll, K. F., 1977, Absolute rate
constants in free-radical polymerization-III. Deter­
mination of propagation and termination rate constants
for styrene and methyl metbacrylate. J. Macronwl. Sci.
Chem. All 967.
Martin J. R., Johnson J. F. and Cooper A. R., 1972,
Mechanical properties of polymers: the influence of mol­
ecular weight and molecular weight distribution. J. macro­
nwl. Sci. Revs. Macronw/. Chem. C8 57.
Nunes R. W., Martin J. R. and Johnson J. F., 1982, Inftuence
of molecular weight and molecular weight distribution on
mechanical properties. Polym. Engng Sci. 22 205.
Perry R. H. and Chilton C. H., Eds, 1973, Chemical Engineers'
Handbook, 5th edn. McGraw-Hill, New York.
Sacks M. E., Lee S. I. and Bjesc:nberger J. A.• 1973, Effect of
temperature on molecular weight distributions: batch,
Soo NG
chain polymerization. Chem. Engng Sci. 28 241.
Sebastian D. H. and Biesenberger J. A., 1976, A study of chain
addition polymerizations with temperature variations:
thermal runaway and instability in styrene pol­
ymerizatio�n experimental study. Polym. Engng Sci. 16
117.
Sebastian D. H. and Biesc:nberger J. A., 1978, Nonisothermal
behavior and thermal runaway phenomena in chain ad­
dition copolymerization. ACS Symp. Ser. 65 173.
Stickler M. and Meyc;rhoff G., 1978, Die Thennische
Polymerisation von Methylmethacrylat-1. Pol-
ymerisation in Substanz, Makronwlek. Chem. 179 2729.
Stolin A. M., Merzhanov A.G. and Malkin A. Ya, 1979, Non­
isothermal phenomena in polymer engineering and science:
a review-I. Non-isothermal polymerization. Polym.
Engng Sci. 19 1065.
Tobolsky A. V. and Baysal B., 1953, A review of rates of
initiation in vinyl polymerization: styrene and methyl­
methacrylate. J. polym. Sci. 11 471.
Trommsdorff E., Koelhe H. and Lagally P., 1947, Zur
polymerisation des Metbacrysaeuremethylesters. Makro­
nwlek Chem. l 169.
Weast R. C.and Astle M. J., 1981, Handbook of Chemistry and
Physics, 62nd edn. CRC Press.