Pure & Appi. Chern., Vol. 53, pp. 617—626. 0033/4545/81/03061700$02.

OO
Pergamon Press Ltd. 1981. Printed in Great Britain.
© IUPAC

FREE RADICAL POLYMERIZATION KINETICS — REVISITED

Kenneth F. O'Drlscoll
Dept. of Chemical Engineering
University of Waterloo
Waterloo , Ont • , CANADA

Abstract Recent developments in the kinetic modellingof free radical poly-

merization are reviewed withanemphasisonworkinthe author's laboratory.
At low conversions the chain length dependence of the termination rate
constant has been experimentally demonstrated and theoretically modelled.
The impact of this chain length dependence on other aspects of free radical
polymerization kinetics has also been explored both theoretically and ex-
perimentally. At higher conversions the gel effect has beenmodelledwith
some success by using the concept of chain entanglement to describe the influence
of high polymer concentrationontheterminationreactionrate. The state of
knowledge concerning copolymerization kinetics is briefly discussed.

INTRODUCTION

The free radical polymerization of vinyl monomers is probably the most.common synthetic
reaction carried out by man. As such, it has been studied long and well. In broad outline
the nature of the reactions involved has been well understood for more than three decades,
and this has led many to believe that the subject is one in which there is not much new to
be found. However, interest does continue in this area because its complexity has so far
prevented anyone from giving a complete description of the course of a polymerization in
quantitative terms. Interest in modelling free radical polymerization kinetics has also
received added stimulus in recent years becuase of the need for kinetic models for computer
control of processes (1).

A complete kinetic model must be able to describe the individual reaction rates for initia-
tion, propagation, chain transfer and termination of the propagating chain. Only from such
information can one deduce the molecular weight distribution or other important properties
of the polymer chain. Conversely, only by using the molecular weight distribution (MWD) or
at least averages of it can one hope to sort out the highly coupled influences of the
various elementary reactions on the overall rate and extent of polymerization.

The common use of gel permeation chromatography (GPC) to determine MWD in the last decade
has provided the polymerization kineticist with what is now an indispensable tool for
experimental work. As a result of its use the "classical" kinetic model, based almost
exclusively on rates and number average degrees of polymerization, must be modified. In
this paper, the work done toward such a modification during the last six years in the
author's laboratory is reviewed and summarized.

When a vinyl monomer undergoes free radical polymerization the simplest possible system
consists of an initiator which is soluble in the monomer and a resultant polymer which
is also soluble in its monomer. A wide variety of techniques have been employed to measure
the kinetics of polymerization for such systems using characteristics such as shrinkage,

617
 618 KENNETH F. O'DRISCOLL

temperature rise, refractive index change, etc. which occur as polymerization does.
Necessarily, the techniques are different for different extents and degrees of polymeri-
zation. One can, however, use the results of investigations from many different laboratories
to synthesize a composite view of a simple polymerization. Such a description is given in
Figure 1.

0.I 1.0 10 I00

% CONVERSION

Fig. 1. Rate of Polymerization as a function of conversion

At the beginning there is a short period when the rate is changing and then the reaction
reaches a steady state; as the reaction advances to 10% or so, both the rate and degree of
polymerization are observed to pass through minima. This is regarded as the onset of the
"gel effect" a period when the increasing viscosity of the polymerizing system causes the
reaction to accelerate. The rate acceleration is matched by an increase in the cumulative
molecular weight of the polymer (especially the weight average) and there is a concomitant
broadening of the MWD. Ultimately, the rate of polymerization passes through a maximum
and, as the mixture becomes glassy, the rate falls to nearly zero. Since there exists an
equilibrium monomer concentration the rate must go to zero at some conversion less than 100%.

This description is semi—quantitative at best and some features of it may be missing or

difficult to observe in a given polymerization. We shall however employ it for discussion
purposes and use the words "low" and "high" to refer to conversion (of monomer to polymer)
before and after the onset of the gel effect, respectively.

THE SIMPLEST KINETIC MODEL

What has come to be called the "classical" kinetic model of free radical polymerization
kinetics involves the unimolecular decomposition of initiator, I, to produce two free
radicals, R•, a fractioñof which, f, then add monomer, M, to initiate the active polymer
chain, P.. The active chain either propagates and grows by adding more monomer or stops
its growth by undergoing chain transfer with some species Y or by terminating in a bimole-
cular reaction with another active chain. These reactions and associated rate expressions
are set forth in Scheme 1. Note that species Y may be initiator, monomer, polymer, solvent,
impurity, or deliberately added chain transfer agent.
 Free radical polymerization kinetics 619

Scheme 1

Reaction Rate expression

Initiation I + 2R kd[I]

R' +M+P1• k.[R•][M]

Propagation n' + M + + kp[Pni[M)

Transfer P • + Y + P + R' k [P •][Y]
n n tr,Y n
Termination P • + p •. +p ktC[P fl•] [Pm•]
n m n+m
"n + m kd[P•][P•]

To solve this simplest model of polymerization for useful kinetic expressions it is commonly
assumed that:

i, the radical reactivity is independent of chain length,

ii, the rate at which chains are initiated equals the rate at which they are terminated,

iii, radicals generated in chain transfer rapidly reinitiate propagating chains, and

iv, chains are very long so that the amount of monomer consumed in initiation and chain
transfer reactions is small compared to that in propagation.

Using these assumptions the rate and degree of polymerization are given by (2)
—d[M]/dt =
(k2fkd[I}/kt)1"2[M} (1)

l/n = [Y]/[M] + k2 R/[]2

ktr,y/kp (2)

The predictions of these and other equations which can be derived for the classical mode,
have been sustained in many investigations carried out at low conversions and over some
concentration variation. However, there exist many reported instances where the predictions
of the classical model have failed (3). In particular this is true at high initiator
concentration, in solution polymerization and, as expected, at high conversion of monomer
to polymer. Some of these failures are readily attributed to the need for inclusion of,
e.g., primary radical termination at high initiator concentration, or specific complex
formation not included in the simple model. However, a combination of theoretical and
experimental work in our laboratory has shown that a more fundamental problem is the first
assumption above: the termination rate constant is a function of chain length.

LOW CONVERSION KINETICS

A number of earlier workers had suspected that k was a function of chain length, but their
experiments were either done where the dependency was weak or their tools were not adequate
to the task. In particular, one might have expected that some of the chain lifetime
measuring techniques, such as the rotating sector, might have revealed the chain length
dependence. However, it is the case for all such techniques that the measurement of chain
lifetime gives only the ratio k /k , and separation of these two variables is usually done
2 pt
using a value of k /k, determined in a separate set of experiments — often done in a

separate laboratory. Therefore, if chain length dependence does exist, such experiments
will not reveal it; instead the data from them ought to be quite scattered and may explain
the enormous scatter of results one sees in literature tabulations (4).
620 KENNETH F. O'DRISCOLL

The experimental work which has established the chain length dependence of the kt is based
on a linear analog of the rotating sector measurement, which we call spatially intermittent
polymerization (SIP) (5). In the SIP apparatus the monomer plus photoinitiator is pumpi
at a measured speed through a glass capillary onto which UV light shines through a set of
evenly placed slits in a mask. Thus the monomer passes alternately through light and dark
regions in which the radical concentration rises and falls. By controlling the pump speed,
the rate of polymerization is varied. A mathematical description of this process is
formally identical to that of the rotating sector experiment, with the pump speed replacing
the (inverse) frequency of rotation. The major advantage of the SIP technique is that the
effluent from the reactor may be collected, at very low conversion, thus providing a sample
large enough for use in a GPC directly for the determination of both conversion and NWD.

(60
350

(20

LOG X,

Fig. 2 Effect of chain length on k and k for methylmethacrylate at 25°C (5)

Results typical of those we have obtained using the SIP are shown in Figure 2. It should
be noted that the chain length dependence is quite strong below a chain length of 100,
but is almost negligible above 1000. Given that most polymerizations yield a degree of
polymerization averaging 1000 or more the question might be asked, is the chain length
dependence unimportant? To answer this it must be remembered that every chain that grows
to be 1000 units in length spends lOX of its time being less than 100 units long; during
that time, its probability of terminating is much higher and therefore the distribution
of lengths of active chains is skewed, as is the final MWD, and the rate of polymerization
is also affected.

To explore further the phenomena affected by the chain length dependence of we developed
a model whose roots can be traced to the work of Benson and North (6) and, more recently
Hone et al. (7). This model begins with the premise that two radical containing polymer
coils (which will ultimately react to form dead polymer) must first undergo translational
diffusion to collide and then must undergo segmental diffusion to bring the chain ends into
a very small volume element where reaction (combination or disproportionation) takes place.
The results of this model (slightly simplified) are expressed in the following equations:

kt = F1(,T)F2(a,N) (3)

where is the segmental friction coefficient, c is the linear chain expansion coefficient
and N is the radical chain length. The detailed form of F1 works out to be approximately
 Free radical polymerization kinetics 621

F1 = l/(aN) (4)

at 300 K while F2 is more complicated:

F2 = (l-0.37/4(l-l.84N02/N2))26 (5)

In eqn. 4 am is the hydrodynamic radius of a segment, N is the chain stiffness and n is

the solvent viscosity.

Since the parameters in eqns 4 and 5 are either known or experimentally accessible, it
is possible to use the model to test experimental data for conformity and, finding this to
be true, to design and interpret further experiments. For example, the data in Table 1
show a dependence of k on the ester length for a number of alkyl methacrylates which is
completely predicted by the variance in F1 calculable from known (or estimated) values of
the parameters of eqn. 4(8).

TABLE 1
Comparison of Experimental and
Calculated Termination Rate Constants of Alkyl Methacrylates (8)

?nomer a
—
Ester of AMA
m AMA,am 1it K AMA/No Mi K Kt MMA x 102
oisej
[
Expt'l Calcd
Methyl (MMA) 0.53 1.0 1.0 100.0 100.0
Butyl 0.91 1.60 0.911 47.6 40.0
Octyl 1.92 1.96 1.093 12.45±4.25 12.88
Dodecyl 3.96 2.84 1.62 2.86±0.9 2.91
Cetyl 7.52 3.95 2.35 0.762±0.19 0.76

The conversion dependence of kt is also predictable by this model in terms of the influence
of polymer on the coil sizer as expressed by the chain expansion coefficient, (9, 10). In
general, anything which reduces the solvent's thermodynamic quality towards the polymer
will cause the coil to shrink. In the smaller coil, the chain end will be found more
easily by the other chain end, and therefore k will be larger. Therefore any reduction
in solvent quality should result in a reduced rate of polymerization. As a polymerization
advances the presence of more polymer makes the solution a thermodynamically poorer one
and therefore produces a slower rate and a lower degree of polymerization. The latter
was first noted experimentally by Schulz and Harborth more than 30 years ago (11), while
North and Reed observed the decrease in k 20 years ago (12); the latter's data are
quantitatively interpreted by an extension of eqn 3 as shown in Figure 3. Furthermore,
the linear dependence of k on polymer concentration seen in Figure 3 is quite general
at low polymer concentrations, and the experimentally observed slopes of the line are
predicted to be simply proportional to chain length. Verifiction of this prediction
has been seen for methylmethacrylate and, as shown in Figure 4, also for styrene.

Other applications of this model have shed some light on the importance of primary
radical termination (13), the influence of pressure on the termination reaction (14, 15)
and the influence of the chain length dependence of k on the quantitative measurement
of chain transfer coefficients (16). From this work has arisen the appreciation of a
somewhat disturbing fact: the use of experimental variables in free radical polymerization
is not as simple as we once thought it was Whenever a variable is changed (such as the
initiator or chain transfer agent concentration) the rate of polymerization is\changed;
this causes a change in the degree of polymerization which in turn causes a change in k.
622 KENNETH F. O'DRISCOLL

Therefore the interpretation of the effect of the change in the original variable is
confounded by the change in k. Separating these two concurrent changes is possible
but requires more care than has usually been given to the interpretation of results
using the "classical" model or variants on it.

0
0 a,

VOLUME FRACTION, Vp 0 1.0 20

Fig. 3 Effect of polymer on k in MMA Fig. 4 Effect of polymer molecular weight

polymerization. Solid line predicted on slopes of plots such as Fig. 3

by theory (9). for styrene (10).

If one takes care to do all experiments at similar degrees of polymerization or to correct

data for changes in chain length then results which are both accurate and precise can be
obtained (17). Figure 5 for example shows the temperature dependence of k and k for
the two most studied monomers, styrene and methylmethacrylate (MMA), which may be
compared with the enormous scatter seen in the tabulated literature (4). Another example
may be seen in Figure 6 where the lumped rate constant for primary radical concentration
obtained with experiments done at similar degrees of polymerization shows the expected
Arrhenius dependence, whereas those data obtained with no attention to chain length gave
a curvature which led to a qualitative misunderstanding of the termination process.

400
—U E
300

40

200
b
o!. 35

KtxlO6

30

100

75
25

50

3.2 33 3A 3.5
io- riic'

Fig. 5 Variation of k and k with temper- Fig. 6 Variation of PRT constant with
ature for styrene and NMA at constant temperature at constant (filled
chain length (17). circles) and variable (open circles)
chain length (13)
 Free radical polymerization kinetics 623

HIGH CONVERSION KINETICS

Given the quantitative understanding of the influence of segmental mobility and polymer
coil size on the kinetics of low conversion polymerization, the question then arises can
we develop a similar understanding of the nature of the gel effect, the autoacceleration
that appears as the reaction mixture becomes viscous and leads rapidly to high conversion.
To do so we begin by recognizing that the onset of the gel effect represents a change in
the nature of the diffusion process which is the rate controlling step in the termination
reaction (18). This is shown schematically in Figure 7 where the rate constant for the
overall process of termination, kt, is viewed as a sum of the constants describing
segmental diffusion, k5, and translational diffusion, kT. As described above, k5 is
expected to increase linearly with conversion, conversely translational diffusion is
expected to decrease with increasing reaction viscosity and, when chain entanglement
becomes dominant, will do so in an explosive manner. This leads to the observed maximum

9
k

Cl
0

GEL

0 I0 20
CON VERS ION

Fig. 7 Schematic showing combined effects of segmental and translational

diffusion processes on k as function of conversion.

To model this phenomenon, we have proposed (19) that the active polymer chains in moderately
concentrated solution are either small enough to be regarded as mobile or large enough
to be regarded as entangled and thus of restricted mobility. The small chains are regarded
as having the 'sane value of kt as in more dilute solution, while the large entangled
chains terminate with a rate constant k . Cross termination between large and small
te
chains is regarded as occurring with a rate constant which is the geometric mean of k
and k. The chain size which separates "small" from "large" is n and given by

Kc = (6)

where Kc is a chain entanglement constant, familiar to those who study concentrated

polymer solutions and polymer melts, and is the volume fraction of polymer in the
reaction mixture. Applying conventional steady state approximations this model yields
for the rate of polymerization,

R /R
p p,o
—l = ((l/) — l)exp(—ncv) (7)

where R is the rate of polymerization (given in eqn 1) and v is the kinetic chain
length, both of which would be obtained in the absence of the gel effect and
PAAC 53:3 - B
624 KENNETH F. O'DRISCOLL

8 = (kte/kt)2 (8)

z
2
U) _______
2
>
z0
03
2
2
U U
w
I.-

(J (. 04 04 04 0
TINE (mm.) FRACTIONAL CONVERSION

Pig. 8 Bulk polymerization of methyl Fig. 9 Molecular weight variation with

methacrylate; solid line predicted conversion in polymerization of

by eqn. 7. methyl methacrylate. Solid line

predicted by theory (19).

Using the same parameters the leading moments of the MWD can also be described. Figures
8 and 9 show typical examples of the success of this model, which has only the two
adjustable parameters Kc and 8 in describing the gel effect quantitatively. The temperature
dependence of these two parameters has been measured and the model rewritten in a some-
what more convenient form so that non—isothermal polymerizations could be described (20).
To do so, two other parameters were defined in terms of the existing kinetic constants:

C1 = [((1 + e)/c2K)([M] /—2e)J (9)

C2 (2k/k[M})K/(1 + e) (10)

where e is the fractional shrinkage accompanying polymerization. In these terms the

rate of polymerization becomes

R/R—l = (11)
(C1g1—l) exp— (C2g2)
where g1 and g2 are functions only of conversion; it should be emphasized that C1 and
C2 are not functions of anything except the physical and chemical characteristics of
the monomer.

An interesting consequence of eqn. 11 is that it can be used to understand why styrene

has so much weaker a gel effect than methylmethacrylate does. Note that C2 appears in
a negative exponential term and that the experimental value for k1"2/k is about six-
fold smaller for styrene than for MMA. In plain language this means that if a poly-
merization of styrene is carried out at the same initial rate as one for MMA the latter
will produce a higher molecular weight polymer and thus become entangled at a lower
conversion the gel effect will start sooner and be more pronounced because of the greater

entanglement.

Useful though these equations have been, the model has proved to have some very real
limitations. It does not properly describe polymerizations carried out in solution,
probably because it does not incorporate any measure of the free volume change which
occurs during the polymerization. Such a change must be different in solution polymer—
izaton from what it is in bulk. Also the model ignores any contribution from primary
radical termination which has been calculated to become of importance as the chain end
 Free radical polymerization kinetics 625

concentration increases and the monomer concentration decreases during the advanced
stages of the gel effect.

A major limitation of the model, one which it was never intended to handle, is the
highest level of conversion where, if the temperature is low enough, the monomer—polymer
mixture becomes glassy. Under such conditions, the propagation and initiation reactions
become diffusion controlled. An approach to treating this problem has been presented by
Hamielec (21). Even if the reaction temperature is higher than the polymer glass transition
temperature, the highest level of conversion must ultimately involve the reversibility
of the propagation reaction. Little work has been done in this area where the kinetics
are controlled by the thermodynamics. However, in an era when trace amounts of suspect
chemicals are of concern in our environment, more attention must be given to poly-
merization at the highest possible level of conversion.

COPOLYMERIZATION KINETICS

If our quantitative understanding of the details of homopolymerization kinetics is as

limited as outlined above, that of copolymerization is woeful. Relative rates of
copolymerization have usually been expressed in terms of reactivity ratios, thousands
of which have been measured and tabulated. Unfortunately, most of the measurements have
been done with a complete disregard for the non—linear character of the equations used
and the statistical validity of most methods for deriving the reactivity ratios is poor.
This problem, it has been recently shown (22), can easily be overcome by proper design
of experiments. Unfortunately, this realization does nothing to improve the quality of
the many data in the literature.

Absolute rates of free radical copolymerization have been described using equations
analogous to homopolymerization, the major difficulty being how to describe the
termination reaction. Rudin has surveyed the various treatments (23). It is clear that,
whichever may be the proper way to describe the termination reaçion in copolytnerization,
it must recognize the diffusion control in the process. For this reason, the use of a
phi factor, which implies chemical control, is not acceptable. The problem of finding
a correct description at low conversion which embodies the segmental motion of a chain
of variable and random composition remains unsolved (24).

At high conversion, recent work by Johnson, Karmo and Smith (25) has shown that the
simple Mayo—Lewis model of four propagation reactions, each with its own propagation rate
constant, may not be able to explain the drift of composition with conversion in the
copolymerization of styrene with methylmethacrylate. This observation has been confirmed
in our laboratory (26) and also noted in the copolymerization of styrene with butyl—
methacrylate (27). It also appears in terpolymerizations of similar monomers. Pichot
et al. have shown that a similar problem exists in acrylonitrile—styrene copolymerization
(28) and have suggested that selective solvation may be the cause. Whatever the cause may prove to
be it appears that there is still a lot to learn about free radical polymerization kinetics.

ACKNOWLEDGEMENT

Suppott of the research described in this review by the Natural Science and Engineering
Research Council of Canada is appreciated. The writing of this paper would not have been
possible without the extensive experimental and theoretical contributions of Drs. M Izu,
J. Cardenas, H. Mahabadi and J. Cardenas. Permission from the appropriate publishers to
reproduce Figs. 2—9 is also acknowledged.
626 KENNETH F. 0'DRISCOLL

REFERENCES

1. J.N. Hendersonand T.C. Bouton, "Polymerization Reactors and Processes", Ed. ACS
Symp. Series 104, 1979.

2. G.C. Eastmond, Chap. 1 in "Comprehensive Chemical Kinetics" V. l4A, C.H. Bamford

and C.F.H. Tippet, Ed. Elsevier, Amsterdam (1976).

3. G.E. Scott and E. Senogles, Rev. Macromol. Chem. 10, 49, (1973).

4. R. Korus and K.F. O'Driscoll, Polymer Handbook p. 11—46, 2nd Ed., J. Brandrup and
E.H. Immergut Eds., Wiley, N.Y. (1974).

5. K.F. O'Driscoll and H.K. Mahabadi, J. Polym. Sci. Polym. Chem. Ed., U, 869 (1976).

6. S.W. Benson & A.M. North, J.Amer. çhern. Soc. 81, 1339, (1959).

7. K. Hone, I. Mito and H. Kambe, J. Polym. Sci. 2 A—i, 7, 2561, (1969).

8. H.K. Mahabadi and K.F. O'Dniscoll, J. Polym. Sci. Polym. Lett. Ed. 14, 671, (1976).

9. H.K. Mahabadiaand K.F. O'Dniscoll, Macromol. 10, 55, (1977).

10. H.K. Mahabadi and K.F. O'Driscoll, J. Polym. Sic. Polym. Lett. Ed. 16, 351, (1978).

11. G.V. Schulz and G. Harborth, Makromol. Chem. 1, 106, (1947).

12. A.M. North and G.A. Reed, Trans. Faraday Soc. 57, 859, (1961).

13. H.K. Mahabadi and K.F. O'Driscoll, Makromol. Chem. 178, 2629, (1977).

14. K.F. O'Driscoll, Makromol. Chem. 178, 899, (1977).

15. K.F. O'Driscoll, Makromol. Chem. 180, 2053, (1979).

16. H.K. Mahabadi, Ph.D. Thesis, Univ. of Waterloo, 1976.

17. H.K. Mahabadi and K.F. O'Driscoll, J.Macrornol. Sci. Chem. All, 967, (1977).

18. J. Dionisio, H.K. Mahabadi, K.F. O'Driscoll, E. Abuin and E.A. Lissi, J.PoijTm.
Sci._Polym. Chem. Ed. 17, 1891, (1979).

19. J.N. Cardenas and K.F. O'Driscoll, J. Polym. Sci.Polym. Chem. Ed. 14, 883, (1976).

20. K.F. O'Driscoll, J.M. Dionisio and H.K. Mahabadi, p. 361 in ref. 1.

21. F.L. Marten and A.E. Hamielec, ACS Symp. Series 104, 43, (1979).

22. R.C. McFarlane, P.M. Reilly and K.F. O'Driscoll, J. Polym. Sci. Poiym. Chem. Ed. 18,
251, (1980).

23. S.S.M. Chiang and A. Rudin, J.Macromol. Sci. A9, 237, (1975).

24. K. Ito and K.F. O'Driscoll, J. Polym. Sci. Polym. Chem. Ed. 17, 3913, (1979).

25. M. Johnson, T.S. Karmo and R.R. Smith, Eur. Polym. J. 14, 409, (1978).

26. J.M. Dionisio and K.F. O'Driscoll, J. Po]ym. Sci. Poly. Lett. Ed. 17, 701, (1979).

27. 5. Balke and R.D. Patel, Recent Advances in Size Exclusion Chromatogrphy,
ACS Symposium Series (in press), (1980).

28. C. Pichot, A. Guyot and C. Strazielle, J. Polym. Sci. Polym. Chem. Ed. 17, 2269,
(1979).