Vous êtes sur la page 1sur 15

See

discussions, stats, and author profiles for this publication at:


http://www.researchgate.net/publication/229366456

Kinetic simulation of living carbocationic


polymerizations. II. Simulation of living
isobutylene polymerization using a
mechanistic model

ARTICLE in EUROPEAN POLYMER JOURNAL · JANUARY 2005


Impact Factor: 3.24 · DOI: 10.1016/j.eurpolymj.2004.08.006

CITATIONS DOWNLOADS VIEWS

13 63 123

5 AUTHORS, INCLUDING:

Kimberley McAuley Gabor Kaszas


Queen's University Goodyear
142 PUBLICATIONS 2,473 CITATIONS 48 PUBLICATIONS 624 CITATIONS

SEE PROFILE SEE PROFILE

Available from: Kimberley McAuley


Retrieved on: 23 June 2015
EUROPEAN
POLYMER
European Polymer Journal 41 (2005) 1–14
JOURNAL
www.elsevier.com/locate/europolj

Kinetic simulation of living carbocationic polymerizations.


II. Simulation of living isobutylene polymerization
using a mechanistic model
Judit E. Puskas a,*, Sohel Shaikh b, Kevin Z. Yao c,
Kim B. McAuley c, Gabor Kaszas d
a
Macromolecular Engineering Research, Department of Polymer Science, The University of Akron, Akron,
OH 44325-3909, USA
b
Department of Chemical and Biochemical Engineering, The University of Western Ontario, London, ON, Canada N6A
c
Department of Chemical Engineering, Queen’s University, Kingston, ON, Canada K7L 3N6
d
Lanxess Inc., 1265 Vidal St. South, Sarnia, ON, Canada N7T 7M2

Received 14 July 2004; received in revised form 30 July 2004; accepted 2 August 2004
Available online 25 September 2004

Abstract

This paper discusses the kinetic simulation of TiCl4––coinitiated living carbocationic isobutylene (IB) polymeriza-
tions governed by dormant-active equilibria, using a mechanistic model. Two kinetic models were constructed from
the same underlying mechanism: one using a commercial simulation software package (Predici), and the other using
the method of moments. Parameter estimation from experimental batch reactor data with Predici yielded a rate con-
stant of propagation kp = 4.64 · 108 ± 2.75 · 108 L/mol s, with no constraints imposed. This agrees with kp data meas-
ured with diffusion clock and competition methods, but disagrees with kinetically obtained kp values. Estimation of rate
constants with Predici and the GREG parameter estimation software packages revealed that it was difficult to estimate
the complete set of kinetic parameters, due to correlated effects of the parameters on model predictions. Estimability
analysis confirmed that some of the strongly correlating parameters could not be estimated simultaneously using the
available experimental data. Using kp = 6 · 108 ± 2.75 · 108 L/mol s measured by Mayr, and using starting estimates
of other rate constants defined by experimentally observed correlations, yielded the set of rate constants required for
the simulations. Both kinetic models yielded good agreement with experimental data, with the exception of Mw values
that slightly diverged from the theoretically predicted MwMn = constant relationship. This may indicate the occur-
rence of a minor side reaction. However, the kp/k1 = 17.5 L/mol average run length calculated from measured and sim-
ulated MWD data agrees well with earlier literature values.
 2004 Published by Elsevier Ltd.

Keywords: Living carbocationic polymerization; Isobutylene; Modeling; Kinetics; Mechanism; Run length

*
Corresponding author. Tel.: +1 330 972 6203; fax: +1 330 972 5290.
E-mail address: jpuskas@uakron.edu (J.E. Puskas).

0014-3057/$ - see front matter  2004 Published by Elsevier Ltd.


doi:10.1016/j.eurpolymj.2004.08.006
2 J.E. Puskas et al. / European Polymer Journal 41 (2005) 1–14

Nomenclature

I initiator, 2-chloro-2,4,4-trimethyl-pentane k0 rate constant of intermediate dissociation,


(TMPCl) s1
IB isobutylene K0 k0/k0, equilibrium constant of intermediate
LA Lewis acid (TiCl4––titanium tetrachloride) formation, L/mol
M monomer, isobutylene (IB) k1 ionization rate constant of form active spe-
PIB polyisobutylene cies with monomeric gegenion, s1
MeCl methyl chloride k1 deactivation rate constant of active species
MeCHx methylcyclohexane with monomeric gegenion, s1
Hx hexane K1 k1/k1, equilibrium constant of the forma-
DtBP di-tert-butylpyridine, proton trap tion of active species with monomeric
ED electron pair donor gegenion
[] concentrations of individual species, mol/L k2 ionization rate constant of form active spe-
[ ]0 initial concentrations of individual species, cies with dimeric gegenion, L/mol s
mol/L k2 deactivation rate constant of active species
MW molecular weight with dimeric gegenion, s1
Mn number average molecular weight K2 k2/k2, equilibrium constant of the forma-
Mw weight average molecular weight tion of active species with dimeric gegenion,
DPn number average chain length L/mol
DPw weight average chain length u0 zeroth moment of dormant polymer chains
MWD M w =M n , molecular weight distribution uc0 zeroth moment of intermediate complex
I * LA initiator/Lewis acid intermediate ud0 zeroth moment of polymer chains with
I+LA active initiator with monomeric gegenion dimeric counterions
Iþ LA 2 active initiator with dimeric gegenion um0 zeroth moment of polymer chains with
PIB-Cl dormant tertiary chloride capped polyiso- monomeric counterions
butylene chain ut0 total zeroth moment (u0 + uc0 + ud0 + um0)
Pn * LA polymer/Lewis acid intermediate with chain u1 first moment of dormant polymer chains
of n, n P 3 uc1 first moment of intermediate complex
Pn-Cl dormant, tert-chloride capped polymer ud1 first moment of polymer chains with dimeric
chain of length n, where n P 3 counterions

Pþ n LA active growing chain of length n, with mono- um1 first moment of polymer chains with mono-
meric Lewis acid gegenions n P 3 meric counterions

Pþ n LA2 active growing chain of length n, with dimer ut1 total first moment (u1 + uc1 + ud1 + um1)
Lewis acid gegenions n P 3 u2 second moment of dormant polymer chains
kp propagation rate constant of polymeriza- uc2 second moment of intermediate complex
tion, L/mol s ud2 second moment of polymer chains with
k 00p overall propagation rate constant of polym- dimeric counterions
erization, L3/mol3 s um2 second moment of polymer chains with
k0 rate constant of intermediate formation, L/ monomeric counterions
mol s ut2 total second moment (u2 + uc2 + ud2 + um2)

1. Introduction uniform polymers with living/controlled carbocationic


polymerization [3]. By the mid-1990s, living/controlled
Living polymerization is characterized by the ab- radical polymerization was also a reality [4]. Living
sence of significant chain transfer or irreversible chain polymerization provides control over the molecular
termination reactions. Living conditions were first weight (MW) and the architecture of the polymer
achieved in anionic polymerization [1]. Essentially liv- produced, and therefore is a preferred method of
ing conditions in carbocationic systems had been macromolecular engineering [5–7]. In both living carbo-
achieved experimentally as early as 1974 [2]. Systematic cationic and radical polymerizations there is a reversi-
research in Kennedys group in the 1980s led to the suc- ble equilibrium between dormant and active chain
cessful production of high molecular weight and nearly ends [7–10].
J.E. Puskas et al. / European Polymer Journal 41 (2005) 1–14 3

Pn Pn I + LA

In the living carbocationic polymerization of isobutylene


k-0 k0 Path B
(IB), it has been demonstrated that the active initiating Path A
and propagating centers are paired ions [9,11].
I*LA
PIB-Cl þ LA PIB == LA + LA
k-1 k-2 k2
where PIB-Cl is the dormant PIB chain capped with a k1
I+LA- I+LA2-
tertiary chloride group, and LA denotes the Lewis acid
coinitiator. The polymerization rate of IB is first-order
+M kp +M kp
in monomer, but under certain conditions apparent zero
order has also been reported [12,13]. The rate is first-
+M +M
order in initiator for a variety of initiating systems kp Pn+LA- Pn+LA2-
kp
[14–16]. This implies that the propagation is a simple k2
k1 k-1 + LA
bimolecular reaction between active chains and mono- k-2
mer. However, the polymerization rate has shown both
Pn*LA
first-order and second-order dependence on Lewis acid
(TiCl4) [15–26]. This close to second-order behavior
k0 k-0
was attributed to the fact that TiCl4 may form bimolec-
ular complex gegenions [25,26]. It was also suggested
that the propagation involves both monomeric and Pn + LA
dimeric counterions [27]. The rate of polymerization
Scheme 1. Comprehensive mechanism proposed for living IB
decreases with increasing temperature. This phenome-
polymerization.
non manifests itself as an ‘‘apparent negative activation
energy’’ [14,22,28,29]. These facts indicate that the
polymerization rate is a complicated function of all
the factors during a polymerization. When interpreting kp
the influences of these factors on the polymerization K1,eq=k1/k-1 +M
rate, the equilibrium between active and dormant chains
plays a key role. The shift of the equilibrium is governed k1 kp k-1
by many factors, including concentrations of initiator, I + LA I* Pn* Pn + LA
k-1 +M
k1
Lewis acid, monomer, additives, such as proton trap
and electron pair donors (EDs), as well as temperature Scheme 2. Simplified model of living IB polymerization.
and solvent polarity [14–26].
The comprehensive mechanistic model, shown in
Scheme 1, was conceived by Puskas group [15,25,26] 2. Kinetic background
for 2-chloro-2,4,4-trimethylpentane (TMPCl)/TiCl4 ini-
tiated living IB polymerization. In this model system, 2.1. The proposed scheme
the structure of the TMPCl initiator mimics that of the
growing polymer chains. It has been demonstrated that The proposed comprehensive mechanism [15] in-
if ki  kp is assumed for this system, it greatly simplifies volves a series of consecutive and competitive reactions.
the kinetic derivations [17]. Subsequently this assump- Firstly, the initiator and Lewis acid form an intermedi-
tion was verified experimentally; kp = 6 ± 4 · 108 L/mol s ate, I * LA, through an equilibrium involving rate con-
was measured by the diffusion clock method [30] and stants k0 and k0. This intermediate species can
with kp/k1 = 16.5, kp  3ki [17,23]. In previous papers, undergo two reactions as shown by the Paths A and
a Predici simulation of a simplified version of the ki- B. In Path A, this intermediate species becomes a mono-
netic model, as shown in Scheme 2, was described meric active initiator through an equilibrium involving
[25,26]. The rate constants that were used in the simple rate constants k1 and k1. The monomeric active initia-
model were combinations of the true kinetic constants tor can polymerize monomers, leading to growth of the
in the proposed comprehensive model. The current pa- active chain. An active growing chain of any length, n,
per is particularly aimed at deconvoluting appropriate can transform back to a polymer/Lewis acid intermedi-
values for the true constants, regardless of the different ate through an equilibrium involving k1 and k1. In Path
experimental conditions (e.g., different initiator and B, the initiator/Lewis acid intermediate can incorporate
Lewis acid concentrations), and using them in the com- an additional Lewis acid to form a dimeric active initia-
prehensive mechanistic model to simulate living IB tor through an equilibrium involving rate constants k2
polymerization. and k2. The dimeric active initiator can then propagate
4 J.E. Puskas et al. / European Polymer Journal 41 (2005) 1–14

by adding monomer. It has been shown that the dimeric nature of the proposed intermediates, I * LA and
gegenion must form by a two-step process [11,24,31]. Pn * LA. Several investigators have suggested the exist-
Halogenated titanium compounds are known to form ence of various intermediates and suggested pathways
neutral dimers in a variety of crystal structures and for their formation. For example, the formation of
also under cryogenic conditions, but with a very lim- polarized (stretched or activated, more-covalent-than-
ited stability range [32,33]. The existence of neutral ionic) dipole intermediates in the reaction of TMPCl
dimeric Ti2Cl8 under dilute polymerization conditions or PIB-Cl with TiCl4 (Winstein spectrum) was sug-
([TiCl4]0  102–101 mol/L) has never been demon- gested by various researchers [17,45]. Plesch suggested
strated experimentally. In Scheme 1, the dimer forma- the formation of a monomer-solvated carbocation
tion follows a two-step course as suggested earlier intermediate [43] and in this case high monomer con-
[11,24,31]. In the first step, the initiator/Lewis acid or centration would result in first-order propagation with
polymer/Lewis acid intermediate dissociate forming the intermediate. Sigwalt argued that this suggestion
monomeric counteranions, which in turn can react with was not convincing since low kp values were also ob-
additional Lewis acid to form dimeric counteranions. served at low monomer concentrations [46,47], and in
The species with dimeric counteranion can propagate turn he suggested a two-step propagation with the
with monomer, or can also release Lewis acid and trans- formation of solvated carbocationic intermediates
form back to an initiator/Lewis acid or polymer/Lewis [48]. With this, the apparent second-order rate con-
acid intermediate by an equilibrium reaction involving stant obtained in kinetic experiments would be
k2 and k2. The polymer/Lewis acid intermediate, kp,app = KSM * kp where KSM is the constant for the
Pn * LA, formed from both Paths A and B can undergo equilibrium between solvated carbocation and mono-
equilibrium reactions to produce dormant chains, Pn, mer-complexed solvated carbocation, this latter pro-
and free Lewis acid. pagating by unimolecular rearrangement. Scheme 1
The mechanism in Scheme 1 accounts for a number proposes the formation of polymerization-active carbo-
of important experimental observations: the depend- cations via intermediates involving the initiator and/or
ence of polymerization rate on Lewis acid can be either the dormant polymer chain and the Lewis acid. It
first or second order, or in between, depending on the may be reasonable to assume that these intermediates
actual Lewis acid concentration and other experimental are polarized species. Following the formation of
 
conditions [17–26]; the true propagation rate constant Pþ þ
n == LA or Pn == LA2 propagation could proceed
kp, can be as high as 109 L/mol s as determined by dif- via a collision of these species with monomer as shown
fusion clock methods [30,34–36] while the kinetically- in Scheme 1, or via a two-step reaction as suggested by
determined rate constant (a combination of K0K1kp Sigwalt: first by monomer solvation, followed by prop-
and K0K2kp in Scheme 1) can be as low as 104 L/mol s. agation via rearrangement. In this latter case the kp in
Table 1 lists the high and low values of kp obtained by Scheme 1 would also represent a composite rate con-
various research groups using different experimental stant involving the monomer solvation equilibrium
techniques. The inconsistency in reported kp values constant and the true rate constant of propagation
has been discussed by Plesch [41,43] and still remains (KSM * kp). Scheme 1 and Sigwalts model show similar-
unresolved [44]. While the mechanistic model shown ities in terms of reasoning that the discrepancy between
in Scheme 1 provides a plausible solution to this appar- kinetically obtained rate constants and those measured
ent discrepancy, as yet there has been no direct experi- by the diffusion clock method could be due to various
mental confirmation of the existence and the chemical pre-equilibria.

Table 1
kp values in carbocationic IB polymerizations obtained by various methods
kp (L/mol s) Initiating system Solvent T/C Reference
3
6 · 10 AlBr3/TiCl4 Heptane 14 [37]
7.9 · 105 Light/VCl4 In bulk 20 [38]
1.2 · 104 Et2AlCl/Cl2 CH3Cl 48 [39]
9.1 · 103 Ionizing radiation CH2Cl2 78 [40,41]
1.5 · 108 Ionizing radiation In bulk 78 [42]
6 · 108 R-Cl/TiCl4 CH2Cl2 78 [30]a
7 · 108 IB n-mer/TiCl4 Hexanes/CH3Cl 80 [34]a
4.7 · 108 TMP-Cl/TiCl4 Hexanes/CH3Cl 80 to 40 [35]a
1.7 · 109 TMP-Cl/BCl3 Hexanes/CH3Cl 80 to 40 [36]a
a
Diffusion clock/competition experiments.
J.E. Puskas et al. / European Polymer Journal 41 (2005) 1–14 5

As mentioned earlier, this paper discusses the decon- Table 2


volution of kinetically obtained rate constants using the Reactions
mechanistic model of Scheme 1. I + LA M I * LA k0, k0
I * LA M I+LA k1, k1
2.2. Kinetic modeling I  LA þ LA $ Iþ LA 2 k2, k2
Iþ LA þ M ! Pþ n LA

ki  kp
Based on Scheme 1, a kinetic model was developed I LA2 þ M ! Pn LA
þ  þ
2 ki  kp
 
using Predici, an open-ended integrated polymerization Pþ þ
n LA þ M ! Pnþ1 LA kp
Pn LA2 þ M ! Pnþ1 LA
þ  þ
kp
software that is based on the discrete Galerkin h-p 
2
Pþn LA ! Pn  LA k1
Method [49]. It allows unrestricted input of reaction rate þ
Pn  LA ! Pn LA 
k1
constants, different types of polymer species, monomers Pþ 
n LA2 ! Pn  LA þ LA k2
and reactants, reaction steps and polymerization condi- Pn  LA þ LA ! Pþ 
n LA2 k2
tions such as temperature and pressure [50]. During a Pn * LA ! Pn + LA k0
typical dynamic simulation, the formation or consump- Pn + LA ! Pn * LA k0
tion of any of the individual species can be followed
either with respect to time or conversion. It also has
parameter estimation capabilities; based on a given material balance equations for each individual species.
kinetic model and sufficient experimental data, rate con- Because we have used TMPCl, essentially a dimer PIB,
stants can be estimated. Within the Predici environment, as the initiator in the experiments, we refer to the first
the series of reactions shown in Table 2, was written. polymeric species formed by propagation as P3 (Table
Independently, a set of ordinary differential equa- 3: Eqs. (g), (i) and (m)).
tions (ODEs) were derived as shown in Table 3 [51]. The method of moments was used to determine the
The model equations are composed of a set of dynamic weight and number average molecular weights of the

Table 3
Material balance equations for the comprehensive model
Low molecular weight species:
d½I
¼ k 0 ½I ½LA þ k 0 ½I  LA (a)
dt
d½I  LA
¼ k 0 ½I ½LA  k 2 ½LA ½I  LA  ðk 0 þ k 1 Þ½I  LA þ k 1 ½Iþ LA þ k 2 ½Iþ LA
2 (b)
dt
þ 
d½I LA
¼ k 1 ½I  LA  ðk 1 þ k p ½M Þ½Iþ LA (c)
dt
d½Iþ LA 2
¼ k 2 ½I  LA ½LA  ðk 2 þ k p ½M Þ½Iþ LA 2 (d)
dt
d½M X 1 X1
¼ k p ½M ð½Iþ LA þ ½Pþ LA þ ½Iþ LA
i¼3 i 2 þ ½Pþ LA
i¼3 i 2 Þ (e)
dt
d½LA X 1 X 1
¼ k 0 ð½I  LA þ ½P  LA Þ  k 0 ½LA ð½I þ
i¼3 i
P Þ  k 2 ½LA ð½I  LA
i¼3 i
dt P1  P1 þ  (f)
þ i¼3 ½Pi  LA Þ þ k 2 ð½I þ LA2 þ i¼3 ½Pi LA2 Þ

High molecular weight species:



d½Pþ3 LA
¼ k p ½M ð½Iþ LA  ½Pþ  þ 
3 LA Þ  k 1 ½P3 LA þ k 1 ½P3  LA (g)
dt
þ 
d½Pi LA   
¼ k p ð½Pþ þ þ
i1 LA  ½Pi LA Þ  k 1 ½Pi LA þ k 1 ½Pi  LA i ¼ 4    1 (h)
dt
þ 
d½P3 LA2
¼ k p ½M ð½Iþ LA þ  þ 
2  ½P3 LA2 Þ  k 2 ½P3 LA2 þ k 2 ½P3  LA ½LA (i)
dt
þ 
d½Pi LA2   
¼ k p ½M ð½Pþ þ þ
i1 LA2  ½Pi LA2 Þ  k 2 ½Pi LA2 þ k 2 ½P3  LA ½LA i ¼ 4    1 (j)
dt
d½P3  LA  
¼ k 1 ½Pþ þ
3 LA  ðk 0 þ k 1 þ k 2 ½LA Þ½P3  LA þ k 2 ½P3 LA2 þ k 0 ½P3 ½LA (k)
dt
d½Pi  LA  
¼ k 1 ½Pþ þ
i LA  ðk 0 þ k 1 þ k 2 ½LA Þ½Pi  LA þ k 2 ½Pi LA2 þ k 0 ½Pi ½LA i ¼ 4    1 (l)
dt
d½P3
¼ k 0 ½P3  LA  k 0 ½P3 ½LA (m)
dt
d½Pi
¼ k 0 ½Pi  LA  k 0 ½Pi ½LA i ¼ 4    1 (n)
dt
6 J.E. Puskas et al. / European Polymer Journal 41 (2005) 1–14

Table 4
Zeroth, first and second moments of polymer species
Intermediate complexes Polymer chains with monomeric counterions
P P
Zeroth moment uc0 ¼ ½I  LA þ 1 i¼3 ½Pi  LA um0 ¼ ½Iþ LA þ 1 þ
i¼3 ½Pi LA

P P1
First moment uc1 ¼ 2½I  LA þ 1 i¼3 i½Pi  LA um1 ¼ 2½I LA þ i¼3 i½Pi LA
þ  þ
P 2 P 2 þ 
Second moment uc2 ¼ 22 ½I  LA þ 1 i¼3 i ½Pi  LA um2 ¼ 22 ½Iþ LA þ 1 i¼3 i ½Pi LA

Polymer chains with dimeric counterions Dormant polymer chains


P1 þ  P
Zeroth moment ud 0 ¼ ½Iþ LA 2 þ i¼3 ½Pi LA2 u0 ¼ ½I þ 1 i¼3 ½Pi
P1 P1
First moment ud 1 ¼ 2½I LA2 þ i¼3 i½Pþ LA
þ 
2 u1 ¼ 2½I þ i¼3 i½Pi
P1 2 þ  P 2
Second moment ud 2 ¼ 22 ½Iþ LA2 þ i¼3 i ½Pi LA2 u2 ¼ 22 ½I þ 1 i¼3 i ½Pi

Table 5 types of polymeric species. Instead, the total of all of the


Moment balance equations second moments can be determined from zeroth and first
dum0 moments:
¼ k 1 um0 þ k 1 uc0 (a)
dt
dud 0 dut2
¼ k 2 ½LA uc0  k 2 ud 0 (b) ¼ k p ½M ð2um1 þ um0 þ 2ud 1 þ ud 0 Þ ð3Þ
dt dt
duc0
¼ k 1 um0  ðk 1 þ k 0 þ k 2 ½LA Þuc0
dt (c) The number average and weight average degree of
þk 2 ud 0 þ k 0 ½LA u0
polymerization can be expressed by three leading mo-
du0
¼ k 0 uc0  k 0 ½LA u0 (d) ments. As the polymer chains are a combination of four
dt
dum1 types of polymer species, sums of the individual mo-
¼ k p Mum0  k 1 um1 þ k 1 uc1 (e) ments are required for the calculation of degree of
dt
dud 1 polymerization:
¼ k p ½M ud 0  k 2 ud 1 þ k 2 ½LA uc1 (f)
dt
duc1 ut1 um1 þ ud 1 þ uc1 þ u1
¼ k 1 um1  ðk 1 þ k 0 þ k 2 ½LA Þuc1 (g) DPn ¼ ¼
dt ut0 um0 þ ud 0 þ uc0 þ u0
þk 2 ud 1 þ k 0 ½LA u1
du1 2½I 0 þ ½M 0  ½M
¼ k 0 ½LA u1 þ k 0 uc1 (h) ¼ ð4Þ
dt ½I 0
d½M
¼ k p ½M ðum0 þ ud 0 Þ (i) ut2 ut2
dt
d½LA DPw ¼ ¼ ð5Þ
¼ k 0 uc0  k 0 ½LA u0  k 2 ½LA uc0 þ k 2 ud 0 (j) ut1 um1 þ ud 1 þ uc1 þ u1
dt
The material balance equations, moment equations, and
equations for degree of polymerization required to sim-
polymeric species. These moments are defined in Table 4 ulate the polymerization process are in Table 3 (Eqs.
and the corresponding moment differential equations are (a)–(d)) Table 5 (Eqs. (a)–(i)) and Eqs. (1)–(5). The dy-
given in Table 5 [51]. namic model equations were solved using DDASAC
From the moment definitions, the total of these zer- [52].
oth moments (ut0 = u0 + uc0 + um0 + ud0) is the total
concentration of polymer chains and other initiator- 2.3. Determination of initial estimates of rate constants
derived species present, which equals the initial concen-
tration of initiator. The total of the first moments 2.3.1. Experimental kinetic measurements
(ut1 = u1 + uc1 + um1 + ud1) is the total number of moles Based on Scheme 1, selected rate constants can be de-
of monomer units in the polymer chains and initiating rived experimentally from initiator and monomer con-
species, which equals the moles of monomer units ini- sumption data, to be used as initial estimates for the
tially in the initiator, plus the monomer consumed: parameter estimation. Assuming that experimental
conditions have been selected that the polymerization
ut0 ¼ ½I 0 ð1Þ
proceeds either predominantly by Path A or predomi-
nantly by Path B, simplified expressions were derived
ut1 ¼ 2½I 0 þ ½M 0  ½M ð2Þ
from Scheme 1 to describe the time-varying behavior
Notice that balances on the second moments of chain of the initiator [I] and monomer [M] concentrations:
length distribution are not included in Table 3 as it is the detailed derivations are given in Appendix A (Path
unnecessary to calculate second moments for the various A) and Appendix B (Path B).
J.E. Puskas et al. / European Polymer Journal 41 (2005) 1–14 7

Path A (TMPCl); Kim and Faust [55]: k1 = 5.0 · 107 s1
½I 0 (TMPCl) and k1 = 3.4 · 107 s1 (PIB-36mer). Using
ln ¼ K 0 k 1 ½LA 0 t ð6Þ kp = 6 · 108 L/mol s we get K0K1 = 5.7 · 109 L/mol and
½I
K0K2 = 8.7 · 108 L2/mol2, which confirms that both
½M 0 equilibria are dramatically shifted toward the dormant
ln ¼ k p K 0 K 1 ½I 0 ½LA 0 t ð7Þ polymer chains Pn-Cl. From K0k1 = 0.22 L/mol s and
½M
K0K1 = 5.7 · 109 L/mol we get k1 = 3.9 · 107 s1. This
Employing Eqs. (6) and (7), K0k1 = 0.22 L/mol s and agrees very well with published capping rate coefficients,
K0K1kp = 3.4 L2/mol2 s were calculated from previously thus we assumed this value for k2 and determined
published experimental initiator and monomer con- K0k2 = (k0/k0) Æ k2 = 3.38 L2/mol2 s.
sumption data in Path A [15]. As a result, the following correlations have been
established:
Path B C1 : k 1 K 0 ¼ k 1  ðk 0 =k 0 Þ ¼ 0:22 L=mol s ð10Þ
½M 0
ln ¼ K 0 K 2 k p ½I 0 ½LA 20 t ð8Þ
½M C2 : k 2 K 0 ¼ k 2  ðk 0 =k 0 Þ ¼ 3:38 L2 =mol2 s ð11Þ
In Path B, that is, under conditions normally used in liv- The correlation between the parameters will influence
ing IB polymerizations, initiation is nearly instantaneous the estimability of individual parameters.
and initiator consumption cannot be followed experi-
mentally. After a very fast initiation period, [Pn] = [I]0. 2.4. PREDICI parameter estimation
In classical living polymerization this simplification is
routinely used in formulating rate equations for mono- For the PREDICI parameter estimation, the reaction
mer consumption. Also, in case of instantaneous ini- scheme in Table 2 was used. Table 6 lists initial rate con-
tiation the monomer consumed during initiation is stant values and sets of estimated parameters. The initial
neglected. In contrast, in Path A, initiation and propa- values for k0, k0, k1 and k2 were established from the
gation was found to proceed simultaneously, so these correlations given in (10) and (11) by trial and error, fit-
simplifications cannot be considered. In Path B, experi- ting experimental data [54]. The parameter estimation
mental data have revealed a fractional order of 1.76 in routine in PREDICI is based on the damped Gauss–
TiCl4 [15], which is in good agreement with the range Newton method [56,57] and has the ability to converge
of 1.7–2.2 reported by other researchers [11,16,19,24]. well even with bad starting values. Seven experimental
½M 0 data sets were used in the simulation [15]. First, all
ln ¼ k 00p ½I 0 ½LA 1:76 ð9Þ parameters were estimated simultaneously. The simula-
½M
tion did converge with kp = 6 * 108 L/mol s as a starting
Here k 00p is the experimentally measured apparent rate estimate without any constraints imposed, although with
constant. This indicates that in reality, both Paths A high error within the 90% confidence intervals; the data
and B proceed simultaneously, with Path B dominating. are listed as Set 1 in Table 6. It is interesting to note that
Using Eq. (9), constant k 00p ¼ 52 L3/mol3 s was obtained kp converged to a value of 4.64 * 108 ± 2.75 * 108 L/mol s.
from experimental monomer consumption data [15]. In contrast, the simulation was unable to converge with
With the interpretation in Eq. (8), K0K2kp = 52 L3/ kp = 104 L/mol s as a starting estimate. This further sup-
mol3 s. Eqs. (6)–(8) were derived with several assump- ports the high kp values obtained in diffusion clock and
tions and should be viewed as simplified rate expres- competition experiments [30,35,36]. The values for k1
sions, but they provide means to establish initial and k2 converged to 1.77 and 2.75 * 107 s1, close to
estimates for the simulations. the initial values. However, k0, k0, k1 and k2 carried very
Selected rate constants in IB polymerizations have large errors and did not satisfy the correlations estab-
been measured or derived independently. For instance, lished in (10) and (11). Next we fixed kp = 6 * 108 L/mol s
‘‘diffusion clock’’ and competition experiments methods and k1 = k2 = 3.9 * 107 s1, and estimated k0, k1 and
yielded kp  108 L/mol s [30,35,36]. As discussed earlier, k2. Set 2 shows that the error is still high. In the next step
this high value is contradicting earlier data, presented we enforced the correlations defined in Eqs. (10) and (11)
in Table 1, with the exception of kp obtained from irra- using interpreter functions in Predici [57]. Set 3 shows
diation-initiated IB polymerization. Industrial experi- that k0 and k0 still carry very high error. As was deter-
ence supports the close to diffusion-limited high value mined from the estimability analysis, these parameters
[53], therefore we accepted this value. Later it will be were correlated and hence could not be estimated simul-
shown that parameter estimation also supports a high taneously. Therefore in Set 4, we fixed the value of k0 to
kp value. Published capping rate constants, although de- 3.9 * 107 s1. Set 4 shows the estimated k0 with low error
rived from differing interpretations, were in the range of within the 90% confidence interval. The parameters in
3–5 · 107 s1 (Puskas and Peng [26]: k1 = 3.9 · 107 s1 Set 4 were used for the simulations with Predici. It should
8 J.E. Puskas et al. / European Polymer Journal 41 (2005) 1–14

Table 6
Initial parameter values and parameter estimates used in simulations
Parameter Initial values Set 1 ± CIa (estimating all) Set 2 ± CI (estimating k0, k1, k2)
1(a): Estimates without constraints
k0 (L/mol s) 1.56 * 108 2.75 * 108 ± 2.79 * 108 7.96 * 107 ± 3.97 * 107
1
k0 (s ) 3.90 * 107 1.13 * 108 ± 7.62 * 108 3.90 * 107
k1 (s1) 5.50 * 102 1.20 * 102 ± 1.23 * 102 1.72 * 102 ± 1.23 * 102
k1 (s1) 3.90 * 107 1.71 * 107 ± 1.77 * 107 3.90 * 107
k2 (L/mol s) 8.45 * 101 5.88 ± 3.93 3.50 ± 1.47
k2 (s1) 3.90 * 107 5.90 * 107 ± 2.75 * 107 3.90 * 107
kp (L/mol s) 6.00 * 108 4.64 * 108 ± 2.75 * 108 6.00 * 108

Set 3 ± CI (estimating k0, k0) Set 4 ± CI (estimating k0)


1(b): Estimates with constraints
k0 (L/mol s) 1.72 * 108 ± 3.84 * 108 1.51 * 108 ± 2.03 * 107
1
k0 (s ) 4.48 * 107 ± 1.16 * 108 3.90 * 107
k1b (s1) 5.73 * 102 5.68 * 102
k1 (s1) 3.90 * 107 3.90 * 107
k2b (L/mol s) 8.80 * 101 8.73 * 101
k2 (s1) 3.90 * 107 3.90 * 107
kp (L/mol s) 6.00 * 108 6.00 * 108
a
CI = 90% confidence interval.
b
The values of k1 and k2 are dictated by correlations defined in Eqs. (10) and (11).

be noted here that the parameters estimated for k0, k0,


k1 and k2 in all four sets gave reasonable agreement with 0.05 Predici
ODE
experimental data. However, the model is very sensitive
to kp/k1. Since kp and k1 are highly correlated we 0.04
had good agreement between simulation and experimen-
[TMPCl], mol/L

tal data with either low or high kp data as long as kp/k1 0.03
was kept constant at 16.5 L/mol. We do not intend to
argue that agreement between experimental and simu-
0.02
lated data alone can prove any proposed mechanisms.
Kinetic simulation can only disprove a proposed mecha-
nism if it shows serious discrepancy between experimen- 0.01
tal and simulated plots. In our particular case, if we
accept independently measured capping constants in 0.00
the range of 107 s1, good agreement can only be found 0 200 400 600 800
for high propagation rate constant (kp  108 L/mol s). time (s)

Fig. 1. Comparison of experimental and simulated initiator


2.5. Parameter estimability analysis using the ODE model
consumption plots for Path A: [IB]0 = 2.0 mol/L, [TMPCl]0 =
0.05 mol/L, [DtBP]0 = 0.007 mol/L. Hx/MeCl=60/40 v/v; Exper-
Attempts were also made to estimate the kinetic imental: [TiCl4] = 0.05 mol/L (m), [TiCl4] = 0.025 mol/L (s);
parameters using the same batch reactor data [15] the Temperature: 80 C; Simulation parameters: Set 4 (Table 1).
ODE version of the model in Tables 3–5 and the GREG
parameter estimation software [52]. Before using GREG
to estimate the parameters, it was confirmed that the Pre- and two Path A runs with [TMPCl]0 = 0.05 mol/L and
dici model and the ODE model yielded practically iden- [TiCl4]0 = 0.0125 mol/L), and were supplied to the
tical simulation results when using the same parameters; GREG parameter estimation software [58].
representative examples for a Path A simulations are Attempts to estimate all of the model parameters
shown in Figs. 1–3. In all other figures, only the PRE- simultaneously failed, even after the model was re-
DICI results are shown. Pooled variances for measured parameterized in an attempt to improve the numerical
molecular weight and initiator and monomer concentra- conditioning of the parameter estimation problem (i.e.,
tions were determined from replicate experiments at two expressing the model in terms of natural logarithms to
different experimental settings (four replicate Path B runs ensure that all parameters estimated would have the
with [TMPCl]0 =0.004 mol/L and [TiCl4]0 = 0.032 mol/L, same order-of-magnitude size). Simulations revealed
J.E. Puskas et al. / European Polymer Journal 41 (2005) 1–14 9

To further investigate the correlated effects of the


2.0
Predici parameters on the model predictions and to determine
ODE
which subsets of the model parameters can be esti-
mated together, a detailed parameter estimability anal-
1.5 ysis [59] was performed, after including the correlations
in Eqs. (10) and (11) in the model. It was found that,
[M], mol/L

among the four parameters, k0, k0, k1 and k2, it is


1.0 impossible to estimate any two parameters simultane-
ously. k0 and k0 are correlated through the kinetic
mechanism (K0 = k0/k0), and k1 and k2 are related
0.5
by the correlations defined in (10) and (11). Hence,
only one of these parameters can be included in the
set of parameters to estimate using the current set of
0.0
0 200 400 600 800 experimental data. If additional measurements were
time (s) available, e.g., measurements of the concentration of
the I * LA or Pn * LA intermediates, then more kinetic
Fig. 2. Comparison of experimental and simulated monomer parameters could be estimated. Without any additional
consumption plots for Path A: [TMPCl]0 = 0.05 mol/L, [IB]0 = measurements to compare with model predictions,
2 mol/L, [DtBP]0 = 0.007 mol/L. Hx/MeCl=60/40 v/v; Experi- however, multiple sets of values for these four param-
mental: [TiCl4] = 0.05 mol/L (m), [TiCl4] = 0.025 mol/L (s);
eters will give the same model predictions as the
Temperature: 80 C; Simulation parameters: Set 4 (Table 1).
parameter estimates in Set 4, which were used to sim-
ulate the experimental results.

100 2.6. Simulation of living IB polymerization using the


proposed model of Scheme 1
Predici
80 ODE
Figs. 1 and 2 show experimental [15] and simulated
initiator and monomer consumption plots for Path A.
60 The agreement is very good. As discussed above, initia-
DPn, DPw

tor and monomer consumption is simultaneous under


the experimental conditions used. As a consequence,
40
the degree of polymerization increases only slightly with
conversion, as shown in Fig. 3. The Mn prediction is
20 quite good, while Mw is slightly overestimated. This
could be attributed to the very low molecular
weights––loss of fractions during precipitation is inevita-
0 ble. The MWD is slightly lower than 2 during the entire
0.0 0.2 0.4 0.6 0.8
reaction.
Conversion
In Path B, the simulated monomer consumption
Fig. 3. Experimental and simulated molecular weight vs
shows very good agreement with experimental data as
conversion plots for Path A: Experimental: Mn (s), Mw (m), can be seen from Fig. 4. The predicted Mn also shows
[TMPCl]0 = 0.05 mol/L, [TiCl4]0 = 0.025 mol/L, [IB]0 = 2 mol/L, a good fit to experimental data (Figs. 5 and 6). However,
[DtBP]0 = 0.007 mol/L. Hx/MeCl=60/40 v/v; Temperature: the Mw value becomes progressively underpredicted
80 C; Simulation parameters: Set 4 (Table 1). with increasing conversion. This points to the existence
of possible side reactions. Therefore linear semilogarith-
mic monomer consumption plots together with linear
that nearly identical simulation results could be obtained Mn––conversion plots, which are currently considered
if certain parameters were adjusted in pairs. For exam- as proof for living conditions, are insufficient to demon-
ple, simulation results indicate that if the forward rate strate living conditions. The additional criterion of
constant k0 and the reverse rate constant k0 are both MwMn = constant [5] should also be considered when
doubled, there are no significant changes in the model evaluating living polymerizations. This is evident from
predictions. The equilibrium position of the reaction is the simulated plots obtained with an ideal model, but
maintained, and since both reactions are already very the experimental data deviates from this prediction.
fast, increasing their rates, together, has no discernable Introduction of various possible side reaction scenar-
effect on the molecular weights or reactant concentra- ios such as slow initiation, termination, transfer and
tions. Similar conclusion applies to kp/k1. slow protic initiation into our model was unsuccessful.
10 J.E. Puskas et al. / European Polymer Journal 41 (2005) 1–14

2.0 600

Predici Predici
500
1.5

400
[M], mol/L

DPn, DPw
1.0 300

200
0.5

100

0.0 0
0 2000 4000 6000 8000 0.0 0.2 0.4 0.6 0.8
time (s) Conversion

Fig. 4. Comparison of experimental and simulated monomer Fig. 6. Experimental and simulated molecular weight vs
consumption plots for Path B: [TMPCl]0 = 0.004 mol/L, [IB]0 = conversion plots for Path B: Experimental: Mn (s), Mw (m),
2 mol/L, [DtBP]0 = 0.007 mol/L. Hx/MeCl=60/40 v/v; Experi- [TMPCl]0 = 0.004 mol/L, [TiCl4]0 = 0.064 mol/L, [IB]0 = 2 mol/
mental: [TiCl4] = 0.064 mol/L (m), [TiCl4] = 0.04 mol/L (s); L, [DtBP]0 = 0.007 mol/L. Hx/MeCl=60/40 v/v; Temperature:
Temperature: 80 C; Simulation parameters: Set 4 (Table 1). 80 C; Simulation parameters: Set 4 (Table 1).

Mn and Mw values. More detailed studies are necessary


600
to find the cause of the Mw divergence.
Predici
500 2.7. Calculating the average run length l0 from Path A

400 Puskas et al. [9] introduced a method to calculate the


average run length l0 , defined as the average number of
DPn, DPw

300
monomer units incorporating in one active ionization
period, from MWD data using Eq. (12)
200 DPw l0
¼1þ ð12Þ
DPn ðDPn Þ
100
where l0 ¼ ðk p =k 1 Þ½M 0 . Eq. (13) shows the form for
intermediate conversion.
0
0.0 0.2 0.4 0.6 0.8  
Conversion l ¼ 1 þ l0  1 1 þ ½M ð13Þ
2 ½M 0
Fig. 5. Experimental and simulated molecular weight vs
conversion plots for Path B: Experimental: Mn (s), Mw (m), Subsequently Müller et al. [60] introduced another
[TMPCl]0 = 0.004 mol/L, [TiCl4]0 = 0.04 mol/L, [IB]0 = 2 mol/L, method:
[DtBP]0 = 0.007 mol/L, Hx/MeCl=60/40 v/v; Temperature:  
80 C; Simulation parameters: Set 4 (Table 1).
DPw 1 2
¼1þ 1 ð14Þ
DPn b X
where X = conversion and b = k1/kp[I]0.
When TMP-Cl and PIB 36-mer were used as initiators, it Storey recently showed that average run length num-
was found that kp  3 * ki [23], but simulations con- bers could also be obtained from ‘‘rapid monomer con-
ducted using a ki value of 1–2 * 108 L/mol s did not show sumption’’ (RCM) data [28]. We now will show that l0
any noticeable difference when compared to simulations can also be obtained from Path A in IB polymerization.
with kp = ki. When slow irreversible termination was At the end of the reaction in Path A, when the initiator
introduced, a better fit was obtained for Mw in certain is consumed, polydispersity of 1.84 was obtained exper-
experiments [54], but in others led to discrepancies in imentally and simulation yielded a polydispersity of 1.88
the monomer consumption and Mn data. Introducing (Fig. 3). The conditions in Path A experiments, when
chain transfer caused a simultaneous reduction in both initiator and monomer consumption is simultaneous,
J.E. Puskas et al. / European Polymer Journal 41 (2005) 1–14 11

can be considered as the first ‘‘run’’ of a living polymer- that can polymerize, Eqs. (A.1) and (A.2) may be added
ization; initiation is complete only at the end of the reac- together to get:
tion. Using DPw/DPn = 1.88 and the DPn value of 40 at
full initiator conversion, l0 ¼ 35 was obtained from Eq. d½I d½I  LA dð½I þ ½I  LA Þ
þ ¼
(13). Using l0 ¼ ðk p =k 1 Þ½M 0 [8], kp/k1 = 17.5 L/mol is dt dt dt
calculated. This is in excellent agreement with that pub- ¼ k 1 ½I  LA þ k 1 ½Iþ LA ðA:3Þ
lished by Kaszas and Puskas [17] (kp/k1 = 16.5 L/mol)
[I] + [I * LA] = [I]u, where [I]u is the concentration of
and Schlaad et al. [23] (kp/k1 = 16.4 L/mol). Thus l0
unreacted initiator, thus Eq. (A.4) defines the rate of
can be obtained experimentally under Path A condi-
initiator consumption and polymer formation:
tions, where initiation will compete with propagation.
Under these conditions Mn and Mw will remain approx- d½I u d½Pn
 ¼ ðA:4Þ
imately constant, with the polydispersity for the most dt dt
probable distribution of MWD = 2. The MWD will start If the reactions involving k0 and k0 are very fast,
to narrow only after initiation is complete. then we will have an initial rapid equilibration, after
which:
3. Conclusions k0
½I  LA ¼ ½I ½LA ðA:5Þ
k 0
In conclusion, we can state that the reaction mecha-
nism proposed in Scheme 1 reconciles seemingly contra- so that Eq. (A.3) becomes:
dictory experimental findings (e.g., shifting TiCl4 order d½I u
and discrepancy in rate constant values obtained with ¼ k 1 K 0 ½I ½LA þ k 1 ½Iþ LA ðA:6Þ
dt
diffusion clock methods and kinetic measurements)
Now, if [I+LA] is small due to the fast propagation
and can be used to simulate experimental results with
reaction of the ionized species, and [LA]  [LA]0, we
rate constants that satisfy the correlations imposed by
have:
the mechanism and/or independently measured rate con-
stants. This by no means can be considered as proof of d½I u
¼ k 1 K 0 ½I ½LA 0 ðA:7Þ
the proposed mechanism. The existence and chemical dt
nature of the proposed intermediates must be proven Expanding (A.7) with 1/[I]u on both sides and substitut-
independently. However, in the absence of another ing (A.5) gives:
model that satisfies all the contradictory requirements
this model does not seem to be unreasonable. In addi- 1 d½I u k 1 K 0 ½I ½LA 0 k 1 K 0 ½LA 0
 ¼ ¼ ðA:8Þ
tion, large-scale production data (confidential) demon- ð½I u Þ dt ð½I þ ½I  LA Þ ð1 þ K 0 ½LA 0 Þ
strates that the rate constant of propagation must be Assuming K0[I]0  1 (A.8) can be simplified to give
very high. (A.9):
1 d½I u
Acknowledgment  ¼ k 1 K 0 ½I ½LA 0 ðA:9Þ
ð½I u Þ dt

Financial support of Bayer Inc., Canada and Integrating (A.9) between the limits of [I]u,0 = [I]0 and
NSERC Canada is gratefully acknowledged. [I]u,t = [I]t + [I * LA]t
Z ½I u;t Z t
d½I u
¼ k 1 K 0 ½LA 0 dt ðA:10Þ
½I u;0 ½I u t¼0
Appendix A. Path A
we get the simplified rate Eq. (A.10)
Initiator consumption rate. From material balances
on the initiator and the intermediates, we have: ½I 0
ln ¼ k 1 K 0 ½LA 0 t ðA:11Þ
½I u;t
d½I
¼ k 0 ½I ½LA þ k 0 ½I  LA ðA:1Þ This is equivalent to Eq. (6) in the text.
dt
Propagation rate. For Path A, the rate of propaga-
d½I  LA tion is described by Eq. (A.12):
¼ k 0 ½I ½LA  k 0 ½I  LA  k 1 ½I  LA
dt
d½M
þ k 1 ½Iþ LA ðA:2Þ Rp ¼  ¼ kp½M ð½Iþ LA þ ½Pþ 
n LA Þ ðA:12Þ
dt
To obtain the net rate of consumption of both of these The consumption rate of the active propagating species
initiator-derived species together to form ionized species is given in Eq. (A.13):
12 J.E. Puskas et al. / European Polymer Journal 41 (2005) 1–14

dð½Iþ LA þ ½Pþ 


n LA Þ
½I þ ½Pn  ½I 0 ðA:25Þ
¼ k 1 ½I  LA  k 1 ½Iþ LA
dt
þ k 1 ½Pn  LA  k 1 ½Pþ  ½LA 0
n LA ½LA  ðA:26Þ
1 þ K 0 ½I 0
ðA:13Þ

The combined change in initiator and polymer concen- With K0[I]0  1, [LA]  [LA]0.
trations with time can be written as follows: Substituting (A.25) and (A.26)into Eq. (A.20) gives
the general form of the differential polymerization rate
dð½I þ ½Pn Þ equation:
¼ k 0 ½I ½LA þ k 0 ½I  LA
dt
 k 0 ½Pn ½LA þ k 0 ½Pn  LA ðA:14Þ d½M ½I 0 ½LA 0
Rp ¼  ¼ k p K 1 K 0 ½M ðA:27Þ
dt 1 þ K 0 ½LA 0
Since the reactivities of the initiator and the growing
Assuming K0[LA]0  1 again, we can simplify and inte-
chains are about equal, the sums of the concentrations
grate the rate equation to get (A.28):
of the active cationic species and also the dormant spe-
cies are constant and the quasi-steady state assumption ½M 0
ln ¼ k p K 0 K 1 ½I 0 ½LA 0 t ðA:28Þ
(QSSA) can be applied: ½M
dð½Iþ LA þ ½Pþ 
n LA Þ (A.28) is equivalent to Eq. (7) in the text.
ffi0 ðA:15Þ
dt

dð½I þ ½Pn Þ Appendix B. Path B


ffi0 ðA:16Þ
dt
With this and assuming fast equilibria, the concentration Propagation rate. The rate of polymerization for Path
of active cationic species is derived as follows: B can be derived similarly to Path A:

½Iþ LA þ ½Pþ 


n LA ¼ K 1 ð½I  LA þ ½Pn  LA Þ ðA:17Þ d½M
Rp ¼  ¼ k p ½M ð½Iþ LA þ 
2 þ ½Pn LA2 Þ ðB:1Þ
dt
½I  LA þ ½Pn  LA ¼ K 0 ð½I þ ½Pn Þ½LA ðA:18Þ
dð½Iþ LA þ 
2 þ ½Pn LA2 Þ
Substituting (A.18) into (A.17), we have ¼ k 2 ½I  LA ½LA
dt
½Iþ LA þ ½Pþ 
n LA ¼ K 1 K 0 ð½I þ ½Pn Þ½LA ðA:19Þ  k 2 ½Iþ LA2

where K1 = k1/k1 and K0 = k0/k0. þ k 2 ½Pn  LA ½LA



Substituting (A.18) into (A.12) yields  k 2 ½Pþn LA2 ðB:2Þ
d½M
Rp ¼  ¼ k p K 1 K 0 ½M ð½I þ ½Pn Þ½LA ðA:20Þ dð½I þ ½Pn Þ
dt ¼ k 0 ½I ½LA þ k 0 ½I  LA
dt
Since the concentration of active cationic species  k 0 ½Pn ½LA þ k 0 ½Pn  LA ðB:3Þ
([I+LA] and ½Pþ 
n LA ) is very small, simplified mass
balances for the initiator and Lewis acid can be written dð½Iþ LA þ 
2 þ ½Pn LA2 Þ
as follows: ffi0 ðB:4Þ
dt
½I 0 ¼ ½I þ ½Pn þ ½I  LA þ ½Pn  LA ðA:21Þ
dð½I þ ½Pn Þ
ffi0 ðB:5Þ
½LA 0 ¼ ½LA þ ½I  LA þ ½Pn  LA ðA:22Þ dt
Substituting (A.17) into (A.20) and (A.21) yields: Since the reactivities of the initiator and the polymer are
½I 0 close to equivalent, the rate of consumption of the initi-
½I þ ½Pn ¼ ðA:23Þ ating species is equal to the rate of the formation of the
1 þ K 0 ½LA
growing species.
and
½Iþ LA þ 
2 þ ½Pn LA2 ¼ K 2 ð½I  LA þ ½Pn  LA Þ½LA
½LA 0 ðB:6Þ
½LA ¼ ðA:24Þ
½I 0
1 þ K0
1 þ K 0 ½LA ½I  LA þ ½Pn  LA ¼ K 0 ð½I þ ½Pn Þ½LA ðB7 ¼ A7Þ

If the concentration of free Lewis acid is small, i.e., ½Iþ LA þ  2


2 þ ½Pn LA2 ¼ K 2 K 0 ð½I þ ½Pn Þ½LA ðB:8Þ
K0[LA]  1, (A.23) and (A.24) can be simplified:
J.E. Puskas et al. / European Polymer Journal 41 (2005) 1–14 13

where K2 = k2/k2. [14] Storey RF, Maggio TL. Macromolecules 2000;33:681.


Substituting (B.8) into (B.1) yields: [15] Puskas JE, Lanzendörfer MG. Macromolecules 1998;31:
8684.
d½M
Rp ¼  ¼ k p K 1 K 0 ½M ð½I þ ½Pn Þ½LA 2 ðB:9Þ [16] Wu Y, Tan Y, Wu G. Macromolecules 2002;35:3801.
dt [17] Kaszas G, Puskas JE. Polym React Eng 1994;2(3):251.
The concentrations of the ionic species are very small, [18] Kennedy JP, Majoros I, Nagy A. Adv Polym Sci 1994;
thus the simplified mass balances established in Eqs. 112:1.
[19] Györ M, Wang HC, Faust R. J Macromol Sci Chem A
(A.20) and (A.21) and Eqs. (A.22)–(A.25) will apply.
1992:639.
With these we get the form given below: [20] Storey RF, Chisholm BJ, Brister LB. Macromolecules
1995;28(12)4055.
d½M ½I 0 ½LA 20
Rp ¼  ¼ k p K 1 K 0 ½M ðB:10Þ [21] Puskas JE et al. In: Puskas JE, Long TE, Storey RF,
dt ð1 þ K 0 ½I 0 Þ2 editors. In situ monitoring of monomer and polymer
synthesis. New York: Kluwer Academic/Plenum Publish-
Assuming K0[LA]0  1 again and integrating, we get ers; 2003. p. 37.
(B.11): [22] Paulo C, Puskas JE, Angepat S. Macromolecules 2000;33:
4634.
½M 0 [23] Schlaad H, Kwon Y, Faust R, Mayr H. Macromolecules
ln ¼ k p K 0 K 1 ½I 0 ½LA 20 t ðB:11Þ
½M 2000;33:743.
[24] Storey RF, Donnalley AB. Macromolecules 2000;33:53.
Eq. (B.11) is equivalent to Eq. (9) in the text. [25] Puskas JE, Lanzendörfer MG, Peng H, Michel AJ, Brister
LB, Paulo C. In: Puskas JE, editor. Ionic polymerizations
References and related processes. Netherlands: Kluwer Academic/
Plenum Publishers; 1999. p. 143.
[1] Szwarc M. Carbanions, living polymers and electron [26] Puskas JE, Peng H. Polym React Eng 1999;7(4):553.
transfer processes. New York: Wiley; 1968. [27] Hadjikyriacou S, Faust R. Macromolecules 1995;28:
[2] (a) Pepper DC. Makromol Chem 1974;175:1077; 7893.
(b) Johnson AF, Young RN. J Polym Sci Polym Symp [28] Storey RF, Thomas QA. Macromolecules 2003;36:5065.
1976;56:211. [29] Fodor Z, Bae YC, Faust R. Macromolecules 1998;31:4439.
[3] Kennedy JP, Kelen T, Tüdös FJ. J Macromol Sci Chem A [30] Roth M, Mayr H. Macromolecules 1996;29:6104.
1982;18(9):1189; [31] Moreau M. IP 97. In: Proceedings of the International
Faust R, Fehervari A, Kennedy JP. J Macromol Sci A Symposium on Ionic Polymerization and Related Proc-
1982;18(9):1209; esses, Paris, 1997. p. 135.
Puskas JE, Kaszas G, Kennedy JP, Kelen T, Tüdös FJ. J [32] Kistenmacher TJ, Stucky GD. Inorg Chem 1971;10:122.
Macromol Sci Chem A 1982;18(9):1245; [33] Rytter E, Kvisle S. Inorg Chem 1985;24:640.
Puskas JE, Kaszas G, Kennedy JP, Kelen T, Tüdös FJ. J [34] Schlaad H, Kwon Y, Sipos L, Faust R, Charleaux B.
Macromol Sci Chem A 1982;18(9):1229; Macromolecules 2000;33:8225.
Kennedy JP. J Polym Sci Polym Chem 1999;37:2285. [35] Sipos L, De P, Faust R. Macromolecules 2003;36:8282.
[4] Georges MK, Veregin RP, Kazmaier PM, Hamer GK. [36] Mayr H, Schimmel H, Kobayashi S, Kotani M, Praba-
Macromolecules 1993;26:2987; rakaran TR, Sipos L, et al. Macromolecules 2002;35:4611.
Wang JS, Matyjaszewski K. Macromolecules 1995;28: [37] Marek M, Chmelir M. J Polym Sci 1967;22:177.
7901; [38] Toman L, Marek M. Makromol Chem 1976;22:3325.
Kato M, Kamigaito M, Sawamoto M, Higashimura T. [39] Magagnini PL, Cesca S, Giusti P, Priola A, Di Maina M.
Macromolecules 1995;28:1721. Makromol Chem 1977;178:2235.
[5] Puskas JE, Kaszas G. Carbocationic polymerization. En- [40] Ueno K, Yamaoka H, Hayashi K, Okamura S. Int J Appl
cyclopedia of polymer science and technology, V5. Wiley- Radiat Isotopes 1966;17:595.
Interscience; 2003. p. 382. [41] Plesch PH. Prog Reaction Kinetics 1993;18:1.
[6] Puskas JE, Kaszas G. Prog Polym Sci 2000;25:403. [42] Taylor RB, Williams F. J Am Chem Soc C 1969;22:177.
[7] Matyjaszewski K, Sawamoto M. In: Matyjaszewski K, [43] Plesch PH. Macromolecules 2001;34:1143.
editor. Cationic polymerizations: mechanisms, synthesis and [44] Puskas JE, Shaikh S. Macromol Symp [in press].
applications. New York: Marcel Dekker; 1996. p. 265. [45] Majoros I, Nagy A, Kennedy JP. Adv Polym Sci 1994;
[8] Kaszas G, Puskas JE, Kennedy JP, Chen CC. J Macromol 112:1.
Sci Chem A 1989;26(8):1099. [46] Lorimer JP, Pepper DC. Prod R Soc Lond A 1976;351:551.
[9] Puskas JE, Kaszas G, Litt M. Macromolecules 1991; [47] Vairon JP, Charleaux B, Moreau M. In: Puskas JE, editor.
24:5278. Ionic polymerizations and related processes. Nether-
[10] Sigwalt P. Makromol Chem Makromol Symp 1991;47:179. lands: Kluwer Academic/Plenum Publishers; 1999. p. 177.
[11] Storey RF, Choate Jr KR. Macromolecules 1997;30:4799. [48] Sigwalt P, Moreau M, Polton A. Makromol Chem
[12] Kaszas G, Puskas JE, Kennedy JP. J Makromol Chem Macromol Symp 2002;183:35.
Makromol Symp 1988;13/14:473. [49] Wulkow M. PREDICI Simulation Package for Poly-
[13] Roth M, Patz M, Preter H, Mayr H. Macromolecules reactions, CiT, GmbH, 1996.
1997;30:722. [50] Wulkow M. Chem Eng Technol 1996;5:393.
14 J.E. Puskas et al. / European Polymer Journal 41 (2005) 1–14

[51] Liu Q. MSc (Eng) thesis, Queens University, 2001. [57] Wulkow M. PREDICI Simulation Package for Poly-
[52] Stewart WE, Caracotsios M, Sorensen JP. DDASAC reactions: Users Manual Ver.5.18.5, CiT GmbH, 2001.
Software Package Documentation, University of Wiscon- [58] Stewart WE, Caracotsios M, Sorensen JP. GREG: Gen-
sin, 1994. eralized REGression Package for Non-linear Parameter
[53] Puskas JE, Schmidt J, Collart P, Verhelst M. Kautsch Estimation 1994, University of Wisconsin.
Gummi Kunst 1995;12:866. [59] Yao KZ, Shaw B, Kou B, McAuley KB, Bacon D. Polym
[54] Shaikh S, Peng H, Puskas JE. Polym Prepr 2002;35: React Eng 2003;11(3):563.
5320. [60] Müller AHE, Litvinenko G, Yan D. Macromolecules
[55] Kim MS, Faust R. Macromolecules 2002;36:8282. 1996;29:2339.
[56] Wulkow M. Macromol Theory Simul 1996;5:393.

Vous aimerez peut-être aussi