Structural and mechanistic aspects of the thermal degradation of
poly(vinyl chloride)
W.H. Starnes Jr.*
Departments of Chemistry and Applied Science, College of William and Mary, P.O. Box 8795,
Williamsburg, VA 23187-8795, USA
Received 15 July 2002; accepted 17 July 2002
Abstract
A critical review of the title subject supports the following major conclusions. Thermal dehydrochlorination of
poly(vinyl chloride) (PVC) begins with internal allylic chloride and tertiary chloride structural defects formed
during polymerization. The tertiary chloride is associated with 2,4-dichloro-n-butyl, 1,3-di(2-chloroethyl), and
chlorinated long branches. Mechanisms for the formation of all of the labile defects are well established.
‘Carbonylallyl’ structures and certain isotactic conformers of ordinary monomer units are unimportant as initiators
of thermal dehydrochlorination. Both the initiation and the subsequent formation of conjugated polyene sequences
occur via carbenium chloride ion pairs or by a closely related concerted four-center quasi-ionic route. Six-center
concerted processes, pathways involving free radicals, and other mechanistic schemes suggested recently are not
involved in polyene elongation. However, during thermal degradation, ordinary monomer units are converted into
internal allylic chloride defects by a mechanism that may include the abstraction of hydrogen by triplet cation
diradicals derived from polyene intermediates. Cyclization reactions seem likely to contribute to the termination of
polyene growth. When PVC is thermolyzed in blends with other polymers, unusual kinetic phenomena are
detected that remain to be fully explained. q 2002 Elsevier Science Ltd. All rights reserved.
Keywords: Poly(vinyl chloride); Thermal degradation; Dehydrochlorination; Degradation initiation; Transfer to monomer;
Unstable structures; Polyenes; Transfer to polymer; Free-radical mechanisms; Ion-pair/quasi-ionic mechanism; Six-center
mechanisms; Polymer blends
Contents
1. Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .2134
2. Thermal degradation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .2135
2.1. Initiation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .2135
2.1.1.
Possible initiation sites. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .2135
2.1.2.
Allylic chloride formation via the two mechanisms for transfer to monomer. . . . . . . . . . . . . . . . . . . .2136
0079-6700/02/$ - see front matter q 2002 Elsevier Science Ltd. All rights reserved.
PII: S 0 0 7 9 - 6 7 0 0 ( 0 2 ) 0 0 0 6 3 - 1
Prog. Polym. Sci. 27 (2002) 2133–2170
www.elsevier.com/locate/ppolysci
* Tel.: þ1-757-221-2552; fax: þ 1-757-221-2715.
E-mail address: whstar@wm.edu (W.H. Starnes).
2.1.3.
Tertiary chloride formation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .2143
2.1.4.
Mechanism and relative importance of initiation by internal allylic and tertiary chloride . . . . . . . . . . .2145
2.1.5.
Initiation by ‘carbonylallyl’ structures? . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .2148
2.1.6.
Initiation by ordinary monomer units . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .2149
2.2. Polyene growth and termination . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .2151
2.2.1.
Overview . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .2151
2.2.2.
Free-radical mechanisms . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .2151
2.2.3.
Ion-pair/quasi-ionic mechanism . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .2155
2.2.4.
Six-center concerted mechanisms . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .2159
2.2.5.
Other proposed mechanisms . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .2164
2.3. Degradation of polymer blends . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .2164
3. Concluding remarks . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .2166
Acknowledgement . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .2166
References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .2167
1. Introduction
Poly(vinyl chloride) (PVC) is extraordinarily useful as a commercial material. Among the
thermoplastics, it ranks second only to polyolefins in total worldwide production volume. Remarkably,
it has achieved this status despite its molecular instability toward heat, an instability that is much more
pronounced than those of all of its major competitors. In a technological sense, this difficulty has been
overcome to a large degree, for otherwise the usage of PVC would never have reached its current level.
Yet further improvements are much to be desired, not only with regard to the property of stability per se,
but also with respect to how this property is achieved, in terms of the requisite monetary costs, health and
environmental considerations, and the effects of thermal stabilization on other useful properties of the
resin.
Because of their technological relevance and intrinsic scientific significance, the thermal degradation
and stabilization of PVC have been the subjects of much research. Unfortunately, they also have
attracted an uncommon amount of mechanistic speculation that frequently has not been informed by a
comprehension of well-known chemical facts. As a result, much of the literature in this field is not only
highly confusing, but also grossly misleading, because it implies that much less is known about the
subject matter than actually is the case.
In its earliest stages, the thermal degradation of ordinary (head-to-tail) PVC involves the sequential
loss of hydrogen chloride molecules accompanied by the generation of conjugated polyene sequences
(Eq. (1))
. During the past fifty years
– ðCH
2
CHClÞ
n
– !
K
– ðCHyCHÞ
n
– þ nHCl
ð1Þ
exceedingly numerous attempts have been made to understand the mechanism of this superficially
simple process. In that regard, some of the major questions addressed have been as follows. At what
labile structures in PVC does HCl loss begin? How are these structures formed during the synthesis of
the polymer? What are the chemical mechanisms for the initiation and propagation of polyene growth?
Why does the growth of the polyenes terminate before HCl loss is complete? How do atmospheric
oxygen and HCl influence the structural consequences and the rate of the degradation?
The subsequent reactions of the polyene sequences have been of great interest, as well. However,
W.H. Starnes Jr. / Prog. Polym. Sci. 27 (2002) 2133–2170
2134
from a technological standpoint, the most important aspects of thermal degradation have been its
prevention, in a chemical sense, and the circumvention of the deleterious consequences on polymer
properties to which it leads (e.g. undesirable coloration caused by the presence of long polyene
sequences
). The goals of prevention and circumvention have usually been approached by the use of
additives, whose mechanisms of action still are being discussed, though at times in ways that are highly
conjectural.
The present review deals with various aspects of the title subject that have been of special interest to
the author and his associates during the past several years. Thermal stabilization is also mentioned
occasionally but is not covered in depth. This review is, in fact, an updated version of parts of a similar
survey that appeared in 1995
. It repeats and reconsiders much of the material presented there
,
frequently with the aid of new information. However, no attempt has been made here to achieve
exhaustive coverage of the voluminous literature in this field, especially those parts of it relating mostly
to technological matters, to degradation at very high temperatures, and to degradation under the
influence of molecular oxygen and various types of irradiation.
Since 1989, various aspects of the thermal degradation and stabilization of PVC have been discussed
in several reviews
.
closely resembles
, while
are essentially
identical.
is primarily a translation of much of
into Spanish.
1
2. Thermal degradation
2.1. Initiation
2.1.1. Possible initiation sites
Numerous investigations with model compounds for PVC have shown that the reaction depicted in
Eq. (1) probably is much more rapid than it would be if polyene growth began exclusively from ordinary
monomer units
. Thus, most of the thermal degradation has been considered to begin, at least at
the outset of the process, from thermally labile structural segments of the polymer chains
. The
nature, number, relative stability, and formation mechanisms of these ‘structural defects’ have,
therefore, been of great interest to researchers and technologists.
Various labile defects have been suggested to occur in PVC. In recent years, the leading candidates
have been structures containing allylic or tertiary halogen
, – CO(CHyCH)
n
CHCl – ðn $
1Þ moieties resulting from incidental air oxidation
, and GTTG
2
isotactic triads derived from
ordinary monomer units
. Adventitious a,b-unsaturated ketone segments have been suggested
to function as true catalysts for the thermal dehydrochlorination of ordinary monomer units
.
Even though a universal consensus still remains to be reached, most researchers in this field now
consider internal allylic and tertiary chloride segments to be of prime importance
. In
subsequent sections of this review, the mechanisms for the formation of those two types of defect are
examined in detail, and comments are made about some of the other theories of degradation initiation.
1
A list of minor corrections of
is available upon request from its senior author, who was not provided by the publisher
with an opportunity to proofread the paper before it was published.
W.H. Starnes Jr. / Prog. Polym. Sci. 27 (2002) 2133–2170
2135
2.1.2. Allylic chloride formation via the two mechanisms for transfer to monomer
Until recently, the ethyl-branch structure 1 was believed to be an important tertiary-chloride-
containing moiety in commercial PVC.
Evidence for its occurrence was obtained from the
13
C NMR spectra of a sample of PVC that had been
prepared by polymerization at 100 8C (a temperature above the usual commercial range) and then
dechlorinated with tri-n-butyltin hydride or tri-n-butyltin deuteride prior to analysis
. Samples of
PVC that were made at two commercial temperatures (40 and 82 8C) have now been characterized
structurally in the same way and shown to contain, instead of 1, the EB ethyl-branch segment that
appears in
. At first glance, this structural variation might seem to have no great significance.
However, more careful scrutiny shows convincingly that its mechanistic implications actually are quite
profound.
Hjertberg and So¨rvik reported microstructural data for a number of PVC specimens that were made
isothermally at four temperatures within the commercial range (45, 55, 65, and 80 8C) and at several
constant pressures of vinyl chloride (VC) below the pressure at saturation
. Branch concentrations of
these ‘subsaturation’ resins were determined by
13
C NMR inspection following reductive dechlorination
with Bu
3
SnH
. We noted later
that at every temperature used
, the sum of the chloromethyl
(MB;
) and EB branch concentrations of these polymers ð
r
MB
þ
r
EB
Þ
; per monomer unit, had a
Fig. 1. Chemical consequences of head-to-head emplacement of monomer during the polymerization of vinyl chloride (VC),
where P
z
is the head-to-tail macroradical, and the ks are rate constants. (Reprinted from
with permission from Wiley –
VCH.)
W.H. Starnes Jr. / Prog. Polym. Sci. 27 (2002) 2133–2170
2136
characteristic value that was different for each temperature and independent of monomer pressure. We
also pointed out
that, at each temperature,
r
EB
increased at the expense of
r
MB
when the monomer
pressure was lowered. These observations then were used to demonstrate
that the chemistry
occurring after head-to-head VC addition is described completely by the scheme in
According to
, the head-to-head radical, 2, does not add to monomer. Instead, it rearranges
quantitatively into 3 by a 1,2 shift of the b chloro substituent. A second 1,2-Cl shift converts 3 into the
doubly rearranged radical 4. Addition of monomer to 3 and 4 gives the MB and EB segments,
respectively. Importantly, radicals 3 and 4 also undergo chain-transfer reactions with VC that create
chloroallylic end groups (A1, A2) and 1,2-dichloroethyl radicals whose addition to the monomer
perpetuates the kinetic chain. With both 3 and 4, the transfer step is kinetically first-order in VC
concentration. It does not involve the unimolecular b cleavage of these radicals into chloroallylic chain
ends and kinetically free chlorine atoms that then form ClCH
2
C
z
HCl radicals by adding to monomer in a
subsequent step. The structural findings
also show
that (a) the interconversion of 3 and 4 is not
a quasi-equilibrium process (i.e. the condition k
22
@ ðk
3
þ k
5
Þ[VC] does not apply), and (b) the relative
amounts of VC addition and b scission that 3 and 4 experience are identical within the probable limits of
the experimental error (i.e. k
1
=k
4
¼ k
3
=k
5
).
Moreover, in a direct refutation of recurring literature statements, kinetic analysis reveals, as well,
that C
M,HH
, the constant for transfer to monomer following head-to-head emplacement, is, in fact, a true
constant whose value thus is independent of the concentration of vinyl chloride
. This value can be
equated to the number of chloroallylic chain ends (A1, A2) per monomer unit, and it is defined
mathematically by Eq. (2)
C
M
;HH
¼ k
0
p
k
4
=ðk
p
þ k
0
p
Þðk
1
þ k
4
Þ ¼ k
0
p
k
5
=ðk
p
þ k
0
p
Þðk
3
þ k
5
Þ
ð2Þ
In VC polymerization, the number-average molecular weight (M
n
) of the polymer being formed is
controlled almost exclusively by chain transfer to the monomer
. Yet the instantaneous M
n
decreases
drastically with decreasing monomer concentration
, notwithstanding the constancy, now
established
, of C
M,HH
. These apparently contradictory observations can be reconciled convincingly,
however, by invoking the concurrent operation of a second mechanism for transfer to monomer whose
possible incursion was considered several years ago
. This ‘auxiliary’ transfer process
is delineated in
. It starts with the inter- or intramolecular delivery of a methylene
hydrogen from a completed monomer unit to the macroradical, P
z
. The resultant polymeric R
z
radical
may add to monomer, as shown, in order to start the growth of a long-branch arrangement that has a
‘tertiary hydrogen’ on its branch-point carbon. Some NMR evidence for the presence of such a structure
has been published
. Of much greater mechanistic import, however, is the other depicted fate of R
z
,
which is a transfer reaction with VC that gives both an internal allylic chloride structure (IA) and a
radical whose addition to monomer starts the growth of a new polymer chain (note the close analogy
with the transfer steps in
). As is required by M
n
changes
, the relative importance of the
auxiliary mode of transfer would increase with decreasing concentration of VC, because the abstraction
of hydrogen by P
z
competes directly with the major reaction of this radical, which is its addition to
monomer.
Conclusive evidence has now been obtained for the operation of the auxiliary transfer process. Since
essentially all of the polymer molecules are formed by transfer to monomer, the simultaneous
occurrence of auxiliary transfer and the mechanism of
requires the presence of , 1 double bond in
W.H. Starnes Jr. / Prog. Polym. Sci. 27 (2002) 2133–2170
2137
toto per number-average PVC molecule
. The chloroallylic termini (A1, A2) and the IA
segments can be determined quantitatively by
1
H NMR measurements when a field strength of 500 MHz
is used
. Application of such a procedure did, in fact, reveal a total double-bond content that was
always very close to one per molecule in a group of suspension PVC resins that were made at 12 different
commercial temperatures ranging from 32 to 82 8C
. Moreover, the same result was forthcoming
for a number of PVC specimens that were prepared under subsaturation conditions at 55 or 80 8C, even
though, in this case, the auxiliary transfer pathway was actually the major one that was followed at the
lowest VC pressures used
. From these numerical findings, it is apparent that the R
z
radical in
, like radicals 3 and 4, does not experience unimolecular b scission in order to form the IA structure
and a kinetically free chlorine atom. If it did react in that way, then additional IA moieties would arise
from the ancillary chain reaction comprising Eqs. (3) and (4), and the total double-bond content thus
should be observably greater than the value of one per molecule that was actually found
.
PVC þ Cl
z
! HCl þ – CHClC
z
HCHCl –
ð3Þ
– CHClC
z
HCHCl – ! Cl
z
þ IA
ð4Þ
The overall monomer transfer constant, C
M,total
, is given by Eq. (5). Here, C
M,aux
refers to the auxiliary
transfer in
and hence can be set equal to the number of IA groups per monomer unit.
C
M
;total
¼ C
M
;HH
þ C
M
;aux
ð5Þ
In the literature, molecular-weight data for PVC have been used to develop a number of Arrhenius-type
expressions that permit the calculation of C
M,total
for various temperatures of polymerization
.
compares some values of C
M,total
computed in that way with those found with Eq. (5) from
double-bond contents that we determined by
1
H NMR. The polymers analyzed spectroscopically were
the conventional suspension resins, referred to above, that contained , 1 double bond in toto per
Fig. 2. Auxiliary mechanism for transfer to monomer during the polymerization of vinyl chloride (VC), where P
z
is the head-to-
tail macroradical, and the ks are rate constants
. (Reprinted with permission from
. Copyright 1998 American
Chemical Society.)
W.H. Starnes Jr. / Prog. Polym. Sci. 27 (2002) 2133–2170
2138
number-average molecule
. Agreement between the tabulated sets of C
M,total
values is remarkably
good.
Because the number-average degree of polymerization, (DP)
n
, is controlled by transfer to monomer,
C
M,total
can be equated to (DP)
n
21
. For that reason, Eq. (6) ensues immediately from Eq. (5). When the
formation rate of radical R
z
(in
) is equated to its rate of disappearance, a steady-state kinetic
treatment converts Eq. (6) into Eq. (7), where the brackets indicate molar concentration, and A is the
approximately constant term [^ (2 – 5)%] that Eq. (8) defines
ðDPÞ
21
n
¼ C
M
;HH
þ C
M
;aux
ð6Þ
ðDPÞ
21
n
< C
M
;HH
2
k
h
k
s
k
p
ðk
0
a
þ k
s
Þ
þ
k
s
ðk
h
A þ k
00
b
Þ
k
p
ðk
0
a
þ k
s
Þ½VC
ð7Þ
½PVC þ ½VC < A
ð8Þ
As required by Eq. (7), experimental plots of (DP)
n
21
vs. [VC]
21
were pleasingly linear
. Thus their
slopes and intercepts could be used to determine C
M,HH
with the aid of Eq. (9), where
r
BA
is the number
of IA groups, per monomer unit, formed via the ‘backbiting’ process whose rate constant is k
00
b
(
.
C
M
;HH
< intercept þ
slope
A
2
r
BA
½VC
A
ð9Þ
That process involves a 1,6 hydrogen transfer and produces IA moieties that are converted, to the
extent of 50%, into monoalkylcyclopentane (MCP) chain ends upon reduction with Bu
3
SnH
.
The MCP concentration, as determined by
13
C NMR, could then be used to calculate
r
BA
; as well
Table 1
Comparison of C
M,total
values (adapted with permission from
. Copyright 1996 American Chemical Society)
Polymerization temperature (8C)
C
M,total
£ 10
3
Literature
a
1
H NMR
b
82
3.1 ^ 0.5
2.7 ^ 0.4
80
2.9 ^ 0.5
2.4 ^ 0.4
70
2.2 ^ 0.3
2.2 ^ 0.3
61
1.7 ^ 0.3
1.6 ^ 0.2
57
1.5 ^ 0.2
1.2 ^ 0.2
56
1.4 ^ 0.2
1.5 ^ 0.2
53
1.3 ^ 0.2
1.1 ^ 0.2
52
1.3 ^ 0.2
1.1 ^ 0.2
49
1.1 ^ 0.1
1.3 ^ 0.2
40
0.8 ^ 0.1
0.8 ^ 0.1
36
0.7 ^ 0.1
0.6 ^ 0.1
36
0.7 ^ 0.1
0.8 ^ 0.1
32
0.6 ^ 0.1
0.6 ^ 0.1
a
Mean values obtained from four different Arrhenius expressions
; deviations are the average deviations from the mean.
b
Values calculated from published data
; deviations are the estimated experimental errors.
W.H. Starnes Jr. / Prog. Polym. Sci. 27 (2002) 2133–2170
2139
as the
r
BA
½VC
=A term in Eq. (9), which was shown by kinetic analysis to have a constant value at a
given temperature
Calculations based upon the available experimental data showed that the
r
BA
½VC
=A term was small
enough to be ignored
. Thus the first two terms in Eq. (9) were used to calculate C
M,HH
at several
temperatures, and the results are shown in
, where they are compared with those obtained by
1
H
NMR
. Also included in this table, in parentheses, is another set of C
M,HH
values obtained as the
intercepts of linear plots of (DP)
n
21
vs. [PVC]/[VC], where [PVC] is the molar concentration of
polymerized monomer units.
2
The C
M,HH
values based on molecular weights were used in Eq. (6) to calculate several values of
C
M,aux
. In
, the latter values are compared with the analogous
1
H NMR numbers. In view
of the excellent agreement revealed by the sets of data in
, the evidence for the occurrence
of all of the transfer reactions discussed above now seems overwhelming. Some additional remarks must
be made here, however, about the quantitative determination of the alkene segments in PVC.
During polymerization, as the pressure of vinyl chloride decreases, increasing numbers of
chloroallylic end groups are converted into internal allylic structures by reactions that have now been
identified conclusively
as those in Eqs. (10) and (11)
. Since these reactions do not change
the total number of double bonds per monomer unit, they have no effect on C
M,total
values found by
1
H
NMR. Nevertheless, they obviously will alter the individual NMR-based values of C
M,HH
and C
M,aux
unless corrective measures are taken. Fortunately, a simple procedure for making the corrections is on
hand
P
z
þ A1
=A2 ! PH þ –CH
2
CH¼CHC
z
HCl
ð10Þ
– CH
2
CH¼CHC
z
HCl !
VC
!
etc
:
IA
ð11Þ
Another caveat has to do with the analytical method that is used to determine the double bonds.
Table 2
Comparison of C
M,HH
values
Polymerization temperature (8C)
C
M,HH
£ 10
3
By
1
H NMR
a
From M
n
data
b
80
1.8 ^ 0.3
1.7 ^ 0.3 (1.8 ^ 0.3)
65
1.4 ^ 0.2
1.3 ^ 0.2 (1.3 ^ 0.2)
55
0.95 ^ 0.1
0.8 ^ 0.1 (0.9 ^ 0.1)
45
0.7 ^ 0.1
0.6 ^ 0.1 (0.7 ^ 0.1)
a
; deviations are the estimated experimental errors.
b
; deviations are the estimated experimental errors. See text for discussion.
2
When k
h
[PVC] @ k
00
b
applies, as was the case here, steady-state kinetics show that reliable values of C
M,HH
can also be
obtained in this way
.
W.H. Starnes Jr. / Prog. Polym. Sci. 27 (2002) 2133–2170
2140
High-field
1
H NMR is, in fact, the only technique that consistently gives trustworthy results
.
Chemical analyses are generally unsatisfactory and, if used at all, must be calibrated with NMR data.
In a recent investigation
, allylation with allyltrimethylsilane was reported to detect an ‘active
chlorine’ content (allyl þ tertiary chloride) of 1.7 ^ 0.1 per number-average molecule in a commercial
PVC specimen (Geon 110 £ 377) whose concentration of allyl chloride was said to be 1.5 per molecule
(as determined by
1
H NMR at a field strength of 300 MHz). In contrast,
1
H NMR analysis at 500 MHz
indicates an allyl chloride concentration of 0.9 per molecule for the same commercial polymer
. That
value seems very likely to be correct, within the probable limits of error, in view of its agreement with
the value of , 1 per molecule that is now to be expected for all virgin PVC resins
(see above). The
discrepancy can be explained by noting that the authors in
used an erroneous M
n
value of
36 600 in their calculations. When the correct M
n
of ca. 21 000
is used instead, their allyl chloride
value becomes 0.9 per molecule and is almost identical to the total corrected ‘active chlorine’ content
found by allylation (1.0 ^ 0.1 per molecule). Yet PVC does contain tertiary chloride, which is
associated primarily with dichlorobutyl branches and has been reliably determined
to occur there at
the level of 0.8 per molecule in Geon 110 £ 377
. Thus it seems that in this case, at least, the
allylation method found only about half of the active chlorines actually present. What fraction of them
was allylic cannot be deduced from the allylation results.
In another recent study, high-field
1
H NMR analysis gave a total double-bond content of 1.8 £ 10
23
per monomer unit for a suspension PVC prepared at 56 8C
. This value translates into a concentration
of 1.2 per molecule if one uses the M
n
expected for polymers made at this temperature (ca. 4.3 £ 10
4
), rather than the reported M
n
of 6.1 £ 10
4
, which is obviously much too high. Moreover, the
labeling in
suggests that the actual total double-bond content might have been only
1.4 £ 10
23
per monomer unit because the total CHyCH intensity was not corrected for the contribution
from the alkene protons of A1/A2. In that event, the experimental double-bond content would
become 1.0 per molecule, in perfect agreement with the value predicted.
For many years, molecular-weight changes resulting from the cleavage of chains by ozone have been
used to determine the number of internal double bonds and polyene sequences in PVC. However, in a
relatively recent investigation where internal double-bond contents obtained by ozonolysis and by
1
H
NMR were compared, the ozonolytic values were found to be much too low (by factors of ca. 4 and 9!)
for two fractions of PVC with M
n
values of 15 500 and 19 000, respectively
. Better agreement was
Table 3
Comparison of C
M,aux
values (adapted with permission from
. Copyright 1998 American Chemical Society)
Polymerization temperature (8C)
P/P
o
a
C
M,aux
£ 10
3
By
1
H NMR
b
From M
n
data
c
80
1.00
0.6 ^ 0.1
0.6 ^ 0.1
80
0.59
1.55 ^ 0.2
1.75 ^ 0.2
55
0.92
0.2 ^ 0.1
0.4 ^ 0.1
55
0.77
0.5 ^ 0.1
0.7 ^ 0.1
55
0.59
1.15 ^ 0.2
1.1 ^ 0.2
a
(Actual VC pressure)/(saturation VC pressure).
b
Values from
c
Values from Eq. (6)
W.H. Starnes Jr. / Prog. Polym. Sci. 27 (2002) 2133–2170
2141
obtained with samples having higher molecular weights
, but even so, now it is quite apparent that
ozonolytic double-bond values are highly suspect, in general.
The mechanisms shown in
and in Eqs. (10) and (11) predict a continual increase in the number
of internal double bonds with decreasing monomer concentration. This prediction has been verified by
1
H NMR data
. Yet some ozonolytic double-bond values published recently fail to reveal such a
trend
. Thus these values undoubtedly are incorrect, and their lack of correlation with thermal
stability
is not surprising.
From the numerous high-field NMR analyses performed by us and other workers over a period of
many years, it now is clear that the IA and A1/A2 moieties are the only alkene structures that occur in
significant concentrations in commercial PVC specimens. Several other alkene structures were proposed
in the older literature, but their presence has been disproved convincingly by the many NMR studies
done more recently at high magnetic field strengths.
The allylic isomer, – CH
2
CHClCHyCH
2
, of the (A1/A2) end groups is ordinarily undetectable in
whole polymers, though low concentrations of it can be found in PVC fractions with low molecular
weights
. However, major amounts of this isomer were observed in some novel PVC samples that
were made by initiation with di-tert-alkylmagnesium etherates
. These polymers also contained
significant amounts of cyclopropyl, trans-2-tert-butylcyclopropyl, and trans-t-BuCHyCHCH
2
– chain
ends, none of which is ever found in conventional PVC resins
. A new internal trans-alkene moiety,
– CHClCH
2
CHyCHCH
2
CHCl – , was detected as well and suggested to arise from the reactions in Eqs.
(12) and (13)
. Polymerizations involving both free-radical and quasi-anionic components were
thought to be responsible for these unusual findings
.
P
z
þ – CHClCH
2
CHClCHyCH
2
! – CHClCH
2
CHClC
z
HCH
2
CHCl –
ð12Þ
– CHClCH
2
CHClC
z
HCH
2
CHCl – þ VC ! – CHClCH
2
CHyCHCH
2
CHCl – þ ClCH
2
C
z
HCl ð13Þ
A theoretical model developed recently for the bulk and suspension polymerization of vinyl chloride
successfully reproduced some experimental degrees of polymerization for 50 8C and a conversion range
in which a saturation concentration of monomer was maintained
. The model assumes that the
monomer transfer constant has the same numerical value in both the monomer-rich and the polymer-rich
phases of the system. However, the approach developed by us and outlined above shows that the values
of C
M,total
in the two phases cannot be one and the same, because the unequal concentrations of monomer
in these phases must lead to different values of C
M,aux
. Thus a closer look at the effect of phase change on
C
M,total
is required. Relevant information for polymerization at 50 8C is lacking, but that process should
closely resemble polymerization at 55 8C, for which data are available.
Comparison of Eqs. (6) and (7) shows that C
M,aux
is a linear function of [VC]
21
in the polymer-rich
(polymer) phase. When such a line was constructed for 55 8C by using [VC] values determined
previously
and the corresponding average values of C
M,aux
, its extrapolation gave a
C
M,aux
value of ca. 0.15 £ 10
23
at a saturation concentration of monomer. Since C
M,HH
is about
0.9 £ 10
23
at the temperature of interest
, it follows from Eq. (5) that C
M,total
is about
1.05 £ 10
23
at 55 8C in the monomer-saturated polymer phase.
Inspection of
allows one to write Eq. (14), where
r
IA
represents the total number of IA groups
per monomer unit.
r
BA
=ð r
IA
2
r
BA
Þ ¼ k
00
b
=k
h
½PVC
ð14Þ
W.H. Starnes Jr. / Prog. Polym. Sci. 27 (2002) 2133–2170
2142
Results reported previously
show that in the polymer phase, the values of
r
BA
and
r
IA
are ca.
0.4 £ 10
23
and 1.1 £ 10
23
, respectively, when [PVC] is 20 M and [VC] is 1.5 M at 55 8C
.
Hence, from Eq. (14), the approximate value of k
00
b
/k
h
is 10. If a similar value applies, as expected, to the
monomer-rich (monomer) phase, wherein the concentration of PVC is minuscule (, 10
24
M
), then
it can be seen from Eq. (14) that, in this phase, essentially all of the IA groups are formed by the
backbiting route (i.e.
r
BA
<
r
IA
; because r
BA
=ð r
IA
2
r
BA
Þ @ 1). In either phase,
r
BA
is inversely
proportional to [VC]
, a situation that leads directly to Eq. (15), where the subscripts ‘M’ and ‘P’
refer to the monomer and polymer phases, respectively.
ð
r
BA
Þ
M
=ð r
BA
Þ
P
¼ ½VC
P
=½VC
M
ð15Þ
At 55 8C, [VC]
M
is 13.4 M, and as noted already, ð
r
BA
Þ
P
is roughly 0.4 £ 10
23
when [VC]
P
is 1.5 M. Use
of these numbers in Eq. (15) gives a ð
r
BA
Þ
M
value of only 0.04 £ 10
23
that is also the value of C
M,aux
in
the monomer phase. Thus it follows from Eq. (6) that, in this phase, the value of C
M,total
is essentially
determined by that of C
M,HH
, which is 0.9 £ 10
23
, as noted above. Therefore, under a saturation pressure
of monomer at 55 8C, the values of C
M,total
in the two phases differ by only about 10%, a conclusion that
tends to justify the reported use
of a single transfer constant at 50 8C. It should be realized, however,
that the numerical effects of phase composition on C
M,total
may be much larger under conditions (higher
temperature, monomer starvation
) that tend to increase the relative importance of C
M,aux
.
Dawkins et al.
have provided additional recent support for the transfer process in
by showing that two linear VC oligomers that they isolated from a commercial PVC resin
were diastereomeric tetramers whose chain ends were – CH
2
CHClCH
2
Cl and the trans isomer
of – CH
2
CHyCHCH
2
Cl.
2.1.3. Tertiary chloride formation
The existence of tertiary-chloride-containing structures in PVC has been known for many years
. Long-branch points (arrangement LB) result from chain transfer to polymer via the mechanism
in Eq. (16)
.
ð16Þ
However, the LB concentrations of commercial resins are quite low
. Most of the tertiary chloride
occurs, instead, in the 2,4-dichloro-n-butyl branch (BB) moiety that arises from the reactions shown in
Eqs. (17) and (18)
P
z
! – CH
2
C
z
ClCH
2
CHClCH
2
CH
2
Cl
5
ð17Þ
ð18Þ
Conclusive evidence for the presence of the BB structure was obtained several years ago from the
13
C
NMR spectrum of PVC that had been exhaustively dechlorinated with Bu
3
SnD
W.H. Starnes Jr. / Prog. Polym. Sci. 27 (2002) 2133–2170
2143
More recently, the experimental BB contents of many PVC samples have been used to develop a
simple equation that predicts the BB concentration of the polymer from the numerical values of only two
variables: the temperature and the average concentration of monomer during polymerization
. The
equation is based theoretically on the mechanism of Eqs. (17) and (18) and confirms its operation
quantitatively. The equation also shows that the activation energy for the backbiting step that creates 5
(Eq. (17)) is greater than that for normal chain propagation by 4.4 kcal/mol
. In ordinary suspension
resins that are made at constant temperatures ranging from 40 to 82 8C, the BB concentration ranges
from 1.0 to 2.4 per thousand monomer units
.
When monomer concentrations during polymerization are very low, the situation with regard to
tertiary chloride is more complex. Under these circumstances, radical 5 is not converted exclusively into
structure BB. Instead, the radical (6) formed from 5 via Eq. (19) is intercepted to some extent by another
backbiting reaction (Eq. (20)) that forms radical 7
. That radical then experiences sequential addition
of monomer, as shown in Eq. (21), in order to yield a di-2-chloroethyl branch (DEB) structure that
contains two tertiary chloride moieties
Two aspects of the DEB formation are of particular interest
. First, at 55 – 80 8C, the rate (relative
to addition) of the second backbiting reaction (Eq. (20)) is greater than the relative rate of the first
backbite (Eq. (17)) by a factor of 15 – 16.
ð19Þ
ð20Þ
ð21Þ
Second, the abstraction of hydrogen a in 6 to give 7 is faster than the intramolecular abstraction of
hydrogen b by a factor of 3 – 4, at least (no abstraction of hydrogen b actually was detected). Possible
explanations of these findings have been considered
The DEB contents of whole polymers are relatively insignificant. Nevertheless, in polymer fractions
that are made at very low monomer levels, these contents are probably large enough to cause reductions
in heat stability that are quite substantial
.
The ethyl-branch structures in PVC still are occasionally assumed to incorporate tertiary chloride and
to contribute to thermal instability for that reason
. However, as noted previously
,
commercial resins actually contain the EB moiety instead, and it is not an unstable group.
Branch contents were reported recently for a series of suspension PVC resins made at 55 8C and at VC
conversions of 6.3 – 93.5%
. The branch values were determined by
13
C NMR after reductive
dechlorination with Bu
3
SnH. As expected
, the number of butyl branches increased with increasing
conversion
, especially at conversions that would have led to decreasing concentrations of monomer
in the monomer-swollen polymer phase. Moreover, the sum of the methyl and ethyl branch contents was
roughly constant (except at the lowest conversion used), while the ethyl/methyl ratio rose as the
W.H. Starnes Jr. / Prog. Polym. Sci. 27 (2002) 2133–2170
2144
conversion increased
. Those observations also are in line with expectations
, though the
scattering of the data
was much greater than in earlier work
. A possible reason for the scatter is
incomplete dechlorination, which is known to result on occasion
from the reduction procedure
that was used
. In that event, appreciable amounts of partially reduced branches are formed, even
when the total level of reduction is very high, and the concentrations of these branches must be added to
those of their fully dechlorinated counterparts in order to obtain the correct concentrations of the
branches in the original (unreduced) polymer
In another recent study, branch concentrations were reported for a polymer whose extent of reductive
dechlorination was only ca. 90%
. Those concentrations were not corrected for partial reduction and
thus are unreliable.
A review article authored by Maddams
asserts that a paper by Starnes et al.
concludes that
tertiary chloride is not present at the long-branch points of PVC. The assertion referred to
is
incorrect, as a brief perusal of
will show.
2.1.4. Mechanism and relative importance of initiation by internal allylic and tertiary chloride
The literature contains abundant evidence for the thermal dehydrochlorination of allylic and tertiary
chlorides in an ionic or quasi-ionic manner. Much of this evidence has to do with the effects of
conditions and structure on reaction rate, and with the failure of other suggested mechanisms to account
for the facts. A few of the important older references in this area are
. The ionic mechanism
is depicted in Eq. (22), where intermediate 8 is best regarded as an ion-pair instead of free ions, in view
of the low polarity of the medium.
ð22Þ
The quasi-ionic mechanism is shown in Eq. (23).
ð23Þ
It involves the concerted loss of HCl in a single step, via a four-center transition state with a large
amount of charge separation in the breaking C – Cl bond. Catalysis of the dehydrochlorination by HCl, a
well-established phenomenon
, can be accounted for most economically by the ion-pair scheme in
Eq. (24) or the concerted process in Eq. (25).
ð24Þ
ð25Þ
W.H. Starnes Jr. / Prog. Polym. Sci. 27 (2002) 2133–2170
2145
Further support for an ionic or quasi-ionic pathway appears in
, which contains
dehydrochlorination rate constants for compounds 9 – 14.
For each of these model substances, the rate increases significantly upon going from o-dichlorobenzene
to benzophenone, a solvent of higher polarity
. This solvent causes a similar rate enhancement for
PVC itself
, a result which implies that the dehydrochlorination of the polymer is ionic or quasi-ionic
as well.
The data in
were obtained by acid – base titrimetry for which the HCl was swept by argon into
the vessel used for titration. In o-dichlorobenzene, the k value for tertiary chloride 9 shows only slight
variations in different runs when the argon flow rate is changed. However, the ks for allylic chlorides 10
and 11 decrease significantly when the flow rate is raised. As a result, these allylic chlorides are more
reactive than 9 at the lowest flow rate used for them and less reactive than 9 at the highest flow rate. The
higher the rate of argon flow, the lower the HCl concentration in the reaction mixtures. Therefore, it is
Table 4
Dehydrochlorination rate constants for model compounds. (Reprinted with permission from Table 2, p. 112 of Ref.
published by Harwood Academic Publishers)
Model
k
a
£ 10
5
(min
21
)
o-Cl
2
C
6
H
4
Ph
2
CO
9
780
b
5800
1100
950
c
10
3120
b
3300
600
290
c
11
2000
4900
860
c
12
6
21
13
7
22
14
2
20
a
At 170 ^ 0.5 8C with an argon flow rate of 0.14 ml/s unless noted otherwise; reproducibilites were # (^ 7%).
b
Argon flow, ! 0.14 ml/s (too slow for accurate measurement).
c
Argon flow, 1.3 ml/s.
W.H. Starnes Jr. / Prog. Polym. Sci. 27 (2002) 2133–2170
2146
apparent that, during dehydrochlorination, allylic chlorides are more susceptible than a tertiary chloride
to HCl catalysis, and that the reactivities of these two types of structure can, in fact, be inverted by
changing the HCl concentration
. Work reported later by other researchers also supports the stronger
response of allylic chloride dehydrochlorination to the presence of HCl
In PVC formulations that are undergoing thermolysis, the HCl concentration will be determined by
factors such as temperature, viscosity (which will affect the rate of HCl diffusion), specimen
morphology, and the ability of basic stabilizers to scavenge HCl. Thus the HCl concentration is very
hard to predict or measure, and for that reason, the relative stability of internal allylic and tertiary
chloride structures in the polymer, in any given situation, is likely to remain obscure. The relative total
contributions of the two types of structure to degradation initiation will also depend, of course, on their
relative concentrations.
Through mathematical modeling studies that recently were reviewed
, Troitskii et al.
have provided support for the proposition that internal allylic and tertiary chloride are the only structural
defects that are important initiators. The early stages of experimental rate curves for thermal
dehydrochlorination were well reproduced by the modeling when reasonable (and, in some cases,
experimentally measured) values were used for the concentrations and dehydrochlorination rate
constants of the two types of defect, the rate constant for dehydrochlorination of the ordinary monomer
units, and the average length of the conjugated polyene sequences that were formed
. All of the
results pertained to degradation that, owing to the effective removal of HCl, did not exhibit autocatalysis
.
Could free radicals be involved in the thermolysis of simple allylic and tertiary chlorides? Evidence
for their absence was obtained by heating dilute solutions of 9 or trans-4-chloro-5-decene in
toluene
. The temperature was 180 8C, and the experiments were carried out in sealed tubes under
vacuum after thorough degassing. Analyses performed by GC/MS showed that even at conversions of
ca. 50%, no products derived from toluene had been formed, at an estimated detection sensitivity of at
least 0.1 mol%. If chlorine atoms or any other free radicals had been produced, they would have been
likely to abstract benzylic hydrogen from the solvent in order to generate benzyl radicals, which should
have been converted, at least to some extent, into bibenzyl or other coupling products.
A lack of evidence for kinetically free radicals does not exclude the mechanism of Eq. (26), in which a
radical pair (15) disproportionates within the solvent or medium cage where it is formed. That
mechanism is ruled out, however, by the relatively low energies of activation that have been found for
the thermal dehydrochlorination, in the condensed phase, of model aliphatic chlorides
. These
energies are far less than the homolytic dissociation energies of aliphatic C – Cl bonds.
ð26Þ
Bacaloglu et al.
have argued that trans allylic chlorides do not dehydrochlorinate at the
temperatures where PVC decomposes. Instead, the degradation of these structures is said to involve their
isomerization into cis allylic chlorides that then lose HCl exclusively in a concerted six-center process
. The evidence that repudiates this mechanism is discussed in Section 2.2.4.
W.H. Starnes Jr. / Prog. Polym. Sci. 27 (2002) 2133–2170
2147
2.1.5. Initiation by ‘carbonylallyl’ structures?
Minsker et al. believe that the thermal stability of PVC is essentially unaffected by chloroallyl groups
per se
. They suggest that these structures are converted rapidly and practically quantitatively into
carbonylallyl moieties (16) when the polymer is exposed to air under ambient conditions
They also argue that 16 is the only defect structure that contributes appreciably to thermal instability
Kinetic studies with model compounds showed that the thermal stability of trans-16 ðn ¼ 1Þ actually
is quite high—so high, in fact, that this structure cannot be regarded as a labile defect
. However, a
model for cis-16 ðn ¼ 1Þ was much less stable than the ordinary monomer units and underwent
dehydrochlorination to form a furan which, in the polymer, would appear as segment 17
. That
structure was shown to be very unstable, as well, and to dehydrochlorinate in a way that would initiate
the growth of a conjugated polyene sequence with a furan group at one end
. Nevertheless, the
properties of cis-16 obviously are irrelevant if this group does not occur in the polymer.
Neither we nor, to the best of our knowledge, any other researchers have obtained convincing NMR
evidence for the presence of 16 or furanoid moieties in PVC. According to Minsker et al., their
concentrations would be quite low. Even so, modern NMR instrumentation would be able to detect them
easily
if they were present at the reported
levels. In 1991, 125 MHz
13
C NMR was used to
search for these structures
in a specific sample of PVC that had been analyzed by Professor Minsker
and was stated by him to contain 16 at the level of 0.05/(1000 C). The NMR work was done with
Bu
3
SnH- or Bu
3
SnD-reduced specimens of this polymer, and the results provided no evidence
whatsoever for the presence of reduced 16, reduced furans, or any other structures into which
they or their progenitors might have been transformed
. However, the supposed precursor of
16 ðn ¼
1Þ
; –CH
2
CHyCHCHCl –
, was shown to occur in the unreduced resin at the level of
0.1/(1000 C)
In view of these observations and the other negative evidence in the literature, it would appear that
the moment of truth with regard to the presence of 16 in PVC has arrived (it actually arrived many years
ago
In a related investigation
, the IR spectrum of a PVC film cast from DMF solution was found to
contain rather strong bands at 3442 and 1065 cm
21
. They were attributed to – OOH groups and thus
seem to have been indicative of extensive oxidation. After the film was heated under vacuum at 190 8C,
these absorptions were replaced by a weak band at 2851 cm
21
and an extremely weak one at 1680 cm
21
that were ascribed to aldehyde C – H stretching and ‘keto allylic stretching’, respectively. These results
were considered to indicate the formation of carbonylallyl moieties that initiated dehydrochlorination.
However, no additional evidence to support this conclusion was presented.
Cisoid – COCHyCH– structures have been said to be true catalysts for the dehydrochlorination of
PVC
. Even so, as is the case for 16, conclusive evidence for their presence in commercial resins
is lacking. In a relevant model-compound study, the dehydrochlorination of 7-chlorotridecane at 180 8C
was actually strongly retarded by cis/trans mixtures of the linear enones C
5
H
11
CHyCHCOC
7
H
15
and
W.H. Starnes Jr. / Prog. Polym. Sci. 27 (2002) 2133–2170
2148
C
3
H
7
CHClCHyCHCOC
4
H
9
. This effect conceivably could have resulted from the likely ability
of these enones to scavenge the catalyst, HCl, via its electrophilic addition to the conjugated system.
2.1.6. Initiation by ordinary monomer units
It has been apparent for some time that the ordinary monomer units in PVC can initiate its thermal
dehydrochlorination to some extent
. Studies with model compounds indicate that the
dehydrochlorination rates of the ordinary units are typically less than those of internal allylic and
tertiary chloride structures by some two to three orders of magnitude
. Nevertheless, the overall
contributions of the ordinary units may be significant, because their total concentration obviously is
much greater than those of the conventional labile defects.
The ground-state energies of the m dyads and mm triads in PVC are slightly higher than those of the r,
mr, and rr structures
. Moreover, in some conformers of the isotactic and heterotactic species,
acceleration due to backside participation by neighboring chlorine helps to explain the reactivity order
isotactic . heterotactic . syndiotactic that has been found for certain reductive dechlorinations and
may apply to thermal dehydrochlorination, as well
. Even so, liquid-phase thermal stabilities of
2,4,6-trichloroheptane and the stereoisomers of 2,4-dichloropentane
do not suggest that such effects
can cause any of the ordinary units in PVC to decompose at rates that are competitive with those of
internal allylic and tertiary chloride moieties. Nevertheless, Milla´n et al. have argued for many years that
the only important labile structure in PVC is, in fact, the GTTG
2
conformer (18, denoted as GTTG by
some workers) of the mm triads that occur in mmr segments of the polymer
.
This conformer is supposedly removed selectively by the nucleophilic displacement of its central
substituent by benzenethiolate anion, a process that is said to lead to greatly improved stability
.
However, the proposed destabilization by 18 has been questioned by many researchers, and the
following points are instructive in this regard.
(1) When the temperature of polymerization is changed, isotacticity and the concentrations of the
conventional labile defects change in parallel ways. Thus the improved stability resulting from
benzenethiolate treatment could actually result from the deactivation of conventional defects by labile
chloride displacement
. Treatment of PVC with trimethylaluminum gave a polymer whose
improved stability was fully explained by the replacement of allylic and tertiary chloride by methyl
groups
. Importantly, the extent of this methylation was much less than the concentration of the
GTTG
2
mm triad
(2) The structural features of 18 that might cause it to be highly unstable are by no means apparent.
Breakage of its C4 – Cl bond cannot be assisted easily by backside Cl participation but would be favored
by the relief of two 1,3 axial Cl – H interactions. Yet the same statement can be made about the central
C – Cl bond of an rr triad in its preferred TTTT conformation. Hence, the rate of cleavage of the C4 – Cl
bond in 18, by any reasonable mechanism, should be about the same as that for cleavage of the analogous
bond in the rr species. Milla´n et al. have argued, however, that the thermally labile C – Cl bonds in 18 are
actually those at C-2 and C-6
. If the proposed environment of 18 in the polymer
is correct,
W.H. Starnes Jr. / Prog. Polym. Sci. 27 (2002) 2133–2170
2149
one of these two bonds will involve the central carbon of an mm triad in its GTGT conformation, while
the other will involve the central carbon of an mr triad with the TGTT conformation. These two
conformations comprise a major fraction of the total triad content of the polymer
.
Thus the supposed lability of C2 – Cl and C6 – Cl in 18 implies that a very sizable fraction of the ordinary
monomer units should be highly unstable as well. This is obviously not the case.
(3) In conventional PVC resins, some 15 – 20% of the triads have the mm configuration
.
Experimental and theoretical studies have shown that 20 – 30% of these triads are of the GTTG
2
type
. Hence the equilibrium concentration of this conformer in the polymer is ca. 3 – 6%, rather than
0.2 – 0.8%, as proposed by Milla´n et al.
. In PVC, the energy barrier to conformational
interconversion is so low
that conformational equilibrium undoubtedly is maintained at the
temperatures where PVC degrades
. Therefore, selective chemical modification of 0.2 – 0.8% of
the polymer can reduce the equilibrium concentration of 18 to only a minor extent and thus would have
little or no effect on thermal stability, if 18 were, in fact, the only unstable structure.
(4) An activation energy of 7 kcal/mol has been reported for the dehydrochlorination of 18
. This
value seems much too low to be associated with a simple thermolysis reaction of a neutral nonradical
species. It may relate, instead, to a physical phenomenon such as HCl diffusion.
(5) Several research groups have been unable to confirm the postulated destabilization of PVC by
isotacticity. Rogestedt and Hjertberg found that the correlation of stability with isotacticity was much
weaker than its correlation with the content of conventional labile structures
. From a mathematical
modeling study, Behnisch et al. concluded that isotacticity would actually increase stability
. In
computer simulations whose results were compared with
13
C NMR data, Radiotis and Brown obtained
no evidence for the preferential dehydrochlorination of isotactic stereosequences
. However, they
considered their observations to be inconclusive because of experimental limitations
. Another
simulation study was reported by Kurskii
. It took into account the possible termination of polyene
growth by cyclization and concluded that even in the most favorable cases, isotacticity can decrease
stability by no more than 30%
. Troitskii et al. detected no correlation of isotacticity with initial or
stationary rates of dehydrochlorination in experiments performed at 200 8C with efficient removal of
HCl
. The same result was obtained for 200, 240, and 250 8C from a computer modeling approach
that reproduced the experimental rate curves up to 30 – 40% conversion
.
(6) Stability variations were found for PVC that had been subjected to a variety of treatments which
were designed to deactivate or destroy the GTTG
2
conformer without causing other changes that would
reduce the rate of formation of HCl
. These treatments included the supposed selective complexation
of the offending conformer with polyesters or esters with low molecular weights and removal of the
conformer by stretching the polymer at 90 8C or annealing it thermally
. Depending upon conditions,
the treatments led to stabilization, destabilization, or no effect on stability. An explanation of all of the
data reported will not be attempted here. Nevertheless, it should be noted that the stability might well
have been altered by several alternative factors. These include such things as deactivation of HCl by
carbonyl groups, via hydrogen bonding, changes in medium polarity caused by the introduction of
additives, and changes in crystallinity, which could have affected the degree of catalysis by HCl.
In the light of the foregoing comments, it should now be apparent that the proposed instability of
structure 18 has some very serious problems. This is not to say, however, that tacticity has no effect at all
on the rate of PVC dehydrochlorination. Raising the syndiotacticity can increase the overall rate by
decreasing the rate of polyene termination
. It apparently does so by tending to prevent the
formation of cis double bonds and thus inhibiting intramolecular polyene cyclization
. Moreover,
W.H. Starnes Jr. / Prog. Polym. Sci. 27 (2002) 2133–2170
2150
because of their higher free-energy content, isotactic monomer units may indeed be found, eventually, to
be slightly more reactive as initiators than syndiotactic units. Nonetheless, large variations in rate of
initiation that are caused by tacticity now seem highly improbable.
Except, perhaps, at very low conversions of monomer
, the tacticity of PVC that is prepared in the
absence of solvents is believed to be controlled exclusively by the temperature of polymerization. Thus,
if stability depended only upon tacticity, the stabilities of polymers made at high conversions and at the
same temperature should be the same. Yet this result was not observed for several PVC samples that
were prepared at a number of constant temperatures in homogeneous PVC – VC mixtures at various
constant monomer levels. At every temperature used, lowering the monomer concentration decreased
the stability while increasing the amounts of internal allylic and tertiary chloride
, as was to be
expected
. Such results would comprise apparently incontrovertible evidence against
exclusive initiation by ordinary monomer units if it were known with certainty that the variations of
monomer content left tacticity unchanged. This point is under investigation
2.2. Polyene growth and termination
2.2.1. Overview
In the following sections, two kinds of ‘growth’ are considered. One of these is the
propagation reaction that causes a specific polyene sequence to lengthen after its first double
bond has been formed. The other type of growth has to do with the creation of additional polyene
sequences which arise from labile defects that were absent from the original polymer and thus are
generated after degradation begins. As used here, ‘termination’ refers to any process that stops the
growth of a given polyene sequence.
2.2.2. Free-radical mechanisms
2.2.2.1. Radical formation. There are several potential sources of radicals in thermally degrading PVC.
Two trivial possibilities that can be avoided, at least in theory, are residual initiator from the
polymerization and peroxide groups resulting from the occurrence of accidental air oxidation, either
during polymerization or thereafter. These adventitious peroxides obviously can serve as precursors of
radicals during subsequent thermal degradations that occur in essentially anaerobic environments. Chain
cleavage during processing (mechanochemical chain scission) under high shear gives radicals
, but
the present author has been informed by representatives of several PVC manufacturers that this cleavage
is unimportant in industrial operations.
Free radicals might also be produced (but to an unknown extent) when benzene is formed via the
thermal cyclization of conjugated hexatriene sequences
. Moreover, under certain conditions, the
one-electron reduction of four-membered cyclic chloronium cations is a potential radical source
.
However, in thermally degrading PVC, the most important route to radicals under most, if not all,
conditions is probably the thermal excitation of conjugated polyenes (and polyenyl cations; see Section
2.2.2.2) into diradical states
, as in Eq. (27). The rate of this process increases considerably with
increasing polyene length
A suggested source of radicals that is devoid of credibility appears in Eq. (28). This reaction was
proposed by two research groups
who erroneously assumed that a reported energy requirement
for Eq. (27) of 36 kcal/mol was actually the energy input needed for Eq. (28). In fact, however, when
W.H. Starnes Jr. / Prog. Polym. Sci. 27 (2002) 2133–2170
2151
C – C s bonds are formed from sp
2
orbitals, as in the case of conjugated polyenes, they are extremely
strong. For example, the dissociation energy of the C
2
– C
3
bond of 1,3-butadiene, as deduced from the
heats of formation of this diene and the vinyl radical
, is 117 kcal/mol. The dissociation energy
would not be significantly less for the C – C single bonds in long polyenes, as the resulting vinylic
radicals obviously would not be stabilized by conjugation with the p-electron systems.
– ðCHyCHÞ
n
– !
K
– C
z
HðCHyCHÞ
n21
C
z
H –
ð27Þ
– ðCHyCHÞ
n
CHyCH – CHyCHðCHyCHÞ
m
– !
K
– ðCHyCHÞ
n
CHyC
z
H þ
z
CHyCHðCHyCHÞ
m
–
ð28Þ
Activation enthalpies ranging from 24.5 to 38.9 kcal/mol have been found for Eq. (27) when n ranges
from 9 to 3, and as expected, this process has been shown to result in cis – trans isomerization rather than
C – C scission
.
2.2.2.2. Radical degradation pathways. Many years ago, Arlman
proposed a thermolysis scheme for
polyene growth that involves the continual concerted loss of HCl molecules from a polyenyl free radical.
A single step of the Arlman process is shown in Eq. (29). However, if the Arlman mechanism is to be
operative
– ðCHyCHÞ
n
C
z
HCHClCH
2
–
19
!
k
el
– ðCHyCHÞ
nþ1
C
z
H – þ HCl
ð29Þ
19 !
k
b
– ðCHyCHÞ
nþ1
CH
2
– þ Cl
z
ð30Þ
Eq. (29) must compete successfully with Eq. (30). The probability that this is the case has now been
assessed quantitatively
with recourse to total delocalization energies for polyenes and polyenyl
radicals that are available in the literature
. Details of the relevant arguments are too lengthy to
present in this review, but the conclusions to which they lead are clear-cut: at 175 8C when n in radical 19
ranges from 0 to 3: (1) the rate constant for elimination (k
el
) is less than that for b scission (k
b
) by factors
of 10
12
– 10
2
; (2) constant k
el
is smaller by at least 2 – 3 orders of magnitude than the experimental rate
constant for polyene growth in PVC. These findings clearly suffice to exclude the Arlman scheme from
further consideration.
The chlorine-atom b scission of Eq. (30) is involved in another venerable mechanism proposed for the
thermal growth of PVC polyenes. This is the radical chain reaction whose propagation steps are depicted in
. One strong argument against this scheme emerged from the analysis just referred to
,
which predicted, inter alia, a drastic decrease of k
b
with increasing values of n in 19. As a result, when n was
only 3 at 175 8C, k
b
was already found to be less by an order of magnitude than the experimentally
Fig. 3. Radical-chain mechanism for the growth of a conjugated polyene sequence during PVC thermolysis
. (Reprinted
with permission from Scheme 3, p. 116 of Ref.
, published by Harwood Academic Publishers.)
W.H. Starnes Jr. / Prog. Polym. Sci. 27 (2002) 2133–2170
2152
determined value of the rate constant for polyene growth in PVC
. This observation implies that the
longer polyenes that exist in the thermolyzed polymer cannot have been formed by the route in
.
Another major objection to this pathway is based on the low selectivity of Cl
z
in H abstraction
reactions and the relatively high concentration of nonallylic hydrogens during the early stages of PVC
thermolysis. These factors tend to disfavor the allylic H abstraction that
depicts. Therefore, to the
extent that nonallylic methylene hydrogen were abstracted by Cl
z
, a chain reaction could ensue which is
analogous to that in
but would lead to isolated alkene linkages instead of a polyene sequence
.
Low viscosity of the medium will facilitate the diffusion of Cl
z
and thus also reduce the likelihood of
the mechanism of
. Admittedly, Cl
z
is so reactive that even in highly fluid media, it may abstract
hydrogen from aliphatic substrates in ‘cage’ reactions whose rates can compete, to some extent, with the
rate of Cl
z
diffusion
. Nevertheless, a quantitative reactivity analysis has shown that even during the
thermolysis of solid PVC, the scheme in
predicts that the rates of dehydrochlorination and new
polyene sequence formation should be similar
. In actual fact, the former rate is about 20 times faster
than the latter one at 190 8C
. Thus the evidence against the mechanism of
appears
conclusive.
Still, the lack of involvement of free radicals in the polyene growth reaction does not necessarily
preclude their intervention at some stage of thermal dehydrochlorination. It was pointed out previously
that radicals may abstract methylene hydrogen from ordinary monomer units in order to form new
radicals that then produce new thermally labile IA defects via Eq. (4). It was also suggested that in the
first step of this sequence, the abstracting radicals might be those resulting from the thermal excitation of
PVC polyene sequences
. Significant information that relates to this route to IA structures has become
available, as will now be shown.
Troitskii and Troitskaya have made some important contributions in this area that were reviewed very
recently
. A comparison of mathematical kinetic models with experimental observations indicated
that the well-known autocatalysis of dehydrochlorination by HCl begins with an interaction of HCl with
a polyene sequence
. This interaction appeared to be facilitated by increasing polyene length and to
lead to an unspecified chemical species that could reinitiate dehydrochlorination by attacking an
ordinary monomer unit
. A later study
implicated free radicals in autocatalysis, for it showed
that the fullerene C
60
, an effective radical trap
that did not scavenge HCl
, was able to retard
dehydrochlorination when HCl was present but not when it was removed from the system by freezing.
More evidence for radicals came from an ESR investigation of thermally degraded PVC
, which
revealed that the spin concentration was much greater when the removal of HCl was inefficient and
autocatalysis had occurred. The observed increase of stable spins was believed to be consistent with a
higher concentration of reactive radicals during autocatalysis. Finally, a detailed mechanism for
autocatalysis was formulated
with recourse to the energies of the first excited singlet and triplet
states of polyenes and polyenyl cations and/or the energy barriers for rotation about the central double
bond of these species.
The suggested mechanism is shown in
. Its initial progenitive species, ion-pair 20, results from
the HCl-assisted cleavage of the C – Cl bond of a polyenyl chloride (as in Eq. (24)). Thermal excitation of
the cation in 20 gives the triplet cation diradical in ion-pair 21, whose discharge by proton transfer yields
HCl and the neutral diradical, 22. The possible abstraction of a nonallylic methylene hydrogen by 22 was
considered but rejected
. Instead, 22 was thought to react with HCl by atom transfer, as shown, in
order to generate Cl
z
, which then attacked the polymer, presumably as in Eq. (3)
. The formation of
Cl
z
from ion-pair 20 in one step (Eq. (31)) was considered to be an alternative possibility
. It must be
W.H. Starnes Jr. / Prog. Polym. Sci. 27 (2002) 2133–2170
2153
noted, however, that Eq. (31) apparently is without precedent
20 ! –
CH
2
ðCHyCHÞ
n
C
z
H – þ HCl þ Cl
z
ð31Þ
and that the formation of Cl
z
from HCl and 22 is exceedingly doubtful because of the high bond strength
of HCl and the probable low reactivity of the resonance-stabilized diradical. This species would be more
likely to abstract hydrogen from a nonallylic methylene group, whose C – H bonds undoubtedly are much
weaker than the bond in HCl.
Recent research in the present author’s laboratory has provided additional insight into the reactions of
free radicals that lead to autocatalysis. In our work
, all-trans-b-carotene (23), which incorporates a
conjugated polyene system containing 11 double bonds, was used to mimic a PVC polyene sequence.
Protonation of 23 would give a polyenyl cation whose structure and chemical properties should resemble
those of the family of cations shown in structure 20.
At 180 8C under a slow stream of flowing argon, 4 wt% of 23 had no immediate effect on the rate of
PVC dehydrochlorination but caused the rate to autoaccelerate strongly after ca. 1.5 h. The temporal
profile of this autoacceleration closely resembled that reported
for the development of free spins
when PVC was heated with 23 at 180 8C under nitrogen. Moreover, in our system, most of the
autoacceleration was prevented by either (a) the addition of 10 wt% of the radical scavenger, 2,6-di-tert-
butyl-p-cresol, or (b) an eightfold increase of the argon flow rate, which would have greatly reduced the
steady-state concentration of HCl. Neither of these two changes had any substantial effect on the initial
rate of dehydrochlorination. Therefore, taken together, all of the observations just described are
consistent with a mechanism for autocatalysis that involves the following steps: (1) slow formation of a
polyenyl cation (cf. 20) via the protonation of 23 by HCl (at the outset, this process would be faster than
the generation of polyenyl cations from PVC, owing to the relatively high initial concentration of 23 as
compared to those of the developing PVC polyenes); (2) reversible thermal conversion of the 23 cation,
perhaps via a singlet diradical species, into a triplet cation diradical (cf. 21); (3) hydrogen abstraction by
the latter species from an unactivated methylene group of the polymer; (4) creation of the initiating
species, IA, via Eq. (4).
Hydrogen abstraction from PVC by a neutral diradical formed from 23 (cf. structure 22) is
Fig. 4. Mechanism proposed by Troitskii and Troitskaya
for the thermal conversion of conjugated polyenyl cations into
radical intermediates.
W.H. Starnes Jr. / Prog. Polym. Sci. 27 (2002) 2133–2170
2154
conceptually possible. However, its importance seems negligible in view of the failure of 23 to increase
the initial loss rate of HCl. There are at least two factors that may contribute to this behavior. One is that
the equilibrium concentration of the neutral diradical is relatively low because the thermal excitation
energy needed to produce this species is much greater than that required to form the corresponding
cation diradical
. Also, it may be that the lower electrophilicity of the neutral diradical simply makes
it less reactive than the charged species for the abstraction of nonallylic methylene hydrogen.
Experiments with 23 and 4-chloroheptane have provided further support for the mechanism of
autocatalysis that we currently favor. After 8 h at 180 8C under argon, regardless of whether 23
(10 wt%) were present or absent, the alkyl chloride experienced only 0.3% of dehydrochlorination.
When the alkyl chloride was heated with 10 wt% of phosphoric acid under the same conditions, the
extent of dehydrochlorination rose to 4.6%, evidently because the acid functioned (though rather
inefficiently) as an electrophilic catalyst for the loss of HCl cf. (Eq. (24)). However, this
dehydrochlorination percentage quadrupled in parallel runs where both 23 (10 wt%) and phosphoric
acid (10 wt%) were present. The synergism between these additives is consistent with their
interaction to form a cation diradical that removed hydrogen, to some extent, from the 3- and 5-
methylene moieties of the 4-chloroheptane. The resulting substrate radical would have been
converted by b scission into cis- and trans-3-heptene cf. (Eq. (4)).
Cation diradical chemistry may also help to explain some of the kinetic peculiarities that have
been unearthed in studies of PVC thermolysis. One of these phenomena pertains to rates of gaseous
HCl evolution from dilute solutions of PVC and 23 that were heated under argon at 195 8C
.
Increasing amounts of 23 decreased the rate monotonically until the concentration of this additive
had reached a certain level. This result was reasonably ascribed to the scavenging of HCl by its
addition to the carotenoid polyene system. Unexpectedly, however, a further increase of the
additive concentration caused a significant rate enhancement that was not explained
. Perhaps
the explanation has to do with hydrogen abstraction from the polymer by one or more isomeric
carotenoid cation diradicals whose concentration only rose to the level of kinetic significance when
the largest amount of 23 was used.
The peroxide-promoted dehydrochlorination of PVC at elevated temperatures was studied recently
with the aid of an ESR technique that allowed radical concentrations to be monitored quickly and
continually during degradation
. Rates of dehydrochlorination were assumed to measure polyene
growth by a radical pathway, and these rates were divided by the observed concentrations of radicals in
order to obtain rate constants for the growth reaction
. One of the difficulties with such an approach
is that rates determined by experiment for the loss of HCl actually include kinetic terms for initiation and
termination as well as propagation
. Also, in view of several arguments made earlier, the
concentrations of the radicals in thermally degrading PVC cannot be equated to the number of growing
polyene sequences. The radical signals detected by ESR arise primarily from products of degradation
instead of reactive intermediates.
2.2.3. Ion-pair/quasi-ionic mechanism
Nowadays, most investigators of the thermal dehydrochlorination of PVC accept a mechanism for
polyene propagation that involves ion pairs (Eq. (22)) or a highly polarized (quasi-ionic) four-center
transition state (Eq. (23)). The ion-pair mechanism is shown in
. Unlike the four-center
process, it is not disallowed by orbital symmetry. It must be noted, however, that forbiddance by
W.H. Starnes Jr. / Prog. Polym. Sci. 27 (2002) 2133–2170
2155
symmetry is merely an indication of a high activation energy for a process, rather than an absolute
proscription against its occurrence.
Section 2.1.4 presents evidence that validates an ion-pair/quasi-ionic mechanism for the thermal
dehydrochlorination of labile structural defects. Most, if not all, of this evidence can be applied to the
polyene growth reaction. Additional pertinent literature is cited in
, including a book chapter
in which the ion-pair pathway is supported by ab initio molecular orbital calculations and
experimental observations that relate the results of these calculations to the molten polymer phase. In a
subsequent quantum chemical study, solid PVC was modeled by surrounding a chloride anion with four
molecules of 1,3-dichlorobutane
. That approach provided direct computational evidence for ionic
dehydrochlorination in the solid polymer
. Further verification of the ion-pair/quasi-ionic
mechanism has come from some new dehydrochlorination studies with model compounds that are
discussed in Section 2.2.4.
Bacaloglu et al. have provided several arguments against the ‘ionic’ degradation of PVC
.
Most of these arguments relate to a mechanism involving free ions, and none offers persuasive evidence
against an ion-pair or quasi-ionic process. Actually, insofar as the present author is aware, the
intervention of free ions has never been considered seriously. That this is the case is shown, for example,
by the following statement in
: “…it seems unlikely that free ions will ever achieve a high
concentration in PVC”.
Bacaloglu and Fisch (BF) have reported the results of semiempirical molecular orbital calculations on
several structures that could be involved in the ionic dehydrochlorination of the polymer
. Since
these calculations dealt exclusively with free ions in the gaseous state, they yielded enthalpies for C – Cl
ionization that were very large—large enough, in fact, to imply that the ionic dehydrochlorination was
impossible
. However, what this argument failed to consider is that most of the energy required to
form free ions from neutral substrates is used to effect the separation of opposite charges to an infinite
distance
. Ion-pair formation requires much lower enthalpy inputs, and in the case of simple
chloroalkanes, for example, the energies for C – Cl heterolysis have been estimated to decrease by
99 kcal/mol when the process gives an ion-pair instead of free ions (for tert-butyl chloride, the reported
energies are 56 and 155 kcal/mol, respectively)
. Moreover, in the condensed phase, solvation
causes the formation of ion pairs to become even more favorable.
Experimental and theoretical evidence show that under some conditions, four-membered cyclic
chloronium cations are intermediates in the dehydrochlorination of PVC
. Nevertheless, BF
report calculations which purport to demonstrate that such a species does not form because its stability is
only slightly greater than that of its open (acyclic) isomer
. Again the argument is invalidated by the
Fig. 5. Ion-pair mechanism for the growth of a conjugated polyene sequence during PVC thermolysis
. (Reprinted with
permission from Scheme 4, p. 117 of Ref.
, published by Harwood Academic Publishers.)
W.H. Starnes Jr. / Prog. Polym. Sci. 27 (2002) 2133–2170
2156
authors’ failure to consider stabilization by either ion pairing or solvation.
3
The same objection applies a
fortiori to the contention of Bacaloglu et al.
that the ion-pair mechanism does not account for HCl
catalysis because the formation enthalpy of the HCl
2
2
anion (said to be only 3 – 4 kcal/mol
) fails to
compensate for a supposed activation enthalpy for C – Cl ionization of 140 – 180 kcal/mol. The reaction
of HCl with Cl
2
actually is exothermic by ca. 14 kcal/mol
. Moreover, the catalysis conceivably
involves the alternative process shown in Eq. (25).
The authors of
cite publications in which bimolecular kinetics were found for reactions of
nucleophiles with various aliphatic chlorides, and they apparently assume that such kinetics rule out the
possibility of substrate ionization. Yet that is by no means the case. If substrates equilibrate with ion
pairs that then react with nucleophiles, the kinetics are cleanly bimolecular unless the concentration of
the nucleophile is so high that it upsets the equilibrium
. One of many examples of this situation
was provided by Sneen and Carter
, who found bimolecular kinetics for reactions of allylic
chlorides with phenoxide ion in 60, 40, or 25% aqueous dioxane (40, 60, and 75% water, respectively).
In all of these solvents, the reactions proceeded through ion pairs
. By applying the Grunwald –
Winstein equation to the kinetic data for 1-chloro-2-butene
, BF deduced that an analogous PVC
chain end could not react by an ion-pair route. This analogy is flawed, however, because the reaction of
the chlorobutene was not a unimolecular process. Moreover, the argument actually is irrelevant, as the
1-chloro-2-butenyl end group is not a significant contributor to the thermal instability of PVC
.
Increases in solvent polarity are well known to accelerate the thermal dehydrochlorination of the
polymer, but the rate enhancements are less than those expected for a reaction involving free ions
.
As was noted by BF, such a solvent effect is more in keeping with the intervention of a polar transition
state
. This could be either the one that leads to ion-pair 8 or the four-center transition state in
Eq. (23). Some unusually large effects of solvents have been reported by Minsker, who observed a rate
increase of ca. 40-fold upon going from “n-dichlorobenzene” (sic; presumably m-dichlorobenzene) to
tributyl phosphate at 150 8C
In
, the conversions of triphenylmethyl chloride and tropenyl chloride into ion pairs in
chlorinated solvents were considered as possible models for the formation of ion pairs from PVC. This
approach was criticized by BF, who neglected to mention that
actually rejected the use of
triphenylmethyl chloride because of a lack of some relevant thermodynamic information. In the case of
tropenyl chloride, comparisons with calculated cation energies suggested that ionization to form ion
pairs might occur in PVC with allyl chloride structures containing as few as one or two double bonds
. It was pointed out, however, that the situation might be complicated by a special coulombic
stabilization of the tropenylium ion-pair arising from its presumably centrosymmetric geometry
.
This objection was reiterated by BF, but the actual effects of geometry on the relative stabilities of the
ion pairs still remain to be established and may be insignificant
.
cites several papers
containing experimental evidence for the presence of ion pairs in thermally degraded PVC.
A further objection of BF
to dehydrochlorination via ion pairs was the lack of evidence for the
presence of cyclopentadiene rings in the thermally degraded resin. It was argued that such structures
would have resulted from the very facile cyclization of conjugated polyenyl cations. Yet there is also no
3
states incorrectly that calculations in
showed that the cyclic cation is more stable than the open one by
23 – 25 kcal/mol. In fact, this was an energy difference favoring the open isomer that emerged from the use of basis sets that
were too small to be satisfactory. Calculations with larger basis sets did reveal, however, that the cyclic ion is the more stable
species by ca. 11 kcal/mol
.
W.H. Starnes Jr. / Prog. Polym. Sci. 27 (2002) 2133–2170
2157
conclusive evidence for the absence of cyclopentadienes from the polymer, owing to the experimental
difficulties involved in elucidating the microstructure of degraded PVC. Furthermore, it is important to
note that the rapid cyclization of polyenyl cations occurs in extremely polar media (e.g. 96% sulfuric
acid) that allow these species to exist primarily as free ions rather than in ion pairs
. Differences in
the reactivities of free and encumbered carbocations can be very large, and the reactivity tends to
decrease dramatically with encumbrance
. This effect is observed, in particular, for reactions that
lead to charge separation
, as would be exemplified by the intramolecular cyclization of the cation
in a polyenyl chloride ion-pair. Therefore, if cyclopentadiene moieties really do not exist in
dehydrochlorinated PVC, their absence may simply be due to the inability of polyenyl chloride ion pairs
to cyclize at a rate that competes successfully with proton transfer during E1 elimination.
Bacaloglu et al. have averred that the ion-pair mechanism ‘does not postulate any interruption
reactions of the degradation chain’
. This statement is astonishing in view of the numerous
discussions in the literature, by the present author and other ion-pair proponents, of various possible
mechanisms for the termination of polyene growth. These mechanisms include, inter alia, the
intramolecular cyclization of polyene segments to form cyclohexadiene moieties
,
intermolecular Diels – Alder cycloaddition
, Friedel – Crafts alkylation
, HCl readdition
, and the inhibition of proton abstraction by the chloride anion when the polyenyl cation attains a
certain length
. In order to account for a polyene length distribution that was reported by BF, the
termination/propagation rate ratio must be the same for every polyene, regardless of its size
.
Termination via the intramolecular production of cyclohexadienes may be able to satisfy this
requirement for all of the growing polyenes except the ones that are too short to cyclize. However, firm
conclusions about the specific mechanism(s) for polyene termination probably will have to await the
definitive elucidation of the microstructure of thermolyzed PVC.
Iva´n et al.
have presented another argument against the ion-pair/quasi-ionic mechanism for
PVC dehydrochlorination. In experiments performed at 200 8C, they observed a rate constant for chain
propagation (polyene growth) that was larger by ‘orders of magnitude’ (the actual value was not
reported) than that for the initial dehydrochlorination of internal allylic chloride units
. This
difference was said to rule out propagation by stepwise dehydrochlorination at chloroallyl polyene ends.
Instead, the propagation was stated to occur via ‘an extraordinarily rapid zipping process’
in which
chloroallyl moieties were not involved at all. However, in the condensed phase at elevated temperatures,
chloroallylic compounds with two conjugated double bonds are well known to dehydrochlorinate much
more rapidly than simple monoenyl allylic chlorides. For example, published Arrhenius parameters
show that at 200 8C, the dehydrochlorinations of 6-chloro-2,4-octadiene and 7-chloro-3,5-nonadiene are
some 200 – 300 times faster than that of 8-chloro-6-tridecene
. Thus, instead of excluding the
stepwise growth of polyenes at chloroallyl sequence ends, the rate difference found by Iva´n et al.
actually strongly supports it. Importantly, such a mechanism also seems to be the only one that allows a
full explanation of various kinetic phenomena, reported by Iva´n and co-workers, that occur during the
thermal stabilization of PVC by various metal stearates
A mathematical modeling study by Troitskii and Troitskaya
provides additional evidence for the
much more rapid dehydrochlorination of conjugated dienyl allylic chlorides as compared to those with
one double bond. This kinetic difference is quite in keeping with expectations based on the ion-pair/
quasi-ionic mechanism, as the pentadienyl cation is stabilized by an energy of conjugation which is
much greater than that of the allyl cation
.
W.H. Starnes Jr. / Prog. Polym. Sci. 27 (2002) 2133–2170
2158
2.2.4. Six-center concerted mechanisms
BF have proposed two six-center concerted schemes for the thermal dehydrochlorination of PVC. The
first of these
, shown in
, was a revised and extended version of a
mechanism suggested by Amer and Shapiro
. According to BF, the all-trans polyene structure 24 is
thermally stable at the temperatures where the polymer loses HCl. Thus, as such, the chloroallyl end of
24
does not dehydrochlorinate at all (i.e. the direct conversion of 24 into 27
0
does not occur). Instead,
under the influence of HCl catalysis, 24 experiences allylic rearrangement into two isomers (25 and 26)
that contain an isolated double bond. The all-trans structure, 26, also is stable to HCl loss. However, structure
25
, which contains an isolated cis double bond, dehydrochlorinates very rapidly in a concerted six-center
manner (see arrows) in order to form the fully conjugated all-trans arrangement, 27. That structure then
is isomerized catalytically into homologues of 25 and 26, and the process continues as before.
This mechanism predicts a continual increase, during degradation, in the total concentration of the
structures that incorporate an isolated trans double bond (i.e. 26 moieties with various values of n ).
Proton NMR evidence for such an increase was reported by BF, who used the intensity of the vinyl
proton absorption at 5.65 – 6 ppm as the measure of concentration
. However, there are good
reasons for believing that vinyl protons at the ends of conjugated PVC polyenes may also absorb in this
range. That possibility is suggested, for example, by the spectra of several tert-butyl-capped linear
polyenes
and the downfield shifts that should be caused by sec-chloro substituents, as deduced from
comparisons of the spectra of linear monochloro alkenes
and linear alkene hydrocarbons
.
Fig. 6. Six-center concerted mechanism proposed by BF for the thermal dehydrochlorination of PVC (adapted with permission
from
. Copyright 1996 American Chemical Society).
W.H. Starnes Jr. / Prog. Polym. Sci. 27 (2002) 2133–2170
2159
The spectrum of trans,trans-2,4-hexadiene contains a multiplet at 5.5 – 5.6 ppm
that should be
shifted into the 5.65 – 6 ppm region by a chloro substituent on C-1. Moreover, absorption in this region
might also arise from 1,3-cyclohexadienes formed by polyene cyclization
, because the unsubstituted
cyclic diene itself absorbs at 5.75 – 5.9 ppm
. Finally, new isolated alkene linkages may be
produced directly from ordinary monomer units (Section 2.2.2.2), rather than by the isomerization of
conjugated polyene moieties. For all of these reasons, the observed increase in absorption at 5.65 – 6 ppm
cannot be regarded as evidence for the operation of the mechanism in
That mechanism also was believed to be supported by some activation enthalpy – entropy correlations
which were interpreted to mean that, in condensed media, cis allylic chlorides dehydrochlorinate via the
same six-center process that accounts for their relatively rapid degradation in the vapor phase
.
However, it was noted later that the large negative entropies of activation found for degradation in
liquids of low polarity
are quite consistent with the solvation of polar species involved in the ion-
pair/quasi-ionic mode of dehydrochlorination
. It was also pointed out that in the liquid phase, the
thermolysis rates of cis and trans allylic stereoisomers actually do not differ greatly (see the data for
compounds 10 – 13 in
. Furthermore, with the aid of kinetic and thermodynamic
information, it was possible to predict that at 170 8C, the conversion of 24 into 27, via 25, would be
slower by a factor of ca. 10
6
than the direct conversion of 24 into 27
0
. In other words, if the ion-
pair/quasi-ionic mechanism were operative, then the allyl chloride end of 24 would be quite reactive
under conditions where the homoallyl chloride terminus did not react at all.
Molecular orbital calculations by BF
and various thermodynamic considerations
suggested that the six-center mechanism would be most likely to operate when the value of n in 24 was 1
. Thus, as a further test of this mechanism, the dehydrochlorination rates of allyl chloride 10 and
homoallyl chloride 28 were compared at 170 8C
. As expected, the reaction of 10 was quite rapid.
However, 28 was perfectly stable and was estimated from the rate data to be less reactive than 10 by a
factor of 10
2
– 10
3
, at least. BF evidently considered these results to be conclusive evidence against the
scheme in
, but they then proposed another six-center concerted mechanism, which involves the
rearrangement of the allyl chloride end of 24 via a 1,3-chlorine shift
Like its predecessor, the new concerted mechanism has several fatal flaws. Its propagation steps are
shown in
, which does not depict the supposedly stable structures containing an isolated trans
double bond. Although these were not identified explicitly by BF, they presumably were considered to
be the all-trans isomers of 30 having various values of n. Their supposed stability toward the steady-state
concentration of HCl contrasts vividly with the postulated lability of 29 and 31, whose stable conjugated
systems should actually retard their isomerization. This paradoxical situation was, apparently, neither
recognized nor resolved.
In an attempt to obtain supporting evidence for the mechanism of
, BF studied the thermal
dehydrochlorination of a mixture of chloroallylic trans-tetradecenes at 150 8C
. They observed the
gradual buildup and decay of the cis isomers of the starting materials and concluded that these were
steady-state intermediates through which all of the dehydrochlorination proceeded via the route in
.
The starting trans isomers supposedly were quite stable toward the elimination of HCl and were said to
W.H. Starnes Jr. / Prog. Polym. Sci. 27 (2002) 2133–2170
2160
serve only as precursors for the reactive cis materials. Remarkably, the published rate data in
were not taken into account.
The cis-chlorotetradecenes reached their maximum concentration at 30 min, which was the time
required for a total conversion of some 35%
. Afterwards, the cis/trans ratio of the chlorides
remained at a constant level of about 1:5.4 until the degradation was complete
. Thus the cis
chlorides cannot be regarded as typical steady-state intermediates, which are present in extremely low
concentrations that are attained very quickly. The concentration – time profiles of the dehydrochlorina-
tion actually are best explained by a mechanism that BF considered but then discarded
. This
mechanism is shown in Eq. (32), where C and T represent the cis and trans allylic chlorides
alkeneðsÞ ˆ
k
T
T Y
K
C !
k
C
alkeneðsÞ
ð32Þ
(both of which were mixtures in the BF study
); the ks are rate constants; and K is the equilibrium
constant for cis – trans interconversion.
From the data of BF
, it follows that the equilibrium was fully established after 30 min and that
the value of K was 1/5.4, which corresponds to a free-energy difference of 1.4 kcal/mol between the (less
stable) cis and trans isomers. This difference agrees well with the stability difference of 1.2 kcal/mol (at
150 8C) found for cis- and trans-3-hexene
. According to Eq. (32), the dehydrochlorination rates of
T
and C are k
T
[T] and k
C
K[T], respectively, where the brackets represent concentrations. The total rate
is given by the sum of these two terms, and the ratio k
C
K/k
T
represents the amount of reaction occurring
from C relative to that from T. The value of this ratio can be estimated roughly by setting k
C
/k
T
equal to
3, which is the approximate reactivity ratio for 11:10 in o-dichlorobenzene at a somewhat higher
temperature (170 8C), as deduced from initial rates (
). This assumption gives a k
C
K/k
T
value of
0.56, which indicates that the amounts of dehydrochlorination occurring directly from T and C were 64
and 36%, respectively. Isomer T was the major source of HCl because of its relatively high
concentration.
Recently, additional kinetic data for model compounds
have provided further evidence against
the very rapid six-center concerted dehydrochlorination of linear cis allylic chlorides in the liquid phase.
The experiments were carried out at 170 8C in o-dichlorobenzene solutions that were bubbled with argon
Fig. 7. New six-center concerted mechanism proposed by BF
for the thermal dehydrochlorination of PVC.
W.H. Starnes Jr. / Prog. Polym. Sci. 27 (2002) 2133–2170
2161
at the highest possible rate in order to minimize catalysis by HCl (the argon flow rate used was 1.25 ml/s;
cf. the data in
, which were obtained with the same apparatus). Analysis by GC/MS showed that
under these conditions, isomerization of the chlorides was relatively slow. Hence the rate constants,
which were determined from the initial slopes of kinetic plots, pertained to mixtures in which the starting
isomer was always by far the major reacting component.
In these experiments, both cis- (32) and trans-1-chloro-2-nonene (33) yielded a k value of
6 £ 10
25
min
21
, in good agreement with the values for 12 and 13 in
, but only half the k value of
12 £ 10
25
min
21
found for both 3-chloro-1-nonene and 3-chloro-1-undecene
. This difference
shows that the 3-chloro-1-nonene could not have reacted via compound 32, which was the only possible
progenitor of a six-center concerted process. Moreover, the ks found for cis- and trans-6-chloro-4-
nonene differed by a factor of only 1.5 (the values were 890 £ 10
25
and 580 £ 10
25
min
21
,
respectively), and when the argon flow rate was reduced to 0.15 ml/s (thereby increasing the steady-state
concentration of HCl), the trans compound 33 actually became more reactive than the corresponding cis
isomer, 32 (with respective k values of 14 £ 10
25
and 8 £ 10
25
min
21
)
Perhaps the strongest evidence to be obtained thus far against the mechanism of
has emerged
from a detailed study of the thermal dehydrochlorination of allylic chloride 34
. Most of the
experiments with this compound were performed without solvent in order to facilitate the analysis of
reaction mixtures by GC/MS. Volatilization losses of reactant and products were prevented by the use of
a relatively low reaction temperature (140 8C) and argon flow rate (0.15 ml/s). A highly flexible
substance, 34 has no obvious steric constraints to retard or favor its dehydrochlorination by either the
ion-pair/quasi-ionic mechanism or the route in
The stability of an allylic cation is not increased appreciably by the attachment of an alkyl substituent
to the central carbon. In fact, under fully ionizing conditions (99.5% formic acid at 44.6 8C), the
solvolysis rate of methallyl chloride is actually less than that of allyl chloride itself, though by only a
factor of 2
. Therefore, if both 34 and 10 thermolyze in the ion-pair/quasi-ionic manner, their rates
of dehydrochlorination should not differ greatly. In keeping with this expectation, the k values found for
34
and 10 were 15 £ 10
25
and 24 £ 10
25
min
21
, respectively, at 140 8C
ð33Þ
Eq. (33) shows the dehydrochlorination products formed from the cyclic chloride at temperatures
ranging from 140 to 170 8C
. At 140 8C, the respective yields of dienes 35 (two stereoisomers)
and 36 were 70 and 30%. For steric reasons, 35 could not have been formed directly from 34 in a six-
center concerted process. The only conceivable route to 35 involving such a process has three steps: (1)
W.H. Starnes Jr. / Prog. Polym. Sci. 27 (2002) 2133–2170
2162
allylic rearrangement of 34 into chloride 37, (2) six-center dehydrochlorination of 37 to form diene 36,
and (3) HCl-promoted rearrangement of 36 into diene 35. This route to 35 was ruled out, however, by
experiments which showed that diene 36 was perfectly stable under the degradation conditions that were
used. The diene remained unchanged, in fact, at 140 8C in the presence of 5-chloro-5-methylnonane,
whose degradation at this temperature was faster than that of 34 and thus was supplying a higher
concentration of dissolved HCl
Since the six-center mechanism does not produce 35, there is no compelling reason to invoke it
for the slower process that gives 36. Formation of both 35 and 36 can be rationalized easily in
terms of ion-pair or quasi-ionic eliminations that are accompanied by the interconversion of 34, 37,
and the configurational stereoisomer of the latter substance. One of the possible routes to 35 is 1,4
elimination from 34 via an ion-pair process. Furthermore, if the six-center mechanism were a major
contributor to the dehydrochlorination of acyclic chlorides such as 10, then the simultaneous
operation of the two types of mechanism would cause the dehydrochlorination of such chlorides to
be much faster than that of 34. Yet, as discussed earlier, the reactivity of 10, relative to that of 34,
is not abnormally high and is quite in accord with the exclusive dehydrochlorination of both
chlorides by the ion-pair/quasi-ionic route.
The six-center dehydrochlorination of cis allylic chlorides may occur in the vapor phase.
Nevertheless, in condensed media, it is not fast enough to compete successfully with the ion-
pair/quasi-ionic process, whose operation now is made possible by the solvation of polar species
. Six-center concerted dehydrochlorination does not occur in the thermal degradation of
PVC.
Fisch, Bacaloglu et al. have argued that thermal dehydrochlorination occurs primarily at the surfaces
of PVC primary particles because molecules at these surfaces have a much higher conformational
mobility than those in the particle interiors
. This high mobility supposedly favors
formation of the cis chloroallyl moieties that are needed for polyene growth
. Optimum
stabilization by substances that deactivate allylic chlorides therefore requires the association of such
stabilizers with primary particle surfaces
. All of these assertions are, however, doubtful. cis
Chloroallyl structures actually are not key intermediates, and their preferred formation at surfaces is
highly unlikely at the temperatures where PVC degrades, as these are well above the glass transition
temperature (T
g
) of the resin. There is no conclusive evidence to show that the interior and surface T
g
s
Fig. 8. Polaron mechanism for the growth of a conjugated polyene sequence during PVC thermolysis (adapted with permission
from Scheme 5, p. 119 of Ref.
, published by Harwood Academic Publishers.)
W.H. Starnes Jr. / Prog. Polym. Sci. 27 (2002) 2133–2170
2163
are appreciably different. Hence, above T
g
, the cis structures should be formed with essentially equal
probability in all amorphous parts of the polymer. Moreover, preferential degradation at surfaces is
inconsistent with the results of published studies (inter alia
) which have shown that under
equilibrium conditions, the rate of dehydrochlorination actually decreases when the surface/mass ratio is
raised. The ostensible reason for this behavior is that increases in surface area facilitate the loss of HCl
by diffusion and thus reduce the concentration of this catalyst inside polymer specimens.
Any stabilization technology that requires the operation of the new six-center mechanism should be
regarded skeptically, and it certainly should not be considered seriously for use in commercial products.
2.2.5. Other proposed mechanisms
An ion-radical (polaron) chain mechanism for the thermal dehydrochlorination of PVC was recently
reviewed in depth
shows the initiation and propagation steps of this mechanism when
the propagating species is a growing cation radical. The overall mechanism involves several suggested
phenomena that are either unsupported by experiment or are at odds with well-known chemical facts.
Some of these phenomena are (a) deactivation of chlorine loss from the – CHClC
z
H – moiety by the
radical that is present, (b) facile direct loss of HCl from – CHClCH
2
C
z
H – or – CHClCH
2
C
þ
H – species,
(c) oxidation of polyenes or polyenyl free radicals by HCl, (d) oxidation or reduction of polyenes by
organic solvents, and (e) thermal stabilization by the replacement of an allylic methylene hydrogen with
an organotin functionality that is attached to the polymer by a stable C – Sn bond.
The loss of HCl from PVC polarons may indeed occur. However, there is no compelling
evidence to show that it is faster than the dehydrochlorination of neutral polyenyl chlorides,
whose concentration during degradation seems likely to be much greater than that of polaron
species. Thus a polaron mechanism for dehydrochlorination would appear to be only a very minor
process, at most. Nevertheless, the possible presence of polarons under some conditions is of interest in
connection with the cation diradical mechanism for degradation chain transfer that was described in
Section 2.2.2.2.
Another mechanism proposed for the thermal dehydrochlorination of PVC is a ‘one-step unzipping’
process
. According to this scheme, the excess thermal energy remaining after ejection of a
single molecule of HCl (DH
activation
2 DH
reaction
) stays in the polymer molecule and causes the loss of
more HCls in a single step. For a free-radical elimination, about 7.5 molecules of HCl are released
simultaneously, but this number increases to 22 when an ionic process occurs
. The original
excess energy is, however, very unlikely to remain localized in a segment of a single PVC molecule, as
has been pointed out previously
2.3. Degradation of polymer blends
Owing to the unique and useful properties that they possess, many blends of PVC with other polymers
are of considerable interest technologically. Yet the thermal degradation of PVC in such mixtures is not
well understood. There are reasons for believing, however, that this degradation may be influenced by a
number of physical and chemical factors that are specific to blended systems. These include such things
as miscibility of the blends, polarity of the medium, PVC aggregation, HCl solubility, HCl diffusion rate,
interaction of HCl with polar groups, and interaction of PVC with intermediates or products that are
derived from the degradation of the other polymer. Even the techniques used to create blends may play a
W.H. Starnes Jr. / Prog. Polym. Sci. 27 (2002) 2133–2170
2164
significant role by influencing homogeneity, morphology, and the nature and extent of PVC aggregation
.
The recent literature on the thermal stability of PVC blends is too extensive to allow its
comprehensive coverage in this review. What will be done here, instead, is simply to mention a few
examples that illustrate some of the various kinds of effects on stability that have been observed in
these systems. Brief surveys of some of the research in this general area are available elsewhere
.
Polymers containing nitrile groups (e.g. nitrile rubber
and polyacrylonitrile
generally destabilize PVC. This effect has been attributed to the retarded diffusion of HCl, owing to the
interaction of this catalyst with the nitrile substituents
. Such an interaction would produce imine
hydrochloride moieties, which are likely to be weaker catalysts than HCl itself. However, their
concentration may increase to a relatively high level when sufficient nitrile groups are present originally.
Interestingly, when it was introduced at the level of only 10 wt%, hydrogenated nitrile rubber slightly
reduced the rate of dehydrochlorination at 180 8C
. Perhaps in this case, the ability of the nitrile
groups to scavenge HCl was more important than catalysis by imine hydrochlorides, because the
maximum achievable concentration of the latter species was low. The literature contains at least one
report of a thermal stabilizer for PVC that may function in part via HCl scavenging by a nitrile function
.
In homogeneous solid binary blends, poly(methyl methacrylate) (PMMA) destabilizes PVC, but the
opposite effect is observed when the blends are heterogeneous
. Moreover, in solution, the
behavior of such blends is even more complex. If a rather poor and weakly basic solvent for PVC, 1,2,3-
trichloropropane, is used, PMMA increases the rate of HCl loss monotonically when it is introduced in
all proportions
. In contrast, when the good and relatively basic solvent, cyclohexanone, or
the poor but relatively basic solvent, benzyl alcohol, is used, low concentrations of PMMA decrease the
rate (to an extent that is very striking in the case of benzyl alcohol), while increasingly higher
concentrations then cause monotonic rate enhancements that are quite significant
. In order to
account for these observations, the researchers considered factors such as PVC aggregation and kinetic
effects of MMA monomer that was formed by degradation
. Aggregation does indeed
seem likely to be important, as it has been shown to have major effects on the stability of PVC in
solutions that contain no other polymers
The degradation occurring in blends that incorporate polyolefins (polyethylene, polypropylene,
polyisobutylene) is especially interesting. Because these polymers do not possess reactive
functional groups, their effects on PVC stability probably can be ascribed entirely to phenomena
that are physical rather than chemical. Owing to their incompatibility with PVC, they have no
effect on its thermal dehydrochlorination in solid blends that are made without good mixing
. On the other hand, in solid blends that are prepared by ‘elastic-deformation grinding’ (a
technique that causes more intimate mixing), the addition of polyethylene accelerates the formation
of HCl
. This effect was ascribed to an interphase interaction of the two components
, but its precise nature is obscure. Could it be connected with a decrease in the rate of
diffusion of HCl?
In solution, the results produced by polyolefins are similar to those that are caused by PMMA. In
weakly basic solvents where dehydrochlorination is slow, polyolefins strongly increase the rate
. However, the opposite behavior is observed in solvents with significantly higher
basicities, where dehydrochlorination is originally faster
. In both types of solvents,
W.H. Starnes Jr. / Prog. Polym. Sci. 27 (2002) 2133–2170
2165
these kinetic effects have been linked to PVC association and segmental dissolution that are caused by
the polyolefins, and this kind of rationale seems to be consistent with the inability of polyolefins to
influence the rate at all in situations where phase separation has been forced to occur by decreasing the
temperature or increasing the PVC concentration
. Even so, a detailed mechanistic
explanation of the connection between the physical phenomena and the kinetics has not been presented
thus far.
In other recent work, IR and Raman spectroscopies, as well as differential scanning calorimetry, were
used to investigate the structures and stabilities of blends of PVC with poly(N-vinyl-2-pyrrolidone)
(PVP), which are miscible in all proportions
. Thermal dehydrochlorination of the PVC
component was accelerated by PVP in a process that was thought to involve two stages
: (a) E1 or
E2 elimination of HCl (with the amido group of PVP serving as a base), followed by (b) the HCl-
promoted formation of polyenes and more HCl, via an ion-pair process.
Compatible blends of PVC with some chlorinated polyurethanes have been studied as well. Their
thermolysis involved the simultaneous destruction of both polymers at 200 – 300 8C, but in some cases
the thermal stability was enhanced by a mechanism that was suggested to involve the specific interaction
of carbonyl moieties with certain parts of the PVC structure
The recent literature also describes some new polymers that act as thermal costabilizers when they are
blended with PVC. These materials are methacrylate-based copolymers in which a number of epoxy or
hydroxyl groups are present
. The epoxy groups presumably scavenge HCl and serve as agents for
the displacement of labile halogen, while the polyhydroxy resins are thought to function as complexing
agents for metal-containing prodegradant species
3. Concluding remarks
Many of the fundamental aspects of PVC thermolysis now are well understood. The microstructures
and concentrations of the thermally labile structural defects have been determined, and their
mechanisms of formation have been firmly established. Polyene growth has been shown to occur by an
ion-pair or quasi-ionic route. Nevertheless, the degradation of PVC does involve free radicals, which
now seem likely to intervene in the transfer process that generates new labile defects and leads to
autoacceleration of the loss of HCl. Cyclization reactions probably play a major role in arresting polyene
propagation. Yet the detailed mechanism of polyene termination remains elusive and is an important
area for future study. Other major mechanistic conundrums that remain to be resolved are the complete
details of degradative chain transfer to polymer and the chemistry that is responsible for the remarkable
effects on PVC stability that occur in polymer blends. Basic research on all of these problems would
appear to be worth pursuing.
Acknowledgement
Generous support for the preparation of this review was provided by the National Science Foundation
under Grant No. DMR-9610361.
W.H. Starnes Jr. / Prog. Polym. Sci. 27 (2002) 2133–2170
2166
References
[1] Iva´n B, Kelen T, Tu¨do˜s F. In: Jellinek HHG, Kachi H, editors. Degradation and stabilization of polymers, vol. 2. New
York: Elsevier; 1989. chapter 8.
[2] Starnes Jr. WH, Girois S. Polym Yearbook 1995;12:105.
[3] Starnes Jr. WH. In: Salamone JC, editor. Polymeric materials encyclopedia, vol. 9. New York: CRC Press; 1996. p.
7042 – 8.
[4] Yassin AA, Sabaa MW. J Macromol Sci, Rev Macromol Chem Phys 1990;30:491.
[5] Minsker KS. Polym Yearbook 1994;11:229.
[6] Minsker KS. Int J Polym Mater 1994;24:235.
[7] Iva´n B. Adv Chem Ser 1996;249:19.
[8] Milla´n J, Martı´nez G. Rev Plast Mod 1997;73:570.
[9] Naqvi MK. Handbook Engng Polym Mater 1997;317.
[10] Zaikov GE, Gumargalieva KZ, Pokholok TV, Moiseev YV. Oxid Commun 1997;20:323.
[11] Troitskii BB, Troitskaya LS. Int J Polym Mater 1998;41:285.
[12] Troitskii BB, Troitskaya LS. Polym Yearbook 1999;16:237.
[13] Zaikov GE, Gumargalieva KZ, Pokholok TV, Moiseev YV, Zaikov VG. Polym Plast Technol Engng 2000;39:567.
[14] Bacaloglu R, Fisch MH, Kaufhold J, Sander HJ. In: Zweifel H, editor. Plastics additives handbook, 5th ed. Cincinnati:
Hanser Gardner; 2001. chapter 3.
[15] Martı´nez G, Go´mez-Elvira JM, Milla´n J. Polym Degrad Stab 1993;40:1.
[16] Milla´n JL, Martı´nez G, Go´mez-Elvira JM, Guarrotxena N, Tiemblo P. Polymer 1996;37:219.
[17] Luka´sˇ R, Pra´dova´ O. Makromol Chem 1986;187:2111.
[18] Luka´sˇ R. Makromol Chem, Macromol Symp 1989;29:21.
[19] Starnes Jr. WH, Schilling FC, Plitz IM, Cais RE, Freed DJ, Hartless RL, Bovey FA. Macromolecules 1983;16:790.
[20] Starnes Jr. WH, Wojciechowski BJ, Velazquez A, Benedikt GM. Macromolecules 1992;25:3638. Erratum 1992;
25:7080.
[21] Hjertberg T, So¨rvik EM. ACS Symp Ser 1985;280:259.
[22] Starnes Jr. WH, Wojciechowski BJ. Makromol Chem, Macromol Symp 1993;70/71:1.
[23] Xie TY, Hamielec AE, Wood PE, Woods DR. Polymer 1991;32:1098.
[24] Starnes Jr. WH, Chung H, Wojciechowski BJ, Skillicorn DE, Benedikt GM. Polym Prepr (Am Chem Soc, Div Polym
Chem) 1993;34(2):114.
[25] Starnes Jr. WH, Chung H, Wojciechowski BJ, Skillicorn DE, Benedikt GM. Adv Chem Ser 1996;249:3.
[26] Starnes Jr. WH, Zaikov VG, Chung HT, Wojciechowski BJ, Tran HV, Saylor K, Benedikt GM. Macromolecules 1998;
31:1508.
[27] Hjertberg T, So¨rvik EM. Polymer 1983;24:673.
[28] Carenza M, Palma G, Tavan M. J Polym Sci, Polym Symp 1973;42:1031.
[29] Starnes Jr. WH, Zaikov VG. Unpublished results.
[30] Starnes Jr. WH, Chung H, Pike RD, Wojciechowski BJ, Zaikov VG, Benedikt GM, Goodall BL, Rhodes LF. Polym Prepr
(Am Chem Soc, Div Polym Chem) 1995;36(2):404.
[31] Pi Z, Kennedy JP. J Polym Sci, Part A: Polym Chem 2001;39:307.
[32] Cozens RJ. Private communication of information supplied by The Geon Company.
[33] Starnes Jr. WH, Wojciechowski BJ, Chung H, Benedikt GM, Park GS, Saremi AH. Macromolecules 1995;28:945.
[34] Mayeda S, Tanimoto N, Niwa H, Nagata M. J Anal Appl Pyrolysis 1995;33:243.
[35] Rogestedt M, Hjertberg T. Macromolecules 1993;26:60.
[36] Xie TY, Hamielec AE, Rogestedt M, Hjertberg T. Polymer 1994;35:1526.
[37] Darricades-Llauro MF, Michel A, Guyot A, Waton H, Pe´tiaud R, Pham QT. J Macromol Sci, Chem 1986;23:221.
[38] Benedikt GM, Goodall BL, Rhodes LF, Kemball AC. Macromol Symp 1994;86:65.
[39] Benedikt GM, Cozens RJ, Goodall BL, Rhodes LF, Bell MN, Kemball AC, Starnes Jr. WH. Macromolecules 1997;30:
10.
[40] Talamini G, Visentini A, Kerr J. Polymer 1998;39:1879.
[41] Xie TY, Hamielec AE, Wood PE, Woods DR. J Vinyl Technol 1991;13:2.
W.H. Starnes Jr. / Prog. Polym. Sci. 27 (2002) 2133–2170
2167
[42] Dawkins JV, Moody CJ, Price D, Castle L, Howarth OW. Macromolecules 1995;28:2985.
[43] Starnes Jr. WH. Dev Polym Degrad 1981;3:135.
[44] Starnes Jr. WH, Schilling FC, Plitz IM, Cais RE, Bovey FA. Polym Bull (Berlin) 1981;4:555.
[45] Mao S, Ohtani H, Tsuge S, Niwa H, Nagata M. Polym J (Tokyo) 1999;31:79.
[46] Starnes Jr. WH, Villacorta GM, Schilling FC, Plitz IM, Park GS, Saremi AH. Macromolecules 1985;18:1780.
[47] Hjertberg T, Wendel A. Polymer 1982;23:1641.
[48] Rao PVC, Kaushik VK. Polym Test 1999;18:429.
[49] Maddams WF. In: Owen ED, editor. Degradation and stabilisation of PVC. New York: Elsevier; 1984. chapter 4.
[50] Boughdady NM, Chynoweth KR, Hewitt DG. Aust J Chem 1991;44:567.
[51] Mayer Z. J Macromol Sci, Rev Macromol Chem 1974;10:263.
[52] Mayer Z, Obereigner B, Lı´m D. J Polym Sci, Part C: Polym Symp 1971;33:289.
[53] Starnes Jr. WH, Plitz IM, Hische DC, Freed DJ, Schilling FC, Schilling ML. Macromolecules 1978;11:373.
[54] Starnes Jr. WH. Polym Mater Sci Engng 1988;58:220.
[55] Boughdady NM, Chynoweth KR, Hewitt DG. Aust J Chem 1991;44:581.
[56] Troitskii BB, Troitskaya LS. J Polym Sci, Part A: Polym Chem 1990;28:2695.
[57] Troitskii BB, Troitskaya LS. J Polym Sci, Part A: Polym Chem 1993;31:75.
[58] Troitskii BB, Troitskaya LS. Eur Polym J 1997;33:1289.
[59] Troitskii BB, Troitskaya LS, Kurskii YA, Kronman AG, Yakhnov AS, Novikova MA, Ganyukhina TG, Denisova VN.
Eur Polym J 1997;33:1179.
[60] Haynie SL, Villacorta GM, Plitz IM, Starnes Jr. WH. Polym Prepr (Am Chem Soc, Div Polym Chem) 1983;24(2):3.
[61] Minsker KS, Kolesov SV, Zaikov GE. Degradation and stabilization of vinyl chloride-based polymers. New York:
Pergamon Press; 1988.
[62] Panek MG, Villacorta GM, Starnes Jr. WH, Plitz IM. Macromolecules 1985;18:1040.
[63] Wojciechowski BJ, Starnes Jr WH, Benedikt GM. Abstracts, 43rd Southeast Regional Meeting of the American
Chemical Society, Richmond, Virginia, November 1991. Abstract No. 375.
[64] Chatterjee N, Basu S, Palit SK, Maiti MM. J Polym Sci, Part A: Polym Chem 1994;32:1225.
[65] Starnes Jr. WH, Plitz IM. Unpublished work.
[66] Flory PJ, Pickles CJ. J Chem Soc, Faraday Trans 2 1973;69:632.
[67] Boyd RH, Kesner L. J Polym Sci, Polym Phys Ed 1981;19:375.
[68] Smith GD, Ludovice PJ, Jaffe RL, Yoon DY. J Phys Chem 1995;99:164.
[69] Raghavachari K, Haddon RC, Starnes Jr. WH. J Am Chem Soc 1982;104:5054.
[70] Milla´n J-L, Martı´nez G, Mijangos C, Go´mez-Elvira JM. Makromol Chem, Macromol Symp 1989;29:185.
[71] Milla´n J, Martı´nez G, Jimeno ML, Tiemblo P, Mijangos C, Go´mez-Elvira JM. Makromol Chem, Macromol Symp 1991;
48/49:403.
[72] Behnisch J, Zimmermann H, Anders H. Polym Degrad Stab 1985;13:113.
[73] Hjertberg T, Martinsson E, So¨rvik E. Macromolecules 1988;21:603.
[74] Rogestedt M, Hjertberg T. Macromolecules 1992;25:6332.
[75] Radiotis T, Brown GR. J Macromol Sci, Pure Appl Chem 1997;34:743.
[76] Kurskii YA. Polym Sci USSR 1997;39:471.
[77] Starnes Jr. WH, Edelson D. Macromolecules 1979;12:797.
[78] Behnisch J, Zimmermann H. Int J Polym Mater 1992;16:143.
[79] Cuthbertson MJ, Bowley HJ, Gerrard DL, Maddams WF, Shapiro JS. Makromol Chem 1987;188:2801.
[80] Scott G. Polym Degrad Stab 1995;48:315.
[81] Troitskii BB, Troitskaya LS. Eur Polym J 1999;35:2215.
[82] Doering W von E, Sarma K. J Am Chem Soc 1992;114:6037.
[83] Knoll K, Schrock RR. J Am Chem Soc 1989;111:7989.
[84] Montaudo G, Puglisi C. Polym Degrad Stab 1991;33:229.
[85] Chien JCW, Uden PC, Fan J-L. J Polym Sci, Polym Chem Ed 1982;20:2159.
[86] Lide DR, editor. CRC handbook of chemistry and physics, 82nd ed. Boca Raton, FL: CRC Press; 2001. p. 5 – 35. see also
p. 9 – 70.
[87] Arlman EJ. J Polym Sci 1954;12:547.
W.H. Starnes Jr. / Prog. Polym. Sci. 27 (2002) 2133–2170
2168
[88] Starnes Jr. WH. In preparation.
[89] Winkler DE. J Polym Sci 1959;35:3.
[90] Stromberg RR, Straus S, Achhammer BG. J Polym Sci 1959;35:355.
[91] Skell PS, Baxter III HN. J Am Chem Soc 1985;107:2823.
[92] Martinsson E, Hjertberg T, So¨rvik E. Macromolecules 1988;21:136.
[93] Troitskii BB, Troitskaya LS. Eur Polym J 1995;31:533.
[94] Troitskii BB, Troitskaya LS, Yakhnov AS, Lopatin MA, Novikova MA. Eur Polym J 1997;33:1587.
[95] Morton JR, Preston KF, Krusic PJ, Hill SA, Wasserman E. J Am Chem Soc 1992;114:5454.
[96] Troitskii BB, Troitskaya LS, Yakhnov AS, Novikova MA, Denisova VN, Cherkasov VK, Bubnov MP. Polym Degrad
Stab 1997;58:83.
[97] Starnes Jr. WH, Ge X. In preparation.
[98] Tran VH, Guyot A, Nguyen TP, Molinie´ P. Polym Degrad Stab 1992;37:209.
[99] Hoang TV, Guyot A. Polym Degrad Stab 1988;21:165.
[100] Yu Q, Zhu S, Zhou W. J Polym Sci, Part A: Polym Chem 1998;36:851.
[101] Fisch MH, Bacaloglu R. J Vinyl Addit Technol 1995;1:233.
[102] Haddon RC, Starnes Jr. WH. Adv Chem Ser 1978;169:333.
[103] Meier RJ, Kip BJ. Proceedings of the 14th International Conference on Advances in the Stabilization and Degradation of
Polymers, Luzern, Switzerland; May 1992. p. 147.
[104] Bacaloglu R, Fisch M. Polym Degrad Stab 1995;47:33.
[105] Maccoll A. Chem Rev (Washington, DC) 1969;69:33.
[106] McDaniel DH, Vallee´ RE. Inorg Chem 1963;2:996.
[107] More O’Ferrall RA. In: Patai S, editor. The chemistry of the carbon – halogen bond, part 2. New York: Wiley; 1973.
chapter 9.
[108] Sneen RA, Carter JV. J Am Chem Soc 1972;94:6990.
[109] Sorensen TS. Carbonium Ions 1970;II:807.
[110] Keating JT, Skell PS. Carbonium Ions 1970;II:573.
[111] Bacaloglu R, Fisch M. Polym Degrad Stab 1994;45:301.
[112] Iva´n B, Kennedy JP, Kelen T, Tu¨do˜s F, Nagy TT, Turcsa´nyi B. J Polym Sci, Polym Chem Ed 1983;21:2177.
[113] Bacaloglu R, Fisch M. Polym Degrad Stab 1994;45:315.
[114] Bacaloglu R, Fisch M. Polym Degrad Stab 1994;45:325.
[115] Bacaloglu R, Fisch M. Polym Degrad Stab 1995;47:9.
[116] Bacaloglu R, Fisch MH. J Vinyl Addit Technol 1995;1:241.
[117] Starnes Jr. WH, Wallach JA, Yao H. Macromolecules 1996;29:7631.
[118] Amer AR, Shapiro JS. J Macromol Sci, Chem 1980;14:185.
[119] Pouchert CJ, Behnke J,, editors, edition I. The Aldrich library of
13
C and
1
H FT NMR spectra, vol. 1. Milwaukee, WI:
Aldrich Chemical Co; 1993.
[120] Starnes Jr. WH. Polym Prepr (Am Chem Soc, Div Polym Chem) 1996;37(1):697.
[121] Fisch MH, Bacaloglu R. J Vinyl Addit Technol 1999;5:205.
[122] Fisch MH, Bacaloglu R. Plast Rubber Compos 1999;28:119.
[123] Stull DR, Westrum Jr. EF, Sinke GC. The chemical thermodynamics of organic compounds. New York: Wiley; 1969. p.
319.
[124] Li Y, Starnes Jr. WH. Presented at the 4th National Graduate Research Polymer Conference, Hattiesburg, MS, June
2000. Abstract P2-9.
[125] Starnes Jr. WH, Zaikov VG, Payne LB, Li Y, Ge X. Polym Prepr (Am Chem Soc, Div Polym Chem) 2001;42(1):404.
[126] DeWolfe RH, Young WG. Chem Rev (Washington, DC) 1956;56:753.
[127] Fisch MH, Bacaloglu R, Biesiada K, Brecker LR. J Vinyl Addit Technol 1999;5:45.
[128] Luther H, Kru¨ger H. Kunststoffe 1966;56:74.
[129] Patel K, Velazquez A, Calderon HS, Brown GR. J Appl Polym Sci 1992;46:179.
[130] Tran VH. J Macromol Sci, Rev Macromol Chem Phys 1998;38:1.
[131] Meier RJ, Kip BJ. Polym Degrad Stab 1992;38:69.
[132] Braun D, Bo¨hringer B, Eidam N, Fischer M, Ko¨mmerling S. Angew Makromol Chem 1994;216:1.
W.H. Starnes Jr. / Prog. Polym. Sci. 27 (2002) 2133–2170
2169
[133] Kolesov SV, Kulish EI, Zaikov GE, Minsker KS. J Appl Polym Sci 1999;73:85.
[134] Kolesov SV, Kulish EI, Minsker KS. Polym Sci, Ser A 2000;42:213. Transl of Vysokomol Soedin, Ser A 2000;
42:306.
[135] Minsker KS. Polym Sci, Ser B 2000;42:44. Transl of Vysokomol Soedin, Ser B 2000;42:372.
[136] Minsker KS, Kolesov SV, Kulish EI, Zaikov GE. Polym Yearbook 1996;13:5.
[137] Abdel-Naby AS. J Vinyl Addit Technol 1999;5:159.
[138] Kolesov SV, Kulish EI, Minsker KS. Polym Sci, Ser B 1995;37:311. Transl of Vysokomol Soedin, Ser B 1995;37:1084.
[139] Kulish EI, Kolesov SV, Sigaeva NN, Chirko KS, Volodina VP, Minsker KS. Polym Sci, Ser A 1998;40:809. Transl of
Vysokomol Soedin, Ser A 1998;40:1304.
[140] Kulish EI, Kolesov SV, Minsker KS, Zaikov GE. Polym Sci, Ser A 1998;40:813. Transl of Vysokomol Soedin, Ser A
1998;40:1309.
[141] Arinshtein AE, Kulish EI, Kolesov SV, Minsker KS. Polym Sci, Ser A 2000;42:1138. Transl of Vysokomol Soedin, Ser
A 2000;42:1743.
[142] Kulish EI, Kolesov SV, Minsker KS, Zaikov GE. Int J Polym Mater 1994;24:123.
[143] Dong J, Fredericks PM, George GA. Polym Degrad Stab 1997;58:159.
[144] Kaynak A, Bartley JP, George GA. J Macromol Sci, Pure Appl Chem 2001;38:1033.
[145] Pielichowski K, Hamerton I. Eur Polym J 2000;36:171.
[146] Braun D, Belik P, Richter E. Angew Makromol Chem 1999;268:81.
W.H. Starnes Jr. / Prog. Polym. Sci. 27 (2002) 2133–2170
2170