Polymer Degradation and Stability 35 (1992) 77-86
Thermal oxidation of anhydride cured epoxies. 1--mechanistic aspects H. M. Le Huy, V. Bellenger, & J. Verdu ENSAM, 151 boulevard de l'Hrpital, 75 013 Paris, France
&
M. Paris EDF/DER, 1 avenue du Gal de Gaulle, 92 170 Clamart C~dex, France (Received 8 October 1990; accepted 25 October 1990)
The thermal ageing of various anhydride cured epoxies was studied in the temperature range 140-200°C, in vacuum and in air. Except for a slight post-crosslinking process, the main structural changes are due to oxidation. two steps can be distinguished. In the first, whose duration does not exceed 300h at 140°C, the monomeric anhydride is reformed presumably from monoacid dangling groups. When this anhydride is saturated (hexahydrophthalic anhydride), it is evolved. When it is unsaturated (methyltetrahydrophthalic anhydride), it is bound to the network by radical additions to the double bonds. This latter process is responsible for the increase in the glass transition temperature. At the end of this first step, the rates of weight loss and T~ change decrease strongly, and the ageing becomes dominated by the oxidation of the regular structural units of the polymer. This process leads essentially to the formation of anhydrides different to those formed in the first step.
INTRODUCTION
Bowen2 for instance will not be considered here. (b) At low temperature, below the glass transition temperature, Tg, the chemical processes are generally very slow and physical ageing I is the main cause of property changes. There have only been a few studies on the physical ageing of epoxies. 4,5 (c) In the intermediate range--typically 100200°C for the polymers under study-----oxidation by atmospheric oxygen generally plays the key role. Oxidation is attested by the build-up of IR bands typically representative of oxygenated structures. 6-8 It proceeds with weight loss, 9-1~ a decrease of the flexural and tensile strength, 9-H and a change of various physical properties including dielectric strength 9 or Tg. 10'12 As for the majority of industrial thermosets, 13'~4 post-crosslinking can predominate in the early period of exposure. 15 It is
Anhydride cured epoxies are widely used in electrical applications, for example in dry transformers, due to their good mechanical and electrical properties. Durability considerations are often crucial in such applications, which explains the interest of various research workers in ageing mechanisms. At least three temperature domains can be more or less clearly distinguished: (a) At very high temperatures---typically above 250°C for the moderately stable materials under study--it is generally recognized that purely thermolytic processes (whose rate depends essentially on the dissociation energy of covalent bonds3), play a predominant role. This case, which was studied by Polymer Degradation and Stability 0141-3910/91/$03-50 © 1991 Elsevier Science Publishers Ltd. 77
H. M. Le Huy et al.
78
very difficult to make a synthesis of the proposed mechanisms, because the resin formulation varies from one author to another, and structural irregularitie,~- mainly ether structures16.17---can be involved in the ageing process. Forster et al.12 proposed, from semi-empirical calculations of molecular orbitals, chain scission at the carbon-oxygen bonds. R--O---CH 2----CH - - C H2--~-O C - - R ' I I1 O O
ester link
O--CHE--CH--CH2-FO
CH/R'
?I
The ester group appears to be more thermostable than the ether group, which seems to be consistent with the calculations but neglects the role of the oxidation chain mechanism. Noskov and Alekseev, 7 suggest that chain scission will occur at the fl position of the alcohol group O
I
II
R--O--CH2--CH--CH2--O
02
R--O
O II + CH
C--R'
/
All the networks under study were prepared from the same diepoxide, diglycidyl ether of bisphenol A (DGEBA), having an epoxide index of 5.27molkg -x, from Ciba-Geigy. Two anhydrides, methyltetrahydrophthalic anhydride (MTHPA) and hexahydrophthalic anhydride (HHPA) were used. -
CO
ether link
\R"
OH
Materials
3HE
I
R
EXPERIMENTAL
O O ICI_._CH2__O___~__R,
HO In the case of epoxide-amine networks it was also suggested that the secondary alcohol could be an especially oxidizable group. TM It is noteworthy that hydroxyl groups, which play a key role in both the preceding studies, must in principle disappear during crosslinking. Since their concentration depends strongly on processing conditions, they must, in principle, strongly influence the polymer durability. Finally, detailed studies of the crosslinking process showed that, in the temperature region of interest, weight loss due to anhydride evaporation with a rate strongly dependent on the initial anhydride/epoxide (A/E) molar ratio, 16 must also be taken into consideration. The aim of the present work is to make a comparative study of the thermal oxidation of various epoxide/anhydride systems differing essentially in the hardener structure.
MTHPA
HHPA
In the case of MTHPA, flexibilized samples were used in addition to the quasi-stoichiometric D G E B A / M T H P A sample. Two flexibilisers were studied, namely PG which is a diacid terminated polypropylene glycol, and NPG which is a diacid terminated neopentyl glycol. The composition was expressed by two ratios namely, A/E, the anhydride/epoxy molar ratio of reactive functions and F/A, the flexibilizer/anhydride molar ratio. The systems under study are described in Table 1. They were moulded for 5 h at 80°C and post-cured for 10 h at 140°C. Samples of various thicknesses were prepared. In some samples A2, copper chloride (CuCI2) was incorporated by diffusion into acetone swelled films. Its concentration was about 1% by weight. Acetone was removed by evaporation under vacuum.
Agemg The samples were aged in air at various temperatures, in ventilated ovens regulated at +2°C and, in certain cases, in vacuum (~1 mm Hg).
Characterization The imaged and aged samples were studied by FTIR spectrophotometry (Perkin-Elmer 1710), thermal gravimetry and DSC (Perkin-Elmer DSC 2). The T~ was taken at the inflexion point of the DSC thermogram. In some cases, the rubbery modulus was determined using an Instron 1193 tensile tester with a temperature controlled chamber allowing measurements to be made at
Thermal oxidation of anhydride cured epoxies. I
79
Table 1. Characteristics o f the systems under study
Code
Anhydride
Flexibilizer
A/E
F/A
T~ (°C)
[OH] (mol kg-])a
A1 A2
MTHPA MTHPA MTHPA MTHPA MTHPA MTHPA MTHPA MTHPA HHPA
PG PG PG NPG NPG NPG NPG none none
0"56 0"71 0"90 0"92 0.73 1"06 0-81 0-96 1.00
0'25 0"25 0.25 0"20 0-20 0-09 0-09 0 0
48 63 62 97 105 113 120 130 136
2"2 2"0 1-7 2"2 2"6 1"5 1"8 0.9 1.1
A3
Bll B12 B21
B22 C D
a Determined from the IR absorbance at 3520 cm ~ using a molar absorptivity e
O H = 50 kg mol -~ cm -~.
(Tg+50K). UV-visible Lambda 5.
Yellowing was studied using a spectrophotometer, Perkin-Elmer
~-TIR spectrophotometry 1.0
"0.0
0.0 4000
RESULTS
2000 wave
1000 number
400
(cm")
The changes observed after exposure in vacuum at 200°C for almost 1000 h were insignificant. The changes which occurred during exposure in air are illustrated in Fig. 1 and can be summarized as follows: (a) A band located at 1780cm -1 increases quickly (in M T H P A systems) in the early period of exposure but its growth rate decreases to reach a low stationary value after 50-300h, depending on the temperature of exposure. Various chemical treatments lead to its disappearance and involve the following spectral changes:-(i) With ethanol and water, acidic bands appear at 3200, 1050 and 880 cm -1.
.2.2 1.0
0.0 4m
. . . . . . . .
2~oo . . . . . . . . . wave
number
IU . . . . .
4oo
(cm -~)
Fig. 1. IR spectra of the system A2 (a) before, (b) after ageing for 200 h
,
4000
,
r
i
,
,
,
,
,
,
,
,
,
,
,
j
,
.
2000 wave
.
.
.
.
1000 number
,
,
,
0,0
400
(cm-')
Fig. 1.----contd. (c) after ageing for 1500 h at 180°C in air.
80
H. M. Le Huy et al. A1780 / A1610
0
o
A1780 / AI610
~ooo
:o'oo
~o'oo
time (h)
Fig. 2. Effect of 1% CuCI2 on the anhydride build-up in sample A2 at 140"C (rq) A2, and (A) A2 + CuCI2.
(ii) With ammonia, an amide peak appears at 1690 cm -1. The species found are undoubtedly anhydrides. Their formation is appreciably accelerated by the presence of copper chloride (Fig. 2). (b) Bands located at 1850 and 900cm -1, presumably due to another type of anhydride, appear after a relatively long induction period, typically 150 h at 200°C, 500 h at 180°C, 1400 h at 160°C and 5000 h at 140°C for samples A2. (c) The concentration of alcohols (3500 cm -1) and acids (1680 and 3220cm -~) increases quickly in the early period of exposure and then increases or decreases slowly depending on the composition of the sample. (d) The bands associated with aliphatic CH groups (2880 and 1460cm-1), esters (1250, 1045cm -1) and ethers ( l l l 0 c m -~) decrease and disappear progressively. (e) In the case of systems containing MTHPA, the bands associated with double bonds (1655 and 662 cm -a) disappear progressively.
time (h) Fig. 4. Growth of the anhydride IR band for the samples (A) A2, ( ~ ) B12, (+) B22, (11) C, and (rq) D at 180°C in air.
Concerning the effect of sample composition, the following observations can be made: (f) The amplitude of spectrophotometric changes is, for PG systems, an increasing function of the molar ratio of anhydride (Fig. 3). (g) The nature and concentration of fiexibilizer seems to have little influence on the build-up rate of anhydride. In contrast, it depends strongly on the nature of the anhydride: it is considerably higher for MTHPA than for H H P A (Fig. 4).
Thermal gravimetry Weight loss was observed in all cases. The results can be summarized as follows: (a) Weight loss results essentially from oxidation. In the case of sample A2 for instance, it reaches a value of 0.6% in vacuum against 6% in air for 1500 h exposure at 200°C. (b) As for the build-up of anhydride, the rate of weight loss is high in the early period of
A3520 / AI610
weight loss (%)
0,5 5
t
~.'I~
o~
/o f . / . Lt-"-a" 0,0
''~ ''°
,_.... ~'-
...t~_ ~ . . . . -~--
---'-" "~""
" ~
.-x . . . . . .
] ~ ' time (h)
Fig. 3. Change of the hydroxyl absorbance at 3520cm -~ during exposure in air at 180°C for the samples (A) A1, (D) A2, and ( ~ ) A3.
0
I
loo
i
i
i
I
son
I
a
I
i
I
looo
I
i
J
I
I
~-
ls~ t(h)
Fig. 5. W e i g h t loss at ]80°C in air f o r the 2-mm t h i c k P G samples (A) A1, ( x ) A2, ( 0 ) A3.
Thermal oxidation of anhydride cured epoxies. I weight loss (%)
81
Tg (*C) 250 []
J
o
150"
a
•
.,"
1000
2000
o
A
50 0
Fig. 6. Weight loss at 180°C in air for the following 4-mm thick systems (k) A2, (rq) D, (~) B12, (+) B22, and (I) C. exposure, then decreases and reaches a low stationary value after 100-500 h, depending on the temperature. (c) The influence of the A / E molar ratio is illustrated by Fig. 5. The rate of weight loss is lowest for the sample of highest crosslink density (A2). (d) In the presence of CuCI2, the rate of weight loss is almost twice as high. (e) The rate of weight loss is noticeably higher for PG than for NPG samples, and for H H P A than for M T H P A samples (Fig. 6).
Glass transition temperature lead
......
•
3000
time (h)
DSC measurements observations:
[]
to
the
following
(a) A slight but measurable post-crosslinking effect can be observed after exposure at 200°C in vacuum. In the case of sample A2, for instance, Tg can increase from 63 to 73°C in the first 300h of exposure and then remains practically constant over several hundred hours. The rubbery modulus varies in a parallel way from 12.8 to 14.7 MPa. (b) For M T H P A systems, the Tg of films or samples taken from the surface zone of the exposed plaques, increases considerably more in air than in vacuum, reaching values close to 200°C for sample A2. (c) This increase of Tg is also accelerated in the presence of CuCI2. Tg reaches 150°C after about 500 h in the presence of CuCI2 compared with 3000 h in undoped samples.
1~0
2000
3~0
Fig. 7. Increase of the glass transition temperature at 180°C for samples (A) A2, (O) B12, (+) B22, (11) C, and (D) D. ( . . . . . ) Exposure in vacuum for A2, and (. . . . . ) exposure in vacuum for D. (d) The rate of increase of Tg in the early period of exposure is high for M T H P A systems and very low for the H H P A one (Fig. 7).
DISCUSSION Weight loss In the case of D G E B A - P A (phthalic anhydride cured system), weight loss results essentially from splitting-off anhydride from monoesters as confirmed by analysis of volatile products.16 This mechanism explains well the coincidence of the minimum rate of weight loss with the maximum crosslink density when the structural variable is the A / E molar ratio. 16 By analogy, it could be concluded that, in the cases under study, weight loss arises essentially from the same mechanism. The shape of the kinetic curves for weight loss is consistent with the fact that t h e precursors (monoesters) are in limited concentration. In the simplest case, the rate of weight loss would be proportional to their concentration, which would lead to a first order process. It will be shown in the second paper of this series ~9 that the first step (autoretardated) of weight loss is effectively a first order process. The fact that weight loss is faster for H H P A than for M T H P A can be explained by the participation of the double bond of the latter in oxidative crosslinking reactions as confirmed by the disappearance of the corresponding IR bands.
82
H. M. Le Huy et al.
O
II
,,,~O--C
O II C~O---H
, ,w,OH + O--C~
k___/
O II ,,~,O--C
O II ~,O----C
O II C--OH ~O' )
0
O II C---OH
0 H C--OH
'~O---(~ 02
'
~O--O" N~O
,,,~OH
(n)
J
+
,w,O~O---O,,~, This process explains well the following facts: (a) That the weight loss is lower for MTHPA than for HHPA since in the former case, the anhydride molecule is linked by covalent bonds to the network whereas it is free in the second case. (b) That the rate of anhydride build-up is higher for MTHPA than for HHPA since in the latter case, the anhydride formed is removed by evaporation. (c) That Tg increases considerably more in MTHPA than in HHPA systems as a result of the crosslinking process. This point will, however, require detailed discussion (see below). As expected from this hypothesis, the rate of weight loss is minimum for the A/E molar ratio 0.71, which leads to the maximum crosslink density, i.e. for the lowest monoester concentration. The only point which is in disagreement with the results above is the fact that weight loss is considerably accelerated in the presence of oxygen whereas, for DGEBA-PA, only small differences were observed between air and vacuum e x p o s u r e s . ~6 Direct and indirect pathways can be suggested for the effect of oxygen on this process.
Direct pathway (oxidation): 0
0
,,,,,,CH2__C~ ~
[[ C---O~H 02 ). many steps
O ,w,C H - - O - - ~
O II C--O H
o'.
~
(m)
/~ scission
0
II
,,~,C~Ho [[ + "O--C
O--C~C=O
0
H
C--O--H
? ? -.,o HO
C~C~OH
<
PH
Thermal oxidation of anhydride cured epoxies. I
In addition, crosslinking by intermolecular anhydridization cannot be completely excluded.
Indirect pathway (hydrolysis): oxidation
~ H20
0 II ~,O----C
83
0 II C~OH H~O
,w,O, C Q O
II
0
II C
,,~,OH + HO
0
II C--O H
--H20
I
O
OH
O
OH
~,O---C
~O
II
--H20)
I
or)
o=_C°c_o Although both hypotheses seem to be reasonable from a mechanistic viewpoint, it must be recognized that their consistency with the kinetic data remains to be established.
0
it
0
intermolecular anhydridation
I
Crosslinking As shown in the preceding section, crosslinking must first be associated with the initial presence of double bonds in the hardener moiety, which can explain well the difference between (unsaturated) MTHPA and (saturated) HHPA. Crosslinking by double bonds suppresses the monoacid dangling chain ends as shown by reaction (11). When it occurs on diester sections, it leads to a considerable increase in the crosslink density since, theoretically, four new crosslinks must be formed for anhydride moiety:
>o
O
O
O ~O---C
O C~O,,Z(
(v)
(w) However, such a process should be relatively improbable at low conversions owing to the low concentration of monoacids.
Anhydride formation Two types of anhydrides are formed. Attempts at extraction by normal solvents, for instance diethylether, which has a noticeable swelling effect on the network, showed clearly that both anhydrides are bound to the polymer. The first type, which appears in the early period of exposure, is presumably formed from monoesters initially present as proposed in scheme 11. The second type appears after a relatively large induction period. It seems reasonable to attribute it to radical oxidation of the polymer. In fact, ester containing molecules containing abstractable hydrogen atoms in the tr position must lead
84
H. M. Le Huy
to anhydrides: R\
Crosslinks appear at especially oxidation sensitive sites, because the corresponding hydrogen is carried by a tertiary carbon in the tr position to an oxygen atom, and thus combines two causes of high abstractability by radicals. 2° On the basis of standard oxidation chain mechanisms, 21 the following scheme may be proposed:
C--A--C--O--R' o~ , il II many.e~
CH--O
0
et al.
0
1 R--C---O C--A---C--O---R' if II fl O O O H
I
R', RH +
~,C--O C H 2 - - - ~ - - C H 2 - - G - - ~ II O O
(w! t)
~'C--O C H 2 - - C ' ~ C H 2 ~ II
O
i
I I
O C=O
~
0 2 (VU b)
OOH
I
( PH
'~'C---O---CH2---C.- - - C H 2 ~ O
(Via¢)
,~C~O
II
O
O I C--~O mono or
bimolecular~(Vl| d)
branching ~ " ~
I
O I C--~O (VII e)//~non terminating /d../bimolecu|ar combination
O"
,.,~--o--c
OO" CHE---C--CH
I H2--C--CH2--O--~t O I ~O
~,,,~ OH
~C~O II o
,~,~O--CH2-~ + "CH2--O-~ O O ,
CHE---C--CH2---O-o
~=o~/
(VIII b
III c) ,.,,.,
,w~C--O--C H 2 - - C - - - C H 2 - - ~ C-~O O 0
CH2---C--CH2--O-~ 0
0 O
~,~--O--C H2--C---C H 2 - - ~ [ ~ 0
o
,w,C--O CH; It 0
i
0 I ~0
+ "0 C '~' II 0
Thermal oxidation of anhydride cured epoxies. 1 In this scheme, the alkoxyl radicals (PO') would play a key role. The existence of an induction period for the second type of anhydrides suggests that they result essentially from bimolecular branching reactions which become efficient only above a critical hydroperoxide concentration. A redox catalyst such as copper chloride must in this case accelerate significantly the whole oxidation process 22 as experimentally observed. The alkoxyl radicals can be involved in various reaction pathways. They can abstract hydrogen to give alcohols (Villa). These alcohols could participate in further crosslinking processes by reaction with polymer bound anhydrides or simply by acid esterification. Their bimolecular combination (ViHb) seems to be the main possible method of chain termination, at least favored when they arise from a cage combination of peroxyls, z3 POO" + P O O ' - - 2 P O " + 02
R--CO~ - - R" + CO2
O--CH2
o~
O R--C--O II O
CH2---OOH ~ -"2°
The stability is expected to be lower for PG (propylene glycol) than for NPG (neopentyl glycol) for the same reasons as cited in the case of crosslink oxidation since PG also contains a tertiary carbon in the ¢r position to an ether group. The especially low stability of PG systems appears clearly in weight loss curves (Fig. 6). It is noteworthy that oxidation events which occur at the extremities of the polyether chain can also lead to anhydrides. H "~'A----C--O II O
I
0 2. PH
C CH2--O--> I manys,o~ CH3 O
R--C--O---C--H II li O O (xl)
It should be noted that these mechanisms are in principle independent of the anhydride structure which agrees with the fact that the 'stationary' rate of anhydride build-up is the same for M T H P A and H H P A (Fig. 4).
.
I ,~A----C----O---C--C H 2--O--II I O CH3
1 ~,A--C--O II O
C----CH3 + "CHz~, II O
Similarly
R ,~,A----C---O
II
O
I CH2---C-i
(PG OR NPG)
o2.PH , many steps
R'
(X)
The oxidation of---CH2 macroradicals resulting of process (Ville) would also lead presumably to an anhydride. R--C
Influence of flexibilizers
(IX)
It is noteworthy that they cannot participate in disproportionation, 2~ as is frequently proposed for hydrocarbon polymers. The resulting peroxide bridge is relatively unstable at high temperature, so that its thermolysis (Villc) is another pathway to the formation of PO" radicals. The other reaction pathways involving alkoxyls are presumably essentially fl scissions leading to anhydrides and primary --CH~ macroradicals (Ville) and (Villi), or ketones and --C---O" macroradicals ( v i l l d ) . These latter II O are able to abstract hydrogen atoms to give acids, or to undergo decarboxylation.
85
~,A----C
II
O
0
R I C - - H + "C-II I
O
R'
If a part of the flexibilizer is incorporated in dangling chains, its degradation must lead to an increase of the crosslink density (Fig. 8) which can play a significant role in the increase of Tg. Indeed, such chain scission events would increase the rate of weight loss since the formation of a volatile fragment needs, in principle, a single event, against two in a network subchain. Attempts to characterize this process by rubber elastic measurements were unsuccessful owing to the heterogeneous distribution of oxidation products in the sample thickness.~9
86
H. M. Le Huy et al.
Fig. 8. Schematization of a chain scission process occurring in dangling chains; (a) and (b) are voltaile fragments.
CONCLUSIONS In the t e m p e r a t u r e range 140-200°C, the ageing of anhydride c u r e d epoxies is largely ~d o m i n a t e d by oxidation processes. Ir~ t h e early period of exposure, oxidation favours anhydride m o n o m e r reformation, presumably f r o m m o n o a c i d dangling groups. W h e n this anhydride initially contains unsaturation, it participates in oxidative crosslinking reactions which are responsible for the anhydride retention and for a noticeable increase in the Tg. W h e n the majority of reactive groups participating in these processes is consumed, the oxidation of the regular structural units of the p o l y m e r b e c o m e s p r e d o m i n a n t . This is essentially characterized by anhydride formation. Various reaction pathways can be envisaged for this process.
ACKNOWLEDGMENTS Special thanks are due to 'Electricit6 de France' which supported this study financially.
REFERENCES 1. Madorsky, S. L., Thermal Degradation of Organic Polymers. Interscience, New York, 1964. 2. Bowen, D. O. Modern Plastics, 45 (1967) 127. 3. Struik, L. C. E. Physical Ageing in Polymers and Other Amorphous Materials. Elsevier, Amsterdam, 1978.
4. Henry, D. Hast. Mod. Elast. (Paris), 30 (1978) 45. 5. Sabra, A. Th~se d'Etat, INSA, Lyon, France, 1985. 6. Alekseeva, I. A., Alekseev, G. A., Zupnik, A. E., Tishenin, Yu. V. & Sinitsyna, T. A. Soviet Elect. Eng., 49 (1965) 52. 7. Noskov, A. M. & Alekseev, G. A. Plast. Massy, 12 (198~ 51. 8. Beneo, C. Tech. Ital., 36(4) (1971) 73. 9. Hubmer, E., Martin, R. J., Steffen, R. & Wittker, K. A. IEEE Int. Syrup. on Electrical Insulation, Chicago, 251, 1987. 10. Bozzo, R., Centurioni, L., Coletti, G., Motori, G. & Sandrolini, F. IEEE Int. Symp. on Electrical Insulation, Montreal, 5, 275, 1984. 11. Kessenikh, R. M., Pokholkov, Yu. P. & Petrov, A. V. Izv Tansk. Politekh. Inst., 204 (1971) 73. 12. Forster, P., Hauschildt, K. R., Huber, J., Piecha, G., Pohl, D. & Wilhelm, D. Poly. Deg. and Stab., 24(4) (1989) 269. 13. Merral, C. T. & Neeks, A. C. J. Appl. Polym. Sci., 16 (1972) 3389. 14. Smith, A. R. & Wale, C. P. Chimie et lndustrie (Paris), 105 (1974) 1179. 15. Kessenikh, R. M., Pokholkov, Yu. P. & Petrov, A. V. Soviet Phys. J., 9 (1966) 144. 16. Fisch, W., Hoffrnan, W. & Koskikalio, J. J. Appl. Chem., 6 (1956) 429. 17. Stevens, G. V. J. Appl. Polym. Sci., 26 (1981) 4259, 4779. 18. Bellenger, V. & Verdu, J. J. Appl. Polym. Sci., 30 (1985) 363. 19. Le Huy, H. M., Bellenger, V., Paris, M. & Verdu, J. Poly. Deg. and Stab., 34 (1991) 171. 20. Korcek, S., Chenier, J. H. B., Howard, J. A. & Ingold, K. V. Can. J. Chem., 50 (1972) 2285. 21. Reich, L. & Stivala, S. Autoxidation of Hydrocarbons and Polyolefins. Marcel Dekker, New York, 1969. 22. Kamiya, Y. & Niki, E. In Aspects of Degradation and Stabilization of Polymer, ed. H. H. G. Jellinek. Elsevier Applied Science Pub., Barking, UK, 1978, 79-149. 23. Russel, G. A. J. Amer. Chem. Soc., 79 (1957) 3871.