Thermal Properties
literature Cited
Organic polymers are relatively poor conductors of heat, which makes it difficult to perform operations requiring heat transfer. Elastomers as a class reflect this poor heat transfer both during processing and vulcanization, especially of thick or heabily loaded stocks, and in dissipation of heat during dynamic operations. Thermal diffusivity and thermal conductivity ( 4 , 13, 74) are about 15yogreater for gum vulcanizates than those measured for comparable stocks of natural rubber, suggesting that this elastomer should exhibit improvement in those operations where heat conduction and dissipation are factors. I t is questionable, however, that these differences, while in the proper direction, are sufficiently large to have practical significance.
(1) Amberg, L. O., Robinson, A. E.: IND.ENG.CHEM.53, 368 (1961). (2) Am. SOC.Testing Materials, “ASTM Standards on Rubber Products,” p. 220, December 1954. (3) Dingle, A. D., Rubber World 143, 93 (1960). (4) Frensdorff, H. K., private communication. (5) Gladding, E. K., Fisher, B. S., Collette, J. W.. IND.ENG. CHEM..PROD.RES.DEVELOP. 1.65 (1962) (6) -Gresham, \V. F.! Hunt, M. (tdE. I: du P m t de Nemours & Co.), L . S. Patent 2,933,480 (April 19: 1960). (7) Mazzanti, G., Valvassori, A , Pajaro, G.. Chim. e. ind. (Milan) 39, 743, 825 (1957). (8) Natta. G.. Atloew. Chem. 68. 393 (1956) : .Clodern Plastics 34, . ‘ ‘169 (1956). (9) Natta, G.. Chim.e . ind. (.Milan) 39, 654, 733 (1957). (10) Natta, G., Rubber and Plastics Age 38, 495 (1957). (11) Natta. G., Crespi: G., Rubber A g e (.\’ea! York) 87, S o . 3, 459 (1960). (12) Natta, G., Valvassori, A , , hiazzanti, G., Sartori, G., Chim. e. ind. (Milan) 40, 717,896 (1958). Natta G., Mazzanti, G., Valvassori. .A,. Pajaro, G., Ibid., “1’1, 764 (i959). 114) Rehner. J.. Jr.. J . Polvmer Sci. 2. 263-74 (1947) (15j Schoenbeck, M;A, private communication. (16) Yerzley, F. L., Rubber Chem. @ Technol. 13, 149 (1940). (1 7 ) Ziegler, K., Holzkamp, E., Breil. H., Martin, H., Anfew. Chem. 67, 426 (1955).
Acknowledgment
,
I
~I
The authors thank S. W. Caywood, E. P. Goffinet, R. P. Madrulli, G.LV. Smith, and R. E. Tarney for their contributions to the synthesis of polymer; M. R. DeBrunner and C. A. Young for providing many of the analytical control techniques ; H. E. Huckins, Jr., and T. F. O’Brien for supplying engineering data; and P. A. Peffer, Jr., and M. A. Schoenbeck for assisting in the laboratory evaluation of the product.
,
\
I
RECEIVED for review September 25, 1961 ACCEPTED February 13, 1962 Division of Petroleum Chemistry, 140th Meeting, ACS, Chicago, Ill., September 1961.
ETHYLENE-PROPYLENE RUBBER VULCANIZATION WITH ARALKYL PEROXIDE AND COAGENTS A. E. ROBINSON, J. V. M A R R A , A N D L. 0. A M B E R G Hercules Powder Co., Wilmington, Del.
The efficiency of the dicumyl peroxide vulcanization reaction in ethylene-propylene rubber is an inverse, approximately exponential funciion of propylene concentration. The principal efficiency-limiting reaction is scission of the elastomer molecule facilitated b y the sequential occurrence of propylene segments. Efficiency can b e increased significantly by using selected coagents which, b y addition to polymer radicals, effect suppression of the scission reaction and/or introduce a chain cross-linking reaction capability. Sulfur is a unique coagent which, although it increases efficiency, gives disproportionately high vulcanizate performance. The sulfur effect is ascribed partly to the inherent quality of sulfide bridges. Trisulfide bridges are indicated, on the average.
THYLENE-PROPYLENE
RUBBER
(EPR) has been described
E in terms of its performance characteristics both by Natta (6) and by the authors (7), who concur basically on its merits. Before such evaluations were possible, however, a difficult vulcanization problem was faced. The same paraffin inertness which is advantageous in the vulcanized product ruled out most of the conventional vulcanization technology. Work in these laboratories indicates that, with few promising alternatives, aralkyl and alkyl peroxides provide the best practical solution. This article presents some interpretations and conclusions based on exploratory work on the EPR vulcanization process 78
I & E C P R O D U C T RESEARCH A N D D E V E L O P M E N T
with dicumyl peroxide, The supporting evidence is derived mainly from cross-link density analysis by the solvent swelling method of Kraus ( 5 ) . T h e copolymers described were prepared in these laboratories.
Raw Polymer Characteristics
In view of the large variety of copolymers possible in the ethylene-propylene system, a brief description of some of the pertinent raw polymer variables is helpful. Dependent primarily on the method of preparation, copolymers of molecular
weight between 100,000 and 1,000,000 (approximate limits) containing upward of about 25 mole % propylene comprise a family of true elastomers. Their basic configuration is probably linear; they contain no significant amount of unsaturation. Under selected conditions of polymerization, the comonomer distribution is probably random, though sequences of one monomer necessarily occur where one monomer predominates in the over-all composition. From the performance standpoint, the selection of an optimum copolymer is a matter of compromise. Strength, for example, varies more or less directly, and processability inversely, with molecular weight. Processability also varies directly with propylene concentration, while vulcanization efficiency (discussed below) varies inversely. A very satisfactory compromise for many uses is the copolymer of weight average molecular weight about 250,000 with 33 mole propylene. In the following discussion reference is to the 33 mole %, Bw250,000 copolymer, except where another is specified.
70
Cross linking
Mechanism. Though the process of peroxide vulcanization is neither new nor specific for EPR, the subsequent discussion is facilitated by definition of the steps involved. Figure 1 shows the generally accepted reaction sequence with dicumyl peroxide. First, the peroxide undergoes homolytic cleavage, producing cumyloxy radicals a t a rate governed by temperature alone. Some of the cumyloxy radicals then abstract hydrogen from the rubber, producing rubber radicals which ultimately couple to form cross links. T h e remaining cumyloxy radicals decompose to acetophenone and methyl radicals. Presumably, the latter also abstract rubber hydrogen. Cumyl alcohol, acetophenone, and methane are the principal reaction products. Table I gives some typical product analyses which show the molar ratios of ketone (and methane) to alcohol found. Although the data account for only half the peroxide, they suggest that the ratio does not vary greatly with ethylene-propylene composition in comparison with the difference found between EPR and styrene-butadiene rubber (SBR) (i.e., 1.17 to 1.65 us. 0.18). I t would be necessary to write many different radical reactions to illustrate all that may occur in EPR during vulcanization. Two, however, are sufficient (and necessary) to account for the observed results. These are radical coupling and betascission, as shown in Figure 2.
HOMOLYllC CWVAGE OF PEROXIDE
Table 1.
Analysis of EPR Vulcanization By-products 25 50 100 SBR 0,286 0.286 0.286
EPR, mole % propylene Initial peroxide, mmoles By-products, mmoles Cumyl alcohol Acetophenone Methane Ketone to alcohol ratio Material balance,
0.130 0,192 0.172 1 .48 57
COUPLING
- c-c+c-c-c
C I
-1
[-c=I-c-cz-l BETA-SCISSION
CUMYLOXY RADICAL DECOMPOSITION
[
-
-I
-
C 1 C-c-c-c-c-c
c4-yc-c
-
C
-.
- c c - c-csi-c-c-I C-c-c-+C~
-1
Figure 2.
-1
-C-c-iS?-c
[
-c-c-c-c:-c
C
Initiation and transfer reactions
0.18
OTHER RUBBER RADICALS
[
Figure 1.
0.091 0.150 0.197 1.65 42
Coupling, apparently, is the principal cross-linking reaction in EPR. I t is a bimolecular reaction of almost negligible energy requirements and no doubt occurs every time two rubber radicals approach within a reaction diameter. The rubber radical concentration, however. probably does not ordinarily exceed about 10-6M (estimated) at any time during vulcanization. Thus, coupling would appear not to be a kinetically favored reaction. Despite this, however, extensive cross linking does actually occur. Presumably. the probability of coupling is significantly enhanced by extensive radical migration (by the hydrogen abstraction process) and by the likelihood that two rubber radicals derived from the same molecule of peroxide are in close original proximity (cage effect). Beta-scission of the rubber molecule is apparently the main competing reaction, as evidenced by the drastically reduced molecular weights of the soluble fractions of incompletely cured amorphous polypropylene vulcanizates. Scission is a unimolecular reaction of finite, if low, energy requirements. Figure 2 shows a tertiary radical with tertiary carbons in both beta positions. O n the basis that high tertiary hydrogen reactivity and a high proportion of secondary hydrogens are offsetting factors, radicals are probably formed in EPR almost randomly. I t is possible, therefore, that any EPR radical which may be written may actually occur, each having a characteristic proclivity for scission. I t would appear that the most favorable configuration for scission is a tertiary radical bounded by tertiary carbons in the beta positions on both sides (Figure 2), which configuration has a high probability of occurrence in EPR of high propylene content. To the extent that copolymer scission occurs, it is wasteful of radicals in the sense that extra cross linkages are required to compensate the detrimental effects of chain breakdown. In the vulcanization temperature range, 280" to 330" F., no advantage of high temperature has been observed. I t might have been expected that high temperature would favor cross
[
RUBBER RADICAL K)RMATK)N
0,149 0.175 0.209 1 .17 57
CI
C
C CI
-
I
Rubber radical reactions VOL 1
NO. 2
JUNE 1 9 6 2
79
linking (bimolecular) over scission (unimolecular) by virtue of giving a higher steady-state radical concentration, but the same factor would probably also favor the reaction of higher activation energy (scission). Efficiency. T h e cross-linking efficiency of dicumyl peroxide in EPR varies as a function of the copolymer composition. T h e relationship of the peroxide requirement for vulcanization to the propylene content of the copolymer is shown by the solid curve in Figure 3, where the left-hand ordinate indicates the minimum amount of peroxide which will give a 95% or more insoluble rubber matrix. The amount is shown to vary approximately exponentially from 1.5 parts of peroxide per 100 parts by weight of copolymer (p.h.r.) a t 25 mole yo propylene to about 18 parts a t 90% propylene. Cross-linking efficiency (100 times the ratio of the actual cross link density, determined by the swelling method, to the theoretical maximum density based on one theoretical cross link per molecule of peroxide) was approximately 70% at 25 mole % propylene, about 65% at 33% propylene, and less than 10% at 90% propylene. I n amorphous polypropylene, the peroxide requirement is about 23 parts (Figure 4). ,411 test stocks were of the same weight average molecular weight. Amorphous polypropylene was used extensively in the exploratory phase of this investigation because it provided a sensitive medium for testing the scission reaction and its suppression. The cross-link density of the minimally cured matrices indicated in Figure 3 is less than that required to give form-stable vulcanizates of acceptable physical properties. [Theoretically, a minimal cure state is attained when n molecules of rubber are connected by n - 1 ideally distributed cross links. The minimally cured matrices of Figure 3 contain 10n to 20n linkages, based on the swelling test measurement ( 5 ) . The cure state required for form-stable vulcanizates of good strength, average modulus. and low set entails an estimated 40 cross links per molecule.] For example, a t 3370 propylene, 2 parts of peroxide are indicated for the minimal cure, but up to 2 additional parts are customari1)- required for practical performance. The peroxide requirement may be reduced significantly by the use of efficiency-increasing coagents. LYhen less than the indicated minimum level of peroxide was used without the benefit of such coagents, partly cured vulcanizates resulted. I n partly cured amorphous polypropylene vulcanizates, the molecular weight reduction found in the soluble fractions clearly demonstrates the effect of scission. A close correlation of the minimum peroxide demand with the estimated number average length of propylene sequences in the copolymer is shown in Figure 3 by the dotted curve. Sequence length was calculated from the equation L = (1 P ) -I, where L is the number average length and P the mole fraction of propylene. The expression is an approximation based on a general equation (3) which assumes random copolymerization, The approximation reportedly holds for degrees of polymerization above about 100.
20
LL
E
‘5
r
1 IO
F
I
5
EFFICIENCY 65% 25
30
40
0 90
50 60 70 80 COPOLYMER COMPOSITION, MOLE % PROWLENE
Figure 3. Peroxide composition
requirement
vs.
copolymer
Correlation with propylene sequence length
AMORPHOUS POLYPROWLENE
PEROXIDE CONC.
Figure 4.
EPR (50MOLE % PROPYLENE)
(EQUALS COAGENT CONC.),
PHR
Coagent evaluation test results
A. Diallyl phthalate B. Quinone dioxime C. Dinitrosobenzene D. Triallyl cyanurate E. Divinyl adipate --- Peroxide alone
-
Efficiency-Improving Coagents
I t is possible to improve the peroxide cross-linking efficiency in EPR appreciably by the use of coagents. As would be expected, the greatest efficiency increases have been observed where the original efficiency is poorest-that is, in high propylene copolymers and in amorphous polypropylene. Figure 4 shows some typical coagent evaluation test results with amorphous polypropylene and 50 mole % copolymer. Coagents (singly) and peroxide, in equiweight proportion, were 80
I & E C P R O D U C T RESEARCH A N D DEVELOPMENT
incorporated a t three levels. The test compounds were cured 30 minutes a t 320’ F. and the insolubility of the vulcanizates in toluene (per cent gel) was measured. The results show that gel values of 80 to 100% were obtained in amorphous polypropylene with coagent in the 5- to 10-p.h.r. range, where gel values in the absence of a coagent were only 5 to 3076. I n the copolymer, in which peroxide (alone) was 55 to 6070 efficient, significant but less dramatic increases in the gel value were noted. A listing of some of the more effective coagents encountered is given in Table 11. Agents have been grouped according to which of two apparent mechanisms might be expected to explain their effect best, although with several agents either mechanism may seem plausible. The proposed coagent reaction mechanisms are shown in Figure 5. I t is suggested that all effective coagents first react
WITH QUINONE DlOXlME
was used. Under the same conditions of extraction and analysis, considerable amounts of cumyl alcohol and acetophenone were obtained. T h e optimum molar ratio of coagent to peroxide has usually been near unity, but, as suggested in Figure 4, it must be determined separately for each coagent. 1[R3-1
[~4.1
+
H H R’C-C-Y-C-C-R3
RH = RUBBER,
Figure
5.
R4H
[ R * 1 = RUBBER RADICAL
Coagent reaction mechanisms
additively with rubber radicals. T h e agents typified by quinone dioxime would appear to provide effective competition for the backbone scission reaction by virtue of a relatively fast addition to rubber radicals to form resonance-stabilized intermediate radicals. T h e stabilized intermediates may couple (Le., cross-link), but they probably lack sufficient energy to abstract hydrogen from the polymer to propagate a chain reaction. Thus, the theoretical cross link productivity of the process remains one cross link per two radicals, and cross-linking efficiency, any increase in which may be attributed to suppression of the scission reaction, may not exceed 100%. Coagents typified by divinyl adipate would form intermediate radicals with little if any resonance stabilization. These more reactive intermediate radicals! which are subject to reversal and are, therefore, probably less effective in suppressing scission, would have sufficient energy to abstract hydrogen from the polymer and propagate a chain reaction. For improvement of cross-linking efficiency, multifunctionality of such agents has been found to be necessary. The monofunctional agents, vinyl butyrate and vinyl stearate, at equivalent concentrations, were found to have no effect on cross-linking efficiency, although some scission reduction might have been expected. By the above mechanism, divinyl adipate-type coagents, which may promote a chain process (Le., increase cross-link productivity), might conceivably increase the cross-linking efficiency well beyond 100%. However, divinyl adipate has been found to increase efficiency only up to 10070, as indicated in Figure 4. No verified instance of cross-linking efficiency greater than 1 0 0 ~ ohas been found. Apparently, steric factors and chain-terminating agents in the polymer ( e g . , oxygen) prevent a genuinrly catalytic peroxide crosslinking process. Aside from the observed improvements of efficiency, the main supporting evidence for the proposed mechanisms is the immunity of the coagents to extraction. By infrared and ultraviolet spectrophotometric analysis of CS2 and iso-octane extracts of vulcanizates (specifically, those with quinone dioxime and triallyl cyanurate coagents), it has been found that less than 5Y0 of the coagents can be removed by exhaustive extraction (one month a t 25’ C.) except where excess coagent
Table 11.
Efficiency-Improving Coagentsa
Quinone dioxime Dinitrosobenzene Diphenylguanidine Grouped according e j e c t ; see Figure 5. a
t o most
Sulfur Effect
R1-C-&Y3-C-R3
Divinyl adipate Triallyl cyanurate Diallyl phthalate plausible mechanism f o r explaining their
Elemental sulfur is apparently unique in its effect in EPR. DiGiulio and Ballini ( 4 ) have shown that sulfur is highly effective in suppressing side reactions such as scission. Sulfur appears also to be responsible for some peroxide waste. When sulfur and dicumyl peroxide (in the ratio of 3 gram-atoms per mole) were reacted in bulk-that is, in the absence of any diluent-only traces of the expected radical decomposition products were found. The major products were acetone and $-cumylphenol. These are the products of a proton-catalyzed ionic cleavage reaction of dicumyl peroxide, a reaction which is not productive of rubber cross linkages. \Vhen the decomposition was repeated a t a dilution of 10% sulfur-plus-peroxide in n-decane, significant amounts of ionic cleavage products were again found, though the major products were the anticipated radical cleavage products. Some wasteful decomposition of peroxide is, thus, apparentlv unavoidable with the sulfur-peroxide combination. The acid catalyst for ionic decomposition is presumably derived from sulfur. although the only combined sulfur forms identified were dimethyl sulfide and disulfide. Nevertheless, EPR vulcanizates with sulfur have consistently out-performed those made with any sulfur-free formulation. Seemingly, the sulfur effect on performance characteristics may be attributed partly to the inherent nature of the sulfur cross linkages. T h e stoichiometry of the EPR-peroxide-sulfur interaction is indicated in Table 111, which shows the relationship of bound sulfur (2) to total sulfur a t different levels of dicumyl peroxide. At 4 parts of peroxide, 1 part OC sulfur was quantitatively incorporated. KO lower amount of peroxide tested gave quantitative sulfur incorporation. Bound sulfur levels with less than 4 parts of peroxide, representing different degrees of a sulfur-rich peroxide-poor situation, suggest that the upper limiting stoichiometry of sulfur to peroxide in the three-way interaction is about 1 to 3 by weight. or about 3 gram-atoms of sulfur per mole of peroxide. T h e existence of trisulfide bridges o n the average is, therefore, indicated. Rehner, Wanless, and Wei (7) have reported that bridges up to pentasulfide form under model conditions. Perhaps the best evidence for the conclusions which have been presented is that in Table IV showing the peroxide requirement for equivalent modulus in identically loaded vulcanizates with several cure modifications of the sort described. Compared with a basic 3.5 parts peroxide stock, both efficiency-improving coagents (triallyl cyanurate and divinyl adipate) are shown to have reduced the peroxide requirement to 2.5 parts. Sulfur, in the absence of another agent, did not
Table 111. Peroxide, P.H.R.
4 3 2 1
Bound Sulfur in
Bound S , P.H.R. 1 .oo 0.86 0.66
EPR Vulcanizates Total S, P.H.R. 1 .oo 1 .oo 1.oo 1 .oo
0.28
VOL.
1
NO. 2
JUNE 1 9 6 2
81
Table IV.
Tensile Properties with Selected Formulations Cure cycle: 40 minutes at 310’ F.
EPR 100 100 HAF black 50 50 Dicumyl peroxide, p.h.r. 3,5 2.5 .,. 3.0 Divinyl adiphate, p.h.r. Triallyl cyanurate, p.h.r. , . . ... Sulfur, p.h.r. ... ... 300% modulus, p.s.i. 1350 1350 Tensile strength, p.s.i. 2260 2100 Elongation at break, % 460 375
100 100 100 50 50 50 2.5 3.5 2.5 ... ... 1 .o 1.5 , . . ... , . . 0.9 0.5 1350 1350 1350 2100 2800 3180 425 510 530
literature Cited
(1) Amberg, L. O., Robinson, A. E., IND.ENG.CHEM.53, 368 (1961) ; Rubber Plastics Age 42, 875-9 (July 1961). (2) . , Am. SOC.Testing Materials, Philadelphia, Pa., ASTM D297-60T, Sect. 18-20. (3) Davis, W. E., Applied Mathematics Division, Hercules Powder Go.. Wilmington. Del.. derivation based on W. Feller. “Introduction to Probvability Theory and its Application,” ’Vol. I, pp. 56-9. Wilev. New York. 1950.
(7) Rehner, J., Jr., Wanless, G. G., Wei, P. E . , ference on Elastomers, 1961. lower the peroxide requirement but did significantly enhance tensile strength. The combination of sulfur and divinyl adipate had the additive effect of both lowering the peroxide requirement and enhancing strength.
RECEIVED for review September 27, 1961 ACCEPTED April 6, 1962 Division of Petroleum Chemistry. 140th Meeting, ACS, Chicago, Ill., September 1961.
CATALYSTS FOR T H E POLYMERIZATION OF
IS0 PR EN E T O fr~n~=l,4=PO LY IS0 PR EN E (SY N T H ET IC BALATA) J. S.
L A S K Y , H. K . G A R N E R , A N D R . H. E W A R T
Research Center, United States Rubber Go., Wayne, iV. J .
A catalyst prepared from aluminum triethyl and vanadium trichloride has been reported in the literature to polymerize isoprene to frons- 1,4-poIyisoprene. The efficiency of this heterogeneous catalyst is increased by a factor of 10 or more by supporting the vanadium trichloride on clay to increase its surface area. A further and even more substantial improvement in efficiency was achieved by adding a tetraalkyl titanate as a third catalyst component. This gave a highly active soluble catalyst system for the preparation of synthetic balata. Properties of synthetic balata are compared to those of natural balata.
t r a w l ,4-polyisoprene, from natural sources has been known for many years. I t is a moderately hard, tough, crystalline thermoplastic material whose major use today is in golf ball covers. A catalyst prepared from aluminum triethyl and vanadium trichloride for the polymerization of isoprene to synthetic balata has been disclosed ( 4 , 5 ) . The authors’ work with this catalyst has led to substantial improvements ( 3 ) ,which are reported on here. A new, highly efficient soluble catalyst system for the preparation of synthetic balata and the properties of the polymer so produced are also discussed. ALATA,
Supported Vanadium Trichloride
After extensive work with the aluminum trialkyl-vanadium trichloride catalyst described by Natta, the best efficiencies in the conversion of isoprene to synthetic balata were only 1.30 to 1.60 grams of polymer per gram of VC13 per hour. The catalyst was heterogeneous, and the limiting factor seemed to be the surface area available. Since VC14 is a liquid which is soluble in organic solvents, it seemed likely that a more finely divided catalyst could be obtained, using this as a starting material in place of the solid, crystalline, insoluble VCl,. Furthermore, Natta (6) had found 82
I&EC P R O D U C T RESEARCH A N D DEVELOPMENT
VC1, to be lar more efficient than VCl, in a catalyst for conversion or butadiene to the all-trans polymer. I n contrast, aluminum trialkyl-vanadium tetrachloride catalysts proved in extensive testing to be less efficient (maximum efficiency, 20 grams of polymer per gram of VC14) than aluminum trialkylvanadium trichloride catalysts (maximum efficiency, 80 to 100 grams of polymer per gram of VCl,). Furthermore, the polymer prepared using VC14 had a higher 3,4- content and a lower trans content than that prepared using Vel,, and also conlained gelled material that could not be broken down by milling. This illustrates again the major difference in polymerization behavior that is often observed with minor changes in monomer structure when stereospecific catalysts are used. I n an effort to increase the active surface, supports were used for the vc13. The VClB could be deposited on the surface of a nonporous inert carrier such as titanium dioxide or kaolin by heating a slurry of the supporting material in a benzene solution of vanadium tetrachloride at reflux for 1 to 2 hours. The VC14 smoothly decomposed into VCI, and chlorine, and the VC1, deposited on the surface as formed. When the reaction was complete, the supported VCl, was filtered off in an inert atmosphere, washed, and dried, and was then ready for use. T h e yield of VC13 based on VClr usually ran between 80 and
90%.
