Correlation between Structure and Thermal Stability of Epoxy Resins GEnr~atD F. L. Ermvas A number o[ polycarboxylic acids, anhydrides, phenols, amines and catalysts, were evaluated as curing agents for epoxy resins and the resulting resins screened for thermal stability. As the basic resin molecule, Bisphenol A was also replaced by other diphenols. The thermal behaviour of these resins was related to their structure. In spite of a certain instability of the ester bond, the anhydride-cured resins gave the highest heat distortion and the best ageing stability. The BFr-amine complex is effective as a catalyst, especially with resins having more than two epoxy groups. Phenols and amines as curing agents gave low heat distortion values. O~ the anhydrides investigated pyromellitic dianhydride is best, followed by maleic anhydride. Chlorendic anhydride, which gives last curing and high heat distortion, has low ageing stability. Replacement of Bisphenol A as. the basic resin by naphthalene, diphenyl or diphenylsulphone derivatives might raise the heat distortion. However, structures like Resin B seem to be more promising.
Tile EPOXY resins are distinguished, among thermosetting resins, by their ease of handling, thermal stability, chemical resistivity and their outstanding mechanical and electrical properties. The preparation of epoxy resins is performed normally by reaction of polyphenols, preferably diphenols of the type represented by 4,4'-dihydroxydiphenylmethane (Bisphenol A), With epichlorohydrin. This reaction yields the diglycidylether, or low molecular weight polymers, of Bisphenol A which have epoxide end-groups. These 'basic resins', as the reaction products of polyphenols and epichlorohydrin will be called throughout this paper, can be cured to thermosetting resins by heating in a solvent or a melt with compounds having two or more active hydrogens, such as carboxylic acids or anhydrides, amines, phenols, or certain types of catalysts, such as tertiary amines, Lewis acids, etc. The mechanism of the reaction is extremely complex and has been explored by numerous investigators 1-9. Carboxylic, phenolic and amino groups can react with epoxide groups to form secondary alcohols, which in turn can react with carboxylic acids and anhydrides as well as with epoxide groups. In the latter ease another hydroxyl group is the reaction product. In addition, catalytic action of reactants may cause epoxide groups to form polyether chains by poly-addition, or dimerization and trimerization products. As a result, it is difficult to predict the influence of certain structures, in either the basic resin or the curing agent, on the thermal stability of the cured resin. The purpose of this paper is to survey these relationships. Weight loss and heat distortion were used as evaluation methods. Weight loss studies provide an estimate of the degree of disintegration within the resin; however, there is no direct relationship between weight loss and thermal stability. Some changes within or between the molecules 304
STRUCTURE AND THERMAL STABILITY OF EPOXY RESINS
may occur with only negligible weight loss or with an increase in weight (crosslinking, additional polymerization, absorption of oxygen). Therefore it is difficult to predict what amount of weight loss will be critical. Some resins may still be serviceable after a weight loss of up to 30 per cent, whereas other materials show heavy destruction with much lower weight losses. The Vicat method was used for determining heat distortion, because this procedure allows testing of very small samples with a high degree of accuracyx°. With this method, a load of 5 kg is applied on a 1 mm 2 area of the material by means of a steel needle with a fiat tip. The temperature is raised 50°C per hour, The Vicat temperature is the temperature at which a penetration of 1-0 mm has been reached. The Vicat heat distortion temperature is a measure of mechanical stability at higher temperatures, and gives a general indication of hardness and flexural strength. It allows conclusions on the degree of crosslinking and the strength and stability of binding forces to be made. High heat distortion values require strong secondary bonds and/or thermally stable primary bonds, in connection with close crosslinking. It is to be expected (as can be seen in the curves) that the Vicat temperature rises during curing and subsequent ageing to a maximum and then falls off with increasing destruction of primary bonds. This might suggest that the weight loss should be measured and calculated from the time at which the maximum Vicat temperature is obtained. But this point can be found only with a large number of measurements. Furthermore, the material may already have been damaged by loss of substance between the onset of gelation and the end of the curing process. EXPERIMENTAL
The basic resin of Bisphenol A, used in the major part of this study, which had a softening point of 61°C and an epoxide equivalent of 470, was made to react with the curing agent by melting. The calculation of the necessary amount of curing agent was based on the assumption that one active hydrogen or one anhydride group corresponds to one epoxide group (see also p. 292). Catalysts were applied in an amount of 1 per cent of the basic resin. The molten mixture was poured into dishes of aluminium foil 60mm in diameter which had been greased with silicone .oil. Curing proceeded for 48 h at 140°C. If this curing time was insufficient to obtain a solid state at this temperature, the curing was continued at 160, 180 or 200 °C. The resin discs so obtained had a thickness of 3-5 mm. Three samples of about 15 × 15 mm were cut from these discs. Two of these three samples were placed on a wire screen and aged in a laboratory oven at 230°C. The weight loss, as percentage of the green weight, was determined after 1/2, 24, 48 h (two samples) and 200h (one sample). An examination of the condition of the samples after ageing (swelling, cracking, melting) supplied additional information on thermal stability. The third sample was used for the determination of the Vicar heat distortion temperature1~ after curing, but with a penetration of the needle 305
G E R H A R D F. L. E H L E R S
of only1° 0"25mm instead of 1.0mm. With the two aged samples, the Vicat temperature after 48 and 200 h ageing was determined. CURING
Carboxylic acids and anhydrides The carboxy and anhydride groups can react with epoxy groups, as well as with alcoholic hydroxyls already present or formed by reaction. On the other hand, the glycidyl groups are partially consumed by reacting with hydroxyl groups. The optimal molecular amount of acid or anhydride groups per epoxide group 12,1~, therefore is not 1, but between 0"6 and 1"0, depending upon the curing agent and the curing conditions. In these experiments one anhydride group per epoxide group was used. It is interesting that results often are not changed much by changing the amount of curing agent within these limits, as was found in the case of chlorendic anhydride. However, preliminary studies showed that phthalic anhydride as a curing agent gives the lowest weight loss during ageing with 0.5 mole of anhydride per mole of epoxide, maleic anhydride with 1-0 mole. Mole ratios between 0"5 and 1"0 have not been examined. Phthalic anhydride gives an optimum Vicat green temperature with an anhydride/epoxide mole ratio of 1"0, while with maleic anhydride the Vicat temperature after prolonged ageing rises with increasing amount of anhydride, at least up to a mole ratio of 5. At the same time, the. weight loss increases rapidly; however, it is less than the calculated excess weight of the maleic anhydride. This surprising fact about the rising of heat distortion temperature can best be explained by assuming that excess of maleic anhydride links with itself and with the resin. It might be possible that a poly-anhydride chain forms between two epoxide or hydroxyl groups of the resin molecules: --O--CH2 0 0 0 0 0 0 CH2--O'--
This structure, offering sensibility to decarboxylation and oxidation, might explain the high weight loss during thermal ageing. Polymerization through double bonds seems, at first, to be less probable. If, however, Vicat temperatures of resins cured with the normal amounts of maleic anhydride (I), citraconic anhydride (lI) and succini~ anhydride (lII) are compared, HC, COOH HaC'C-COOH H2C'COOH
for the first two curing agents, especially maleic anhydride, higher values are found and these even rise during ageing. Because the structure of maleic anhydride differs from that of succinie acid only in the double bond, and no excess of anhydride groups is available for poly-addition reactions, a polymerization through the double bonds, preferably at temperatures above 200"C should also be considered. Certain evidence for this is given 306
STRUCTURE AND THERMAL STABILITY OF EPOXY RESINS by a recent publication" wherein the author discusses the polymerization of double bonds by catalytic m o u n t s of epoxy compounds. Phthalic anhydride as a curing agent gives resins with higher Vicat temperatures than succinie acid, because of its more rigid structure. On the other hand, the benzene ring effects greater distance between resin chains and this may be responsible for a slightly higher sensitivity of the resin to deterioration, i.e. for higher weight loss. Weight loss is lower and heat distortion is higher with hexahydrophthalic anhydride. This would not normally be expected because the thermal stability of this molecule is lower and the structure is more flexible. An explanation for this behaviour cannot be given at this time. Strong secondary bonding forces may be the reason for the high heat distortion of resins cured with chlorendic anhydride. However, the high Vicat temperatures fall off rapidly during thermal ageing, and the weight loss is considerable. This fact seems to indicate that the structure of this hardener has insutficient thermal stability due either to its inner fivemembered ring structure or its high chlorine content. Heating of the cured resin for a short time at 350°C results in acid deterioration products. This resin, like other resins with high chlorine content, has a very low adhesion to the aluminium dishes used. High functionality, close crosslinking and stable structure are the reasons for the high Vicat temperatures and the favourable weight losses with pyromellitic dianhydride as a curing agent. The optimal heat distortion will not be reached without prolonged ageing. Figure 1 shows examples of Vicat temperature and weight loss curves of anhydride-cured resins.
/ 8 C 0Ia n w
24 48 200h~ Ageing time at 230°C
, (b) n 24 48 200~ Ageing time at 230°C
Figure /--Anhydride curing agents. Resin A with : I, maleic anhydride; 2, maleic anhydride (in 5-~old excess); 3, phthalic anhydride; 4, chlorendic anhydride; 5, pyromellitic dianhydride 307
GERHARD F. L. EHLERS
Polyhydric phenols Diphenols cure mainly by extension of the resin chain and less by crosslinking. Even polyphenols with three or four phenolic hydroxyls seem to form a relatively loose network, because the phenolic groups react only with epoxide end-groups and not with secondary alcoholic hydroxyls within the resin chain, as is the case with anhydrides. A certain degree of crosslinking might occur by the reaction of epoxides with the alcoholic hydroxyls formed during the process. Therefore, the phenolic cured resins, even those hardened with polyvalent compounds such as phloroglucinol and 1,1,2,2-tetra(p-hydroxyphenyl)ethane, have low Vicat temperatures. On the other hand, weight losses with phenols normally are lower than with anhydride hardeners; this may indicate that the stability of the ether linkage is higher than that of the ester bond in anhydrides. Although polyhydric phenols do not seem suitable as curing agents for high .temperature resistant epoxides, some interesting conclusions on the effect of structure may be reached from the test results. Diphenols with hydroxyl groups so arranged that they are opposite in the molecule give rise to resins with somewhat lower Vicat temperatures than those with the hydroxyl groups in other positions. Resorcinol is more favourable than hydroquinone, 1,6-dihydroxynaphthalene better than 1,5or 2,7-dihydroxynaphthalene, and 3,3"-dihydroxybiphenyl better than
4,4'-dihydroxybiphenyl. One explanation, but not necessarily the only one, for the higher Vicat temperatures of the compounds indicated might be a twisting effect of the chains with a possible interlinkage of different chains. Such an enchainment of different molecules might result in more rigid structures. However, so far this is merely an assumption. Another interesting fact can be obtained from the results. 4,4"-diaminodiphenylsulphone was "caused to react with basic resin in a 1 : 1 ratio (amino/epoxide group), so that the same structure would be expected as with 4,4"-dihydroxydiphenysulphone, except that one resin contains the - - C H . O H - - C H 2 - - N H - - linkage, the other the ---CH.OH---CH~----O--bond. Results, especially after ageing, showed about equal thermal stability of both linkages. Furthermore, it seems that substitution in the diamine occurred in the 4,4'- and not in the 4,4- position, otherwise lower Vicat temperatures would be expected. Vicat temperatures and weight loss with polyhydric phenols as hardeners are of the same order of magnitude regardless of whether the basic structure of the curing agent is benzene, naphthalene, biphenyl, diphenyldimethyl methane or diphenylsulphone (see Figure 2). This indicates again the lack of crosslinking because the difference in the results is considerably greater when these different structures are applied in the basic resin with anhydride as curing agents.
Amines As was indicated by some preliminary studies and experiments, normally both hydrogen atoms in a primary amine react with the epoxide. Although 308
STRUCTURE AND THERMAL STABILITY OF EPOXY RESINS
(b) 3 ~
~- 601. I ~ i 2/-, 48 200 h Ageing time at 230°C
24 z,8 200h Ageing time at 230~C
Figure 2--Phenolic curing agents. Resin A with: 1, hydroquinone; 2, resor-
cinol; 3, phloroglucinol; 4, 1,5-naphthalenediol; 5, 1,6-naphthalenediol; 6, 4,4'dihydroxydiphenylsulphone
N,N-diallylmelamine gives a higher heat distortion with an excess of this compound (namely 1 amino group per epoxide, instead of 0"5), this probably is due to polymerization of the allyl groups. In general, no higher heat distortion values were obtained with amines as curing agents. As in the case with phenols, the possibilities of reaction are lower than with acids and anhydrides, although every primary amino group has two reactive hydrogens. All of these will react only with epoxide groups, normally at the ends of the resin chains, while carboxyl and anhydride groups, as was mentioned earlier, partially form half-esters with alcoholic hydroxyls and react further with glycidyl groups; the unchanged glycidyls form ether linkages with other alcoholic groups, thus forming a dense crosslinking. However, because of the double functionality of the amino groups, the Vicat temperatures of comparable structures are somewhat higher with the amines as curing agents than with phenols. Because of the branching properties of the amino group, 'twisting effects' might not be seen as clearly as with phenols, m-PhenyIenediamineand 2,6-diaminopyridine gave higher heat distortions than p-phenylenediamine; but with diaminodiphenylsulphone, just the reverse effect happens: resins cured with 3,3"-diaminodiphenylsulphone have lower Vicat temperatures and lower weight losses than those cured with 4,4-diaminodiphenylsulphone. Ethylenediamine and diethylenetriamine as hardeners show very low thermal stability. This may, to a lesser extent, be due to the structure; the aliphatic amines react very fast and probably form a gel with gaps and reactive groups included; this gel might be sensitive to thermal degradation. Of the aromatic amines investigated the following indicate an order of decreasing Vicat temperatures and, nearly in the same way, increasing weight loss: 4,4"-diaminodiphenylsulphone, benzidine, 4,4"-diaminodiphenylmethane and p-phenylenediamirte. The SO2-bond seems to be rather rigid and gives rise to strong secondary forces; the benzidine molecule is more rigid than is the diphenylmethane, while p-phenylenediamine, as a rather short linkage, does not contribute too much to the rigidity of the system. 2,4,6-Toluenetriamine seems to give a loose network similar to that of phloroglucinol, in spite of its functionality of 6. Among other nitrogen compounds, melamine cures slowly and results in low Vicat temperatures in spite of its high functionality. Results for amine-cured resins are given in Figure 3. 309
GERHARD F. L. EHLERS
24 48 200h Ageing time at 230°C
24 48 200h Ageing time at 230°C
Figure 3--Amine curing agents. Resin A with: l, dicthylenetriamine; 2, mphenylenediamin¢; 3, p-phenylenediamin¢; 4, 2,4,6-triaminotoluen¢; 5, benzidine; 6, 3,3'-diaminodiphenylsulphone; 7, 4,~'-diaminodiphenylsulphone; 8, N,N-diallylmelamine; 9, BF~monocthylamine complex
Ageing time at 230°C
Ageing time at 230°C
Figure 4---Various basic resins cured with maleic anhydride: 1, Resin B; 2, Resin B cured with BFa-monoethylamine complex, 3, Resin C; 4, basic resin from 1,5-naphthalenediol; 5, basic resin from 1,6onaphthalenediol; 6, basic resin from 4,4"-dihydroxybiphenyl; 7, basic resin from 3,Y-dihydroxybiphenyl;. 8, basic resin from 4,4"-dihydroxydipheoylsulphone
STRUCTURE AND THERMAL STABILITY OF EPOXY RESINS
Catalysts A number of experiments with catalytic amounts of curing agents indicated a BF3-monoethylamine complex, as the most promising catalyst, although different in its effect. Bisphenol A as basic resin (Resin A) gave no superior products (Figure 3), but Resin B (see below) cured to a resin with a very high Vicat temperature (Figure 4). Owing to its structure, Resin B, by homopolymerizing, gives a much closer network than is possible with resins from Bisphenol A. CURING OF OTHER T Y P E S OF B A S I C R E S I N S Besides Resin A, some other types of basic resins were investigated. One of these was 1,1,2,2-tetra(p-glycidyloxyphenyl)-ethane (Resin B);
CH2--.CH--CH2--O ~ : : : ~ --~--0-\o / k_k__Z/ Resin B
CH2--CH--CH2 \0 /
another 3,4-epoxy-6-methylcyclohexylmethyl-3,4-epoxy-6-methyleyclohexanecarboxylate (Resin C). These were cured with different curing agents.
0 II °
Resin C Both compounds react very fast with pyromeUitic dianhydride, resulting in premature gelation. With other curing agents, Resin B forms products with high heat distortion, the optimum of which will not be reached before prolonged ageing. Resin C also provides a good heat distortion, but thermal stability during ageing is lower, probably because of its ester bond. Epoxy-resins from 1,5;naphthalenediol, 4,4'-dihydroxybiphenyl and 4,4"-dihydraxydiphenylsulphone indicate good heat distortion values, but the first two also have high weight losses. The materials 3,4,6 and 7 in Figure 4(b) show a second knee in their curves at approximately 40 h at 230°C. This probably can be attributed to the evaporation of some material, which did not react with the curing agent, during the first phase of the ageing process. A surprising effect was found using ~-pinene oxide, dipentene oxide and allylglycidyl ether as components for the Resin A-maleie anhydride system. Because these compounds contain only one epoxide group per molecule, a plasticizing effect was expected. Actually, however, the heat distortion was very high, especially after ageing. A few examples can be seen in Figure 5. With --pinene oxide as a component, an optimum in heat distortion seems to 311
GERHARD F. L. EHLERS
(o) 24 /,8
200h Ageing time at. 230 °C
200h Ageing time at 230 °C
Figure 5--Effect of some monofunctional epoxy compounds: 1, Resin A (100 %); 2, Resin A (90%), ~,-pinene oxide (10%); 3, Resin A (80%), =-pinene oxide (20%); 4, Resin A (70%), =-pinene oxide (30%); 5, Resin A (60%), ¢-pinene oxide (40%); 6, Resin A (60%), allyl glycidyl ether (40%). Curing agent: maleie anhydride
be obtained (after prolonged ageing) with about 10 per cent, while allylglycidy! ether still gives high Vicar temperatures with 40 per cent. A polymerization through the double bonds of these compounds, probably including those of maleic anhydride, can be presumed (see p. 306). ~-Pinene oxide (I) may have formed a double bond (II) after reacting with a carboxyl group15. which is formed in the first step of curing with anhydrides.
---" ~C/OH HaC/
~CH 3 (If)
CONCLUSIONS For the preparation of thermally stable epoxy resins, especially for laminating purposes, rigid structures of basic resins with four epoxide groups and rigid types of dianhydrides are promising. Suitable location of the active groups in the molecule (such as the 1,2,6,7- position in a 312
S T R U C T U R E A N D T H E R M A L STABILITY OF EPOXY RESINS
naphthalene molecule) may contribute to twisting effects. Arrangements of active groups as in triazine derivatives or phloroglucinol do not seem to be promising, at least in connection with the normal glycidyl ether structure --O~CH2---CH--CH2--, because curing with anhydrides forms 6-
membered aliphatic links in an arrangement which results in a rather loose network. Closely crosslinked parallel chains are more desirable. Epoxide groups within ring systems reduce the number of atoms in the aliphatic linkages to two with anhydrides and to one with phenols and amines. Structures such as (III) and (IV), therefore, should yield high thermal stability.
An interesting effect is the tendency of maleic anhydride to polymerization during curing and ageing. This, as well as the polymerization through the double bonds of compounds such as butadiene monoxide, dipentene oxide, ~-pinene oxide, allyl glycidyl ether, seems to be due to the catalysing effect of epoxy groups. Thus, epoxy resins with heat distortions much higher than those described here might be obtained by the use of polyvalent, bulky epoxy monomers in conjunction with appropriate amounts of polymerizable, thermally stable 'curing agents and additives.
Materials Central, Wright Air Development Division, Air Research and Development Command, Wright Patterson Air Force Base, Ohio, U.S.A. (Received 21st August, 1959. Revised version received 2nd February, 1960) REFERENCES 1 NARRA¢Orr, E. S. Brit. Plast. 1953, 26, 120-3 2 Fxscn, W. and HOFN^NN, W., J. Polym. Sci. 1954, 12, 497-502 a SI-~cIrrER, L. and W Y N S ~ , J., Industr. Engng Chem. 1956, 48, 86 4 SNECHTER, L., WYNSTR^, J. and KURKJY, R. P. lndustr. Engng Chem. 1956, 48, 94 s DEARBORN,E. C., Fuoss, R. M. and Wnrre, A. F. J. Polym. Sci. 1955, 16, 201 D^MUSIS, A. Paper presented at American Chemical Society Symposium, Atlantic City, September 1956 SMI~, D. D., MURCH, R. M. and PIERCE, O. R. lndustr. Engng Chem. 1957, 49, 1241 8 PEERMAN, D. E., TOLBERG, W. and FLOYD, D. E. Industr. Engng. Chem. 1957, 49, 1091 FLOYD, D. E., P~I~RNAN,D. E., WrrrcOFF, K. J. appl. Chem. 1957, 7, 250 to E m . ~ s , G. F. L. Bull. Amer. Soc. Test. Mat. Feb. 1959, No. 236, 54 11STEPnE~SON, C. E. and WmLaOtnRN, A. H. Bull. Amer. Soc. Test. Mat. Sept. 1957, No. 224, 28
G E R H A R D F. L. E H L E R S le Hooker Electrochemical Company. Bull. No. 43. 1~ WEISS, H. K. lndustr. Engng Chem. 1957, 49, 1089 14 BEHNKE, E. Kunststoff-Rundschau 1959, 6, 217 ~* Becco Chemical Division. Bull. No. 82