Theoretical study on mechanism of epoxy-amine cure ...

Theoretical Study on Mechanism of EpoxyAmine Cure Reactions 1

2

1

Jan-Eric Ehlers , Nelson G. Rondan , Lam K. Huynh and Thanh N. 1 Truong * 1

Henry Eyring Center for Theoretical Chemistry, Department of Chemistry, University of Utah, 315 South 1400 East, Room 2020, Salt Lake City, Utah 84112. 2

Material Science & Information Research, The Dow Chemical Company, B-1226, Freeport, Texas 77541.

INTRODUCTION Epoxy compounds have a wide range of applications due to their 1 outstanding properties. As all epoxy end-use products require reaction with a curing agent, often an amine, the elucidation of the reaction mechanism is of special interest and has a long history. It was well accepted that the general reaction occurs via a nucleophilic attack of the amine nitrogen on the terminal carbon of the epoxy function. The 2 mechanism has been proven to be a SN -type and thus the reaction rate obeys second-order kinetics (cf. Scheme 1).

COMPUTATIONAL METHODS All electronic structure calculations were carried out using the 16 Gaussian 03 program package. A hybrid non-local Density 17 Functional Theory B3LYP level of theory with the 6-31G(d,p) basis set has been used for locating stable structures for reactants (R), transition states (TS) and products (P). Frequencies of the stationary points were also calculated at the same level of theory. The reaction pathways from reactants to products were also verified by Intrinsic 18 , 19 Reaction Path (IRC) calculations. Solvation models such as the 20 Polarizable Continuum Model (PCM) and Conductor-like PCM 21 , 22 (CPCM, COSMO) were used for selected reactions, e.g reactions going through TS 7 and 12 (cf. Figs. 1 & 5). The equilibrium constants and rate constants for selected reactions were also carried out at the Transition State Theory (TST) level employing the kinetic module of the web-based Computational 23 Science and Engineering Online (CSE-Online) environment. RESULTS AND DISCUSSION Model System. Fig. 1 shows molecular structures of the reactants, complexes, and transition states of the reaction involving 2methoxymethyloxirane, methylamine and propan-2-ol as a model system. This model system is believed to provide a reasonable description for the first polymerization step in epoxy-amine curing reactions, in which propan-2-ol represents both a model hydrogendonating catalyst and a product hydroxyl group and methylamine a curing agent.

Scheme 1. General accepted mechanism of epoxy-amine curing. The first step is assumed to be the rate determining step (r.d.s.) whereas at which the proton transfer is fast compared to the nucleophilic attack. The sign | denotes the transition state. The reaction may involve both primary and secondary amines as 2 well as hydrogen bonding catalysts and promoters. Narracott Chapman, Isaacs and Parker were amongst the first researchers suggesting several possible reaction pathways that may occur during 3 the curing process and evaluating them according to their importance. Shechter and coworkers reported the importance of catalysis caused 4 by hydroxyl groups and Smith extended his studies to any potentially 5 catalytic impurity. Horie and co-workers included the case in which the primary amine reacts twice with an epoxy molecule to form a tertiary 6 amine in their investigations and Mijovic and coworkers considered 7 hydrogen bond complexes with reactant amine molecules as possible. These models have often been cited and used during the past decades. They have been refined or slightly altered or extended up to 8,9 present time, especially the Horie model. Other approaches were 10 also applied, however most studies are based on a kinetic model with rate parameters fitted to data from thermometric measurements to 11 , 12 elucidate the mechanism. A major disadvantage of this deduction is that it cannot examine individual elementary reaction; the system can only be examined as an ensemble. A prime example is the possibility of cyclic ter-molecular transition states with an epoxy and 13 two amine molecules, described in a review by Rozenberg. A cyclic and acyclic transition state of the same stoichiometry cannot be distinguished by thermometric measurements although their reaction pathway for this specific reaction differs considerably; the amine addition via a cyclic transition state is believed to complete in one step whereas for the acyclic pathway occurs via an intermediate and an extra proton transfer step is necessary. Due to the advantage of the hydrogen-bond network the reaction pathways involving cyclic transition states are often considered as favorable pathway in 14 , 15 literature. Theoretical calculations are able to elucidate the epoxy-amine cure mechanism. Although it is not possible to model a realistic polyaddition reaction in bulk phase, using the continuum solvation model in addition to investigation of gas-phase reactions provides a viable approach to understand the chemistry more detailed than experimental methods allow at present time.

Figure 1. Molecular structures of the stationary-point complexes. ROH denotes propan-2-ol, RNH2 for methylamine and E for methoxymethyloxirane; HB for hydrogen bond complex, TS for transition state and c for cyclic complex. Precursor Hydrogen Complexes. Complexes 1 to 4 (cf. Fig. 1) show hydrogen-bonding complexes between two aliphatic alcohol functional groups 1, two aliphatic amine functional groups 2, an alcohol group and an epoxy oxygen 3 and between an alcohol group and an epoxy oxygen 4, respectively. It is found that the OH...O hydrogen bond in 1 is considerably stronger than the NH...N bond in 2, resulting in a shorter bond length and higher equilibrium constants. The

Proceedings Published 2006 by the American Chemical Society

calculated equilibrium constant at 298 K of 2.45 L/mol is in good 24 agreement with experimental values (2.45 L/mol). The trend for complex 1 and 2 can also be recognized comparing hydrogen complexes 3 and 4. The interaction with an epoxy oxygen results in a slightly smaller binding energy. The equilibrium constant for complex 3 at 298 K of 1.35 L/mol can be compared to an experimental value for the model system isopropanol/phenylgycidyl ether (experimental value 24 1.08 L/mol). Acyclic Transition States. Geometries of possible acyclic TSs’ were given in Fig. 2. It is found that the TS of the attack of an amine on an epoxy molecule is stabilized by both a hydrogen bond between the amine hydrogen H1 and the ether oxygen O2 and a hydrogen bond between the epoxy oxygen O1 and a hydrogen donor. Comparing the TS geometries from 5 to 7, the self-promoting reaction 6 and the alcohol-catalyzed reaction 7 show more reactant-like transition state geometries (e.g. the atom distance C1-O1 and the bond angle O1-C2C1) leading to smaller calculated barrier heights (102, 76, 62 kJ/mol for 5, 6 and 7, respectively). This indicates the hydrogen-bond with the epoxy oxygen playing an important role in lowering the barrier height. From experimental studies Mijovic reported rate constants are in -3 -2 -1 -1 the order of between 10 and 10 (Lmol min ) at 393K (120°C) for both primary and secondary amines for the alcohol-catalyzed reaction 7, 25 of phenylglycidyl ether and aniline. Matejka obtained the same order of magnitude for the system diglycidylaniline aniline at 373K 26 , 27 (100°C). Comparing these values with the calculated rate constant -3 of 3.92x10 at 400K for the alcohol-catalyzed reaction justifies the adequacy of the model system and accuracy of the level of theory. The calculated activation energy Ea values for the catalyzed reactions from 61 to 76 kJ/mol are in the same range of experiments 9,11, 28 from 53-65 kJ/mol whereas for the un-catalyzed reaction it is not. However, it should be mentioned that it is hardly possible to obtain impurity-free reactants that may catalyze a reaction. More important, the cure reaction is self-accelerated by its product hydroxyl groups and an auto-catalysis cannot be ruled out as well as a self-promotion of the amine thus an un-catalyzed reaction is not likely exist in nature. In addition it should be noted that the gas phase used in computation does not simulate real conditions accurately.

(complex 3 leading complex 7) is supposed to be the most probable pathway whenever an alcohol catalyst is present.

Scheme 2. Un-catalyzed, self-promoted and alcohol-catalyzed pathway for the epoxy-amine polymerization. IM abbreviated for intermediates and P for products. Cyclic Transition States. The suggestion of cyclic TSs in literature justifies a more detailed investigation. Cyclic transition states exhibit a one-step mechanism in which the bond cleavage and bond formation of the epoxy-amine system can occur simultaneously with a hydrogen transfer, thus directly leading to a product as shown in Scheme 3. Though, comparing the acyclic with the cyclic transition states the latter exhibit higher steric crowding and lack the etheroxygen amine-hydrogen bond (H1-O2 in Fig. 2). Geometries of the cyclic TS were given in Fig. 3

Scheme 3. Comparison of acyclic and cyclic transition state reaction pathways.

(a)

(b)

(c)

Figure 2. Transition states for (a) the non-catalyzed pathway 5 TS_RNH2-E; (b) the self-promoted pathway 6 TS_RNH2-E-H2NR; and (c) the alcohol-catalyzed pathway 7 TS_RNH2-E-HOR. Key: dark gray, carbon; light gray, hydrogen; red, oxygen; blue, nitrogen. A direct comparison of the data of the un-catalyzed complex 5 with the alcohol-catalyzed complex 7, illustrated in the middle and lower row of Scheme 2 leads to the suggestion that hydrogen-donating impurities in the epoxy-amine mixture are able to accelerate the overall curing reaction rate, with a greater impact at lower temperatures. The same statement is almost true for the self-promoted reaction pathway which is illustrated in the upper row of Scheme 2. However, although the energy barrier is considerably lower than for the un-catalyzed pathway 5, the formation constants Kc for the hydrogen precursor complexes have to be taken into account. Complex 4 which leads to the TS complex 6 is more unlikely to form than complex 3 by almost two orders of magnitude. From these circumstances it may be deduced that the un-catalyzed pathway (complex 5) and the self-promoted pathway (complex 4 leading to complex 6) are occurring parallel in absence of any added catalyst whereas the alcohol-catalyzed pathway

(a)

(b)

(c)

Figure 3. Cyclic transition states for (a) the non-catalyzed pathway 8 TS_c-RNH2-E; (b) the self-promoted pathway 9 TS_c-RNH2-E-H2NR and (c) the alcohol-catalyzed pathway 10 TS_c-RNH2-E-HOR. In Fig. 3, the complex 8 represents a four-member cyclic transition state. The N1-C1 bond formation, C1-O1 bond cleavage and O1-H1 bond formation can be shown to be a concerted mechanism by Intrinsic Reaction Coordinate (IRC) calculations. The TS geometry indicates a late transition state relative to both the other cyclic and


acyclic complexes, together with increased steric crowding resulting in a high energy barrier. The large calculated barrier of 181 kJ/mol for this pathway seems to be consistent with previous suggestion that such a 29 complex is not reactive . Complexes 9 and 10 represent the selfpromoted and alcohol-catalyzed pathway with the calculated barrier height of 149 and 140 k/Jmol, respectively. Both complexes are relatively similar to each other. The actual reaction, i.e. bond cleavage and formation, may preferably occur in a four-center ring structure, involving O1, C1, N1 and H3, with a H3-O1 atom distance of 2.151 Å and 2.234 Å for complex 9 and 10, respectively. Scheme 4 clarifies the reaction mechanism and demonstrates that in this case the alcohol and second amine molecules, respectively, are merely spectator molecules. Figure 4. Comparison of the influence of different solvents on the relative absolute energy for the reactant species methylamine (A), epoxy-propan-2-ol complex 3 (HB_ROH-E) and the resulting transition state 7 (TS_RNH2-E-HOR).

Scheme 4. In a six-member cycle the actual reaction may occur in a four-member cycle. Comparing the cyclic with the acyclic pathways suggests that possible cyclic reaction pathways play little role in the epoxy-amine curing reactions. The cyclic arrangement forces an unfavorable nucleophilic syn-attack, resulting in steric crowding and hence a considerably higher activation barriers than for acyclic transition states, leading to reaction rates several orders of magnitude lower than for the acyclic species. In addition to high activation barriers the activation entropy argues against the formation of the cyclic complexes. Influence of Solvation. The influence of solvation was studied by employing the CPCM solvation model for three different solvents; benzene represents nonpolar solvents, tetrahydrofuran medium polar and methanol highly polar solvents. It was found that increasing solvent polarity causes the transition state moving toward the entrance channel and thus leading to lower barrier. In addition, a high polar solvent shields the solute oxygen and carbon charges from each other, hence lowering the Coulomb attraction force. The same effect also stretches the forming N-C bond and the hydrogen bonds, especially the weaker H1-O2 bond. The effect of solvation on both reactants and transition state is pointed out in Fig. 4. It displays their solvation free energies as functions of the dielectric constant εr. Since the rate determining step of the epoxy-amine curing reaction is a charge separation type II SN2 reaction and thus as solvent polarity increases solvation free energies of the TS increases (in the absolute magnitute, more negative in Fig. 4) faster than those of reactants. Consequently, solvent effects lower the activation energy. These results are consistent with earlier studies on 30 this topic. Primary versus Secondary Amines. There exists much controversy about the ratio of rate constants for primary and secondary amines. Most authors agree that the ratio is somewhere between 0.5 31 (e.g. Horie kinetic model) and unity. Major effects that may decrease the reaction rate were believed to be: 1) the steric effects caused by bulky substituents and 2) the substitutions effects. In the latter case it is assumed that a secondary amine has different electronic effects compared to that of primary one.

Figure 5. Transition state for the alcohol-catalyzed pathway of a secondary amine 12 TS_R2NH-E-HOR. Table 1. Comparison of the Influence of Solvation Between Primary and Secondary Amines 7 TS_RNH2-E-HOR 12 TS_R2NH-E-HOR gas phase CPCM/methanol gas phase CPCM/methanol 394.5 482.8 397.5 462.0 ν| d(N1-C1) 2.002 2.146 2.025 2.144 d(H1-O2) 1.994 2.181 2.011 2.132 d(H2-O1) 1.712 1.733 1.712 1.737 d(C1-O1) 1.951 1.845 1.934 1.849 d(C2-O1) 1.380 1.403 1.383 1.403 d(C1-C2) 1.476 1.461 1.475 1.462 86.1 80.2 85.2 80.4 ∠(O1-C2-C1) 110.7 108.1 111.1 108.5 ∠(C2-C1-N1) kfwd (300 K) 3.06x10-6 5.55x10-3 1.42x10-5 1.78x10-4 -3 -2 kfwd (400 K) 3.92x10 2.28 1.31x10 1.13x10-1 61.6 51.0 58.3 53.5 ΔV | gas phase 4.6 3.3 k12/k7400 MeOH 0.032 0.050 k12/k7400 In contrast to these assumptions the data of Table 1 suggests that solvent effects may be a reason for secondary amines to react slower than primary species. While the barrier height in gas phase is lower for secondary amines it is the opposite in a polar solvent like methanol. At this it is unambiguous that this effect arises from the solvent as the difference in steric demand is minimal for methyl- and dimethylamine, Fig. 5 depicts the transition state which shows that steric effects play a 32 minor role, and the substitution effect is rather small. The inclusion of solvation makes a reasonable comparison with experimental values possible. Although the rate constant ratio from calculation is much smaller than reported in experiments due to a different reaction environment than in real cure reactions (e.g. a less polar environment) the calculations qualitatively correct predict that secondary amines react slower than primary amines. Since the barrier height is underestimated at a DFT/B3LYP level of theory and the dielectric constant of methanol might be greater than that of a bulk cure reaction


in an experiment the rate constants are overestimated. Nevertheless, both factors can be improved. CONCLUSIONS The epoxy-amine curing reaction has been examined by using a system of three model compounds and computational methods at DFT/RB3LYP level of theory. The model system clarifies the importance of catalytic compounds for the curing reaction. Hydrogen precursor complexes have been examined that are involved in the actual reaction. The polymerization reaction can be divided into noncatalytic, self-promoted and alcohol-catalyzed types as well as cyclic and acyclic transition states. It is assumed that these pathways describe the major reaction types for this reaction. It emerges that the acyclic alcohol-catalyzed pathway exhibits the lowest activation barrier and consequently shows the fastest reaction rate of all examined pathways. It is noteworthy that the curing reaction generates a hydroxyl-group from the epoxy oxygen so that the polymerization is accelerated with proceeding conversion and thus the alcohol-catalyzed pathway always dominates in later stages of conversion. In absence of a catalyst the curing reaction may proceed via both the acyclic noncatalytic and self-promoted pathway at which the latter one exhibits the faster reaction rate. Due to auto-acceleration of the reaction by product hydroxyl-groups these pathways are only important at early stages of conversion. Although cyclic transition states exhibit the advantageous feature of direct proton transfer they are not of great importance for the curing reaction because of their high activation barrier due to steric crowding and entropy of activation since three molecules have to form a cycle. It is noteworthy that the reaction may favor a four-member cycle instead of the suggested six-member cycle in earlier studies, probably because this pathway involves only one proton transfer instead of two (cf. Scheme 3 with Scheme 4). Investigations on the influence of solvation have also been made. The magnitude of the activation barrier of the reaction shows a dependence on the polarity of the solvent since two charges are separated during the nucleophilic attack of the amine molecule; their Coulomb attraction which counteracts on the reaction is partially shielded by solvent molecules. Though, the observed effect is negligible when going from medium polar (e.g. tetrahydrofuran) to high polar solvents (e.g. methanol). Although solvation obviously has an observable effect on the reaction it must be emphasized that for the prediction of possible reaction pathways gas phase calculations yield qualitatively correct results so that solvation effects may be neglected in order to decrease calculation time. Finally, it is found that an increase steric crowding and substitution effects are considered to be the main influence on the reaction rate; however computational results show that another influence on the reaction rate decrease arises from solvation effects. ACKNOWLEDGEMENT This work is supported in part by the National Institutes of Health (Grant # NCRR 1 S10 RR17214-01) for computer time. J. E. is grateful to the German Academic Exchange Service for a scholarship for the exchange between the Chemistry Departments of the Technische Universität Braunschweig, Germany and the University of Utah, USA. The authors would like to thank the Utah Center for High Performance Computing for computer time support.

1 2. 3. 4. 5. 6.

7. 8. 9. 10.

11. 12. 13. 14. 15.

16.

17. 18. 19. 20. 21. 22. 23. 24. 25 26. 27.

28. 29. 30.

31. 32.

Mijovic, J.; Fishbain, A. Wijaya, J., Macromolecules 1992, 25(2), 979. Cole, K. C. Macromolecules 1991, 24(11), 3093. Blanco, M.; Angeles Corcuera, M.; Riccardi, C. C.; Mondragon, I., Polymer 2005, 46(19), 7989. e.g. an approach similar to the Two-Fluid-Model for glassy state transition: Vinnik, R. M.; Roznyatovsky, V. A. J Therm. Anal. Calorim. 2003, 73(3), 807. Paz-Abuin, S.; Pellin, M. P.; Paz-Pazos, M.; Lopez-Quintela, A. Polymer 1997, 38(15), 3795. Zvetkov,V. L. Macromolecular Chemistry and Physics 2002, 203, 467. Rozenberg, B. A. Adv. Polym. Sci. 1985, 75, 113. Noyori hydrogenation - Noyori, R.; Hashiguchi, S. Accounts. Chem. Res. 1997, 30(2), 97. Meerwein-Ponndorf-Verley reaction - Meerwein, H.: Schmidt, R. Ann. 1925, 444, 221-38. Ponndorf, W. Angewandte Chemie 1926, 39, 138-143. Verley, A. Bulletin de la Societe Chimique de France 1925, 37, 537. Frisch, M. J.; Trucks, G. W.; Schlegel, H. B.; Scuseria, G. E.; Robb, M. A.; Cheeseman, J. R. and co-workers Gaussian 03, Revision C.02, Gaussian, Inc., Wallingford CT, 2004. Becke, A. D. Journal of Chemical Physics 1993, 98(7), 5648. Gonzalez, C.; Schlegel, H. B. J. Chem. Phys. 1989, 90(4), 2154. Gonzalez, C.; Schlegel, H. B. J. Phys. Chem. 1990, 94(14), 5523. Miertus, S.;Scrocco, E.;Tomasi, J. Chem. Phys. 1981, 55(1), 117. Barone, V.; Cossi, M. J. Phys. Chem. A 1998, 102(11), 1995. Cossi, M.; Rega, N.; Scalmani, G.; Barone, V. J. Comp. Chem. 2003, 24(6), 669. http:///www.cse-online.net references cited in Rozenberg, B. A. Adv. Polym. Sci. 1985, 75, 113. Mijovic, J.; Wijaya, J. Macromolecules 1992, 27(26), 7589. Matejka, L.; Dusek, R. Macromolecules 1989, 22(7), 2911. It is important to mention that the alcohol-catalyzed reaction is assumed as ter-molecular reaction resulting in a third-order rate 2 -2 -1 constant (L mol s ). In contrast, in this study it is assumed that the epoxy and alcohol molecule react in a pre-equilibrium (cf. -1 Scheme ) which results in a second-order rate constant (Lmol s 1 ). This difference might of course make it questionable if the rate constants are comparable at all with each other. Vinnik, R. M.; Roznyatovsky, V. A. J Therm. Anal. Calorim. 2004, 76(1), 285. Vinnik, R. M.; Roznyatovsky, V. A. J Therm. Anal. Calorim. 2003, 73(3), 819. Vedenyapina, N. S.; Kuznetsova, V. P.; Ivanov, V. V.; Zelenetskii, A. N.; Rakova, G. V.; Plokhotskaya, L. A.; Ponomarenko, A. T.; Shevchenko, V. G.; Enkolopyan, N. S. Izvestiya Akademii Nauk SSSR, Seriya Khimicheskaya 1976, 9, 1956 e.g. Grillet, A. C.; Galy, J.; Pascault, J. P.; Bardin, J.; Polymer 1989, 30(11), 2094-2103. as an indicator: pKa = 10.66 for methylamine and pKa = 10.73 for dimethylamine..

REFERENCES Harper, C. A. Handbook of Plastics, Elastomers and Composites. 2nd ed.; McGraw-Hill: New York, 1992, p. 10.55. Narracott, E. S. British Plastics and Moulded Products Trader 1953, 26, 120. Chapman, N. B.; Isaacs, N. S.; Parker, R. E. Journal of the Chemical Society, Abstracts 1959, 1925. Shechter, L.; Wynstra, J.; Kurkjy, R. P. J. Ind. Eng. Chem. (Washington D. C.) 1956, 48, 94. Smith, I. T. Polymer 1961, 2, 95. Horie, K.; Hiura, H.; Sawada, M.; Mita, I.; Kambe, H. J. Polym. Sci. Pol. Chem.: Part A-1 1970, 8(6), 1357.
