J. MATER. CHEM., 1994, 4( 12), 1793-1797 Comparative Kinetic Analyses for Epoxy Resins cured with Imidazole-Metal Complexes Gabriel J. Buist,a Ian Hamerton,*a Brendan J. Howlin," John R. Jones," Shuyuan Liu" and John M. Bartonb aDepartment of Chemistry, University of Surrey, Guildford, Surrey, UK GU2 5XH b Structural Materials Centre, Non-metallics, Defence Research Agency, Farnborough, Hampshire, UK GU74 6TD Proton nuclear magnetic resonance ('H NMR) spectroscopy experiments were carried out for a commercial epoxy resin system in [2H6]DMS0 solution and in bulk. Good agreement was obtained between the secondary rate constants found for the epoxide polyetherification reaction when monitored by Fourier-transform infrared (FTIR) or 'H NMR spectroscopy in the bulk over the range of temperatures studied.This highlights the usefulness of the high temperature bulk NMR technique in kinetic studies of polymerization. Epoxy resins are an extremely versatile family of thermosetting polymers and a wide variety of curing agents may be used in their processing.' The use of imidazoles as curing agents is well and they are known to produce cured products with good physico-mechanical properties. Further- more, at elevated temperatures they exhibit a fast catalytic action upon the process of epoxide homopolymerization (polyetherification), which occurs through the opening of the oxirane ring (Fig. 1). Imidazoles are known to display poor storage stability and, in the past, work has been directed at overcoming this pr~blem.~ Elsewhere we have the use of copper complexes of the adducts of phenyl glycidyl ether (PGE)and 2-ethyl-4-methylimidazole (EMI) (designated 1 and 2, respectively in Fig.1) as curing agents for epoxy resins, which display improved solubility and storage stability enabling their formulation in one-pot compositions. Whilst remaining stable at ambient and low temperature, the complex r 1 dissociates at ca. 120°C to produce the adduct 4. The imidazole-induced cure reaction is depicted in €ig. 2. It involves two consecutive reactions: (i) the activation of the lone pair of electrons on the imidazole ring to atrack the oxirane ring, resulting in the opening of the epoxy group (kl); and (ii) the base (alkoxide) catalysed homopolymerization of the epoxy ring (polyetherification) and subsequent network formation, a first-order reaction under the experimental con- ditions (k2).The properties of cured epoxy resins are very much dependent on the chemical structure and extent of crosslinking within the resin network, hence it is necessary to have a full understanding of the processes of resin network formation. To date many traditional methods have been used to study epoxy cure mechanisms including FTIR' and differential scanning calorimetry (DSC),9 although little work has been carried out into the kinetic aspects of the curing reaction of epoxies catalysed in this manner.Jones el ~1."~" CUCI,1 YN:eO-CH2-CH-CH2-N CH2I CH3 4 1 r 1 /\c b ?H0dHZC-CH-CHZ-0 \ / \ 0-CH2-CH-CH2-0 CH3 CH3ardn 3 Fig.1 Structures of the compounds employed in this study (with 'H NMR designations) J. MATER. CHEM.. 1994, VOL. 4 C?H dCH3T = 120°C ~O-CH,-CH-CH,-N /=(YH 00-CH, -CH-CH,--N CUCI;! -CH, CH,II CH3CH3 -4 41 oligomer formation 4 + epoxy 3 ring-opened intermediate 5 0-YH-y2 YACH3-7 p, ACH3 -NyN;CHp-CH-y CH2-CH-k2 * N CH2-CH--NYFd' c' CHZ CH2 I I CH3 CH3 polyetherif icat ion 5 + epoxy 3 and network formation Fig. 2 Proposed mechanism of epoxy cure involving the metal-imidazole complexes developed a radiochemical tracer method for the study of epoxide cure kinetics that gives accurate, quantitative infor- mation about the reactants, intermediates and products. Bouillon et all2 studied polymerization mechanisms of PGE cured with BF, complexes of aniline using gel-permeation (GPC) and high-performance liquid chromatography (HPLC).However, to date NMR spectroscopy, which is probably the most versatile of all the spectroscopic techniques, does not appear to have been introduced into this area. NMR techniques have improved dramatically in recent years and it is now possible to study complex reaction mixtures. Most dynamic NMR kinetic studies are based on equilibrium systems. However, for an irreversible reaction, it is possible to monitor the reaction quantitatively using the chemical shifts of the reactants, intermediates and products, which can be resolved by integration of the NMR signals as a function of time.The reaction chosen here is that between the copper(I1) complexes of the adduct of PGE and EM1 1 and 2 and a commercial prepolymer of bisphenol-A diglycidyl ether (3). Experimental Equipment 'H NMR spectra were obtained both in [2H6]DMS0 and in the bulk at a range of temperatures by using a Bruker AC-300 high field FT NMR spectrometer operating at 300.13 MHz (a minimum of 64 scans were collected). A ceramic NMR tube spinner and a Bruker B-VT 1000 variable-temperature unit were used for the measurements made at elevated temperatures in the range 120-165 "C (as described elsewhere).6 The sample tube was loaded into the probe after the desired temperature had been reached and measurements made at preprogrammed time intervals using an automated routine with the spectrometer operating in the unlocked mode for bulk (solvent-free) samples and the free-induction decay (FID) signals saved.The ratio of the integrals of the epoxide methylene protons (the doublet at 6 x1.56 corresponding to Hd in Fig. 1) and an internal standard, the methyl proton in the isopropylidene bridge (a singlet at 6 =0.6 corresponding to Ha), was determined and converted to epoxide concen- tration ([El). This ratio was referenced against that of the signals at the beginning of the cure of the MY750 prepolymer (3) to provide a measure of conversion during cure. where I(Hd/Ha) is the ratio of the integral of H, (epoxide) and Ha (methyl group as internal standard) signals, lois the l(Hd/Ha) value at the beginning of the cure and x is the fractional conversion of the epoxide.IR spectra were recorded using a Perkin-Elmer 1750 FT-IR spectrometer interfaced with a Perkin-Elmer 7300 computer; the samples were presented as thin films (ca. 0.025 mm) on KBr plates in a thermostated cell holder. FTIR measurements were made at preprogrammed time intervals using an auto- mated routine with the spectrometer (samples were scanned 24 times at a resolution of 2cm-'). The ratio of the peak heights (in absorbance) of the epoxide ring stretch (914 cm-') and the C-C aromatic ring stretch (1605 cm-' ') as an internal standard, was calculated and converted to epoxide ring con- centration as follows: where A914/1605 is the ratio of the absorbances at 914 cm-I (epoxide) and 1605 cm-' (aromatic ring as internal standard), A, is the A914/1605 value after full curing (assume the conver- J.MATER. CHEM., 1994, VOL. 4 sion is 100% at this stage) and A, is the A914/1605 value at the beginning of the cure. Materials The commercial epoxy prepolymer MY750 (Fig. 1, 3) was provided by Ciba-Geigy. PGE and EM1 were obtained from Aldrich Chemical Co. and purities were determined using 'H NMR. Copper(I1) chloride (Aldrich) gave satisfactory analyt- ical results. The preparation of the 1 : 1 adduct of PGE and EM1 ( PGE-EMI) (4) and the corresponding copper complex, Cu(PGE-EMI),CI, (1), have already been reported else-where' as have the corresponding 2: 1 adduct (PGE,-EMI) and its complex Cu(PGE-EMI2),CI2 (2). The adducts were dispersed directly in the resin, while the complexes (10 mol% in MY750) were dissolved in AnalaR acetone prior to mixing (the solvent being removed under vacuum at room tempera- ture and the resulting mixture injected into a number of 5 mm NMR tubes, these were then stored in a freezer prior to use).Results Solution 'H NMR Study At elevated temperatures (100-140 "C) the viscosity of the epoxy resins is greatly reduced resulting in a dramatic improvement in the resolution of the NMR spectra, especially for those of high concentration. The chemical shifts of the starting materials, intermediates and the products are well resolved (Fig. 3) and hence it is possible to monitor the extent of the reaction by direct integration of these signals.However, because of the limitations imposed by the sensitivity of the 'H NMR method, it is necessary to use a greater acquisition time and it becomes more appropriate to study those reactions with first-order rate constants < s-'. Fig. 3 depicts a stacked plot for the progressive cure of a sample of commercial epoxy prepolymer MY750 (3) initiated 3.5 3 2.5 2 1.5 s Fig. 3 'H NMR spectra of MY750-Cu( PGE-EMI),Cl, in [2H6]DMSO as a function of reaction time at 150"C (Hb not shown) with Cu(PGE-EMI),Cl, (13 mol%) in C2H6]DR/BS0 (as an 80% solution) at 149 "C. In this curing reaction (Fig. 2) protons H, and H, in the initial prepolymer (see dthsignations in Fig.1)are shifted to new positions (H,, and Hd,respectively) as cure proceeds; it is possible to resolve protons €I,, Hd and H,, and Ha in the spectra. Proton Ha does not participate in the reaction and so its chemical shift remains tinchanged (hence it may be used as a reference signal in a qitantitative ratio analysis). The fractional conversion of the resin at each stage of the reaction can be determined from the relative integrations of signals Hb and H, against the hiitial (0% conversion) and infinity ( 100% apparent conversion) reference spectra. Fig. 4 depicts the fractional conversion of epoxy resin referenced to Ha in this manner as a function of reaction time. The rate constant of the cure reaction was deriwd from a linear regression analysis of the logarithm of the relative integrations of Hb, H, and Hb,,respectively, as a function of time (Fig.5). From these plots it is seen that the dala deviate from linearity. This may be attributed to reduced solubility due to polymerization and crosslinking. It is inevitable that some reduction in resolution will result as the reactio 11mixture solidifies and if we neglect those unreliable data from what is effectively the solid state then the correlation coefficients are greater than 0.99. The results show that as expected this is a first-order reaction, k2=4.03 x lop5s-' (at 150 "C). Bulk 'H NMR Study In an earlier publication6 we demonstrated that for d sample containing 10 mol% of curing agent, recorded at a temperature of 163 "C over the course of 10 h, the narrow lines and good 1.OOr -tlmin Fig.4 Fractional conversion derived from ratio of H, (0)and Hd(m)from 'H NMR in ['H6] DMSO at 150"C -1.0 -h m< 5-c -2.0-h I"IE E -3.0-r (61 L 1 1-4.0-20 20 60 100 tlmin Fig. 5 First-order kinetic plot of MY750-Cu( PGE-EMI),Cl, in ['HJ DMSO at 150 "C: (a)In(HJH,), (b)In(H,/H,) -41 . I 0 130 260 390 520 650 t/min Fig. 6 Consecutive first-order plots of MY750-Cu( PCE-EMI)&l2 in (bulk 'H NMR) at 150°C: (a) ln[E], (b)ln(differences). Table 1 'H NMR kinetic data for MY750-Cu(EMI-PGE),C12 at a range of temperatures (in the absence of solvent) T/"C k,"/min -' k/ /min -' 130 0.0134 5.54 x 10-4 135 0.0168 7.68 x 10-4 140 0.0243 1.11 x 10-3 150 0.0269 1.70 x 10-3 163 0.0385 2.59 x 10-3 ~~~~~~~ ~ a k, and k2 refer to reaction steps in Fig.2. I-121 U 0.0023 0.0024 0.0025 1/T Fig. 7 Arrhenius plots of (a) the oligomer formation reaction (k,) and (b)the etherification reaction (k,) resolution make the calculation of epoxide concentration and of the extent of cure simple and straightforward. When the logarithm of the epoxide concentration is plotted against time (Fig. 6) the results appear as two consecutive first-order processes whose rates are sufficiently different so as to be able to calculate both rate constants (Table l), an important advantage of the method over the solution NMR technique (from which only k2 could be calculated). The plot for k, was obtained from the differences between the extrapolation of the linear portion of the h[E] plot and the points between t =O and 110 min.These reactions have first-order rate con- stants (k,=4.48 x s-', kZ= 1.70 x s-' at 150 "C) that are conveniently slow so that the time taken for spectrum acquisition (109 s per spectrum) does not cause a problem. A series of measurements were made at different temperatures enabling Arrhenius parameters to be obtained (Fig. 7). FTIR Kinetics FTIR spectra were recorded for a sample containing 10 mol% of curing agent at a temperature of 140°C over the course of 10h. When the logarithm of the epoxide concentration is plotted against time (Fig. 8) the results appear as two con- J.MATER. CHEM., 1994, VOL. 4 -1.2 hf -1.4 2 v< -C -1.6 -1.8 tlmin Fig.8 Consecutive first-order plots of the cure of MY750-Cu(PGE-EMI),C12 (10 mol%) at 140 "C by FTIR Table 2 Curing reaction parameters obtained from several analytical techniques monitoring curing agent kl/10-6 k,/lO-' technique T/"C (molo/o) S-' S-' 'H NMR 150 13 -4.03 (C2H61DMSO)'H NMR 140 10 4.05 1.85 (fused state) FTIR 140 10 -1.88 secutive first-order processes, but unlike high-temperature 'H NMR spectroscopy, these rates are not sufficiently resolved to be able to calculate both rate constants. The first-order rate constant for the polyetherification reaction (k2= 1.88 x lo-' s-l at 140°C) is in close agreement with that obtained using high-temperature 'H NMR spectroscopy (Table 2).Discussion Tmidazoles are added to epoxy systems to catalyse the homo- polymerization of epoxide groups (polyetherification), but unmodified imidazoles begin to react when mixed with epox- ides (cure occurs slowly at room temperature) making them unsuitable for use in one-pot compositions. Transition metals have been used to prepare complexes of imidazoles that react very slowly at room temperature, but have exhibited a rapid cure at elevated temperatures. The complexation of a trans- ition metal (in this case copper) effectively arrests the poly- etherification reaction at ambient temperature until the tem- perature is elevated to effect cure. A further advantage over existing imidazole curing agents is the efficiency of com-plexation.After a period of time at the cure temperature (e.g. 120"Cor 140"C)the epoxy-complex mixture can be quenched at room temperature with no further advancement of cure.7 This development makes the tailoring of a commercial epoxy system possible thus facilitating more complex cure schedules in a more controllable manner. At present the exact mechan- ism by which the reaction is quenched is unclear. Re-association is a distinct possibility, but it must be borne in mind that the copper plays no further part in the reaction once the cure has been initiated. Hence, the possibility of association with the alkoxide (RO-) groups formed during the reaction (Fig. 2) can not be discounted at this stage.The cure process is summarized by the following scheme: J. MATER. CHEM., 1994, VOL. 4 A4CuC1, 4A + CuC12 A+E*AE AE + ELAEE AEE+E-%AEEE AE ... E,-, + ELAE... E, where A represents the PGE-imidazole adduct (l),E rep-resents PGE and AE, AEE etc. represent successive oligomers. We assume that the dissociation occurs rapidly and plays no part in the kinetics of cure. The k, step involves imidazole- induced opening of the oxirane ring, the k, and succeeding steps involve alkoxide-induced etherification. As a first approximation we assume k, =k, =. . . k, because these steps are essentially the same, and involve oligomers of increasing chain length. On this basis the rate of loss of PGE is given by: -dCEl/d~=k1CAl~El +k,CEI"AEI +[AEE] + .. . + CAE.. .En-,I) Let CAIo represent the initial concentration of adduct, then: [A] =[A10 -(CAE] + [AEE] + . . . +CAE.. .En- 11) Hence, -d CElldt =kl CAlCEl+ k2 CEI(CAl0-CAI) In the initial stage of cure the k2 term is negligible, and d[E]/dt is equal to d[A]/dt: -d[A)/dt=k, [A][E] The epoxide is present in excess, so first-order kinetics are expected for this stage. When all of the adduct has been consumed, the k, term is negligible: -d CElldt =k2 [El CAI0 Again, first-order kinetics are expected, this time because CAIo is a constant. When ln[E] (from the NMR measurements) is plotted against time, the initial points fall on a curve, but the remainder of the plot is linear (Fig. 6). The appearance of the plot suggests that a consecutive first-order treatment is appli- cable.The linear portion was extrapolated back to the ln[E] axis, and the differences between [El calculated from the extrapolated line and [El observed were calculated. The plot of ln(differences) against time is linear (Fig. 6) showing that, to a good approximation, the treatment is justified. The rate constants were calculated from the slopes of the linear plots; the ratio k,:k, is 15 or greater (it varies with temperature), therefore it is reasonable to assume that the epoxide is present in excess throughout the first stage. Arrhenius plots for kl and k2 (Table 1 and Fig. 7) give respective activation energies of 45 and 60 kJ mol-'. Conclusions The results of this work show that 'H NMR spectroscopy is valuable for monitoring the kinetics of polymerization in 'real time', prior to gelation, of a diepoxide with a soluble organometallic catalyst.Experiments were carried out in solution in C2H6]DMS0 and in bulk. There is an advantage in monitoring the reaction in bulk as there is no precipitation of material due to non-compatibility of the polymer with the solvent. This study shows that good agreement can be found between the secondary rate constants found for the reaction when monitored by FTIR or 'H NMR spectroscopy in the bulk (Table 2). Again good agreement is found over the range of temperatures studied and this points to the usefulness of the high-temperature bulk NMR technique in kinetic studies of polymerization.The work of Shuyuan Liu was generously supported by the Structural Materials Centre, Defence Research Agency, Farnborough, Hampshire, UK. The commercial epoxy pre- polymer (MY750) was kindly donated by Mr. Ian Gurnell and Mrs. Debbie Stone of Ciba-Geigy (UK) Duxford, Cambridgeshire, UK. We are grateful to Dr. Y Sun (Robotics Centre, University of Surrey, Guildford, Surrey, UK) for help with graphics presentation. References 1 W. R. Ashcroft, in Chemistry and Technology of Epoxy Resins, ed. B Ellis, Blackie Academic and Professional, London and Glasgow, 1993, ch. 2, pp. 58-59. 2 M. Ito, H. Hata and K. Kamagata, J. Appl. Polym. Sci., 1987, 33, 1843. 3 R. J. Jackson, A. M. Pigneri and E. C. Gaigoci, SAMPE J., 1987, 23, 16. 4 J. M. Barton, Br. Pat., 2135316B, 1984. 5 J. M. Barton, G. J. Buist, I. Hamerton, B. J. Howlin, J. R. Jones and S. Liu, J. Muter. Chem., 1994,4, 379. 6 J. M. Barton, G. J. Buist, I. Hamerton, B. J. Howlin, J. R. Jones and S. Liu, Polym. Bull., 1994,33,215. 7 J. M. Barton, I. Hamerton, B. J. Howlin, J. R. Jones and S. Liu, Polym. Bull., 1994,33, 347. 8 F. Ricciardi, M. M. Jonillie, W. A. Romanchick and A. R. Griscavage, J. Polym. Sci., Polym. Lett., 1982,20, 127. 9 J. M. Barton and P. M. Shepherd, Makromol. Chem., 1975, 176, 919. 10 J. R. Jones, C. Poncipe, J. M. Barton and W. W. Wright, Br. Polym. J., 1986, 18, 312. 11 G. J. Buist, A. J. Hagger, J. R. Jones, J. M. Barton and W. W. Wright, Polym. Commun., 1988,29, 5. 12 N. Bouillon, J-P. Pascault and L. Tighzert, Makromol. Chem., 1990,191,1417. Paper 4/04276J; Received 13th July, 1994