Polymer International 40 (1996) 63-71 Proton NMRStudies of Molecular-Level Mobility in Vulcanised and Epoxidised Natural Rubber Robin K. Harris,* Barry J. Say Department of Chemistry, University of Durham, South Road, Durham DH1 3LE, UK & Soon Ng Department of Chemistry, University of Malaya, Pantai Valley, 50603 Kuala Lumpur, Malaysia (Received 12 October 1995; revised version received 21 November 1995; accepted 22 December 1995) Abstract: Proton spin-lattice relaxations in the laboratory and rotating frames, together with transverse relaxation (free induction decay), have been measured for natural and epoxidised rubber, plus a series of vulcanised samples. The data have been analysed by fitting procedures to give characteristic times which are discussed in terms of the mobility of the various parts of the samples, and of their compositions (including the effects of parafin wax and oil additives). Interpretation has been aided by recording free induction decays following a period of spin-locking. Carbon-13 magic-angle spinning spectra have also been measured. K e y words: NMR, relaxation, rubber, vulcanisation, epoxidisation, additives, mobility. INTRO D UCTlON mation on network structures has been obtained, viz. accelerated sulphur vulcanisation results in primarily polysulphidic structures with sulphurisation allylic to the double bond, cis-trans isomerisation is common, and small-to-moderate levels of double-bond migration can O C C U ~ . ~ In this research a proton NMR relaxation study has been carried out on NR, carbonxarbon crosslinked NR, and epoxidised NR (ENR 50), as well as various vulcanised rubbers, cured with and without paraffin additives. Produced by the chemical modification of NR,6 ENR is now an established commercial polymer with a unique set of properties (high strength due to an ability to undergo strain crystallisation, an increased glass transition temperature and an increased solubility parameter) which are reflected in the vulcanisates with increased oil resistance, enhanced adhesive properties, a high degree of damping, and reduced gas p e r m e a t i ~ n . ~ The paraffin additives in the ENR formulation provide ozone protection. The chemical structures of NR and ENR are shown below. Sulphur vulcanisation of rubber results in the formation of a network structure consisting of crosslinks, network chains (macromolecular segments bound at each end by a crosslink) and chain ends (macromolecular segments bound at one end only to a crosslink).' The relation between density of crosslinks and density of network chains depends on the crosslink functionality and the occurrence of chain entanglements. The proportion of chain ends increases when chain scission occurs during vulcanisation.'*2 Both dicumyl peroxide (DICUP) as a curative3 and gamma irradiation4 produce a network in natural rubber (NR) that contains only carbon-carbon crosslinks and is accompanied by cis-trans isomerisation, double-bond migration and main-chain scission. As expected, crosslinking induces motional restrictions in the polymer chain. From various experimental techniques, especially solid-state I3C NMR, definitive infor- * To whom correspondence should be addressed. 63 Polymer International 0959-8103/96/$09.00 0 1996 SCI. Printed in Great Britain R . K . Harris, B. J . Say, S . N g 64 I Y NR IS 8 ENR A list of possible structures occurring upon sulphurisation of cis-polyisoprene' is shown below. I 5, I (hans) The sites of sulphurisation in ENR are the same as in NR, i.e. at the methyl and methylene carbons allylic to the double bond.g The double bonds may be isolated, having epoxide groups on either side, or have one adjacent epoxide group, the latter situation being twice as frequent as the former in the case of ENR 50. Earlier 'H NMR studies of crosslinked rubber used linewidths to show the existence of motional restriction^,^.'^ low-frequency interactions in the kHz range,' and moderately rapid segmental motions." Koenig et al. have investigated' by solid-state I3C NMR relaxation techniques the effect of crosslinking on the molecular dynamics, and the results are summarised as follows. The V-curves in the plot of nT, versus 1/T for all polyisoprene carbon signals and for the detectable - signals arising from the network unit (carbons in polyisoprene units containing sulphur bridges, transpolyisoprene and cyclic sulphide structures) exhibit similar general trends with increased curing time, namely curve broadening, shift of the minima to higher temperatures, and increase of the Tlminvalues. All the backbone carbons show quantitatively similar effects of crosslinking on the spin-lattice relaxation. In lightly crosslinked NR, motional restrictions are more pronounced for the methyl group than for the main-chain carbons. The experimental data suggest that crosslinking affects isotropic motion much more than librational motion but does not influence rotational motion. The data also indicate qualitatively that concepts of multiple correlation times and/or a distribution of correlation times are applicable. Krejsa & Koenig have shown14 that in accelerated sulphur-vulcanised cis-polyisoprene, the proton T2 relaxation process exhibits biexponential behaviour, consistent with a slow motion related to crosslinked regions and a fast motion related to mobile regions. The correlation time difference between these regions is roughly an order of m a g n i t ~ d e . ' ~ Nine samples have been studied in this research. The basic materials are a natural rubber (NR) and the epoxidised derivative ENR 50. Samples of NR which underwent vulcanisation with 2.5 and 0-3 parts of sulphur to TABLE 1. Composition of vulcanisates (in parts by weight) and glass transition temperatures Constituent NR ENR 50 Zinc oxide Stearic acid Sulphur Permanax TO" Santoflex 13' MOR" CBSd DICUP" TMTD' Paraffin wax Paraffin oil T,("C) ' VNR-I 100 VNR-II 100 4 1 2.5 0.5 4 1 0.3 0.5 0.6 3.0 CNR VENR JB/27/4 JB/27/5 JB/27/6 100 5 2 1.2 100 5 2 1.5 3 3 100 5 2 1.5 3 3 100 5 2 1.5 3 3 1.5 1.5 1.5 100 4 1 0.5 2.0 3.0 2.0 -28.5 4 5 -25.8 4 10 -30.8 4 20 -30.9 Permanax TO (antioxidant) : Poly-2,2,4-trimethyl-I ,2-dihydroquinoline. Santoflex 13 (antiozonant) : N - ( I ,3-Dimethylbutyl)-N'-phenyl-p-phenylenediamine, MOR : 2-Morpholinothiobenzothiazole. CBS: N-Cyclohexylbenzothiazole-2-sulphenamide. DICUP: Dicumyl peroxide. This causes direct C-C crosslinking of the polyisoprene NR chains. TMTD: Tetramethylthiuram disulphide. POLYMER INTERNATIONAL VOL. 40, NO. 1. 1996 Molecular-level mobility of vulcanised and epoxidised natural rubber 100 parts of rubber (VNR-I and VNR-I1 respectively), together with a C-C crosslinked material (CNR), have also been examined. Sample VNR-I1 probably contains mainly monosulphide links between polyisoprene chains, whereas VNR-I will have polysulphide links. The epoxidised rubber was also subjected to vulcanisation (sample VENR, with 1.2 parts of sulphur per hundred parts of ENR 50). The final three vulcanised epoxidised samples contain paraffin wax, plus oil in various proportions. All materials except NR and ENR 50 contained a number of additives (see Table 1). The aim of this study is to evaluate via NMR relaxation the effects of epoxidisation, vulcanisation and of wax and oil additives on the mobility regimes of natural rubber. Three types of proton relaxation have been measured, viz. spin-lattice relaxations in the laboratory and rotating frames (TI and TIPrespectively) and transverse relaxation (via free induction decays). These processes were mathematically analysed in the case of TIP and FID to yield multicomponent fits. Carbon-13 magic-angle spinning spectra were also acquired as checks on sample composition. EXPERIMENTAL Preparation of materials The samples of natural rubber and vulcanisates were obtained from the Rubber Technology Centre of the Rubber Research Institute of Malaysia, Kuala Lumpur. They were prepared according to standard procedures' and the composition of each sample is shown in Table 1. The materials were not specially made for this work, and therefore contain a range of constituents as for commerical use (see Table 1). However, the only difference between the three samples containing parafin wax is in the proportion of paraffin oil. The natural rubber (NR) is of SMR-L grade (Standard Malaysian Rubber-light coloured). Sample ENR 50 is 50 mol% epoxidised. The cure temperature was 140°C for all vulcanisations, and the minimum cure time was 60 min (except for sample CNR, for which it was 80min). Nuclear magnetic resonance Proton relaxation measurements were carried out using a purpose-built spectrometer operating at 60 MHz, which has been described in a previous publication.'' The glass NMR tube was of 7.5mm 0.d. The samples were difficult to manipulate at room temperature and were therefore placed in the tube in a frozen state (liquid nitrogen temperature), being subsequently warmed to ambient probe temperature for the NMR experiments. All measurements were carried out at 298 K f 1 K (the temperature of the house air supply). POLYMER INTERNATIONAL VOL. 40, NO. 1, 1996 65 Spin-lattice relaxation times (TI) were obtained by the inversion-recovery method, while spin-lattice relaxation times in the rotating frame (TIP)were measured by observing (following a solid echo 90-2-909,) the decay of initial magnetisation generated by a 90" pulse (duration 1.5ps) and then spin-locked by an r.f. magnetic field of strength corresponding to 4 0 k H ~ . ' ~ .As '~ the resulting decays were not single-exponential in the case of Tip, the times of observation were chosen so as to afford the best characterisation of the decay, i.e. more closely spaced in the initial, rapidly decaying part. Transverse magnetisation relaxation was investigated, either by observation of the F ID following a solid echo (in the cases where decay was so rapid as to avoid any effects of B , inh~rnogeneity)'~,'~ or by a combination of F I D and Carr-Purcell-Meiboom-Gill (CPMG) meas u re me n t~ ,'~ where there was both a rapidly decaying part, and longer-lived components ( > 2 ms). All quoted measurements were carried out with a delay between successive cycles of the relevant pulse sequence equal to at least 10 times the longest TI component in order to ensure full relaxation of all spins. Typically 16 points were used for Tl plots, 112 for Tlp and 1024 for FIDs, each point being the average of 8, 16 and 32 co-added determinations respectively. The B , field was adjusted to observe the signals as close as possible to resonance as judged from lack of oscillations on the 'imaginary' receiver channel. Quadrature data pairs were collected simultaneously, allowing phase correction of the averaged signals. Data analysis was carried out by nonlinear least squares fitting as described in a previous publication,' involving Gaussian, Weibullian and exponential functions in the case of FIDs. The Weibullian function" used here is exp[-t/T')"] where n can take any value: n = 1 gives an exponential, n = 2 a Gaussian, 1 < n < 2 must be intermediate, and n < 1 looks like an exponential whose time constant increases with time (whence the term stretched exponential). All relaxation times are quoted to two significant figures and fractions to half a percentage point. It must be remembered that fitting to multiexponentials is an ill-conditioned situation, and that there is a high covariance between parameters, so the exact values quoted do not necessarily represent any intrinsic properties of the system. Care should be taken particularly with the values associated with the Gaussian components. The magnitudes of the populations are significant to the quoted accuracy but the time constants cannot be interpreted to generate second moments. Carbon-13 spectra were obtained at high resolution using magic-angle spinning (typical rates c. 3 kHz) and high-power proton decoupling. A Varian VXR 300 spectrometer operating at 75.43 MHz for carbon and at ambient probe temperature was employed, together with a Doty probe accepting 7mm 0.d. rotors. Singlepulse sequences were used, with 90" pulse angles and recycle delays of either 1 s or 30s. The number of FIDs R. K . Harris, B. J . Say, S . N g 66 that at 6, = 26.9ppm to C-9 and C-15, with the shoulder between these signals attributable to C-8 (see Ref. 20). Addition of paraffin wax and paraffin oil produced further resonances in the aliphatic region, as expected. In the case of the sample with most paraffin oil (JB/27/ 6) (Fig. lb) there was also a small decrease in linewidth of the ENR 50 lines (with an incipient splitting of the signal at 6, 135), and the lines not present in ENR 50 (and therefore assigned to the paraffin oil) were notably sharp-significantly sharper than for JB/27/4 and JB/ 27/5. There is some evidence here that the oil agglomerates do give more mobile domains when present in high quantities, which is also evident in the proton relaxation data (see below). Our relaxation data are reported in Tables 2 and 3. The latter includes results for pure paraffin wax and paraffin oil. The first row in Table 3 is the same as the last in Table 2, and is included for comparison purposes. Spin-lattice relaxation is, in all solid cases, well represented by a single exponential. Fitting to two exponentials was attempted in several cases, but the quality of fit was not sufficiently improved to warrant usage of the data. Of course, the paraffin oil, being liquid and therefore having inefficient spin diffusion, did not have single-exponential Tlrelaxation. The values of TI for the vulcanised (but not epoxidised) samples do not significantly differ from that of natural rubber itself, indicating relatively light crosslinking. However, epoxi- accumulated for each spectrum varied between 50 and 2000. RESULTS AND DISCUSSION The proton-decoupled 13C NMR spectrum of the natural rubber sample consisted of sharp signals (linewidths c. 0-5ppm) at chemical shifts 6, = 135.1, 125.6, 32.7, 27.0 and 23.8ppm, assigned to C-4, C-3, C-1, C-2 and C-5 respectively. There were also a few peaks of minor intensity. The spectrum of the epoxidised form (ENR 50) was more complex and showed substantially broader lines (Fig. la). A new signal appeared at 6, = 22.8ppm. There were no discernible differences on vulcanisation (VENR). Increase in the recycle delay from 1 s to 30s increased the relative intensities of the lines at 6, = 135.1, 60.3 and 22.8 ppm, presumably because these correspond to carbons with long spin-lattice relaxation times. These carbons are assigned to C-7 (quaternary), C-11 (quaternary) and C-12 (mobile methyl) respectively. The epoxy methyl (C-12) would appear to have a longer Tl than the isoprenic methyl (C-13) (assumed to give a signal indistinguishable from C-5) and is therefore presumably more mobile (lower barrier to C-CH, internal rotation). The signals from the quaternary and methyl carbons were appreciably sharper than the rest. The CH, carbons did not give clear single lines; the peak at 6, = 32.5ppm can be assigned to C-14, and nA (a) I 150 - I I I I I 100 I I I I I I I I I I 0 50 6c /PPm Fig. 1. Carbon-13 magic-angle spinning NMR spectra with high-power proton decoupling at 75.4MHz and ambient probe temperature of (a) epoxidised natural rubber (sample ENR SO), and (b) sample JB/27/6, which contains added paraflin wax and paraflin oil. Spectrometer operating conditions: Repeated single-pulse operation with 90" pulse angle and 1 s pulse delay; number of transients 2000 for (a) and 1280 for (b); spin rate 2930 Hz. POLYMER INTERNATIONAL VOL. 40, NO. 1, 1996 67 Molecular-level mobility of vulcanised and epoxidised natural rubber TABLE 2. Proton-spin relaxation data for natural rubber and derivative materials 72 Natural rubber (NR) Epoxidised natural rubber (ENR 50) 250 18 8.7 8.7 1.2 0.48 0.21 90 11 3.5 (79.5%) (20.5%) (1.5%) (7.5%) (42.5%) (48.5%) (1 .O%) (87.5%) (11 .So/,) Vulcanised natural rubber, S, 2.5 pphrb (VNR-I) 81 Vulcanised natural rubber, S, 0.3 pphrb (VNR-II) 77 59 (1.5%) 12 (91.5%) 3.3 (7.0%) C-C crosslinked natural rubber (CNR) 74 60 (2.0%) 14 (92.0%) 2.7 (6.OYO) 290 Vulcanised epoxidised natural rubber (VENR) 55 4.9 0.57 0.21 (2.0%) (1 .OYO) (33.5%) (63.5%) 1.2 (92.5%) w1 .7 0.23 (7.5%) W0.6 0.29 (11.OYo) E 0.12 (77.0%) E 0.055 (1 2.0%) E 10 (5.0%) E 5.0 (66.5%) E 2.5 (9.5%) E 0.51 (17.0%) E 0.013 (2.0%) G 12 (6.OYO) E 5.8 (63.0%) E 0.34 (28.0%) E 0.014 (3.0%) G 18 (4.5%) E 6 (64.0%) E 0.57 (30.0%) E 0.010 (1 .5%) G 0.42 (2.5%) E 0.12 (47.5%) E 0.051 (45.5%) E 0.017 (4.5%) G W = Weibullian; E =exponential; G =Gaussian. For the Weibullian the exponent is quoted following the W. Parts per hundred of rubber. a ' TABLE 3. Relaxation data for vulcanised epoxidised samples and paraffin additives Sample Paraffin oila T, (ms) 290 VENR J 812714 5 270 J 812715 10 240 J 812716 20 230 Paraffin wax 174 Paraffin oilc 160 (27.0%) 65 (73.0%) a Parts per hundred of rubber. ' E =exponential; G = Gaussian Liquid. POLYMER INTERNATIONAL VOL. 40, NO. 1, 1996 TI, (ms) 55 (2.0%) 4.9 (1 .OYO) 0.57 (33.5%) 0.21 (63.5%) 39 (3.0%) 4.5 (5.0%) 0.57 (38.0%) 0.21 (54.0%) 51 (5.5%) 4.5 (6.0%) 0.64 (31.5%) 0.23 (57.0%) 52 (12.5%) 4.2 (6.5%) 0.64 (29.5%) 0.23 (51.5%) 19 (45.0%) 8.7 (55.0%) 62 (52.0%) 17 (48.0%) T* (ms)b 0.42 (2.5%) E 0.12 (47.5%) E 0.051 (45.5%) E 0.017 (4.5%) G 0.43 (8.6%) E 0.12 (49.0%) E 0.046 (34.0%) E 0.016 (8.0%) G 0.74 (9.0Yo) E 0.14 (45.0%) E 0.053 (38.0%) E 0.017 (8.0%) G 0.95 (18.0%) E 0.14 (45.0%) E 0.052 (31 .O?'o) E 0.018 (6.00/,) G 0.22 (3.0%) E 0.015 (97.0%) G long R . K . Harris, B. J . Say, S . N g 68 disation causes a significant increase in TI, giving values within a narrow range for all epoxidised samples. Rotating-frame relaxation, on the other hand, is clearly not single-exponential for any of the samples, and we have fitted the experimental measurements to two, three or four exponentials, making judgements as to the optimum number necessary in each case. Figure 2 shows an example. Such judgements are mathematical and are deliberately not based on any particular physical model. However, it is significant that all epoxidised samples required four exponentials, whereas the vulcanised (but not epoxidised) materials were satisfactorily fitted to three, and natural rubber needed only two (as did paraffin wax). It is well understood that the analysis of NMR lineshapes or FID forms is far from straightforward. Although the shape attributed to rigid solids is generally Gaussian, critical examination always shows the situation to be more complex. In the case where the motions in the sample conform to a single-exponential correlation function an exponential FID (Lorentzian lineshape) is observed, which can be theoretically interpreted. When the motions become more complex, i.e. either an exponential autocorrelation function is no longer appropriate or there is a distribution of correlation times, the situation is rather intractable. Various authors have taken an ad hoc approach to the matter, suggesting model functions for the FID, while there have also been attempts to achieve a theoretical prediction for the lineshape in certain systems. Our own approach has been to use a sum of simple functions to represent the FID, on the understanding that if such a sum represents the FID to within experimental error, then the true functions must be linear combinations of the fitted functions, also to within experimental error. It is extremely important that the quoted values (as for the results given herein) are not taken to correspond with specific parts of the sample unless there is some corroborating evidence or good theoretical reasoning. The general form of the FID of the uncured rubber is best represented (with our set of functions) by a Weibullian (intermediate between a Lorentzian and a Gaussian) plus a small contribution from a stretched exponential (i.e. a Weibullian with an exponent of < 1). This is consistent with our expectations for elastomer.’l The FIDs of the vulcanised samples show a small Gaussian component with a time constant which is appropriate to residual static dipolar interactions, indicating the presence of regions where the molecular framework is effectively rigid. However, the majority of the decay in these cases becomes considerably longer, indicating an increase in mobility for most of the sample. Our first postulate in this case is that, along with crosslinking, some chain scission occurs in vulcanisation. Tl is largely unaffected by this procedure while Tlp decreases for the bulk of the sample, indicating an increase in the efficiency of relaxation of 30% to 80%. A small longer-time component also appears, which may well be related to the crosslinked regions. The effect of epoxidisation, on the other hand, is entirely different. The FID becomes shorter by nearly an order of magnitude and is best represented as a sum of three exponentials. This alone would indicate a sig- W.?+-*P.,-Ce+ d ~ Residuals x 1 6 2 ms , / 0% i . 0 . 4 ms Fig. 2. Analysis of TIPresults for JB/27/6. Trace (i) shows the acquired data for the free induction decay and (as a solid line) the longest exponential component. Trace (ii) shows the difference derived from trace (i) and the second longest exponential component. Traces (iii) and (iv) are derived in the same way. In each case the abscissa scale is successively magnified by a factor of five. The residuals are differences (magnified by a factor of 16) between the acquired data and the sum of the four components. The ordinate scales are intensities normalised to the zero-time full transverse magnetisation. Estimated errors on the points are within the size of the symbols used, as can be seen from the plot of the residuals. POLYMER INTERNATIONAL VOL. 40, NO. 1, 1996 69 Molecular-level mobility of vulcanised and epoxidised natural rubber nificant reduction in the mobility of the rubber. The TIP decreases substantially and TI increases. Both these observations are also consistent with a decrease in the mobility of the rubber chains. Vulcanisation of this material produces a small Gaussian component in the FI D and a long TIPcomponent as in previous examples of crosslinking. The remaining samples (Tables 1 and 3) differ in the amount of paraffin oil which has been blended in. As the percentage of paraffin oil increases, so does the amplitude of the ‘tail’ of the FID. Figure 3 shows an example of our mathematical treatment of the FID. All these samples have the same amount of paraffin wax added, and the increase in the Gaussian percentage over VENR is consistent with the wax retaining the structure of its ‘solid’. The effect of the paraffin oil on TIPis to increase the amplitude of the long-time component as the amount of oil increases. In an attempt to further elucidate these matters we carried out a two-dimensional experiment on sample JB/27/6, acquiring FIDs following various spin-lock times. The FIDs were analysed as before, using as a starting point the parameters for the immediately preceding FID. The results of these analyses are presented in Table 4 and shown graphically in Fig. 4. It is immediately apparent that the time constants of the FID components vary with spin-lock time. This emphasises the inherent simplification in our model for the FID shape. However, the components are sufficiently different that the same qualitative model can be applied to all the FIDs. Also, it will be noted that while the sums of squares of residuals are fairly constant down the series, the values do increase around 4 and 8ms of spin-lock time. This is probably due to trace signals from ‘lost components’ which are too small to identify precisely but which are large enough to affect the SDSQ. , , ‘ ,’ ,’ 60% -,’ ’ / 40% -” ,’ ,’ 80% -,’ ,/ /’ ’ 20% -.’ . , / I 0 . 0 4 ms , , I A* 0% ? - 0 . 0 4 ms Fig. 3. Analysis of the FID components for JB/27/6. The data are displayed as in Fig. 2 except that the timescale is identical for traces (iii) and (iv). + x rl 20% - 0 - Long E x p o n e n t i a l - Medium Exponential - Short Exponential - Gaussian 2% 1% 10 ms Fig. 4. Decay of the amplitude of the various FID components for JB/27/6 as a function of spin-lock time. The percentages are plotted on a logarithmic scale and refer to the initial amplitude of the FID for a zero spin-lock time. POLYMER INTERNATIONAL VOL. 40, NO. 1, 1996 R. K . Harris, B. J . Say, S . N g 70 TABLE 4. Analysis of free induction decays from sample JB/27/6 following various spin-lock times and a solid echo Spin-lock time (ms) Long exponential Amplitude T2 (ps) 95.6 94.1 91.6 89.8 85.7 84.2 83.1 81.4 71.2 60.6 52.1 37.8 25.3 13.4 8.6 989 101 5 101 7 1043 1085 1092 1093 1098 1239 1411 1451 1480 1604 1505 1366 0.0 0.1 0.2 0.4 0.5 0.6 0.8 1 .o 2.0 4.0 8.0 20 40 80 100 Mid-exponential Amplitude 245.5 190.2 147.4 89.7 69.8 54.7 35.3 20.6 Short-exponential T2 (p) Amplitude 133 138 139 141 146 148 140 132 165 103 63.1 27.3 19.7 14.5 5.3 4.1 2.0 T2 ( p ) 48.7 47.9 47.7 46.1 44.9 46.0 45.6 52.8 39.0 SDSQ" Gaussian Amplitude 35.9 29.2 27.5 23.9 21.6 21 .I 21.3 20.0 16.1 11.7 8.2 5.3 5.1 4.0 3.4 T2 ( p ) 16.7 14.9 14.5 13.3 13.4 13.1 13.3 12.7 12.2 10.8 10.2 9.4 9.4 9.4 9.4 139 118 127 120 117 127 126 116 133 345 439 268 150 139 132 SDSQ =Sum of deviations squared for the fitting procedure. The dependence of the amplitudes for each component on spin-lock time was then analysed as a multiexponential decay. The results are presented in Table 5 and plotted as solid lines in Fig. 4. The long-exponential FID component does not show a simple behaviour under spin-locking, but appears to be lost with two time constants of 51 ms and 2.3ms, which we attribute to free and restricted paraffin oil, the latter being embedded more thoroughly in the rubber matrix. The two intermediate components of the FID are lost with submillisecond time constants and are attributed to the epoxidised rubber. These components form major parts of the samples which have undergone epoxidisation. The Gaussian component has a range of decay times. The longest is longer than any which were observed in the direct measurement of T I P .The directly observed longest relaxation time TIPis thus a combination of relaxation times. The use of this two-dimensional TABLE 5. Multiexponential spin-locked relaxation of FID components" ~ FID component Relaxation characteristics Amplitude (%) Gaussian Short exponential Medium exponential Long exponential a 6.3 (1.2%) 18 (3.3%) 11.4 (2.1Yo) 163.6 (30.2%) 244.6 (45.1YO) 58.3 (10.7%) 36.8 (6.8%) From the data in Table 4. Time constant (ms) 170 3.5 0.17 0.22 0.40 51 2.3 approach (FID following spin-locking) allows the better discernment of the behaviour of the rigid component. CONCLUSIONS The molecular-level effects of vulcanisation, epoxidisation and the addition of oil/wax additives have been investigated by studying the 13C MAS spectra and proton relaxation times. These provide evidence for some increase of chain mobility on vulcanisation, possibly associated with chain scission, whereas epoxidisation causes a reduction in mobility. Analysis of free induction decays following spin-locking enables further conclusions to be made about the mobility of the sample, with a small part shown to be highly rigid. ACKNOWLEDGEMENTS We are grateful to the UK Science and Engineering Research Council for the 13C spectra obtained under the National Service in Solid-state NMR, based at Durham. We thank Dr A. M. Kenwright for helpful discussions. Grateful thanks are due to Mr Lim Hun So0 and Dr K. Muniandy of the Rubber Research Institute of Malaysia, Kuala Lumpur, for providing all the natural rubber samples and vulcanisates used in this study and for valuable discussions concerning these materials. REFERENCES 1 Saville, B. & Watson, A. A., Rubber Chem. Technol.,40 (1967) 100. 2 Koenig, J. L. & Patterson, D. I., Elastomerics (November 1986) 21. 3 Patterson, D. J., Koenig, J. L. & Shelton, J. R., Rubber Chem. Technol.,56 (1983) 971. POLYMER INTERNATIONAL VOL. 40, NO. 1, 1996 Molecular-level mobility of vulcanised and epoxidised natural rubber 4 Patterson, D. J. & Koenig, J. L., Appl. Spectrosc., 41 (1987) 441. 5 Krejsa, M. R. & Koenig, J. L., Rubber Chem. Technol., 66 (1993) 376, and references cited therein. 6 Baker, C. S. L., Gelling, I. R. & Newell, R., Rubber Chem. Technol., 58 (1985) 67. 7 Gelling, I. R., J . Nat. Rubber. Res., 6 (1991) 184. 8 Krejsa, M. R. & Koenig, J. L., Rubber Chem. Technol., 65 (1992) 427. 9 Gelling, I. R. & Morrison, N. J., Rubber Chem. Technol., 58 (1985) 243. 10 Zaper, A. M. & Koenig, J. L., Rubber Chem. Technol., 60 (1987) 252. 11 Schaefer, J., Chin, S. H. & Weissman, S. I., Macromolecules, 5 (1972) 798. 12 Schaefer, J., in Topics in Carbon-13 N M R Spectroscopy, Vol. 1, ed. G. C. Levy. John Wiley, New York, 1974, pp. 149-208. POLYMER INTERNATIONAL VOL. 40, NO. 1, 1996 71 13 Andreis, M., Liu, J. & Koenig, J. L., Rubber Chem. Technol., 62 (1989) 82. 14 Krejsa, M. R. & Koenig, J. L., Rubber Chem. Technol., 65 (1992) 956. 15 Harris, R. K., Kenwright, A. M., Royston, A. & Say, B. J., Meas. Sci. Technol., l(1990) 1304. 16 Kenwright, A. M. & Say, B. J., in N M R Spectroscopy of Polymers, ed. R. Ibbett. Chapman and Hall, London, 1993, pp. 231-274. 17 McBrierley, V. J. & Packer, K. J., Nuclear Magnetic Resonance in Solid Polymers. Cambridge University Press, Cambridge, 1993. 18 Dadayli, D., Harris, R. K., Kenwright, A. M., Say, B. J. & Siinnetcioglu, M. M., Polymer, 35 (1994) 4083. 19 Weibull, W., J. Appl. Mech., 18 (1951) 293. 20 Davey, J. E. & Loadman, M. J. R., Br. Polym. J.,16 (1984) 134. 21 Kaufmann, S. & Bunger, D. J., J . Magn. Reson., 3 (1970) 218.