Unsupported SiO2-based organic–inorganic membranes Part 1.—Synthesis and structural characterization Sandra Dire`,*a Eva Pagani,a Florence Babonneau,b Riccardo Ceccatoa and Giovanni Carturana aDipartimento di Ingegneria dei Materiali, Universita` di T rento, v.Mesiano 77, 38050 T rento, Italy bChimie de la Matie`re Condense�e, Universite� Pierre etMarie Curie/CNRS, 4, place Jussieu, Paris, France Tetraethoxysilane (TEOS) and methyltriethoxysilane (MTES) have been used to prepare hybrid SiO2-based membranes.These self-supported materials were obtained from controlled polymerization reactions for various TEOS/MTES molar ratios ensuring the achievement of crack-free disks 8 cm in diameter and 10–40 mm in thickness. The rheological behaviour of precursor solutions was studied and gelling times were determined.The whole process, from starting solution to xerogel, was followed by FTIR spectroscopy, viscosity measurements and multinuclear solid-state NMR, and is discussed in terms of the hydrolysis–condensation kinetics of tetrafunctional and trifunctional silicon alkoxides. Density, shrinkage, elastic modulus (E), modulus of rupture (MOR) and elongation at break were all determined and related to preferential structural arrangements of networks according to the TEOS/MTES ratio.Hybrid organic–inorganic materials are currently under inves- solution at pH=1.5. Various water/alkoxide ratios were calcutigation as potential multifunctional and high-performance lated for the different compositions, so that H2O/SiOEt=0.5. materials: organic modified ceramics (ORMOCERs) properties No solvent was used except for the pure Si(OEt)4 gel (ethanol).are exploited in the fields of adhesives, sealing and modified Details concerning the labelling of samples, compositions and glass surface materials,1–3 sensors4 and artificial membranes.5–8 preparation conditions are reported in Table 1. As regards membrane technology, the extension of appli- Solutions were stirred at room temperature, using storage cation fields to more and more sophisticated processes parallels times determined by solution rheology, affected by the the development of the new class of inorganic membranes, TEOS/MTES ratio.Each solution (2.4 ml) poured into poly- which can also compete with organic polymers in terms of styrene Petri dishes (diameter 8 cm) and covered by polymer chemical and thermal durability and separation performances. foil afforded clear, homogeneous gel membranes in ca. 10 days, Sensors and membranes are involved with mass transport during which time several pinholes on the foil were made. phenomena due to concentration, pressure or electrical field Samples of thickness 10–40 mm were allowed to dry in air at gradients: any chemical modification causing changes of net- constant temperature and humidity (Fig. 1). work structure and surface features affects detection or separation properties so that hybrid materials are expected to improve the development of inorganic membranes. Characterization techniques The sol–gel process can lead easily to silica-based hybrid The rheological properties of the gelling solutions were deter- materials containing SiMO and SiMC bonds: intrinsic molecu- mined at 24.8±0.2°C with a cone/plate system using a MC- lar gels with both organic and inorganic moieties are obtained 10 Physica viscosimeter.Values of shear stress, shear rate, by using R¾nSi(OR)4-n as precursors.9–11 Note that the con- temperature and viscosity were collected for the various solu- ditions chosen for hydrolysis–condensation of these precursors tions during ageing.affect the structural features and physical properties of the Solid-state NMR measurements were performed on a MSL final products.12,13 Obviously, with a constant CH3/Si/O ratio, 400 Bruker spectrometer. Solid samples were spun at 4 kHz. the ordinary synthesis of hybrid SiO2 materials, from mixtures For the 29Si single-pulse experiments (SPE), pulse width (2 ms: of Si(OR)4, as cross-linking agent and di- or tri-functional h=30°) and relaxation delays (30 s) were chosen to take into methylsilanes, may not lead to the same product as that account the long T1 relaxation times. 29Si MAS NMR spectra obtained from pure CH3Si(OR)3 ; in particular, the different were also recorded using cross polarization (CP) techniques spatial location of CH3MSi bonds may result in different and variable contact times.The spectra were analysed using porosities and pore size distributions. This apparent complithe WINNMR and WINFIT programs.14 13C CP MAS NMR cation may modify mechanical properties: in particular, concentration and distribution of CH3MSi groups may influence spectra were recorded using 2 ms contact time.Recycle delays the achievement of self-carrying items. of 10 s were used for all CP spectra. In this initial work, the sol–gel process was used to prepare FTIR spectra were recorded on a Nicolet 5DXC instrument self-supported, hybrid organic–inorganic membranes from mix- equipped with both horizontal attenuated total reflectance tures of silicon tetraalkoxide and CH3Si(OEt)3: results on the (HATR) and diffuse reflectance (DR) accessories. In the HATR development of the gel structure from these precursors and on mode, a known volume of sol was poured directly onto the the characterization of the resulting xerogels with various surface of the ZnSe crystal, closed with an appropriate cap; 64 Si(OEt)4/CH3Si(OEt)3 ratios, will be presented in this paper, scans at 45° to the IR incident radiation were collected.For while gas-separation properties and surface characterization DR measurements, xerogels were powdered and dispersed in will be reported in a future paper. KBr, accumulating 64 scans for each spectrum. Densities were measured in water and hexane at 25°C by Archimedes’ method; linear shrinkage was calculated by meas- Experimental uring the xerogels’ diameters and expressing them as percent- Synthesis of Si(OEt)4/MeSi(OEt)3 gels ages referring to the Petri dish diameter.MOR, Young’s modulus (E) and maximum elongation at break (eR ) were Various mixtures of Si(OEt)4 (TEOS) and CH3Si(OEt)3 (MTES) were hydrolysed at room temperature with an HCl obtained with a three-point bending test, using an Instrom J. Mater.Chem., 1997, 7(1), 67–73 67Table 1 Sample labels and compositions Sample label, composition, H2O/Si ethanol/Si TxMy molar ratio molar ratio molar ratio [Si]0/mol l-1 T100 100 TEOS 2 0.5 3.47 T70M30 70 TEOS/30 MTES 1.85 — 4.01 T50M50 50 TEOS/50 MTES 1.75 — 4.12 T30M70 30 TEOS/70 MTES 1.65 — 4.24 M100 100 MTES 1.5 — 4.42 Fig. 1 T70M30 xerogel disk testing machine with a load cell of 100 N and a testing rate of 1 mm min-1. Results The hydrolysis and condensation process of MTES and TEOS precursors at various ratios was studied using different techniques, in order to compare the advance of the gelling process and of the gel network structure. IR spectra Fig. 2(a) shows FTIR spectra recorded in the HATR mode in the interval 4000–650 cm-1 for T100, T50M50 and M100 samples, 15 min after preparation of the solutions.In the highfrequency field, CMH stretching vibrations at 3000–2850 cm-1 and a wide band centred at 3350 cm-1 due to OMH stretching were present. The bending vibrations of the Et and Me groups were observed in the interval 1500–1300 cm-1; for compositions containing MTES, a precise signal due to SiMCH3 stretching was recorded at 1276 cm-1.The 1200–1000 cm-1 region showed overlapping signals of OMSiMO bonds in different species. The signal due to SiMOEt bonds at 954 cm-1 was recognized clearly in the T100 spectrum, absorption due to ethanol being observed at 880 cm-1; the M100 spectrum displays peaks at 916 cm-1 attributed to SiMOH and SiMO- bonds,15 and that of ethanol at 880 cm-1.The signal at 955 cm-1 was virtually lost, suggesting that the hydrolysis of ethoxide groups in this sample was almost completed. In the T50M50 spectrum, an intermediate situation appeared with peaks at955, 920 and 880 cm-1.Spectra recordedafter 120 min are shown ig. 2(b). In the T100 sample, the band at 3350 cm-1 revealed large amounts of water and ethanol, which decreased as the MTES content increased, in agreement with a decrease of the signal at 1637 cm-1 [d(OH)].The M100 sample was characterized by almost complete hydrolysis, the SiMCH3 signal at 1276 cm-1 and the SiMOH one at 916 cm-1 still being visible. Instead, a remarkable concentration of SiMOEt was observed in T100, while T50M50 showed intermediate behaviour. The M100 spectrum recorded between Fig. 2 HATR-FTIR spectra recorded (a) 15 min and (b) 120 min after 120 min and gelling time (220 min) remained virtually preparation of the T100, T50M50 and M100 sols unchanged, whereas slow, progressive evolution was found for 68 J. Mater. Chem., 1997, 7(1), 67–73T100 in the interval 120–330 min, i.e. T100 gelling time.This between the viscosity increase and composition; samples with higher concentrations of MTES displayed a viscosity increase supports the fact that hydrolysis–condensation processes still occur for T100 when the MTES sample had already gelled. at 238–250 min ageing time, whereas TEOS-rich compositions showed a net increase after 327–342 min, the slope of the curve Fig. 3 reports 4000–400 cm-1 spectra recorded in the DR mode, which revealed the spectral window between 650 and being appreciably smaller. A trend discontinuity was observed for T30M70, whose transition occurred at 226 min. 400 cm-1. TxMy films, stored at the same temperature and humidity, show that adsorbed water increased as %TEOS Shear stress vs. shear rate and viscosity vs. shear rate plots are shown in Fig. 5. M100 lost Newtonian flow behaviour increased (bands at 3300 and 1630 cm-1). In the 3000–2800 cm-1 interval, CMH stretching vibrations of after 200 min [Fig. 5(a)], while T100 showed it up to 305 min [Fig. 5(b)]; in both cases, further ageing produced yield behav- residual OEt and Me groups were present; the intensity of the peak at 1272 cm-1 [n(SiMCH3 )] decreased as the MTES iour [Fig. 5(c) and (d)].17 TxMy compositions displayed intermediate behaviour. concentration decreased. Other features were the intensity inversion of peaks in the 1200–1000 cm-1 interval and the lowering of the peak at 940 cm-1 (SiMOH and SiMOEt) as NMR results the amount of MTES increased. The position of the peak Solid-state NMR (MAS NMR) can provide direct information corresponding to the angular deformation d(SiMOMSi) and on the local environment of different structural units, and thus related to the silica units arrangement,16 at 466 cm-1 in the on the degree of condensation of the network.Fig. 6 shows T100 spectrum, was shifted to lower frequencies as %MTES the 29Si MAS NMR spectra of T100, T50M50 and M100 increased. recorded as single-pulse experiments (SPE); Table 2 shows the percentages of various units obtained after simulation of SPE Viscosity measurements spectra, as well as the degree of condensation of the network.Fig. 4 reports the evolution of solution viscosities of the TxMy It confirms what was published previously,18 that the degree samples. Viscosity vs. time diagrams indicate a relationship of condensation decreases with the average functionality of the precursors. It is interesting to note that the degree of condensation of the T50M50 sample is intermediate between those of the T100 and M100 samples.Series of 29Si CP MAS NMR spectra with variable contact times were also recorded on the three samples. Analysis of the variation of magnetization vs. contact time was carried out according to the simplest model describing CP between two spin reservoirs, one for dilute spins, S, and one for abundant spins, I, using the well known formula:19 MS(tc)=cI cS M0S 1 1-lG1-exp[-(1-l) tc TIS ]Hexp(- tc T 1rI) with l=TIS/T1rI.M0S is the magnetization at the equilibrium in the static field B0, TIS is the CP standard time which is related to the strength of the I–S dipolar coupling and T1rI is the relaxation time of the abundant spins in the rotating frame, which will cause a loss of magnetization for long contact time.cI and cS are the magnetogyric ratios for spins I and S, respectively. This formula applies for S spins which are partially decoupled from the protons. The TSiH and T1rH fitted values are reported in Table 2 as well as the percentages of the various units detected by CP, and estimated from the M0S value. For T100 and M100 samples, the agreement factor for the fitting procedure, R, is > 0.99, showing a good agreement between the experimental behaviour and the theoretical model.In the case of the T50M50 CP MAS spectrum, a lower agreement factor is obtained (R>0.97) which may be related to the presence of several local environments for the different T and Q sites.Fig. 3 DRIFT spectra of the TxMy xerogels Two interesting features can be pointed out from this study. First, a comparison between the quantitative analyses of the various sites carried out from the SPE and CP spectra shows perfect agreement for the M100 and T50M50 samples, and a discrepancy for the T100 sample.In this last sample, the number of Q4 units is underestimated in the CP technique, suggesting that the analysis of the CP dynamics measures only the amount of Q4 units close to Q3 or Q2 units. In contrast, in the T50M50 system, the Q4 units should be in close proximity to T units, which allows their detection via the CP technique. A second interesting point is the difference in the TSiH values corresponding to the Q units, in the T100 and T50M50 samples.Their decrease in the T50M50 sample indicates stronger 1H–29Si dipolar coupling for these units and could be related to the close proximity of the T units. These results strongly suggest a good chemical homogeneity of the Fig. 4 Viscosity vs. time diagrams of the TxMy samples: 1, M100; #, T100; %, T50M50; +, T30M70; ×, T70M30 T and Q units within the T50M50 sample.J. Mater. Chem., 1997, 7(1), 67–73 69Fig. 5 Viscosity vs. shear rate [(a) and (b)] and shear stress vs. shear rate [(c) and (d)] plots of M100 [(a) and (c)] and T100 [(b) and (d)] samples present in the 1H MAS NMR spectrum of the T50M50 xerogel [Fig. 7(b)]. Physical and mechanical results The density, shrinkage and mechanical features of the xerogel membranes were also studied.Densities are reported in Table 3; a linear decrease as the percentage of MTES increased was observed. Shrinkages, calculated as described in the Experimental section, displayed a linear relationship with MTES content (Fig. 8), proving the absence of linear shrinkage for M100. Mechanical measurements performed on thin bars of the xerogel were obtained by a three-point bending test; the results are reported in Table 4.The elastic modulus, E, is an intrinsic property depending on the bond density and related to the material structure. The E value of T100 agrees with reported values of silica xerogels prepared under acidic conditions, i.e. 5–10 GPa.20 The decrease in elastic modulus as the organic load increased accounted for the lower cross-linking consequent upon the introduction of a trifunctional precursor.The consequent reduction in SiMO bond density also affected the maximum elongation at break eR , which increased with MTES concentration. Fracture surfaces were featureless and disks appeared dense and macroscopically homogeneous, as shown in Fig. 9. Discussion For a two-step acid–base process, van Bommel et al.21 reported that, under acidic and neutral conditions, the hydrolysis rate Fig. 6 29Si MAS NMR spectra of M100, T50M50 and T100 xerogels of alkyl-substituted silicon alkoxides is faster than for TEOS; moreover, condensation already occurs in the acid step. These observations have been confirmed recently:22a after the addition 13C CP MAS NMR and 1H MAS NMR spectra (Fig. 7), recorded on M100 and T50M50 samples, show the presence of water, under acidic conditions, the hydrolysis of alkylsubstitutedalkoxides with short alkyl chains is almost complete of SiMMe groups with signals at d -3.7 and 0.4, respectively. Moreover, these spectra indicate an incomplete hydrolysis– within the first few minutes and the condensation degree increases quickly during the first 2 h and then slows. condensation process.As a matter of fact, the signals due to residual SiMOEt groups are present in the 13C CP MAS NMR Results obtained for TxMy samples are consistent with these works, based on viscosity and NMR results:21,22 the HATR- spectra (d 18.4 and 58.1 for M100, d 17.6 and 59.9 for T50M50), and a signal attributed to SiMOH terminal units (d 4.1) is FTIR spectra indicate that MTES is hydrolysed faster than 70 J.Mater. Chem., 1997, 7(1), 67–73Table 2 29Si solid-state NMR data units obtained (%) degree of sample sitea d TSiH/ms T1r/ms SPE CP condensation M100 T3 -65.2 1.7 52 88 86 0.96 T2 -57.6 1.1 48 12 14 T100 Q4 -110.9 8.7 82 49 25 Q3 -101.4 4.2 27 44 64 0.85 Q2 -92.1 3.6 21 7 11 T50M50 T3 -63.1 2.2 123 45 42 T2 -55.3 1.6 70 5 10 Q4 -108.9 5.1 177 21 20 0.89 Q3 -100.7 2.2 113 25 25 Q2 -91.7 1.5 65 4 3 a Tn and Qn: n=number of bridging oxygens.Fig. 7 1H MAS NMR and 13C CP MAS NMR spectra of (a) M100 and (b) T50M50 xerogels Table 3 TxMy density results Table 4 Mechanical characterization of TxMy samples sample density/g cm-3 sample E/GPa eR (%) MOR (MPa) T100 1.84±0.04 T100 5.2±0.6 7 40±10 T70M30 1.61±0.03 T50M50 2.6±0.3 19 50±8 T50M50 1.48±0.02 M100 0.7±0.1 43 30±10 T30M70 1.34±0.04 M100 1.29±0.05 Fig. 9 Fracture surface of the T50M50 bar (SEM image) Fig. 8 Linear shrinkage vs. %MTES of TxMy disks J. Mater. Chem., 1997, 7(1), 67–73 71TEOS, since the concentration of ethoxy groups for M100 is gels (Table 2) and the lower agreement factor between the experimental and simulated spectra.The decrease in TSiH of Q reduced markedly after 15 min; moreover, hydrolysis seems to be complete in 120 min, and extensive condensation of silanols units in T50M50, compared also with results reported for silica gels prepared from TEOS in acid conditions,24 could be may be deduced from the negligible evolution of the M100 spectra from 120 min to gelling time.The different behaviour ascribed to the formation of an intimate mixture between silica gel and organic modified moieties.25 TSiH values of 1 ms have in the reactivities of M100 and T100 under acidic conditions may be ascribed to the inductive effect of the CH3 group, been reported for T units in which the Si atom is surrounded by three bridging oxygens.26 The slight increase in TSiH of T leading to activated species 1 which is lower in energy than 2, ultimately resulting in faster hydrolysis–condensation reactions units in T50M50, owing to a low effectiveness of magnetization transfer for short contact times, could be related to a different for MTES.23 mobility of T units in the network.Peeters et al.27 recently studied the functionality of hybrid gels obtained from mixtures of TEOS and various trifunctional organosilanes and concluded that, in spite of the enhanced condensation of Q and T units, the total number of network bonds decreases with increasing substitution level. This general This interpretation is consistent with the short time required behaviour may be extended to our results on the physical and for M100 viscosity increase.The fast condensation reaction mechanical properties of the gels. Density values decrease from leads to extended oligomers, which collapse to the gel network 1.84 (T100) to 1.29 g cm-3 (M100), as a result of a less well by condensation of few residual SiMOH bonds: indeed, the interconnected network. The same fact is observed for the E gelling time of M100 is shorter than that of T100, in which and eR values: these parameters are related to bond density the viscosity increase takes much longer, suggesting slower, and reflect increased network mobility as %MTES continuous SiMOH condensation.17 The gelling kinetics for increases.13,28 Linear shrinkage is maximum for T100 and M100 and T100 solutions having intermediate compositions decreases with increasing MTES content.Shrinkage may be do not lead to behaviours corresponding to the sum of M100 ascribed to byproduct release and condensation between and T100 behaviours multiplied by the molar fractions of residual SiMOH groups after gel formation; the latter process relevant precursors: for instance, the viscosity trend of T50M50 is prominent in our case, owing to the experimental conditions is very close to that of M100 (Fig. 4). This observation deserves employed. The possibility of affording a siloxane bond depends some specific comments. If gelling time is considered as a on the concentration of terminal SiMOH groups. Solid-state kinetic parameter which substantiates the occurrence of a NMR spectra show that the M100 xerogel displays slow closed SiO2 network, i.e.the occurrence of a structure where availability of SiMOH, since the T2 units (OH and OEt most oxygens bridge two Si atoms, parameter t, defined as: terminal units) only amount to 12% and terminal silanols may be not present, in agreement with our 1H MAS NMR and t=Si functionality gelling time [Si]0 FTIR results. These facts and the hydrophobicity of sample M100,29 preventing hydrolysis of residual SiMOR groups by where [Si]0=silicon concentration in the starting solutions adsorbed water, accounts for the absence of linear shrinkage (Table 1), Si functionality=number of SiMOR in the starting observed in M100.In contrast, the shrinkage of TxMy xerogels precursor, quantifies the rate required for engaging the SiMO increases with %TEOS, owing to the greater concentration of bonds of the precursors in SiMOMSi units per volume unit.SiMOH groups and to the possible involvement of moisture Values of t for each sample are shown in Fig. 10 vs. MTES to complete SiMOR hydrolysis. mol%. Clearly, MTES accelerates the occurrence of the gel; in fact t increases above the value expected if the process resulted Provincia Autonoma di Trento is greatly acknowledged for from the sum of the individual gelling processes of TEOS and financial support.MTES (Fig. 10, dotted line). From the viewpoint of reaction mechanism, this fact implies that terminal SiMOH groups References derived from TEOS are involved more rapidly in condensation to SiMOMSi upon reaction with SiMOH groups derived from 1 H.Scholze, J. Non-Cryst. Solids, 1985, 73, 669. MTES. Indeed, partial cocondensation between T and Q units 2 H. Schmidt, H. Scholze and G. Tunker, J. Non-Cryst. Solids, 1986, may be presumed in the T50M50 gel, considering the variations 80, 557. 3 J. Wen, V. J. Vasudevan and G. L. Wilkes, J. Sol–Gel Sci. T echnol., in TSiH of T and Q units compared to T100 (Q) and M100 (T) 1995, 5, 115. 4 P. Lacan, P. Le Gall, I. Rigola, C. Lurin, D. Wettling, C. Guizard and L. Cot, Sol–Gel Optics II, Proc. SPIE Vol. 1758, ed. J. D. Mackenzie, SPIE, Washington, DC, 1992, p. 464. 5 A. Kaiser, H. Schmidt and H. Bottner, J.Membr. Sci., 1985, 22, 257. 6 C. Guizard, N. Ajaka, M. P. Besland, A. Larbot and L. Cot, in Polyimides and Other High T emperature Polymers, ed. M. K. M. Abadie and B.Sillion, Elsevier, Amsterdam, 1991, p. 537. 7 C. Guizard and P. Lacan, in Proc. 1st Eur. Workshop on Hybrid Organic Inorganic Materials, Bierville, November 8–10, 1993, ed. C. Sanchez and F. Ribot, CNRS, Paris, 1993, p. 153. 8 T. Okui, Y. Saito, T. Okubo and M. Sadakata, J. Sol–Gel Sci. T echnol., 1995, 5, 127. 9 S. Dire`, F. Babonneau, C. Sanchez and J. Livage, J. Mater. Chem., 1992, 2, 239. 10 Z. Zhang, Y. Tanigami, R. Terai and H. Wakabayashi, J. Non- Cryst. Solids, 1995, 189, 212. 11 F. Babonneau, L. Bois, J. Maquet and J. Livage, in Eurogel 91, ed. S. Vilminot, R. Nass and H. Schmidt, Elsevier, Amsterdam, 1992, p. 319. 12 W. G. Fahrenholtz and D. M. Smith, Mater. Res. Soc. Symp. Proc., 1992, 271, 705. Fig. 10 Parameter t vs. MTES mol(%) (t=Si functionality 13 H.H. Huang, B. Orler and G. Wilkes, Macromolecules, 1987, 20, 1322. ×[Si]0/gelling time) [$, experimental; #, calculated (mixture rule)] 72 J. Mater. Chem., 1997, 7(1), 67–7314 Programs from Bruker Spectrospin, Wissembourg, France. 1994, 346, 365; (b) S. Prabakar, R. A. Assink, N. K. Raman and C. J. Brinker,Mater. Res. Soc. Symp. Proc., 1994, 346, 979. 15 R. M. Almeida, T. A. Guitton and C. G. Pantano, J. Non-Cryst. Solids, 1990, 121, 193. 23 C. J. Brinker and G. W. Scherer, Sol–Gel Science, Academic Press, New York, 1990, ch. 3. 16 F. Babonneau, K. Thorne and J. D. Mackenzie, Chem. Mater., 1989, 1, 554. 24 K. L.Walther, A.Wokaun and A. Baiker,Mol. Phys., 1990, 71, 769. 25 F. Babonneau, Mater. Res. Soc. Symp. Proc., 1994, 346, 949. 17 M. D. Sacks and R. Sheu, J. Non-Cryst. Solids, 1987, 92, 383. 18 R. H. Glaser, G. L. Wilkes and C. E. Bronnimann, J. Non-Cryst. 26 G. S. Carajava, D. E. Leyden, G. R. Quinting and G. E. Maciel, Anal. Chem., 1988, 60, 1776. Solids, 1989, 113, 73. 19 M. Mehring, Principles of High Resolution NMR in Solids, 27 M. P. J. Peeters, W. J. J. Wakelkamp and A. P. M. Kentgens, J. Non-Cryst. Solids, 1995, 189, 77. Springer-Verlag, Berlin, 1983, p. 129. 20 M. J. Muratagh, E. K. Graham and C. G. Pantano, J. Am. Ceram. 28 R. H. Glaser and G. L. Wilkes, Polym. Bull., 1988, 19, 51. 29 C. Della Volpe, S. Dire` and E. Pagani, J. Non-Cryst. Solids, 1996, Soc., 1986, 69, 775. 21 M. J. van Bommel, T. N. M. Bernards and A. H. Boonstra, J. Non- in press. Cryst. Solids, 1991, 128, 231. 22 (a) L. Delattre and F. Babonneau, Mater. Res. Soc. Symp. Proc., Paper 6/03554J; Received 21stMay, 1996 J. Mater. Chem., 1997, 7(1), 67–73 73