Role of titanium in TiO, :SiO, sol-gels: an X-ray diffraction study? Jane S. Rigden,"" Robert J. Newport," Mark E. Smith," Peter J. Dirken" and Graham Bushnell-Wyeb "Physics Laboratory, University of Kent at Canterbury, Canterbury, Kent, UK CT2 7NR bCCRL Daresbury Laboratory, Daresbury, Warrington, UK WA4 4AD Transmission X-ray diffraction has been used to study a series of powdered silica :titania sol-gel glasses with titania contents ranging from a 'pure' silica sample through to high titania levels where phase separation is predicted to occur. Analysis of the data reveals a change in second- and third-neighbour coordination numbers with increasing Ti content and confirms that for low titanium contents the sol-gels are atomically mixed. The lower titanium content sol-gels have also been studied as spun thin films using shallow-angle X-ray diffraction.Comparison with the transmission studies shows an increase in disorder in the silica network when the material is in the form of a coating rather than in the bulk. An increase in the number of Si- 0-H bonds is also suggested. Mixed silica :titania materials are of significant technological importance. Silica glasses containing a few mol% TiO, are used as ultra-low thermal expansion (ULE)glasses' and mixed titanium :silicon oxides are important as catalysts and catalytic support materials.' In the optical industry they can be pro- duced as anti-reflective thin film coatings, or with tailored or graduated refractive indices.' The properties of titania :silica binaries, however, are strongly dependent on their chemical composition, homogeneity and texture; homogeneity at the atomic level is especially important.The different rates of hydrolysis of Ti alkoxides and Si alkoxides can mean that phase separation occurs as Ti-rich and Si-rich regions are formed: this significantly reduces the usefulness of the material. Sol-gel synthesis, based on hydrolysis and subsequent conden- sation of metal alkoxide precursors, is a relatively new method that combines atomic level mixing with a high degree of porosity .' Although SiO, shows four-fold Foordination in the bulk with an Si-0 distance of ca. 1.6; A, TiOz is six-coordinated with Ti-0 distance of ca. 1.93 A. A short Ti-0 distance of 1.82 A has been observed in EXAFS studies of silica gels and glasses with low TiOz levels, suggesting that the titanium is in four-fold coordination with oxygen.170NMR spectroscopy has confirmed that atomic mixing occurs in such glasses by revealing the presence of Ti-0-Si bond^;^.^ in contrast, OTi, and OTi4 features in the NMR spectra' of glasses with higher TiOz content (ca. 41 mol%) indicate that they are phase separated. Although silica and silica :titania binaries have been studied in their crystalline phases using X-ray diffracti~n,~.~ little work has been completed on the gels in their amorphous state. Transmission X-ray or neutron diffraction can reveal structural information averaged over an entire sample, and is therefore a useful method for studying bulk materials.However, conven- tional techniques cannot be used to study thin films or coatings, owing to the difficulty in separating properties of the film from those of the substrate. Shallow-angle X-ray diffraction, where sample penetration depths are lessened by reducing the incident angle of radiation onto the film, is a relatively new technique which can enable scattering from a thin film, and therefore structural information, to be isolated. The results presented herein represent an X-ray diffraction study of four SiOz :TiO, sol-gel glasses, with titanium content varying from 0 to ca. ?Presented at the Second International Conference on Materials Chemistry, MC', University of Kent at Canterbury, 17-21 July 1995.5 atom%. The samples are studied in bulk form using trans- mission diffraction and as thin films using shallow angle diffraction. Samples Four sol-gel glasses were prepared by hydrolysis of alkoxides with water and ethanol mixtures in the approximate ratios 1:2: 7.5, with varying titania contents., Sample 1, labelled 'pure silica', contained no titania; samples 2 and 3 contained small amounts of titania and were labelled 'low titania' and 'high titania', respectively; sample 4 contained very high titania levels and was in the domain in which phase separation was predicted to occur, and this sample was therefore labelled 'phase separated'. Note that all four samples contained substan- tial residual amounts of volatiles such as ethanol.Table 1 shows the compositional information, including mass and electron densities. It is immediately clear that the sample labelled 'phase separated' was compositionally very different from the samples which were expected to be atomically mixed. Whilst the other samples maintained a hydrogen :carbon ratio of ca. 2.7 : 1, the H:C ratio for sample 4 has risen to ca. 6: 1, with much less carbon but much more oxygen held in the system. The silica sol-gel and the two lowest titanium content mixed gels, were used to produce thin films by the spin coating method;' the 'phase separated' sample was not used to create a thin film as it was not possible to produce a uniform coating which adheres to the substrate with a non-atomically mixed sol.An excess of liquid was dispensed onto the surface of the substrate, and then the sample was rotated at low speed so that the liquid flowed radially outwards, driven by centripetal force. Any surplus liquid flowed to the edge of the substrate and dripped off. As the film thinned, the rate of removal of liquid slowed as the viscosity increased; in the final stages most of the thinning occurred by evaporation of volatiles.* The method produced a very uniformly thin coating, and the process could be repeated several times to build up a thicker film or, for example, to produce layers with slightly differing qualities. In the samples used herein, six layers of film were deposited to produce a film ca. 1 pm thick on a polished silicon wafer.Although the underlying physics and chemistry that govern polymer growth and gelation are the same for films and bulk sol-gels, several factors in the evolution of thin films mean that, structurally, the two forms can be quite different.' In bulk J. Muter. Chem., 1996, 6(3), 337-342 337 Table 1 Compositional information for the four SiO, :Ti02 samples sample composition (atom%) no. label Ti Si 0 1 pure silica 0.0 4.0 16.2 2 low titania 0.4 4.7 19.7 3 high titania 0.84 3.8 17.0 4 phase separated 4.9 7.0 38.2 systems evaporation usually occurs after gelation, whereas in thin films the deposition and evaporation processes happen simultaneously, and this results in a competition between compaction of the structure caused by evaporation and the stiffening (and therefore resistance to compaction) of the mate- rial caused by the condensation process.The short duration of deposition and evaporation/drying in thin films means that considerably less cross-linking occurs than in bulk gels, which generally results in more compact, dried structures.' Also, thin films are constrained by their geometry, and the continued shrinking causes stresses; this is particularly true for films made by spinning methods. It is likely, therefore, that there is a marked difference between the structure of sol-gels in thin films and in the bulk; it is expected that the rapid gelation of thin films will result in a more disordered material than in the bulk with a lower concentration of volatiles.Transmission Diffraction and the Shallow-angle Technique Both transmission and shallow-angle X-ray diffraction measurements were carried out on Station 9.1 at the Synchrotron Radiation Source at the CCRL Daresbury Laboratory, UK. The intrinsically highly parallel nature of the beam provided by a synchrotron source is of advantage over conventional focused laboratory X-ray sources for the shallow- angle technique in that the associated serious geometric aber- ration effects are avoided. Further, the high intensity beam provided by a synchrotron source is necessary for the relatively weak scattering from the small volume of amorphous material sampled in the shallow-angle geometry; also the availability of relativ5ly hard X-rays allows a wide dynamic range (up to ca. 18A-').The shallow-angle technique was first developed by Lim and Ortizg in 1987, who studied polycrystalline iron oxide layers on glass substrates, i.e. sharp Bragg peaks on a diffuse Debye Scherrer background. The method and analysis has recently been developed and used to study a variety of both amorphous and crystalline thin films. The refractive index of materials at X-ray wavelengths is less than unity; consequently, at incident angles below a critical value, ac, total external reflection occurs. Below a, limited penetration is achieved uiathe evanescent mode, and is exponentiallx damped: in principle sampling depths of ca. 10 to ca. 1000A may be achieved. Above a, the penetration depth increases rapidly with incident angle, inversely with the wavelength of the radiation, and is limited by photoelectric absorption; it is this region where shallow-angle diffraction can be used in a practi- vertxal crystal horizontal monochromator kapton shts 1 f0ll wtute beam -C H densitylg cm-3 electrons per A3 21.6 58.6 2.25 0.736 20.2 55.5 2.45 0.789 21.0 57.3 2.65 0.856 7.0 42.7 3.10 0.954 cal way to isolate scattering from a thin film mounted on a substrate. The conventional (transmission) X-ray diffraction arrange- ment12 is modified to produce the shallow-angle configuration, as shown schematically in Fig.1. The white beam from the synthrotron source is monochromated to a wavelength of 0.7 A by a channel cut crystal and proceeds through a pair of slits which define the incident-beam profile; a narrow slit profile of 100 pm x 10 mm is used in shallow-angle work to limit off-sample contamination scattering from the straight through beam at the lowest incident angles, where the beam's 'footprint' will be at its largest.The sample is set at a fixed, small angle a, to the incident X-rays. An iterative procedure of height and angle adjustment is used to define the zero-angle for the sample;fo~'' this procedure is very important given the small angles used in data collection. It is also essential that the sample is smooth and flat: any significant irregularity in the film thickness or sample curvature will produce a high uncertainty in a, and hence in the collected scattering profile.The scattered radiation passes through an arrangement of horizontal and vertical slits to the detector. A long slit package limits the viewed area and reduces the angular spread of scattered radiation incident on the detector and results in a resolution of ca. 0.07'. Data is collected sequentially at angles 28=2-130"; this is later converted to the scattering vector Q = 4n/A sin 8. X-Ray Diffraction Theory and Data Analysis Transmission diffraction Preliminary data reduction for transmission diffraction is carried out in the usual way,13 and includes correction for dead-time losses, the polarisation of the incident X-ray beam, and removal of the 8 dependence resulting from the changing sample volume ill~minated.~ The sample container and back- ground scattering is subtracted using a suitably corrected 'empty cell' data set; correction for sample absorption is then carried out.For a system of N identical atoms the scattered intensity (in electron units), is given by:" wherefis atomic form factor, Q is the scattering vector with magnitude lQl= (4n/A)sin 8 (for a scattering angle 28 and incident X-ray wavelength A), and ramis the distance vector between the positions of atoms n and m. This equation represents both the intra-atomic (self scattering) and the inter- atomic scattering (interference term) of the system. In the experimental data a third term is also collected which includes the inelastic scattering produced by the system; this can be calculated using tables14 and removed.When there is more than one atom type, an approximate monitor method can be used to calculate the scattered intensity by Fig. 1 Experimental configuration for the shallow-angle diffraction choosing a convenient 'unit of composition', uc, for the arrangement material. We can then define an average scattering factor per 338 J. Muter. Chem., 1996,6( 3), 337-342 electran? c fm UCf,=-c2, UC where the sum over uc represents the weighted sum over the atoms of atomic number 2,. The form factor for each atom type can then be approximated by fm=Kmfe,where K, will be approximately equal to 2,. In each case K, will vary with scattering vector Q,and the validity of this treatment depends on the error in treating K, as an average over the entire Q range involved.For any fixed displacement Y, the electron density averaged over all directions is given by pj(r) where the subscript j represents the atom type; this shows fluctuations from the average electron density of the sample, pe. For an amorphous material with no preferred orientation, spherical symmetry can be assumed and the integral over volume can be reduced to: The left-hand side of the equation is readily obtainable from experiment and is known as the structure factor S(Q).We then define an interference function i(Q)by rm Qi(Q)=4.n I Kjr[pj(r)-pel sin Qr dr (4)Jo uc Inverting this by Fourier transformation we obtain the com- bined radial distribution function, which may be rearranged to give the total pair distribution function g(r)= uc=1+ JV 2.n2rpec Kj (5)pe CKj uc uc In practice, it is not possible to measure X-ray diffraction data directly in electron units; an absolute intensity measurement is therefore obtained by scaling the data to oscillate about the theoretical self scattering term, and hence producing the S(Q).Shallow-angle diffraction It is not possible to treat shallow-angle X-ray diffraction data in the same quantitative manner as transmission diffraction data: basic data reduction accounts for detector dead-time, changes in incident-beam current and beam polarisation effects. A further correction is needed to account for the fact that the collected X-ray beam is actually scattered from the refracted beam within the sample; this produces a small shift in the measured scattering angle, 28." More sample-specific correc- tions such as sample absorption and multiple scattering are not included in the reduction procedure for the shallow-angle technique; these corrections are complicated by unknown factors in the sample geometry which make it difficult to determine the actual spread of penetration depths into the sample and/or substrate and the contributions from each.This situation could be clarified somewhat if the incident X-rays did not penetrate the substrate at all; this can be achieved by using thicker films, reducing the incident angle, or increasing the incident X-ray wavelength; however, increasing 1decreases the Q-range and therefore the real-space resolution, and there is little to be gained by reducing cri much below the detector slit resolution.Progress in the longer term is likely to depend on the use of indirect data reduction tools based on Monte Carlo methods. Subtraction of the background scattering in the shallow- angle geometry is also problematic, as there is no direct method of removing the sample and measuring the 'background' scatter. It must therefore be assumed that the background scattering may be approximated by a smooth curve, and can therefore be removed, along with the atomic form factor, by fitting a Chebyshev polynomial through the data. While this method produces an 'interference function' which shows the same peak positions as would be derived by following standard procedures for transmission geometry data,g there is no practi- cable method of converting the data to electron units, and therefore it is not possible to produce absolute coordination numbers from the real-space information.Transmission diffraction Fig. 2 shows the experimental S(Q) data for. the four silica: titania samples, confined to a Q-range <10 A-' for clarity. It is clear that the general shape of all four curves is the same, with only the 'phase separated' curve showing a significant increase in the intensity of the oscillations. Since X-ray scat- tering results from interaction with electrons, scattering will be dominated by correlations involving heavier atoms, i.e.Si, Ti and 0.It can be seen from Table 1 that samples 1-3 show very similar compositions and only a very small amount of titanium is present; sample 4, however, contains significantly more titanium which has resulted in much stronger scattering. Fig. 3 shows the corresponding g(r) curves: despite the similarity between these data and the S(Q) data for the lower titanium content samples, small changes are highlighted when corrections for sample composition are included in the trans- ition to g(r)and the data is normalised to the number of bond pairs. Clear differences in the shape of the curves for the four samples are evident, particularly in the region of theosecond and third neighbours. The first main peak at ca.1.61 A shows 12 3 4 5 6 7 8 9 10 CUB-' Fig. 2 Experimental S(Q) data for the four silica: titania samples measured in transmission geometry I I I 1 0.4 ;L. I I 1 I I 1 2 3 4 5 6 7 r/a Fig. 3 Experimental g(r) data for the four samples in transmission geometry J. Mater. Chem., 1996,6(3), 337-342 339 slight differences in position, but the peaks are similar in width and height, indicating that all four samples have a very similar first-neighbour coordination number Since the S(Q) data covered relatively short dynamic range in this case (0 45-14 A-’), and a heavy windowing function16 was used in the Fourier transform to avoid termination errors, the reso- lution of the real-space tats is relatively low, all correlations between ca 14 and 1 9 A, therefore, are contained within the first peak Thi! region is expected to include correlations from Si-Oo at 1 61 A, the short four-coordinate Ti-0 distance at 182 A (where present), and the C-0 and C-C correlations at 1 43 and 1 53 A, respectively, of any residual ethanol present in the sample The low resolution of the data means that it is not possible to distinguish between first-neighbour Si-0 and Ti-0 distances if the network remains four-coordinated, how- ever, if phase separation occurred and regions of pure six- coordinated tit?nia were formed, the longer Ti-0 correlation length of 193 A should be visible in the g(r), resulting in a wider first peak for the ‘phase separated‘ sample Although there is a slight widening of the peak in the ‘phase separated’ sample, it is not evident whether this is statistically significant Clear differences are visible in the g(r) data between ca 22 and 34A, however, corresponding to the 0-Si-0, 0-Ti-0, Si-0-Si, Si-0-Ti or Ti-0-Ti correlations (Table 2) Although the interatomic distances for four-coordinated silica and six- coordinated titania are known, there is some uncertainty concerning the distances within an atomically mixed silica tit- ania amorphous network The distortion which will be pro- duced when a titanium atom with a long Ti-0 distance is substituted into the silica network may result in a slight shortening of the Si-0 distance or, more probably, a distortion in the bond-angle distribution The distances in Table 2 were calculated assuming bond angles and distances are the same in the atomically mixed case as in pure silica, the values therefore will be only an indication of the possible interatomic distances The g(r) dtta for the ‘pure silica’ sample show! a strong peak at 3 12 A with a distinct shoulder at ca 2 65 A, as is the case for the first peak, the silica data appear to show interatomic distances which are slightly larger than those expected this may arise from the relatively poor statistical quality of the original, exploratory data for this sample, which limits the real-space resolution when a severe window function is applied Despite this problem, the data show clearly that correlations in this region result from two separate main interatomic distances, those of 0-Si-0 and Si-0-Si The introduction of a very small amount of titanium in the ‘low titania’ sample results in an immediate change in the shape of the g(r)curve with the two peaks becoming of similar intensity, but with a plateau formed between them This observation suggests that the network has become more complex, with a range of correlations forming This is consistent with a small number of 0-Ti-0 and/or Si-0-Ti bonds forming in the network, indicating the presence of atomically mixed SiO, Ti0, regions The findings for the ‘high titania’ sample follow this trend, with more correlations occurring at longfr distances corresponding to mixed corredations at or above 3 A, and less at the 0-Si-0 distance of 2 6 A The ‘phase separated’ sample does not appear, on averaging over the entire sample, to be structurally different from the lower titania samples The same trend of a slight shift to longer distances is observed, but there is no clear direct evidence for 0-Ti-0 correlations at 2 46 and 2 79 A,as might be expected if phase separation had occurred (although the peaks do appear to be slightly wider, which would be consistent with phase separation) It is a!so interesting to observe the formation of a peak at ca 0 95 A in the ‘phase separated’ sample, which is also visible in the ‘high titania’ sample although it is less intense This peak may result from the 0-H bond distance, being present in either volatiles such as ethanol, or in the main silica titania network Since there is little carbon in the ‘phase separated’ sample (see Table 1) it is likely that the 0-H groups are forming within the silica titania network, and are increasing in number as the amount of titanium in the sample is increased This might suggest that 0-H groups are becoming network terminators to reduce the stresses formed by the distortion of the network at titanium sites, or are forming at the boundary between regions of phase separated silica or titania Shallow-angle diffraction Fig 4 shows the corrected data for the three sol-gel samples after fitting and subtracting a polynomial from the data The scattering from the samples containing titania look very similar, but the ‘pure silica’ sample shows a much stronger scattenng across the whole Q-range Small Bragg peaks due to the $icon substrate are visible at approximately 6 0, 9 5 and 13 5 A-’ in the ‘pure silict’ and ‘low titania’ samples, although only the peak at ca 6A-’ appears to be present in the ‘high titania’ sample, this is likely to be due to the fact that the higher titania film is more electron-dense and therefore the incident beam is attenuated more quickly This observation, coupled with the fact that the silicon peaks are very small, suggests that the penetration depth covered by the incident X-rays is only just greater than the thickness of the films and so penetration into the silicon wafer is very small The sharpness of t>e first sol-gel peak in the ‘pure silica’ data at ca 1 9A-1 may, in addition, indicate contamination from an underlying residual silicon Bragg reflection The interference functions reveal the similanties between the scattering from all three samples after the first major peak It is $ear that the visible Bragg peaks, particularly the one at ca 6 A-l, represent a significant problem if analysis were to continue by way of conventional direct Founer transform to a pair distribution function, their presence could lead to strong pure silica -low htwa ----Oo6 hgh titama0 05I1fl -0 04Y5 003 n a 002 v h9 0010 000 -0 01 -0 02 Fig.4 Experimental S(Q) data for the three lower titania sol-gel samples using the shallow-angle technique Table 2 Interatomic distances for different types of correlations SIO, SIO, Ti02 Ti0, four-coordinate atomically mixed six-coordinate correlatiop 0-Si-0 SI-0-s1 0-Ti-0 Si- 0-Ti 0-Ti-0 TI-0-Ti dist ance/A 2 62 3 06 ca 30 ca 33 2 46, 2 19 3 03 340 J Muter Chem , 1996,6(3), 337-342 silicon correlations in g(r).However, the rapid decay of the dat? to the asymptotic value, after the firstosharp peak at ca. 1.8 A-l and a small second peak at ca. 4.5 A-l, indicates that all three samples show a high degree of disorder. This is demonstrated further in Fig. 5 where scattering from the ‘high titania’ sample in thin-film form (shallow-angle geometry) is compared to scattering from the bulk (transmission geometry); both data sets are at a similar stage of data reduction.Both curves are dominated by a first sharp peak primarily associated with Si-0 first-neighbour correlations, but the bulk sample also shows definite second and third peaks; for the thin-film sample it is very difficult to determine any distinct higher-order correlations, although some evidence of residual structure in that region is visible. Owing to the contamination by silicon Bragg reflections, and the large amount of statistical noise in the data produced by scattering from very small volumes, the information available from a Fourier transform-ation into real-space is limited; Fig. 6 shows the Fourier transform of the S(Q) function for the ‘high titania’ sample as an example of the r-space information obtainable.The S(Q) data is initially weighted by a sharpening function, and then a heavy windowing function is used in the Fourier transform to reduce ripples resulting from statistical noise or termination effects. The strong correlation visible at ca. 1.5 A is associated with the Si-0 distante; in bulk silica the Si-0 first-neighbour distance is 1.61,A; however, the silicon-oxygen distance is reduced to 1.50A when taken out of the confines of the silica network, for example when part of an Si(OH)4 unit. This may be further evidence that the silica network has become more disordered when in a thin film, and, contrary to the case for the bulk material, there are few long silicon-oxygen chains I N1 shallow angle -I-‘I 0.1s I transmission----I A .-v) c3 $ 0.10 v h9-0.05 -----------_---_____ -.--I I I I I I0.000 2 4 6 8101214 0iA-’ Fig.5 A comparison of I(Q) data for the ‘high titania’ sample, meas-ured in both transmission and shallow-angle geometries 0.95 I ,w, II I 1 1 I 012345678 r/A Fig. 6 Example Fourier transform of the shallow-angle S(Q) data for the ‘high titania’ sample and more hydrogen atoms terminating the network. There is little order apparent in the g(r) after the first main peak; in particular, interatomic distances resulting from 0-Si-0 (2.6 A) and Si-0-Si (3.0 A) bonds which am prominent in g(r) data from the bulk silica :titania sol-gels are not visible here. Discussion and Conclusions Preliminary X-ray diffraction data on four silica :titania sol-gel glasses, using a conventional 8:28 ‘powder’ transmission geometry, with compositions ranging from 0 to 42mol% titanium, has provided evidence to suggest that for lower levels of Ti, up to ca.20 mol%, the silica and titania are atomically mixed in fourfold coordination, in agreement with NMR data.3 For higher atomic percentages of titanium, there is no direct evidence for the existence of phase separated areas of six-coordinated Ti, although this cannot be discounted; the increas-ing O-H coordination number in the higher titanium samples does, however, indicate that phase separation has indeed occurred at these elevated Ti concentrations. The bulk pair correlation function data shown was produced using a heavy windowing function in the Fourier transform, which results in lower resolution real-space data.The method used to calculate g(r)also assumes that the self-scattering term for each of the atom types does not vary far from that produced for one ‘atomic unit’ of the material; when atoms of very differing atomic number are present in a sample this is a poorer approximation, and this may cause some error in the relative intensities of the peaks in g(r). Nevertheless, we are able to conclude that: (i) there is an increase in the number of 0-Ti-0 and Ti-O-Si bonds with increasing Ti content below ca. 20mol%; this indicates that the material is atomically mixed; (ii) for higher Ti contents (above ca. 40 mol%) there is no direct evidence for Ti-O-Ti bonds or phase separation, although there is a slight increase in the average first neighbour distance; and (iii) the number of O-H bonds increases with increasing Ti content suggesting that there is more network termination; this may be indicative of phase separation.In addition, the more novel shallow-angle X-ray diffraction method has been used to examine three silica: titania sol-gel thin films. Although difficulties arising from contamination from the substrate can reduce the quantitative nature of the final data, it is still possible to make clear qualitative statements about the structure of the films in comparison with their bulk counterparts. There is evidence for a higher degree of disorder in the silica network of the thin films, with many Si-0 bonds but a reduction in the number of rigid Si-0-Si chains com-pared to the bulk.An apparent shortening of the Si-0 first-neighbour distance may indicate that there is a tendency for more Si-O-H bonds to form, possibly associated with cracks in the stressed films; this is consistent with observations of the properties of sol-gel thin films discussed earlier.8 It was not possible to measure quantitatively the small differences in structure between the ‘pure silica’ sample and those containing titania; this might be expected considering the low titania content and the fact that it bonds substi-tutionally into the silicon network at these low titanium levels. Moreover, it is unlikely that anything other than major differences could be observed with the shallow-angle technique until a method is devised of allowing an exact background subtraction process, and this remains a limitation to the usefulness of the technique when applied to the study of amorphous materials.The penetration depth of the X-rays into the sample/substrate assembly would also have to be reduced, using either thicker samples (in which case the films might be structurally different from those used in real appli-cations), or higher wavelengths. The method also collects data from a range of scattering volumes, dependent on the penetra-tion and film thickness; this results in a range of absorption and multiple scattering effects and therefore a blurring of the J.Mater. Chem., 1996, 6(3), 337-342 341 data The method does, however, have some potential in the study of amorphous thin films and coatings where sample processing or treatment induces more substantial structural effects (eg crystallisatiam, phase separation or the loss of volatiles from a single sample) that can be followed zn sztu We acknowledge the CCRL Daresbury Laboratory for the provision of Synchrotron beamtime on Station 9 1 References C J Bnnker and G W Scherer, Sol-Gel Science, Academic Press, Boston, 1990 M Itoh, H Hatton, and K J Tanabe, J Catal, 1974,35,225 M E Smith and H J Whitfield, J Chem SOC Chem Commun, 1994,723 P J Dirken, M E Smith and H J Whitfield, J Phys Chem, 1995, 99,395 5 T J Bastow, A F Moodie, M E Smith and H J Whitfield, J Mater Chem ,1993,3,697 6 J-J Cheng and D-W Wang, J Non-Cryst Solids, 1988,100,288 7 M Emih, L Incoccia, S Mobiho, G Fagherazzi and M Guglielmi, J Non-Cryst Solids, 1985,74, 129 8 D E Bornside, C W Macosko and L E Scnven, J Imaging Technol, 1987,13,122 9 W P G Lim and C Ortiz, J Mater Res, 1987,2,471 10 T M Burke, D W Huxley, R J Newport and R Cernik, Rev Sci Instrum ,1992,63, 1150 11 T M Burke, PhD Theszs, University of Kent, 1994 12 G Bushnell-Wye and R Cernik, Rev Scz Instrum, 1992,63,999 13 D W Huxley, PhD Thesis, University of Kent, 1991 14 International Tables for Crystallography Vol Ill, ed C H Macgillavry and G D Rieck, Kynoch Press, Birmingham, vol 3,1968 15 B E Warren, X-ray diffraction, Dover Publications, New York, 1990 16 J Waser, Rev Mod Phys ,1953,25,671 Paper 5/05332C, Received 9th August, 1995 342 J Mater Chem , 1996, 6(3), 337-342