J. MATER. CHEM., 1991,1(3), 381-386 38 1 Electrochromic Nb,O, and Nb,O,/Silicone Composite Thin Films prepared by Sol-Gel Processing G. Rob Lee and Joe A. Crayston* Department of Chemistry, Purdie Building, University of St Andrews, Fife KY16 9ST, UK Alcoholysis of NbCI, in EtOH gives a viscous solution of NbCI,(OEt),-, which may be used to spin-coat optically transparent electrodes (indium tin oxide, ITO). The electrode is converted to Nb205 by immersion in 1 mol dm-3 H2S04.Electron microscopic examination of the resulting 5-10 pm thick films revealed extensive shrinkage and cracking on gel drying. These films proved to be electrochromic (A,,, ca. 800 nm) but unstable to voltammetric cycling in MeCN-(0.5 mol dm-3)LiCI04. Greater stability (>30 cycles) was achieved with a composite Nb205/ silicone electrode, prepared from a solution containing Nb :Si :H20 in the molar ratio 5 : 13:82.This increase in durability was achieved with no sacrifice in electrochromic efficiency. The electrochromic colouration efficiency (6 cm2 C-') compares favourably with that for sputtered Nb205 films. The colouration-bleaching cycle of the composite is complete in <40s, and the electrochromical kinetics of colouration and bleaching follow a diffusional model with DLi=6 x lo-' cm2s-' for Nb20, and 3.2 x lo-' cm2s-' for the composite. The mechanism of the electrochromic response and its decay are discussed. Keywords: Thin film; Electrochromism; Sol-gel processing; Niobium oxide An electrochromic material is one in which a visible absorp- tion band can be introduced reversibly by the passage of current through the material, or by application of an electric field. These materials are of interest for display devices.' A major requirement of an electrochromic material is that it exhibits mixed conduction, i.e. both electronic and ionic conductivity.Typically, a metal redox centre is generated by injecting electrons, and for charge neutrality to be preserved cations must be admitted into the material. Since it is usually the movement of ions which is rate-limiting, materials with fast-ion conduction at room temperature are desired (usually with H+, Li' or Na' ions).' Often such materials are amorphous and of low density so as to provide an open framework for rapid ionic diffusion.Amorphous materials also have the advantages of flexibility of composition, high concentrations of vacancies for ionic motion and are generally more easily prepared as the thin coatings necessary for display devices. Materials which have been investigated include Wo3,3 Prussian blue,4 and organic molecules such as alkyl viologen~.~ Colouration in transition-metal oxides such as W03 is believed to arise from intervalent charge-transfer bands at W5+ sites, and whilst W03 has been extensively studied, other oxides, such as Nb205 have received much less attention. Thin Nb205 films have been prepared by thermal6 (500 "C) or anodic7 oxidation of the metal, and by oxidation of sputtered NbN,.' Unfortunately, the metal oxidation routes preclude the use of the electrochromic films in most display devices where the metal oxide is sandwiched between optically transparent electrodes; however, they can be used in reflec- tance mode.Finally, it is also possible to sputter Nb205 thin fiims.9 In this paper we describe a new route to electrochromic Nb205 films using the sol-gel method. This method involves the controlled hydrolysis of high-purity metal alkoxides. It has the advantages of control of porosity and structure, convenience and the ability to coat large objects with adherent The method has already been employed successfully in the synthesis of V205, TiO2I2 and Wo3I3 electrochromic devices. The sol-gel process can only be used to make crack- free films of thickness 10.5 pm.Typical electrochromic efficiencies are 10 mC cm-' for V205, corresponding to an absorbance of ca. 0.5 in a film of thickness 0.5 pm." For greater display densities a method for producing thick (5 pm), crack-free electrochromic films with fast response times would be desirable. We report the fabrication of thick-film Nb205 electrodes and their stabilisation by overcoating with a sili- cone-based gel to give a more durable composite electrode. This composite is related to the organically modified ceramics or silicates (ORMOCERS or ORMOSILS) introduced by Schmidt.I4 Experimental Electrode Fabrication Nb2O5/I??3 Electrode A5cmxlcmx0.3cm IT0 glass electrode was ultrasonically cleaned then coated with a 50 mm3 drop of 1 mol dm-3 NbC15 (Aldrich) in spectroscopic EtOH (Aldrich), and this was then spun at 1250rpm for 3s on an upturned Pine Electrodes Rotator, so that 2 cm2 was covered. The electrode was then placed in 1 mol dm-3 HzS04 for 1 min for complete hydrolysis and gelation.This gave an even coating ca. 5 pm thick as measured by micrometer and scanning electron microscopy (SEM). The edges of the film were ca. 5-10 pm thicker than the central part. Excess water was very gently blown off with a low-velocity air stream at an angle of ca. 10" to the gel. The edges were stripped off and replaced by silicone grease in order to mask the exposed underlying ITO. The final Nb205/IT0 electrode area was 1.2 cm'. NbzOs/Silicone Composite Electrode An Nbz05/IT0 electrode was prepared as above, but without the silicone grease applied to the exposed ITO, and was then immersed for 3 s in a composite Si-containing sol.The com- posite sol contained 10.0 cm3 0.4 mol dm-3 NbCIS (Aldrich) in spectroscopic-grade EtOH (Aldrich), 2.3 cm3 3-(2,3-epoxy- propoxy)propyl(trimethoxy)silane (Glymo) (96% Aldrich) and 1.17 cm3 deionised water. The resulting solution thus con- tained Nb :Si :H20in the molar ratios 5 :13:82. The electrode was then spun at 1250 rpm for 30 s to remove excess sol and dried at 60 "C for 3 h. Electrode Characterization SEM was performed on a JEOL JSM 35CF (15 keV) at the Gatty Marine Lab., which had a scanning EDAX attachment. Further characterization included Fourier transform infrared (FTIR) (Perkin-Elmer 1710) and Raman spectra (Spex Rama- log).Chloride analysis was performed according to a standard digestion and potentiometric titration method." Niobium analysis was performed according to a standard combustion met hod. ' Electrochemical and Optical Measurements Cyclic voltammetry and chronoamperometry were carried out using a combined pulse/scan generator and potentiostat (Pine Instruments RDE4 or homebuilt) and a Graphtec XY recorder. The three-electrode cell configuration is shown in Fig. l(a). Electrolytes used were, 1 mol dmd3 H2S04 or 0.5 mol dm-3 BuiNPF6 (Aldrich recrystallised) in HPLC grade acetonitrile (Aldrich). All potentials are reported us. SCE (standard calomel electrode).The cell was N2 purged before use. Chronoabsorptiometry data and spectra were recorded using the cell depicted in Fig. l(b), which could be inserted directly into the cell block of a Philips PU8720 UV-VIS Spec- trometer. Here a Pt quasireference was used instead of the SCE. Results and Discussion Nbz05Films Nb205 films were prepared by the spin coating onto IT0 optically transparent electrodes of 1 mol dm- NbC1,-EtOH solution, which was subsequently hydrolysed in acid. The NbC1,-EtOH solution has been shown to contain chiefly NbCl(OEt), and NbC12(OEt)3.'6 Thus the hydrolysis follows: NbCl,(OEt),-,+ 5H20++Nb205 +(5 -x)EtOH+xHCl X-Ray diffraction, EDAX, C1- and Nb analysis confirmed that the resulting film was hydrated amorphous Nb,O,; no evidence could be found for remaining Nb-Cl bonds after hydrolysis.The FTIR spectrum of the gel, after drying at 150°C, showed that all the EtOH and EtO-had been removed by the hydrolysis reaction and subsequent drying. A very broad band (600-700cm-') due to antisymmetric Nb-0 stretching in the distorted Nb06 octahedral7 is observed. A shoulder at ca. 900cm-' may be assigned to v(Nb=O). Three weak bands observed at 1180, 1120 and 1050 cm-' correspond to bands claimed to be diagnostic of ageing of Nb oxide precipitates.I8 Specifically, the 1180 and 1050cm-' bands are assigned to the Nb-(OH)-Nb group.'* The Raman spectrum of the as-formed gels contain bands due to the EtO group, for example v(C0) at 1090 cm-l. Weak bands below 1000 cm -(932,760 465 cm- I) correspond to those of B-Nb2O5.I7 No bands due to Nb-C1 stretching insulator -cuvette (a) light (b1 Fig. 1 Cells used to study sol-gel-processed Nb,05 films: (a)electro-chemical measurements; (b) absorbance measurements using the same electrode assembly J.MATER. CHEM., 1991, VOL. 1 vibrations were observed. After heating the gel to 900 "Cthe Raman spectrum is identical to that of H-Nb205. After these thick films (5-10 pm) had been dried at room temperature it was found that substantial cracking and peeling occurred, pulling the film from the electrode. Electron microscopy revealed that the gels had suffered shrinkage and cracking to give isolated islands of ca. 10 pm in diameter [Fig. 2(a)]. It is known that such shrinkage in dried sol-gel processed films is caused by capillary stresses due to surface tension." For this reason the soft hydrated amorphous Nb205 films proved unstable to potential cycling in cells containing non-aqueous solvent.SEM upon cycled films that failed [Fig. 2(b)J shows a similar set of cracking and islands of ca. 10 pm diameter to those found in the dried films. The films proved to be a little more stable when 1 mol dm-3 H2S04 (aq.) was used as the solvent/electrolyte. In order to increase the longevity of the electrochromic films, several modifications of the procedure were attempted. (a)Using the method of Alquier et a!.," 1 mol dm-3 NbC15 in ethylene glycol was used as the sol precursor to gelation. Although electrochromic films were produced by this method, their durability was actually less than that of the Nb205*xH20 films described above.(b) Mixed Nb-Si gels were produced by the addition of the trialkoxysilane, 3-(2,3-epoxypropoxy)propyl(trimethoxy)silane(Glymo) to the niobium chloroalkoxide precursor, followed by hydrolysis. Glymo is known to form flexible rubbery gels suitable for applications such as soft contact-lens material^.'^ However, as the niobium chloroalkoxides have a widely different hydrolysis rate to Glymo it was found that at Nb :Si ratios of >I hydrolysis led to particulate Nb205 formation, coated in a cracked hydrolysed Glymo film, whereas at Nb :Si ratios of I1, transparent gels were obtained with dramatically reduced electronic and ionic conductivities.No electrochromic response was observed for any of these films. (c) Finally the composite system, described below, gave more robust electro- chemically active films. Nb205Films with Glymo Composite A 0.05 cm3 drop of 1 mol dm-3 NbC15 in spectroscopic-grade EtOH was spread over 2cm2 of IT0 glass, spun on an upturned rotating-disc electrode, typically at 1250 rpm for 3 s. The IT0 electrode was then immersed in 1 mol dmm3 H2S04 for 1 min to catalyse the hydrolysis and gelation of the sol. The second coating of Nb :Si :H20 solution of molar ratio 5 :13 :82 was applied by dipping the electrode in the solution, spinning for 3 s at 1250 rpm and drying at 60 "C for 3 h. These films showed significantly greater durability in non- aqueous solvents but broke up on immersion in water/acid.The films were rougher in morphology [Fig. 2(c)] than the uncoated Nb205 films but did not break up on drying. EDAX revealed that the Si in the second coating had penetrated throughout the full 5 pm thickness of the film; thus, we prefer to describe the electrode as a composite rather than a hetero- geneous two-layer structure. Electrochemical failure was accompanied by the same type of microcracking as in the uncoated Nb205 films [Fig. 2(d)]. During drying the role of the Glymo may be one or both of the following: (1) it provides flexibility during the drying process thus preventing large capillary forces from cracking the material; (2) the epoxide chains on the silicon may act as a hydrophobic barrier, forming less porous pore walls and thus trapping H20 in the metal oxide and hence slowing the drying process.A full range of composite molar ratios were tested for the second coating, as were spinning speeds and times; however, most of the films thus produced failed for a number of reasons: reproducibility, film cracking and loss of electrochromic properties. J. MATER. CHEM., 1991, VOL. 1 Fig. 2 SEM photographs of (a) Nb205*xH20 film after drying at ambient temperature for 4 h; (b) Nb20S*xH20 film after failure during electrochemical potential cycling (five cycles) in 0.5 mol dm-3 LiC10,MeCN; (c) Nb20=,*xH20 film with Nb/Si/H20 (8: 12:80 mole ratio) composite after drying at 60"C for 3 h; (d) film as in (c) after failure during prolonged potential cycling in 0.5 mol dm-3 LiClO,/MeCN Cyclic Voltammetry Cyclic voltammograms (CVs) run at 100 mV s-l of both the coated and uncoated films on IT0 electrodes in 0.5 mol dm-3 LiC104-MeCN are shown in Fig.3(a)and 3(b).The potential range covers the potentials controlling the colouration and bleaching processes. The onset of the blue colouration is observed when the potential reaches -0.4 V. Both CVs exhibit a very large cathodic current at potentials less than -0.4V but this is presumably due to H2 evolution at the ITO/ 'L'& .-Nb205 *xH20 interface rather than colouration current. On +1 .o +1 .o the reverse sweep an anodic peak is observed at -0.5 V vs.SCE , Icassociated with the bleaching of the electrode. Roughly 50% of the composite electrodes required 2-3 cycles of 'breaking in' or 'condhbning' before maximum contrast was observed between the bleached and coloured absorption.This is a common phenomenon seen in electrochromic films due to the solvent opening of the pores. There' appeared to be no apparent morphological differences between the samples; how- ever, the composite sol has to diffuse through the pores and undergo gelation within them. This would give rise to smaller pore sizes and presumably partly closed pores, which would need to be broken into. Similar CVs are observed for the uncoated Nb205 films in 1 mol dm-3 H2S04(aq); however, the coated films are unstable in aqueous media. The optical absorption changes on colouration of the Fig.3 Cyclic voltammetry (100mVs-') of (a) Nb205 film inNb205 films at -0.875 V (Fig.4) are dominated by the broad 0.5 mol dm-3 LiClO,/CH,CN; (b) composite Nb,O,/silicone elec- structureless band which peaks at A >800 nm. This absorption trode. The cross indicates zero current and potential 384 0.6 h .-8 0.4 C3 -? Y 0.3 C 9 sntu '0.2 0.1 10.0 400 450 500 550 600 650 700 750 800 Ilnm Fig.4 UV-VIS spectral changes in NbzO5/ITO absorbance in (a) the bleached state (0 V us. SCE) and (b) the coloured state (-0.875 V us. SCE). Medium 1.0 mol dm-3 H2S04, film thickness ca. 5 pm is probably associated with Nb bronze formation xLi +xe-+Nb2OSS Li,Nb20S Simultaneous injection of electrons and cations is necessary to maintain charge neutrality.The reaction is related to the same process in crystalline Nb bronzes.20 The kinetics of intercoversion between bleached and coloured states is described below. Durability For the uncoated Nb20s films the absorbance maximum of the coloured state (-0.875 V) decays very rapidly with the number of cycles (100 mV s-') as shown in Fig. 5, while simultaneously the background absorbance of the bleached 0.29 0 Oa30!0.28 + 0.27..-E 0.26. C 0.25. x E 0.24' a, 0.231 /.A n / 12345678910. cycles Fig. 5 Durability of the electrochromic response. Dependence of the absorbance at 800nm in the coloured state (upper traces) and the bleached state (lower traces) as a function of the number of cycles: pure Nb205 electrode (--a); composite Nb,O,/silicone electrode (-sample 1; ----sample 2).Media: Nb205, 1 moldm-' H2S04; composite Nb,O,/silicone, 0.5 mol dm-3 LiClO,/MeCN J. MATER. CHEM., 1991, VOL. 1 state increases, possibly as a result of incomplete bleaching, or more likely owing to the increased light-scattering proper- ties of the cracked film. The two states become indistinguish- able after only four cycles, and the electrode suffers the dramatic peeling referred to earlier [Fig. 2(b)]. As already noted, different behaviour is observed for the Glymo Nb2OS composite film (Fig. 5). Here, there is often a rise in absorbance of the coloured state to reach a maximum at about the third cycle (breaking in) followed by a steady decay. Alternatively, a steady decay from the initial absorbance maximum on the first cycle is observed.For both cases after prolonged cycling (typically 30 cycles) the two states are indistinguishable and the electrode is deemed to have failed. Visually, at this point, only the outer edges of the electrode have been observed to retain any of the blue bronze formation. The morphology of the film suggests cracking and peeling [Fig. 2(d)] in a similar way to that of the uncoated electrode. We propose two possible mechanisms for the rapid failure of these films: (i) peeling due to H2 evolution on the cathodic sweep; (ii) rapid dehydration of the hydrated oxide layer, either by H20 diffusion into the non-aqueous solvent, or by H20 binding as water of hydration to the Li+ ions cycling in and out of the film.It is interesting to note that uncycled Nb20S films left in MeCN for a few hours will, upon removal, show similar cracking patterns to those shown in Fig. 2(b), lending credence to (ii). Those left in sulphuric acid tended to peel in large area pieces from the IT0 implying that here dehydration of the oxide film was not so much of a problem as the formation of a water layer between the film and the ITO. That the composite dramatically increases cycling lifetime can thus be explained. If (i) is operating then the composite may be slowing down the diffusion of H+ (however, we show later that the diffusional coefficients for Li' in the two electrodes are comparable); if (ii) is operating then the com- posite may be serving as a barrier to H20/MeCN exchange and dehydration, while still allowing Li' into the film.If this is so then it may be possible to eliminate film failure during cycling by the predrying of the films at a reduced drying rate. This is achieved (and commonly in industry) by drying the films in a stream of air saturated with the solvent used in the sol. Kinetics of NbzOSElectrocbromic Films in LiCIO,/MeCN The kinetics of colouration and bleaching may be studied by stepping the potential between the bleached state (OV us. SCE) and the coloured state (-0.8 V us. SCE) and back, whilst monitoring the time dependence of the current (chronoamper- ometry) and the absorbance at 800 nm (chronoabsorptio-metry).Chronoamperometric results for the Nb2OS and the Nb205 silicone composite electrodes [Fig. 6(a) and (b)] revealed that for both electrodes the full cycle is complete in ca. 40 s. In each case the bleaching current decays considerably faster than the colouration current. Furthermore, the colour- ation charge (area under the i-t curve) is much larger than the bleaching charge. This is probably due to hydrogen evolution on the cathodic potential excursion. Note also that the colouration current rise time of the composite is very sluggish, owing to the high initial electronic resistance of the bleached composite film. In addition, chronoabsorptiometric data are shown for the composite electrode [Fig.6(c)].The Coulombic efficiency (q)for electrodes is given as: J. MATER. CHEM., 1991, VOL. 1 -0.8 V Fig.6 Current-time curve for (a) pure Nb205 and (b) composite Nb,O,/silicone electrodes on stepping the potential from the bleached state (0 V us. SCE) to the coloured state (-0.8 V us. SCE) as in the left- hand side of the diagram, and the back to the bleached state (right- hand side). 0.5 mol dm -LiClO,/MeCN; (c) chronoabsorptiometric curve (absorbance us. time) for the composite Nb,O,/silicone electrode where AA is the difference in absorbance between bleached and coloured states, a is the area of the electrode (cm2) and Q is the total charge passed during colouration. For the composite electrode the charge passed during colouration is 4.9 mC.Thus the Coulombic efficiency for this 0.6 cm2 electrode is 6 cm2 C-'. This value is comparable to the ca. 10 cm2 C-' reported earlier for sputtered Nb205.21 Colouration Kinetics Fig. 7 shows for both the Nb205 and the composite films that the decaying part of colouration i-t curve follows a t-f. dependence. For the composite electrode, during the initial few seconds when the colouration current rise-time is sluggish owing to high initial resistance, the curve deviates from this t-f. dependence. If small overvoltages were employed the t-f. 0:O 0:2 0.4 0:s 0:s 1:O 1.2 1.4 116 f1/2/S-1/2 Fig. 7 Typical i-t-* plots for the decaying portion of the colouration current transients of pure Nb20, (m) and composite electrodes (0) dependence is explained by taking Li' injection at the film/ electrolyte interface as the rate-determining step." However, since we used large overvoltages the current is limited by the diffusion of Li' within the film. The diffusion coefficient (DLi) may be obtained from the slope of the i us. t-f.graphs because in diffusion-controlled systems the Cottrell equation applies: Q Df.d2i( t)=tot.tl-dtf.' D Qtot is the total charge (in C) passed during the colouration step, t is the time taken (s) and d is the thickness of the film (cm). Such plots should extrapolate to the origin; however, the plot for the composite electrode (Fig. 7) intercepts the negative current axis at infinite times. This may reflect lower than expected current values owing to the resistance of the electrode.For each of the two types of electrode the derived diffusion coefficients are independent of the particular cycle taken. These results (Table 1) suggest that diffusion is slower in the com- posite than in the pure Nb205 film. A possible explanation for this is that the silicone layer has diffused and gelled in the very same pores and grain boundaries through which the Li' is trying to move. Finally, the diffusion coefficients for both elec- trodes compare favourably (Table 1) with those obtained for Li20/Nb205 films' and sputtered Nb205 films.g Absorbance us. time transients for the composite electrode may be analysed by assuming that the absorbance is pro- portional to the charge passed.Thus by integrating eqn.(2) we obtain A= 2vSt' + A0 (3) where q is the electrochromic efficiency, S is the slope of i (t-*) graph and A. is the absorbance at t =0. However, plotting A us. tf.gives a sigmoidal curve indicating a more complex depen- den~e,~~similar to that obtained for sputtered films.' Kinetics of Bleaching According to Faughnan et bleaching of W03 films should be enhanced by migration of ions through the field developed across the electrochromic film, and thus the current is expected to decay as t3 rather than as the tf.of the diffusional model. log i us. log t plots for bleaching of our uncoated and composite- coated films appear to show that for the uncoated Nb205 a diffusional process is occurring [Fig.8(a)], whereas for the composite film a migratory process is the dominant form [Fig. 8(b)].The reason for this difference is not clear and further work will be directed at trying to answer this point. The log-log plots should indicate the point (tf) at which the film is completely bleached,24 but for our electrodes the tfpoint is not clearly defined. However, we may obtain a minimum value for tffrom the point at which the curve deviates from the -3 slope. This will give a maximum value for D. In the migration model this point is given by2'*24 where p is the volume charge density of the cations (obtained from the charge per cm3passed during colouration), d is film thickness, E, is the relative permittivity of Nb205, E, is the Table 1 Derived Li' diffusion coefficients for the colouration cycle DLi+/cm2s-' pure Nb20, composite Nb,O,/silicone 5050 Li20/Nb205* sputtered Nb205' 6.0x 3.2 x 3.4 x 1.5 x lo-* J.MATER. CHEM., 1991, VOL. 1 Table 2 Derived Li+ diffusion coefficients for the bleaching cycle in cm2 s-pure Nb205 diffusion model migration model 1.0x 10-7 3.5 x -0.4 -0.2 0.0 0.2 0.4 0.6 0.8 log (t/s) -0.4 -0.2 0.0 0.2 0.4 0.6 0.8 log (Us) Fig. 8 log-log plots of the bleaching current vs. time for: (a)the pure Nbz05 electrode (0, 1st cycle; +, 2nd cycle); (b) the composite electrode (0,1st cycle; + is 3rd and is 5th) permittivity of free space, p is the ion mobility in the film (D=RTp/F) and Vis the applied bleaching voltage across the film.For example, using p = 1.75 C cm- 3, d =5 x cm, E, = 1 15, E, =8.85 x 10-l4 F cm-V= 0.8875 V, tf=2.5 s we obtain the values of D shown in Table2. These are compared to values obtained by a simple diffusional model: d2 tf =5 Clearly the diffusional model appears to be more in keeping with the data obtained from the colouration measurements. Assuming absorbance is proportional to the charge passed and integrating the current-time relations for both models, we should find for the diffusional model that Acct* and for the migration model that Acct*. However, a log (A-A,) us. log t plot does not reflect either of these simple relationship^,^^ and must await a better model. Conclusions We have shown that it is possible to make large-area thick- film displays from sol-gel-processed Nb205, but that these composite Nb,O,/silicone diffusion model migration model 1.0 x 10-7 5.8 x cells are extremely unstable to cycling and thus their durability is limited.It is possible to create more durable cells by the application of a mixed Nb/Si second coating. This coating does not appear to detract from the electrochromic properties of the Nb205 films, as these composite films show similar Li colouration diffusion properties to the uncoated film and + similar Coulombic efficiencies to earlier reported Nb205 films.2' This would indicate that the durability of thick sol- gel-processed films can be increased without the loss of their electrochromic properties, i.e. Coulombic efficiencies and response times are similar to those of Nb205 films created by sputtering techniques.We thank BP for a research studentship (G.R.L.), Irwin Davidson for the SEM photographs, Dr. T. J. Dines (Dundee University) for providing the Raman spectra and Dr. Ian Thompson (BP)for helpful discussions. References 1 M. Green and K. Kang, Displays, 1988, 9(4), 166. 2 W. A. England, M. G. Cross, A. Hamnett, P. J. Wiseman and J. B. Goodenough, Solid State Ionics, 1980, 1,231. 3 B. Reichman and A. J. Bard, J. Electrochem. SOC., 1979,126, 583. 4 K. Itaya and I. Uchida, Acc. Chem. Res., 1986, 19, 162. 5 R. E. Malpas and A. J. Bard, Anal. Chem., 1980,52, 109. 6 B. Reichman and A. J. Bard, J. Electrochem. SOC., 1980,127,241. 7 (a) S.Rigo and J. Siejka, Solid State Commun., 1974, 15, 259; (b) T. Hurless, H. Bentzen and S. Homkjol, Electrochim. Acta, 1987, 32, 1613; (c) R. M. Toresi and F. C. Nant, Electrochim. Acta, 1988,33, 1015; (d) C. K. Dyer and J. S. L. Leach, J. Electrochem. Soc., 1978, 125, 23. 8 R. Cabanet, J. Cliaussy, J. Mazuer, G. De la Bouglise, J. C. Jubert, G. Barral and C. Montella, J. Electrochem. SOC., 1990, 137, 1444. 9 e.g. N. Machida, M. Tatsumisaga and T. Minami, J. Electrochem. SOC.,1986, 133, 1963. 10 C. J. Brinker and G. Scherer, Sol-Gel Science, Academic Press, New York, 1989. 11 J. Livage, Prog. Solid State Chem., 1988, 18, 259. 12 M. Nabari, S. Doeff, C. Sanchez and J. Livage, Mater. Sci. Eng., 1989, 133, 203. 13 (a) A. Chemseddine, R. Monneau and J. Livage, Solid State Zonics, 1983, 5, 357; (b) C. Sanchez, J. Livage, M. Henry and F. Babonneau, J.Non-Cryst. Solids, 1988,100,65; (c)0.Kamaguchi,D. Tonihisa, H. Kawabata and K. Shimizer, J. Am. Ceram. SOC., 1987, 70, C94; (d)K. Yamanaka, J. Appl. Phys., 1981, 20, L307. 14 H. Schmidt, Inorganic and Organometallic Polymers, ACS Sym- posium Series 360, Washington DC, 1988, ch. 27, p. 333. 15 A. I. Vogel, Textbook of Quantitative Inorganic Analysis, Long-man, London, 1971. 16 G. R. Lee and J. A. Crayston, to be submitted for publication. 17 A. A. McConnell, J. S. Anderson and C. N. R. Rao, Spectrochim. Acta, Part A, 1976, 32, 1067. 18 M. Dartiguenave and Y. Dartiguenave, Bull. SOC.Chim. Fr., 1968, 171. 19 C. Alquier, M. T. Vandenborne and M. Henry, J. Non-Cryst.Solids, 1986, 79, 383. 20 R. J. Cava, D. W. Murphy and S. M. Zahurak, J. Electrochem. SOC.,1983, 130, 2345. 21 R. D. Rauh and S. F. Cogan, Solid State Ionics, 1988, 28-30, 1707. 22 R. S. Crandall and B. W. Faughnan, Appl. Phys. Lett., 1976, 28, 95. 23 G. R. Lee and J. A. Crayston, 1990, unpublished results. 24 B. W. Faughnan, R. S. Crandall and M. A. Lampert, Appl. Phys.Lett., 1975, 27, 275. Paper 0105 1201; Received 14th November, 1990