Sol-gel-derived vanadium and titanium oxides as cathode materials in high-temperature lithium polymer-electrolyte cells Andrew Davies? Richard J. Hobson,*a Michael J. Hudson,*a William J. Macklinb and Robin J. Neatb aDepartment of Chemistry, University of Reading, Whiteknights, Reading, UK RG6 2AD bApplied Electrochemistry Department, AEA Industrial Technology, Harwell Laboratory, Oxfordshire, UK OX11 ORA Binary and ternary vanadium- and titanium-containing xerogels have been prepared by hydrolysis of the metal isopropoxides and subsequent condensation. Powder X-ray diffraction (XRD) has been used to show that pure gel-derived titanium(1v) oxide possesses a structure resembling poorly crystalline anatase, whereas for all the vanadium-containing materials prepared in isopropyl alcohol solution the data are consistent with some two-dimensional (2D) order and a similar short-range arrangement of V05 moieties to that in crystalline V205.In contrast, th: vanadium oxide xerogel obtained from an aqueous milieu shows evidence only of one-dimensional order, interlayer distance 14.2 A. Thermal analysis and XRD have been used to show that all the vanadium-containing gels lose water in three stages and that no structural change occurs until the third stage of water loss, which occurs simultaneously with crystallisation. The oxides have been employed as the active component of the cathode in lithium polymer-electrolyte cells operating at 120 "C and their cycling performance has been investigated. The binary oxides showed no improvement in performance over similar crystalline materials whereas the ternary materials, whether chemical or physical mixtures, showed good reversibility and gave observed energy densities which compared favourably with that of V6013 in a similar cell.This improvement in performance has been attributed to the preferential reduction of the Ti" over VIV near the low-voltage limit which prevents a reorganisation of the microstructure of the material. The discovery by Armand et a!.' that poly(ethy1ene oxide) (PEO) and certain lithium salts (e.g. LiC104, LiCF3S03) reacted to give polymeric solid electrolytes has led to the development of high-energy-density lithium polymer-electro- lyte batterie~~-~ in which the cathode-active material was a lithium insertion or intercalation compound.Previously we have shown5 that vanadium phosphate glasses may be employed as reversible cathode-active materials in secondary lithium polymer-electrolyte cells. The glasses gave higher open- circuit voltages than V6013and exhibited comparable average cell voltages on cycling (2.1-2.4 V) but their specific capacities were lower than that of V6OI3 (ca. 330mA hg-l) at the capacity plateau. This resulted in lower observed energy densi- ties (500-600 W h kg-' for glasses containing 278 mol% V205)after 25 cycles than those observed for V6013(ca.770 W h kg-I). The poorer performance of the glasses can be attri- buted to the presence of the electrochemically inert phosphate component. This suggested that an amorphous V205-based material in which all the components were electrochemically active might give a utilisation comparable to that of V6013.This has previously been achieved in liquid-electrolyte cells with a cathode containing amorphous V205 prepared by the fast-roller quenching method.6 There has, in recent years, been a great deal of interest in the preparation of amorphous materials, particularly oxides, using the sol-gel pro~ess.~The technique is based on the preparation of a colloidal suspension, a sol, and its transition into a gel from which the amorphous material can be obtained. Sols are commonly prepared by the hydrolysis of alkoxides, the rate of which is controlled either by the amount of solvent present, usually an alcohol, or by the addition of a catalyst such as acetic acid, whereas the extent of hydrolysis depends upon the controlled addition of water.Hydrolysed species link t Present address: Cookson Technology Centre, Sandy Lane, Yarnton, Oxfordshire, UK OX5 1PF. together via a condensation reaction, (l), ( RO), -M -OR' + HO-M (OR), --S (RO),-lM-O-M(OR),-,+R'OH; R'=H or R (1) and as the hydrolysis-condensation or polycondensation pro- ceeds larger molecules are formed which ultimately make up the colloidal particles of the sol. Gelation is said to occur7 when a solid skeleton is formed enclosing a liquid phase. Subsequent evaporation of the solvent at room temperature gives the xerogel, the first completely solid material obtained without any heat treatment. The xerogel frequently contains water or solvent and needs heating in order to obtain a pure amorphous oxide, during which the occupied pores may collapse leading to considerable shrinkage of the material.Vanadium oxide gel-derived materials have been studied by Livage and co-workers,8-12 who prepared their gels by poly- condensation of decavanadic acid, obtained from sodium meta- vanadate using cation exchange. The materials were found to possess a layered structure with 1D order, the interlayer separation depending upon the degree of The local ordering within the layers was to that of orthorhombic V205. The subject area has been thoroughly reviewed.13 Hirashima et a1.14-16 used alkoxide precursors to prepare vanadium and mixed vanadium-titanium oxide gels containing up to 20 mol% Ti02.The crystallisation tempera- ture increased and dc conductivity of the xerogels decreased with increasing TiO, ~ontent.'~?~~ On the basis of XRD evi- dence it was suggestedl6 that these materials also possessed layered structures but with some regularity within the layers, with the titanium component possibly acting as a network former within the structure.15 Some studies of sol-gel-derived materials as cathodes in rechargeable lithium polymer-electrolyte cells have been made. For example, V205 xerogel, prepared by polyconden-sation of decavanadic acid, has been shown to give promising re~ersibility'~when a lower limit of 2.2 V was employed during cycling, which corresponds to an Li/V molar ratio of 0.55.J. Muter. Chern., 1996, 6(l), 49-56 However, this depth of discharge led to low capacities and hence only modest energy densities; to achieve a similar performance to V6013 it is necessary for cathodes containing amorphous vanadium oxide to show reversibility following discharges down to 1.7 V. Minnett and Owen'' reported such a study of V,05-TiO, cathode-active materials derived from alkoxide precursors. The materials tested, most of which were annealed at 400°C and consequently were of low crystallinity, gave disappointing capacities in lithium polymer-electrolyte cells. This included V6013, which gave 225 mA h g-' on first discharge compared with 417 mA h g-' for Li8V6013, which suggests that the low capacities may have been a function of cell design or discharge current density" rather than deficiencies in the materials.The cycling data" suggested that better reversibility was obtained with the mixed materials than for V205 alone, an interesting effect which was noted but not considered in any detail. Clearly, additional studies are required in order fully to evaluate the potential of amorphous alkoxide-derived V205 materials as alternatives to V6013,employing secondary lith- ium polymer-electrolyte cells of a design and with a charge- discharge regime which allows high utilisations to be obtained (e.g. 80% theoretical capacity5 for V6013). The effect of Ti02 on the reversibility of lithium insertion into vanadium-based materials is also of particular interest and warrants further investigation.It is unclear whether the titanium acts as a structure modifier or whether its effect is electronic in origin because, unlike phosphorus5 for example, Ti" may be reduced and thus participate in the lithium-insertion process. This paper describes the preparation and characterisation of a series of sol-gel-derived V205-Ti02 materials, including the binary oxides, and their evaluation as cathode-active materials in secondary lithium polymer-electrolyte cells. The synthetic methods were based on the alkoxide hydrolysis- polycondensation method described by Hirashima et u1.'4-16 which is reported to give single-phase material in which the titanium dioxide modifies the vanadium oxide str~cture.'~ Experimenta1 Syntheses The syntheses described in this section give typically 1-2 g of product which were used for initial investigations.For batches of material used in the cell-testing experiments the amounts were scaled-up about sixfold. Vanadium oxide: 'alcohol route'. The method was based on that described previo~sly.'~,~~ Vanadyl triisopropoxide (Alfa) (10 cm3) was added to isopropyl alcohol (20 cm3) with stirring. To the stirred mixture a solution of water (10.6cm3) in isopropyl alcohol (177 cm3) was added dropwise. The stop- pered flask was allowed to stand and a yellow gelatinous precipitate formed within + h which turned green after 24 h. The green precipitate was filtered off, washed with isopropyl alcohol and air dried.No attempt was made to prepare a monolithic gel because it has been shown previously16 that the isopropoxide does not give a monolithic gel when hydrolysed. Vanadium oxide: 'aqueous route'. Vanadyl triisopropoxide (10 cm3) was added with vigorous stirring to distilled water (100 cm3). The inhomogeneous red gel which formed immedi- ately was dispersed by shaking the flask vigorously for 5 min. The resulting red sol was left to stand for 24 h with the flask stoppered. The dark-red gel was then removed from the flask and air dried in order to obtain the solid xerogel which was green. Titanium(1v) oxide. The sol-gel route described by Yoldasl' was used as the basis for the preparation of amorphous TiO, except that Ti(OPri)4 (Aldrich) was employed instead of Ti(OEt),.Titanium(1v) isopropoxide (8.0 g, 0.028 mol) was added with stirring to isopropyl alcohol (115 cm3, 1.53 mol). Nitric acid (0.1 cm3 conc., 0.0016 mol) was diluted with water (2.5 cm3, 0.14 mol) and added to the stirred mixture, then the flask was stoppered and allowed to stand for 24 h to give a transparent, monolithic gel. Mixed vanadium-titanium oxides. Vanadyl triisopropoxide and titanium(1v) isopropoxide were mixed together in the appropriate amounts (Table 1) and added to isopropyl alcohol (20 cm3). Water (5 cm3) as a solution in isopropyl alcohol (85 cm3) was then added. The flasks were stoppered and left for 24 h, after which the resulting gel was removed and air dried in order to obtain the xerogel.Details of the various gel compositions that were prepared are given in Table 1. Techniques Reduced vanadium (V'") and total vanadium contents in the xerogels were determined by potassium permanganate titration and atomic absorption spectroscopy (AAS), respectively. The titanium content of the pure TiO, xerogel was also determined using AAS but this technique was unsatisfactory for analysis of Ti in the mixed gels owing to interference from vanadium. The vanadium-titanium molar ratios of the mixed oxide samples were established using quantitative energy-dispersive X-ray analysis (EDXA) using a JEOL JXA 840 scanning electron microscope. Simultaneous thermogravimetric and differential thermal analyses (TG and DTA) were carried out at a heating rate of 10°C min-l in atmospheres of static air and flowing nitrogen using a Stanton Redcroft STA 1000 instrument equipped with data manipulation software.Powder XRD patterns were recorded using graphite-monochromated Cu-Kcr radiation on a Philips PW1710 diffractometer controlled by a Citrons Cray 112 system running Sietronics 112 software. Data were col- lected for the 20 range 4-64" at 2" min-' in steps of 0.04". IR spectra were measured between 1400 and 300cm-' as KBr discs using a Perkin-Elmer 1720-X FTIR spectrometer. Densities were obtained by pycnometry using toluene or cyclohexane. Table 1 Compositions of the mixed V,O,-TiO, xerogels used found mass/g molar ratio amount (mol%) amount (mol%) VO(OP+), Ti (OPr' ),H,O/( V +Ti) V,O, TiO, V,O, TiO, V(mass%) v'v/vto' 1 5.00 0.15 13.2 95 5 94 6 46.3 0.18 2 5.00 0.32 12.8 90 10 89 11 45.8 0.18 3 5.00 0.73 12.0 82 18 80 20 44.2 0.19 4 5.20 1.01 11.1 75 25 76 24 42.7 0.17 5 4.00 1.oo 13.9 70 30 73 27 42.5 0.20 50 J.Muter. Chem., 1996, 6(l), 49-56 Cell fabrication and cycling The xerogels were ground and sieved to give a maximum particle size of 53 pm and then heated at 200°C for 2 h in order to remove all loosely bound water. The efficacy of the heat treatment was tested using TG; a material was accepted for use in a cathode when it showed an overall mass loss of < 1% below 200°C. For the purposes of calculating the amount of material to be included in the composite cathode the oxides were considered to have the stoichiometries V205 and Ti02.Solid-state cells were constructed and cycled as described previously' using PEO-LiC10, (CEO units]/[Li] = 12) as the electrolyte. Cycling was carried out galvanostatically between 3.5 and 1.7V for all cells except those containing only Ti02 for which limits of 2.5 and 1.2 V were used. To allow direct comparisons to be made with crystalline materials studied previously the theoretical capacities of the cells containing V205 were calculated using the theoretical capacity of V6013, 417 mA h g-l, whereas for those cells containing only TiOz the value for LiTi02, 335 mA h g-', was adopted.Results and Discussion Appearance and characterisation of xerogels and gels Vanadium oxide. A monolithic gel could not be obtained from VO (OPr'), using water-propan-2-01 mixtures, confirming the results of previous work.16 An aqueous medium gave a red monolithic gel after 24 h, showing that a large excess of water is required for gelation," and evaporation of the solvent from this gel gave a green xerogel. A greater proportion of vanadium was reduced (Table 2), during the preparation in alcohol solution, providing evidence for reduction by the solvent, I t 0 100 200 300 400 500 600 100 95 n be-903 6 85 80 Table 2 Compositions of the V,O, xerogels source V,,, (mass%) V'"/v,,, alcohol route 46.3 0.19 aqueous route 47.2 0.11 although the xerogel prepared by the aqueous route shows that reduction may also be effected either by the alcohol liberated during gel formation or via the mechanism suggested by Hirashima et ~1.~'~which relates the reduction to the process of gel formation.The simultaneous TG and DTA curves of the alkoxide- derived V205 xerogels (Fig. 1) were all similar. When the nitrogen purge was replaced by static air the only difference was a slight gain in mass accompanied by a very weak exotherm at about 430°C due to the oxidation of VIV to Vv. The TG curves show three stages of mass loss, (Table 3 and Fig. l), the boundaries and mid-points of which were delineated using the first derivatives. Stage 1, ca. 2O-15O0C, is accompanied by a large endotherm and can be attributed to loss of loosely bound, probably hydrogen-bonded water.Stage 2, ca. 150-270°C, involves too gradual a mass loss to be detected by DTA. However, data collected at a slower ramp rate, 6°C min-', showed no significant change in the relative mass lost in this region, suggesting that it is not simply the result of the overlap of the low- and high-temperature pro- cesses. It is likely that propan-2-01, formed from residual isopropoxide groups from the alkoxide precursor, is lost at this stage. Stage 3 is characterised by a rapid loss of mass which occurs almost concurrently with crystallisation (T,,= 315-325 "C) to give orthorhombic VzOs, see below. The endo- therm associated with the mass loss 'interrupts' the exotherm 100 95 nE!M 1fi 85 80 0 100 200 300 400 500 100 95 0 100 200 300 400 500 600 0 100 200 300 400 500 600 TPC Fig.1 TG (i)and DTA (ii) curves of the V,O, xerogels. (a)Aqueous route; (b) alcohol route; both recorded at a ramp rate of 10°C min-' in an atmosphere of flowing nitrogen; (c) the same as (a)except that it was recorded in static air; (d) is identical to (b)except that a ramp rate of 6 "C min-' was used. J. Muter. Chern., 1996,6( l), 49-56 51 Table 3 Thermal analysis of the Vz05 xerogels ~~~ ~~ ~~ alcohol route T range/ mass "C loss (Yo.)" assignmentb ~~~ ~ ~~~~ ~ stage 1 20- 160 16.2 2.1 HzO stage 2 160-280 2.4 0.3 H20 stage 3 280-325 2.6 0.3 H20 total 20-325 21.2 2.7 H,O Percentages given are averages of two runs and have an absolute estimated aqueous route T range/ "C mass loss (%)o assignmentb 20-145 14.3 1.8 H20 145-265 2.5 0.3 H20 265-330 3.4 0.4H20 20-330 20.2 2.6 H20 error of &0.3%.To one decimal place. The figures quoted represent the maximum amount of bound water assuming that none of the observed mass loss is due to organic material. assigned to the crystallisation (Fig. 1). When the ramp rate was decreased, both the loss of water and the concomitant crystallisation occurred at a slightly lower temperature (cu. 10°C), a change which can be ascribed to a kinetic effect: that both events were equally affected confirms that they are linked. Therefore, the water lost at this stage is either very strongly bound and associated with the structure of the material (for example, water derived from hydroxy groups), or trapped within the amorphous oxide.The amounts of each type of water associated with the two xerogels are similar with the largest relative difference being observed for the third stage. The endotherms associated with this loss of tightly bound water are also different, with the material prepared by the aqueous route showing evidence that the strongly bound water exists in two distinct sites within the structure (at least one of which is thought to be between the layers), whereas for the xerogel prepared in alcohol solution only one type of strongly bound water exists. Three stages of mass loss have also been described for the xerogel prepared by the ion-exchange meth~d.~ However, in that case the loss at stage 2 was better defined and coincided with a structural change, see below.The XRD pattern of the xerogel prepared by the aqueous route [Fig. 2(a)] indicates a lack of long-range order in tke structure. There is a very intense peak at 28 =6.2", d =14.2 A, 1. .l....l....l....l....l,.r,l. 10 20 30 40 50 60 281degrees Fig.2 XRD patterns of V20, gel (aqueous route): (a) as prepared; (b) after heating in air at 200 "C followed by further very broad peaks of low in!ensity centred around 28= 12.5" and 25.1", d= 7.07 and 3.54 A, respectively, which are probably the second- and fourth-order reflections relative to the first intense peak and indicate a degree of 1D order in the structure of the gel.Previous workers,'-13 who examined vanadium oxide gels and xerogels prepared by the ion-exchange method, suggested that the gels possessed layered structures but lacked long-range order within the layers. They assigned c as the interlayer axis and the peaks in Fig. 2 are labellFd accordingly. The interlayer spacing in this case is 14.2 A. Upon heating to 200°C (near the mid-point of stage 2) the XRD pattern of the aqueous xerogel becomes better resolved [Fig. 2(b)] with all axial peaks up to (006) clearly visible, indicating that the ordering along the c axis becomes better defined. However, there is still no evidence of any long-range order within the layers.Additionally, no shift in peFk positions is observed so that the interlayer spacing, 14.2A, does not change during this heat treatment. Therefore, it appears that the water lost at stage 2 for the alkoxide-derived materials is within, rather than between, the layers. This suggests that there is a fundamental difference between this material and that prepared by ion exchange, which exhibits a decrease in inter- layer spacing as loosely bound water 1,s lost upon heating.' Clearly, an interlayer distance of 14.2 A is too large for the layers to be connected by chains of V...O-V bonds, as in orthorhombic V,05, so the most strongly bound water, which is released when the temperature is raised to around 300°C, probably has a structural role since its loss leads to a concomi- tant crystallisation of the material to give highly crystalline orthorhombic V20,.The XRD pattern of the xerogel prepared in alcoholic solution [Fig. 3(a)] has a poorer signal-to-noise ratio than that of the xerogel prepared by the aqueous route [Fig. 2(a)], suggestive of a more amorphous material. The diffraction pattern of the former [Fig. 3(a)] is similar to that obtained by Hirashima et uE.16 from a material prepared using VO(OEt),, which it was suggested had a similar layer structure to the gel prepared by the ion-exchange method. There are, however, some features in the pattern which were not explained, in particular, the range of broadness and intensity for different (001) peaks, and two peaks which could not be indexed, the relatively intense peak at 20=2$0", d=3.43 A, and a broad peak around 28=61.0", d= 1.52 A.A comparison with the hkO reflections of crystalline V,05 [Fig. 3(c)] suggests that the xerogel may have some 2D order rather than a 1D layer structure, with a similar short-range arrangement of V05 moieties2' to that of crystalline V205.Heating the xerogel at 200°C for 2 h has little effect on the diffraction pattern [Fig. 3(b)], so the loss of the loosely bound water does not affect the structure, whereas the most strongly bound water, which is released at around 300"C, may have a structural role since its loss leads to crystallisation to give orthorhombic V205[Fig. 3(d)]. The differences observed between the xerogels prepared via the aqueous and alcohol routes and their heat-treated deriva- 52 J.Muter. Chem., 1996, 6(l), 49-56 l..I....I....I....l...,1....1. 10 20 30 40 50 60 29ldegrees Fig. 3 XRD patterns of V205gel (alcohol route): (a) as prepared; and after heating in air at (b) 200 "C; (d) 450 "C. (c) Calculated hkO peaks of V20,. (a) and (b) have been smoothed using a simple three- point algorithm. tives are interesting. Chemical and thermochemical analyses show that the only major difference between the two xerogels is their VIV content (Table 2). It is possible, therefore, that this difference is responsible for the difference in structure, since the presence of V" implies either a deficit of oxygen or the presence of some protons, both of which could lead to greater intralayer ordering.The bands in the IR spectra of both vanadium oxide xerogels are at lower frequencies and less well defined than those of crystalline V205 (Table 4). The retention of the V=O stretch- ing band in the gel spectra around lOOOcm-' indicates that the vanadium is present in a similar distorted square-pyramidal environment to that in orthorhombic V205, which suggests that the microstructure of the xerogels may be similar to that of the crystalline material. The shift to lower frequencies of the V=O and V-0-V stretching bands may be connected with a decrease in the average vanadium oxidation state. The spectra are better defined than those of the V205-P205 glasses' of highest V205 content, which have a V20s-like microstruc- ture, suggesting a greater degree of short-range ordering in the gels than in the glasses, which is consistent with the XRD data.Titanium oxide. Ti02 gels have been obtained previo~sly'~ using Ti(OEt), as a precursor. Very similar conditions were employed in this work with Ti(OPr'), as starting material except that an acid/alkoxide ratio of 0.057 rather than 0.2 was found to be required. A translucent xerogel was obtained from this system after evaporation of the solvent at room tempera- ture. The simultaneous TG and DTA curves of the xerogel are shown in Fig. 4. The large initial endotherm is coincident with Table 4 IR spectra of V205xerogels and crystalline V205 sample wavenumber/cm- cryst. V20, 1023 828 599,478 xerogel (alc.) 1005 762 535 xerogel (aq.) 997 759 534 100 95 75 70 65 I 0 100 200 300 400 500 600 TPC Fig.4 TG (i) and DTA (ii) traces of TiO, xerogel a mass loss of ca. 31% up to about 200 "C which is, presumably, mainly due to the loss of loosely bound water and residual organic material. Between 200 and 400°C there is a region of slow mass loss of cu. 5% which is probably due to the loss of chemically bound water. The exotherm, onset at 414 "C,corre-sponds to increased ordering in the material, see below. The total observed mass loss of 36% is equivalent to a formula of Ti02.2.6H20 for the xerogel, with the chemically bound water accounting for 0.34 mol of H,O per TiOz formula unit.XRD patterns of the TiO, gels before and after heat treat- ment are shown in Fig. 5. All show broad peaks in the same positions, which are coincident with those of anatase (Fig. 5), with the exception of an additional low-intensity peak at 28= 30.8'. Even the sample heated above the exotherm around 430°C shows poorly resolved peaks, although the degree of definition does increase with the temperature of heat treatment. Thus all the materials, including the xerogel, have poorly crystalline anatase-type structures. A. IAl A 10 20 30 40 50 60 20/degeea Fig. 5 XRD patterns of TiO, gel: (a) as prepared; and after heating in air for 2 h at (b) 200°C; (c) 450°C. (d) XRD pattern of anatase. J. Muter. Chem., 1996,6(l), 49-56 Mixed vanadium-titanium oxides.Green monolithic gels were obtained for all of the V205-Ti02 mixtures with no significant changes in composition during the process (Table 1). The 30mol% TiOz xerogel appeared to be slightly inhomo- geneous, which is probably an indication that complete gelation did not occur at this composition, particularly since a gel could not be obtained from a mixture containing, on mixing, 40 mol% TiO,. The fraction of reduced vanadium (V'v/V,,,) was relatively constant with, presumably, the reduction being brought about by a similar mechanism to that of the Vz05 gel, see earlier. The simultaneous TG and DTA curves of the V20s-Ti02 gels (Fig. 6) are similar to those of the Vz05 gels (Fig. 1). Single-crystallisation exotherms are observed, at least for the gels containing less than 20mol% TiOz, which is strong evidence for a series of single-phase materials. In addition, as the TiO, content increases so the crystallisation temperature increases and becomes less well defined, and the loss of the strongly bound water at stage 3 becomes spread over a wider temperature range (Table 5). The trend of increasing crystal- lisation temperature with increasing TiOz content was also observed by Hirashima et However, the overall amount ~1.~~7~~ of associated water (19.3-21.1%) is similar to that of the V205 loo. 95* E,.3 a 85.80. 0 100 200 300 400 500 600 100 95 n s '90 85 80 0 100 200 300 400 500 600 TIT Fig. 6 TGA (ij and DTA (ii) traces of mixed V,O,-TiO, gels: (a) 6 mol% TiO,; (b)20 mol% TiO, Table 5 Thermal analysis of mixed V205-Ti02 xerogels stage 3 TiO, (mol%) T rangePC T,*/"c 6 300-350 347 11 300-365 371 20 290-405 409 24 290-425 -27 290-440 -I""I""I""I""I""l' I....l....l....I....I.... I. 10 20 30 40 50 60 WdW-Fig. 7 XRD patterns of a V,O,-TiO, gel: (a) 20 mol% TiO,; (b) the same gel after heating in air at 450°C. (a) has been smoothed using a simple three-point algorithm. *, (101j reflection of anatase Ti02. gels, which is not consistent with the suggestion" that TiO, replaces water in the structure. The XRD patterns of the V205-TiO, xerogels (Fig. 7) are identical to that of the Vz05 xerogel prepared in alcohol (Fig. 3), implying an identity of structure.No peaks belonging to the TiOz gel (Fig. 5) were observed in any of the mixed-gel patterns which provides further evidence that the mixed gels are single-phase materials. The lack of dependence of the structure on TiO, content suggests that the Ti atoms replace V atoms within the 2D layers without inducing interlayer ordering. Xerogels heated at 200°C showed unchanged XRD patterns, like their Vz05 counterparts, whereas XRD patterns of samples heated past their crystallisation temperatures (Fig. 7) showed peaks due to orthorhombic V205 and, for the gels containing >20mol% TiOz, a peak at 28=25.3" corre- sponding to the (101) reflection of anatase TiOz. Thus crystal- lisation gave the expected two products. The anatase component in the other crystallised gels is presumably too small to detect using XRD. Performance in high-temperature lithium polymer-electrolyte cells Vanadium oxides.The discharge profiles of cells containing binary vanadium oxides showed monotonic curves on all cycles, characteristic of a single-phase lithiation of an amorph- ous material. The oxide prepared by the aqueous route gave virtually 100% of the theoretical capacity on the first cycle, whereas that prepared by the alcohol route discharged to only 80% of the theoretically expected value. This is consistent with the greater extent of reduction of Vv to V" during preparation of the latter material (Table 2). However, the absolute capacit- ies achieved by the 20th cycle are very similar (cu. 60% of the theoretical) so that a similar VIV/Vv ratio has been achieved in both materials by this stage.All the cells showed the same cycling behaviour (Fig. 8), with a steady capacity decline over the first ten cycles followed by a region of slower decline. This behaviour is in contrast to 54 J. Muter. Chem., 1996, 6(l), 49-56 6 12 18 24 30 cycle number Fig. 8 Cycling performance of a cell containing sol-gel-derived V205 (alcohol route) the results obtained in similar cells for vanadium phosphate glasses5 which showed a region of steadily increasing capacity after the initial decline. The initial decrease for the gel-derived materials may be due to some lithium becoming immobilised in the structure but this is unlikely to continue over many cycles.Recent ~ork~~,~~ has shown that lithium may be inserted into V205 reversibly, although not topotactically, to give an Li/V molar ratio of between 0.9 and 1.0, and that irreversibility past this value may be due to the slow reoxidation of the V"' formed after a structural rearrangement. The theoretical capacity of the cells studied here is based on that of V6013 (417mA hg-') and is equivalent to an Li/V molar ratio of 1.33; so to achieve this capacity some reduction to V"' is necessary. The decline in capacity for the gel-derived vanadium oxides observed in this study is similar to that previously shown for crystalline V205,6 so it is likely that a similar mechanism operates. The cells all gave open-circuit voltages of cu. 3.5 V and the average voltages on discharge were consistently in the region of 2.3 V.This gives a maximum theoretical energy density of 960 W h kg-', which is superior to that of cells containing V6013(880 W h kg-I). However, the decline in capacity with cycling means that the observed energy density consistently decreased with continued cycling. Titanium oxide. The cells containing the gel-derived TiO, gave monotonous discharge curves, unlike crystalline anatase, in which no discharge occurs until a plateau24 at 1.8 V, with a capacity on the first discharge in excess of 100% of the theoretical value based on LiTi02. This is consistent with previous reports that poorly crystalline anatase-type structures allow free diffusion of lithium ions.25 In contrast, Minnett and Owen1* were able to contain a capacity of only 2 mA h g-' on the first cycle, which is less than 1 YOof the theoretical value. A large loss of capacity was observed on the second dis- charge, followed by further small losses before steady cycling was observed at around 10% of the theoretical capacity.The cycling behaviour is similar to that observed for anatase, which has been shown to retain lithium when discharged to 1.2V forming, irreversibly, a new phase.24 Significantly, anatase has been shown to be reversible along the plateau at 1.8V, becoming irreversible only when discharge proceeds beyond this. The average cell voltage on the first discharge was 1.7V, giving a theoretical energy density of 570 W h kg-', which is the same as that of anata~e.,~ The observed energy density (650 W h kg-') exceeded the theoretical value by virtue of the higher than estimated capacity (see earlier).However, sub- sequent discharges gave average voltages of only 1.4 V which, when coupled with the low capacities achieved, gave very low observed energy densities, for example 110 W h kg-' (second discharge) and 53 W h kg-' (20th discharge). Chemically mixed vanadium-titanium oxides. The cycling performances of representative cells with cathodes containing mixed oxides are shown in Fig. 9. An initial drop in capacity was observed over the first few cycles in each case which was complete by the sixth cycle. The capacity achieved on this cycle was then retained on further cycling up to an experimental limit of 30 cycles.The initial capacity decline was smallest with the 6mol% TiO, cathode, such cells eventually cycling at above 80% of theoretical capacity. This behaviour is in contrast to that of the cells with only V205 as the cathode-active material, see earlier, which continue to lose capacity (Fig. 8). This strongly suggests that it is the TiOz component of the cathodes that is influencing the reversibility. The initial capacity decline observed for all of the cells may be due to the immobilisation of some lithium ions, a view which is consistent with the increasing loss in capacity with increasing titanium content (Fig. 9). The interruption of the V-0-V chains in the structure by Ti" would be expected 100 80 60 40 20 0 4 a 12 16 20 100 80 0 6 12 18 24 30 100 (c> " 10 20 30 cycle number Fig.9 Cycling performances of cells with ternary sol-gel-derived V,05-TiO, cathode-active materials: (a) 6 mol% TiO,; (b) 20 mol% TiO,; (c)physical mixture containing 10 mol% TiO, J. Muter. Chem., 1996,6(l),49-56 55 to hinder the electron transfer along these which, in turn, would reduce lithium ion mobility. Additionally, the titanium atoms may physically restrict the lithium ion passage by inducing disruptions to the local microstructure. The second region of capacity decline (Fig. 8) is slower and does not occur when Ti0, is also present in the cathode active materials [Fig.9(u) and (b)].This effect could be due to the preferential reduction of Ti at low voltages or a structural modification induced by the Ti. To investigate this phenomenon some cells were constructed using physical mixtures of sol-gel-derived V205 and TiO,. Physically mixed vanadium and titanium oxides. The cycling performance of just such a cell is shown in Fig. 9(c). After an initial sharp drop in capacity the cell shows an excellent capacity retention from the fifth cycle onwards. In fact, the mixed cathode is reversible at close to 80% of theoretical capacity (330 mA h g-') which is very similar to the behaviour of V6OI3in the same type of cell. This synergism is the same as was observed for the chemically prepared ternary oxides, demonstrating that the contribution of the Ti-containing com- ponent is not of a structural nature.The most probable protective mechanism is the preferential reduction of the Ti" over V" near the low-voltage limit, which prevents the 'over- reduction' of vanadium to V"' with a concomitant reorganis- ation of the microstructure of the material, see earlier. Overall performance and energy densities The open-circuit voltages observed for all of the cells containing chemically or physically mixed V,O, and Ti02 sol-gel-derived oxides was around 3.5 V, compared with 2.8 V for cells contain- ing V6013,and the average voltages observed on discharge remained steady at around 2.3 V for each successive discharge. This gives a theoretical energy density of cu.960 W h kg-' for each material if the theoretical capacity is based on that of V6013. This value is greater than that of V6OI3(880 W h kg-') because of the improved average cell voltage. The energy densities for some of the reversible cells studied in this work with mixed V205 and TiO, cathode active materials are given in Table 6. The tenth discharge is chosen as a typical later discharge when the cells had all settled into a period of steady reversible cycling and a comparison with the first discharge shows the drop in energy density caused by the initial capacity decline. The observed energy densities compare favourably with that of V6OI3,with the cells using the oxides containing 6mol% TiOz showing a significant improvement. This is due to improved average voltages on discharge combined with high reversible cell capacities.Table 6 Observed energy densities of the reversible cathode materials energy densityfw h kg-' V205 content first tenth material (mol%) discharge discharge V205-Ti02 94 1005 800 V205-Ti02 73 850 570 V20, +TiO," 90 880 690 V6013 - 880 610 " Physical mixture. Conclusion The ternary materials showed good reversibility when employed as cathode components in lithium polymer-electro- lyte cells up to the experimental test limit of 30 cycles. Since a similar performance was obtained from a cell containing a physical mixture of oxides it appears that the preferential reduction of Ti" over V" near the low-voltage limit is a key factor.Observed energy densities were, in many cases, higher than have been obtained with similar cells incorporating V6OI3. The mechanism of protection against cathode over- reduction observed in this work shows great promise for future developments in lithium battery research. Materials previously thought unsuitable because of irreversibility at low voltages might still find favour as cathodes if a second material can be found with suitable electrochemical behaviour. We thank SERC and AEA Industrial Technology, Harwell Laboratory, for a CASE award to A.D. References 1 M. B. Armand, J. M. Chabagno and M. J. Duclot, in Fast Ion Transport in Solids, ed. P. Vashishta, J. N. Mundy and G. K. Shenoy, North Holland, New York, 1979,p. 131.2 M. B. Armand, Solid State Ionics, 1983,9110, 745. 3 A. Hooper and B. C. Tofield, J. Power Sources, 1984,11,33. 4 M. Gauthier, D. Fauteux, G. Vassort, A. Belanger, M. Duval, P. Ricoux, J-M. Chabagno, D. Muller, P. Rigaud, M. B. Armand and D. Deroo, J. Electrochem. SOC.,1985,132,1333. 5 A. Davies, R. J. Hobson, M. J. Hudson, W. J. Macklin and R. J. Neat, J. Muter. Chem., 1994,4, 113. 6 Y. Sakurai, S. Okada, J. Yamaki and T. Okada, J. Power Sources, 1987,20, 173. 7 C. J. Brinker and G.W. Scherer, Sol-Gel Science: The Physics and Chemistry of Sol-Gel Processing, Academic Press, London, 1990. 8 P. Aldebert, N. Baffier, N. Gharbi and J. Livage, Muter. Res. Bull., 1981,16, 669. 9 J. J. Legendre and J. Livage, J.Colloid Interface Sci., 1983,94,75. 10 J. J. Legendre, P. Aldebert, N. Baffier and J. Livage, J. Colloid Interface Sci., 1983,94, 84. 11 P. Aldebert, H. W. Haesslin, N. Baffier and J. Livage, J. Colloid Interface Sci., 1984,98,478. 12 J. Livage, P. Barboux, J. C. Badot and N. Baffier, in Better Ceramics Through Chemistry III, ed. C. J. Brinker, D. E. Clark and D. R. Ulrich, MRS, Pittsburgh, 1988, pp. 167-177. 13 J. Livage, Chem. Muter., 1991,3, 578, and references therein. 14 H. Hirashima, T. Koyama and T. Yoshida, Yogyo-Kyokui Shi, 1985,93, 554. 15 H. Hirashima, S. Kamimura, R. Muratake and T. Yoshida, J. Non-Cryst. Solids, 1988,100,394. 16 H. Hirashima, K. Tsukimi and R. Muratake, J. Ceram. SOC.Jpn., 1989,97,232. 17 K. West, B. Zachau-Christiansen, M. J. L. Ostergard and T. Jacobsen, J. Power Sources, 1987,20, 165. 18 M. G. Minnett and J. R. Owen, J. Power Sources, 1990,32,81. 19 B. E. Yoldas, J. Muter. Sci., 1986,21, 1087. 20 C. Sanchez, M. Nabavi and F. Taulelle, in Better Ceramics Through Chemistry 111, ed. C. J. Brinker, D. E. Clark and D. R. Ulrich, MRS, Pittsburgh, 1988, p. 93. 21 R. Enjalbert and J. Galy, Acta Crystallog., Sect. C, 1986,42, 1467. 22 J. M. Cocciantelli, J. P. Doumerc, M. Pouchard, M. Broussely and J. Labat, J. Power Sources, 1991,34, 103. 23 C. Cartier, A. Tranchant, M. Verdaguer, R. Messina and H. Dexpert, Electrochim. Acta, 1990,35,889. 24 W. J. Macklin and R. J. Neat, Solid State Ionics, 1992,53-56,694. 25 K. Kanamura, K. Yuasa and Z. Takehara, J. Power Sources, 1987, 20. 127. Paper 51024276; Received 18th April, 1995 J. Muter. Chem., 1996, 6(l), 49-56