J. MATER. CHEM., 1991, 1(3), 339-346 Room-temperature Electrochemical Reduction of YBa,Cu,O, -Solid-state and Solution Chemical Results Michael Schwartz,ta Yosef Scolnik," Michael Rappaport,b Gary HodesC and David Cahen*a a Departments of Structural Chemistry, Nuclear Physics and Materials Research, The Weizmann Institute of Science, Rehovot 76100, Israel The oxygen content of polycrystalline samples of YBa,Cu,O,-, can be reduced quantitatively, in a controlled fashion, by electrochemical techniques at room temperature in propylene carbonate. Upon reduction, the propylene carbonate undergoes an unusual reaction at the YBa,Cu,O,-, cathode to produce propanal, apparently because of the production of an active oxygen species on the surface. The reduced materials have been characterized by X-ray powder diffraction, electrical resistivity and magnetic susceptibility.The large-grained, reduced pellets are found to be inhomogeneous with respect to x. The reduced materials exhibit a broadened transition to the superconducting state. This effect is ascribed to the formation of metastable phases formed during reduction. After a low-temperature anneal, 80, 60 and 20 K transition temperatures are observed. These results indicate that T, is a continuous function of oxygen content, but a discontinuous function of oxygen ordering. Keywords: Electrochemical reduction; Y-Ba-C u-0 system; Superconductivity Among the superconducting cuprates, YBazCu307 --x (123) is the one most readily amenable to controlled solid-state chemi- cal studies owing to the relative ease of its (de)oxygenation at moderate-to-high temperatures.This property leads to a continuous change in the oxygen content from seven (x=O) to six (x = l).',' Results from a variety of techniques, including re~xygenation,~*~high-temperature solid-state electro~hemistry,~*~inelastic relaxation measurements,'o and polarized light microscopy (to follow the position of the orthorhombic/tetragonal phase boundary)," $ show that, not- withstanding appreciable differences between values arrived at by different methods, the chemical diffusion coefficient for oxygen in this material is (very) high. For example, values of 2.2 XIO-'~cm2 s-' (at ca. 200 "CI,~ 5 xIOb8cm2 s-' (at ca. 550"C),87.5~10-~cm's-' (at ca.300°C)" and even 6~10-~cm's-' (at ca. room temperature)," have been measured or deduced. This oxygen lability is the more interesting because it has been shown that there is a marked non-linear dependence of T, and of the crystallographic parameters on oxygen con- These results have been obtained on samples that are prepared at temperatures >350 "C by annealing with Zr as an oxygen getter" or at even higher temperatures by vacuum annealingI3 and q~enching.'~ Careful preparation of oxygen-deficient 123 by these methods has led to the obser- vation of a 'plateau' behaviour of T, with decreasing oxygen content, in which two T,s of ca. 90 and 60K are observed for samples with x between 0 and 0.6.129'3 Electron diffraction and single-crystal X-ray diffraction data have been used to suggest the occurrence of long-range ordering of oxygen vacancies in the oxygen sublattice to form two The occurrence of ordering is supported by results from thermodynamic ~alculations.'~*'~ However, the existence of a distinct, microscopically homogeneous 60 K phase has been questioned on the basis of high-resolution electron microscopy t Present address: Xsirius Superconductivity Materials Ltd., studies." A study of reduced samples prepared by solid-state electrochemical techniques at 500 "C showed the presence of many phases, each with only slightly different oxygen contents." The behaviour of oxygen in this material can be rational- ized, on the basis of its crystal structure,21 which is derived from the ideal perovskite structure, AB03, with Cu occupying the B sites, and Ba and Y the A sites.The combination of Y and Ba ordering and oxygen deficiency leads to a tripled unit cell relative to the basic perovskite unit cell. In the fully oxygenated compound the oxygen vacancies (two per unit cell) are ordered with oxygens missing from the Y plane[(f,0, f) and (0, f,3)sites] and from the basal plane [(f,0 0) site]. These ordered vacancies lead to planes of square-pyramidal CuO, units perpendicular to c and to chains of square planar CuO, units parallel to b (Fig. 1). Powder ,Cu-0 chain 0 cu 0BaII 0 ll Oy 0 Jerusalem, Israel. C $ In ref. 11, use is made of the fact that the oxygen out-diffusion, at sufficiently high temperatures, leads to a phase transformation kafrom an orthorhombic to a tetragonal phase of 123 and that the boundary between these phases can be visualized clearly, when a sample is observed by polarized light microscopy.Fig. 1 Structure of YBa,Cu,O,-, (after ref. 21) neutron diffraction studies on oxygen-deficient material have shown that the most labile oxygen comes from the (O,+,O) ~ite.'~*~~Loss of this oxygen breaks up the chains, and in the end-member x= 1, no chains are present. Changes in oxygen content also play a role in the electronic J. MATER. CHEM., 1991, VOL. I of the total oxidative power of the samples [as expressed in (Cu-0)' units and assuming three Cu per formula unit and an essentially 1OO% pure material].32 Lao~,Sr0.,CoO3 -x was prepared according to ref.29 and was a single phase as determined by X-ray powder diffraction. The parameter x structure. Above T,, 123 has metallic hole cond~ctivity.~~ was deduced to be 0.01 from iodometric titration. Reduction of the oxygen content to below ca. 6.4 leads to a new semiconducting, tetragonal phase.24 In the fully oxygen- ated material (x=O), the average, formal valence of Cu is +2.33 (assuming Y3+, Ba2+ and 02-)and in the semicond- ucting phase (x = I), the average formal Cu valence is +1.67. Formal valencies in such a complex, metallic oxide should be used with caution and it is probably more realistic to describe the holes in the conduction band as belonging to (Cu-0)' units rather than to discrete Cu3+-02- or Cu2+-O- units.Regardless of the formal description, it is clear that the number of charge carriers is governed by the oxygen content. We report here solid-state and solution chemical results of room-temperature electrochemical reduction of this material The bulk electrochemical reductions were performed in a single-compartment cell using propylene carbonate-tetra-butylammonium perchlorate (0.1 mol dm- 3, as the electrolyte. The samples were in the form of pressed pellets, ca. 100 mg, 8 mm diameter by 0.5 mm thick, with a density of 4.6- 5.0 gcm-3. The pellets were pressed onto an undersize Pt disc which served as the current collector. The Pt disc and wire were encased in a glass tube to prevent direct contact with the electrolyte.Two Pt wires were used as counter and quasi-reference electrodes. Ar was bubbled through the solu- tion during the reduction. Standard electrochemical equip- ment, including a digital coulometer, was used. The cathodic limit for propylene carbonate-tetrabutylammonium perchlor-(preliminary accounts of this work have been pre~ented~~-~~). ate decomposition on Pt was more negative than -1 V and We thought that such a reduction could be possible because of the relatively high oxygen diffusivities near room tempera- ture that have been observed in other oxygen-deficient perov- ~kites.~~,~'Thus, oxygen-deficient 123 could be prepared from fully oxygenated 123 at room temperature using electrochemi- cal reduction according to YBa2Cu307-x +nye- +YBa2Cu307 -x-y +y[O"-] (1) The square brackets around 0"-and the use of the parameter n indicate that we do not know the exact form in which oxygen diffuses through or in which it leaves the solid.By working at room temperature we hoped to restrict large-scale structural rearrangements, which can occur by higher tem- perature preparation and to allow for the formation of phases that are not stable at higher temperatures. In addition, if n is known, electrochemical reduction permits precise control of the degree of reduction via coulometry, which is not always possible with higher-temperature techniques. Because of the impracticality of using gas/solid reactions to exploit room-temperature oxygen mobility, we explored liquid/solid electrochemical reactions, as was done for the 'normal' oxygen-deficient perovskites Nd0.50Sr0.-x2850C~03 and Lao,50Sro.50C003 -x29 in aqueous solvents. The known reactivity of 123 with H20 required the use of non-aqueous electrolyte^.^^ We chose propylene carbonate because of its wide working potential window and because we found that it is chemically inert towards 123. Some experiments were done also in acetonitrile. We note that electrochemical tech- niques at room temperature have been used also to intercalate H and Li into 123.31 Experimental 123 was prepared by heating a mixture of the precipitated nitrates. The mixture was heated to dryness three times to remove any excess acid and then heated overnight at 120 "C.The powder was ground thoroughly, pressed into pellets and fired twice at 950°C for 12h in O2 and then cooled at 50 "C h-' to room temperature with one grinding in between. After these firings, the powder was reground, passed through a 400 mesh (37 pm) sieve, repressed and refired as above except that a 12 h anneal at 500 "C was added. Typically an elongated grain size resulted, with a long dimension of 5-10 pm, as determined by scanning electron microscopy. X-Ray powder diffraction showed the orthorhombic pattern of 123.l Some batches contained a small amount of other phases. The starting material showed, reproducibly, an oxygen content of 6.97f0.02 as determined by iodometric titration was not approached during reductions or during the separate measurements of the diffusion coefficient. The reductions were carried out potentiostatically.The voltage was set at -1 V relative to the Pt quasi-reference electrode. The potential of the Pt quasi-reference electrode was found to be ca. 250 mV more positive than the saturated calomel electrode and stable over the course of a few hours. Initial current densities were of the order of 1 mA cm-2 and the current decayed during the course of the experiment. Initially, there was a fast decay which may be due to adsorbed oxygen. This decay accounted for no more than 30mC of charge passed, i.e. less than 2% of the total charge needed to reduce the oxygen content by 0.05, equivalent to x =0.001. The initial voltage at the counter electrode was > +2 V relative to the reference electrode. The reduction was allowed to continue until the desired amount of charge had been passed as measured by coulometry.Typical times were 4-10 h. Reduced samples were washed with acetone and stored in uacuo. Iodometric titration was used to analyse for oxygen content.32 Simply soaking the pellets in the electrolyte had no effect on their electrical, magnetic and structural properties. These experiments were also repeated with samples of 5Sr0.5c003* Some samples were annealed after reduction by sealing them in fused silica ampoules in a vacuum of better than 1 x lop2 Pa. These samples were heated at 150 or 300 "C for 24 h and allowed to cool in the furnace.Experiments to study the reaction products in solution were performed in a two-compartment cell in which the counter electrode was separated from the working electrode by a sintered glass frit. The working electrode was prepared by attaching a wire to the 123 (or Lao~5Sro~,Co03~x) pellet with Ag paste, covering this area with insulating epoxy to prevent contact with the electrolyte. After an initial Ar purge, the electrolyte was blanketed by Ar so as to minimize loss of any gaseous species dissolved in the electrolyte. Reduction conditions were the same as above. After a certain degree of reduction (x=0.10-0.19, cyclic voltammetry of the solution in the working electrode compartment was performed using a Pt wire in place of the 123 electrode.Both propylene carbonate and acetonitrile solutions were used. The solutions in the working electrode compartment were analysed by gas chromatography and mass spectrometry using a Finnegan model #4500 GC/MS equipped with a Supelco SBP5 30m capillary column. Semi-quantitative gas chromatography analyses were performed with a Tracor 560 GC with an OV1 1-101 column of 1.5 m. An effective value for the chemical diffusion coefficient of J. MATER. CHEM., 1991, VOL. 1 oxygen in polycrystalline 1237, at room temperature, was determined from the time decay of the current under potentio-static condition^,^' using acetonitrile in addition to propylene carbonate, both with tetrabutylammonium perchlorate as the electrolyte. Electrical resistivity measurements were performed using the van der Pauw four-probe method with measuring currents of 1-50mA, and contact was made using Ag paint [ca.(1-2) x cm2 area]. Minimum detectable resistance was R. Magnetic susceptibility measurements were done using the a.c. mutual inductance technique at a frequency of 75 Hz and a field of ca. 0.33 G. The samples were cooled in the field. D.c. magnetic susceptibility measurements were also performed on a vibrating sample magnetometer and found to be similar to the a.c. susceptibility results. X-Ray powder diffraction patterns were obtained with Cu-Kcc radiation. Results Reduced Samples Structural Evidence for Electrochemical Reduction of 123 The X-ray diffraction patterns in the region of the (200) and (020)+(006) reflections for an unreduced and highly reduced sample are shown in Fig.2. These reflections are unambiguous in that each peak is directly related to a lattice constant. Comparison of the two patterns shows that the (200)reflection is broadened while the (020)+(006) reflection is not broadened further upon reduction. In addition, the (200) reflection is shifted to lower Bragg angles, indicating that the lattice parameter a has increased. Because the X-ray powder diffrac-tion was performed on pieces of pellets that had been thor-oughly ground, the bulk of the sample, rather than only those regions near the surface exposed to the electrolyte, was probed. We note that it was possible, via heavy reduction, to obtain samples that showed no more splitting of the (200) and the (020)+(006) peaks (x FZ 6.2).Because of the above-mentioned broadening of the peaks upon reduction (of bulk samples), this is only suggestive of a transition to a tetragonal phase. 48.0 47.5 47.0 46.5 46.0 281" Fig. 2 X-Ray powder diffraction patterns of bulk (pellet) YBazCu30,-, in the 20 region 48.0-46.0", Cu-Kcr radiation and a 20 scan rate of 0.25 " min-'. (a) Starting material, x=O.O5; (b)x= 0.29, before annealing; (c)x =0.29, after annealing. Dashed lines are at the centre of full-width at half-maximum for the x=O.O5 sample. The arrows represent the expected shifts in the (200) and (006) reflections for a sample with x=O.29 based on data from ref. 12 and 14 341 Stoichiometric Evidence for Electrochemical Reduction of 123 A second result that supports oxygen removal from the bulk is the exact agreement between y in eqn.(1) as determined by coulometry (using n=2) and the final oxygen content of the sample as derived form iodometric titration. Finally, it was observed that the density of the pellet was important. Typical densities were 4.6-5.0 g cmP3(ca. 75% of the theoretical density). For pellets with much higher densities, ca. 5.8 g cmP3 (91% of the theoretical density), the current decayed very rapidly and reduction was limited to 1-3% of the labile oxygen (x =0.01-0.03). Samples with lower density have greater porosity and, therefore, a larger effective surface area for contact with the liquid electrolyte.The rapid current decay that was observed with dense pellets can then be understood by realizing that in a wet electrochemical set-up actual electron transfer occurs at all the grain surfaces that are contacted by the electrolyte, rather than only at the surface of the pellet. This situation is quite different from that in (high-temperature) solid-state electrochemical experiments, where as dense a pellet as possible is needed to obtain the highest possible area of direct contact between solid electrode and electrolyte [cf. ref. 9(b)]. Magnetic Susceptibility and Electrical Resistivity of Reduced Samples Plots of the temperature dependence of the magnetic suscepti-bility and electrical resistivity have been published in prelimi-nary communication^.^^-^^ Representative plots are shown in Fig.3 and 4 and the data are summarized in Table 1. The data show that the transition to the superconducting state is broadened relative to the starting material [cf.also ref. 12(b)]. The degree of broadening increases for larger amounts of Omom 0.04 TIK Fig. 3 Plots of electrical resistivity us. temperature for two samples of reduced YBazCu30,-, x, x=0.14, before annealing; +, x =0.10, after annealing at 300 "C 2.5, ~0.04 G hl.04 : Rc 0.02 Q-0.01 G 0.0 0.00OS5-0 2.0 40 60 80 100 TIK Fig. 4 Plots of electrical resistivity us. temperature for a sample of reduced YBa2C~30,-x.x , x=0.20 before annealing; +, x=0.29 after annealing at 300 "C J.MATER. CHEM., 1991, VOL. 1 Table 1 Effect of room-temperature reduction and subsequent low-temperature anneal on the superconducting transition temperature (from electrical resistivity measurements) and u lattice parameter of YBa2Cu30, -x onset temperature/K before after before after anneal anneal anneal anneal 0.05 0.05 83 84 0.10 0.10 85 81, 59 0.15 0.23 80 79, 57 0.20 0.29 75 61, 20 0.22 78 0.29 70 'From ref. 12; 'from ref. 14. reduction. Except for the unreduced material, the electrically measured transition to the superconducting state is broadened as the current is increased (10-50 mA). There is also a slight decrease in the onset temperature for superconductivity (Table 1). In addition, increased reduction leads to thermally activated normal-state electrical behaviour. The effective acti- vation energy in the normal state increases with increasing reduction (from In p us.l/Tplots). Reduced and Annealed Samples After reduction, some samples were annealed at 150 and 300 "C under conditions described in the Experimental. For samples with x20.10, additional oxygen loss occurred upon annealing. This suggests that oxygen lability increases with increasing reduction, possibly owing to lattice destabilization as a result of second (and higher) nearest-neighbour effects (i.e.decreasing distance between oxygen vacancies). Initial and final oxygen contents are given in Table 1. The X-ray diffrac- tion patterns were also sharper than those obtained for the unannealed samples [Fig.2(c)]. The a values remain larger with respect to unreduced samples and are summarized in Table 1. They agree well with those obtained for samples reduced by high-temperature methods with similar oxygen content.l2-I4 Plots of magnetic susceptibility (x) and of electrical resis- tivity (p) as a function of temperature (T)for the annealed samples have been presented Representative p-T plots are shown in Fig. 3 and 4 and the results are summarized in Table 1. After the 150 "C anneal, a break in the transition to the superconducting state is observed at ca. 70 K. After the 300 "C anneal, the X-T measurements show distinct transitions to the superconducting state. For samples with x<O.25, two transitions at T,zt:O and 80 K are found.For x=O.29, one T, at ca. 60 K is observed. The electrical resistivity measurements agree with the magnetic measure- ments except that a T,z20 K is also osberved for the x =0.29 sample (cf: also the recent results of Jorgensen et al. on the effects of (sub)room temperature anneals of quench-reduced samples33). The fact that this T, is observed by the electrical resistivity measurements and not by the magnetic suscepti- bility measurements indicates that only a small amount (but above the percolation limit) of the sample has this T,. The presence of the distinct 60 and 80 K transitions in the reduced, annealed samples is similar to the 'plateau' behaviour observed in oxygen-deficient samples prepared by other method^.'^,'^,^^ After the low-temperature annealing treatment, the normal- state electrical resistivity of the reduced samples is thermally activated.The degree of activation decreases as T, is approached. Similar effective activation energies for samples with different final oxygen contents are found over similar temperature ranges. transition width/K before anneal 5 22 25 37 Solution Reactions a lattice parameter/pm after anneal 38 1.7(4) 382.4(8) 382.6(8) 383.2(4) 382.7(2)" 383.52' To complement the results obtained from the measurements on the electrode, experiments were performed to elucidate the oxygen chemistry occurring at the surface during reduction, i.e. to identify the product(s) in the reaction y[02-1 +electrolyte-+products (2) After a certain degree of reduction (x =0.10-0.19, cyclic voltammograms were performed on the solution.As shown in Fig. 5, a species is formed which is not present initially. This species is clearly not 02,which has a very distinct cyclic voltammogram on Pt in propylene carbonate [Fig. 5(c)]. GC/ MS analyses were then performed on the solution from the working electrode compartment and a reaction product, ident- ified as propanal, was observed. The amount of propanal formed increased with the amount of charge passed. Semiquantitative analyses indicated that the amount of pro- panal was consistent with the amount of oxygen produced as determined by coulometry, i.e. it is a major product. However, exact figures cannot be obtained since the stoichiometry of eqn.(2) is not known. The electroactive species observed in the cyclic voltammogram is not propanal as the cyclic voltam- mograms of the unknown were different from cyclic voltam- mograms of propanal. The unknown may be due to Cog-(see Discussion later). Attempts to precipitate Cog-with Li+ $. -lo@ +10 pA f ' ~ ' I ' " l ' ~ ' 1 -1 .o 0 1.o potential/V vs. Pt Fig. 5 Cyclic voltammograms of propylene carbonate-tetrabutyl- ammonium perchlorate (0.1 mol dm- 3, electrolyte in working elec- trode compartment using Pt wires as the working and quasi-reference electrodes: (a) before reduction of 123; (b) after reduction of 123 (x= 0.15); (c) saturated with O2 J.MATER. CHEM., 1991, VOL. 1 from the reaction solution were not successful. It was also observed that neither trace amounts of HzO nor saturation of the electrolyte with oxygen had any effect upon the reduction. In addition, the presence of an oxidizing species in the counter-electrode compartment was detected using I -. In order to determine whether the formation of the propanal was due to reaction with the oxygen from 123 or due to some catalytic effect of the metal oxide and the electrical potential difference, a sample of YBazCU306 (1236) was used in place of 123. The 1236 was held at the same potential as that used for the reduction of 123 for the same period of time. Only very small amounts of current could be passed and the total charge accumulated was equivalent to x=O.Ol. This can be understood by realizing that if 123 does not contain labile oxygen, it is no longer a mixed ionic/electronic conductor but just an electronic one, which then becomes a blocking elec- trode to the liquid electrolyte. Under these conditions reaction (1) cannot take place.This lack of reactivity is not due to a density effect since the density of the 1236 pellets was 4.5 g cm-3 (72% of the theoretical density). Cyclic voltammo- grams were unchanged from the initial background scans and subsequent GC/MS analysis of this solution showed no traces of propanal or any other reaction products. This experiment was also repeated using Lao.5Sro.5C003 -x as the working electrode. Current could be passed as in the case of 1237 and after reduction equivalent to x=0.07, GC/ MS analyses of the solution in the working electrode compart- ment were made.There were three reaction products: the two major products were identified as propanal and tributylamine from their mass spectra; the other product could not be identified. Diffusion-coefficient Measurements The effective chemical diffusion coefficient for oxygen in polycrystalline 123, B, measured at room temperature by the method of ref. 29, was found to be 4+_2 x cmz s-' irrespective of grain size, pellet density, porosity, surface treatment, organic solvent or applied voltage, as described in detail elsewhere. 34 a Discussion Effects of Reduction on 123 All experimental observations on material that has been reduced electrochemically at room temperature indicate bulk oxygen loss.While the density dependence of the reduction process and the agreement between y in eqn. (1) as determined by coulometry and by iodometric titration could be explained by either a near surface or a bulk electrochemical reduction of the material (but not by a chemical decomposition), these results, taken together with the X-ray diffraction data, show that a bulk reduction occurs. If only near-surface reduction were to take place, we would expect to see X-ray diffraction from the non-reduced core. Using accepted estimated sensi- tivity limits for X-ray powder diffraction, and the 10 pm average grain diameter found from scanning electron microscopy on our samples, we calculate that unreduced cores with diameters no less than 3 pm should have been detected in the X-ray diffraction patterns.Smaller unreduced cores can constitute d 3% of the sample. In addition, the persistence of significant amounts of unreduced cores in the grain can be excluded since they would be detected in the magnetic suscep- tibility and electrical resistivity measurements of the unan- nealed samples as a sharp drop followed by a In our samples, only a continuous drop is observed. (Recent results on fine-grained, polycrystalline thin films, obtained from Tel Aviv University, show near-uniform changes in the XRD patterns upon r.t. reduction, again supporting the occurrence of bulk reduction.34b Because the reduction is carried out at room temperature, the issue of oxygen mobility at this temperature is a crucial one.Straight extrapolation of high-temperature values obtained from solid-state electrochemical measurements, sug- gests a value for 0" of ca. 10-zocmz s-'.~This is similar to what can be estimated by extrapolating the 600-300 "C self- diffusion data, obtained from tracer studies on single crystals and oriented polycrystalline samples,36 to room temperature. Such a small diffusion coefficient would limit oxygen loss to a surface region only and therefore cannot explain our results. Extrapolation of the values obtained between 900 and 300 "C from the earlier men_tioned polarized light microscropy experi- mects" suggests a D of 7 x lo-'' cmz s-' at 300 K.Estimates of D based on the amount of oxygen removed over time from the average grain size of our samples give values on the order of 10-'o-lO-'z cm_' s-'. These estimates agree with our measurements of D (according to the method of ref. 29) at room temperature. These results should be compared with the values of 1.4x10-" and 5 x1O-l5 cm2 s-' at 25 "C for in other perovskites, Ndo~50Sro~50C00328and 50C~03Lao.50Sro, -x,z9 respectively. All these estimates assume that diffusion of the oxygen species within the grain to the surface is the rate-limiting step, rather than subsequent chemi- cal reactions at the surface with species in the electrolyte. This assumption is based, inter ah, on our measurements of the effective diffusion coefficient (see Results section)34 and fits with the shell model that will be described below.Clearly, the room-temperature reduction process leads to materials that are not homogeneous, as shown by the (h 0 0), (h k 0) and (h k 1) peaks in X-ray powder diffraction patterns and by broadened transitions to the superconducting state. However, careful analysis of the experimental results can yield information on the electronic and structural properties, which are important for understanding superconductivity in these materials. Both the shift observed in the a parameter and the broadening indicate that a number of domains/phases with different a parameters form. Such inhomogeneity is to be expected because of the low temperature of the reduction process.This shift in the a parameter is consistent with oxygen removal, as shown by diffraction studies on oxygen-deficient material prepared by other high-temperature When we compare the lattice parameters reported for such samples with our samples of similar average oxygen content, we find that the shifts for our reduced, unannealed samples show regions with (and an averaged value of) a larger and a c value that is significantly smaller than found in the former (assuming minimal shifts in b, cf. Fig. 2). This then suggests that room-temperature reduction produces samples contain- ing regions with a structure different from those obtained in more homogeneous samples prepared by high-temperature reduction.The effect of these metastable structures is seen in the occurrence of the 20 K transition after annealing. Indeed, the observation of a distinct 20 K transition alongside the well known 60 K one may indicate the existence of another 'plateau' of stable phases. This idea is supported by the recent results of Jorgensen et which confirm our report and support our deductionz7 for the existence of a stable, ordered 20 K phase. The 20 K transition is not observed (initially) in samples prepared by higher-temperature technique^.'^-'^ This may be an indication of the lower stability of this phase as compared to those with higher T,. Evidence for metastable regions of disordered oxygen within a crystallite has also been obtained from electron micros~opy.'~ Models for Room-temperature Reduction of 123 The broadening in the superconducting transitions can have several explanations.One possibility is physical degradation 344 of the grain boundaries leading to weak links. This can provide one simple explanation for the observed normal-state thermal activation of the electrical resistivity in strongly reduced samples. However, this can be discarded as the sole cause for several reasons. First, a grain-boundary effect cannot explain the results from the X-ray powder diffraction and magnetic susceptibility experiments, which indicate a bulk effect. In addition, samples in which the grain boundaries have been damaged by ion-beam irradiation3’ show electrical behaviour at T, which is substantially different from the behaviour observed in our samples.The ion-beam irradiated samples show a sharp drop in electrical resistivity at T,and then a tail. In some severely damaged samples, only a dip at the original T, is observed and the resistivity increases with increasing temperature. In such samples, ion-beam irradiation has no effect on the crystallinity. In comparison, X-ray powder diffraction showed that the bulk crystallinity of our samples had been affected (Fig. 2). Also, the annealed samples in their normal states showed thermally activated behaviour with effective activation energies similar to those found for the as- reduced samples over comparable temperature intervals, Finally, we tested to see if the conditions used for post- reduction annealing are sufficient to promote sintering.This was checked by annealing at 300 “Cfor 24 h an unsintered pellet of 123. Although the transition to the superconducting state was observed by magnetic susceptibility on this sample, neither a full transition nor a sharp drop in resistivity was observed in the electrical-resistivity measurements, indicating that the grain boundaries are not reconstructed under these annealing conditions. These results confirm that grain-bound- ary degradation cannot be the sole cause for the broadening. The most likely major cause for the broadened transitions upon reduction is the presence of a residual gradient of oxygen content within each grain, which will disrupt long-range order.This is consistent with the low temperature of preparation and the X-ray powder diffraction data. This gradient can arise from the fact that oxygen mobility at room temperature is still insufficient for complete equilibration of the oxygen concentration in our relatively large-grained, room-tempera- ture reduced samples. (It is worthwhile here to point out that room- temperature reduction of ca. 1 pm grain-size polycrys- talline films does not lead to broadening of the X-ray diffrac- tion peaks.34 During reduction oxygen loss will occur near the surface of the grains where most of the electrical potential drop (A4) occurs. As a result of initial oxygen loss, a chemical potential gradient (Ap) is set up which extends deeper into the grain.The resulting gradient in electrochemical potential (AD) will decrease with increasing depth into the grains. After reduction (A$=O), a finite Ap will remain. From the exper- imentally observed change in rest potential (ca. lo2 mV) during tens of minutes after reduction, we deduce that this Ap can lead to some further migration of oxygen, thus decreasing the final gradient, although this may be predominantly a surface effect. The possibility of such room-temperature migration of oxygen, under the influence of a chemical potential gradient alone, is underscored by the recent results of Jorgensen et .~~~1 and by the transmission electron microscopy obser- vations of Miiller et aL3’ As Ap decreases, this process will become slower until it finally becomes negligible on our timescale, leaving a grain with gradually decreasing oxygen content towards the surface (Fig.6). The changes in the physical properties upon reduction can be understood by comparison with results obtained for more homogeneously reduced samples. T, is expected to be higher (and the normal-state resistivity is expected to be more metallic) further inside each grain, where less reduction has occurred. This leads to a shell model, shown in idealized fashion in Fig. 7. A somewhat related model, to explain the J. MATER. CHEM., 1991, VOL. 1 *t I b--VO gt 0 distance into particle ‘ ‘ distance into particle Fig. 6 Effect of electrochemical reduction on oxygen content x and chemical potential Ap as a function of time of reduction and distance into an individual particle.V, is an oxygen vacancy in the 123 lattice, 0, is an oxygen atom on one of the possible oxygen sites in the basal plane, t, and t2 are successively increasing times of reduction and p, is the chemical potential of oxygen Fig. 7 Schematic illustration of the idealized shell model for 123, reduced at room temperature. The change in density of shading represents changes in oxygen content in a particle experimentally found lack of dependence on [O], i.e. x in YB~,CU~O~-~of the out-diffusion rate of oxygen, has been suggested by Tu et aL3*In the shell model more shells undergo the transition to the superconducting state, as the temperature is gradually lowered, and the resistivity of each grain gradually decreases.Full superconductivity is reached when the surface (most reduced) of each grain becomes superconducting (although tunnelling and proximity effects could lead to full superconductivity while the immediate surface region is still not superconducting). The observed effect of changes in the current density on the electrically measured transition to the superconducting state can be construed as support for this model. Another model that could explain the gradual decrease in resistivity of the reduced samples with decreasing temperature involves differences in overall oxygen content from one grain to another. According to this model, one grain might be completely superconducting, while another would not yet be superconducting, again leading to a gradual decrease in resistivity with decreasing temperature.However, there appears to be no obvious physical reason for such an abrupt change of properties between grains. The first model suggests a range of T,,i.e. not only at 60 and 90 K (similar behaviour was noted also by Namgung et QZ.,~’ using samples quenched in Hg). We attribute the occur- rence of such intermediate T, values to the lack of full relaxation of the lattice (cf: discussion of Fig. 2, above) and the absence of long-range order in the a direction with respect to oxygen vacancies [cf: also ref. 12(b) for a related idea]. Again, this is an effect of the low temperature of preparation. We note that results from high-resolution electron microscopy studies on oxygen-deficient 123 have been interpreted to suggest that superconductivity can occur even in samples that show very-short-range ordering of oxygen and vacancies.” J.MATER. CHEM., 1991, VOL. 1 Subsequent annealing of the room-temperature reduced samples allows for long-range order as deduced from the observation of definite 60 and 90 K transitions. Structural Implications of the Model Neutron powder diffraction data of oxygen deficient samples have shown that the most labile oxygen is located on the (O,+, 0) site.14 From this, together with the observation that for our reduced samples it is mainly the u lattice parameter that is affected, we deduce that the gradient in oxygen content leads to increasingly larger sections of uninterrupted -Cu(l)-0-Cu(1)-0-chains and longer distances between interrupted chain fragments deeper into the core of the grains.Long-range order of the type observed by X-ray diffraction in reduced samples'' or even the (probably shorter- range) order observed by electron diffraction in higher-tem- perature quench-reduced samplesI6 is not expected to be observed here owing to the low temperature of preparation. Near(est)-neighbour effects may lead to some degree of local ordering of oxygen vacancies but this would not be detected in the X-ray powder diffraction. The low temperature of preparation will also limit any structural rearrangements to local relaxation of the atoms. The broadening in the X-ray powder diffraction (the scattering is primarily due to the heavy atoms) can then be explained by the different degrees of relaxation of the heavy atoms due to the oxygen gradient.40 There is the additional consideration that not all grains have to respond in exactly the same manner to reduction.(Strain and small particle size can also cause broadening in X-ray powder diffraction peaks but cannot account for the shifts in a). Further structural studies are necessary to clarify these points. Upon annealing the sample, oxygen lability is increased owing to the higher temperatures, and structural rearrange- ment to stable phases occurs on a very short timescale, as reflected in the observation of distinct transitions to the superconducting state at 60 and at 20 K.This type of struc- tural rearrangement is probably an oxygen vacancy ordering, similar to that observed in samples reduced by other method^.".^^ The recent results of Jorgensen et confirm that such rearrangement results from the metastability of the disordered 0 sublattice. The differences in behaviour between unannealed and annealed samples with the same oxygen content can then be attributed primarily to the ordering of the oxygens. Reduction Products in the Electrolyte The oxygen that is extracted electrochemically from 123 reacts further with propylene carbonate to form propanal. The reaction can be described by the following equation zOCOCH2CH(CH3)0+YBa2Cu307--x +2ze--+zCH3CH2CH0+zC0; -+YBa2Cu307--x -(3) The cathodic decomposition of propylene carbonate has been studied previously and it was found that on metal electrodes (Pt, Ni, Li etc.)41*42 and propylene carbonate decomposes to propene and C0:- according to the following equation: OCOCH2CH(CH3)O+2e-+CH3CH=CH2+COi-(4) It is conceivable, although not likely, that propylene carbonate decomposes chemically, OCOCH2CH(CH3)O+CH3CH2CH0+C02 (5) However, this decomposition does not seem to be catalysed by 1236. The observed absence of measurable amounts of propanal using 1236 could simply be due to the very small amounts of current that we succeeded in passing, giving concentrations of propanal below the detection limit.How- ever, the main purpose of this control experiment was to show that the reaction is not a chemical reaction catalysed by an oxide biased with an electrical potential. There are also no reports of this type of decomposition, except by heating at temperatures >150 "C, in which case other products are formed besides propanal and carbon At room temperature, propylene carbonate undergoes an alkaline hydrolysis to yield propylene glycol, not pr~panal.~' Based on this discussion, we propose the following reactions [02-] +Cog- (6) +OCOCH2CH(CH3)b-+CH3CH2CH0 or [02-]+H20+20H- (74 OH-+OCOCH2CH(CH3)O- CH3CH2CH0+HCO, (74 The observation of the formation of tributylamine along with propanal during the reduction of Lao.5Sr,.5C003 -x is indica- tive of the presence of a reactive oxygen species on the surface of the oxides, since it is known that tetraalkylammonium salts react with strong bases to form trialkylamine~.~~ Under the conditions of the present experiments, [02-1 may be reacting directly or it may react first with trace amounts of water in the solvent to form OH-, which then reacts with the propylene carbonate and the tetrabutylammonium ion.The agreement between the values of y in eqn. (1) as calculated by coulometry and by iodometric titration, using oxygen di-anions [n=2, in eqn. (l)], shows that, when the oxygen leaves 1237 during reduction, it is not in a peroxide or superoxide form. The fact that only propanal was observed with 1237 may be due to differences in catalytic activity between Lao.5Sro~5Co03--x and 1237, and in particular the presence of La and Co instead of Y and Cu.The reaction of a 123 cathode with propylene carbonate to produce propanal and C0:-is analogous to the high- temperature loss of oxygen in 123. This process can be written in terms of the defect chemistry of oxygen47 (at least in the region above the metal-insulator transition) as [2zhib, YBa2Cu3O7]7+zO2 +[zV& YBa2Cu307 -,] (8) where hib is a hole in the valence band and V, is a dipositive oxygen vacancy. The square brackets enclose electrically neutral entities. The equivalent reaction for the electrochemi- cal reduction of 123 and subsequent reaction with propylene carbonate can be written as I I2ze-+[2zh;,, YBa2Cu307]+zOCOCH2CH(CH3)0 -+[zV& YBa2Cu307-,I +zCH3CH2CH0+zC0:-(9) where the electrons are introduced electrochemically.In both instances YBa2C~307-x is in the same final electronic state. As noted above, the formation of propanal is not an electro- chemical reaction, but is due to the presence of a reactive oxygen species on the surface of the 1237 electrode. Comparison between eqn. (8) and (9) (high-temperature DS. room-temperature methods of reduction of 1237) shows both the similarity of the two methods in that the stoichiometry of 1237 is identical (although there is a structural difference) and the difference in the fate of the oxygen. This difference, taken together with the difference between the products of propylene carbonate decomposition on 123 on the one hand and those obtained on metals or on 1236 on the other hand is significant and confirms earlier observations of 123 as an 346 active catal~st~~-~' and suggests its use in other catalytic and electrocatalytic systems.Conclusion We have found that the oxygen content of 123 can be reduced at room temperature in an electrochemical cell with a propyl- ene carbonate-tetrabutylammonium perchlorate electrolyte. This reduction is accompanied by an unusual reaction of the propylene carbonate on the 123 electrode which can be attributed to the relatively high oxygen mobility in this material and different surface chemistry as compared to a metal electrode. The as-reduced samples are inhomogeneous (probably with respect to an oxygen gradient), and appear to contain regions of new metastable structures, distinct from those obtained by high-temperature reduction.Comparison of the physical properties of the samples, as-reduced, with samples after annealing show that T, is a continuous function of oxygen content. The formation of samples with a T, of 60 K (and 20K) can be ascribed then to the formation of structures at temperatures that are high enough to allow for ordering of the oxygen vacancies. The distinct 20 K transition appears to be accessible only uia initial room-temperature reduction or by low-to-room-temperature rearrangement of quench-reduced samples.33 This work was supported by the United States-Israel Binational Science Foundation, Jerusalem.We wish to thank N. Fleischer for electrochemical advice, S. Reich for use of electrical resistivity equipment. A. Tishbee for help with the GC/MS analysis, and I. Felner for the VSM measurements. References 1 (a) P. K. Gallagher, H. M. O'Bryan, S. A. Sunshine and D. W. Murphy, Mater. Res. Bull., 1987, 22, 995; (b)A. Manthiram, J. S. Swinnea, Z. T. Sui, H. Steinfink and J. B. Goodenough, J. Am. Chem. SOC., 1987, 109, 6667. 2 L. H. Greene and B. G. Bagley, in Physical Properties of High Temperature Superconductors II, ed. D. M. Ginsburg, World Scientific, Singapore, 1990, in the press. 3 G. S. Grader, P. K. Gallagher, J. Thomson and M. Gurvitch, Appl. Phys. A, 1988, 45, 179. 4 K. N. Tu, S. I. Park and C. C. Tsuei, Appl. Phys. Lett., 1987, 51, 2158. 5 (a) S.I. Park, C. C. Tsuei and K. N. Tu, Phys. Rev. B, 1988, 37, 2305; (b)K. N. Tu, C. C. Tsuei, S. I. Park and A. Levi, Phys. Rev. B, 1988,38, 772. 6 D. S. Ginley, P. J. Nigrey, E. L. Venturini, B. Morosin and J. F. Kwak, J. Mater. Res., 1987, 2, 732. 7 A. Yoshida, H. Tamura, S. Morohashi and S. Hasuo, Appl. Phys. Lett., 1988, 53, 811. 8 (a) E. J. M. O'Sullivan and B. P. Chang, Appl. Phys. Lett., 1988, 52, 1441; (b)R. Beyers, E. M. Engler, P. M. Grant, S. S. P. Parkin, G. Lim, M. L. Ramirez, K. P. Roche, J. E. Vazquez, V. Y. Lee and R. D. Jacowitz, Physica C, 1988, 153-155, 590. 9 (a) D. J. Vischjager, P. J. van der Put, J. Schram and J. Schoonman, Solid State Ionics, 1988,27, 199; (b)J. MacManus, D. Fray and J.Evetts, Physica C, 1990, 169, 193. 10 G. Cannelli, R. Cantelli and F. Cordero, Phys. Rev. B, 1988, 38, 7200. 11 D. Shi, J. Krucpzak, M. Tang, N. Chen and R. Bhadra, J. Appl. Phys., 1989,66,4325. 12 (a) R. J. Cava, B. Batlogg, C. H. Chen, E. A. Reitman, S. M. Zahurak and D. Werder, Phys. Rev. B, 1987, 36, 5719; (b) R. J. Cava, A. W. Hewat, E. A. Hewat, B. Batlogg, M. Marezio, K. M. Rabe, J. J. Krajewski, W. F. Peck Jr. and L. W. Rupp Jr., Physica C, 1990, 165,419. 13 W. E. Farneth, R. K. Bordia, E. M. McCarron 111, M. K. Craw-ford and R. B. Flippen, Solid State Commun., 1988, 66, 953. J. MATER. CHEM., 1991, VOL. 1 14 W.K. Kwok, G. W. Crabtree, A.Umezawa, B. W. Veal, J.D. Jorgensen, S. K. Malik, L. J. Nowicki, A. P. Paulikas and L.Nunez, Phys. Rev. B, 1988, 37, 106. 15 R. M. Fleming, L. F. Schneemeyer, P. K. Gallagher, B. Batlogg, L. W. Rupp and J. V. Waszczak, Phys. Rev. B, 1988,37, 7920. 16 M. A. Alario-Franco, C. Chaillout, J. J. Capponi and J. Chenavas, Mater. Res. Bull., 1987, 22, 1685. 17 L. T. Wille, A. Berera and D. de Fontaine, Phys. Rev. Lett., 1988, 60,1065. 18 Y. Kubo and H. Igarashi, Jpn. J. Appl. Phys., 1987,26, L1988. 19 M. Hervieu, B. Domenges, B. Raveau, M. Post, W. R. McKinnon and J. M. Tarascon, Mater. Lett., 1989, 8, 73. 20 R. Beyers, B. T. Ahn, G. Gorman, V. Y. Lee, S. S. P. Parkin, M. L. Ramirez, K. P. Roche, J. E. Vazquez, T. M. Gur and R. A. Hug-gins, Nature (London), 1989, 340, 619. 21 F. Beech, S. Miraglia, A. Santoro and R.S. Roth, Phys. Rev. B, 1987,35, 8778. 22 B. Rupp, P. Fischer, E. Porschke, R. R. Arons and P. Meuffels, Physica C, 1988, 156, 559. 23 Z. Z. Wang, J. Clayhold, N. P. Ong, J. M. Tarascon, L. H. Greene, W. R. McKinnon and G. W. Hull, Phys. Rev. B, 1987,36, 7222. 24 A. Santoro, S. Miraglia, F. Beech, S. A. Sunshine, D. W. Murphy, L. F. Schneemeyer and J. V. Waszczak, Mater. Res. Bull., 1987, 22, 1007. 25 M. Schwartz, M. Rappaport, G. Hodes and D. Cahen, Physica C, 1988, 153-155, 1457. 26 M. Schwartz, D. Cahen, M. Rappaport and G. Hodes, Solid State Ionics, 1989, 32-33, 1137. 27 M. Schwartz, M. Rappaport, G. Hodes, S. Reich and D. Cahen, Mater. Lett., 1989, 7, 41 1. 28 T. Kudo, H. Obayashi and T. Gejo, J. Electrochem. SOC., 1975, 122, 159.29 F. R. van Buren, G. H. J. Broers, A. J. Bouman and C. Boesveld, J. Electroanal. Chem., 1978, 88, 353. 30 R. L. Barns and R. A. Laudise, Appl. Phys. Lett., 1987, 51, 1373. 31 H. Eickenbusch, W. Paulus, E. Gocke, J. F. March, H. Koch and R. Schollhorn, Angew. Chem. Int. Ed. Engl., 1987, 26, 1188. 32 K. Kishio, J. Shimoyama, T. Hasegawa, K. Kitazawa and K. Fueki, Jpn. J. Appl. Phys., 1987, 26, L1228. 33 J. D. Jorgensen, S. Pei, P. Lightfoot, H. Shi, A. P. Paulikas and B. W. Veal, Physica C, 1990, 167, 571. 34 (a)Y. Scolnik, E. Sabatini and D. Cahen, Physica C, in the press; (b)Y. Scolnik and D. Cahen, manuscript in preparation. 35 A. E. White, K.T. Short, D. C. Jacobson, J. M. Poate, R.C. Dynes, P. M. Mankiewich, W. J. Skocpol, R.E. Howard, M. Anzlowar, K. W. Baldwin, A. F. J. Levi, J. R. Kwo, T. Hsieh and M. Hong, Phys. Rev. B, 1988,37, 3755. 36 S. J. Rothman, J. 1. Routbort and J. E. Baker, Phys. Rev. B, 1989, 40,8852. 37 J-H. Muller, W. Mertin and R. Gruehn, Supercond. Sci. Technol., 1990, 3, 273. 38 K. N. Tu, N. C. Yeh, S. I. Park and C. C. Tsuei, Phys. Rev. B, 1989, 39, 304. 39 C. Namgung, J. T. S. Irvine and A. R. West, Physica C, 1990, 168, 346. 40 T. Wieder, K. Thomas and H. Gartner, Appl. Phys. A, 1988, 46, 165. 41 G. Eggert and J. Heitbaum, Electrochim. Acta, 1986, 31, 1443. 42 G. Eichinger, J. Electroanal. Chem., 1976, 74, 183. 43 C. K. Mann, in Electroanalytical Chemistry, ed. A. J. Bard, Marcel Dekker, New York, 1969, vol. 3, p. 119. 44 R. Jasinski, in Advances in Electrochemistry and Electrochemical Engineering, ed. P. Delahay and C. W. Tobias, Wiley, New York, 1971, vol. 8, p. 253. 45 A. H. Saadi and W. H. Lee, J. Chem. SOC.B, 1966, 1. 46 J. March, Advanced Organic Chemistry, Wiley, New York, 3rd edn., 1985, pp. 906-909. 47 J-H. Park and P. Kostic, Mater. Lett., 1988, 6, 327. 48 S. Hansen, J. Otamiri, J-0. Bovin and A. Anderson, Nature (London), 1988, 334, 143. 49 N. Mizuno, M. Yamato and M. Misono, J. Chem. SOC., Chem. Commun., 1988, 13, 887. 50 E. A. McNamara, S. A. Montzka, R. M. Barkley and R. E. Siev- ers, J. Chromatogr., 1988, 452, 75. Paper 0/04644B; Received 16th October, 1990