J O U R N A L O F C H E M I S T R Y Materials Synthesis, structure, electrical and magnetic properties of the new nonstoichiometric perovskite phase, Ca2MnNbOc Angela Kruth,a Mitsuharu Tabuchi,b Ulrich Guthc and Anthony R. Westa aUniversity of Aberdeen, Department of Chemistry, MestonWalk, Aberdeen, UK AB24 3UE bOsaka National Research Institute, 1-8-31 Midorigaoka, Ikeda, Osaka 563, Japan cUniversity of Greifswald, Department of Chemistry, Institute of Physical Chemistry, Soldtmannstraße 23, 17489 Greifswald, Germany Received 26 June 1998, Accepted 6th August 1998 The new phase Ca2MnNbOc has an orthorhombic, GdFeO3 structure with B site disorder of Mn and Nb.The oxygen content c varies from 5.86 to 6.00 depending on heat treatment conditions. Treatment in high pressure O2 is required for full oxidation.Magnetic susceptibility measurements indicate mixed valence (2+, 3+) ofMn for c=5.87 and Curie–Weiss paramagnetism at high temperatures. SQUID measurements indicate spin glass-like behaviour below 15 K. Ac impedance shows semiconducting behaviour, E=0.31(2) eV. On heating in O2-poor atmospheres, resistive surface layers and grain boundary eVects are seen, which point to a p-type conduction mechanism.stack together with a layer of A-atoms in between. A homolo- Introduction gous series, AnBnO3n+2 forms, with n=4 in this case, but in Perovskites and perovskite-related oxides oVer a wide range other phases n-values range from 1 to 6.12–14 The brownmillerof interesting electrical properties, e.g. giant magnetoresistance, ite structure of Ca2Fe2O5 and Ca2FeAlO5 is derived from superconductivity and oxide ion conductivity. Mixed oxide perovskite with ordering of oxygen vacancies.15,16 The oxygen ion–electron conducting perovskites have possible uses as vacancies give tetrahedral coordination of iron resulting in electrode materials in oxygen sensors and solid-oxide fuel cells alternate sheets of FeO6 octahedra and FeO4 tetrahedra. and may have major advantages over porous noble-metal CaMnO3-d also has oxygen vacancies, in the compositional electrodes because they are stable to temperatures well above range 0d0.5, which order at certain compositions; Mn has 1250 °C1 and charge-transfer reactions can take place every- mixed 3+/4+ valence and occupies a mixture of octahedra where at the electrode/gas interface, rather than at the triple and square-pyramids which form ordered intergrowth strucjunctions: electrolyte/electrode/gas.In some non-stoichio- tures where the proportion of square-pyramidal Mn3+ relative metric perovskites, electrochemical reactions occur at specific to octahedral Mn4+ increases with d. In CaMnO2.5, d=0.5, vacant surface sites;2–4 these materials exhibit catalytic activity the Mn is 3+ and in five-fold square-pyramidal coordination for processes such as the oxidation of CHx and NOx.only, which is distinct from brownmillerite which contains a Perovskites based on elements of the first transition series mixture of tetrahedra and octahedra.17,18 display a wide variety of defect-related phenomena which have On replacing Fe3+ in brownmillerite, Ca2Fe2O5, by Nb5+, their origin in the unfilled 3d electron shell.The relative ease an extensive oxygen-deficient, cubic perovskite solid solution, with which electrons can be removed from or added to some Ca2Fe2-xNbxOc: 0.45x0.65 was obtained.19 Its oxygen metals, especially Fe and Mn, means that they often exist in content is variable, e.g. 5.55c5.92 for x=0.6. It has high numerous valence states, depending on temperature and electronic conductivity and probably also has high oxide ion oxygen partial pressure; this also gives rise to a potentially conductivity since the oxygen content, c, readjusts rapidly on large number of phases, which themselves may deviate from varying either temperature or oxygen partial pressure and exact stoichiometry in terms of both cation and oxygen could therefore have uses as a mixed conducting electrode in content.5 Non-stoichiometry was until recently explained in devices involving gas–solid interfacial reactions.terms of point defect equilibria.6,7 For most compounds with This project is based on the Ca2Fe2O5–CaMnO3–Ca2Nb2O7 sizeable deviations (>1%) in stoichiometry, however, the con- system with the initial objective of synthesising perovskitecept of point defect equilibria is inadequate. Alternative models related phases with variable oxygen content, high mixed elecare needed which involve ordering of point defects or new tronic and oxide ion conductivities and stability under both structural features which eliminate point defects.8 Defects in reducing and oxidising conditions over large temperature perovskite oxides can arise from cation deficiency in either A ranges. We attempt to control the oxidation state of Fe, Mn or B sites or oxygen deficiency; the resulting vacancies often and associated oxygen content by both adjusting the niobium form ordered superstructures.concentration at B-cation sites and post-reaction heat treat- Tofield and Scott9 suggested three possible models to ments at diVerent oxygen pressures.Initial investigations were accommodate oxygen excess in the perovskite structure: carried out on the binary joins, Ca2Fe1.4-xMnxNb0.6Oc and (i ) introduction of interstitial oxygen at either (D00) or (BBB) Ca2Mn2-xNbxOc. As a first step, a variety of characterisation positions of the cubic unit cell; this, however, is energetically studies on the new phase, Ca2Mn2-xNbxOc with x=1.0, were unfavourable; (ii ) creation of cation vacancies at A and/or B made and are described here.sites, leaving a perfect oxygen sublattice and (iii ) formation of new, oxidised perovskite-related phases. For example, in LaMnO3+l and La2TiCoO6+l, oxygen excess is accommo- Experimental dated by vacancies in both A and B sites.10,11 In Cu2Nb2O7 Starting materials CaCO3 (99%, AnalaR) and Nb2O5 (99.9%, and La2Ti2O7 with large oxygen excess, new perovskite-related Aldrich) were dried at 200 and 600 °C, respectively. MnO2 structures form: the parent, cubic perovskite is sliced parallel to (110) to give slabs of composition (An-1BnO3n+2)2 which (99%, Aldrich) was used directly from the bottle since a J.Mater. Chem., 1998, 8(11), 2515–2520 2515thermogravimetry (TG) study confirmed its stoichiometry, were determined by ac impedance spectroscopy over the frequency range 0.03 Hz to 13 MHz using a Hewlett-Packard during a two-stage decomposition to Mn2O3 at ca. 400–700 °C and Mn3O4 at ca. 900–1000 °C. Appropriate quantities of 4192A LF Impedance Analyser.Pellets, 8 mm diameter and 1–2 mm thick, were prepared by cold-pressing powders at ca. MnO2, CaCO3 and Nb2O5 giving ca. 5 g total weight were mixed in acetone for at least 10 minutes, dried and heated in 200 MPa, sintered in air overnight at 1400 °C and quenched to room temperature. To modify the oxygen content, pellets Pt boats at 900 °C for 12 h, 1200 °C for 20 h and at 1200 °C to 1350 °C in steps of 50 °C for periods of 20 h each, with were heated at 1100 °C for 5 h and slowly cooled to room temperature (2 K min-1) in either O2 or N2.One pellet was regrinding between each heating period. Samples were finally heated at 1400 °C for 72 h. annealed at 600 °C and ca. 40 atm O2. Electrodes were attached by coating pellet faces with a thin InGa (151) alloy layer and To modify the oxygen stoichiometry, samples were heated at diVerent temperatures in air, O2 or Ar and either quenched gold strips held to the sample by pressure.or cooled slowly to room temperature. AMorris High Pressure Furnace, HPS-3210 was used to anneal samples, wrapped in Results and discussion Au foil, under high O2 pressure.Two experimental conditions were used which gave pressures at 600 °C of either ca. 40 or The new phase, Ca2MnNbOc, was obtained on reaction of the oxides in a stepwise heating programme; phase-pure ca. 60 atm. After heating at 600 °C for 10 h, samples were cooled slowly under pressure. samples were obtained finally after heating at 1400 °C, 72 h. This new phase was easily recognised since its XRD pattern Oxygen contents and the oxidation state of Mn were determined from a combination of TG and magnetic measure- is very similar to that of Ca2FeNbO619 and, therefore, it also has an orthorhombic, GdFeO3 structure.ments. By TG, the oxygen contents were determined from the weight loss on reduction in 10% H2–90% N2 using a Stanton Oxygen contents, c, were determined by H2-reduction TG; typical traces are shown in Fig. 1.In all cases, no reduction Redcroft TG-DTA 1500 instrument. A Shimadzu MB-3 magnetic balance was used to measure magnetic susceptibility occurred up to at least 400 °C but reduction was complete by 950–1150 °C. Samples with higher oxygen contents (c,d ) values which were converted to eVective magnetic moments and then compared with theoretical values obtained from the started to lose weight at lower temperatures than the others.Although weight losses could be determined accurately for ‘spin-only’ formula: (NH4)2Mn(SO4)2·6H2O was used as calibration standard. SQUID measurements were carried out all samples, in order to convert these to variations in oxygen content, it was necessary to know the stoichiometry of either below 100 K with magnetic fields of 5000 and 10 Oe. Phase identity, stability and purity were determined by the final, fully reduced state or that of at least one of the starting compositions.The assumption was made initially that powder X-ray diVraction, Ha�gg-Guinier camera, Cu-Ka1 radiation. For indexing and lattice parameter refinement, a Philips the H2-reduction would aVect only the oxidation state of Mn and that the final state in all cases would be Mn2+, correspond- DiVractometer PW1710 and Stoe Stadi P software were used, KCl internal standard, 20 to 80° 2h, Cu-Ka1 radiation.A Stoe ing to an overall final stoichiometry of ‘Ca2MnNbO5.5’; H2 reduction of CaMnO3-d had shown the product to be Stadi P transmission diVractometer was used for structural studies (Rietveld refinement) of the fully oxidized phase, CaMnO2, containing Mn2+17 and we assumed a similar behaviour for our samples.Using this assumption, the starting Ca2MnNbO6 with data collected over the range 8 to 113° 2h and step width 0.02°. The refinement was carried out using compositions were evaluated. Two samples heated in high pressure O2 had c values close to 6, viz. 6.02(3) after 40 atm the package PFSR (pattern–fitting structure–refinement) with programs CDF, CDE, RVI and RVR. First, the lattice param- O2 at 600 °C and 5.97(3) after 60 atm O2; within errors, it is concluded that these were fully oxygenated with c=6.00. An eters and halfwidths of the reflections were refined, e.g. lattice constants, 2h zero point, halfwidth parameters and back- oxygen content of 6 is the highest that can be expected for this structure type; a large number of phases with analogous ground.After convergence of profile refinement, the structural parameters were refined, e.g. overall scale factor, atomic formulae and GdFeO3 structure are known and all have an oxygen content of 6. This result therefore further supports coordinates and isotropic thermal vibration parameters.DTA was used to look for any phase transitions over the range 25 the correctness of the initial assumption concerning the stoichiometry of the H2-reduced samples. to 1300 °C in air, at heating and cooling rates of 8 K min-1. Electrical properties in air over the range -60 to 200 °C The oxygen content of Ca2MnNbOd varied according to the Fig. 1 Weight loss and corresponding oxygen content, c, during H2 reduction of samples prepared under diVerent heating conditions: (a) slowly cooled from 1100 °C in Ar, (b) quenched from 1400 °C in air, (c) slowly cooled from 1100 °C in O2 and (d) heated at ca. 40 atm O2 and 600 °C; during heating 2 and cooling & cycle. 2516 J. Mater. Chem., 1998, 8(11), 2515–2520Table 1 X-Ray powder diVraction data for Ca2MnNbO5.87 orthorhombic; a=5.4527(5) A° , b=5.5622(5) A° , c=7.7062(8) A° h k l Int.d(obs)/A° d(calc)/A° 1 1 0 30 3.8933 3.8938 0 0 2 14 3.8540 3.8531 1 1 1 3 3.4753 3.4753 0 2 0 25 2.7808 2.7811 1 1 2 100 2.7388 2.7388 2 0 0 25 2.7267 2.7263 2 1 0 2 2.4467 2.4481 1 2 1 2 2.3597 2.3586 2 1 1 3 2.3333 2.3332 1 0 3 2 2.3252 2.3236 1 1 3 2 2.1441 2.1442 1 2 2 2 2.0834 2.0839 Fig. 2 Influence of annealing temperature in air on oxygen content, 2 1 2 2 2.0653 2.0663 c, in Ca2MnNbOc . 2 2 0 25 1.9462 1.9469 0 0 4 13 1.9265 1.9266 2 2 1 3 1.8873 1.8876 1 3 0 4 1.7552 1.7554 final heat treatment conditions; lowest values were obtained 2 2 2 5 1.7375 1.7377 at high temperatures, e.g. 5.87 in air at 1400 °C and 5.86 in 1 1 4 6 1.7266 1.7268 Ar at 1200 °C.The oxygen content as a function of temperature 1 3 1 2 1.7111 1.7115 in air is shown in Fig. 2. The data were obtained on samples 1 3 2 12 1.5973 1.5974 quenched from diVerent temperatures and show that oxygen 0 2 4 8 1.5837 1.5837 3 1 2 22 1.5758 1.5764 content varies over the range ca. 400 to 700 °C but is essentially 2 0 4 16 1.5731 1.5734 constant at higher and lower temperatures. 1 3 3 1 1.4487 1.4493 Indexed powder XRD data for one composition, 0 4 0 2 1.3906 1.3905 Ca2MnNbO5.87, are given in Table 1. The data index on an 2 2 4 10 1.3694 1.3694 orthorhombic unit cell with a Ó2, Ó2, 2 relation to a cubic 4 0 0 3 1.3626 1.3632 perovskite-like subcell. Lattice parameters show a small 3 2 3 2 1.3094 1.3091 4 1 1 1 1.3052 1.3049 increase in c with decreasing c from 7.683 A° for c=6.00 to 3 3 0 2 1.2977 1.2979 7.706 A° for c=5.87; a and b change little with c. 4 0 2 2 1.2856 1.2851 Ca2MnNbO6 for Rietveld refinement was synthesised in 60 2 4 0 1 1.2389 1.2387 atm oxygen. The starting model was the orthorhombic per- 3 3 2 3 1.2301 1.2300 ovskite-related GdFeO3 structure.20 A Pearson VII profile 2 4 1 3 1.2234 1.2230 function with exponent m=2.0 was applied.The profile and 0 4 3 6 1.2224 1.2229 1 1 6 3 1.2194 1.2197 structural parameters allowed to refine were scale factor, 2h 2 4 2 4 1.1793 1.1793 zero point, cell constants, background, halfwidth polynomial, 4 2 2 4 1.1663 1.1666 atomic positions for all atoms and thermal vibration param- 0 2 6 1.1660 eters for Ca and Mn/Nb. After convergence, R-factors were: 3 3 3 4 1.1584 1.1584 R( p)=0.0343, R(wp)=0.0443 and R(I,hkl)=0.1332.Final 1 3 5 1.1582 values for atomic positions and thermal parameters are given 0 4 4 2 1.1279 1.1275 4 0 4 2 1.1129 1.1128 in Table 2 and Table 3 shows calculated CaMO and 1 5 2 4 1.0489 1.0488 Nb/MnMO bond lengths. The XRD pattern for Ca2MnNbO6 2 4 4 2 1.0421 1.0419 and the diVerence between observed and calculated profiles 1 3 6 2 1.0367 1.0365 are shown in Fig. 3. 4 2 4 6 1.0331 1.0331 Magnetic measurements were carried out for 5 1 2 4 1.0310 1.0311 Ca2MnNbO5.87; Fig. 4(a) shows the field dependence of mag- 3 1 6 1.0307 netization at 83 and 293 K between 1.8 and 12.5 kOe. No spontaneous magnetization was observed down to 83 K. The magnetic susceptibility was 4.17×10-5 cm3 g-1, which is Fe analogue structures, A2FeXO65A=Ca, Sr, Ba; X=Nb, Ti, Sb.21–25 suYciently large to ignore the contribution of closed shell diamagnetism, 3.33×10-7 cm3 g-1.Temperature dependence Ac impedance data were recorded isothermally at temperatures in the range 200 to 400 K for pellets that had of the inverse molar susceptibility, xm-1, reveals a Curie– Weiss paramagnetism down to 83 K, Fig. 4(b), with meff= been given a range of post-sinter heat treatments. Two general patterns of behaviour were seen. Samples heat-treated in O2- 5.341(6) mB and h=+21 K. The positive Weiss temperature suggests ferromagnetic coupling below 83 K. The observed meff rich atmospheres showed the simplest response, Fig. 6 (a), (b); the Z vs. Z¾ complex plane plot showed a main arc and a value is intermediate between the spin only values of Mn2+, meff,Mn2+=Ó35 mB=5.92 mB, and Mn3+, meff,Mn3+=Ó24 mB= much smaller, poorly-resolved, low frequency arc.Z, M spectra showed essentially coincident peaks at high frequency 4.90 mB, suggesting that this sample is a Mn2+/3+ mixed valence compound. This iconsistent with the TG value for with associated capacitance 8–10 pF, indicative of bulk or intragranular response.The low frequency Z shoulder peak the oxygen content, c=5.87, which gives a 26574 ratio for Mn2+5Mn3+. had a capacitance of ca. 10 nF, attributable to either a thin surface layer or an electrode–sample interfacial eVect. Hence, SQUID measurements made for two applied fields over the temperature range 4–100 K are shown in Fig. 5.Above 20 K, the impedance data were dominated by the bulk response of the sample and there is little evidence of resistive grain the data sets are essentially superposable but diverge dramatically at lower temperature. In particular, the zero field cooled boundary eVects (which would have capacitances in the range 0.1 to 1 nF). data at 10 Oe show a sharp magnetisation maximum at 15 K which transforms to an almost temperature-independent mag- For samples heated in air or N2/Ar, the impedance response was more complex, Fig. 6(c), (d). The Z vs. Z¾ plots showed netisation in field cooled data. Such features are strongly indicative of spin glass-like behaviour of the kind reported in clear evidence of several overlapping arcs, but also dramatic J. Mater. Chem., 1998, 8(11), 2515–2520 2517Table 2 Values for atomic coordinates and thermal vibration parameters for Ca2MnNbO6 in Rietveld refinement WyckoV Atom position Occ. x/a y/b z/c Uij Ca 4c 1.0 -0.0074(11) 0.0050(5) 0.2500 0.0150(1) Mn 4b 0.5 0.5000 0.0000 0.0000 0.0025(4) Nb 4b 0.5 0.5000 0.0000 0.0000 0.0025(4) O(1) 4c 1.0 0.0915(15) 0.4684(19) 0.2500 0.0030 O(2) 8d 1.0 -0.2946(14) 0.2947(12) 0.04531(8) 0.0030 Table 3 Bond distances, D, and bond angles, a, for Ca2MnNbO6 CaMO D/A° OMMn/NbMO a/° CaMO(1) 2.303(10) (1×) O(2)M Mn/NbMO(2) 87.5(3) CaMO(2) 2.358(7) (2×) O(2)M Mn/NbMO(2) 89.0(3) CaMO(1) 2.416(11) (1×) O(2)M Mn/NbMO(1) 89.3(3) CaMO(2) 2.618(8) (2×) O(2)M Mn/NbMO(1) 90.7(3) CaMO(2) 2.700(7) (2×) O(2)M Mn/NbMO(1) 91.0(3) O(2)MMn/NbMO(1) 92.5(3) Mn/NbMO D/A° Mn/NbMO(1) 1.992(2) (2×) Mn/NbMO(2) 1.999(7) (2×) Mn/NbMO(2) 2.013(7) (2×) Pbnm (no. 62); a=5.445 A° , b=5.555 A° , c=7.683 A° . R( p)=0.0343, R(wp)=0.0443, R(I,hkl)=0.1332. Fig. 5 SQUID measurement for Ca2MnNbO5.87: at 10 (&; %) and 5000 Oe ($; #). changes occurred on polishing the pellet surfaces prior to attaching electrodes (c). The bulk response was again seen at highest frequencies in the M spectra, as the low frequency tail of a peak at 107 Hz (d), but the corresponding Z spectra and therefore the Z vs.Z¾ plots were completely dominated by large, lower frequency impedances. The sensitivity of the impedance response to polishing suggests that resistive surface layers are present, even after polishing; however, there is a large impedance, attributable to grain boundaries from its capacitance value of ca. 0.2 nF, which dominates the overall sample impedance. The conclusion from these observations is that resistive grain boundaries and surface layers appear on heating samples in less oxidising atmospheres. In order to assess possible variation in bulk resistance as a Fig. 3 Rietveld refinement: (a) observed pattern and (b) diVerence consequence of diVerent heat treatments, data for the frebetween observed and calculated patterns. quency maxima in the M spectra are compared in Fig. 7. A major advantage of comparing fmax data for a range of samples is that fmax is independent of sample geometry: R and C are both influenced by geometry, but inversely to each other and therefore, the geometrical terms cancel in fmax.26 In addition, C values vary little from sample to sample for the bulk response of similar-sized pellets. Hence fmax, given by 2pfmaxRC=1, provides a direct measure of R-1 and hence of s.The data for four diVerent heat treatments fall on a single straight line with activation energy 0.31(2) eV. This suggests that the bulk conduction mechanism, the conducting species and the magnitude of the conductivity are essentially independent of oxygen content for this range of compositions.Given the complete absence of any electrode polarisation eVects in the low frequency region of the impedance response and the low value of the activation energy, it is concluded that the conduction species are electrons rather than ions, and therefore are associated with the mixed valency of Mn.Conductivity data were extracted from the fmax data. These are plotted in Arrhenius format in Fig. 8, together with total pellet conductivity values, obtained from the low frequency intercepts of the Z vs. Z¾ plots on the Z¾ axis. The pellet Fig. 4 Magnetic data for Ca2MnNbO5.87: (a) field dependence of magnetization and (b) Curie–Weiss plot. conductivity values show a large variation, unlike the bulk 2518 J.Mater. Chem., 1998, 8(11), 2515–2520Fig. 6 (a) Complex impedance plane and (b) combined M, Z spectroscopic plot at 252 K for Ca2MnNbO6, annealed in O2 at 40 atm, 600 °C; (c) complex impedance plane and (d) combined M, Z spectroscopic plot at room temperature for Ca2MnNbO5.87, quenched in air from 1400 °C; $ polished pellet, #, unpolished pellet.values which are almost independent of sample history. For pellets treated in O2-rich atmospheres, the grain boundary resistance was relatively small and the total and bulk conductivities are almost coincident, Fig. 8. For the samples treated in air or N2, however, very large grain boundary resistances are evident, Fig. 6(c), (d), and these dominate completely the pellet conductivities.In such O2-poor atmospheres, the samples lose O2 from their surfaces by the reaction 2O2-�O2+4e Since the surface/grain boundary conductivity decreases as a consequence, it is concluded that the conduction mechanism, in these regions at least, is p-type. The similarity in activation energy for the bulk and total conductivities in unreduced samples, Fig. 8 (slow cool from 1100 °C, 1 atm O2 and from 600 °C, 40 atm O2), indicates that a similar conduction mechanism occurs in both bulk and grain boundary regions and Fig. 7 Arrhenius plots of bulk (filled symbols) and total (open therefore that the bulk conductivity is also p-type. The insensi- symbols) conductivities of Ca2MnNbOc , prepared by diVerent heat tivity of the bulk conductivity to sample history indicates that treatments; &, %: slow cool from 1100 °C in nitrogen; ,, (: the number of p-type carriers is high, associated with the high quenched from 1400 °C in air; $, #: slow cool from 1100 °C in 1 atm oxygen; +, 6: slow cool from 600 °C in 40 atm oxygen.concentration of Mn3+ ions. Conclusions Ca2MnNbO6 has the orthorhombic GdFeO3 structure. This is not surprising given the existence of similar structures in Ca2FeNbO6,27 Sr2CrTaO628 and Pb2ScNbO6.29 The similar sizes of Mn3+ and Nb5+ permit disorder of these elements over the octahedral B sites.Although Mn3+ is Jahn–Teller active, there is no evidence of distortions in the (Mn,Nb)O6 octahedra, presumably because there is insuYcient Mn present to exert a cooperative distortion. The possible existence of oxygen vacancies in the GdFeO3 structure is often speculated upon but is not well documented.Here, we find oxygen contents as low as 5.87 in samples heated Fig. 8 Arrhenius plots of bulk (filled symbols) and total (open at high temperatures and in atmospheres of low oxygen partial symbols) conductivities of Ca2MnNbOc, prepared by diVerent heat pressure. We believe that the oxygen content may be reduced treatments: &, %: slow cool from 1100 °C in nitrogen; ,, (: further on H2 reduction, although in fully reduced (to Mn2+ quenched from 1400 °C in air; $, #: slow cool from 1100 °C in 1 atm oxygen; +, 6: slow cool from 600 °C in 40 atm oxygen.at c=5.50) samples, the products are multiphase and therefore, J. Mater. Chem., 1998, 8(11), 2515–2520 2519Chem., Proceedings of the Second European Conference, Veldhoven, the oxygen content can not be reduced as low as 5.50; further ed.J. Schoonman, Elsevier, Amsterdam, 1982, p. 247. work on this is in progress. 4 S. Shin, Y. Hatakeyama, K. Ogawa and. Shimomura, Mater. The defect structure in the oxygen-deficient samples is not Res. Bull., 1979, 14, 133. known, but it must involve reduction in coordination of some 5 A.Atkinson, Adv. Ceram., 1987, 23, 3. of the (Nb,Mn)O6 octahedra. NbO6 octahedra are not easily 6 F. A. Kroger, The chemistry of the imperfect crystal, North- Holland, Amsterdam, 1964. reduced and oxygen loss may, therefore, be confined to the 7 G. G. Libowitz, Prog. Solid State Chem., 2, Pergamon Press, MnO6 octahedra, as occurs with CaMnO3 which exhibits a Oxford, 1965.coordination number of 5 for Mn on oxygen loss. Interestingly, 8 C. N. R. Rao, J. Gopalakrishnan and K. Vidyasagar, Indian it is diYcult to fully oxygenate Ca2MnNbOc and an oxygen J. Chem. A, 1984, 23, 265. content of 6.00 is achieved only under high oxygen pressures; 9 B. C. Tofield and W. R. Scott, J. Solid State Chem., 1974, 10, 183.this may reflect a reluctance of the Jahn–Teller active Mn3+ 10 C. N. R. Rao, A. K. Cheetham and R. Mahesh, Chem. Mater., 1996, 8, 2421. ion to occupy an octahedral site that is constrained to be 11 R. Mahesh, K. R. Kannon and C. N. R. Rao, J. Solid State undistorted by the adjacent NbO6 octahedra. Chem., 1995, 114, 294. Ca2MnNbOc provides another example of spin glass-like 12 R.Portier, A. Carpy, M. Fayard and J. Galy, Phys. Status Solidi behaviour, similar to that seen in A2FeXO6: A=Ca, Sr, Ba; A, 1975, 30, 683. X=Nb, Ti, Sb.21–25 It is a p-type hopping semiconductor with 13 M. Hervieu, F. Studer and B. Raveau, J. Solid State Chem., 1977, an activation energy, 0.31 eV, that is largely independent of 22, 273. 14 K. Scheunemann and H. K. Muller-Buschbaum, J.Inorg. Nucl. oxygen content and appears to be associated with the Mn3+ Chem., 1975, 37, 1875; 2261. ions that are present in high concentration. In reduced samples, 15 J. Berggren, Acta Chem. Scand., 1977, 25, 3616. there appears to be preferential oxygen loss from sample 16 J. C. Grenier, M. Pouchard and P. Hagenmuller, Structure surfaces and grain boundaries, as evidenced by a dramatic Bonding, 1981, 47, 1.decrease in conductivity in these regions. This is probably 17 K. R. Poeppelmeier, M. E. Leonowitz and J. M. Longo, J. Solid because the concentration of Mn3+ ions is greatly reduced, in State Chem., 1982, 44, 89. 18 K. R. Poeppelmeier, M. E. Leonowitz, J. C. Scalon, J. M. Longo favour of Mn2+ ions, and hence the p-type carrier concenand W. B. Yelon, J.Solid State Chem., 1982, 45, 71. tration is small. The bulk of the samples is largely uninfluenced 19 J. A. Chavez-Carvayar, T. C. Gibb and A. R. West, J. Mater. by this eVect (although the conductivity of the sample heated Chem., 1996, 6, 1957. in N2 is reduced, which may signal the outset of a significant 20 S. Geller and E. A.Wood, Acta Crystallogr., 1956, 9, 563. reduction in p-type carrier concentration in the bulk for this 21 R. Rodrý�guez, A. Ferna�ndez, A. Isalgue�, J. Rodrý�guez, A. Labarta, sample). Whilst it is widely accepted that oxidation/reduction J. Tejada and X. Obradors, J. Solid State Phys., 1985, 18, L401. 22 T. C. Gibb, P. D. Battle, S. K. Bollen and R. J. Whitehead, processes generally commence at surfaces, the present material J. Mater. Chem., 1992, 2, 111. provides a particularly clear example of the oxygen concen- 23 T. C. Gibb, J. Mater. Chem., 1993, 3, 441. tration gradients that must often occur; this eVect is readily 24 T. C. Gibb, A. J. Herod and N. Peng, J. Mater. Chem., 1995, 5, 91. apparent here because of the sensitivity of the electrical 25 P. D. Battle, T. C. Gibb, A. J. Herod, S.-H. Kim and P. H. Munns, properties to the degree of reduction. J. Mater. Chem., 1995, 5, 865. 26 J. T. S. Irvine, D. C. Sinclair and A. R. West, Adv. Mater., 1990, 2, 132. References 27 J. A. Chavez-Carvayar, PhD Thesis, Aberdeen, 1995. 28 G. Patrat, M. Brunel and F. DeBergevin, J. Phys. Chem. Solids, 1 B. A. Boukamp, M. P. van Dijk, K. J. de Vries and 1976, 37, 285. A. J. Burggraaf, Adv. Ceram., 1987, 23, 447. 29 F. Galasso and W. Darby, Inorg. Chem., 1965, 4, 71. 2 G. Bronoel, J. C. Grenier and J. Reby, Electrochim. Acta, 1980, 25, 1015. 3 J. C. Grenier, M. Pouchard and P. Hagenmuller, Solid State Paper 8/04865G 2520 J. Mater. Chem., 1998, 8(11), 2515&nd