J. MATER. CHEM., 1994, 4( 5), 653-656 High-temperature Thermoelectric Properties of In,O,-based Mixed Oxides and their Applicability to Thermoelectric Power Generation Michitaka Ohtaki, Daisuke Ogura, Koichi Eguchi and Hiromichi Arai* Department of Materials Science and Technology, Graduate School of Engineering Sciences, Kyushu University, 6-1 Kasuga koen, Kasuga, Fukuoka 876 Japan The thermoelectric properties of mixed oxides In,O,~MO,(MO, =Cr203, Mn203, NiO, ZnO, Y203, Nb205,SnOJ are investigated in terms of the thermoelectric materials at high temperature. The Seebeck coefficients, S, of all the samples have negative values, and those of In,O,.SnO, and In,O,.ZnO increase linearly with temperature, attaining values of -90 and -210 pV K-' at 1000 "C, respectively.The electrical conductivities, B, of these oxides are significantly high. The power factor S2a of the oxides has constantly positive temperature coefficients up to over 1000 "C. Rather low thermal conductivities, IC, of the sintered bodies of the oxides, ca. 1.7 W m-' K-' at room temperature, lead to the largest value of the thermoelectric figure of merit Z=0.4 x lov4K-' for M =Sn at 1000 "C, and the Z value increases toward higher temperatures. The Seebeck effect converts thermal energy directly to electrical energy through the thermoelectromotive force gener- ated by a temperature gradient along a solid material. Thermoelectric power generation via the Seebeck effect was proposed in 1940 and was under intensive research until 1960s. However, an energy conversion efficiency sufficient for practical application was not established, and research on thermoelectric power generation ebbed away except in aero- space technologies.Very recently, however, thermoelectric technology has again attracted interest since efficient energy utilization is becoming more and more important owing to concerns about energy resources and the global environment. The efficiency of the thermoelectric energy conversion depends on both the Carnot efficiency qc of the system and the thermoelectric figure of merit Z=S2a/rc of materials constituting the thermoelectric power generating devices, where S, 0and IC are the Seebeck coefficient, and the electrical and thermal conductivities, respectively.Since qc improves the greater the temperature difference over which the device operates, durability to high-temperature operation is desirable for thermoelectric materials. To date, Si-Ge alloys,' several metal ~halcogenides,2,~transition-metal disilicides,- and some boron corn pound^^*^ have been developed as materials for high-temperature thermoelectric power generation. However, practical utilization has been limited because many of them require complicated surface protection to prevent oxidation or vaporization, and some have essential tempera- ture limits due to phase transitions at high temperature. Metal oxides at their common oxidation state seem to be eminently advantageous for high-temperature operation in air, and many of these oxides have high electrical conductivit- ies.Also, there have been extensive studies on the Seebeck coefficients of oxide materials since the Seebeck coefficient is an important and fundamental parameter for evaluating the electrical transport properties of materials. In the studies on physical and chemical sensors, there have been some attempts to utilize the Seebeck effect of oxide materials to detect temperature changes. For example, it has been reported that a 'Seebeck sensor' consisting of an SnO, pellet with a Pt catalyst deposited on one side could detect several flammable gases with voltage output generated by the combustion heat of the gases: although only the magnitude of the Seebeck coefficient is of interest since high electrical conductivity is unnecessary in such applications.Nonetheless, it is surprising that there have apparently been no reports attempting to apply the oxide materials to thermo- electric power generation. While high-temperature super- conducting cuprates were once of interest in application to thermoelectric refrigerators," they had essentially poor performance''. l2 mainly due to their extremely small carrier mobilities (as small as lo-' cm2 V-' s-'). Metal oxides are generally considered to have rather small carrier mobility, partly because of their highly ionic character. However, some oxides have considerably large mobilities. For instance, the reported value of the Hall mobility of Sn02 single csy~tals'~ is 240 cm2 V-'S-' ,and this is much larger than the value14 of 30 cm2 V-'S-' (this is the same as the drift mobility of copper) for ReO,, which shows completely metallic behaviour and the highest electrical conductivity of all the known oxides in the normal state.Indium oxide (In203)is also well known as a highly conductive oxide, and is widely used as sputtered or vapour-deposited thin films for transparent electrodes as indium tin oxide (ITO) which contains several percent of Sn. The high electrical conductivities of these oxides can be attributed to their rather large carrier mobility.'5 In the present paper, we have investigated the thermoelectric proper- ties of several In203-based mixed oxides in terms of the thermoelectric power generation particularly at high tempera- tures.As far as we are aware, this is the first time that metal oxides have been proposed as candidate materials for high- temperature thermoelectric power generation. Experimental Sample Preparation The mixed oxides In203-M0, (MO, =Cr203,Mn203, NiO, ZnO, Y,03, Nb205,SnO,) were prepared from equimolar (on the molecular formula basis) mixtures of the corresponding single metal oxide powders (>99.9%). The powders were mixed and pulverized in a nylon-lined ball mill for 24 h. The powder mixture was pressed into a pellet and sintered at 1400 "C for 10 h in air. The heating and cooling rate was 200 "C h-l. The crystal phases in the samples thus obtained were determined from a powder X-ray diffraction (XRD) study using Cu-Ka radiation.Measurement of Thermoelectric Properties The samples for the electrical measurements were cut out from the sintered pellets as rectangular bars of ca. 15 mm x 5 mm x 3 mm, and polished with Sic emery papers. The measurements of the electrical conductivity and the Pt-Rh/Pt thermocouple -13% +A//I I I ll R sheet Fig. 1 Schematic side view of an experimental set-up for the simul- taneous measurement of the Seebeck coefficient and the electrical conductivity Seebeck coefficient were simultaneously carried out in air from room temperature to 1000 "C. The experimental set-up for the simultaneous measurement is schematically shown in Fig. 1. The CJ values were measured by the dc four-probe technique by using each Pt leg of the thermocouples as a current lead.The current was supplied with a programmable digital-regulated dc power supply, and the potential differences between two voltage probes were measured on a digital voltmeter. The S values were obtained from the least-squares regressions of the thermoelectromotive force as a function of a temperature difference of <5 K applied by a heater at each temperature of measurement. All the measurements were carried out after attaining the steady-state temperature at each step. The thermal conductivity was determined from the thermal diffusivity and the specific heat capacity obtained by the laser flash measurement on ULVAC TC-7000 for sample disks 10 mm in diameter and 1-2 mm in thickness.The data were calibrated with a standard sample of a sapphire single crystal. Results and Discussion Crystal Phases in In,O,-based Mixed Oxides The XRD study revealed that most of the samples of In,O,-MO, consist of two crystal phases, the In,03 phase and the single oxide phase of the counterpart MO,, except for samples combined with Y203, Nb,O5 and Sn0,. The samples of In2O3.Y2O3 and In203.Nb205 were confirmed to be single phases of InYO, and InNbO,, respectively. The mixed oxide In203.Sn02 consists of the In,07 phase and a fairly large amount of an unknown phase, showing no SnO, phase as seen in Fig. 2. Although some diffraction lines in 40 50 60 28ldegrees Fig. 2 Powder X-ray diffraction patterns of In,O,.SnO, sintered at 1400 "C.The lines denoted as unknown (0)can be assigned to the intermediate compound reported in ref. 18. 0 =In203; A = In,Sn,O, -x J. MATER. CHEM., 1994, VOL. 4 Fig. 2 can be assigned to In2Sn207-,,16 most reports on the system 1n2O3-SnO, have discussed that the equilibrium phase at the composition of In,O,/SnO, =1:1 is a simple mixture of a cubic In,O, phase and a tetragonal SnO, phase. Whereas a rhombohedra1 In4Sn3OI2 phase was reported by Bates et all7 to exist at around the composition of In/Sn= 1:1, such a phase has not been confirmed to date. Recently, however, Enoki et all8 reported that in the In,O,-SnO, binary system an intermediate compound exists above 1300 "C in the composition range 47.9-59.3 mol% of SnO,, although the precise composition and the detailed crystal structure of the compound are not clear yet.As indicated in Fig. 2, the diffraction lines of the unknown phase found in our study were very similar to those of the intermediate compound reported there. However, further study should be required to elucidate the phase composition of the mixed oxide In,O,.SnO, in detail. Electrical Transport Properties The Arrhenius plots of the electrical conductivities of In203-M0, shown in Fig. 3(a) indicate that whereas the sample of neat In,O, exhibits metallic behaviour, all the mixed oxides except In,O,.ZnO are semiconducting and have CJ increasing with temperature. The extremely low conductivit- ies for M=Y and Nb would be due to the new mixed oxide phases which could have electrical properties completely different from In,O, and SnO,. On the other hand, the mixed oxide In,03-Sn0, has the highest electrical conductivity of all the samples.In,O,-ZnO is also highly conductive, showing metallic behaviour similar to In203, and the CT values are slightly smaller than those of In203. The slight decrease in CJ would be explained by the lower electron density in In,O,.ZnO due to substitution of trivalent In ions by divalent Zn ions, in accordance with the slight increase in the negative values of the Seebeck coefficient plotted in Fig. 3(b). However, it should be noted that although In,0,.Sn02 also shows a monotonic increase in negative values of S as a i-6 0 1 2 3 4 103 T-W' A A ' - 80 200 0 400 0600 800 31000 77°C Fig.3 Thermoelectric properties of In,03.MOx. (a) The Arrhenius plots of the electrical conductivities, and (b)the temperature depen- dence of the Seebeck coefficients. MO,: @, SnO,; 0, Cr,03;ZnO; 0,A,Mn203; A,NiO; 0,Nb,O,; a, Y203 J. MATER. CHEM., 1994, VOL. 4 function of temperature up to over 1000 "C similar to those of In,O,.ZnO and In,03, the a values of In,O,.SnO, also increase with temperature and exhibit semiconducting behav- iour. Whereas the temperature dependence of cr for M=Sn appears to have a transition point at ca. 500 "C, the relation- ship between log c and 1/T is not linear even in the tempera- ture region higher than the transition point. This means that the steep increase in a above 500 "C cannot be ascribed to the intrinsic region in which both electrons and positive holes are generated by a thermal energy larger than the bandgap energy.Fig. 4 clearly shows that there is a linear relationship between log aT and 1/T, and that the conduction mechanism of In,O,.SnO, is consistent with the hopping conduction of electrons for which the electrical conductivity can be expressed as CT cc1/T exp(-E,/kBT) where E, is the activation energy of the hopping conduction and kB is the Boltzmann constant. In hopping conduction, the thermally activated carrier mobility can lead to a con- stantly increasing electrical conductivity as seen in Fig. 3(a). The activation energy was determined from Fig. 4 as 0.17 and 0.06 eV in the higher and lower temperature regions from the transition, respectively.The power factor S'CTof the samples for the thermoelectric conversion was calculated from the values in Fig. 3, and is shown in Fig. 5. The mixed oxides with SnO, and ZnO had rather large power factor values which increased almost linearly, similarly as for neat In,O,. However, at 1000 "C the values for M=Sn and Zn obviously exceed that of In,O,, owing to the much steeper increase above 800 "C. Changes in the thermoelectric properties of (In203)1 -x (SnO,), with varying composition, x were examined and the maximum value of the power factor was obtained at ca. x=O.5. An increase in x more of than 0.5 resulted in a marked decrease in a, whereas an increase of <0.5 caused 6.0 7 1 4.0 a 1 2 3 4 lo3 T-IIK-' Fig.4 Relationship between log oT and 1/T for In,O,.SnO,. 1.5 I 0.5 -0 00 A' ' o.OLndLAA''' 0 200 400 600 800 1000 77°C Fig. 5 Temperature dependence of the power factor of In,O,.MO,. MO,: 0,SnO,; Ci,ZnO; 8,Cr203; A, Mn203;A,NiO unfavourably small values of S. While the amount of the unknown phase in (In203)1-x (SnO,), estimated in a prelimi- nary study correlated to some extent with the magnitude of the power factor, the behaviour of these two values was not exactly the same, i.e. the amount of unknown phase was greatest for x=O.6. Thus, the maximum of the power factor would be attributable to a mixture of the unknown phase and the In20, phase which attains a optimal combinaticm of S and CT. Thermal Transport Properties and Thermoelectric Figure of Merit The thermal conductivity of In,O,.MO, observed at room temperature was 1.58-1.75 W rn-' K-' for M=In, Sn, Ni, Mn, and Nb, indicating no marked differences among them.Rather low relative density of the samples, namely 65-70%, is presumably responsible for these low and similar values of IC.The temperature dependence of the thermal conductivity of In,O,.SnO, is shown in Fig. 6. The plots reveal that although IC of In,O,.SnO, increases slightly above 500 "C,the value was as low as 3.1 W m-l K-' even at 750 'C. The temperature dependence of the figure of merit Z=S2a/lc of In,O,.SnO, is also plotted in Fig. 6 in which data at and above 800 "C were calculated from the extrapolated kalues of IC.The oxide exhibited an almost linear increase in 2 and attained a value of 0.4 x K-l at 1000 "C without any levelling off.The electrical conductivity of semiconductors increases steeply above the temperature at which the materials show a transition to intrinsic conduction. The thermoelectric: power accordingly diminishes quickly at temperatures higher than the transition point. The figure of merit of the conventional thermoelectric materials thereby usually has a maximum showing an optimal temperature, and decreases beyond this point. However, the temperature dependence of 2 for In,O,.SnO, shows constantly positive temperature coefficients in the whole temperature range up to 1000 "C, and the value is still increasing toward higher temperatures.These results are apparently due to a monotonic increase both in the electrical conductivity and in the thermoelectric power even at very high temperature. Such a tendency has also been reported for B,-,C, for which the hopping conduction mechanism has been ~onfirmed.~,'~ In,03-Sn02 is not single phase, and thus careful investigation of the conduction mechanism of the oxide is necessary; the hopping nature of the conduction revealed in Fig. 4 is probably responsible for the notable thermoelectric properties particularly at high temperatures. Moreover, the higher limit of the operating temperature will be examined in further study; however, In,O,.SnOz and In,O,.ZnO were substantially stable even at 1400 "C'in air..-c 0z -0E 0.2 -4 .c0 o.o<o 0 200 400 600 800 1000 77°C Fig. 6 Thermal conductivity and the figure of merit of In,C),SnO,. J. MATER. CHEM., 1994, VOL. 4 Conclusions The thermoelectric properties of the In,O,-based mixed oxides were investigated. In particular, the mixed oxides In,03-SnOz and In,O,-ZnO at high temperatures exhibited large values of the power factor for thermoelectric power generation. These results are attributed to both the high electrical conductivities and the constantly increasing Seebeck coefficients with increas- ing temperature up to lo00 "C.The promising thermoelectric properties of In,O,.SnO, are presumably due to the hopping conduction of electrons in the oxide with large carrier mobility.The thermal conductivity of In,O,-SnO, was as low as 1.7 W m-' K-' at room temperature, and 3.1 W m-' K-' even at 750 "C. The mixed oxide consequently attained the figure of merit 2 of 0.4 x K-' as the largest value of the samples at lo00 "C. Although the 2 value of the oxide is at present rather small compared with the state of the art of conventional thermoelectric materials, the metal oxides presented here would be hopeful as materials for high-temperature thermo- electric power generation because of their notable thermoelec- tric properties and their excellent durability at high temperatures in air. The authors thank Mr. Yasuhiro Yamada of the Government Industrial Research Institute, Kyushu, for his kind cooperation on the laser flash measurement of IC.One of the authors (M.O.) is grateful to the Kazuchika Okura Memorial Foundation for financial support of this work. References 1 C. M. Bhandari and D. M. Rowe, Contemp. Phys., 1980,21,219. 2 J. C. Bass and N. B. Elsner, Proc. 3rd Int. Conf. Therm. Energ. Conv.,University of Texas at Arlington, Arlington, 1980, p. 8. 3 J. F. Nakahara, T. Takeshita, M. J. Tschetter, B. J. Beaudry and K. A. Gschneidner Jr., J.Appl. Phys., 1988,63,2331. I. Nishida, Phys. Reu. B, 1973,7,2710. I. Nishida and T. Sakata, J.Phys. Chem. Solid, 1978,39,499. T. Kojima, Phys. Status Solidi (a), 1989,111,233. C. Wood and D. Emin, Phys. Rev. B, 1984,29,4582. S. Yugo, T. Sat0 and T. Kimura, Appl. Phys. Lert., 1985,46,842. J. F. McAleer, P. T. Moseley, P. Bourke, J. 0. W. Norris and R.Stephan, Sensors Actuators B, 1985,8,251. 10 W. J. Macklin and P. T. Moseley, Muter. Sci.Eng., 1990, B7, 11 1. 11 T. 0.Mason, Mater. Sci.Eng., 1991, B10,257. 12 W. J. Macklin and P. T. Moseley, Muter. Sci.Eng., 1991, B10,260. 13 C. G. Fonstad and R. H. Rediker, J. Appl. Phys., 1971,42,2911. 14 T. P. Pearsall and C. A. Lee, Phys. Rev. B, 1974,10,2190. 15 S. J. Wen, C. Couturier, J. P. Chaminade, E. Marquestaut, J. Claverie and P. Hagenmuller, J. Solid State Chem., 1992, 101, 203. 16 Powder Diffraction Files, Inorganic Phases, 39-1058, JCPDS International Centre for Diffraction Data 1991. 17 J. L. Bates, C. W. Gril€in, D. D. Marchant and J. E. Garnier, Am. Ceram. SOC.Bull., 1986,65,673. 18 H. Enoki, J. Echigoya and H. Suto, J. Muter. Sci., 1991,26,4110. 19 G. A. Samara, D. Emin and C. Wood, Phys. Rev. B, 1985,32,2315. Paper 3/07624E; Received 30th December, 1993