THERMOELECTRIC PROPERTIES OF THE HALF-HEUSLER COMPOUND (Zr,Hf)(Ni,Pd)Sn ABSTRACT

Transcription

1 THERMOELECTRIC PROPERTIES OF THE HALF-HEUSLER COMPOUND (Zr,Hf)(Ni,Pd)Sn V. M. Browning a, S. J. Poon b, T. M. Tritt c, A.L Pope c, S. Bhattacharya c, P. Volkov b, J. G. Song b, V. Ponnambalam b, A. C. Ehrlich a a Naval Research Laboratory, Washington, DC 2375 b University of Virginia, Charlottesville, VA 2294 c Clemson University, Clemson, SC ABSTRACT Recent measurements of the thermoelectric transport properties of a series of the half- Heusler compound ZrNiSn are presented. These materials are known to be bandgap intermetallic compounds with relatively large Seebeck coefficients and semimetallic to semiconducting transport properties. This makes them attractive for study as potential candidates for thermoelectric applications. In this study, trends in the thermoelectric power, electrical conductivity and thermal conductivity are examined as a function of chemical substitution on the various fcc sub-lattices that comprise the half-heusler crystal structure. These results suggest that the lattice contribution to the thermal conductivity may be reduced by increasing the phonon scattering via chemical substitution. The effects of these substitutions on the overall power factor and figure-of-merit will also be discussed. INTRODUCTION In recent years there has been a renewed interest in the development of novel thermoelectric materials with improved properties for applications such as thermoelectric cooling and power generation. In general, those materials which look most promising exhibit a large Seebeck effect, S, good electrical conductivity, σ, and poor thermal conductivity, κ. Together, these quantities are used to determine the dimensionless figure-of-merit, ZT = S 2 σt/κ, which provides a measure for performance of a given material in a thermoelectric device application. Due to their relatively large Seebeck coefficients, narrow band semiconductors and semi-metals are one of the most widely studied classes of materials for thermoelectrics. The ability to chemically dope these materials means that it is often possible to achieve reduced κ values in polycrystalline samples while preserving the overall power-factor, S 2 σ, thereby increasing ZT. One class of materials which is being investigated for use in thermoelectric power generation applications at intermediate temperatures are the half-heusler compounds (MNiSn, M=Zr,Hf,Ti). These intermetallic compounds are known to exhibit Seebeck coefficients on the order of 1-3 µv/k at room temperature and exhibit semi-conducting electrical transport [ 1,2,3 ]. High temperature resistivity measurements suggest the presence of a gap in the density of states at the Fermi surface on the order of.1 to.2 ev [ 2 ]. Evidence suggesting that the thermopower continues to increase above room temperature in these materials makes them potentially attractive for use in thermoelectric applications above room temperature. However, the thermal conductivity of these materials has been found to be prohibitively large in wellordered samples. Therefore, recent studies of this system for thermoelectric applications have focussed on reducing the thermal conductivity by introducing additional phonon scattering mechanisms.

2 The structure of the half-heuslers is that of MgAgAs [ 4,5,6 ] which consists of three filled and one vacant interpenetrating fcc sublattices. The vacancy sub-lattice separates the half- Heusler compounds from the Heusler compounds, MNi 2 Sn, which contain an additional Ni fcc sublattice. The ability to substitute isoelectronic atoms with differing masses on the various sublattices has been suggested as a possible approach to reducing the lattice contribution to the thermal conductivity [ 7 ]. In particular, C. Uher et al. [ 8,9 ] have reported on the effects of substitution on the Zr sublattice in ZrNiSn and have demonstrated significant reductions in κ at low temperatures. Unfortunately, the effects on the room temperature values of κ have been minimal. Uher et al. s studies have also demonstrated an unusual dependence of the transport properties on annealing conditions. Their results showed that high temperature anneals resulted in a substantial increase in thermal conductivity accompanied by a corresponding decrease in the electrical conductivity. The effects of annealing on the electrical transport are in agreement with earlier studies which showed that increasing disorder resulted in an overall decrease in resistivity and an onset of a metal-insulator transition at low temperature [2]. While the increase in thermal conductivity can be understood in terms of increased ordering of the crystal structure, the decrease in electrical conductivity is not so well understood. These results underscore the need for further study of the effects of processing conditions on the thermoelectric transport properties of these materials. In the present study, the effects of differential mass scattering are further explored via chemical substitutions on both the Zr and Ni sub-lattices in ZrNiSn.. Early measurements of samples made under non-optimum processing conditions exhibited thermal conductivity values as low as 25-3 mw/cm-k [ 1 ]. Unfortunately, these early results have not been reproduced in samples made under improved conditions. However, the weight loss in the ingots from which both sets of samples were obtained was found to be less than 1% of the nominal weight. That is, the different results seen are not attributable to compositional variation. In an effort to determine whether the thermal conductivity can systematically be reduced to the values observed in the early samples, it is necessary to understand the effects of chemical disorder in the observed transport properties. Due to their similar size, the Zr and Sn atoms are known to substitute for each other on their respective lattice sites [ 11 ]. The substitution of more massive, isoelectronic atoms in this system may, therefore, affect the transport properties both by modulating the degree of Zr/Sn site substitution and by directly affecting the band structure and the electron and phonon scattering mechanisms. The results reported here look specifically at the effects on the thermoelectric transport properties of simultaneous substitutions of Hf and Pd for Zr and Ni, respectively. In addition, the effects of varying Sn composition are studied as well as the effects of the addition of boron which, due to its small size, is expected to substitute interstitially. Trends in the thermal conductivity, resistivity, and thermopower values are examined as a function of chemical composition. These results suggest that this system has not yet been optimized in terms of its potential for thermoelectric applications. Preliminary results on a similar study of chemically disordered TiNiSn samples will also be presented. EXPERIMENT Sample Preparation Ingots of nominal composition were prepared by melting together elements (>99.9% purity) in an arc furnace. Phase purity was assessed via x-ray diffraction studies performed on a

3 Scintag X-ray diffractometer using Cu K α_ radiation. The observed peaks in the diffraction spectra were indexed successfully to the MgAgAs structure. Of concern is whether the samples 25 2 (22) Intensity (a.u.) (111) (2) (311) (222) (4) (331) (42) two theta (deg.) (422) (333 or 511) (44) (531) (442 or 6) Figure 1: X-ray diffraction scan of Ti.5 Hf.5 Ni.85 Pt.15 Sn.95 alloy indexed to the MgAgAs structure. remain single phase with chemical substitution. It was found that in all samples, the observed peaks were consistent with a single phase material. A sample scan demonstrating that phase purity is maintained even in chemically substituted samples is shown for Ti.5 Hf.5 Ni.85 Pt.15 Sn is shown in Figure 1. These results are consistent with those reported for pure ZrNiSn [11] and TiNiSn [3]. In addition to x-ray diffraction measurements, thermal analysis studies using a Perkin Elmer DTA7 indicated thermal stability of the compounds up to 125 C. Thermal stability at high temperatures is essential in these materials if they are going to be of use in thermoelectric devices at moderate to high temperatures. Although the thermal analysis results suggest that these materials retain structural stability up to temperatures > 12 C, annealing studies suggest that the transport properties may drastically change at temperatures above 8 C, thus restricting their use to temperatures below this value. Experimental Procedures Characterization of the thermoelectric transport properties of the samples was carried out in collaboration between facilities at the Naval Research Laboratory and Clemson University. Measurements taken at the Naval Research Laboratory were accomplished using a transient

4 voltage ZT-meter technique. In this technique, the Peltier effect is used to generate a temperature gradient across the sample. The resulting Seebeck component of the sample voltage is monitored as a function of the Peltier induced T, and the Seebeck coefficient, S, is determined by the linear slope of the S( T) curve. This experimental configuration also allows measurement of the thermal conductivity which is determined by the relation: κ = (S*T*I/ T)*C, where S is the thermopower, T is the sample temperature, I is the applied current and T is the resulting Peltier induced temperature gradient. The quantity C is a correction factor which accounts for losses due to radiation and thermal conductance of the leads. Complete details of this technique can be found elsewhere [ 12 ]. Thermopower measurements taken at Clemson University were accomplished by applying a temperature gradient on the order of 1% of the sample temperature via a small chip heater attached to the sample and monitoring the induced Seebeck voltage. Thermal conductivity measurements were taken using an absolute technique in which heat is introduced to the sample via a resistive heater. The temperature gradient of the sample is monitored as a function of input power to the heater to determine κ. Both the NRL and Clemson resistivity measurements were taken using a standard 4-probe technique. RESULTS AND DISCUSSION In order to determine the feasability of differential mass scattering as a means of optimizing the thermoelectric performance of the ZrNiSn compound, the various thermoelectric transport properties of a series of samples with varying compositions were measured. Specifically, this study focussed on the following compositions: ZrNiSn, ZrNi.97 Sn, Zr.7 Hf.3 Ni.7 Pd.3 Sn, Zr.8 Hf.2 Ni.7 Pd.3 Sn.97, Zr.5 Hf.5 Ni.7 Pd.3 Sn.95, and ZrNiSn with 1% boron. Measurements reported for the ZrNiSn and ZrNiSn with 1% boron were taken on ascast samples. The remaining samples were subjected to a 12 hour anneal at 9 C prior to measurement. Figure 2 shows the temperature dependent resistivities measured for these samples. In general, with increasing chemical disorder the resistivity decreases in magnitude. In addition, the overall temperature dependence of the resistivity varies widely from apparently activated behavior in the pure ZrNiSn sample to a crossover between non-metallic to metallic behavior in a boron substituted sample. In the sense that increased disorder enhances the electrical conductivity, our results are in agreement with previous annealing and substitutional studies. The general form for the resistivity in systems exhibiting activated behavior is: ρ = ρ (T)exp( ρ /k B T). In the simplest case, where carriers are thermally excited over an energy gap, ρ, at the Fermi surface, ρ (T) is taken as a constant, and ρ is determined from the slope of the ln ρ versus 1/T curve. In disordered systems, localization effects become significant and ρ(t) exhibits a power law behavior which is dependent on the mechanism of localization. Disorder can localize carriers either through Anderson localization or, in the presence of strong electronphonon coupling, through the formation of polarization induced lattice distortions known as polarons. Figure 3 shows the resistivity plotted as lnρ versus 1/T. Clearly over this temperature range, the data cannot be explained by simple thermal activation over an energy gap, suggesting that localization of carriers is a contributing factor in the observed transport properties. Evidence for a metal-insulator transition in some of the samples further supports the presence of a localization potential in this system.

5 Although the trends in the resistivity data with increasing disorder are not completely understood, earlier theoretical work provides some insight to two possible mechanisms which may be responsible for the observed trends in the transport properties. Band structure 1.1 ρ (Ω-cm) Figure 2: Resistivity versus temperature for a series of chemically substituted ZrNiSn samples: ZrNiSn ( ), ZrNi.97 Sn ( ), Zr.7 Hf.3 Ni.7 Pd.3Sn ( ), Zr.8 Hf.2 Ni.7 Pd.3 Sn.97 ( ), Zr.5 Hf.5 Ni.7 Pd.3 Sn.95 ( ), and ZrNiSn with 1% boron ( ) ln ρ /temperature (K -1 )

6 Figure 3: Temperature dependent resistivity plotted as ln (ρ) versus 1/T for ZrNiSn( ) and Zr.5 Hf.5 Ni.7 Pd.3 Sn.95 ( ). calculations of these materials are in agreement with experiment in that they predict semiconducting transport properties for well ordered samples. Using the pseudopotential totalenergy method, S. Ogut and K.Rabe [ 13 ] have predicted the existence of an indirect semiconducting gap near the Fermi level of approximately.5 ev. The existence of this gap is attributed to a hybridization of the Sn p and Zr d -levels which is mediated by the Ni sub-lattice. The magnitude of the calculated gap is several times larger than that observed in ZrNiSn at high temperatures. However, they are able to account for this discrepancy by allowing for inter-site substitution between the Zr and Sn sub-lattices in their calculations. It is interesting to note that by allowing for this inter-site substitution, their calculations also predict that with approximately 15% site substitution, the gap closes, thereby forming a semi-metal. This result is in quite good agreement with experiments showing the existence of a metal-insulator transition in ZrNiSn samples with inter-site substitutions on the order of 2% as evidenced by x-ray diffraction studies [ 11 ]. It has also been suggested that the vacant Ni sub-lattice may serve as a localization potential for charge carriers [ 2 ]. In this scenario, it is argued that a narrow band, associated with the ordered Ni vacancy sublattice, exists near the Fermi level and beneath the conduction band. It is postulated that this vacancy band localizes charge carriers within the gap. Positron annihilation experiments in NiMnSb which reveal the presence of localization centers at vacancy sites [ 14 ] are cited as experimental evidence for this model. Presumably, any increase in disorder would have a deleterious effect on the ability of the fcc vacancy sub-lattice to serve as a localization potential for charge carriers. The trends in the resistivity may be explained within the framework of either model and it is likely that both mechanisms are, to some degree, contributing to the experimental observations. The electrical resistivity is sensitive to both carrier concentration and mobility and, therefore, it is difficult to determine the degree to which carriers are localized based on electrical transport measurements alone. In principal, thermopower measurements may help to determine whether carrier concentration or carrier mobility is dominating the electronic transport. In a thermally activated system, the thermopower is expected to exhibit a temperature dependence of the form: S(T) = k B /e[( S /k B T) + B]. Since thermopower is predominantly sensitive to carrier concentration while resistivity is equally sensitive to carrier concentration and mobility, one can distinguish localized hopping transport from thermal activation of carriers over an energy gap by comparing the relative magnitude of S and ρ. If ρ >> s, then carrier mobility is dominating the transport properties. Figure 4 plots the temperature dependent thermopower for the various ZrNiSn compounds. None of the samples exhibited a strictly 1/T temperature dependence over the entire range suggesting, in agreement with the resistivity measurements, that the observed transport properties cannot be explained by assuming either strictly thermal activation over an energy gap or localization of charge carriers as the dominant mechanism. While the significant decrease in the thermopower with disorder suggests an increase in the states available for electric transport, the temperature dependence does not support thermal activation of carriers over the measured temperature range. Clearly further study of the transport properties is needed to fully understand this system. In order to determine whether this system is likely to prove useful for thermoelectric applications, one also needs to know the thermal conductivity. The effect of chemical disorder on the thermal conductivity is shown in Figure 5. The multinary compounds

7 Zr.8 Hf.2 Ni.7 Pd.3 Sn.97 and Zr.7 Hf.3 Ni.7 Pd.3 Sn exhibited a significant reduction in the thermal conductivity values compared to the pure ZrNiSn sample. Interestingly, the addition of 1% boron did not significantly change the room temperature value of the thermal conductivity. One possible explanation for this result is that any decrease in lattice thermal conductivity may be Seebeck coefficient (µv/k) Figure 4: Thermopower versus temperature for chemically substituted ZrNiSn: ZrNiSn ( ), ZrNi.97 Sn ( ), Zr.7 Hf.3 Ni.7 Pd.3 Sn ( ), Zr.8 Hf.2 Ni.7 Pd.3 Sn.97 ( ), Zr.5 Hf.5 Ni.7 Pd.3 Sn.95 ( ), and ZrNiSn with 1% boron ( ) κ (mw/cm-k)

8 Figure 5: Thermal conductivity versus temperature for chemically substituted ZrNiSn: ZrNiSn ( ),Zr.7 Hf.3 Ni.7 Pd.3 Sn ( ), Zr.8 Hf.2 Ni.7 Pd.3 Sn.97 ( ), Zr.5 Hf.5 Ni.7 Pd.3 Sn.95 ( ), and ZrNiSn with 1% boron ( ). offset by an increase in the electronic contribution as evidenced by the order of magnitude decrease in resistivity shown in Figure 2. Figure 6 shows the overall net effect of the chemical substitution on the power factor, S 2 /ρ. These results show that all of the substitutions lead to a net decrease in the power factor. When one takes into account the thermal conductivities, one sees that the figure-of-merit, ZT, is also decreased by these substitutions as shown in Figure 7. The small values of ZT at room temperature together with the trends in the data are not encouraging in terms of prospective thermoelectric applications. With this in mind, recent studies have focussed on another of the half-heusler compounds TiNiSn. Figure 8 shows the power factor for this system for several compositions. In contrast to the ZrNiSn study, these results show a net increase in power factor with chemical complexity. Further clarification of whether this system may prove useful for thermoelectric applications awaits further study including measurement of the thermal conductivity properties. CONCLUSION In summary, we have studied the thermoelectric properties of a series of ZrNiSn samples with various chemical substitutions in an effort to assess the feasability of optimizing these properties for potential thermoelectric applications. Although, the properties varied greatly with composition, no net increase in the figure-of-merit, ZT, was achieved. However, recent results obtained by Uher et al. [8] suggest that substition with indium may enhance ZT. In addition preliminary measurements of chemically substituted TiNiSn suggest that this system may hold promise for enhanced ZT values in chemically disordered materials.

9 1 1-5 power factor (W/K 2 -cm) Figure 6: Temperature dependence of the power factor for chemically substituted ZrNiSn: ZrNiSn ( ), ZrNi.97 Sn ( ), Zr.7 Hf.3 Ni.7 Pd.3 Sn ( ), Zr.8 Hf.2 Ni.7 Pd.3 Sn.97 ( ), Zr.5 Hf.5 Ni.7 Pd.3 Sn.95 ( ), and ZrNiSn with 1% boron ( ) ZT Figure 7: Figure-of-merit, ZT, versus temperature for chemically substituted ZrNiSn: ZrNiSn ( ), ZrNiSn w/1% boron ( ), Zr.8 Hf.2 Ni.7 Pd.2 Sn.97 ( ), Zr.7 Hf.3 Ni.7 Pd.3 Sn ( ).

10 2 1-5 power factor (W/K 2 -cm) Figure 8: Power factor as a function of temperature for chemically substituted TiNiSn: TiNiSn ( ), TiNi 1.3 Sn ( ), Ti.5 Hf.5 Ni.85 Pt.15 Sn.95 ( ).

11 ACKNOWLEDGEMENTS V. M. Browning gratefully acknowledges the financial support of both DARPA and ONR. S. J. Poon acknowledges support from NSF #DMR In addition, T. M. Tritt acknowledges financial support from DARPA/ARO #DAA G , ONR #N and ONR #N One of us (A.L. Pope) acknowledges Clemson University for partial finanical support through the Dean's Scholars Award Program. REFERENCES 1. F. G. Aliev, N. B. Brandt, V. V. Moshchalkov, V. V. Kozyrkov,R. V. Skolozdra, and A. I Belogorokhov, Z. Phys. B 75, 167 (1989). 2. F. G. Aliev, Physica B 171, 199 (1991). 3. B. A.Cook, J. L. Harringa, Z. S. Tan, and W. A. Jesser in Proceedings XV International Conference on Thermoelectrics, (IEEE Cat. No. 96TH8169, New York, 1996), p W. Jeitschko, Metall. Trans. 1, 3159 (197). 5. T. T. M. Palstra, G. J. Nieuwenhus, R. M. F. Vlastuin, J. A.Mydosh, J. van den Berg and K. H. J. Bushow, J. Magn. & Magn. Mater. 67, 331 (1987). 6. F. G. Aliev, N. B. Brandt, V. V. Kozyrkov, V. V. Moshchalkov, R. V. Scolozdra and Yu. V. Stadnik, Fiz. Nizk. Temp. 12, 498 (1987). 7. P. G. Klemens in Proc. Phys. Soc. (London) A 68, 1113 (1955). 8. C. Uher, J. Yang, S. Hu, D. T. Morelli, and G. P. Meisner, (submitted to Phys. Rev. B). 9. C. Uher, S. Hu, J. Yang, G. P. Meisner, and D. T. Morelli, in Proc. XVI Int. Conf. On Thermoelectrics (in press). 1. V.M. Browning (unpublished results). 11. F. G. Aliev, N. B. Brandt, V. V. Kozyrkov, V. V. Moshchalkov. R. V. Scolozdra, Yu. V. Stadnyk, and V. V. Pecharskii, Pis ma v Zh. Eksp. Teor. Fiz. 45, 535 (1987). 12. R. J. Buist in CRC Handbook of Thermoelectrics, edited by D. M. Rowe (CRC Press, Boca Raton, 1995) p S. Ögut and K. M. Rabe, Phys. Rev. B 51, 1443 (1995). 14. K. E. H. M. Hannssen and P. E. Mijnarends, Phys. Rev. B 34, 59 (1986).