- Email: [email protected]
Microstructure and thermoelectric properties of WSi2-added CrSi2 composite
Journal of Alloys and Compounds 690 (2017) 652e657
Contents lists available at ScienceDirect
Journal of Alloys and Compounds journal homepage: http://www.elsevier.com/locate/jalcom
Microstructure and thermoelectric properties of WSi2-added CrSi2 composite Masashi Mikami*, Yoshiaki Kinemuchi National Institute of Advanced Industrial Science and Technology, Nagoya 463-8560, Japan
a r t i c l e i n f o
a b s t r a c t
Article history: Received 3 February 2016 Received in revised form 3 August 2016 Accepted 18 August 2016 Available online 20 August 2016
CrSi2-based composite materials with submicrometer-sized grains were fabricated by the powder metallurgical process using ball-milling and pulse-current sintering. By the solid-state reaction of submicrometer-sized fine powders of CrSi2 and WSi2 during the sintering, the composites consisting of Cr-rich hexagonal (Cr,W)Si2 phase and W-rich tetragonal (W,Cr)Si2 phase were obtained. The grain size of CrSi2 phase was effectively reduced by the WSi2 addition because of the suppression of grain growth during sintering. Thermal conductivity was then decreased with the WSi2 content by the enhancement of phonon scattering resulting from the reduction of grain size, the introduction of the CrSi2/WSi2 interface and the crystal lattice distortion induced by the W substitution for CrSi2 phase. Seebeck coefficient and electrical resistivity were also decreased due to the addition of metallic WSi2 phase and the deformation of CrSi2 electronic structure resulting from the W substitution. The WSi2 addition had then little effect on the thermoelectric power factor. Consequently, the thermoelectric figure-of-merit, ZT, was enhanced by the reduction of thermal conductivity and reached 0.3 around 700 K. © 2016 Elsevier B.V. All rights reserved.
Keywords: Thermoelectric materials Silicide Composite materials Powder metallurgy
1. Introduction Thermoelectric devices have recently attracted renewed interest in their potential application to clean energy-conversion systems. In particular, energy recovery from vast amounts of waste heat are targeted for thermoelectric power generation because the temperature and the energy scale of most heat sources are too low to reuse efficiently with other conventional energy conversion systems. The conversion efficiency of thermoelectric devices depends mainly on the thermoelectric performance of the material, which is evaluated using the thermoelectric figure of merit, ZT ¼ (S2/rk)T, where S is the Seebeck coefficient, r is the electrical resistivity, k is the thermal conductivity, and T is the absolute temperature. The telluride materials, such as Bi-Te or Pb-Te system, exhibit relatively high ZT values and are suited for thermoelectric power generation devices. However, it seems to have run into difficulty for widespread use because of its higher cost and the limited supply of raw materials, especially Te. Transition metal silicides have been studied intensively for thermoelectric application because of its abundance of raw
* Corresponding author. E-mail address: [email protected] (M. Mikami). http://dx.doi.org/10.1016/j.jallcom.2016.08.185 0925-8388/© 2016 Elsevier B.V. All rights reserved.
material, low toxicity and high chemical stability. Chromium disilicide, CrSi2, is one of the promising silicide because of its large S and low r, resulting in large power factor, PF (¼S2/r), over 1 mW/ mK2 above room-temperature [1e4]. In addition, the high oxidation resistance in air up to 900 K [5] is a favorable property for the long-term stability as a thermoelectric power generation device. However, the relatively high intrinsic k value of CrSi2 of over 12 W/ mK [4] limits its ZT value to ~0.2. The reduction of k is therefore necessary for efficient thermoelectric power generation. Since k in the CrSi2 system is mainly governed by the lattice part of k, the enhancement of phonon scattering should be effective to reduce total k value. Indeed, k value was effectively decreased by the structural refinement, such as the reduction of grain size [3,4]. In this study, CrSi2-based composites were fabricated by the addition of WSi2. For the selection of additive material, the low r and crystallographic compatibility are desirable in order to avoid the degradation of electrical conductivity by the formation of composite. In addition, the high chemical stability in CrSi2 matrix is also required. From these aspect, metallic or semimetallic silicide materials seem to be promising candidate as the additive material for CrSi2. However, since the 3d transition silicides or the homologous molybdenum silicide easily form a solid solution with CrSi2 [2,6,7], it is difficult to obtain the stable composite structure. Therefore, WSi2 having a limited solid
M. Mikami, Y. Kinemuchi / Journal of Alloys and Compounds 690 (2017) 652e657
2. Experimental CrSi2-based composite materials with the nominal composition of (CrSi2)1x(WSi2)x (x ¼ 0.02e0.10) were fabricated by the powder metallurgical process. Commercially available CrSi2 (99%, Rare Metallic Co., Ltd.) and WSi2 (99%, Rare Metallic Co., Ltd) powders were used as the raw material. Although the purity of these raw materials was rather low, almost all impurity materials are oxides, which are usually stable in silicide. Therefore, it can be expected that the effect of impurity phase on thermoelectric properties of samples is negligible. Appropriate amounts of CrSi2 and WSi2 powders were mixed and milled in a planetary ball mill (FRITSCH GmbH, PULVERISETTE 5/4) with a 500-ml-capacity Cr steel pot and carbide balls. In the milling system, the pot rotates on its axis against the direction of orbital motion. The pot was back-filled with a purified Ar gas atmosphere after creating a vacuum. A total mass of 800 g of 5-mm-diameter WC balls was inserted with 30 g of mixed powder into the pot. The ball-milling was performed at 250 rpm for 10 h with a 10 min rest every half-hour. After the ballmilling process, well-mixed and finely-pulverized powders with submicrometer-size were obtained. The prepared powder was sintered using pulse-current sintering (PCS). The powder was put into a graphite mold and sintered at 1373 K for 10 min in vacuum under a uniaxial pressure of 50 MPa. The heating and cooling rates were 100 K/min. For comparison, CrSi2-only sintered samples were also prepared from the raw CrSi2 powder with an average particle size of several mm and from the as-milled CrSi2 powder synthesized by the same ball-milling process. The obtained bulk samples, with a typical size of 10 mm in diameter and 2 mm in thickness, were used to measure k and then cut into a bar shape with a typical size of 2 2 9 mm3 to measure r and S. Crystalline phase analysis was performed using X-ray diffraction (XRD) with Cu Ka radiation. Microstructural observation and compositional analysis was conducted using the scanning electron microscopy (SEM) and the attached energy dispersive X-ray spectrometry (EDX). S and r were simultaneously measured in a He atmosphere. r was evaluated by a conventional four-probe DC technique. S was calculated from a plot of the thermoelectric voltage versus the temperature difference. k was evaluated from the density (D), thermal diffusivity (a), and heat capacity (Cp) with the relationship k ¼ D a Cp. D was measured using the Archimedes method, and a and Cp were evaluated using the laser flash method. Density functional theory was performed based on the linearized augmented plane wave method using Wien2k code [8]. For the pure CrSi2, experimental structure was adopted, while the structural optimization was carried out for W doped CrSi2. Here, doping was performed to the super cell structure of 1 1 2, and chemical composition was fixed to the Cr2WSi6 in which doping site was selected to the Cr (1/2, 1/2, 1/3) in the super cell. The number of kpoint in the Brillouin zone of 2560, the size of basis set (rkmax) of 8.0, and the exchange-correlation functional of generalized
gradient approximation were adopted. The calculation selfconverged within the energy of 0.1 mRy. 3. Results and discussion The XRD measurement was performed on the cross-section of sintered (CrSi2)1x(WSi2)x composites to investigate the relation between the composition x and the crystal structure of the obtained samples. As shown in Fig. 1, the stoichiometric CrSi2-only sintered sample made from the as-milled powder formed a C40 hexagonal structure with space group of P6222 as reported in the literature. Although no trace of an impurity phase, such as oxides, was detected, weak diffraction perks from chromium monosilicide CrSi was observed. The Rietveld analysis had revealed that the presence of a secondary phase of the CrSi phase at 2.7 at%. In addition, the CrSi phase was also observed in every composite samples around 3 at%. However, it is reasonable to assume that the secondary phase could not disturb the investigation of WSi2 addition effect on CrSi2 phase because the CrSi phase was present in every samples at an almost constant proportion. For the composite samples, diffraction peaks from CrSi2 phase tend to monotonically shift toward the lower 2q angle with the amount of WSi2 content. The lattice parameter of CrSi2 phase calculated from the 2q angle of diffraction peaks then increased with the amount of the WSi2 addition, as shown in Fig. 2. This tendency reflects that the crystallographic structure of CrSi2 phase was enlarged by the partial substitution of W, which has a larger atomic radius than Cr. In addition, diffraction peaks from the tetragonal WSi2 phase were detected in x 0.04, as shown in Fig. 1. Moreover, the peak position was slightly higher than that of the WSi2 powder, which was used as a raw material. Although it is difficult to precisely calculate lattice parameter from these weak diffraction peaks, this result implies that the W site in the WSi2 phase was also partially substituted by Cr having the smaller atomic radius, resulting in the shrinkage of
WSi2 CrSi
x=0.10
x=0.08
Intensity (arb. units)
solubility for CrSi2 was chosen as an additive material in this study. The composites consisting of submicrometer-sized fine grains were prepared by the powder metallurgical process using ballmilling and pulse-current sintering in order to enhance the phonon scattering at grain boundaries. In addition, the heavyelement W doping effect on the CrSi2 crystal structure, which is also effective for the reduction of k by the mass-difference scattering in the crystal lattice, was investigated. It was found that the appreciable reduction of k by the WSi2 addition without a serious deterioration of electrical aspect of thermoelectric property improved the ZT value of the CrSi2-based material.
653
x=0.06
x=0.04
x=0.02
CrSi2
20
40
60
80
100
2θ (deg) Fig. 1. XRD patterns (Cu Ka radiation) diffracted from the CrSi2 sintered sample made from the milled powder and sintered composites with the nominal composition of (CrSi2)1x(WSi2)x.
654
M. Mikami, Y. Kinemuchi / Journal of Alloys and Compounds 690 (2017) 652e657
Fig. 2. Relation between the nominal WSi2 content and lattice parameter calculated from 2q angle of peaks in XRD patterns of the sintered (CrSi2)1x(WSi2)x composite.
crystal structure of WSi2 phase. Since these peak shifts were not observed in the mixed powders after the ball-milling process, the element substitutions should be induced by the solid-state reaction of CrSi2 and WSi2 during the sintering. The SEM observation was performed on the as-milled powders and the cross-section of sintered samples. A typical SEM image of the as-milled powder was shown in Fig. 3a). The powders were consisted of submicrometer-sized particles and the dependence of particle size on the WSi2 content was not observed. Fig. 3b) and c) show the microstructure of sintered samples made from the raw CrSi2 powder consisting of several mm particles and from the asmilled CrSi2 powder. Because of the difference of particle size of the powders, grain size in sintered sample was reduced from several mm to around 1 mm. In addition, the nanometer-sized dot structure, which appears dark spots on SEM image, was formed inside the grains of sintered sample made from the as-milled CrSi2 powder. Although formation mechanism and crystallographic phase were unclear, it is presumed that the nano-dot structure should be attributed to compositional variation by the pulverization and/or unavoidable slight oxidation during the milling process and the handling of powders in air. The nano-dot structure was then observed in every sample made from milled powders, as shown in Fig. 3e)-i). One should be noticed that grain size of sintered sample was effectively decreased by the WSi2 addition. For instance, grain size was reduced from around 1 mm to 300 nm by the small amount of 2 at% addition of WSi2, as shown in Fig 3e). Although the clear segregation of WSi2 at grain boundary of CrSi2 was not observed, the WSi2 with the high melting point of 2440 K is assumed to suppress the grain growth of CrSi2 during sintering because the grain size is almost the same as the particle size of the milled powder. As shown in Fig. 3d), the sparsely-distributed bright spots were observed in a low magnification SEM image of the sintered sample with the nominal composition of (CrSi2)0.98(WSi2)0.02. EDX mapping analysis suggested that these bright spots were attributed to the WSi2 phase grains with micrometer size, although it was difficult to determine the composition precisely. Therefore, this sample also formed a CrSi2-WSi2 composite, whereas clear diffraction peaks from WSi2 phase was not detected on the XRD pattern. The number density of the micrometer-sized WSi2 grains then increased with the WSi2 content. In addition, nanometersized WSi2 grains were also observed in the sintered (CrSi2)1x(WSi2)x samples with x 0.04, as shown in Fig. 3f)-i). The number density of the nanometer-sized WSi2 grains increased
Fig. 3. SEM images of a) the as-milled powder with the nominal composition of (CrSi2)0.9(WSi2)0.1 and the fracture cross-sections of the sintered sample made from b) the raw powder of CrSi2, c) the milled powder of CrSi2, d-i) (CrSi2)1x(WSi2)x for 0.04 x 0.10. j) Chemical compositions of (CrSi2)1x(WSi2)x composites measured by EDX. Each dashed line represents the nominal composition.
with the WSi2 content while the size is nearly independent of the WSi2 content. The chemical compositions in sintered composites were almost identical to those of the nominal compositions, as shown in Fig. 3j). Furthermore, the absence of contaminant from ball-milling media, such as Fe, was confirmed by EXD analysis. Therefore, compositional variation and contamination during ball-milling process was sufficiently suppressed and negligibly low. These results of XRD, SEM and EDX analysis indicate that the compositionally controlled composites consisting of the Cr-rich hexagonal (Cr,W)Si2 phase and the W-rich tetragonal (W,Cr)Si2 phase were obtained. In order to clarify the effect of the WSi2 addition on the thermoelectric performance of CrSi2 compound, thermoelectric
M. Mikami, Y. Kinemuchi / Journal of Alloys and Compounds 690 (2017) 652e657
16
DOS / state eV-1 UC-1
Electrical Resistivity (μΩm)
15
10
CrSi2 CrSi2 x=0.02 W0.02 0.04 W0.04 W0.06 0.06 W0.08 0.08 W0.10 0.10
5
500
700
900
a) CrSi2 (C40)
14
total Cr Si
12 10 8
6 4 2 0
0 300
655
-4
1100
2
4
DOS / state eV-1 UC-1
16
b) Cr2WSi6 (C40)
14
total Cr Si W
12 10 8
6 4 2 0 -4
-2
0
2
4
E / eV Fig. 5. The calculated density of state for a) pure CrSi2 and b) W-substituted CrSi2 phase.
200
Seebeck coeficient (μV/K)
properties were evaluated. Fig. 4 shows r of the sintered composites incorporating different amounts of WSi2 as a function of temperature. The every r-T curve has a broad maximum around 700 K. This temperature dependence of r is essentially identical to the previously reported electrical conduction property of CrSi2 phase having the degenerate semiconducting nature [6]. In the lower temperature exhaustion region, r increases with temperature due to the decrease in carrier mobility induced by the acoustic phonon scattering. In the higher temperature intrinsic region, r decreases with temperature because of the increase in carrier concentration resulting from electron carriers thermally excited across the bandgap. On the other hand, the value of r monotonically decreases with the WSi2 content. Since the number of valence electrons of W is the same as that of Cr, carrier doping effect by the element substitution on the CrSi2 phase could not be expected. Instead, the addition of metallic WSi2 phase having the low r of 0.1e0.8 mUm [9] should play the main role in the reduction of r in the composites. For the further consideration, in order to assess the W substitution effect on electronic properties of CrSi2 phase, the electronic structure calculation was conducted. As shown in Fig. 5, the calculated density of state (DOS) suggested that the band gap of CrSi2 phase decreases by the W substitution mainly due to the dispersion of electronic state especially around valence band-edge related to Cr. Therefore, this narrowing of band-gap also caused the reduction of r in the composites. Fig. 6 represents S of the sintered composites as a function of temperature. Every S-T curve also exhibits a broad maximum around 650 K. This temperature dependence of S can also be attributed to the degenerate semiconducting nature of CrSi2 phase [6]. In the lower temperature exhaustion region, the linear increase in S with temperature can be roughly explained according to the Mott's theory [10], which suggests that the value of S is proportional to temperature. In the higher temperature intrinsic region, the electron carriers thermally excited across the bandgap counteract the Seebeck effect produced by holes, resulting in the decrease in S with increasing temperature. In addition, the value of S monotonically decreases with the WSi2 content especially in the lower temperatures. This decrease in S should be caused by the addition of metallic WSi2 phase having the low S value of less than 10 mV/K below 600 K [11]. In the higher temperatures, since the S value in WSi2 phase increases rapidly with increase in temperature from 20 mV/K to 110 mV/K in the temperature range from 700 K to 1100 K [11], the S reduction effect by the WSi2 addition decreases with rising temperature, as shown in Fig. 6.
0
E / eV
Temperature (K) Fig. 4. Temperature dependence of electrical resistivity in the sintered (CrSi2)1x(WSi2)x composites.
-2
150
100 CrSi2 CrSi2 W0.02 x=0.02 W0.04 0.04 W0.06 0.06 W0.08 0.08 W0.10 0.10
50
0
300
500
700
900
1100
Temperature (K) Fig. 6. Temperature dependence of Seebeck coefficient in the sintered (CrSi2)1x(WSi2)x composites.
For the consideration of W substitution effect on S of CrSi2 phase, the electronic structure calculation offers some insight into possible behaviors of S. Fig. 5 represents the deformation of DOS by the W substitution especially in the electronic state related to Cr. In comparison with the pure CrSi2 phase, the narrowing of band gap resulting from the increase in DOS around conduction band edge and the lowering the slope of DOS around valence band edge can be observed in DOS of W-substituted CrSi2 phase. The narrowing band
M. Mikami, Y. Kinemuchi / Journal of Alloys and Compounds 690 (2017) 652e657
gap intensify the bipolar contribution of electron carriers excited thermally across the bandgap, resulting in the decrease in positive S of holes. The lowering the slope of DOS around valence band edge also reduce S according to Mott's theory [10], which suggests that the value of S is proportional to the slope of the DOS. Therefore, the W substitution for Cr site in CrSi2 phase also attributed to the monotonic decrease in S with the W content. These results of r and S evaluation revealed that the WSi2 addition on the CrSi2 has little effect on the electrical aspect of thermoelectric performance assessed by PF, as shown in Fig. 7. Every PF-T curve also has a broad maximum around 600 K mainly due to the temperature dependence of S. The PF value reaches 1.8 mW/mK2 around 600 K. To further evaluate the thermoelectric performance of the WSi2added CrSi2 sintered composites, the k value was measured as depicted in Fig. 8. In the lower temperature region, the k values in all the samples tended to decrease with temperature resulting from the decrease in the thermal diffusivity. In the higher temperature above 700 K, the k values increased with temperature because of the increase in Cp and the electronic part of k. Compared to the reported k value, which is indicated by a solid line in Fig. 8, for a stoichiometric coarse-grained CrSi2 compound prepared by arcmelting [4], the CrSi2-only sintered sample made from the milled powder exhibited much lower k value because of the enhancement of phonon scattering at grain boundaries induced by the microstructure refinement as mentioned above. Moreover, the k value was further reduced by the WSi2 addition and reached 4.8 W/mK for the sintered composite with the nominal composition of (CrSi2)0.90(WSi2)0.10 at 300 K. Since the intrinsic electronic part of k in the WSi2 phase can be roughly estimated to be 60 W/mK at 300 K from the reported r value of 0.13 mUm [9] using the Wiedemann-Franz law, the total k value should be greater than that value, which is much higher than the k value of CrSi2 phase. Therefore, the addition of WSi2 phase itself could not expect to reduce k of the CrSi2 material. From the view point of microstructure refinement, the composite effects induced by the WSi2 addition, such as the further reduction of grain size and the introduction of the CrSi2/WSi2 interfaces, should enhance the phonon scattering by the increase in the number of grain boundaries and interfaces, resulting in the reduction of k. In addition, from the crystal-structural viewpoint, the crystal lattice distortion and point defects induced by the W-substitution for Cr site in the CrSi2 phase should also reduce k by the increase in phonon scattering effect, such as the mass-difference scattering. It is
Thermal conductivity (W/mK)
656
CrSi2 CrSi2 W0.02 x=0.02 W0.04 0.04 W0.06 0.06 W0.08 0.08 W0.10 0.10
Ref. 4) Coarse-grained CrSi2
10
5
0
300
500
700
900
1100
Temperature (K) Fig. 8. Temperature dependence of thermal conductivity of the sintered (CrSi2)1x(WSi2)x composites and of a stoichiometric coarse-grained CrSi2 compound prepared by arc-melting in Ref. [4] (solid line).
noteworthy that these modifications for the reduction of k have little influence on the electrical conductivity, which is beneficial for the enhancement of thermoelectric performance. Finally, the ZT value was calculated from the measured thermoelectric properties. Fig. 9 depicts a significant increase in the ZT value due to the WSi2 addition especially in the middle temperature range from 500 K to 900 K. The ZT value then reached 0.29 for the composite with the nominal composition of (CrSi2)0.90(WSi2)0.10 around 750 K. This enhancement of thermoelectric performance is mainly due to the large reduction of k. 4. Conclusions The effect of WSi2 addition on the thermoelectric properties of CrSi2 sintered material was investigated. By using powder metallurgical process, the microstructured composites consisting of the Cr-rich hexagonal (Cr,W)Si2 phase and the W-rich tetragonal (W,Cr) Si2 phase were obtained. The composite effects induced by the WSi2 addition effectively reduced k of CrSi2 material, whereas the metallic nature of WSi2 phase enhanced the electrical conductivity. This contrasting effect on the thermal and electronic conduction can effectively enhance the ZT value of CrSi2-based material. This
2.0
0.3
1.5 0.2
ZT
Power factor (mW/mK2)
15
1.0 CrSi2 CrSi2 W0.02 x=0.02 W0.04 0.04 W0.06 0.06 W0.08 0.08 0.10 W0.10
0.5
0.0
300
500
700
900
CrSi2 CrSi2 W0.02 x=0.02 W0.04 0.04 W0.06 0.06 W0.08 0.08 W0.10 0.10
0.1
1100
Temperature (K) Fig. 7. Temperature dependence of the thermoelectric power factor of the sintered (CrSi2)1x(WSi2)x composites.
0.0 300
500
700
900
Temperature (K)
1100
Fig. 9. Temperature dependence of the thermoelectric figure-of-merit ZT of the sintered (CrSi2)1x(WSi2)x composites.
M. Mikami, Y. Kinemuchi / Journal of Alloys and Compounds 690 (2017) 652e657
result suggests that the addition of metallic or semimetallic materials with nanometer-sized microstructure is an effective way to improve the thermoelectric performance.
[5] [6]
Acknowledgments
[7]
This work was partly supported by the Naito Science & Engineering Foundation (No grant number).
[8]
References
[9]
[1] Z.J. Pan, L.T. Zhang, J.S. Wu, Effects of Al doping on the transport performances of CrSi2 single crystals, Scr. Mater 56 (2007) 245e248. [2] Z.J. Pan, L.T. Zhang, J.S. Wu, Effects of V doping on the transport performances of CrSi2 single crystals, Scr. Mater 56 (2007) 257e260. [3] T. Dasgupta, A.M. Umarji, Role of milling parameters and impurity on the thermoelectric properties of mechanically alloyed chromium silicide, J. Alloys Compd. 461 (2008) 292e297. [4] S. Perumal, S. Gorsse, U. Ail, M. Prakasam, B. Chevalier, A.M. Umarji,
[10] [11]
657
Thermoelectric properties of chromium disilicide prepared by mechanical alloying, J. Mater. Sci. 48 (2013) 6018e6024. J. Ma, Y. Gu, L. Shi, L. Chen, Z. Yang, Y. Qian, Synthesis and thermal stability of nanocrystalline chromium disilicide, J. Alloy Compd. 376 (2004) 176e179. I. Nishida, T. Sakata, Semiconducting properties of pure and Mn-doped chromium disilicides, J. Phys. Chem. Solids 39 (1978) 499e505. Y. Ohishi, A. Mohamad, Y. Miyazaki, H. Muta, K. Kurosaki, S. Yamanaka, Thermoelectric properties of Cr1-xMoxSi2, J. Phys. Chem. Solid. 87 (2015) 153e157. P. Blaha, K. Schwarz, G.K.H. Madsen, D. Kvasnicka, J. Luitz, in: K. Schwarz (Ed.), WIEN2k, an Augmented Plane Wave þ Local Orbitals Program for Calculating Crystal Properties, Techn. Universitat Wien, Austria, 2001. F. Nava, E. Mazzega, M. Michelini, O. Laborde, O. Thomas, J.P. Senateur, R. Madar, Analysis of the electrical resistivity of Ti, Mo, Ta, and W monocrystalline disilicides, J. Appl. Phys. 65 (1989) 1584e1590. N.F. Mott, H. Jones, The Theory of the Properties of Metals and Alloys, Clarendon Press, Oxford, 1936. T. Kajikawa, K. Shiraishi, T. Sugiyama, M. Omori, and T. Hirai, Thermoelectric properties of high temperature transition metal silicides prepared by spark plasma sintering method, Proc. 18th Inter. Conf. Thermoelectrics. (1999) 601e604.
Contents lists available at ScienceDirect
Journal of Alloys and Compounds journal homepage: http://www.elsevier.com/locate/jalcom
Microstructure and thermoelectric properties of WSi2-added CrSi2 composite Masashi Mikami*, Yoshiaki Kinemuchi National Institute of Advanced Industrial Science and Technology, Nagoya 463-8560, Japan
a r t i c l e i n f o
a b s t r a c t
Article history: Received 3 February 2016 Received in revised form 3 August 2016 Accepted 18 August 2016 Available online 20 August 2016
CrSi2-based composite materials with submicrometer-sized grains were fabricated by the powder metallurgical process using ball-milling and pulse-current sintering. By the solid-state reaction of submicrometer-sized fine powders of CrSi2 and WSi2 during the sintering, the composites consisting of Cr-rich hexagonal (Cr,W)Si2 phase and W-rich tetragonal (W,Cr)Si2 phase were obtained. The grain size of CrSi2 phase was effectively reduced by the WSi2 addition because of the suppression of grain growth during sintering. Thermal conductivity was then decreased with the WSi2 content by the enhancement of phonon scattering resulting from the reduction of grain size, the introduction of the CrSi2/WSi2 interface and the crystal lattice distortion induced by the W substitution for CrSi2 phase. Seebeck coefficient and electrical resistivity were also decreased due to the addition of metallic WSi2 phase and the deformation of CrSi2 electronic structure resulting from the W substitution. The WSi2 addition had then little effect on the thermoelectric power factor. Consequently, the thermoelectric figure-of-merit, ZT, was enhanced by the reduction of thermal conductivity and reached 0.3 around 700 K. © 2016 Elsevier B.V. All rights reserved.
Keywords: Thermoelectric materials Silicide Composite materials Powder metallurgy
1. Introduction Thermoelectric devices have recently attracted renewed interest in their potential application to clean energy-conversion systems. In particular, energy recovery from vast amounts of waste heat are targeted for thermoelectric power generation because the temperature and the energy scale of most heat sources are too low to reuse efficiently with other conventional energy conversion systems. The conversion efficiency of thermoelectric devices depends mainly on the thermoelectric performance of the material, which is evaluated using the thermoelectric figure of merit, ZT ¼ (S2/rk)T, where S is the Seebeck coefficient, r is the electrical resistivity, k is the thermal conductivity, and T is the absolute temperature. The telluride materials, such as Bi-Te or Pb-Te system, exhibit relatively high ZT values and are suited for thermoelectric power generation devices. However, it seems to have run into difficulty for widespread use because of its higher cost and the limited supply of raw materials, especially Te. Transition metal silicides have been studied intensively for thermoelectric application because of its abundance of raw
* Corresponding author. E-mail address: [email protected] (M. Mikami). http://dx.doi.org/10.1016/j.jallcom.2016.08.185 0925-8388/© 2016 Elsevier B.V. All rights reserved.
material, low toxicity and high chemical stability. Chromium disilicide, CrSi2, is one of the promising silicide because of its large S and low r, resulting in large power factor, PF (¼S2/r), over 1 mW/ mK2 above room-temperature [1e4]. In addition, the high oxidation resistance in air up to 900 K [5] is a favorable property for the long-term stability as a thermoelectric power generation device. However, the relatively high intrinsic k value of CrSi2 of over 12 W/ mK [4] limits its ZT value to ~0.2. The reduction of k is therefore necessary for efficient thermoelectric power generation. Since k in the CrSi2 system is mainly governed by the lattice part of k, the enhancement of phonon scattering should be effective to reduce total k value. Indeed, k value was effectively decreased by the structural refinement, such as the reduction of grain size [3,4]. In this study, CrSi2-based composites were fabricated by the addition of WSi2. For the selection of additive material, the low r and crystallographic compatibility are desirable in order to avoid the degradation of electrical conductivity by the formation of composite. In addition, the high chemical stability in CrSi2 matrix is also required. From these aspect, metallic or semimetallic silicide materials seem to be promising candidate as the additive material for CrSi2. However, since the 3d transition silicides or the homologous molybdenum silicide easily form a solid solution with CrSi2 [2,6,7], it is difficult to obtain the stable composite structure. Therefore, WSi2 having a limited solid
M. Mikami, Y. Kinemuchi / Journal of Alloys and Compounds 690 (2017) 652e657
2. Experimental CrSi2-based composite materials with the nominal composition of (CrSi2)1x(WSi2)x (x ¼ 0.02e0.10) were fabricated by the powder metallurgical process. Commercially available CrSi2 (99%, Rare Metallic Co., Ltd.) and WSi2 (99%, Rare Metallic Co., Ltd) powders were used as the raw material. Although the purity of these raw materials was rather low, almost all impurity materials are oxides, which are usually stable in silicide. Therefore, it can be expected that the effect of impurity phase on thermoelectric properties of samples is negligible. Appropriate amounts of CrSi2 and WSi2 powders were mixed and milled in a planetary ball mill (FRITSCH GmbH, PULVERISETTE 5/4) with a 500-ml-capacity Cr steel pot and carbide balls. In the milling system, the pot rotates on its axis against the direction of orbital motion. The pot was back-filled with a purified Ar gas atmosphere after creating a vacuum. A total mass of 800 g of 5-mm-diameter WC balls was inserted with 30 g of mixed powder into the pot. The ball-milling was performed at 250 rpm for 10 h with a 10 min rest every half-hour. After the ballmilling process, well-mixed and finely-pulverized powders with submicrometer-size were obtained. The prepared powder was sintered using pulse-current sintering (PCS). The powder was put into a graphite mold and sintered at 1373 K for 10 min in vacuum under a uniaxial pressure of 50 MPa. The heating and cooling rates were 100 K/min. For comparison, CrSi2-only sintered samples were also prepared from the raw CrSi2 powder with an average particle size of several mm and from the as-milled CrSi2 powder synthesized by the same ball-milling process. The obtained bulk samples, with a typical size of 10 mm in diameter and 2 mm in thickness, were used to measure k and then cut into a bar shape with a typical size of 2 2 9 mm3 to measure r and S. Crystalline phase analysis was performed using X-ray diffraction (XRD) with Cu Ka radiation. Microstructural observation and compositional analysis was conducted using the scanning electron microscopy (SEM) and the attached energy dispersive X-ray spectrometry (EDX). S and r were simultaneously measured in a He atmosphere. r was evaluated by a conventional four-probe DC technique. S was calculated from a plot of the thermoelectric voltage versus the temperature difference. k was evaluated from the density (D), thermal diffusivity (a), and heat capacity (Cp) with the relationship k ¼ D a Cp. D was measured using the Archimedes method, and a and Cp were evaluated using the laser flash method. Density functional theory was performed based on the linearized augmented plane wave method using Wien2k code [8]. For the pure CrSi2, experimental structure was adopted, while the structural optimization was carried out for W doped CrSi2. Here, doping was performed to the super cell structure of 1 1 2, and chemical composition was fixed to the Cr2WSi6 in which doping site was selected to the Cr (1/2, 1/2, 1/3) in the super cell. The number of kpoint in the Brillouin zone of 2560, the size of basis set (rkmax) of 8.0, and the exchange-correlation functional of generalized
gradient approximation were adopted. The calculation selfconverged within the energy of 0.1 mRy. 3. Results and discussion The XRD measurement was performed on the cross-section of sintered (CrSi2)1x(WSi2)x composites to investigate the relation between the composition x and the crystal structure of the obtained samples. As shown in Fig. 1, the stoichiometric CrSi2-only sintered sample made from the as-milled powder formed a C40 hexagonal structure with space group of P6222 as reported in the literature. Although no trace of an impurity phase, such as oxides, was detected, weak diffraction perks from chromium monosilicide CrSi was observed. The Rietveld analysis had revealed that the presence of a secondary phase of the CrSi phase at 2.7 at%. In addition, the CrSi phase was also observed in every composite samples around 3 at%. However, it is reasonable to assume that the secondary phase could not disturb the investigation of WSi2 addition effect on CrSi2 phase because the CrSi phase was present in every samples at an almost constant proportion. For the composite samples, diffraction peaks from CrSi2 phase tend to monotonically shift toward the lower 2q angle with the amount of WSi2 content. The lattice parameter of CrSi2 phase calculated from the 2q angle of diffraction peaks then increased with the amount of the WSi2 addition, as shown in Fig. 2. This tendency reflects that the crystallographic structure of CrSi2 phase was enlarged by the partial substitution of W, which has a larger atomic radius than Cr. In addition, diffraction peaks from the tetragonal WSi2 phase were detected in x 0.04, as shown in Fig. 1. Moreover, the peak position was slightly higher than that of the WSi2 powder, which was used as a raw material. Although it is difficult to precisely calculate lattice parameter from these weak diffraction peaks, this result implies that the W site in the WSi2 phase was also partially substituted by Cr having the smaller atomic radius, resulting in the shrinkage of
WSi2 CrSi
x=0.10
x=0.08
Intensity (arb. units)
solubility for CrSi2 was chosen as an additive material in this study. The composites consisting of submicrometer-sized fine grains were prepared by the powder metallurgical process using ballmilling and pulse-current sintering in order to enhance the phonon scattering at grain boundaries. In addition, the heavyelement W doping effect on the CrSi2 crystal structure, which is also effective for the reduction of k by the mass-difference scattering in the crystal lattice, was investigated. It was found that the appreciable reduction of k by the WSi2 addition without a serious deterioration of electrical aspect of thermoelectric property improved the ZT value of the CrSi2-based material.
653
x=0.06
x=0.04
x=0.02
CrSi2
20
40
60
80
100
2θ (deg) Fig. 1. XRD patterns (Cu Ka radiation) diffracted from the CrSi2 sintered sample made from the milled powder and sintered composites with the nominal composition of (CrSi2)1x(WSi2)x.
654
M. Mikami, Y. Kinemuchi / Journal of Alloys and Compounds 690 (2017) 652e657
Fig. 2. Relation between the nominal WSi2 content and lattice parameter calculated from 2q angle of peaks in XRD patterns of the sintered (CrSi2)1x(WSi2)x composite.
crystal structure of WSi2 phase. Since these peak shifts were not observed in the mixed powders after the ball-milling process, the element substitutions should be induced by the solid-state reaction of CrSi2 and WSi2 during the sintering. The SEM observation was performed on the as-milled powders and the cross-section of sintered samples. A typical SEM image of the as-milled powder was shown in Fig. 3a). The powders were consisted of submicrometer-sized particles and the dependence of particle size on the WSi2 content was not observed. Fig. 3b) and c) show the microstructure of sintered samples made from the raw CrSi2 powder consisting of several mm particles and from the asmilled CrSi2 powder. Because of the difference of particle size of the powders, grain size in sintered sample was reduced from several mm to around 1 mm. In addition, the nanometer-sized dot structure, which appears dark spots on SEM image, was formed inside the grains of sintered sample made from the as-milled CrSi2 powder. Although formation mechanism and crystallographic phase were unclear, it is presumed that the nano-dot structure should be attributed to compositional variation by the pulverization and/or unavoidable slight oxidation during the milling process and the handling of powders in air. The nano-dot structure was then observed in every sample made from milled powders, as shown in Fig. 3e)-i). One should be noticed that grain size of sintered sample was effectively decreased by the WSi2 addition. For instance, grain size was reduced from around 1 mm to 300 nm by the small amount of 2 at% addition of WSi2, as shown in Fig 3e). Although the clear segregation of WSi2 at grain boundary of CrSi2 was not observed, the WSi2 with the high melting point of 2440 K is assumed to suppress the grain growth of CrSi2 during sintering because the grain size is almost the same as the particle size of the milled powder. As shown in Fig. 3d), the sparsely-distributed bright spots were observed in a low magnification SEM image of the sintered sample with the nominal composition of (CrSi2)0.98(WSi2)0.02. EDX mapping analysis suggested that these bright spots were attributed to the WSi2 phase grains with micrometer size, although it was difficult to determine the composition precisely. Therefore, this sample also formed a CrSi2-WSi2 composite, whereas clear diffraction peaks from WSi2 phase was not detected on the XRD pattern. The number density of the micrometer-sized WSi2 grains then increased with the WSi2 content. In addition, nanometersized WSi2 grains were also observed in the sintered (CrSi2)1x(WSi2)x samples with x 0.04, as shown in Fig. 3f)-i). The number density of the nanometer-sized WSi2 grains increased
Fig. 3. SEM images of a) the as-milled powder with the nominal composition of (CrSi2)0.9(WSi2)0.1 and the fracture cross-sections of the sintered sample made from b) the raw powder of CrSi2, c) the milled powder of CrSi2, d-i) (CrSi2)1x(WSi2)x for 0.04 x 0.10. j) Chemical compositions of (CrSi2)1x(WSi2)x composites measured by EDX. Each dashed line represents the nominal composition.
with the WSi2 content while the size is nearly independent of the WSi2 content. The chemical compositions in sintered composites were almost identical to those of the nominal compositions, as shown in Fig. 3j). Furthermore, the absence of contaminant from ball-milling media, such as Fe, was confirmed by EXD analysis. Therefore, compositional variation and contamination during ball-milling process was sufficiently suppressed and negligibly low. These results of XRD, SEM and EDX analysis indicate that the compositionally controlled composites consisting of the Cr-rich hexagonal (Cr,W)Si2 phase and the W-rich tetragonal (W,Cr)Si2 phase were obtained. In order to clarify the effect of the WSi2 addition on the thermoelectric performance of CrSi2 compound, thermoelectric
M. Mikami, Y. Kinemuchi / Journal of Alloys and Compounds 690 (2017) 652e657
16
DOS / state eV-1 UC-1
Electrical Resistivity (μΩm)
15
10
CrSi2 CrSi2 x=0.02 W0.02 0.04 W0.04 W0.06 0.06 W0.08 0.08 W0.10 0.10
5
500
700
900
a) CrSi2 (C40)
14
total Cr Si
12 10 8
6 4 2 0
0 300
655
-4
1100
2
4
DOS / state eV-1 UC-1
16
b) Cr2WSi6 (C40)
14
total Cr Si W
12 10 8
6 4 2 0 -4
-2
0
2
4
E / eV Fig. 5. The calculated density of state for a) pure CrSi2 and b) W-substituted CrSi2 phase.
200
Seebeck coeficient (μV/K)
properties were evaluated. Fig. 4 shows r of the sintered composites incorporating different amounts of WSi2 as a function of temperature. The every r-T curve has a broad maximum around 700 K. This temperature dependence of r is essentially identical to the previously reported electrical conduction property of CrSi2 phase having the degenerate semiconducting nature [6]. In the lower temperature exhaustion region, r increases with temperature due to the decrease in carrier mobility induced by the acoustic phonon scattering. In the higher temperature intrinsic region, r decreases with temperature because of the increase in carrier concentration resulting from electron carriers thermally excited across the bandgap. On the other hand, the value of r monotonically decreases with the WSi2 content. Since the number of valence electrons of W is the same as that of Cr, carrier doping effect by the element substitution on the CrSi2 phase could not be expected. Instead, the addition of metallic WSi2 phase having the low r of 0.1e0.8 mUm [9] should play the main role in the reduction of r in the composites. For the further consideration, in order to assess the W substitution effect on electronic properties of CrSi2 phase, the electronic structure calculation was conducted. As shown in Fig. 5, the calculated density of state (DOS) suggested that the band gap of CrSi2 phase decreases by the W substitution mainly due to the dispersion of electronic state especially around valence band-edge related to Cr. Therefore, this narrowing of band-gap also caused the reduction of r in the composites. Fig. 6 represents S of the sintered composites as a function of temperature. Every S-T curve also exhibits a broad maximum around 650 K. This temperature dependence of S can also be attributed to the degenerate semiconducting nature of CrSi2 phase [6]. In the lower temperature exhaustion region, the linear increase in S with temperature can be roughly explained according to the Mott's theory [10], which suggests that the value of S is proportional to temperature. In the higher temperature intrinsic region, the electron carriers thermally excited across the bandgap counteract the Seebeck effect produced by holes, resulting in the decrease in S with increasing temperature. In addition, the value of S monotonically decreases with the WSi2 content especially in the lower temperatures. This decrease in S should be caused by the addition of metallic WSi2 phase having the low S value of less than 10 mV/K below 600 K [11]. In the higher temperatures, since the S value in WSi2 phase increases rapidly with increase in temperature from 20 mV/K to 110 mV/K in the temperature range from 700 K to 1100 K [11], the S reduction effect by the WSi2 addition decreases with rising temperature, as shown in Fig. 6.
0
E / eV
Temperature (K) Fig. 4. Temperature dependence of electrical resistivity in the sintered (CrSi2)1x(WSi2)x composites.
-2
150
100 CrSi2 CrSi2 W0.02 x=0.02 W0.04 0.04 W0.06 0.06 W0.08 0.08 W0.10 0.10
50
0
300
500
700
900
1100
Temperature (K) Fig. 6. Temperature dependence of Seebeck coefficient in the sintered (CrSi2)1x(WSi2)x composites.
For the consideration of W substitution effect on S of CrSi2 phase, the electronic structure calculation offers some insight into possible behaviors of S. Fig. 5 represents the deformation of DOS by the W substitution especially in the electronic state related to Cr. In comparison with the pure CrSi2 phase, the narrowing of band gap resulting from the increase in DOS around conduction band edge and the lowering the slope of DOS around valence band edge can be observed in DOS of W-substituted CrSi2 phase. The narrowing band
M. Mikami, Y. Kinemuchi / Journal of Alloys and Compounds 690 (2017) 652e657
gap intensify the bipolar contribution of electron carriers excited thermally across the bandgap, resulting in the decrease in positive S of holes. The lowering the slope of DOS around valence band edge also reduce S according to Mott's theory [10], which suggests that the value of S is proportional to the slope of the DOS. Therefore, the W substitution for Cr site in CrSi2 phase also attributed to the monotonic decrease in S with the W content. These results of r and S evaluation revealed that the WSi2 addition on the CrSi2 has little effect on the electrical aspect of thermoelectric performance assessed by PF, as shown in Fig. 7. Every PF-T curve also has a broad maximum around 600 K mainly due to the temperature dependence of S. The PF value reaches 1.8 mW/mK2 around 600 K. To further evaluate the thermoelectric performance of the WSi2added CrSi2 sintered composites, the k value was measured as depicted in Fig. 8. In the lower temperature region, the k values in all the samples tended to decrease with temperature resulting from the decrease in the thermal diffusivity. In the higher temperature above 700 K, the k values increased with temperature because of the increase in Cp and the electronic part of k. Compared to the reported k value, which is indicated by a solid line in Fig. 8, for a stoichiometric coarse-grained CrSi2 compound prepared by arcmelting [4], the CrSi2-only sintered sample made from the milled powder exhibited much lower k value because of the enhancement of phonon scattering at grain boundaries induced by the microstructure refinement as mentioned above. Moreover, the k value was further reduced by the WSi2 addition and reached 4.8 W/mK for the sintered composite with the nominal composition of (CrSi2)0.90(WSi2)0.10 at 300 K. Since the intrinsic electronic part of k in the WSi2 phase can be roughly estimated to be 60 W/mK at 300 K from the reported r value of 0.13 mUm [9] using the Wiedemann-Franz law, the total k value should be greater than that value, which is much higher than the k value of CrSi2 phase. Therefore, the addition of WSi2 phase itself could not expect to reduce k of the CrSi2 material. From the view point of microstructure refinement, the composite effects induced by the WSi2 addition, such as the further reduction of grain size and the introduction of the CrSi2/WSi2 interfaces, should enhance the phonon scattering by the increase in the number of grain boundaries and interfaces, resulting in the reduction of k. In addition, from the crystal-structural viewpoint, the crystal lattice distortion and point defects induced by the W-substitution for Cr site in the CrSi2 phase should also reduce k by the increase in phonon scattering effect, such as the mass-difference scattering. It is
Thermal conductivity (W/mK)
656
CrSi2 CrSi2 W0.02 x=0.02 W0.04 0.04 W0.06 0.06 W0.08 0.08 W0.10 0.10
Ref. 4) Coarse-grained CrSi2
10
5
0
300
500
700
900
1100
Temperature (K) Fig. 8. Temperature dependence of thermal conductivity of the sintered (CrSi2)1x(WSi2)x composites and of a stoichiometric coarse-grained CrSi2 compound prepared by arc-melting in Ref. [4] (solid line).
noteworthy that these modifications for the reduction of k have little influence on the electrical conductivity, which is beneficial for the enhancement of thermoelectric performance. Finally, the ZT value was calculated from the measured thermoelectric properties. Fig. 9 depicts a significant increase in the ZT value due to the WSi2 addition especially in the middle temperature range from 500 K to 900 K. The ZT value then reached 0.29 for the composite with the nominal composition of (CrSi2)0.90(WSi2)0.10 around 750 K. This enhancement of thermoelectric performance is mainly due to the large reduction of k. 4. Conclusions The effect of WSi2 addition on the thermoelectric properties of CrSi2 sintered material was investigated. By using powder metallurgical process, the microstructured composites consisting of the Cr-rich hexagonal (Cr,W)Si2 phase and the W-rich tetragonal (W,Cr) Si2 phase were obtained. The composite effects induced by the WSi2 addition effectively reduced k of CrSi2 material, whereas the metallic nature of WSi2 phase enhanced the electrical conductivity. This contrasting effect on the thermal and electronic conduction can effectively enhance the ZT value of CrSi2-based material. This
2.0
0.3
1.5 0.2
ZT
Power factor (mW/mK2)
15
1.0 CrSi2 CrSi2 W0.02 x=0.02 W0.04 0.04 W0.06 0.06 W0.08 0.08 0.10 W0.10
0.5
0.0
300
500
700
900
CrSi2 CrSi2 W0.02 x=0.02 W0.04 0.04 W0.06 0.06 W0.08 0.08 W0.10 0.10
0.1
1100
Temperature (K) Fig. 7. Temperature dependence of the thermoelectric power factor of the sintered (CrSi2)1x(WSi2)x composites.
0.0 300
500
700
900
Temperature (K)
1100
Fig. 9. Temperature dependence of the thermoelectric figure-of-merit ZT of the sintered (CrSi2)1x(WSi2)x composites.
M. Mikami, Y. Kinemuchi / Journal of Alloys and Compounds 690 (2017) 652e657
result suggests that the addition of metallic or semimetallic materials with nanometer-sized microstructure is an effective way to improve the thermoelectric performance.
[5] [6]
Acknowledgments
[7]
This work was partly supported by the Naito Science & Engineering Foundation (No grant number).
[8]
References
[9]
[1] Z.J. Pan, L.T. Zhang, J.S. Wu, Effects of Al doping on the transport performances of CrSi2 single crystals, Scr. Mater 56 (2007) 245e248. [2] Z.J. Pan, L.T. Zhang, J.S. Wu, Effects of V doping on the transport performances of CrSi2 single crystals, Scr. Mater 56 (2007) 257e260. [3] T. Dasgupta, A.M. Umarji, Role of milling parameters and impurity on the thermoelectric properties of mechanically alloyed chromium silicide, J. Alloys Compd. 461 (2008) 292e297. [4] S. Perumal, S. Gorsse, U. Ail, M. Prakasam, B. Chevalier, A.M. Umarji,
[10] [11]
657
Thermoelectric properties of chromium disilicide prepared by mechanical alloying, J. Mater. Sci. 48 (2013) 6018e6024. J. Ma, Y. Gu, L. Shi, L. Chen, Z. Yang, Y. Qian, Synthesis and thermal stability of nanocrystalline chromium disilicide, J. Alloy Compd. 376 (2004) 176e179. I. Nishida, T. Sakata, Semiconducting properties of pure and Mn-doped chromium disilicides, J. Phys. Chem. Solids 39 (1978) 499e505. Y. Ohishi, A. Mohamad, Y. Miyazaki, H. Muta, K. Kurosaki, S. Yamanaka, Thermoelectric properties of Cr1-xMoxSi2, J. Phys. Chem. Solid. 87 (2015) 153e157. P. Blaha, K. Schwarz, G.K.H. Madsen, D. Kvasnicka, J. Luitz, in: K. Schwarz (Ed.), WIEN2k, an Augmented Plane Wave þ Local Orbitals Program for Calculating Crystal Properties, Techn. Universitat Wien, Austria, 2001. F. Nava, E. Mazzega, M. Michelini, O. Laborde, O. Thomas, J.P. Senateur, R. Madar, Analysis of the electrical resistivity of Ti, Mo, Ta, and W monocrystalline disilicides, J. Appl. Phys. 65 (1989) 1584e1590. N.F. Mott, H. Jones, The Theory of the Properties of Metals and Alloys, Clarendon Press, Oxford, 1936. T. Kajikawa, K. Shiraishi, T. Sugiyama, M. Omori, and T. Hirai, Thermoelectric properties of high temperature transition metal silicides prepared by spark plasma sintering method, Proc. 18th Inter. Conf. Thermoelectrics. (1999) 601e604.