2019 Volume 60 Issue 6 Pages 1072-1077
In the present study, self-assembled submicrometer-sized composite structure of bismuth-antimony (Bi–Sb) alloy and indium antimonide (InSb) were fabricated by melt-spinning (MS) technique and spark plasma sintered (SPS) process with the eutectic composition. The power factor reached 2.6 mW/mK2 at 300 K. The thermal conductivity decreased by 35% compared with the Bi–Sb alloy. The results suggest that this MS-SPS procedure is promising for reducing the thermal conductivity with maintaining the electrical properties.
Fig. 6 The temperature dependences of thermal conductivities, (a) total thermal conductivity, (b) electronic thermal conductivity, and (c) lattice thermal conductivity.
Thermoelectric (TE) materials can directly convert heat into electricity and have been considered to be an alternative probability to address the environmental problems and the upcoming energy crisis.1,2) However, the present usage is limited to specific applications due to its low conversion efficiency.3) The conversion efficiency of a TE device is determined by the dimensionless figure of merit ZT, and is defined as ZT = S2ρ−1Tκ−1, where S is the Seebeck coefficient, ρ is the electrical resistivity, κ = κe + κl is the thermal conductivity which composed of electronic thermal conductivity κe, and lattice thermal conductivity κl, and T is the absolute temperature.4) Apparently, high electrical properties (power factor, PF = S2ρ−1) and low thermal conductivities are necessary to obtain large ZT values for TE materials. Nevertheless, the high power factor is usually accompanied by a high thermal conductivity because the Seebeck coefficient, the electrical conductivity and the electronic thermal conductivity are interrelated to each other. Therefore, it is challenging to enhance ZT value by decoupling these parameters, which means reducing the thermal conductivity without degradation of the power factor.5)
Both bismuth (Bi) and antimony (Sb) are semimetals from group V metals with a similar crystal structure of A7 structure (rhombohedral) of $R\bar{3}m$ space group.6) Bi100−xSbx alloys form a solid solution over the entire composition range because Bi and Sb have similar chemical properties and lattice parameters.7) Bi–Sb alloy is a n-type material (typically S = −100 µV K−1 at 300 K) and semiconducting inside the range 0.07 < x < 0.22. The best ZT = 0.33 at 300 K along the trigonal direction was reported for single crystal with 0.09 < x < 0.16.8–11) Nonetheless, the single crystals are mechanically weak, having difficulty in cleaving and growing, along with slow growth rate.11,12) Poly-crystalline materials might be suitable but so far all have been reported to have lower ZT than the single crystals. The performance degradation is mainly caused by heterogeneity in the Bi/Sb composition inside the sample, which easily occurs at the cooling process. Hence several preparation methods of melt-quenching,11) arc melting,12,13) mechanical alloying,14) and rapid solidification with spark plasma sintering15) process have been developed in order to enhance the homogeneity and ZT performance. Nevertheless, conventional water-quenching can not provide sufficient cooling rate, and oxidation and contamination occur in the material during the mechanical alloying process. These problems can be prevented in the rapid solidification process such as melt-spinning (MS) method.
Recently, we have synthesized nanometer-sized eutectic composite using the MS method combined with subsequent spark plasma sintering (SPS) process for Si–CrSi2 system.16,17) Very fine microstructure was obtained using both the natural phase separation at eutectic composition and liquid quenching by MS. The thermal conductivity was reduced and ZT was enhanced by the fabrication method. We focused on the Bi–Sb alloy - indium antimonide (InSb) system, which has a typical eutectic point between Bi–Sb and InSb as shown in the phase diagram in Fig. 1. The eutectic phase has a lower melting point than its components. InSb is a narrow band gap semiconductor with a band gap of 0.18 eV at room temperature and possesses a high electron mobility18) and also a high thermal conductivity.19) In this work, we prepared the eutectic n-type Bi0.88Sb0.12/InSb TE material using MS technique and consequently sintered by SPS. The TE properties of the composites samples were measured, and the relationship between TE transport properties and microstructure was discussed.
Pseudobinary phase diagram of the BiSb–InSb system.
As starting materials, Bi (5N), In (5N) and Sb (5N) were used to prepare a master ingots with a nominal composition of (Bi0.88Sb0.12)90(InSb)10. The material have been carefully weighted out with the stoichiometric and loaded in the silica tubes. The silica tubes were sealed under vacuum using oxygen/hydrogen torch and transferred in the vertical furnace. The silica tubes have been heated with the melting condition (823 K, 5 K/min for 12 hours holding) and then rapidly quenched by putting the heat treated silica tubes into cold water. The obtained quench-melted (QM) ingot was then melted in a silica tube with a 0.6 mm diameter nozzle by induction heating. The MS technique was used with the melted ingot ejected under an argon (Ar) atmosphere of 0.1 MPa onto a copper roller that was operated at wheel linear speed of 2000 RPM and 4000 RPM. The obtained ribbon samples are represented as MS2000 MS4000, where the number indicates the rotation speed of the roller. Subsequently, the QM, MS2000 and MS4000 ribbon sample were roughly crushed in a tungsten carbide mortar for 1 min by hand into fine powders. The obtained powder sample was placed into a graphite die for SPS in Ar flow at 473 K for 10 min under axial pressure of 50 MPa. The forming bulk samples are denoted as MS2000-SPS, and MS4000-SPS. A sample with same composition and same SPS conditions was prepared by QM and SPS to compare the morphology, denoted as QM-SPS.
2.2 CharacterizeThe samples phase identification was characterized by powder X-ray diffraction (XRD) analysis using an X-ray diffractometer (Ultima IV, Rigaku Co.) with Cu Kα radiation at room temperature. The powder lattice parameters were deduced from XRD peak positions using a least-squares refinement method. The microstructure of all the sample was observed using a field-emission scanning electron microscope (FE-SEM, JEOL JSM-6500F) and energy dispersive X-ray spectrometry (EDX). The TE properties of the bulk sample were measured. The ρ and S of the samples were determined simultaneously in a He atmosphere at temperatures ranging from 300 K to 773 K by a four-point technique and a static direct-current method using a ZEM-3 instrument system (ULVAC-RIKO, Inc.). The κ, was determined using a formula of κ = αCPd, where α is the thermal diffusivity, CP is the heat capacity, and d is the sample density. The α, was measured by laser flash method (Netzsch LFA-457) from 300 K to 773 K. CP = 3nR, where n is the number of atoms per formula unit and R is the gas constant. The calculated room temperature heat capacity and measurement of the thermal diffusivity are shown in Table 1. d was calculated from the measured weight and dimensions of the samples.
Figure 2 shows the powder XRD patterns for the QM, MS and SPS samples. The powder XRD pattern confirms that all the patterns were well consistent with those of Bi and InSb. As shown in the Bi–Sb peaks, a simultaneous shift to higher angle was observed compared with peaks of pure Bi (JCPDS 01-085-1331). The peak split was not observed. These results confirm the solid solution formation for the Bi–Sb phase and volume reduction by the Sb addition. The volume change agrees with the smaller lattice parameters for pure Sb, as shown in Table 2. The peak width of the XRD pattern is noticeably broadened in the MS2000 and MS4000 samples as compared with the other SPS samples, indicating that melt-spinning technique forms the very fine microstructure. The XRD peaks sharpen for MS2000-SPS and MS4000-SPS, indicating grain growth occurred in the sintering process. Among the sintered samples, peaks for QM-SPS sample was broader and shows tailing. It may be caused by the heterogeneity in the Bi/Sb composition due to the insufficient cooling rate at the fabrication process. The lattice parameters of Bi–Sb phase calculated from the XRD patterns are shown in Table 1, which lies between the reference values for pure Bi and Sb.20) The MS-SPS samples show smaller lattice parameters than QM-SPS samples. The difference may be affected by the sample heterogeneity as mentioned above. Relative density was estimated from the volume fraction and calculated theoretical density of each phase. The volume fraction of InSb phase is about 18%. The relative densities for composite samples are higher than 97% of the theoretical values.
Powder XRD pattern of (Bi0.88Sb0.12)90(InSb)10 samples.
To identify the atomic distribution in the sample, EDS analysis was carried out for MS4000-SPS sample and the images are shown in Fig. 3. The Bi–Sb alloy and InSb phases are clearly separated and the elements are homogeneously distributed in each phase. The grey matrix areas in SEM image correspond to Bi–Sb phase and the dark region represents the InSb phase as shown in Fig. 3(a). The microstructure observation confirms Bi–Sb alloy-InSb dual phase structure with spherical InSb embedded in the matrix of Bi–Sb alloy. The eutectic structure shape was determined by the volume fraction, surface tension, growth rate, and so on. The low volume fraction of InSb phase (18% < 28%) may generate that microstructure.
Surface image of MS4000-SPS sample: (a) SEM image and EDX mapping images of (b) In, (c) Sb, and (d) Bi, respectively. The white scale bars represent 5 µm.
In the SEM images of Fig. 4, it is found that the MS2000-SPS and MS4000-SPS samples, which were subjected to melt-spinning, have finer microstructure than of the comparative QM-SPS sample. Although grain growth occurred during the sintering process, the size of InSb particle is 1∼2 µm for MS2000-SPS sample and 0.5∼1 µm for MS4000-SPS sample. Very fine microstructure was successfully generated by MS-SPS process for the sample at the eutectic composition.
Scanning electron micrograph of the (Bi0.88Sb0.12)90(InSb)10 samples (A) QM-SPS, (b) MS2000-SPS, and (c) MS4000-SPS.
Figure 5 shows the temperature dependence of Seebeck coefficient S, electrical resistivity ρ, and power factor PF for Bi0.88Sb0.12 and (Bi0.88Sb0.12)90(InSb)10 composite samples. For all the samples and over the entire temperature range, the sign of the S is negative and ρ values increased with increasing temperature. The composite samples have larger absolute S and higher ρ than Bi0.88Sb0.12. It would be caused by existence of semiconducting InSb phase, which possess larger S and higher ρ values. By comparing QM-SPS and MS4000-SPS, larger S value is observed for the MS sample. The rapid cooling rate at the MS process could improve the homogeneity in Bi–Sb phase, which increase the band gap and enhance the S value. On the other hand, the enhancement in S is not observed for MS2000-SPS sample, indicating that the cooling rate was insufficient in spite of the MS process. Same tendency was reported previously,15) considerably high cooling rate is needed for the Bi–Sb alloy. The higher cooling rate decreases the InSb particle size as shown in Fig. 4, which increases electron-interface scattering. Hence MS samples show higher electrical resistivity than QM-SPS sample. However, the electrical resistivity of MS2000-SPS sample is higher than that of MS4000-SPS sample. Additional mechanisms may alter the electrical resistivity upon the carrier concentration. It is notable that PF of MS4000-SPS sample is nearly equal to that of QM-SPS in spite of the stronger electron-interface scattering, reached 2.6 mW/mK2 at 300 K. It indicates that the interfaces between Bi–Sb and InSb phases are clean because natural eutectic phase separation occurs in this fabrication procedure, which maintain the electrical properties as TE materials.
Temperature dependence of (a) electrical resistivity, (b) Seebeck coefficient, (c) power factor for Bi0.88Sb0.12 and composite samples.
Figure 6 shows the temperature dependence of thermal conductivities. The total thermal conductivity was lowered for the composite samples compared to Bi0.88Sb0.12. The sample MS4000-SPS has minimum thermal conductivity among the samples, 30–35% lower than Bi–Sb alloy. Electronic thermal conductivity was calculated using the Wiedemann-Franz law κe = L · ρ−1 · T, where L is the Lorenz number. Lorenz number depends on the Fermi level, hence estimated from the Seebeck coefficient. The values of around 2 × 10–8 W·Ω·K−2 are used. κl was obtained by subtracting κe from κ. The sample MS4000-SPS shows lower κ than other composite samples due to the fine microstructure, which provide frequent phonon-interface scattering. Nevertheless, the composite samples have higher lattice thermal conductivities than Bi0.88Sb0.12 alloy. It would be caused by the high thermal conductivity of InSb phase, around 18 W/mK at room temperature.19) Further refining is needed to enhance the phonon-interface scattering and ignore the high thermal conduction in InSb phase.
The temperature dependences of thermal conductivities, (a) total thermal conductivity, (b) electronic thermal conductivity, and (c) lattice thermal conductivity.
The temperature dependence of ZT in each sample is shown in Fig. 7. MS4000-SPS sample shows the highest ZT among other samples in entire temperature range. The ZT decreases with temperature, the maximum value of 0.23 is obtained at room temperature. By comparison with QM-SPS sample, the ZT value improves by about 25% in all temperature range. It is achieved by reduction of thermal conductivity without degradation of power factor. It indicates that this microstructure refining technique, using liquid quenching and phase separation at the eutectic composition, is promising method to improve the ZT value. On the other hand, the ZT value is almost same as that for Bi0.88Sb0.12 alloy. Appropriate materials selection and more rapid cooling rate are needed to achieve further improvement of thermoelectric performance.
Temperature dependence of ZT for Bi0.88Sb0.12 and composite samples.
We successfully fabricated Bi0.88Sb0.12/InSb eutectic alloy and their microstructure and thermoelectric properties were investigated. The power factor reached a maximum value of 2.6 mW/mK2 at 300 K for the MS4000-SPS sample. Although the ZT is as same as the eutectic-free Bi–Sb matrix with the value of 0.23 at 300 K, we could reduce the thermal conductivity by 35%. Furthermore the thermal conductivity reduces while maintaining the power factor for the composite samples. Combination of rapid cooling by melt spinning and eutectic-phase separation is a promising method to synthesize high performance thermoelectric composites.
