Thermoelectric Properties of Ni0.05Mo3Sb5 ... - Wiley Online Library

DOI: 10.1002/ejic.201501063

Full Paper

Thermoelectric Materials

Thermoelectric Properties of Ni0.05Mo3Sb5.4Te1.6 with Embedded SiC and Al2O3 Nanoparticles Nagaraj Nandihalli,[a] Quansheng Guo,[a] Stéphane Gorsse,[b] Atta Ullah Khan,[b] Takao Mori,[c] and Holger Kleinke*[a] Abstract: First, a large sample of Ni0.05Mo3Sb5.4Te1.6 was prepared by heating the elements in the stoichiometric ratio. β-SiC nanoparticles were added in volume fractions, f, of 0.01, 0.02, and 0.034 to three different portions of the material, and a portion without SiC nanoparticles was retained as a reference. All four samples were subjected to consolidation by hot-pressing. Furthermore, Al2O3 nanoparticles were added in volume fractions of f = 0.01, 0.0216, and 0.0325 to three other portions, again retaining a fourth protion as reference. These four samples were consolidated by spark-plasma sintering. The thermoelectric transport properties of these composites were characterized from 325 to 740 K. For the sample with 0.01 volume

fraction of SiC, there was an enhancement in figure of merit by 18 % compared with the reference sample, mainly due to an 18 % reduced thermal conductivity. The 9 % reduction in thermal conductivity of the sample with 0.01 volume fraction of Al2O3 was not enough to compensate the loss in the power factor of 18 %, leading to a decrease in the figure of merit. Microstructural information obtained by SEM, TEM, and BET was used to elucidate the phase and transport properties. The spark-plasma-sintered bulk sample has a figure of merit that is 35 % higher than the bulk sample consolidated through hotpressing.

Introduction

tronics, for which reliability is of the utmost importance, by using electronic-thermal-control (ETC) systems.[11] The energy efficiency of a thermoelectric device is determined by its dimensionless figure of merit, ZT, and is defined as ZT = TS2σκ–1,[12] in which T is the temperature, S is the Seebeck coefficient, which is the potential difference ΔV per unit temperature difference ΔT along the sample and is expressed as – ΔV/ΔT, σ is the electrical conductivity, and κ the thermal conductivity. κ consists of two major parts: κl, the lattice or phonon contribution, and κe, the contribution from heat-carrying charge carriers. κe is related to σ according to the Wiedemann– Franz law: κe = L0Tσ, in which L0 is the Lorenz number (2.45 × 10–8 V2 K–2 for metals and degenerate semiconductors).[13] Thus, κ = κl + κe. To enhance ZT, TE materials should have a large Seebeck coefficient, a high σ, and small κ. The value of σ can be increased by increasing the doping concentration. However, by doing so, the Seebeck coefficient will normally decrease,[14] and an increase in σ leads to an increase in κe.[15] Thus, it is a challenge to balance all the interdependent properties to increase the ZT value of TE materials. The development of nanotechnology has led to improvements in ZT values, for example, from bulk values of ZT = 1 to beyond ZT = 2 for 2D quantum wells and 1D quantum wires, mainly due to the reduced mean free path of heat-carrying acoustic waves.[16–18] However, some of these nanostructures demand very expensive ultra-high-quality thin films fabricated through molecular beam epitaxy or chemical vapor deposition and are not scalable for commercial production.[19] Another increasingly popular method for improving ZT is based on the formation of nanostructure by ball-milling. For example, in ball-

Buoyed by recent progress in theoretical[1] and experimental[2–5] studies of thermoelectric (TE) transport properties, researchers around the world are giving more attention to TE materials. These materials have the ability to convert thermal energy into electrical energy or to develop a temperature difference between two ends of the material when electrical energy is supplied. In addition, this energy conversion is very reliable.[6] In spite of considerable efforts to improve the economic and environmental performance of automotives, 55 % of the energy released in the internal combustion engine will be carried away by the exhaust gas.[7] If we are able to scavenge thermal energy from the exhaust gas and convert that into electrical energy by using TE materials and use the electrical energy produced to improve vehicle kinetics, the economics of gas consumption could be improved, an improvement that could make a big difference in terms of overall natural gas consumption worldwide. TE materials could also be used to tap into renewable energy such as solar energy[8–10] and to cool high-speed elec[a] Department of Chemistry and Waterloo Institute for Nanotechnology, University of Waterloo, Waterloo, Ontario, N2L 3G1, Canada E-mail: [email protected] http://kleinke.uwaterloo.ca [b] CNRS, Université de Bordeaux, ICMCB, and Bordeaux INP, ICMCB, UPR 9048, 33600 Pessac, France [c] National Institute for Materials Science, Tsukuba, Ibaraki Prefecture 305-0047, Japan Supporting information for this article is available on the WWW under http://dx.doi.org/10.1002/ejic.201501063. Eur. J. Inorg. Chem. 2016, 853–860

853

© 2016 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim

Full Paper milled “BiSbTe alloys”, an increase in ZT was achieved mainly because of reduced lattice thermal conductivity.[20] Another approach is to incorporate chemically and physically inert nanoparticles into bulk TE materials to arrest the coherent propagation of heat-carrying acoustic waves without affecting the electrical conductivity and Seebeck coefficient.[21,22] Another approach, inspired by metallurgical concepts, relies on the control of phase transformation by using thermal treatments to generate in situ a fine distribution of the secondary phase that gives the desired transport properties.[23–27] In pursuit of ZT enhancement, we have chosen bulk Ni0.05Mo3Sb5.4Te1.6 as a suitable candidate for nanocomposite synthesis on the basis of both its high thermal conductivity (4.0 W m–1 K–1) and its ZT of 0.96 at 1000 K.[28,29] Refractory materials such as silicon carbide (SiC, bulk m.p. 2973 K) and alumina (Al2O3, bulk m.p. 2945 K) are very inert at high temperatures. The incorporation of 0.4 % SiC nanoparticles into the Bi0.3Sb1.7Te3 matrix led to an enhancement of the ZT from 1.2 to 1.33 at 373 K, which amounts to an improvement of 11 %.[30] Similarly, adding nano-SiC with a volume fraction of 0.0024 to Bi2Te3, consolidated through spark-plasma sintering (SPS), enhanced the ZT of the bulk by 18 % by contributing to the reduction of thermal conductivity.[31] In addition, SiC nanoparticles have augmented the mechanical properties of some TE materials such as Bi0.5Sb1.5Te3, Bi2Te3, and SiC-whiskers-reinforced alumina.[32,33] Therefore we have studied composites of Ni0.05Mo3Sb5.4Te1.6 with nanoinclusions of β-SiC and Al2O3.

Figure 1. SEM images of hot-pressed Ni0.05Mo3Sb5.4Te1.6/0.034 SiC. (a) Surface with 20–30 μm voids, (b) SiC aggregates on bulk particles, (c) clumped aggregates of SiC, and (d) SiC aggregates between the bulk particles.

Results and Discussion Figure 1 shows the microstructures of a sample of Ni0.05Mo3Sb5.4Te1.6 with a volume fraction, f, of SiC of 0.034 after hot-pressing. The surface shows some voids of 20–30 μm (Figure 1, a). The bulk has a wide range of particle sizes, ranging from 2 to 10 μm (Figure 1, b). The SiC nanoparticles occur in agglomerated form because nanoparticles have the general tendency to agglomerate due to their high surface energy (Figure 1, c,d).[34] It is evident that the process of nanoinclusion needs to be further optimized, noting that we had already sonicated the nanopowder in acetone to separate nanoparticles in an attempt to minimize this issue. TEM micrographs of the sample Ni0.05Mo3Sb5.4Te1.6/0.034SiC are displayed in Figure 2. Part a of Figure 2 shows two different parts of the sample: SiC aggregates of up to 500 nm are visible along with clean grain boundaries, which are preferred over rough boundaries to not impact the electron transport too much. Figure 2 (b–f ) show elemental mappings of silicon, nickel, molybdenum, antimony, and tellurium, all obtained by using the aforementioned TEM in energy-dispersive X-ray spectroscopy (EDX) mode. The distribution of the elements Ni, Mo, Sb, and Te appears to be very homogeneous throughout the sample, which indicates that no element precipitated at the grain boundaries or reacted with SiC, thus leaving the bulk intact. Figure 3 displays the SEM images of the Ni0.05Mo3Sb5.4Te1.6 sample with Al2O3 of f = 0.0325. Figure 3 (a) at a magnification of 100 μm shows a smooth surface. At a higher magnification Eur. J. Inorg. Chem. 2016, 853–860

www.eurjic.org

Figure 2. TEM images of hot-pressed Ni0.05Mo3Sb5.4Te1.6/0.034SiC. (a) SiC aggregates on the surface of bulk particles and (b,c,d,e,f) elemental maps of Si, Ni, Mo, Sb, and Te respectively.

(Figure 3, b,c), voids of 2–5 μm can be seen. In addition, Al2O3 aggregates of around 400 nm can be observed (Figure 3, d). In comparison, the SiC sample has larger voids and a rougher surface than Al2O3.

854


Full Paper Table 2. Thermoelectric properties of Ni0.05Mo3Sb5.4Te1.6 Ni0.05Mo3Sb5.4Te1.6/SiC composites at 325 and 740 K. f –1

–1

σ [Ω cm ] σ0 [Ω–1 cm–1] S [μV K–1] PF[a] [μW cm–1 K–2] κ [W m–1 K–1] κ0 [W m–1 K–1] κl [W m–1 K–1] ZT

and

the

0 (bulk)

0.010

0.020

0.034

1338/876 1585/1036 61/137 4.8/16.5 5.23/4.00 6.20/4.70 4.30/2.88 0.030/0.31

1322/836 1630/1029 60/138 4.8/16.6 4.11/3.30 5.07/4.03 3.16/2.16 0.036/0.38

776/553 1006/717 60/140 2.5/10.1 3.20/2.35 4.15/3.00 2.63/1.61 0.030/0.31

664/458 868/712 63/140 1.9/8.5 2.90/2.31 3.79/3.01 2.44/1.74 0.025/0.27

[a] PF = power factor. Table 3. Thermoelectric properties of Ni0.05Mo3Sb5.4Te1.6 Ni0.05Mo3Sb5.4Te1.6/Al2O3 composites at 325 and 740 K. f –1

Figure 3. SEM images of spark-plasma-sintered Ni0.05Mo3Sb5.4Te1.6/0.034 Al2O3. (a) Surface morphology, (b) voids of 2 μm and Al2O3 aggregates on the bulk particles, (c) voids and bulk particles at higher magnification, and (d) Al2O3 aggregates of 400 nm.

Listed in Table 1 are the surface areas and cumulative pore volumes for different volume fractions of SiC and Al2O3. Comparing the SiC f = 0 and 0.034 samples, there is an increase of 80 % in the surface area and a 280 % increase in the cumulative pore volume for the 0.034 sample. Comparing the Al2O3 f = 0 and 0.0325 samples, 90 and 230 % increases in surface area and cumulative pore volume were observed. These findings are in qualitative agreement with the results of the SEM studies. Similarly, for the SiC f = 0 and 0.034 samples, the relative densities are 94.2 and 90.7 % respectively, whereas for the Al2O3 f = 0 and 0.0325 samples, the relative densities are significantly higher, namely 98.5 and 95.0 %, respectively (see Tables 4 and 5 in the Experimental Section).

–1

σ [Ω cm ] σ0 [Ω–1 cm–1] S [μV K–1] PF[a] [μW cm–1 K–2] κ [W m–1 K–1] κ0 [W m–1 K–1] κl [W m–1 K–1] ZT

0 (bulk)

0.010

0.0216

0.0325

1780/1115 1861/1165 66/139 7.71/21.5 4.84/3.80 5.06/3.80 3.64/2.25 0.050/0.42

1554/1038 1732/1149 55/130 4.8/17.6 4.46/3.47 4.95/3.40 3.36/2.13 0.033/0.40

1583/1027 1780/1155 57/135 5.2/19.0 4.45/3.46 5.01/3.44 3.37/1.90 0.034/0.41

1495/991 1731/1169 62/132 5.7/17.4 4.33/3.45 5.01/3.50 3.23/2.00 0.038/0.37

and

[a] PF = power factor.

Previously we reported p-type carrier concentrations in the order of 4 × 1021 cm–3 for Mo3Sb5.4Te1.6 and Ni0.06Mo3Sb5.4Te1.6.[35] With such a high carrier concentration, the electrical conductivity decreases with increasing temperature, as the mobility decreases, as observed for all eight samples displayed in Figure 4 and is typical of thermoelectric materials. Electrical conductivity values in excess of 103 Ω–1 cm–1 are expected for the Mo3(Sb,Te)7 family of compounds at room temperature.[29,36,37] At 325 K, the σ values decrease with increasing SiC content from 1338 Ω–1 cm–1 (no SiC) to 664 Ω–1 cm–1 (3.4 % SiC). Assuming an error of ±5 %, most of the differences are significant.

Table 1. Surface areas and BJH (Barrett–Joyner–Halenda) cumulative pore volumes. Composite

Surface area [m2 g–1]

Cumulative pore volume [cm3 g–1]

0.58 0.65 0.78 1.05

3.38 × 10–3 4.30 × 10–3 4.80 × 10–3 12.9 × 10–3

0.47 0.49 0.53 0.89

1.84 × 10–3 1.61 × 10–3 1.96 × 10–3 5.06 × 10–3

SiC f f f f

= = = =

0 0.010 0.020 0.034

Figure 4. Temperature dependence of the electrical conductivity of (a) the bulk and SiC and (b) the bulk and Al2O3 composites.

Al2O3 f f f f

= = = =

0 0.010 0.0216 0.0325

The thermoelectric properties of the composite samples are summarized in Tables 2 and 3 at both the minimum and maximum measurement temperatures. They will be discussed below along with their respective data plots. Eur. J. Inorg. Chem. 2016, 853–860

www.eurjic.org

This decrease of 50 % in the value of σ for the SiC f = 0.034 sample can be explained in terms of microstructural changes coupled with charge-carrier scattering from the nanoinclusions. As the volume fraction increases, the degree of aggregation not only increases but also the cumulative pore volume and surface area are enhanced, as the BET data demonstrate. This leads to a huge degradation in mobility due to charge-carrier scattering. For the SiC f = 0 and 0.01 samples, σ tends to follow a T–0.5 dependence across the entire temperature range. However, as

855


Full Paper the content of SiC increases to f = 0.02 and 0.034, this dependence starts to diminish to T–0.2. A similar trend was observed in the hot-pressed CoSb3 bulk with 0 and 0.3 wt.-% ZrO2 (nano). Although both composites have similar relative densities of 95.6 and 95.4 %, there is a decrease in the electrical conductivity of about 72 % for the sample with 0.3 wt.-% ZrO2 (nano) compared with the bulk sample at 325 K.[38] In our previously reported bulk/MWCNT (multiwalled carbon nanotube) composites, we observed that the electrical conductivity follows a T–1.5 dependency, typical of an acoustic-phonon scattering mechanism.[39,40] During consolidation by hot-pressing, the SiC nanoparticles may have suppressed grain growth, thus enhancing the porosity and surface area, and these microstructural features may lead to a change in the charge-carrier scattering mechanism. Also, the similar cumulative pore volumes of these samples may attribute to this deviation. Similarly in La0.75Fe3CoSb12, the samples with 12.4 and 12.8 % porosity showed a drastic reduction in mobility of 47 and 82 %, respectively.[41] The electrical conductivity curves of the Al2O3-containing samples are shown in Figure 4 (b). At 325 K, the bulk sample exhibits σ = 1780 Ω–1 cm–1, and the 3.25 % sample σ = 1495 Ω–1 cm–1. Across the entire temperature range from 325 to 740 K, all the samples follow a T–1.5 dependency, an indication of acoustic-phonon scattering. As the volume fraction increases, the conductivity values converge, falling within the error margin of 5 %. During the SPS process, grain growth can occur due to intense Joule heating, the electrical field diffusion effect, and high pressure. This leads to mass transfer that facilitates smaller particles filling voids, thus reducing porosity.[42] At high temperatures, the σ values start to converge because of the minority carrier effect. The difference in the electrical conductivities of the SiC f = 0 and Al2O3 f = 0 samples at 325 K amounts to 28 %, and at 740 K the difference is still 24 %. This large difference mainly stems from the 98 % density of the spark-plasma-sintered bulk sample of the Al2O3 series compared with the 94 % density of the hot-pressed bulk sample of the SiC series. To analyze the extent of reduction in σ stemming from the different porosity we followed Adachi et al. and applied a correction based on the Maxwell–Eucken equation: σp = σo(1 – p)/ (1 + βp), in which σp and σo are the electrical conductivities with pores and without pores, respectively, p is the porosity (p = 1 – relative density), and β is an empirical value that depends on the shapes of the pores, usually lying between 1 and 3.[43] At 325 K, the pore-corrected electrical conductivities are, by using β = 2 for spherical pores, σo = 1585 Ω–1 cm–1 for 0 % SiC and 868 Ω–1 cm–1 for 3.4 % SiC. Similarly for the Al2O3 series, we obtained σo = 1861 Ω–1 cm–1 for 0 % Al2O3 and 1731 Ω–1 cm–1 for 3.25 % Al2O3. Thus, the pore effect alone cannot completely account for the reduction of the electrical conductivity with increasing nanoadditions. It has been observed that a low relative density of the nanocomposites should cause poor electrical conductivity because of charge-carrier scattering caused by the increased interface density in the nanocomposites.[44–46] Thus, in comparing both composites, the higher magnitude of reduction in electrical conductivity for the SiC composites is in part due to the higher porosity, which can Eur. J. Inorg. Chem. 2016, 853–860

www.eurjic.org

lead to a decrease in mobility. For example, La0.75Fe3CoSb12 showed a systematic reduction in mobility with respect to the amount of porosity. Four samples with 0.3, 0.5, 12, and 15 % porosity were found to have charge-carrier mobilities of 11.5, 8.2, 6.1, and 2.1 cm2 V –1s–1, respectively.[41] The temperature dependence of the total thermal conductivity, κ, is shown in Figure 5. The total thermal conductivities of the bulk and SiC composites decrease steadily with increasing temperature. At 325 K, the thermal conductivity values are 5.23 W m–1 K–1 for the SiC-free sample and 2.90 W m–1 K–1 for the sample with 3.4 % SiC, and at 740 K, the corresponding κ values are 4.00 and 2.31 W m–1 K–1. We have reported similar temperature dependences for the Ni0.05Mo3Sb5.4Te1.6/C60 and Ni0.05Mo3Sb5.4Te1.6/MWCNT composites.[39,47] When a considerable number of pores exist, as in the SiC series, the pores provide more interfaces and a larger surface area and therefore phonon–phonon and phonon–interface scattering become dominant, thus effectively reducing the thermal conductivity.

Figure 5. Temperature dependence of the total thermal conductivity of (a) the bulk and SiC and (b) the bulk and Al2O3 composites.

At 325 K, the κ values for the Al2O3 f = 0 and 0.0325 samples are 4.84 and 4.33 W m–1 K–1, respectively. All the curves except for the one without Al2O3 are very close to each other over the entire temperature range from 325 to 740 K. At 740 K, all the samples with Al2O3 have approximately the same thermal conductivity of around 3.5 W m–1 K–1. For La0.75Fe3CoSb12 (consolidated by SPS), the sample with 0.3 % porosity has κ = 2.6 W m–1 K–1, whereas the sample with 15 % porosity has κ = 1.48 W m–1 K–1 at 300 K.[41] Analogous to the discussion on electrical conductivity, we also applied the Maxwell–Eucken's equation to the thermal conductivity: κp = κo (1 – p)/(1 + βp), in which κp and κo are the experimentally determined thermal conductivity and the hypothetical, pore-free one, respectively. We obtained pore-corrected thermal conductivities for the SiC f = 0 and 0.034 samples of 6.2 and 3.79 W m–1 K–1, respectively, and for the Al2O3 f = 0 and 0.0325 samples of 5.06 and 5.01 W m–1 K–1, respectively. In conclusion, the changes in measured thermal conductivity with increasing nanoaddition can be fully explained by the increasing porosity in the case of the Al2O3 additions, but not in case of the SiC additions. Lattice thermal conductivity, κl, was determined by subtracting κe from the total thermal conductivity, κ; the dependence of κl on temperature is displayed in Figure 6. The values of κe were obtained from the Wiedemann–Franz law: κe = L0σT. We calculated the Lorenz number, L0, from the Seebeck coefficient data assuming our samples follow the single parabolic model

856


Full Paper and mixed scattering mechanisms, that is, alloy scattering and acoustic-phonon scattering, and thus used the scattering parameter λ = 0.[36,48] The results of these calculations are displayed in Figure S1 in the Supporting Information.

The Seebeck coefficient values, S, are positive over the entire temperature range from 325 to 740 K (Figure 7), which indicates that the synthesized compounds are p-type materials, like other members of the Mo3(Sb,Te)7 family. Typically S starts at around 60 μV K–1 and increases almost linearly to 140 μV K–1 at 760 K for samples with an Sb/Te ratio of 5.4:1.6.[29,53] Because S is roughly proportional to T, the diffusive part of the Seebeck coefficient is dominant in these samples.[54]

Figure 6. Temperature dependence of lattice thermal conductivity of (a) the bulk and SiC and (b) the bulk and Al2O3 composites.

Despite the high electrical conductivity, the lattice thermal conductivity, κl, is still the dominant component. A systematic reduction in κl with increasing temperature for both composites indicates that phonon–phonon interactions are dominant, thereby diminishing the thermal conductivity at higher temperatures. For the SiC f = 0 and 0.034 samples, we obtained κl = 4.30 and 2.44 W m–1 K–1, respectively, at 325 K. The reduction in κl from f = 0 to f = 0.034 thus amounts to almost 50 %. This reduction in thermal conductivity is mainly due to the disruption of the coherent propagation of energy-carrying phonons and interphonon interactions. The embedded SiC nanoparticles act as scattering centers and also reduce the cross-sections available for phonons. In the case of the Al2O3 series, the f = 0 and 0.0325 samples exhibit κl = 3.64 and 3.23 W m–1 K–1, respectively, at 325 K. The reduction in κl from f = 0 to f = 0.01 is just 8 %. This decrease is due to the presence of Al2O3 nanoparticles that exert a negative influence on the coherent flow of small wavelength phonons. However, adding more Al2O3 barely reduces κl further, which indicates that high-energy phonons easily cross the boundaries. It has been shown for many composites that incorporating particles with very high thermal conductivity into a bulk matrix of lower thermal conductivity increases the composite's thermal conductivity, depending on the particle size. For example, adding 2 μm diamond particles with κ = 600 W m–1 K–1 to ZnS with κ = 17.4 W m–1 K–1 resulted in an increased thermal conductivity. On the other hand, adding 0.25 μm diamond particles to the same ZnS matrix reduced the thermal conductivity.[49] In our case, κβ-SiC and κAl2O3 are 320 and 30 W m–1 K–1[50] with particle sizes of 45–55 and 13 nm, respectively. As the particle size decreases, the surface area/volume ratio becomes larger, thus increasing the interfacial surface area. An increased surface area reduces the phonon mean free path, aiding the lowering of the thermal conductivity, as observed in case of SiC addition.[51,52] Owing to the small size of Al2O3, one could expect a more significant reduction in thermal conductivity when adding more and more of its nanoparticles. However, all the Al2O3-containing samples exhibit comparable thermal conductivity values. This may be due to the agglomeration of Al2O3 nanoparticles. Eur. J. Inorg. Chem. 2016, 853–860

www.eurjic.org

Figure 7. Temperature dependence of the Seebeck coefficient of (a) the bulk and SiC and (b) the bulk and Al2O3 composites.

In the Al2O3 f = 0 sample, the Seebeck coefficient, S, starts at 66 μV K–1 at 325 K and increases linearly to reach 139 μV K–1 at 740 K, very similar to the SiC f = 0, 0.02, and 0.034 samples. In all the samples with Al2O3, S starts at around 55–60 μV K–1 and increases linearly, similarly to the f = 0 sample, to reach 130–135 μV K–1 at 740 K. Taking into account an estimated error of ±3 %, all the Seebeck curves except for the one without Al2O3 are within the margin of error. The Ni0.05Mo3Sb5.4Te1.6/C60 and Ni0.05Mo3Sb5.4Te1.6/MWCNT composites exhibited the same temperature dependence, and no appreciable change in the Seebeck coefficient was observed with respect to the amount of C60 and CNT.[39,47] In n-type Co0.92Ni0.08Sb2.96Te0.04, the bulk material and samples with 0.5, 1, 2, and 3 mass-% C60 also did not show much difference in the Seebeck coefficient in the entire temperature range from 300 to 870 K.[55] Likewise, adding 4.77 mass-% C60 to CoSb3 had only a minor impact on the Seebeck coefficient.[56] The calculated power factor curves, PF = S2σ, are displayed in Figure 8. From 325 to 740 K, PF increases from around 5 to 16.5 μW cm–1 K–2 with the same slope for the SiC f = 0 and 0.010 samples, similarly to our previous reports on Mo3(Sb,Te)7 compounds.[29] However, the samples with f = 0.020 and 0.034 exhibit significantly lower values throughout the whole temperature range measured.

Figure 8. Temperature dependence of the power factor of (a) the bulk and SiC and (a) the bulk and Al2O3 composites.

857


Full Paper For the Al2O3 f = 0 sample, PF is 7.7 μW cm–1 K–2 at 325 K and increases approximately linearly with temperature reaching 21.5 μW cm–1 K–2 at 740 K. The samples with Al2O3 have significantly lower values (and comparable slopes). Because the thermal conductivities were measured at slightly different temperatures to the electrical properties, fits of the power factor were used to extract PF values corresponding to thermal conductivity measurement temperatures to calculate the figure of merit, ZT. The error margin for ZT was estimated to be 7 % by using the propagation of error method. Typical of Mo3(Sb,Te)7 materials, the ZT values of all the samples investigated here increase rapidly with temperature (Figure 9). From 325 to 740 K, the ZT value for the SiC f = 0.010 sample increases from 0.036 to 0.38, for the SiC f = 0 sample, from 0.030 to 0.31, and for the Al2O3 f = 0 sample, from 0.050 to 0.42. Thus, an improvement of 17 % was observed at 740 K for the 1 % SiC sample compared with its bulk sample. The difference of 35 % between the two bulk samples (f = 0) is quite significant, a consequence of the different consolidation technique, that is, hot-pressing versus spark-plasma sintering.

Figure 9. Temperature dependence of the figure of merit of (a) the bulk and SiC and (b) the bulk and Al2O3 composites.

On the other hand, the ZT values of the Al2O3 f = 0.010, 0.0216, and 0.0325 samples are 0.40, 0.41, and 0.37, respectively, at 740 K, that is, not higher than the bulk sample of this series. In the bulk/MWCNT composite, the 21 % reduction in PF was outweighed by a 40 % reduction in the thermal conductivity, thus leading to a 25 % increase in ZT.[39] It has been suggested that the inclusion of insulating nanoparticles in composites can lead to enhanced ZT, provided that the inclusions do not disrupt electrical conductivity but are scattering phonons.[57] The outcome of dispersing inert oxides in TE materials has been mixed. For example, thermal conductivity was lowered in TE materials containing some inert oxides,[58] but negative effects on the electrical performance have also been reported.[57] These results show that improvements in the TE properties of nanocomposites vary from case to case.

Conclusions We have successfully synthesized and studied the properties of Ni0.05Mo3Sb5.4Te1.6/SiC and Ni0.05Mo3Sb5.4Te1.6/Al2O3 nanocomposites. The addition of 1 % SiC enhanced the thermoelectric figure of merit by 17 % compared with the SiC-free sample. The increased surface area of the SiC f = 0.02 and 0.034 samples resulted in reduced electrical conductivity due to charge-carrier scattering with SiC nanoparticles, pores, and grain boundaries, Eur. J. Inorg. Chem. 2016, 853–860

www.eurjic.org

leaving the Seebeck coefficient unchanged. Also, the enhanced surface area and cumulative pore volume of the SiC samples facilitated interphonon scattering and phonon scattering with grain boundaries, thereby aiding the reduction of the thermal conductivity. The lattice thermal conductivity is still the dominant component in both composites. There is a consistent reduction in lattice thermal conductivity with increasing SiC content over the entire temperature range from 325 to 740 K for the SiC samples. For the Al2O3 f = 0.010, 0.0216, and 0.0325 samples, the magnitude of the reduction of the lattice thermal conductivity compared with the bulk sample is the same across the temperature range of 325 to 740 K, which suggests diffusive phonon scattering due to less effective pores. Concomitant effects of interfacial area, pore volume, and volume fraction of SiC prevail in lowering the phonon mean free path much more efficiently than Al2O3. The larger porosity of the SiC samples also contributes to the phonon scattering. There was no appreciable improvement in the figure of merit for the Al2O3 composites. Post-transport measurement phase analysis of the composites showed no deterioration in the stoichiometry of the constituents in both types of composites. It is worth further investigating the wet chemistry of these systems to coat bulk particles with SiC and Al2O3 nanoparticles homogeneously and reduce thermal conductivity even further. As the figure of merit for the bulk sample consolidated by SPS is 35 % higher than that of the hot-pressed bulk sample, further investigations are needed to optimize the consolidation parameters such as pressure, temperature, and heating rate.

Experimental Section We conducted four solid-state reactions, with 4.6 g of starting material, to obtain enough of the pure phase of Ni0.05Mo3Sb5.4Te1.6. The starting materials were elemental Ni powder (99 % nominal purity, –100+200 mesh), Mo powder (99.95 %, –100 mesh), Sb powder (99.95 %, –100 mesh), and Te powder (99.8 %, –325 mesh), all from Alfa Aesar. These elements were weighed in the stoichiometric ratios inside an Ar-filled glove box and then transferred to silica tubes. These tubes were then evacuated, followed by sealing with an oxygen/hydrogen torch. The as-obtained tubes were heated in a programmable furnace at a rate of 1.5 K/min to 1000 K, then annealed at 1000 K for 10 days, and finally cooled to room temperature at a rate of 1.5 K/min. At this stage, the phase purity of the ground samples was investigated by using Inel's X-ray powder diffractometer with a position-sensitive detector and Cu-Kα1 radiation. The reaction mixtures were reheated using the same temperature profile if noticeable amounts of side-products such as MoTe2 were identified. The SiC nanoparticle powder was purchased from Alfa Aesar, and had a primary particle size of 45–55 nm, according to TEM analysis. Similarly, the Al2O3 nanopowder of 13 nm primary particle size (99.8 % purity) was purchased from Sigma–Aldrich. The bulk Ni0.05Mo3Sb5.4Te1.6 obtained from the four solid-state reactions was thoroughly mixed as follows: All the samples were added to a pestle and mortar and stirred to form one big homogeneous mixture and this mixture was added to a 50 mL vial. The vial was subjected to high-frequency vibrations in a Vortex-Mixture (Fisher-Scientific) for 20 min and then divided into two equally sized heaps. These heaps were again divided into four, of around 2.3 g each, and finally SiC and Al2O3 nanoparticles were added.

858


Full Paper The desired amounts of SiC and Al2O3 nanopowders to give composites with three different volume fractions were suspended in acetone and subjected to sonication to minimize agglomeration during mixing. Small amounts of each suspension were stepwise added to the bulk inside the mortar and stirred until depletion of the whole suspension. In this way, two reference samples without nanoadditions and three samples each with different amounts of SiC and Al2O3 were prepared for a total of eight samples. Consolidation of the SiC composites and its reference sample (“bulk”) was carried out by using a 30 ton Oxy-Gon hot-press. One at a time, all four samples were transferred to a pyrolytic graphite die with a 12.8 mm bore diameter and graphite punches of 12.7 mm in diameter. Hot-pressing was performed by applying 56 MPa of pressure at 925 K for 2 h under argon. This procedure yielded dense disk-shaped pellets with a diameter of 12.7 mm and a thickness of around 2 mm. Their densities were measured by using Archimedes principle and the data are displayed in Table 4.

Figure 10. PXRD patterns of pristine SiC, the composites, and the bulk (from top to bottom).

Table 4. Experimental densities of the Ni0.05Mo3Sb5.4Te1.6/SiC composites at 295 K compared to those of the hypothetical pore-free materials. f 0 (bulk) 0.010 0.020 0.034

Density of pellet, ρ [g cm–3]

Relative density [%]

8.21 7.96 7.93 7.85

94.2 92.8 91.0 90.7

The three Al2O3-containing samples were consolidated by using the spark plasma sintering technique at the National Institute for Materials Science, Japan. The SPS-825 system with 25.5 ton capacity from Fuji Electronic Industrial Co. was used for consolidation. The samples were heated from room temperature to 873 K under 60 MPa pressure, maintained at that temperature for 5 min, and then cooled to room temperature at a rate of 2.25 K/min. Their densities were also determined by using Archimedes' principle (Table 5). Table 5. Experimental densities of the Ni0.05Mo3Sb5.4Te1.6/Al2O3 composites at 295 K compared to those of the hypothetical pore-free materials. f 0 (bulk) 0.010 0.0216 0.0325

Density of pellet, ρ [g cm–3]

Relative density [%]

8.59 8.36 8.27 8.13

98.5 96.5 96.0 95.0

To investigate whether any side-products were formed during SiC inclusion or during consolidation or transport property measurement, we subjected the hot-pressed and sintered samples to powder XRD analysis after all the transport properties had been characterized. Figures 10 and 11 show the XRD patterns of both the SiC and Al2O3 series of composites. No changes were detected. No X-ray peaks representing SiC and Al2O3 appeared in any of the composite samples because of their low content. Similarly, in other composites, Bi0.3Sb1.7Te3 with SiC f = 0.004 and Bi0.5Sb1.5Te3 with SiC f = 0.01, no SiC peaks were reported.[59,60] Microstructural and morphological features were studied by using University of Waterloo's Waterloo Advanced Technology Laboratory (WATLabs)'s Zeiss Ultra|plus SEM coupled with an integrated secondary electron detector in in-lens mode. The micrograin boundaries were studied by TEM using a JEOL 2200 FS microscope equipped with a spherical aberration (Cs) corrector. Electron-transparent specimens were prepared by using Gatan's Precision Ion Polishing System (PIPS) with an operating voltage of 5 kV and a low Eur. J. Inorg. Chem. 2016, 853–860

www.eurjic.org

Figure 11. PXRD patterns of pristine nano-Al2O3, the composites and the bulk (from top to bottom).

incident angle of 8–10° at liquid-nitrogen temperature. Precautions were taken to avoid mechanical and ion-beam damage. Pore volume and interfacial surface area have a large impact on the cross-section available for phonon propagation, and they also affect electrical conductivity. To study these parameters, nitrogen adsorption/desorption measurements were carried out by BET using an Autosorb-1C (Model: AX1C-MP-LP) adsorption analyzer from Quantachrome Instruments. The thermal and electrical properties of the samples were measured between 325 and 740 K. Thermal diffusivity, α, was measured by means of a flash technique[61,62] using an Anter FlashLineTM instrument equipped with a xenon flash lamp to shine light on specimens and an InSb IR detector to monitor the transient rise in temperature; a steady flow of Ar was maintained throughout the experiment inside the furnace to avoid contamination of the samples. Diskshaped pellets with a diameter of 12.7 mm were investigated, prepared by hot-pressing (SiC samples) and spark-plasma sintering (Al2O3 samples). The thermal conductivity, κ, was determined by using κ = αρCP, in which ρ is the density and CP is the specific heat estimated by using the rule of mixtures: (ρCP)composite = (ρCP)nanof + (ρCP)bulk(1 – f), in which f is the volume fraction. To study the electrical properties, the same pellets were cut into quadratic prisms of 11 × 2 × 2 mm using a low-speed diamond saw. The electrical conductivity, σ, and the Seebeck coefficient, S, were measured simultaneously by using four-probe and differential methods, respectively, with the ULVAC-RIKO ZEM-3 system. S was determined from the slope of the thermo-electromotive force, ΔV, versus temperature gradient, ΔT, using three different ΔT values, and an aver-

859


Full Paper age was taken. An inert atmosphere was maintained inside the furnace by using a helium pressure of –0.09 MPa during the measurements.

[29] [30]

Acknowledgments The authors greatfully acknowledge Quan Pang and Chun Yuen Kwok for their help with the BET measurements. Financial support from the Natural Sciences and Engineering Research Council of Canada (NSERC) is appreciated.

[31] [32] [33] [34] [35]

Keywords: Nanostructures · Nanocomposites · Thermoelectric

[36] [37]

properties · Antimony · Tellurium

[38] [1] [2] [3] [4] [5] [6] [7] [8] [9]

[10] [11] [12] [13] [14] [15] [16] [17] [18] [19] [20] [21] [22] [23] [24] [25] [26] [27] [28]

L. Hicks, M. Dresselhaus, Phys. Rev. B 1993, 47, 12727. G. J. Snyder, E. S. Toberer, Nat. Mater. 2008, 7, 105–114. H. Kleinke, Chem. Mater. 2010, 22, 604–611. M. G. Kanatzidis, Chem. Mater. 2009, 22, 648–659. H. Ohta, K. Sugiura, K. Koumoto, Inorg. Chem. 2008, 47, 8429–8436. L. E. Bell, Science 2008, 321, 1457–1461. G. Bourhis, P. Leduc, Oil Gas Sci. Technol. 2010, 65, 39–46. M. Xie, D. M. Gruen, J. Phys. Chem. B 2010, 114, 14339–14342. J. Baxter, Z. Bian, G. Chen, D. Danielson, M. S. Dresselhaus, A. G. Fedorov, T. S. Fisher, C. W. Jones, E. Maginn, U. Kortshagen, Energy Environ. Sci. 2009, 2, 559–588. D. Kraemer, B. Poudel, H. P. Feng, J. C. Caylor, B. Yu, X. Yan, Y. Ma, X. Wang, D. Wang, A. Muto, Nat. Mater. 2011, 10, 532–538. G. Min, D. M. Rowe, Solid-State Electron. 1999, 43, 923–929. E. Altenkirch, Phys. Z. 1909, 10, 560–580. A. H. Wilson, The Theory of Metals, Cambridge University Press, Cambridge, 1953. N. F. Mott, H. Jones, The Theory of the Properties of Metals and Alloys, Dover Publications, Mineola, NY, 1936. L. Lorenz, Ann. Phys. Chem. 1872, 147, 429–451. T. Harman, P. Taylor, M. Walsh, B. LaForge, Science 2002, 297, 2229–2232. R. Venkatasubramanian, E. Siivola, T. Colpitts, B. O'quinn, Nature 2001, 413, 597–602. P. Pichanusakorn, P. Bandaru, Mater. Sci. Eng. R 2010, 67, 19–63. S. M. Lee, D. G. Cahill, R. Venkatasubramanian, Appl. Phys. Lett. 1997, 70, 2957–2959. B. Poudel, Q. Hao, Y. Ma, Y. Lan, A. Minnich, B. Yu, X. Yan, D. Wang, A. Muto, D. Vashaee, Science 2008, 320, 634–638. D. Rowe, C. Bhandari, Appl. Energy 1980, 6, 347–351. C. J. Vineis, A. Shakouri, A. Majumdar, M. G. Kanatzidis, Adv. Mater. 2010, 22, 3970–3980. S. Gorsse, P. B. Pereira, R. Decourt, E. Sellier, Chem. Mater. 2010, 22, 988– 993. S. Gorsse, P. Bellanger, Y. Brechet, E. Sellier, A. Umarji, U. Ail, R. Decourt, Acta Mater. 2011, 59, 7425–7437. N. A. Heinz, T. Ikeda, Y. Pei, G. J. Snyder, Adv. Funct. Mater. 2014, 24, 2135–2153. U. Ail, S. Gorsse, S. Perumal, M. Prakasam, A. Umarji, S. Vivès, P. Bellanger, R. Decourt, J. Mater. Sci. 2015, 50, 6713–6718. P. Bellanger, S. Gorsse, G. Bernard-Granger, C. Navone, A. Redjaimia, S. Vivès, Acta Mater. 2015, 95, 102–110. H. Xu, K. M. Kleinke, T. Holgate, H. Zhang, Z. Su, T. M. Tritt, H. Kleinke, J. Appl. Phys. 2009, 105, 053703/1-053703/5.CCDC 053703 (for XX1), and 053703 (for XX2) contain the supplementary crystallographic data for

Eur. J. Inorg. Chem. 2016, 853–860

www.eurjic.org

[39] [40] [41] [42] [43] [44] [45] [46] [47]

[48] [49] [50] [51] [52] [53] [54] [55] [56] [57] [58] [59] [60] [61] [62]

this paper. These data can be obtained free of charge from The Cambridge Crystallographic Data Centre. H. Xu, K. M. Kleinke, T. Holgate, D. Rossouw, G. Botton, T. M. Tritt, H. Kleinke, J. Alloys Compd. 2010, 504, 314–319. J. Li, Q. Tan, J.-F. Li, D.-W. Liu, F. Li, Z.-Y. Li, M. Zou, K. Wang, Adv. Funct. Mater. 2013, 23, 4317–4323. J.-F. Li, J. Liu, Phys. Status Solidi A 2006, 203, 3768–3773. D. W. Liu, J. F. Li, C. Chen, B. P. Zhang, L. Li, J. Micromech. Microeng. 2010, 20, 125031. P. F. Becher, G. C. Wei, J. Am. Ceram. Soc. 1984, 67, C267–C269. C. B. Murray, C. Kagan, M. Bawendi, Annu. Rev. Mater. Sci. 2000, 30, 545– 610. H. Zhang, J. He, B. Zhang, Z. Su, T. M. Tritt, N. Soheilnia, H. Kleinke, J. Electron. Mater. 2007, 36, 727–731. Q. Guo, H. Kleinke, J. Solid State Chem. 2014, 215, 253–259. X. Shi, Y. Pei, G. J. Snyder, L. Chen, Energy Environ. Sci. 2011, 4, 4086– 4095. C. Stiewe, Z. He, D. Platzek, G. Karpinski, E. Müller, S. Li, M. Toprak, M. Muhammed, Materialwiss. Werkstofftech. 2007, 38, 773–776. N. Nandihalli, S. Gorsse, H. Kleinke, J. Solid State Chem. 2015, 226, 164– 169. H. Muta, K. Kurosaki, S. Yamanaka, J. Alloys Compd. 2003, 350, 292–295. L. Yang, J. Wu, L. Zhang, J. Alloys Compd. 2004, 364, 83–88. M. Omori, Mater. Sci. Eng. A 2000, 287, 183–188. J. Adachi, K. Kurosaki, M. Uno, S. Yamanaka, J. Alloys Compd. 2007, 432, 7–10. J. Martin, G. Nolas, W. Zhang, L. Chen, Appl. Phys. Lett. 2007, 90, 222112. X. Fan, J. Yang, Z. Xie, K. Li, W. Zhu, X. Duan, C. Xiao, Q. Zhang, J. Phys. D 2007, 40, 5975. N. Gothard, T. Tritt, J. Spowart, J. Appl. Phys. 2011, 110, 023706. N. Nandihalli, A. Lahwal, D. Thompson, T. C. Holgate, T. M. Tritt, V. Dassylva-Raymond, L. I. Kiss, E. Sellier, S. Gorsse, H. Kleinke, J. Solid State Chem. 2013, 203, 25–30. V. I. Fistul, J. Blakemore, Heavily doped semiconductors, Plenum Press, New York, 1969. A. Every, Y. Tzou, D. Hasselman, R. Raj, Acta Metall. Mater. 1992, 40, 123– 129. H. O. Pierson, Handbook of Refractory Carbides & Nitrides: Properties, Characteristics, Processing and Apps, Noyes Publications, Westwood, NJ, 1996. Y. Wang, K. Fujinami, R. Zhang, C. Wan, N. Wang, Y. Ba, K. Koumoto, Appl. Phys. Express 2010, 3, 031101. M.-S. Jeng, R. Yang, D. Song, G. Chen, J. Heat Transfer 2008, 130, 042410. F. Gascoin, J. Rasmussen, G. J. Snyder, J. Alloys Compd. 2007, 427, 324– 329. R. D. Barnard, Thermoelectricity in Metals and Alloys, Taylor & Francis, Ltd., London, 1972. T. Itoh, K. Ishikawa, A. Okada, J. Mater. Res. 2007, 22, 249–253. X. Shi, L. Chen, J. Yang, G. Meisner, Appl. Phys. Lett. 2004, 84, 2301–2303. N. Scoville, C. Bajgar, J. Rolfe, J.-P. Fleurial, J. Vandersande, Nanostruct. Mater. 1995, 5, 207–223. F. Brochin, B. Lenoir, X. Devaux, R. Martin-Lopez, H. Scherrer, J. Appl. Phys. 2000, 88, 3269–3275. J. Li, Q. Tan, J. F. Li, D. W. Liu, F. Li, Z. Y. Li, M. Zou, K. Wang, Adv. Funct. Mater. 2013, 23, 4317–4323. D. W. Liu, J. F. Li, C. Chen, B. P. Zhang, J. Electron. Mater. 2011, 40, 992– 998. W. Parker, R. Jenkins, C. Butler, G. Abbott, J. Appl. Phys. 1961, 32, 1679– 1684. R. D. Cowan, J. Appl. Phys. 1963, 34, 926–927.

Received: September 16, 2015 Published Online: February 3, 2016

860
