Synthesis and thermoelectric properties of Nd-single filled p-type skutterudites
Wu Hong1, 2, Shaheen Nusrat1, Yang Heng-Quan1, Peng Kun-Ling1, 2, Shen Xing-Chen1, 2, Wang Guo-Yu2, Lu Xu1, Zhou Xiao-Yuan1, †
Department of Applied Physics, Chongqing University, Chongqing 400044, China
Chongqing Institute of Green and Intelligent Technology, Chinese Academy of Sciences, Chongqing 400714, China

 

† Corresponding author. E-mail: xiaoyuan2013@cqu.edu.cn

Abstract

We report the synthesis of Nd-filled and Fe substituted p-type NdxFe3.2Co0.8Sb12 (x = 0.5, 0.6, 0.7, 0.8, and 0.9) skutterudites by the solid-state reaction method. The influences of Nd filler on the electrical and thermal transport properties are investigated in a temperature range from room temperature to 850 K. A lowest lattice thermal conductivity of is obtained in Nd0.8Fe3.2Co0.8Sb12 at 673 K, which results from the localized vibration modes of fillers and the increase of grains boundaries. Meanwhile, the maximum power factor is for the Nd0.9Fe3.2Co0.8Sb12 sample at 668 K. Overall, the highest dimensionless figure of merit zT = 0.87 is achieved at 714 K for Nd0.9Fe3.2Co0.8Sb12.

1. Introduction

A great deal of effort has been devoted to developing highly efficient thermoelectric (TE) materials, which can convert waste heat into useful electricity.[1] The thermoelectric performance of a material is evaluated by the dimensionless figure of merit ( , where S, T, σ, and κ are the Seebeck coefficient, absolute temperature, electrical conductivity, and thermal conductivity, respectively.[2] The skutterudites based thermoelectric materials are of particular interest since they possess high thermoelectric performance, good thermal stability and excellent mechanical properties. They are therefore considered as a promising candidate for waste heat recovery for automobiles. The thermoelectric performance of pure skutterudite is limited by its rather high thermal conductivity, primarily the lattice part.[3] The skutterudite is represented as a general formula with a body centered cubic phase structure, where M (e.g., Co, Rh, and Fe) refers to the transition metal and X represents the pnictogen atom (e.g., As, P, and Sb). There are two voids per unit cell with sufficient size, which can be easily accommodated with other metal atoms, then the general formula for the compound becomes , where R is the filler metal atom, which is weakly bounded with the surrounding atoms in the matrix.[4] Slack et al. proposed that the lattice thermal conductivity could be greatly reduced when the filler atoms are located in the voids of skutterudites, which is due to the strong phonon scattering caused by localized vibration modes of filler atoms.[5,6] In the last decade, the strategies for reducing the thermal conductivity of skutterudites have involved the manipulation of filling atoms with the different types of fraction modes and vibration modes.[7]

However, most of the work has focused on the n-type skutterudites, which shows high zT values over 1.7.[8] The highest zT value of the p-type skutterudite is much lower than that of the n-type one. From a practical point of view, both n-type and p-type are equally important for power generation devices. Up to now, enhancing the thermoelectric performance of the p-type skutterudites has been a challenging task. Typically, the p-type skutterudite can be attained by the partial substitution of Fe for Co in a pure CoSb3 skutterudite compound, where Fe atoms have less valence electrons than Co atoms. Usually, the Fe-to-Co ratio can be tuned on the transition metal site together with void filling, while the Seebeck coefficient can be enlarged by pushing the Fermi level into the heavier transition metal d-bands. However, the large effective mass of holes in skutterudite and stronger alloy scattering after Fe substitution deteriorate the carrier mobility significantly, thereby causing a diminished power factor, which is about half the value of the corresponding n-type skutterudite. In addition, the bipolar conduction caused by electron-hole pairs moving across the band gap at high temperatures leads to a higher thermal conductivity. These factors are hindering improvement to the thermoelectric performance in p-type skutterudite.[9]

Many researchers have tried to fill the voids with different rattlers. Tang et al. reported that the single phase barium-filled p-type skutterudite compound and achieved a zT value of 0.9 at 750 K in Ba0.27Fe0.98Co3.02Sb12.[10] Tan et al. studied the Ce-filled p-type skutterudites and achieved the highest zT value of 1.0 at 800 K in Ce1.1Fe4Sb12.[12] Zhu et al. investigated the double filled Yb–La and attained the highest zT value of 1.0 at 700 K in Yb0.25La0.60Fe2.7Co1.3Sb12.[13,14] Herein, we report the systematic study on the thermoelectric performance of Nd-filled p-type skutterudites with compositions of NdxFe3.2Co0.8Sb12 (x = 0.5, 0.6, 0.7, 0.8, and 0.9) synthesized by the traditional method of melting-annealing and further followed by a spark plasma sintering process. It is observed that the electrical conductivity decreases while the Seebeck coefficient increases with increasing the Nd filling fraction. The thermal conductivity is reduced by the Nd fillers through the rattling effect along with the nanostructured precipitate through strengthened phonon scattering. A maximum zT value of 0.87 is achieved for Nd0.9Fe3.2Co0.8Sb12 at 714 K due to its high power factor and low thermal conductivity.

2. Experimental details

Nd-filled p-type skutterudites with compositions of NdxFe3.2Co0.8Sb12 (x = 0.5, 0.6, 0.7, 0.8, and 0.9) were synthesized by the solid-state reaction method. In a typical process, high-purity elemental materials of Nd (ingot 99.99%), Fe (pieces 99.99 %), Co (slugs and foils 99.99%), and Sb (chunk 99.99%) were mixed according to the stoichiometric ratios and placed in carbon coated quartz tubes inside a glove box. The sealed quarts tubes were heated up to 1373 K with a ramp rate of , maintained for 27 h and then quenched in saturated salt water. The resulting chunks were ground into powder samples and they were cold-pressed into cylinders with a diameter of 10 mm. These cylinders were sealed in tubes again for further annealing at 923 K for 7 days. The annealed ingots were ground into fine powders and loaded into graphite dies with an inner-diameter of 10 mm for spark plasma sintering (SPS-625) at 833 K for 11 min under a pressure of 45 MPa. The sintered samples were cut and polished into bars (2 mm×3 mm×10 mm) for electrical transport property measurement and disks (10 mm in diameter and 2 mm in thickness) for thermal conductivity measurement, respectively.

The phase structures and purities of the as-annealed ingots were characterized by a powder x-ray diffraction (XRD) apparatus (PANalytical X’pert powder with Cu Kα radiations). The morphologies and homogeneities of the samples were investigated by the field emission scanning electron microscopy (JSM-7800F, JEOL). The electrical conductivity and Seebeck coefficient were measured by the static direct-current method using a commercial equipment LSR-3. The total thermal conductivities of the samples ranging from 300 K–850 K were calculated from the relationship , where λ is the thermal diffusivity measured using the laser flash method (LFA 457) and is the specific heat measured using the Netzsch 404 F3. The densities of the samples were obtained using the Archimedes method, and relative densities are higher than 95% of the theoretical value. Room temperature carrier concentration was determined using a home-made Hall measurement system with the magnetic field of ±1 T.

3. Results and discussion

The XRD patterns of the NdxFe3.2Co0.8Sb12 (x = 0.5, 0.6, 0.7, 0.8, and 0.9) are presented in Fig. 1(a). All the peaks are well matched with the standard reference data of the JCPDF-65-8978 card with skutterudites phase except that a trace of Sb impurity phase is found at x = 0.9. All samples remain in the body centered cubic structure with the space group of Im-3. In addition, the lattice parameters for each sample are shown in Fig. 1(b). The lattice parameter slightly increases with filler (Nd) content increasing (see Fig. 1(b)), which can enlarge the icosahedron in the unit cell.

Fig. 1. (color online) (a) XRD patterns and (b) lattice parameters of NdxFe3.2Co0.8Sb12 (x = 0.5, 0.6, 0.7, 0.8, and 0.9) as-sintered samples.

The SEM images of the as-sintered bulk of Nd0.9Fe3.2Co0.8Sb12 sample are shown in Fig. 2. From the SEM images, we observe that the crystallized grains are closely packed, which indicates the high density of bulk material. The majority of the grain sizes range from hundreds of nanometers to a few micrometers. Also, there exist many nanoparticles in the matrix as shown in Fig. 2(b). The non-uniform distribution in grain sizes and nanoparticles can help to scatter phonons with different mean free paths. Also, the tight junction feature of crystallized grains is beneficial to charge carrier transportation.

Fig. 2. (color online) [(a) and (b)] SEM images of Nd0.9Fe3.2Co0.8Sb12 and [(c) and (d)] Nd0.7Fe3.2Co0.8Sb12 as-sintered sample with different resolutions.

The temperature-dependent electrical properties of the NdxFe3.2Co0.8Sb12 (x = 0.5, 0.6, 0.7, 0.8, and 0.9) filled skutterudites are displayed in Fig. 3(a). With increasing Nd concentration, the electrical conductivity decreases due to the reduction in the carrier concentration, i.e., the carrier concentration for the x = 0.5 sample is , which is reduced to for the x = 0.9 sample as listed in Table 1. The carrier concentration decreases with increasing Nd content because the electrons are donated from the filler atoms. The Nd0.5Fe3.2Co0.8Sb12 sample shows a typical highly-degenerate semiconductor behavior, where the electrical conductivity gradually decreases with increasing temperature over the whole temperature range. However, the other samples show an increase trend in electrical conductivity at high temperatures ( ), which may be attributed to the bipolar effect. Generally, with increasing temperature, the electrical conductivity decreases, which is attributed to stronger electron scattering at higher temperatures. Also with further increasing temperature, the decrease in electrical conductivity is comparatively slow. This is probably due to the gradually increased carrier concentration from thermal excitation or the variation in the carrier scattering mechanism.[15]

Fig. 3. (color online) Temperature-dependent electrical properties for NdxFe3.2Co0.8Sb12 (x = 0.5, 0.6, 0.7, 0.8, and 0.9) samples, showing (a) electrical conductivities; (b) Seebeck coefficients, and (c) power factors.
Table 1.

Room temperature transport properties of the NdxFe3.2Co0.8Sb12 (x = 0.5, 0.6, 0.7, 0.8, and 0.9) skutterudites.

.

Figure 3(b) shows the temperature dependence of the Seebeck coefficient (S) of NdxFe3.2Co0.8Sb12 (x = 0.5, 0.6, 0.7, 0.8, and 0.9) filled skutterudites. The sample has a positive sign of Seebeck coefficient, which represents that the majority of charge carriers are holes. In the entire temperature range, the absolute value of the Seebeck coefficient first increases with increasing temperature, and then it starts to decrease due to the presence of an intrinsic semiconducting behavior at high temperature.[16] The Seebeck coefficient of p-type NdxFe3.2Co0.8Sb12 is enhanced with increasing concentration of Nd, and the Seebeck coefficient for Nd0.9Fe3.2Co0.8Sb12 reaches a maximum value of at 670 K. Figure 3(c) demonstrates the temperature-dependent power factor (PF) of NdxFe3.2Co0.8Sb12 (x = 0.5, 0.6, 0.7, 0.8, and 0.9) filled skutterudites. The maximum power factor reaches up to for the Nd0.9Fe3.2Co0.8Sb12 sample at 690 K, mainly resulting from the high Seebeck coefficient.

The temperature-dependent total thermal conductivities of NdxFe3.2Co0.8Sb12 (x = 0.5, 0.6, 0.7, 0.8, and 0.9) filled skutterudites are shown in Fig. 4(a). The total thermal conductivity can be divided into the lattice thermal conductivity ( ) and the electronic thermal conductivity ( ), i.e., . The electronic component of the thermal conductivity ( ) is due to the heat transported by electrons while the lattice component of the thermal conductivity ( ) arises from the heat transported by the phonons. Here can be calculated from the Weidman–Franz law: , where the Lorenz number is selected as the metallic limit value of .[17] It is observed that the electrical thermal conductivity ( ) accounts for about 50% portion of the total thermal conductivity due to the high electrical conductivity. The calculated lattice thermal conductivities of all samples are displayed in Fig. 4(b). It is clearly seen that the thermal conductivity decreases after Nd filling because of the strong phonon scattering from the vibrational modes of the filling atoms. It is not surprising that the strategy of filling atoms is an effective way to reduce the lattice thermal conductivity, thereby greatly suppressing the total thermal conductivity. As a result, the lowest thermal conductivity is found to be in Nd0.7Fe3.2Co0.8Sb12. Here the lattice thermal conductivities of all Nd-filled samples are less than in a temperature range from room temperature to 650 K. In general, the filling fraction limitation is determined by the electronegativity, ionic radius and valance states. As is well known, the lattice thermal conductivity follows the relation when the Umkklap phonon scattering process dominates and decreases with the increasing filling fraction.[18] It cannot be ignored that the grain bounties and nanoparticles displayed in the SEM images still scatter phonons with the wavelengths ranging from the middle to long wavelength at high temperature. The lowest lattice thermal conductivity is found to be at x = 0.8. The low lattice thermal conductivity can be mainly attributed to phonon–phonon scattering assisted with point defect Fe/Co with mass and strain fluctuation, boundary scattering particularly. However, the lattice thermal conductivity of Nd0.9Fe3.2Co0.8Sb12 increases suddenly due to the Sb impurity phase with high thermal conductivity. The phenomenon is common in other TE materials containing second phases, such as the SnO2 impurity phase with high in SnSe polycrystal.[19]

Fig. 4. (color online) Temperature-dependent thermal conductivities for NdxFe3.2Co0.8Sb12 (x = 0.5, 0.6, 0.7, 0.8, and 0.9) samples, showing (a) thermal conductivities and (b) lattice thermal conductivities.

Figure 5 shows the temperature-dependent figure-of-merit (zT) calculated from the electrical conductivity, Seebeck coefficient, and thermal conductivity. The maximum zT value is found to be 0.87 in Nd0.9Fe3.2Co0.8Sb12 at 690 K as a result of the high power factor and low thermal conductivity as compared with those in other samples. The zT values of all samples increase with temperature and attain a maximum value in a temperature range of 665 K–690 K. As shown in Fig. 6, the zT value of our work is comparable to the previously reported values for the single filled skutterudites. Therefore, we conclude that the zT value of the p-type filled skutterudites could be improved through optimizing the carrier concentration and reducing the thermal conductivity by filling Nd and Fe for Co.

Fig. 5. (color online) Figure-of-merits for NdxFe3.2Co0.8Sb12 (x = 0.5, 0.6, 0.7, 0.8, and 0.9) samples.
Fig. 6. (color online) Plots of the maximum zT versus the filling fraction for YbyFexCo4−xSb12,[20] NdyFe3Co1Sb12,[21] BayFe3Co1Sb12, YbyFe2.7Co1.3Sb12, LayFe2.7Co1.3Sb12, (Ce1−xYbx)yFe3CoSb12,[22] (YbxLa1−x)yFe2.7Co1.3Sb12,[23] and this work (NdyFe3.2Co0.8Sb12).
4. Conclusions

Nd-filled p-type skutterudites NdxFe3.2Co0.8Sb12 (x = 0.5, 0.6, 0.7, 0.8, and 0.9) have been successfully synthesized by the solid-state reaction method, and their TE transport properties are investigated. Using Nd as a filler in skutterudites, the lowest value of lattice thermal conductivity, at 630 K, is attained in the Nd0.8Fe3.2Co0.8Sb12 sample. The electrical conductivity of the samples decreases with increasing Nd concentration, which results from the diminished carrier concentration. A maximum zT value of 0.87 is achieved in Nd0.9Fe3.2Co0.8Sb12 at 690 K due to its high power factor and low thermal conductivity.

Reference
[1] Skutterudites Uher C 2001 Semiconductors and semimetals 69 139
[2] Slack G A Tsoukala V G 1994 J. Appl. Phys. 76 1665
[3] Zhao L D Dravid V P Kanatzidis M G 2014 Energy Environ. Sci. 7 251
[4] Sales B Mandrus D Williams R K 1996 Science 272 1325
[5] Bell L E 2008 Science 321 1457
[6] Morelli D T Meisner G P Chen B Hu S Uher C 1997 Phys. Rev. 56 7376
[7] Sales B Mandrus D Chakoumakos B Keppens V Thompson J 1997 Phys. Rev. 56 15081
[8] Rogl G Grytsiv A Bauer E Rogl P Zehetbauer M 2010 Intermetallics 18 57
[9] Nordstrom L Singh D J 1996 Phys. Rev. 53 1103
[10] Tang X Chen L Goto T Hirai T Yuan R 2002 J. Mater. Res. 17 2953
[11] Zhou C Morelli D Zhou X Wang G Uher C 2011 Intermetallics 19 1390
[12] Tan G Zheng Y Tang X 2013 Appl. Phys. Lett. 103 183904
[13] Zhou L Qiu P Uher C Shi X Chen L D 2013 Intermetallics 32 209
[14] Chandra M R Mueller E Kim I H 2012 J. Appl. Phys. 111 023708
[15] Duan B Zhai P Liu L Zhang Q Ruan X 2012 J. Mater. Sci.: Materials in Electronics 23 1817
[16] Qiu P Liu R Yang J Shi X Huang X Zhang W 2012 J. Appl. Phys. 111 023705
[17] Xu C Duan B Ding S Zhai P Li P 2013 Physica B: Conden Matter 425 34
[18] Sales B Mandrus D Chakoumakos B Keppens V Thompson A 1997 Phys. Rev. 56 15081
[19] Peng K L Wu H Yan Y C Guo L J Wang G Y Lu X Zhou X Y 2017 J. Mater. Chem. 5 14053
[20] Liu R H Qiu P F Chen X H Huang X Y Chen L D 2011 J. Mater. Res. 26 1813
[21] Guo L J Cai Z W Xu X L Peng K L Wang G W Wang G Y Zhou X Y 2015 J. Nanosci. Nanotechnol. 5 1
[22] Liu R H Yang J Chen X H Shi X Chen L D Uher C 2011 Intermetallics 19 1747
[23] Liu R H Yang J Chen X H Shi X Chen L D Uher C 2013 Intermetallics 32 209