Simple and efficient synthesis of nanograin structured single phase filled skutterudite for high thermoelectric performance

Acta Materialia 142 (2018) 8e17

Contents lists available at ScienceDirect

Acta Materialia journal homepage: www.elsevier.com/locate/actamat

Full length article

Simple and efficient synthesis of nanograin structured single phase filled skutterudite for high thermoelectric performance Sanghoon Lee a, 1, Kyu Hyoung Lee b, 1, Young-Min Kim a, e, Hyun Sik Kim c, d, G. Jeffrey Snyder d, Seunghyun Baik e, f, **, Sung Wng Kim a, * a

Department of Energy Science, Sungkyunkwan University, Suwon, 16419, Republic of Korea Department of Nano Applied Engineering, Kangwon National University, Chuncheon, 24341, Republic of Korea Materials R&D Center, Samsung Advanced Institute of Technology, Suwon, 16678, Republic of Korea d Department of Materials Science and Engineering, Northwester University, Evanston, IL 60208, USA e Center for Integrated Nanostructure Physics, Institute for Basic Science (IBS), Suwon, 16419, Republic of Korea f School of Mechanical Engineering, Sungkyunkwan University (SKKU), Suwon, 16419, Republic of Korea b c

a r t i c l e i n f o

a b s t r a c t

Article history: Received 21 June 2017 Received in revised form 22 August 2017 Accepted 20 September 2017 Available online 21 September 2017

Filled skutterudites are promising mid-to-high temperature range thermoelectric materials for power generation, however, a traditional melt-solidification process followed by annealing (TMA) and powder metallurgical sintering requires a long processing time more than 10 days to ensure the structural and compositional homogeniety of materials with a high thermoelectric conversion efficiency zT. To address this, we herein report a simple and efficient synthesis of high-performance n- and p-type filled skutterudites that successfully produces a complete single phase from single to multiple filled materials in a day. The nanograin (~440 nm) structured bulks are prepared from the combined process of temperatureregulated melt spinning (MS) using ingots and short-time spark plasma sintering (SPS). The controlled phase evolution and transformation by adjusting rapid solidification and densification conditions are demonstrated by a comprehensive analysis including structure refinement and atomic-scale observation, verifying the desired occupancy and random distribution of filling elements, respectively. The maximum zT values of filled skutterudites fabricated here were 1.48 ± 0.17 at 800 K for n-type In0.12Yb0.20Co4.00Sb11.84 and 1.15 ± 0.13 at 750 K for p-type Ce0.91Fe3.40Co0.59Sb12.14, which are comparable to the highest zT values reported for filled skutterudites fabricated by TMA-based processes. Superior reproducibility achieved in shortened processing time enables the present synthetic process to be utilized for commercial manufacturing process that can be readily applied to massive production of bulk filled skutterudites for high-performance thermoelectric power generators. © 2017 Acta Materialia Inc. Published by Elsevier Ltd. All rights reserved.

Keywords: Thermoelectric Multiple filled skutterudites Rattling Filling fraction Melt-solidification process

1. Introduction Filled skutterudites with a chemical composition RxT4Pn12 (R ¼ rare earth, actinide, alkaline-earth, or alkali metal, T ¼ transition metal, Pn ¼ pnictogen atom) are compounds with cubic lattice in which R atoms (fillers or rattlers) fill in voids (nanosized cages) at the body-centered position [1]. Among them,

* Corresponding author. ** Corresponding author. Center for Integrated Nanostructure Physics, Institute for Basic Science (IBS), Suwon, 16419, Republic of Korea. E-mail addresses: [email protected] (S. Baik), [email protected] (S.W. Kim). 1 These authors contributed equally to this work. https://doi.org/10.1016/j.actamat.2017.09.044 1359-6454/© 2017 Acta Materialia Inc. Published by Elsevier Ltd. All rights reserved.

filled skutterudite CoSb3- and FeSb3-based alloys are potential thermoelectric materials especially for mid-to-high temperature power generation applications such as automotive thermoelectric generators (ATEGs) due to their high thermoelectric figure of merit, zT (¼ S2sT/ktot, where S is Seebeck coefficient, s is electrical conductivity, and ktot is total thermal conductivity at a given absolute temperature T), originated from extremely low lattice thermal conductivity (klat) due to the rattling effect of fillers [24]. It has been regarded that the rattling effect by multiple fillers can be more effective for the further reduction of klat due to the resonant phonon scattering [46]. Traditional melt-solidification (MS) processes generally produce large grain structured ingots with a complex phase formation behavior, in which the Co-Sb and Fe-Sb binary systems with a

S. Lee et al. / Acta Materialia 142 (2018) 8e17

peritectic phase transformations solidify into various secondary phases including constituent elements and binary alloys such as CoSb2 (or FeSb2), RSb, and RSb2 [711]. It is thus generally accepted that the long-time annealing process is indispensable to ensure homogeneity in composition and structure, especially for multiple filled skutterudites [4,8,9]. Indeed, a very high zT value of 1.7 at 850 K in n-type CoSb3-based multiple filled skutterudites was obtained by melt-solidification and subsequent 7 days annealing at 750  C [4]. In these considerations, the long-time melt-solidification-based fabrication process has been regarded as one of the main obstacles for cost-effective practical thermoelectric device applications. In search for a simple and efficient method with shortened processing time, several approaches using ingots fabricated by traditional melt-solidification (TMA) processes have been developed [1012], in which key aims are to reduce the grain size to nanoscale and to obtain a homogenous dispersion of various nanoscale secondary phases by pulverizing ingots. This can trigger the transformation to the filled skutterudite compounds during the sintering through the reaction between Co (or Fe), CoSb2 (or FeSb2), Sb, and R-Sb binary alloys. Although the high-energy ball milling of ingots followed by hot pressing produced Ce and Nd double filled ptype Ce0.45Nd0.45Fe3.5Co0.5Sb12 in 2 days, the nanoinclusions of secondary phases have been inevitable [13], making difficult to precisely control the thermoelectric performance and ensure the performance reliability. Another route is the rapid solidification process such as gas atomization and MS followed by spark plasma sintering (SPS) [10,11,14]. Indeed, the n-type Yb0.3Co4Sb12 and ptype CeFe4Sb12 filled skutterudites were prepared without longtime annealing process. However, these materials are not single phase compounds but composites with nanoinclusions, implying that the reproducibility and high temperature stability issues might be aroused along with thermoelectric performance degradation due to the non-stoichiometric contents of constituent atoms [10,11]. For a more simple and efficient synthesis of high-performance single phase filled skutterudites that is always more favorable for a cost-effective large-scale industrial production. Thus we focused on the clarification of phase transformation behavior at every step of process based on the quantitative analysis that allows the reproducibility of reliable and high-performance materials. We conceived that the compositional and dimensional control of melt spun ribbons with the homogeneous distribution of secondary phases is an essential prerequisite to accelerate the evolution of single phase filled skutterudites. This conceptual strategy would be well implemented by the control of melting and consolidating conditions in non-equilibrium rapid solidification followed by pressure induced sintering process. In the present study, we demonstrate for the first time that the complete single phase of both n- and p-type single, double, and multiple filled skutterudites are fabricated in a day by temperatureregulated MS followed by SPS process. We intentionally set the melting temperature of MS process at ~1250  C, which is slightly higher than solidus temperature (~1050  C) of skutterudites, to suppress the formation of undesired secondary phases in meltspun ribbons. From the manipulation of phase formation behavior, we can trace the transformation history from precursor powders with various secondary phases to complete single phase bulk filled skutterudites. Notably, this approach is feasible for both n- and p-type nanograined filled skutterudites and yields the high zT values of 1.48 ± 0.17 at 800 K for n-type In0.12Yb0.20Co4.00Sb11.84 and 1.15 ± 0.13 at 750 K for p-type Ce0.91Fe3.40Co0.59Sb12.14, which are comparable to the highest values ever reported filled skutterudites. It is noted that the present simple and efficient synthesis of filled skutterudites is verified by the comprehensive qualitative

9

analysis confirming the actual contents of fillers from structural refinement and chemical composition analysis, and by the direct observation of atomic structure confirming the random distribution of fillers. Finally, the high-performance of nanograined filled skutterudites is attributed to the enhanced power factor (S2s) and reduced klat. 2. Methods 2.1. Synthesis High purity Co (99.9%, ingot, Kojundo Chemical), Fe (99.99%, grains, Kojundo Chemical), Sb (99.999%, shot, CLCDM), Ce (99.9%, grains, RND Korea), Yb (99.99%, grains, RND Korea), and In (99.99%, shot, RND Korea) were commercially obtained and used as received. The mixtures of raw materials with the appropriate ratio were loaded into a vacuum-sealed (~105 torr) quartz tube, and the contents were melted in a box furnace for 12 h at 1150  C then water quenched. The acquired ingots were pulverized into powders, and ribbons were prepared from the powders by using MS at intentionally controlled melting temperatures (~1250 ± 40  C) to manipulate the phase formation behavior. The powders (~6 g) were loaded into a graphite tube with a 0.35 mm nozzle and held under Ar atmosphere (30 kPa). After that, they were melted by a temperature monitoring induction system, and were injected under a pressure of 40 kPa Ar onto a Cu wheel (~300 K) rotating with linear speed of 50 m s1. In the present study, we obtained ~6 g melt spun ribbons from the compacted powders of ~8.5 g (yield ~70%) by optimization of MS processing variables. The obtained melt-spun ribbons were ground into powders and then sintered by SPS at 687  C for 10 min under 80 MPa and at 587  C for 10 min 40 MPa for n- and p-type materials, respectively. Highly dense (relative density > 96%) disk-shaped pellets with 20 mm in diameter and 8 mm in thickness were obtained. The ingot for TMA-SPS N0 sample was prepared by a typical melt-solidification and annealing at 740  C for 336 h. The ingot was pulverized and densified at 687  C for 10 min under 80 MPa using SPS [15]. 2.2. Characterizations Phase formation behavior of the melt-spun ribbons and SPSed bulks were investigated by X-ray diffraction (XRD) using a Smartlab (9 kW, Rigaku, Japan) with Cu Ka radiation (l ¼ 1.5418 Å). The actual compositions of the compacted samples were obtained by Field emission electro probe micro analyzer (FE-EPMA, JXA-8500F, 15 kV, JEOL) and also calculated from Rietveld refinement. The microstructures of the melt-spun ribbons and SPSed bulks were investigated using scanning electron microscope (SEM, JSM-7600F, JEOL) and electron backscatter diffraction (EBSD). Atomic structure of fillers into 2a site was confirmed by scanning transmission electron microscope (STEM, JEM-AEM 200F, 200 keV, JEOL). 2.3. Measurements The s and S were measured using a bar-type sample (16 mm  3 mm  3 mm) from 300 to 800 K with a ZEM-3 (Ulvac, Japan) in a He atmosphere. The thermal conductivity (k ¼ rs  Cp  a) was calculated from the separate measurements. Sample density (rs) was measured by the Archimedes principle (AlfaMiracle, MD-300S). Temperature dependences of heat capacity (Cp) and thermal diffusivity (a) were collected using a DSC (NETZSCH, DSC 200 F3) and a laser flash method (Ulvac, TC1200RH), respectively. The data of Hall effects were measured at a magnetic field of 2 T by a home-made equipment based on the van der Pauw method [16]. All measured TE transport properties,

10

S. Lee et al. / Acta Materialia 142 (2018) 8e17

which were acquired at the same dimension and configuration, are obtained within the experimental error of S (~4%), s (~4%), and k (~4.5%). Thus, we assume total uncertainty of zT as 12%.

3. Results and discussions 3.1. Phase formation behavior and microstructure Previous studies on the synthesis of filled skutterudites via rapid solidification process have reported single filled skutterudites only [10,11]. Furthermore, the phase formation behavior in the rapidly solidified melt-spun ribbons has not been investigated though the processing parameters of melting and solidification often affect the generation of compositional fluctuation related to the volatile Sb and the unexpected various nanoscale secondary phases. We thus targeted the preparation of melt-spun ribbons of In and Yb double filled skutterudites with appropriate secondary phases that can be transformed to a complete single phase during a short processing time of SPS. The melt-spun ribbons of double filled skutterudite with a nominal composition of In0.2Yb0.2Co4Sb12 were prepared by using melt-quenched ingots. To avoid the formation of unexpected secondary phases during the MS, we intentionally regulated the processing temperature at ~1250  C by considering Co-Sb binary phase diagram (Fig. 1a), which is the low limit temperature for the melt with injectable viscosity. As shown in the XRD pattern for melt-spun In0.2Yb0.2Co4Sb12 ribbons (Fig. 1b), they contain various phases consisting of Co, Sb, Co-Sb binary (CoSb, CoSb2, and CoSb3) and R-Sb binary (InSb and YbSb2) alloys. It should be noted that filling elements (In and Yb) form only binary phases of InSb and YbSb2. This controllable and reproducible phase formation is totally different from the reported melt-spun ribbons of filled skutterudites that were precipitated to various unidentified phases containing filling elements [11,14]. To elucidate the phase transformation to filled skutterudites, we performed differential scanning calorimetry (DSC) analysis and XRD measurement for the melt-spun In0.2Yb0.2Co4Sb12 ribbons. As shown in Fig. 1c, there are two conspicuous exothermic peaks at ~400  C and ~450  C during the heating (heating rate ~10  C min1) of the ribbons and no peaks are detected in the reheating, implying that the phase transformation into filled skutterudite was completed below 500  C. XRD pattern for the obtained powders by interrupting the measurement at 437  C between two peaks (Fig. 1b) gives a direct evidence for the phase transformation behavior. After the first exothermic reaction, the relative intensities

of CoSb, CoSb2, and Sb largely decrease while that of CoSb3 increases. It is noted that the peaks for InSb are almost disappeared while the peak intensity of YbSb2 decreases. This clearly indicates that the first and second exothermic peaks correspond to the reactions of Co þ CoSb þ CoSb2 þ Sb þ InSb / InxCoSb3 and CoSb2 þ Sb þ YbSb2 / YbyCoSb3, respectively. After two exothermic reactions, a complete single phase of In and Yb double filled skutterudite is evolved as verified from the XRD pattern (bottom of Fig. 1b). The actual filling fractions obtained from the structural and chemical analyses are described in Table 1. It is noteworthy that the actual filling fraction of In (12%) of MS-SPS N1 sample (In0.12Yb0.2Co4Sb11.84) is much higher than that (7%) of TMASPS N0 sample (In0.07Yb0.2Co4Sb11.93). Considering the lower In filling fraction in TMA-based process due to the formation of secondary phases such as InSb and CoSb2 connected with the simultaneous generation of In antisite defects at void and Sb sites, the present MS-based simple process offers an efficient formation of filled skutterudites by the optimization of MS processing parameters [6,17]. Indeed, all the filled skutterudites fabricated here are successfully transformed to the complete single phase alloys by the MS and subsequent short processing time of SPS as summarized in Table 1. Fig. 2a shows the photograph of melt-spun ribbons with 8e10 mm thick, 0.5e1 mm wide, and 5e15 mm long. The contact surface of melt-spun ribbons has randomly shaped grains with the characteristic size of several dozen nanometers (Fig. 2b) due to a high cooling rate (~106 K s1) inhibiting the grain growth [18], which accelerated the phase evolution of filled skutterudite during SPS. It is notable that the processing time for a phase transformation to single phase filled skutterudite, which usually takes 10 days in a TMA-based process is completed in a short SPS process (~10 min). This may be ascribed to the significantly reduced diffusion distance in several dozen nanometers, improved dispersibility, and controlled nanoscale phase formation in melt-spun ribbons. Fig. 2d shows the SEM image for the fractured surface of SPSed bulk sample shown in Fig. 2c. The grain size observed in the SPSed bulk is around 200e600 nm (Fig. 2d) which is approximately one order of magnitude smaller than that of filled skutterudite prepared by TMA-SPS process. It is noted the no inclusions and secondary phases are observed in SEM (Fig. 3(a) and (b)) and TEM images (Fig. 3(c) and (d)), indicating that the complete single phase of In and Yb double filled n-type and Ce filled p-type skutterudites are successfully fabricated. While, as clearly shown in Fig. 3(e) and (f), the SPSed bulks fabricated by using a conventional (high

Fig. 1. (a) Co-Sb binary phase diagram and phase formation behavior in temperature-regulated melt-spinning process. (b) XRD patterns of as-prepared and as-heated melt spun ribbons and sintered pellet. The as-heated melt spun ribbon was obtained by quenching the heated as-prepared melt spun ribbon at 437  C. (c) Differential scanning calorimetry (DSC) analysis showing the two exothermic peaks at (1) 402  C and (2) 447  C corresponding to each reaction (1) and (2), respectively.

S. Lee et al. / Acta Materialia 142 (2018) 8e17

11

Table 1 Comparison between nominal and actual composition studied in this work. The actual composition was obtained by FE-EPMA and Rietveld analysis in sintered pellets with reliability (Rwp) at room temperature. Type of samples

Sample #

n-type

TMA-SPS

p-type

MS-SPS

N0 N1 N2 P1 P2 P3 P4

Nominal composition

In0.2Yb0.2Co4Sb12.6 In0.2Yb0.2Co4Sb12.6 Ce0.15In0.2Yb0.2Co4Sb12.6 Ce0.9Fe3.4Co0.6Sb12 Ce0.9Fe3.4Co0.6Sb12.6 Ce0.8Yb0.1Fe3.4Co0.6Sb12 Ce0.8Yb0.1Fe3.4Co0.6Sb12.6

Actual composition FE-EPMA

Rietveld

Rwp

In0.07Yb0.20Co4Sb11.93 In0.12Yb0.20Co4Sb11.84 Ce0.14In0.17Yb0.19Co4Sb11.84 Ce0.90Fe3.4Co0.59Sb11.53 Ce0.91Fe3.4Co0.59Sb12.14 Ce0.79Yb0.10Fe3.4Co0.59Sb11.46 Ce0.78Yb0.11Fe3.4Co0.59Sb11.92

In0.07Yb0.19Co4Sb12 In0.12Yb0.20Co4Sb11.94 Ce0.14In0.17Yb0.19Co4Sb11.95 Ce0.89Fe3.4Co0.6Sb11.69 Ce0.89Fe3.4Co0.6Sb11.91 Ce0.78Yb0.10Fe3.4Co0.6Sb11.56 Ce0.79Yb0.12Fe3.4Co0.6Sb11.86

6.11 5.09 6.23 5.11 6.28 7.77 8.28

Fig. 2. (a) Photo of melt spun ribbons (4 g). (b) SEM image of contact surface of melt spun ribbon. (c) Photo of spark plasma sintered pellets (C1, C2, and C4) and elements (C3) for module. The samples for thermal and electrical transport measurement are cut from C1. (d) SEM image of fractured surface of sintered pellet, showing no nanosized inclusions and secondary phases.

temperature melting in MS process) MS-SPS process contain lots of nanoinclusions as reported in literatures [10,11,14,1924]. The elemental mapping of the p-type Ce-filled skutterudite in the STEM mode using characteristic X-ray signals acquired by energydispersive X-ray spectroscopy (EDS) also confirms a complete single phase without defects such as nanoprecipitates and substitutional antisite defects (Fig. 4). This synthetic process ensures commercial-scale mass production and reproducible high thermoelectric performance. To demonstrate the validity of the present process for filled skutterudites by confirming the fillers in the short processing time of SPS, we observed and analyzed the atomic structure by using STEM. 3.2. HAADF STEM analysis Atomic structure imaging of filled skutterudites was performed by using aberration-corrected high-angle annular dark field (HAADF). STEM combined with model-based image simulations. In the filled skutterudite, the fillers occupy 2a sites in the cages among the Sb-icosahedron [2527]. To resolve an atom-by-atom configuration of the skutterudite structure including the fillers, the HAADF STEM images which produce atomic number (Z) contrast due to its incoherent imaging nature were acquired along the [100]

orientation of the specimens. Fig. 5a and b show Z-contrast HAADF STEM images obtained from n- and p-type skutterudites with the nominal compositions of Ce0.15In0.2Yb0.2Co4Sb12.6 (N2) and Ce0.9Fe0.6Co3.4Sb12.6 (P2), respectively. Simulated HAADF STEM images and the corresponding atomic models are displayed alongside the respective experimental images (Methods). As observed in both HAADF STEM images (Fig. 5a and b), a notable structural feature is that the 2a void sites are populated by the fillers within their occupation probability, which is very similar to the previous work on TMA-based process [4,9,15,2830]. For the ntype MS-SPS N2 sample (Fig. 5a), we detect diffusive and inhomogeneous intensities in the void sites due to small occupancy, while the Sb atoms show relatively uniform bright contrast. The analysis on the observed atomic structures shows how the fillers are distributed. The intensity profiles extracted from three void sites in n-type multiple filled Ce0.15In0.2Yb0.2Co4Sb12.6 skutterudite (denoted with A1 to A3 in Fig. 5a) clearly demonstrate the random filling behavior of rattlers (see the lower panels in Fig. 5a). Given the relatively different intensities of the fillers as compared with those of the Sb atoms, we can estimate that the site-wise relative occupancy of the fillers roughly varies from 0 (A1, not filled) to 40% (A3) in maximum. Note that a small bump of intensity at the empty void site (A1) was observed. This can be attributed to effects of thermal diffuse scattering from adjacent atoms and side

12

S. Lee et al. / Acta Materialia 142 (2018) 8e17

Fig. 3. SEM images of (a) n-type (In0.2Yb0.2Co4Sb12.6) or (b) p-type (Ce0.9Fe3.4Co0.6Sb12.6) pellets and TEM images of (c) n-type (In0.2Yb0.2Co4Sb12.6) or (d) p-type (Ce0.9Fe3.4Co0.6Sb12.6) pellets fabricated by temperature-regulated MS and SPS. TEM images of (e) n-type (In0.2Yb0.2Co4Sb12.6) or (f) p-type (Ce0.9Fe3.4Co0.6Sb12.6) pellets fabricated by conventional MS and SPS.

lobes of the electron probe, which can produce some artificial bright contrast at positions where no atoms exit. Likewise, the intensities of the Ce at void sites in p-type filled Ce0.9Fe0.6Co3.4Sb12.6 skutterudite are all different as shown in Fig. 5b. Ce fillers occupy the void sites in the undeformed Fe3.4Co0.6Sb12 network structure in p-type filled skutterudite, well-matching with the simulated crystal structure as shown in Fig. 4. The site-wise relative occupancy of the fillers is approximately measured to vary from 20 (B1) to 50% (B3) in maximum. Due to relatively high concentration of the fillers, the empty 2a sites were not observed in this sample. It is noteworthy that the random distribution of fillers is commonly observed in both the present n- and p-type filled skutterudites. This is the direct observation to clearly demonstrate the random distribution of fillers, which can be linked to the theoretically proposed resonant scattering effect resulting in lower klat value

benefiting from an intensified phonon scattering [9,15,28]. Meanwhile, the observed undeformed CoSb3 and FeSb3 network structures in both n- and p-type filled skutterudites that are well matched with the simulated structures can be relevant to a high power factor as a characteristic feature of PGEC materials. 3.3. Thermoelectric properties Fig. 6a shows the temperature dependence of s for n-type filled skutterudites. The s values decrease with temperature for all samples, indicating the degenerated semiconducting property. To clarify the effect of fillers on the electronic and thermal transport properties, we evaluated the contents of constituent elements by using FE-EPMA as listed in Table 1. The actual compositions, especially contents of fillers, of n- and p-type filled skutterudites

S. Lee et al. / Acta Materialia 142 (2018) 8e17

13

Fig. 4. The STEM-EDS elemental mapping images of each element (a) Sb, (b) Co, (c) Ce, (d) Fe, and (e) combined one from (a) to (d) in the STEM mode using characteristic X-ray signals acquired by energy-dispersive X-ray spectroscopy (EDS) for the image of Fig. 4(b). (f) Mapping image of combined elements viewed down the [001] zone axis.

Fig. 5. (a) High-angle annual dark field scanning transmission electron microscopy (HAADF STEM) image (upper) for N2 sample with model and simulation along [001] orientation. Intensity profiles of three different void sites (lower panel A1 e A3). Orange and red circles in the panel represent the intensities of antimony and fillers, respectively. (b) HAADF STEM image (upper) for P2 sample with model and simulation along [001] orientation. Intensity profiles of three different void sites (lower panel B1 e B3). Orange and blue circles in the panel represent the intensities of antimony and filler, respectively. (For interpretation of the references to colour in this figure legend, the reader is referred to the web version of this article.)

determined by FE-EPMA are well matched with the calculated compositions by structural refinement using Rietveld analysis in Fig. S1a and Table 1. The s values increase as the fillers (In, Yb, and Ce) occupied the voids due to the increase in carrier concentration (n) originated from the donation of valence electrons into skutterudite framework. However, the m was inevitably deteriorated probably due to the change of scattering mechanism by fillers (Table 2) from dominant ionized impurity scattering at low n to acoustic phonon scattering at high n [4,9]. Notably, significantly enhanced s values of MS-SPS N1 sample compared to TMA-SPS N0 sample is attributed to the higher n from the enhanced filling fraction of In and rather small decrease in m (Table 2) from the appropriate grain size in Fig. S2 and undeformed CoSb3 network structure in Fig. 5 for carrier transmitting. The S values for all n-type samples were negative and the absolute values increase with temperature. As expected, the S values decrease with the content of fillers due to the increased n (Fig. 6b). On the other hand, S values observed in MS-SPS N1 and N2 samples are rather large considering the typical decreasing behavior of single filled skuttrudites with n. To clarify these variations of electronic transport behavior, we have calculated the density of states (DOS) effective mass m*d value that is one of the main factors determining S from the following equation [3134]:

S¼

8p2 k2B *
p 2=3 m T 3n 3eh2 d

where kB, e, and h are the Boltzmann constant, elementary charge, and the Planck constant, respectively. The most remarkable change is the enhanced m*d values at 300 K for MS-SPS N1 (~4.5m0) and N2 (~4.9m0) samples, which are larger than that (~3.5m0) of TMA-SPS N0 sample (Table 2). As a result, the maximum power factor value of 5.8 mW m1 K2 at 800 K for MS-SPS N1 sample was achieved (Fig. 6c) due to its optimized n and m*d values. Thus, for a material design of filled skutterudites accommodating various fillers with different charge states, the combination of filling elements is crucial to obtain a high power factor originated from the optimum n and m*d values. The thermal conductivity (ktot) values for all MS-SPS n-type filled skutterudites are lower than 3.5 W m1 K1 in the whole measured temperature range as shown in Fig. 6d mainly due to the extremely low klat values (Fig. 6e). To elucidate the effects of fillers and nanograin on phonon scattering, klat values are obtained by subtracting the electronic (kele) and bipolar (kbi) contributions from the ktot values. The kele values are calculated using the WiedemanneFranz law (kele ¼ LsT). We adopted Lorenz numbers (L) as a

14

S. Lee et al. / Acta Materialia 142 (2018) 8e17

Fig. 6. Temperature dependence of (a) electrical conductivity (s), (b) Seebeck coefficient (S), (c) power factor (sS2), (d) total thermal conductivity (ktot), (e) lattice thermal conductivity (klat), and (f) dimensionless figure of merit (zT) for n-type filled skutterudites. The inset of (e) is the contribution of bipolar thermal conduction (kbi).

Table 2 Room temperature thermoelectric transport parameters (klat, n, m, and effective mass) for MS-SPS n- and p-type filled skutterudites. Those for TMA-SPS n-type filled skutterudite are shown for comparison. Type of samples

Sample #

Filling fraction (FE-EPMA)

Klat ( W m1 K1)

n (1020 cm3)

m (cm2 V1 s1)

Effective mass (m0)

n-type

TMA-SPS

p-type

MS-SPS

N0 N1 N2 P1 P2 P3 P4

0.263 0.319 0.498 0.903 0.909 0.892 0.887

2.578 1.808 1.629 1.218 1.123 1.194 1.122

2.98 5.57 7.34 9.26 15.05 10.66 17.40

32.4 25.8 19.8 9.9 7.2 9.0 6.5

3.5 4.5 4.9 5.0 6.5 5.2 6.9

function of temperature (1.90e1.94  108 V2 K2 at 300 K and 1.75e1.78  108 V2 K2 at 800 K), which are approximated from the following equation:

=

2

ðhÞ

2

#2 ! ðhÞ

=

"  r þ 5 2 Frþ    r þ 3 2 Frþ 1 ðhÞ 2 5

ðhÞ

3

=

=

 2  r þ 7 2 Frþ   r þ 3 2 Frþ

=

= =

kB e

1

=

 L¼

2

where r is the scattering parameter, Fn(h) is the n-th order Fermi integral, and h is Fermi energy, respectively. Details of the L calculation have been described elsewhere [32,35,36]. The temperature dependence of kbi values (inset of Fig. 6e) are estimated from the 1000/T plot for ktot - kele [3740]. There is little contribution from kbi at low temperatures where the acoustic phonon scattering is predominant, while the contribution from kbi gradually increases with the temperature from 600 K. Besides the enhanced electronic transport properties, the filling of rattlers also has a positive effect on the reduction of klat as clearly shown in Fig. 6e. The klat values decrease with filling fraction and reduction of klat is intensified by the resonant phonon scattering effect in the presence of randomly distributed multiple fillers (Fig. 5) with a different localized vibration frequency [1,4,9,41]. Therefore, the combination of filling elements with different rattling frequencies should be prioritized for a material design of filled skutterudites to

achieve broad-spectrum phonon scattering for low klat realizing a high zT. On the other hand, we estimated klat values form measured ktot values for un-filled Co4Sb12.6 fabricated by TMA-SPS (average grain size ~2500 nm) and MS-SPS (average grain size ~440 nm) with different grain sizes. From the calculated klat values (dash lines) for un-filled Co4Sb12.6 samples with smaller grain sizes (100 and 10 nm) by using Debye-Callaway model as shown in Fig. S3, it is clearly shown that the klat values can be reduced by decrease in grain size. Detailed method for klat calculation is described in the Supplementary Information. Meanwhile, considering the differences in the actual fraction of fillers, the significantly reduced klat values of MS-SPS N1 and N2 samples compared with the TMA-SPS N0 sample is ascribed to the additional phonon scattering contribution from the highly dense grain boundaries of nanograined structure obtained by the present MS-SPS process (Fig. 2d). Lattice thermal conduction is effectively suppressed in the whole measured temperature range by the grain boundary phonon scattering, while the m values of MS-SPS samples are maintained at the similar level of previous reported values for samples fabricated by TMA-based process. Due to this synergetic effect of wide-range frequency phonon scattering by fillers and grain boundaries combined with activated carrier transmitting (high power factor) by controlled electronic transport parameters (n, m, and m*d), a very high zT value of 1.48 ± 0.17 at 800 K is obtained

S. Lee et al. / Acta Materialia 142 (2018) 8e17

15

Fig. 7. Temperature dependence of (a) electrical conductivity (s), (b) Seebeck coefficient (S), (c) power factor (sS2), (d) total thermal conductivity (ktot), (e) lattice thermal conductivity (klat), and (f) dimensionless figure of merit (zT) for three different N1, N1-1, and N1-2 samples. The inset of (e) is the contribution of bipolar thermal conduction (kbi).

for In0.12Yb0.2Co4Sb11.84 filled skutterudites, which is one of the highest zT values reported so far [15,17,2830]. Most of all, it should be highlighted that this superior performance is realized in a large-scale (~20 mm in diameter and ~8 mm in thickness) bulk by the present simple MS-SPS processes. The zT values for all MS-SPS samples increase with temperature and reach to 1.48 ± 0.17 at 800 K in double filled skutterudite (N1), suggesting the possibility for a higher zT by the optimization of combination and amount of filling elements in multiple filled skutterudites. Fig. 7 presents the temperature dependence of thermoelectric properties of three batches of N1 sample. N1, N1-1, N1-2 samples prepared from different batch melt spun ribbons to confirm the feasibility of our temperature-regulated MS process. The variation of s, S, and power factor are ~2%, ~4%, and ~6%, respectively, confirming the reproducibility in the process. Moreover, we verify the reliable electronic transport properties (s, S, and power factor) via static annealing test as shown in Fig. 8. These results suggest that our temperature-regulated MS and SPS process is promising for the large-scale production of filled skutterudites. We also demonstrate the feasibility of our process for p-type filled skutterudites from actual compositions and thermoelectric properties. Ce single filled

and Ce and Yb double filled skutterudites with different Sb contents are prepared. We also fabricated the Ce filled skutterudite by using conventional MS and SPS. As shown in Table 1, Ce and Yb are successfully occupied at void sites, indicating that our temperature-regulated MS process is also valid for p-type filled skuttrudites. The filling behavior of random distributed rattlers is confirmed by the observation of atomic structure as shown in Fig. 5b. Fig. 9a shows the temperature dependence of s for MS-SPS p-type filled skutterudites. The s values of Ce and Yb double filled skutterudites (P3 and P4) are slightly higher than those of Ce single filled skutterudites (P1 and P2) due to the higher n (Table 1). Meanwhile, a higher Sb content in p-type skutterudites resulted in a higher n, indicating that the Sb content is a critical parameter in composition for the enhancement of electronic transport properties. It should be noted that we obtained the controllable and reproducible Sb contents in the final bulks, which guarantee the reliability of our MS process by the strict control of Sb volatilization via precise regulation both of melting temperature and Ar partial pressure in MS chamber. The S values of MS-SPS p-type filled skutterudites show the trade-off relationship with s (Fig. 9b) [4244], while those of P2 and P4

Fig. 8. Durability test by ZEM3 during 24 h with N1 sample. Durable time dependence of (a) electrical conductivity (s), (b) Seebeck coefficient (S), and (c) power factor (sS2).

16

S. Lee et al. / Acta Materialia 142 (2018) 8e17

Fig. 9. Temperature dependence of (a) electrical conductivity (s), (b) Seebeck coefficient (S), (c) power factor (sS2), (d) total thermal conductivity (ktot), (e) lattice thermal conductivity (klat), and (f) dimensionless figure of merit (zT) for p-type filled skutterudites. The inset of (e) is the contribution of bipolar thermal conduction (kbi). Properties for Ce0.9Fe3.4Co0.6Sb12.6 fabricated by conventional MS and SPS are also shown for comparison.

samples are rather larger than those of P1 and P3 (Table 2). Since the n values of P2 and P4 are higher than those of P1 and P3, the larger S values of P2 and P4 can be attributed to the increased m*d values (~6.5m0 for P2 and ~6.9m0 for P4) compared to those of P1 and P3 (~5.0m0 for P1 and ~5.2m0 for P3) as summarized in Table 2, suggesting that the variation of Sb content modifies the band structure [8,17]. The highest power factor in the present p-type filled skutterudites was obtained to be of 3.68 mW m1 K2 at 750 K for P2 sample. These high electronic transport properties are originated from the formation of complete single phase (Fig. 5b), whereas the TE properties of the Ce filled skutterudite fabricated by conventional MS and SPS are significantly deteriorated due to the presence of nanoprecipitates (Fig. 3d). The ktot values for all MS-SPS p-type filled skutterudites are lower (<2.8 W m1 K1) than those of MS-SPS n-type samples (<3.4 W m1 K1) as shown in Fig. 9d due to the lower electronic contribution originated from the lower s, yielding the very low klat values ranging from 0.5 W m1 K1e1.2 W m1 K1. Although a small fluctuation in klat according to the compositions is observed probably due to the random distribution of fillers, the present simple MS-SPS process yields a very high zT value in p-type filled skutterudites with 1.15 ± 0.13 at 750 K for Ce0.91Fe3.40Co0.59Sb12.14 (P2). 4. Conclusion Complete single phases of n- and p-type single to multiple filled skutterudites with outstanding zTs have been fabricated by using melt spinning and spark plasma sintering without long time annealing process. Temperature-regulated melt spinning significantly accelerates the phase evolution of filled skutterudite during a short processing time of spark plasma sintering by the homogeneous dispersion of nanosized binary compounds. The filling behavior of the samples, which is confirmed by direct observation of atomic structure, reveals that the fillers randomly occupy the void in the skutterudite framework. Thus, as opposite to long-time

traditional melt-solidification-based processes, the electronic and thermal transport properties could be easily manipulated by adjusting the combination of fillers and amount of filling fraction. Large size bulks with nanograined structure attain the maximum zT values of 1.48 ± 0.17 at 800 K for n-type In0.12Yb0.20Co4.00Sb11.84 and 1.15 ± 0.13 at 750 K for p-type Ce0.91Fe3.40Co0.59Sb12.14, which are comparable to the highest zT values reported for filled skutterudites, due to the controlled atomic and nanoscale structures. This simple and efficient synthesis gives a pathway to overcome the barrier for large scale industrial production of high-performance thermoelectric materials. Acknowledgments This work was supported by the National Research Foundation of Korea (NRF) grant funded by the Korea government (MSIT) (No. 2017R1A2A1A17069289) and IBS-R011-D1 and also supported by a National Research Foundation of Korea (NRF) grant funded by the Korean Government (MSIP) (NRF-2015R1A5A1036133). Appendix A. Supplementary data Supplementary data related to this article can be found at https://doi.org/10.1016/j.actamat.2017.09.044. References [1] W. Jeitschko, D. Braun, LaFe4P12 with filled CoAs3-type structure and isotypic lanthanoid-transition metal polyphosphides, Acta Crystallogr. Sect. B 33 (1977) 3401e3406. [2] M. Rull-Bravo, A. Moure, J. Fernandez, M. Martin-Gonzalez, Skutterudites as thermoelectric materials: revisited, Rsc Adv. 5 (2015) 41653e41667. [3] J.R. Salvador, J.Y. Cho, Z. Ye, J.E. Moczygemba, A.J. Thompson, J.W. Sharp, J.D. Koenig, R. Maloney, T. Thompson, J. Sakamoto, H. Wang, A.A. Wereszczak, Conversion efficiency of skutterudite-based thermoelectric modules, Phys. Chem. Chem. Phys. 16 (2014) 12510e12520. [4] X. Shi, J. Yang, J.R. Salvador, M. Chi, J.Y. Cho, H. Wang, S. Bai, J. Yang, W. Zhang, L. Chen, Multiple-filled skutterudites: high thermoelectric figure of merit

S. Lee et al. / Acta Materialia 142 (2018) 8e17

[5] [6]

[7] [8]

[9] [10]

[11]

[12] [13]

[14]

[15]

[16] [17]

[18]

[19]

[20]

[21]

[22]

[23]

[24]

through separately optimizing electrical and thermal transports, J. Am. Chem. Soc. 133 (2011) 7837e7846. E.S. Toberer, A. Zevalkink, G.J. Snyder, Phonon engineering through crystal chemistry, J. Mater. Chem. 21 (2011) 15843. Y. Qiu, L. Xi, X. Shi, P. Qiu, W. Zhang, L. Chen, J.R. Salvador, J.Y. Cho, J. Yang, Y.c. Chien, S.-w. Chen, Y. Tang, G.J. Snyder, Charge-compensated compound defects in Ga-containing thermoelectric skutterudites, Adv. Funct. Mater. 23 (2013) 3194e3203. G. Nolas, J. Cohn, G.A. Slack, Effect of partial void filling on the lattice thermal conductivity of skutterudites, Phys. Rev. B 58 (1998) 164e170. J. Yu, W.-y. Zhao, P. Wei, D.-g. Tang, Q.-j. Zhang, Effects of excess Sb on thermoelectric properties of barium and indium double-filled iron-based ptype skutterudite materials, J. Electron. Mater. 41 (2012) 1414e1420. S. Ballikaya, G. Wang, K. Sun, C. Uher, Thermoelectric properties of triple-filled Ba x Yb y in z Co4Sb12 skutterudites, J. Electron. Mater. 40 (2010) 570e576. G. Tan, W. Liu, S. Wang, Y. Yan, H. Li, X. Tang, C. Uher, Rapid preparation of CeFe4Sb12 skutterudite by melt spinning: rich nanostructures and high thermoelectric performance, J. Mater. Chem. A 1 (2013) 12657. H. Li, X. Tang, Q. Zhang, C. Uher, Rapid preparation method of bulk nanostructured Yb0.3Co4Sb12þy compounds and their improved thermoelectric performance, Appl. Phys. Lett. 93 (2008) 252109. Q. Li, Z. Lin, J. Zhou, Thermoelectric materials with potential high power factors for electricity generation, J. Electron. Mater. 38 (2009) 1268e1272. Q. Jie, H. Wang, W. Liu, H. Wang, G. Chen, Z. Ren, Fast phase formation of double-filled p-type skutterudites by ball-milling and hot-pressing, Phys. Chem. Chem. Phys. 15 (2013) 6809e6816. H. Li, X. Tang, X. Su, Q. Zhang, Preparation and thermoelectric properties of high-performance Sb additional Yb0.2Co4Sb12þy bulk materials with nanostructure, Appl. Phys. Lett. 92 (2008) 202114. J. Graff, S. Zhu, T. Holgate, J. Peng, J. He, T.M. Tritt, High-temperature thermoelectric properties of Co4Sb12-Based skutterudites with multiple filler atoms: Ce0.1In x Yb y Co4Sb12, J. Electron. Mater. 40 (2011) 696e701. H. Wang, Z.M. Gibbs, Y. Takagiwa, G.J. Snyder, Tuning bands of PbSe for better thermoelectric efficiency, Energy Environ. Sci. 7 (2014) 804e811. Y. Tang, Y. Qiu, L. Xi, X. Shi, W. Zhang, L. Chen, S.-M. Tseng, S.-w. Chen, G.J. Snyder, Phase diagram of IneCoeSb system and thermoelectric properties of In-containing skutterudites, Energy Environ. Sci. 7 (2014) 812e819. A.G. Gillen, B. Cantor, Photocalorimetric cooling rate measurements on a Ni5wt% Al alloy rapidly solidified by melt spinning, Acta Metall. 33 (1985) 1813e1825. H. Li, X. Tang, Q. Zhang, Microstructure and thermoelectric properties of YbFilled skutterudites prepared by rapid solidification, J. Electron. Mater. 38 (2009) 1224e1228. J.R. Salvador, R.A. Waldo, C.A. Wong, M. Tessema, D.N. Brown, D.J. Miller, H. Wang, A.A. Wereszczak, W. Cai, Thermoelectric and mechanical properties of melt spun and spark plasma sintered n-type Yb- and Ba-filled skutterudites, Mater. Sci. Eng. B 178 (2013) 1087e1096. G. Tan, Y. Zheng, X. Tang, High thermoelectric performance of nonequilibrium synthesized CeFe4Sb12 composite with multi-scaled nanostructures, Appl. Phys. Lett. 103 (2013) 183904. D.R. Thompson, C. Liu, J. Yang, J.R. Salvador, D.B. Haddad, N.D. Ellison, R.A. Waldo, J. Yang, Rare-earth free p -type filled skutterudites: mechanisms for low thermal conductivity and effects of Fe/Co ratio on the band structure and charge transport, Acta Mater. 92 (2015) 152e162. L. Guo, Y. Zhang, Y. Zheng, W. Yao, G. Wang, G. Wang, X. Zhou, Super-rapid preparation of nanostructured Nd x Fe3CoSb12 compounds and their improved thermoelectric performance, J. Electron. Mater. 45 (2015) 1271e1277. L. Guo, G. Wang, K. Peng, Y. Yan, X. Tang, M. Zeng, J. Dai, G. Wang, X. Zhou, Melt spinning synthesis of p -type skutterudites: drastically speed up the

17

process of high performance thermoelectrics, Scr. Mater. 116 (2016) 26e30. [25] G.S. Nolas, M. Kaeser, R.T. Littleton IV, T.M. Tritt, High figure of merit in partially filled ytterbium skutterudite materials, Appl. Phys. Lett. 77 (2000) 1855e1857. [26] P.N. Alboni, X. Ji, J. He, N. Gothard, T.M. Tritt, Thermoelectric properties of La0.9CoFe3Sb12eCoSb3 skutterudite nanocomposites, J. Appl. Phys. 103 (2008) 113707. [27] T. He, J.Z. Chen, H.D. Rosenfeld, M.A. Subramanian, Thermoelectric properties of indium-filled skutterudites, Chem. Mater 18 (2006) 759e762. [28] S. Ballikaya, N. Uzar, S. Yildirim, J.R. Salvador, C. Uher, High thermoelectric performance of In, Yb, Ce multiple filled CoSb3 based skutterudite compounds, J. Solid State Chem. 193 (2012) 31e35. [29] R.C. Mallik, Thermoelectric properties of InzCo4Sb12 skutterudites, Metals Mater. Int. 14 (2008) 223e228. [30] A. Grytsiv, P. Rogl, H. Michor, E. Bauer, G. Giester, In y Co4Sb12 skutterudite: phase equilibria and crystal structure, J. Electron. Mater. 42 (2013) 2940e2952. [31] G.J. Snyder, E.S. Toberer, Complex thermoelectric materials, Nat. Mater. 7 (2008) 105e114. [32] D. Suh, S. Lee, H. Mun, S.-H. Park, K.H. Lee, S. Wng Kim, J.-Y. Choi, S. Baik, Enhanced thermoelectric performance of Bi0.5Sb1.5Te3-expanded graphene composites by simultaneous modulation of electronic and thermal carrier transport, Nano Energy 13 (2015) 67e76. [33] Z. Xiong, X. Chen, X. Huang, S. Bai, L. Chen, High thermoelectric performance of Yb0.26Co4Sb12/yGaSb nanocomposites originating from scattering electrons of low energy, Acta Mater. 58 (2010) 3995e4002. [34] P.-X. Lu, Q.-H. Ma, Y. Li, X. Hu, A study of electronic structure and lattice dynamics of CoSb3 skutterudite, J. Magn. Magn. Mater. 322 (2010) 3080e3083. [35] J.S. Dyck, W. Chen, C. Uher, L. Chen, X. Tang, T. Hirai, Thermoelectric properties of the n-type filled skutterudite Ba0.3Co4Sb12 doped with Ni, J. Appl. Phys. 91 (2002) 3698e3705. [36] S.I. Kim, K.K. Lee, H.A. Mun, H.S. Kim, S.W. Hwang, J.W. Roh, D.J. Yang, W.H. Shin, X.S. Li, Y.H. Lee, G.J. Snyder, S.W. Kim, Dense dislocation arrays embedded in grain boundaries for high-performance bulk thermoelectrics, Science 348 (2015) 109e113. [37] Y.-L. Pei, Y. Liu, Electrical and thermal transport properties of Pb-based chalcogenides: PbTe, PbSe, and PbS, J. Alloys Compd. 514 (2012) 40e44. [38] L.D. Zhao, H.J. Wu, S.Q. Hao, C.I. Wu, X.Y. Zhou, K. Biswas, J.Q. He, T.P. Hogan, C. Uher, C. Wolverton, V.P. Dravid, M.G. Kanatzidis, All-scale hierarchical thermoelectrics: MgTe in PbTe facilitates valence band convergence and suppresses bipolar thermal transport for high performance, Energy & Environ. Sci. 6 (2013) 3346. [39] R. Liu, X. Chen, P. Qiu, J. Liu, J. Yang, X. Huang, L. Chen, Low thermal conductivity and enhanced thermoelectric performance of Gd-filled skutterudites, J. Appl. Phys. 109 (2011) 023719. [40] S. Wang, J. Yang, T. Toll, J. Yang, W. Zhang, X. Tang, Conductivity-limiting bipolar thermal conductivity in semiconductors, Sci. Rep. 5 (2015) 10136. [41] Y. Qiu, J. Xing, X. Gao, L. Xi, X. Shi, H. Gu, L. Chen, Electrical properties and microcosmic study on compound defects in Ga-containing thermoelectric skutterudites, J. Mater. Chem. A 2 (2014) 10952. [42] Y. Tang, S.-w. Chen, G.J. Snyder, Temperature dependent solubility of Yb in YbeCoSb3 skutterudite and its effect on preparation, optimization and lifetime of thermoelectrics, J. Materiomics 1 (2015) 75e84. [43] J.-K. Lee, S.-M. Choi, W.-S. Seo, Y.-S. Lim, H.-L. Lee, I.-H. Kim, Thermoelectric properties of spark plasma sintered InxYbyLa0.3-x-yCo4Sb12 skutterudite system, Renew. Energy 42 (2012) 36e40. [44] S. Ballikaya, C. Uher, Enhanced thermoelectric performance of optimized Ba, Yb filled and Fe substituted skutterudite compounds, J. Alloys Compd. 585 (2014) 168e172.