High Pressure Synthesis of p-Type CeyFe4−xCoxSb12 Skutterudites

Co-substituted p-Type CeFe4−xCoxSb12 skutterudites were successfully synthesized with a high pressure synthesis method. The structure, composition, and thermoelectric properties were investigated. The obtained CeyFe4−xCoxSb12 samples show the skutterudite structure of Im3¯ symmetry. The hole concentration decreases with elevating Co substitution level, leading to increased Seebeck coefficient and electrical resistivity. Meanwhile, the filling fraction of Ce decreases, which is unfavorable for reducing the lattice thermal conductivity. As a result, the thermoelectric performance of CeyFe4−xCoxSb12 deteriorates with higher Co content. The maximal ZT of 0.91 was achieved at 763 K for the optimal Ce0.92Fe4Sb12 sample.


Introduction
Thermoelectric materials have attracted great research interest during the past few decades due to their prospective applications in energy conversion to alleviate the greenhouse effect and environmental pollution [1]. In general, the performance of thermoelectric materials is characterized by the dimensionless figure of merit ZT where S, T, ρ, and κ are the Seebeck coefficient, temperature, electrical resistivity, and thermal conductivity, respectively. To make the thermoelectric device competitive with other energy conversion devices, thermoelectric materials with high ZT in the working temperature range must be fabricated, which is a practical challenge because of the entanglements between S, ρ, and κ. Improvement of one parameter usually adversely influences others. To unravel the entanglements, Slack proposed the concept of phonon-glass electron-crystal (PGEC) in 1995 [2]. Skutterudites, a prototype of the PGEC concept, are among the most promising thermoelectric materials working in the intermediate temperature region [3]. Various approaches have been taken to improve the thermoelectric properties of skutterudites, such as void filling [4][5][6][7][8][9][10][11][12] and lattice atom substituting [13][14][15][16][17][18][19]. Compared with n-type CoSb 3 -based skutterudites for which ZT records are constantly refreshed [4][5][6], the developments in p-type skutterudites relatively lag behind [9,10]. Elemental filled FeSb 3 -based skutterudites are one of the most important p-type skutterudites. In this class of materials, κ is greatly suppressed due to the rattling filler atoms (with a near unit filling fraction), and a low ρ is ensured by the high hole concentration (>10 21 cm´3); meanwhile, a moderate S is maintained due to the large effective mass of hole [20]. Further enhancement of ZT for elemental filled FeSb 3 -based skutterudites might be possible through enhancing power factor PF and suppressing thermal conductivity.
Substituting framework atoms with electron-richer atoms (such as Co/Ni for Fe) would be a possible method. By such kind of substitutions, the hole concentration can be lowered to benefit enhancement of S. In addition, the impurity scattering of phonons due to mass fluctuation may contribute to further lower κ.
The effects of Co/Ni substitution for Fe on the thermoelectric properties of CeFe 4 Sb 12 were investigated previously [21][22][23][24]. In these studies, the time-consuming solid state reaction method was employed to fabricate the samples, which usually takes several days. Recent works have highlighted pressure as a fundamental thermodynamic variable in synthesizing elemental filled CoSb 3 [25][26][27] and Te-doped CoSb 3 [28][29][30]. High pressure can lower the reaction temperature and facilitate the synthesis of metastable phase by shifting reaction equilibrium [31]. In this work, Co-substituted CeFe 4 Sb 12 skutterudites were synthesized by using a time-saving high pressure synthesis (HPS) method. We found that the actual filling fraction of Ce decreases slightly with elevating Co substitution level, and a notable decrease in the hole concentration is observed. The hole concentration decrease shows a more pronounced effect on the electrical resistivity, leading to a significantly reduced power factor in the sample with the highest Co level. As a result, the thermoelectric performance deteriorates with higher Co content. The optimal Ce 0.92 Fe 4 Sb 12 showed a maximal ZT of 0.91 at 763 K.

Experimental Methods
Ce (99.9%), Fe (99.5%), Co (99.8%), and Sb (99.999%) powders were mixed according to the atomic ratio of 1:(4´a):a:12 (a = 0, 0.5, and 1, respectively). The mixture was filled into a mold and shaped into a cylinder with a cold press method. The as-prepared cylinder was inserted into an h-BN crucible and loaded into a high pressure apparatus for HPS experiments. The first step of HPS was carried out at 3 GPa and 1100 K for 0.5 h. The intermediate product was ground into powders under argon protection, shaped and subject to the second step of HPS (3 GPa and 873 K for 3 h). The obtained product was ground into powders, washed in acidic solution to remove excess Ce, and sintered into dense pellets by using spark plasma sintering (SPS) at 60 MPa and 853 K for 15 min. Rectangular blocks (2ˆ2ˆ8 mm 3 ), disks (ø 6 mmˆ1 mm), and slices (2.5ˆ6ˆ0.4 mm 3 ) were cut from the final pellets for the electrical transport, thermal conductivity, and Hall coefficient measurements, respectively.
The chemical composition and crystal structure of the final products were characterized with electron probe microanalysis (EPMA, JEOL JXA-8230, JEOL, Tokyo, Japan) and X-ray diffraction (XRD, Rigaku D/MAX/2500/PC, Rigaku, Tokyo, Japan), respectively. The Seebeck coefficient and electrical resistivity were measured with a ZEM-3 apparatus (Ulvac-Riko, Yokohama, Japan) and the thermal conductivity was measured with a TC-9000H apparatus (Ulvac-Riko, Yokohama, Japan). The uncertainty in S, ρ, and κ measurements is within 5%. The Hall coefficient measurements were performed at room temperature with a Physical Property Measurement System (PPMS, Quantum Design, San Diego, CA, USA). The heat capacity measurements were carried out between 5 and 25 K with the heat capacity option of PPMS.

Results and Discussion
XRD patterns from Ce y Fe 4´x Co x Sb 12 samples are shown in Figure 1. All these samples are dominated with a skutterudite structure of Im3 symmetry. XRD results also reveal trace amounts of FeSb 2 and Sb, which are less than 5% as determined by EPMA analyses. A systematical shift of diffraction peaks to higher diffraction angles occurs with increasing Co content, as illustrated in Figure 1b for the magnified (013) peak, indicating that the lattice parameter decreases with elevating Co level. The actual content of Co (x) and filling fraction of Ce (y) in Ce y Fe 4´x Co x Sb 12 samples determined by EPMA measurements are listed in Table 1. The highest filling fraction of Ce of 0.92 is reached in the sample without Co substitution (Ce 0.92 Fe 4 Sb 12 ), which is at the same level as that of 0.91 achieved in an ambient-pressure synthesized sample [11]. Cobalt atom has one more valence electron and smaller atomic size than those of iron atoms. By substituting Co on Fe sites, extra electrons from Co are introduced to the framework meanwhile the lattice parameter is reduced. Both effects contribute adversely to the void-filling process of Ce. Therefore, the higher Co substitution level, and the lower Ce filling fraction [24], and both contribute to a smaller lattice parameter. The actual filling fraction of Ce in our HPS samples decreases slightly with increasing Co substitution level, and is higher than that in the ambient-pressure synthesized sample with the same Co level [24], demonstrating an advantage of high pressure for achieving higher filling fraction.
Materials 2016, 9,257 3 Co are introduced to the framework meanwhile the lattice parameter is reduced. Both effects contribute adversely to the void-filling process of Ce. Therefore, the higher Co substitution level, and the lower Ce filling fraction [24], and both contribute to a smaller lattice parameter. The actual filling fraction of Ce in our HPS samples decreases slightly with increasing Co substitution level, and is higher than that in the ambient-pressure synthesized sample with the same Co level [24], demonstrating an advantage of high pressure for achieving higher filling fraction. Room temperature electrical transport properties for different samples are summarized in Table  1. All samples exhibit positive Seebeck coefficient, indicating the majority charge carriers of holes. The hole concentration deduced from the measured Hall coefficient decreases notably with elevating Co substitution level, which is due to Co substitution introducing extra electrons into the system. In addition, the hole concentration in our Ce0.85Fe3.03Co0.97Sb12 sample is 9.5 × 10 20 cm −3 , less than half of that in the ambient-pressure synthesized Ce0.9Fe3CoSb12 (nominal composition) [21]. The difference can be attributed to the higher filling fraction of Ce in our HPS sample, and accounts for the significant increase in the electrical resistivity of our HPS Ce0.85Fe3.03Co0.97Sb12 sample. The hole concentrations can also be calculated by simple charge counting based on the chemical compositions, which differ from those listed in Table 1. A similar difference was previously reported [11,32], and might be related to the multiple bands near the Fermi level in elemental-filled FeSb3 materials where a simple relationship between carrier concentration and Hall coefficient does not exist. Note that the carrier mobility is continually suppressed with increasing Co level, indicating a decreased hole mean free path caused by impurity scattering.
The temperature dependent thermoelectric properties of CeyFe4-xCoxSb12 samples are shown in Figure 2. The electrical resistivity increases with elevating Co substitution level (Figure 2a), which can be attributed to the decreased hole concentration compensated by the additional electrons from cobalt. The most Co-substituted Ce0.85Fe3.03Co0.97Sb12 exhibits the largest electrical resistivity, which starts to decrease at high temperature due to thermally activated carriers. The other two samples with  Room temperature electrical transport properties for different samples are summarized in Table 1. All samples exhibit positive Seebeck coefficient, indicating the majority charge carriers of holes. The hole concentration deduced from the measured Hall coefficient decreases notably with elevating Co substitution level, which is due to Co substitution introducing extra electrons into the system. In addition, the hole concentration in our Ce 0.85 Fe 3.03 Co 0.97 Sb 12 sample is 9.5ˆ10 20 cm´3, less than half of that in the ambient-pressure synthesized Ce 0.9 Fe 3 CoSb 12 (nominal composition) [21]. The difference can be attributed to the higher filling fraction of Ce in our HPS sample, and accounts for the significant increase in the electrical resistivity of our HPS Ce 0.85 Fe 3.03 Co 0.97 Sb 12 sample. The hole concentrations can also be calculated by simple charge counting based on the chemical compositions, which differ from those listed in Table 1. A similar difference was previously reported [11,32], and might be related to the multiple bands near the Fermi level in elemental-filled FeSb 3 materials where a simple relationship between carrier concentration and Hall coefficient does not exist. Note that the carrier mobility is continually suppressed with increasing Co level, indicating a decreased hole mean free path caused by impurity scattering.
The temperature dependent thermoelectric properties of Ce y Fe 4´x Co x Sb 12 samples are shown in Figure 2. The electrical resistivity increases with elevating Co substitution level (Figure 2a), which can be attributed to the decreased hole concentration compensated by the additional electrons from cobalt. The most Co-substituted Ce 0.85 Fe 3.03 Co 0.97 Sb 12 exhibits the largest electrical resistivity, which starts to decrease at high temperature due to thermally activated carriers. The other two samples with lower Co substitution level show a heavily doped semiconductor behavior and the resistivities increase with elevating temperature due to stronger scattering. As the hole concentration decreases with higher Co substitution level, an increase in the Seebeck coefficient occurs (Figure 2b). While the most Co-substituted Ce 0.85 Fe 3.03 Co 0.97 Sb 12 shows an early onset of intrinsic regime of conduction occurring at ca. 700 K, the Seebeck coefficient for the other two samples increases monotonically with elevating temperature. The tendencies of the Seebeck coefficient with respect to temperature and Co substitution level are consistent with those of the electrical resistivity.
Materials 2016, 9,257 4 lower Co substitution level show a heavily doped semiconductor behavior and the resistivities increase with elevating temperature due to stronger scattering. As the hole concentration decreases with higher Co substitution level, an increase in the Seebeck coefficient occurs (Figure 2b). While the most Co-substituted Ce0.85Fe3.03Co0.97Sb12 shows an early onset of intrinsic regime of conduction occurring at ca. 700 K, the Seebeck coefficient for the other two samples increases monotonically with elevating temperature. The tendencies of the Seebeck coefficient with respect to temperature and Co substitution level are consistent with those of the electrical resistivity. The temperature dependent power factor is presented in Figure 2c. Compared with the increase of S as a function of Co substitution level, the increase of  is more significant, leading to decreased peak values of PF with increasing Co substitution level. The maximal PF of 3200 µ W m −1 K −2 is achieved for Ce0.92Fe4Sb12, which is comparable to the best values of other single elemental filled FeSb3 skutterudites [11]. We note a remarkably reduced PF in Ce0.85Fe3.03Co0.97Sb12 sample. In this sample, the actual filling fraction of Ce is higher than that in the ambient-pressure synthesized sample with the same Co level [24], leading to a lower hole concentration and a significantly increased electrical resistivity.
The temperature dependence of thermal conductivity is shown in Figure 2d for CeyFe4-xCoxSb12 samples. The variation tendency of  with respect to temperature is similar to that of CeyFe4-(x/2)Nix/2Sb12 samples [24]. The total thermal conductivity is a sum of the lattice thermal conductivity (L) and the carrier contribution c, c can be evaluated with the WiedemannFranz law, The temperature dependent power factor is presented in Figure 2c. Compared with the increase of S as a function of Co substitution level, the increase of ρ is more significant, leading to decreased peak values of PF with increasing Co substitution level. The maximal PF of 3200 µW m´1 K´2 is achieved for Ce 0.92 Fe 4 Sb 12 , which is comparable to the best values of other single elemental filled FeSb 3 skutterudites [11]. We note a remarkably reduced PF in Ce 0.85 Fe 3.03 Co 0.97 Sb 12 sample. In this sample, the actual filling fraction of Ce is higher than that in the ambient-pressure synthesized sample with the same Co level [24], leading to a lower hole concentration and a significantly increased electrical resistivity.
The temperature dependence of thermal conductivity is shown in Figure 2d for Ce y Fe 4´x Co x Sb 12 samples.
The variation tendency of κ with respect to temperature is similar to that of Ce y Fe 4-(x/2) Ni x/2 Sb 12 samples [24]. The total thermal conductivity is a sum of the lattice thermal conductivity (κ L ) and the carrier contribution κ c , κ " κ L`κc (3) κ c can be evaluated with the Wiedemann´Franz law, where L is the Lorenz number and a value of 2ˆ10´8 WΩ K´2 is used here [33]. The temperature dependent κ L can then be estimated and is shown as the inset of Figure 2d. The minimum κ L of 0.71 Wm´1 K´1 is reached in Ce 0.92 Fe 4 Sb 12 at 700 K. For all the samples, κ L first decreases with increasing temperature, and then increases slightly at high temperature due to the bipolar effect. The onset of bipolar effect shifts to higher temperature for the sample with lower Co substitution level, which can be attributed to the higher hole concentration suppressing the bipolar effect. For Ce y Fe 4´x Co x Sb 12 samples, it is plausible that both impurity atoms (Co) and rattling atoms (Ce) contribute to phonon scattering and reduce κ L . However, κ L increases with elevating Co substitution level. Similar results were previously observed in Ni-doped CeFe 4 Sb 12 systems [24]. These results indicate that impurity scattering of phonon due to mass/force constant fluctuation of Fe and Co is less significant for our samples, and the reduction of κ L is dominated by the rattling fillers: higher filling fraction leads to lower κ L . One of the effective techniques to confirm such kinds of localized vibrational modes is the low temperature heat capacity (C P ) measurement [34][35][36], which was performed on Ce 0.92 Fe 4 Sb 12 in the temperature range of 5´25 K. The heat capacity as a function of temperature is well described by the sum of an electronic term, γT, and phonon terms, which are the Einstein term for the loosely bonded filler atom and the Debye T 3 term for the 16 atoms of the framework. The extra Debye T 5 term can be neglected at low temperature [32,37]. As shown in Figure 3, C P is well fitted by where γ = 0.281 J K´2 mol´1, β = 0.00159 J K´4 mol´1, (corresponding to a Debye temperature of 269 K), A = 22.05 J K´1 mol´1, and x = Θ E /T with an Einstein temperature Θ E = 89.4 K. The fitting quality is significantly good with χ 2 = 5.62ˆ10´3 and R 2 = 0.99996. The Einstein temperature determined in this work is comparable with that (86 K) reported previously [38]. In addition, the filling fraction determined from A/3R (R is the ideal gas constant) is 0.88, which is very close to the value of 0.92 from EPMA measurement and indicates successful filling of Ce atoms into FeSb 3 voids.
Materials 2016, 9,257 c = LT/ρ (4) where L is the Lorenz number and a value of 2 × 10 −8 WΩ K −2 is used here [33]. The temperature dependent L can then be estimated and is shown as the inset of Figure 2d. The minimum L of 0.71 Wm −1 K −1 is reached in Ce0.92Fe4Sb12 at 700 K. For all the samples, L first decreases with increasing temperature, and then increases slightly at high temperature due to the bipolar effect. The onset of bipolar effect shifts to higher temperature for the sample with lower Co substitution level, which can be attributed to the higher hole concentration suppressing the bipolar effect. For CeyFe4-xCoxSb12 samples, it is plausible that both impurity atoms (Co) and rattling atoms (Ce) contribute to phonon scattering and reduce L. However, L increases with elevating Co substitution level. Similar results were previously observed in Ni-doped CeFe4Sb12 systems [24]. These results indicate that impurity scattering of phonon due to mass/force constant fluctuation of Fe and Co is less significant for our samples, and the reduction of L is dominated by the rattling fillers: higher filling fraction leads to lower L.
One of the effective techniques to confirm such kinds of localized vibrational modes is the low temperature heat capacity (CP) measurement [34][35][36], which was performed on Ce0.92Fe4Sb12 in the temperature range of 525 K. The heat capacity as a function of temperature is well described by the sum of an electronic term, γT, and phonon terms, which are the Einstein term for the loosely bonded filler atom and the Debye T 3 term for the 16 atoms of the framework. The extra Debye T 5 term can be neglected at low temperature [32,37]. As shown in Figure 3, CP is well fitted by where γ = 0.281J K −2 mol −1 , β = 0.00159 J K −4 mol −1 , (corresponding to a Debye temperature of 269 K), A = 22.05 J K −1 mol −1 , and x = Θ E /T with an Einstein temperature Θ E = 89.4K. The fitting quality is significantly good with χ 2 = 5.62 × 10 −3 and R 2 = 0.99996. The Einstein temperature determined in this work is comparable with that (86 K) reported previously [38]. In addition, the filling fraction determined from A/3R (R is the ideal gas constant) is 0.88, which is very close to the value of 0.92 from EPMA measurement and indicates successful filling of Ce atoms into FeSb3 voids.  Figure 4 shows ZT of CeyFe4-xCoxSb12 samples calculated from the measured thermoelectric properties. For the optimal Ce0.92Fe4Sb12 sample, ZT increases with temperature and reaches the highest value of 0.91 at 763 K, which is slightly higher than that achieved in ambient-pressure synthesized sample [11]. It is noted the thermoelectric performance of our HPS CeyFe4-xCoxSb12  Figure 4 shows ZT of Ce y Fe 4´x Co x Sb 12 samples calculated from the measured thermoelectric properties. For the optimal Ce 0.92 Fe 4 Sb 12 sample, ZT increases with temperature and reaches the highest value of 0.91 at 763 K, which is slightly higher than that achieved in ambient-pressure synthesized sample [11]. It is noted the thermoelectric performance of our HPS Ce y Fe 4´x Co x Sb 12 samples deteriorates with increasing Co substitution level, which differs from those ambient-pressure synthesized samples. For example, the highest ZT of 0.9 was achieved in Ce 0.9 Fe 3 CoSb 12 [21] and CeFe 3.8 Co 0.2 Sb 12 [22]. In our samples, the actual filling fraction of Ce only slightly decreases with increasing Co content, which leads to a much smaller hole concentration and an undesirable increase in ρ compared with those ambient-pressure synthesized samples. Consequentially, an enhancement in thermoelectric performance does not occur through substituting framework atoms with electron-richer atoms alone in our HPS samples.
Materials 2016, 9,257 samples deteriorates with increasing Co substitution level, which differs from those ambientpressure synthesized samples. For example, the highest ZT of 0.9 was achieved in Ce0.9Fe3CoSb12 [21] and CeFe3.8Co0.2Sb12 [22]. In our samples, the actual filling fraction of Ce only slightly decreases with increasing Co content, which leads to a much smaller hole concentration and an undesirable increase in  compared with those ambient-pressure synthesized samples. Consequentially, an enhancement in thermoelectric performance does not occur through substituting framework atoms with electronricher atoms alone in our HPS samples. Nonetheless, HPS methods employed in this study demonstrates an advantage over the solid state reaction method at ambient pressure. HPS significantly accelerates the fabrication process and shortens the total reaction duration to several hours, and allows us to tune the filling fraction in a wider range. In addition, the synthetic pressure and temperature conditions can easily be realized in industrial production [39], another advantage for massive production for thermoelectric applications. Combined with the strategies of multiple filling/substituting and nanostructuring (through, e.g., ball milling and severe plastic deformation) [9,10,40,41], further enhancement of ZT values of p-type skutterudites could be realized with the HPS method.

Conclusions
p-type CeyFe4-xCoxSb12 skutterudites were successfully synthesized by using an efficient HPS method within several hours. The highest filling fraction of Ce (y = 0.92) was reached in the Co-free sample. With increasing Co substitution level, both hole concentration and Ce filling fraction decrease. The enhancement in S due to a smaller hole concentration is offset by the increase in . Moreover, the suppression of L seems dominated by the rattling filler atoms, and lower filling fraction of Ce with higher Co content is unfavorable to reduce L. The maximal ZT of 0.91 was achieved in the optimal Ce0.92Fe4Sb12 sample, and the thermoelectric performance of CeyFe4-xCoxSb12 deteriorates with higher Co content.  Nonetheless, HPS methods employed in this study demonstrates an advantage over the solid state reaction method at ambient pressure. HPS significantly accelerates the fabrication process and shortens the total reaction duration to several hours, and allows us to tune the filling fraction in a wider range. In addition, the synthetic pressure and temperature conditions can easily be realized in industrial production [39], another advantage for massive production for thermoelectric applications. Combined with the strategies of multiple filling/substituting and nanostructuring (through, e.g., ball milling and severe plastic deformation) [9,10,40,41], further enhancement of ZT values of p-type skutterudites could be realized with the HPS method.

Conclusions
p-type Ce y Fe 4´x Co x Sb 12 skutterudites were successfully synthesized by using an efficient HPS method within several hours. The highest filling fraction of Ce (y = 0.92) was reached in the Co-free sample. With increasing Co substitution level, both hole concentration and Ce filling fraction decrease. The enhancement in S due to a smaller hole concentration is offset by the increase in ρ. Moreover, the suppression of κ L seems dominated by the rattling filler atoms, and lower filling fraction of Ce with higher Co content is unfavorable to reduce κ L . The maximal ZT of 0.91 was achieved in the optimal Ce 0.92 Fe 4 Sb 12 sample, and the thermoelectric performance of Ce y Fe 4´x Co x Sb 12 deteriorates with higher Co content.