1. INTRODUCTION
Skutterudite is a natural mineral discovered in the Skutterud region of Norway, with the basic chemical formula of (Fe,Co,Ni)As3. Synthetic skutterudite belongs to the crystallographic space group Im3 ¯ and possesses a unit cell, as shown in Fig 1 [1]. The unit cell of skutterudite consists of 32 atoms arranged in 8MX3 groups, where M represents transition metals, such as Fe, Co, Ni, Ru, Rh, Pd, Os, Ir, and Pt, and X represents pnictogen elements, such as P, As, and Sb. The physical and chemical properties of skutterudite are determined by the elements occupying the M and X sites, and an improvement in thermoelectric performance can be expected through compositional optimization.
Skutterudite has two voids (□) per unit cell; thus, it can be expressed by the chemical composition formula □2M8X24 [2-5]. In general, the formula □M4X12 is represented as a half unit cell including one void [6]. If a guest element fills this void, it can be expressed as RM4X12 (R: alkaline-earth or rare-earth element). This is referred to as filled skutterudite [3-5,7], and the element occupying the void is called a filler or rattler. The element filling the void can enhance phonon scattering through independent lattice vibrations at specific resonant frequencies, thereby reducing the lattice thermal conductivity. This phenomenon is known as the rattling effect [3,4,8,9]. In addition to reducing lattice thermal conductivity through void filling, the thermoelectric performance of skutterudite can be improved by optimizing its carrier concentration through doping or partial substitution.
Generally, skutterudite is classified as n-type [Co4Sb12] based on cobalt, and p-type [Fe4Sb12]4-based on iron. Co4Sb12 is electronically stable with a valence electron count (VEC) of 72, whereas [Fe4Sb12]4-is electronically unstable with a VEC of 68, indicating a deficiency of four electrons [10]. Fig 2 shows the equilibrium phase diagram of the Fe-Sb binary system [11]. Only two kinds of phases are stable: the marcasite phase (FeSb2) formed through a peritectic reaction at 738 °C (1011 K) and the ε-phase formed at 1019 °C (1292 K). However, Fe-based skutterudite, Fe4Sb12 (FeSb3), cannot be formed solely from Fe and Sb. By filling the voids in skutterudite, it is possible to stabilize the compound in the form of R4+[Fe4Sb12]4-. However, most filling elements have a valence state of R2+ or R3+, resulting in a deficiency of one or two valence electrons. Consequently, it exists as an unstable phase while simultaneously exhibiting high carrier concentration [12]. A high carrier concentration causes the Fermi level (EF) to be positioned deep within the valence band (EV), leading to a small number of available energy states. Consequently, this leads to a low Seebeck coefficient and power factor [13]. Therefore, phase stabilization is required in the form of R3+[Fe3CoSb12]3- or R2+[Fe3NiSb12]2-by charge compensation with Co or Ni at the Fe site and optimization of carrier concentration.
In this study, single fillers of La/Ce/Pr/Nd/Yb were introduced into the voids of skutterudite, and p-type skutterudites were fabricated by substituting Co at the Fe site for charge compensation. The effects of the filling and charge compensation on the thermoelectric properties were compared. Furthermore, p-type skutterudites were prepared in bulky sintered bodies for the production (application) of thermoelectric modules, and the uniformity and reproducibility of their thermoelectric performance were evaluated.
2. SYNTHESIS AND ANALYSIS
P-type RyFe4-xCoxSb12 (y = 0.8-1.0; x = 0-1.0) skutterudites filled with rare-earth elements (R: La, Ce, Pr, Nd, and Yb) and charge-compensated with Co at the Fe site were synthesized by encapsulated melting [14-20]. The constituent elements La (purity 99.95%), Ce (purity 99.9%), Pr (purity 99.9%), Nd (purity 99.9%), Yb (purity 99.9%), Fe (purity 99.95%), Co (purity 99.95%), and Sb (purity 99.999%) were weighed according to their respective compositions. They were then charged into a quartz tube coated with carbon, sealed in a vacuum state, melted at 1323 K for 10 h, and then quenched in water. To control the reaction of the constituent elements, a two-step temperature increase interval was set: the temperature was increased to 873 K at a rate of 3 K/min and subsequently increased to 1323 K at a rate of 1 K/min. The obtained ingots were again sealed in a vacuum state and homogenized by heat treatment at 823-873 K for 24 h. After crushing the heat-treated ingots to 75 μm or less, they were charged into a graphite mold with inner diameters of 10 mm (10 Φ: specimens for measuring thermoelectric properties) and 37 mm (37 Φ: specimens for evaluating uniformity/reproducibility of thermoelectric performance) and then sintered by hot pressing (HP) at 873-898 K for 1 h under a pressure of 70 MPa (Fig 3).
X-ray diffraction (XRD; Bruker, D8-Advance) was conducted to analyze the phases of the samples. A Cu-Kα characteristic X-ray was radiated, and the standard diffraction data for skutterudite (ICDD PDF# 56-1123) were used to identify the phase. The Seebeck coefficient (α), electrical conductivity (σ), and thermal conductivity (κ) were measured in the temperature (T) range of 323-823 K to evaluate the power factor (PF = α2σ) and dimensionless figure of merit (ZT = α2σκ-1T) for thermoelectric performance. The HP compacts were processed to dimensions of 3 × 3 × 9 mm, and the Seebeck coefficient and electrical conductivity were measured using the ZEM-3 device (Advance Riko). After measuring the thermal diffusivity of a specimen processed to a diameter × thickness of 10 × 1 mm with a laser flash TC-9000H (Advance Riko) system, the thermal conductivity was evaluated using the density and specific heat of the sample.
3. THERMOELECTRIC PERFORMANCE
The element filling the voids of the skutterudite induces lattice scattering at a specific resonant frequency, determined by Equation 1 [21]:
where ωo is the resonant frequency, k is the spring constant, and m is the atomic mass of the filling element. In other words, a lower resonant frequency can be obtained from a filling element with a higher atomic mass. The resonant frequencies of the elements used as fillers in this study are listed in Table 1 [15,22,23]. Therefore, by filling with rareearth elements that are heavier than alkaline-earth elements, the lattice thermal conductivity (κL) can be effectively reduced by phonon scattering at a low resonant frequency.
In addition, obtaining a low lattice thermal conductivity is possible when the filling element has a smaller ionic radius [24], as represented by Equation 2:
where WL is the lattice thermal resistivity, WM is the thermal resistivity of the unfilled skutterudite, WPD is the thermal resistivity owing to point defects, WR is the thermal resistivity owing to resonance scattering, rcage is the void radius of the skutterudite, and rion is the ionic radius of the filling element. That is, a smaller ionic radius of the filling element leads to a higher lattice thermal resistivity (lower lattice thermal conductivity).
Meanwhile, filling an element with a smaller ionic radius among rare-earth elements with a high atomic mass is effective in reducing the lattice thermal conductivity, but structural stability cannot be ignored. In the periodic table of elements, excluding Eu and Yb, the atomic mass of rare-earth elements increases and ionic radius decreases as the atomic number increases. Elements filling the voids of skutterudite undergo thermal vibrations while being loosely coupled with the host elements in the lattice. Elements with an ionic radius greater than the void radius (189.2 pm) cannot fill the voids, while when elements with a small ionic radius are filled, the bonding strength is weakened; thus, the rattling effect cannot be expected because they escape from the voids during thermal vibration. Therefore, to achieve effective phonon scattering and phase stabilization in skutterudite materials it is crucial to fill elements with both a high atomic mass and appropriate ionic radius [21,25].
Table 1 summarizes the characteristics of the charge compensation elements used for phase stabilization and electrical property improvement in a p-type skutterudite. By charge-compensating with Co or Ni for Fe, phase stabilization in the form of R3+[Fe3CoSb12]3- or R2+[Fe3NiSb12]2- and optimization of carrier concentration are possible.
In this paper, the thermoelectric properties of p-type RyFe4-xCoxSb12 sintered pellets with a diameter of 10 Φ fabricated by the EM-HP process are summarized and compared. The PF (Fig 4) and ZT (Fig 5) values of the samples without charge compensation and the optimal sample with charge compensation for each filled skutterudite are briefly plotted and compared [14-20].
3.1. LayFe4-xCoxSb12
The PF increased as the temperature increased and then became saturated. LaFe4Sb12 exhibited a maximum PF = 2.44-2.46 mWm-1K-2 at 623-723 K. For charge-compensated La0.9Fe3CoSb12, the PF decreased to 2.18 mWm-1K-2 at 723 K. This is because the carrier (hole) concentration is reduced by charge compensation; thus, the electrical conductivity decreases. The ZT increased with increasing temperature, and La0.9Fe3CoSb12, which had a high PF and low thermal conductivity, exhibited the highest value of ZT = 0.67 at 723 K. Compared to LaFe4Sb12 (ZT = 0.56 at 823 K), the ZT was improved by the partial filling and charge compensation. Qui et al. [24] reported a ZT = 0.75 at 800 K for LaFe4Sb12 prepared by hermetic melting and spark plasma sintering. The slightly lower ZT value for LaFe4Sb12 in this study was presumed to be due to the formation of secondary phases during the synthesis process.
3.2. CeyFe4-xCoxSb12
The PF of CeFe4Sb12 was in the range of 2.67-2.71 mWm-1K-2 at 723-823 K. However, the PF of Ce0.9Fe3CoSb12 decreased to 1.97 mWm-1K-2 at 723 K due to the decrease in carrier concentration by charge compensation. In addition, the onset temperature of the intrinsic transition shifted to lower temperatures, and the PF decreased at temperatures above 723 K. In the case of CeFe4Sb12, ZT increased with increasing temperature and exhibited a maximum ZT = 0.70 at 823 K. Meanwhile, in the case of Ce0.9Fe3CoSb12, ZT exhibited a maximum value of 0.63 at 723 K and then decreased. For Ce-filled skutterudites, it was considered that Co should not be charge-compensated for Fe, or should be substituted with x < 1.
3.3. PryFe4-xCoxSb12
All PryFe4-xCoxSb12 samples exhibited similar values, except for PrFe3CoSb12, which had the lowest PF. PrFe4Sb12 exhibited a maximum PF of 3.02 mWm-1K-2 at 823 K, whereas Pr0.8Fe3CoSb12 exhibited 2.87 mWm-1K-2 at 723 K. In this study, the PF values of PryFe4-xCoxSb12 were higher than those of LayFe4-xCoxSb12 (PF = 2.46 mWm-1K-2 at 623 K for LaFe4Sb12) and CeyFe4-xCoxSb12 (PF = 2.71 mWm-1K-2 at 823 K for CeFe4Sb12). This indicates that the Pr filling was more effective in increasing the PF than the La and Ce filling. PrFe4Sb12 achieved a ZT value of 0.80 at 823 K. However, the thermoelectric performance was improved by Co charge compensation; thus, Pr0.8Fe3CoSb12 exhibited the highest ZT = 0.89 at 723 K. This value was similar to the ZT = 0.87 at 750 K for PrFe4Sb12 fabricated by Qiu et al. [24].
3.4. NdyFe4-xCoxSb12
The maximum PF of NdFe4Sb12 was 2.91 mWm-1K-2 at 823 K, but Nd0.9Fe3.5Co0.5Sb12 exhibited an increased value of 3.00 mWm-1K-2 at 723 K. The PF values of NdyFe4-xCoxSb12 were similar to those of PryFe4-xCoxSb12 and therefore higher than those of LayFe4-xCoxSb12 and CeyFe4-xCoxSb12. It was judged that the filling of Nd or Pr was more effective in increasing the PF than the filling of La or Ce. NdFe4Sb12 exhibited a ZT = 0.78 at 823 K, and Nd0.9Fe3.5Co0.5Sb12 exhibited the highest ZT = 0.91 at 723 K, thereby improving the thermoelectric performance through Nd filling and Co charge compensation. Qiu et al. [24] reported a maximum ZT = 0.85 at 750 K for NdFe4Sb12.
3.5. YbyFe4-xCoxSb12
The PF of YbFe4Sb12 was 2.10 mWm-1K-2 at 723-823 K, but it increased to 2.40-2.41 mWm-1K-2 at 723-823 K for Yb0.9Fe3CoSb12 due to Co charge compensation. The PF values of YbyFe4-xCoxSb12 were similar to those of LayFe4-xCoxSb12 and CeyFe4-xCoxSb12 and were lower than those of PryFe4-xCoxSb12 and NdyFe4-xCoxSb12. To enhance the PF values of p-type skutterudites, the filling effect of Yb was similar to that of La/Ce but smaller than that of Pr/Nd. As the temperature increased, the ZT values of YbyFe4-xCoxSb12 increased and reached a maximum between 723 and 823 K. YbFe4Sb12 exhibited ZT = 0.37 at 823 K; however, the ZT value for Yb0.9Fe3CoSb12 increased to 0.56 at 823 K.
4. UNIFORMITY AND REPRODUCIBILITY
To evaluate the uniformity and reproducibility of the thermoelectric performance of bulky skutterudite materials, sintered pellets with a size of 37 mm-diameter × 12 mmthickness (37 Φ × 12 t) were fabricated using the method described in Section 2. In this study, for Nd0.9Fe3.5Co0.5Sb12 with the best thermoelectric performance among the RyFe4-xCoxSb12 sintered pellets with a diameter of 10 Φ, the 37 Φ × 12 t sintered body were processed into five (1A-1E) cylinders (10 Φ × 12 t) by electrical discharge machining (Fig 6), and samples for XRD phase analysis and thermoelectric property measurement were prepared. After additionally fabricating a 37 Φ × 12 t sintered compact using the same process, the reproducibility of the thermoelectric performance of another five samples (2A-2E) was verified.
Fig 7 presents the XRD patterns for each position/batch of the bulky compacts (37 Φ × 12 t) of Nd0.9Fe3.5Co0.5Sb12 skutterudite. All samples were synthesized in the skutterudite phase, and the diffraction patterns matched the standard diffraction data for skutterudite (PDF# 56-1123). Thus, the uniformity and reproducibility of the skutterudite phase were confirmed.
Fig 8 shows the thermoelectric PF values of Nd0.9Fe3.5Co0.5Sb12 skutterudite bulky compacts (37 Φ × 12 t). In the first batch, a maximum PFmax = 2.81-3.19 mWm-1K-2 (average PFave = 3.04 mWm-1K-2 and standard deviation = 0.1731) was obtained at 723 K. In addition, for the second batch, PFmax = 3.09-3.19 mWm-1K-2 (PFave= 3.15 mWm-1K-2 and standard deviation = 0.0545) was achieved at 723 K. As shown in Section 3-4, Nd0.9Fe3.5Co0.5Sb12 with a diameter of 10 Φ exhibited PFmax = 2.99 mWm-1K-2 at 723 K. In Fig 8(b), the PF values of 10 Φ-Nd0.9Fe3.5Co0.5Sb12 are compared.
Fig 9 shows the dimensionless figures of merit of the Nd0.9Fe3.5Co0.5Sb12 bulky compacts (37 Φ × 12 t). In the first batch, the maximum ZTmax = 0.82-0.86 (average ZTave = 0.84 and standard deviation = 0.022) was obtained at 723 K. In addition, in the second batch, ZTmax = 0.86-0.91 (ZTave = 0.89 and standard deviation = 0.024) was achieved at 723 K. As shown in Section 3-4, Nd0.9Fe3.5Co0.5Sb12 with a diameter of 10 Φ exhibited ZTmax = 0.91 at 723 K. In Fig 9(b), the ZT values of 10 Φ-Nd0.9Fe3.5Co0.5Sb12 are compared.
5. SUMMARY
Rare-earth-filled and Co-charge-compensated p-type skutterudite compounds were synthesized and consolidated using the EM-HP process. The skutterudite phase stabilization and thermoelectric properties were investigated based on differences in valence, atomic mass, and ionic radius among the filling elements of La/Ce/Pr/Nd/Yb and the charge compensation of Co for Fe. The thermoelectric performances of the sintered samples (10 Φ × 12 t and 37 Φ × 12 t) of RyFe4-xCoxSb12 skutterudites are summarized in Table 2. The substitution of Co for Fe not only contributed to phase stabilization of the skutterudite by charge compensation but also reduced the lattice thermal conductivity through phonon scattering, and decreased electrical conductivity due to a decrease in carrier concentration. However, the addition of rareearth elements and Co doping increased the power factor and thermoelectric figure of merit, effectively improving the thermoelectric performance of p-type skutterudites. In addition, bulky samples of the best-performing p-type skutterudite, Nd0.9Fe3.5Co0.5Sb12, were produced to validate the uniformity and reproducibility of the thermoelectric performance.