Effect of ZnO and SnO2 Nanolayers at Grain Boundaries on Thermoelectric Properties of Polycrystalline Skutterudites

Nanostructuring is considered one of the key approaches to achieve highly efficient thermoelectric alloys by reducing thermal conductivity. In this study, we investigated the effect of oxide (ZnO and SnO2) nanolayers at the grain boundaries of polycrystalline In0.2Yb0.1Co4Sb12 skutterudites on their electrical and thermal transport properties. Skutterudite powders with oxide nanolayers were prepared by atomic layer deposition method, and the number of deposition cycles was varied to control the coating thickness. The coated powders were consolidated by spark plasma sintering. With increasing number of deposition cycle, the electrical conductivity gradually decreased, while the Seebeck coefficient changed insignificantly; this indicates that the carrier mobility decreased due to the oxide nanolayers. In contrast, the lattice thermal conductivity increased with an increase in the number of deposition cycles, demonstrating the reduction in phonon scattering by grain boundaries owing to the oxide nanolayers. Thus, we could easily control the thermoelectric properties of skutterudite materials through adjusting the oxide nanolayer by atomic layer deposition method.


Introduction
The regulation of thermal energy is a global issue and critical for reducing global warming, which causes climate change. Various renewable and sustainable-energy-related technologies, such as energy storage and conversion materials, have been researched as alternatives to fossil fuel energy technologies. The thermoelectric (TE) technology, which allows the direct conversion of thermal energy into electrical energy without generating any pollutant, is considered the key solution to control the abundant waste heat energy [1][2][3]. The conversion efficiency of TE power generation devices is determined by the thermoelectric dimensionless figure of merit zT, which can be calculated by the equation zT = σ · S 2 · T/κ, where S is the Seebeck coefficient, σ is the electrical conductivity, T is the absolute temperature, and κ is the total thermal conductivity of the materials. The κ includes electronic (κ elec ) and lattice (κ latt ) contributions, with κ elec governed by Wiedemann-Frantz law (κ elec = L · σ · T, where L is Lorenz number) and κ latt determined by phonon scattering. Therefore, it is necessary to develop a facile method to enhance the electronic transport properties by reducing the thermal transport properties to achieve high TE performance [4][5][6].
Various TE materials have been studied in terms of their operating temperature [3]. Among them, skutterudite (SKD)-based compounds are considered one of the best candidates for medium-to-high-temperature applications owing to their good performance, stability, and mechanical properties [7,8]. Filled SKD materials with a composition of R x T 4 Pn 12 (R = rare earth, actinide, alkali, and alkaline-earth metal; T = transition metal; Pn = pnictogen atom) composed of a cubic lattice of T 4 Pn 12 and R elements as fillers (or rattlers), where the filler element tends to rattle, resulting in the effective scattering of lattice phonons, exhibit a low κ, representing the phonon-glass-electron-crystal model. Co 4 Sb 12 -based SKD materials show n-type conduction, and a variety of filler materials have been explored for Co 4 Sb 12 -based compounds [9][10][11][12][13][14][15]. Single elements such as In or Yb-filled n-type SKD materials have been investigated to effectively modulate the TE properties in terms of phonon scattering. He et al. reported a zT value of 1.2 for an In 0.25 Co 4 Sb 12 n-type SKD material [9] and showed that the limit for In incorporation is~0.22. A Yb-filled SKD has shown a high zT value of 1.5 for a high filling fraction of~0.3 for Co 4 Sb 12 [11]. Moreover, multiple filler elements are substantially used for broad-frequency phonon scattering, leading to reach high TE performance over 1.4 [13,14]. However, the Co 4 Sb 12 mother matrix had a filling limitation for each filler element, which rendered the scattering of more phonons difficult. In addition to the control of filler composition, a different strategy such as grain boundary engineering is also required to enhance the phonon scattering of SKD materials to obtain high-performance TE materials.
Secondary phase incorporation is one of the promising technologies to enhance TE properties [15][16][17][18][19][20]. Much effort has been devoted to incorporating nanostructures such as nanoparticles or nanosheets into TE matrices. The atomic layer deposition (ALD) method, [21,22] which is typically used for thin film deposition, could be the best candidate for evaluating the effect of thin coating layer. Furthermore, the excellent conformality could be achieved even on the complex shaped substrate or powder materials [23]. The recent paper reveals that the introduction of ZnO on the grain boundary of Bi 0.4 Sb 1.6 Te 3 significantly reduces the κ latt without deteriorating the electronic transport properties [18]. The effect of nanolayers at the grain boundaries introduced by ALD on the electronic and thermal transport properties of other TE materials needs to be studied.
In this study, we prepared SKD-based complex TE materials by the ALD technique and investigated the effects of nanolayer coating at the grain boundaries of SKD materials. We successfully synthesized ZnO-and SnO 2 -nanolayer-coated In 0.2 Yb 0.1 Co 4 Sb 12 SKD powders by the ALD method and investigated the effect of the number of ALD cycles on the electronic and thermal transport properties.

Materials and Methods
An In 0.2 Yb 0.1 Co 4 Sb 12 ingot was prepared by a typical solid-state reaction method. High-purity Co (99.998%, Alfa Aesar, Ward Hill, MA, USA), Sb (99.999%, Alfa Aesar, Ward Hill, MA, USA), In (99.999%, Alfa Aesar, Ward Hill, MA, USA), and Yb (99.9%, Alfa Aesar, Ward Hill, MA, USA) were weighed to obtain the stoichiometric composition of In 0.2 Yb 0.1 Co 4 Sb 12 and charged into a carbon-coated quartz tube. The tube was vacuum-sealed to below 9.5 × 10 −3 Torr, and the raw materials were melted using induction melting furnace for 20 min. The ingot was sealed again and annealed under 700 • C for one week. The ingot was pulverized using a high-energy ball mill (8000D, SPEX, Metuchen, NJ, USA) for 5 min. The resulting powder was sieved to obtain fine powder.
ZnO and SnO 2 were deposited on the In 0.2 Yb 0.1 Co 4 Sb 12 powder by ALD (iSAC Co. Ltd., iOVd100) in the temperature range of 150-350 • C using SnCl 4 (99.995%, iChems Co. Ltd., Hwaseong, Korea) and (C 2 H 5 ) 2 Zn (iChems Co. Ltd., Hwaseong, Korea), respectively, and H 2 O was used as an oxidant. Each deposition cycle consisted of a pulse of precursors, N 2 purging, H 2 O pulse, and a second N 2 purging process. Each purging process lasted for 20 s to allow the sufficient exhaustion of residual gases and physisorbed precursors. The precursor was supplied to the chamber with a 50 sccm flow of N 2 gas. The deposition pressure was maintained as~0.6 Torr, and 50 sccm N 2 gas was continuously supplied to the chamber during the process. Each coating was applied to the In 0.2 Yb 0.1 Co 4 Sb 12 powder with 1, 3, 5, and 100 ALD cycles.
Polycrystalline bulk samples were prepared by spark plasma sintering at 873 K for 5 min under 50 MPa in vacuum. Our bulk samples have the high density more than 96% of the theoretical density of SKD. The temperature-dependent S and σ parameters were measured from room temperature to 860 K (ZEM-3 Advanced-RIKO, Yokohama, Japan). The κ values of the samples were calculated by the equation κ = ρ s · C p · λ, where ρ s , C p , and λ are the theoretical density, heat capacity, and thermal diffusivity, respectively. The λ values were measured using a laser flash installation (Laser Flash Analysis, LFA 467, Netzsch, Selb, Germany) from room temperature to 700 K. The microstructures of the ZnO and SnO 2 coatings were analyzed by bright-field transmission electron microscopy (TEM, JEM-F200, JEOL, Tokyo, Japan).  [18,19]. Nanomaterials 2020, 10, x 3 of 8 supplied to the chamber during the process. Each coating was applied to the In0.2Yb0.1Co4Sb12 powder with 1, 3, 5, and 100 ALD cycles. Polycrystalline bulk samples were prepared by spark plasma sintering at 873 K for 5 min under 50 MPa in vacuum. Our bulk samples have the high density more than 96% of the theoretical density of SKD. The temperature-dependent S and σ parameters were measured from room temperature to 860 K (ZEM-3 Advanced-RIKO, Yokohama, Japan). The κ values of the samples were calculated by the equation κ = ρs • Cp • λ, where ρs, Cp, and λ are the theoretical density, heat capacity, and thermal diffusivity, respectively. The λ values were measured using a laser flash installation (Laser Flash Analysis, LFA 467, Netzsch, Selb, Germany) from room temperature to 700 K. The microstructures of the ZnO and SnO2 coatings were analyzed by bright-field transmission electron microscopy (TEM, JEM-F200, JEOL, Tokyo, Japan). Figure 1 shows the high-resolution TEM (HRTEM) images of the ZnO and SnO2-coated In0.2Yb0.1Co4Sb12 SKD powders after 100 cycles of ALD. As can be seen, the thickness of the ZnO layers after 100 ALD cycles is ~26 nm (Figure 1a), while that of the SnO2 layer after 100 ALD cycles is ~15 nm (Figure 1c). From the coating thicknesses, the ZnO and SnO2 ALD deposition rates were calculated as 2.60 Å/cycle and 1.50 Å/cycle, respectively. The polycrystalline ZnO and SnO2 layers were homogeneously coated on the In0.2Yb0.1Co4Sb12 SKD powder (Figure 1b,d). In addition, the selected area electron diffraction (SAED) patterns (inset images in Figure 1b,d) of the powder show concentric rings and well defined spots corresponding to the (100), (101), and (110) planes of hexagonal wurtzite ZnO and the (111) plane of orthorhombic SnO2, respectively. Thus, compared with other ALD methods reported in literature, our ALD method could synthesize homogeneous coating layers of ZnO and SnO2 on the In0.2Yb0.1Co4Sb12 SKD powder [18,19].  Figure 2 shows the high-angle annular dark field (HAADF) and energy dispersive X-ray spectroscopy (EDS) mapping images of the ZnO and SnO2-coated In0.2Yb0.1Co4Sb12 SKD powders after  Figure 2 shows the high-angle annular dark field (HAADF) and energy dispersive X-ray spectroscopy (EDS) mapping images of the ZnO and SnO 2 -coated In 0.2 Yb 0.1 Co 4 Sb 12 SKD powders after 100 ALD cycles. The bright region corresponds to the heavy elements of the In 0.2 Yb 0.1 Co 4 Sb 12 matrix, whereas the dark region corresponds to the light elements of the ZnO or SnO 2 coating layers (Figure 2a,d). The EDS mapping images of ZnO- (Figure 2b,c) and SnO 2 -coated In 0.2 Yb 0.1 Co 4 Sb 12 SKD powders (Figure 2e,f) indicate the uniform coating of ZnO and SnO 2 on each In 0.2 Yb 0.1 Co 4 Sb 12 particle. This confirms that the facile ALD coating technology can produce a uniform coating of ZnO and SnO 2 layers on the In 0.2 Yb 0.1 Co 4 Sb 12 SKD powder. We expect that the homogeneous coating of oxide layer would affect the TE properties of the In 0.2 Yb 0.1 Co 4 Sb 12 SKD material. Nanomaterials 2020, 10, x 4 of 8 100 ALD cycles. The bright region corresponds to the heavy elements of the In0.2Yb0.1Co4Sb12 matrix, whereas the dark region corresponds to the light elements of the ZnO or SnO2 coating layers ( Figure  2a,d). The EDS mapping images of ZnO- (Figure 2b,c) and SnO2-coated In0.2Yb0.1Co4Sb12 SKD powders (Figure 2e,f) indicate the uniform coating of ZnO and SnO2 on each In0.2Yb0.1Co4Sb12 particle. This confirms that the facile ALD coating technology can produce a uniform coating of ZnO and SnO2 layers on the In0.2Yb0.1Co4Sb12 SKD powder. We expect that the homogeneous coating of oxide layer would affect the TE properties of the In0.2Yb0.1Co4Sb12 SKD material.  Figure 3 shows the temperature-dependent electrical transport properties (σ and S) of pristine and ZnO and SnO2-coated In0.2Yb0.1Co4Sb12. The σ of pristine In0.2Yb0.1Co4Sb12 decreased with an increase in temperature, exhibiting a metallic conduction behavior (Figure 3a). However, the conduction became less metallic with the coating of a ZnO layer, and the σ of the ZnO-coated In0.2Yb0.1Co4Sb12 after 100 ALD cycles decreased with increasing temperature, exhibiting a semiconducting behavior. With increasing number of ALD cycles, the σ decreased over the entire temperature range because of the possible electron scattering by the oxide nanolayers formed at the grain boundaries. However, interestingly, the S changed only slightly even after 100 ALD cycles (Figure 3b). This indicates that the oxide coating only affects the carrier mobility. The S at room temperature varied from 100 µV/K to 112 µV/K, while the σ at room temperature decreased from 1988 S/cm for pristine In0.2Yb0.1Co4Sb12 to 410 S/cm for ZnO-coated In0.2Yb0.1Co4Sb12 after 100 ALD cycles. Therefore, the power factor is mainly governed by the σ in this case; hence, pristine In0.2Yb0.1Co4Sb12 exhibited the highest power factor of 2.18 mW/mK 2 (Figure 3c,f). For SnO2 coatings, the temperature dependences of σ and S are similar to those observed for ZnO coatings, while SnO2coated In0.2Yb0.1Co4Sb12 exhibited a slight metallic conduction behavior even after 100 ALD cycles (Figure 3d). The S did not change significantly with increasing number of ALD coating cycles ( Figure  3e). Overall, the results indicate that the coating of ZnO and SnO2 is unfavorable for the electrical transport properties of In0.2Yb0.1Co4Sb12.  Figure 3 shows the temperature-dependent electrical transport properties (σ and S) of pristine and ZnO and SnO 2 -coated In 0.2 Yb 0.1 Co 4 Sb 12 . The σ of pristine In 0.2 Yb 0.1 Co 4 Sb 12 decreased with an increase in temperature, exhibiting a metallic conduction behavior (Figure 3a). However, the conduction became less metallic with the coating of a ZnO layer, and the σ of the ZnO-coated In 0.2 Yb 0.1 Co 4 Sb 12 after 100 ALD cycles decreased with increasing temperature, exhibiting a semiconducting behavior. With increasing number of ALD cycles, the σ decreased over the entire temperature range because of the possible electron scattering by the oxide nanolayers formed at the grain boundaries. However, interestingly, the S changed only slightly even after 100 ALD cycles (Figure 3b). This indicates that the oxide coating only affects the carrier mobility. The S at room temperature varied from 100 µV/K to 112 µV/K, while the σ at room temperature decreased from 1988 S/cm for pristine In 0.2 Yb 0.1 Co 4 Sb 12 to 410 S/cm for ZnO-coated In 0.2 Yb 0.1 Co 4 Sb 12 after 100 ALD cycles. Therefore, the power factor is mainly governed by the σ in this case; hence, pristine In 0.2 Yb 0.1 Co 4 Sb 12 exhibited the highest power factor of 2.18 mW/mK 2 (Figure 3c,f). For SnO 2 coatings, the temperature dependences of σ and S are similar to those observed for ZnO coatings, while SnO 2 -coated In 0.2 Yb 0.1 Co 4 Sb 12 exhibited a slight metallic conduction behavior even after 100 ALD cycles (Figure 3d). The S did not change significantly with increasing number of ALD coating cycles (Figure 3e). Overall, the results indicate that the coating of ZnO and SnO 2 is unfavorable for the electrical transport properties of In 0.2 Yb 0.1 Co 4 Sb 12 . Nanomaterials 2020, 10, x 5 of 8 The measured κ and κlatt values are shown in Figure 4. As shown in Figure 4a,b, the samples exhibit similar κ values of 3.5 W/mK at room temperature. However, because of the decrease in σ due to the coatings, an increase in κlatt is expected despite the unchanged κ. Figure 4c

exp | | 116
(1) The calculated κlatt for both the coatings increased with increasing number of ALD cycles. The insets of Figure 4c,d show the κlatt at room temperature. As can be seen, the κlatt gradually increased from 2.35 to 2.85 W/mK with an increase in the number of ZnO coating cycles to five.
The κlatt is represented by Debye-Callaway model as Equation (2): where ν, τtot, kB, ħ, θa, and z are the phonon group velocity, total phonon relaxation time, Boltzmann constant, reduced Planck's constant, Debye temperature, and ħω/kBT, respectively. The τtot in Equation (2) is estimated from the individual relaxation time for different defect structures (τi) by Equation (3), while the inverse relaxation time for boundary phonon scattering (τB −1 ) is expressed as Equation (4).
Here, αt and d are the grain boundary transmission coefficient and grain size, respectively. Therefore, based on the underlying physics, changes in αt and d would lead to a modification of κlatt. In Equation (4), the αt was fitted to the experimental κlatt (symbols in Figure 4c,d) with assuming αt = 1.0 for In0.2Yb0.1Co4Sb12, while d is assumed as ~500 nm. For ZnO-coated samples, the αt becomes systematically larger to 1.11, 1.22 and 1.40 for. For SnO2 coated samples, the αt also increases The measured κ and κ latt values are shown in Figure 4. As shown in Figure 4a,b, the samples exhibit similar κ values of 3.5 W/mK at room temperature. However, because of the decrease in σ due to the coatings, an increase in κ latt is expected despite the unchanged κ. Figure 4c,d show the calculated κ latt by deducting κ elec calculated by Wiedemann-Frantz law (κ elec = L · σ · T), where L is the Lorenz number, which can be estimated by Equation (1) [24].
The calculated κ latt for both the coatings increased with increasing number of ALD cycles. The insets of Figure 4c,d show the κ latt at room temperature. As can be seen, the κ latt gradually increased from 2.35 to 2.85 W/mK with an increase in the number of ZnO coating cycles to five.
The κ latt is represented by Debye-Callaway model as Equation (2): where ν, τ tot , k B ,h, θa, and z are the phonon group velocity, total phonon relaxation time, Boltzmann constant, reduced Planck's constant, Debye temperature, andhω/k B T, respectively. The τ tot in Equation (2) is estimated from the individual relaxation time for different defect structures (τ i ) by Equation (3), while the inverse relaxation time for boundary phonon scattering (τ B −1 ) is expressed as Equation (4).
Here, α t and d are the grain boundary transmission coefficient and grain size, respectively. Therefore, based on the underlying physics, changes in α t and d would lead to a modification of κ latt . In Equation (4), the α t was fitted to the experimental κ latt (symbols in Figure 4c,d) with assuming α t = 1.0 for In 0.2 Yb 0.1 Co 4 Sb 12 , while d is assumed as~500 nm. For ZnO-coated samples, the α t becomes systematically larger to 1.11, 1.22 and 1.40 for. For SnO 2 coated samples, the α t also increases Nanomaterials 2020, 10, 2270 6 of 8 systematically larger to 1.11, 1.18 and 1.29 for one, three, and five cycles. Although the fitting at high temperatures shows a discrepancy for SnO 2 -coated samples, the general trend of κ latt increase is maintained. The κ value generally affect both the intrinsic κ value of the certain material and the phonon scattering on the defect structures (e.g., grain boundaries). In our system, the intrinsically high κ values of ZnO and SnO 2 (~50 and~100 W/mK, respectively, at room temperature) seems to increase α t . Nanomaterials 2020, 10, x 6 of 8 systematically larger to 1.11, 1.18 and 1.29 for one, three, and five cycles. Although the fitting at high temperatures shows a discrepancy for SnO2-coated samples, the general trend of κlatt increase is maintained. The κ value generally affect both the intrinsic κ value of the certain material and the phonon scattering on the defect structures (e.g., grain boundaries). In our system, the intrinsically high κ values of ZnO and SnO2 (~50 and ~100 W/mK, respectively, at room temperature) seems to increase αt. The temperature-dependent zT values of ZnO-and SnO2-coated In0.2Yb0.1Co4Sb12 are shown in Figure 5. The zT values of pristine In0.2Yb0.1Co4Sb12 and that coated with ZnO with one, three, and five cycles of ALD are 0.69, 0.62, 0.63, and 0.63, respectively, at 700 K (Figure 5a). For SnO2-coated In0.2Yb0.1Co4Sb12, the zT decreased to 0.58, 0.53, and 0.55 after one, three and five ALD cycles, respectively (Figure 5b). The change in zT value by introducing ZnO and SnO2 nanolayer does not seem positive effect in our work due to the high intrinsic κ values of selected oxide materials. However, we suggest the ALD-coated structure to control the physical properties in SKD TE materials. We believe that the fine tuning of the composition of coating layer or condition of ALD process could realize the high TE performance in SKD materials. The temperature-dependent zT values of ZnO-and SnO 2 -coated In 0.2 Yb 0.1 Co 4 Sb 12 are shown in Figure 5. The zT values of pristine In 0.2 Yb 0.1 Co 4 Sb 12 and that coated with ZnO with one, three, and five cycles of ALD are 0.69, 0.62, 0.63, and 0.63, respectively, at 700 K (Figure 5a). For SnO 2 -coated In 0.2 Yb 0.1 Co 4 Sb 12 , the zT decreased to 0.58, 0.53, and 0.55 after one, three and five ALD cycles, respectively (Figure 5b). The change in zT value by introducing ZnO and SnO 2 nanolayer does not seem positive effect in our work due to the high intrinsic κ values of selected oxide materials. However, we suggest the ALD-coated structure to control the physical properties in SKD TE materials. We believe that the fine tuning of the composition of coating layer or condition of ALD process could realize the high TE performance in SKD materials.
In0.2Yb0.1Co4Sb12, the zT decreased to 0.58, 0.53, and 0.55 after one, three and five ALD cycles, respectively (Figure 5b). The change in zT value by introducing ZnO and SnO2 nanolayer does not seem positive effect in our work due to the high intrinsic κ values of selected oxide materials. However, we suggest the ALD-coated structure to control the physical properties in SKD TE materials. We believe that the fine tuning of the composition of coating layer or condition of ALD process could realize the high TE performance in SKD materials.

Conclusions
We synthesized polycrystalline In 0.2 Yb 0.1 Co 4 Sb 12 SKD with ZnO and SnO 2 oxide nanolayers at grain boundaries by an ALD process. The ALD method enables the coating of oxide nanolayers on the surfaces of SKD powders. With increasing numbers of deposition cycles, the electrical conductivity gradually decreased, while the Seebeck coefficient changed only slightly. The oxide nanolayers formed at the grain boundaries scatter the carriers, thus decreasing the mobility. In contrast, the lattice thermal conductivity increased with increasing number of deposition cycles, demonstrating the reduction in phonon scattering by grain boundaries owing to the highly thermally conductive oxide nanolayers. Thus, our work does not show the positive effect of oxide nanolayer coating on TE performance but gives a facile strategy for controlling electrical/thermal transport properties on a simple ALD process.