Improvement in the thermoelectric performance of highly reproducible n-type (Bi,Sb)₂Se₃ alloys by Cl-doping

Abstract

(Bi,Sb)₂Se₃ alloys are promising alternatives to commercial n-type Bi₂(Te,Se)₃ ingots for low-mid temperature thermoelectric power generation due to their high thermoelectric conversion efficiency at elevated temperatures. Herein, we report the enhanced high-temperature thermoelectric performance of the polycrystalline Cl-doped Bi_2−xSb_xSe₃ (x = 0.8, 1.0) bulks and their sustainable thermal stability. Significant role of Cl substitution, characterized to enhance the power factor and reduce the thermal conductivity synergetically, is clearly elucidated. Cl-doping at Se-site of both Bi_1.2Sb_0.8Se₃ and BiSbSe₃ results in a high power factor by carrier generation and Hall mobility improvement while maintaining converged electronic band valleys. Furthermore, point defect phonon scattering originated from mass fluctuations formed at Cl-substituted Se-sites reduces the lattice thermal conductivity. Most importantly, spark plasma sintered Cl-doped Bi_2−xSb_xSe₃ bulks are thermally stable up to 700 K, and show a reproducible maximum thermoelectric figure of merit, zT, of 0.68 at 700 K.

---

1. Introduction

Bi₂Te₃-based alloys are the only commercialized thermoelectric materials for solid-state cooling and low-mid temperature (473–873 K) power generation, and their ingot-type materials are widely used due to a high thermoelectric figure of merit (zT = S²σT/κ_tot, where S, σ, κ_tot, and T are the Seebeck coefficient, electrical conductivity, total thermal conductivity, and the absolute temperature, respectively) of about 1.0 near room-temperature.¹ However, ingots of Bi₂Te₃-based alloys have a poor mechanical reliability (fracture strength of ∼10 MPa) because of 00l-oriented structure weakly bonded by van der Waals forces,² which limits their wider applications such as in automobile thermoelectric generator (ATEG). To address this, polycrystalline bulk form materials have been intensively studied, and an improved mechanical strength (∼80 MPa) with a higher zT ∼1.1 at 300 K has been obtained in micro-grained p-type Bi_2−xSb_xTe₃ prepared by ball milling (BM) and spark plasma sintering (SPS).² Its n-type counterpart with a comparable mechanical strength and zT is required to construct thermoelectric module with improved mechanical reliability as well as high performance, however, no marked improvement in zT from n-type micro-grained materials was achieved in BMed and SPSed Bi₂Te_2.7Se_0.3 (∼0.63 at 300 K).³ Furthermore, a severe reproducibility problem was also found in this polycrystalline sample owing to the uncontrollable defect structures such as vacancies (Te- or Se-site) and antisite defects. Polycrystalline bulk of n-type Bi₂Te_2.7Se_0.3 with high zT (∼0.98 at 300 K) and improved reproducibility has been demonstrated by combining a cold deformation and a hot extrusion benefitting from precise control of point defects,⁴ however, a simpler and yet easily scalable approach is always sought after.

Bi₂Se₃ is a narrow-bandgap layered semiconductor (space group R3̄m-D⁵_3d) with tetradymite structure and it has singly degenerate conduction band. The conduction band minimum (CBM) is observed at the center of the Brillouin zone (Γ-point) and the second conduction band is located 150–250 meV (Z-point) above the CBM,^5–7 thus the zT of pristine Bi₂Se₃ is very low (<0.1 at 300 K) mainly due to low S (∼−40 μV K⁻¹ at 300 K).⁸ High κ_tot ∼2.4 W m⁻¹ K⁻¹ at 300 K is another reason for the low zT of Bi₂Se₃. Very recently, Te-free (Bi,Sb)₂Se₃-based alloys have been received attention as promising alternatives to Bi₂(Te,Se)₃-based alloys especially for low-mid temperature power generation applications. High zT values of ∼1.0 and ∼1.4 were obtained at 800 K in micro-grained I- and Br-doped BiSbSe₃, respectively.^5,9 One main origin of the high thermoelectric performance of these compounds was the simultaneous improvement of electronic (enlarged S) and thermal transport properties (reduced lattice thermal conductivity (κ_lat)) due to a structural transition.⁵ A phase transition triggered the convergence of conduction band resulting in largely increased density of states (DOS) effective mass . Additionally, κ_lat was reduced due to the phonon softening and substantial lattice anharmonicity benefitting from weakened interchain interaction in orthorhombic phase. The high zTs were obtained by enhanced power factor (S²σ) of I- and Br-doped BiSbSe₃ (by an order of magnitude), resulted from the increase in carrier concentration (n_c). However, intrinsic drawbacks of I- and Br-doped BiSbSe₃, which included solubility limit of I (∼3 at%) and Hall mobility (μ_H) deterioration with Br doping, should be resolved to enhance the low σ value (<300 S cm⁻¹ at 300 K) in order to increase the efficiency of the thermoelectric power generation module. Moreover, the thermal stability and reproducibility of high-temperature thermoelectric performance was still elusive, demanding a clear criterion for the temperature limit of the module.

Chlorine (Cl) is a commonly used doping element especially at Se-site of various selenides such as In₄Se_3−x, PbSe, AgPb₁₈SbSe₂₀, and SnSe₂ to increase n_c.^10–13 Moreover, a large difference in atomic mass between Se (M_Se = 78.971) and Cl (M_Cl = 35.45) is advantageous to further reduce κ_lat by mass-defect phonon scattering. In this work, we fabricated the polycrystalline bulks of Cl-doped Bi_2−xSb_xSe₃ (x = 0.8, 1.0) and evaluated their electronic and thermal transport properties in an effort to develop (Bi,Sb)₂Se₃-based alloys with high σ and zT, at the same time. We found that Cl was an effective doping element to facilitate the carrier transport in Bi_1.2Sb_0.8Se₃, thus high μ_H of ∼27.3 cm² V⁻¹ s⁻¹ was observed even in highly Cl-doped Bi_1.2Sb_0.8Se_2.76Cl_0.24 with high n_c (∼9.0 × 10¹⁹ cm⁻³). Compared to I-doped Bi_1.2Sb_0.8Se_2.91I_0.09 (σ ∼80 S cm⁻¹ and zT ∼0.53 at 700 K),⁵ higher σ (∼165 S cm⁻¹ at 700 K) and higher zT (∼0.67 at 700 K) were obtained in Bi_1.2Sb_0.8Se_2.76Cl_0.24. Moreover, we confirmed the thermal stability of the sample by a cyclic measurement.

2. Experimental methods

High purity elements of bismuth (Bi shot, 99.99%, 5N Plus), antimony (Sb shot, 99.99%, 5N Plus), selenium (Se shot, 99.99%, iTASCO), and antimony trichloride (SbCl₃ crystalline, 99.99%, Sigma Aldrich) were used as starting raw materials. Stoichiometric (BiSbSe_3−yCl_y (y = 0, 0.12, 0.18, 0.24) and Bi_1.2Sb_0.8Se_3−zCl_z (z = 0, 0.12, 0.18, 0.24, 0.3)) amount of these materials was weighted and sealed into a quartz tube under vacuum (∼10⁻³ Pa), and the mixtures were melted at 1173 K for 12 h. After melting, the quartz tubes were water quenched and annealed at 673 K for 48 h. The acquired ingots were pulverized into powders via BM, and compacted bulks (13 mm in diameter and 11 mm in thickness) were fabricated by using SPS for 2 min at 773 K under a pressure of 40 MPa.

Phase formation behavior in SPSed bulks was analyzed by X-ray diffraction (XRD, SmartLab (9 kW), Rigaku, Japan) with CuKα radiation (λ = 1.5418 Å). The microstructures of the fractured surface of the SPSed bulks were investigated by scanning electron microscopy (SEM, JSM-7600F, JEOL, Japan). The temperature dependences of S and σ were measured using a commercial measurement system (ZEM-3, Ulvac-Riko, Japan) from 300–700 K under a He atmosphere. The κ_tot (=D × C_p × ρ, where D, C_p, and ρ are the thermal diffusivity, specific heat capacity, and the density, respectively) was calculated from the separate measurement of D and ρ. Temperature-dependent D was measured by laser flash method (TC-1200RH, Ulvac-Riko, Japan) from 300–700 K under a vacuum and ρ was measured by the Archimedes principle (MD-300S, Alfa Miracle, Japan). Temperature dependence of C_p was estimated from the reported values.⁵ The rectangular bar-type sample (10 mm × 3 mm × 3 mm) and square plate-type sample (10 mm × 10 mm × 1 mm) were cut in a plane perpendicular and parallel to the SPS press direction, respectively. In this manner, electronic (S, σ) and thermal (D) transport properties can be measured in the same direction. The Hall coefficient (R_H) was measured by the van der Pauw method via a commercial Hall effect measurement system (8400 Series, LakeShore, USA) at room temperature. The n_c and μ_H were calculated by n_c = e⁻¹R_H⁻¹ and μ_H = σR_H.

3. Results and discussion

In the present study, we selected two different matrixes; (1) BiSbSe₃ with an orthorhombic phase and (2) Bi_1.2Sb_0.8Se₃ with orthorhombic and rhombohedral phases.⁵ Fig. 1a shows the XRD patterns for BiSbSe_3−yCl_y samples. All the peaks can be indexed as a pure orthorhombic structure of Sb₂Se₃ without any secondary phases, suggesting the Cl substitution at Se-site. Structure factors including lattice parameters of BiSbSe₃-based compounds with orthorhombic phase (Fig. S1†) and those of Bi_1.2Sb_0.8Se₃-based compounds with mixed (orthorhombic and rhombohedral) phases (Fig. S3†) were extracted by the Rietveld refinement (GSAS II suite) after refinements with different structural models at the condition of convergence with the best pattern match. The slight decrease in lattice parameters (a and c) of BiSbSe₃ after the Cl doping is another evidence for Cl substitution due to the smaller ionic radius of Cl⁻ (167 pm) when compared to that of Se²⁻ (184 pm) (Fig. S1†). On the other hand, as reported in the previous report,¹⁴ both orthorhombic and rhombohedral phases are clearly detected in Bi_1.2Sb_0.8Se_3−zCl_z samples as shown in Fig. 1b. Peaks for the rhombohedral structure of Bi₂Se₃ were observed at 2θ ∼18.58° and ∼29.35°. The strongest intensity of (402) indicates the preferred crystal orientation generated during the SPS process. Oriented grain structure is also found in SEM images for the fractured surface of SPSed Bi_1.2Sb_0.8Se_3−zCl_z (Fig. 1c and S2†). The mole fraction of rhombohedral phase estimated by Rietveld refinement is about 0.74, and this value does not show significant change with Cl doping contents (see the Table S1†). And slight decrease in cell volume by Cl-doping is observed in orthorhombic phase (Fig. S3†). Cl-related impurity phase (Bi₃Se₄Cl) was observed in Bi_1.2Sb_0.8Se_2.7Cl_0.3, suggesting that the solubility limit of Cl for Se-site is about 8 at% in Bi_1.2Sb_0.8Se₃ (Fig. S4†).

Fig. 1 XRD patterns for the SPSed bulks of (a) BiSbSe_3−yCl_y (y = 0, 0.12, 0.18, 0.24) and (b) Bi_1.2Sb_0.8Se_3−zCl_z (z = 0, 0.12, 0.18, 0.24). (c) SEM image for the fractured surface of Bi_1.2Sb_0.8Se_2.76Cl_0.24.

The temperature dependences of σ for both BiSbSe_3−yCl_y (y = 0.12, 0.18, 0.24) and Bi_1.2Sb_0.8Se_3−zCl_z (z = 0.12, 0.18, 0.24) samples are plotted in Fig. 2a. All thermoelectric transport properties (σ, S, and κ) are measured perpendicular to SPS pressing direction since the electrical transport is dominant along the in-plane direction. The σ values of BiSbSe₃ and Bi_1.2Sb_0.8Se₃ are effectively increased by Cl-doping. Interestingly, the σ values of Cl-doped Bi_1.2Sb_0.8Se₃ are higher than those of Cl-doped BiSbSe₃ in the whole measured temperature range. The σ values of the BiSbSe_2.76Cl_0.24 are 132 S cm⁻¹ and 61.2 S cm⁻¹ at 300 K and 700 K, respectively, while those of Bi_1.2Sb_0.8Se_2.76Cl_0.24 are 397 S cm⁻¹ and 159 S cm⁻¹ at 300 K and 700 K, respectively. To clarify this, we estimated the n_c and μ_H of both Cl-doped BiSbSe₃ and Bi_1.2Sb_0.8Se₃ at 300 K (Fig. 2b). The improvement in σ by Cl-doping is resulted from the increase of μ_H as well as n_c both in BiSbSe₃ and Bi_1.2Sb_0.8Se₃. It is noted that μ_H values of Cl-doped Bi_1.2Sb_0.8Se₃ are much higher than those of Cl-doped BiSbSe₃. The μ_H values of BiSbSe_3−yCl_y (y = 0.12, 0.18, 0.24) at 300 K is ranged from 8.44 to 9.01 cm² V⁻¹ s⁻¹, whereas that of Bi_1.2Sb_0.8Se_2.88Cl_0.12 is ∼50.4 cm² V⁻¹ s⁻¹. Moreover, the μ_H value of highly Cl-doped BiSbSe_2.76Cl_0.24 is retained in value about 27.3 cm² V⁻¹ s⁻¹ despite of the high n_c ∼9.0 × 10¹⁹ cm⁻³. This high μ_H has been also reported in I-doped Bi_1.2Sb_0.8Se₃.⁵

Fig. 2 (a) Temperature dependence of electrical conductivity and (b) carrier concentration and Hall mobility for BiSbSe_3−yCl_y (y = 0.12, 0.18, 0.24) and Bi_1.2Sb_0.8Se_3−zCl_z (z = 0.12, 0.18, 0.24).

Unexpected difference between electronic transport properties of Cl-doped Bi_1.2Sb_0.8Se₃ and those of I-doped Bi_1.2Sb_0.8Se₃ is observed in S. Fig. 3a depicts the temperature dependences of S for both BiSbSe_3−yCl_y (y = 0.12, 0.18, 0.24) and Bi_1.2Sb_0.8Se_3−zCl_z (z = 0.12, 0.18, 0.24) samples. The S values of the all samples are negative in the whole measured temperature range, indicating n-type semiconductors. The large |S| values of Cl-doped BiSbSe₃ samples due to the convergence of conduction band by phase transition are well demonstrated both in I-doped and Br-doped BiSbSe₃.^5,9 To investigate the change in band structure by Cl-doping especially in Bi_1.2Sb_0.8Se₃, we calculate the by using measured S and n_c at 300 K based on the following eqn (1):¹

where k_B, e, and h are the Boltzmann constant, elementary charge, and Planck constant, respectively. The values are listed in Table 1 together with those for I-doped BiSbSe₃ and Bi_1.2Sb_0.8Se₃ samples, which are estimated from eqn (1) by using previously reported data.⁵ Large values of 1.55m₀ and 1.67m₀ are obtained both in BiSbSe_2.82Cl_0.18 and BiSbSe_2.91I_0.09, respectively, with pure orthorhombic phase benefiting from converged electronic band valleys. However, the values of Cl-doped Bi_1.2Sb_0.8Se₃ and those of I-doped Bi_1.2Sb_0.8Se₃ are smaller than those of Cl- and I-doped BiSbSe₃ mainly due to the large mole fraction of rhombohedral phase (Table S1†) with singly-degenerate conduction band. It is noted that value of Bi_1.2Sb_0.8Se_2.76Cl_0.24 reaches value about 0.90m₀, which results in a large S even in Bi_1.2Sb_0.8Se₃ systems. Fig. 3c shows the Pisarenko plots (n_c–|S|) for both BiSbSe_3−yCl_y (y = 0.12, 0.18, 0.24) and Bi_1.2Sb_0.8Se_3−zCl_z (z = 0.12, 0.18, 0.24) samples at 300 K. Those for I-doped BiSbSe₃, I-doped Bi_1.2Sb_0.8Se₃, and Br-doped BiSbSe₃ samples are also shown for comparison.^5,9

Fig. 3 Temperature dependence of (a) Seebeck coefficient and (b) power factor for BiSbSe_3−yCl_y (y = 0.12, 0.18, 0.24) and Bi_1.2Sb_0.8Se_3−zCl_z (z = 0.12, 0.18, 0.24). (c) Pisarenko plot for Cl-, I-, and Br-doped BiSbSe₃ and Cl- and I-doped Bi_1.2Sb_0.8Se₃ at 300 K.

Table 1 The density-of-states effective mass values of Cl-doped BiSbSe₃ and Bi_1.2Sb_0.8Se₃. Those of I-doped BiSbSe₃ and Bi_1.2Sb_0.8Se₃, which are estimated from the reported data ref. 5, are also shown for comparison

Compositions (nominal)	(m0)	Compositions (nominal)	(m0)
BiSbSe2.88Cl0.12	1.17	BiSbSe2.97I0.03	1.50
BiSbSe2.82Cl0.18	1.55	BiSbSe2.94I0.06	1.63
BiSbSe2.76Cl0.24	1.38	BiSbSe2.91I0.09	1.67
Bi1.2Sb0.8Se2.88Cl0.12	0.54	Bi1.2Sb0.8Se2.97I0.03	0.33
Bi1.2Sb0.8Se2.82Cl0.18	0.66	Bi1.2Sb0.8Se2.94I0.06	0.65
Bi1.2Sb0.8Se2.76Cl0.24	0.90	Bi1.2Sb0.8Se2.91I0.09	0.43

---

As clearly shown in Fig. 3c, similar value of S is obtained in Bi_1.2Sb_0.8Se_2.76Cl_0.24 despite of the large increase in n_c when compared to that of I-doped Bi_1.2Sb_0.8Se₃ samples. Resultantly, a maximum power factor values of ∼3.19 μW cm⁻¹ K⁻² and ∼5.61 μW cm⁻¹ K⁻² at 300 K and 700 K, respectively, are obtained in Bi_1.2Sb_0.8Se_2.76Cl_0.24 (Fig. 3b), which ensures the enhanced zT especially at higher temperatures. This beneficial characteristic feature for the realization of highly-efficient low-mid temperature thermoelectric power generation system is only found in Cl-doped Bi_1.2Sb_0.8Se₃ among other (Bi,Sb)₂Se₃-based alloys.

Fig. 4a shows the temperature dependence of κ_tot for BiSbSe_3−yCl_y (y = 0.12, 0.18, 0.24) and Bi_1.2Sb_0.8Se_3−zCl_z (z = 0.12, 0.18, 0.24) samples. The κ_tot values of Cl-doped Bi_1.2Sb_0.8Se₃ are higher than those of Cl-doped BiSbSe₃. The room temperature κ_tot values of both Cl-doped BiSbSe₃ and Bi_1.2Sb_0.8Se₃ are ∼0.58–0.62 W m⁻¹ K⁻¹ and ∼0.72–0.83 W m⁻¹ K⁻¹, respectively. This is considered to be related to the increased electronic contribution (κ_ele) originated from the higher σ of Cl-doped Bi_1.2Sb_0.8Se₃. On the other hand, as shown in Fig. 4a, the κ_tot of all the samples gradually decrease with temperature, suggesting the small contribution of bipolar thermal conduction (κ_bp). We estimated the κ_lat and κ_ele by using the relationship of κ_tot = κ_ele + κ_lat. Details for the calculation are described in Section 6 of ESI.†

Fig. 4 Temperature dependence of (a) total thermal conductivity and (b) lattice thermal conductivity for BiSbSe_3−yCl_y (y = 0.12, 0.18, 0.24) and Bi_1.2Sb_0.8Se_3−zCl_z (z = 0.12, 0.18, 0.24).

Fig. 4b shows the temperature dependence of κ_lat for both Cl-doped BiSbSe₃ and Bi_1.2Sb_0.8Se₃. The temperature dependence of κ_lat shows roughly κ_lat ∝ T^−0.5, indicating the additional point defect phonon scattering from the mass difference between host Se (M_Se = 78.971) and dopant Cl (M_Cl = 35.45) by Cl substituted at Se-site. The κ_lat values of Cl-doped BiSbSe₃ are lower than those of Cl-doped Bi_1.2Sb_0.8Se₃ mainly due to the soft bonding in orthorhombic phase, however, κ_lat reduction effect by Cl-doping is relatively small compared to that of Cl-doped Bi_1.2Sb_0.8Se₃ due to cumulative phonon scattering by soft bonding and point defect. Thus the significantly reduced κ_lat (∼0.56 W m⁻¹ K⁻¹ at 300 K and ∼0.39 W m⁻¹ K⁻¹ at 700 K) is obtained in Bi_1.2Sb_0.8Se_2.76Cl_0.24 mainly due to the intensified mass-defect phonon scattering. The slightly higher κ_lat of BiSbSe_2.82Cl_0.18 than that of BiSbSe_2.88Cl_0.12 is considered to be related with the difference in preferred orientation (Fig. 1b).

Fig. 5a and b show the temperature dependent zT of BiSbSe_3−yCl_y (y = 0.12, 0.18, 0.24) and that of Bi_1.2Sb_0.8Se_3−zCl_z (z = 0.12, 0.18, 0.24), respectively. The Cl-doping effectively enhances the zT both in BiSbSe₃ and Bi_1.2Sb_0.8Se₃ due to improvement of electronic and thermal transport properties. High zT of Cl-doped BiSbSe₃ with pure orthorhombic phase is mainly due to the enlarged benefitting from the increased valley degeneracy and flattened band, which results in a larger S. Reduced κ_lat by the bond softening in orthorhombic phase is another origin for high zT of Cl-doped BiSbSe₃. On the other hand, higher zT found in Cl-doped Bi_1.2Sb_0.8Se₃ despite of high rhombohedral phase fraction (∼0.74) is attributed to the simultaneous improvement of electronic (enlarged and improved μ_H) and thermal (reduced κ_lat) transport properties by Cl-doping. The highly-reproducible maximum zT reaches in value about 0.68 ± 0.04 at 700 K for three different Bi_1.2Sb_0.8Se_2.76Cl_0.24 samples. Moreover, high σ values of 397 S cm⁻¹ at 300 K and 159 S cm⁻¹ at 700 K make this material a promising candidate for practical applications.

Fig. 5 Temperature dependence of dimensionless figure of merit zT for (a) BiSbSe_3−yCl_y (y = 0.12, 0.18, 0.24) and (b) Bi_1.2Sb_0.8Se_3−zCl_z (z = 0.12, 0.18, 0.24).

We verify the thermal stability of Bi_1.2Sb_0.8Se_2.76Cl_0.24 via the cyclic measurement of zT within temperature range from 300 K to 700 K (Fig. 6a) and remeasurement of power factor after annealing at 800 K for 10 h (Fig. 6b). Fig. 6a and b indicate that the Cl-doped Bi_1.2Sb_0.8Se₃ alloys are chemically stable up to 700 K.

Fig. 6 (a) Cyclic measurement of temperature dependence of dimensionless figure of merit zT for Bi_1.2Sb_0.8Se_2.76Cl_0.24. (b) Variation in temperature dependent power factor of Bi_1.2Sb_0.8Se_2.76Cl_0.24 after annealing at 800 K for 10 h.

4. Conclusions

In summary, temperature dependent thermoelectric transport properties of Te-free Cl-doped BiSbSe₃ and Bi_1.2Sb_0.8Se₃ are systematically investigated. Improved Seebeck coefficient and electrical conductivity are simultaneously obtained due to the enlarged density-of-states effective mass by high content Cl-doping (∼8 at%), while maintaining intrinsic high mobility of Bi_1.2Sb_0.8Se₃-based alloys. This provides the optimized power factor of ∼5.61 μW cm⁻¹ K⁻² at 700 K. Additionally, despite of the weaker phonon scattering owing to decreased bond softening effect in Bi_1.2Sb_0.8Se₃ compared to that in BiSbSe₃, lattice thermal conductivity is effectively reduced in value about 0.39 W m⁻¹ K⁻¹ at 700 K by 8 at% Cl-doping from the intensified mass-defect phonon scattering. This synergetic effect contributes to a high electrical conductivity of ∼159 S cm⁻¹ and the high zT ∼0.68 at 700 K.

Conflicts of interest

There are no conflicts to declare.

Acknowledgements

This work was supported by the National Research Foundation of Korea (NRF) grant funded by the Korean Government (MSIP) (NRF-2017R1A2B3011949) and Global Frontier Program through the Global Frontier Interface Materials (GFHIM) of the National Research Foundation of Korea (NRF) funded by the Ministry of Science, ICT & Future Planning (2013M3A6B1078870).

References

1. G. J. Snyder and E. S. Toberer, Nat. Mater., 2008, 7, 105.
2. J. Jiang, L. Chen, S. Bai, Q. Yao and Q. Wang, Scr. Mater., 2005, 52, 347.
3. K. H. Lee, S. I. Kim, H. Mun, B. Ryu, S. M. Choi, H. J. Park, S. Hwang and S. W. Kim, J. Mater. Chem. C, 2015, 3, 10604.
4. S. J. Jung, B. H. Lee, B. K. Kim, S. S. Lim, S. K. Kim, D. I. Kim, S. O. Won, H. H. Park, J. S. Kim and S. H. Baek, Acta Mater., 2018, 150, 153.
5. S. Wang, Y. Sun, J. Yang, B. Duan, L. Wu, W. Zhang and J. Yang, Energy Environ. Sci., 2016, 9, 3436.
6. S. K. Mishra, S. Satpathy and O. Jepsen, J. Phys.: Condens. Matter, 1997, 9, 461.
7. P. Larson, V. A. Greanya, W. C. Tonjes, R. Liu, S. D. Mahanti and C. G. Olson, Phys. Rev. B: Condens. Matter Mater. Phys., 2002, 65, 85108.
8. W. Liu, K. C. Lukas, K. McEnaney, S. Lee, Q. Zhang, C. P. Opeil, G. Chen and Z. Ren, Energy Environ. Sci., 2013, 6, 552.
9. X. Liu, D. Wang, H. Wu, J. Wang, Y. Zhang, G. Wang, S. J. Pennycook and L. D. Zhao, Adv. Funct. Mater., 2019, 29, 106558.
10. J. S. Rhyee, K. Ahn, K. H. Lee, H. S. Ji and J. H. Shim, Adv. Mater., 2011, 23, 2191.
11. Q. Zhang, H. Wang, W. Liu, H. Wang, B. Yu, Q. Zhang, Z. Tian, G. Ni, S. Lee, K. Esfarjani, G. Chen and Z. Ren, Energy Environ. Sci., 2012, 5, 5246.
12. Q. Zhang, Y. Lan, S. Yang, F. Cao, M. Yao, C. Opeil, D. Broido, G. Chen and Z. Ren, Nano Energy, 2013, 2, 1121.
13. S. I. Kim, S. Hwang, S. Y. Kim, W. J. Lee, D. W. Jung, K. S. Moon, H. J. Park, Y. J. Cho, Y. H. Cho, J. H. Kim, D. J. Yun, K. H. Lee, I. Han, K. Lee and Y. Sohn, Sci. Rep., 2016, 6, 19733.
14. V. G. Kuznetsov, K. K. Palkina and A. A. Reshchikova, Izv. Akad. Nauk SSSR, Neorg. Mater., 1968, 4, 670.

---

Footnotes

† Electronic supplementary information (ESI) available. See DOI: 10.1039/d0ra04065g

‡ These authors contributed equally to this study.

---