Transport properties and enhanced thermoelectric performance of aluminum doped Cu₃SbSe₄

Abstract

The electrical transport and thermoelectric properties of Cu₃Sb_1−xAl_xSe₄ (x = 0, 0.01, 0.02 and 0.03) compounds are investigated in the temperature range of 300–600 K. The results indicate that with increasing Al content from x = 0 to x = 0.03, hole concentration increases monotonically from 8.04 × 10¹⁷ to 1.19 × 10¹⁹ cm⁻³ due to the substitution of Al³⁺ for Sb⁵⁺, thus leading to a large decrease in the electrical resistivity of Cu₃Sb_1−xAl_xSe₄. Meanwhile, the increase in hole concentration leads to a transition from a non-degenerate (x = 0) to a partial degenerate (x = 0.01, 0.02) and then to a degenerate state (x = 0.03). The power factor (PF) of all the Al-doped Cu₃Sb_1−xAl_xSe₄ samples is remarkably improved due to the optimization of hole concentration. Lattice thermal conductivity κ_L of the heavily doped sample (x = 0.03) is reduced. As a result, a large thermoelectric figure of merit ZT = 0.58 is obtained for Cu₃Sb_0.97Al_0.03Se₄ at 600 K, which is around 1.9 times as large as that of the un-doped Cu₃SbSe₄.

---

1. Introduction

Thermoelectric (TE) materials have attracted more and more attention because of their applications as power generators and heat pumps.¹ The efficiency of a TE material is generally characterized by the dimensionless figure of merit ZT, defined as ZT = (S²T)/ρ(κ_C + κ_L), where ρ, S, κ_C, κ_L and T are the electrical resistivity, Seebeck coefficient, thermal conductivity from carrier contribution, lattice thermal conductivity and absolute temperature, respectively.² Copper-based multinary semiconductors have recently received much attention due to their good electronic transport properties and relatively low intrinsic thermal conductivity, which contributes to being good TE materials (e.g. Cu_2.10Cd_0.90SnSe₄, ZT = 0.65 at 700 K, Cu₂Sn_0.90In_0.10Se₃, ZT = 1.14 at 850 K, Cu₃Sb_0.97Ge_0.03Se_2.8S_1.2, ZT = 0.89 at 650 K).^3–10

The ternary semiconductor of Cu₃SbSe₄, as a narrow band gap semiconductor (0.13–0.42 eV)¹¹ with a unit cell four times larger than ZnSe (n = 8 for Cu₃SbSe₄ versus n = 2 for ZnSe), was firstly synthesized by Wernick and Benson¹² with high Seebeck coefficient and low thermal conductivity at room temperature (the Cu/Se atoms form the structural framework and the rest Sb atoms have the rattling behaviors similar to the resonators¹³). However, the un-doped Cu₃SbSe₄ compound has low hole concentration and relatively large electrical resistivity, which leads to low TE performance. As reported in early work,¹⁴ the ZT of Cu₃SbSe₄ is too small to be used in practice. Hence, it is a key issue to reduce its electrical resistivity and to optimize its PF (PF = S²/ρ). Earlier work done by Skoug et al. showed that doping with either Ge or Sn at the Sb site can improve the thermoelectric performance of Cu₃SbSe₄.¹⁵ Wei et al.¹⁶ present a quite systematic study of Cu₃SbSe₄, which is done in advance of current work, and have achieved promising ZT = ∼0.7 at 673 K for Cu₃Sb_0.98Sn_0.02Se₄.

In our previous work,¹⁷ we showed that the thermoelectric performance of Cu₃SbSe₄ could be improved by doping the element Bi, where the equivalent substitution of Sb⁵⁺ with Bi⁵⁺ led to a limited increase in the hole concentration. In the present work, Al, a cheap, environment-friendly and abundant element, is used as the dopant and the inequitable substitution of Al³⁺ for Sb⁵⁺ will introduce hole into the host, giving rise to a large increase in hole concentration in this p-type compound Cu₃SbSe₄. Our experiments demonstrate that Al-doping can effectively adjust the hole concentration and optimize PF. Moreover, κ_L of the heavily doped sample (x = 0.03) is reduced. As a result, ZT = 0.58 is achieved for Cu₃Sb_0.97Al_0.03Se₄ at 600 K, which is about 1.9 times larger than that of the Cu₃SbSe₄.

2. Experimental

Samples of Cu₃Sb_1−xAl_xSe₄ (x = 0, 0.01, 0.02 and 0.03) were prepared by direct melting of the pure elements (Cu, 99.99%; Sb, 99.99%; Al, 99.99%; Se, 99.99%) in evacuated quartz ampoules that were pumped under vacuum of 10⁻² Pa. Before that, the quartz ampoules were under ultrasonic treatment for two hours to insure the cleanliness of the surface. The samples were placed in a digitally controlled furnace and slowly heated (20 °C h⁻¹) for 2700 min up to 900 °C, held for 12 hours, then cooled to 500 °C (1 °C min⁻¹) and quenched in water at room temperature. After quenching, the samples were annealed at 300 °C for 48 hours to promote homogeneity. The resulted ingots were pulverized into powders (The size distribution is around 1–2 μm) in an agate mortar. The bulk samples were obtained by spark plasma sintering (SPS) for 5 min at the temperature of 673 K under the pressure of 50 MPa. In each charge for the SPS machine, about 4.5 g powders were employed.

The phase of the obtained samples was checked by using X-ray diffraction (Philips diffractometer, Cu Kα radiation). The specimens for transport measurements were cut from the bulk samples by using a diamond saw. Electrical resistivity and Seebeck coefficient were measured by the ZEM-3 system from ULVAC. The thermal diffusivity, D, was measured using the laser flash method (Netzsch, LFA-457). The specific heat, C_p, was determined by differential scanning calorimetry (DSC Pyris Diamond). The resulting thermal conductivity was calculated from the measured thermal diffusivity D, specific heat C_p, and density d from the relationship κ = DdC_p. The density of the bulk materials was measured by Archimedes' methods using alcohol as the medium, and all were above 90% of the theory density, as shown in Table 1. Thin plate samples were used for Hall measurements, which were conducted by CVM-200 Hall effect measurement system.

Table 1 List of lattice parameters (a and c), Hall mobility μ, carrier concentration p and relative density d_r at room temperature for Cu₃Sb_1−xAl_xSe₄ (x = 0, 0.01, 0.02 and 0.03)

Cu3Sb1−xAlxSe4	a^a (Å)	c^b (Å)	μ^c (cm^2 V^−1 s^−1)	p^d (cm^−3)	dr^e (%)
a a is lattice parameter.b c is lattice parameter.c μ is Hall mobility.d p is carrier concentration.e dr is relative density, defined as dr = d/d0, where d is measured density of samples and d0 (= 5.86 g cm^−3) is theoretical density of Cu3SbSe4.
x = 0	5.654	11.196	94.1	8.04 × 10^17	91
x = 0.01	5.635	11.175	83.0	2.50 × 10^18	91
x = 0.02	5.618	11.156	61.0	4.12 × 10^18	92
x = 0.03	5.601	11.137	52.5	1.19 × 10^19	91

---

3. Result and discussions

The XRD patterns of all the samples Cu₃Sb_1−xAl_xSe₄ (x = 0, 0.01, 0.02 and 0.03) at room temperature are shown in Fig. 1. The main diffraction peaks correspond well to the standard JCPDS card (no. 01-085-0003) of Cu₃SbSe₄with tetragonal structure (space group I4̄2m, a = 5.661 Å, c = 11.28 Å),¹³ indicating that the doped specimens have the same crystallographic structure as that of the Cu₃SbSe₄ phase and no obvious impurity phase is observed. A Reitveld refinement approach is employed to calculate the lattice parameters from the XRD data. The result reveals that lattice parameters a and c decrease monotonically with increasing Al content, as shown in Table 1, which is due to the substitution of Al³⁺ with smaller ionic radius (0.39 Å) for Sb⁵⁺ possessing larger ionic radius (0.62 Å).

Fig. 1 XRD patterns at room temperature for Cu₃Sb_1−xAl_xSe₄ samples: (a) x = 0, (b) x = 0.01, (c) x = 0.02 and (d) x = 0.03.

Fig. 2 shows the SEM image of fracture surface of bulk sample Cu₃Sb_0.97Al_0.03Se₄. One can see from the image that the typical grain sizes are in the range of 1–5 μm. In addition, there are a large number of nanoparticles embedded at grain boundaries, indicating that the dispersed distribution of the nanoparticles will be beneficial to inhibiting lattice thermal conductivity of the bulk specimens due to extra enhanced phonon scattering. Our observation indicates that the nanoparticles exist in both doped samples and un-doped samples (not shown here).

Fig. 2 SEM image of fracture surface of bulk sample Cu₃Sb_0.97Al_0.03Se₄.

The thermoelectric properties of Cu₃Sb_1−xAl_xSe₄ (x = 0, 0.01, 0.02 and 0.03) are shown in Fig. 3. As displayed in Fig. 3(a), the electrical resistivity ρ of Cu₃SbSe₄ decreases with increasing temperature, indicating that the un-doped sample exhibits semiconductor-like behavior. However, in the case of the Al-doped compounds, ρ shows different temperature dependences. ρ of sample with x = 0.01 initially increases slightly with increasing temperature and reaches a peak value of 3.12 × 10⁻⁴ Ω m at 375 K and then decreases with further increasing temperature. With increasing Al content to x = 0.02, ρ almost remains the same value in the temperature range from 325 K to 375 K and then decreases with further increasing temperature, reaching 1.3 × 10⁻⁴ Ω m (at 600 K) which is comparable to that of the sample with x = 0.01. However, ρ in the case with x = 0.03 almost remains unchanged in the whole temperature range. In addition, Al doping causes the decrease in the magnitude of the electrical resistivity.

Fig. 3 Temperature dependences of (a) electrical resistivity, (b) Seebeck coefficient, (c) power factor, and (d) total thermal conductivity for Cu₃Sb_1−xAl_xSe₄: (a) x = 0, (b) x = 0.01, (c) x = 0.02 and (d) x = 0.03. The inset of (a) is the logarithmic plot of ρ versus 1/T for Cu₃SbSe₄.

In order to examine the temperature behavior of the resistivity for Cu₃SbSe₄, logarithm of the resistivity ρ as a function of reciprocal of temperature is given in the inset of Fig. 3(a). It can be seen that good linear relationships between ln ρ and 1/T exist in the high temperature range for Cu₃SbSe₄. The existence of a linear relationship between ln ρ and 1/T means that the resistivity can be described by using a thermally activated expression in corresponding temperature regimes, written as:

where C is an consistent and k_B Boltzmann constant, E_g band gap. By best fitting of the experimental data to formula (1), one can obtain E_g = 0.3 eV for Cu₃SbSe₄, which is almost equal to the energy gap E_g = 0.29 eV for Cu₃SbSe₄ reported by Wei et al.,¹⁶ indicating that the un-doped sample is an intrinsic semiconductor. The activation energy E_a = E_g/2 = 0.15 eV, which is similar to the result (0.13 eV) of Wei et al.¹⁶ determined from resistivity.

Our further measurements (see following text) demonstrate that Seebeck coefficient and Hall coefficient of Cu₃SbSe₄ are positive in the whole temperature investigated, indicating the major carriers are holes. Since the substitution of Al³⁺ for Sb⁵⁺ occurs, the doping of Al is expected to introduce holes into the host. Therefore, the Al substitution for Sb will give rise to a large increase in hole concentration. The hole concentrations p at room temperature for Cu₃Sb_1−xAl_xSe₄ (x = 0, 0.01, 0.02 and 0.03) are calculated according to the measured Hall coefficient and the results are listed in Table 1. One can see that the hole concentration increases from 8.04 × 10¹⁷ cm⁻³ for x = 0 to 2.50 × 10¹⁸ cm⁻³ for x = 0.01, 4.12 × 10¹⁸ cm⁻³ for x = 0.02 and 1.19 × 10¹⁹ cm⁻³ for x = 0.03, respectively. Al-doping leads to the increase in hole concentration, which explains the smaller resistivity upon doping.

All the samples are p-type semiconductors, as verified by the positive Seebeck coefficients, which are shown in Fig. 3(b). The Seebeck coefficient S for Cu₃SbSe₄ decreases with the increasing temperature, and after about 550 K, it shows a slight increase with increasing temperature. In contrast, S for samples with x = 0.01 and 0.02 decreases much slowly with increasing temperature and its value drops with increasing content of aluminum. This behavior is similar to that of the electrical resistivity, which also indicates that the Al-doped compounds become partially degenerate. Specially, one notices that S for sample with x = 0.03 increases monotonically with temperature. This is the typically characteristic of a heavily degenerate semiconductor.

The temperature dependences of power factor PF (= S²/ρ) of Cu₃Sb_1−xAl_xSe₄ (x = 0, 0.01, 0.02 and 0.03) are shown in Fig. 3(c). The values of PF for all the Al-doped compounds are larger than that for the un-doped one. Specially, the PF of sample with x = 0.03 nearly increases linearly with temperature, and it reaches 1.05 × 10⁻³ W m⁻¹ K⁻² at 600 K, which is about twice as large as that of Cu₃SbSe₄.

The temperature dependences of the total thermal conductivity κ, lattice thermal conductivity κ_L and carrier contribution κ_c for Cu₃Sb_1−xAl_xSe₄ (x = 0, 0.01, 0.02 and 0.03) compounds are presented in Fig. 3(d) and 4, respectively. The results of the specific heat, C_p, determined by DSC for Cu₃SbSe₄ from 300 K to 600 K are shown in Fig. 5. The values of C_p range from 0.31 to 0.37 J g⁻¹ K⁻¹, which are in agreement with the result of Skoug et al.⁹ but smaller than the data reported by Wei et al.¹⁶ As to Fig. 3(d), for all the samples, κ decreases with increasing temperature in the whole temperature range investigated. In comparison, κ of the Al-doped samples is slightly lower at room temperature and then becomes higher than that of the un-doped Cu₃SbSe₄ with the increasing temperature due to the increased contribution from κ_c (the inset of Fig. 4). Lattice thermal conductivity κ_L is estimated by subtracting the carrier thermal conductivity κ_c from κ, where the Wiedemann–Franz relation with a Lorentz constant L₀ is applied for estimating κ_c (κ_c = L₀T/ρ). Here, L₀ is set to 1.5 × 10⁻⁸ V² K⁻², 2 × 10⁻⁸ V² K⁻² and 2.44 × 10⁻⁸ V² K⁻² for a non-degenerate semiconductor (x = 0), a partial degenerate semiconductor (x = 0.01, 0.02) and a degenerate semiconductor (x = 0.03), respectively.^18,19 As compared to the results of Wei et al.,¹⁶ our lattice thermal conductivity κ_L (at 300 K) for un-doped one is around ∼30% lower than that reported by Wei et al.,¹⁶ which could, on the one hand, come partly from more porosities in our samples than theirs (the relative densities d_r for each sample are given in Table 1). On the other hand, the large number of dispersed nanoparticles (as shown in Fig. 2) will make great contribution to inhibiting lattice thermal conductivity of the bulk specimens due to extra phonon scattering at these nanoparticles, leading to smaller κ_L. As shown in Fig. 4, κ_L of samples with x = 0.01 and 0.02 is similar to that of the un-doped sample. However, κ_L for x = 0.03 is smaller than that of the pure sample especially at high temperatures.

Fig. 4 Temperature dependences of the lattice thermal conductivity κ_L and the carrier thermal conductivity κ_c (the inset) for Cu₃Sb_1−xAl_xSe₄: (a) x = 0, (b) x = 0.01, (c) x = 0.02 and (d) x = 0.03.

Fig. 5 Temperature dependences of the specific heat C_p determined by DSC for Cu₃SbSe₄.

Fig. 6 shows the temperature dependence of ZT for Cu₃Sb_1−xAl_xSe₄ (x = 0, 0.01, 0.02 and 0.03). ZT of the Al-doped compounds Cu₃Sb_1−xAl_xSe₄ (x = 0.01, 0.02 and 0.03) is larger than that of un-doped Cu₃SbSe₄. The maximum ZT reaches 0.58 for Cu₃Sb_0.97Al_0.03Se₄ at 600 K, which is around 1.9 times as large as that of the un-doped Cu₃SbSe₄ and is larger than the ZT value reported by Wei et al.¹⁶ for Sn-doped Cu₃SbSe₄ (ZT = ∼0.5 at 600 K). The elevation of ZT for Cu₃Sb_0.97Al_0.03Se₄ results mainly from both its enhanced PF due to the optimization of hole concentration and reduced κ_L.

Fig. 6 Temperature dependence of ZT for Cu₃Sb_1−xAl_xSe₄: (a) x = 0, (b) x = 0.01, (c) x = 0.02 and (d) x = 0.03.

4. Conclusion

The effects of Al-doping on the thermoelectric properties of Cu₃Sb_1−xAl_xSe₄ (x = 0, 0.01, 0.02 and 0.03) compounds obtained by combining melting method with SPS are investigated in this work. Experiments show that Al-doping leads to large decreases in the electrical resistivity due to great increase in hole concentration. With increasing Al-doping content the samples transform from non-degenerate (x = 0) to partial degenerate (x = 0.01, 0.02) and then to degenerate state (x = 0.03). PF of all the Al-doped samples is remarkably improved. Moreover, heavy Al-doping (x = 0.03) leads to reduced lattice thermal conductivity. The largest ZT_max = 0.58 is achieved at 600 K for Cu₃Sb_0.97Al_0.03Se₄, which is about 1.9 times larger than that of the un-doped compounds. Our study demonstrates that Al-doping is an effective and environment-friendly way to improve the thermoelectric performance of Cu₃SbSe₄.

Acknowledgements

Financial support from National Natural Science Foundation of China (nos 11174292, 11374306, 51101150, 50972146 and 10904144) is gratefully acknowledged.

References

1. B. Poudel, Q. Hao and Y. Ma, Science, 2008, 320, 634–638.
2. J. P. Heremans, V. Jovovic and E. S. Toberer, Science, 2008, 321, 554–557.
3. M. L. Liu, I. W. Chen, F. Q. Huang and L. D. Chen, Adv. Mater., 2009, 21, 3808–3812.
4. X. Shi, L. Xi and J. Fan, Chem. Mater., 2010, 22, 6029–6031.
5. X. Y. Shi, F. Q. Huang, M. L. Liu and L. D. Chen, Appl. Phys. Lett., 2009, 94, 122103.
6. E. J. Skoug, J. D. Cain and D. T. Morelli, J. Electron. Mater., 2012, 41, 1232–1236.
7. D. Li, R. Li and X. Y. Qin, CrystEngComm, 2013, 15, 7166.
8. A. Suzumura, M. Watanabe, N. Nagasako and R. Asahi, J. Electron. Mater., 2014, 43, 2356–2361.
9. E. J. Skoug, J. D. Cain and D. T. Morelli, Appl. Phys. Lett., 2011, 98, 261911.
10. E. J. Skoug, J. D. Cain and D. T. Morelli, J. Appl. Phys., 2011, 110, 023501.
11. T. R. Wei, F. Li and J. F. Li, J. Electron. Mater., 2014, 43, 2229–2238.
12. C. An, Q. Liu and K. Tang, J. Cryst. Growth, 2003, 256, 128–133.
13. C. Yang, F. Huang, L. Wu and K. Xu, J. Phys. D: Appl. Phys., 2011, 44, 295404.
14. H. Katsumi and K. Masaru, Jpn. J. Appl. Phys., 1968, 7, 54–59.
15. E. J. Skoug, J. D. Cain and P. Majsztrik, Sci. Adv. Mater., 2011, 3, 602–606.
16. T. R. Wei, H. Wang, Z. M. Gibbs, C. F. Wu, G. Jeffrey Snyder and J. F. Li, J. Mater. Chem. A, 2014, 2, 13527.
17. X. Y. Li, D. Li and H. X. Xin, J. Alloys Compd., 2013, 561, 105–108.
18. B. C. Sales, D. Mandrus and B. C. Chakoumakos, Phys. Rev. B: Condens. Matter Mater. Phys., 1997, 56, 23.
19. J. Zhang, H. J. Liu and L. Cheng, J. Appl. Phys., 2014, 116, 023706.

---