Suppressing adverse intrinsic conduction of Bi₂Te₃ thermoelectric bulks by Sb and Cu co-substitutions via HPHT synthesis

Abstract

Chemical substitution combined with high-pressure tuning (high pressure and high temperature, HPHT) was applied to synthesize Bi₂Te₃ polycrystalline bulks. The synthesis time was sharply shortened to about half an hour, which is a distinct advantage over conventional synthesis methods for mass production. Acceptor-like antisite and substitutional defects owing to Sb and Cu substituting Bi sites were responsible for the increase in carrier concentration. Pressure-induced multiple textures and microstructures contributed to phonon scattering, including grain boundaries, lattice distortions, and dislocations. A maximum ZT value of 1.20 was achieved for Cu_0.005Bi_0.5Sb_1.495Te₃ at 473 K. Moreover, an available ZT value of 1.17 was obtained in a wide temperature range of 423–523 K. The increases in carrier concentration and band gap effectively suppressed the adverse intrinsic conduction and delayed the onset of the bipolar effect. Cu-substituted samples exhibit new vibrational modes in Raman spectroscopy, which implies that the substitutions induced significant changes in lattice vibrations.

---

1. Introduction

It is noteworthy that thermoelectric conversion is gradually occupying a significant position in the recovery of exhaust heat and solid-state cooling, on account of several encouraging features: stability, reliability and environment-friendliness. Nonetheless, a high conversion efficiency (ZT = S²σT/κ) is not easily obtained, as it requires a high power factor (S²σ) and low thermal conductivity (κ) simultaneously.¹ Owing to the interdependence and adverse interactions of the three parameters (S, σ, and κ), many strategies have yielded small improvements in the past decades. Currently, nanostructuring and band engineering provide a powerful impetus for an increase in ZT via phonon scattering effects,² quantum confinement effects,³ convergence of conduction bands,⁴ and distortion of the electronic density of states.⁵ By employing a superlattice in Bi₂Te₃/Sb₂Te₃, Venkatasubramanian et al. achieved a ZT value of 2.4 at 300 K,⁶ which represents remarkable progress in contrast to traditional unidirectional solidification. As reported by Poudel et al., a ZT value of up to 1.4 at 100 °C was obtained for nanostructured bismuth antimony telluride bulk by ball milling and subsequent hot pressing.⁷ Recently, full-spectrum phonon scattering was utilized to reduce the lattice conductivity of Bi_0.5Sb_1.5Te₃ dramatically so as to obtain an outstanding ZT value of 1.86 at 320 K.⁸ However, it appears, unfortunately, that these approaches seem inapplicable for mass production and scalable applications owing to their intensive energy consumption, limited production output, and complexity of the fabrication process. Therefore, it is of considerable urgency to find manufacturing methods that are practicable for commercial production.

Because of their detrimental intrinsic conduction, commercially available Bi₂Te₃-based materials are extensively used near room temperature. With the purpose of shifting their peak thermoelectric efficiency to elevated temperature, it is necessary to suppress their detrimental intrinsic conduction. Moreover, owing to their quintuple-layered structure and van der Waals bonds, Bi₂Te₃ single crystals exhibit poor mechanical properties and anisotropic features. Therefore, polycrystalline solids are extensively investigated via chemical tuning and progressive manufacturing techniques.

Owing to antisite defects, the substitution of Bi by Sb has the potential to tune the carrier concentration.⁹ In addition, substitution by Cu is proposed to optimize the chemical composition.^10,11 The difference in electronegativity between Cu (1.9) and Bi (2.02) is smaller than that between Cu and Te (2.1), which makes Cu atoms occupy Bi sites preferentially. As the diffusion of intercalated Cu atoms along the basal plane gives rise to aging problems,¹² we intend to incorporate Cu atoms into Bi sites. Besides, the interaction of Cu atoms with common defects in Bi₂Te₃ alloys can suppress uncontrollable Te vacancies, which improves reproducibility effectively.¹⁰ Although some research on diffusion in Cu electrodes has been conducted for thermoelectric devices, we mainly focus on the thermoelectric performance of materials in this study.

In addition to chemical substitution, pressure tuning is deemed to be an alternative strategy, but in situ measurements on as-synthesized samples show severe reversibility.^13,14 However, the high pressure and high temperature technique (HPHT) is effective not only for synthesizing pure-phase bulks directly from elements in a short time, but also for tuning their thermoelectric properties simultaneously.^15,16 Besides, the synthesis time is sharply reduced to about half an hour, which is advantageous for scalable production.

In this paper, Cu_xBi_0.5Sb_1.5−xTe₃ bulks were successfully prepared by the HPHT method. The effects of substitutions by Sb and Cu on the thermoelectric transport properties are assessed. How the bipolar effect is suppressed by co-substitution is discussed in detail. The increase in ZT is mainly ascribed to a decrease in electrical resistivity.

2. Experimental procedure

2.1 Sample preparation

p-Type Cu_xBi_0.5Sb_1.5−xTe₃ samples were synthesized from highly pure powder (5 N, 200 mesh) Cu, Bi, Sb and Te. Stoichiometric amounts of the elements were ground in an Ar-filled glove box to avoid oxidation, and then the well-mixed powders were densified into a cylinder (φ 10.5 × 4 mm). The as-pressed bulk was assembled in high-temperature and high-pressure conditions by a China-type large-volume cubic high-pressure apparatus (CHPA, SPD-6x1200). The assembled chamber was heated to 890 K for 25 min at 2 GPa. The temperature was measured by a Pt-RH/Pt-Rh6-type thermocouple junction, which was fixed at the surface of the sample chamber. The pressure was measured via the change in the resistance of standard materials. To eliminate the impact of anisotropy, the as-synthesized bulks were ground thoroughly into powders and then sintered into a cylinder (φ 10.5 × 12 mm) at 623 K for 5 min at 2 GPa. After sintering, a bar of 2 × 2 × 8 mm and a disk of φ 10 × 2 mm were cut along the 12 mm height direction to make sure that the measurements were in the same direction.

2.2 Characterization

The phase structure was characterized by X-ray diffraction (D/Max-RA) in the 2θ range of 5° to 70°. The morphology of the fractured surface was observed by field emission scanning electron microscopy (JEOL JSM-6700F). The investigation of microstructures was performed by high-resolution transmission electron microscopy (JEOL JEM-2200FS). The Seebeck coefficient (S) and electrical resistivity (ρ) were measured simultaneously from 323 to 523 K by ZEM-3 commercial equipment (Ulvac-Riko). The thermal diffusivity (λ) was directly measured by the laser flash method with a commercial system (Netzsch LFA-427) and the specific heat capacity (C_p) was indirectly determined by DSC conversion using a standard sample (sapphire) (STA PT-1750). The volume density (d) was measured by the Archimedes method. The total thermal conductivity (κ) was then calculated as the product of the thermal diffusivity, specific heat capacity and volume density (κ = λC_pd). Raman spectra of the bulks were obtained in the range of 50 to 550 cm⁻¹ at room temperature with a He–Ne laser (671 nm) to demonstrate the effect of substitutions on the lattice vibrations.

3. Results and discussion

3.1 XRD

X-ray diffraction (XRD) patterns of bulks with different Cu contents are presented in Fig. 1. Compared with the rhombohedral structure of Bi_0.5Sb_1.5Te₃ (JCPDS 49-1713), all diffraction patterns demonstrate a well-crystallized single phase without any detectable impurity phase. No preferential grain orientation of (0 0 l) planes is revealed compared with standard diffraction data. This means that all samples have no anisotropic features in crystalline granules after the grinding and sintering procedure. The lattice parameters decrease with an increase in Cu content from x = 0 to x = 0.05. The reduction can be ascribed to the smaller atomic size of Cu compared with Bi. Evidently, the Cu_xBi_0.5Sb_1.5−xTe₃ bulks were successfully synthesized by HPHT in only about half an hour.

Fig. 1 (a) XRD patterns of Bi₂Te₃-based bulks with various Sb and Cu contents. (b) Lattice parameters of samples with different Cu contents.

3.2 SEM and HRTEM

FESEM images show a representative morphology in Fig. 2. The samples have abundant boundaries and a wide crystal size distribution ranging from small particles to tens of μm. The lamellar granules are distributed randomly with no distinct preferred orientation, which is consistent with the XRD patterns. The samples have no pores or cracks and exhibit a high relative density of 97.9% (theoretical density = 6.878 g cm⁻³). The typical layered feature (indicated by the arrow) originated from a two-dimensional nucleation mechanism and slippage of the basal plane due to shear forces under high pressure. In brief, the well-developed granules confirm that the simple and rapid HPHT method is favorable for synthesis in just 30 min.

Fig. 2 FESEM image of fractured surfaces of a Cu_0.005Bi_0.5Sb_1.495Te₃ sample.

HRTEM analysis was performed to characterize the microstructures. Grains with clear boundaries are observed and some twin boundaries are marked in Fig. 3a. The measured interplanar distances are about 0.23, 0.31 and 0.33 nm, which correspond to (1 0 10), (0 1 5) and (0 0 9) planes, respectively. A fast Fourier transform (FFT) image corresponding to a twin-crystal region is shown (Fig. 3b). The measured quintuple neighboring distance of 1.03 nm is indexed to (0 0 3) planes, corresponding to a crystal structure of quintuple layers connected by van der Waals forces (Fig. 3c). Both the electron diffraction pattern in the FFT image and the above interplanar distances confirm the rhombohedral structure of Bi_0.5Sb_1.5Te₃ (PDF #49-1713). Besides, the HRTEM image (Fig. 3d) shows crystal distortions including dislocations and lattice curvatures, which are caused by compression and shear forces at high pressure. The inverse fast Fourier transform (IFFT) image (Fig. 3e) clearly identifies lattice distortions. Based on relaxation time theory, the overall combined relaxation time is given as:

τ⁻¹_c = τ⁻¹_U + τ⁻¹_B + τ⁻¹_N + τ⁻¹_e-ph + τ⁻¹_D + τ⁻¹_l (1)

where τ_U, τ_B, τ_N, τ_e-ph, τ_D, and τ_l are relaxation times corresponding to scattering from the Umklapp process, boundary process, normal phonon–phonon process, electron–phonon process, dislocations and lattice curvatures, respectively. The microstructures-induced strain fields will serve as an effective phonon scattering center and decrease the relaxation time and phonon mean free path.¹⁷ In brief, the hierarchical microstructures on all scales contribute to a broad phonon scattering spectrum.

Fig. 3 Typical HRTEM images of Cu_0.005Bi_0.5Sb_1.495Te₃ samples. (a) Clear twin boundaries. (b) FFT image corresponding to the twin crystals in (a). (c) Representative quintuple layered structure of Bi₂Te₃. (d) Lattice defects: dislocations and lattice curvatures. (e) IFFT image corresponding to (d).

3.3 Thermoelectric properties

The dependences on temperature of electrical transport properties vary notably with the composition (Fig. 4). The positive Seebeck coefficients reveal typical p-type transport behavior with holes as the dominant carriers (Fig. 4a). The Seebeck coefficients decrease with an increase in the Cu content near room temperature. Because the relationship between the Seebeck coefficient and the carrier concentration can be described as follows:where h is the Planck constant and k_B is the Boltzmann constant, the reduction in the Seebeck coefficient is ascribed to the increase in carrier concentration due to substitution by Cu.

Fig. 4 Dependence on temperature of electrical transport properties of Bi₂Te₃ samples: (a) Seebeck coefficient, (b) electrical resistivity, (c) power factor.

For the Cu-free sample, the Seebeck coefficient first slightly increases with an increase in temperature, reaching a maximum value of 195 μV K⁻¹ at 373 K, and then decreases dramatically.

The occurrence of the peak at 373 K is ascribed to the onset of the bipolar effect. Owing to the low carrier concentration of the Cu-free sample, intrinsic thermal excitation above 373 K makes more minority carriers cross the band gap. Both majority carriers and minority carriers will contribute to thermoelectric transport. According to the two-band theory, the Seebeck coefficient can be expressed as the following formula:

where the subscript symbols p and n represent p-type and n-type carriers, respectively. Because of the opposite sign of S_n and S_p, the Seebeck coefficient must be smaller than with only p-type carriers. The Seebeck coefficient decreases with an increase in minority carriers according to the relation of eqn (3).

In contrast to what is observed for the Cu-free sample, the Seebeck coefficients of Cu-doped samples increase almost linearly initially. The Cu_0.005Bi_0.5Sb_1.495Te₃ and Cu_0.01Bi_0.5Sb_1.49Te₃ samples exhibit peaks near 473 K and 523 K, whereas the Cu_0.05Bi_0.5Sb_1.45Te₃ sample displays no maximum in the investigated temperature range. The higher the concentration of extrinsic majority carriers is, the higher is the temperature required to obtain sufficient intrinsic minority carriers in order to induce the bipolar effect.¹⁸ Therefore, the temperature value at S_max shifts to higher temperatures with an increase in the Cu content. Alloying with Sb can modify the band gap of Bi₂Te₃, as the band gap of Sb₂Te₃ (E_g = 0.28 eV) is larger than that of Bi₂Te₃ (E_g = 0.13 eV). The band gap can be estimated from the value of S_max at T_max:

E_g = 2eS_maxT_max (4)

The obtained results for the band gaps are 0.145, 0.171 and 0.196 eV with an increase in the Cu content from x = 0 to 0.005 and 0.01, respectively. This means that the band gap is widened by substitution with Sb and Cu, and the details of the band structure are modified by chemical substitution to some extent.⁹

The positive dependence on temperature of the electrical resistivity indicates typical degenerate behavior (Fig. 4b). A distinct increase in electrical conductivity is observed for Cu-doped samples. The abnormal value of electrical resistivity for Cu_0.01Bi_0.5Sb_1.49Te₃ requires detailed data for further investigation. For the Cu-free sample, the bipolar effect is responsible for the decrease in electrical resistivity at high temperatures (above 473 K), which is due to intrinsic excitation. It is noteworthy that the temperature at the peak in electrical resistivity is higher than that for the Seebeck coefficient, as the Seebeck coefficient is more sensitive to the minority carrier concentration. However, the electrical resistivity of Cu-doped samples increases with temperature up to 523 K. Besides, as more Cu content is doped, the electrical resistivity displays a more moderate increasing tendency with an increase in temperature.

To put it simply, the mechanism of the increase in carrier concentration is described as follows. Except for various intrinsic defects in Bi₂Te₃, many acceptor-like antisite defects of Sb_Te are introduced by alloying with Sb.

As the Bi sites are substituted by Cu, each Cu_Bi defect contributes one more hole.

With an increase in the Cu content, a higher concentration of holes will suppress the generation of minority carriers and thus delay the onset of the bipolar effect.¹⁸

An assessment of the effect of substitution by Sb and Cu on electrical transport is elucidated in Fig. 4c. The optimized Cu content is x = 0.005, which yields the highest power factor of 29.2 × 10⁻⁴ W m⁻¹ K⁻² at 323 K. It is almost twice as high compared with the Cu-free sample.

Fig. 5 illustrates the dependence on temperature of thermal conductivity and ZT for Cu_xBi_0.5Sb_1.5−xTe₃ samples. The thermal conductivity increases with the Cu content (Fig. 5a). The values for Cu_xBi_0.5Sb_1.5−xTe₃ (x = 0, 0.005, 0.01, 0.05) at 323 K are 1.00, 1.06, 1.31, and 1.80 W m⁻¹ K⁻¹, respectively. The thermal conductivity of the Cu-free bulk first decreases with an increase in temperature, reaches a minimum at 473 K, and then gradually rises. This behavior is obviously affected by the onset of bipolar diffusion, which is sufficient to compensate for the reduction in lattice thermal conductivity owing to the Umklapp process. In contrast, all the Cu-doped samples exhibit monotonic decreases in thermal conductivity, which implies the suppression of bipolar conduction.

Fig. 5 Variations in (a) thermal conductivity and (b) ZT value as a function of temperature.

Electronic thermal conductivity is calculated by the Wiedemann–Franz law (κ_e = LσT). The Lorenz number L depends on the reduced Fermi energy, temperature and scattering parameter. However, the common Lorenz number for degenerate semiconductors (2.45 × 10⁻⁸ V² K⁻²) will overestimate electronic thermal conductivity and yield a physically impossible value of lattice thermal conductivity for HPHT samples. Similarly, samples with a unique electronic density of states or resonant levels have a reduced electronic thermal conductivity, where the Wiedemann–Franz law is also inapplicable.^19,20 Therefore, pressure-induced variations in band structure make the Wiedemann–Franz law unsuitable for samples prepared under high pressure.

The peak value of ZT appears at a relatively elevated temperature (473 K) compared with other values ranging from 300 to 380 K (Fig. 5b).^{7,9,10,19,21,22} The temperatures at maximum values of ZT for samples prepared by different methods are summarized in Table 1 for comparison. A maximum ZT value of 1.20 was achieved for Cu_0.005Bi_0.5Sb_1.495Te₃ at 473 K, which mainly benefits from the distinct decrease in electrical resistivity. In addition, an average ZT value of 1.17 was achieved in a wide temperature range of 423–523 K. As the conversion efficiency of a TE device is described as follows:

a higher temperature at the peak value of ZT and a wide temperature range with high values of ZT can increase the thermoelectric conversion efficiency, which is favourable for practical applications in power generation.

Table 1 Temperature at maximum value of ZT for samples prepared by different methods

Sample	ZTmax	Tmax	Reference
BiSbTe	1.4	373 K	7
Bi0.3Sb1.7Te3	1.3	380 K	9
Cu0.01Bi2Te2.7Se0.3	1.1	373 K	10
Bi2Te3	1.1	340 K	19
Bi0.4Sb1.6Te3	1.38	383 K	21
Bi2Te3	1.35	300 K	22
(CuBiSb)2Te3	1.2	473 K	This work

---

Because precise values of the lattice thermal conductivity were not obtained, Raman scattering was investigated to identify the effect of substitution on the lattice vibrations. Based on theoretical lattice dynamics, the rhombohedral (R3̄m) structure contains 15 lattice dynamical modes at q = 0, including 3 acoustic modes and 12 optical modes. The 12 optical modes are classified as:

Γ = 2E_g + 2A_1g + 2E_u + 2A_1u (8)

where E_g and A_1g are Raman-active.²³ The E_g modes represent vibrations along the layers (in-plane), whereas the A_1g modes represent vibrations along the c-axis perpendicular to the layers (out-of-plane) (Fig. 6b). Owing to symmetry restrictions, the Raman spectra display four peaks: E_g(1), A_1g(1), E_g(2), and A_1g(2). However, the lowest-frequency mode E_g(1) is rarely present.²⁴ In this work, three aspects are discussed below. Firstly, the relative intensity and frequency of the Raman patterns vary distinctly with the composition (Fig. 6a). After doping of Cu into the Bi₂Te₃ matrix, some new lattice vibration modes occur due to Cu–Te bonding (marked by *). However, owing to the lack of Raman modes for Cu–Te compounds in the literature, we cannot identify the new peaks. Secondly, the assignment of Raman modes and change in frequencies were analyzed (Fig. 6c). All spectra were decomposed into individual profiles in the wavenumber range from 50 to 250 cm⁻¹ (the inset of Fig. 6a is an example). The strong peaks around 96, 114, and 133 cm⁻¹ are assigned to A_1g(1), E_g(2), and A_1g(2) modes, respectively. The weak peaks between 150 and 210 cm⁻¹ are associated with unknown Raman modes (U), which are probably related to contamination on the surface such as Bi₂O₃.²⁵ Most Raman modes reveal a slight increase in frequency with an increase in Cu content (Fig. 6c). Thirdly, the alterations in Raman modes reveal changes in lattice dynamics and thus thermal conductivity, which result from atomic displacement, fluctuations in mass and strain fields caused by substitution.

Fig. 6 (a) Raman spectra of the synthesized bulks. The appearance of new modes is marked by *. The inset is an example of Gaussian fitting to the experimental data. (b) Diagram of the Raman-active vibrational modes. (c) Dependence on Cu content of the Raman mode frequencies.

4. Conclusions

Polycrystalline Cu_xBi_0.5Sb_1.5−xTe₃ bulks were successfully synthesized by the HPHT technique in a short time of 30 min. A small amount of Cu content gave rise to enhanced thermoelectric performance. A maximum ZT value of 1.20 was achieved at 473 K, and an average ZT value of 1.17 was obtained in the temperature range of 423–523 K for Cu_0.005Bi_0.5Sb_1.495Te₃. The simple, rapid and scalable HPHT method demonstrates the potential for large-scale commercial production. The increase in carrier concentration and widened band gap effectively suppress the adverse bipolar effect on transport properties. This strategy may further stimulate the shifting of peak conversion efficiency to higher temperatures.

Acknowledgements

This work was financially supported by the National Natural Science Foundation of China (51171070, 51071074).

Notes and references

1. T. M. Tritt, Science, 1999, 283, 804–805.
2. W. Kim, J. Zide, A. Gossard, D. Klenov, S. Stemmer, A. Shakouri and A. Majumdar, Phys. Rev. Lett., 2006, 96, 045901.
3. L. Hicks and M. Dresselhaus, Phys. Rev. B: Condens. Matter Mater. Phys., 1993, 47, 12727.
4. W. Liu, X. Tan, K. Yin, H. Liu, X. Tang, J. Shi, Q. Zhang and C. Uher, Phys. Rev. Lett., 2012, 108, 166601.
5. J. P. Heremans, V. Jovovic, E. S. Toberer, A. Saramat, K. Kurosaki, A. Charoenphakdee, S. Yamanaka and G. J. Snyder, Science, 2008, 321, 554–557.
6. R. Venkatasubramanian, E. Siivola, T. Colpitts and B. O'quinn, Nature, 2001, 413, 597–602.
7. B. Poudel, Q. Hao, Y. Ma, Y. Lan, A. Minnich, B. Yu, X. Yan, D. Wang, A. Muto and D. Vashaee, Science, 2008, 320, 634–638.
8. S. I. Kim, K. H. Lee, H. A. Mun, H. S. Kim, S. W. Hwang, J. W. Roh, D. J. Yang, W. H. Shin, X. S. Li and Y. H. Lee, Science, 2015, 348, 109–114.
9. L. P. Hu, T. J. Zhu, Y. G. Wang, H. H. Xie, Z. J. Xu and X. B. Zhao, NPG Asia Mater., 2014, 6, e88.
10. W. S. Liu, Q. Zhang, Y. Lan, S. Chen, X. Yan, Q. Zhang, H. Wang, D. Wang, G. Chen and Z. Ren, Adv. Energy Mater., 2011, 1, 577–587.
11. J. Cui, W. Xiu and H. Xue, J. Appl. Phys., 2007, 101, 123713.
12. S. Fujimoto, S. Sano and T. Kajitani, J. Alloys Compd., 2007, 443, 182–190.
13. D. Polvani, J. Meng, N. Chandra Shekar, J. Sharp and J. Badding, Chem. Mater., 2001, 13, 2068–2071.
14. J. Meng, N. C. Shekar, J. Badding, D. Chung and M. Kanatzidis, J. Appl. Phys., 2001, 90, 2836–2839.
15. H. Sun, X. Jia, L. Deng, P. Lv, X. Guo, Y. Zhang, B. Sun, B. Liu and H. Ma, J. Mater. Chem. A, 2015, 3, 4637–4641.
16. Y. Zhang, X. Jia, L. Deng, X. Guo, H. Sun, B. Sun, B. Liu and H. Ma, J. Alloys Compd., 2015, 632, 514–519.
17. K. Biswas, J. He, Q. Zhang, G. Wang, C. Uher, V. P. Dravid and M. G. Kanatzidis, Nat. Chem., 2011, 3, 160–166.
18. W. S. Liu, B. P. Zhang, J. F. Li, H. L. Zhang and L. D. Zhao, J. Appl. Phys., 2007, 102, 103717.
19. M. Saleemi, M. S. Toprak, S. Li, M. Johnsson and M. Muhammed, J. Mater. Chem., 2012, 22, 725–730.
20. A. Minnich, M. Dresselhaus, Z. Ren and G. Chen, Energy Environ. Sci., 2009, 2, 466–479.
21. Y. Dou, X. Qin, D. Li, Y. Li, H. Xin, J. Zhang, Y. Liu, C. Song and L. Wang, RSC Adv., 2015, 5, 34251–34256.
22. X. Tang, W. Xie, H. Li, W. Zhao, Q. Zhang and M. Niino, Appl. Phys. Lett., 2007, 90, 12102.
23. R. Vilaplana, O. Gomis, F. Manjón, A. Segura, E. Pérez-González, P. Rodríguez-Hernández, A. Munoz, J. González, V. Marín-Borrás and V. Muñoz-Sanjosé, Phys. Rev. B: Condens. Matter Mater. Phys., 2011, 84, 104112.
24. Y. Kim, X. Chen, Z. Wang, J. Shi, I. Miotkowski, Y. Chen, P. Sharma, A. L. Sharma, M. Hekmaty and Z. Jiang, Appl. Phys. Lett., 2012, 100, 071907.
25. S. Narang, N. Patel and V. Kartha, J. Mol. Struct., 1994, 327, 221–235.

---