Synthesis and characterization of kesterite Cu₂ZnSnTe₄ via ball-milling of elemental powder precursors

Abstract

A homologous series of kesterite light absorber materials Cu₂ZnSnX₄ (CZTX; X: S, Se, Te) can be used for realization of multi-junction solar cells. A stable member of the series (i.e. Cu₂ZnSnTe₄ (CZTTe)) was synthesized using a mechano-chemical route from its elemental precursors. Detailed characterization of the as synthesized as well as annealed CZTTe samples are being reported. XRD, Raman, TEM and SEM-EDS studies as well as Raman spectra confirm the formation of single-phase, stoichiometric, kesterite CZTTe nano-crystalline powder. The bandgaps of various samples of CZTTe were found in the range of 0.84–0.88 eV as confirmed by UV-Vis-NIR spectroscopy. The low band gap < 1 eV, coupled with the high absorption coefficient of ∼10⁴ cm⁻¹ suggests the possible use of this material as the absorber layer in the bottom cell of all kesterite multi-junction solar cells.

---

Introduction

Materials in the kesterite family of compounds with the nominal formula, CZTX (Cu₂ZnSnX₄; X: S, Se), and primarily, those in which the sulphur is partially or completely replaced by selenium have been widely studied and reported in the literature. These materials are utilized as a light absorber in solar PV applications. Maximum power conversion efficiencies of 8.4% ¹ and 9.7% ² for CZTS and CZTSe respectively have been reported to date. The champion efficiency (12.6%) of solar cells composed of a light absorber in this family of materials, was reported with a solid solution of CZTS and CZTSe.^3–6 One of the reasons for the higher efficiency when using the solid solution is, the bandgap of the solid solution matches the energy corresponding to maximum efficiency in the Shockely–Queisser plot for single absorber solar cells. Such bandgap tuning in this system is possible due to; (a) systematic change in bandgap with change in anion species; the valence band maximum is composed mainly of anion np orbitals and (b) formation of solid solutions of CZTX (X: S, Se) over the entire range of compositions.

The global focus till date has been on the thin film solar cells based on absorber materials such as CZTS, CZTSe and their solid solution (i.e. CZTSSe). However, other members in this family of materials, for instance, CZTO and CZTTe have not been explored that well; very few reports on these compounds exist.^7,8 As noted above, the valence band maximum in the kesterite family of materials is primarily composed of anion np orbitals hybridized with the Cu s orbitals, and the conduction band minimum is primarily composed of anion ns, np orbitals hybridized with Sn s states.⁹ Therefore, the trend in the bandgap exhibited by the materials in the kesterite family with Cu, Zn and Sn (CZTX) should be E_g(CZTS) > E_g(CZTSe) > E_g(CZTTe). Such a trend in bandgap has been experimentally observed in the structurally similar Cu, Zn and Ge based kesterites (Cu₂ZnGeX₄; band gap sequence: Cu₂ZnGeS₄ > Cu₂ZnGeSe₄ > Cu₂ZnGeTe₄ ¹⁰). Based on these arguments, we can infer that a material composed of earth-abundant elements, structurally similar to already existing solar cell absorber materials, and more importantly with a bandgap in the red region of the electromagnetic spectrum can be realized, provided CZTTe could be synthesized. In addition to its application as photovoltaic absorber material, CZTTe can also be used as a potential thermo-electric material.^8,11 Moreover, all-kesterite multi-junction solar cells can be realized with CZTX (X: S, Se, Te) materials forming the different bandgap layers in multi-junction solar cells, just as in the case of InAs and GaAs materials. However, in contrast to InAs/GaAs solid solutions for multi-junction solar cell applications, the CZTX family of materials has the advantage that it is primarily composed of earth-abundant component elements, and thin films of these materials can be obtained via cheap, scalable processes.

Mechano-chemical reaction via ball-milling is considered as a low cost and scalable route for synthesizing alloys such as complex alkali metal hydrides,¹² LiFePO₄ ¹³ and solar PV materials like CZTS,^14–18 CIGS,¹⁹ CdTe²⁰ and CuInSe₂.²¹ Since, in addition to near room temperature occurrence of solid state reaction, ball-milling offers a particle size reduction up to nanoscale, its seems suitable for the synthesis of Cu₂ZnSnTe₄ nanoparticles.

In this paper we report on the synthesis of the quaternary compound Cu₂ZnSnTe₄ via ball-milling approach and study its structural characteristics and optical properties.

Experimental details

Raw materials

Elemental powders of Cu (2.6689 g), Zn (1.3730 g), Sn (2.4929 g) and Te (10.7184 g) (Sigma-Aldrich) with purity > 99.5% were fed in to the ball-mill bowl of 80 ml capacity. 1-Butanol, which is widely used in printing ink and coating industries, was added to the mixture as a wet medium to enhance the efficiency of ball-milling and to prevent the agglomeration of particles in the mixture.¹⁵

Synthesis procedure

Ball-milling of the crudely mixed elemental powders of Cu, Zn, Sn and Te (grey colored mixture) was carried out at 450 rpm for 30 h duration using a planetary ball-mill (Fritsch Pulverisette 6). The ball to powder mixture weight ratio was maintained at 5 : 1. After each hour of rotation, the ball-milling operation was paused and the instrument was allowed to cool down for 15 minutes. In addition, the direction of rotation was also reversed hourly. After completion of the experiment, the initial grey colored mixture transformed to a dark black colloidal solution. The as-synthesized dispersion was dip coated on to a glass substrate to obtain a thin film of CZTTe. These films were then subjected to an annealing treatment in N₂ atmosphere at 500 °C for 5 minutes; a heating rate of ∼10 °C per minute was used and the samples were allowed to cool down to room temperature, before removal from the furnace.

Material characterization

As-synthesized and annealed sample of CZTTe were characterized by X-ray diffraction patterns using Panalytical X'Pert-Pro (Cu K-α radiation, 1.5418 Å) diffractometer. Jobin-Yvon-Horiba (HR800UV) Raman spectrometer at ambient conditions equipped with a laser (514.5 nm) was used to obtain the Laser Raman Spectrum. TEM diffraction patterns were obtained on a Philips CM 200, operating at 20–200 kV accelerating voltage and having a resolution of ∼2.4 Å. EDS (Energy Dispersive X-ray Spectroscopy) analysis was done using Field Emission Gun-Scanning Electron Microscope (FEG-SEM), JSM-7600F model operating at an accelerating voltage of 0.1 to 30 kV, magnification range of 25–1 000 000× and having resolution 1.0–1.5 nm (15 kV). Thickness of the films were measured by a zeta 3D microscope. Optical characterization of the thin films made of as-synthesized as well as annealed samples was carried out using Perkin-Elmer Lambda-950 spectrometer.

Results and discussions

The XRD patterns of the CZTTe (Cu₂ZnSnTe₄) samples prepared after ball-milling for 30 h duration (as-synthesized [bottom panel] and annealed at 500 °C [top panel]) are shown in Fig. 1. The XRD patterns of the as-synthesized CZTTe samples exhibit major peaks at 2θ = 25.56°, 42.22° and 49.86° corresponding to d-spacing of 3.480, 2.138 and 1.827 Å respectively. The XRD pattern corresponding to a CZTTe sample that has been subjected to an annealing treatment exhibits the three major peaks at 2θ = 25.56°, 42.22° and 49.91° corresponding to d-spacing of 3.480, 2.138 and 1.825 Å. The JCPDS cards used to identify the peaks are card no: 01-081-5256 and 01-081-7520, corresponds to the simulated data for CZTTe possibly in kesterite phase. Our data exhibits a close match with the peak positions of the simulated XRD database of CZTTe. Since the XRD peaks of CZTTe closely matches with the peaks of Cu₂SnTe₃ and ZnTe, as such simply on the basis of XRD pattern, confirmation of phase pure CZTTe synthesis is not justified.

Fig. 1 XRD patterns of as synthesized as well as annealed sample of CZTTe.

In both the XRD patterns of the as-synthesized and annealed films shown in Fig. 1, the three major peaks could correspond to the (112), (220) and (312) planes of kesterite CZTTe according to PDF card no.: 01-081-5256 and 01-081-7520. Assuming that the quaternary kesterite phase has been synthesized in both cases, the lattice parameters for the CZTTe were determined to be a = b = 6.05 ± 0.01 Å, c = 12.1 ± 0.1 Å. The XRD peak positions and hence the lattice parameters, for both the as-synthesized and the annealed samples is within a standard deviation from the mean values. The differences between the as synthesized and annealed samples, however is seen as a decreased FWHM and increased intensity in annealed sample. This can be attributed to the increase in crystallinity upon annealing. The calculation of crystallite size using FWHM of the various peaks was performed using Williamson–Hall (W–H) equation.²² (β_hkl cos θ) vs. (4 sin θ) curve was plotted for the preferred orientation peaks of CZTTe in kesterite phase, where β_hkl is the FWHM, corrected for the instrument broadening. The slope and y-intersect of the fitted line represent the strain and crystallite size, respectively. Fig. 2 shows the Williamson–Hall plot for the as synthesized and annealed CZTTe samples. Lower slope of the line corresponding to annealed sample in W–H plot, implies the strain relief during short annealing step (i.e. 5 min@500 °C). By considering the lower magnitude of the intercept (kλ/D_av; k = 0.9 is constant, λ is wavelength for Cu target for XRD instrument and D_av is the average crystallite size), it can be concluded that the annealing step assisted in achieving significant grain growth at rapid rate. The calculated values of the crystallite sizes are given in Table 1.

Fig. 2 Williamson–Hall plot for CZTTe sample prepared by 30 h ball-milling; with and without annealing.

Table 1 XRD analysis of as-synthesized and annealed CZTTe samples, showing the calculated crystallite sizes corresponding to major peaks. D_av1 and D_av2 are calculated average crystallite size for the as-synthesized and annealed samples respectively

(hkl)	CZTTe
	As-synthesized			Annealed
	2θ (deg.)	βhkl (deg.)	Dav1 (nm)	2θ (deg.)	βhkl (deg.)	Dav2 (nm)
(112)	25.56	0.40	25.23	25.56	0.13	81.62
(220)	42.22	0.51		42.22	0.15
(312)	49.86	0.59		49.91	0.17

---

As mentioned earlier, XRD peaks for CZTTe, cannot be distinguished from Cu₂SnTe₃ and ZnTe, Raman spectroscopy becomes essential for phase confirmation. The Raman spectrum analysis has been discussed in detail to resolve the phase assemblage issue and confirm the phase purity of the CZTTe powders. Fig. 3 shows the Raman spectra of the as-synthesized and annealed CZTTe samples.

Fig. 3 Raman spectra of CZTTe sample prepared by 30 h ball-milling; with and without annealing.

The pattern corresponding to the as-synthesized material shows peaks at 119.3 and 135.9 cm⁻¹. Similar to our findings in case of CZTS and CZTSe reported elsewhere,¹⁴ the Raman spectra, peaks for CZTTe are also slightly blue shifted after annealing of the sample and the peaks appears at 122.1 and 137.2 cm⁻¹. The most intense Raman peak in the spectra corresponds to the strongest A1 mode which originates due to the motion of the Te atom. However, the Cu, Zn and Sn atoms remain in the state of rest, as discussed in the literature of similar compounds, e.g. as in the case of Cu₂ZnSnS₄, Cu₂ZnSnSe₄, and Cu₂ZnSiTe₄.^23–25 The Raman peak positions of the kesterite CZTTe phase are not being reported so far in the literature. However, analysis of both the XRD and Raman spectra together can be used to unambiguously determine the phase assemblage.

In order to confirm the phase pure synthesis of kesterite CZTTe through ball-milling route, a detailed Raman analysis was done. In Raman spectra, the signature peak of quaternary phase is attributed to the main vibrational A1 symmetry modes. The main Raman vibration frequency (ν) depends directly on force constant f_M–X (where, M: Cu, Zn, Sn; X: S, Se, Te) and inversely on reduced mass (μ) as per eqn (1). Higher values of μ and smaller values of f_M–X imply a lower frequency for the characteristic peak of the quaternary phase. Since μ_M–Te > μ_M–Se > μ_M–S and f_M–Te < f_M–Se < f_M–S,²⁶ the main Raman peak for CZTTe should be at a lower frequency than that of CZTS and CZTSe.¹⁴ Following Tomlinson (1981),²⁷ a simplified expression for the Raman vibration frequency (ν) can be written as:

where, f is the effective force constant and μ is reduced mass. For any quaternary compound A₂BCD₄ the reduced mass (μ) can be calculated as:where, M_A, M_B, M_C and M_D are molecular weight of elements A, B, C and D respectively. For estimation of the intense Raman peak positions (A1 mode) of CZTTe, the available data for a similar compound Cu₂ZnSiTe₄ (CZSiTe) was used as a reference. The same value of effective force constants for both CZSiTe and CZTTe was assumed, as reported earlier in case of Se based compounds.²⁸ Therefore, the Raman frequency for Cu₂ZnSnTe₄ can be estimated using the following equation:

The A1 mode frequencies for CZSiTe as reported in literature are 134 and 151 cm⁻¹.²⁵ Using the Raman frequency values for CZSiTe and eqn (1)–(3), the values of A1 mode frequency obtained for CZTTe are ν_CZTTe ≈ 117 and 132 cm⁻¹. As per our experimental findings (Fig. 3), the Raman peaks appears at 119.3 & 122.1 cm⁻¹ for as synthesized and 135.9 & 137.1 cm⁻¹ for annealed CZTTe. The estimated values of the Raman vibration frequencies for CZTTe, agree reasonably well with the experimental finding of the present study. The difference in the estimated and observed frequency is slightly high in case of annealed CZTTe sample, since the reported data for CZSiTe used in the calculation are possibly corresponding to disordered kesterite structure as discussed in literature.²⁵ In addition to the estimation of the Raman peak positions, correlation of the observed Raman spectra of CZTTe with that of compounds with similar compositions such as Cu₂ZnSiX₄ (CZSiX; X: S, Se, Te) and Cu₂ZnSnY₄ (CZTY; Y: S, Se) can help in confirming the findings of the present study. Similar approach of assigning Raman peaks to novel or less reported compounds is described by Levcenko et al. and Rincón et al.^25,28

Quaternary compounds in the Cu, Zn, Si based chalcogenide family of materials, CZSiS, CZSiSe, and CZSiTe have been considered for further discussion on Raman peak positions. The most intense Raman peak of CZSiTe appears at a frequency lower than that of CZSiS and CZSiSe,^25,29 which is also the trend observed in the present study and discussed above. To further validate the Raman peak positions for CZTTe, another approach was used. If the frequency of the most intense peak (A1 mode) of quaternary chalcogenides is defined as ν_CZXY (where X: Si or Sn (denoted as T in CZTY) and Y: S, Se or Te) then the peak shifting ratio can be used as a parameter to correlate the observed Raman peak position of CZTTe. By considering the reported values of the quaternary compounds and the Raman peaks observed in the present study, it can be observed that: ν_CZSiSe/ν_CZSiTe ≈ ν_CZTSe/ν_CZTTe (≈1.6) and ν_CZSiS/ν_CZSiTe ≈ ν_CZTS/ν_CZTTe (≈2.8–2.9).^14,25 This means the ratio of peak shifting in case of CZSiSe, CZSiTe and that for CZTSe, CZTTe are almost equal, which further justifies the assignment of observed Raman peak to CZTTe, in the present study. Further, the ratios of Raman peak frequency (A1 mode) of CZSiS, CZTS and CZSiSe, CZTSe closely matches with that of CZSiTe and CZTTe (i.e., ν_CZSiS/ν_CZTS = ν_CZSiSe/ν_CZTSe = ν_CZSiTe/ν_CZTTe ≈ 1.1).

Considering the XRD patterns again as shown in Fig. 1, the observed XRD peaks positions suggest the following phase assemblages in the material (i) phase-pure CZTTe or (ii) a mixture of CZTTe, Cu₂SnTe₃, and ZnTe. On top of it, the Raman spectra (Fig. 3) and its analysis discussed is indicative of the fact that the peaks corresponding to ZnTe (203, 404 cm⁻¹)³⁰ and Cu₂SnTe₃ (such as 167, 176 and 190 cm⁻¹)³¹ are not present. Therefore, we can conclude that the only possible phase consistent with both our XRD and Raman data is Cu₂ZnSnTe₄.

TEM studies were carried out to further validate our findings in XRD and Raman spectra. The BF images in Fig. 4(a) correspond to the ball-milled, annealed powder of CZTTe. HREM image as shown in Fig. 4(b) of one such CZTTe particle exhibiting lattice fringes corresponds to the (220) (d = 0.215 nm) and (200) (d = 0.304 nm) planes of CZTTe. Consistent with our findings in the XRD pattern (Fig. 1), the three strongest reflections for the kesterite structure are also seen in the ring SAED patterns Fig. 4(c), emphasizing the polycrystalline nature of the samples. The simulated SAED ring pattern (left portion of Fig. 4(c)) was obtained using the standard Crystallographic Information File (CIF) of CZTTe, which is available in ICSD database based on theoretical calculation for structure of CZTTe (ICSD code: 656156). The plane corresponding to each ring were assigned using diffraction ring profiler software.³² It can be clearly seen that the obtained ring pattern (right portion of Fig. 4(c)) matches very closely with the simulated pattern.

Fig. 4 TEM bright field image (a), HREM (b) and SAED analysis showing with diffraction ring simulation (c) of CZTTe sample prepared during 30 h ball-milling.

To determine the stoichiometry of the synthesized compound, analysis of the EDS (SEM) data of the sample was performed for the as-synthesized sample at two random spots. The results are shown in Table 2. It can be seen that the sample is very close to stoichiometry as expected, illustrating that the ball-milling preserved the stoichiometry of the elemental precursor mixture.

Table 2 EDS analysis of as-synthesized CZTTe prepared after 30 h of ball milling. The readings were taken at two different spots

Atomic percentage (%)
	Cu	Zn	Sn	Te
Stoichiometric	25	12.5	12.5	50
Spot-1	25.08	12.23	11.98	50.71
Spot-2	25.27	12.58	12.34	49.81

---

To determine the band gap of CZTTe sample, as synthesized nano crystalline CZTTe powder was dip coated on a glass substrate to form a film of thickness ≈ 2 μm. Optical transmission (T) and reflection (R) measurements for as synthesized as well as annealed film of CZTTe were taken at room temperature. The optical absorption coefficient (α) was determined using the formula^33,34

where t is the thickness of the film. Fig. 5 shows the (αhν)² vs. hν curve (i.e. Tauc plot) of a CZTTe thin film of as synthesized sample as well as annealed thin film at 500 °C. The optical band gap of the samples was obtained by extrapolating the linear region of the plot (αhν)² versus photon energy (hν). The bandgaps of the as synthesized and annealed samples were found to be 0.84 and 0.88 eV respectively. A slight shift in the bandgap upon annealing is due to the improved crystallinity, as discussed elsewhere for similar compounds.^14,17,35 The absorption coefficient α was found to be of the order of 10⁴ cm⁻¹ in the visible region of the solar spectrum, which makes the CZTTe material suitable for light absorbing application.

Fig. 5 Tauc plot showing band-gap of CZTTe samples.

Conclusions

In conclusion, the synthesis of Cu₂ZnSnTe₄, the stable material with the lowest bandgap in the quaternary kesterite CZTX series of compounds, has been achieved via ball-milling process. The obtained powders were characterized using XRD, SEM, TEM, Raman and UV-Vis-NIR spectroscopy. Both XRD and Raman spectra revealed the formation of phase-pure CZTTe compound. From an analysis of the XRD peak positions, the lattice parameter for CZTTe was determined to be a = b = 6.05 ± 0.01 Å, c = 12.1 ± 0.1 Å, showing that the c/a ratio is very close to that of the ideal kesterite structure. In conjunction with the XRD data, the detailed analysis of the observed Raman spectra has been used to unambiguously determine the phase assemblage, ruling out the presence of other phases. These conclusions are further corroborated by the HREM and SAED data. The optical band gap of CZTTe was determined to be in the range of 0.84–0.88 eV. Appropriate band gap with the high absorption coefficient of 10⁴ cm⁻¹ suggests CZTTe as a suitable material for PV applications.

Acknowledgements

The authors would like to thank Sophisticated Analytical Instrument Facility (SAIF) at IIT Bombay for providing the access to the facilities of FEG-TEM and FEG-SEM. Author PS would like to acknowledge the IITB-ISRO space technology cell at IIT Bombay for funding the research work through grant 15ISROC002. KRB would like to thank IRCC, IIT Bombay for partial funding of this project through the grant 12IRCCSG014.

Notes and references

1. B. Shin, O. Gunawan, Y. Zhu, N. A. Bojarczuk, S. J. Chey and S. Guha, Prog. Photovoltaics, 2013, 21, 72–76.
2. G. Brammertz, M. Buffière, S. Oueslati, H. Elanzeery, K. Ben Messaoud, S. Sahayaraj, C. Köble, M. Meuris and J. Poortmans, Appl. Phys. Lett., 2013, 103, 163904.
3. T. K. Todorov, K. B. Reuter and D. B. Mitzi, Adv. Mater., 2010, 22, E156–E159.
4. D. A. R. Barkhouse, O. Gunawan, T. Gokmen, T. K. Todorov and D. B. Mitzi, Prog. Photovoltaics, 2012, 20, 6–11.
5. T. K. Todorov, J. Tang, S. Bag, O. Gunawan, T. Gokmen, Y. Zhu and D. B. Mitzi, Adv. Energy Mater., 2013, 3, 34–38.
6. W. Wang, M. T. Winkler, O. Gunawan, T. Gokmen, T. K. Todorov, Y. Zhu and D. B. Mitzi, Adv. Energy Mater., 2014, 4, 1301465.
7. H. Shen, X. D. Jiang, S. Wang, Y. Fu, C. Zhou and L. S. Li, J. Mater. Chem., 2012, 22, 25050.
8. C. Sevik and T. Çağın, Appl. Phys. Lett., 2009, 95, 112105.
9. S. Chen, X. G. Gong, A. Walsh and S.-H. Wei, Appl. Phys. Lett., 2009, 94, 041903.
10. D. Chen and N. M. Ravindra, J. Alloys Compd., 2013, 579, 468–472.
11. K. Wei and G. S. Nolas, J. Solid State Chem., 2015, 226, 215–218.
12. L. Zaluski, A. Zaluska and J. O. Ström-Olsen, J. Alloys Compd., 1999, 290, 71–78.
13. H. C. Kang, D. K. Jun, B. Jin, E. M. Jin, K.-H. Park, H.-B. Gu and K.-W. Kim, J. Power Sources, 2008, 179, 340–346.
14. D. Pareek, K. R. Balasubramaniam and P. Sharma, Mater. Charact., 2015, 103, 42–49.
15. Z. Zhou, Y. Wang, D. Xu and Y. Zhang, Sol. Energy Mater. Sol. Cells, 2010, 94, 2042–2045.
16. Q. M. Chen, Z. Q. Li, Y. Ni, S. Y. Cheng and X. M. Dou, Chin. Phys. B, 2012, 21, 038401.
17. Q. M. Chen, X. M. Dou, Z. Q. Li, S. Y. Cheng and S. L. Zhuang, Adv. Mater. Res., 2011, 335–336, 1406–1411.
18. B. Pani and U. P. Singh, J. Renewable Sustainable Energy, 2013, 5, 0531311–0531319.
19. C. P. Liu and C. L. Chuang, Powder Technol., 2012, 229, 78–83.
20. C. E. M. Campos, K. Ersching, J. C. de Lima, T. A. Grandi, H. Höhn and P. S. Pizani, J. Alloys Compd., 2008, 466, 80–86.
21. S. Mehdaoui, N. Benslim, O. Aissaoui, M. Benabdeslem, L. Bechiri, A. Otmani, X. Portier and G. Nouet, Mater. Charact., 2009, 60, 451–455.
22. V. Mote, Y. Purushotham and B. Dole, J. Theor. Appl. Phys., 2012, 6, 1–8.
23. A. Khare, B. Himmetoglu, M. Johnson, D. J. Norris, M. Cococcioni and E. S. Aydil, J. Appl. Phys., 2012, 111, 083707.
24. D. Dumcenco and Y. Huang, Opt. Mater., 2013, 35, 419–425.
25. S. Levcenko, A. Nateprov, V. Kravtsov, M. Guc, A. Pérez-Rodríguez, V. Izquierdo-Roca, X. Fontané and E. Arushanov, Opt. Express, 2014, 22, A1936.
26. V. Kumar, J. Phys. Chem. Solids, 2000, 61, 91–94.
27. R. D. Tomlinson, Phys. Rev. B: Condens. Matter Mater. Phys., 1981, 23, 6288–6293.
28. C. Rincón and F. J. Ramírez, J. Appl. Phys., 1992, 72, 4321–4324.
29. M. Guc, S. Levcenko, V. Izquierdo-Roca, X. Fontane, M. Y. Valakh, E. Arushanov and A. Pérez-Rodríguez, J. Appl. Phys., 2013, 114, 173507.
30. F. Fauzi, D. G. Diso, O. K. Echendu, V. Patel, Y. Purandare, R. Burton and I. M. Dharmadasa, Semicond. Sci. Technol., 2013, 28, 045005.
31. G. Marcano, C. Power, C. Rincón, I. Molina and L. Nieves, Solid State Commun., 2011, 151, 451–455.
32. L. Zhang, C. M. B. Holt, E. J. Luber, B. C. Olsen, H. Wang, M. Danaie, X. Cui, X. Tan, V. W. Lui, W. P. Kalisvaart and D. Mitlin, J. Phys. Chem. C, 2011, 115, 24381–24393.
33. X. Lin, J. Kavalakkatt, K. Kornhuber, S. Levcenko, M. C. Lux-Steiner and A. Ennaoui, Thin Solid Films, 2013, 535, 10–13.
34. J. I. Pankove, Optical Processes in Semiconductors, Devor Publications, Newyork, 1975.
35. T. Toyama, T. Konishi, Y. Seo, R. Tsuji, K. Terai, Y. Nakashima, R. Maenishi, A. Arata, S. Yudate, Y. Tsutsumi and S. Shirakata, Jpn. J. Appl. Phys., 2015, 54, 015503.

---

Footnotes

† Electronic supplementary information (ESI) available. See DOI: 10.1039/c6ra09112a

‡ Authors contributed equally to this work.

---