Temperature dependent transient surface photovoltage spectroscopy of a Cu1.95Zn1.1Sn0.96Se4 kesterite single phase powder

An off-stoichiometric but single phase Cu1.95Zn1.1Sn0.96Se4 kesterite powder was investigated by temperature dependent transient surface photovoltage (SPV) spectroscopy. SPV signals excited at different wavelengths were transformed into SPV spectra that depended on the response time of measurement. Shallow electronic states and states with transition energies at 0.83 eV or 0.78… 0.9 eV were distinguished. The temperature dependence of the band gap of Cu1.95Zn1.1Sn0.96Se4 was obtained. Results were discussed on the basis of defects in Cu-poor and Zn-rich kesterite.

Knowledge about defect states and the temperature dependence of the band gap (E g ) of Cu 2 ZnSnSe 4 (CZTSe) and related kesterite materials is of practical and fundamental interest regarding applications in solar cells 1,2 and a deeper understanding of the structure 3 and defects 4,5 of CZTSe. The band gap of CZTSe is about 1 eV and depends sensitively on preparation conditions 6 and defects. 4 Orderdisorder effects in the Cu-Zn layers of CZTSe have been shown by neutron and X-ray diffraction 7 and influenced E g as well. 8 Recently, E g of CZTSe was measured by spectral dependent ellipsometry over a wide temperature range 9 whereas defect states were not considered in the analysis. In Cu 2 ZnSn(S,Se) 4 (CZTSSe) kesterite based solar cells, the open circuit voltage is mainly limited by tail states 10 and the highest solar energy conversion efficiency can be reached for Zn-rich absorbers. 11 Further, off-stoichiometric but single phase kesterite has been discovered for Zn-rich but Cu-poor CZTSe. 12 In this work, defect states and the temperature dependence of E g in an offstoichiometric Cu 1.95 Zn 1.1 Sn 0.96 Se 4 single phase kesterite powder have been investigated by transient surface photovoltage (SPV) spectroscopy. In this method, surface photovoltage (SPV) transients are excited at numerous photon energies and the SPV values at fixed response time(s) are converted into spectra. Therefore, photo-generated electrons and holes can be excited over a large variety of optical transitions including delocalized and localized states (see Figure 1). The relaxation of charge carriers separated in space sensitively depends on defects states involved. The transient SPV spectroscopy allows, in contrast to conventional SPV spectroscopy 13 investigated with a Kelvin probe under cw illumination 14 or with a fixed capacitor under modulated light, [15][16][17] to distinguish spectral fingerprints of processes with relaxation times between ten(s) of ns and ms.  SPV transients were measured in the fixed capacitor arrangement 13 whereas the capacitor was formed between the carbon pad with the fixed Cu 1.95 Zn 1.1 Sn 0.96 Se 4 powder covered with a mica sheet and the quartz electrode coated with SnO 2 :F. SPV transients were excited with pulses (duration time 5 ns) of a tunable Nd:YAG laser (EKSPLA, NT 342/1/ UVE) at wavelengths between 720 and 2260 nm. SPV transients were recorded with a sampling oscilloscope (GAGE, CS14200, used sampling rate of 100 Ms/s) by applying a logarithmic read-out 18 and averaging over 40 transients at each wavelength (repetition rate 1 Hz). The temperature was varied between À120 and 120 C. The pressure in the home-made chamber was 3 Â 10 À6 mbar. As remark, SPV signals disappeared irreversibly during heating at 140 C. Figure 2 shows SPV transients excited with photon energies slightly above the band gap of Cu 1.95 Zn 1.1 Sn 0.96 Se 4 and measured at different temperatures. The sign of the SPV signals was negative at short times and changed to positive at longer times for measurements at lower temperatures. The SPV signals at 20 ns (after switching on the laser pulse, Dt 1 ) were À30.5 mV (À80 and 0 C) and À27 mV (80 C) and still increased slightly at times longer than Dt 1 within tens of ns at low temperatures.
A negative sign of SPV signals corresponds to preferential separation of photo-generated electrons towards the external surface. Fast charge separation is dominated by charge transport in a surface space charge region. Therefore, the grains of the investigated Cu 1.95 Zn 1.1 Sn 0.96 Se 4 powder can be considered as a p-type semiconductor. The change of the sign of the SPV signals at longer times gave evidence for trapping of holes at surface states or de-trapping of electrons from states close to the external surface. The further increase of SPV signals in time corresponds to an increase in the distance between the centers of positive and negative photo-generated charge, probably due to electron transport within a very thin region near the surface with a high density of defects.
The SPV transients decayed faster with increasing temperature for temperatures below 80 C. For example, the times at which the SPV signals decreased to the half amplitude (s 1/2 ) were about 0.44, 0.14, and 0.09 ls for À80, 0, and 80 C, respectively. The activation energy of the reciprocal s 1/2 (E A s ) was about 55 meV. This low value of E A s points to the importance of shallow states such as the copper vacancy (V Cu ) 4 for the relaxation of large negative SPV signals.
SPV transients measured at À120 C and excited at different wavelengths are presented in Figure 3. The SPV signals at Dt 1 were À26.5, À11, À2.7, and À0.7 mV for excitation at 1100, 1200, 1410, and 2010 nm, respectively. Therefore, the SPV signals decreased strongly in the region of E g of Cu 1.95 Zn 1.1 Sn 0.96 Se 4 (between 1200 and 1410 nm) but did not disappear for excitation with photon energies significantly lower than E g . The latter can be explained by excitation of mobile charge carriers from defect states. The change of the sign from negative to positive disappeared for the SPV transient excited at 2010 nm, which gave evidence for direct excitation of, for example, electrons from occupied defect states near the external surface to un-occupied states in the bulk.
The inset of Figure 3 shows SPV spectra converted at response times of 0.02, 0.1, 1, 10, and 100 ls (Dt 1 to Dt 5 , respectively) in a wide spectral range. The SPV signals at Dt 1 and Dt 2 were very similar, amounted to about À1 mV even at the lowest photon energies, increased to À3.8 mV between 0.83 and 0.90 eV, decreased practically to zero at 0.97 eV and increased strongly to À34 mV between about 1.00 eV and 1.10 eV, i.e., near the band gap. The SPV signals at Dt 3 amounted as well to about À1 mV at the lowest photon energies, increased slightly to À1. to þ2.2 mV at þ2.6 mV at 0.96 eV (Dt 4 ) and 0.93 eV (Dt 5 ). The SPV signals at Dt 4 changed to only À0.5 mV near the band gap. In contrast, the SPV signals at Dt 5 did not show a signature of the band gap.
The transition setting on at around 1.00 eV corresponds to the band gap of the Cu 1.95 Zn 1.1 Sn 0.96 Se 4 kesterite phase. The transition at 0.83 eV can be assigned to Zn Sn if regarding the excess of Zn and the energy obtained from theoretical studies. 4 The low SPV signals setting on at the lowest photon energies can be related to defects deeper in the band gap such as tin vacancies (V Sn ) 4 or zinc interstitials (Zn i ) 4 or to surface states. The transition setting on between 0.79 and 0.83 eV led to an opposite sign of the SPV signals and to very long relaxation times, i.e., charge separation was not related to the surface space charge region in this case. The reason for transitions leading to SPV signals with a positive sign is probably a very thin defect layer with distributed trap states near the surface. Depending on the thickness of a related defect layer, an upward band bending very close to the surface can be a reason for the change of the sign of the SPV signals. Band bending at CZTSSe/CdS interfaces 19 or at grain boundaries between CZTSSe crystallites, 20 for example, has been demonstrated by photoelectron spectroscopy 19 or KPFM (Kelvin probe force microscopy), respectively. 20 Figure 4 shows the negative values of the SPV signals measured at 20 ns for À100 and 120 C in a spectral range between 0.97 and 1.07 eV. The investigation of the temperature dependence of the band gap (E g (T)) of Cu 1.95 Zn 1.1 Sn 0.96 Se 4 kesterite in terms of the onset energy or of a Tauc gap 21 is vague due to the superposition of different processes. The region near the E g had to be approximated by a superposition of a Gaussian with a temperature-independent peak energy (E t about 0.992-0.993 eV) and a temperature-dependent onset of fundamental absorption (E on about 1.008 and 0.995 at À100 and 120 C, respectively).
The Arrhenius plot of the amplitude of the Gaussian is given in Figure 5. The thermally activated part resulted in an activation energy E A G equal to about 0.26 eV. This means that the defect state with the optical transition at E t is coupled with a transfer of charge from or into a deep state the occupation of which is thermally activated. The state at E t and the deep state are probably related to the same but differently charged defect.
The value of E on can be practically treated as the band gap of Cu 1.95 Zn 1.1 Sn 0.96 Se 4 kesterite. The temperature dependence of E on of Cu 1.95 Zn 1.1 Sn 0.96 Se 4 kesterite is compared with the temperature dependence of E g of Cu 2 ZnSnSe 4 kesterite (values of E g after Choi et al. 9 ) in Figure 6. The values of E on of Cu 1.95 Zn 1.1 Sn 0.96 Se 4 kesterite are larger than those of E g of Cu 2 ZnSnSe 4 kesterite. Up to about 50 C, the decrease in E on of Cu 1.95 Zn 1.1 Sn 0.96 Se 4 kesterite with increasing temperature was only À0.08 meV/K, which was FIG. 4. Negative transient SPV spectra near the band edge obtained at 20 ns for À100 (circles) and 120 C (triangles). The spectra were approximated by Gaussians fixed in energy (thin and thick solid lines for À100 and 120 C, respectively) and parts shifting towards higher energy with increasing temperature (stars and squares for À100 and 120 C, respectively) where the arrows mark the onset energies. much less than for the temperature dependent decrease in E g of Cu 2 ZnSnSe 4 kesterite (about À0.13 meV/K) in the considered range. At temperatures above 50 C, the values of E on tended to saturate around 0.955 eV which was close to E t . In addition, the full width at half maximum of the Gaussian tended to increase from about 12 to 16 meV in the temperature range between À50 and 50 C. Therefore, the temperature dependence of E on of Cu 1.95 Zn 1.1 Sn 0.96 Se 4 kesterite was strongly influenced by defect states near the band gap. For comparison, high concentrations of different defect pairs can strongly influence E g of kesterites as well. 4 Electronic transitions in crystallites of a Cu 1.95 Zn 1.1 Sn 0.96 Se 4 single phase kesterite powder have been studied by temperature dependent transient surface photovoltage spectroscopy. This method allows distinguish shallow and deep electronic states in relation to the relaxation time of charge carriers separated in space. Electronic transitions could be correlated with results of theoretical analysis on defects in Cu poor but Zn rich kesterite. 4 The bandgap of Cu 1.95 Zn 1.1 Sn 0.96 Se 4 kesterite was higher than for Cu 2 ZnSnSe 4 kesterite and the temperature dependence of E on of Cu 1.95 Zn 1.1 Sn 0.96 Se 4 kesterite was lower than for Cu 2 ZnSnSe 4 kesterite. The analysis of defect states by transient surface photovoltage spectroscopy and their deeper understanding will be useful not only for the further improvement of kesterite solar cells but also for the investigation of any other photoactive material.