The three A symmetry Raman modes of kesterite in Cu2ZnSnSe4 · spectroscopy. The main peaks at 170 cm −1, and 195 cm are found to have A symmetry. The Raman signal at 170 cm−1

The three A symmetry Raman modes of kesterite in Cu2ZnSnSe4

Rabie Djemour,1,* Alex Redinger,1 Marina Mousel,1 Levent Gütay,1,4 Xavier Fontané,2 Victor Izquierdo-Roca,2 Alejandro Pérez-Rodríguez,2,3 and Susanne Siebentritt1

1Laboratory for photovoltaics, University of Luxembourg, Belvaux, Luxembourg 2Catalonia Institute for Energy Research (IREC), C. Jardins de les Dones de Negre 1, 08930 Sant Adrià del Besòs

(Barcelona, Spain), Barcelona, Spain 3IN2UB, Departament d´Electrònica, Universitat de Barcelona, C. Martí i Franquès 1, 08028 Barcelona, Spain

4Current address: Energy and Semiconductor Research Laboratory, Department of Physics, University of Oldenburg, Oldenburg, Germany

*[email protected]

Abstract: We investigate CZTSe films by polarization dependent Raman spectroscopy. The main peaks at 170 cm−1, and 195 cm−1 are found to have A symmetry. The Raman signal at 170 cm−1 is found to be composed of two modes at 168 cm−1 and 172 cm−1. We attribute these three Raman peaks to the three A symmetry modes predicted for kesterite ordered Cu2ZnSnSe4. The main Raman peak is asymmetrically broadened towards lower energies. Possible sources of the broadening are tested through temperature and depth dependent measurements. The broadening is attributed to phonon confinement effects related to the presence of lattice defects.

©2013 Optical Society of America

OCIS codes: (170.5660) Raman spectroscopy; (160.6000) Semiconductor materials.

References and links

1. S. Siebentritt and S. Schorr, “Kesterites - a challenging material for solar cells,” Prog. Photovolt. Res. Appl. 20(5), 512–519 (2012).

2. D. B. Mitzi, O. Gunawan, T. K. Todorov, K. Wang, and S. Guha, “The path towards a high-performance solution-processed kesterite solar cell,” Sol. Energy Mater. Sol. Cells 95(6), 1421–1436 (2011).

3. T. K. Todorov, J. Tang, S. Bag, O. Gunawan, T. Gokmen, Y. Zhu, and D. B. Mitzi, “Beyond 11% Efficiency: Characteristics of State‐of‐the‐Art Cu2ZnSn (S, Se) 4 Solar Cells,” Adv. Energy Mater. (2012).

4. U. Rau, D. Abou-Ras, and T. Kirchartz, Advanced Characterization Techniques for Thin Film Solar Cells (Wiley-VCH, 2011).

5. A. Pérez Rodríguez, L. Calvo-Barrio, J. Alvarez-Garcia, J. R. Morante, V. Bermudez, O. Ramdani, J. Kurdi, P. P. Grand, L. Parissi, and O. Kerrec, “Raman scattering microcrystalline assessment and device quality control of electrodepo sited CuIn(S,Se)(2) based solar cells,” Thin Solid Films 516(20), 7021–7025 (2008).

6. T. C. Damen, S. P. S. Porto, and B. Tell, “Raman Effect in Zinc Oxide,” Phys. Rev. 142(2), 570–574 (1966). 7. B. Eckert, H. O. Albert, H. J. Jodl, and P. Foggi, “Raman studies of sulfur crystal (α-S8) at high pressures and

low temperatures,” J. Phys. Chem. 100(20), 8212–8219 (1996). 8. C. Persson, “Electronic and optical properties of Cu2ZnSnS4 and Cu2ZnSnSe4,” J. Appl. Phys. 107(5), 053710

(2010). 9. J. Paier, R. Asahi, A. Nagoya, and G. Kresse, “Cu2ZnSnS4 as a potential photovoltaic material: A hybrid

Hartree-Fock density functional theory study,” Phys. Rev. B 79(11), 115126 (2009). 10. S. Botti, D. Kammerlander, and M. A. L. Marques, “Band structures of Cu2ZnSnS4 and Cu2ZnSnSe4 from

many-body methods,” Appl. Phys. Lett. 98(24), 241915 (2011). 11. M. I. Aroyo, J. M. Perez-Mato, C. Capillas, E. Kroumova, S. Ivantchev, G. Madariaga, A. Kirov, and H.

Wondratschek, “Bilbao crystallographic server: I. Databases and crystallographic computing programs,” Z. Kristallogr. 221(1_2006), 15–27 (2006).

12. M. I. Aroyo, A. Kirov, C. Capillas, J. M. Perez-Mato, and H. Wondratschek, “Bilbao crystallographic server. II. Representations of crystallographic point groups and space groups,” Acta Crystallogr. A 62(2), 115–128 (2006).

13. N. B. M. Amiri and A. Postnikov, “Electronic structure and lattice dynamics in kesterite-type Cu2ZnSnSe4 from first-principles calculations,” Phys. Rev. B 82(20), 205204 (2010).

14. T. Gürel, C. Sevik, and T. Cagin, “Characterization of vibrational and mechanical properties of quaternary compounds Cu2ZnSnS4 and Cu2ZnSnSe4 in kesterite and stannite structures,” Phys. Rev. B 84(20), 205201 (2011).

#186930 - $15.00 USD Received 12 Mar 2013; revised 24 May 2013; accepted 28 May 2013; published 17 Jun 2013(C) 2013 OSA 1 July 2013 | Vol. 21, No. S4 | DOI:10.1364/OE.21.00A695 | OPTICS EXPRESS A695

15. A. Khare, B. Himmetoglu, M. Johnson, D. J. Norris, M. Cococcioni, and E. S. Aydil, “Calculation of the lattice dynamics and Raman spectra of copper zinc tin chalcogenides and comparison to experiments,” J. Appl. Phys. 111, 083707 (2012).

16. M. Grossberg, J. Krustok, K. Timmo, and M. Altosaar, “Radiative recombination in Cu2ZnSnSe4 monograins studied by photoluminescence spectroscopy,” Thin Solid Films 517(7), 2489–2492 (2009).

17. Z. Chen, L. Han, L. Wan, C. Zhang, H. Niu, and J. Xu, “Cu2ZnSnSe4 thin films prepared by selenization of co-electroplated Cu-Zn-Sn precursors,” Appl. Surf. Sci. 257(20), 8490–8492 (2011).

18. N. Vora, J. Blackburn, I. Repins, C. Beall, B. To, J. Pankow, G. Teeter, M. Young, and R. Noufi, “Phase identification and control of thin films deposited by co-evaporation of elemental Cu, Zn, Sn, and Se,” J. Vac. Sci. Technol. A 30(5), 051201 (2012).

19. K. W. Adu, H. R. Gutiérrez, U. J. Kim, and P. C. Eklund, “Inhomogeneous laser heating and phonon confinement in silicon nanowires: A micro-Raman scattering study,” Phys. Rev. B 73(15), 155333 (2006).

20. K. W. Adu, Q. Xiong, H. R. Gutierrez, G. Chen, and P. C. Eklund, “Raman scattering as a probe of phonon confinement and surface optical modes in semiconducting nanowires,” Appl. Phys. A. Mmate. 85, 287–297 (2006).

21. S. Piscanec, M. Cantoro, A. C. Ferrari, J. A. Zapien, Y. Lifshitz, S. T. Lee, S. Hofmann, and J. Robertson, “Raman spectroscopy of silicon nanowires,” Phys. Rev. B 68(24), 241312 (2003).

22. H. Münder, C. Andrzejak, M. G. Berger, U. Klemradt, H. Luth, R. Herino, and M. Ligeon, “A detailed sRaman study of porous silicon,” Thin Solid Films 221(1-2), 27–33 (1992).

23. H. Richter, Z. P. Wang, and L. Ley, “The one phonon raman spectrum in microcrystalline silicon,” Solid State Commun. 39(5), 625–629 (1981).

24. I. H. Campbell and P. M. Fauchet, “The effect of microcrystal size and shape on the one phonon Raman spectra of crystalline semiconductors,” Solid State Commun. 58(10), 739–741 (1986).

25. X. Fontané, V. Izquierdo-Roca, E. Saucedo, S. Schorr, V. O. Yukhymchuk, M. Y. Valakh, A. Pérez-Rodríguez, and J. R. Morante, “Vibrational properties of stannite and kesterite type compounds: Raman scattering analysis of Cu2(Fe,Zn)SnS4,” J. Alloy. Comp. 539, 190–194 (2012).

26. D. Dumcenco and Y.-S. Huang, “The vibrational properties study of kesterite Cu2ZnSnS4 single crystals by using polarization dependent Raman spectroscopy,” Opt. Mater. 35(3), 419–425 (2013).

27. E. Rudigier, T. Enzenhofer, and R. Scheer, “Determination of the quality of CuInS2-based solar cells combining Raman and photoluminescence spectroscopy,” Thin Solid Films 480-481, 327–331 (2005).

28. A. L. Stancik and E. B. Brauns, “A simple asymmetric lineshape for fitting infrared absorption spectra,” Vib. Spectrosc. 47(1), 66–69 (2008).

29. P. K. Sarswat, M. L. Free, and A. Tiwari, “Temperature-dependent study of the Raman A mode of Cu2ZnSnS4 thin films,” Phys. Status Solidi B 248, 2170–2174 (2011).

30. M. Grossberg, J. Krustok, J. Raudoja, K. Timmo, M. Altosaar, and T. Raadik, “Photoluminescence and Raman study of Cu2ZnSn(SexS1-x)(4) monograins for photovoltaic applications,” Thin Solid Films 519(21), 7403–7406 (2011).

1. Introduction

Kesterite materials (Cu2ZnSnSe4 and Cu2ZnSnS4) have recently found increased interest as absorbers in thin film solar cells [1, 2]. In this paper we will concentrate on the selenide Cu2ZnSnSe4 (CZTSe) compound. This is motivated on the one hand by the record efficiencies in these technologies which have been achieved in Se rich pentanary compounds (11.1% [3]) on the other hand by the consideration that quaternary pure selenide or sulfide systems are easier to study.

Raman spectroscopy is an optical non-contact and non-destructive characterization technique often used to monitor the phases present and their crystal qualities [4]. The phonon energies probed in Raman spectroscopy are strain and composition dependent making it an ideal technique for thin film solar cell absorber characterisation [4, 5]. Raman spectroscopy probes optical phonons at the Γ point by measuring the energy of inelastically scattered laser light. Due to momentum conservation, Raman spectroscopy usually probes the k = 0 phonons. In this contribution, polarization Raman measurements are performed to assign symmetries to the observed phonon modes. Additionally, micro Raman measurements were done at different depths in the absorber. The main mode measured shows an asymmetry towards lower energies. The effects leading to this asymmetry are discussed leaving the phonon confinement model as the most likely explanation.


2. Sample preparation & Raman measurements

CZTSe absorbers with power conversion efficiencies of up to 6.2% were prepared via co-evaporation of Cu, Zn, Sn and Se on molybdenum coated soda lime glass in a molecular beam epitaxy system at 320°C. The samples have been post annealed in a graphite box at 500°C in a tube furnace. In order to prevent the decomposition of the material, excess Se and SnSe powders were provided. Further details of sample preparation are found in [6].

Micro Raman (µRaman) measurements were made at different depths in the absorber. The depth resolution was achieved by sputtering craters to different depths. The sputter damaged layer was etched away with bromine methanol for 15s (0.02M). The μRaman measurements were performed in a home built confocal setup equipped with a liquid helium cooled cryostat. The laser excitation spot size is estimated to be a micrometer in diameter. The excitation source is an Argon ion laser that can be equipped with an etalon to provide a single laser line. The resolution of the setup is estimated by measuring the single laser line full width half max; we obtain a FWHM of 3.2 cm−1. The same FWHM was found by measuring the narrow lines of sulphur reported at 214.4 cm−1 and 218 cm−1 in the literature [7]: the two lines were observed separately with an individual FWHM of 3.2 cm−1.

The spectra shown here are integrated 5 μm long line scans taken at an excitation laser power of about 2.5 mW. This setup is equipped with an additional light microscope enabling the user to monitor the exact spot measured by Raman spectroscopy. The polarized macro Raman measurements were performed on a T64000 Horiba Jobin-Yvon spectrometer. Polarization dependent measurements were done under different parallel (h z|xx|z i) and perpendicular (h z|xy|z i) [6] polarization configurations. A spot size of 100 μm was chosen to ensure the excitation of a high number of randomly orientated microcrystalline grains. In order to minimise the presence of thermal effects in the spectra, the excitation power on the sample surface was kept below 10 mW. All Raman measurements were performed with an excitation wavelength of 514.5 nm.

3. Literature review of phonon calculations

In CZTSe, a number of crystal structures with different band gaps are possible. The most probable ones according to calculations by various methods are kesterite and stannite [8, 9] [10]. All recent calculations, no matter what method used, agree that the binding energy is lower by a few millielectron volts per atom for the stannite compared with the kesterite [1]. For these crystal structures, the number of possible phonons and their symmetries are given by group theoretical analysis of the space group. The zone centre phonon representation for

the kesterite structure (space group: I 4) is: Γ = 3A + 6B + 6E. All these 15 phonon modes are Raman active modes [11, 12]. For the stannite structure the zone centre phonon

representation (space group: I 42m) is: Γ = 2A1 + A2 + 2B1 + 4B2 + 6E . Of these 15 modes only the A2 mode is not Raman active [11, 12]. The difference between these zone centre phonon representations is the number of A symmetry modes i.e. three Raman active A symmetry modes for kesterite and two A1 symmetry Raman active modes for stannite. This is therefore a possible distinction tool between kesterite and stannite. The energies of the corresponding zone centre phonons have been calculated by DFT.

Both DFT calculations and experimentally acquired A/A1 symmetry Γ point phonon energies are summarized in Table 1. The exact values of the calculated phonon energies differ but for the kesterite structure Amiri et al. [13] and Gürel et al. [14] agree that the two modes with the highest energy are separated by around 11 cm−1. While Khare et al. [15] and Gürel et al. [14] agree that the two lower energy A modes are separated by around 4 cm−1. The two stannite A modes are separated by 15 cm−1 to 28 cm−1 depending on the reference.


Table 1. Summary of DFT and experimentally acquired Raman active A symmetry mode energies (in cm−1) for CZTSe

Kesterite - calculated Stannite - calculated Experimental

symmetry references symmetry references references

[14] [13] [15] [14] [13] [15] [16] [17] [18] (this work)

A 196.2 203 213.1 A1 194.6 203 213.8 196 196 194 195

A 183.6 192 193 A1 180 175 184.5 173 173 171 171

A 181 178 188.1 167 168

In Raman spectroscopy an asymmetric broadening of the main peak is often observed which can be explained by two types of effects: those leading to an asymmetry by themselves like the confinement of the phonons by lattice defects or small crystallites, and effects that lead to a shift of the Raman mode energies and therefore need to be present in an asymmetric distribution in the probing volume. These effects are inhomogeneous laser heating [19, 20] [21], a distribution of defects or nanocrystallite sizes, a strain distribution [22], and composition gradients [16].

The Phonon Confinement Model (PCM) [23, 24] describes the relaxation of the q-conservation rule by the confinement of phonons to small crystallites or small coherent volumes limited by lattice defects, even point defects, which break the translational symmetry. In this case, the correlation length model gives an estimation of the average distance between defects in the scattering volume (for further details see chapter 14 of [4]). Confinement in real space is leading to a relaxation in reciprocal space i.e. the activation of non-centre phonons. The symmetry of the broadening of the Raman modes depends on the phonon density of states around the Γ point which is given by the dispersion relation of the considered mode around the Γ point. For example a maximum in the dispersion relation leads to asymmetric contributions at lower energies.

4. Raman spectroscopy of CZTSe

Figure 1 shows an overview of the CZTSe related Raman modes. The measured Raman spectra of our CZTSe samples are composed of 7 modes. The three main ones are at 168 cm−1, 171 cm−1 and 195 cm−1 and weaker modes at (51, 77, 231 & 244) cm−1. The Raman signal in the energy range of 170 cm−1 is often discussed in literature as one mode even if the signal often suggests a second mode (see Table 1). This is strongly supported by the higher Full Width at Half Maximum (FWHM) of this peak that is about twice that of the main peak at 195 cm−1. Figure 2 shows a micro Raman measurement restricted to the energy range 145-220 cm−1. The fit clearly shows that the broader Raman signal at 170 cm−1 spectral region is composed of two Raman peaks at 168 cm−1 (5 cm−1 FWHM) and 172 cm−1 (6 cm−1 FWHM). The FWHM of these modes is then comparable to the one of the main peak at 195 cm−1 (4 cm−1 FWHM).

Figure 1 further shows polarization dependent measurements: the relative intensity of the main modes at 168-172 cm−1 and 195 cm−1 to the other weaker modes changes with the polarisation.

The polarization measurements are done on polycrystalline films, i.e. on randomly oriented crystals. There is not necessarily a need to measure on single crystals for the identification of the fully symmetric A modes. In the case of the kesterite crystalline structure

(space group I 4) the intensity of the modes with A symmetry decreases when the spectra are measured in perpendicular configuration in relation to the values obtained when working in parallel configuration. This results from the structure of the Raman tensor of A symmetry modes, where all elements away from the diagonal are zero. On the other hand, the intensity of E symmetry modes increases when working in perpendicular configuration. In the case of


B symmetry modes, in principle both behaviors (decrease or increase of intensity of modes) can take place. However, previous polarization measurements performed on polycrystalline powder Cu2ZnSnS4 kesterite samples (which is a very similar compound with the same crystal structure) show that only A symmetry modes decrease when moving from perpendicular to parallel configuration [25] which is consistent with the results from Dumcenco et. al. [26]. This strongly supports the identification of the modes with decreasing intensity in perpendicular configuration with A symmetry modes.

No change in shape is observed for the modes at 168 cm−1 and 172 cm−1 when normalizing the polarization dependent measurements as seen in the inset in Fig. 1. The proposed two modes thus have the same polarization dependent behaviour, and therefore are both A symmetry modes. We propose that these three A symmetry modes are the three kesterite CZTSe modes since the mode energies are in accordance with the DFT calculations predicting one A mode of higher energy and two A modes of lower energy only separated by a few cm−1 (see Table 1).

Fig. 1. Raman spectra from CZTSe absorber taken under parallel (black) and perpendicular (red) polarization configurations. Inset shows normalized spectra in the energy range 150-180cm−1.

Fig. 2. Micro Raman spectrum of CZTSe absorber restricted to the spectral area of 145-220 cm−1 (black). The main mode at 195cm−1 is fitted with two lorentzians(a)) or an asymmetrically broadened lorentzian(b)) (blue). The residual is plotted red. The other modes at 168, 172 cm−1 are fitted with two lorentzians (grey line).The green line shows the sum of the fits.

Looking closer at the Raman spectra, one sees that the main Raman peak at (195 cm−1) is asymmetrically broadened towards lower energies. The best fit was achieved using a broadened Lorentzian of the form [27, 28]:


( )( )

02

0

0

ω

ωx1 4

ω ωωx

ω ω1 α

F

II

F

= +

−=− −

(1)

with the parameters: 0ω = 195.5; α 0.26 0.02; F 4.3 0.02= ± = ± , where 0ω is the peak

position, I0 is the peak intensity, F is the FWHM and α is the broadening parameter responsible for the asymmetry.

The influence of a further low intensity mode was discarded as a reason for the asymmetric broadening. In fact using two Lorentzians did not improve the fit quality for the sample shown here. The two Lorentzian fit shown in Fig. 2 shows a positive residual on the low energy side of the main mode and a negative one on the other side. This shows that adding a second Lorentzian is not sufficient to fit the asymmetry of the peak opposed to the asymmetric peak shape that perfectly fits the measured data. The asymmetric peak shape not only has less fit parameters than two Lorentzians but also shows better fit results. Therefore we favor the asymmetric shape for fitting. Another supporting argument is that the calculation of the phonon dispersion points to the same broadening of the main mode in the case of phonon confinement, as discussed below.

In some samples, prepared in the same way (see Figs. 3-4) we do see a contribution of a further mode at 190 cm−1 which we attribute to a different crystal modification, like stannite or disordered kesterite or Cu-Au ordering (as seen in the CZTS [15, 25]). The peak is not due to SnSe2 since this compound is observed at 183 cm−1 under the same measurements conditions on CZTSe samples grown under basically the same conditions. Because the occurrence of the peak at 190 cm−1 can be clearly discerned, as discussed in the following, the peak shape in those samples, where it does not occur (like the one in Fig. 2) is clearly due to broadening and not due to a second line.

Figure 3 shows the fitting of the main Raman mode of CZTSe only with the asymmetrically broadened Lorentzian mentioned earlier (Eq. (1). For this fit only those parts of the spectrum where the contribution of the main CZTSe mode is dominant are used: The spectral range where the CZTSe Raman modes of lower intensity are situated is disregarded and the spectral range between peaks is used to describe the background.

Fig. 3. Raman spectrum of CZTSe absorber restricted to the spectral area of 140-280 cm−1. The main mode is fitted with an asymmetrically broadened Lorentzian peak shape. The thin lines show the fit (blue) and residuals (red) of the data excluding the low intensity CZTSe peaks marked in grey, thick lines show fits and residuals with the low energy shoulder of the main mode (marked in blue) discarded when fitting.


Opposed to the fit with the asymmetric peak shape shown in Fig. 2 b) where the residual shows no special features, fitting the main Raman mode shown in Fig. 3 does not lead to a satisfactory fit (i.e. a fit without both a strong negative and positive residual (thin lines for fit and residuals in Fig. 3)). The bad quality of this fit is also seen in both ends of the spectral range fitted where the fit is higher than the measurement. This fit shown in Fig. 3 is significantly enhanced when the low energy shoulder of the main mode (marked in blue in Fig. 3) is discarded when fitting (thick lines for fit and residuals in Fig. 3). The fit in the central part of the peak is improved, as seen by the considerably smaller residual. The positive residual that peaks at around 190 cm−1 in this case is in fact a further Raman signal which is attributed to the separate phase.

To understand the origin of the broadening, we take into account that it is expected that the sample dependent effects (i.e. a distribution of defects or nanocrystallite sizes, a distribution of strain, composition gradients) leading to the asymmetry are likely to vary with the sample depth. However the depth resolved Raman measurements show that the fitted asymmetry parameter varies within the fit uncertainties.

In the following, the relevance of these effects on our samples and measurements is discussed. The assumption of a temperature gradient in the probing volume of the laser spot means that we have a low heat dissipation leading to a stronger heating in the middle of the laser spot. Therefore changing the laser heating power would lead to an observable temperature effect. This was experimentally tested: Fig. 4 shows μRaman spectra taken at the same spot of the CZTSe absorber with different laser powers of 0.5 mW and 2.5 mW. A lower laser power of 0.2 mW was also tested with the same result (but with significantly lower signal to noise ratio). The only difference is a difference in background (see: inset in Fig. 4), Whereas the temperature related effects are mainly peak shifting and asymmetric broadening [29]. This can be seen in Fig. 5 where the same spot on the absorber was measured at 10K and at 300K. Increasing the temperature of the sample results in a broadening and a red shift of the Raman modes. This temperature related shifting can be used to estimate the temperature increase due to the excitation laser irradiation. Assuming a linear dependence, 30K is estimated as an upper limit of the temperature increase of the probed volumes for the different laser powers. There is no peak shift and no broadening of the spectra measured at room temperature with different laser powers. Therefore we take the uncertainty given by fitting the peaks position (in Fig. 4) of 0.2 cm−1 as estimation for the peak shift due to different laser powers. This leads to the conclusion that the thermal conductivity is high enough to dissipate the power deposited by the excitation laser. Therefore we conclude that inhomogeneous laser heating effects do not play a role in the asymmetrical broadening of our spectra.

Fig. 4. Normalized straight line background corrected micro Raman spectra of CZTSe absorber restricted to the spectral area of 150-220 cm−1. Both measurements were taken at the same spot at different excitation laser powers (black curve: 0.5 mW/micro spot; red curve: 2.5 mW/micro spot). Inset shows data prior to background subtraction.


Fig. 5. Two micro Raman spectra of CZTSe absorber restricted to the spectral area of 150-220 cm−1. Both measurements were taken at the same spot at different temperatures (black curve: 300 K; blue curve: 10 K).

Strain leads to a shift of the Raman mode positions. A gradient in strain in the probed volume would lead to an asymmetry of the peaks. To see the effect of strain we compared measurements of the top of the absorber to the back of the absorber by mechanically removing the deposit. We observe a red shifting in the energy of the modes when the deposit is removed. This shifting is attributed to strain release. The fitted asymmetry value is comparable (within the errors of the fitting) for both measurements with a tendency towards higher values for the mechanically removed (i.e. relaxed) samples. Thus, in our samples inhomogeneous strain can be discarded as a reason for the asymmetric broadening: Lowering of the asymmetry is expected if strain is released, and if strain would play a significant role in the broadening of Raman modes. An asymmetric distribution of defects in the scattering volume is possible and cannot be excluded. However, the similar asymmetry values observed from spectra measured at different depths from the layers strongly suggest that this is not the origin of the observed asymmetry of the main Raman peak.

Changes in composition lead to a shifting of the modes while a gradient of composition within the probing volume would also lead to an asymmetric broadening. Composition dependent shifting in peak position was only seen in the pentanary system with varying S/Se content [30]. This does not apply here since the presented samples are pure CZTSe. Additionally, no clear correlation was seen between the composition and the main modes peak positions for different samples (composition measured by 20 keV EDX) and depth regions (depth dependence of composition measured by SIMS (not shown here)). So this effect is excluded from playing a role. So, we are left with phonon confinement as the source of the asymmetry.

Applying the PCM to the calculated phonon dispersions by Gürel et. al. [14], one can predict an asymmetry of the main Raman mode towards lower energies which is in accordance with our experimental findings. In fact, the calculated density of states around the Γ point for the highest energy A modes for both kesterite and stannite show an asymmetry towards low energies [14]. For the stannite the phonon dispersion relation of the A1 mode has a local maximum at the Γ point. For the kesterite the dispersion relation of the A mode of the highest energy is a bit more complicated: it has a maximum at the Γ point going from Γ to X, P, and N, with a minimum at the Γ point going to Z. At first glance, a saddle point seems contradictory to prior statements; this is not the case since the slope of the decreasing phonon energies (Γ to X or P) is higher than the slope of the increasing phonon energies (from Γ to Z).

The A modes of lower energy for both kesterite (two A modes) and stannite (the A1 mode) exhibit the opposite phonon dispersion behaviour around the Γ point i.e. a minimum at the Γ


point. This would lead to an asymmetry towards higher energies. However, the overlap between the two peaks in the 170 cm−1 region make a fit to the broadened Lorentzian unreliable. Therefore a second confirmation for the phonon confinement model is not seen in the lower energy A modes. Nevertheless, the asymmetry of the main mode agrees perfectly with the asymmetry of the phonon density around the zone center, and is thus a strong support for the asymmetry of the Raman peak being due to phonon confinement.

5. Conclusions

In summary, we experimentally attribute the main Raman modes (modes with highest Raman signal at 168 cm−1, 172 cm−1 and 195 cm−1) to the group theory symmetry, in this case A symmetry, by their polarization dependence. Fitting the Raman signal at 170 cm−1 shows that it is composed of two Raman modes, therefore we propose that the three A modes seen here are the three A modes of kesterite CZTSe. The main Raman peak is asymmetrically broadened to lower energies. This is attributed to the confinement of the phonons because of the presence of lattice defects in the scattering volume. Other effects that could lead to such broadening are discussed and excluded from playing a major role.

Acknowledgments

The authors acknowledge the financial support by the Luxembourgish Fonds National de la Recherche and the use of the SEM and XRD apparatus through the CRP Lippmann (Luxembourg).The research leading to these results was partially supported by Spanish “Ministerio de Economia y Competitividad” (MINECO) project KEST-PV (ref. ENE2010-121541-C03-01/02). Authors from IREC and the University of Barcelona belong to the M-2E (Electronic Materials for Energy) Consolidated Research Group and the XaRMAE Network of Excellence on Materials for Energy of the “Generalitat de Catalunya”. V. I.-R. thanks the MINECO, Subprogram Juan de la Cierva (ref. JCI-2011-10782).


Documents

The three A symmetry Raman modes of kesterite in Cu2ZnSnSe4 · spectroscopy. The main peaks at 170 cm −1, and 195 cm are found to have A symmetry. The Raman signal at 170 cm−1