The effect of annealing temperature on Cu₂ZnGeSe₄ thin films and solar cells grown on transparent substrates

Kesterite-type material Cu₂ZnSn(S,Se)₄ (CZTSSe) has gained interest to be used as absorber for thin-film solar cells in the last years. This material has the advantage of consisting of elements of low toxicity and high abundancy in the Earth's crust, having a high absorption coefficient, and a tunable band gap energy E_g between 1.0 eV (Cu₂ZnSnSe₄) and 1.6 eV (Cu₂ZnSnS₄). Efficiencies of 12.5% and 12.7% have been achieved for Cu₂ZnSnSe₄ (CZTSe) [1] and CZTSSe [2, 3], respectively. However, these are still very low compared to other materials used in thin film photovoltaic devices, as e.g. Cu(In,Ga)Se₂ and CdTe [4]. The main limitation of the kesterite technology is its rather high V_OC deficit (E_g/q− V_OC) [5]. The formation of a deep Sn-related defect has been postulated as one of the reasons of the high V_OC deficit of these solar cells [6–8]. The partial substitution of Sn by Ge increases the band gap energy of the Cu₂ZnSn_1−x Ge_x Se₄ (CZTGSe) [9], leading to an enhanced V_OC and increase in device performance for low concentrations of Ge (x around 0.22) [10]. However, the increase in the Ge content led to a decreased efficiency due to higher V_OC deficit [10]. Recently, Choubrac et al [11] have shown the full potential of the total substitution of Sn by Ge in kesterite-based photovoltaic devices, achieving an efficiency of 8.5% and V_OC values reaching 65% of the theoretical limit for Cu₂ZnGeSe₄ (CZGSe) solar cells. To achieve such high V_OC, different strategies were implemented, i.e. chemical etching of the surface of the absorber layer and a heat treatment of the final device. Previously, an efficiency of 7.6% was achieved by the optimization of the CdS buffer layer to enhance the buffer/CZGSe interface [12]. Benhaddou et al [13] have shown the importance of the annealing conditions and surface treatments on CZGSe grown by selenization of sputtered metallic precursors improving the suppression of secondary phases formation and crystalline quality, obtaining a maximum efficiency of 5.6% [13], and later of 6.5% [14].

Very recently, Khelifi et al [15] have investigated the possible use of CZGSe-based devices as top cell in a tandem solar cell. They determined that the optimal E_g of the kesterite-based top cell for a four-terminal tandem device with crystalline Si as a bottom cell would be in the range of 1.5–1.7 eV. Furthermore, CZGSe-based top cells with E_g around 1.45 eV to 1.55 eV would be well suited together with CISe bottom cells with E_g of 1.0 eV [16] posing to be an option for S-free tandems, reducing the chalcogen-need to only Se in a potential chalcogenide thin-film tandem device. For that purpose, it is important to increase the transparency of the top cell maintaining a high performance. Additionally, semi-transparent kesterite solar cells start to attract the attention of the community, not only to be used for a cost-efficient tandem device, but also for bifacial solar cells and advanced building integrated photovoltaic (BIPV) concepts, further increasing their range of applications. Espindola-Rodriguez et al [17] reported bifacial CZTSSe solar cells with an efficiency of 7.7% and E_g = 1.09 eV using FTO as back contact and a Mo nanolayer between the kesterite and FTO. Later, the use of a transition metal oxides, such as TiO₂ and V₂O₅ together with a Mo:Na nanolayer improved the back interface of CZTSe, CZTS and CZTSSe solar cells deposited on FTO, obtaining efficiencies of 6.1%, 6.2%, and 7.9%, respectively [18]. Following this idea, we recently prepared CZTGSe thin films by co-evaporation onto Mo/V₂O₅/FTO/glass and subsequently annealed in Se-atmosphere, achieving a device performance of 5.6%, E_g = 1.28 eV and a transmittance of 30% in the near-infrared range [19]. Other groups also investigated semi-transparent kesterite solar cells grown on ITO. However, in the latter devices, in-diffusion towards the CZTSSe absorber layer and the formation of SnO_x seem to limit the conversion efficiency [20].

In this work, we propose to combine wide band gap CZGSe thin films and transparent FTO substrates. CZGSe thin films are deposited by co-evaporation onto Mo/V₂O₅/FTO/glass back contact structures followed by a thermal annealing treatment. We investigated the influence of the annealing temperature of the absorber layer on its transparency, structural, morphological, vibrational and compositional properties, as well as on the properties of the completed solar cells. To the best of our knowledge, a record efficiency of 5.3% (active area: 5.8%) for a 1.47 eV band gap CZGSe-based solar cells grown on FTO substrates is achieved.

2.1. Deposition of Cu₂ZnGeSe₄ thin films and device fabrication

2.2. Characterization techniques

Table 1 summarizes the composition of CZGSe thin films after the two different thermal treatments. The increase of maximum annealing temperature up to 525 °C has an important impact on the composition of the CZGSe layers. The increase of temperature influences the [Cu]/([Zn] + [Ge]) and [Zn]/[Ge] atomic ratios significatively. The sample annealed at 480 °C presents a Zn-poor and Cu-poor composition, which is more pronounced at the surface region of the absorber layer. However, the samples heated at 525 °C are characterized by a Cu-poor and Zn-rich composition, optimal conditions to achieve high-efficient kesterite solar cells [25].

In order to evaluate the effect of the maximum temperature on the diffusion of the elements through the absorber layer, GD-OES depth profiles were measured (see figure 1). A significant impact of the annealing temperature on the distribution of elements throughout the CZGSe/back contact structure becomes apparent as well. CZGSe annealed at 480 °C presents a very inhomogeneous distribution of all the elements, while the absorber heated at 525 °C is characterized by a very uniform distribution of Cu, Zn, Ge and Se. The sample annealed at the lowest temperature presents a significant drop of Cu and Zn and an increase of Ge, Se and Na signals in a region near the surface. The Na depth-profiles of both samples are characterized by a higher Na concentration near the back contact and the surface of the absorbers, following a U-shape, with the exception of a slightly higher Na content near the region where a significant drop of Cu signal is observed in the absorber treated at lower temperature. This effect is in agreement with the affinity of Na for Cu vacancies, V_Cu, which is widely observed before in CIGSe solar cells [26]. In addition, the surface of the sample annealed at 480 °C is characterized by increased Se and Ge contents and decreased Zn and Cu signals. The Sn-signal in the GD-OES-profiles comes from the FTO back contact.

Zoom In Zoom Out Reset image size

Figure 2 shows the GIXRD patterns of both CZGSe samples. For comparison, a diffraction pattern of FTO is also plotted in figure 2(a). As it can be observed in figure 2(a), CZGSe (00-052-0867) is the main phase formed together with FTO corresponding to the back contact. In addition, the presence of a MoSe₂ (04-004-8782) phase at the back interface cannot be ruled out, especially for the sample heated at 525 °C. Diffraction peaks at around 13.7°, 14.9°, 25.4°, 35.4° and 41.4° are assigned to a monoclinic GeSe₂ (04-003-1033) secondary phase in case of the sample annealed at 480 °C. The sample heated at 525 °C only presents the 002 Bragg peak of this phase at around 15° with much lower intensity. Taking into consideration the EDX and GD-OES measurements, the presence of a GeSe₂ phase was expected in the case of the 480 °C sample, which is now confirmed by the GIXRD patterns. Figures 2(b) and (c) show an enlarged plot of the GIXRD patterns in the range of 10°–29° diffraction angle measured at GI angles of 1° and 4° of each sample, clearly identifying the Bragg peaks corresponding to the GeSe₂ phase. More Bragg peaks with higher intensities related to GeSe₂ can be distinguished in figure 2(b) showing that this phase is segregated at the surface of the sample annealed at 480 °C. However, the presence of that secondary phase is not enhanced at the surface of the sample annealed at 525 °C, which is in good agreement with the EDX measurements. Due to the higher volatility of GeSe₂, the excess of GeSe₂ evaporates away from the film at the highest temperature, forming almost only CZGSe phase for the absorber annealed at 525 °C. There is a Ge/Se incorporation/loss phenomena caused by the excess Se and GeSe₂ present in the annealing atmosphere. In contrast to results presented in [13], no more secondary phases are segregated at temperature higher than 480 °C.

Zoom In Zoom Out Reset image size

A multiwavelength Raman scattering analysis was performed in order to further assess the formation of kesterite and possible secondary phases. For the latter, the spectra were recorded with excitation wavelengths of 442 nm and 785 nm, and no peaks of any secondary phases were found. Measurements under the 532 nm excitation allowed to assess the kesterite phase under non-resonant conditions and avoiding overlap with any possible secondary phases (figure 3). Moreover, the semi-transparent contacts enable to excite the absorber from both the front side (through the ITO layer) and the back side (through the FTO layer), allowing to obtain information about the surface and the back interfaces, similar as performed in [17, 19]. The spectra measured from the front side showed only relatively sharp CZGSe peaks, denoting the good crystalline quality of both absorbers independent of the annealing temperature. It is worth noting, that the strong changes in the cation's ratios of the two analysed samples that were found in the EDX measurements are expected to have significant influence on the Raman spectra, as they indicate changes in the type and quantity of point defects, as was previously found in similar CZTSe compounds [27, 28]. However, no such changes were found, which implies that the chemical compositions of the kesterite phase of both samples are in fact quite similar, while the main difference in the EDX results seems to be associated to the presence of the secondary GeSe₂ phase that was found by XRD.

Zoom In Zoom Out Reset image size

Therefore, we conclude that the annealing temperature does not significantly influence the formation of the CZGSe kesterite phase itself but the segregation of extra Ge and Se in secondary phase, which only appears at low temperature. The spectra measured from the back side of the absorber show intense peaks attributable to the MoSe₂ phase [29], formed at the back interface of the absorber, together with peaks of CZGSe. The relative intensity of MoSe₂ Raman peaks is higher in the sample annealed at 525 °C, indicating a thicker MoSe₂ layer at this back interface (see dotted spectra in figure 3). As detected by GD-OES, the Na-signal is also higher compared to the Mo-signal in the sample heated at higher temperature in comparison to the sample annealed at 480 °C. The higher Na content near the Mo contact might enhance the formation of a thicker MoSe₂ layer in this region. Previously, Na was found to act as a catalyst, promoting the formation of MoSe₂ layer at the back interface of CIGSe-based solar cells [30], which is also observed in the previous work with CZTGSe-based devices [19].

Comparison of the Raman spectra of the CZGSe phase measured at the front and the back side of the absorber layers did not show any significant difference for either sample. This implies that the CZGSe kesterite phase is quite homogenous throughout the sample depth, while the changes in elemental distribution found from GD-OES, mainly for the sample annealed at 480 °C, are related to the formation and inhomogeneous distribution of GeSe₂.

Figure 4 shows cross-sectional SEM images of both samples. The most striking observation is the larger grain size for the absorber annealed at higher temperature, as expected because the growth of the grains is thermally activated [31]. Apart from that, in both samples the absorber morphology is characterized by a compact structure, free of pin-holes.

Zoom In Zoom Out Reset image size

Solar cell devices were finished using the absorber layers investigated. Figure 5 shows J–V characteristics of the best CZGSe-based solar cells with the absorber layers deposited at 480 °C and 525 °C measured in the dark and under illumination. It can be observed that the device annealed at 480 °C presents a more significant cross-over between the illuminated and dark curves at J_cross = 5.5 mA cm⁻², while J_cross increases up to 6.6 mA cm⁻² when CZGSe was annealed at 525 °C. Mitzi et al [32] reported that this cross-over can be linked to a current blocking back contact between Mo and CZTSSe, which can suppress the hole transport. As stated above, we observed that the higher annealing temperature resulted in a higher accumulation of Na near the Mo back contact that allows for a thicker MoSe₂ layer. The different MoSe₂ layer formed at the back interface in both devices might therefore be responsible for the different behaviour of these J–V curves. The fact that the cross-over in the dark and illumination curves is still present in the 525 °C sample might indicate that the back contact is not completely optimized yet. On the other hand, this cross-over behaviour could also be due to a non-optimal kesterite/buffer interface. It has been observed that the presence of GeSe₂ secondary phase becomes more important at the surface of the CZGSe annealed at 480 °C, which presents the most significant cross-over.

Zoom In Zoom Out Reset image size

Table 2 shows the PV parameters of the best devices shown in figure 5. A higher efficiency of 5.3% is obtained for the device with CZGSe deposited at 525 °C, mainly due to a higher V_OC and fill factor FF. The lower cross-over effect and the improved V_OC in this device can explain the better FF. A maximum performance of 4.2% is achieved for the CZGSe-based device with the kesterite layer annealed at 480 °C. Therefore, the improvement of the device interfaces (increase of the thickness of MoSe₂ layer at the back and reduction of the GeSe₂ secondary phase at the front) and the higher quality of the bulk absorber layer (larger grain size and significant decreased of GeSe₂ phase) by increase of the annealing temperature, are fundamental to further enhance the device performance.

Figure 6(a) shows EQE spectra of the devices shown in figure 5 and table 2. A higher EQE starting from approximately 490 nm is measured for the CZGSe annealed at higher temperature, in agreement with the higher short circuit current density J_SC. The integrated J_SC determined from EQE appears in table 2, leading to an active area efficiency of 5.8% for the device annealed at 525 °C. EQE curves start to decrease starting from around 500 nm in both devices, indicating losses due to optical reflection and/or carrier collection suggesting that the CdS/CZGSe interface and CZGSe bulk material are not fully optimized, especially for the absorber synthesized at lower temperature. The stronger drop of EQE in the case of the CZGSe layer annealed at 480 °C can be related to the significant presence of GeSe₂ secondary phase, being detrimental for the carrier collection. IQE was also determined from the reflectance measurements of the solar cells. An increased spectral response until 800 nm wavelength is observed for both solar cells (see figure 6(a)). Short circuit current densities of 15.5 mA cm⁻² and 17.8 mA cm⁻² are calculated from IQE measurements for CZGSe-based solar cells with the absorber layer annealed at 480 °C and 525 °C, respectively. Therefore, these results suggest that the use of an antireflection coating could enhance the device efficiency up to 4.7% and 6.4% for CZGSe heated at 480 °C and 525 °C respectively. EQE spectra were measured under reverse bias (−1 V), increasing the EQE response in the range above 500 nm accordingly (see figure 6(b)). The difference between the EQE spectra without bias and with the reversed bias indicates a poor carrier collection, because of the increased depletion width under −1 V [33], which is more significant for the device with the absorber layer annealed at 480 °C with smaller grain size. This implies that the collection efficiencies are probably limited by very low carrier lifetime, which is found to be a more general problem limiting the V_OC in kesterite-based devices [32].

Zoom In Zoom Out Reset image size

The optical band gap energy E_g of CZGSe has been determined from EQE using two methods (figure 6(c)). The first one was from the linear fit of (E × EQE)² versus E, and the second from the inflection point of EQE at long wavelengths (minimum of dEQE/dE), as shown in figure 6(c). As reported in literature [34], the derivative method gives similar values to that obtained by optical absorption measurements, while the values obtained by the linear fit method are more accurate in samples with band gap energy gradient, which is not the case here. Slightly higher E_g are obtained by using the linear fitting method for both devices, but all the values are very consistent and lie between 1.45 and 1.47 eV. Therefore, the band gap energy of both absorber layers is not responsible for the difference in V_OC between both devices. As mentioned before, the presence of high amounts of GeSe₂ was detected in the CZGSe synthesized at lower temperature, which could influence the lower absorption in the bulk of the absorber layer and limit the V_OC and final device performance. The lower V_OC of this device agrees with the higher difference in EQE (−1 V)/EQE (0 V); however, more investigation is necessary to conclude about the origin of that.

Figure 7 shows the total, straight-through and diffuse transmittance measurements of both CZGSe absorber layers grown on Mo/V₂O₅/FTO/glass structure. Although the straight-through transmission is lower for the absorber annealed at 525 °C, the diffuse one is higher because of its larger grain size, leading to a higher total transmittance. The larger grain size increases the roughness of the surface leading to an increase of the light scattering. It should be noted that this scattered light can be still absorbed by the bottom solar cell in tandem devices, and therefore it is not lost. This result shows that higher annealing temperature has beneficial effect to the CZGSe-based devices without significant losses in total transmitted light.

Zoom In Zoom Out Reset image size

Obviously, highly transparent kesterite-based solar cells are desirable to be used for a tandem device, for bifacial cells or for BIPV applications. Therefore, it is necessary to find an optimal compromise between high efficiency and high transparency, which is a great challenge. Different strategies can be envisioned to enhance device performance and transparency. Among others, we point out the following strategies:

• Consideration of other back contact configurations, as investigated by Espindola-Rodriguez et al [17].
• Optimization of the Na-supply, for example, the thickness of the NaF layer, which probably plays an important role in the development of an optimized back interface, as observed in CIGSe-based solar cells [35].
• Optimization of the kesterite absorber layer: reduction of the thickness that allows for higher transparency without losing device efficiency [36], alternative thermal processes, increase of E_g by addition of S [37, 38] to gain in transparency.
• Alternative buffer layers to enhance the buffer/kesterite interface and band alignment [39].
• Optimization of the window layer (reduced absorption).
• Introduction of an anti-reflecting coating to reduce optical losses.

The optimization of these parameters can be crucial to achieve the goal of high performance and transparency together. This work has shown some first steps, indicating the potential of semi-transparent CZGSe-based devices with a promising future, but there is still much to do and investigate.

CZGSe thin films have been grown by co-evaporation followed by an annealing process at maximum temperature of 480 °C or 525 °C. The effect of the annealing temperature on the transmittance, compositional, structural and vibrational properties of the absorber layer grown on a semi-transparent Mo/V₂O₅/FTO/glass structure has been investigated. The CZGSe layer annealed at 525 °C is characterized by a larger grain size, a Cu-poor and Zn-rich composition and a uniform distribution of the elements throughout the absorber layer. However, the CZGSe treated at 480 °C presented a Zn-poor composition, the co-existence of kesterite with a GeSe₂ secondary phase, the latter also being present in the absorber annealed at higher temperature, but in less proportion. Raman spectroscopy confirmed the formation of a MoSe₂ layer at the back interface, with a higher thickness for the CZGSe annealed at 525 °C. Moreover, good crystalline quality of the kesterite phase was found for both samples from the spectra measured at the front and back side of the absorber. This yields that annealing temperature was mainly influencing the size of the grains, the formation of the secondary phase at the front interface and a MoSe₂ phase at the back, rather than had a significant influence to the CZGSe kesterite phase itself. J–V characteristics of the solar cells showed a reduced cross-over between the illuminated and dark curves, as well as a higher spectral response when CZGSe was synthesized at the highest temperature. The origin for that could be in the differently thick MoSe₂ layers formed between the kesterite and the back contact; however, other explanations as e.g. a reduction of the amount of the GeSe₂ phase in the bulk and at the surface region of the absorber layer and the larger grain size can play an important role as well. A higher total-area efficiency of 5.3% (active area: 5.8%) and E_g = 1.47 eV was achieved for the CZGSe annealed at 525 °C. A higher total transmittance was also obtained for the absorber heated at the highest temperature. Presenting the first efficient semi-transparent CZGSe-based solar cell is a crucial first step of the development.

This work was supported by Spanish Ministry of Science, Innovation and Universities Project WINCOST (ENE2016-80788-C5-2-R), Project RTI (2018-096498-B-I00, AEI/MICINN/ERDF) and European Project INFINITE CELL (H2020-MSCA-RISE-2017-777968). Authors from IREC belong to the SEMS (Solar Energy Materials and Systems) Consolidated Research Group of the 'Generalitat de Catalunya' (Ref 2017 SGR 862). MG and MP acknowledge the financial support from Spanish Ministry of Science, Innovation and Universities within the Juan de la Cierva (IJC2018-038199-I) and Ramón y Cajal (RYC-2017-23758) Fellowships respectively. MGP also acknowledges funding from AEI/MICINN (PTA2019-016763-I). The authors acknowledge the service from the MiNa Laboratory at IMN-CSIC, and funding from CM (project S2018/NMT-4291 TEC2SPACE), MINECO (project CSIC13-4E-1794) and European Union (FEDER, FSE). The authors would also like to thank Professor Rosalia Serna from IO-CSIC for helpful discussion.

All data that support the findings of this study are included within the article (and any supplementary files).