The successful development of chalcopyrite-type semiconductors such as Cu(In,Ga)Se2 has led to commercialization by an increasing number of companies. Favorable optoelectronic properties yield efficiencies close to 20% on the single cell and well above 10% on the module level. Despite or even as a direct consequence of its recent and near-future success, the scarcity and increasing prices of indium eventually could limit the production growth for this type of solar cells.
Seeking alternatives not containing indium, the related thin-film material Cu2ZnSnS4 (CZTS) shows a very similar crystal structure with a direct band gap expected in the range of 1.4–1.5 eV 1, while containing solely abundant and nontoxic elements. Although this kesterite-type semiconductor material has not been studied much so far and little is known about its electronic properties, several successful photovoltaic devices have been demonstrated 2, 3, with the highest efficiency of 6.7% reported by Katagiri et al.4, 5. They prepared copper-poor and zinc-rich CZTS absorber layers by a co-sputtering technique using three different targets (Cu, SnS, and ZnS), followed by an annealing step in an H2S atmosphere lasting several hours.
Sequential evaporation processes in which SnS, ZnS, and CuS layers are alternatively deposited and subsequently annealed allow rapid formation of kesterite 6. However, films often show strong inhomogeneities and multiple phases 7. In this paper, we report on a different route, using a fast (16 min) coevaporation of all elements in a single-stage process. In this process a copper-rich growth mode has been used drawing on the experience with CuInS2-solar cell preparation where such growth conditions have been found to lead to the best electronic properties of the absorber layers and solar conversion efficiencies in contrast to the copper-poor growth generally preferred for the deposition of Cu(In,Ga)Se2 and also CZTS based solar cells so far. Evaporation rates and times were found by adjustment of parameters deduced from in-situ controlled sequential evaporation processes.
Polycrystalline CZTS thin films were deposited on molybdenum-coated soda lime glass using thermal coevaporation of copper, tin, and zinc sulfide source materials. Sulfur was evaporated using a cracker source with the effusion cell heated to 210°C and the cracker zone heated to 500°C.
Sulfur partial pressure was measured by an ion gauge and was controlled by the effusion cell temperature. The sulfur partial pressure was kept at 2–3 × 10−3 Pa during the entire deposition process. The substrate was heated up to 550°C with a ramp of 50 K/min and maintained for 10 min at this temperature before the copper, tin, and zinc sulfide sources were opened simultaneously for 16 min. Subsequently, the substrate was cooled down at a rate of 15 K/min to 200°C. After that the sulfur source was closed and the substrate was naturally cooled down to room temperature.
Solar cells of 0.5 cm2 area were fabricated in a standard process previously optimized for CuInS2, an absorber material with a similar band gap of Eg ∼ 1.5 eV 8.
The CZTS thin films were analyzed by grazing incidence X-ray diffraction (GIXRD) using a PANalytical XPertPro MPD system (CuKα1,2 radiation) and an incident angle of 0.5°. Scanning electron microscopy on cross-sections was used to analyze the film morphology and thickness. Energy-dispersive X-ray spectroscopy (EDX) mappings and linescans were also performed on cross-sections using an acceleration voltage of 7 kV. For both analyses a LEO1530 (Gemini) with a field emission cathode was used. The overall chemical composition was determined by EDX from top in a LEO440 SEM with hairpin cathode using an acceleration voltage of 12–20 kV. The characteristic L-lines of zinc and tin and the K-lines of copper and sulfur were used for the quantitative composition determination. The J–V characteristics of solar cells under illumination were measured with a solar simulator under standard test conditions without light soaking. External quantum efficiencies were analyzed using monochromatic illumination under short-circuit conditions.
RESULTS AND DISCUSSION
The scanning electron microscope (SEM) image of a cross-section of the as-deposited kesterite absorber layer is depicted in Figure 1. It shows a relatively homogeneous film all the way from the back contact to the top surface, with no apparent secondary phases. A small contrast change at the top surface might indicate a thin CuS-layer which is typically found for copper-rich growth conditions in chalcopyrite thin-film deposition. This has been checked by GIXRD analysis shown in Figure 2. The Bragg peaks in the diffraction pattern indeed indicate two types of crystal structures, one can be attributed to CZTS (kesterite, tetragonal, Inorganic Crystal Structure Database (ICSD) #171983) and the other one to CuS (covellite, hexagonal, PDF: 00-006-0464). From the GIXRD analysis alone, it is not possible to exclude the occurrence of additional secondary phases, because the Bragg peaks of Cu2SnS3 and ZnS could be hidden under the observed kesterite and covellite peaks due to the close relations between the structures of these compounds. The elemental composition has been checked by EDX analysis as shown in Table I. A Cu/(Zn + Sn) ratio of 1.4 and a Zn/Sn of 1 was determined for the as-grown films.
Table I. Chemical composition of CZTS film from coevaporation process determined by EDX.
In order to remove the CuS secondary phase the film was etched with a KCN solution, as is routinely performed on copper-rich grown Cu(In,Ga)S2 thin films 8. A GIXRD pattern of the etched film is also shown in Figure 2 (upper curve). It can be seen that the Bragg peaks pertaining to CuS have vanished indicating that the KCN etch has completely removed this secondary phase. Indeed EDX analysis of the etched films included in Table I now shows a Cu/(Zn + Sn) ratio of 1 and Zn/Sn = 1 indicating that stoichiometric kesterite films have been obtained.
To investigate the depth homogeneity of the film an EDX line scan was performed as shown in Figure 3. It can be seen that the copper, zinc, tin, and sulfur concentrations stay constant within the accuracy of the measurement, as indicated by the almost constant net counts of the four elements over the depth of the sample. No particular layering as observed in earlier sequential evaporation experiments 7 is visible. The net counts only give qualitative information about the depth distribution not about the chemical composition. As the energy levels of sulfur K-line and the molybdenum L-line overlap an apparent residual Mo signal from the back contact is detected throughout the absorber.
From the GIXRD measurement, the spatial uniformity, the stoichiometric chemical composition and the constant Zn/Sn before and after KCN etching, we exclude significant amounts of secondary phases in the as-grown films other than CuS.
Figure 4 shows a plan view of the CZTS layer after KCN etching. Grain sizes up to 1 µm can be detected. The somewhat porous morphology of the layer likely results from the etching of CuS-secondary phases which covered and enclosed the CZTS crystallites in the as-grown film.
The KCN-etched CZTS absorber layers were processed to solar cells by applying a 50 nm CdS buffer layer by chemical bath deposition, followed by a window layer of intrinsic (Zn,Mg)O and an aluminum-doped ZnO deposited by magnetron sputtering. A nickel/aluminum front contact grid (5% shading) was evaporated on top of the window layer. This solar cell device configuration has been developed and used for CuInS2-based solar cells and modules and has not been specifically optimized for the CZTS absorber layers. Figure 5 depicts the J–V characteristics of the best device measured at standard test conditions. This device showed a total area efficiency of 4.1% (effective area efficiency 4.3%) with an open-circuit voltage of 541 mV, a short-circuit current density of 13.0 mA/cm2, and fill factor of 59.8%. To our knowledge this is the highest efficiency obtained for a coevaporated CZTS-device up to date. To gain further insights in the device performance and loss mechanisms the external quantum efficiency (EQE) was measured on the same solar cell as shown in Figure 6.
The EQE shows a steep increase around 350 nm related to the absorption edge of the ZnO window layer, a maximum value of about 60% at wavelengths between 400 and 500 nm and a subsequent broad decline for wavelengths above 520 nm. The optical gap of the CZTS absorber layer can be estimated from this EQE measurement, if the absorption coefficient for the material is modeled assuming a direct band gap semiconductor with parabolic bands close to the band edge. As shown in the inset of Figure 6 a band gap of 1.51 ± 0.01 eV is obtained from the linear extrapolation of (h·ν ln(1−EQE))2 vs. h·ν9. This value is in very good agreement with two recent theoretical calculations putting the value of the band gap in CZTS at 1.5 eV 10, 11. For wavelengths larger than the estimated optical gap (820 nm) significant photocurrent collection is observed in the EQE. This is likely due to substantial band tailing due to large amount of lattice disorder in the CZTS film. The collection length can be estimated by analyzing the electrical characteristics of the Mo/CZTS/CdS/ZnO device with the SCAPS-1D device simulation program 12. Using material parameters consistent with those commonly used for device simulation of Cu(In, Ga)Se2 or CuInS2 solar cells 13, the J–V and EQE characteristics can be qualitatively reproduced by assuming a bulk diffusion length below 100 nm corresponding to a density of recombination centers in the range of 1018 cm−3, and a space charge width of around 180 nm, corresponding to a net acceptor level of about 2 × 1016 cm−3.
It has been shown that near-stoichiometric single-phased polycrystalline CZTS thin films can be prepared by a fast coevaporation process. Copper-rich growth conditions lead to the segregation of a CuS secondary phase which can be removed by KCN etching. Solar cell devices made from these CZTS layers show the currently highest device performance of CZTS based thin film solar cells fabricated by a fast coevaporation process with a maximum efficiency of 4.1%, open-circuit voltage of 541 mV, short-circuit current density of 13.0 mA/cm2, and fill factor of 59.8%.
The authors would like to thank the baseline team of the Division Materials Research for Solar Energy Technology at the HZB for providing substrates and the processing of solar cells. Furthermore, thanks to Melanie Nichterwitz for SEM cross-section images, EDX mapping, and linescans.