Reactive close-spaced sublimation processed CuSbSe₂ thin films and their photovoltaic application

Thin film solar cells have attracted tremendous attention over the past few decades due to their advantages such as less consumption of raw materials and advantages for building integration. Traditional cadmium telluride (CdTe), copper indium gallium selenide (CIGS) and emerging lead halide perovskite thin film solar cells are the outstanding representatives, which have achieved an impressive certified power conversion efficiency of 22.1%, 22.6%, and 22.7%,¹ respectively. However, the scarcity of In and Te, the toxicity of Cd and Pb, and the instability of lead halide perovskites are intrinsic properties for these materials, which seriously restrict their large-scale commercial production. Thus, researchers never stop to look for new materials for thin film photovoltaic applications. Representative new absorber materials include Cu₂ZnSnS₄,^2,3 Cu₂SnS₃,⁴ Sb₂Se₃,^5,6 CuSbS₂,^7,8 and CuSbSe₂.^9,10

CuSbSe₂ is a low-toxic, low-cost, and earth abundant material with a proper band gap of ∼1.1 eV, which is quite suitable for single junction solar cells with over 30% theoretical power conversion efficiency (PCE). It is chemically similar to the well-known CuInSe₂ by replacing In³⁺ with Sb³⁺. Furthermore, the trivalent Sb with lone-pair 5s² electrons causes an even larger absorber coefficient.¹¹ Owing to all these advantages, various methods have been applied to prepare CuSbSe₂ films, such as one-step electrodeposition and annealing,¹² selenization of Sb-Cu metal precursors,¹³ hydrazine-based solution spin-coating and annealing,^10,14 and co-sputtering with Sb₂Se₃ and Cu₂Se.^9,15 An evaporation based method is simpler, convenient, and easy to be industrialized, which is widely used in thin film preparation. The representational close-spaced sublimation (CSS) method not only produced the first CdTe solar cell with more than 10% efficiency¹⁶ but also mainly resulted in the consequent growth of efficiency.¹⁷ However the evaporation based method for CuSbSe₂ film deposition has not been exploited, possibly because it is easy to decompose into Sb₂Se₃ and Cu₂Se or Cu₃SbSe₃ at high temperatures.¹⁵ 

Herein, we introduce reactions into the traditional CSS method and develop a new method called reactive close-spaced sublimation (RCSS) to deposit CuSbSe₂ films. The pre-sputtered Cu layer was used as a substrate and Sb₂Se₃ powder was evaporated using the traditional CSS method. The formation process was revealed by X-ray diffraction (XRD) analyses: Sb₂Se₃ partially decomposes during the evaporation, releases Se reacting with Cu to form Cu₂Se, and the newly produced Cu₂Se further reacts with Sb₂Se₃ to generate CuSbSe₂. A series of CuSbSe₂ films were prepared at different substrate temperatures (from 380 to 500 °C) and their phase purity, morphology, and electrical properties were characterized. The film prepared at 440 °C shows the best electrical properties with 7.80 × 10¹⁶ cm⁻³ carrier concentration and 133 Ω·cm resistivity. We further built photovoltaic devices with the substrate structure of AZO/CuSbSe₂/CdS/i-ZnO/ZnO:Al/Au and obtained a high PCE of 3.04%.

In the initial trials of the CSS method, we used CuSbSe₂ powder as the source, but only got some Sb₂Se₃ on the substrate and all Cu elements stayed in the source powder. Fortunately, a unique CuSbS₂ thin film preparation method inspired us, where the authors heated a Sb₂S₃/Cu multilayer in a vacuum to let Cu diffuse into Sb₂S₃, which reacted with Sb₂S₃ and eventually produced CuSbS₂ films.¹⁸ Then we pre-sputtered a layer of the Cu film on a normal substrate and changed the source powder to Sb₂Se₃, which is known for its easy evaporation and high vapor pressure.^19,20 The details of the whole CuSbSe₂ film deposition process are as follows:

The substrate was a 2 mm thick soda-lime glass with 1000 nm AZO (ZnO:Al) coating layer (Zhuhai Kaivo Optoelectronic Technology Co., Ltd.). The Cu layer (100 nm) was deposited via a magnetron sputtering system (Beijing Technol Science Co., Ltd.). Before sputtering, all the substrates were successively cleaned by ultrasonification in detergent, isopropanol, ethanol, and de-ionized (DI) water for 30 min each, then dried by nitrogen flow. The target was a 99.99% pure copper rectangular target with the size of 500 × 100 × 8 mm, and the distance between the substrate and target was 120 mm. To ensure film uniformity, the substrate holder had a target centered left-right movement during the sputtering. After the chamber evacuated to a base pressure below 8 × 10⁻⁴ Pa, 99.99% pure Ar was introduced at a rate of 100 sccm to maintain a working pressure of 0.1 Pa. 40 kHz pulse power was used for sputtering with a power of 1.5 kW. The deposition process was carried out under room temperature and continued for 10 min.

Then the RCSS process was performed in the equipment (MTI, Hefei, China) whose photograph and schematic representation are displayed in Fig. 1. Before the deposition, the AlN ceramic plate and quartz holder were annealed at 1000 °C for 1 h to remove possible contaminants and absorbed moisture. Then 0.5 g Sb₂Se₃ powder (99.999% purity, Jiangxi Ketai Advanced Materials, Co., Ltd.) was uniformly sprinkled on the AlN ceramic plate, which was placed on the quartz holder, as shown in Fig. 1(a). The substrate was placed on the quartz holder upside down, the Cu side facing Sb₂Se₃ powder. The distance between the substrate and source was about 0.8 cm. When the chamber pressure decreased to be lower than 10 mTorr, the two heaters were turned on at the same time, and the bottom one increased to 500 °C, while the top one heated the substrate to different temperatures (380, 400, 420, 440, 460, 480, or 500 °C). Both of the heaters would reach the respective setting temperature in 1 min, then maintained it for 6 min, and finally stopped at the same time.

To reveal the phase evolution mechanism from Cu to CuSbSe₂ in the RCSS process, a time-progression analysis was applied. Three samples named A2, A4, and A6 were prepared with a Sb₂Se₃ evaporation time of 2, 4, and 6 min, respectively. All these samples employed Cu layers of the same thickness of 100 nm and the same substrate temperature of 420 °C. X-ray diffraction (XRD, X’Pert Pro, PAN analytical, with Cu Kα radiation, λ = 1.541 78 Å) was applied and the patterns are shown in Fig. 2. Disregarding diffraction peaks of AZO, the main ingredient in the A2 sample is Cu₂Se and a small amount of CuSbSe₂. This means that after 2 min evaporation and reaction, the Cu layer entirely turned into Cu₂Se and started forming CuSbSe₂. The source of Se which selenized Cu into Cu₂Se was from the partial decomposition of Sb₂Se₃ as expressed in Eq. (1) during the evaporation process,²¹ 

From A2 to A6, the intensity of diffraction Cu₂Se peaks at 13.05°, 26.19°, and 43.96° progressively decreased and fully disappeared in A6, while the intensity of CuSbSe₂ gradually increased. This indicated that Cu₂Se gradually transformed into CuSbSe₂ as the evaporation time went on and 6 min was sufficient to form a single-phased CuSbSe₂ film. The evaporation process of Sb₂Se₃ has been carefully studied via the temperature dependent vapor-pressure equilibria of Sb₂Se₃, element Sb, and element Se.¹⁹ The result demonstrated that Sb₂Se₃ has a high vapor pressure of about 400 Pa at 500 °C. Se has an even larger vapor pressure, but the vapor pressure of Sb was far lower than that of both of them. Thus, we can conclude that during the process, part of Sb₂Se₃ was decomposed, and then the Se vapor reached the substrate and reacted with Cu as expressed in Eq. (2). Sb would largely stay in the source powder due to its low vapor pressure; similarly, Cu stayed in the CuSbSe₂ powder as mentioned above. The generated Cu₂Se would further react with remaining undecomposed Sb₂Se₃ vapor to form CuSbSe₂ as expressed in Eq. (3),

Substrate temperature is very influential to obtain a high quality film. Therefore, a series of samples with the substrate temperature ranging from 380 to 500 °C were prepared and the corresponding XRD patterns are displayed in Fig. 3(a). For all of these samples, the source temperature was fixed at 500 °C and the evaporation time was 6 min. Except for traces of Sb₂Se₃ (JCPDS: 00-015-0861) at low temperatures (380 and 400 °C), there were no other secondary phases such as unreacted Cu₂Se or the decomposition product, Cu₃SbSe₃,¹⁴ in any of the film. This suggests that the RCSS method can be utilized to synthesize single-phased CuSbSe₂ films. The secondary phase of Sb₂Se₃ that showed diffraction peaks at 15.03° and 16.87° was owing to the excessive temperature difference between the substrate and source. The vapor pressure of Sb₂Se₃ could be regarded as a fixed value when the source temperature was fixed at 500 °C. At the same source temperature, the lower the substrate temperature, the more Sb₂Se₃ vapor condensed onto the substrates. From 380 to 420 °C, as the substrate temperature rose up, the peak intensity of Sb₂Se₃ was decreased, which means less and less Sb₂Se₃ secondary phase in the film. This was also consistent with Andriy Zakutayev’s calculations and experiments, who showed that under fixed Sb₂Se₃ equilibrium vapor pressure, single-phased CuSbSe₂ could only exist in a certain temperature range. The film would contain some Sb₂Se₃ below this range, while above this range CuSbSe₂ would decompose into Cu₃SbSe₃.¹⁵ 

The morphology of these samples [Figs. 3(c)–3(g)] was characterized by field-emission scanning electron microscopy (FE-SEM, FEI Nova NanoSEM450). At 380 °C, the grains were small and parts of them seemed like lying rods or sheets. Because CuSbSe₂ has a two dimensional crystal structure, (002) and (004) peaks with lying orientation (CuSbSe₂ nanosheets stacking in parallel with the substrates) showed high intensity in this XRD pattern. With the increase of temperature, the orientation re-organized from [002], [004] to [013], which suggests CuSbSe₂ nanosheets incline upward against the substrate. In the corresponding SEM image of the 400 °C film, we could see the re-orientated grains. The film of 440 °C showed the best morphology with clear grain boundaries and a large grain size of about 1 μm. As the temperature continued to rise up, the film started melting a bit. The grain boundaries became blurred in the SEM image of the 480 °C film. In larger scale SEM images, we could clearly see many dendritic streaks, which were possibly formed by the partially melted CuSbSe₂ under the action of gravity, as the samples were placed upside down.

To study the electrical properties of CuSbSe₂ films, sister samples with a substrate temperature from 380 to 500 °C were prepared on frosted glass for Hall effect measurements (Ecopia HMS-5500, with gold electrodes). The corresponding substrate temperature dependent doping density and film resistivity are shown in Fig. 3(b). All films were measured as p-type from their positive signals. The hole concentration was estimated as 4.86 × 10¹⁷ cm⁻³ for the 380 °C sample, then slowly decreased with the increase of the substrate temperature to 7.80 × 10¹⁶ cm⁻³ at 440 °C, possibly because of the crystallinity improvement at higher temperature. After a slight increase at 460 °C, it rapidly rose to 1.79 × 10¹⁸ cm⁻³ for the sample prepared at 500 °C, possibly due to the presence of a minuscule amount of the decomposition product, Cu₃SbSe₃. Cu₃SbSe₃ is a p-type degenerate semiconductor and its presence is too small to be detected by XRD. The resistivity shows a reverse U-bend trend and the highest value was 133 Ω·cm at 440 °C, conforming to the lowest hole concentration.

The best carrier concentration for materials applied in single junction solar cells is known at around 10¹⁶ cm⁻³. However, it was a challenge to prepare CuSbSe₂ films with the desired 10¹⁶ cm⁻³ doping density due to the easily formed heavily p-type doped Cu₃SbSe₃ impurities and the low formation energy of acceptor defect V_Cu.^10,15 Compared with the doping density in hydrazine solution processed CuSbSe₂ films (10¹⁷–10¹⁹ cm⁻³),¹⁴ the hole concentration of our film prepared by the RCSS process is reduced by at least an order of magnitude. One possible reason could be the suppression of CuSbSe₂ decomposition into Cu₃SbSe₃ by the presented sufficiently high Sb₂Se₃ vapor. Another reason may be the appropriate amount of Se provided by the decomposition of Sb₂Se₃. Based on the results of density functional theory (DFT) simulation,¹⁰ under Se-poor, Cu-rich, and Sb-rich conditions, the donor defect Cu_i and accepter defect V_Cu have comparable formation energy. V_Cu is considered to be the main acceptors leading to high carrier concentration in p-type CuSbSe₂ films. While under this Se-poor condition, Cu_i has considerable compensation effects and decreases the concentration of free holes. The growth of carrier concentration at high substrate temperatures was tentatively attributed to the increased Se vapor supplied by enhanced Sb₂Se₃ decomposition at high temperatures. This is excellent news for tunable carrier concentration (10¹⁶–10¹⁸ cm⁻³) in single-phased CuSbSe₂ films.

X-ray photoelectron spectroscopy (XPS, ESCALab250) was implemented to analyze the chemical nature of CuSbSe₂ films. Results of the 440 °C sample, the one with best morphology and lowest doping density, are presented in Fig. 4. A Cu 2p doublet at the binding energy of 932.0 eV (2p_3/2) and 951.8 eV (2p_1/2) with a separation of 19.8 eV corresponded to Cu⁺, which is in good agreement with the literature results.^7,22 For antimony, two pairs of Sb 3d peaks were observed. The lower energy pair at the binding energy of 528.8 eV (3d_5/2) and 538.2 eV (3d_3/2) with the separation of 9.4 eV was consistent with Sb³⁺ in CuSbSe₂.²² Another pair of peaks at higher energy of 529.4 eV (3d_5/2) and 538.8 eV (3d_3/2) with the same separation of 9.4 eV were resulted from Sb₂O₃. The slight oxidization of the film surface was frequently observed in thermally evaporated Sb₂Se₃ films under similar conditions.²³ The O 1s peak at 529.9 eV overlapped with Sb 3d also confirmed the tiny oxidization of our film surface as the XPS was quite sensitive to the state of the film surface. The selenium core-level spectrum showed the Se 3d_5/2 peak at 53.5 eV and the Se 3d_3/2 peak at 54.4 eV with a separation of 0.9 eV, which was consistent with Se⁻².^19,22 The XPS spectrum further confirmed that the RCSS method successfully prepared CuSbSe₂ films with the normal valence state of Cu⁺Sb³⁺Se₂⁻².

To further fabricate photovoltaic devices with the substrate structure of AZO/CuSbSe₂/CdS/i-ZnO/ZnO:Al/Au [showed in Fig. 5(a)], the following steps were performed: (i) CdS layer (60 nm) grown by the chemical bath deposition method as reported in the literature;¹⁹ (ii) i-ZnO layer (50 nm) and ZnO:Al layer (700 nm) successively deposited by RF magnetron sputtering; and (iii) gold electrode grids (60 nm) deposited by thermal evaporation through a fingerlike metal mask. Finally, cells with an active device area of 0.40 cm² were scribed mechanically. The device performance shows a close relationship to the substrate temperature. At a high temperature of 480–500 °C, the devices had low a FF around 25% and an efficiency lower than 0.5% likely due to the presence of the Cu₃SbSe₃ secondary phase and overly high doping density. For devices prepared at a low substrate temperature of 380–420 °C, they demonstrated a slightly better efficiency of around 1%. Devices produced between 440 °C and 460 °C showed the highest average efficiency of 2.8% as CuSbSe₂ films produced at this temperature show optimal doping density and high crystallinity, thus enabling better performance. Our champion device obtained an encouraging conversion efficiency of 3.04%, with an open-circuit voltage (V_OC) of 0.34 V, a short-circuit current density (J_SC) of 18.84 mA/cm², and a fill fact (FF) of 47.34%. The corresponding current density versus voltage (J-V) curves in dark and under simulated AM 1.5G irradiation with a power density of 100 mW/cm² are shown in Fig. 5(b). The performance is slightly better than our hydrazine-processed CuSbSe₂ devices with an efficiency of 2.70% (V_OC = 0.36 V, J_SC = 20.52 mA/cm², and FF = 36.68%).¹⁴ Both of them may be encumbered with the low quality and thick window layer of our sputtered ZnO:Al. We can expect a further improvement for the quality of CuSbSe₂ films and a higher conversion efficiency record, in pace with the deepening of research and systematically optimizing each functional layers.

As a newly developed method, the RCSS method shows its tremendous competence and potential to prepare CuSbSe₂ thin films as it combines the advantages of evaporation and the metal selenization method. It not only simplifies the fabrication procedure but also produces excess Sb₂Se₃ vapor to prevent decomposition and secondary phase formation, a general challenge for the Cu-Sb-Se(S) system because of its complicated phase diagram. This method is applicable to other I-V-VI group materials such as CuSbS₂, CuBiS₂, AgSbS₂, and so on, which are widely used in photovoltaic, photodetector, thermoelectric, and photocatalytic fields. Ag or Cu can be deposited on the substrate, and Bi₂S₃ or Sb₂S₃ can be used as the reactive evaporation source.

In summary, a new reactive close-spaced sublimation (RCSS) method was developed to fabricate ternary chalcogenide CuSbSe₂ thin films. In this method, the evaporated Sb₂Se₃ serves as dual sources of Sb₂Se₃ and Se because of its partial decomposition. The reaction process was confirmed as Cu is first selenized into Cu₂Se and then Cu₂Se reacts with Sb₂Se₃ to generate CuSbSe₂. Using this method, high quality, single-phased CuSbSe₂ films with a doping density of ∼10¹⁶ cm⁻³ could be prepared at an optimal substrate temperature of 440–460 °C. An encouraging efficiency of 3.04% was obtained for the substrate CuSbSe₂ photovoltaic device, showing the potential of the RCSS method.

This work was financially supported by the National Natural Science Foundation of China (No. 61725401), by the National Key R&D Program of China (Nos. 2016YFA0204000 and 2016YFB700702), the HUST Key Innovation Team for Interdisciplinary Promotion (Nos. 2016JCTD111 and 2017KFXKJC003), and by the Special Fund for Strategic New Development of Shenzhen, China (No. JCYJ20160414102210144). The authors thank the Analytical and Testing Center of HUST and the facility support of the Center for Nanoscale Characterization and Devices, WNLO.