Scalable and efficient Sb₂S₃ thin-film solar cells fabricated by close space sublimation

Solar cells convert solar energy directly into electricity and provide one of the most effective ways to generate renewable, sustainable, and affordable energy.^1,2 Recently, a group of non-cubic chalcogenide photovoltaics shed light on the potential thin film solar cell application.^3–5 The orthorhombic Sb₂Se₃ thin film solar cell can achieve power conversion efficiency (PCE) up to 7.6% with inexpensive low-temperature evaporations, e.g., vapor transport deposition (VTD), demonstrating the great potential of these materials.^6–10 Compared to the low bandgap of antimony selenide Sb₂Se₃ (i.e., 1.1 eV), antimony sulfide Sb₂S₃ possesses a wide bandgap of ∼1.7 eV, as well as desired photovoltaic behavior.^11,12 The Sb₂S₃ solar cells with the solution-based process with PCE of ∼7.5% have also been achieved.¹³ These materials exhibit unique crystal structures with quasi-one-dimensional ribbons bonded by weak van der Waals force, e.g., (Sb₄X₆)_n ribbons in Sb₂X₃,¹⁴ where X = (Se, S). The ribbon-like morphologies lead to highly anisotropic charge transport, different from traditional cubic solar cell materials with isotropic transport (e.g.„ CdTe, CIGS, and perovskites).^3,8,15

To date, a variety of growth technologies have been developed for Sb₂S₃ film deposition to achieve high-quality absorber with the ribbons aligned with the buffer layer for desired charge transport.³ For example, Sb₂S₃ solar cells can achieve PCE of 3.5% with rapid thermal evaporation (RTE).^15–18 Recently, high efficiency Sb₂S₃ solar cells using the solution-based process were reported. For instance, sensitized solar cells have PCE of 6.4% using a solution-based technology^18–20 and solution processing for highly efficient Sb₂S₃ with PCE of 6.78%–7.5% with varying starting precursors.⁵ Sb₂S₃ solar cells fabricated by atomic layer deposition (ALD) in planar structure can achieve PCE of ∼5.7%.²¹ However, it is highly desired to employ the physical vapor deposition technologies to integrate the Sb₂S₃ into the commercial thin-film solar cell manufacturing. In particular, fast deposition techniques with independent control of the substrate and source temperature are highly demanded for integrating Sb₂S₃ into the high-throughput manufacturing technology. The close space sublimation (CSS) process is well known for CdTe manufacturing which directly vaporizes the solid source materials and allows high-throughput deposition on moving substrates.^10,22,23 The CSS technique can greatly reduce the cost and improve the absorber quality.¹⁴ Recently, the CSS technology has been successfully applied to Sb₂Se₃ thin film solar cells with PCE of 6.8%.^14,24

In this work, we experimentally investigated the Sb₂S₃ growth and the performance of CSS-grown Sb₂S₃ solar cells. It is demonstrated that Sb₂S₃ can be successfully grown using the CSS technique. The best device performance is 3.8% PCE. A theoretical study based on first-principles density functional theory (DFT)^25,26 was performed to elucidate its electronic and optical property. In addition, the study of intrinsic point defects and deep energy states provides a deeper understanding of the device performance. Our work further paves the way of antimony chalcogenide based emerging cost-effective and earth-abundant thin-film photovoltaics for solar energy applications.

About 60-nm thick CdS window layers were deposited on a cleaned fluorine-doped SnO₂ coated soda-lime glass (FTO TEC 10, Pilkington, USA) by chemical bath deposition (CBD) at a bath temperature of 65 °C. The CdS films were annealed in air at 400 °C for 30 min to improve the crystallization with CdCl₂ solution treatment. Sb₂S₃ thin films were grown in Ar ambient using a commercial CSS system (MTI, USA). The high purity Sb₂S₃ powder (99.999%, Alfa Aesar, USA) was placed on the bottom AlN plate, and the CdS coated FTO substrate was loaded on the top AlN plate (10 mm distance from the Sb₂S₃ powder). To optimize the growth conditions of Sb₂S₃ absorbers, the substrate and source temperatures were varied from 300 °C to 400 °C and 500 °C to 580 °C, respectively, at chamber base pressure about 5 mTorr, and the deposition pressure is about 15 mTorr. The Sb₂S₃ thin film thickness can be tuned by varying the deposition time from 30 to 100 s, then switching off the halogen lamp heater, and cooling the film naturally to room temperature. The as-grown Sb₂S₃ films were rinsed with de-ionized water, and then graphite and Ag paste was screen printed on the Sb₂S₃ (with an active area of 0.08 cm²) to define the solar cells. The finished solar cell’s current-voltage (J-V) curve was characterized using a solar simulator (Newport, Oriel Class AAA 94063A, 1000 W xenon light source) with a source meter (Keithley 2420) at 100 mW/cm² AM 1.5G irradiation. A calibrated Si-reference cell and meter (Newport, 91150 V, certificated by NREL) was used to calibrate the solar simulator before each measurement. The external quantum efficiency (EQE) data were obtained by using a solar cell spectral response measurement system (QE-T, Enli Technology, Co., Ltd). The film thickness was determined by using the surface profilometer (Dektak V220). The structure of the grown films was characterized by using an X-ray diffraction (XRD) system (X’Pert). The film morphologies and chemical composition were determined by using the scanning electron microscope (SEM, JEOL 7000) with Electron-dispersive Spectroscopy (EDS) attached to the SEM. The Raman experiments were conducted on a single stage Raman spectrometer with a solid-state laser (Horiba LabRam HR, 532 nm wavelength). The absorbance and transmittance spectra were measured using a UV-Vis spectrometer (Shimadzu UV-1800). The atomic force microscopy (AFM) and conductive AFM images were recorded on a grounded Sb₂S₃ sample using atomic force microscopy (AFM, Park XE70). The topography and current images were simultaneously recorded in contact mode using a Pt/Ir coated contact probe (ANSCM-PT from AppNano, Inc.). The cantilever spring constant was about 3 N/m, and resonance frequency was ∼60 kHz.

Atomistic, electronic structures and optical properties of Sb₂S₃ crystals were calculated using first-principles DFT^25,26 as implemented in the Vienna Ab Initio Simulation Package (VASP).²⁷ The Perdew-Burke-Ernzerhof (PBE)²⁸ form of exchange-correlation functional within the generalized gradient approximation (GGA)²⁹ and a plane wave basis set with a 400 eV energy cutoff were employed for VASP calculations. To account for van der Waals (vdW) interactions between (Sb₄S₆)_n ribbons, we adopted optB86-vdW functional.^30–32 The structural optimization and electronic relaxation were calculated in a Γ-centered Monkhorst-Pack³³ k-point sampling grid of 4 × 12 × 4. The maximal residual force of each atom is less than 0.01 eV Å⁻¹, and the convergence criterion for electronic relaxation was set to 10⁻⁶ eV. As the DFT-GGA often underestimates the bandgap, we adopt the modified Becke-Johnson (MBJ)³⁴ exchange potential and hybrid HSE06 functional³⁵ for electronic structure and optical property calculations. We used a 12 × 28 × 12 k-point grid in the MBJ calculations and 3 × 9 × 3 k-point grid in the HSE06 calculations. The MBJ calculation yields a bandgap close to the HSE06 result for Sb₂S₃. For the defect calculations, we increase the supercell by 2 × 6 × 2 which contains 480 atoms and apply an energy correction scheme which includes potential alignment and image-charge correction. There is a trade-off between computational cost and errors caused by the finite-size effect.^36,37 We checked the cell-dependent defect formation energies and found that the error caused by the finite-size effect is large in the 1 × 3 × 1 supercell but becomes much smaller in the 2 × 6 × 2 supercell, suggesting that it is necessary to employ a large supercell. The cutoff energy for defect calculations is set to 300 eV. More details of the calculations and methods can be seen in the supplementary material.

Since Sb₂S₃ has a low melting point (T_m ∼ 550 °C) and high vapor pressure, we deposit Sb₂S₃ with a relative lower source temperature than that for Sb₂Se₃ (T_m ∼ 608 °C).³⁸ As shown in Fig. 1(a), the XRD spectrum for the Sb₂S₃ film was indexed as orthorhombic structure with Pbnm space group, with respect to the Joint Committee for Powder Diffraction Standards (JCPDS) 42-1393.³⁹ The as-grown Sb₂S₃ film displays strong (211)-preferred orientation. The substrate temperature during the growth significantly impacts the orientation of the Sb₂Se₃ due to its orthorhombic crystalline structure, which is similar to the previous report.¹⁶ This (211)-preferred orientation can benefit charge transfer along the (Sb₄S₆)_n ribbons, like the (Sb₄Se₆)_n ribbons in Sb₂Se₃ film.^8,14 The Raman spectrum is shown in Fig. 1(b), suggesting a good crystallinity. The two major peaks located at 139 and 292 cm⁻¹ are ascribed to the Sb and symmetric vibrations of Sb₂S₃ pyramidal units with C_3v symmetry, respectively.^40,41 Other weak Raman peaks are indexed at 184 and 247 cm⁻¹. To understand the Sb₂S₃ electronic structure, Figs. 1(d) and S1 show the calculated electronic structure and orbital-resolved projected density of state (PDOS) using first-principles DFT with MBJ³⁴ and HSE06 functional³⁵ (calculation details can be found in the supplementary material). The PDOS indicates that the valence band is dominated by S-p orbitals, and the conduction band is dominated by Sb-p and Sb-s orbitals. Moreover, the DFT-MBJ results predicted that the indirect and direct bandgaps of Sb₂S₃ are 1.71 and 1.74 eV, respectively, which agree with the experimental optical bandgap (1.56 and 1.71 eV; see more details later).

Figure 2(a) shows the top view of the as-grown Sb₂S₃ film, indicating a dense and pin-hole free morphology with polyhedron grains. The grain size distribution is non-uniform with some extremely large grains (∼1 µm) and some small grains (∼300 nm). This may be due to the fast deposition process during the CSS deposition (∼1 µm·min⁻¹). The cross-sectional image as shown in Fig. 2(b) indicates that the Sb₂S₃ thickness is ∼400 nm with a 50 nm CdS layer. The chemical composition was determined using the energy dispersive X-ray spectroscopy (EDS), as shown in the inset of Fig. 2(b), where the atomic ratio between S and Sb is 60.4:39.6, close to the stoichiometry of 3:2 in Sb₂S₃. The chemical distribution across the film is characterized using EDS mapping, as shown in Fig. 2(c), where the Sb and S elements distributed uniformly without segregation. To ascertain the uniformity of the grain size and the localized conductivity, we performed additional characterization using the atomic force microscopy (AFM) and conductive AFM (cAFM), as shown in Figs. 2(d)–2(f). The topography and current images are 5 × 5 µm². The conductive AFM image was measured with a sample bias of +6 V. The line profile for the topography height and conductivity in Fig. 2(f) shows that the grain boundary area (valley in the height profile) has a higher resistivity while the grain (peak of the height profile) is more conductive, in line with the observation in Sb₂Se₃ but very different from other cubic photovoltaic materials such as CdTe.¹⁴ This is possibly due to the (Sb₄S₆)_n ribbons grown perpendicular to the substrate, where their vdW boundaries without dangling bonds suppress carrier recombination at the grain boundaries.

The optical absorption properties were characterized using UV-Vis spectroscopy, as shown in Fig. 3(a), where the strong absorption edge starting around 750 nm can also be observed in the transmittance plot [Fig. 3(b)]. The direct and indirect bandgaps can be determined as 1.71 and 1.56 eV by extracting the Tauc fitting, as shown in Fig. 3(c) and the inset. The experimental direct bandgap is in good agreement with the theoretical calculation, as shown in Table S2. Figure 3(d) and the inset are the calculated absorption coefficients of Sb₂S₃ compared with Sb₂Se₃. It is shown that both materials exhibit excellent light absorption, while Sb₂S₃ shows slightly better absorption than that of Sb₂Se₃ in a higher energy range. Overall, the optical absorption calculation demonstrates that Sb₂S₃ is a promising absorber material for non-cubic chalcogenide solar cells.

To elucidate the surface state of the CSS grown Sb₂S₃ film, the X-ray photoelectron spectroscopy (XPS) measurement was performed, and the results are shown in Fig. 4. The binding energies (BE) of all the peaks were corrected with C1s energy at 284.6 eV. The full spectrum survey is given in Fig. 4(a), showing the presence of Sb 3d and S 2p peaks. The high-resolution XPS spectra of Sb 3d are displayed in Fig. 4(b) with peaks at 539.9 eV for Sb 3d_3/2, and 532 eV for Sb 3d_5/2, respectively. These two peaks correspond to Sb³⁺ oxidation states in Sb₂S₃. In the high-resolution S2p spectrum, four peaks were observed at 161.5, 163.1, 164.0, and 165.2 eV, respectively. These peaks correspond to the S 2p_3/2 and S 2p_1/2, as shown in Fig. 4(c).

The Sb₂S₃ solar cell was fabricated in an FTO/CdS/Sb₂S₃/graphite cell structure, as shown in Fig. 5(a), and their device performance was measured accordingly. Figures 5(b) and 5(c) show the energy level alignment and the current-voltage (J-V) curve, and the external quantum efficiency (EQE) of the device, respectively. The typical device performance is listed in Table I. The best device performance is 3.83% with the open circuit voltage of 0.66 V, short circuit current density (J_sc) of ∼13 mA cm⁻², and Fill Factor (FF) of ∼44.65%. Therefore, the CSS-grown Sb₂S₃ film device performance is on par with the thermal evaporation deposited Sb₂S₃ solar cells, though still lower than that of the dye-sensitized solar cells.^{12,13,15,21,38}

To understand the defect chemistry inside the CSS deposited Sb₂S₃, we perform first-principles defect formation energy calculations. Here, we consider five types of intrinsic defects, including Sb vacancy (V_Sb), S vacancy (V_S), S interstitial (S_i), and Sb and S antisites (Sb_S and S_Sb), and the calculation details can be found in the supplementary material. Figures 6(a) and 6(b) show the formation energy of each defect type under S-rich and S-poor condition, respectively, and Fig. 6(c) shows the corresponding transition levels of charged defects. As shown in Fig. 6(c), V_Sb, S_i, and S_Sb are p-type defects (labeled in red); V_S and Sb_S are n-type defects (labeled in blue). Only V_Sb and S_Sb show relatively shallow acceptor levels (0/-1), i.e., transition levels located near valence band maximum (VBM). The ionization energy of these two levels is located at ∼0.07 eV and ∼0.28 eV, respectively. However, the formation energy of Sb vacancy is relatively high, which is above 2 eV in S rich condition, while S_Sb has a formation energy of ∼1 eV, much lower than V_Sb. Thus, S_Sb is potentially the dominant defect type in Sb₂S₃ that contributes to the p-type conduction. As for the n-type defects, V_S exhibits a rather deep donor level (2/0) far below the conduction band minimum (CBM), indicating that it is difficult to ionize, thus rarely contributing to n-type conductivity. In addition, Sb_S shows the n-type transition level (∼0.2 eV); nevertheless, it is unlikely to form in the S-rich condition due to its high formation energy above 2 eV. Overall, Sb₂S₃ exhibits p-type conductivity and S_Sb is the dominant native acceptor defect species. The loss of S will lead to the low device performance by diminishing Sb₂S₃ intrinsic p-typeness because V_S and Sb_S have lower formation energy, while V_Sb has higher formation energy in S-poor condition, compared to the S-rich condition. The calculation results are in agreement with the previous report.⁴² These observations are similar to the defect study of the Sb₂Se₃ system.⁴³ In summary, the intrinsic p-type conductivity of Sb₂S₃ originates from S_Sb, and the S chemical potential significantly impacts the conduction. It is expected that the higher carrier concentration can be achieved in the S-rich condition, although our calculation results suggest that the S vacancies are still likely to form even in the S-rich case.

In summary, we demonstrate the Sb₂S₃ thin film solar cells with 3.8% PCE successfully fabricated using the scalable close space sublimation deposition technique. The microstructure and optical properties of the CSS grown Sb₂S₃ have been characterized experimentally and studied using theoretical approaches. The detailed electronic structure and defect calculations provide insights into the conduction mechanism inside the Sb₂S₃ films, which suggests that the device can be further improved by controlling the S environment during film deposition or through post-deposition sulfurization. Our results show that Sb₂S₃ is suitable for large-scale manufacturing at low cost using CSS-like fast vapor transport technology and it is feasible to integrate Sb₂S₃ into the conventional manufacturing process for thin film solar cells.

See supplementary material for the lattice parameters and the calculation details of defect formation energy diagram and transition levels of charged defects.

L.P.G. and F.Y. acknowledge the support from the University of Alabama Research Grant Committee fund and the ORAU junior faculty Enhancement Award. B.Z. and X.Q. acknowledge the startup fund from Texas A&M University and the support from Texas A&M Energy Institute.