Discovery of two new Cu–Sn chalco–halides for potential solar absorber applications

Abstract

We explore multiple-cation chalco–halide phase fields evaluated by their synthetic accessibility using machine learning models. Exploratory synthesis guided by computational tools leads to the discovery of two new compounds; CuSn₂SI₃ and Cu_0.35Sn_5.29S₂I₇, their structures, and electronic and optical properties are reported herein. This is the first report of a stable quaternary compound in the Cu–Sn–S–I phase field. The two new compounds show related crystal structures where Sn₄S₂I₄ layers are a common structural motif in both. These Sn₄S₂I₄ layers are connected by Cu₂I₂ layers and disordered Cu–Sn–I layers, forming the three-dimensional structures of CuSn₂SI₃ and Cu_0.35Sn_5.29S₂I₇ respectively. Electronic band structure calculations using density functional theory show the presence of a direct band gap in CuSn₂SI₃ and suggest anisotropic transport, in line with the layered structure of the compound. A mixture of the two compounds with ∼86% CuSn₂SI₃, shows a band gap in the visible region, close to 2.1 eV and a significant photo-induced charge carrier mobility of ∼1.3 cm² V⁻¹ s⁻¹. This demonstrates Cu–Sn chalco–halides can form a promising phase space to explore for solar absorber materials, with further design and tuning of band gap.

---

Introduction

The need for efficiently and sustainably harvested solar energy to meet the rising energy requirements of humankind is paramount. In recent years, lead halide perovskites and related materials have attracted the attention of multidisciplinary researchers, and have emerged among leading candidate materials for future energy applications,^1–4 demonstrating a dramatic increase in power conversion efficiencies over the last decade.^5,6 While their low cost of production from simple solution-based methods⁷ and the excellent device performance make them attractive for commercial applications, the presence of the toxic element lead (Pb) and the poor long term stability of these materials, which often incorporate organic cations, have been limiting factors in their development.^8–10 Replacing Pb²⁺ with isoelectronic Sn²⁺ or Ge²⁺ in the halide perovskites, while reducing toxicity, compromises stability even further.⁸ However, the presence of the heavy s² cation has been found to be critical in ensuring high performance,^11,12 by giving rise to electronic energy levels that are tolerant to high defect concentration. In addition to this, effective application of lead halide materials in devices, especially as multijunction solar cells, requires multiple materials with strong absorption onsets at different energies within the solar spectrum.

Multiple anion compounds, especially those that contain chalcogenide and halide anions (chalco–halides) have emerged as promising materials for various applications, due to the additional flexibility in tuning the chemical structure, dimensionality and material properties that the different anions provide.^13,14 In the context of solar absorbers, these materials are expected to be more stable than halides because of the stronger covalent bonds formed by the chalcogenide anions, thus providing a way to improve the stability of halide-based materials while retaining their excellent absorption and optical properties.¹⁵ However, most of the chalco–halides studied for solar absorber applications are lower dimensional ternary compounds¹⁶ (e.g. SbSI,¹⁷ BiSeI,¹⁸ etc.). Such a reduction in dimensionality in two-dimensional (2D) or one-dimensional (1D) structures tends to give rise to less dispersive electronic bands, leading to lower charge carrier mobilities. Conversely, such flat bands could also lead to high absorption coefficients, arising from the high density of states near the band gap, or formation of tightly bound excitons due to quantum confinement.^19–21 Exploration of the chalco–halide chemical space could offer additional degrees of freedom to control the electronic and structural dimensionality to maximise both absorption as well as charge carrier mobilities.

Several copper containing compounds, especially chalcogenides, such as Cu₂ZnSnS₄ (CZTS) and CuIn_xGa_1−xSe₂ (CIGS) are established solar absorber materials. One of the reasons these materials are promising for optoelectronic applications is due to the ability of low valence, d¹⁰ cations such as Cu⁺, Ag⁺, etc. to promote defect tolerance by formation of shallow defects.^11,22–25 Chalco–halide materials containing multiple cations, particularly d¹⁰ cations are highly underexplored for solar absorber applications,¹⁵ with CuBiSCl₂ being a notable exception, and the combination of multiple anions could enable access to more diverse structure types and offer greater control of material properties. Other known quaternary chalco–halides with a d¹⁰ and an s² cation often contain the Bi³⁺ cation. Only a handful of other s² cation containing quaternary compounds are known including CdSnSX₂ (X = Cl, Br),²⁶ Hg₂PbS₂I₂,²⁷ Cu₅SbS₃I₂,²⁸ CdSbS₂X (X = Cl, Br),²⁹ etc. Exploration of such space thus presents an opportunity to identify new and potentially functional materials.

In this article we report on the synthetic exploration of Cu containing chalco–halide phase fields, guided by computational tools and the discovery of two new quaternary compounds. We investigate the structure, optical properties, and electronic band structure of CuSn₂SI₃ to evaluate its performance as a candidate solar absorber material.

Experimental and computational methods

Determining phase fields

An unsupervised machine learning algorithm, based on a variational autoencoder (VAE) was used to assess the synthetic accessibility of the phase fields, similarly to the approaches developed in ref. 30 and 31. We first select compounds for which a finite (non-zero) band gap is reported in the materials platform for data science (MPDS) database³² resulting in 68 823 entries. The phases were then aggregated into phase fields, i.e., Ag₃AlF₆ and AgAlF₄ represent the same phase field Ag–Al–F, which resulted in 40 785 unique phase fields as the training data, which introduces a bias towards elemental combinations exhibiting a finite band gap (Fig. S1).

Two VAE models were built based on different schemes for element descriptions – Mat2Vec³³ and Local structure-induced Atomic Features (LEAFs)³⁰ – reflecting knowledge of their existence and characteristic structural motifs of local atomic environment, respectively. During training, the VAE reduces the dimensionality of the 200- and 37-dimensional elemental descriptors, for Mat2Vec and LEAFs respectively, aligning similar values in the latent space, and through such encoding and decoding learns the patterns of elemental combinations characteristic of the synthetic accessibility.

The candidate phase fields consisted of all possible combinations of the elements Cu-M-Ch-X-X′with M being a cation with s² electronic configuration among Sn, Bi, Pb, and Sb, and Ch stands for the chalcogen anion S or Se, X are halide anions among F, Cl, Br, and I, and X′ was from {Ch, X}, which are then subjected to the trained VAEs for ranking: the lower the score the more similar is the combination to the training data in the latent space of the model, and we hypothesise it is more likely to be synthetically accessible, i.e. more likely to contain synthetically stable compounds.

Points distribution algorithm

The points distribution algorithm attempts to evenly space out suggested compositions whilst avoiding known phases in a phase field.³⁴ The phase field is first discretised using a 2D orthogonal grid of approximately 500 points spanning all charge balanced compositions. A Markov Chain Monte Carlo (MCMC) method is then used to identify a suitable solution consisting of a small subset (five) of these points. At the heart of the MCMC method is the function used as a proxy to the energy which is used to evaluate how good a solution is. The proxy energy function attempts to jointly maximise all pairwise distances between suggested points as well as all pairwise distances between suggested points and known phases. The MCMC method begins with a random initial distribution of points. This distribution is then perturbed by randomly updating a selection of the points to new positions. The new distribution is then accepted with a probability based on the improvement in the proxy energy function. The process is then repeated 50 000 times, and the best distribution obtained during the entire run is used as the initial sampling suggestion. For this solution, the discrete grid is constrained to exclude any compositions with less than 5% (molar) of either copper or tin.

Exploratory synthesis

The initial compositions as suggested by the points distribution algorithm, equally spaced out from each other and known compounds were: C0–Cu_2.44Sn_1.13S_2.27I_0.16, C1–Cu_1.68Sn_1.74S_2.58, C2–Cu_0.63Sn_2.31S_2.19I_0.87, C3–Cu_1.98Sn_0.95S_0.81I_2.26, and C4–Cu_1.81Sn_1.29S_1.49I_1.41. The precursors used were copper(I) iodide (CuI, 99.998% metals basis, Puratronic), copper(I) sulfide (Cu₂S, 99.5% metals basis, Thermofisher Scientific), tin(II) iodide (SnI₂, ultra dry, 99.999% metals basis, Thermofisher Scientific), and tin(II) sulfide (SnS, 99.5%, Thermofisher Scientific). The precursors were weighed out in the required stoichiometry of the chosen initial compositions and ground well to mix using a mortar and pestle, inside an argon-filled glovebox. The ground mixtures were then sealed in evacuated quartz tubes and heated to a temperature of 400 °C (T₁), at a rate of 5 °C per minute. After dwelling at 400 °C for 12 hours, the reactions were slowly cooled down to room temperature at a rate of 6.3 °C h⁻¹. A similar set of reactions were carried out with the same initial compositions, and protocol but with the temperature set to 600 °C (T₂) at a rate of 5 °C per minute. After dwelling at 600 °C for 12 hours, the reactions were slowly cooled down to room temperature at a rate of 6.3 °C h⁻¹. Finally, a third set of reactions were run at 700 °C (T₃). The lowest temperature of 400 °C was chosen such that it is above the melting point of SnI₂ (320 °C) to aid reaction of the precursors. Subsequent higher temperatures were chosen to facilitate melting of the rest of the precursors with higher melting temperatures while remaining below the boiling point of SnI₂ (714 °C), to avoid issues with possible over pressurization of the ampoule and loss of iodides. The reactions were labelled using the initial composition and the temperature used, e.g., C1T3 for initial composition C1–Cu_1.68Sn_1.74S_2.58 run at the temperature T₃–700 °C.

Powder X-ray diffraction (XRD) was carried out on all fifteen samples, and single crystals were harvested from the reactions that had highest intensity of unindexed peaks (C3T2 and C4T3). Single crystal X-ray diffraction studies on such crystals, revealed two new structures – CuSn₂SI₃ and Cu_0.35Sn_5.29S₂I₇. Once the two new structures were solved, the powder X-ray diffraction patterns of all the exploratory reactions could be satisfactorily fitted with a mixture of precursors, known ternaries and binaries, and the two new structures, indicating no other new compounds have been formed with the compositions and reaction conditions we used. The reaction products of exploratory syntheses were mixtures of various known and newly found phases, not single phase pure compounds. This is expected as the initial compositions were predicted for exploration and not as target compositions, and conditions were not optimised. With the compositions obtained from single crystal XRD, we attempted to synthesise CuSn₂SI₃ and Cu_0.35Sn_5.29S₂I₇ as high purity powders via solid state synthesis.

Solid state synthesis of CuSn₂SI₃

The precursors CuI, SnI₂, and SnS were weighed out in the molar ratio 1 : 1 : 1, and ground well to form a homogenous mixture using a mortar and pestle, in an argon filled glovebox. The powder was then pressed into a 10 mm pellet using a stainless-steel die and a uniaxial press with a weight of 2 tons for about 1 minute. The pellet was then sealed in an evacuated quartz tube and heated up to 250 °C with a rate of 60 °C h⁻¹. After dwelling at 250 °C for 24 hours, it was slowly cooled down to room temperature at a rate of 5 °C h⁻¹. This resulted in a mixture of CuSn₂SI₃, Cu_0.35Sn_5.29S₂I₇, and unreacted SnI₂. Hence the process was repeated with intermediate grinding and pelletizing, as well as with increasing dwell time at 250 °C to one week. After three repeated reactions the products formed were ∼85% (by weight) CuSn₂SI₃ as estimated from powder XRD. Further firing of the reaction did not make any discernible difference in the purity. Attempts to make Cu_0.35Sn_5.29S₂I₇ by similar methods always resulted in a mixture of phases-Cu_0.35Sn_5.29S₂I₇, CuSn₂SI₃, and unreacted precursors, with samples containing <70% Cu_0.35Sn_5.29S₂I₇. Hence, we did not carry out any optical property measurements on Cu_0.35Sn_5.29S₂I₇.

Powder samples prepared through solid state synthesis were ball milled in a planetary ball mill with a zirconia pot and 5 mm diameter balls (1 : 10 weight for sample to balls), at 300 rpm for 20 minutes, to obtain particle sizes of several micron. Powder XRD and energy dispersive X-ray spectroscopy (EDX) confirmed non-degradation of sample in the process. The milled powder was used for making cold pressed pellets and used for terahertz spectroscopy measurements.

Single crystal X-ray diffraction

X-ray diffraction measurements on orange-red, needle shaped, single crystals picked from the exploratory synthesis attempts C3T2 and C4T3 were carried out in-house using a Rigaku MicroMax-007 HF X-ray diffractometer equipped with a Mo K_α rotating-anode microfocus source and a Saturn 724 + detector. The measurements were performed with the sample at 300 K for CuSn₂SI₃ and at 100 K for Cu_0.35Sn_5.29S₂I₇. Cell refinement and data reduction were performed using the CrysAlis^Pro software. The structure was solved and refined using SHELX,³⁵ implemented through Olex2.³⁶ The crystal structures were visualised using the software VESTA³⁷ for creating the images.

Powder X-ray diffraction

Preliminary phase identification of powder samples was done using powder X-ray diffraction using an in-house Rigaku Smartlab diffractometer with a Mo K_α source (λ = 0.7107 Å) in the Debye–Scherrer geometry. Powder samples were sealed in 0.3 mm diameter borosilicate capillaries inside an Ar-filled glovebox for measurements. Synchrotron powder X-ray diffraction measurements were carried out at the beamline I11 at the Diamond Light source, Didcot,³⁸ UK. Rietveld refinement on powder XRD data was carried out using TOPAS Academic V5 software.

UV-visible spectroscopy

UV-visible spectra on powder samples were measured using an Agilent Cary 5000 spectrometer in the diffuse reflectance mode. The reflectance data were transformed using Kubelka–Munk equations to obtain absorption data.

Scanning electron microscopy (SEM)

Imaging and EDX measurements were done on a Tescan S8000 microscope using a X-Max^N detector from Oxford instruments. Powder and crystals were mounted on an aluminium stub using a carbon tape. The sample was coated using carbon to avoid charging. Analysis of EDX spectra was done using Aztec software.

X-ray photoemission spectroscopy (XPS)

X-ray photoemission spectra on powder samples were measured at HarwellXPS, Didcot, UK. Measurements were carried out using a Kratos Axis SUPRA XPS fitted with a monochromatic Al K_α X-ray source (hν = 1486.6 eV), a spherical sector analyzer, and 3 multichannel resistive plate, 128 channel delay line detectors. All data were recorded at 150 W and a spot size of 700 × 300 µm². Survey scans were recorded at a pass energy of 160 eV, and high-resolution scans were recorded at a pass energy of 20 eV. Electronic charge neutralization was achieved using a magnetic immersion lens. All sample data were recorded at a pressure below 10⁻⁸ Torr and temperature of 294 K. Data were analysed using CasaXPS v2.3.26PR1.0. The background was fitted using a Shirley background and peaks fitted using a Gaussian–Lorentzian (GL(50)) peak shape. The spectra were calibrated with respect to the Fermi edge of a Ni foil reference.

First principles studies

First principles density functional theory (DFT)^39,40 calculations, as implemented in the plane-wave basis code Quantum Espresso,^41,42 were carried out for the primitive cell of CuSn₂SI₃. The PBE exchange correlation functional⁴³ was used with norm-conserving fully relativistic Vanderbilt pseudopotentials⁴⁴ as available in the PseudoDojo database.⁴⁵ We used kinetic-energy and charge-density cutoffs of 100 Ry and 400 Ry, respectively. We performed unconstrained variable-cell relaxations on the experimentally measured structure using a uniform Γ-centred k-point grid of 8 × 8 × 3 after which the unit cell volume expanded by 11% (see Table. S10), without significant changes to the relative atomic positions as compared to the unrelaxed structure. We note that structural optimisation with PBEsol yields a unit cell volume in better agreement with the experimental structure than that obtained using standard PBE (2% vs. 11%, respectively). However, a comparison of computed band structures for both the experimental and PBE relaxed structures (see SI Fig. S12), shows relatively minimal differences in the conduction and valence band edge states. Subsequent calculations presented in the main manuscript are performed on the PBE relaxed structure. We compute the density of states using a uniform 16 × 16 × 6 k-point grid (772 irreducible k-points), including spin–orbit coupling (SOC). The effective mass tensor is calculated as the inverse of the second derivative of the energy with respect to the wave-vector, with derivatives being calculated using the finite difference method with a 5-point stencil and an optimised increment of 0.01 Bohr⁻¹. The effective mass tensor is computed and diagonalised using the effective mass calculator (emc) code.^46,47 We also included spin–orbit coupling for the relaxed structure using a q-point grid and k-point grid of 4 × 4 × 2.

Charge-carrier mobility measurements

The effective electron–hole sum mobility at THz frequencies was measured using the optical-pump terahertz-probe (OPTP) method for a pressed pellet of CuSn₂SI₃ of known thickness (See SI for full details of the experimental technique and data analysis). In the OPTP method, charge carriers are photoexcited in a semiconductor with ultrafast laser pulses (wavelength λ = 400 nm), and the corresponding change in conductivity is measured through changes in the transmission of a single-cycle terahertz pulse. A value for the electron–hole sum mobility may be extracted from such data, provided the initial charge-carrier excitation density is known.^48–50 The pellet of CuSn₂SI₃ was pressed at room temperature using the pre-synthesized powder described above to achieve uniform densification. The resulting pellet measured 0.49 mm in thickness, 5 mm in diameter, and had a mass of 51.9 mg, corresponding to a calculated density of approximately 5.40 g cm⁻³.

Results and discussion

Computationally guided phase field determination and exploration

We employ a computationally guided approach to identify the previously unexplored phase field of Cu–Sn–S–I as promising and explore it experimentally resulting in the discovery of two new compounds. This was attained through a two-step process.

1. Identifying Cu–Sn–S–I phase field. Using chemical knowledge and promising characteristics for solar absorbers, we first select a subset of element combinations denoted as Cu-M-Ch-X′-X, with M being a cation with s² electronic configuration among Sn, Bi, Pb, and Sb, and Ch stands for the chalcogen anion S or Se, X are halides anions among F, Cl, Br, and I, and X′ was from {Ch, X}. This way we incorporated an s² and a d¹⁰ cation, known for imparting defect tolerance, while two cations provide greater extent of tuneability. The choice of chalco–halide compounds was inspired by the fact that many ternary lead and tin halides have shown excellent optical absorption and charge carrier mobilities, but poor stability in ambient atmosphere/moisture. The addition of chalcogen anions to these halides are known to improve the stability. With this element selection, there still is a large number of choices to explore, and we used an unsupervised machine learning algorithm, based on a variational autoencoder (VAE) to assess the synthetic accessibility of the phase fields, similarly to the approaches developed in ref. 30 and 31. This algorithm ranks the candidate fields with its likelihood to have a stable compound. The candidate phase fields, Cu-M-Ch-X-X′, were subjected to the trained VAEs for ranking. A Pareto front optimising the results of the two VAE models we used (Mat2Vec and LEAFs, see methods for details) is shown in Fig. 1 for Ch = S.

Fig. 1 Phase field and initial composition selection for exploratory synthesis. The VAE models of synthetic accessibility are trained for phase fields for which a band gap value greater than zero is reported in MPDS. Two different models are trained based on Mat2Vec and LEAFs descriptions of constituent elements. At the prediction stage, the Pareto front of the scores from the two models for the candidate combinations of elements is presented, where a lower score denotes higher synthetic accessibility. For the selected highest ranked (lowest score) phase field, a quaternary cut of the Cu–Sn–S–I phase field is explored with the precursors used for the synthesis at the corners. Known ternaries are shown by red points along the boundaries. A points distribution algorithm suggested the five points shown in green as possible starting points, away from known compounds, for exploratory synthesis.

Among the ranked test fields according to both models, the top five combinations contained Sn²⁺. Chloride containing combinations were also consistently found higher ranked than other halides. We selected a quaternary phase field derived from the top ranked quinary candidate Cu–Sn–S–I–Cl phase field (circled in Fig. 1) in order to begin experimental investigation. The four-element field Cu–Sn–S–I does not have any reported quaternary compounds and is among the less explored fields with a low number of known ternaries and thus was chosen for exploratory synthesis attempts. The known ternary compounds in this phase field are listed in Table S1.

2. Exploratory synthesis to identify new compounds. Once the phase field was chosen, we need to explore it in a systematic manner, by attempting synthesis with varying initial composition and reaction conditions. A points distribution algorithm was used to suggest five plausible initial compositions for experimental synthesis within the quaternary cut of the phase space as restricted by the choice of precursors – Cu₂S, CuI, SnI₂, and SnS. The algorithm ensures that the separation between these five compositions and from the precursors is maximised whilst also avoiding known phases. These five compositions, C0–C4, were weighed out and reacted at different temperatures (see Methods). After this, all products were analysed with powder X-ray diffraction. Samples which had unidentified peaks were then subject to single crystal XRD, where numerous single crystals from each batch were screened to find a unit cell not reported in ICSD. Two new structures were discovered this way, with the compositions of CuSn₂SI₃ and Cu_0.35Sn_5.29S₂I_7.

Structure determination and description

The crystal structure of CuSn₂SI₃ (Fig. 2) was solved from single crystal X-ray diffraction data with C2/m (#12) symmetry, and unit cell parameters (obtained from powder XRD) of a = 14.0369(1) Å, b = 4.30723(5) Å, c = 14.4467(1) Å, α = 90° β = 97.007(5)°, and γ = 90° (Tables S11–S17). The structure is formed of three different cation environments (Fig. 2(b)): a slightly distorted tetrahedral copper environment and two different capped trigonal prismatic coordination environments with tin at the centre. The Sn1 site occupies an SnS₃I₄ environment in which the Sn²⁺ cation is displaced from the centre yielding two shorter (3.4546 Å) and two longer (3.7579 Å) Sn–I distances. Sn2 occupies an SnSI₆ environment in which the S²⁻ anion caps a single face with a range of Sn–I distances of 3.1285–3.7563 Å, again reflecting the off-centre Sn²⁺ cation position.

Fig. 2 (a) The three-dimensional crystal structure of CuSn₂SI₃. (b) Polyhedral environments for each cation; one CuI₄ tetrahedral site, and the SnS₃I₄ and SnSI₆ capped trigonal prismatic environments. (c) CuI₄ tetrahedra share edges along b to form Cu₂I₂ layers. (d) The Cu₂I₂ layers (purple shading) and Sn₄S₂I₄ layers (green shading) are stacked along the c axis to form the three-dimensional structure of CuSn₂SI₃. The red dashed circles highlight longer Sn–I distances (3.7528(8) Å) which provide connectivity along the a axis within the Sn₄S₂I₄ layer. (e) The Sn polyhedra face share to form Sn₄S₂I₄ layers.

These environments combine in CuSn₂SI₃ to produce Sn₄S₂I₄ layers and Cu₂I₂-like layers which stack alternately along the c axis to give the three-dimensional structure (Fig. 2(d)). Fig. 2(c–e) show the top-down view of both layers. The Sn₄S₂I₄ units, formed via face-sharing of the SnS₃I₄ and SnSI₆ environments (Fig. 2(e)), are connected along the a axis of CuSn₂SI₃ by Sn–I distances of 3.7528(8) Å (red dotted circle in Fig. 2(d)), slightly longer than the typical ∼3.6 Å Sn–I distances in SnI₂,⁵¹ or Sn₂SI₂.⁵² These distances contribute sufficient valence in bond valence sum^53,54 (BVS) calculations, increasing the valence for both the Sn1 and Sn2 sites closer to the formal +2 oxidation state expected for Sn (Table S2–S4). These Sn₄S₂I₄ structural units are the same as found in Sn₂SI₂.⁵² and Sn₄SI₆,⁵⁵ and similar motifs are observed in other chalco–halide compounds such as Pb₅S₂I₆.⁵⁶ The presence of stereo chemically active lone pair cations can favour the formation of lower dimensional or layered structures, as observed in other s² compounds.²⁹

The crystal structure of Cu_0.35Sn_5.29S₂I₇ (Fig. 3) was, like CuSn₂SI₃, solved from single crystal XRD in the monoclinic space group C2/m with unit cell parameters (at 100 K) of a = 14.1572(8) Å, b = 4.3847(2) Å, c = 17.1829(11) Å, α = 90° β = 110.303(7)°, and γ = 90° (Table S18–S20). The structure consists of the same Sn₄S₂I₄ infinite layers as seen in CuSn₂SI₃ (Fig. 3(f and g)), thus the in-layer a and b unit cell parameters are comparable between the two materials. These Sn₄S₂I₄ layers stack alternately along the c axis with thicker Cu_0.35Sn_1.29I₃ layers (purple layer in Fig. 3(f)) to give the material composition Cu_0.35Sn_5.29S₂I₇, which accounts for the increased c lattice parameter. This thicker Cu_0.35Sn_1.29I₃ layer is disordered (Fig. 3 (c–e)), with partial occupancy of two slightly distorted tetrahedral copper environments (Cu1 occupancy 0.096(4); Cu2 occupancy 0.078(4)) and one octahedral SnI₆ environment (occupancy of 0.644(2)) (Fig. 3(b)) in which the Sn²⁺ cation is displaced from the centre yielding one shorter (3.0567 Å), four similar (∼3.099 Å), and one longer (3.2032 Å) Sn–I distances. It is the occupancy of the Sn octahedral site that prevents the complete filling of the neighbouring face-shared Cu tetrahedral sites as seen in ordered CuSn₂SI₃, driving the disorder within this layer. The Sn2 site in Cu_0.35Sn_5.29S₂I₇ occupies a similar SnSI₆ environment as in CuSn₂SI₃, however, the Sn1 site occupies a bicapped trigonal prismatic SnS₃I₅ environment involving one extra I⁻ compared to CuSn₂SI₃ because of a smaller S–Sn–I angle (Fig. 3(g)).

Fig. 3 (a) The three-dimensional crystal structure of Cu_0.35Sn_5.29S₂I₇. (b) Polyhedral environments for each cation; two CuI₄ tetrahedra, a SnI₆ octahedron, a SnSI₆ capped trigonal prism, and a SnS₃I₅ bicapped trigonal prism. (c–e) show the arrangement of the tetrahedral Cu and octahedral Sn sites to form the Cu_0.35Sn_1.29I₃ layer. (f) The layered structure of Cu_0.35Sn_5.2S₂I₇, comprising Cu_0.35Sn_1.29I₃ layers (purple shading) and Sn₄S₂I₄ layers (green shading). (g) The Sn₄S₂I₄ layers are similar to those in ordered CuSn₂SI₃ (Fig. 2) with reduced S–Sn–I angles in the Sn₄S₂I₄ unit which changes the environment of Sn1 to a bicapped trigonal prism.

Similar modifications of the M₄S₂X₄ structural motif are observed in Pb₂BiS₂I₃ and Sn₂BiS₂I₃,⁵⁷ while disordered and mobile Cu⁺ cations are also a common feature among many Cu-based chalco–halides such as Cu₃Bi₃S₅I₂,⁵⁸ Cu₃Bi₂S₃I₃,⁵⁹ and Cu₄BiSeI₄.⁶⁰

The structures of CuSn₂SI₃ and Cu_0.35Sn_5.29S₂I₇, with C2/m symmetry and comparable a and b lattice parameters, consist of motifs which are common to both materials, such as the Sn₄S₂I₄ infinite layers. This can be understood via comparison of their anion arrangements (Fig. 4) which are related between the two materials. The anion nets of both CuSn₂SI₃ and Cu_0.35Sn_5.29S₂I₇ combine alternating regions of well-known hexagonal and square nets ({3⁶} and {4⁴} in Schläfli notation, respectively), separated by five-coordinate S²⁻ anions ({3³.4²}) as shown in Fig. 4(a) and (b). There is one additional row of I⁻ anions in Cu_0.35Sn_5.29S₂I₇ compared to CuSn₂SI₃ (Fig. 4(b)), consistent with the increased I⁻ to S²⁻ ratio in Cu_0.35Sn_5.29S₂I₇. These nets stack in similar ways in both structures such that the extended layered motifs become visible within the ab plane and alternate along the c axes of both structures (Fig. 4(c) and (d)). The additional row of I⁻ anions in Cu_0.35Sn_5.29S₂I₇ (Fig. 4(b)) expands the c lattice parameter compared to CuSn₂SI₃. The environments generated by the S₂I₄ layers are occupied exclusively by Sn²⁺ in both materials to yield the Sn₄S₂I₄ infinite layers in the ab plane (green shading in Fig. 4(e) and (f)). Cu⁺ occupies environments within the I⁻ only layers in both materials (purple shading in Fig. 4(e) and (f)).

Fig. 4 Comparison of the anion nets in (a) CuSn₂SI₃ and (b) Cu_0.35Sn_5.29S₂I₇ with anion rows labelled based on Schäfli notation. The stacking of these nets in (c) and (d) yields layered motifs in the ab plane of S₂I₄ and I⁻ anions which alternate along the c axis. The orange line shown in both (c) and (d) represents the single-layer anion nets shown in (a) and (b). The structures of (e) CuSn₂SI₃ and (f) Cu_0.35Sn_5.29S₂I₇ shown along the same projection as in (c) and (d) formed from Sn₄S₂I₄ (shaded green) and Cu₂I₂-like (shaded purple) layers.

The Sn₄S₂I₄ unit which exists in both CuSn₂SI₃ and Cu_0.35Sn_5.29S₂I₇ is the building block of the ternary Sn₂SI₂ itself and exists in other ternaries such as Sn₄SI₆, and also analogous M₄S₂X₄ units (where X = halogen, M = Pb²⁺, Bi³⁺, Sb³⁺) exist in Pb₅S₂I₆,⁵⁶ BiSI,⁶¹ SbSI,⁶² and CdBiS₂Cl.²⁹ The capped trigonal prismatic coordination of cations is common to many of these compounds showcasing the robust nature of these structural units. However, in the case of most other known structures, the addition of a different cation (i.e. M-M′-X-X′) leads to it being incorporated into the structural unit either expanding or disrupting the motif, as in cases of CdBiS₂Cl, Cu₃Bi₂S₃I₃,⁵⁹ In₂BiSe₄I,⁶³ and Hg₂PbS₂I₂ ²⁷ (Fig. S2 and Table S5). However, in the case of CuSn₂SI₃ and Cu_0.35Sn_5.29S₂I₇, Cu⁺ is not incorporated into the Sn₄S₂I₄ units and leads to the formation of separate structural layers. The two new compounds discovered here represent some of the few sulphide-iodide quaternary compounds with group 14 s² cation (Sn²⁺ or Pb²⁺). The absence of a large number of similar compounds could be attributed to the large size difference between the S²⁻ and I⁻ anions.

A CuSn₂SI₃ powder sample was prepared by solid state synthesis as described in the Methods. A Rietveld refinement against synchrotron powder X-ray diffraction data is shown in Fig. 5. It confirms the majority phase formed (with 86% weight percent) is the desired CuSn₂SI₃, with the primary impurity being Cu_0.35Sn_5.29S₂I₇. Repeated firings and different temperatures did not increase the purity of the samples. The structural and refinement parameters are given in the SI Tables S14 and S15. Powder XRD data measured at 100 K is also shown in the SI (Fig. S3).

Fig. 5 Synchrotron powder X-ray diffraction pattern of as synthesised CuSn₂SI₃ sample measured at room temperature. CuSn₂SI₃ constitutes 86% by weight of the sample, with the disordered Cu_0.35Sn_5.29S₂I₇ being the primary impurity and the binary iodides present in small amounts. Black circles show the measured data and the red line is the calculated pattern from a Rietveld refinement. The difference between the two are shown as a blue line and the Bragg peak positions of the phases present are plotted as well.

The elemental analysis of the crystals and powder sample were carried out using Energy Dispersive X-ray Spectroscopy in a scanning electron microscope. The resulting atomic percentages of Cu, Sn, S, and I are plotted in ternary diagrams in Fig. S4 in SI, along with an image of the crystal in Fig. S5. The measured compositions form a cluster of points close to the target composition. The oxidation states of the various elements present in the CuSn₂SI₃ sample were confirmed using the core level XPS measurements, and are shown in SI Fig. S6–S10. The binding energies obtained from fitting the spectra were compared with the values from the literature to confirm the oxidation states of the different species present in the compound (Tables S6–S9). The values obtained agree with the assignment of the oxidation states Cu⁺, Sn²⁺, S²⁻, and I⁻ in CuSn₂SI₃.

Optical property measurements

To gain an understanding of the optical properties and band gap of the material, UV-visible spectroscopy was carried out on the powder sample of CuSn₂SI₃. The measured diffuse reflectance data was converted using the Kubelka–Munk function^64,65 and plotted in Fig. 6. The absorption falls sharply above ∼500 nm, indicating an absorption onset around that energy range, consistent with the orange-red colour of the sample. To estimate the optical band gap, a Tauc analysis⁶⁶ was performed assuming a direct band gap, as shown in Fig. 6(b). The linear fit to the plot of (F(R_∞)hν)² versus hν gives an intercept to the x-axis, at 2.1 ± 0.2 eV (error defined as the standard deviation obtained from multiple measurements from the same batch of powder sample). However, the powder sample is not phase pure and the impurities present could be contributing to the absorption edge and hence the band gap values obtained this way. We eliminate the possibility that the absorption onset is coming from either CuI or SnI₄, as CuI has a much wider gap of ∼3 eV^67,68 and by measuring samples with added SnI₄ to examine the effect this has on the absorption spectra (Fig. S11). Although SnI₄ has a bandgap close to where we see the absorption edge, we conclude ∼2% of SnI₄ cannot give rise to the absorbance we observe from the powder sample of CuSn₂SI₃. This leaves the determined band gap to be from a mixture of CuSn₂SI₃ and Cu_0.35Sn_5.29S₂I₇. Although the measured optical absorption onset suggests a band gap of approximately 2.1 eV, which is too high for use in a single junction solar cell, one cannot rule out the potential for use in a multijunction cell, for example in conjunction with perovskites, where absorption occurs across the visible energy ranges and record high efficiencies are reported.⁶⁹ Also, it is to be noted that this is the first report in the Cu–Sn chalco–halide family of compounds. Further exploration and attempts to tune the band gap by examining analogues and by chemical substitution are possible avenues for future research.

Fig. 6 The pseudo-absorption spectrum for CuSn₂SI₃ measured using UV-Vis spectroscopy. (a) The measured diffuse reflectance data is plotted after transforming as Kubelka–Munk function F(R_∞) as a function of wavelength. (b) A Tauc plot for the absorption spectra assuming a direct band gap. Red dotted line shows the linear fit giving a band gap of E_g = 2.1 ± 0.2 eV.

Electronic band structure calculations

The electronic band structure of CuSn₂SI_3, shown in Fig. 7(a), is calculated from DFT (PBE) + SOC along a path Y₂ (−0.5, 0.5, 0) – Γ (0, 0, 0) – M₂ (−0.5, 0.5, 0.5) – Y₂ in reciprocal lattice units. See Fig. S12(a) and S12(b) for the full path. We find a direct band gap, with dispersive bands in the directions corresponding to the Cu–I or Sn–S–I planes of the structure, and non-dispersive otherwise (for example along Y₂–M₂). As shown in Table 1, we calculate a band gap of ∼1.0 eV within semi-local DFT (PBE) + SOC, and of ∼2.5 and ∼1.8 eV using the PBE0 ⁷⁰ and the HSE⁷¹ hybrid functionals, respectively. The differences between these calculated band gaps and the experimental band gap estimated from optical absorption experiments are consistent with previous benchmarking data reported in the literature.⁷²

Fig. 7 . (a) The electronic band structure calculated for the ordered phase of the relaxed structure of CuSn₂SI₃ using DFT(PBE) + SOC along the closed path Y₂–Γ–M₂–Y₂ in the Brillouin zone (See SI Fig. S12(a) and S12(b) for the band structure along the full path for the experimental as well as the relaxed structure, respectively, with the Brillouin zone shown in Fig. S12(c)). The orbital projected k-resolved DOS in the Brillouin zone and the partial DOS for each atom is shown in SI Fig. S13. 7(b) The charge density distribution corresponding to the valence band top (VBT) (red shaded regions) and the conduction band bottom (CBB) (green shared regions) at the point Y₂ in reciprocal space that also corresponds to the direct bandgap.

Table 1 Comparison of the experimentally obtained bandgap with predictions from DFT + SOC and DFT + SOC with Hybrid functionals PBE0 and HSE, for the volume relaxed structure of CuSn₂SI₃

Method	Bandgap (eV)
Experiment	2.1 ± 0.2
DFT	1.0
PBE0	2.45
HSE	1.79

---

To better understand the origin of flat bands between Y₂ and M₂, we computed the charge density distribution for the valence band top (VBT) and the conduction band bottom (CBB) at these specific points in the Brillouin zone (see Fig. 7(b) for Y₂). The red and green iso-surfaces show the charge density at the VBT and the CBB, respectively. The charge density corresponding to the VBT is clearly centred around the Cu₂I₂ layers. Along the crystallographic c-axis, each such layer is spatially well separated from the next one by the Sn₄S₂I₄ layer in between leading to minimal overlap between the different Cu₂I₂ layers. This leads to the formation of a flat band at the VBT along the corresponding direction in k-space which is Y₂–M₂. Similarly, the charge surfaces for the CBB are primarily centred around the Sn atoms in the Sn₄S₂I₄ layers and separated along the c-axis, leading to a flat band at the CBB as well. Further, the atom-projected band structure shown in SI Fig. S13 with corresponding atom and orbital resolved density of states (DOS) confirms that the contributions to the VBT arise primarily from Cu and I atomic orbitals, while the CBB consists mostly of Sn with very small contributions from S and I orbitals. The presence of flat bands along with dispersive bands in the Y₂–Γ–M₂ direction indicates a possibility of highly anisotropic electron and hole transport. To quantify this, we calculated and compared the effective masses of the holes and electrons at the high symmetry points Y₂ and M₂ along the three directions, as tabulated in Table 2.

Table 2 Effective masses at the valence band top and the conduction band bottom for the high symmetry k-points M₂ and Y₂ for CuSn₂SI₃ in units of electron mass (m_e) along the specified direction of eigenvectors in crystal units in the reciprocal space

High-symmetry point	m* (me) (Harmonic mean)	m1(me) (direction)	m2 (me) (direction)	m3 (me) (direction)
M2 (−0.5, 0.5, 0.5)	VBT (hole) 1.31	0.45 (1, −1, 0)	>>1 (1, 1, 0.09)<	>> 1 (0.04, 0.04, 1)
	CBB (electron) 0.45	0.21 (1, −1, 0)	0.56 (1, 1, −0.04)	>>1 (0.16, 0.16, 1)
Y2 (−0.5, 0.5, 0)	VBT (hole) 1.40	0.48 (−1, 1, 0)	>>1 (1, 1, −0.21)	>>1 (0.32, 0.32, 1)
	CBB (electron) 0.46	0.21(1, −1, 0)	0.59 (1, 1, 0.03)	>>1 (0.1, 0.1, 1)
Γ(0, 0, 0)	VBT (hole) 2.58	0.92 (1, −1, 0)	>>1 (1, 1, −0.07)	>> 1 (0.19, 0.19, 1)
	CBB (electron) 0.72	0.423 (1, −1, 0)	1.9 (1, 1, 0.91)	0.78 (−0.85, −0.85, 1)

---

The effective mass calculations at Y₂ and M₂ show clear anisotropy in three different directions. The direction along the crystallographic c-axis (m₃) has the heaviest effective masses for both holes and electrons. In the other two orthogonal directions the effective masses are much smaller at a fraction of a free electron mass indicating a possible high mobility of charge carriers. Also, the average effective mass of the holes is ∼3 times larger than the average effective mass of the electrons. It is to be noted that, in semiconductors with anisotropic and non-parabolic bands, the experimental effective mass often depends on the carrier concentration^73,74 which we are not considering in our calculations here. Despite highly anisotropic chemical and electronic structures many materials like Sb₂Se₃,⁷⁵ BiSI,⁷⁶ etc. have been successfully incorporated into solar cells with high efficiency with careful synthesis of oriented thin films. Thus, the anisotropic electronic bands do not rule out the potential of CuSn₂SI₃ for solar absorber applications.

An understanding of the contributions to the band edges coming from different elements is critical to further understand the photo-induced charge carrier transport and to predict ways to tune the band gap. We compare the DOS near the Fermi energy obtained from DFT calculations with an experimentally measured X-ray photoemission spectroscopy valence band spectrum for CuSn₂SI₃ in Fig. 8. The calculated density of states has been processed with a Gaussian smearing of 0.25 eV accounting for the lifetime broadening in the measured data and spectrometer resolution,^77,78 to aid comparison with the experimental data. The background was subtracted from the measured XPS data and scaled to match with the DOS.

Fig. 8 Comparison of the total density of states as obtained from DFT + SOC and the valence band spectrum obtained from X-ray photoemission spectroscopy measurements. The DFT data has an experimentally motivated Gaussian smearing of 0.25 eV.

The total DOS and the XPS valence band data agree reasonably well. As discussed before, the VBT contributions come primarily from Cu and some overlap with I, according to DFT calculations. The measured XPS spectrum shows a broad and asymmetric peak, centered around ∼2.5 eV. The calculations roughly capture this peak position, with the asymmetric peak centered around ∼2 eV with multiple shoulders. The smaller peak further away from the Fermi energy (∼7.5 eV) is also reproduced, although with a much higher intensity and with a slight shift. The disparities in intensity could be arising, in part, from the different photoionization cross sections for different atomic orbitals, which are not accounted for in DOS calculations. We also note that the relative intensity of the main peak and the peak at ∼1 eV in calculations (which is a barely visible shoulder in experimental data) is better reproduced with PBE0 and HSE DOS, as shown in SI Fig. S14. This is expected, as the semi-local DFT not only underestimates the band gap, but also band widths.⁷² Nevertheless, the general agreement between DFT and XPS data shows that the calculations have captured the band edge properties of the compound very well.

Charge carrier mobility

Terahertz (THz) photoconductivity measurements were performed to assess the mobility and dynamics of charge carriers in CuSn₂SI₃. Fig. 9 shows the change in photoconductivity in a cold-pressed pellet of CuSn₂SI₃ as a function of time after photoexcitation with 35 fs-laser pulses at a photon energy of 3.1 eV. From the value of the initial amplitude immediately after excitation, and under knowledge of the absorbed photon density (see SI for details) a value of the effective electron–hole sum mobility of φµ = 1.3 ± 0.2 cm² (V s⁻¹) was extracted. This value is comparable to THz mobilities measured for other newly emerging solar absorbers, such as Cu_4x(AgBi)_1−xI₄,⁷⁹ (AgI)_x(BiI₃)_y rudorffites⁸⁰ and AgBiS₂,⁸¹ but about an order of magnitude lower than those typical for tin iodide perovskites.⁸² The conductivity signal decays within the first ∼10 ps after photoexcitation, suggesting rapid recombination or trapping of charge carriers, which may be mediated by both point defects and impurity phases. Since both types of defects may also influence the charge-carrier mobility through scattering events, further material purification and refinement is likely to offer increases in both mobility and lifetime of charge carriers.

Fig. 9 Differential transmission of the THz electric field amplitude, proportional to THz transient photoconductivity, recorded using optical-pump terahertz-probe spectroscopy for a CuSn₂SI₃ cold-pressed pellet. Curves were recorded following excitation with 400 nm (3.1 eV) light pulses of 200 mJ cm⁻² fluence. A value of φµ = 1.3 ± 0.2 cm² V⁻¹ s⁻¹ was extracted for the effective electron–hole sum mobility from the peak value after excitation under knowledge of the density of absorbed photons.

Conclusions

We discover two new semiconducting Cu-containing chalco–halides, CuSn₂SI₃ and Cu_0.35Sn_5.29S₂I_7, through exploratory synthesis guided by computational tools. These compounds form a three-dimensional structure with related anion lattices. Layers of Sn₄S₂I₄ are connected by different layers forming the two different compounds, highlighting the structural versatility and possibility to explore more such structures. CuSn₂SI₃ has a band gap in the visible region of ∼2 eV and moderate charge-carrier mobilities comparable to those of recently emerging new solar absorber materials in the metal halide and chalcogenide space, making this a possible candidate material for solar absorber applications, e.g., in multi-junction cells with further design and further tuning of the bandgap. Electronic band structure calculations using DFT suggest CuSn₂SI₃ to be a direct band gap material with flat bands formed due to lack of overlap between the Cu₂I₂ and Sn₄S₂I₄ layers, and highly anisotropic effective masses. These compounds open a relatively less explored class of d¹⁰–s² cations containing chalco–halides as a path for making semiconductors with promising electronic properties.

Author contributions

B. K., T. D. M., and M. J. R. developed the project direction. B. K. discovered and synthesized the materials. L. M. D., T. D. M., and J. B. C. advised on the synthesis. B. K., J. P. S., C. M. R., J. B. C., and L. M. D. characterized the crystal structure. D. B. carried out DFT calculations under the supervision of M. R. F. A. V. developed ML models. D. R. developed the points distribution algorithm under the supervision of M. S. D. UV-vis measurements were done by B. K. under the supervision of T. D. M. D. M. and L. H. performed Terahertz spectroscopy measurements and analysis. M. Z. performed SEM-EDX measurements and analysed the data. C. J. H. did X. P. S. measurements and advised on analysis. B. K. developed the initial draft with inputs from D. B., A. V., and D. M. All authors were involved in discussions and evaluations of the drafts during the writing process. M. J. R. directed the project and acquired funding.

Conflicts of interest

There are no conflicts to declare.

Data availability

CCDC 2503028 and 2503027 contain the supplementary crystallographic data for this paper.^83a,b

The data that support the findings of this research are openly available on the University of Liverpool Data Repository (https://doi.org/10.17638/datacat.liverpool.ac.uk/3072).

Supplementary information (SI): literature information; Structure comparison of CuSn₂SI₃ and Cu_0.35Sn_5.29S₂I₇; Synchrotron PXRD; Elemental analysis; Sample images; XPS data; UV-visible spectroscopy; electronic band structures; Terahertz spectroscopy; crystallographic data. See DOI: https://doi.org/10.1039/d5ta06204g.

Acknowledgements

We thank Dr. Krzystof Pawlak (Materials Innovation Factory, University of Liverpool) for assistance with UV-Vis measurements. We acknowledge funding from the Engineering and Physical Sciences Research Council (EPSRC) under EP/X038777/1 and EP/V026887. The Leverhulme Trust via the Leverhulme Research Centre for Functional Materials Design is thanked for support of J.P.S. This work was possible due to use of the shared analytical equipment of the Materials Innovation Factory (created as part of the UK Research Partnership Innovation Fund (Research England) and co-funded by the Sir Henry Royce Institute). X-ray photoelectron (XPS) data were acquired at the EPSRC National Facility for XPS (“HarwellXPS”, EP/Y023587/1, EP/Y023609/1, EP/Y023536/1, EP/Y023552/1 and EP/Y023544/1). We thank Diamond Light Source for access to beamline I-11 (CY37989). M.R.F and D.B acknowledge computational resources provided by the Texas Advanced Computing Center (TACC) at UT Austin using the Frontera system through the Frontera Pathways allocation DMR23008.

References

1. M. A. Green, A. Ho-Baillie and H. J. Snaith, The emergence of perovskite solar cells, Nat. Photonics, 2014, 8, 506–514.
2. N.-G. Park, Perovskite solar cells: an emerging photovoltaic technology, Mater. Today, 2015, 18, 65–72.
3. H. J. Snaith, Present status and future prospects of perovskite photovoltaics, Nat. Mater., 2018, 17, 372–376.
4. A. K. Jena, A. Kulkarni and T. Miyasaka, Halide Perovskite Photovoltaics: Background, Status, and Future Prospects, Chem. Rev., 2019, 119, 3036–3103.
5. M. M. Lee, J. Teuscher, T. Miyasaka, T. N. Murakami and H. J. Snaith, Efficient Hybrid Solar Cells Based on Meso-Superstructured Organometal Halide Perovskites, Science, 2012, 338, 643–647.
6. H. Min, D. Y. Lee, J. Kim, G. Kim, K. S. Lee, J. Kim, M. J. Paik, Y. K. Kim, K. S. Kim, M. G. Kim, T. J. Shin and S. Il Seok, Perovskite solar cells with atomically coherent interlayers on SnO₂ electrodes, Nature, 2021, 598, 444–450.
7. Z. Li, T. R. Klein, D. H. Kim, M. Yang, J. J. Berry, M. F. A. M. van Hest and K. Zhu, Scalable fabrication of perovskite solar cells, Nat. Rev. Mater., 2018, 3, 1–20.
8. Z. Xiao, Z. Song and Y. Yan, From Lead Halide Perovskites to Lead-Free Metal Halide Perovskites and Perovskite Derivatives, Adv. Mater., 2019, 31, 1803792.
9. J. Zhuang, J. Wang and F. Yan, Review on Chemical Stability of Lead Halide Perovskite Solar Cells, Nano-Micro Lett., 2023, 15, 84.
10. F. De Angelis, The Prospect of Lead-Free Perovskite Photovoltaics, ACS Energy Lett., 2021, 6, 1586–1587.
11. R. E. Brandt, J. R. Poindexter, P. Gorai, R. C. Kurchin, R. L. Z. Hoye, L. Nienhaus, M. W. B. Wilson, J. A. Polizzotti, R. Sereika, R. Žaltauskas, L. C. Lee, J. L. MacManus-Driscoll, M. Bawendi, V. Stevanović and T. Buonassisi, Searching for “Defect-Tolerant” Photovoltaic Materials: Combined Theoretical and Experimental Screening, Chem. Mater., 2017, 29, 4667–4674.
12. A. M. Ganose, C. N. Savory and D. O. Scanlon, Beyond methylammonium lead iodide: prospects for the emergent field of ns² containing solar absorbers, Chem. Commun., 2017, 53, 20–44.
13. H. Kageyama, K. Hayashi, K. Maeda, J. P. Attfield, Z. Hiroi, J. M. Rondinelli and K. R. Poeppelmeier, Expanding frontiers in materials chemistry and physics with multiple anions, Nat. Commun., 2018, 9, 772.
14. U. V. Ghorpade, M. P. Suryawanshi, M. A. Green, T. Wu, X. Hao and K. M. Ryan, Emerging Chalcohalide Materials for Energy Applications, Chem. Rev., 2023, 123, 327–378.
15. C. Ming, Z. Chen, F. Zhang, S. Gong, X. Wu, J. Jiang, T. Ye, Q. Xu, K. Yang, L. Wang, X. Cao, S. Yang, S. Zhang, Y. Zhang, J. Shi and Y.-Y. Sun, Mixed Chalcogenide-Halides for Stable, Lead-Free and Defect-Tolerant Photovoltaics: Computational Screening and Experimental Validation of CuBiSCl₂ with Ideal Band Gap, Adv. Funct. Mater., 2022, 32, 2112682.
16. J. He, X. Hu, Z. Liu, W. Chen and G. Longo, Prospect for Bismuth/Antimony Chalcohalides-Based Solar Cells, Adv. Funct. Mater., 2023, 33, 2306075.
17. K. T. Butler, S. McKechnie, P. Azarhoosh, M. van Schilfgaarde, D. O. Scanlon and A. Walsh, Quasi-particle electronic band structure and alignment of the V-VI-VII semiconductors SbSI, SbSBr, and SbSeI for solar cells, Appl. Phys. Lett., 2016, 108, 112103.
18. A. M. Ganose, S. Matsumoto, J. Buckeridge and D. O. Scanlon, Defect Engineering of Earth-Abundant Solar Absorbers BiSI and BiSeI, Chem. Mater., 2018, 30, 3827–3835.
19. J. Wang and S. Y. Quek, Isolated flat bands and physics of mixed dimensions in a 2D covalent organic framework, Nanoscale, 2020, 12, 20279–20286.
20. G. Sethi, M. Cuma and F. Liu, Excitonic Condensate in Flat Valence and Conduction Bands of Opposite Chirality, Phys. Rev. Lett., 2023, 130, 186401.
21. G. Pasquale, Z. Sun, K. Čerņevičs, R. Perea-Causin, F. Tagarelli, K. Watanabe, T. Taniguchi, E. Malic, O. V. Yazyev and A. Kis, Flat-Band-Induced Many-Body Interactions and Exciton Complexes in a Layered Semiconductor, Nano Lett., 2022, 22, 8883–8891.
22. S. B. Zhang, S.-H. Wei, A. Zunger and H. Katayama-Yoshida, Defect physics of the CuInSe₂ chalcopyrite semiconductor, Phys. Rev. B, 1998, 57, 9642–9656.
23. A. Zakutayev, C. M. Caskey, A. N. Fioretti, D. S. Ginley, J. Vidal, V. Stevanovic, E. Tea and S. Lany, Defect Tolerant Semiconductors for Solar Energy Conversion, J. Phys. Chem. Lett., 2014, 5, 1117–1125.
24. A. Nicolson, S. R. Kavanagh, C. N. Savory, G. W. Watson and D. O. Scanlon, Cu₂SiSe₃ as a promising solar absorber: harnessing cation dissimilarity to avoid killer antisites, J. Mater. Chem. A, 2023, 11, 14833–14839.
25. G. Volonakis, M. R. Filip, A. A. Haghighirad, N. Sakai, B. Wenger, H. J. Snaith and F. Giustino, Lead-Free Halide Double Perovskites via Heterovalent Substitution of Noble Metals, J. Phys. Chem. Lett., 2016, 7, 1254–1259.
26. M.-Y. Ran, S.-H. Zhou, W. Wei, B.-J. Song, Y.-F. Shi, X.-T. Wu, H. Lin and Q.-L. Zhu, Quaternary Chalcohalides CdSnSX₂ (X = Cl or Br) with Neutral Layers: Syntheses, Structures, and Photocatalytic Properties, Inorg. Chem., 2021, 60, 3431–3438.
27. R. Blachnik, W. Buchmeier and H. A. Dreisbach, Structure of lead(II) dimercury(II) diiodide disulfide, Acta Crystallogr. C, 1986, 42, 515–517.
28. A. Pfitzner, CuI)₂Cu₃SbS₃ : Copper Iodide as Solid Solvent for Thiometalate ions, Chem.–Eur. J., 1997, 3, 2032–2038.
29. L. Wang, Y.-C. Hung, S.-J. Hwu, H.-J. Koo and M.-H. Whangbo, Synthesis, Structure, and Properties of a New Family of Mixed-Framework Chalcohalide Semiconductors: CdSbS₂X (X = Cl, Br), CdBiS₂X (X = Cl, Br), and CdBiSe₂X (X = Br, I), Chem. Mater., 2006, 18, 1219–1225.
30. A. Vasylenko, J. Gamon, B. B. Duff, V. V. Gusev, L. M. Daniels, M. Zanella, J. F. Shin, P. M. Sharp, A. Morscher, R. Chen, A. R. Neale, L. J. Hardwick, J. B. Claridge, F. Blanc, M. W. Gaultois, M. S. Dyer and M. J. Rosseinsky, Element selection for crystalline inorganic solid discovery guided by unsupervised machine learning of experimentally explored chemistry, Nat. Commun., 2021, 12, 5561.
31. A. Vasylenko, D. Antypov, V. V. Gusev, M. W. Gaultois, M. S. Dyer and M. J. Rosseinsky, Element selection for functional materials discovery by integrated machine learning of elemental contributions to properties, npj Comput. Mater., 2023, 9, 1–10.
32. A. Blokhin and P. Villars, https://mpds.io.
33. V. Tshitoyan, J. Dagdelen, L. Weston, A. Dunn, Z. Rong, O. Kononova, K. A. Persson, G. Ceder and A. Jain, Unsupervised word embeddings capture latent knowledge from materials science literature, Nature, 2019, 571, 95–98.
34. D. Ritchie, M. W. Gaultois, V. V. Gusev, V. Kurlin, M. J. Rosseinsky and M. S. Dyer, Probabilistic Isolation of Crystalline Inorganic Phases, J. Chem. Inf. Model., 2025, 65, 13226–13237.
35. G. M. Sheldrick, A short history of SHELX, Acta Crystallogr. A, 2008, 64, 112–122.
36. O. V. Dolomanov, L. J. Bourhis, R. J. Gildea, J. a. K. Howard and H. Puschmann, OLEX2: a complete structure solution, refinement and analysis program, J. Appl. Crystallogr., 2009, 42, 339–341.
37. K. Momma and F. Izumi, VESTA 3 for three-dimensional visualization of crystal, volumetric and morphology data, J. Appl. Crystallogr., 2011, 44, 1272–1276.
38. A. A. Coelho, Whole-profile structure solution from powder diffraction data using simulated annealing, J. Appl. Crystallogr., 2000, 33, 899–908.
39. P. Hohenberg and W. Kohn, Inhomogeneous Electron Gas, Phys. Rev., 1964, 136, B864–B871.
40. W. Kohn and L. J. Sham, Self-Consistent Equations Including Exchange and Correlation Effects, Phys. Rev., 1965, 140, A1133–A1138.
41. P. Giannozzi, S. Baroni, N. Bonini, M. Calandra, R. Car, C. Cavazzoni, D. Ceresoli, G. L. Chiarotti, M. Cococcioni, I. Dabo, A. Dal Corso, S. de Gironcoli, S. Fabris, G. Fratesi, R. Gebauer, U. Gerstmann, C. Gougoussis, A. Kokalj, M. Lazzeri, L. Martin-Samos, N. Marzari, F. Mauri, R. Mazzarello, S. Paolini, A. Pasquarello, L. Paulatto, C. Sbraccia, S. Scandolo, G. Sclauzero, A. P. Seitsonen, A. Smogunov, P. Umari and R. M. Wentzcovitch, QUANTUM ESPRESSO: a modular and open-source software project for quantum simulations of materials, J. Phys. Condens. Matter, 2009, 21, 395502.
42. P. Giannozzi, O. Andreussi, T. Brumme, O. Bunau, M. Buongiorno Nardelli, M. Calandra, R. Car, C. Cavazzoni, D. Ceresoli, M. Cococcioni, N. Colonna, I. Carnimeo, A. Dal Corso, S. de Gironcoli, P. Delugas, R. A. DiStasio, A. Ferretti, A. Floris, G. Fratesi, G. Fugallo, R. Gebauer, U. Gerstmann, F. Giustino, T. Gorni, J. Jia, M. Kawamura, H.-Y. Ko, A. Kokalj, E. Küçükbenli, M. Lazzeri, M. Marsili, N. Marzari, F. Mauri, N. L. Nguyen, H.-V. Nguyen, A. Otero-de-la-Roza, L. Paulatto, S. Poncé, D. Rocca, R. Sabatini, B. Santra, M. Schlipf, A. P. Seitsonen, A. Smogunov, I. Timrov, T. Thonhauser, P. Umari, N. Vast, X. Wu and S. Baroni, Advanced capabilities for materials modelling with Quantum ESPRESSO, J. Phys. Condens. Matter, 2017, 29, 465901.
43. J. P. Perdew, K. Burke and M. Ernzerhof, Generalized Gradient Approximation Made Simple, Phys. Rev. Lett., 1996, 77, 3865–3868.
44. D. R. Hamann, Optimized norm-conserving Vanderbilt pseudopotentials, Phys. Rev. B, 2013, 88, 085117.
45. M. J. van Setten, M. Giantomassi, E. Bousquet, M. J. Verstraete, D. R. Hamann, X. Gonze and G.-M. Rignanese, The PseudoDojo: Training and grading a 85 element optimized norm-conserving pseudopotential table, Comput. Phys. Commun., 2018, 226, 39–54.
46. A. Fonari, C. Sutton, Effective Mass Calculator , 2012, https://github.com/afonari/emc.
47. A. Faghaninia, Handy-VASP-DFT-Calculations-Scripts, 2016, https://github.com/albalu/Handy-VASP-DFT-Calculations-Scripts/blob/master/README.md.
48. C. Q. Xia, J. Peng, S. Poncé, J. B. Patel, A. D. Wright, T. W. Crothers, M. Uller Rothmann, J. Borchert, R. L. Milot, H. Kraus, Q. Lin, F. Giustino, L. M. Herz and M. B. Johnston, Limits to Electrical Mobility in Lead-Halide Perovskite Semiconductors, J. Phys. Chem. Lett., 2021, 12, 3607–3617.
49. M. Righetto, S. Caicedo-Dávila, M. T. Sirtl, V. J.-Y. Lim, J. B. Patel, D. A. Egger, T. Bein and L. M. Herz, Alloying Effects on Charge-Carrier Transport in Silver–Bismuth Double Perovskites, J. Phys. Chem. Lett., 2023, 14, 10340–10347.
50. C. L. Davies, J. Borchert, C. Q. Xia, R. L. Milot, H. Kraus, M. B. Johnston and L. M. Herz, Impact of the Organic Cation on the Optoelectronic Properties of Formamidinium Lead Triiodide, J. Phys. Chem. Lett., 2018, 9, 4502–4511.
51. R. A. Howie, W. Moser and I. C. Trevena, The crystal structure of tin(II) iodide, Acta Crystallogr., Sect. B, 1972, 28, 2965–2971.
52. J. Fenner, Die Kristallstruktur des Zinnsulfidjodids, Naturwissenschaften, 1976, 63, 244.
53. N. E. Brese and M. O'Keeffe, Bond-valence parameters for solids, Acta Crystallogr., Sect. B, 1991, 47, 192–197.
54. I. D. Brown, The Chemical Bond in Inorganic Chemistry: the Bond Valence Model, Oxford University Press, 2006.
55. J. Fenner, Die Kristallstruktur des Zinnsulfidiodids Sn₄ S I₆, Z. Anorg. Chem., 1978, 33, 479–481.
56. B. Krebs, Die Kristallstrukturen von Pb₄SeBr₆, Pb₅S₂I₆ und Pb₇S₂Br₁₀, Z. Anorg. Allg. Chem., 1973, 396, 137–151.
57. S. M. Islam, C. D. Malliakas, D. Sarma, D. C. Maloney, C. C. Stoumpos, O. Y. Kontsevoi, A. J. Freeman and M. G. Kanatzidis, Direct Gap Semiconductors Pb₂BiS₂I₃, Sn₂BiS₂I₃, and Sn₂BiSI₅, Chem. Mater., 2016, 28, 7332–7343.
58. M. Ohmasa and K. Mariolacos, The crystal structure of (Pb_1−x,Bi_x)Bi₂Cu₂Cu_2−xS₅I₂ (x = 0.88), Acta Crystallogr., Sect. B, 1974, 30, 2640–2643.
59. T. Balić-Žunić, K. Mariolacos, K. Friese and E. Makovicky, Structure of a synthetic halogen sulfosalt, Cu₃ Bi₂ S₃ I₃, Acta Crystallogr. B, 2005, 61, 239–245.
60. A. Heerwig, A. Isaeva and M. Ruck, Copper(I) d10···d10 Interactions and Diselenide(1–) Radical Anions in Mixed Valent Selenides Cu4–BiSe4I, Z. Anorg. Allg. Chem., 2011, 637, 1131–1136.
61. A. Audzijonis, R. Žaltauskas, R. Sereika, L. Žigas and A. Rėza, Electronic structure and optical properties of BiSI crystal, J. Phys. Chem. Solids, 2010, 71, 884–891.
62. A. Kikuchi, Y. Oka and E. Sawaguchi, Crystal Structure Determination of SbSI, J. Phys. Soc. Jpn., 1967, 23, 337–354.
63. T. Rosenthal, M. Döblinger, P. Wagatha, C. Gold, E.-W. Scheidt, W. Scherer and O. Oeckler, Structural Features and Physical Properties of In₂Bi₃Se₇I, InBi₂Se4I, and BiSeI, Z. Anorg. Allg. Chem., 2011, 637, 2239–2245.
64. P. Kubelka and F. Munk, Ein Beitrag zur Optik der Farbanstriche, Z. Für Techn Phys., 1931, 12, 593–601.
65. P. Makuła, M. Pacia and W. Macyk, How To Correctly Determine the Band Gap Energy of Modified Semiconductor Photocatalysts Based on UV–Vis Spectra, J. Phys. Chem. Lett., 2018, 9, 6814–6817.
66. J. Tauc, R. Grigorovici and A. Vancu, Optical Properties and Electronic Structure of Amorphous Germanium, Phys. Status Solidi B, 1966, 15, 627–637.
67. W. Peng, L. Li, S. Yu, H. Zheng and P. Yang, Structure, binding energy and optoelectrical properties of p-type CuI thin films: The effects of thickness, Appl. Surf. Sci., 2020, 502, 144424.
68. M. Nakamura, S. Inagaki, Y. Okamura, M. Ogino, Y. Takahashi, K. Adachi, D. Hashizume, Y. Tokura and M. Kawasaki, Band structures in orientation-controlled CuI thin films under epitaxial strain, Phys. Rev. B, 2022, 106, 125307.
69. M. T. Hörantner, T. Leijtens, M. E. Ziffer, G. E. Eperon, M. G. Christoforo, M. D. McGehee and H. J. Snaith, The Potential of Multijunction Perovskite Solar Cells, ACS Energy Lett., 2017, 2, 2506–2513.
70. C. Adamo and V. Barone, Toward reliable density functional methods without adjustable parameters: The PBE0 model, J. Chem. Phys., 1999, 110, 6158–6170.
71. A. V. Krukau, O. A. Vydrov, A. F. Izmaylov and G. E. Scuseria, Influence of the exchange screening parameter on the performance of screened hybrid functionals, J. Chem. Phys., 2006, 125, 224106.
72. S. E. Gant, J. B. Haber, M. R. Filip, F. Sagredo, D. Wing, G. Ohad, L. Kronik and J. B. Neaton, Optimally tuned starting point for single-shot GW calculations of solids, Phys. Rev. Mater., 2022, 6, 053802.
73. H. A. Huy, B. Aradi, T. Frauenheim and P. Deák, Calculation of carrier-concentration-dependent effective mass in Nb-doped anatase crystals of TiO₂, Phys. Rev. B, 2011, 83, 155201.
74. J. Tang, L. Y. Deng, C. B. Tay, X. H. Zhang, J. W. Chai, H. Qin, H. W. Liu, T. Venkatesan and S. J. Chua, Determination of carrier concentration dependent electron effective mass and scattering time of n-ZnO thin film by terahertz time domain spectroscopy, J. Appl. Phys., 2014, 115, 033111.
75. A. Mavlonov, T. Razykov, F. Raziq, J. Gan, J. Chantana, Y. Kawano, T. Nishimura, H. Wei, A. Zakutayev, T. Minemoto, X. Zu, S. Li and L. Qiao, A review of Sb₂Se₃ photovoltaic absorber materials and thin-film solar cells, Sol. Energy, 2020, 201, 227–246.
76. N. T. Hahn, J. L. Self and C. B. Mullins, BiSI Micro-Rod Thin Films: Efficient Solar Absorber Electrodes?, J. Phys. Chem. Lett., 2012, 3, 1571–1576.
77. T. J. Whittles, Electronic Characterisation of Earth-Abundant Sulphides for Solar Photovoltaics, Springer International Publishing, Cham, 2018.
78. J. W. Pinder, G. H. Major, D. R. Baer, J. Terry, J. E. Whitten, J. Čechal, J. D. Crossman, A. J. Lizarbe, S. Jafari, C. D. Easton, J. Baltrusaitis, M. A. van Spronsen and M. R. Linford, Avoiding common errors in X-ray photoelectron spectroscopy data collection and analysis, and properly reporting instrument parameters, Appl. Surf. Sci. Adv., 2024, 19, 100534.
79. L. R. V. Buizza, H. C. Sansom, A. D. Wright, A. M. Ulatowski, M. B. Johnston, H. J. Snaith and L. M. Herz, Interplay of Structure, Charge-Carrier Localization and Dynamics in Copper-Silver-Bismuth-Halide Semiconductors, Adv. Funct. Mater., 2022, 32, 2108392.
80. S. Lal, M. Righetto, B. W. J. Putland, H. C. Sansom, S. G. Motti, H. Jin, M. B. Johnston, H. J. Snaith and L. M. Herz, The Role of Chemical Composition in Determining the Charge-Carrier Dynamics in (AgI)_x(BiI₃)_y Rudorffites, Adv. Funct. Mater., 2024, 34, 2315942.
81. M. Righetto, Y. Wang, K. A. Elmestekawy, C. Q. Xia, M. B. Johnston, G. Konstantatos and L. M. Herz, Cation-Disorder Engineering Promotes Efficient Charge-Carrier Transport in AgBiS2 Nanocrystal Films, Adv. Mater., 2023, 35, 2305009.
82. V. J.-Y. Lim, A. M. Ulatowski, C. Kamaraki, M. T. Klug, L. Miranda Perez, M. B. Johnston and L. M. Herz, Air-Degradation Mechanisms in Mixed Lead-Tin Halide Perovskites for Solar Cells, Adv. Energy Mater., 2023, 13, 2200847.
83. (a) CCDC 2503028: Experimental Crystal Structure Determination, 2026,  DOI:10.25505/fiz.icsd.cc2q0lv6; (b) CCDC 2503027: Experimental Crystal Structure Determination, 2026,  DOI:10.25505/fiz.icsd.cc2q0lt5.

---