Eilidh L. Quinn,1,† Hugh Lohan,1,2,† Elita Tmava,1 Shiling Dong,1 Aron Walsh,2 and Robert L. Z. Hoye,1,*

1. Inorganic Chemistry Laboratory, University of Oxford, South Parks Road, Oxford, OX1 3QR, United Kingdom
2. Department of Materials and Centre for Processable Electronics, Imperial College London, Exhibition Road, London, SW7 2AZ, United Kingdom 

†These authors contributed equally to this work
Email: robert.hoye@chem.ox.ac.uk (R.L.Z.H.)

Main Text: 
Raw data is stored in txt, origin, or VASP standard output files. It can be opened/parsed in Origin, Adobe Illustrator, MATLAB, python, or Microsoft Excel.

Fig. 1 
a, Proposed crystal structure of CsBiSCl2 by Huang et al. (17) and Quarta et al. (33).
b, Energy above the Cs-Bi-S-Cl convex hull of CsBiSCl2 structures selected for DFT refinalization. Shown are the lowest energy structure (yellow star), tetragonal perovskite structure (blue star), and remaining CsBiSCl2 structures with 70 meV/atom of the hull (yellow circles). Data obtained by Hugh Lohan. 
c, Phonon dispersion curve of the lowest-energy CsBiSCl2 structure from AiRSS. Lack of imaginary modes (i.e., no negative frequencies) demonstrates the dynamic stability of this structure. Data obtained by Hugh Lohan. 
d, Comparison of the simulated powder X-ray diffraction (PXRD) pattern for the lowest energy Pnma structure and the CsBiSCl2 PXRD pattern reported by Huang et al. (17) Comparison with tetragonal perovskite in Fig. S6, SI. Data obtained by Hugh Lohan and Eilidh Quinn. 


Fig. 2
a, Illustration of the synthesis procedure for preparation of the DMABiS2 intermediate. Created in BioRender by Eilidh Quinn (https://BioRender.com/xo6o9y9)
b, Images of the dark brown product powder and the orange precipitate observed in the filtrate during the washing stage. Data obtained by Eilidh Quinn
c, PXRD pattern of synthesised DMABiS2 material. The light blue bars indicate the references patterns of Bi2S3 (ICSD #30775), while the light green bars represent the reported pattern by Huang et al.  for their claimed DMABiS2 intermediate. (17) Data obtained by Elita Tmava and Eilidh Quinn.
d, Infrared spectra of the dark brown powder and DMACl reference material. The peaks assigned to the ammonium N-H stretch are highlighted. The increase in peak wavenumber for the dark brown powder sample is attributed to the reduction in electronegativity of the environment leading to stronger N-H bonds. Data obtained by Elita Tmava and Eilidh Quinn 
e, The effect of solvent washing volume on the mean atomic percentage of iodine from EDS measurements of dark brown product powders. The error of atomic percentage is calculated using procedure described in Fig. S12, S1. Data obtained by Shiling Dong and Elita Tmava


Fig. 3
a, PXRD pattern of thin films synthesised according to Huang et al. (solid circles), with the corresponding Pawley refinement to Cs3Bi2I9 and Bi2S3 (red line); Rwp = 9.1874 GoF = 1.2124. Inset Photograph of the as-prepared thin film on 1.4 cm2 glass. Data obtained by Elita Tmava and Eilidh Quinn. 
b, UV-Vis absorption spectrum of prepared thin films and Cs3Bi2I9 reference spectrum, (52) with the characteristic Cs3Bi2I9 excitonic peaks labelled. Data obtained by Elita Tmava. 
c, PXRD pattern of solid residue from precursor solution from the attempted synthesis of CsBiSCl2 (solid circles), with the corresponding Pawley refinement to Bi2S3 and CsCl (red line); Rwp = 12.9970 GoF = 2.1417, unidentified peaks are labelled by *. Insert: Photograph of the solid residue in a glass vial. Data obtained by Elita Tmava and Eilidh Quinn. 
d, PXRD pattern of material from attempted solid-state synthesis of CsBiSCl2 (solid circles), with the corresponding Pawley refinement to CsCl, Bi2S3, and CsBi3S5 (red line); Rwp = 10.9693 GoF = 1.7417. Data obtained by Hugh Lohan. 


Supplementary Information: 

Supplementary Table S1
Convex hull structures for refinalized DFT dataset. The tetragonal perovskite structure is included for reference. Data obtained by Hugh Lohan.

Supplementary Fig. S1
Mean absolute error (MAE) and root mean squared error (RMSE) of bespoke EDDP. Top row: Error over the ranges 0-0.2 eV, 0-1 eV and 0-10 eV (left to right) for full Cs-Bi-S-Cl testing dataset. Bottom row: error over the ranges 0-0.2 eV, 0-1 eV and 0-10 eV (left to right) for CsBiSCl2 structures only. To facilitate plotting, the zero point for each composition is defined by the lowest energy structure of that composition. Data obtained by Hugh Lohan. 

Supplementary Fig. S2
Unphysical structure removed from EDDP structure search database. Anion clusters and unrealistically short bond lengths as seen above were a frequent issue in early training of the EDDP. Reducing the hard spheres (OVERLAP) reduced the prevalence of clustered structures. Data obtained by Hugh Lohan. 

Supplementary Fig. S3
Structural density of states for 'hot-AiRSS' before and after molecular dynamics runs. The distribution is broadened into the high energy regions, suggesting some melting occurred. We also note the lowest-energy structure of CsBiSCl2 (Pnma space group) was rediscovered an additional 8 times, but no lower energy structures were found. Data obtained by Hugh Lohan. 

Supplementary Fig. S4
Band structure for the lowest-energy Pnma structure of CsBiSCl2 interpolated from a uniform k-point grid (6×12×3) using AMSET (left). (1) Projected density of states plotted using sumo (right). (2) We estimate the indirect and direct bandgaps to be 1.9 eV and 2.0 eV respectively. These were calculated using the r2SCAN functional with spin-orbit coupling and D3(BJ) dispersion correction. Due to the bandgap problem of (meta-)GGA, these likely provide a lower limit estimate of the bandgap. By detailed balance, this material is unsuitable for single-junction solar cells, although this would be in range for use in indoor-photovoltaics or a tandem top-cell. We note that, for precise bandgaps, a calculation along high-symmetry lines through k-space is standard. Data obtained by Hugh Lohan. 

Supplementary Fig. S5
MatterSim Benchmark. Structural density of states for 216 DFT and 500 MatterSim structures. Data obtained by Hugh Lohan. 

Supplementary Fig. S6
Visual comparison between the pattern reported by Huang et al. (3) and the tetragonal perovskite pattern found from random structure searching (Table S1). Data obtained by Eilidh Quinn and Hugh Lohan.

Supplementary Fig. S7
a, Powder X-ray diffraction (PXRD) pattern of dark brown product (solid circles) with Pawley refinement to Bi2S3 (red line); Rwp = 27.3685, GoF = 3.12564. Data obtained by Elita Tmava and Eilidh Quinn
b, PXRD pattern of dark brown product (solid circles) with Pawley refinement to Bi2S3 and DMAI (red line); Rwp = 25.7725, GoF = 2.98988. There is an improvement in the fit when the powder is refined to both Bi2S3 and DMAI, suggesting both phases to be present in the dark brown product. However, there are still some peaks that remain unidentified. Data obtained by Elita Tmava and Eilidh Quinn.

Supplementary Fig. S8 
Scanning electron microscopy (SEM) image of the dark brown intermediate product prepared using the reported synthesis method. (3) 
a, Micrograph what appears to be two distinct phases: a layered solid, and smaller particles on the surface. Data obtained by Shiling Dong.
b, Another micrograph that more clearly shows a layered structure consistent with Bi2S3. Data obtained by Shiling Dong. 
c, Micrograph with labels indicating where energy dispersive X-ray spectroscopy (EDX) measurements were taken. Data obtained by Shiling Dong. 

Supplementary Table S2
The elemental composition obtained by fitting each EDX spectrum identified in Fig. S8. We have excluded C and O from the calculation as they are mainly from the carbon tape used for mounting and electrical grounding, as well as adventitious contamination. We normalized the sum of the remaining elements to a total of 100%. It should be noted that the extracted elemental ratios are inadequate to establish the precise stoichiometry of the phases, since the effective interaction depth of the electron beam extends from tens of nanometers to several micrometers, thereby encompassing signals from multiple particles. Nevertheless, spectra 3 and 6 exhibit relatively high iodine and nitrogen contents, consistent with their assignment to DMAI present on the smaller surface domains. Data obtained by Elita Tmava and Shiling Dong. 

Supplementary Fig. S9 
PXRD pattern of dark brown intermediate product (solid circles) with Pawley refinement to BiSI (red line); Rwp = 42.0746, GoF = 4.7876. The poor fitting confirms that the dark brown product is not BiSI. Data obtained by Elita Tmava and Eilidh Quinn. 

Supplementary Fig. S10 
Infrared spectrum of the orange precipitate showing no distinct functional group features. Data obtained by Elita Tmava and Eilidh Quinn. 

Supplementary Fig. S11
a, SEM image of the thin film prepared. Data obtained by Shiling Dong. 
b, close-up to show the two different sized microstructures in greater detail. Data obtained by Shiling Dong. 
c, SEM image of the part of the thin film sample that underwent EDX measurements, with the regions spectra were taken from indicated. Data obtained by Shiling Dong. 

Supplementary Table S3 
Elemental composition for each spectrum identified in Figure S11. Spectrum 1, which is located on the micron-sized crystal, has a Bi:S ratio of 1:1.2, which is close to that of Bi2S3, albeit with I-based species included (see comment in Supplementary Note 2). Spectrum 2 and 3 are from the smaller-sized microfeatures and have no S present. In all cases, Cl was not observed in the film. Data obtained by Elita Tmava and Shiling Dong. 

Supplementary Fig. S12
EDX elemental analysis results and typical fitted EDX spectra of (a) Bi2S3 and (b) BiCl3. Carbon and oxygen elements are involved in the fitting of the spectra, but excluded from the elemental analysis because they mainly come from the carbon tape and adventitious contaminants. Data obtained by Shiling Dong. 





