Structural changes during water-mediated amorphization of semiconducting two-dimensional thiostannates

The amorphization of semiconducting two-dimensional thiostannates was studied using X-ray total scattering and pair distribution function analysis. The local structure and light absorption properties are retained, while the amorphization is associated with disorder of the thiostannate nanosheet stacking.

AEPz-SnS-1: a=b=13.3037(3) Å, c=19.2195(5) Å, R F =0.24. trenH-SnS-1: a=b=13.2457(4) Å, c=19.0860(8) Å, R F =0.10. 1.2 PXRD data of AEPz-SnS-1 and trenH-SnS-1 dispersed in ethanol PXRD patterns of samples dispersed in ethanol for 1 h were collected on a STOE STADI P diffractometer using CuK α1 radiation (1.54056Å) at room temperature. Ground powder samples were distributed on a piece of sticky tape, and data were collected in transmission geometry. 2. Synchrotron total scattering data of water-mediated samples Figure S3: Synchrotron total scattering data of pristine and water-treated all data (left) and low Q (right). Figure S4: Synchrotron total scattering data of pristine and water-treated trenH- all data (left) and low Q (right).   (Baur & Khan, 1971, Filsø et al., 2017, compared with the experimental PDF of trenH-SnS-1.  The isostructural thiostannate layers in AEPz-SnS-1 and trenH-SnS-1 has P6 3 /mmc symmetry and contains four atoms in the asymmetric unit. Refinement details are shown in Table S1 Table S3: Refinement parameters of trenH-SnS-1 fits using anisotropic atomic displacement parameters.  (5) 13.285 (12) 18.97 (7) (2)    Total scattering patterns of pristine, non-stirred and stirred samples (in water 24 h) of trenH-SnS-1 (Fig.  S10a). As the non-stirred sample contains a larger amount of precursor SnO 2 (different trenH-SnS-1 batches) and thus some peaks related to this phase, we will only compare I(Q) at Q < 1.8 Å -1 , where no SnO 2 peaks are present (Baur & Khan, 1971). Evidently, more peaks remain in the non-stirred sample compared to that of the stirred sample, indicating a higher degree of order remaining in the non-stirred sample. The PDFs of the two samples ( Fig. S10b) show retention of the local [Sn 3 S 7 2-] n structure. In the PDF of the nonstirred sample, peaks related to the SnO 2 impurity is mainly observable in the high-r region (r > 10 Å), where the thiostannate correlations are weak in the water-treated sample.

Correlation sorting script
A correlation-sorting MATLAB script was written as a tool to visualize refined correlations against the experimental PDFs. Output from refinements in PDFgui is used as input for the visualization script.
When using a cif-file in PDFgui as a structural starting model for PDF data analysis, an expansion of the structure from the original space group (here P6 3 /mmc) to P1 is made, such that each atom in the unit cell is given a unique index. There are two thiostannate layers in one unit cell of AEPz-SnS-1 and trenH-SnS-1, and the unique atomic ID (e.g. "SN (#1)") allows identifying in which layer an atom in question is placed. This allows distinguishing between bonds/correlations within the layers ("intralayer correlations") and between different layers ("interlayer correlations").
The bond correlation output from PDFGui has the format: SN (#1) -S (#18) = 2.40104 (0.0975516) Å From the PDFgui expansion it is known that the atoms "SN (#1)" and "S (#18)" are in the same layer, and that the distance between the atoms is 2.40104 Å, with an error of 0.098 Å. By a "string compare", the script recognizes a correlation between Sn(1) and S(18), and that both atoms are within the same layer. In addition to identifying (1) the correlation length and (2) whether the bond is intralayer or interlayer, we (3) apply a weighting to the correlation reflecting the scattering power of the two atoms in the correlation pair. Thereby, a Sn-Sn correlation will be weighted higher than an S-S correlation in the histogram. In the above example, a weighting is added by: Z(Sn) * Z(S) = 50 * 16 = 800. An S-S correlation will use the weighting of 16 2 , where Sn-Sn uses 50 2 . Each weighted correlation length is saved in a list for plotting in a histogram.
A bar in the resulting histograms is a sum of the weighted multiplicity of different correlations within the distance interval defined as one bar. The script identifies and sorts all correlations between 0-14.5 Å, while it (in its current form) is unable to extend beyond 14.5 Å. At larger distances, there is a risk of mixing intraand interlayer correlations, as the atomic ID system will start to repeat itself, as layer n and n+2 are equivalent in the structure (as there are two thiostannate layers in one unit cell).

Debye refinements and simulations
In addition to the data analysis methods presented in the manuscript we performed Debye refinements of the PDF data in the program Diffpy-CMI by using the structure file obtained from the refinements in PDFgui. By doing so, we aimed at obtaining information on the crystallite shape and size. Only one Atomic Displacement Parameter (ADP) of S and Sn was varied (i.e. all Sn and all S atoms, respectively, were treated identically). The crystallite size was altered by introducing a supercell, and the shape and size were changed by varying the length along the a, b and c unit cell axes. The best refinement was evaluated from the lowest R w factor. This method suggested crystallite sizes of 1-4 unit cells. As this result is unphysical (e.g. PXRD of the pristine samples reveal microsized crystallites), the method was not explored further.
In a different approach to describe the data, PDFs were calculated in Diffpy-CMI by varying the crystallite sizes and comparing the resulting PDFs to the experimental data. We calculated a series of PDFs where only the layer size (dimension in the ab-plane) was changed (while the size along c was fixed). This series was complemented by a series of simulated PDFs in which only the number of layers along c was changed. By comparing the two series of patterns, we aimed at identifying peaks that only changed in either case. The calculated PDFs were compared to the experimental data. However, both data series presented changes in the PDF intensities at similar distances, which complicated the peak assignment for both samples.

TEM
TEM images were acquired on a Tecnai Spirit electron microscope equipped with a TWIN lens system operating at 120 kV, and using a Veleta CCD side mounted camera. TEM reveals formation of small particles in addition to larger particles identified by SEM (manuscript Fig. 4).

XPS -spectra and atomic concentrations
Representative XPS spectra are shown in Fig. S15-S18. The relative atomic concentrations were determined using a Shirley background and deconvolution.
In the S 2p spectra of both pristine samples ( Fig. S15c and S17c, respectively) an additional peak is observed besides the sulfide peak from the [Sn 3 S 7 2-] n layers. The additional peak corresponds to oxidized sulfur (possibly sulfate) (Moulder & Chastain, 1992) and disappears by water treatment. The two Sn 3d peaks arise from the spin-orbit coupling (3d 5/2 and 3d 3/2) and are assigned to Sn 4+ , as confirmed by literature (Price et al., 1999, Hyeongsu et al., 2018, He et al., 2013. All N is spectra present multiple components, indicative of multiple nitrogen sites. In Fig. S15e and S16e, the spectra of pristine and a water treated sample of trenH-SnS-1 are seen. Two components have been fitted, at binding energies of 399.8 and 401.7 eV (area ratio of 2.3:1 for the pristine sample) assigned to primary and tertiary amines, respectively, of the tren molecule. In the post water treatment, the primaryto-tertiary amine ratio has decreased to 1.3:1. Conversely, the N peaks (tertiary and primary/secondary amine) from AEPz, in Fig. S17 and S18, almost retains the same area ratio of 1:1.8 for the pristine sample and 1:1.7 for the water treated sample.
In all C spectra three components are found at binding energies of 285.5, 286.4 and 288.3 eV, which we suggest to arise from the C sp 3 , C-N/C-O and carbonyl groups, due to the tren/AEPz molecules and surface contaminants.
The O 1s spectra have not been deconvoluted owing to the high contamination of the sample surface.
In Table S5-8 the areas of the integrated peaks are tabulated along with their respective binding energies. Table S9 displays the averaged composition of the samples. The composition presented in the manuscript is based on the values in Table S9.       1.24 ± 0.14 0.60 ± 0.14 1.17 ± 0.01 0.57 ± 0.09