Nanocrystals of Lead Chalcohalides: A Series of Kinetically Trapped Metastable Nanostructures

We report the colloidal synthesis of a series of surfactant-stabilized lead chalcohalide nanocrystals. Our work is mainly focused on Pb4S3Br2, a chalcohalide phase unknown to date that does not belong to the ambient-pressure PbS–PbBr2 phase diagram. The Pb4S3Br2 nanocrystals herein feature a remarkably narrow size distribution (with a size dispersion as low as 5%), a good size tunability (from 7 to ∼30 nm), an indirect bandgap, photoconductivity (responsivity = 4 ± 1 mA/W), and stability for months in air. A crystal structure is proposed for this new material by combining the information from 3D electron diffraction and electron tomography of a single nanocrystal, X-ray powder diffraction, and density functional theory calculations. Such a structure is closely related to that of the recently discovered high-pressure chalcohalide Pb4S3I2 phase, and indeed we were able to extend our synthesis scheme to Pb4S3I2 colloidal nanocrystals, whose structure matches the one that has been published for the bulk. Finally, we could also prepare nanocrystals of Pb3S2Cl2, which proved to be a structural analogue of the recently reported bulk Pb3Se2Br2 phase. It is remarkable that one high-pressure structure (for Pb4S3I2) and two metastable structures that had not yet been reported (for Pb4S3Br2 and Pb3S2Cl2) can be prepared on the nanoscale by wet-chemical approaches. This highlights the important role of colloidal chemistry in the discovery of new materials and motivates further exploration into metal chalcohalide nanocrystals.


Support Material a) Direct synthesis of Pb 4 S 3 Br 2 NCs
The Pb4S3Br2 nanocrystals (NCs) used in this work have been synthesized via a modification of the heat-up method previously published by our group. 1 In a standard synthesis, 0.2 mmol of PbBr2 and 0.2 mmol of Pb(SCN)2 were dissolved in a mixture of 10 mL ODE and 250 μL of OLAM and OA at 110°C in a 25 mL three-necked flask. This temperature was not exceeded during the solubilization, to avoid triggering the thermal decomposition of SCNions. Once the two solids were completely dissolved the solution appeared clear and pale-yellow. At that point, the flask was lifted from the heating-mantle, which was overheated to an inner temperature of 250°C to grant the fast increase of temperature required for the synthesis. The flask, till that point kept at ≈110°C by hovering it at the rim of the mantle, was again inserted and as a result the solution quickly heated up (≈ 20°C / min), consequently turning from light-yellow to bloody red above 150°C. As reported in our previous paper, the following reaction is taking place: 1 2 SCN -+ 3O2 → S 2-+ SO2 + 2 CO2 + N2 As a consequence, S 2becomes suddenly available in the reaction system and co-precipitates with Pb 2+ and Brforming the reported Pb4S3Br2 phase.    Figure S3. Timeline of the reaction for the synthesis of Pb4S3Br2 NCs by optical absorption. Absorption spectra of the aliquots shown in Figure S1. The spectra were not normalized to give additional information on how the absorbance increases while the NCs nucleate and grow. All the samples were prepared by diluting 50 μL of the crude reaction batch in 900 μL of toluene inside a 10 mm-path cuvette. The spectra collected from 175 -180 -190°C aliquots show the characteristic effect of scattering, since particles grew too large to form a stable colloidal suspension in toluene once the samples cooled to room temperature and were diluted.  NCs which underwent overheating (T≈200°C). The sharp peaks do not belong to Pb 4 S 3 Br 2 , whose pattern is barely visible above the baseline, but are instead attributed to yet unreported PbBr 2 layered structures. We suspect that they share similarities with the PbI 2 impurity observed for of Pb 4 S 3 I 2 NCs (see Figure S32). 2 The peak shift towards lower angles (larger periodicities) is probably due to a structural distortion.

Further accretion of Pb4S3Br2 NCs
Large-size Pb4S3Br2 NCs were needed to attempt an ab initio NC structure solution via single particle 3D-ED. A batch of NCs was synthesized according to the reported procedure, and the growth was quenched at 170°C. The particles were not recovered from the reaction batch; instead, the flask was reheated to 170°C and kept stirring at this constant temperature. In a separated flask, 0.6 mmol of PbBr2 The compositions of both pseudospherical and platelet-shaped NCs were measured via TEM-EDX and were found in good agreement with each other, suggesting that NCs and nanoplatelets share the same stoichiometry. When comparing the two measurements, one must take into account that the two morphologies will be differently affected by the surface termination and preferential growth effects, thus making it unlikely that the overall elemental compositions will be identical for the two different NC morphologies. b) XPS analysis on the Pb 4 S 3 Br 2 NCs as synthesized. Figure S8. XPS analysis on the Pb4S3Br2 NCs as synthesized. a) Wide scan, showing signals due to Pb, S, Br (from the inorganic core of the NCs) together with C and O ones, arising from the organic ligands and possibly also from environmental contaminations, which is common for on samples exposed to moist air. b-d) High resolution scans on the S 2p, Pb 4f and Br 3d energy regions. Figure S8 reports the XPS data collected on a representative Pb4S3Br2 sample. As shown in the wide scan of Figure S8a, the sample is characterized by the presence of Pb, S and Br, together with C and O.
These last two elements can be attributed to the organic ligand shell surrounding the inorganic core of the as-synthesized NCs. However, we cannot completely exclude the additional presence of environmental contaminations, due to exposure of the sample to moist air before the XPS data acquisition. Figure S8b 3 The Br 3d spectrum in Figure S8d shows the presence of a single doublet, with the main component centered at 68.0 ± 0.2 eV. This position is typical of bromides. 3 Similar results were obtained on all the investigated batches; the following table reports the results of the quantitative analysis.

c) EDX analysis on as synthesized Pb 4 S 3 Br 2 NCs.
One of the NC suspensions used to prepare the solar cells was analyzed via SEM-EDX to measure the      Figure S10 shows the outcome of the four TGA experiments we used to provide an estimate of the stoichiometry of the NCs based on a "bulk" measurement. We want to point out that the first mass loss, occurring in a temperature range of 110 -300°C, represents the desorption/decomposition of the organic fraction in the sample (solvent, ligands) and is therefore heavily sample-dependent. For example, the sample in Figure S11 had not been washed, and the first mass loss accounted for a huge 87% of the total.
The bottom left sample in figure S10 contained instead the largest particles that we were able to prepare (∼30 nm diameter) and had been cleaned carefully. Indeed, the first mass loss in this case accounted for less than 10% of the total.
The second mass loss, combined with the XRD analyses of the sample after the first mass loss and after the second mass loss, carries the compositional information. The stoichiometry was measured from each dataset according to the following method. We first accurately determined the residual relative masses according to the graphical method shown in Figure S12. Figure S12. Accurate determination of the initial and final points of the mass loss step. The plateau regions before and after the mass loss were approximated as straight lines tangent to the plot (red lines) as well as the transition slope (green line). The residual relative masses before and after the weight loss were measured at the intersection points between the green and the red lines.

S13
We know from XRPD that, after the first mass loss, the sample contains a mixture of and , in an unknown molar ratio : , that is . Assuming that in the first mass loss only the organic fraction is removed, the stoichiometry of our initial NCs, in terms of and , is . In the second mass loss, decomposes to and , with the latter sublimating, as the only final product left is PbS. The overall sequence of the two reactions, can then be written as: We determine by the mass loss of the second step, since only sublimates in this step. We then determine , since we now know experimentally both and the total mass of left at the end of the second step. From and we then determine the stoichiometry of our initial NCs. This estimate was done on four separate TGA+XRD analyses, run on four different samples. The resulting stoichiometries are reported in the following table.      Indexing.
The determination of the cell parameters, the first crucial step of the ab initio structure solution, was carried out by EXPO2014, 4 providing to the indexing program N-TREOR09 5 28 observed peak positions in the 12°-89° 2 range. Due to the large broadness of the experimental peaks and the consequently unavoidable low accuracy of peak positions, and in order to be able to find the correct cell among the possible ones, the indexing process was carried out in a non-default way by increasing the tolerance con-

Space group determination.
The space group determination process by EXPO2014 exploits the information on cell parameters, integrated intensities (extracted in the space group having the largest Laue symmetry and no extinction conditions, i.e., in Pmmm) and expected unit cell content, to carry out a statistical analysis on suitably weighted integrated intensities, in order to detect systematic absences and calculate a probability value associated to the different possible extinction symbols compatible with the identified crystal system [7][8][9] .
Due to the broad and overlapping peaks in the NC powder pattern and, consequently, to the large errors on the integrated intensities estimates, the results of the statistical study were unreliable and prevented the correct identification of the space group. In case of the 3D-ED data, the greater accuracy of the integrated intensities enabled to detect the absence of: i) The class of reflections (0 k l) with k+l odd (i.e., the presence of a glide n normal to the a axis) ii) The class of reflections (h k 0) with h odd (i.e., the presence of a glide a normal to the c axis) This led to the identification of the space group Pnma. This result, derived by the investigation based on 3D electron diffraction data, was actively used for the structure solution step by powder diffraction data.

EXPO2014 estimates the integrated intensities of reflections by alternating the application of the Le
Bail algorithm 10 to least-squares cycles minimizing the residual between observed and calculated profiles.
The default profile function used to describe the peak shape is the Pearson VII. The refined variables are scale factor, background coefficients, 2 -zero shift, peak asymmetry, FWHM parameters, the m parameter of the Pearson VII function and the unit cell parameters. Thanks to a useful property of the Le Bail algorithm, to improve the integrated intensities estimates EXPO2014 can fruitfully take advantage of some prior information 11 , including the expected positivity of the Patterson function 12 that is particularly effective in case of structures with heavy atoms; for that reason, it was exploited in case of the NC powder diffraction pattern.
Structure solution and structure model optimization.
The default structure solution process by EXPO2014 is based on the application of Direct Methods (DM), providing 20 sets of phases processed by the ALLTRIAL procedure 4 , that, for each set of stored phases, automatically performs a preliminary refinement and structure model optimization by a Fourier recycling approach 13 ; the corresponding 20 structure models were carefully analyzed by a graphical inspection. The culated structure factor modulus, respectively). The asymmetric unit of the crystal structure determined by the ab initio structure solution from XRPD data (whose main details are provided in Table S7) and its local environment are shown in Figure S18a and S18b, respectively. The crystal structure reveals itself very similar to that one determined by 3D electron data (see Figure S19, showing a view of the overlay of the crystal structures determined by XRPD and 3D electron diffraction data, represented by red and blue rods respectively). A comparison of the two structure models reveals a root mean square deviation (RMSD) of 0.534 Ǻ between them. Figure S18. Ab initio structure solution from XRPD data. The asymmetric unit a) and its surroundings b) as obtained from the XRPD data-based ab initio structure solution. 95.98 S1-Pb2-Pb1-S1 0.02 S2-Pb2-S1-Pb1 −51.95 Pb2-S2-Pb3-Br1 116.46 Pb1-Pb2-S1-Pb1 0.02 Pb2-S1-Pb1-Pb2 0.02 Figure S19. Visual assessment of matching between XRPD and 3D-ED models. Overlay of the asymmetric unities of the structure models obtained by 3D-ED (blue) and by XRPD (red) ab initio structure solution. The root mean square deviation (RMSD) for the two structure models was found to be 0.534 Ǻ. The outcome of the refinement is summarized in the table below. Notes on the fit: refining the anisotropic thermal factors was needed in order to reach a good match between the data and the calculated profile. Only null or positive values were allowed.
The following crystal structures are available to the reader in the form of .CIF files;  DFT-relaxed structure as shown in Figure 4d S28 o) XPS and UPS analyses ligand-exchanged Pb 4 S 3 Br 2 NCs for devices.
As reported in the main text, to make a film of Pb4S3Br2 NCs conductive and use it as active layer in a photovoltaic device, we performed a solid-ligand exchange with 1-ethyl-3-methylimidazolium iodide (EMII, see details in the main manuscript). The effectiveness of the ligand replacement as well as the suitability of the energy level alignment from the ligand exchange in the layered stack were investigated via XPS and UPS. Figure S26 shows        al. 18 The good fit of the overall profile confirms that we obtained the expected phase, but evidences the presence of a crystalline impurity in non-negligible amounts, clearly indicated by the peaks found in the residual profile. A good match with the pattern of PbI 2 flakes is observed. 2 Note on the fit: the anisotropic broadening factors were included in order to take into account the anisotropic shape observed in NCs under the electron microscope. The relatively high Goodness of Fit parameter (lower is better) depends mainly on the presence of the non-fitted impurity peaks.

r) Direct synthesis and structural characterization of Pb 3 S 2 Cl 2 NCs
Conditions similar to those described in paragraph S1.a were applied for the synthesis of Pb3S2Cl2 NCs: 0.2 mmol of PbI2 and 0.2 mmol of Pb(SCN)2 were dissolved in a mixture of 10 mL ODE and 250 μL of OLAM and OA at 110°C in a 25 mL three-necked flask. The dissolution of PbCl2 was slower than that of PbBr2 and sometimes was incomplete. Thus, we introduced an additional filtering step: we let the batch cool down to ∼50°C, decanted it for a few seconds and then filtered the liquid with a 0.2 μm PTFE syringe filter. The clear liquid was then heated up again to 110°C and the procedure continued unchanged.
NCs formed upon heating up were generally smaller than those of Pb4S3Br2 (if compared at the same quench temperature, e.g. 170°C) and were recovered via ethyl-acetate assisted precipitation. One advantage of this approach is that we also eliminated the large PbS NCs that often formed during the synthesis and that were found in the precipitate. Differently from the case of Pb4S3I2, no ternary Pb-S-Cl phases matching with our material could be found in crystal structure databases (COD, ICSD). However, an unexpected help in its identification came from the work of Rabenau et. al. (dated 1969), in which they reported two tentative new phases, named "2PbS + PbBr2" and "2PbS + PbCl2". 19 According to their report, the first phase was obtained pure in form of a precipitate by first dissolving PbS in concentrated HBr and subsequently diluting the solution with water.
They measured its stoichiometry and collected the XRPD pattern but could not solve the structure. The second phase was obtained via a similar protocol, involving HCl instead of HBr, and produced a red precipitate which rapidly decomposed to form a black powder. Thus, they could not collect it pure; however, they measured the XRPD of the mixture, finding that it was mainly composed by PbS but also contained a residual which produced a pattern similar to that of "2PbS + PbBr2". Hence, they hypothesized S35 the existence of a "2PbS + PbCl2" phase. Sixty-one years later, Ni et. al. reported the solid-state synthesis of the high-pressure chalcohalide Pb3Se2Br2 (4 GPa, 700°C), which features an equivalent stoichiometry and a pattern very close to that of our chlorine-based NCs 20 (after proper shifting of the peaks to take into account the different sizes of the cells). Figure S34 compares the XRPD we measured that reported for "2PbS + PbCl2" and for Pb3Se2Br2, suggesting that we obtained NCs of the Pb3S2Cl2 phase that Rabenau et. al. hypothesized many years ago. Note on the fit: only the unit cell parameters and the scale factor were refined. All the remaining parameters were kept as originally determined by Ni et.al. in their work. 20 The overall fit match is good, but the residual indicates that the peaks are slightly misplaced. The cause is most likely a slight distortion from the original cubic symmetry. However, due to the peak broadening the information needed to address this deformation was lost, and we preferred to keep a pseudocubic description for the structure. We plan to investigate this aspect in further specific studies.  18,20 The calculations were performed starting from the models obtained by Rietveld fit as reported in sections S.q and S.r, after relaxing the unit cell content while keeping the cell parameters fixed. Since Cl and S atoms in Pb 3 S 2 Cl 2 share randomly the same Wyckoff sites, a stoichiometric cell with random distribution of the two elements was considered. Calculations were performed without including spin orbit coupling (SOC) to better estimate the bandgap.