Results and Discussion

The charge-ordered Cs₂In(I)In(III)Cl₆ (In⁺:In³⁺ = 1:1) single crystals were synthesized using a traditional solid-state reaction method. For a typical synthesis, an equivalent molar ratio of InCl₃ and InCl was mixed together with 2 equivalent amounts of CsCl, all of which were ground carefully in an Ar glovebox. Afterward, the fine powder was loaded into an evacuated quartz tube. The sealed ampoule was heated at 650 °C for 60 h in a furnace and then slowly cooled down to room temperature. Colorless block-like crystals were harvested as shown in SI Appendix, Fig. S1.

Similar to other charge-ordered perovskites such as CsTlCl₃ (26), the crystal structure of Cs₂In(I)In(III)Cl₆ is very complicated, which requires a comprehensive suite of structural characterization techniques to obtain its accurate crystal structure model. Synchrotron-based single-crystal X-ray diffraction (SCXRD) data obtained at 100 K indicates that there are a set of strong reflection data with weak superreflections in the unwrapped images (SI Appendix, Fig. S2). If only considering the strong reflections, we can deduce a tetragonal cell with space group P4/m and a = 5.3915(4) Å, c = 5.5123(6) Å. Based on this analysis, a structure model with a typical perovskite framework with disordered In positions was obtained. However, the simulated powder X-ray diffraction (PXRD) based on this structure model only matches the strongest reflections in the experimental PXRD data, indicating further work is necessary to model the weaker reflections among the strong reflections. Considering the weaker reflections together, a tetragonal supercell with space group I4/m and a = 17.2604(12) Å, c = 11.0113(16) Å can be obtained. Further analysis using rotation electron diffraction (RED) collected from a small piece of nanosized crystal crushed from the bulk crystal also confirmed that this supercell is more suitable for describing the crystal structure of Cs₂In(I)In(III)Cl₆ (SI Appendix, Fig. S3) (30, 31). As a result, we used this unit cell for data reduction, structure determination, and refinement based on our best SCXRD dataset. The detailed structure parameters are listed in SI Appendix, Tables S1 and S2. The final structure refinement turned out to converge successfully with a low R value. The simulated PXRD pattern based on this model well matched the experimental PXRD data (see below for detailed discussion). Energy-dispersive spectroscopy shown in SI Appendix, Fig. S4 determined the atomic ratio of Cs:In:Cl to be around 1:0.95:2.99, which was consistent with the double perovskite structure chemical formula.

The final structure model of Cs₂In(I)In(III)Cl₆ turns out to be heavily disordered. The structure model shows common structure features as the reported charge-ordered structure of CsTlCl₃ and Cs_1.17In_0.81Cl₃ (26, 32), but with its own signature. In our structure model, the In, Cl, and Cs ions are found to be located at disordered positions with various occupancies. Among all of the 4 individual In positions, In(I)3 and In(I)4 are refined to be at a few disordered positions with different occupancies as shown in Fig. 1 A and B. The exact coordinational polyhedra are hard to be concluded here due to the disorder complexity, as the ionic radius of In(I) is very similar to Cs⁺, and significantly different from In(III). The coordination sphere of In(I) with Cl might be more similar to Cs⁺ instead of being located as octahedra like In(III)Cl₆. As there is a lack of enough accurate statistic experimental data on In(I)–Cl coordinational sphere in both CCDC and ICSD database, here we prefer not to make an absolute conclusion on the In(I)–Cl coordinational sphere, but just use a red ball to represent In(I) position when illustrating our structure model. Unlike the disordered location of In(I), both of the 2 independent In(III) atoms are refined to be located at the ordered positions, while the coordinated Cl atoms around the equatorial plane of the octahedra are all disordered into 2 positions. These disorders introduce the 45° rotation of the In(III)Cl₆ octahedra as highlighted in green and yellow in Fig. 1C. The partial occupancies of Cl2A/Cl2B, Cl5A/Cl5B, and Cl7A/Cl7B are equal to 0.79/0.21, 0.79/0.21, and 0.77/0.23, respectively. The rotation of the In(III)Cl₆ results in 2 disordered structural components as shown in Fig. 1 D–F and Fig. 1 G–I, with a ratio around 0.78/0.22 as determined by the ratio of disordered Cl atom components. The general crystal structure model of component 1 shown in Fig. 1 D–F is very much the same as a double perovskite, while the rotation of the In(III)Cl₆ octahedra as highlighted in blue introduces the supercell feature, which is very similar to the reported Cs_1.17In_0.81Cl₃. The small difference between component 1 and Cs_1.17In_0.81Cl₃ is the atomicity of In(I) position as illustrated with a red ball in the figure. In our structure, based on the comprehensive elemental analysis and refinement results, In(I) positions should be fully occupied with In, while mixed occupancies of Cs⁺ and In⁺ have been observed in Cs_1.17In_0.81Cl₃. The other 22% of the single crystal is contributed by component 2 resulting from the 45° rotation of the In(III)Cl₆ octahedra as shown in Fig. 1 G–I. In component 2, the structure model is deviated far away from a typical perovskite structure. Although it is not fully understood why the In(III)Cl₆ octahedra preferred to be partially disordered with a rotation feature, it should be the reason for the 8-fold-like extra reflection circles in the unwrapped image of the SCXRD pattern (SI Appendix, Fig. S2). Moreover, compared with Cs_1.17In_0.81Cl₃, this structure difference actually contributes to the different physical and chemical properties of our material and results in significant differences in the PXRD patterns.

• Download figure
• Open in new tab
• Download powerpoint

Fig. 1.

Single-crystal structure of Cs₂In(I)In(III)Cl₆. (A and B) Disordered In(I)3 (A) and In(I)4 (B) positions in the In(I)Cl₆ octahedra. (C) Disordered Cl ion positions around the In(III)Cl₆ octahedra. (D–F) Disordered component 1 viewing along [001], [110], and [100] zone axis, respectively. (G–I) Disordered component 2 viewing along [001], [110], and [100] zone axis, respectively. Color scheme: red balls, In(I) atoms; green and yellow balls, Cl atoms; blue and amber, In(III) atoms; purple, Cs atoms.

PXRD collected with laboratory source Cu Kα radiation in Fig. 2A shows that the experimental result is well consistent with the simulated pattern, indicating that the sample is in a single pure phase. Also, we reproduced the crystal of the reported Cs_1.17In_0.81Cl₃ and find the experimental and simulated PXRD patterns are significantly different from ours (SI Appendix, Fig. S5). As most of the In(I)-based halides materials are susceptible to oxidation (33), the structure stability of the as-synthesized Cs₂In(I)In(III)Cl₆ was investigated under ambient condition with PXRD, as shown in Fig. 2B. The framework of Cs₂In(I)In(III)Cl₆ is stable in dry air (relative humidity of 35%) even over 18 d. The air stability is further confirmed by Raman and absorption spectra (SI Appendix, Figs. S6 and S7). The air stability is significantly enhanced compared with the reported Cs_1.17In_0.81Cl₃. We have confirmed the stability difference by tracking the PXRD, absorption, and appearance evolution in the same condition as shown in SI Appendix, Figs. S8 and S9. Since In⁺ can be easily oxidized, and InCl₃ is highly hygroscopic, the formation of the charge-ordered perovskite structure could significantly enhance the stability. To investigate the In⁺ charge stability, ex situ synchrotron-based soft X-ray photoelectron spectroscopy (SXPS) has been conducted (Fig. 2 C and D). The results indicate that the surface of the as-synthesized fresh Cs₂In(I)In(III)Cl₆ crystal has an In⁺/In³⁺ ratio of about 0.55:1, while the sample after exposure in air shows only pure In³⁺, suggesting that the valence state of In⁺ is not stable against oxidation. Although the PXRD peaks indicate that the crystallographic crystalline phase remains the Cs₂In(I)In(III)Cl₆ charge-ordered phase, In⁺ might have already been oxidized to In³⁺ on the surface, and prevents the framework from collapsing through self-limiting surface oxidation of the crystals (34, 35). This is also consistent with the weaker peak intensity and peak broadening of the PXRD pattern after a few days of air exposure.

• Download figure
• Open in new tab
• Download powerpoint

Fig. 2.

Structure and valence state stability. (A and B) PXRD patterns of (A) as-synthesized Cs₂In(I)In(III)Cl₆ powder and (B) that after exposure to air for 0 to 18 d. (C and D) SXPS of fresh sample and that after exposure to air for 1 d.

Previous studies on the charge-ordered halide perovskites such as CsAuCl₃ and CsTlCl₃ claimed that the wide bandgap semiconducting properties in these crystals originated from the strong coupling of the charge carriers (25, 29). Herein, we utilize optical and electrical measurements to explore the properties of our Cs₂In(I)In(III)Cl₆ crystals. SI Appendix, Fig. S10 shows the absorbance spectrum at room temperature, indicating that the material has an optical bandgap of around 3.0 eV. From photoluminescence measurement (SI Appendix, Fig. S11), a broad emission peak centered at about 500 nm with very low intensity appears at room temperature. Further lowering the temperature down to 4 K, the emission intensity is increased 2-fold of its original intensity, but still quite broad and weak, indicating poor light-emitting property of the material. Furthermore, electrical characteristics of this crystal were studied via current (I)–voltage (V) measurements. Symmetric and linear I–V output characteristics in the dark (I_dark), under white-light illumination (I_light), and the photocurrent (I_ph = I_light – I_dark) are shown in SI Appendix, Fig. S12, respectively, indicating the semiconducting behavior of this material without external stress.

To test possible semiconductor-to-metal phase transition of this crystal at high pressures, we explored its optical properties under pressure using a 4-post symmetric DAC coupled with a hydrostatic pressure medium. DAC has been widely used to investigate the structural and electronic properties of halide perovskites under high pressures (10, 36⇓⇓⇓⇓–41). Raman measurement of Cs₂In(I)In(III)Cl₆ at ambient pressure (0 GPa) and room temperature (295 K) showed 1 prominent peak at 279 cm⁻¹, which could be assigned to the In(I)–Cl–In(III) mode (26, 32) and 2 other minor peaks at 143 and 189 cm⁻¹ (SI Appendix, Fig. S6). The Raman spectra did not show any significant temperature dependence from 295 K down to 77 K (SI Appendix, Fig. S13). We then performed pressure-dependent Raman spectroscopy at room temperature. The main Raman peak at 279 cm⁻¹ gradually shifts to higher wavenumber with the increase of pressure (Fig. 3A and SI Appendix, Fig. S14) and experiences an abrupt intensity drop at 13.2 GPa and finally disappears at 21.9 GPa, suggesting a metallic phase at and beyond this pressure. Meanwhile, the 2 minor Raman peaks at lower frequencies were too weak to be detected under high pressure. The Raman peaks fully recover after the pressure is released; hence the phase transition is reversible. As shown in Fig. 3B, the Raman peak shifts almost linearly with pressure at a rate of dω/dP = 3.7 cm⁻¹/GPa before it collapses at about 22 GPa. The integrated area also shows a drastic drop at ∼10 GPa and becomes nondetectable at >22 GPa (10). The Raman vibration modes softening can be regarded as an important evidence of a charge-ordering to non–charge-ordering phase transition (23, 28, 29). To better demonstrate the pressure-induced semiconductor–metal phase transition, pressure-dependent optical absorbance measurement at room temperature has been performed. As shown in Fig. 3C, the absorbance spectrum of Cs₂In(I)In(III)Cl₆ red shifts with pressure up to 13.9 GPa, above which absorption increases abruptly with a broadband metallic feature. In the raw transmittance spectra (SI Appendix, Fig. S15), the transmitted light intensity abruptly drops at 13.9 GPa and remains low up to 34.7 GPa. The pressure-dependent bandgap evolution of Cs₂In(I)In(III)Cl₆ is shown in Fig. 3D. The bandgap starts at about 2.9 eV under ambient pressure, and then gradually decreases to about 2.0 to 2.1 eV until 13.9 GPa and closes completely under higher pressures. A strong absorption was also observed in MAPbI₃ and charge-ordered CsAuI₃ at the transition pressure, consistent with the simultaneous electrical conductivity change near the same pressure, suggesting a semiconductor–metal phase transition (10, 27, 28). The good agreement between optical bandgap closure and electrical conductivity change detected in these 2 similar materials implies that experimentally the optical band gap closure is a reliable and robust clue of a metallic phase, especially when combined with Raman spectra change.

• Download figure
• Open in new tab
• Download powerpoint

Fig. 3.

Semiconductor–metal transition under high pressures. (A and B) Raman peak blue shifts and intensity (normalized integrated area) weakens with the increase of pressure and disappears completely at about 22 GPa. (C and D) Absorbance spectra (C) and bandgap shrink with the increase of pressure (D). After about 14 GPa, a broadband metallic absorption feature appears.

The pressure-induced bandgap changes and semiconductor–metal transition of the charge-ordered Cs₂In(I)In(III)Cl₆ were further interpreted by density functional theory (DFT) calculations. We built a 5-atom unit cell of perovskite structure with space group P4/m and a 10-atom unit cell of double perovskite structure with space group I4/m for the hypothesized non–charge-ordered CsInCl₃ (purely In²⁺) and charge-ordered Cs₂In(I)In(III)Cl₆, respectively. The calculated band structure and density of states are shown in SI Appendix, Fig. S16. The results show that, without external stress, the charge-ordered perovskite possesses an electronic bandgap of about 2.3 eV, while the non–charge-ordered system lies in a metallic state of 0-eV bandgap. Particularly, under ambient pressure, the charge-ordered semiconducting phase possesses a lower energy than the non–charge-ordered state, suggesting that the charge-ordered phase is the ground state. The calculated bandgap of the charge-ordered phase gets narrower with the increase of pressure and decreases smoothly to 0 eV at around 13 GPa (Fig. 4A). As pressure increases, the energies of both non–charge-ordered and charge-ordered phases increase, and the difference between the energies of the 2 phases decreases (Fig. 4B). We further analyzed the In–Cl bond length in the 2 phases under increasing pressure (SI Appendix, Fig. S17). The length difference between the 2 types of In–Cl bonds in the charge-ordered phase becomes smaller with the increase of pressure and their values approach the In–Cl bond length in the non–charge-ordered phase. The electronic delocalization of the In⁺ and In³⁺ centers contributes to the bandgap change and semiconductor-to-metal transition. To summarize, by combining the Raman peaks disappearance, bandgap closure, and DFT simulations, we conclude that the charge-ordered perovskite undergoes a gradual and reversible semiconductor–metal phase transition within the pressure range of 14 to 22 GPa.

• Download figure
• Open in new tab
• Download powerpoint

Fig. 4.

DFT modeling on the pressure-induced phase transition. (A and B) Bandgap (A) and average energy per atom changes (B) of the charge-ordered and non–charge-ordered states at high pressures. The charge-ordered system equilibrium energy (at 0 GPa) is lower than the non–charge-ordered case.