Metastable alloying structures in MAPbI3−xClx crystals

Chlorine incorporation engineering has been widely used in optoelectronic devices based on methylammonium lead iodide (MAPbI3) perovskites. However, the characteristics of I/Cl alloying structures in MAPbI3−xClx mixed-halide perovskites and their influences on the optoelectronic properties have been issues of a long-standing controversy. Here, we present a detailed study of the I/Cl alloying structures in MAPbI3−xClx (x = 0.0 to 0.3) single crystals. We found that a small amount of Cl can substitute for the iodide of the PbI3 inorganic lattice, leading to a phase transition from the tetragonal to cubic phase and anomalous cation dynamics evolution. Analyses based on time-dependent X-ray diffraction, 207Pb NMR, and 2H NMR indicate that the alloying structures of the MAPbI3−xClx crystals are metastable and decompose over time. In addition, the photocurrent response measurement of MAPbI3−xClx proved a close correlation between the alloying structures and photoelectric properties of the material. This work sheds light on the essential understanding of the I/Cl alloying structure and provides a plausible explanation for the controversy regarding the role of chloride ions in optoelectronic devices. A more complete understanding of alloy formation in an exciting class of optoelectronic materials has been developed by researchers in China, which could help improve solar-cell efficiency. Lead halide perovskites are used in low-cost, lightweight and flexible electronic devices. The properties of the perovskite can be modified to suit a specific application by doping, exchanging some atoms in the crystal for others, but this can make the structure unstable. Xue Lu Wang, Ye-Feng Yao and colleagues from the East China Normal University, Shanghai, tracked atomic structure changes in perovskite methylammonium lead iodide as the iodine atoms were partially replaced by chlorine atoms. The team showed that the state of chloride in MAPbI3−xClx crystals is metastable and the crystal lattice can transform from a cubic to a tetragonal structure over time. The characteristics of I/Cl alloying structures in MAPbI3-xClx mixed-halide perovskites and their influences on the optoelectronic properties have been issues of long-standing controversy. In this work, by using on the time-dependent X-ray diffraction, 207Pb and 2H NMR, the authors investigated the I/Cl alloying structures and its long term stability in MAPbI3-xClx (x = 0.0 to 0.3) single crystals. It was found that Cl can substitute for I of the PbI3 inorganic lattice, leading to the tetragonal to cubic phase transition. Moreover, it revealed that the alloying structures of the MAPbI3-xClx crystals are metastable and may decompose over time.


Introduction
The development of organic-metal halide perovskite materials with the formula ABX 3 , where A is generally either methylammonium (MA) or formamidinium (FA), B is Pb or Sn, and X is I, Br, or Cl, has rapidly progressed in recent years. These materials have been employed in a wide array of applications, such as in solar cells [1][2][3] , photodetectors 4 , light-emitting diodes (LEDs) 5 , transistors 6 , and lasers 7 , owing to their excellent optoelectronic performances, such as adjustable bandgaps 8 , high absorption coefficients 9 , and long charge-carrier diffusion lengths 10 . The power conversion efficiencies of perovskite solar cells have significantly increased in recent years, and a high power conversion efficiency of 25.2% was obtained last year for a solar cell based on a structurally optimized organic-metal halide perovskite 11 . Recently, halide alloying engineering has been developed as a critically important strategy for promoting the performance of optoelectronic devices based on organic-metal halide perovskite materials 12 . This strategy has been particularly applied to mixed-halide APbI 3−x Cl x perovskite materials [13][14][15][16] , and progress in this area has rapidly developed. This strategy has been demonstrated to not only improve the optoelectronic properties of the resulting perovskite materials but also enhance their structural stability and even be applicable to bandgap modulation 17,18 . However, despite this promising progress, a number of fundamental issues regarding the halide alloying engineering process remain unresolved. A particularly important issue involving how Cl atoms are incorporated into MAPbI 3−x Cl x materials as a halide alloying agent is a subject of strong debate 19,20 .
On one side of the above-discussed debate, many reports have provided evidence indicating that no Cl ions are embedded in the inorganic PbI 3 lattice. This suggests that Cl ions influence the optoelectronic properties of MAPbI 3−x Cl x perovskites only in an indirect way, and numerous indirect enhancement effects have been proposed. For example, Chen et al. 19 attributed the optoelectronic performance enhancement of MAPbI 3−x Cl x perovskites to the improved heterojunction interfaces induced by the Cl ion incorporation process. Lian et al. 20 demonstrated that Cl ion incorporation could improve the crystallinity of MAPbI 3−x Cl x perovskite materials and in turn increase the crystal growth rate. However, no Cl ions were detected in the crystal lattice. Fan et al. 21 proposed that residual PbCl 2 passivated defects in MAP-bI 3−x Cl x perovskite thin films and thereby improved their electrical properties. On the other side of the debate, a similar number of reports have provided evidence that Cl ions can be embedded in the inorganic PbI 3 lattice. For example, Zhang et al. 22 demonstrated the presence of Cl ions in the inorganic lattice of MAPbI 3−x Cl x nanofibers fabricated using a solution-assembly method. Zhang et al. 23 reported that Cl can be maximally doped into the MAPbI 3−x Cl x perovskite crystal lattice using an ion exchange−decomposition process. Min et al. 24 demonstrated that Cl ions can be introduced into the FAPbI 3 crystal lattice through the substitution of FA sites with methylenediammonium dichloride ions. In addition, the possible crystal structures of MAPbI 3−x Cl x perovskite materials have been investigated by ab initio computer simulation studies, and the potential optoelectronic properties of the proposed structures have been evaluated 25,26 . However, previous studies mainly focused on perovskite films, whose lack of stability and the difficult distinction between the surface and bulk are great challenges for accurate detection by various instruments. Moreover, few studies have rigorously examined the I/Cl alloying structures and their long-term stability.
The present study seeks to address this long-standing controversy by presenting a detailed study of the I/Cl alloying structures observed in a series of MAPbI 3−x Cl x (x = 0, 0.05, 0.1, 0.3, 0.4, and 0.5) single crystal perovskite samples over time. Here, single-crystal samples are much more ideal than powder samples for the proposed study owing to their greatly reduced crystal defect concentrations and fewer crystal boundaries 10,20 . The structure of the single crystals is observed to change from tetragonal to cubic with increasing x. Characterization of the samples by means of X-ray diffraction (XRD), X-ray photoelectron spectroscopy (XPS), 207 Pb nuclear magnetic resonance (NMR), and 2 H NMR reveals that a small amount of Cl can be incorporated into the iodide sites of the inorganic PbI 3 lattice, leading to a phase transition from the tetragonal to cubic phase and a series of anomalous cation dynamics evolutions including a slower rotation rate and a more disordered dynamics tendency. Analyses based on time-dependent XRD, 207 Pb NMR, and 2 H NMR indicate that the I/Cl alloying structures of the MAPbI 3−x Cl x crystals are metastable and disappear over time at room temperature. In addition, we explored the photoelectric properties of MAPbI 3−x Cl x single-crystal photodetectors and found that the photocurrent response obviously decreases over time, even though Cl incorporation helps boost the photocurrent response. This long-term instability indicates that different conclusions regarding the effect of Cl ions on the properties of MAPbI 3−x Cl x perovskites may arise over different time scales, which provides a plausible explanation for the long-standing controversy regarding the effect of Cl ions on the performance of optoelectronic devices based on MAPbI 3−x Cl x. Crystallization of CH 3 NH 3 PbI 3-x Cl x /CD 3 NH 3 PbI 3-x Cl x single crystals MAI, MACl, and PbI 2 were mixed and dissolved at different molar ratios into GBL as the precursor solution. DMF was chosen as the second solvent when the concentration of MACl exceeded 0.2 M because of the low solubility of MACl in GBL. During dissolution, the precursor solution was ultrasonicated and heated to 60°C. Next, a filter (0.2 µm) was used to filter invisible impurities. After the above solution was heated and kept at 110°C for 12 h, MAPbI 3-x Cl x single crystals were harvested. The X-ray 2θ scan on the maximal facet of the typical MAPbI 3-x Cl x single crystals showed two sharp diffraction peaks, suggesting a well-structured singlecrystalline nature ( Supplementary Fig. S1). Deuterated MAPbI 3-x Cl x was synthesized using a similar approach to that mentioned above, except that CD 3 NH 3 I and CD 3 NH 3 Cl were used as cation sources.

Materials
It is noteworthy that the crystals could be obtained by the above method only for x ≤ 0.5. For x > 0.5, no crystal was synthesized. However, we found that a new trace MAPbCl 3 phase grew together with the double halogen perovskite component when x was above 0.3, as proven by the XRD pattern in Supplementary Fig. S2 and 207 Pb NMR spectrum in Supplementary Fig. S3. As a result, this article focuses only on the samples with x ≤ 0.3.
Crystallization of MAPbI 3-x Br x single crystals CH 3 NH 3 I, CH 3 NH 3 Br, and PbI 2 were mixed and dissolved at different molar ratios into GBL as the precursor solution, and a filter (0.2 µm) was used to filter invisible impurities. After the above solution was heated and kept at 100°C for 12 h, MAPbI 3-x Br x single crystals were harvested. The static 207 Pb NMR spectra of the MAP-bI 3−x Br x (x = 0, 0.2, 1, 3) samples are shown in Supplementary Fig. S4.

Crystallization of MAPbBr 3 and MAPbCl 3 /CD 3 NH 3 PbI 3 single crystals
To grow perovskite CH 3 NH 3 PbBr 3 single crystals, MABr and PbBr 2 (1:1) were mixed and dissolved in DMF, with the concentration controlled at 1.0 M. A filter (0.2 µm) was used to filter invisible impurities. After the above solution was heated and kept at 80°C for 6 h, MAPbBr 3 single crystals were harvested. To grow the single-crystalline perovskite CH 3 NH 3 PbCl 3 , MACl, and PbCl 2 (1:1) were mixed and dissolved in a mixed solution of DMSO and DMF (1:1), with the concentration controlled at 2.0 M. A filter (0.2 µm) was used to filter invisible impurities. After the above solution was heated and kept at 80°C for 6 h, MAPbCl 3 single crystals were harvested. Deuterated MAPbI 3 was synthesized using a similar approach to that mentioned above except that CD 3 NH 3 I and CD 3 NH 3 I were used as cation sources.
Powder X-ray diffraction (XRD) patterns XRD measurements were carried out on a Rigaku Smart Lab (X-ray source: Cu Kα; λ = 1.54186 Å) operated at 40 kV and 30 mA. MAPbI 3-x Cl x powders were made by grinding a large piece of crystal into a fine powder.

XPS spectra
XPS spectra were measured by an X-ray photoelectron spectrometer (PHI Quantro SXM, ULVAC-PHI) with an Ar ion gun sputtering the surface. Before measuring the sample, the crystal surface was polished in a glove box filled with inert nitrogen gas to expose a fresh surface.
Static 207 Pb solid-state NMR 207 Pb NMR experiments were carried out on a 300 M NMR instrument operating at 62.73 MHz for 207 Pb. A Bruker two-channel static PE probe with a homemade 4 mm coil was used to record the spectra. The 207 Pb spectra were acquired using a single pulse excitation pulse sequence. The excitation pulse in the experiments was 2.5 μs at an RF field strength of 100 kHz. The recycle delay was 1 s. All 207 Pb NMR spectra were referenced to PbMe 4 (δ ( 207 Pb) = 0 ppm) by setting the 207 Pb peak of nonspinning solid MAPbCl 3 measured at 293 K to −647.5 ppm.
1 H solid-state MAS NMR 1 H NMR spectra were recorded with a 600 M NMR instrument operating at 600.44 MHz. The spinning rate was set to 10 kHz. The recycle delay was set to 2 s. All 1 H chemical shifts were calibrated using adamantane (δ = 1.85 ppm).
H NMR experiments were carried out on a 300 M NMR instrument operating at 46.07 MHz for 2 H. A Bruker two-channel static PE probe with a homemade 4 mm coil was used to record the spectra. The 2 H spectra were acquired using the solid echo sequence. The 2 H pulse width was 2.5 μs at an RF field strength of 100 kHz. The 2 H patterns were simulated via the weblab (http:// weblab.mpip-mainz.mpg.de/weblab/).

Preparation and characterization of MAPbI 3−x Cl x single crystals
The materials and methods employed for preparing the MAPbI 3−x Cl x single crystal samples used in the present study are outlined in detail in the Materials and Methods. The preparation procedures are briefly described as follows. The inverse temperature crystallization (ITC) method was used to prepare MAPbI 3 (i.e., x = 0) single crystal perovskites 27,28 . For MAPbI 3−x Cl x single crystal samples, x was varied by adjusting the proportions of methylamine hydrochloride (MACl) and methylamine hydroiodide (MAI) employed during synthesis while retaining a constant concentration of PbI 2 . Here, MACl was first added into a pure gamma-butyrolactone (GBL) solvent. However, this often yielded a turbid solution due to the poor solubility of MA in GBL, as shown in Fig. 1A. Nonetheless, the solution became transparent again once MAI and PbI 2 were added, as shown in Fig. 1B. Second, the solution was heated to 383 K to promote the growth of single crystals. Representative single-crystal MAPbI 3−x Cl x samples are shown in Fig. 1C and Supplementary Fig. S1 against a cm 2 square inscribed background to indicate the sample sizes. We note that MAPbCl 3 impurity phases are synthesized together with MAPbI 3−x Cl x crystals for x > 0.3 (Supplementary Figs. S2 and S3). Hence, x = 0.3 was the largest value considered in the present study. Notably, reflecting the influence of Cl − doping on the crystal structure, the shapes of the single crystal samples change with increasing Cl − concentration. The changes in the crystal shape do not indicate a change in the crystal symmetry or the formation of a polycrystalline component. Similar phenomena have been observed in MAPbI 3 single crystals 29 . Figure 1D, E presents the XRD patterns and enlarged views of the patterns in the 2θ range from~28 to 29°, respectively, for MAPbI 3−x Cl x (x = 0, 0.05, 0.1, and 0.3) single crystal samples at room temperature. We note the presence of a clear splitting of the two peaks indicative of the (004) and (220) facets of the MAPbI 3 sample at 2θ values ranging from 28 to 29°. This is a characteristic feature of the diffraction peaks obtained for MAPbI 3 perovskite single crystals in the tetragonal phase at room temperature 30 . These (220) and (004) peaks are observed to shift gradually toward increasing 2θ values with increasing x in the range of 0.05-0.3. This gradual shift indicates that Cl doping causes the dimensions of the crystal lattice cell to progressively decrease with increasing x. We also note that the (220) and (004) diffraction peaks gradually merge with increasing x, indicating that the MAPbI 3−x Cl x single crystal lattice structure gradually changes from the tetragonal phase to the cubic phase with increasing x, as verified in past studies 31 .
The atomic concentrations of Cl atoms in the MAP-bI 3−x Cl x single crystal samples were evaluated using highresolution XPS. Prior to conducting XPS analysis, the surface of each single crystal sample was mechanically polished to avoid the influence of residual solvent on the XPS results. Figure 1F presents the XPS Cl 2p core-level spectra of the Cl atoms in the samples, which are composed of the Cl 2p 1/2 and Cl 2p 3/2 doublet states. We note that the Cl 2p signal intensity clearly increases with increasing x. The atomic Cl concentrations in the samples were determined relative to the atomic Pb concentrations from the areas under the Cl 2p and Pb 4f curves of the XPS core-level spectra obtained at multiple points of each sample surface and scaled according to their respective sensitivity factors. According to the chemical formula MAPbI 3−x Cl x , the atomic ratio of Pb/Cl should be equal to x. The Cl/Pb atomic ratios obtained through four repeated tests on the same batch of samples are plotted in Fig. 1G with respect to x. Clearly, the Cl/Pb atomic ratio increases with increasing x, as expected. However, the Cl/ Pb atomic ratio in the samples does not scale according to x, and only a Cl/Pb atomic ratio of 0.138 ± 0.021 is obtained for x = 0.3, which is much less than the expected value. This indicates that a substantial proportion of Cl ions failed to enter the single crystal and were accumulated in the precursor during crystal growth, further inhibiting crystal growth. The results also explain why the sample size decreases with increasing x (Fig. 1C).  also given in Fig. 2A. The spectra of the Cl-and Br-doped materials include curves (solid red lines) based on the two Gaussian curves (dashed black lines) given in each spectrum. The 207 Pb ssNMR spectrum of the MAPbI 3 crystal sample presents a very broad symmetrical peak with a chemical shift (δ) of 1423 ppm. In contrast, the spectrum of the MAPbI 2.7 Cl 0.3 sample has a main peak consistent with that of the MAPbI 3 spectrum, and a clear shoulder peak is observed at 1117 ppm. The origin of this shoulder peak can be assigned based on a comparison with the shoulder peak in the 207 Pb ssNMR spectrum observed for the MAPbI 2.9 Br 0.1 single crystal sample at 1176 ppm, which was assigned to a PbI 5 Br alloying structure in a past study 32,33 . Here, the PbI 5 Br alloying structure includes five iodine atoms and one bromine atom connected to a Pb atom to form an octahedron. We thus tentatively assign the shoulder peak of the MAPbI 2.7 Cl 0.3 spectrum to the PbI 5 Cl unit. If this assignment is valid, then it would indicate that Cl atoms can replace I atoms in the crystal lattice in a manner similar to that observed in the MAPbI 2.9 Br 0 . 1 crystal. Details regarding the 207 Pb ssNMR spectrum analysis are outlined in Section S2, in conjunction with Supplementary Figs. S4-6.
An alternative method of detecting changes in the lattice structure of MAPbI 3 perovskite materials is to probe the dynamic changes in the MA cations embedded in the inorganic lattice by means of 2 H NMR analysis 34 . This approach relies on the strong coupling between the MA cations and the surrounding crystal lattice 35 . Here, small changes in the inorganic lattice cell have been reported to induce significant changes in the dynamics of the embedded MA cations 26,36 . For this purpose, MAPbI 3−x Cl x (x = 0, 0.05, 0.1, 0.2, and 0.3) samples were synthesized in which the methyl group was fully deuterated ( Supplementary Fig. S7). In addition, powder samples were used in the 2 H NMR experiments to ensure an adequate signal. Figure 2B presents the solid-state 2 H NMR spectra (solid black lines) of the deuterated samples. The figure also includes spectra (solid red lines) simulated according to a method discussed later. For MAPbI 3 , the 2 H pattern exhibits a typical Pake doublet lineshape, and the lineshapes of the signals vary significantly with increasing x, indicating gradual changes in the dynamics of the MA cations in the samples, which, as discussed, can be ascribed to changes in the inorganic lattice cell environment of the MA cations.
The motion of MA cations in organic-metal halide perovskite crystals has been studied by several groups 34,37,38 . These past studies have demonstrated that details regarding the dynamic changes in the MA cations in these perovskite samples can be obtained by analyzing the lineshapes of 2 H NMR spectra. Therefore, we applied the MA cation motion model proposed by Franssen et al. 37 Table S1. This process is illustrated here in Fig. 2C and described briefly as follows. An MA cation in the tetragonal phase of MAPbI 3 can undergo rotation about two axes, including intrinsic axial rotation (rotation 1) about the Cn axis along with the C-N bond formed between the CH 3 and NH 3 components of the MA cation and secondary axial rotation (rotation 2) at a rate Ω about the Rc axis, which lies at a specific angle θ with respect to the Cn axis. The θ values are assumed to be distributed about a mean value θ 0 according to a Gaussian distribution with standard deviation σ, as indicated by the equation and graph in Fig. 2C. Applying this model with 2σ = 1.5°to the 2 H spectrum obtained for the MAPbI 3 sample in Fig. 2B yields a θ 0 of 58.9°, and the difference between the doublet peaks of the Pake lineshape (v QS ) is 4 kHz. We note that the simulated 2 H NMR spectrum is in good agreement with the actual spectrum. For MAPbI 2.95 Cl 0.05 , the observed 2 H NMR spectrum lineshape undergoes obvious changes, where the separation between the two peaks of the Pake doublet lineshape decreases significantly. Applying the model with 2σ = 1.5°to the 2 H spectrum MAPbI 2.95 Cl 0.05 yields a θ 0 of 57.5°. The slight reduction in θ 0 with increasing x stems from the slight phase transition tendency from the tetragonal phase to the cubic phase revealed by XRD in Fig. 1E. With a further increase in x, we note that the 2 H patterns gradually transform into a single broad peak for MAPbI 3−x Cl x (x = 0.1, 0.2, and 0.3). The lineshape analysis results indicate that the appearance of a broad peak can be attributed to a decrease in Ω, accompanied by an increase in σ. Based on the strong coupling between the MA cations and the inorganic lattice, we can conclude that these changes in the motion of MA cations are associated with the lattice contraction and lattice distortion caused by the random and partial substitution of iodine ions with Cl ions in the inorganic lattice. We also note that compared with the XRD and 207 Pb NMR results, the 2 H NMR spectra appear to exhibit a much greater sensitivity to structural changes in the inorganic framework, indicating the great potential of this technique for probing structural changes in MAPbI 3 perovskite materials. Figure 3A presents the time-dependent XRD patterns (solid black lines) of the MAPbI 2.7 Cl 0.3 single crystal sample obtained over a 2θ range of 27.6°to 28.8°at 0, 5, 10, and 20 days after synthesis. The samples were stored in a nitrogen-filled glove box between measurements to eliminate the influence of water vapor and oxygen as much as possible. The solid red line in the figure corresponds to the XRD pattern of the MAPbI 3 single crystal sample originally shown in Fig. 1E. In comparison with the XRD pattern of the MAPbI 3 single crystal sample, we note that the XRD pattern of the originally cubic MAP-bI 2.7 Cl 0.3 single-crystal shown in Fig. 3A gradually transitions over the 20-day period to an XRD pattern very similar to that observed for the MAPbI 3 single crystal sample with a tetragonal structure. This strongly indicates that the MAPbI 2.7 Cl 0.3 crystal underwent a gradual solid-phase transition from the cubic phase to the tetragonal phase. Figure 3B presents the time-dependent 207 Pb NMR results obtained for the MAPbI 2.7 Cl 0.3 single crystal sample at 0, 5, 10, and 20 days after synthesis. As before, the samples were stored in a nitrogen-filled glove box at all times between testing. The solid red line in the figure corresponds to the 207 Pb NMR spectrum of the MAPbI 3 single crystal sample originally shown in Fig. 2A  gradually transitions over the 20-day period to a symmetrical 207 Pb NMR spectrum very similar to that observed for the MAPbI 3 single crystal sample, with equivalent peak shifts of 1423 ppm. Here, the shoulder observed in the as-synthesized sample was greatly diminished after 5 days and completely disappeared after 10 days. As discussed above, the shift in the main peak and the appearance of the shoulder can be related to the formation of PbI 5 Cl alloying structures. Accordingly, the reverse shift of the main peaks and the disappearance of the shoulder thus strongly indicate that the PbI 5 Cl alloying structures gradually changed and disappeared over time. Figure 3C presents the time-dependent 2 H NMR results obtained for the deuterated MAPbI 2.7 Cl 0.3 powder sample at 0, 5, 10, and 20 days after synthesis. As before, the samples were stored in a nitrogen-filled glove box at all times between testing. The solid red line in the figure corresponds to the 2 H NMR spectrum of the MAPbI 3 deuterated powder sample originally shown in Fig. 2B. In comparison with the 2 H NMR spectrum of the MAPbI 3 sample, we note that the 2 H NMR spectrum of the MAPbI 2.7 Cl 0.3 single-crystal shown in Fig. 3C gradually transitions over the 20-day period to a standard Pake doublet lineshape similar to that observed for the MAPbI 3 single crystal sample. Here, the broad peak observed in the as-synthesized sample became considerably less broad after 5 days, and Pake doublet peaks began to emerge after 10 days. However, we note that the lineshape of the 2 H NMR spectrum acquired after 20 days failed to fully return to that of the MAPbI 3 spectrum, indicating that some residual distortion remains in the inorganic lattice of the sample.

Time-dependent structural characterization of MAPbI 3−x Cl x materials
The 2 H NMR spectra of the as-synthesized MAP-bI 2.7 Cl 0.3 sample were also acquired at different temperatures (i.e., 294, 310, 330, and 340 K; see Supplementary Fig. S12 in Supporting Information). It was observed that the lineshapes of the 2 H NMR spectra remained nearly unchanged before and even after exceeding the phase transition temperature of MAPbI 3 (337 K). This result indicates that the MAPbI 2.7 Cl 0.3 structure does not undergo a similar phase transition as that of MAPbI 3 . This may also be associated with the doping of Cl atoms in the inorganic lattice.
The combined observations above clearly reveal many intriguing details regarding the structural transformations of MAPbI 2.7 Cl 0.3 over time. As discussed, the timedependent XRD results indicate that the MAPbI 2.7 Cl 0.3 crystal undergoes a gradual solid-phase transition from the cubic phase to the tetragonal phase within 20 days. In addition, while the time-dependent changes in the chemical shift and lineshape of the 207 Pb spectrum strongly indicate that the PbI 5 Cl alloying structures present in the as-synthesized sample gradually disappear over time, the 207 Pb NMR results cannot reveal whether the PbI 5 Cl alloying structures have completely disappeared after 20 days due to the limited sensitivity of 207 Pb NMR signals. The time-dependent XRD results suffer from a similar limitation in that XRD is not sensitive to local changes in the crystalline structure. Fortunately, this issue is clarified to some extent by the fact that the lineshape of the 2 H NMR spectrum acquired after 20 days did not return fully to that of the MAPbI 3 spectrum. As discussed, this indicates that some residual distortion remained in the inorganic lattice of the sample. Accordingly, this suggests that some amount of alloying structures remains in the sample but not at a sufficiently high level to sustain the cubic phase of the crystal or to greatly affect the local Pb environment overall. Clearly, 2 H NMR represents a more sensitive technique for probing changes in the inorganic lattice of MAPb 3 halide perovskite crystals due to Cl doping.

Photoresponse of photodetectors based on MAPbI 3−x Cl x single-crystal absorbers
The influence of Cl doping on the optoelectronic properties of MAPbI 3−x Cl x materials was illustrated by fabricating five planar-type photodetectors with appropriately shaped and finished MAPbI 3−x Cl x (x = 0, 0.05, 0.1, 0.2, and 0.3) single crystal absorbers. An image of a representative photodetector device is shown in Fig. 4A. Here, interdigitated Au wire electrodes (0.5 mm in length, 40 µm in width, and approximately 100 nm thick) were evaporated on the well-polished (100) plane of each single crystal. The effective absorber area of each detector was 0.1 mm 2 . The current responses were measured under a bias voltage of 1 V with illumination provided by a 405 nm wavelength continuous laser beam with an optical power density of 0 to 100 mW/cm 2 . Figure 4B shows the photocurrent responses of the photodetectors employing the as-synthesized single-crystal absorbers as a function of the incident light intensity. The results demonstrate that the photocurrent increased almost linearly with increasing incident light intensity for each single crystal photodetector. Meanwhile, the photocurrent of the photodetectors at a given incident light intensity was positively correlated with the Cl doping ratio (i.e., x), where the photocurrent increased with increasing x. Figure 4C presents the photocurrents of the devices obtained with respect to time under continuous laser illumination operating with an optical power density of 70 mW/cm 2 . As anticipated, all of the photodetectors exhibited a very good photoresponse, where the photocurrent immediately increased and decreased at the beginning and end of each laser pulse, respectively. We also note that the photocurrent obtained at each laser pulse noticeably increased with increasing x. For example, the photocurrent of the device based on MAPbI 3 was only 3 µA, whereas that of the device based on MAPbI 2.7 Cl 0.3 was 23 µA.
However, a comparison of the photocurrents with respect to the optical power density of the incident illumination for photodetectors based on the as-synthesized MAPbI 3 , MAPbI 2.95 Cl 0.05 , and MAPbI 2.7 Cl 0.3 singlecrystal absorbers, shown by the filled black squares in Fig. 4D, E, and F, respectively, and the same photodetectors after storage in a nitrogen-filled glove box for 20 days (filled red circles) demonstrates that the optoelectronic properties of the Cl-doped perovskite materials were not stable and degraded after 20 days of storage to photocurrents essentially commensurate with that of MAPbI 3 . The photoresponse data obtained for photodetectors based on MAPbI 2.9 Cl 0.1 and MAPbI 2.8 Cl 0.2 behaved similarly ( Supplementary Fig. S13). In contrast, the photoresponse data obtained for the MAPbI 3 -based photodetector exhibited no obvious changes after storage for 20 days, which demonstrates the stability of the MAPbI 3 structure. The above results support the previous time-dependent analyses based on XRD, 207 Pb NMR, and 2 H NMR, which demonstrated that the crystal lattices of the MAPbI 3−x Cl x (x = 0.05, 0.1, 0.2, and 0.3) single crystals changed from cubic to tetragonal after 20 days, and the Cl alloying structures in the Cl-doped materials gradually disappeared.
As discussed, the roles of Cl ions in MAPbI 3−x Cl x perovskites are a very controversial issue. A number of theoretical studies have demonstrated that the substitution of Cl for I in the inorganic lattice can have an inherently positive effect on the electronic properties of MAPbI 3−x Cl x perovskites by reducing electron-phonon coupling and electron−hole recombination 25 , enhancing the intensity of the p orbital of I ions in the HOMO state 39 , or enhancing electron localization and the charge-carrier diffusion length 26 . However, other studies have demonstrated that Cl substitution may have a negative effect on the properties of MAPbI 3−x Cl x perovskites 19,22 . For example, the disassociation of Cl atoms from the inorganic lattice can generate a large amount of anionic defects in the lattice, which serve as deep-level defects that may strongly hinder the generation and diffusion of charge carriers 40 . Here, the long-term instability of MAPbI 3−x Cl x alloying structures reported in the present work sheds some light on this controversial issue. In the as-synthesized state of MAPbI 3−x Cl x single crystals, the substitution of Cl for I contributes to enhanced electronic properties of the Cldoped perovskites and explains the origin of their enhanced optoelectronic properties. Meanwhile, the Cl atoms gradually dissociate from the inorganic lattice over time and may result in the generation of anionic vacancies in the lattice. As a result, the optoelectronic properties of the MAPbI 3−x Cl x single crystals deteriorate over time. This long-term instability indicates that different conclusions regarding the effect of Cl ions on the optoelectronic properties of MAPbI 3−x Cl x perovskites may arise over different time scales, which provides a plausible explanation for the long-standing controversy regarding the effect of Cl ions on the performance of optoelectronic devices based on MAPbI 3−x Cl x .
In addition, the stability issues of mixed-halide perovskites, particularly APbI 3−x Br x materials, have attracted some attention recently 41,42 . The results revealed that APbI 3−x Br x can reversibly segregate into I-rich and Br-rich domains under continuous visible illumination 43,44 . For MAPb 3−x Cl x , new Cl-rich domains, such as MAPbCl 3 , were not discovered in either the XRD patterns or the 207 Pb NMR spectra. We thus hesitate to directly correlate the instability of MAPb 3−x Cl x over time to the halide segregation present in APbI 3−x Br x . However, the halide segregation phenomenon is possibly related to some metastable states present only under illumination. Therefore, an in situ study of MAPb 3−x Cl x is worth investigating if one wants to probe the halide segregation phenomenon. Such an in situ study is ongoing in our laboratory.

Conclusion
In this article, we report a detailed study of MAPbI 3−x Cl x (x = 0.0 to 0.3) single crystal structures based on a combination of XRD, XPS, and solid-state NMR. The results revealed that a small amount of Cl atoms could substitute for iodide atoms in the inorganic lattice of Cl-doped crystals. This substitution distorted the inorganic lattice, resulting in a transformation from the tetragonal phase to the cubic phase and a cation dynamics change with increasing x. However, the Cl/I alloying structures were demonstrated to be unstable over a long term of~20 days. Time-dependent XRD and solid-state NMR results demonstrated that the Cl atoms gradually dissociated from the Cl/I alloying structures. This Cl dissociation induced a counter transformation from the cubic phase to the tetragonal phase. The photocurrents obtained for photodetectors based on as-synthesized MAPbI 3−x Cl x (x = 0, 0.05, 0.1, 0.2, and 0.3) absorbers linearly increased with increasing x, but the photocurrents obtained for the same photodetectors after 20 days of storage degraded to photocurrents commensurate with that of MAPbI 3 , which supported the results of analyses based on the time-dependent structural transformations of MAPbI 3−x Cl x single crystals occurring over long time periods. This indicates that the optoelectronic properties of the Cl-doped perovskite materials are strongly correlated with the presence of Cl/I alloying structures. Moreover, the observed long-term instability indicates that different conclusions regarding the effect of Cl ions on the optoelectronic properties of MAPbI 3−x Cl x perovskites may arise over different time scales, which provides a plausible explanation for the long-standing controversy regarding the effect of Cl ions on the performance of optoelectronic devices based on MAPbI 3−x Cl x .

Data availability
All data needed to evaluate the conclusions in the paper are present in the paper and/or the Supplementary Materials. Additional data related to this paper may be requested from the authors.