Unassisted self-healing photocatalysts based on Le Chatelier’s principle

Single-particle observations of crystal destruction and self-healing

Halide perovskites are unstable in the presence of humidity^24,25, oxygen^26,27, and heat^28,29. However, when they are in a dynamic equilibrium state, the effects of these stimuli may be mitigated because the saturated solution acts as a passivating medium. Therefore, we employed single-particle fluorescence microscopy to directly monitor structural changes in individual perovskite crystals in response to light irradiation. This technique allows the evaluation of nanoscopic heterogeneities in chemical reactions at solid-liquid interfaces, which are typically obscured in ensemble-averaging bulk measurements^30,31,32.

Figure 2a depicts the experimental setup for single-particle observation using a home-built inverted fluorescence microscope system^33,34. MAPbBr_xI_3−x microcrystals (where MA⁺ = CH₃NH₃⁺) were spin-coated onto a cleaned cover glass to serve as seed crystals. A supersaturated perovskite solution was prepared by heating a suspension containing the target crystals at 80 °C, which was immediately dropped onto the substrate. As the temperature of the solution decreased, the perovskite crystals reprecipitated owing to the reduced solubility. Few studies have been conducted on the synthesis of perovskites in aqueous media^35,36,37. Therefore, we investigated the correlation between the halide composition of the prepared solution and that of the resulting crystals. Interestingly, we found that iodide anions tended to be incorporated into the crystals more readily than bromide anions (Supplementary Fig. 1a). A better correlation was observed between the iodide/lead ratios of the prepared solution and those of crystals (Supplementary Fig. 1b). A discussion of this tendency is provided in Supplementary Note 1. After allowing the crystals to grow sufficiently, we initiated imaging of the crystal morphology and spectroscopic measurements (Supplementary Fig. 1c). Figure 2b–f illustrates optical images of the obtained MAPbBr_xI_3−x crystals under dynamic equilibrium. The halide compositions were determined based on data from X-ray diffraction (XRD)³⁸ (Supplementary Fig. 1d). Photoluminescence (PL) imaging of individual perovskite crystals with various halide compositions revealed that the size of the MAPbBr₃ and MAPbI₃ crystals remained unchanged, or slightly decreased, or increased during 405-nm continuous wave (CW) laser irradiation (Supplementary Figs. 2 and 3). However, in the case of the mixed-halide MAPbBr_2.8I_0.2 crystals, the crystal morphology changed drastically when the excitation density was at least 780 mW·cm⁻² or higher at the sample surface (Fig. 2g–i and Supplementary Movie 1), and we refer to this morphological change as crystal destruction. When the damaged crystals in the aqueous solution were left in the dark after crystal destruction, self-healing behavior was observed (Fig. 2j, k and Supplementary Movie 1). This finding indicates that the proposed self-healing mechanism based on dynamic equilibrium applies to perovskites.

a Experimental setup for single-particle measurements. The experimental setup is based on a wide-field fluorescence microscope system. Both PL and transmitted light were captured using the same objective lens. b–f Optical images of mixed-halide MAPbBr_xI_3−x perovskites with various x-values in aqueous solution. g–k Optical transmission images in aqueous solution. The inset labels indicate the time after the start of photoirradiation. Excitation was provided for the first 300 s, after which photoirradiation was stopped. A 405-nm CW laser (ca. 780 mW·cm⁻² at the sample surface) was used as the excitation source.

Mechanisms of crystal destruction and self-healing

Next, we investigated the mechanisms underlying the observed crystal destruction and self-healing behaviors. To determine why only the mixed-halide MAPbBr_2.8I_0.2 crystals were destroyed, we conducted color imaging of the perovskites under laser irradiation. Figure 3a, b–d displays the optical transmission image of a MAPbBr_2.8I_0.2 crystal before laser irradiation and the PL images of MAPbBr_2.8I_0.2 under 405-nm CW laser irradiation, respectively. Initially, the MAPbBr_2.8I_0.2 crystal exhibited a weak green emission. As the laser irradiation time increased, the edges of the crystal began to exhibit red emission. Finally, a bright red emission was observed from the entire crystal. Spectral measurements of a specific region (approximately 2 µm² in area) of the crystal revealed that MAPbBr_2.8I_0.2 initially exhibited PL at approximately 560 nm. As the irradiation time increased, however, this peak exhibited a slight blue shift and a broad emission band appeared at longer wavelengths. The intensity of the latter component increased and red-shifted, eventually reaching a maximum at approximately 700 nm (Fig. 3e and Supplementary Fig. 4). These characteristic PL changes were observed repeatedly (Supplementary Fig. 5) and were attributed to light-induced phase segregation³⁹. To explain the origin of phase segregation, several models have been proposed⁴⁰, including those based on the miscibility gap due to thermodynamically unstable mixed-halide states, polaron-induced lattice strains that increase the mixing enthalpy of halide anions, and electric field-driven ion migration induced by the trapping of charge carriers at defects. In any cases, based on the band structures of perovskites, photogenerated charge carriers migrate from bromide-rich domains to iodide-rich domains with a narrower band gap, leading to the appearance of red-shifted PL (Supplementary Fig. 6). Similar PL features related to phase segregation have been reported elsewhere^41,42,43; however, these studies did not involve conditions of aqueous dynamic equilibrium. We also conducted the same measurements on MAPbBr_0.3I_2.7 and MAPbBr_1.3I_1.7 (Supplementary Fig. 7). MAPbBr_1.3I_1.7 exhibited a similar red-shifted PL, followed by crystal destruction and self-healing reactions. In contrast, MAPbBr_0.3I_2.7 did not show such behavior and retained their original morphology without any damage. Furthermore, we examined whether phase segregation induces the destruction of crystals with extreme halide compositions. In this experiment, a supersaturated HI (HBr) solution was dropped onto a MAPbBr₃ (MAPbI₃)-coated cover glass. Hereafter, we refer to each resulting crystal as MAPbI₃ (MAPbBr₃) with a small amount of bromide (iodide) based on the PL peak at approximately 760 nm (540 nm) (Supplementary Figs. 8a and 9a). Upon exposure to light, the morphology of MAPbI₃ with a small amount of bromide remained unchanged (Supplementary Fig. 8b, c), whereas the MAPbBr₃ crystals with a small amount of iodide were destroyed in a manner similar to that of MAPbBr_2.8I_0.2 (Supplementary Fig. 9c, d). As illustrated in Supplementary Fig. 9b, the PL spectrum of MAPbBr₃ with a small amount of iodide not only exhibits dominant green PL at approximately 540 nm but also weak red PL at approximately 700 nm, indicating that light-induced phase segregation occurred even with an extremely small amount of iodide anions. These results indicate phase segregation induces the crystal destruction. Therefore, self-healing effect is not exclusive to the MAPbBr_2.8I_0.2, it can be applicable to other composition perovskites.

Optical transmission and PL images of MAPbBr_2.8I_0.2 in aqueous solution before (a) and during photoirradiation (b–d). The inset labels show the time elapsed from the start of light irradiation. A 405-nm CW laser (ca. 1.21 W·cm⁻² at the sample surface) was used as the excitation source. e Spectral changes of the MAPbBr_2.8I_0.2 crystal in aqueous solution under irradiation. A 405-nm pulsed laser (ca. 8×10⁻¹⁷ J·pulse⁻¹) was used as excitation source. f Schematic illustration of the crystal destruction mechanism induced by halide phase segregation. g Pb 4 f XPS spectra of MAPbBr_2.8I_0.2 before and after irradiation, and after self-healing. A 405-nm LED (ca. 350 mW·cm⁻²) was used as excitation source. h XRD patterns of MAPbBr_2.8I_0.2 in aqueous solution before and after irradiation. A 405-nm LED (ca. 350 mW·cm⁻²) was used as excitation source. Asterisks indicate characteristic peaks of metallic Pb (PDF card: 00-004-0686). i Temporal changes in the Pb⁰ peak area under irradiation (yellow-shaded region) and after stopping irradiation (gray-shaded region). The error bars represent the standard deviation.

Based on the aforementioned results, we conclude that phase segregation is essential for crystal destruction (Fig. 3f). Prior to light irradiation, halide anions are uniformly distributed within the crystal. Upon photoexcitation of the mixed-halide perovskites, phase segregation mainly occurs in the regions near the surface, considering the limited penetration depth ( ~ 120 nm) of the excitation light⁴⁴. Holes are trapped by chemical species such as halide anions in solution²³, leading to the accumulation of excess electrons in the localized iodide-rich regions. The accumulated electrons reduce the perovskites, leading to crystal destruction and morphological changes. Such significant crystal destruction was not observed in air, where no halide anions were present. The MAPbI₃ with a small amount of bromide did not decompose because the charge carriers remained within the iodide-rich regions that constitute most of the crystal (Supplementary Fig. 10). This proposed mechanism is supported by simultaneous imaging of the PL color and crystal morphology of MAPbBr_2.8I_0.2 crystals (Supplementary Fig. 11). Upon laser irradiation, MAPbBr_2.8I_0.2 emitted red PL, attributed to the iodide-rich domains formed by phase segregation. Once crystal destruction occurred, the regions near the damaged areas began to emit green PL, corresponding to bromide-rich domains, whereas red PL persisted in areas distant from the damage.

To assess the electronic states of MAPbBr_2.8I_0.2, we performed X-ray photoelectron spectroscopy (XPS). The Pb 4 f XPS spectra in Fig. 3g exhibit only Pb²⁺ peaks before photoirradiation (black line), whereas Pb⁰ peaks appeared after photoirradiation (red line). The formation and disappearance of metallic Pb were further confirmed by in-situ XRD measurements and scanning electron microscopy-energy dispersive X-ray spectrometry (SEM-EDS) (Fig. 3h and Supplementary Fig. 12). In the XRD pattern, Pb⁰ peaks were clearly observed, and no broad amorphous peaks appeared after photoirradiation. Therefore, we consider that most of the Pb⁰ is in the crystalline phase. Figure 3i illustrates the temporal change in the Pb peak area. The perovskites were irradiated for 15 min, after which the light was turned off. With increasing photoirradiation time, the Pb peak area also increases. Photodegradation resulting from the generation of Pb⁰ in mixed-halide perovskites via phase segregation has been reported previously^45,46, suggesting that our proposed mechanism is plausible. Such crystal destruction and self-healing reactions were further investigated using the diffuse reflectance spectra of MAPbBr_2.8I_0.2 (Supplementary Fig. 13). After photoirradiation, a broad band, possibly attributed to the absorption of photogenerated metallic Pb, was observed. This band nearly disappeared when the sample was kept in the dark.

Regarding the self-healing behavior in aqueous solution, we consider the dynamic equilibrium of perovskites as follows²¹:

In aqueous solution, various chemicals such as bromide anions, iodide anions, and phosphinic acids are present, with the latter being particularly important for reducing oxidative species such as triiodide anions. The standard redox potentials of the relevant reactions and the band structures of MAPbX₃ (X = Br, I) are summarized in Supplementary Fig. 14. The Pb⁰ generated in the perovskite-saturated aqueous solution is oxidized to Pb²⁺ according to the following reaction:

Protons are the dominant electron acceptors in aqueous solution because of the presence of phosphinic acid. The Gibbs free energy change (ΔG) of Eq. (2) is determined by the following equation:

where n is the number of electrons involved in the reaction, F is Faraday’s constant, and ΔE is the potential difference between the related half-reactions. The calculated ΔG is −24.3 kJ·mol⁻¹⁴⁷, indicating that the oxidation of metallic Pb⁰ to Pb²⁺ proceeds spontaneously when Pb⁰ is generated by the destruction of perovskites. Due to the high concentration of halide anions in the aqueous solution, lead complexes form⁴⁸.

Based on Le Chatelier’s principle, the formation of [PbBr_xI_3−x]⁻ causes a shift in the dynamic equilibrium, resulting in the production of MAPbBr_xI_3−x(s), as expressed in Eq. (1). Consequently, the perovskite material self-heals. This self-healing behavior was also observed using in situ XRD measurements and diffuse-reflectance spectra. Upon stopping the light irradiation, the peak area associated with Pb⁰ gradually decreased and eventually disappeared completely (Fig. 3i). After leaving damaged perovskites in the solution under dark condition, absorption of photons with lower than bandgap energy was drastically decreased, indicating that self-healing reactions are proceeded (Supplementary Fig. 13). Supplementary Fig. 15 shows PL spectra before photodamaging, after photodamaging, and after self-healing reactions. In this figure, the PL wavelength, which reflects the halide composition of perovskites, remained unchanged after the self-healing process. Hence, the photodamaging and self-healing processes do not affect the Br/I ratio of the crystal. Based on the formation of MAPbBr₃ from Pb powder in an aqueous HBr solution containing MABr as the organic cation source (Supplementary Fig. 16), the self-healing formation of mixed-halide perovskites is thermodynamically and kinetically feasible. This self-healing behavior, achieved without external stimuli, has significant implications for the development of novel catalysts that utilize a dynamic equilibrium reaction system.

Photocatalytic hydrogen evolution

Upon light exposure, the MAPbBr_2.8I_0.2 powder in a saturated aqueous solution exhibited a color change from reddish brown to black, due to the formation of metallic Pb (Fig. 4a and Supplementary Fig. 13)⁴⁹. Moreover, numerous bubbles formed on the particles. As illustrated in Supplementary Fig. 14, the conduction band minimum of MAPbX₃ is more negative than the potential for proton reduction, enabling photocatalytic hydrogen generation as described in the following equations^23,36,50:

a Optical images of the MAPbBr_2.8I_0.2 powder in saturated aqueous solution (left) before and (right) after 470-nm LED light irradiation (ca. 125 mW·cm⁻²) for 24 h. b Total amount of hydrogen generated by perovskites, indicating the catalytic activity of MAPbBr₃ (black line and symbols), MAPbI₃ (red line and symbols), MAPbBr_1.3I_1.7 (purple line and symbols), MAPbBr_2.2I_0.8 (green line and symbols) and MAPbBr_2.8I_0.2 (blue line and symbols). The yellow-shaded region indicates the period of light irradiation, while the gray-shaded region indicates the period of darkness. c The hydrogen production of MAPbBr_2.8I_0.2 during intermittent irradiation. d Schematic illustration of degradation and self-healing reactions in aqueous solution under dynamic equilibrium. The self-healing reaction occurs spontaneously once the perovskites are damaged, and this cycle can continue repeatedly.

In addition, the photodamaging reaction occurred as follows:

In the present dynamic equilibrium system, the self-healing reactions include hydrogen generation, as expressed in Eq. (2). Generally, photocatalysts are only active under photoirradiation. However, if our proposed mechanism holds true, perovskites can produce hydrogen not only when photoexcited but also under dark conditions. To test this hypothesis, we evaluated the photocatalytic hydrogen production activity using gas chromatography. Figure 4b compares the amounts of hydrogen gas produced during 5 h of visible light-driven HX splitting and 18.5 h of storage in the dark. In these experiments, the perovskites were able to produce hydrogen for 5 h, indicating that they can stably exist in the solution without complete crystal destruction for a longer duration compared to microscopic measurements. This is likely due to the difference in excitation power densities at the sample surface under each condition: ca. 780 mW·cm⁻² for microscopic measurements (Fig. 2) and ca. 125 mW·cm⁻² for photocatalytic measurements (Fig. 4). The photocatalytic activity of mixed-halide perovskites is significantly higher than those of the single-halide perovskites, MAPbBr₃ and MAPbI₃. One possible reason for the higher photocatalytic activity of the mixed-halide perovskites is light-induced phase segregation. During this process, iodide-rich domains are formed near the surface, enabling efficient charge transport to these domains and facilitating photocatalytic reactions³⁶. This behavior explains how the original (undamaged) perovskites are effective photocatalysts. Hydrogen generation from the pre-irradiated samples held in the dark suggests that the damaged perovskites remained active during the self-healing reactions. As shown in Supplementary Fig. 17, the hydrogen generation rates under light are gradually decreased as light irradiation time increased, indicating that photocatalytic hydrogen evolution and self-healing-induced hydrogen evolution (Eq. (2)) occurred simultaneously. Another photocatalytic activity test supports our hypothesis. After initiating the photocatalytic reaction, we intermittently stopped the light irradiation. As illustrated in Supplementary Fig. 18, the system continued to exhibit stable hydrogen generation even in the dark. In our tests, the amount of Pb²⁺ in the MAPbBr_2.8I_0.2 crystals was approximately 20.5 µmol. However, the perovskites are not completely reduced to Pb⁰, as evidenced by their continued PL (Supplementary Fig. 11). This suggests that they retain part of their perovskite structure even after crystal destruction. Therefore, the actual amount of photogenerated Pb⁰ should be less than 20 µmol. Nonetheless, the total amount of hydrogen produced under dark conditions reached approximately 24.5 µmol (9.08, 7.98, and 7.51 for the first, second, and third cycles, respectively, Fig. 4c), again highlighting the self-healing capability of MAPbBr_2.8I_0.2 as a photocatalyst. The estimation of the potential lifetime of the photocatalytic reaction system, based on the amount of chemicals in the solution, is discussed in Supplementary Note 2. The hydrogen production activity of mixed-halide perovskites tends to decrease as the ratio of iodide anions in the crystals increases (Fig. 4b). This may be because the increased iodide content leads to a larger area of phase-segregated iodide-rich regions. Consequently, less efficient electron transfer occurs compared to MAPbBr_2.8I_0.2, resulting in lower hydrogen production activity and reduced susceptibility to destruction. This explanation is supported by microscopic observations (Supplementary Figs. 7–9).

However, with an increasing number of cycles, their photocatalytic activity gradually decreased. This could be attributed to insufficient self-healing processes. Supplementary Fig. 19 shows SEM-EDS images of MAPbBr_2.8I_0.2 prepared under various conditions. Before irradiation, each component—Br, I, and Pb—was homogeneously distributed across the entire crystals. After photoirradiation, I-rich and Pb-rich regions, corresponding to phase-segregated domains and photoreduced metallic Pb, respectively, were observed. Even after being left in their saturated solution, these I-rich and Pb-rich regions remained, suggesting that self-healing reactions were not completed within the experimental timescale. Furthermore, the solution may become supersaturated during irradiation due to the decomposition of crystals and the dissolution of their components.

Figure 4d summarizes the overall reaction scheme. Initially, perovskites stably exist in an aqueous solution by achieving dynamic equilibrium at the solid liquid interfaces. Upon photoexcitation, the accumulated electrons reduce Pb²⁺ in the perovskites to Pb⁰, resulting in significant morphological changes (Fig. 2i). After stopping light irradiation, self-healing reactions occur along with hydrogen generation. The reconstructed (self-healed) perovskites again function as photocatalysts for continued use.

Supplementary Fig. 20 illustrates the cycle of deciduous trees⁵¹. When temperatures cool, dormancy is induced, leading to the transfer of nitrogen (N), an energy source for plants, from leaves to stems. Subsequently, the leaves are shed. The stored nitrogen is then used during bud break in the next cycle to produce new growth. The self-healing mechanism of perovskites demonstrated in this study is analogous to this process. In the damaged state, akin to the dormant state in plants, perovskites store energy in the form of charges in the metallic Pb⁰ regions. The self-healing reaction, which is comparable to bud break and leaf growth in plants, is driven by the utilization of both stored chemical energy and thermal energy. The similarity between the processes in perovskites and plants suggests that biomimicry is a powerful strategy for achieving efficient energy utilization.