Introduction

In the past decade, the vast latent capacity of halide perovskites in solar energy utilization has been reported; e.g., perovskite solar cells showed a power conversion efficiency of 25.5%1. In addition to solar-to-electricity conversion, the high light harvesting capability, superior carrier transport, and suitable band position of halide perovskites motivate the wide applications of these materials in photocatalysis and photoelectrochemical (PEC) fields2,3,4,5,6. Compared to particulate photocatalytic systems, PEC systems feature semiconductor photocatalysts and external biases and thereby exhibit significant advantages in high-efficiency charge separation and collection as well as easy catalyst recycling7. Recently, halide perovskite polycrystalline thin films (PCTFs) have been exploited for PEC water splitting and CO2 reduction8,9,10,11,12,13,14,15,16,17. The use of organic solvent instead of water not only is expected to improve the stability of photoelectrode18, but also enables a broader set of organic synthesis reactions, achieving higher Faradaic efficiency and value-added chemicals19,20,21,22,23,24,25. Compared to PCTFs, perovskite single-crystal thin films (SCTFs) inherit the advantages of both polycrystalline films and bulk single crystals26, such as low trap-state densities, long carrier diffusion lengths, well-defined thicknesses and, more importantly, high stability, which thus has pinned great expectation as more promising photoelectrodes for PEC applications.

2,5-Dimethoxy-2,5-dihydrofuran (DMDF) has been considered a key intermediate for producing pyridines, pyridazines, pyrroles, benzenoid compounds, and coumarin analogs27,28,29,30. In its traditional chemical synthetic method, a large amount of bromine is employed as an oxidant, the mass usage of which raises issues of environmental pollution29. As an alternative approach to produce DMDF, electrochemical oxidation of furan can be mildly implemented in NH4Br alcoholic solution. In 2017, Sayama and co-worker provided another more direct approach based on a PEC system using BiVO4/WO3 as a photoanode with a Br+/Br mediator under visible light irradiation30. PEC synthesis has been found to significantly lower the required applied potential; however, the resulting photocurrent density is slightly unsatisfactory (<0.55 mA cm−2). Currently, it is reasonable to expect that halide perovskite SCTFs could deliver advanced PEC performance towards DMDF production, considering their remarkable photoelectric properties. However, the utilization of perovskite-based PEC cells for organic synthesis remains unexplored.

Herein, methylammonium lead bromide MAPbBr3 SCTF grown on a conductive FTO glass substrate is demonstrated as an efficient photoanode in a PEC cell to afford a high turnover rate for artificial photosynthesis of DMDF from furan. The surface is the most vulnerable to defects and passivation of single-crystal surface defects is the most important task. The subsequent deposition of Al2O3 and Ti overlayers significantly promoted the operating stability and photocatalytic activity of MAPbBr3 SCTFs. As an encouraging result, the highest photocurrent of 7.8 mA cm−2 (0.8 V vs. Ag/AgCl) is achieved for MAPbBr3 SCTF/Al2O3/Ti. Such work paves the way for the future design of high-performance perovskite SCTF-based solar-driven photosynthesis systems.

Results and discussion

PEC characterizations of MAPbBr3 SCTF and PCTF

MAPbBr3 SCTFs were grown in situ on an FTO/TiO2 substrate through our previously reported space-limited crystallization method with some modifications31,32. The as-prepared MAPbBr3 SCTF exhibits a smooth and pinhole-free surface with a lateral size of 11.5 mm × 7.6 mm and a film thickness of approximately 14 µm (Fig. 1a, b and Supplementary Fig. 1a). In contrast, a 300-nm-thick MAPbBr3 PCTF prepared by spin-coating (Fig. 1c and Supplementary Fig. 1b) is composed of small crystals accompanied by abundant sharp grain boundaries (Fig. 1d). The configuration of the PEC cell system is depicted in Fig. 1e, where MAPbBr3 SCTF (or PCTF) on an FTO/TiO2 substrate and a Br/Br+ redox couple in an acetonitrile/methanol mixed solution function as the photoanode and electrolyte, respectively. Ultraviolet photoelectron spectra (UPS, Supplementary Fig. 2a, b) show that the valence band maximum (VBM) of the MAPbBr3 SCTF is approximately 1.45 V vs. SHE, capable of bromine oxidation (0.7 V vs. SHE, Supplementary Fig. 2c)28. Upon light excitation, the Br in the electrolyte will be oxidized into Br+ species by photogenerated holes, which then serve as oxidizing reagents to transform furan into DMDF. The resultant saturated photocurrent densities of MAPbBr3 PCTF and SCTF are 0.7 and 1.2 mA cm−2 at 0.20 V vs. Ag/AgCl (Fig. 1f), respectively. Such a 70% enhancement in photocurrent density for MAPbBr3 SCTF is mainly ascribed to its high crystallinity, high charge mobility and low trap density31. In addition, it can be clearly seen that the photocurrent density of the MAPbBr3 PCTF PEC cell decreases rapidly within ~3 min, while that of the MAPbBr3 SCTF retains 54% of the initial value even after 30 min (Fig. 1g). This phenomenon is further supported by the stability of the materials in acetonitrile/methanol electrolyte (Supplementary Fig. 3). MAPbBr3 PCTF decomposed within 25 min, while MAPbBr3 SCTFs exhibit negligible degradation even after 20 h, further demonstrating the better stability of SCTFs. The diffusion length of MAPbBr3 SCTF (5.6 µm)31 is much higher than that of perovskite-based thin films (0.33 µm)33. Electric fields can help carriers migrate from a 14 µm thick single crystal to the surface and then participate in chemical reactions. On the premise of ensuring light absorption, further reducing the thickness of single crystal film is beneficial for shortening the carrier transmission distance and further improving the PEC performance. However, it is still challenging to control the perovskite single crystal films with large size and several microns thickness. In terms of influence of thickness on stability of perovskite film, thicker films tend to exhibit better stability (Supplementary Fig. 4). However, grain boundaries in PCTF have also been widely proven to primarily bear the brunt of perovskite degradation due to massively distributed trap states34. Therefore, the stability of the SCTF-based PEC cell can be mainly ascribed to the advantages of being few defect states.

Fig. 1: Comparison of MAPbBr3 SCTF with PCTF.
figure 1

Optical microscopy and SEM images of MAPbBr3 SCTF (a, b) and PCTF (c, d). e Schematic configuration of the proposed working principle of MAPbBr3 SCTF-based (or PCTF-based) PEC cells. f Linear sweep voltammetry (LSV) of MAPbBr3 SCTF-based and PCTF-based photoelectrodes. g Chronoamperometric trace of MAPbBr3 SCTF-based and PCTF-based photoelectrodes recorded at an applied potential of 0.2 V vs. Ag/AgCl.

Investigation of the Al2O3 passivation effect for MAPbBr3 SCTFs

Halide perovskite materials have been widely proven to have a low trap formation energy and reconfigurable surface lattice35,36, and therefore, surface trap states are easily formed as charge carrier recombination centers37. Herein, we deposit an ultrathin Al2O3 layer onto a MAPbBr3 SCTF through atomic layer deposition (ALD) to further boost the PEC performance by reducing the amounts of surface trap states (see section “Experimental details” in the “Methods” section).

The presence of Al2O3 is first verified by the additional Al 2p peak at 74.8 eV in the X-ray photoelectron spectroscopy (XPS) spectra (Fig. 2a). Furthermore, characteristic signals of Al–CH3, (OH)–Al=O and Al–O–Al species are observed at 1256, 1080, and 888 cm−1 in the Fourier transform infrared (FTIR) spectra after ALD treatment, again proving the existence of Al2O3 (Supplementary Fig. 5a). During the process of ALD deposition of Al2O3, TMA reacts with CH3NH3+ of MAPbBr3 and results in CH3NH2 release accompanied by the formation of Br3Pb–Al(CH3)–PbBr3 intermediates (Fig. 2d)38, which further react with O2 to produce Al2O3. Further insight into the chemical modification on the perovskite surface was provided by comparing the surface XPS spectra of N 1s, Pb 4f, and Br 3d before and after Al2O3 deposition (Fig. 2a–c). The intensity of the N 1s peak is significantly weakened due to CH3NH2 release. Furthermore, both Pb 4f7/2 and Pb 4f5/2 peaks (Fig. 2c) shift to lower binding energy values by 0.37 eV after the ALD process because of the Pb-rich surface defects bonded with adsorbed oxygen39,40, which can additively suppress nonradiative recombination on the perovskite surface. The surface microstructure of the MAPbBr3 SCTF is well retained after Al2O3 deposition, as revealed from the scanning electron microscopy (SEM) image (Supplementary Fig. 5b). Identical X-ray diffraction (XRD) patterns are observed in MAPbBr3 SCTF and MAPbBr3 SCTF/Al2O3 with the same growth orientation along the (h00) crystal plane, which not only reveals that the crystal phase of perovskite is well preserved during the ALD process, but also implies that the deposited Al2O3 is thin/amorphous (Supplementary Fig. 6).

Fig. 2: Al2O3 passivation effect investigation.
figure 2

XPS spectra of pristine MAPbBr3 SCTF and MAPbBr3 SCTF/Al2O3: Al 2p and Br 3d (a), N 1s (b), and Pb 4f (c). d Proposed passivation mechanism of Al2O3 on the MAPbBr3 SCTF surface. e PEC behavior of MAPbBr3 SCTF/Al2O3-based photoelectrodes as a function of Al2O3 deposition thickness. Al2O3-thickness-dependent steady-state photoluminescence (PL) spectra (f) and time-resolved PL decay curves (g).

A batch of MAPbBr3 SCTF/Al2O3 (0, 1, 5, and 10 nm) were fabricated by adjusting the ALD deposition cycles to unlock the effect of the Al2O3 layer on the PEC performance. Al2O3 thickness is achieved by controlling the number of ALD deposition cycles. Approximately 0.1 nm of aluminum oxide was deposited per cycle (for instance, 50 cycles, namely 5 nm). As shown in Fig. 2e, the saturated photoanode current density of the pristine MAPbBr3 SC film is ~1.2 mA cm−2 at 0.4 V vs. Ag/AgCl (0.53 V vs. SHE, Supplementary Fig. 7), which gradually increases with thickening Al2O3 and reaches a maximum value of 5.5 mA cm−2 at 5 nm of Al2O3. The enhanced PEC performance affirms that the ultrathin Al2O3 layer would not hinder photogenerated holes injection from perovskite to the electrolyte due to charge tunnelling. Comparing with pristine MAPbBr3 SC, coating a thin Al2O3 overlayer with a thickness of 1–10 nm resulted in slight overpotential increasement. This is because that deposited Al2O3 can isolate the photoelectrode from the electrolyte, therefore the photocorrosion phenomenon of MAPbBr3 SC being oxidized can be effectively suppressed. Meanwhile, with the increase of the thickness of Al2O3 overlayer, the passivation of effect of MAPbBr3 SC surface is more sufficient, so that the thicker 10 nm Al2O3 can reduce the overpotential comparing to the thinner Al2O3 (1 and 5 nm). The specific passivation mechanism will be discussed later. However, further thickening of the Al2O3 layer (e.g., 10 nm) attenuated the photoanodic current density, probably due to the aggrandized charge transport resistance as a result of its intrinsic insulating properties. The PEC performance of MAPbBr3 SCTF/Al2O3 (5 nm) is ~10 times higher than that of a previously reported BiVO4/WO3 photoanode with a maximum photocurrent of 0.55 mA cm−2 at 0.5 V vs. SHE)28. We have also investigated the influence of Al2O3 thickness on the stability. The stability of the electrode gradually improved with the increase of the thickness of Al2O3 from 0 to 10 nm (Supplementary Fig. 8), since the deposited Al2O3 can isolate photoelectrode from the electrolyte.

To identify the underlying reason for the increased PEC performance after coupling with an ultrathin Al2O3 layer, the optical properties of MAPbBr3 SCTF/Al2O3 were investigated. As shown in Fig. 2f, the PL emission intensity of the MAPbBr3 SCTF is significantly amplified after depositing 1-nm-thick Al2O3, and is further increased upon thickening of the Al2O3 layer to 5 and 10 nm. Time-resolved PL (TRPL) decay plots at the emission maximum (Fig. 2g) also demonstrate that all MAPbBr3 SCTF/Al2O3 samples decay sluggishly compared with pristine MAPbBr3 SCTF. Specifically, the fitted average PL decay time of MAPbBr3 SCTF/Al2O3 (5 nm) is twofold longer than that of pristine MAPbBr3 SCTF (Supplementary Table 1). This enhanced PL intensity and prolonged PL lifetime can be ascribed to the suppression of nonradiative recombination of electron-hole pairs in MAPbBr3 SCTF/Al2O3. In addition, spatial and temporal imaging PL emission was monitored with a sample size of 100 µm × 100 µm. Figure 3a, c show confocal fluorescence images of pristine MAPbBr3 SCTF and MAPbBr3 SCTF/Al2O3 (5 nm). Apparently, greater brightness is observed for MAPbBr3 SCTF/Al2O3, corresponding to stronger PL emission intensity. More importantly, brightness contrast appears with apparent darker spots in the images of pristine MAPbBr3 SCTF, but these spots are well dispelled after Al2O3 deposition, thus affording a uniform PL emission image (Fig. 3c). These darker spots on MAPbBr3 SCTF could be assigned to the region with higher surface trap-state density, and thus with active nonradiative charge carrier recombination41. Agreeing well with the TRPL test, fluorescence lifetime imaging microscopy (FLIM, Fig. 3b, d) data suggest a longer average PL lifetime for MAPbBr3 SCTF/Al2O3 (553.5 ns) than for pristine MAPbBr3 SCTF (263.8 ns). These results strongly affirm the passivation role of Al2O3, which effectively suppresses the trap states on the perovskite surface42.

Fig. 3: Fluorescence spectroscopic measurements.
figure 3

Confocal fluorescence intensity images (left) and fluorescence lifetime imaging microscopy (FLIM) images (right) of pristine MAPbBr3SCTF (a, b) and MAPbBr3SCTF/Al2O3 (c, d).

Different from the PL tests, transient reflection (TR) spectroscopy (Fig. 4a) can solely monitor the surface luminous and nonluminous species and provide detailed insight into the surface charge carrier dynamics, avoiding interferential signals from the bulk37,43. Both pristine MAPbBr3 SCTF and MAPbBr3 SCTF/Al2O3 feature a bandgap of 2.36 eV, corresponding to a UV–Vis absorption onset at 525 nm (Supplementary Fig. 9). Therefore, a 400 nm laser pulse with corresponding penetration depths of 80 nm was employed to excite only MAPbBr344, and broadband continuum white light in the range of 450–700 nm was used as the probe pulse to monitor the reflection difference. Figure 4b, c show the pseudocolour TR spectra of pristine MAPbBr3 SCTF and MAPbBr3 SCTF/Al2O3, both characterized by an intense signal at approximately 520 nm. Accordingly, transient TR profiles recorded at a delay time of 5 ps also show a typical antisymmetric peak centred at 520 nm (Fig. 4d, e). To clearly reveal the transient absorption (TA) difference upon light excitation, the inverse Hilbert transform (iHT) was performed according to the Kramers–Kronig relationship45. As displayed in Fig. 4d, e, the resultant TA curves after iHT exhibit a ground state bleaching peak centred at ~520 nm, agreeing well with the absorption onset. Due to the restricted time window (≤7 ns) for the TR tests, the decay dynamics are mainly derived from the diffusion and recombination processes of surface photo-induced charge carriers46. The decay kinetics probed at 520 nm (Fig. 4f) were analysed and interpreted and show a very sluggish decay for MAPbBr3 SCTF/Al2O3 compared to that of the pristine MAPbBr3 SCTF (141 ps vs. 70 ps) due to the retardative surface trap-assisted recombination afforded by the ultrathin Al2O3 layer.

Fig. 4: Transient reflection spectroscopic measurements.
figure 4

a Diagram of transient reflection (TR); TR spectra pseudocolour images, and TR spectra recorded at a delay time of 5 ps of pristine MAPbBr3 SCTF (b, d) and MAPbBr3 SCTF/Al2O3 (c, e). Kinetic fit of the data of these two SCTFs (f).

To further investigate the underlying effect of the Al2O3 layer on the electrical properties of the MAPbBr3 SCTF, Kelvin probe force microscopy (KPFM) was performed to probe the spatial charge transfer and separation on the perovskite. After subjection to the ALD 5 nm procedure, the MAPbBr3 SCTF surface appears slightly rougher with an increase in the root-mean-square value from 0.7 to 5.7 nm (Fig. 5a, d). Continuously distributed nanosized particles can be clearly observed in Fig. 5d, further corroborating that an ultrathin compact Al2O3 layer was uniformly deposited onto the MAPbBr3 SCTF, which is associated with island growth mechanism during ALD deposition of Al2O338. The KPFM images and the corresponding line profiles of the surface photovoltage (SPV) difference are displayed in Fig. 5b, c, e, f. MAPbBr3 behaves as an n-type semiconductor according to UPS analysis, thus resulting in a space charge layer at the surface with an inherent electric field direction from the interior towards the surface47. The positive SPV values of MAPbBr3 SCTF and MAPbBr3 SCTF/Al2O3 indicated the photoinduced holes could transfer from the interior to surface of the sample. The SPV value of MAPbBr3 SCTF/Al2O3 is approximately 150 mV, twice higher than that of pristine MAPbBr3 PCTF (75 mV), indicating richer density of holes accumulated onto the surface, which is beneficial for enhancing the PEC performance.

Fig. 5: Kelvin probe force microscopy.
figure 5

AFM topography images, KPFM images and CPD profiles (subtracting the potential in the dark from that under light irradiation) of pristine MAPbBr3 SC film (ac) and MAPbBr3 SCTF/Al2O3 (df).

Investigation of the Ti layer effect on the PEC performance of MAPbBr3 SCTFs

The above experimental results reveal that the Al2O3 passivation layer significantly enhances the PEC performance and the stability of the photoelectrode to some extent. However, due to its ultrathin thickness, long-term PEC stability is still challenging to maintain. Conductive overlayers have been widely proven to enhance the operating stability and even catalytic activity of semiconductor materials8,13. Thus, a 145-nm-thick Ti layer with a particle size of approximately 30 nm (Supplementary Fig. 10) was introduced onto MAPbBr3 SCTF/Al2O3 as a protective layer to avert direct interaction with the electrolyte. The MAPbBr3 SCTF/Al2O3/Ti shows almost the same XRD pattern as MAPbBr3 (Supplementary Fig. 11a). The reasons beyond the failure to observe the characteristic XRD pattern of Ti layer probably can be concluded as follows: (a) the high crystallinity of perovskite single crystal leads to its diffraction signals completely overwhelms the Ti signals; (b) amorphous nature of the deposited Ti layer. In order to find out the underlying reasons, we have tried to deposit Ti layer on the surface of quartz. The XRD results (Supplementary Fig. 11b) show that there are two new phases, which can be assigned to the Ti (PDF#01-074-7075) and rutile TiO2 (PDF#04-008-7856) with low crystallinity. The formation of TiO2 is ascribed to the oxidation of titanium by residual oxygen in the system during the evaporation process48. XPS spectrum (Supplementary Fig. 12) reveals that metallic Ti0+ (at 454.1 and 460.2 eV), Ti3+ (at 456.9 and 462.4 eV) and Ti4+ (at 458.7 and 464.4 eV) species are present, implying a TiOx-rich surface49. Figure 6a displays the current-potential curves of the MAPbBr3 SCTF/Al2O3/Ti-based PEC cell under chopped illumination. After incorporating the Ti layer on the MAPbBr3 SCTF/Al2O3 photoanode, the onset potential (~0.2 mA cm−2)50 was negatively shifted by 120 mV. The photocurrent was saturated for MAPbBr3/Al2O3 (5 nm) at 0.4 V vs. Ag/AgCl with 5.5 mA cm−2. In contrast, the photocurrent is not saturated for MAPbBr3 SCTF/Al2O3/Ti throughout all the test voltages from −0.4 to 0.8 V, reaching 7.8 mA cm−2 at 0.8 V vs. Ag/AgCl. To confirm the catalytic activity of Ti layer, the electrochemical behavior of FTO/Ti and FTO was also studied (Supplementary Fig. 13). Compared with the weak anodizing current of the FTO electrode, the FTO/Ti electrode exhibited remarkable electrocatalytic activity towards bromine oxidation with a near-zero overpotential and an onset potential of 0.73 V vs. SHE (0.60 V vs. Ag/AgCl). This result indicates that holes can easily be injected into the electrolyte and immediately participate in the chemical reaction due to the co-catalyst function of the Ti layer.

Fig. 6: Ti layer effect on the PEC performance.
figure 6

a PEC behaviors of MAPbBr3 SCTF/Ti-based and MAPbBr3 SCTF/Al2O3/Ti-based photoelectrodes. The projected DOS of b TiO2−x and c TiO2. DFT-calculated adsorption of Br on the surfaces of d TiO2−x and e TiO2. f, g Corresponding electron density difference, where yellow indicates a gain of electrons and cyan indicates loss of electrons.

Density functional theory (DFT) calculations were conducted to shed light on the impact of Ti species on the electronic properties of the materials51. As shown in Supplementary Fig. 14, the calculated density of state (DOS) of Ti is continuous owing to its intrinsic metallic nature. Figure 6b, c illustrate that Ti3+ produces a higher density of oxygen vacancies and rearranges the band structure. The interaction between the Br atom and the three-atomic-layer p(3x) surface was probed (Fig. 6d, e). The calculated Br adsorption energy is more negative for TiO2 (−3.61 eV) than for TiO2−x (−0.57 eV), implying that TiO2−x has more potential as a Br oxidation catalyst. Upon adsorption, Br atoms suffer from charge redistribution and form chemical bonds with Ti atoms. The length of the Ti–Br bond on the TiO2−x surface (2.59 Å) is estimated to be shorter than that on TiO2 (2.67 Å), agreeing well with its advantageous Br adsorption owing to the low-coordinated oxygen surface. The charge difference analysis in Fig. 6f, g displays a more significant charge transfer on the TiO2−x surface than on the TiO2 surface. Based on Mulliken charge analysis, the Br atom is −0.26|e| on TiO2−x but only −0.06|e| on TiO2. A similar trend has been observed based on Bader charge analysis, i.e., −0.54|e| on TiO2−x and −0.26|e| on TiO2. The increased electron cloud density of Br favours the harvesting of holes during the subsequent oxidation process. Similar phenomena have also been observed in defect-rich TiO2, where the oxygen vacancies and Ti3+ ions facilitate more efficient charge transfer52 and serve as an intrinsic active species to increase the electrochemical53, photocatalytic, and PEC activities54. The current calculation results are consistent with the hypothesis that the introduction of catalytic active sites on the surface of the photoelectrode promotes the further improvement of PEC performance.

In addition to affording superior PEC performance, Ti overlayer can also protect the internal Al2O3 layer and perovskite layer. In order to further mitigate the influence of the electrolyte penetration from the edge, silicone sealant (Kafuter K-705) is used to encapsulate the edge of electrode. The resulted photoelectrode exhibited impressive stability for DMDF evolution, as displayed in Supplementary Fig. 15. The Br-mediated photogeneration of DMDF was carried out at 0.2 V vs. Ag/AgCl, which is lower than the potential required to initiate the electrochemical oxidation of bromine. The photocurrent density is well maintained even after 6 h of continual electrolysis at 0.2 V. The formation of DMDF is confirmed via mass spectrometry (Supplementary Fig. 16), and the Faraday efficiency is calculated to be as high as 93%, further promising its infinite potential as an effective and stable photoelectrode.

In summary, a MAPbBr3 SCTF-based photoanode could drive the conversion of furan to DMDF has been demonstrated. Combined investigations with FLIM and TR as well as KPFM characterizations demonstrated that an ultrathin Al2O3 layer effectively suppressed trap-assisted nonradiative charge recombination, resulting in a doubled surface charge carrier lifetime. Furthermore, the subsequent deposited Ti layer not only stabilized the MAPbBr3 SCTF by isolating perovskite from the electrolyte but also facilitated the bromine-mediated oxidation of furan owing to the Ti3+ catalytic active site. As a result, the highest photocurrent of 7.8 mA cm−2 at 0.8 V vs. Ag/AgCl was realized for the MAPbBr3 SCTF/Al2O3/Ti photoanode. The photoanode retained remarkable stability for more than 6 h under chopped light illumination. Considering the lead-related-toxic issues55,56, stable lead-free perovskite SCTFs should deserve more future research focus towards environment-friendly and efficient photocatalysts, whose application will be potentially expanded to currently still challenging biomass conversion and organic synthesis reactions in the near future.

Methods

Synthesis of methylammonium bromide (MABr)

A mixture of methylamine solution (30–33 wt% in methanol, Aladdin) and hydrobromic acid (48 wt% in water, Aladdin) with a molar ratio of 1:1.5 was stirred in an ice-water bath for 2 h. After that, MABr powder was precipitated by rotary evaporating the solvent at 50 °C and then purified by recrystallization. The collected MABr powder was dried in a vacuum oven for 24 h.

Fabrication of MAPbBr3 single-crystal thin films (SCTFs)

MAPbBr3 SCTFs were grown onto TiO2-coated FTO glass substrates according to our previously reported method31,32. First, FTO glass substrates (7 Ω square−1, Nipponsheet Glass) were sequentially cleaned by ultrasonication in detergent water, alcohol, and acetone for 30 min per step. A compact TiO2 layer was subsequently deposited on the cleaned FTO glass via spin-coating of a TiO2 colloidal solution at 4 k rpm for 30 s, followed by an annealing process at 500 °C for 30 min. The annealed TiO2-coated FTO substrate was further immersed into a 0.04 M TiCl4 solution at 70 °C for 30 min before a secondary annealing process at 500 °C.

A space-limited model was assembled by sandwiching a thin polytetrafluoroethylene rectangular ring frame (20 μm in thickness) into two pieces of TiO2-coated FTO glasses. Then, the MAPbBr3 precursor solution, prepared by dissolving equimolar PbBr2 and MABr in DMF, was circulated into the limited space between the two substrates by a peristaltic pump (BT100-2J, Longer Pump). Note that the bottom FTO glass was drilled with two holes to facilitate the circulation of the perovskite precursor solution. The middle part of the model was locally heated to 95 °C for crystallization and then maintained at 80 °C for crystal growth. After ~72 h, the two FTO glasses were separated, with MAPbBr3 SCTFs grown in situ on the conductive substrate. The precipitated polycrystals were removed by γ-butyrolactone, and the residual solution was cleaned and removed by vacuum drying.

Fabrication of MAPbBr3 PCTF

Equimolar MABr and PbBr2 were dissolved in DMF/DMSO (3:1, v–v) at a concentration of 1.2 M. An appropriate amount of perovskite precursor solution was spin-coated onto the substrate with a ramp rate of 3000 rpm, and 200 µL of chlorobenzene was dropped onto the spinning substrate 10 s after the spin-coating began. The obtained film was annealed at 120 °C for 10 min.

Fabrication of MAPbBr3 SCTF/Al2O3/Ti

An ultrathin aluminum oxide layer was grown on the surface of MAPbBr3 SCTF by an atomic layer deposition (ALD) machine (ALD-SC6-PE, Syskey) with TMA and plasma oxygen as the element sources. In each ALD cycle, TMA was dosed into the chamber for 80 ms and purged out by Ar gas for 8 s, followed by dosing plasma O2 into the chamber for 3 s and purging out by Ar gas for 8 s. Approximately 0.1 nm of aluminum oxide was deposited per cycle. The ALD chamber was kept at a substrate temperature of 80 °C to achieve effective deposition of Al2O3. Another thin layer of Ti was further deposited on the aluminum oxide layer by the thermal evaporation deposition method (SKY Technology Development).

Characterizations

XRD patterns were collected on a Rigaku Miniflex 600 X-ray diffractometer with Cu Kα radiation (λ = 1.5418 Å). X-ray photoelectron spectroscopy (XPS) was performed by means of a Thermo-VG Scientific ESCALAB 250 X-ray photoelectron spectrometer. The cross-sectional surface morphology and film thickness were examined by high-resolution field emission scanning electron microscopy (FE-SEM, Gemini 500). Atomic force microscopy (AFM) and scanning KPFM measurements were performed on a Bruker Dimension Fast Scan AFM system. UV–Vis absorption was obtained on an ultraviolet–visible spectrophotometer (Shimadzu UV-3600). Photoluminescence (PL) spectra were obtained on a fluorescence spectrophotometer (Edinburgh FLS980) with an excitation wavelength of 406 nm. Confocal fluorescence images and fluorescence-lifetime imaging microscopy (FLIM) images were recorded on a Zeiss LSM880 NLO microscope with an excitation wavelength at 405 nm. Transient reflectance spectra were measured on a Helios (Ultrafast Systems LLC) pump-probe setup in reflection mode. The pump beam at 400 nm was generated in a collinear optical parametric amplifier (OPerA Solo, Coherent) pumped by 800 nm fundamental pulses (100 fs, 1 kHz repetition rate, Astrella-Tunable-V-F-1k, Coherent). A sapphire crystal was used to generate a broad white light continuum probe spectrum. The TR signal changes were recorded in the visible region from 450 to 700 nm.

DFT calculations

DFT calculations were carried out by means of the CP2K package57. The PBE functional58 with Grimme D3 correction59 and Goedecker-Teter-Hutter (GTH) pseudopotentials60,61 and DZVPMOLOPT-GTH basis sets62 were utilized to describe the system and the molecules, respectively. Unrestricted Kohn–Sham DFT was performed as the electronic structure method in the framework of the Gaussian and plane waves method62,63. A plane-wave energy cut-off of 500 Ry was employed. The adsorption energy (Eb) is defined as:

$${{E}}_{\mathrm{b}} = E_{{\mathrm{Br/sur}}} - 1/2{{E}}_{{\mathrm{Br}}_2}(g) - {{E}}_{{\mathrm{sur}}},$$
(1)

where EBr/sur, \({{E}}_{{\mathrm{Br}}_2}\), and Esur are the surface Br adsorption energy and the individual electron density of the molecule and surface, respectively. The charge density difference (Δρ) is defined as

$$\Delta \rho = \rho _{{\mathrm{Br/sur}}} - \rho _{{\mathrm{Br}}} - \rho _{{\mathrm{sur}}},$$
(2)

where ρBr/sur, ρBr, and ρsur are the electron density of surface-adsorbed Br and the individual electron density of Br and surface, respectively.

Photoelectrochemical (PEC) measurements

PEC performance was measured on an electrochemical workstation (CHI660E, CH Instruments) at room temperature. A conventional three-electrode system with a Pt mesh as a counter electrode and Ag/AgCl(sat KCl) as a reference electrode was adopted. To prevent dissolution and decomposition of perovskite, excessive MABr (2.02 g) was added to a mixed solution of anhydrous acetonitrile (45 mL) and anhydrous methanol (5 mL) as an anolyte. To prepare the target product of 2,5-dihydro-2,5-dimethoxyfuran, an additional 2.5 mL of furan (Innochem) was added. For the catholyte, 1-butyl-1-methylpyrrolidinium bis(trifluoromethylsulfonyl)imide (3.47 g, 99%, Shanghai Cheng Jie Chemical) and deionized water (0.5 mL) were added to anhydrous acetonitrile (50 mL). The linear sweep voltammogram (LSV) curves were measured at a scanning rate of 10 mV s−1 from −0.4 V to 0.8 V vs. Ag/AgCl under simulated solar illumination (Zolix, Gloria-X150A, 100 mW cm−2, AM 1.5G filtered, calibrated with a standard Si solar cell). The light on/off was regulated by the light chopping equipment (Perfect Light, PFS40A). After the photoelectrocatalysis test, 2 mL H2O was added into 1 mL of anode electrolyte and the solution was extracted with 0.5 mL ethyl acetate. The 2,5-dihydro-2,5-dimethoxyfuran in the top organic layer was quantitatively analyzed by gas chromatography using commercial 2,5-dihydro-2,5-dimethoxyfuran (TCI Shanghai) diluted with ethyl acetate as an external standard.