Dimethylammonium iodide stabilized bismuth halide perovskite photocatalyst for hydrogen evolution

Metal halide perovskites have emerged as novel and promising photocatalysts for hydrogen generation. Currently, their stability in water is a vital and urgent research question. In this paper a novel approach to stabilize a bismuth halide perovskite [(CH3)2NH2]3[BiI6] (DA3BiI6) in water using dimethylammonium iodide (DAI) without the assistance of acids or coatings is reported. The DA3BiI6 powder exhibits good stability in DAI solutions for at least two weeks. The concentration of DAI is found as a critical parameter, where the I- ions play the key role in the stabilization. The stability of DA3BiI6 in water is realized via a surface dissolution–recrystallization process. Stabilized DA3BiI6 demonstrates constant photocatalytic properties for visible light-induced photo-oxidation of I- ions and with PtCl4 as a co-catalyst (Pt-DA3BiI6), photocatalytic H2 evolution with a rate of 5.7 μmol⋅h-1 from HI in DAI solution, obtaining an apparent quantum efficiency of 0.83% at 535 nm. This study provides new insights on the stabilization of metal halide perovskites for photocatalysis in aqueous solution.


Introduction
Hydrogen evolution from solar-driven splitting of water or acids has been regarded as a promising approach for the conversion and storage of solar energy, and thus it has been under the scope of research for nearly half a century [1][2][3]. Until today, the typical photocatalysts have been rather inefficient because of their large band gap or severe recombination of photocarriers [4,5]. More recently, metal halide perovskites have been widely employed as efficient photoactive materials in solar cells, with a surge of power conversion efficiency from 3.8% to exceeding 25.2% over the past decade [6,7]. The exceptional performance in solar cells originates from their superior optoelectronic properties, such as wide absorption window and long electron-hole diffusion lengths [8][9][10], which are also highly expected among photocatalysts. For the first time, Nam et al. reported their pioneering work of photocatalytic hydroiodic acid (HI) splitting using methylammonium lead iodide (MAPbI3) [11]. To date, because of the instability of hybrid perovskites in water, all photocatalytic hydrogen evolution reactions on lead based perovskites are carried out in concentrated hydrohalic acids [11][12][13][14][15][16][17][18][19]. Moreover, despite their superior light-harvesting and carrier transporting properties, the high cost and corrosive nature of concentrated hydrohalic acids as well as the environmental and health risks associated with lead make such processes for H2 evolution impractical and have thus largely limited their general applications in photocatalysis [20,21].
Approaches have been developed to replace lead and alleviate the problem of perovskite instability simultaneously using e.g. all-inorganic halide perovskites and bismuth halide perovskites. In particular, bismuth halide perovskites have shown better stability under ambient atmosphere in solar cells in comparison with traditional lead based perovskites [22]. Although some works have demonstrated that the bismuth halide hybrids show catalytic activity in lower concentration of hydrohalic acids or even water [23,24], their stability issue in aqueous solution is still lacking. In addition, encapsulation strategy has been explored to stabilize metal halide perovskites in water [25,26]. However, for coating approach, one needs to consider many aspects such as the thickness, chemistry and charge transport behavior of the shell material to ensure the photogenerated carriers can eventually reach the surface and react with adsorbed species there [27,28]. Considering mentioned above, stabilizing bismuth-based halide perovskites directly in water is a challenging but meaningful route for widening the photocatalytic applications with copious water.
More recent years, the dimethylammonium iodide (DAI)-related methods are widely used to stabilize the perovskite phase [29,30]. It has been reported that the incorporation of DAI into perovskite solar cells can improve the air and thermal stability as well as charge transport property [31][32][33]. Notably, the stabilization with DAI seems more promising than the use of insulating coating materials in photocatalysis because the latter may hinder the charge transfer from perovskite to surface catalytic sites and deteriorate the following redox reactions [34,35]. were used as received.

Synthesis of DAI
DAI was synthesized by adding 32.3 mL of HI solution dropwise in 28.5 mL of (CH3)2NH in a round-bottom flask, and then stirred at 0 °C for 2 h in ice-water. To collect the product, the excess solvent was removed using a rotary evaporator at 60 °C for 2 h. The obtained brown solid was washed with diethyl ether until colorless solid was obtained. Then, the powder was recrystallized in the mixture of diethyl ether and absolute ethanol (1:5 v/v) for 24 h. After filtration, a white powder (DAI) was collected and dried at 60 °C for 24 h. A similar procedure was utilized to synthesize methylammonium iodide (MAI) except that 19.5 mL of CH3NH2 was used.

Synthesis of DA3BiI6 and bismuth triiodide (BiI3)
0.103 g of Bi2S3 was dissolved in 2 mL of HI solution by stirring for 10 min, and the obtained red solution was filtered through a 0.45 μm polytetrafluoroethylene (PTFE) syringe filter. Next, the solution was added into 12 mL of absolute ethanol and 800 μL of DMF and heated at 80 °C (in water bath) under continuous stirring for 2 h. Finally, the obtained dark red solution was transferred into a baker and sealed, then cooled at room temperature overnight for crystallization. The DA3BiI6 powder was collected with a paper filter (2 μm) and washed with chloroform until the solution was colorless. The resultant orange powder was dried at room temperature for 24 h. BiI3 was synthesized by the reaction of bismuth trioxide (Bi2O3) and hydriodic acid with 1:1 molar ratio, followed by centrifuging (8,000 rpm, 3 min) and drying at 60 °C overnight.

Solubility of DA3BiI6 in HI solution
The solubility was determined by analyzing the supernatant fraction of saturated solutions of DA3BiI6 powder. In brief, 250 mg of DA3BiI6 powder was added into 1 mL of HI solution, followed by sonication and shaking simultaneously for 30 min. Next, the upper clear fraction of the liquid was collected and filtered (0.45 μm PTFE), and then kept stand-still overnight. After that, a washed glass slide (weight: W1) was chosen as a carrier. 10 μL of the upper liquid was dropped on it and thermally annealed in oven at 110 °C for 30 min (weight: W2). The solubility was calculated by the mass difference (W2 − W1).

Amount of evolved triiodide (I3 − ) ions
Firstly Notably, the stock solution should be fresh because the I3 − ions may be disproportionated into I3 − and IO3 − ions in water and the KI solution is not stable when exposed to light. Then, the concentrations of I3 − ions before and after light irradiation were determined by the standard curve mentioned above. Owing to the similar absorption range caused by iodobismuthate ions in photocatalytic solutions, a 1/100 dilution with water of stock solution was used to determine the concentration of I3 − ions.

Characterization
The morphology of DA3BiI6 powder was analyzed by field emission scanning electron microscopy (FESEM, Zeiss Sigma FESEM). Powder X-ray diffraction (XRD) was carried out by using PANalytical X'Pert Pro Diffractometer with Cu Kα, at 2θ scan from 5° to 80°. Attenuated total reflection Fourier-transform infrared (ATR-FTIR) spectra were recorded with a Bruker HYPERION 3000 FT-IR Microscope equipped with ATR crystal. X-ray photoelectron spectroscopy (XPS) was obtained by using a Thermo Fisher Scientific ESCALAB 250Xi instrument with Al Kα irradiation. The spectrometer work function used was 4.59 eV. The binding energy of aliphatic C1s peak (284.8 eV) was applied as reference for calibration. Absorption spectra of solutions were collected using a UV-vis spectrophotometer (UV-2600, SHIMADZU). The absorption spectrum of solid was recorded with a spectrophotometer system (Optronic Laboratories, USA) equipped with an integrating sphere. The experimental bandgap was determined from the Tauc plot for indirect transition (hν ~ (αhν) 1/2 , where hν is the photon energy and α is the optical absorption coefficient). The light-emitting diode (LED) spectra were analyzed using Ocean Optics USB4000 spectrometer and the light intensity was measured by the Compact Power and Energy Meter Console (PM100D) equipped with a power sensor head (THORLABS, S310C).

Density functional theory (DFT) calculations
The electronic structure calculations were based on the experimental lattice parameters (a = b = 30.8173 Å and c = 8.8171 Å) of the crystal structure of DA3BiI6 [23]. All the DFT calculations were carried out using Quantum-ESPRESSO package [36]. Total energy calculations were performed within the linear density approximation (LDA), using Perdew-Zunger (PZ) exchange-correlation functional. H (1s, 2p, 3d, 4f), C (2s, 2p, 3d, 4f), N (2s, 2p, 3d, 4f), I (5s, 5p, 5d, 4f) and Bi (6s, 6p, 6d, 5f) orbitals were treated as valence orbitals. Since the crystal unit cell was large, a 2 × 2 × 4 k-point mesh was used to sample the Brillouin-zone of the crystal. The results of the calculations were checked for convergence with respect to the number of k points and the plane wave cut off energy. After the full relaxation of DA3BiI6, followed by self-consistent field calculations for total electronic densities of states (DOS), nonself-consistent field calculations along symmetry directions Г-K-H-Г was performed for its electronic band structure analysis. The reorganization energy (λi, i is hole or electron) was calculated based on Marcus theory [37,38], using the following equation [39] ( ) ( ) where 0 E  is the energy of ion calculation with the geometry of the neutral molecules, E  is the energy of ion in ion geometry. 0 E  is the energy of neutral molecules in ion state, and 0 E is the energy of neutral molecule at ground state geometry.

Photocatalytic reactions
The photocatalytic reactions were performed in a quartz vial (40 mL) equipped with a cooling system (6 °C) under a commercial white LED lamp (Blue Import BIM, LED, 100 W). The light intensity was 76.4 mW·cm −2 at a distance of 48 mm and the total optical power impinging on the solution was determined to be 422 mW. The photooxidation of I − ions to I3 − ions was conducted in 10 mL of DAI solution with or without DA3BiI6 powders and the evolved I3 − ions were determined by UV-vis spectroscopy after irradiation. For hydrogen evolution, the same system was used. Typically, 4 mL of H3PO2 was mixed with different amount of HI, and the overall reaction solution was kept as 10 mL by adjusting the amount of DAI solution. Then, a desired amount of DA3BiI6 and 1 mg of PtCl4 powder were added into the above mixture, followed by degassing with N2 gas for 30 min. The solution was irradiated with the commercial white LED lamp for 1 h under continuous magnetic stirring. The evolved hydrogen was analyzed with a Micro GC (Agilent 490G, MolSieve 5 Å column, with argon gas as carrier). The repeatability tests for H2 evolution were carried out by four consecutive photocatalytic experiments. Each time the reactor was irradiated with the white LED source for 1 h, and then bubbled with N2 for 10 min before the next cycle. After every run (4 h), the solution was kept overnight and then bubbled with N2 for 30 min before the next photocatalytic reaction. The apparent quantum efficiency (AQE) was determined under different monochromatic LED lamp (420, 450, 485 and 535 nm) from multichannel photochemical reaction device (PCX50B Discover, Beijing Perfect light Co. Ltd.). The measured light intensities were 41.7, 44.6, 18.8 and 13.3 mW·cm −2 , respectively. The reflectivity on the quartz-air interface was calculated from the Fresnel equation for normal incidence and thus about 4% of the incident intensity was reflected. The modified incident light intensities were 40.0, 42.8, 18.0 and 12.8 mW·cm −2 without considering the loss of intensity at quartzto-solution interface. The AQE values were calculated according to the following formula [14] ( ) where N(H2) is the number of evolved H2 molecules and N(photons) is the number of incident photons.

Structure and morphology
Here we report a novel solution-processed method for the synthesis of DA3BiI6 powder at 80 °C for 2 h (see details in Experimental section). This method is much easier than those of previously reported in Refs. [23,40,41], which were processed under solvothermal conditions or lasted for three months. The morphology of DA3BiI6 powder characterized by FESEM reveals rod-like crystals with smooth facets ( Fig. 1(a)), similar to those reported using a solvothermal method [23]. According to the XRD analysis ( Fig. 1(b)), the diffraction pattern of the product matches well with the previously reported rhombohedral DA3BiI6 [23,40]. The elemental composition and oxidation states of Bi of the product were further confirmed by XPS (Fig. 1(c)) revealing the presence of C, N, Bi and I with an atomic ratio of 6.05 for I/Bi (inset in Fig. 1(c)), which is consistent with the stoichiometric value in DA3BiI6 as well as the result from energy dispersive X-ray spectroscopy (EDS, Fig. S1 in the Electronic Supplementary Material (ESM)). The O 1s signal at 531.9 eV may originate from adsorbed O2 or ambient moisture [42]. The Bi 4f spectrum shows the characteristic binding energy located at 157.6 and 162.9 eV with a corresponding spin-orbit splitting of 5.3 eV (Fig. 1(d)) indicating Bi 3+ oxidation state in the product [43]. Combining the above analysis, it can be concluded that DA3BiI6 is synthesized successfully using a simple method that does not require autoclaves and elevated pressures as in solvothermal synthesis or long reaction period.

Stability of DA3BiI6 in DAI solution
Due to the ionic characteristics, halide perovskites generally exhibit extreme sensitivity to polar solvents, especially water. For instance, once MAPbI3 is in contact with water, it starts to decompose, forming MA + , I − ions and PbI2. From the viewpoint of chemical reaction, the low solubility of PbI2 in water is the main obstacle to rebuild MAPbI3 molecules, while the formation of PbI3 − ions in water from the reaction of PbI2 and I − ions seems promising to stabilize halide perovskites in water. Zheng et al.
proposed that the removal of I − ions from the surface of MAPbI3 triggers the initial dissolution of MAPbI3 in water [44]. On the other hand, degraded MAPbI3 can be regenerated through a treatment with MAI or I2 solution [45,46]. Such observations suggest that controlling the concentration of DAI might stabilize DA3BiI6 in water. As shown in Fig. 2(a), a series of aqueous DAI solutions, with concentrations up to 0.25 M were prepared and equal amounts of DA3BiI6 powder were added. In water, the DA3BiI6 powder decomposed into BiI3 immediately and the color of the powder changed from orange to black (see Fig. S2 in the ESM) whereas no visibly black-colored substance was observed in the presence of DAI, implying directly that DAI molecules can hamper the degradation of DA3BiI6 in water. The possible reason is that the addition of DAI to water leads to the dissolution of BiI3 and formation of iodobismuthate(III) ions (BiI4 − , BiI6 3− or BiI7 4− ) [47], thus stabilizing DA3BiI6 powder in water. To verify this hypothesis, after 1-day immersion, a 1/10 dilution of solution ( Fig. 2(b)) with acetonitrile was prepared and identified using UV-vis absorption spectroscopy. The spectral changes are displayed in Fig. 2(c). In water, no apparent absorption appeared between 300 and 600 nm, which are the characteristic absorption ranges of iodobismuthate(III) ions, implying the rather high stability of BiI3 in water. Increasing the DAI concentration to 0.05 M, the aqueous solution turned to yellow, which indicates the appearance of iodobismuthate(III) ions. Besides, two peaks appeared to UV-vis spectra located at 350 and 461 nm, which can be assigned to [BiI4]ions and BiI3 (Figs. S3 and S4 in the ESM), respectively [48]. Upon further increase of the DAI concentration (up to 0.15 M), the absorption peaks became stronger, suggesting the increasing amount of iodobismuthate(III) ions. Meanwhile, the peaks gradually shifted to 347 and 483 nm, that are reasonably ascribed to the characteristic peaks of BiI6 3− ions [48]. The difference (347 nm vs. 350 nm and 483 nm vs. 486 nm) for the peak positions may result from the presence of DAI in the aqueous solution, which is known to have a great effect on the absorption spectra [49]. The gradual shift of absorption peaks also suggests a stepwise BiI3→[BiI4] − →[BiI6] 3− transformation with the increase of DAI concentration. Above 0.15 M of DAI in the solution, no further peak shift in the spectra was observed, indicating that only BiI6 3− ions were present. In addition, no peaks above 500 nm appeared, precluding the existence of BiI7 4− ions in our current system [47].
To understand how DAI molecules stabilize DA3BiI6 in water (i.e. to confirm the role of (CH3)2NH2 + and I − ions in stabilization of DA3BiI6), three solutions were prepared: (1) 0.2 M aqueous dimethylamine, (2) 0.2 M aqueous KI, and (3) 0.2 M aqueous MAI. As shown in Fig. S5 in the ESM, the DA3BiI6 powder was deliberately immersed in water to obtain BiI3. Then, the BiI3 powder was collected and transferred into the above three solutions. Note that the black color of the powder turned into red in 0.2 M KI and 0.2 M aqueous MAI, whereas the color of solutions turned to yellow. The characteristic absorption peaks of BiI6 3− ions were also detected ( Fig. 2(d)), implying that the I − ions reacted with BiI3. In contrast, white powder was obtained in 0.2 M dimethylamine aqueous solution and BiI6 3− was not formed according to UV-vis absorption spectra. These observations mean that I − ions play an important role in the stabilization. On the basis of the above results, we could conclude that the I − ions from DAI can suppress the degradation of DA3BiI6 in water, without the use of encapsulation layers or the assistance of an acid.
Another aspect to be studied is related to the long-term stability of DA3BiI6 in DAI solution and the underlying reasons behind the stabilization. To examine the long-term stability, DA3BiI6 was mixed in 0.2 M DAI solution (all experiments reported below are for 0.2 M DAI solution unless otherwise stated) for two weeks, then the powder was collected (denoted as I-DA3BiI6) for further analysis. First, the composition before and after immersion was investigated by using ATR-FTIR spectroscopy, as shown in Fig. 3(a). Before the immersion, the DA3BiI6 powder showed the representative absorption peaks as reported before (see details in Fig. S6 in the ESM) [23]. After two weeks, new peaks at around 3,500 and 3,145 cm −1 appeared, while all other absorption peak positions remained the same. The weak broad peak centred at 3,500 cm −1 is due to O-H vibration of surface adsorbed water [50], whereas the doublet around 3,100 cm −1 can be assigned to the hydrogenbonded N-H (3,145 cm −1 ) and free N-H stretching vibration (3,093 cm −1 ) [41,51]. The adsorbed water does not cause the peak position shift of other absorption bands, especially N-H vibration, which indicates that the water just adsorbed on the surface of I-DA3BiI6 without reaching deeper in the bulk of the particles [50].
We also noted that the color of DA3BiI6 powder changed after immersion from orange to red (inset of Fig. 3(a)). It is well known that surface structural change or phase transformation can result in color difference [52,53]. Due to the ionic character of hybrid perovskites, the ions from halide perovskites can dissolve in water and then recrystallize at room temperature because of the low formation energy. For instance, this approach has been exploited for the synthesis of CsPbBr3 nanosheets from Cs4PbBr6 nanocrystals by aqueous phase exfoliation method [54]. Accordingly, we suspect that the color change in our work may originate from the reconstruction (i.e. dissolution of surface ions and subsequent recrystallization with I − and DA + ions in solution) of the surface of DA3BiI6 when exposed to aqueous DAI. During this process, the morphology and also the phase may be altered due to the difference between dissolution and recrystallization rates leading to altered optical scattering properties and consequently to a change in the color [55]. To verify this hypothesis, FESEM images before and after immersion were analyzed to study the change of microstructure. As displayed in Fig. 3(b), the surface of DA3BiI6 particles is smooth, whereas I-DA3BiI6 exhibits a rough surface and some plate-like particles formed on those (Fig. 3(c)). The rough surface leads to more exposed DA3BiI6 molecules in I-DA3BiI6. This result corroborates the increased intensity of I-DA3BiI6 powder in ATR-FTIR spectrum ( Fig. 3(a)).
Phase transformation may occur due to the different rate of dissolution and recrystallization upon surface DA3BiI6 molecules, thus forming plate-like particles. In addition, owing to the orderdisorder dynamics of the polar dimethylammonium cations, phase transformation is often observed in dimethylammonium metal hybrid compounds [56,57]. In this work, during the cooling process, first the formation of small plate-like particles was observed, and then rod-like products were obtained after keeping the samples at ambient conditions overnight. In DA3BiI6, DA + ions are assumed to originate from the protonated dimethylamine after the hydrolysis of dimethylformamide (generating dimethylamine and formic acid) under acid conditions [23,41]. As described by Snaith et al., the remained formic acid in precursor solution can induce the oriented growth of crystal [58]. Therefore, DA3BiI6 powder using DAI instead of DMF (marked as D-DA3BiI6) was tentatively prepared. As shown in Fig. 3(d), plate-like particles were obtained, similar to the newly formed particles in I-DA3BiI6. To confirm the phase, the XRD patterns of I-DA3BiI6 and D-DA3BiI6 were assessed (Fig. 3(e)). While no significant differences in the 2θ positions were found, the relative intensities of the reflections were quite different. This was likely due to the different morphology of I-DA3BiI6 and D-DA3BiI6 as shown in Figs. 3(c) and 3(d). The I-DA3BiI6 powder still maintained the rod-like shape, which will influence the peak intensity of XRD patterns. Thus, ethanol was used to etch the surface of DA3BiI6 powder (denoted as E-DA3BiI6). Ethanol was chosen, since DA3BiI6 powder dissolves in ethanol slowly instead of complete dissolution immediately as occurs in other common polar solvents such as, dimethylformamide and acetonitrile. The XRD patterns from I-DA3BiI6 and E-DA3BiI6 were further compared, and no indistinguishable change was observed. Thus, it is reasonable to conclude that the plate-like particles formed in I-DA3BiI6 originate from phase transformation of DA3BiI6 powder during the dissolution-recrystallization process. It is also worth mentioning, that common degradation products (BiI3, BiOI, DAI) were not found based on XRD (Fig. S7 in the ESM).
Furthermore, XPS analysis of the powders (Fig. 3(f)) indicates slight binding energy shift (0.1 eV) of Bi 3+ in I-DA3BiI6 and D-DA3BiI6 as compared to DA3BiI6. This may be related to a small change of microenvironments caused by the recrystallization [59]. The presence of BiI3 was further excluded due to the large difference in binding energy (157.6 eV for DA3BiI6 versus 159.6 eV for BiI3, Fig. S8 in the ESM). The additional peaks located at 155.6 and 160.9 eV (Δ = 5.3 eV) can be assigned to metallic bismuth [60], originating from partial reduction of trivalent bismuth after exposure to X-ray. Similar behavior was often observed in other metal halide materials [61][62][63]. The above results demonstrated that the stability of DA3BiI6 powder in DAI solution originates from the phase transformation through a dissolution-recrystallization process of surface DA3BiI6 molecules.

Photophysical and photocatalytic properties
Prior to photocatalytic performance evaluation, the optical properties and electronic structure of DA3BiI6 were studied. As concluded from UV-vis absorption measurements, the DA3BiI6 powder absorbs wavelengths up to ~ 570 nm (Fig. 4(a)). From the Tauc plot, the optical bandgap was determined to be 2.14 eV (the inset in Fig. 4(a)) assuming indirect transitions (no good linearity was observed for direct band gap, as shown in Fig. S9 in the ESM). This claims that the DA3BiI6 is an indirect band gap material [64], which is consistent with our theorical result ( Fig. 4(b)) from DFT calculations based on the reported crystal structure (hexagonal phase, Fig. S10 in the ESM) [23]. The valence band maximum (VBM) and conduction band maximum (CBM) appear on H point and Г point respectively, proving an indirect band gap. The absolute band gap from calculations (2.98 eV) compares reasonably well with the experimental data of 2.14 eV, if not considering the spin-orbit coupling effects in DFT calculations [65,66]. The carrier transport of DA3BiI6 was also predicted by calculating the reorganization energy based on Marcus theory [37,38]. The smaller reorganization energy means the faster carrier transport. The calculated reorganization energy is 73 meV for holes and 26 meV for electrons, respectively. Both are smaller than the reported values of MAPbI3 (410 meV for holes and 90 meV for electrons) [39], suggesting that DA3BiI6 is preferred for efficient transportation of the carrier, and holds the promise for better photocatalytic property than the lead analogs. The electronic structure was further elucidated from valence-band XPS spectrum (Fig. 4(c)) showing the VBM position of DA3BiI6 to be located at 1.2 eV versus Fermi level. Now, by considering the optical bandgap, the CBM is at −3.61 eV [67]. Thus, the CBM and VBM of DA3BiI6 are located at −0.85 and 1.29 V with respect to the normal hydrogen electrode (NHE) [68]. Note that the CBM of DA3BiI6 is at a more negative potential than the potential of proton reduction (Fig. 4(d)), which indicates that the photo-generated electrons would have sufficient driving forces to reduce proton and generate H2. However, since the VBM is more positive than the potential of iodine oxidation as compared to that of water oxidation, the photo-generated holes from DA3BiI6 prefer to oxidize I − ions to I3 − ions instead of oxidizing water to oxygen or I − ions to IO3 − ions [69].
The possibility of photocatalytic H2 evolution in DAI solution was firstly evaluated using a commercial LED as light source (100 W, photocatalytic system and light spectrum shown in the inset of Fig. 5(a) and Fig. S11 in the ESM, respectively). Although no H2 signal was detected after 24 h, we noted that the I − ions were oxidized to I3 − ions ( Although it was demonstrated that photooxidation reaction can be realized on DA3BiI6 in acid-free aqueous solution under visible light irradiation for the first time, the photocatalytic activity is still not adequate for H2 evolution. This possibly originates from the relatively high overpotential for water splitting, which requires four electrons, while only two electrons are needed for acid-splitting [70]. Therefore, photocatalytic H2 generation with DA3BiI6 in the presence of aqueous HI solution was examined. In this study, H3PO2 was added to alleviate the effect of evolved I3 − ions [11], and a series of HI solutions (namely, 1.52, 2.28, 3.04 and 3.8 M) were prepared by adjusting the amount of DAI solution. To optimize the photocatalytic efficiency, PtCl4 powder (1 mg) was added simultaneously as a co-catalyst (denoted as Pt-DA3BiI6) [23]. The hydrogen gas was measured with a micro gas chromatography. In the first place, the solubility of DA3BiI6 powder in HI solution was determined to be about 133 mg·mL −1 (Table S1 in the ESM), indicating forming of a homogeneous solution for all the evaluated cases. As displayed in Fig. 5(b), all examined cases showed photocatalytic H2 evolution activity under visible light irradiation suggesting the possibility of photocatalytic HI-splitting activity for DA3BiI6 powders in aqueous DAI solution. In particular, the amount of H2 evolution increased to the optimal value of 4.25 μmol for 3.04 M of HI solution after 1 h irradiation. Control experiments demonstrated that no H2 signals were detected in the absence of light or the photocatalyst, which means that H2 was formed via the photocatalytic reaction. The amount of DA3BiI6 on photocatalytic activity was also studied as shown in Fig. S14 in the ESM. The best hydrogen evolution activity was 4.55 μmol after 1 h reaction when 50 mg of DA3BiI6 was used. Moreover, Pt-DA3BiI6 showed good structural stability assessed by XRD (Fig. S15 in the ESM) and retained its photocatalytic activity for H2 evolution in repeated experiments with an average evolution rate of 5.7 μmol·h −1 (Fig. 5(c)). The apparent quantum efficiency was determined by applying monochromatic LED lamp (Fig. 5(d)). As shown in Fig. S16 in the ESM, the reaction solution absorbs all the light up to 520 nm, thus assuming that the total power was utilized by the photocatalyst. The H2 evolution rate (μmol·h −1 ) surpasses some reported data over Pt-MAPbX3 (X = I, Br) or Pt-Cs2AgBiBr6 for H2 evolution from concentrated HI solution (Fig. 5(e)) [12,13,15,19]. The smaller calculated reorganization energy as described above and the homogeneous system applied instead of heterogeneous system may contribute to the better performance [23,39].  monochromatic LED irradiation and the AQE values were 0.35%, 0.34%, 0.64% and 0.83% at 420, 450, 485 and 535 nm, respectively, indicating that this photocatalyst can work from the wavelength of 420 nm up to 535 nm. Though the performance is still low, the AQE values exceed the reported data (0.16% at 450 nm) of double perovskite Cs2AgBiBr6/RGO for H2 evolution from HBr-splitting (Fig. 5(d)). Note that as a co-catalyst, the reduced graphene oxide (RGO) displayed ca. 51 times and ca. 23 times higher photocatalytic H2 generation rate than Pt in Cs2AgBiBr6 and CH3NH3PbI3, respectively [12,19]. More importantly, the DA3BiI6 demonstrates hightolerance to water. So, we believe that there is great space to optimze its photocatalytic activity in low concentration of halogen acid solutions or pure water via electronic structure engineering or heterojunction engineering [71].

Conclusions
In summary, a lead-free bismuth halide perovskite DA3BiI6 was synthesized and stabilized in aqueous DAI solution over 2 weeks without the addition of acids. The I − ions from DAI molecules in aqueous solution can alleviate the degradation and thus stabilize the DA3BiI6 powder through a surface dissolution-recrystallization process when the concentration of DAI solution is higher than 0.15 M. The photophysical properties of DA3BiI6 were thoroughly studied by experimental and modelling approaches proving that DA3BiI6 as a photocatalyst is suitable for hydrogen evolution. As a proof of concept, DA3BiI6 was applied as a photocatalyst for photooxidation of I − ions as well as photocatalytic hydrogen evolution from HI in DAI solutions for the first time. Using PtCl4 as a co-catalyst. a H2 evolution rate of 5.7 μmol·h −1 during 4 runs with no significant decrease in activity was obtained, yielding an AQE of 0.83% at 535 nm. The bismuth-based perovskite developed in this work exhibits high-tolerance to water and can broaden the applications of halide hybrid perovskite family in photocatalytic systems. Open Access This article is licensed under a Creative Commons Attribution 4.0 International License, which permits use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons licence, and indicate if changes were made.
The images or other third party material in this article are included in the article's Creative Commons licence, unless indicated otherwise in a credit line to the material. If material is not included in the article's Creative Commons licence and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this licence, visit http://creativecommons.org/licenses/by/4.0/