Superior visible light hydrogen evolution of Janus bilayer junctions via atomic-level charge flow steering

Although photocatalytic hydrogen evolution (PHE) is ideal for solar-to-fuel conversion, it remains challenging to construct a highly efficient PHE system by steering the charge flow in a precise manner. Here we tackle this challenge by assembling 1T MoS2 monolayers selectively and chemically onto (Bi12O17) end-faces of Bi12O17Cl2 monolayers to craft two-dimensional (2D) Janus (Cl2)-(Bi12O17)-(MoS2) bilayer junctions, a new 2D motif different from van der Waals heterostructure. Electrons and holes from visible light-irradiated Bi12O17Cl2 are directionally separated by the internal electric field to (Bi12O17) and (Cl2) end-faces, respectively. The separated electrons can further migrate to MoS2 via Bi–S bonds formed between (Bi12O17) and MoS2 monolayers. This atomic-level directional charge separation endows the Janus bilayers with ultralong carrier lifetime of 3,446 ns and hence a superior visible-light PHE rate of 33 mmol h−1 g−1. Our delineated Janus bilayer junctions on the basis of the oriented assembly of monolayers presents a new design concept to effectively steer the charge flow for PHE.

P hotocatalytic hydrogen evolution (PHE) from water via solar energy and semiconductor photocatalysts provides the prospect of replacing fossil fuels with carbon-free and sustainable hydrogen energy 1 . Yet so far, its realization is still a mirage because PHE efficiency is thwarted to a large extent by the undesirable electron-hole recombination arisen from the random charge flow after the electron-hole separation 2 . The solution of this challenging issue requires our understanding and tuning of charge flow from the microcosmic perspective, preferably at the atomic level, which is still beyond the current state-of-the-art advances on the charge flow steering.
Recent flourishes in graphene studies are provoking tremendous interest in exfoliation-based fabrication of 2D single-layered nanosheets 3 . These advances breathe new life into PHE, as the as-exfoliated 2D monolayers with atomically thin thicknesses offer access to the atomic-level understanding and steering of charge flow [4][5][6][7] . Such monolayers enable atomiclevel control over architectures and electronic structures, thus manipulating the charge flow from the electron-hole separation sites to their surface in a desired route 8,9 . However, individual semiconductor monolayers still encounter substantial charge recombination for PHE, because they lack sufficient hydrogen-evolving sites to manipulate the fates of these surfacereaching electrons 10,11 . To address this issue and also realize the charge flow at the atomic level, it is highly desirable to load monolayers with abundant hydrogen-evolving sites, such as single-layered metallic transition metal dichalcogenides, selectively on electron-accumulated sites of semiconductor monolayers to extract electrons. Such assembly design may steer the separation, transportation and consumption of charge all at the atomic level, but at present remains unexplored.
In this study, we report on the synthesis of novel bilayer junctions by assembling metallic MoS 2 monolayers selectively and chemically on oxygen-deficient (Bi 12 O 17 ) end-faces of Bi 12 O 17 Cl 2 semiconductor monolayers via oxygen-vacancy (OV) chemistry. With the resultant 2D Janus motif of (Cl 2 )-(Bi 12 O 17 )-(MoS 2 ), we can successfully steer the charge separation, transportation and consumption all at the atomic level to construct a highly active PHE system. In such a system, electrons originating from visible light-irradiated Bi 12 O 17 Cl 2 are first driven by the internal electric field (IEF) between (Cl 2 ) and (Bi 12 O 17 ) layers to (Bi 12 O 17 ) end-faces, and further transferred via Bi-S bonds formed between Bi 12 O 17 Cl 2 and MoS 2 to MoS 2 for the final catalytic hydrogen evolution, while IEF also drives holes to (Cl 2 ) end-faces for the oxidation of organic scavengers.

Results
Assembly and characterization of the Janus bilayer junctions. Small-sized MoS 2 monolayers (1L-MS) enriched with metallic phase and large-sized oxygen-deficient Bi 12  ), which had a c ¼ 16.427 Å, a unit cell consisting of two Bi 2 WO 6 monolayers, but a monolayer thickness of 0.5cE0.8 nm as experimentally measured. The use of OV chemistry to craft the aforementioned Janus bilayer junctions (BOC-MS) was inspired by the intriguing capability of OV-deficient metal oxides to induce the selective deposition of metals on their OV sites 14 .
Enlightened by this methodology, we assembled the above two monolayers with a facile hydrothermal processing, where OVs induced the oriented anchoring of 1L-MS on (Bi 12 O 17 ) end-faces of Bi 12 O 17 Cl 2 to form 2D Janus bilayer junctions of (Cl 2 )-(Bi 12 O 17 )-(MoS 2 ). The key to this OV-oriented assembly lies in the metallic characteristic of 1L-MS and the asymmetric structure of 1L-BOC composed of only (Cl 2 ) layers and oxygen-deficient (Bi 12 O 17 ) layers. Transmission electron microscopy (TEM) image (Fig. 1a) of BOC-MS revealed a 2D heterostructured bilayer composed of many small nanosheets tightly wrapping on a large nanosheet, while the observed transparent nature indicated their ultrathin structures. Elemental mapping images (Fig. 1b-e) coupling with X-ray photoelectron spectroscopy (XPS) (Fig. 1f) and X-ray diffraction pattern (Supplementary Fig. 5 and Supplementary Note 5) identified the small and large nanosheets as MoS 2 and Bi 12 O 17 Cl 2 , respectively. S K-edge X-ray absorption near-edge structure spectra (XANES) (Fig. 1g) revealed that the MoS 2 monolayers in the bilayer showed a distorted 1T metallic phase 15 . 3D topographic atomic force microscopy (AFM) images (Fig. 1h,i) and their corresponding height profiles (Fig. 1j,k) (Fig. 1m). These observations clearly demonstrate the occurrence of an oriented assembly. As 1L-BOC has an asymmetric structure consisting of (Cl 2 ) and oxygen-deficient (Bi 12 O 17 ) end-faces and their assembly might be initiated by OVs of (Bi 12 O 17 ) end-faces, we could thus deduce that 1L-MS were anchored selectively on the (Bi 12 O 17 ) end-faces of 1L-BOC. The aberration-corrected high-angular annular dark field scanning TEM (HAADF-STEM) image (Fig. 1n) and energy loss spectroscopy (EELS) elemental maps ( Fig. 1o-s) of their cross-sectional atomic structures provided direct, atomic-resolution evidences that this oriented assembly resulted in 2D Janus bilayer junctions of (Cl 2 )-(Bi 12 O 17 )-(MoS 2 ). Meanwhile, we also found that the amount of MoS 2 loaded on Bi 12 O 17 Cl 2 decreased with OVs quenching, and the assembly behaviour annihilated when OVs were completely removed (Supplementary Figs 5-7,  Supplementary Table 1 and Supplementary Note 6 and 7), confirming that the oriented assembly of 2D Janus (Cl 2 )-(Bi 12 O 17 )-(MoS 2 ) bilayer junctions was driven by the OVs on the (Bi 12 O 17 ) end-faces of 1L-BOC.
Visible light PHE. The successful design of BOC-MS featuring Janus (Cl 2 )-(Bi 12 O 17 )-(MoS 2 ) bilayer junctions allowed us to first explore their PHE activity under visible light (l4420 nm). Visible light irradiation of the solution containing BOC-MS and ascorbic acid (organic scavengers) gave a typical gas chromatograph (GC) signal of H 2 at the retention time of 82 s (Fig. 2a), consistent with the previously reported results 16 . At the optimal MoS 2 loading (10 wt%, Supplementary Fig. 8a) and ascorbic acid concentration (0.3 mol l À 1 , Supplementary Fig. 8b), BOC-MS delivered a PHE rate of 33 mmol h À 1 g À 1 (Fig. 2b) and a quantum yield of around 36% at 420 nm (Fig. 2h), representing the currently achieved state-of-the-art PHE activity among all the MoS 2 -based, monolayer-based and bismuth oxyhalide-based PHE systems (Supplementary Tables 2-4). This activity could last over 100 h without significant decay and the bilayer structures were very stable (Supplementary Fig. 9 and Supplementary Note 8), indicating BOC-MS were highly active and robust for visible-light PHE. The absence of OVs signal in BOC-MS ruled out the possibility that its superior PHE activity was arisen from the OVs (Fig. 2c). During PHE, dehydroascorbic acid and its hydrated compound as the oxidized products of ascorbic acid were detected (Fig. 2d) 17 , indicating that ascorbic acid acted as an efficient hole scavenger to suppress the electron and hole recombination. Even using ascorbic acid as the hole scavenger and having surface OVs, 1L-BOC only exhibited a PHE rate of 0.86 mmol h À 1 g À 1 , 1/38 that of BOC-MS.
This giant PHE activity enhancement by the introduction of 1L-MS ignited our interest in unravelling how 1L-MS assisted 1L-BOC to precisely steer charge flow across the assembled Janus bilayer junctions. Density functional theory calculations of 1L-MS showed its density of states resided across the Fermi level ( Fig. 2e), revealing its metallic characteristic 3 , which indicated that 1L-MS favoured the electron collection. Mott-Schottky plots showed that 1L-BOC had the much more negative conduction band potential than 1L-MS ( Fig. 2f and Supplementary Fig. 10), suggesting that the electron transfer from 1L-BOC to 1L-MS was thermodynamically more favourable. As 1L-MS showed no photoresponse and its assembly onto 1L-BOC could enhance the photoresponse (Fig. 2g), we could conclude that 1L-BOC in the photoexcited BOC-MS solely offered electrons and these electrons were further transferred from 1L-BOC to 1L-MS for PHE (Fig. 2i). This conclusion was confirmed by the observation that the irradiation wavelength-dependent quantum yield of PHE matched well with the photoabsorption edge of 1L-BOC, not with those of 1L-MS or BOC-MS ( Fig. 2h and Supplementary Fig. 11).
IEF-directed charge flow within Bi 12 O 17 Cl 2 monolayers. To clarify these detailed charge flow processes, we first examined the charge flow in 1L-BOC. Charge density contour plots viewed from the {110} facet of Bi 12 O 17 Cl 2 (Fig. 3a) showed that the charge density surrounding (Bi 12 O 17 ) 2 þ layer was higher than that of (Cl 2 ) 2 À layer and their electrostatic potential differences (DE) was calculated to be 6.3 eV (Fig. 3b). This non-uniform charge distribution between (Bi 12 O 17 ) 2 þ and (Cl 2 ) 2 À layers in Bi 12 O 17 Cl 2 nanosheets would polarize the related atoms and orbitals to form a permanent IEF along [001] orientation perpendicular to the (Bi 12 O 17 ) 2 þ and (Cl 2 ) 2 À layers [18][19][20] . Using the Kanata's model (Supplementary Methods), which has been used to successfully identify the dependence of IEF intensity of Bi 3 O 4 Cl on its {001} facet exposure percentages 20 , we found that monolayer engineering could enhance the IEF magnitude of Bi 12 O 17 Cl 2 by sixfolds ( Supplementary Fig. 12). As 1L-BOC possessed an asymmetric structure composed of only (Cl 2 ) 2 À and (Bi 12 O 17 ) 2 þ layers ( Supplementary Fig. 3f), we proposed that its IEF would drive electrons and holes to (Bi 12 O 17 ) 2 þ and (Cl 2 ) 2 À end-faces, respectively. To confirm this atomic-level charge separation, we thus deposited Pt and MnO x on 1L-BOC with H 2 PtCl 6 and Mn(NO 3 ) 2 as the respective precursors under visible light. We determined their deposition sites through the interaction between Pt (or MnO x ) and the atoms at deposition sites via XPS. The Pt deposition resulted in a shift of O 1 s peak position toward lower binding energy without affecting Cl 2p peak position (Fig. 3c,d   These selective photo-depositions were further validated by Raman spectra (Supplementary Fig. 15), which revealed that the intensities of Bi-O and Cl-Cl stretching bands were suppressed by Pt and MnO x depositions, respectively. The atomic-resolution HAADF-STEM images (Fig. 3g,k) and EELS elemental maps ( Fig. 3h-j,l-n) of the cross-sectional atomic structures of 1L-BOC photo-deposited with Pt and MnO x provided direct and visual evidences that Pt was photo-deposited on the (Bi 12 O 17 ) end-faces and MnO x on the (Cl 2 ) end-faces. As Pt and MnO x depositions were induced by electron reduction and hole oxidation, respectively 21 , we could therefore conclude that, in 1L-BOC, IEF drove electrons to (Bi 12 O 17 ) end-faces and holes to (Cl 2 ) end-faces (Fig. 3o,p), achieving the atomic-level steering of charge separation. Although the 1L-BOC were rich in surface OVs, the selective deposition of Pt was independent on the OVs on the (Bi 12 O 17 ) end-surfaces, as evidenced by the oriented anchoring of Pt on OV-free (Bi 12 O 17 ) end-surfaces ( Supplementary Fig. 16). This phenomenon suggested the above charge separation was not induced by OVs, but further verified the role of IEF in steering the atomic-level charge separation. As a result of this directional charge separation, 1L-BOC afforded a carrier lifetime of 136 ns and a PHE rate of 0.86 mmol h À 1 g À 1 (Supplementary Fig. 17 (Fig. 3s). The energy barrier and transfer distance could be decreased by 5 eV and 2.2 Å, respectively, by the introduction of the Bi-S bonds (Fig. 3t), indicating that the electron transfer along Bi-S bond was energetically and spatially more favourable.
To support this theoretical calculation we used femtosecondresolved transient absorption (TA) spectroscopy to decode the build-up dynamics of electrons' excited-state absorption signals. The TA curves of both BOC-MS and BOC-MS-Mix gave three lifetime components (Fig. 3u and Supplementary Fig. 20), among which t 1 , t 2 and t 3 , with relative intensities I 1 , I 2 and I 3 , are rationally assigned to the electron migration along the Bi-S bonds, the electron trapping by surface states and the electron hopping within Van der Waals heterostructures 25,26 . It was found that I 1 increased ca. 17-fold from 5.23% for BOC-MS-Mix to 86.59% for BOC-MS, and its magnitude highly depended on the amounts of Bi-S bonds, confirming the electron transport from 1L-BOC to 1L-MS via the Bi-S bonds within BOC-MS (Fig. 3v). This interfacial electron transfer was ultrafast with a rate of up to 5.8 Â 10 À 11 s À 1 , as calculated by (t BOC-MS ) À Fig. 18c).
Atomic-level charge flow across the Janus bilayer junctions. To further verify the atomic-level charge flow described above and directly image the destination of electrons and holes in the photoexcited Janus bilayer junctions, we further deposited Pt and MnO x on BOC-MS under visible light. As shown in Fig. 4a-e,o-s, Pt preferred to deposit on MoS 2 surfaces (Fig. 4n), while MnO x tended to locate on the opposite surfaces, namely the (Cl 2 ) end-faces of Bi 12 O 17 Cl 2 (Fig. 4y). The Pt and MnO x photodeposition sites were further directly evidenced by the atomic-resolution HAADF-STEM images (Fig. 4f,t) and EELS elemental maps (Fig. 4g-j,u-x) of the cross-sectional atomic structures of BOC-MS photo-deposited with Pt and MnO x . When using H 2 AuCl 4 as the metal precursors, Au still photo-deposited on the MoS 2 surfaces in the Janus bilayers ( Supplementary  Fig. 21). This control experiment strongly supported that the spatially selective photo As shown in Fig. 4k-m, 1L-MS displayed much lower surface potential than 1L-BOC, indicating that the photogenerated electrons from the photoexcited BOC-MS were inclined to accumulate on the 1L-MS 28,29 . This phenomenon could also serve as a direct evidence for the steered electron transfer from 1L-BOC to 1L-MS. Besides, the Janus distribution of electrons and holes in BOC-MS also indicated the existence of a Janus photocatalysis behaviour, including the electron-induced hydrogen evolution on MoS 2 and the hole-mediated photo-oxidation of organic scavengers on (Cl 2 ) end-faces of Bi 12 O 17 Cl 2 ( Supplementary  Fig. 23).

Discussion
Subsequently, we quantified the efficiencies of charge transportations during the above two charge flow processes to probe the origin of the superior PHE performance of BOC-MS by using the following equations (Supplementary Methods).
In equation (1), Z abs (set as '1'), Z bulk , Z 1L-BOC , Z interface , Z surface and Z catalysis are the efficiencies of photoabsorption, bulk charge separation of BOC-MS, bulk charge separation of 1L-BOC, interfacial charge flow from 1L-BOC to 1L-MS, surface charge separation and hydrogen-evolving catalysis, respectively. In equation (2) and equation (3), J abs is the current density converted from the absorbed photons when assuming Z abs ¼ 1, while J ascorbic-acid and J water are the photocurrent densities measured by using ascorbic acid and water as electrolytes, respectively. The establishment of equation (2) is based on the assumption that ascorbic acid oxidation could completely suppress the surface recombination and its catalytic efficiency is 1, similar with the case that Kim et al. 30 used sulfite oxidation to determine the Z bulk of BiVO 4 . These experiments were performed under irradiation of 420 nm monochromatic light, which gave a J abs ¼ 0.788 mA cm À 2 . For 1L-BOC, J ascorbic-acid was measured to be 0.725 mA cm À 2 ( Supplementary Fig. 24), so Z bulk ¼ J ascorbicacid /J abs ¼ 0.92. This means, through the atomic-level charge separation steering by IEF, 1L-BOC enabled 92% of photogenerated electrons and holes to be transferred from the electron-hole separation sites to its (Bi 12 O 17 ) and (Cl 2 ) end-faces, respectively. Despite this ultrahigh Z bulk , the quantum yield of 1L-BOC was still as low as 0.8%, because of its ultralow Z surface (0.8%) and poor Z catalysis (10.8%; Supplementary Fig. 24d). When 1L-MS was selectively and chemically assembled on (Bi 12 O 17 ) end-faces, Z surface and Z catalysis increased by 11.4 and 4.2 times, respectively, as this oriented assembly enabled the Janus distribution of electrons and holes in BOC-MS to prevent the surface recombination and the metallic nature of 1L-MS was highly active for the hydrogen-evolving catalysis. Moreover, another crucial factor for the remarkable quantum yield (36%) of BOC-MS lied in the ultrahigh Z interface (91%) offered by the interfacial Bi-S bonds, by which the separated electrons of 1L-BOC could quickly migrate from (Bi 12 O 17 ) end-faces to 1L-MS, achieving atomic-level charge flow steering from 1L-BOC to 1L-MS.
In conclusion, we have designed Janus (Cl 2 )-(Bi 12 O 17 )-(MoS 2 ) bilayer junctions by assembling MoS 2 monolayers selectively and chemically on (Bi 12 O 17 ) end-faces of Bi 12 O 17 Cl 2 monolayers via OV chemistry. This atomic-level structural and interfacial design allowed us to steer all the charge separation, transportation and consumption at the atomic level. Electrons originating from visible light-irradiated Bi 12 O 17 Cl 2 were first driven by the IEF between (Cl 2 ) and (Bi 12 O 17 ) end-faces to (Bi 12 O 17 ) end-faces, and further transferred via the Bi-S bonds formed between Bi 12 O 17 Cl 2 and MoS 2 monolayers to MoS 2 monolayers to finally catalyse the hydrogen evolution. Meanwhile, the IEF drove holes to (Cl 2 ) end-faces where the organic scavenger was oxidized. Such an atomic-level steering of charge flow offered a visible-light PHE rate of 33 mmol h À 1 g À 1 with a quantum efficiency of 36% at 420 nm, superior to any reported MoS 2 , or monolayer, or bismuth oxyhalide-based PHE systems. This work sheds atomic-level mechanistic insights into the flow of photogenerated electrons and holes, thus paving new ways into the exploration and design of high-performance, cost-effective photocatalysts for hydrogen evolution.

Methods
Preparation of bulk layered Bi 12 O 17 Cl 2 nanosheets. First, 8 mmol of BiCl 3 was dissolved in 80 ml of ethanol with vigorous stirring at ambient temperature. Next, the pH value of the obtained homogeneous solution was adjusted to 13.4 by dropwise adding NaOH (1 mol l À 1 ) solution, yielding yellow precipitations. Then these precipitations were collected, washed thoroughly with deionized water and ethanol for several times, and dried in an oven at 50°C under vacuum. Calcination of the collected yellow powders in muffle furnace at 450°C for 2 h produced layered Bi 12 O 17 Cl 2 nanosheets, which we term BOC.
Preparation of bulk layered MoS 2 nanosheets. Typically, 1.5 mmol of (NH 4 ) 6 Mo 7 O 24 Á 4H 2 O and 36 mmol of thiourea were added in 75 ml of distilled water at room temperature with continuous stirring to give a transparent solution. The resulting mixture solution was then poured into a 100 ml Teflon-lined stainless autoclave. The autoclave was allowed to be heated at 200°C for 24 h under autogenous pressure, and then air cooled to room temperature. The resulting precipitates were collected, then washed with ethanol and deionized water thoroughly, and finally dried at 50°C under vacuum. We term the obtained sample MS.
Preparation of Bi 12 O 17 Cl 2 and MoS 2 monolayers. Organolithium chemistry was used to exfoliate layered MoS 2 (or Bi 12 O 17 Cl 2 ) nanosheets into their single-layered counterpart. Briefly, intercalating lithium into the spaces between each neighbouring MoS 2 (or Bi 12 O 17 Cl 2 ) monolayer-unit was first conducted by immersing 0.2 g of layered MoS 2 (or 0.1 g of Bi 12 O 17 Cl 2 ) nanosheets in 8 ml (10 ml for Bi 12 O 17 Cl 2 ) of n-butyllithium (Caution: n-butyllithium is highly pyrophoric.) under argon atmosphere for 72 h at 100°C. The suspension was then washed with hexane for several times to remove the excess of n-butyllithium. The collected nanosheets with lithium intercalation were re-dispersed in distilled water at 1 mg ml À 1 and sonicated at a low-power sonic bath (40 W) for 40 min to obtain exfoliated nanosheets. After centrifuging the resultant dispersions at 8,000 r.p.m. (8,000 r.p.m. for MoS 2 and 6,000 r.p.m. for Bi 12 O 17 Cl 2 ) for 10 min, the supernatants were collected, removing the unsuccessfully exfoliated nanosheets. The collected supernatants were centrifuged at 12,000 r.p.m for 10 min for several times to remove excess impurity, producing single-layered MoS 2 and Bi 12 O 17 Cl 2 nanosheets with yields of B47 and 26%, respectively. We term the as-prepared MoS 2 and Bi 12 O 17 Cl 2 monolayers 1L-MS and 1L-BOC, respectively.
It should be noteworthy that the surface of 1L-BOC was rich in oxygen vacancies, because, during the lithium-intercalation process, the organic ligands in n-butyllithium would interact strongly with lattice oxygen atoms on Bi 12 O 17 Cl 2 to form coordinate bonds, and during the ultrasonication-mediated exfoliation process, the organic ligands escaped from the interior of Bi 12 O 17 Cl 2 , which inevitably removed a fraction of oxygen atoms from the lattice, leading to the occurrence of surface oxygen vacancies in single-layered Bi 12  were dispersed in a three-neck flask containing 120 ml of distilled water under ultrasonication to obtain transparent solution B. Then the solution A was dropwise added to the solution B under vigorous stirring at ambient temperature to give homogeneous mixture solution. Next, the above mixture was deoxygenated by bubbling argon gas at room temperature for 60 min, subsequently refluxed at 80°C for 2 h under a stirring rate of 200 r.p.m., and finally air cooled to room temperature. The resulting suspension were first centrifuged at 3,000 r.p.m. for 5 min to remove the aggregated precipitations, and further centrifuged at 12,000 r.p.m. for 10 min to obtain the assemble of MoS 2 and Bi 12 O 17 Cl 2 monolayers, which we call BOC-MS.
We treated the synthesized 1L-BOC, 1L-MS and BOC-MS by many times of water-ethanol washing and plasma treatment to eliminate the surface-adsorbed impurities. Furthermore, when measuring their heights via AFM, we diluted the concentrations of the measured solutions to a considerably low level so as to maintain the nanosheets' surface as flat as possible, which would avoid the wrinkling or twisting.
Characterization. The powder X-ray diffraction patterns of the samples were recorded on a Bruker D8 Advance diffractometer with monochromatized Cu Ka radiation (l ¼ 0.15418 nm). The powders were deposited on copper grids with carbon support films for electron microscopy observation. TEM and HRTEM (high-resolution TEM) observations were performed on Hitachi H-7650 and JEOL-2010F with an acceleration voltage of 200 kV. The atomic-resolution HRTEM images and EELS elemental mapping images were investigated by aberration-corrected STEM and EELS carried out in a Nion UltraSTEM (Nion) microscope operating at 200 keV and equipped with a cold-field emission gun, a third-generation C3/C5 aberration corrector and a Gatan Enfinium EEL spectrometer. Energy dispersive X-ray spectroscopy were carried out using a JEM-ARM 200F Atomic-Resolution Analytical Microscope operating at an accelerating voltage of 200 kV. Elemental mappings were collected by a Gatan GIF Quantum 965 instrument. Ultraviolet-visible diffused reflectance spectra of the samples were obtained for the dry-pressed film samples with using a ultraviolet-visible spectrophotometer (UV-2550, Shimadzu, Japan) with BaSO 4 as the reflectance standard. Chemical compositions and states were analysed using XPS (Thermo Scientific ESCLAB 250Xi). All binding energies were referenced to the C 1 s peak (284.6 eV) arising from the adventitious carbon. Atomic concentrations were calculated by normalizing peak areas to the elemental sensitivity factor. Raman measurements were carried out by a confocal laser micro-Raman spectrometer (Thermo DXR Microscope, USA). The laser was 633 nm with a 5 mW. AFM and Kelvin probe force microscope measurements were carried out on an AFM instrument (SPM-9600, Shimadzu). The electrical resistivity study was corrected on a Keithley 4200 station with the computer-controlled four-probe technique. Photoluminescence emission and photoluminescence decay spectra were recorded at room temperature with a fluorescence spectrophotometer (Edinburgh Instruments, FLSP-920). The femtosecond-resolved TA spectra were performed using a modified ExciPro pump-probe spectrometer (CDP) in connection with an amplified femtosecond laser system (Coherent). Electron paramagnetic resonance spectra were performed on a Bruker EMX EPR Spectrometer (Billerica, MA). The Bi:O:Cl molar ratios in Bi 12 O 17 Cl 2 monolayers were detected by IRIS (INTREPID 2) inductively coupled plasma atomic emission spectrometry. The 13 C NMR spectra were recorded on a Bruker AVANCE III 600M system. The Bi L-edge X-ray absorption fine structure spectroscopy was carried out at BL 14W1 beamline at the Shanghai Synchrotron Radiation Facility (SSRF) China. Surface photovoltage spectroscopy was recorded on a lock-in amplifier (SR830-DSP, Stanford Research Systems) synchronized with a light chopper (SR540).
Photocatalytic hydrogen evolution. The photocatalytic hydrogen production experiments were performed at ambient temperature and atmospheric pressure using 300 W Xe arc lamp as the light source. In all, 10 mg of photocatalyst was dispersed in 80 ml of aqueous solution containing 0.3 M ascorbic acid in a 120 ml Pyrex flask. Before irradiation, the suspensions were bubbled with nitrogen for 30 min to remove the dissolved oxygen and to ensure the anaerobic conditions of the reaction system. During the whole reaction process, the aqueous solution with photocatalyst was continuously stirred by a magnetic stirrer. The generated hydrogen gas was analysed with an online GC (C36880-14, RESTEK) equipped with a thermal conductivity detector, where Ar was used as a carrier gas.
Theoretical calculations. Theoretical calculations were performed using density functional theory as implemented in the VASP code. 3D periodic boundary conditions were used to approximate an infinite solid. Exchange-correlation effects were described through the generalized gradient approximation, within the Perdew-Burke-Ernzerhof formalism. The core electrons (Bi:(Xe), Cl:(Ne), O:(He)) were treated within the projector augmented wave method. The energy cutoff is set to be 520 eV, and the atomic positions are allowed to relax until the energy and force are o10 À 4 eV and 10 À 3 eV Å À 1 , respectively.