Nanoarchitectonics on Z-scheme and Mott–Schottky heterostructure for photocatalytic water oxidation via dual-cascade charge-transfer pathways

The bottleneck for water splitting to generate hydrogen fuel is the sluggish oxidation of water. Even though the monoclinic-BiVO4 (m-BiVO4)-based heterostructure has been widely applied for water oxidation, carrier recombination on dual surfaces of the m-BiVO4 component have not been fully resolved by a single heterojunction. Inspired by natural photosynthesis, we established an m-BiVO4/carbon nitride (C3N4) Z-scheme heterostructure based on the m-BiVO4/reduced graphene oxide (rGO) Mott–Schottky heterostructure, constructing the face-contact C3N4/m-BiVO4/rGO (CNBG) ternary composite to remove excessive surface recombination during water oxidation. The rGO can accumulate photogenerated electrons from m-BiVO4 through a high conductivity region over the heterointerface, with the electrons then prone to diffuse along a highly conductive carbon network. In an internal electric field at the heterointerface of m-BiVO4/C3N4, the low-energy electrons and holes are rapidly consumed under irradiation. Therefore, spatial separation of electron–hole pairs occurs, and strong redox potentials are maintained by the Z-scheme electron transfer. These advantages endow the CNBG ternary composite with over 193% growth in O2 yield, and a remarkable rise in ·OH and ·O2− radicals, compared to the m-BiVO4/rGO binary composite. This work shows a novel perspective for rationally integrating Z-scheme and Mott–Schottky heterostructures in the water oxidation reaction.


Introduction
Since pressing energy and environmental issues are afflicting modern society, humans depend more and more on new sources of renewable energy. 1 To this end, converting inexhaustible solar energy into chemical bonds through articial photosynthesis, e.g., water reduction into hydrogen fuel, is desirable for energy storage. 2 Along with the process of photocatalytic water reduction, photocatalytic water oxidation (PWO) determines the overall reaction rate of water splitting. As a thermodynamically uphill process, the PWO reaction is a four-proton/four-electron step with sluggish kinetics, involving the dissociation of O-H chemical bonds as well as the formation of O-O chemical bonds. 3,4 Targeting high-efficiency PWO, monoclinic bismuth vanadate (m-BiVO 4 ) has stood out as one of the nest PWO catalysts since Kudo's report in 1999. 5 To date, m-BiVO 4 has been extensively studied due to its lone-pair effect, local electric eld, appropriate band structure, earth abundance, and low toxicity. [6][7][8] Despite vigorous development of single m-BiVO 4 , its PWO performance still falls short of expectations, owing to hampered carrier transfer and excessive carrier recombination, particularly on the surface and/or interface. 9 Among the booming examples of modied tactics, heterointerface charge modulation has proven to be a useful method to elevate charge separation within photocatalysts. 10 Nevertheless, there is still a lot of room to reconstruct m-BiVO 4 -based composites via heterojunction engineering, 11 which should ne-tune the surface and/or interface properties with regard to the PWO reaction.
The Mott-Schottky heterostructure constructed by coupling semiconductors and reduced graphene oxide (rGO) is supposed to exert the distinct advantages of each component. 12,13 Meanwhile, the heterointerface between adjacent components works for charge transfer and separation. With regard to m-BiVO 4 , rGO possessing surface oxygen-containing groups (OCGs) offers the prerequisites for anchoring m-BiVO 4 . Upon intimate coupling, the rGO component can darken the composite for light absorption, speed up electron transfer for high conductivity, and supply abundant active sites with large specic surface areas. 14 Signicantly, a high conductivity region at the heterointerface of m-BiVO 4 /rGO ensures spontaneous electron transfer over the Mott-Schottky heterojunction. 15,16 Therefore, when the m-BiVO 4 component enters into the excited state, its photogenerated electrons are conducive to ow toward rGO and can diffuse along a carbon network. Even though heterointerface charge modulation is implemented in the current m-BiVO 4 / rGO Mott-Schottky heterostructure, the m-BiVO 4 component with a two-dimensional (2D) layered morphology in fact possesses another bare surface, which is not in contact with the rGO component. In this context, some photogenerated charge carriers still undergo surface recombination on the m-BiVO 4 component, while others migrate to the heterointerface between m-BiVO 4 and rGO. 17,18 Therefore, without hindering the existing interfacial interaction and electron transfer in the m-BiVO 4 /rGO Mott-Schottky heterostructure, how to further overcome this surface recombination on the m-BiVO 4 component remains challenging.
Green plants are known to convert CO 2 and H 2 O into O 2 and carbohydrates through natural photosynthesis, whereby photogenerated electrons are transported under a Z-scheme mode. 19 Mimicking natural photosynthesis to build an articial Z-scheme heterostructure is an optimal route to suppress surface recombination on a single semiconductor. 20,21 The carbon nitride (C 3 N 4 ) candidate, as a typical 2D photocatalytic material, belongs to a redox-complementary component for the direct Z-scheme heterostructure. [22][23][24][25] Once the m-BiVO 4 /rGO (BG) Mott-Schottky heterostructure has been acquired, further coupling of C 3 N 4 with the m-BiVO 4 component may additionally construct the m-BiVO 4 /C 3 N 4 Z-scheme heterostructure. Between m-BiVO 4 and C 3 N 4 , the 2D/2D face-to-face contact heterointerface featuring an effective charge transfer is a promising candidate to tackle surface recombination on the m-BiVO 4 component. 26 However, so far, no studies have established a Zscheme heterostructure based on a Mott-Schottky heterostructure. Given that one surface of m-BiVO 4 is in contact with rGO to form a heterointerface, the other bare surface of m-BiVO 4 provides an opportunity to bear the m-BiVO 4 /C 3 N 4 Zscheme heterostructure. This additional 2D/2D intimate coupling can be predicted according to the large contact area, limited crystal boundary, and strong electronic coupling effect between the dimension-matched 2D layered materials. 27,28 Proting from the synergistic role of the Mott-Schottky and Zscheme heterostructure, especially the spatial separation of carriers and the sustained high redox potentials of components in the latter, the C 3 N 4 /m-BiVO 4 /rGO (CNBG) ternary composite with face-to-face contact is speculated to achieve excellent PWO performance.
Herein, using the m-BiVO 4 /rGO Mott-Schottky heterostructure as a prototype, we also established the m-BiVO 4 /C 3 N 4 Z-scheme heterostructure. The obtained CNBG ternary composite with face-to-face contact is postulated to solve the excessive surface recombination during the PWO reaction. To study the synergistic role of this dual heterostructure, particularly the virtues of Z-scheme electron transfer, extensive investigation of physical structures, chemical status, optoelectrical features, and catalytic performance was performed, and the related PWO mechanism is further illustrated. As expected, the CNBG ternary composite shows over 193% growth in O 2 yield compared with the binary composite and this PWO performance boost is validated by the increase in reactive oxygen species.

Results and discussion
CNBG was fabricated using a multi-step process. In detail, the pristine C 3 N 4 was synthesized via thermal polymerization from pre-heated melamine and chloride precursors, and the BG binary composite was obtained via hydrothermal treatment. The CNBG ternary composite was synthesized by simply mixing C 3 N 4 and BG in deionized water followed by freeze-drying. To investigate the formation of CNBG, the zeta potentials of pristine C 3 N 4 , BG, and CNBG were measured in aqueous solution (Fig. S1 †). The decrease in values from −8.7 mV for BG to −30.9 mV for CNBG is due to the incorporated C 3 N 4 (−41.5 mV), certifying that the CNBG ternary composite was smoothly obtained based on the BG binary composite.
The physical structures were unveiled by microstructure models (Fig. 1a-c), scanning electron microscopy (SEM), and transmission electron microscopy (TEM) images ( Fig. 1d-i). The pristine C 3 N 4 shows irregular, lamellar-like, and stacking layers with a rough surface ( Fig. 1d and g), and these structural characteristics accord with C 3 N 4 reported previously. 29,30 In terms of the BG binary composite, the sheet-shaped BiVO 4 is uniformly distributed on rGO, the size of the BiVO 4 nanosheets exceeds 100 nm, and the surface of rGO seems wrinkled and smooth ( Fig. 1e and h). Aer the incorporation of C 3 N 4 in the BG binary composite, the BG is covered by C 3 N 4 to form the CNBG ternary composite, which is like a "sandwich"-the top layer is the smooth rGO with translucence under electron beam imaging, the bottom layer is the C 3 N 4 with a rough surface, and the m-BiVO 4 nanosheets with size over 100 nm are co-wrapped between the top and bottom layers, in perspective view ( Fig. 1f and i). The face-to-face contact of the three components in CNBG, in particular, is desirable for heterointerface electron transfer, whereas their similar Brunauer-Emmett-Teller (BET) specic surface areas imply that the incorporated C 3 N 4 makes no difference to the overall surface area of CNBG (Fig. S2 †).
Elemental mapping images show the homogeneous distribution of C, N, O, Bi, and V elements across the CNBG (Fig. 1jn). The existing areas of C and N elements are consistent with C 3 N 4 , the shapes of the Bi, O, and V elements are identical to m-BiVO 4 nanosheets, and the partial regions of C and O elements can be attributed to rGO. These images corroborate a ternary hybrid architecture, namely the m-BiVO 4 co-wrapped by C 3 N 4 and rGO with intimate contact. The characterization of BG and CNBG from high-resolution TEM (HRTEM) shows an interplanar spacing of 2.55 Å for the (002) spacing of m-BiVO 4 (insets in Fig. 1h and i), while the lattice fringes of C 3 N 4 cannot be detected due to the poor crystallinity. Therefore, the incorporation of C 3 N 4 shows no impact on the morphology and crystal structure of m-BiVO 4 .
To further reveal the crystal structure information, X-ray diffraction (XRD) peaks for BG and CNBG, except for a peak at 27.8°, are both indexed to the m-BiVO 4 (Fig. 2a), and their Raman bands (100 to 1000 cm −1 ), except for a peak at 734 cm −1 , are also assigned to m-BiVO 4 (Fig. 2b). These results agree with the HRTEM images and validate the m-BiVO 4 being highly crystalline and phase pure, excluding any impact from the incorporated C 3 N 4 on the crystal structure of m-BiVO 4 . Distinct from the BG binary composite, the CNBG ternary composite shows a small peak for the C 3 N 4 (002) plane at 27.8°(in Fig. 2a), which belongs to the interlayer stacking of the conjugated aromatic segments. 31 However, this peak in CNBG is much weaker than that in pristine C 3 N 4 , indicating decreased crystallinity within the C 3 N 4 framework aer its incorporation. A similar phenomenon is apparent in the Raman spectra (Fig. 2b), where the CNBG ternary composite presents a spike at 734 cm −1 and a broad peak (1000 to 2250 cm −1 ) relative to the BG, which are consistent with in-plane bending and disordered graphitic carbon-nitrogen vibrations, respectively . 32 These additional peaks also emerge for pristine C 3 N 4 , suggesting the structure of C 3 N 4 is preserved in CNBG aer its incorporation. All these ndings unambiguously reect that the CNBG ternary composite was subtly constructed, and the sandwich-like structure with face-to-face contact is potentially useful for photocatalytic reactions.
To glean precise information on the heterointerface between m-BiVO 4 and C 3 N 4 in CNBG, the surface chemical status was analyzed by X-ray photoelectron spectroscopy (XPS). The change in binding energy reects the interfacial chemical interaction and the different electron densities in the semiconductorsemiconductor heterojunction. The electron density increases when the binding energy shis negatively, and decreases when the binding energy shis positively. 33 As shown in Fig. 2c, the Bi 4f spectrum of CNBG shows two peaks at 164.2 and 158.9 eV assigned to the binding energies of Bi-O chemical bonds; however, these peaks are negatively shied compared with BG. Parallel negative shis can be found in V 2p and O 1s spectra as well; the binding energies of the V-O and V-C chemical bonds appear at 523.4 and 516.5 eV, 524.5 and 517.4 eV in CNBG, respectively, while these peaks are located at higher values in BG (Fig. 2d). The three peaks at 533.1, 531.2, and 529.8 eV in CNBG (Fig. 2e) correspond to small molecules, bridging hydroxyls, and lattice oxygen, 34 respectively, and together they are reduced relative to those in BG. In contrast, the N 1s spectra show an opposite trend (Fig. 2f). The binding energies of the C-N-H, N-(C) 3    heterostructure in the CNBG ternary composite. The electrons migrate from C 3 N 4 to m-BiVO 4 as a result of the Fermi level equilibrium upon hybridization. 26,35 The C 1s spectra uncover whether the interfacial chemical interaction and the electron transfer are affected by the incorporation of C 3 N 4 (Fig. S3 †). The C 1s spectrum of graphene oxide (GO) was tted into three peaks at 288.3 eV (C]O chemical bonds), 287.4 eV (C-O chemical bonds), and 285.3 eV (C]C chemical bonds). 36 Aer in situ growth of m-BiVO4 on rGO, the C 1s spectrum of BG was divided into two peaks, at 285.8 eV (C-O chemical bonds) and 284.7 eV (C]C chemical bonds). Compared with BG, the C 1s spectrum of CNBG shows an additional peak at 288.6 eV (N]C-N chemical bonds). 37 Conforming to the Bi 4f, V 2p, O 1s, and N 1s spectra results, the C 1s spectral analysis reconrms that the C 3 N 4 was incorporated into CNBG. In addition, there is an overall shi in the peaks over both BG and CNBG relative to GO, indicating the emergence of chemical bonds involving C atoms, i.e., the V-C chemical bonds in the m-BiVO 4 /rGO Mott-Schottky heterostructure for both BG and CNBG (Fig. 2d). As a result, the higher Fermi level of rGO compared with that of m-BiVO 4 results in an electron rearrangement, 38,39 leading to a high conductivity region as well as downward band bending within m-BiVO 4 . When the m-BiVO 4 absorbs visible light to generate electronhole pairs, the rGO can accept electrons from m-BiVO 4 through this high conductivity region in both BG and CNBG. A similar binding energy shi was found between black phosphorus (BP) and m-BiVO 4 /BP in the P 2p spectra in a previous study, 40 indicating a strong interfacial chemical interaction between BP and m-BiVO 4 , as well as the electron transfer from m-BiVO 4 to BP during photocatalytic reactions. Thus, the incorporated C 3 N 4 makes no difference to both the V-C chemical bonds in CNBG and the electron transfer over the Mott-Schottky heterojunction.
When electrons transfer toward rGO, its degree of reduction determines the electron diffusion ability. In the C 1s spectra (Fig. S3 †), the C]O and C-O chemical bonds of GO represent the OCGs, contributing to the in situ growth of m-BiVO 4 . For both BG and CNBG, the C-O peaks shied more than the C]C peaks, causing shorter distances between the C-O and C]C peaks than for GO. This phenomenon indicates that some OCGs of GO turned into V-C chemical bonds in both BG and CNBG. Moreover, the absence of C]O peaks and diminished C-O peaks relative to GO indicate the removal of other OCGs, which agrees with the results of the Raman spectra. As the Raman bands range from 1000 to 2000 cm −1 (Fig. S4 †), both D band (sp 3 carbon defects) and G band (sp 2 carbon atoms) are derived from the carbon network within the rGO. The lower intensity ratio (I D /I G ) in BG (0.79) than in GO (1.37) reveals that the rGO is partially reduced in BG. Although a broad peak induced by disordered graphitic carbon-nitrogen vibrations overlaps with both the D band and G band in CNBG, a similar degree of reduction of rGO in CNBG can be speculated based on the lower I D /I G in BG, as well as a facile composite process in CNBG. Therefore, Raman spectra and C 1s spectra elaborate an effective degree of reduction of rGO in both BG and CNBG, achieving a high conductivity and electron diffusion capacity for the carbon network aer accepting electrons.
An assessment of the energy band structure of the components is crucial to evaluate the origin of the charge carrier behavior. The photoresponse range of the photocatalysts was disclosed by UV-visible (UV-vis) absorption spectra (Fig. 3a), and pristine C 3 N 4 and BG display discrete optical absorption edges attributed to their intrinsic band gap transitions. As for CNBG, there is a ladder-like optical absorption edge resulting from the overlapped band gap transitions of the C 3 N 4 and BG components, which is consistent with the color change of the three photocatalysts (Fig. S5 †). Furthermore, the CNBG exhibits similar absorption features to pristine C 3 N 4 and BG, implying that the C 3 N 4 linked with m-BiVO 4 at the heterointerface is not incorporated into the m-BiVO 4 lattice. This inference agrees with the XRD and Raman results. The optical band gap of C 3 N 4 and BG can be quantitatively analyzed from Tauc plots (i.e., the curve of (ahv) r vs. hv, where r = 2) for direct band gap semiconductors of m-BiVO 4 (ref. 41) and C 3 N 4 . 42 As shown in Fig. 3b, the optical band gap of BG is 2.47 eV and is smaller than that of C 3 N 4 (2.69 eV), which is in harmony with their discrete optical absorption edges. Despite the larger optical band gap of C 3 N 4 than of BG, both C 3 N 4 and BG are able to realize visible light absorption. Hence, for the CNBG ternary composite, the m-BiVO 4 component and C 3 N 4 component concurrently trigger abundant photogenerated carriers upon visible light irradiation. Besides the optical band gap, proper matching of the conduction band (CB) and valence band (VB) positions of the components with the redox potentials of photocatalytic reactions cannot be neglected. Measured by XPS VB spectra (Fig. S6a †), the VB positions of BG and C 3 N 4 are +1.59 eV and +1.29 eV, respectively. Therefore, the VB position of the m-BiVO 4 component in both BG and CNBG meets the thermodynamic demands for the PWO. Based on (ahv) r vs. hv curves and XPS VB spectra, the CB positions of BG and C 3 N 4 were calculated as −0.88 eV and −1.40 eV, respectively. Hence, the approximate band positions of BG and C 3 N 4 were drawn vs. a normal hydrogen electrode (NHE, pH = 7) in Fig. S6b. † Once the C 3 N 4 was incorporated into the BG binary composite, a staggered energy band structure formed with the Fermi level of C 3 N 4 much higher than that of m-BiVO 4 . 23, 43 The electrons of C 3 N 4 would be transferred toward m-BiVO 4 via the Bi-O-N and V-O-N chemical bonds until the Fermi level equilibrium was reached, as corroborated by the Bi 4f, V 2p, O 1s, and N 1s spectral results. This spontaneous electron redistribution creates an electron depletion layer with upward band bending in C 3 N 4 , and an electron accumulation layer with downward band bending in m-BiVO 4 . Hence, an internal electric eld (IEF) with a direction from C 3 N 4 to m-BiVO 4 appears at the heterointerface. Under visible light irradiation, both C 3 N 4 and m-BiVO 4 components became excited states with electron transition from the VB to the CB. The IEF can promote the transfer of electrons in the CB of m-BiVO 4 across the heterointerface to consume holes in the VB of C 3 N 4 , i.e., a Z-scheme electrontransfer pathway. 26,35 In consequence, spatial separation of electron-hole pairs occurs in the heterointerface between m-BiVO 4 and C 3 N 4 in CNBG, preserving the strong reducing electrons in the CB of C 3 N 4 and the strong oxidizing holes in the VB of m-BiVO 4 .
To illustrate the positive effect of the Z-scheme process on PWO activity, the activity was evaluated under visible light irradiation. Controlled tests revealed no O 2 yield without photocatalysts or visible light. The CNBG exhibited an O 2 yield of 1000.8 mmol g −1 (Fig. 3c), over 193% of that of BG, in the rst 5 h, while a gap in the PWO performance between BG and a mixture of BG and C 3 N 4 was small (within experimental error). Therefore, a strong interfacial chemical reaction, together with the Z-scheme process, explains the enhanced performance in CNBG. The pristine C 3 N 4 showed scarce O 2 yield owing to its weak PWO driving force induced by an unsuitable VB position. Also, the apparent quantum efficiency (AQE) was qualitatively analyzed with the help of different band-pass lters (Fig. 3d). The enhanced PWO performance of CNBG explains the maximum of 26.35% at 420 nm and is 1.42-fold higher than that of BG. As the light wavelength continues to increase, the gradually reduced AQE of both BG and CNBG validates that the PWO reaction is dependent on visible light absorption. However, the sharply decreased AQE of CNBG began at 475 nm, compared with BG. Since the m-BiVO 4 component can be excited by visible light even up to 520 nm, whereas visible light beyond 475 nm cannot fully excite the C 3 N 4 component (Fig. 3a), the photogenerated holes in the C 3 N 4 component were decient in annihilating photogenerated electrons in the m-BiVO 4 component. This deduction supports the proposed Z-scheme pathway in CNBG. A comparison of the AQE of CNBG with the applied photocatalysts is provided in Table S1, † certifying that the Z-scheme heterostructure along with the Mott-Schottky heterostructure enables a superior PWO performance for CNBG relative to previous photocatalysts.
Except for the pristine C 3 N 4 with scarce O 2 yield, both BG and CNBG were collected to measure their durability. As shown in Fig. 3e, they both exhibited continuous O 2 yield in each cycle. Considering that loss of photocatalyst in each cycle caused a decrease in performance, the CNBG still maintained over 374% of the performance of BG in the h cycle. XRD patterns and the TEM images conrm that the crystal phase and morphology of CNBG were sustained aer the cycling tests (Fig. S7 †), proving its reusability and photostability. To survey the impact of the incorporated amount of C 3 N 4 on the PWO performance, CNBGs with different C 3 N 4 /BG molar ratios were prepared as reference samples (Fig. 3f). Evidently, the PWO performance continually decreased when the molar ratio Molecular oxygen activation measurements were performed to uncover the inuence of the Z-scheme process on reactive oxygen species. Assisted by 5,5-dimethyl-L-pyrroline-N-oxide (DMPO), $O 2− radicals and $OH radicals were both detected. Electron spin resonance (ESR) signals from the DMPO − $O 2− adduct with light were found for pristine C 3 N 4 , BG, and CNBG (Fig. 4a), with $O 2− radicals from O 2 reduction by electrons; while peaks from the DMPO − $OH adduct with light were observed for BG and CNBG (Fig. 4b), with $OH radicals originating from OH − /H 2 O oxidation by holes. By contrast, almost no DMPO − $OH adduct formed with light for pristine C 3 N 4 , which agrees with its scarce O 2 yield under irradiation and was due to the VB position being too negative thermodynamically to oxidize OH − /H 2 O. Thereby $OH radicals were mainly gathered in the m-BiVO 4 component, while $O 2− radicals were accumulated in the rGO component, or the C 3 N 4 component in CNBG, supporting that the electron transfer follows the Mott-Schottky and Z-scheme pathway. Notably, the stronger signals from two adducts for CNBG than for BG affirm that more photogenerated electrons and holes were formed in CNBG, which can be ascribed to the increased spatial separation of the electron-hole pairs and the improved redox capability. Hence, based on the Mott-Schottky heterostructure of the BG binary composite, an additional Z-scheme heterostructure endows the CNBG ternary composite with formidable molecular oxygen activation ability, as well as superior PWO performance.
Considering the prominent photocatalytic behavior of CNBG, we carried out controlled experiments to probe the reaction mechanism. To begin with photoluminescence (PL) spectra were obtained to survey the electron-transfer pathway (Fig. 5a). The pristine C 3 N 4 and BG displayed two peaks centered at 469 and 511 nm, respectively, assigned to the intrinsic band gap emission. On the other hand, the CNBG showed a broad peak from 445 to 485 nm, which is not matched with the peaks of either pristine C 3 N 4 or BG. The emission from the broad peak was assigned to the multiple heterointerface recombination of electrons in the CB of the m-BiVO 4 component and of holes in the VB of the C 3 N 4 component via Z-scheme electron transfer. In addition, the CNBG showed stronger uorescence intensity than the BG. Although the high conductivity region that favors electron transfer exists over the Mott-Schottky heterostructure, the incorporated C 3 N 4 can accelerate radiative recombination via the Z-scheme process, leading to uorescence enhancement for CNBG. Moreover, time-resolved PL (TRPL) spectra were applied to further investigate the electron-transfer pathway (Fig. 5b). Based on the parameters from biexponential tting (Table S2 †), the uorescence lifetime (s av ) value decreased from 4.86 ns for BG to 3.38 ns for CNBG. Despite a high conductivity region over the Mott-Schottky heterostructure, the decreased s av reconrms an additional Zscheme process that quenches low-energy electrons and holes in CNBG, in accord with the uorescence enhancement described above. Therefore, the strong reducing electrons in the C 3 N 4 component and strong oxidizing holes in the m-BiVO 4 component are maintained with spatial separation, ultimately strengthening the photocatalytic behavior of the CNBG. As for the pristine C 3 N 4 , the strongest uorescence intensity and the shortest s av imply its intrinsic high recombination rate of charge carriers without construction of a Z-scheme heterostructure.
Photoelectrochemical measurements were also employed to gain insight into the reaction mechanism. The electrochemical impedance spectroscopy (EIS) Nyquist plots were recorded under irradiation, and the arc radius was related to the interfacial transfer resistance (Fig. 5c). A smaller arc radius for CNBG than for BG veried the signicance of an additional Z-scheme process in CNBG; namely, the spatial separation of strong reducing electrons and strong oxidizing holes in respective components raised the interfacial transfer efficiency. Measurements of transient photocurrent responses were conducted to evaluate the effect of the appended Z-scheme process on the photoelectron conversion efficiency (Fig. 5d). The photocurrent intensity of CNBG was 2.85 times that of BG, proving that more carriers are generated in double semiconductors and spatially separated via the Z-scheme process in CNBG. The almost unchanging photocurrent intensity with repeated on/off cycles illustrates the ideal photostability of CNBG. In addition to the results of the PL and TRPL spectra, the biggest arc radius and the lowest photocurrent intensity reconrm the serious charge carrier recombination in pristine C 3 N 4 .
Taking these results together, the PWO mechanism over CNBG is summarized as follows (Fig. 6): (1) on constructing the CNBG ternary composite, the sandwich-like hybrid architecture with chemical bonding fullls the requirement for intimate face-to-face contact of the three components; (2) until the Fermi level equilibrium is reached, a high conductivity region exists at the m-BiVO 4 /rGO heterointerface and an IEF appears at the m-BiVO 4 /C 3 N 4 heterointerface; (3) both C 3 N 4 and m-BiVO 4 components achieve excited states, with electron transition from the VB to the CB upon visible light irradiation; (4) the rGO accepts some photogenerated electrons from m-BiVO 4 through a high conductivity region, and these electrons soon rapidly diffuse along the carbon network within rGO; (5) the IEF, meanwhile, promotes other photogenerated electrons in the CB of m-BiVO 4 , consuming the low-energy photogenerated holes in  Nanoscale Advances the VB of C 3 N 4 via the Z-scheme process; (6) the feasible spatial separation of electron-hole pairs sustains plenty of strong oxidizing holes in the VB of m-BiVO 4 , accomplishing the subsequent PWO reaction. As a consequence, an innovative strategy is proposed to develop multicomponent photocatalysts with dual-cascade charge-transfer pathways, aimed at highefficiency articial photosynthesis.

Conclusions
Taking the unresolved surface recombination on the m-BiVO 4 component in the BG binary composite into account, we mimicked natural photosynthesis to establish a new m-BiVO 4 / C 3 N 4 Z-scheme heterostructure based on the m-BiVO 4 /rGO Mott-Schottky heterostructure. As a result, the face-contact CNBG ternary composite was elaborately constructed for PWO. In the heterointerface between m-BiVO 4 and rGO, the rGO accepted photogenerated electrons from m-BiVO 4 through a high conductivity region, achieving further electron diffusion along a carbon network. Noteworthy are the heterointerface between m-BiVO 4 and C 3 N 4 , the IEF-promoted electrons in the CB of m-BiVO 4 consuming holes in the VB of C 3 N 4 upon irradiation, the spatial separation of the strong reducing electrons in the CB of C 3 N 4 , and the strong oxidizing holes in the VB of m-BiVO 4 . With these merits, the CNBG showed an O 2 yield of 1000.8 mmol g −1 , which is over 193% that of BG, together with an AQE of 26.35% at 420 nm, which is 1.42-fold higher than that of BG. In addition, the CNBG possessed excellent reusability and photostability. We have not only incorporated C 3 N 4 into the BG binary composite to resolve carrier recombination on the bare surface of the m-BiVO 4 component, but also unravelled the synergistic role of Mott-Schottky and Z-scheme heterostructures in a ternary composite, thus emulating the interfacial chemistry in articial photosynthesis.

Materials synthesis
All chemicals were of analytical grade and were used without further purication. A GO sol-gel solution was prepared according to a previous study, 45 and the pristine C 3 N 4 was obtained by thermally exfoliating a mixture of pre-heated melamine and chloride according to a previous report. 46 In a typical synthesis of m-BiVO 4 grown in situ on rGO (BG), a GO sol-gel solution (containing 100 mg of GO) was dispersed in 85 mL of deionized water. Then, 631 mg BiCl 3 and 400 mg cetyltrimethylammonium bromide were dissolved into the suspension by stirring for 4 h. Aerwards 234 mg NH 4 VO 3 was dissolved into the suspension and stirred for 1 h. Soon aer, excess 5 M aqueous ethanolamine was added into this suspension until pH 10, then the pH was mediated to 6.2 by adding 2 M hydrochloric acid. The solution was poured into an autoclave and heated at 160°C for 12 h. Aer cooling, the product was centrifuged, washed, and collected by freezedrying.
In a typical synthesis of CNBG, 70 mg of pristine C 3 N 4 was dispersed in 80 mL of deionized water, and 80 mg of pristine BG was added into the mixed solution under stirring for 24 h. The product was centrifuged, washed, and collected by freezedrying. Other reference samples were prepared under the same conditions but with different C 3 N 4 /BG molar ratios ranging from 0.75 to 1.50.

Material characterization
Zeta potentials were measured in deionized water using a Zetasizer Nano ZS90. SEM images were recorded on a JEOL JSM-6360LV electron microscope. TEM and HRTEM were performed on a JEOL JEM-ARM200F electron microscope. The crystal phase was conrmed by X-ray diffraction using a Rigaku D/max 2550VL/PC system. Raman spectra were performed using a LabRAM HR system at 633 nm. XPS spectra were collected on an ESCALAB 250Xi instrument; the binding energies were referred to the C 1s peak from adventitious carbon. UV-vis absorption spectra were recorded with a PerkinElmer Lambda 750S UV-Vis-NIR spectrophotometer. The BET specic surface area was measured by using a nitrogen adsorption apparatus. The ESR spectra were acquired on a Bruker Elexsys E580 spectrometer with DMPO as the spin-trapping reagent. PL spectra were obtained on a Shimadzu RF-5301PC spectrophotometer at 365 nm. The uorescence lifetime was quantied by the TRPL decay spectra on an Edinburgh FLS1000 spectrophotometer at 365 nm.

Photoelectrochemical measurements
EIS spectra and transient photocurrent responses were obtained with 0.5 M Na 2 SO 4 electrolytes in a standard threeelectrode cell on a CHI 600E electrochemical workstation (Chenhua, China). Indium tin oxide (ITO) deposited with synthesized samples, platinum wires, and Ag/AgCl electrodes were utilized as the working electrodes, counter electrodes, and reference electrodes, respectively. To prepare the working electrode, 10 mg of sample and 500 mL of 5 wt% Naon solution were dispersed in 5 mL of water-ethanol solution (V water /V ethanol = 1 : 4) to form a homogeneous slurry, which was then spincoated on 1.5 cm × 3.0 cm ITO and exposed to air. All working electrodes were prepared with the same conditions to guarantee the same loading. The EIS measurements were obtained under a constant potential of 0.6 V with irradiation, and the amplitude of the applied sine wave potential in each case was 5 mV. The transient photocurrent responses as the light switched on or off were measured at a constant potential of 0.6 V. The measurements were completed at room temperature. The potential in each case was 5 mV. The transient photocurrent responses as the light switched on or off were measured at a constant potential of 0.6 V. The measurements were completed at room temperature.

Activity evaluations
The PWO activity of samples was tested with a PLS-SXE300 Labsolar-6A system (Perfectlight, China). A 30 mg portion of the sample was poured into 50 mL of 0.15 M Fe(NO 3 ) 3 solution in a Pyrex reaction vessel equipped with a quartz lid cooled by recirculating water. Soon aer, the reaction setup was degassed to remove air, and the reaction solution was magnetically stirred and then irradiated under visible light. The O 2 yield was recorded using a gas chromatograph (Panna G2090A), which was equipped with a thermal conductivity detector (column temperature at 60°C, inlet temperature at 150°C, test temperature at 125°C). All peaks were examined against corresponding standards, and each test was conducted three times in parallel to obtain an average value.

AQE measurements
The AQE was qualitatively analyzed under the same conditions as the activity evaluations, except for the insertion of band-pass lters (420, 450, 475, 500, 520 nm). The light intensity was measured at seven different points to obtain an average value using a PL-MW2000 optical power meter. The AQE was calculated using the following formula: 31

AQE ¼
Number of reacted holes Number of incident photons where a = 4 for the O 2 yield, M is the amount of O 2 molecule, N A is Avogadro's constant, ħ is Planck's constant, c is the speed of light, S is the irradiation area, P is the irradiation intensity, t is the irradiation time, and l is the wavelength of the light.

Conflicts of interest
There are no conicts to declare.