Exceptional Photocatalytic Hydrogen Peroxide Production from Sandwich‐Structured Graphene Interlayered Phenolic Resins Nanosheets with Mesoporous Channels

Harnessing solar energy to produce hydrogen peroxide (H2O2) from water (H2O) and dioxygen (O2) via artificial photosynthesis is an attractive route. To achieve high solar‐to‐H2O2 conversion efficiency, herein, an interfacial self‐assembly strategy is adopted to pattern mesoporous resorcinol‐formaldehyde resin (MRF) onto reduced graphene oxide (rGO) to form sandwich‐structured rGO@MRF polymeric photocatalysts. The internal graphene layer that mimics the electron transport chain of plant leaf, can effectively transfer electrons, and promote the two‐electron reduction of O2. Moreover, the mesoporous channels mimic the stomata, beneficially boost the fluid velocity, enrichment of O2, and diffusion of H2O2. Consequently, the developed metal‐free material can achieve an exceptional solar‐to‐chemical energy conversion efficiency of 1.23%. This ingenious interface engineering brings new opportunities for the design of efficient artificial photocatalysts.


Introduction
Artificial photosynthesis can convert earth-abundant resources into essential substances, allowing mankind to achieve sustainable development by utilizing solar energy. [1] As one emerging type of artificial photosynthesis, recent years have seen immense interests in photocatalysis to synthesize hydrogen peroxide (H 2 O 2 ) from H 2 O and O 2 . [2] H 2 O 2 have a wide range of applications as an oxidant [3] and emerging energy fuel. [4] In addition, as a sustainable strategy, the photocatalytic synthesis of H 2 O 2 eliminates the multi-step, highenergy input, and environmental impacts of the traditional anthraquinone (AQ) industry. [5] The solar-to-chemical conversion (SCC) efficiency for H 2 O 2 photosynthesis has currently reached ≈1.1%. [6] Several scientific challenges must be overcome to achieve higher solar-to-H 2 O 2 conversion efficiency, particularly for the rational design of photocatalyst, whose composition and structure determine the light absorption, charge separation and the selectivity of reaction products. [7,8] Owing to the high selectivity for two-electron O 2 reduction reactions (ORR) and the limited ability to decompose the formed H 2 O 2, organic semiconductors, e.g., graphitized carbon nitride (g-C 3 N 4 ), [9] covalent triazine frameworks (CTF), [10] covalent organic frameworks (COF) [6a,11] and covalent heptazine frameworks (CHF), [12] exhibit good photocatalysis activity for H 2 O 2 generation. Specifically, the organic resorcinol-formaldehyde (RF) resin that can be prepared from inexpensive and highly available monomers, may create an emerging photocatalyst for scalable H 2 O 2 production. [13] Recently, the pioneering RF resin sphere with donor-acceptor (D-A) molecular structure has been synthesized and it achieved a recorded SCC efficiency of up to 0.5%. [14] Nevertheless, some intrinsic properties of RF resin prevent the further improvement of its photocatalysis activity. For example, despite the D-A structure of RF resin is favorable for charge carrier separation, the poor conductivity of RF resin hinders separated electrons further transfer to the catalyst surface to interact with reactants. Moreover, low surface area of RF resin is not favored for mass transfer to the active catalytic sites in photocatalysis. To tackle these challenges, generation of strongly π-stacking structure [15] or incorporating conductive polythiophene into RF resin [16] have been exploited to improve the electron transfer process, and RF resin with mesoporous structure has also been synthesized to deal with the mass transport in photocatalysis. [6b,17] However, it is still desirable for simultaneously tuning electron and mass transfer of RF resin through a facile strategy.
As the important organ of plants, leaf convert solar energy to chemical energy through the photosynthesis process (H 2 O + CO 2 + solar energy → sugar + O 2 ). [18] To better utilize photogenerated electron for redox reactions, the electron is transferred though an electron transfer chain embedded in the chloroplast membrane. [19] Moreover, on the surface of leaf, the stomata allow for mass exchange to ensure mass balance and smooth photosynthesis by absorbing CO 2 and transporting H 2 O and O 2 . [20] These unique and efficient strategies for electron/mass management in natural leaf bring us commendable inspiration to design RF based photocatalyst for H 2 O 2 production. As a mimicry of leaf's structure and functions, we propose a 2D reduced graphene oxide@mesoporous RF (rGO@MRF) nanosheets with sandwich structure for H 2 O 2 production. The embedded conductive rGO layer between two RF layers functions as an electron transport chain, which can effectively capture conduction band (CB) electrons from RF resins to promote charge separation. Meanwhile, the mesopores of RF resins act as the mass transfer channels by tunning the O 2 adsorption and H 2 O 2 desorption in photocatalysis. As such, this "two birds with one stone" strategy has a potential to achieve high-efficiency H 2 O 2 production. However, there are no photocatalysts that combine graphene and mesoporous phenolic resins so far. In particular, the compound of two components (graphene and resin) into fine mesoporous sandwich-structured catalysts for photocatalytic H 2 O 2 production has not been explored.
Here, the sandwich-structured composite nanosheets of reduced graphene oxide and mesoporous resin (rGO@MRF) were fabricated via an interfacial self-assembly strategy. [21] Unlike pristine RF, the embedded rGO layer in rGO@MRF optimized the electron transfer as well as the oxygen reduction reaction selectivity, which are evidenced by electrochemical impedance spectroscopy (EIS), photocurrent test, photoluminescence (PL) spectroscopy and rotating ring-disk electrode (RRDE) method. Moreover, finite element simulation proved that the mesoporous pores of rGO@MRF promote the circulation and enrichment of O 2 and the timely diffusion of H 2 O 2 . Consequently, the sandwich-structured rGO@MRF nanosheets achieved a record SCC efficiency of 1.23% for photosynthetic H 2 O 2 production under AM1.5G simulated sunlight irradiation. The photocatalyst based on the electron/mass management functions of the photosynthetic leaf may serve as a model for the development of other high-efficiency photocatalytic systems.

Preparation of Photocatalysts
An interfacial self-assembly strategy was adopted to pattern 2D mesoporous resorcinol-formaldehyde resin on rGO nanosheets. [22] As illustrated in Figure 1a, this strategy employs micelles formed by commercial F127 copolymer as the mesoporous soft template, [23] 1, 3, 5-trimethyl benzene (TMB) as the micelle swelling agent, [24] resorcinol and formaldehyde as the polymerized monomers, [6b,25] and the surface of graphene oxide (GO) as the self-assembly substrate, [26] respectively. First, a modified Hummers method was used to prepare GO nanosheets. [27] The smooth and non-mesoporous surface of GO was confirmed by the transmission electron microscope (TEM) ( Figure S1, Supporting Information). In the mixed aqueous solution of F127, TMB and GO, micelles consisting of poly(propylene oxide) (PPO) as the core and poly(ethylene oxide) (PEO) as the corona are tightly packed on GO surfaces through hydrogen bonding interactions. [ 28] Both micelles and GO appear homogeneous in the reaction solution based on direct observations ( Figure S2, Supporting Information). Subsequently, with the absorption of the resin monomers in the PEO domains of F127, ammonia-catalyzed polymerization of resorcinol and formaldehyde was confined between close-packing composite micelles, yielding MRF polymeric network on GO. [29] The final hydrothermal treatment at 250 °C caused GO to transform into rGO, [30] as proved by the Raman characterization where the D (1345 cm −1 ) and G bands (1590 cm −1 ) of rGO appear ( Figure S3, Supporting Information). [22a] Interestingly, a peak centered at 1565 cm −1 appeared, which may be attributed to the ring CC stretching mode of polymeric resin. After hydrothermal treatment, Fourier-transform infrared (FTIR) spectra prove the complete removal of the F127 copolymer, in which the characteristic signals ≈2800 cm −1 disappear ( Figure S4a, Supporting Information). Simultaneously, a large amount of resorcinol in quinoid forms were produced in MRF as evidenced by the functional moieties. A small peak at 1650 cm −1 , corresponding to CO groups (character of quinone), appears in GO@MRF treated at 250 °C ( Figure S4b, Supporting Information). [14] Through the interfacial self-assembly strategy, a serial of rGO@MRF-x [x (wt.%) = GO/ resorcinol × 100; x = 0.2, 0.5, 1.0, 5.0.] composites were successfully fabricated by varying the GO content. Of note, the non-mesoporous rGO@RF ( Figure S5, Supporting Information) was also synthesized without adding F127 and TMB, which proves that the F127/TMB composite micelles are soft templates for formation of mesopores. [6b] As envisioned in Figure 1b, the built-in rGO layer is expected to improve charge separation. Different from the usual sandwich-structured photo catalysts, resorcinol formaldehyde resin mesoporous channels could both enhance the mass transfer and modulate the donor and acceptor distribution, resulting in improved photo synthesis of H 2 O 2 performance.

Characterizations of Photocatalysts
Typically, the morphology characterizations of the as-obtained rGO@MRF-0.5 are presented in Figure 2. The scanning electron microscope (SEM) (Figure 2a and Figure S6a) and TEM images ( Figure S6b, Supporting Information) clearly exhibit a uniform nanosheet-like morphology of rGO@MRF-0.5 with close-packing mesopores. Among them, some SEM images of rGO@MRF-0.5 and rGO@MRF-5.0 edges exhibit a typical twolayer MRF network (Figure 2b; Figure S6c, Supporting Information). Notably, TEM image of some edges on rGO@MRF-0.5 nanosheets expose the internal rGO ( Figure 2c). These details confirm that the rGO has been sandwiched between two layers of MRF. The mesoporous nanosheet have a uniform distribution of C and O elements, which is observed by high-angle annular dark-field scanning transmission electron microscopy (HAADF-STEM) image ( Figure 2d) and corresponding element mapping images (Figure 2e,f). The layered rGO@MRF-0.5 composite is 30-40 nm thick, as measured by atomic force microscopy (AFM) ( Figure S6d,e, Supporting Information). Further, the structural porosity was investigated by nitrogen sorption analysis. The specific surface area (S BET ) of the rGO@MRF-0.5 is calculated to be 50 m 2 g −1 by the Brunauer-Emmett-Teller (BET) method ( Figure S7a, Supporting Information). Probably due to the random stacking of nanosheets, there is no accurate mesopore size but lots of mesoporous distribution ( Figure S7b, Supporting Information). [31] Certainly, mesoporous resorcinolformaldehyde resin spheres (MRFS) were synthesized in the absence of GO (x = 0) to provide the polymerization interface ( Figure S8, Supporting Information). Interestingly, the rGO@MRF-x (x = 0.2, 1.0, 5.0) tended to form 2D nanosheets with increasing the content of GO, as illustrated in Figures S9-S11 (Supporting Information) in detail.
The chemical and physical properties of the as-obtained rGO@MRF-x were investigated. Compared with the non-hydrothermal resin ( Figure S4b, Supporting Information), the FTIR spectra of rGO@MRF-x ( Figure S12, Supporting Information) shows several similar features. A small peak appears at 1650 cm −1 , corresponding to the typical CO group of quinone, indicating the forming of quinone in the composite. [14,15] Besides, the smooth stretching vibration peak of the CH bonds in the benzene ring (1370 and 1300 cm −1 ) and the relatively sharp stretching vibration peak of methylene linker (1450 cm −1 ) display the tight cross-linking between aromatic rings.
[6b] The high-resolution C 1s and O 1s X-ray photoelectron spectroscopy (XPS) spectra of the MRFS and rGO@ MRF-0.5 reveal two components assigned to CO and CO groups ( Figure S13, Supporting Information), which further support the emergence of quinoids. Moreover, the structure of the polymer is verified by powder XRD ( Figure S14, Supporting Information). The (002) plane of graphitic carbon appears at ≈26°, however, all of the mesoporous resin materials investigated exhibited broad diffraction at 2θ = ≈20° (d = ≈4.4 Å), indicative of the amorphous carbonaceous structure. [6b, 14,15] The peak position of the hydrothermal resin migrates to a high angle with the increase of the rGO content, which also proves narrowed distance between the aromatic planes (π-stacking) and the composite structure of the resin and rGO. In short, these characterization results proved that the hydrothermally produced RF resins contains quinoid forms of resorcinol as electron acceptors (A), which are π-conjugated with inherent benzenoid forms of resorcinol as electron donors (D) to constitute D-A couples. [14] In addition, the contact angles between these samples were similar, demonstrating nearly the same affinity for water ( Figure S15, Supporting Information).
The band structure of the prepared catalysts was investigated to understand their thermodynamic properties for relevant photocatalytic redox reactions. The π-conjugated and π-stacked benzenoid-quinoid D-A couples form the basis for RF resins as low-bandgap polymeric semiconductor photocatalysts. [14,15] As shown in Figure 3a, the UV-vis absorption spectra of pristine MRFS exhibits the wide range of visible light absorption. Interestingly, the light absorption below 550 nm decreases with increasing rGO loading. This is may be due to the intercalation and electron enrichment of graphene that decrease the π-stacking between the benzene rings and weaken the photo excited electron leap of D-A pairs, especially the shrinkage or even disappearance of the absorption peak ≈500 nm as rGO increases. [14,32] However, there is an enhanced absorbance in the wavelength range above 550 nm with increasing rGO content. Therefore, the introduction of graphene can adjust the light absorption to be effective for the visible-light response of the composite ( Figure S16, Supporting Information). [33] According to the Tauc plot ( Figure S17a, Supporting Information), its bandgap energy is calculated to be 1.90 eV. Besides, the Mott-Schottky plots at different frequencies for MRFS ( Figure S17b, Supporting Information) exhibit a positive slope, indicating it is a typical n-type semiconductor with a flat band potential of −0.07 V. Figure 3b illustrates the band potentials of the MRF. The conductance band (CB) lower limit (−0.17 V versus reversible hydrogen electrode, RHE) of MRF lies above the potential for the donor level of rGO (−0.08 V vs RHE) and two-electron reduction of O 2 (+0.68 V, RHE). [34] These results indicate that the transfer of photogenerated electrons from MRF to rGO, and the subsequent reduction of oxygen by photoinduced electrons, are thermodynamically favorable. Also, the valence band (VB) upper limit (1.73 V) of MRF lies below the oxidation potential of water (+1.23 V, RHE), indicating the photoinduced hole of MRF is capable for water oxidation. In particular, the possibility of overall water splitting on the material was ruled out, since no H 2 was detected during the actual photocatalytic reaction. In contrast, CB e − is more likely to drive two-electron ORR to produce H 2 O 2 in the presence of O 2 .

Photocatalytic Activity
The photocatalytic reactions were first performed in O 2 saturated water (100 mL) containing catalysts (25 mg) at room temperature (25 °C) under visible light irradiation (λ ≥ 420 nm, Xe lamp). Then, the results of parallel experiments were counted to demonstrate the catalytic activity of each sample. Figure S18a   Figure 4a showcase the rate of photocatalytic H 2 O 2 production on each photocatalyst. Among them, rGO@ MRF-0.5 shows the highest activity of 861 µmol h −1 g −1 , which is 1.6 times that of MRFS without rGO. However, the photocatalytic activity of rGO@MRF-x decreased when a large amount of rGO (x > 0.5) is sandwiched. This may be because the strong absorption of light at λ ≥ 420 nm suppresses the photoexcitation of the bottom catalysts in the aqueous solution during photocatalytic tests. [35] Besides, we noticed that the samples with mesoporous structure and the same rGO content produced H 2 O 2 at higher rates than the samples without mesoporous structure (rGO@RF-0.5 and rGO@RF-5.0). Also, the photocatalytic activity of the physical mixture of MRF and rGO (0.5 wt.%) has no enhancement but attenuation compared with MRFS alone. Because dispersed rGO does not have a close contact with MRF, electrons cannot be transferred and the formed H 2 O 2 might be even decomposed by exposed rGO ( Figure S18b, Supporting Information). Moreover, consistent H 2 O 2 production rate was observed even during the four-cycle reaction (2 h per cycle) with rGO@MRF-0.5, proving its reliable stability for photocatalysis (Figure 4b). To evaluate the SCC efficiency, the catalysis performance of the rGO@MRF-0.5 for H 2 O 2 production was then tested under simulated AM1.5G sunlight (1 sun) irradiation ( Figure S19a, Supporting Information). During the initial stages (<1 h), rate of H 2 O 2 production and the corresponding SCC efficiency gradually decreased (Figure 4c). This is largely due to part of VB h + produced by light-excitation participating the oxidation of the polymer itself. [14] The resin structure can be also oxidized by the H 2 O 2 formed in initial photocatalytic process. Those reductive structures (e.g., certain reducing groups) can be oxidized, while the stable structures are retained. With the reaction proceeding, the organic phenolic resins tend to stabilize and thus SCC efficiency for The reaction conditions were as follows: water (100 mL), O 2 , catalyst (25 mg), λ ≥ 420 nm (Xe lamp), RT. c) The amounts of H 2 O 2 evolution on rGO@MRF-0.5 and the corresponding SCC efficiency under AM1.5G simulated sunlight (1 sun) irradiation. The reaction conditions were as follows: water (150 mL), O 2 , catalyst (400 mg), AM1.5G (1 sun), 50 °C; the irradiated area is 2.5 × 2.5 cm 2 . d) Summarized SCC efficiencies from reported photocatalysts toward H 2 O 2 production.
the H 2 O 2 generation on rGO@MRF-0.5 gradually stabilized after the previous decay. Finally, the efficiency value was maintained at 1.23% during 2 h. The 1.23% conversion efficiency is higher than the photosynthetic efficiency of rapidly growing trees (≈1%), such as poplars. [36] Notably, the rGO@MRF can achieve the highest SCC efficiency at lower catalyst concentrations than the reported powder photocatalyst as far as we know (Figure 4d and Table S2, Supporting Information). Simultaneously, rGO@MRF-0.5 exhibits a strong correlation between its apparent quantum efficiency (AQE) and absorption spectrum under monochromatic light irradiation ( Figure S19b, Supporting Information), demonstrating that H 2 O 2 is generated from bandgap excitation. As the other half reaction, the photocatalytic water oxidation performance of rGO@MRF-0.5 with AgNO 3 as a sacrificial electron acceptor was tested. The results confirm the water oxidation driven by photo-generated h + (Figure S20, Supporting Information). Moreover, O element was traced in these reactions using the isotopic photoreaction experiment. After MnO 2 was added to the resultant solutions, the evolved gases were analyzed by gas chromatography-mass spectrometry (GC-MS). As shown in Figure S21 (Supporting Information), the results obtained after the 24 h reaction show a strong 18  We also investigated the decomposition activity of rGO@MRF-0.5 towards H 2 O 2 . It is found that rGO@MRF-0.5 did not show any decomposition activity for H 2 O 2 ( Figure S22, Supporting Information, red and green line), which is beneficial for the accumulation of H 2 O 2 in photocatalysis.

Promoted Two-Electron Oxygen Reduction Reaction
It is confirmed that H 2 O 2 is generated from O 2 via twoelectron reduction pathway since H 2 O 2 generation is almost completely inhibited in N 2 atmosphere ( Figure S22, Supporting Information, blueline). Unlike MRFS, rGO@MRF-0.5 shows significantly enhanced activity for H 2 O 2 generation, therefore it is reasonable to conclude that rGO plays critical roles for the two-electron oxygen reduction reaction. In order to deeply understand the function of rGO interlayer in improving the photocatalytic activity, the electrochemical properties and charge carrier separation characteristics of rGO@MRF-0.5 and MRFS were investigated. First, electrochemical impedance spectroscopy (EIS) measured under visible light irradiation (Figure 5a) shows that the charge transfer resistance across the electrode/electrolyte interface (R CT ) of the rGO@MRF-0.5 and rGO@MRF-5.0 is 16 and 5 kΩ, which are much smaller than that of MRFS without rGO (23 kΩ). Notably, the work function of rGO is −4.42 eV versus vacuum (−0.08 V vs RHE), [34] which is much more positive than the CB position of MRF (−0.17 V vs RHE). Therefore, the CB e − generated by light-excited MRF spontaneously tend to transfer to rGO, which may facilitate efficient charge separation. [35] To further understand photoinduced electron−hole separation behavior of rGO@MRF-0.5, the material's photocurrent response was measured by loading catalyst on a fluorine tin oxide (FTO) electrode under visible light irradiation. Compared to the MRFS, the rGO@MRF-0.5 demonstrated a significantly enhanced photocurrent (Figure 5b). The high photocurrent from rGO@MRF-0.5 suggests efficient photogenerated charge separation and migration mediated by graphene. Generally, photoluminescence (PL) emission spectroscopy reflects the irradiative recombination of electrons and holes and is used to elucidate the separation and migration of charge carriers. Compared to MRFS, the PL emission intensities for rGO@MRF-0.5 is much lower, indicating enhanced charge separation and transfer behavior of rGO@MRF-0.5 (Figure 5c). Notably, the PL emission intensity of rGO@MRF-5.0 is higher than that of rGO@MRF-0.5, while the excitation wavelength tends to be shorter. Large amount of rGO may suppresses the photoexcitation process of the catalyst due to strong absorption of long wavelengths (λ ≥ 420 nm), which is consistent with the absorbance of rGO@MRF-x (Figure 3a). In more detail, the transient fluorescence spectroscopy (TFS) was used to investigate the photo-excited carrier dynamics in MRFS, rGO@MRF-0.5 and rGO@MRF-5.0. As shown in Figure 5d, the average fluorescence lifetime of the photogenerated charges for rGO@MRF-0.5 (3826.5 ps) is longer than that of MRFS (3642.2 ps) or rGO@ MRF-5.0 (3488.2 ps), indicating that reasonable intercalation of rGO in phenolic resin is conducive to efficient electron-hole separation and prolonged lifetime of charge carriers.
As well as graphene-mediated charge separation, the subsequent ORR was also investigated, especially the conversion process of oxygen species on the sandwiched structure. First, electron paramagnetic spectroscopy (EPR) was carried out to identify the active oxygen species during ORR route by using 5,5-dimethyl-1-pyrroline N-oxide (DMPO) as the spintrap agent. rGO@MRF-0.5 exhibits a typical six characteristic signals for DMPO-·O 2 − under light irradiation, which were absent in the dark ( Figure S23, Supporting Information), illustrating that H 2 O 2 is possibly produced by a sequential twostep ORR. [5] Moreover, the average electron transfer number (n) as a key index to estimate the selectivity of ORR process was evaluated by electrochemical rotating ring-disk electrode (RRDE) method. To detect the generated H 2 O 2 , the potential on the platinum (Pt) ring was fixed at 1.2 V (vs RHE). The linear-sweep voltammograms (LSV) curves of the catalysts were measured on RRDE under O 2 -saturated neutral aqueous solution at 1600 rpm. After calculation, the rGO@MRF-0.5 showed a higher current (Figure 5e) and higher selectivity for two-electron pathway to H 2 O 2 ( Figure S24, Supporting Information). The n value for rGO@MRF-0.5 is 2.48, which is much lower than 3.05 of MRFS (Figure 5f). The n of rGO@MRF-0.5 closer to 2 proves that the rGO interlayer is conducive to the two-electron reduction of O 2 . Besides, Density functional theory (DFT) calculations were used to evaluate the adsorption energy for O 2 over rGO and D-A couples in RF, which is beneficial to predict the reduction position of O 2 in the composites. After structural optimization, molecular structures of rGO ( Figure S25 and −0.10 eV (E ads, RF ) can be obtained, respectively. Therefore, we infer that O 2 is inclined to adsorb onto rGO and further reduce to H 2 O 2 . Such result is in good line with previous published studies, which have revealed that carbonaceous materials including rGO, [37] porous carbon, [38] carbon nanotubes (CNT) [39] and carbon quantum dots (CDs) [40] have the property of selective ORR via two-electron pathway. Thus, in this work, the graphene intercalations promote selective two-electron oxygen reduction reaction rather than simply enhancing charge separation.

Simulation of Mass Transfer in Mesopores
Generally, the porous structure of the material affects the local mass transfer to alter the progress of the catalytic reaction. [41] In this case, finite element analysis (FEA) simulations were adopted to construct the mesoporous structure of the rGO@ MRF-0.5 (see Supporting Information for detailed model), and visualize the transport and distribution of reactant (O 2 ) and product (H 2 O 2 ). [42] To ensure clear visualization, we extract the one-half cross section of a single mesopore equivalently without altering the results. Driven by the pressure gradient, water flows through the mesoporous channel ( Figure S27  Supporting Information). Considering the law of viscosity and the narrowing of the flow channel, [43] the flow velocity is accelerated from the initial 0.235 to 1.88 m s −1 at the center of the mesoporous pore (Figure 6a,b). In this case, an 8-fold increase in fluid velocity at the center of the mesopore may rapidly deliver O 2 and push the generated H 2 O 2 out in time.
In light of convection, the diffusion and fluid flow, concentration distribution trends of O 2 and H 2 O 2 are calculated. Driven by the accelerated flow, O 2 concentration drops abruptly from 0.669 mol m −3 at the mesopore inlet to 0.627 mol m −3 near the graphene layer, and then down to 0.615 mol m −3 at the mesopore outlet ( Figure S28, Supporting Information). Contrary to O 2 , H 2 O 2 rapidly increases from virtually no distribution at the mesopore inlet to 0.081 mol m −3 near the graphene layer, and finally increase to 0.107 mol m −3 at the mesopore outlet ( Figure S29, Supporting Information). Further, a 2D 5 × 5 mesopore array with the mesopore size of 7 nm was modelled to identify the mass transport in multiple mesopores clearly and intuitively. [44] Consistent with the results of single mesoporous channel (Figure 6a,b), the flow velocity shows a trend of accelerating from outside to inside mesopores (Figure 6c) duo to the viscous effect of the liquid. The higher flow velocity can entrain more O 2 per unit, resulting in a much higher concentration of O 2 in the centre of the mesopore (0.632 mol m −3 ) than near the wall of the mesopore (0.612 mol m −3 ) (Figure 6d). In contrast, the concentration distribution of H 2 O 2 near the walls of the mesopores (0.113 mol m −3 ) is much higher than that at the centre of the mesopore (0.073 mol m −3 ) (Figure 6e). This phenomenon indicates that H 2 O 2 as the reaction product may form a certain concentration gradient inside the mesopores, which is favorable for its diffusion out with the accelerated water flow.
The above results clearly demonstrate that the abundant mesoporous distribution on the catalyst surface provides more channels for the fluid acceleration, the enrichment of O 2 and the diffusion of H 2 O 2 . Mesoporous rGO@MRF-0.5 also decomposes H 2 O 2 slower than non-mesoporous rGO@RF-0.5 in the inert gas atmosphere, which may be due to the rapid circulation of H 2 O 2 in the mesoporous channels, which shortens the residence time of H 2 O 2 within the photocatalyst ( Figure S30, Supporting Information). Particularly, the diffusion of H 2 O 2 into the bulk solution is crucial to preventing its further reduction or decomposition on the catalyst surface. [45] The mesoporous channels do offer another pathway for H 2 O 2 diffusion. Experimentally the H 2 O 2 production efficiencies of rGO@MRF-0.5 and rGO@MRF-5.0 increase by 29.4% and 12.9% respectively compared to the non-mesoporous counterparts (rGO@RF-0.5 and rGO@RF-5.0, Figure 4a). Such results could come from of the enhanced mass diffusion caused by the increased flow velocity inside the mesopores. Therefore, similar to plants'  stomata, the mesopores are favorable for promoting the mass transport in rGO@MRF photocatalysis.
In summary, both charge separation and mass transfer behavior of the metal-free photocatalysts are improved by employment of sandwich-structure with the built-in rGO layer and RF resin mesoporous channels. Furthermore, through three step process, 2D rGO@MRF photocatalyst can first enrich O 2 via the mesopores, convert O 2 into H 2 O 2 over the graphene layer with enhanced charge separation ability, and then promptly desorb the produced H 2 O 2 from the catalyst layer. The enhancement of mass transfer would extend the functionality of conventional sandwich photocatalysts. Practically, it brings new thoughts to the sandwich-structured photocatalyst -apart from charge separation, it is useful for taking account of the adsorption/desorption of substrates/reaction intermediates and its selectivity of products.

Conclusion
Inspired by the electron transport chains and stomata in leaves, we intercalate rGO into mesoporous phenolic resins by an interfacial self-assembly strategy to rationally regulate the photocatalytic activity for the H 2 O 2 production. The builtin rGO assists transfer light-excited e − from CB of MRF to O 2 , promoting the separation of photo-generated charges and highly selective two-electron reduction reaction. Meantime, mesoporous channels improve fluid velocity, enrichment of O 2 and diffusion of H 2 O 2 . Under simulated AM1.5G sunlight irradiation, the rGO@MRF-0.5 demonstrates a high SCC efficiency of 1.23% for photosynthesis of H 2 O 2 . This interfacial composite concept of metal-free photocatalysis may provide guidance for the development of other efficient biomimetic photocatalytic systems for solar energy conversion.

Experimental Section
Catalyst Preparation: GO was synthesized using modified Hummer's method. [27] In a typical synthesis of rGO@MRF-x (x = 0.2, 0.5, 1.0, 5.0), 4x mg of GO was dispersed in a mixture of 15 mL of water and 15 mL of ethanol. Then 0.13 g of F127, 0.4 mL of TMB and 1.0 mL of ammonia were added to the reaction mixture under stirring at room temperature. When the mixture became milky white, 400 mg of resorcinol and 0.56 mL of formaldehyde were added to the reaction solution. After stirring for 12 h, the mixture was heated in a sealed Teflon-lined autoclave at 250 °C for 24 h. The rGO@MRF-x was collected by centrifugation and the surfactant was extracted thoroughly by refluxing the as-synthesized materials in 100 mL of methanol containing 1 mL of a concentrated HCl aqueous solution for 12 h at 60 °C (twice). When x = 0, the mesoporous resorcinol-formaldehyde resin spheres (MRFS) were synthesized accordingly.
Photocatalytic Measurement: For photocatalytic production of H 2 O 2 , 25 mg of photocatalysts were well dispersed in an aqueous solution (100 mL). A 300 W Xe lamp (PLS-SXE300, PerfectLight) was used as light source (λ ≥ 420 nm) to trigger the photocatalytic H 2 O 2 generation under continuous flow of O 2 . The temperature of the reaction solution was kept at 25 °C via a circulating water bath during the whole experiment. The amount of produced H 2 O 2 was determined by spectroscopic titration with titanium sulfate solution. H 2 O 2 and titanium sulfate form a yellow precipitate of a titanium complex, which was dissolved in sulfuric acid and then subjected to colorimetric measurement. The reagent addition of the standard curve was shown in Table S1 (Supporting Information). The corresponding standard curves was shown in the Figure S31 (Supporting Information). From the reaction solution, 1 mL of solution was filtered out using filter plugs (apertures ≈ 200 nm). 0.2 mL of concentrated ammonia, 0.1 mL of 5% titanium sulfate, and the filtered reaction solution were mixed and allowed to stand for 2 h. Then, the mixed solution was added with 5 mL of 2 M H 2 SO 4 to dissolve the precipitate of peroxide-titanium complex. Finally, the absorbance at λ = 410 nm was measured. From the equation of the calibration curve ( Figure S31b, Supporting Information), the amount of H 2 O 2 produced in the reaction solution can be determined by following equation: (1) For photocatalytic water oxidation, 10 mg of photocatalysts were well dispersed in an aqueous solution (100 mL) containing AgNO 3 (1 mm) as an electron acceptor. The reactor was evacuated for 30 min before reaction to remove the residual air. A 300 W Xe lamp (PLS-SXE300, PerfectLight) was used as light source (λ ≥ 420 nm) to trigger the photocatalytic O 2 generation. Evolved O 2 was measured through a thermal conductivity detector (TCD) gas chromatograph (5 Å molecular sieve column, Argon as the carrier gas). Blank experiments revealed no appreciable gas evolution without irradiation or photocatalysts.
Analysis of AQE: For the determination of apparent quantum efficiency (AQE), a Xe lamp with different bandpass filters (λ ± 10 nm) was used to illuminate 400 mg of rGO@MRF-0. 5 The average intensity of irradiation was determined to be 10 mW cm −2 by a spectroradiometer, and the irradiation area was controlled at 2.5 × 2.5 cm 2 . Therefore, the number of incident photons (N) was 1.88 × 10 −6 λ mol.
Analysis of SCC Efficiency: For the determination of SCC efficiency, the AM1.5 global spectrum was used with 400 mg of rGO@MRF-0.5 and 150 mL of water under O 2 bubbling at 50 °C. The SCC efficiency was determined using the equation: The free energy for H 2 O 2 formation is 117 × 10 3 J mol −1 , the irradiance of the AM1.5 global spectrum (300-2500 nm) was 0.1 W cm −2 and the irradiated area was controlled at 2.5 × 2.5 cm 2 . The total input power was therefore 0.625 W.
Isotopic Photoreaction Experiment: 10 mg of catalyst (rGO@MRF-0.5) was dispersed into 10 mL of H 2 16 O. The gas in the system was evacuated and then 18 O 2 was added. The reaction was carried out under light irradiation (λ ≥ 420 nm). After 24 h, the system was evacuated with Ar and then MnO 2 was added to decompose the formed H 2 O 2 . And finally, the O 2 generated was detected by GC-MS.
Electrochemical Analysis: EIS measurements were carried out in the three-electrode cell by applying 10 mV alternative signal versus the reference electrode over the frequency range of 0.1 Hz to 100 kHz. The EIS Nyquist plots were fitted and quantified with R S , R CT and C DL . The R S of MRFS, rGO@MRF-0.5 and rGO@MRF-5.0 were 11.18, 9.25, and 53.26 Ω, respectively. The R CT of MRFS, rGO@MRF-0.5 and www.afm-journal.de www.advancedsciencenews.com rGO@MRF-5.0 were 22.62, 15.9, and 5.29 kΩ, respectively. The C DL of MRFS, rGO@MRF-0.5 and rGO@MRF-5.0 were 1.92 × 10 −6 , 1.99 × 10 −6 , and 9.79 × 10 −5 F, respectively. Photoelectrochemical characterizations were conducted on a conventional three-electrode system connected to an electrochemical analyzer. The working electrode was prepared with a fluorine-doped tin oxide (FTO) glass (1×1 cm 2 ). The coating area of the catalyst on the FTO glass is 0.8 cm 2 . The measurements were performed on a catalyst coated FTO glass in 0.1 m Na 2 SO 4 solution at a bias of 0.5 V (vs Ag/AgCl). For the Mott-Schottky measurements, similar strategy was performed on FTO glass by the same doctor blade method. Mott-Schottky measurements were performed at a potential range from 0.6 to −0.2 V versus RHE, with a voltage amplitude of 5 mV and in a frequency of 1000 Hz. Each increase of potential was 0.05 V. The quiet time for each test was 2 s. For RRDE test, a three-electrode system includes an Ag/AgCl electrode was used as the reference electrode, a graphite rod as a counter electrode, and a rotating ring-disk electrode as the working electrode. The RRDE assembly consists of a glassy carbon rotation disk electrode (disk area: 0.247 cm 2 ) and a Pt ring (ring area: 0.186 cm 2 ), with a theoretical collection efficiency of 37%. To prepare the catalyst ink, the catalyst is dispersed in ethanol and 5 wt.% Nafion solution. The mass loading of all catalysts was 0.5 mg cm −2 .
The electrochemical measurements were performed in O 2 -saturated 0.1 m Na 2 SO 4 solution at room temperature. The rotating speed of the working electrode was 1600 rpm throughout the tests. First, a stable cyclic voltammetry (CV) was obtained at a scan rate of 10 mV S −1 , and then linear scan voltammetry (LSV) was performed at a scan rate of 10 mV S −1 in N 2 -saturated and O 2 -saturated electrolyte, respectively. All potentials were converted into those versus the reversible hydrogen electrode (RHE) by using the equation: The selectivity of H 2 O 2 and electron transfer number (n) were calculated based on the equation: The electron transfer number (n) was determined by the ratio of current density for the ring/disk electrode by using the equation: where I r is the ring electrode current, I d is the disk electrode current, and N is the ring electrode collection efficiency (0.37). In order to completely oxidize H 2 O 2 , the ring electrode was set at a constant potential of 1.20 V versus RHE. Computational Method: The Vienna Ab Initio Package (VASP) [46] has been employed to perform all the density functional theory (DFT) calculations within the generalized gradient approximation (GGA) using the PBE [47] formulation. The projected augmented wave (PAW) potentials have been chosen to describe the ionic cores and take valence electrons into account using a plane wave basis set with a kinetic energy cutoff of 400 eV. [48] Partial occupancies of the Kohn−Sham orbitals were allowed using the Gaussian smearing method and a width of 0.05 eV. The electronic energy was considered self-consistent when the energy change was smaller than 10 −5 eV. A geometry optimization was considered convergent when the force change was smaller than 0.02 eV Å −1 . Grimme's DFT-D3 methodology was used to describe the dispersion interactions. [49] The equilibrium lattice constant of hexagonal graphene unit cell separated by a vacuum layer in the depth of 15 Å was optimized, when using a 15 × 15 × 1 Monkhorst-Pack k-point grid for Brillouin zone sampling, to be a = 2.468 Å. It was then used it to construct a rGO monolayer surface model with p(6 × 6) periodicity in the x and y directions and a vacuum layer in the z direction with the depth of 15 Å in order to separate the surface slab from its periodic duplicates. During structural optimizations, the gamma point in the Brillouin zone was used for k-point sampling, and all atoms were allowed to relax. The tetramer was used to simulate the polymer. The adsorption energy (E ads ) of adsorbate A was defined as: where E A/surf , E surf and E A(g) are the energy of adsorbate A adsorbed on the surface, the energy of clean surface, and the energy of isolated A molecule in a cubic periodic box with a side length of 20 Å and a 1 × 1 × 1 Monkhorst-Pack k-point grid for Brillouin zone sampling, respectively. Finite Element Simulation: A finite element method simulation was performed using the COMSOL Multiphysics software to study the interaction between flow and mass transport in the composite mesoporous material region, considering the convective and diffusive effects of O 2 and H 2 O 2 . To simplify the computational simulation, a single mesoporous channel was created (mesopore size: 7 nm) in 3D space based on the geometric architecture of rGO@MRF-0.5. Of note, the model was assumed to be at a fixed position and deformation was ignored. [50] In the middle of the single mesoporous channel, a graphene layer (≈1 nm thick) was inserted. To simulate the inflow and outflow of substances in the non-porous structure, two vacant spaces are set at a height of 10 nm above and below the mesoporous channel, respectively. The simulation model adopts axisymmetric geometric modelling to calculate the fluid velocity and pressure in the mesoporous material region, as well as the spatial concentration distribution of O 2 and H 2 O 2 on both sides of the graphene layer. In this model, the top and bottom are set as the water inlet and outlet, respectively. And the mass transfer of O 2 and H 2 O 2 was mainly driven by two factorsfluid flow and concentration diffusion. The governing equations are as follows: where the dependent variables are velocity (u), pressure (p) and concentration (c i ). ρ is the fluid density, µ is the fluid viscosity, D i is the diffusion coefficient of substance i, and R i is the reaction rate. For the mass transport simulation, the diffusion coefficients of O 2 and H 2 O 2 in water are both set to 10 −9 m 2 s −1 . For the 3D model of a single mesopore, water flows in from the top, passes through the composite mesoporous material region, and flows out from the bottom. The inlet velocity is 0.235 m s −1 to model the flow induced by the rotation of the solution, [41] and the boundary condition at the outlet was zero pressure. The boundary inlet of the model was set as a solution saturated with O 2 . Based on experimental findings, the middle graphene layer was considered as the main O 2 consumption and H 2 O 2 generation site, and the molar ratio of O 2 consumption and H 2 O 2 generation was set to 1/2 (H 2 O + 1/2O 2 = H 2 O 2 ). In detail, the concentration of O 2 at the top inflow boundary is set to 21.4 mg L −1 , that was, 0.66875 mol m −3 , and the concentration of H 2 O 2 was set to 0. The bottom was the outflow boundary condition that the diffusion flux was 0. For the reaction rate in the region where graphene was located, the O 2 consumption rate was set to −50 mol mL −1 s −1 , and the H 2 O 2 production rate was set to 100 mol mL −1 s −1 . During the simulation, the effects of geometric dimensions such as mesopore diameter and spacing on O 2 and H 2 O 2 concentrations were shown. The geometric parameters required for modeling were derived from experimental data such as electron microscope images.