Direct growth of copper(I) oxide nanocubes on graphitic carbon nitrides enhancing aqueous carbon dioxide photoreduction


 Semiconductor hybrid structures containing multiple components have been considered an ideal photocatalyst design to generate long-lived charge-separated states. Particularly for the reactions requiring high activation energies, such as a CO2 reduction reaction (CO2RR), the reaction activity is highly susceptible to the catalyst component and morphology. In this study, we selected g-C3N4 and Cu2O as photocatalytic components having bandgaps suitable for CO2RR. Then, we tried to form good electric junctions between two domains by direct growth of Cu on g-C3N4 using a polyol process. The resulting g-C3N4/Cu2O hybrid was employed as photocatalysts in an aqueous medium without hole acceptors. The catalyst exhibited a noticeable activity (5.4 mmol gcat-1h-1) and quantum yield (3.7%) with a nearly quantitative selectivity for CH4 production, superior to any other photocatalysts for CO2RR. The strong coordination of g-C3N4 to the Cu2O surface could form a conductive junction and induce effective electron transfer enforcing the Z-scheme process for CO2RR in high activity and selectivity. This result ensured the importance of junctions and interfaces in the hybrid catalyst structure to exhibit excellent photocatalytic CO2RR performances.


Introduction
Photocatalytic carbon dioxide reduction reaction (CO 2 RR) is one of the most straightforward approaches to converting carbon dioxide into synthetic fuels using solar energy [1][2][3][4][5] . It pursues reducing greenhouse gases and concomitantly storing renewable energy. Since Inoue et al. reported a pioneering work, photocatalytic CO 2 RR with heterogeneous catalysts have made impressive progress using various semiconductor materials 5,6 . As well as the electrochemical potentials (E = -0.61~-0.08 V vs. normal hydrogen electrode (NHE)) competitive to the hydrogen evolution reaction (HER), there exist multiple factors to be seriously considered, particularly for CO 2 RR: low CO 2 adsorption on the catalyst surface, high overpotential required for the reaction, and low selectivity towards a single product. [6][7][8][9][10][11][12] Metal oxides are suitable for photocatalytic CO 2 RR due to their a nity for CO 2 coordination, generating stable intermediates. Among various oxides, copper(I) oxide, Cu 2 O, is a suitable catalytic material with ptype characteristics 13 . A narrow bandgap (2.2 eV) leads to facile visible light absorption, and the high conduction band edge (-0.28 eV vs. NHE) provides su cient overpotentials yielding reduced products.
However, the relatively high valence band edge (1.92 eV vs. NHE) does not t the oxidation potentials of counter-ions; therefore, heterojunction formation with n-type materials is a common strategy to promote both photoinduced electron and hole transfers to the reactants completing entire reaction circuits 14 . TiO 2 was combined with Cu 2 O and formed p-n junctions, resulting in a considerable enhancement of CO 2 RR [15][16][17] . We also generated a colloidal hybrid structure with p-type ZnO, achieving a quantum e ciency of 1.5% with a nearly quantitative conversion into CH 4 without sacri cial reagents 18 . Although these approaches have been successful on CO 2 RR, the catalysts were highly sensitive to the reaction conditions, and the activity and stability were still limited. It may be caused by the instability of metal oxide surface and junctions; for example, ZnO is vulnerable under a low pH environment.
Recently, graphitic carbon nitrides, g-C 3 N 4 , have attracted much attention as photocatalysts due to their band gaps near the visible range (2.7 eV) and high photochemical stability. Yet, g-C 3 N 4 alone has a low conduction band edge energy and induces fast charge-carrier recombination [19][20][21] . Hence, the hybridization with other semiconductors is a potential solution to overcome these problems. 22 Several researchers have investigated the combination of g-C 3 N 4 and Cu 2 O because their band energies match each other. However, the resulting materials were mainly used for dye decomposition but not for photocatalytic CO 2 RR [23][24][25][26][27] . In these reports, photoinduced excitons were proposed to transfer between distinct domains via the double charge transfer mechanism, making total band gaps narrower and providing potentials only to reduce organic dyes. For supplying su cient overpotentials leading to CO 2 RR, electron transfer between hetero-domains should occur via a Z-scheme pathway. Sun et al.
prepared Cu 2 O particles on g-C 3 N 4 foams and yielded CO in the gas phase reaction 28 . Zhao et al.
encapsulated Cu 2 O nanowire arrays with g-C 3 N 4 , produced ethanol in the gas phase, and proposed the Zscheme mechanism by light irradiation 29 . These catalysts exhibited insu cient catalytic activities yet compared to other hybrid materials, despite the ideal band matching of g-C 3 N 4 and Cu 2 O.
In this work, we notice that uniform junctions and interfaces are critical to promoting effective charge transfer and separation in the hybrid structure. For this purpose, we induced the direct growth of Cu 2 O cubes on g-C 3 N 4 sheets by a polyol process. The resulting g-C 3 N 4 /Cu 2 O hybrid was used as a photocatalyst for CO 2 RR in an aqueous solution without sacri cial reagents. Remarkably, the catalyst exhibited an enormous activity of 5.4 mmol g cat -1 h -1 and a quantum yield of 3.7%, with a CH 4 selectivity of 98%. The measurements revealed that good electric junctions were formed between the g-C 3 N 4 and Cu 2 O domains. The strong coupling through the interface may boost effective photo-induced electron transfer from g-C 3 N 4 to Cu 2 O via the Z-scheme process and provide a high potential enough to reduce CO 2 into CH 4 in a high e ciency.

Results
Hybrid formation and characterization. The g-C 3 N 4 /Cu 2 O hybrid was synthesized by the polyol process ( Fig. 1a). In detail, g-C 3 N 4 nanosheets were prepared by urea pyrolysis at 550 o C 21 . The transmission electron microscope (TEM) image shows crumbled g-C 3 N 4 sheets as nanosized fragments (Fig. 1b). The g-C 3 N 4 sheets were transferred to 1,5-pentanediol in the presence of poly(vinyl pyrrolidone). The copper oxides were directly grown on the g-C 3 N 4 surface by short heating after adding the copper precursor solution. The TEM image indicates that the nanocubes are formed with an average edge size of 25 ± 8 nm, lying on the crumpled sheets ( Fig. 1c and Supplementary Figs. S1 and S2). Notably, all cubes are attached to the sheets without free-standing particles under optimized synthetic conditions. The X-ray diffraction (XRD) pattern of g-C 3 N 4 /Cu 2 O is the sum of the (002) peak at 27° from g-C 3 N 4 (JCPDS No. 87-1526) and the (111) peak at 36° from primitive cubic Cu 2 O (JCPDS No. 77-0199) (Fig. 2a).
The UV-visible absorption spectrum of the hybrid (blue) is also the linear combination of g-C 3 N 4 and Cu 2 O; the high energy peaks are majorly from g-C 3 N 4 (red), and the tail up to 500 nm originates from Fig. 2b). X-ray photoelectron spectroscopy (XPS) provides essential information for bonding nature and oxidation states. Three species were analyzed by the deconvolution in the C 1s region (Fig.   2c). The peak at 284.8 eV (blue) is assignable to the adventitious carbon (C-C), and a shoulder (green) at 286.5 eV and a large peak (red) at 288.1 eV are ascribed as the carbons at C=N and N-C=N bonds in the aromatic CN heterocycles, respectively. In the N 1s region, the prominent peak (red) at 398.6 eV and a broad peak (green) at 400.0 eV correspond to the N atoms of C-N=C and tri-coordinate N-(C) 3  Photochemical CO 2 RR under aqueous medium. The g-C 3 N 4 /Cu 2 O hybrid was used as a photocatalyst for CO 2 RR under an aqueous medium without sacri cial reagents. After removing carbon residues to clean up the surface, the catalyst (10 mg) was dispersed in a CO 2 -saturated buffer solution (pH = 7.4). By irradiation using a 300 W Xe lamp, gas products were collected and analyzed by gas chromatography (GC). During the reaction up to 3 h, the amount of CH 4 linearly increased with the average activity of 53.6 μmol h -1 (5.4 mmol g cat -1 h -1 ) with an exclusive selectivity (98%) of CH 4 (Fig. 3a). Quantum e ciency (QE) was calculated by the number of electrons used for CH 4 evolution (eight electrons for producing each CH 4 molecule from CO 2 ) divided by the number of incident photons, estimated using the 200-540 nm absorption in the UV-visible spectrum 30 . In this experiment, the quantum e ciency was measured to be 3.7%. The production amounts of CO and H 2 were 4.8 and 1.4 μmol h -1 , respectively (Fig. 3b), with no other liquid products detected by nuclear magnetic resonance (NMR) spectroscopy. O 2 was evolved with the rate of 9.8 μmol h -1 by the counter oxidation reaction, far smaller than expected as a pair of CO 2 RR.
This stoichiometric unbalance has commonly been observed in photocatalytic reactions without sacri cial reagents or oxidation catalysts 12,31 .
There have been numerous reports on photocatalytic CO 2 RR, but the reaction conditions were so different that the reaction activities were hard to compare with each other. The reactions were either in CO 2 -saturated aqueous solutions or in the gas phase with CO 2 and water vapor ow. In the aqueous solution, sacri cial hole acceptors were generally used to avoid the kinetic in uence of counter-oxidation reactions, leading to the additional promotion of CO 2 RR. We selected photocatalysts for CO 2 RR without using sacri cial reagents (Supplementary Table S1) 32 The two hour-reaction was repeated three times using the same catalyst in a fresh CO 2 -saturated aqueous solution for the stability test. The average CH 4 evolution rate was unchanged during the reaction cycles ( Fig. 3c). After the three repetitive reactions, the XPS spectrum in the Cu 2p 3/2 region did not signi cantly change the Cu(I) signal except for a slight increase of the Cu(II) peak intensity. The Auger peaks in the Cu LMM region were nearly identical before and after the catalytic cycles ( Supplementary   Fig. S4). The TEM images after the reaction show that the Cu 2 O domains retain an attachment with the g-C 3 N 4 surface without damaging their cubic morphology ( Supplementary Fig. S5).
We carried out a series of control experiments to ensure the photocatalytic CO 2 RR reaction These superior features of the catalysts for CO 2 RR are ascribed to the strong coupling of two distinct domains by heterojunction formation. When each domain was solely used for the reaction under the identical reaction conditions, the CH 4 production rates were as low as 1.7 μmol h -1 for the g-C 3 N 4 sheets and 3.1 μmol h -1 for the Cu 2 O nanocubes, respectively. On the contrary, g-C 3 N 4 /Cu 2 O produced an 18 times larger amount of CH 4 than each material (Fig. 3d). Interestingly, a physical mixture of the two materials exhibited an activity of 0.089 μmol h -1 , even lower than that with the Cu 2 O nanocubes only. The photoinduced electrons in the Cu 2 O cubes may be consumed by the g-C 3 N 4 sheets through collision, which would prevent the direct electron transfer to CO 2 .
In the UV-visible spectrum (Fig. 2b), the absorption tail is extended to the visible region, although most of the absorption occurs in the UV region below 400 nm. The visible activity of the catalyst was measured by light irradiation using a lter cutting off the wavelengths shorter than 420 nm (Supplementary Fig. S8). CH 4 was still a primary product with a rate of 1.9 μmol h -1 and a selectivity of 95%. The g-C 3 N 4 domains have relatively low absorption in the visible region, which signi cantly affects the CO 2 RR activity.
Reaction pathways through heterojunctions. The origin of superior activity for CH 4 production in the g-C 3 N 4 /Cu 2 O catalyst was examined in detail. The estimated band diagram of the heterojunction catalyst is depicted with corresponding potentials vs. NHE (Fig. 4a). The band gaps were estimated to be 2.8 eV for g-C 3 N 4 and 2.2 eV for Cu 2 O, respectively, by the Tauc plots based on the UV-visible absorption spectra ( Fig. 2b and Supplementary Fig. S9) 40  We performed several experiments to evidence the Z-scheme process as a dominant reaction pathway of the g-C 3 N 4 /Cu 2 O catalyst. Photoresponses were measured using the catalyst deposited on the uorinedoped tin oxide (FTO) electrode at -0.25 V vs. Ag/AgCl in a phosphate buffer. The photocurrent density of the hybrid catalyst (red) was measured to be 32.8 μA cm -2 by irradiation, which was three times higher than that of the g-C 3 N 4 (11.1 μA cm -2 , black) (Fig. 4b). It reveals that the photoinduced charge separation occurs more e ciently in the hybrid structure. The visible light irradiation induced an even signi cant photocurrent difference from g-C 3 N 4 , indicating that the charge separation relies on the bandgap of the Cu 2 O domain ( Supplementary Fig. S10). For the direct comparison, the photoresponse of the ZnO-Cu 2 O catalyst was measured under identical circumstances 18 . The photocurrent of g-C 3 N 4 /Cu 2 O was two times larger than that of ZnO-Cu 2 O (14.6 μA cm -2 ), proving the better performances caused by effective charge separation on the proper catalyst combination (Supplementary Fig. S11).
It is known that the interface between the two domains should form an Ohmic contact with low contact resistance for inducing the Z-scheme-type electron transfer 42 . We suggest that such good electric contact is the main reason for the high CO 2 RR activity, as well as matching band positions of the distinct domains. The photoluminescence (PL) spectrum of g-C 3 N 4 by excitation with 375 nm light shows a prominent peak with a maximum at 435 nm (Fig. 4c). The PL intensity of g-C 3 N 4 /Cu 2 O is ten times weaker than that of g-C 3 N 4 , indicating that the excited electrons at the conduction band of g-C 3 N 4 migrate into the other energy levels; i. e., the valence band of Cu 2 O. We also examined photoexcited carrier dynamics by time-correlated single-photon counting (TCSPC) (Fig. 4d). The overall exciton lifetimes were estimated by the amplitude-weighted average of the triexponential decay functions (Supplementary Table   S2). The lifetimes were 2.9 ns for g-C 3 N 4 and 1.9 ns for g-C 3 N 4 /Cu 2 O, respectively. This lifetime decrease in the hybrid catalyst means the rapid transfer of photoexcited electrons from the g-C 3 N 4 to the Cu 2 O domains through the heterojunction 28,45 .
Electrochemical impedance spectroscopy (EIS) analysis provides additional evidence of the charge transfer behavior 26,46 . The semi-arc loops appear in the middle frequency region in the Nyquist plots. The arc diameters are assigned as the charge transfer resistance -the sum of the electrolyte resistance on the electrodes, the electrode resistance, and the contact resistance with the current collector. In our experiments, the electrolyte and contact resistances are expected to be close due to the structural similarity of g-C 3 N 4 and g-C 3 N 4 /Cu 2 O deposited on the electrodes by the same treatment; then, the measured resistance would be directly related to the charge transfer in the catalyst material. The charge transfer resistances were estimated to be 369.5 Ω for g-C 3 N 4 and 171.0 Ω for g-C 3 N 4 /Cu 2 O, indicating rapid interfacial charge separation and transfer in the g-C 3 N 4 /Cu 2 O hybrid structure (Fig. 4e). These photophysical and electrochemical properties univocally represent the facile photoinduced electron transfer from the g-C 3 N 4 to the Cu 2 O domains on g-C 3 N 4 /Cu 2 O, resulting in enforcing the Z-scheme process for CO 2 RR.
We suppose that this effective photoinduced electron transfer is ascribed to the strong interaction between the hetero-domains induced by growing Cu domains directly on the defect sites of g-C 3 N 4 . X-ray absorption near-edge spectroscopy (XANES) was performed to identify the species and their electronic properties. The Cu K-edge spectrum of g-C 3 N 4 /Cu 2 O (red) shows a distinctive shoulder at 8983.4 eV attributed to the 1s to 3d transition and a strong absorption edge at 8996.5 eV for the 1s to 4p transition (Fig. 4f). The two edges shift to the high energy by 0.8 and 1.6 eV, compared to the absorption edges of the Cu 2 O standard (blue). These positive shifts were similarly observed in single Cu atoms deposited on carbon nitrides, indicating that the strong N-coordination on the Cu surface is present at the interface between the two domains 47 . The Cu-N bonding is likely to form at the early stage of the polyol synthesis. The Cu(I) atoms reduced from the copper precursor can be bound on the N atoms at the defects of g-C 3 N 4 , and direct and continuous Cu deposition on these sites leads to growing single-crystalline Cu 2 O domains. As a result, a large number of Cu-N bonding with the defect sites generates quasi-continuous energy levels at the contact interface. This situation is like the conductors, ready to form the Ohmic contact promoting the electron transfer between the two domains, enforcing the Z-scheme process, and inducing the stable charge-separated states [47][48][49] .

Discussion
In conclusion, we induced the direct growth of Cu 2 O nanocubes on g-C 3 N 4 sheets by a polyol process.
The resulting g-C 3 N 4 /Cu 2 O hybrid formed good electric junctions between the two distinct domains. The hybrid was employed as a photocatalyst for CO 2 RR in an aqueous solution. They exhibited enormous activity of 5.4 mmol g cat -1 h -1 with a quantum yield of 3.7% and a selectivity of 98% for CH 4 production. To the best of our knowledge, g-C 3 N 4 /Cu 2 O is one of the most active photocatalysts among any other catalytic materials for CO 2 RR in aqueous media without sacri cial reagents. The catalyst was recyclable three times without changing the catalyst morphology. The photocatalytic reaction mechanism was rationalized as the Z-scheme process. The direct growth of Cu on g-C 3 N 4 led to the strong coordination of g-C 3 N 4 to the Cu 2 O domains, which formed conductive junctions to promote the facile electron transfer from g-C 3 N 4 to Cu 2 O following the Z-scheme, and provided su cient overpotentials for CO 2 RR. It reveals that the primary keys of hybrid photocatalyst design toward high-performance CO 2 RR are: 1) matching band positions and 2) generating good ohmic junctions between two distinct materials. This work recon rms basic principles to fabricate high performance photocatalysts bearing a general form of catalyst-cocatalyst combinations.
Synthesis of g-C 3 N 4 nanosheets. 5.0 g of urea was placed in a semi-closed quartz reactor with ber glasses. The reactor was heated to 550 °C for 2 h under static air in a furnace with a ramping rate of 5 °C min -1 . After the reaction, the reactor was cooled to room temperature. The resulting light-yellow product was thoroughly washed with ethanol and water.
Synthesis of g-C 3 N 4 /Cu 2 O hybrid catalysts. The solution comprising g-C 3 N 4 (0.015 g), PVP (1.5 g, 12.0 mmol), and 1,5-PD (50 mL) was heated to 130 °C under inert conditions. The mixture was heated to 225°C . Subsequently, a solution of copper acetylacetonate (0.040 g, 0.16 mmol) and 1,5-PD (5.0 mL) was added and stirred at the same temperature. After 5 min, the reaction mixture was cooled to room temperature, yielding g-C 3 N 4 /Cu 2 O hybrid catalysts. The resulting light orange product was thoroughly washed with ethanol.
Materials Characterization. TEM and HRTEM images were obtained by using an FEI Tecnai G2 F20 (200 kV). STEM and EDS elemental mapping images were obtained in Talos F200X (200 kV). The specimens were prepared by dropping a few samples dispersed in ethanol on carbon-coated 200-mesh nickel grids (Ted Pella Inc.). XRD patterns were recorded on a Rigaku D/MAX-2500 diffractometer. XPS was obtained by K-alpha X-ray photoelectron spectroscopy (Thermo VG Scienti c). UV-visible absorption spectra were measured from Shimadzu UV-3600 spectrophotometer at room temperature. Steady-state photoluminescence (PL) spectra were obtained from a spectrophotometer (F-7000, Hitachi) at room temperature. Time-resolved photoluminescence (PL) spectra were obtained from a spectrophotometer (Fluotime 300, PicoQuant) with picosecond-pulsed diode lasers (output wavelengths of 375 nm) at room temperature.
X-ray absorption spectroscopy analysis. X-ray absorption ne spectroscopy (XAFS) of hybrid catalysts over Cu K-edge, 8979 eV in copper foil was analyzed at Pohang Accelerator Laboratory (7D-XAFS beamline in PLS-II) using Si (111)  Photocatalytic CO 2 conversion experiments. Before photocatalytic measurements, g-C 3 N 4 /Cu 2 O hybrid catalysts were thoroughly washed with ethanol and water to remove carbon residues. Then, 10 mg of g-C 3 N 4 /Cu 2 O hybrid catalysts were dispersed in water (80 mL). The reactor was a homemade quartz ask with a total volume of 215 mL. Supercritical-uid grade CO 2 gas was used to avoid any hydrocarbon contamination. To make the CO 2 saturation in the reaction medium, CO 2 gas was additionally bubbled at ambient pressure and temperature for 30 min. The photocatalytic CO 2 conversion was conducted by irradiation using a Xe lamp (300 W, Oriel) equipped with a 10 cm IR water lter. During the reaction, the gas product was collected using a needle-type probe passing through a sealed rubber septum. The gas sample was analyzed by an FID detector equipped with a Carboxen 1000 column in gas chromatography.
Mott-Schottky plot measurements. The at-band edge positions of g-C 3 N 4 and g-C 3 N 4 /Cu 2 O were measured by electrochemical impedance measurement. Quantum e ciency calculation. The apparent quantum e ciency was calculated by a method reported by Bharadwaj et al. 30 First, the incident light power was measured using the power meter (Newport 848-R), assuming that the light intensity was uniformly distributed. Second, the emission pro le of the Xe lamp was recorded by using a ber-coupled spectrometer (Ocean Optics USB4000) in the range of 200-700 nm. Third, based on UV-visible absorption spectrum data, the effective absorption for the photocatalytic reaction was assumed in the range of 200-520 nm. The light power (E m ) in the range of 200-520 nm was calculated by multiplying the total light power (200-700 nm) with the fraction of the area (200-520 nm) to the total area (200-700 nm) in the emission spectrum. The energy of a single photon was considered to be the weighted average energy of all photons in the range of 200-520 nm. For this, contribution of a photon of each wavelength (200-520 nm) towards the total energy was calculated as where I λi is the intensity at wavelength λ i , and I total is the sum of intensities of all photons in the wavelength range of 200 to 520 nm obtained from the emission pro le of the lamp. The weighted average energy of single-photon (E a ) is calculated as 49. Li, Y., Li, B., Zhang, D., Cheng, L. & Xiang, Q. Crystalline carbon nitride supported copper single atoms for photocatalytic CO 2 reduction with nearly 100% CO selectivity. ACS Nano 14, 10552-10561 (2020).

Supplementary Files
This is a list of supplementary les associated with this preprint. Click to download. C3N4Cu2OSupplement.docx