WANG Hong-zhi, ZHAO Yue-zhu, YANG Zhong-xue, BI Xin-ze,WANG Zhao-liang, WU Ming-bo

（College of New Energy, State Key Laboratory of Heavy Oil Processing, China University of Petroleum (East China), Qingdao 266580, China）

Abstract： Using CO2 as a renewable carbon source for the production of high-value-added fuels and chemicals has recently received global attention.The photoelectrocatalytic (PEC) CO2 reduction reaction (CO2RR) is one of the most realistic and attractive ways of achieving this, and can be realized effectively under sunlight illumination at a low overpotential.Oxygen-incorporated carbon nitride porous nanosheets (CNs) were synthesized from urea or melamine by annealing in nitrogen or N2/O2 gas mixtures.They were used as the photoanode with Bi2CuO4 as the photocathode to realize PEC CO2 reduction to the formate.The electrical conductivity and the photoelectric response of the CNs were modified by changing the oxygen source.Oxygen in CNs obtained from an oxygen-containing precursor improved the conductivity because of its greater electronegativity, whereas oxygen in CNs obtained from the calcination atmosphere had a lower photoelectric response due to a down shift of the energy band structure.The CN prepared by annealing urea, which served as the source of oxygen and nitrogen, at 550 °C for 2 h in nitrogen is the best.It has a photocurrent density of 587 μA cm?2 and an activity of PEC CO2 reduction to the formate of 273.56 μmol cm?2 h?1, which is nearly 19 times higher than a conventional sample.The CN sample shows excellent stability with the photocurrent remaining constant for 24 h.This work provides a new way to achieve efficient catalysts for PEC CO2 reduction to the formate, which may be expanded to different PEC reactions using different cathode catalysts.

Key words: Oxygen-incorporated；Carbon nitride；Photoelectrocatalytic；CO2 reduction reaction；Formate

1 Introduction

Over the past few decades, global warming and energy shortage are increasing serious with the overuse of fossil fuels.Converting clean energy sources(such as solar energy) to chemical energy, which is easy to further transportation, storage and utilization by photocatalytic (PC) reactions, is a promising approach to alleviate energy and environmental issues[1–2].However, PC reactions usually subject to the low quantum efficiency and reaction activity because of the rapid recombination of charge carriers[3–4].On the other hand, photoelectrochemical(PEC) reduction can integrate and optimize the advantages of both photocatalysis and electrocatalysis(EC) to promote charge separation and improve solar conversion efficiency[5–10].So, the balance between the electrical conductivity of EC process and photoelectric response of PC process has obvious effect on PEC properties.The electrical conductivity represents the transfer rate of charges, and the photoelectric response represents the conversion rate of light energy to electrical energy[11–12].Nowadays, PEC reaction has been used in a lot of areas, such as water splitting[13],ammonia synthesis[14], pollutant degradation[15]and CO2conversion[11].Therein, PEC CO2conversion is the most promising and simplest method so far because it can be useful to control thermodynamics and kinetics of CO2reduction reaction by mimicking natural photosynthesis without high input energy cost, so as to achieve high product selectivity and fast reaction kinetics[16].During the PEC CO2conversion process, the valid catalysts is the key to achieve high activity, high selectivity, and long stability.Therefore,it is highly desirable to construct suitable catalysts with both excellent electrical conductivity and photoelectric response to realize highly active, selective,and stable PEC CO2conversion.

Recently, as a promising visible-light-driven catalyst for photocatalysis, energy conversion, lightemitting devices and tribological coatings, two-dimensional metal-free carbon nitride (CN) has received tremendous attention[17–18], owing to its tunable electronic property, suitable band gap (2.7 eV), facile synthesis, remarkable chemical[19], thermal stability[20], and nontoxicity[21].In 2009, Wang and co-workers first reported g-C3N4as a novel photocatalyst that exhibited photoactivity for H2generation under visible-light irradiation[22].Then, many efforts have been made to synthesize g-C3N4through the thermal treatment of some nitrogen-rich organic precursors, such as cyanamid[23], dicyanamide[24], triazine and heptazine derivatives.It has also been documented that parameters such as the precursor (urea, melamine) and thermal treatment conditions (temperature, reaction time, and atmosphere) used for the preparation of g-C3N4greatly affect its photocatalytic activity[25].However,pristine g-C3N4still suffers from small specific surface area (10-15 m2g?1), poor light absorption at long wavelength, low charge migration rate and high recombination rate of photogenerated electron-hole pairs, resulting in unsatisfactory photocatalytic performance in real practices[26].Many attempts have been devoted to improve the photoresponse of g-C3N4including cationic doping, anionic doping and hybridization with another semiconductor materials[27].Among various approaches, impurity doping is an effective method to enhance the photocatalytic performance through tuning the band gap structure, extending the light absorption, increasing the charge transfer mobility and creating more active sites[28].Shen et al.reported ultrathin boron-doped and nitrogen-deficient g-C3N4nanosheets as H2― and O2― evolving photocatalysts through the strategy of electrostatic self-assembly, and the solar-to-hydrogen efficiency reaches 1.16% under one-sun illumination[29].Xie et al.prepared the ultrathin g-C3N4by a “green” liquid exfoliation route from bulk g-C3N4in water, which induced their extremely high PL quantum yield up to 19.6%through enhancing the intrinsic photoabsorption and photoresponse[30].Lin et al.synthesized silver-decorated ultrathin g-C3N4nanosheets (Ag@U-g-C3N4-NS),which exhibited enhanced electrochemical properties and excellent performance for the degradation of organic pollutants[17].Yu et al.prepared a sulfur-doped g-C3N4(TCN) by simply calcinating thiourea at 520 °C, which showed higher activity for photocatalytic reduction of CO2to CH3OH than that of undoped g-C3N4(MCN)[31].However, besides the photoresponse, the conductivity of the CN also has effects on the PEC activity by changing the transport resistance of photogenerated charges.

Herein, we report the CN catalysts with excellent PEC performance, which are designed and fabricated as photoanode to drive CO2conversion with the assistance of Bi2CuO4cathode to generate formate with high efficiency.The electrical conductivity and the photoelectric response of CN are improved successfully by controlling the oxygen source during the synthesis process.The oxygen from the precursor could improve the conductivity because of the more negative electronegativity.On the other hand, the oxygen from the calcination atmosphere showed side effects on the photoelectric response by changing the energy band structure.Under the optimal conditions,the photocurrent density is 587 μA cm?2and the activity of PEC CO2reduction to formate is 273.56 μmol cm?2h?1(nearly 19 times than that of the conventional sample).

2 Materials and methods

2.1 Materials

Melamine (purity ≥ 99.0% ), urea (purity ≥99.0%), potassium hydrogen carbonate (purity ≥99.5%), sodium sulfate (purity ≥ 99.0%), cupric nitrate (Cu(NO3)2·3H2O), bismuth nitrate (Bi(NO3)3·5H2O) were purchased from Sinopharm Chemical Reagent Co.Nafion solution (5%) were purchased from Aladdin Industrial Corporation.All the chemicals were used as received without further purification.

2.2 Synthesis of CNs

The CN samples were prepared by thermal polymerization.In detail, 2 g melamine mixed with urea powder (0.33 g) was placed in a porcelain crucible with a lid after ground for 30 min.The crucible containing urea was heated in the tube furnace at 550 °C for 120 min with a heating rate of 5 °C min?1under N2gas flow (50 mL min?1).The yellow powder was obtained.The as-prepared product was labeled as CN1.CN2 was prepared with the same procedure as CN1 but thermally polymerized under N2/O2mixed gas.2 g melamine was heated with the same procedure as CN1 but without mixing urea, which was labeled as CN3.CN4 was prepared with the same procedure as CN3 but thermally polymerized under N2/O2mixed gas.

2.3 Synthesis of Bi2CuO4

The Bi2CuO4was synthesized by the previously reported method[32].Cu(NO3)2·3H2O (0.120 8 g,0.5 mmol) and Bi(NO3)3·5H2O (0.485 g, 1 mmol)were dissolved in 20 mL water, and 10 mL KOH(8 mol L?1) was added and kept refluxing for 0.5 h.The mixture solution was added to a 50 mL Teflon reactor and heated to 150 °C for 12 h.After filtration,the precipitate was washed with water to remove unreacted reagents.The obtained powder was further dried under vacuum at 60 °C for 12 h to get Bi2CuO4.

2.4 Material characterization

The crystal structure of products was characterized by X-ray diffraction (XRD) (X’Pert PRO MPD,Holland) with CuKα (λ= 0.154 06 nm).The surface morphology of each catalyst was observed by transmission electron microscopy (TEM, JEM-2100UHR,Japan).The electron paramagnetic resonance (EPR)results were collected on Bruker A200 at room temperature.The pore size distribution was obtained by the Barrett-Joyner-Halenda method using the adsorption branch.The surface functional groups and chemical compositions of products were performed by Fourier transform infrared spectroscopy (FT-IR,Bruker V70) and X-ray photoelectron spectroscopy(XPS, Escalab 250Xi, UK) using AlKα radiation and 500 μm X-ray spot.The optical properties of products were tested by UV–vis diffuse-reflectance spectroscopy (UV?vis DRS UV-2700, Shimadzu, Japan).Band gap determination from UV-vis diffuse reflectance spectra were first transformed to absorption spectra according to the Kubelka-Munk function,

whereRis the relative reflectance of samples with infinite thickness compared to the reference.Moreover,the band gaps of samples were estimated on the basis of the Tauc equation

whereh,v,AandEg represents Planck constant, light frequency, proportionality constant and band gap, respectively, andndepends on the nature of transition in a semiconductor.The values ofEgwere determined from the plot of [F(R)hν]n2againsthvand corresponded to the intercept of the extrapolated linear portion of the plot near the band edge with the hν axis.CNs samples were treated as the semiconductors with allowed indirect transition.The values ofEgwere thus determined from the plot of [F(R)hν]n2against hν.

Photoluminescence (PL) spectra were measured on a fluorospectrophotometer (F97pro, Lengguang Tech, China) with an incident light of 330 nm.The VB-XPS spectrum was carried out to determine the valence band (VB), then the CB potentials of different photocatalysts were calculated according to the following equations:ECB=EVB?Eg, whereECB,EVB, andEgare the conduction band potential, valence band potential (EVBfrom VB-XPS), band gap (Egfrom UV–vis DRS), respectively.

2.5 Electrochemical measurements

Electrochemical measurements were performed on Vertex.C.EIS (Ivium Technologies BV, Netherlands) with a typical three-electrode cell using 0.5 mol L?1sodium sulfate (Na2SO4) aqueous solution(pH = ～ 7) as the electrolyte.The KCl-saturated Ag/AgCl and Pt net were used as the reference electrode and counter electrode, respectively.1 mg of catalysts and 20 μL of Nafion solution (5%) were dispersed in 200 μL of deionized water by ultrasonication for 30 min to form a homogeneous ink.The catalyst ink was painted onto a piece of glassy carbon electrode and dried at atmosphere for more than 8 h to serve as the working electrode.A 300 W Xe lamp(Beijing Perfectlight Technology Co., Ltd.) equipped with a AM1.5G filter was utilized as the light source.

The linear sweep voltammetry (LSV), electrochemical impedance spectroscopy (EIS) from 1 to 10 000 Hz at different applied potentials, transient photocurrent, and Mott–Schottky plots were performed in 0.5 mol L?1sodium sulfate (Na2SO4)aqueous solution electrolyte (pH= ～ 7).All potentials in this study were measured against the Ag/AgCl reference electrode and converted to the RHE reference scale usingE(vs RHE)=E(vs Ag/AgCl)+0.21 V+0.059 1×pH[33].

2.6 Photoelectrochemical experiments

Electrochemical measurements were performed in a three-electrode system with a two-compartment cell connected to an electrochemical workstation (CHI 760E, CH Instrument, China).The cathode containing 30 mL of 0.5 mol L?1KHCO3as the electrolyte and the anode containing 0.5 mol L?1Na2SO4aqueous solution (pH = ～ 7) as the electrolyte were separated by a proton exchange membrane (Nafion 117).

2.7 Analysis of products

The gas phase products of PEC CO2reduction reaction were detected by an on-line gas chromatography (BFRL-3420A, China) with a thermal conductivity detector (TCD) for detecting hydrogen and a flame ionization detector (FID) for detecting carbon monoxide and other hydrocarbons.The liquid phase products were detected by liquid chromatography(LC-2030 Plus, SHIMADZU, Japan).

3 Results and discussion

3.1 Structure and morphology of CNs

The fabrication process of CN catalysts was illustrated schematically in Fig.1a.In detail, the CN catalysts were synthesized by pyrolysis of melamine and/or urea under different atmospheres (N2or mixed gas of N2and O2) at 550 °C.Fig.1b-e are the TEM images of CN1, CN2, CN3 and CN4, respectively.All the pictures show that the CN samples are free-standing nanosheets, and nearly transparent, indicating that the thickness is extremely ultrathin.Besides, the well-defined Tyndall effect of the CN solution indicates the presence of highly monodisperse ultrathin nanosheets in water from the inset digital photograph of Fig.1b-e.All the pictures indicate that the morphology and thickness of the synthesized CN samples are all independent of the preparation process.

Fig.1 (a) Schematic of the synthesis derived from melamine or/and urea.TEM images of (b) CN1, (c) CN2, (d) CN3 and (e) CN4.The inset digital photographs show CN nanosheet solutions

According to the XRD patterns (Fig.2a), only CN1 and CN2 exhibit the typical diffraction signals of(002) at about 27.1°.There are no obvious diffraction signals in the samples of CN3 and CN4.The results suggest that the precursor urea could improve the crystallinity of the samples, and the oxygen from the mixed gas has no effect on the crystallinity.The functional groups on the surface structure of CN samples were further revealed by FT-IR spectra.As shown in Fig.2b, similar characteristic peaks of CN samples can be observed in the four samples.The sharp peaks at 808 cm?1correspond to a breathing vibration mode of the tri-s-triazine units[34].The main absorption bands in the range of 1 150–1 650 cm?1are attributed to the characteristic C―N and C＝N stretching vibration of the skeletal tri-s-triazine units.Moreover, the broad bands ranging from 2 900 to 3 600 cm?1can be assigned to the stretching vibrations of surface amino group and adsorbed ―OH group[35].All the results indicat that the four CN samples were successfully synthesized by pyrolysis of nitrogen-containing carbon precursors.Fig.2c compares the room-temperature EPR spectra of the CN samples.EPR signal intensity is an indication of the concentration of unpaired electrons in flat π-conjugated CN samples[34].Clearly, the CN1 exhibits the highest EPR signal intensity atg=2.005 among all the samples, which can be ascribed to the more unpaired electrons.This result indirectly reflects the existence of more surface structure defects in CN1, which may contribute to separate the photo generated charge during the PEC reaction.The photoluminescence (PL) spectra of all the samples were determined under 330 nm excitation to investigate the recombination efficiency of the photo-generated carriers under radiation.The PL spectra evidently display emission peaks at about 450 nm on account of the recombination of the photo-generated electron–hole pairs.Generally, the stronger PL intensity indicates the faster combination of photogenerated charges[28].As shown in Fig.2d, the peak intensity of CN1 is the smallest among the four catalysts, suggesting that the recombination of electron–hole pairs is obviously suppressed, thus improving charge separation, which is consistent with the result of the transient photocurrent density response.

Fig.2 (a) X-ray diffraction patterns, (b) FT-IR spectra, (c) EPR spectra and (d) steady-state PL spectra of CN samples

3.2 Photoelectrochemical performance of CNs

Due to the n-type conductivity (Fig.S1), the oxygen-incorporated CN catalysts could be characterized as photoanodes using a standard three-electrode system in an aqueous solution.Fig.3a shows the currentpotential characteristics of the four CN samples.The sample CN4 shows the higher current densities than that of CN3 at all potentials, indicating that pyrolysis of melamine in mixed gas (N2and O2) could enhance its conductivity than in inert gas (N2), which may be resulted from the incorporated oxygen atom from O2.In addition to provide more oxygen by the calcination atmosphere, the sample of CN2was synthesized by calcining oxygen-containing precursors (urea).The current density of CN2 is the smallest among the four CN catalysts, which may be resulted from the fact that the structure of CN samples were broken by incorporating excessive oxygen.Besides, the sample CN1 shows the highest current density among all the samples, indicating its optimal electrical conductivity.All the results suggest that the content and source of oxygen can change the conductivity of the CN catalysts, and there is an appropriate oxygen doping amount to obtain the optimum electrical conductivity.In addition to the conductivity, the photo-electric response intensity also affects the performance of PEC CO2RR.We measured the photocurrent density using simulated AM 1.5G illumination to evaluate the photo-electric response signal.As shown in the Fig.3b and Fig.S2, an obvious increase of the photocurrent is observed upon light illumination over the entire voltage range.Beyond the LSV curves, the CV curves (Fig.S3) and EIS curves (Fig.S4) of all the four samples also show the visible difference between the dark and AM 1.5G illumination.In detail, the area of CV curves under light irradiation is larger than that in dark, and the radius in the EIS curves under light irradiation is smaller than that in dark.All the results indicate that the prepared CN samples show excellent light response signal and the separation efficiency of photogenerated charges and transmission could be improved significantly under AM 1.5G illumination.

Fig.3 (a) LSV curves.(b) LSV curves of CN1 under dark (full line) and AM 1.5G illumination (dotted line).(c) Transient amperometric I–t curves at ?0.9 V vs. RHE under AM 1.5G illumination.(d) Photocurrent stability of CN1 photoanode under AM 1.5G illumination

Fig.3c shows the photocurrent–time (I–t) curves of the 4 CN samples at an external bias of ?0.9 V versus reversible hydrogen electrode (RHE) under light irradiation.The traditional CN4 that was prepared by pyrolysis of melamine in N2shows the small photocurrent density of 0.032 mA cm?2.While the photocurrent density of CN3 is about 8.5 times bigger than that of CN4.Besides, the same tendency could be observed comparing the photo current density of CN1 with CN2.The tendency indicates that the oxygen in the CN catalysts from the mixed gas may accelerate the photo charge recombination to reduce the photo current density.On the other hand, the oxygen in the CN catalysts from the oxygen-containing precursor(urea) is beneficial to the separation of the photogenerated charge.In summary, the sample of CN1 shows the highest photo-current density among all the samples, indicating the high separation efficiency of photogenerated charge carriers.The results of LSV andI-tcurves prove that the oxygen in the CN catalysts could change the conductivity and the photoelectric response.In detail, the oxygen from the mixed gas may reduce the photoelectric response and the oxygen from the oxygen-containing precursor may improve the conductivity.As exhibited in Fig.3d, for the CN1 photoanode under AM 1.5G light illumination,the photocurrent of 1.5 mA cm?2at a potential of?0.9 V versus RHE could be maintained for at least 24 h.All these results strongly suggest that the CN1 photoanode possesses excellent performance under continuous light illumination.

3.3 PEC CO2 reduction performance of CNs

As illustrated in Fig.4a, the CN samples with different oxygen contents were adopted as photoanodes, the Bi2CuO4(Fig.S5) was used as the cathode and Ag/AgCl as the reference electrode to realize the PEC CO2reduction to formate.Under light irradiation, electrons in the CNs could be excited to the conduction band (CB) and transfer to the counter electrode of Bi2CuO4via the external circuit to finish the reduction reaction of CO2to formate.As can be seen in Fig.4b, under AM 1.5G illumination, the CN1,CN2, CN3 and CN4 provide PEC CO2reduction to formate activities of 273.56, 118.08, 92.63 and 14.11 μmol cm?2h?1, respectively, at a bias of ?0.9 Vvs.RHE.The results agree well with the photocurrent density of the four catalysts under the same light irradiation (Fig.3c).In other words, the high yield of formate is derived from the excellent photogenerated charge separation and transfer of the CN, which is supported by energy band alignment characterization as discussed below.Such a rate of PEC CO2RR to formate is superior to many other catalysts tested under similar conditions (Table S1).

Fig.4 (a) Schematic illustration of AM 1.5G light induced electron/hole separation and transfer process during PEC CO2RR in CN photoanodes.(b) The PEC performance of CN samples toward CO2 reduction to formate

To understand the excellent PEC CO2reduction of the CN catalysts under AM 1.5G light irradiation,the UV-vis spectra and Tauc plots of CN samples are displayed in Fig.5a and 5b.As shown in Fig.5a, all the samples exhibit an absorption edge at around 480 nm, and the corresponding Tauc plots show that the band gap energies are 2.58 eV (CN1), 2.55 eV(CN2), 2.67 eV (CN3), 2.63 eV (CN4) (Fig.5b).The phenomenon indicates that the prepared CN samples could absorb visible light effectively.So it is hypothesized that the significant difference of photocurrent density of the four catalysts is not induced from the light absorption.To investigate the band positions,XPS valence band spectra (Fig.S6), Mott-Schottky plots (Fig.S1) along with the Tauc plot (Fig.5b) were used to determine the band alignment as shown in Fig.5c.Although the bandgaps of CN samples show the similar results, the CB and VB of the 4 samples down shift with the same trend as the photocurrent density (Fig.5c).This further proves that the oxygen from the mixed gas may reduce the photoelectric response and the oxygen from the oxygen-containing precursor may improve the conductivity.The oxygen contents are measured by XPS (Fig.5d, S7, S8,Table S2).It is observed that CN catalysts are comprised of C, N and O elements.As shown in Fig.4d,the O1s peaks in high-resolution spectra of CN1 and CN2 located at ca.533.3-533.5 eV are attributed to C―O―C bond, which further proves that the oxygen from the oxygen-containing precursor could be doped into the lattice.Furthermore, the peak at 532.6-532.8 eV of CN2 and CN4 simples correspond to adsorbed ―OH[28], and the peak at 531.6-531.9 eV of CN1, CN2 and CN4 samples correspond C―O bond.All the results indicate that the oxygen from the atmosphere could only adsorb on the surface of the CN samples.The N1s spectra in high-resolution spectra are shown in Fig.S8, the major peak centered at 398.4-398.8 eV corresponds to sp2-hybridized nitrogen in C-containing triazine rings (C―N＝C), and the peak at 400.1-400.4 eV corresponds to N―(C)3 or H―N―(C)2[28].

Fig.5 (a) UV–vis absorption spectra, (b) Tauc plots, (c) the band alignment and (d) XPS spectra of O1s for CN samples

4 Conclusions

In conclusion, we report oxygen-incorporated carbon nitride porous nanosheets, which are capable for PEC CO2reduction under AM 1.5G light irradiation using Bi2CuO4as a cathode.By precise control the oxygen source, the oxygen content and doping form could be accurately tailored, which result in the different electrical conductivities and the photoelectric responses.Based on the optimal electrical conductivity and the photoelectric response of CN1, high PEC CO2reduction activity (273.56 umol cm?2h?1)and excellent photocurrent density stability for 24 h under AM 1.5G light irradiation are achieved.It is envised that this study could be extended to the fabrication of PEC reaction cell using other cathode catalysts, which may pave the way to the development of a new class of solar energy conversion materials.

Acknowledgements

This work was financially supported by the National Natural Science Foundation of China(52072409), Natural Science Foundation of Shandong Province (ZR2021QE062), Major Scientific and Technological Innovation Project of Shandong Province (2020CXGC010402), Qingdao postdoctoral applied research project (qdyy20200063), and Taishan Scholar Project (ts201712020).