Phosphorus-Doped Hollow Tubular g-C3N4 for Enhanced Photocatalytic CO2 Reduction

Photocatalytic CO2 reduction is a tactic for solving the environmental pollution caused by greenhouse gases. Herein, NH4H2PO4 was added as a phosphorus source in the process of the hydrothermal treatment of melamine for the first time, and phosphorus-doped hollow tubular g-C3N4 (x-P-HCN) was fabricated and used for photocatalytic CO2 reduction. Here, 1.0-P-HCN exhibited the largest CO production rate of 9.00 μmol·g−1·h−1, which was 10.22 times higher than that of bulk g-C3N4. After doping with phosphorus, the light absorption range, the CO2 adsorption capacity, and the specific surface area of the 1.0-P-HCN sample were greatly improved. In addition, the separation of photogenerated electron–hole pairs was enhanced. Furthermore, the phosphorus-doped g-C3N4 effectively activated the CO2 adsorbed on the surface of phosphorus-doped g-C3N4 photocatalysts, which greatly enhanced the CO production rate of photocatalytic CO2 reduction over that of g-C3N4.


Introduction
Human beings are facing two major challenges today: a huge energy demand and serious environmental problems [1,2].As an inexhaustible clean energy source, solar energy has been widely studied and utilized for decades, and the high-efficiency utilization of solar energy through photocatalysts has become a research hotspot in recent decades [3,4].The photocatalytic reduction of CO 2 into valuable products such as CO and CH 4 is considered a promising technology for alleviating the greenhouse effect [5,6].The discovery of single-layer graphene brought extensive attention to 2D materials, and its unique properties are favored in different research fields [7].For example, as a two-dimensional nonmetal semiconductor material, graphitic carbon nitride (g-C 3 N 4 ) has attracted increasing attention from researchers due to its inexpensive raw materials, simple preparation methods, excellent photoelectric physical structure properties, and chemical stability [8].At present, g-C 3 N 4 has been extensively studied in photocatalytic CO 2 reduction and has shown excellent performance and considerable application prospects [9].However, the large block microstructure and low specific surface area of bulk g-C 3 N 4 prepared directly through traditional thermal polymerization are not conducive to the full exposure of active sites, the migration of photogenerated charges, and the mass transfer process of reactants [10].In addition, bulk g-C 3 N 4 tends to exhibit a narrow light absorption range and severe recombination of photogenerated electrons, resulting in lower catalytic activity [11].
The preparation of g-C 3 N 4 with a specific microstructure has become an available strategy for augmenting the specific surface area and accelerating the transfer of photogenerated electrons and holes.In addition, 3D g-C 3 N 4 with special microstructures of hollow spheres and a 3D network structure can be prepared by using SiO 2 microspheres

Materials
All chemicals were of analytical grade and used without further purification.Melamine and Nafion ® solutions were purchased from Aladdin Industries.NH 4 H 2 PO 4 was purchased from Xilong Technology Co., Ltd.(Shantou, China).

Synthesis of the Photocatalyst
Synthesis of bulk g-C 3 N 4 : 5 g of melamine was placed in a crucible with a lid and transferred to a muffle furnace, which was heated to 550 • C with a heating rate of 5 • C/min, and then kept for 2 h in an air atmosphere.The bulk g-C 3 N 4 was ground and collected after cooling to room temperature.
Synthesis of phosphorus-doped hollow tubular g-C 3 N 4 and hollow tubular g-C 3 N 4 : First, the precursors were synthesized with a hydrothermal method; 5 g of melamine and x g of NH 4 H 2 PO 4 (x = 0.5, 1.0, 1.5, 2.0) were dispersed in 60 mL of deionized water and stirred for 30 min.Then, the mixture was transferred to a stainless steel autoclave with Teflon lining and kept at 180 • C for 12 h.The solid product was separated through centrifugal filtration, washed with deionized water, and then dried at 60 • C for 12 h.Finally, the obtained solid was placed in a covered crucible, and the temperature was raised to 550 • C at a rate of 2.5 • C/min and kept for 2 h.The phosphorus-doped hollow tubular g-C 3 N 4 was ground and collected after cooling to room temperature; it was designated as x-P-HCN (x = 0.5, 1.0, 1.5, 2.0), and the synthesis process is shown in Scheme 1.In addition, the hollow tubular g-C 3 N 4 (HCN) was prepared with the same method without adding the NH 4 H 2 PO 4 .
Materials 2023, 16, x FOR PEER REVIEW 3 of 16 Scheme 1. Schematic illustration of the synthesis of x-P-HCN.

Characterization
X-ray powder diffraction (XRD) was performed on a SMARTLAB3KW X-ray powder diffractometer (Akishima, Japan) equipped with a Cu-Kα radiation source.Raman spec-Scheme 1.Schematic illustration of the synthesis of x-P-HCN.

Characterization
X-ray powder diffraction (XRD) was performed on a SMARTLAB3KW X-ray powder diffractometer (Akishima, Japan) equipped with a Cu-Kα radiation source.Raman spectra were obtained on a LabRam HR 800 laser confocal Raman spectrometer (Paris, Franch).SEM images were recorded on a ZEISS Gemini 300 (Oberkochen, Germany) field-emission scanning electron microscope (FE-SEM).Transmission electron microscopy (TEM) and high-resolution transmission electron microscopy (HR-TEM) images and energy-dispersive spectroscopy (EDS) mappings were acquired on an FEI Talos F200S transmission electron microscope (Waltham, MA, USA).X-ray photoelectron spectroscopy (XPS) was performed on a Thermo Scientific K-Alpha system (Waltham, MA, USA).Time-resolved fluorescence spectra were acquired on an FLS 1000 steady-state/transient fluorescence spectrometer (Livingston, UK).Photoluminescence (PL) spectra were recorded on a Thermo Scientific Lumina fluorescence spectrometer (Waltham, MA, USA).Ultraviolet-visible diffuse reflectance spectra (UV-vis DRS) were recorded using a TU-19 ultraviolet-visible spectrophotometer (Beijing, China).The N 2 adsorption and desorption curves, Brunauer-Emmett-Teller (BET) specific surface area, and pore size distribution of the samples were obtained on a TriStar II system (Norcross, GA, USA).CO 2 temperature-programmed desorption (CO 2 -TPD) was carried out on an Altamira AMI 300 system (Pittsburgh, PA, USA).The samples were treated in a He atmosphere at 300 • C for 1 h to eliminate surface substances contributing to physical adsorption, cooled to room temperature, injected with pure CO 2 for 1 h, and then heated from 50 • C to 450 • C at a rate of 10 • C/min.Fourier transform infrared (FT-IR) spectra and in situ diffuse reflectance infrared Fourier transform spectroscopy (DRIFTS) experiments were performed on a Bruker TENSOR II infrared spectrometer (Karlsruhe, Germany).For in situ DRIFTS, the samples were treated in Ar (30 mL/min) at 300 • C for 1 h to eliminate impurities on the sample surface and then cooled to room temperature.After that, CO 2 (20 mL/min) was passed through deionized water and introduced into the infrared cell for 60 min in the dark.The reaction temperature was maintained at 25 • C using a cooling circulating water system.Then, argon gas was injected into the infrared cell for 20 min to discharge free CO 2 and water vapor, and a 300 W xenon lamp (CEL-HXF300, Beijing China Education Au-light Co., Ltd., Beijing, China) was used as the light source.The IR spectra were collected in the dark for 60 min and under light irradiation for 60 min.

Photocatalytic CO 2 Reduction
First, 30 mg of the photocatalyst was dispersed at the bottom of a 220 mL quartz reactor containing 3 mL of deionized water, and then the photocatalyst at the bottom of the reactor was dried at 60 • C for 8 h.Before the photocatalytic reaction, CO 2 and water vapor were allowed to enter the reactor at a rate of 40 mL min −1 for 0.5 h by bubbling CO 2 gas through deionized water.A 300 W Xe lamp with a 400 nm cutoff filter was used as the light source.Gas products were analyzed with a Shimadzu GC-2030 gas chromatograph (Kyoto, Japan) with a barrier discharge ionization detector (BID).When carrying out the cycle experiment, the quartz reactor containing the photocatalyst was vacuum-dried after each cycle reaction; then, CO 2 and water vapor were reintroduced into the reactor, and the other reaction conditions remained unchanged.

Photoelectrochemical Measurements
Photoelectrochemical measurements were performed on an electrochemical workstation (CHI 760E, Shanghai Chenhua, Shanghai, China) with a three-electrode system.In this system, the Ag/AgCl electrode was used as the reference electrode, the platinum electrode was used as the counter-electrode, and an FTO substrate loaded with the photocatalyst was used as the working electrode.Briefly, the working electrode was prepared as follows: 20 mg of photocatalyst and 20 µL of Nafion ® solution were mixed with 400 µL of absolute ethanol and sonicated for 0.5 h.Then, 20 µL of the above suspension was evenly spread on the FTO to cover an area of 1 cm × 1 cm and dried at room temperature to obtain a working electrode.In addition, 0.5 M Na 2 SO 4 solution was used as the electrolyte in all photoelec-trochemical measurements.The transient photocurrent response was measured with a 300 W Xe lamp (equipped with a 400 nm cutoff filter, CEL-HXF300, Beijing China Education Au-light Co., Ltd., Hangzhou, China) as the light source.Electrochemical impedance spectroscopy (EIS) was performed in the frequency range from 0.01 to 1,000,000 Hz with an amplitude of 5 mV.The Mott-Schottky plot was tested using three frequencies of 1000, 1500, and 2000 Hz.

Results and Discussion
The X-ray diffraction (XRD) patterns of the synthesized samples are presented in Figure 1A.Two peaks at 13 • and 27.3 • were found in all of the samples; these corresponded to the typical (110) and (002) of g-C 3 N 4 , and (100) and (002) were ascribed to the in-plane repeat unit of heptazine and the characteristic interlayer structure of g-C 3 N 4 , respectively [25].Both of the (002) peaks of HCN and x-P-HCN were weaker than those of g-C 3 N 4 , indicating their poor in-plane periodicity, which is in agreement with other carbon nitrides with tubular structures [26].The chemical groups of the catalysts were determined with FT-IR spectra.As shown in Figure 1B, the peak at 808 cm −1 (pink area) of the different catalysts was ascribed to the out-of-plane bending vibration of the heptazine ring, and the peaks around 1408-1638 cm −1 (orange area) corresponded to the stretching vibration of the heptazine-derived repeating unit.Moreover, the peaks around 1240-1320 cm −1 (blue area) were attributed to the stretching vibration of C-N(-C)-C or C-NH-C, and the broad peak at 3000-3500 cm −1 was attributed to terminal uncondensed -NH or -NH2 [27].However, the absorption peaks The chemical groups of the catalysts were determined with FT-IR spectra.As shown in Figure 1B, the peak at 808 cm −1 (pink area) of the different catalysts was ascribed to the outof-plane bending vibration of the heptazine ring, and the peaks around 1408-1638 cm −1 (orange area) corresponded to the stretching vibration of the heptazinederived repeating unit.Moreover, the peaks around 1240-1320 cm −1 (blue area) were attributed to the stretching vibration of C-N(-C)-C or C-NH-C, and the broad peak at 3000-3500 cm −1 was attributed to terminal uncondensed -NH or -NH 2 [27].However, the absorption peaks of P-C or P-N functional groups were not observed, which may have been due to the low doping amount of P [28].The peaks at 400-1300 cm −1 in the Raman spectra (Figure 1C) were attributed to typical heptazine units, and the peaks at 486, 593, 713, 751, 980, and 1256 cm −1 were related to the vibrational modes of CN heterocycles, among which the peaks at 713 and 1256 cm −1 were attributed to the ring breathing mode of s-triazine, and the broad peak at ~1500 cm −1 (green area) was attributed to the D and G bands of the typical graphitic structure [29,30].Notably, the P-doped g-C 3 N 4 obtained with the hydrothermal treatment had the same Raman peaks as those of g-C 3 N 4 , indicating that the hydrothermal pretreatment and incorporation of P did not change the framework of g-C 3 N 4 , which was consistent with the results of the XRD patterns and FT-IR spectra.Furthermore, no new peaks of P species were found because of the low doping amount of P.
The textural properties of different catalysts were revealed by the N 2 adsorptiondesorption isotherms and pore size distributions.As shown in Figure 1D,E and Table S1, the adsorption isotherms of g-C 3 N 4 , HCN, and x-P-HCN were all type II, and the hysteresis loops all belonged to type H3, indicating that the hydrothermal treatment and P doping did not significantly change the mesoporous structure of g-C 3 N 4 .In addition, the BET specific surface area of HCN (7.91 m 2 •g −1 ) obtained after the hydrothermal pretreatment slightly increased compared with that of bulk g-C 3 N 4 (6.81 m 2 •g −1 ).With the increase in NH 4 H 2 PO 4 addition, the specific surface area significantly increased, and the BET specific surface areas of 0.5-P-HCN, 1.0-P-HCN, 1.5-P-HCN, and 2.0-P-HCN reached 9.61, 13.85, 17.55, and 20.01 m 2 •g −1 , respectively.It is worth noting that the specific surface area of g-C 3 N 4 was increased by doping with P, which might have been due to the change in morphology after doping with P. It can be seen from the SEM images (Figures 2 and S1) that a hollow tubular morphology was formed and a large number of nanopores were generated in the P-doped g-C 3 N 4 , and this was able to greatly increase its specific surface area g-C 3 N 4 .This larger surface area can expose more active sites to improve the photocatalytic performance of the reduction of CO 2 into CO.
The CO 2 adsorption capacities of the three catalysts were studied with CO 2 temperatureprogrammed desorption (CO 2 -TPD).The CO 2 adsorption capacity of HCN was higher than that of g-C 3 N 4 , indicating that the hydrothermal pretreatment of the precursor was beneficial for CO 2 adsorption, as shown in Figure 1F.In addition, 1.0-P-HCN exhibited the largest CO 2 adsorption capacity in comparison with those of g-C 3 N 4 and HCN, determining that the CO 2 adsorption of HCN was further enhanced by doping with P, which promoted the photocatalytic reduction of CO 2 into CO [31].
The microstructure and morphology of the photocatalyst are exhibited in SEM images (Figures 2 and S1).g-C 3 N 4 exhibited an irregular structure, while HCN showed a rodlike and nanosheet structure, indicating that the hydrothermal pretreatment of the precursor was able to impact the morphology of g-C 3 N 4 .After doping with P, the x-P-HCN samples displayed a hollow tubular structure, indicating that the morphology of HCN could be regulated by doping with P. The hollow tubular structure was conducive to exposing abundant active sites for CO 2 reduction on the photocatalyst surface and decreased the transfer distance of the photogenerated electrons and holes [32].
TEM and HRTEM images (Figures 3, S2 and S3) were used for a further investigation of the morphology and microstructure of the photocatalyst.An amorphous structure was found in the samples of g-C 3 N 4 , HCN, and 1.0-P-HCN.Notably, the hollow tubular morphology was not observed in the TEM images of 1.0-P-HCN, which might have been because the TEM images were obtained from some local positions of 1.0-P-HCN.In addition, based on the HADDF image and the corresponding EDS elemental mapping of 1.0-P-HCN (Figure 3D), the P was evenly distributed in the catalyst, which confirmed the incorporation of P in g-C 3 N 4 .loops all belonged to type H3, indicating that the hydrothermal treatment and P doping did not significantly change the mesoporous structure of g-C3N4.In addition, the BET specific surface area of HCN (7.91 m 2 •g −1 ) obtained after the hydrothermal pretreatment slightly increased compared with that of bulk g-C3N4 (6.81 m 2 •g −1 ).With the increase in NH4H2PO4 addition, the specific surface area significantly increased, and the BET specific surface areas of 0.5-P-HCN, 1.0-P-HCN, 1.5-P-HCN, and 2.0-P-HCN reached 9.61, 13.85, 17.55, and 20.01 m 2 •g −1 , respectively.It is worth noting that the specific surface area of g-C3N4 was increased by doping with P, which might have been due to the change in morphology after doping with P. It can be seen from the SEM images (Figures 2 and S1) that a hollow tubular morphology was formed and a large number of nanopores were generated in the P-doped g-C3N4, and this was able to greatly increase its specific surface area g-C3N4.This larger surface area can expose more active sites to improve the photocatalytic performance of the reduction of CO2 into CO.The CO2 adsorption capacities of the three catalysts were studied with CO2 temperature-programmed desorption (CO2-TPD).The CO2 adsorption capacity of HCN was higher than that of g-C3N4, indicating that the hydrothermal pretreatment of the precursor was beneficial for CO2 adsorption, as shown in Figure 1F.In addition, 1.0-P-HCN exhibited the largest CO2 adsorption capacity in comparison with those of g-C3N4 and HCN, determining that the CO2 adsorption of HCN was further enhanced by doping with P, which promoted the photocatalytic reduction of CO2 into CO [31].
The microstructure and morphology of the photocatalyst are exhibited in SEM images (Figures 2 and S1).g-C3N4 exhibited an irregular structure, while HCN showed a rodlike and nanosheet structure, indicating that the hydrothermal pretreatment of the precursor was able to impact the morphology of g-C3N4.After doping with P, the x-P-HCN samples displayed a hollow tubular structure, indicating that the morphology of HCN could be regulated by doping with P. The hollow tubular structure was conducive to exposing abundant active sites for CO2 reduction on the photocatalyst surface and decreased the transfer distance of the photogenerated electrons and holes [32].
TEM and HRTEM images (Figures 3, S2, and S3) were used for a further investigation of the morphology and microstructure of the photocatalyst.An amorphous structure was found in the samples of g-C3N4, HCN, and 1.0-P-HCN.Notably, the hollow tubular morphology was not observed in the TEM images of 1.0-P-HCN, which might have been because the TEM images were obtained from some local positions of 1.0-P-HCN.In addition, based on the HADDF image and the corresponding EDS elemental mapping of 1.0-P-HCN (Figure 3D), the P was evenly distributed in the catalyst, which confirmed the incorporation of P in g-C3N4.The surface chemical states of the photocatalyst were revealed with X-ray photoelectron spectroscopy (XPS).As shown in the XPS spectra of C 1s (Figure 4A), the three peaks at 288.5, 286.6, and 284.8 eV were assigned to N-C=N of sp 3 , C-N-C of sp 2 , and C-C of graphite carbon, respectively, and the broad peak at 294.0 eV was attributed to π-π* excitation of interlayer [33].In the N 1s XPS spectra (Figure 4B), three peaks could be observed The surface chemical states of the photocatalyst were revealed with X-ray photoelectron spectroscopy (XPS).As shown in the XPS spectra of C 1s (Figure 4A), the three peaks at 288.5, 286.6, and 284.8 eV were assigned to N-C=N of sp 3 , C-N-C of sp 2 , and C-C of graphite carbon, respectively, and the broad peak at 294.0 eV was attributed to π-π* excitation of interlayer [33].In the N 1s XPS spectra (Figure 4B), three peaks could be observed at 401.6, 400.0, and 399.0 eV, corresponding to C-N-H, sp 2 hybridized N bonded to three atoms (C-N(-C)-C or C-N(-H)-C), and the aromatic N (C-N=C) in the triazine ring, respectively; the broad peak from 403 eV to 406 eV was attributed to the interlayer π-π* excitation [26].Notably, the signal of P was not detected in the XPS spectra of 1.0-P-HCN due to the low doping amount.The light absorption abilities of g-C3N4, HCN, and x-P-HCN were illustrated with UV-Vis diffuse reflection spectra (Figure 4C,D).The light absorption ranges and the corresponding band gaps of g-C3N4 were slightly affected by doping with P. Notably, the 1.0-P-HCN sample exhibited the strongest light absorption ability compared to the other samples, which might have been because the electronic structure of g-C3N4 was changed by doping with moderate P [26].Moreover, the band gaps of g-C3N4, HCN, 0.5-P-HCN, 1.0-P-HCN, 1.5-P-HCN, and 2.0-P-HCN were calculated to be 2.83, 2.78, 2.75, 2.73, 2.78, and 2.80 eV.
The separation and transfer of photogenerated charge carriers in the photocatalyst were revealed with steady-state and time-resolved photoluminescence spectra (Figure 5A,B).Bulk g-C3N4 exhibited a strong fluorescence peak, indicating that it had a high photogenerated electron-hole recombination rate, which is unfavorable for the photocatalytic reduction of CO2 [34].However, the fluorescence peak of HCN was lower than that of bulk g-C3N4, indicating that the hydrothermal treatment had a beneficial effect by allow- The light absorption abilities of g-C 3 N 4 , HCN, and x-P-HCN were illustrated with UV-Vis diffuse reflection spectra (Figure 4C,D).The light absorption ranges and the corresponding band gaps of g-C 3 N 4 were slightly affected by doping with P. Notably, the 1.0-P-HCN sample exhibited the strongest light absorption ability compared to the other samples, which might have been because the electronic structure of g-C 3 N 4 was changed by doping with moderate P [26].Moreover, the band gaps of g-C 3 N 4 , HCN, 0.5-P-HCN, 1.0-P-HCN, 1.5-P-HCN, and 2.0-P-HCN were calculated to be 2.83, 2.78, 2.75, 2.73, 2.78, and 2.80 eV.
The separation and transfer of photogenerated charge carriers in the photocatalyst were revealed with steady-state and time-resolved photoluminescence spectra (Figure 5A,B).Bulk g-C 3 N 4 exhibited a strong fluorescence peak, indicating that it had a high photogenerated electron-hole recombination rate, which is unfavorable for the photocatalytic reduction of CO 2 [34].However, the fluorescence peak of HCN was lower than that of bulk g-C 3 N 4 , indicating that the hydrothermal treatment had a beneficial effect by allowing the avoidance of the recombination of photogenerated electrons and holes.Meanwhile, after doping with P, the fluorescence intensity of g-C 3 N 4 was further decreased compare with that of HCN, indicating that the separation and transfer of the photogenerated charge carrier in HCN could be enhanced by doping with P. Noteworthily, the fluorescence intensity of the 1.0-P-HCN sample was the lowest, demonstrating that the photogenerated electron-hole pairs can be greatly separated after doping with moderate P, which can boost the reaction rate of photocatalytic CO 2 reduction [35].
rescence intensity of the 1.0-P-HCN sample was the lowest, demonstrating that the photogenerated electron-hole pairs can be greatly separated after doping with moderate P, which can boost the reaction rate of photocatalytic CO2 reduction [35].
In addition, the kinetics of photoinduced electrons and holes in the g-C3N4, HCN, and 1.0-P-HCN samples were further investigated with their TRPL spectra, and the fluorescence lifetime was calculated [36].From Figure 5B, the gray, cyan and pink lines are the curves of the original data of g-C3N4, HCN, and 1.0-P-HCN respectively, and the black, blue and red lines are the fitted curves of g-C3N4, HCN, and 1.0-P-HCN respectively.The average photocarrier lifetimes of g-C3N4, HCN, and 1.0-P-HCN were 6.26, 6.90, and 6.73 ns, respectively, indicating that the recombination rates of photoinduced electrons and holes in HCN and 1.0-P-HCN were significantly lower than that in g-C3N4, which further confirmed that the hydrothermal treatment of the precursor and doping with P are available ways to enhance the separation of photogenerated electron-hole pairs [37].However, the average photocarrier lifetime of 1.0-P-HCN was shorter than that of HCN, which might have been because the surface defects were adjusted by P doping to affect the transfer of photogenerated electrons [28].The separation and transfer of photogenerated charge carriers of g-C3N4, HCN, and 1.0-P-HCN were further revealed with photoelectrochemical measurements.As shown by the transient photocurrent densities of g-C3N4, HCN, and 1.0-P-HCN (Figure 5C), the 1.0-In addition, the kinetics of photoinduced electrons and holes in the g-C 3 N 4 , HCN, and 1.0-P-HCN samples were further investigated with their TRPL spectra, and the fluorescence lifetime was calculated [36].From Figure 5B, the gray, cyan and pink lines are the curves of the original data of g-C 3 N 4 , HCN, and 1.0-P-HCN respectively, and the black, blue and red lines are the fitted curves of g-C 3 N 4 , HCN, and 1.0-P-HCN respectively.The average photocarrier lifetimes of g-C 3 N 4 , HCN, and 1.0-P-HCN were 6.26, 6.90, and 6.73 ns, respectively, indicating that the recombination rates of photoinduced electrons and holes in HCN and 1.0-P-HCN were significantly lower than that in g-C 3 N 4 , which further confirmed that the hydrothermal treatment of the precursor and doping with P are available ways to enhance the separation of photogenerated electron-hole pairs [37].However, the average photocarrier lifetime of 1.0-P-HCN was shorter than that of HCN, which might have been because the surface defects were adjusted by P doping to affect the transfer of photogenerated electrons [28].
The separation and transfer of photogenerated charge carriers of g-C 3 N 4 , HCN, and 1.0-P-HCN were further revealed with photoelectrochemical measurements.As shown by the transient photocurrent densities of g-C 3 N 4 , HCN, and 1.0-P-HCN (Figure 5C), the 1.0-P-HCN sample exhibited the highest photocurrent density, showing that the incorporation of P enhanced the separation of photoinduced electron-hole pairs in g-C 3 N 4 , which can enhance the photocatalytic CO 2 reduction performance [38].Furthermore, as shown by the EIS Nyquist plots of g-C 3 N 4 , HCN, and 1.0-P-HCN (Figure 5D), the smallest arc radius of 1.0-P-HCN indicated that it had the smallest charge transfer resistance, which accelerated the transfer of the charge carriers and enhanced the performance of photocatalytic CO 2 reduction [39].
The photocatalytic CO 2 reduction performance of different catalysts under visible light (>400 nm) was investigated.As shown in Figure 6A,B, the CO production rate of HCN was 2.78 µmol•g −1 •h −1 , which was 3.16 times that of g-C 3 N 4 (0.88 µmol•g −1 •h −1 ).Notably, the 1.0-P-HCN sample exhibited the best CO production rate (9.00 µmol•g −1 •h −1 ), which was 3.24 times and 10.23 times those of HCN and g-C 3 N 4 .However, the CO production rates of 1.5-P-HCN and 2.0-P-HCN decreased with the increase in P doping, which might have been because the increased P could be used as a site for the recombination of photogenerated electrons and holes.To show the advantage of 1.0-P-HCN, the photocatalytic CO 2 reduction activity of some reported g-C 3 N 4 and phosphorus-doped g-C 3 N 4 photocatalysts is summarized in Table S2.Compared with the other photocatalysts in Table S2, 1.0-P-HCN showed excellent photocatalytic CO 2 reduction performance without cocatalysts or sacrificial agents under visible light irradiation.This was attributed to the enhanced CO 2 adsorption capacity and enhanced photogenerated carrier separation ability caused by the hollow tubular morphology and phosphorus doping.
The carbon source of the CO product was ascertained by using control experiments.As shown in Figure 6C, without CO 2 , light, or a catalyst, no CO products were detected, indicating that CO was generated through the photocatalytic reduction of CO 2 by 1.0-P-HCN.Notably, when water vapor was not added to the photocatalytic reactor, the rate of photocatalytic CO production was 0.49 µmol•g −1 •h −1 , indicating that water vapor played a vital role in the photocatalytic CO 2 reduction reaction.
The stability of photocatalytic CO 2 reduction with 1.0-P-HCN was investigated through cycling experiments.As shown in Figure 6D, the performance of the photocatalytic reduction of CO 2 into CO did not significantly decrease after three cycles, and the average CO generation rate in the third cycle remained at 90% of that in the first cycle, which indicated that 1.0-P-HCN is stable for the photocatalytic reduction of CO 2 into CO.After three cycles of the reaction, 1.0-P-HCN was characterized via FT-IR, XRD, SEM, TEM, and XPS to clarify the its stability.As shown in Figures S4 and S5, no obvious changes were observed in the XRD patterns and FT-IR spectra of 1.0-P-HCN before and after the reaction, demonstrating that the structure of 1.0-P-HCN did not change after the reaction.In addition, the SEM images (Figure S6), TEM images, and EDS elemental mapping (Figure S7) showed that the morphology of 1.0-P-HCN before and after the reaction was unaltered.Moreover, according to the XPS spectra in Figure S8, the chemical composition and chemical state of 1.0-P-HCN did not change significantly before and after the reaction.The above results indicate that 1.0-P-HCN was stable in the photocatalytic CO 2 reduction reaction.
photocatalytic CO2 reduction activity of some reported g-C3N4 and phosphorusdoped g-C3N4 photocatalysts is summarized in Table S2.Compared with the other photocatalysts in Table S2, 1.0-P-HCN showed excellent photocatalytic CO2 reduction performance without cocatalysts or sacrificial agents under visible light irradiation.This was attributed to the enhanced CO2 adsorption capacity and enhanced photogenerated carrier separation ability caused by the hollow tubular morphology and phosphorus doping.The in situ DRIFTS spectra were used to investigate the reaction pathway and mechanism of photocatalytic CO 2 reduction.As shown in Figure 7A, several peaks could be observed after the co-adsorption of CO 2 and H 2 O vapor onto g-C 3 N 4 , HCN, and 1.0-P-HCN in the dark for 60 min.The four peaks at 1356, 1419, 1497, and 1509 cm −1 corresponded to monodentate carbonate (m-CO 3 2− ) [40][41][42], and the peaks at 1396, 1436, 1457, 1473, 1647, and 1653 cm −1 could be attributed to HCO 3 − [12, [42][43][44].The peak at 1489 cm −1 was attributed to methoxy (CH 3 O) [41], and the peaks at 1521 and 1576 cm −1 were attributable to bidentate carbonate (b-CO 3 2− ) [41,43].In addition, the peak of COOH* (The * represents an adsorbed intermediate species.)was also found at 1541 cm −1 ; COOH* is generally suggested to be a crucial intermediate in the formation of CO [40,44].In addition, HCOO - (1558, 1636, and 1793 cm −1 ) [12,41,45], CO 2 − (1670, 1684, and 1698 cm −1 ) [37,43,44], and chelating-bridged carbonate (c-CO 3 2− , 1715, 1733, 1748, and 1773 cm −1 ) [42,46] can be observed in Figure 7A.Furthermore, the peaks at 1868, 2017, and 2179 cm −1 ascribed to CO* [42,44,46] can be observed in Figure 7B.These carbon species are important intermediates for CO 2 conversion.Notably, the peak intensity of intermediates on 1.0-P-HCN is much stronger than those of g-C 3 N 4 and HCN, indicating that more CO 2 can be adsorbed and activated on the surface of 1.0-P-HCN.The in situ DRIFTS spectra of 1.0-P-HCN with CO2 and H2O vapor under illumination were also recorded to reveal the impact of light on the activation of adsorbed CO2 with 1.0-P-HCN.As shown in Figure 7C,D, the peak intensity of the adsorbed carbon species was greatly enhanced after the light was turned on, indicating that the adsorption for CO2 was increased under light irradiation.In addition, the peak intensity of various CO2 intermediates was increased with the extension of the illumination time.Compared to the in situ DRIFTS spectra in the dark, no new characteristic peaks were detected.However, the ratio of the COOH* peak was greatly enhanced under light irradiation, and the CO* was also increased with the increase in the illumination time.Therefore, the photocatalytic CO2 reduction reaction with 1.0-P-HCN might follow the pathway of CO2→COOH*→CO*→CO.
Mott-Schottky plots were obtained to determine the Fermi level of the photocatalyst.As shown in Figure S9, the positive slope of the Mott-Schottky plots suggests that the photocatalysts were n-type semiconductors.The flat band potentials (Efb) of g-C3N4, HCN, and 1.0-P-HCN were ascertained to be −1.03,−0.97, and −1.08 V (vs.Ag/AgCl, pH = 7).According to Equation (1), where E θ (Ag/AgCl) = 0.197 V at 25 °C [47], the Efb values of g-C3N4, HCN, and 1.0-P-HCN were calculated to be −0.42,−0.36, and −0.47 V (vs.NHE, pH = 0).Theoretically, the value of the Fermi level of the photocatalyst was equal to the flat band potential.Therefore, the Fermi levels of g-C3N4, HCN, and 1.0-P-HCN were considered to be −0.42,−0.36, and −0.47 V (vs.NHE, pH = 0) [12].The in situ DRIFTS spectra of 1.0-P-HCN with CO 2 and H 2 O vapor under illumination were also recorded to reveal the impact of light on the activation of adsorbed CO 2 with 1.0-P-HCN.As shown in Figure 7C,D, the peak intensity of the adsorbed carbon species was greatly enhanced after the light was turned on, indicating that the adsorption capacity for CO 2 was increased under light irradiation.In addition, the peak intensity of various CO 2 intermediates was increased with the extension of the illumination time.Compared to the in situ DRIFTS spectra in the dark, no new characteristic peaks were detected.However, the ratio of the COOH* peak was greatly enhanced under light irradiation, and the CO* was also increased with the increase in the illumination time.Therefore, the photocatalytic CO 2 reduction reaction with 1.0-P-HCN might follow the pathway of CO 2 →COOH*→CO*→CO.
Mott-Schottky plots were obtained to determine the Fermi level of the photocatalyst.As shown in Figure S9, the positive slope of the Mott-Schottky plots suggests that the photocatalysts were n-type semiconductors.The flat band potentials (E fb ) of g-C 3 N 4 , HCN, and 1.0-P-HCN were ascertained to be −1.03,−0.97, and −1.08 V (vs.Ag/AgCl, pH = 7).According to Equation (1), where E θ (Ag/AgCl) = 0.197 V at 25 • C [47], the E fb values of g-C 3 N 4 , HCN, and 1.0-P-HCN were calculated to be −0.42,−0.36, and −0.47 V (vs.NHE, pH = 0).Theoretically, the value of the Fermi level of the photocatalyst was equal to the flat band potential.Therefore, the Fermi levels of g-C 3 N 4 , HCN, and 1.0-P-HCN were considered to be −0.42,−0.36, and −0.47 V (vs.NHE, pH = 0) [12].E(NHE) = E(Ag/AgCl) + E θ (Ag/AgCl) + 0.059pH (1) In addition, the valence band position relative to the Fermi level was measured with the XPS-VB spectra (Figure S10).The valence bands of g-C 3 N 4 , HCN, and 1.0-P-HCN relative to the Fermi level were 2.88, 2.73, and 2.70 eV.Therefore, the valence band positions of g-C 3 N 4 , HCN, and 1.0-P-HCN were calculated to be 2.46, 2.37, and 2.23 V, respectively (vs.NHE, pH = 0).Moreover, as shown in Figure 4D, the band gaps of g-C 3 N 4 , HCN, and 1.0-P-HCN were 2.83, 2.78, and 2.73 eV.Hence, the conduction band positions of g-C 3 N 4 , HCN, and 1.0-P-HCN were calculated to be −0.37,−0.41, and −0.50 V (vs.NHE, pH = 0).The positions of CB were more negative than that of the reduction CO 2 into CO (−0.12 V vs. NHE, pH = 0), indicating that g-C 3 N 4 , HCN, and 1.0-P-HCN were able to reduce CO 2 into CO.The band arrangements of g-C 3 N 4 , HCN, and 1.0-P-HCN are shown in Figure 8A.In Figure 8A, we can find that 1.0-P-HCN had the narrowest band gap, which enhanced the photocatalytic CO 2 reaction under visible light.Moreover, 1.0-P-HCN had the most negative conduction band, which indicated that it had the strongest reducibility among these samples.
In addition, the valence band position relative to the Fermi level was measured with the XPS-VB spectra (Figure S10).The valence bands of g-C3N4, HCN, and 1.0-P-HCN relative to the Fermi level were 2.88, 2.73, and 2.70 eV.Therefore, the valence band positions of g-C3N4, HCN, and 1.0-P-HCN were calculated to be 2.46, 2.37, and 2.23 V, respectively (vs.NHE, pH = 0).Moreover, as shown in Figure 4D, the band gaps of g-C3N4, HCN, and 1.0-P-HCN were 2.83, 2.78, and 2.73 eV.Hence, the conduction band positions of g-C3N4, HCN, and 1.0-P-HCN were calculated to be −0.37,−0.41, and −0.50 V (vs.NHE, pH = 0).The positions of CB were more negative than that of the reduction of CO2 into CO (−0.12 V vs. NHE, pH = 0), indicating that g-C3N4, HCN, and 1.0-P-HCN were able to reduce CO2 into CO.The band arrangements of g-C3N4, HCN, and 1.0-P-HCN are shown in Figure 8A.In Figure 8A, we can find that 1.0-P-HCN had the narrowest band gap, which enhanced the photocatalytic CO2 reaction under visible light.Moreover, 1.0-P-HCN had the most negative conduction band, which indicated that it had the strongest reducibility among these samples.According to the above results of the photocatalytic CO2 reduction test, characterization, and in situ DRIFTS, a mechanism of photocatalytic CO2 reduction with 1.0-P-HCN is proposed (Figure 8B).Under visible light illumination, photogenerated electrons were excited and transitioned from VB to CB of 1.0-P-HCN and were then transferred to the surface of 1.0-P-HCN for the photocatalytic CO2 reduction reaction.The CO2 was first adsorbed and activated on the surface of 1.0-P-HCN and then converted into the adsorbed carbon intermediates; then, the generation of CO followed the pathway of CO2→COOH*→CO*→CO.Furthermore, the P doping of g-C3N4 enhanced its light absorption properties, reduced its charge transfer resistance, improved its CO2 adsorption capacity, and improved the separation efficiency of the photogenerated electron-hole pairs, all of which is conducive to boosting photocatalytic CO2 reduction.

Conclusions
In summary, NH4H2PO4 was used as a phosphorus source in the hydrothermal treatment of melamine for the first time, and phosphorus-doped hollow tubular g-C3N4 (x-P-HCN) photocatalysts were finally prepared for photocatalytic CO2 reduction.x-P-HCN presented much higher photocatalytic CO2 reduction performance than that of g-C3N4.
The production rate of CO with 1.0-P-HCN reached 9.00 µmol•g −1 •h −1 , which was 10.22 times that of bulk g-C3N4.Moreover, the enhanced performance of 1.0-P-HCN might have been due to the increased specific surface area, excellent CO2 adsorption ability, enhanced light absorption capacity, lower charge transfer resistance, and more efficient separation According to the above results of the photocatalytic CO 2 reduction test, characterization, and in situ DRIFTS, a mechanism of photocatalytic CO 2 reduction with 1.0-P-HCN is proposed (Figure 8B).Under visible light illumination, photogenerated electrons were excited and transitioned from VB to CB of 1.0-P-HCN and were then transferred to the surface of 1.0-P-HCN for the photocatalytic CO 2 reduction reaction.The CO 2 was first adsorbed and activated on the surface of 1.0-P-HCN and then converted into the adsorbed carbon intermediates; then, the generation of CO followed the pathway of CO 2 →COOH*→CO*→CO.Furthermore, the P doping of g-C 3 N 4 enhanced its light absorption properties, reduced its charge transfer resistance, improved its CO 2 adsorption capacity, and improved the separation efficiency of the photogenerated electron-hole pairs, all of which is conducive to boosting photocatalytic CO 2 reduction.

Conclusions
In summary, NH 4 H 2 PO 4 was used as a phosphorus source in the hydrothermal treatment of melamine for the first time, and phosphorus-doped hollow tubular g-C 3 N 4 (x-P-HCN) photocatalysts were finally prepared for photocatalytic CO 2 reduction.x-P-HCN presented much higher photocatalytic CO 2 reduction performance than that of g-C 3 N 4 .The production rate of CO with 1.0-P-HCN reached 9.00 µmol•g −1 •h −1 , which was 10.22 times that of bulk g-C 3 N 4 .Moreover, the enhanced performance of 1.0-P-HCN might have been due to the increased specific surface area, excellent CO 2 adsorption ability, enhanced light absorption capacity, lower charge transfer resistance, and more efficient separation of photogenerated electrons and holes of 1.0-P-HCN, which significantly boosted its photocatalytic CO 2 reduction.This work provides an in-depth research perspective for accelerating the photocatalytic CO 2 reduction rate of g-C 3 N 4 .

Figure 6 .
Figure 6.Yield (A) and production rate (B) of the photocatalytic reduction of CO 2 into CO for g-C 3 N 4 , HCN, and x-P-HCN.The CO production rate of photocatalytic CO 2 reduction by 1.0-P-HCN under different conditions (C).Cyclic test of photocatalytic reduction of CO 2 into CO by 1.0-P-HCN (D).

Figure 7 .
Figure 7.In situ DRIFTS spectra of g-C3N4, HCN, and 1.0-P-HCN with CO2 and H2O vapor for 60 min in the dark (A,B).In situ DRIFTS spectra of 1.0-P-HCN with CO2 and H2O vapor under illumination (C,D).

Figure 7 .
Figure 7.In situ DRIFTS spectra of g-C 3 N 4 , HCN, and 1.0-P-HCN with CO 2 and H 2 O vapor for 60 min in the dark (A,B).In situ DRIFTS spectra of 1.0-P-HCN with CO 2 and H 2 O vapor under illumination (C,D).