Photocatalytic Reduction of Cr (VI) over g-C3N4 Photocatalysts Synthesized by Different Precursors

Graphitic carbon nitride (g-C3N4) photocatalysts were synthesized via a one-step pyrolysis process using melamine, dicyandiamide, thiourea, and urea as precursors. The obtained g-C3N4 materials exhibited a significantly different performance for the photocatalytic reduction of Cr(VI) under white light irradiation, which is attributed to the altered structure and occupancies surface groups. The urea-derived g-C3N4 with nanosheet morphology, large specific surface area, and high occupancies of surface amine groups exhibited superior photocatalytic activity. The nanosheet morphology and large surface area facilitated the separation and transmission of charge, while the high occupancies of surface amine groups promoted the formation of hydrogen adsorption atomic centers which were beneficial to Cr(VI) reduction. Moreover, the possible reduction pathway of Cr(VI) to Cr(III) over the urea-derived g-C3N4 was proposed and the reduction process was mainly initiated by a direct reduction of photogenerated electrons.


Introduction
In recent years, with the rapid development of mining, electroplating, leather tanneries, and pigments, a large number of heavy metal compounds have been discharged. Among of them, hexavalent chromium Cr(VI) is one of the most virulent contaminants which can be accumulated by digestion system via the exposure and intake the polluted substance, causing serious illness such as cancer and skin allergy [1][2][3]. Because of its high toxicity, Cr(VI) has been ranked among the top 20 toxic pollutants on Superfund Priority List of Hazardous Substances [4]. Photocatalytic reduction technology is an attractive alternative technology for Cr(VI) reduction because of its acceptable cost, easy operation, and high safety. Specifically, it can directly reduce high toxic Cr(VI) to less harmful Cr(III), which is a necessary trace element for human being and easy to precipitation in aqueous solution (K θ sp (Cr(OH) 3 ) = 6.3 × 10 −31 ) [5][6][7]. As a metal-free semiconductor photocatalyst, graphitic carbon nitride (g-C 3 N 4 ) has been widely studied due to its visible-light-driven, narrow bandgap, non-toxic, low cost, and excellent stability [8,9]. In addition, the conduction band of g-C 3 N 4 level is much more negative than that of Cr(VI)/Cr(III) (1.3 eV vs. NHE), suggesting that the photo-generated electrons in g-C 3 N 4 possess a large thermodynamic driving force to reduce Cr(VI) to Cr(III) [10,11]. In general, g-C 3 N 4 can be synthesized via thermolysis method from cheap available nitrogen-rich precursors such as urea, thiourea, cyanamide, dicyandiamide, and melamine. Due to the different chemical structures of these nitrogen-rich precursors, as well as the influence of foreign chemical elements such as O and S, the g-C 3 N 4 photocatalysts prepared from these nitrogen-rich precursors exhibit different structural and optical properties, and thus display different photocatalytic activities [12][13][14].

Structure and Morphology
Figure 1a shows X-ray diffraction (XRD) patterns of g-C 3 N 4 prepared by different precursor systems. It can be seen from the XRD pattern that the g-C 3 N 4 obtained by the four precursors all have two characteristic peaks. The weak peak near 12.8 • is attributed to intralayer long-range atomic order (100), which is associated with the hydrogen bonds in g-C 3 N 4 [27,28]. The stronger peak near 27.4 • corresponded to the (002) crystal plane of g-C 3 N 4 , which is caused by the interlayer accumulation of the conjugated aromatic system [29]. These characteristic diffraction peaks are consistent with previous reports. In addition, the diffraction peaks of U-CN are broader and lower intensity, indicating that U-CN has the lowest crystallinity. The low crystallinity of U-CN might be that extra O in urea produces H 2 O, CO 2 , and ammonia during calcination, which inhibit the growth of the surface crystal [30]. Furthermore, the (100) crystal plane of U-CN is substantially lower than the other three samples, indicating a reduced hydrogen bond effect in the intralayer of U-CN [20]. Figure 1b presents the FTIR spectra of the g-C 3 N 4 samples to demonstrate their graphitic structures. The absorption peaks observed between 1200-1700 cm −1 correspond to the characteristic breathing modes of aromatic carbon nitride heterocyclic rings [31]. The sharp absorption band at around 801 cm −1 is attributed to the respiratory pattern of triazine units while the board vibration bands at 3000-3500 cm −1 can be ascribed to the uncondensed amine groups and the water molecules adsorbed on the surface [14]. In addition, it can be seen from Figure 1c that U-CN blue-shifted at 801 cm −1 . The main reason might be due to the hydrogen bond-containing which influences the triazine ring stretching in the g-C 3 N 4 structure [16].
condensed amine groups and the water molecules adsorbed on the surface [14]. In addition, it can be seen from Figure 1c that U-CN blue-shifted at 801 cm −1 . The main reason might be due to the hydrogen bond-containing which influences the triazine ring stretching in the g-C3N4 structure [16].
The lower degree of polymerization of U-CN can also be obtained from thermal stability of the different as-prepared g-C3N4. It can be seen from Figure 1d that there is no further significant weight loss up to a temperature of 400 °C. However, weight losses of 21.1%, 29.0%, 33.3%, and 53.5% were observed for the M-CN, T-CN, D-CN, and U-CN between 400 °C to 650 °C, respectively. U-CN can be completely decomposed at 700 °C, while the completely decomposition temperature of T-CN, D-CN, and M-CN are 728 °C, 732 °C, and 746 °C, respectively. The result suggest that U-CN has the worst thermally stability, which might be due to its low degree of polymerization and poor stability of the triazine ring structure. The morphologies of the prepared samples were investigated by SEM. D-CN and M-CN show the typical flat and layered structure with small lamellas wrapped in large particles ( Figure 2a,b), whereas T-CN has an obvious layered structure, with large lamellae and a few fine particles scattered on the surface (Figure 2c). U-CN displays nanosheets morphology with irregular wrinkles (Figure 2d) [32]. It can also be observed that U-CN displays porous structure while the last three sample present large sheet without porous structures. The lower degree of polymerization of U-CN can also be obtained from thermal stability of the different as-prepared g-C 3 N 4 . It can be seen from Figure 1d that there is no further significant weight loss up to a temperature of 400 • C. However, weight losses of 21.1%, 29.0%, 33.3%, and 53.5% were observed for the M-CN, T-CN, D-CN, and U-CN between 400 • C to 650 • C, respectively. U-CN can be completely decomposed at 700 • C, while the completely decomposition temperature of T-CN, D-CN, and M-CN are 728 • C, 732 • C, and 746 • C, respectively. The result suggest that U-CN has the worst thermally stability, which might be due to its low degree of polymerization and poor stability of the triazine ring structure.
The morphologies of the prepared samples were investigated by SEM. D-CN and M-CN show the typical flat and layered structure with small lamellas wrapped in large particles (Figure 2a,b), whereas T-CN has an obvious layered structure, with large lamellae and a few fine particles scattered on the surface (Figure 2c). U-CN displays nanosheets morphology with irregular wrinkles (Figure 2d) [32]. It can also be observed that U-CN displays porous structure while the last three sample present large sheet without porous structures.
The surface area and porous structure of the prepared samples were further studied on the basis of nitrogen gas adsorption-desorption isotherms and pore size distribution curves. As shown in Figure 3, all the samples exhibit a classical type IV isotherms, which is the characteristic of the typical mesoporous materials. The BET surface areas and pore volume of the as-prepared samples were summarized in Table 1. It can be seen that U-g-C 3 N 4 has the largest specific surface area (S BET = 81.060 m 2 g −1 ) and pore volume (V meso = 0.164 cc g −1 ), while the BET surface areas and pore volume of D-CN (S BET = 8.779 m 2 g −1 , V meso = 0.025 cc g −1 ), M-CN (S BET = 8.363 m 2 g −1 , V meso = 0.021 cc g −1 ), and T-CN (S BET = 7.262 m 2 g −1 , V meso = 0.019 cc g −1 ) have little difference. These results are in good agreement with the morphologies of the samples. The presence of the oxygen heteroatom in urea might also play an important role in increasing the BET surface area and pore volume of the U-CN sample. The emission of pyrolysis-generated gases during the thermal condensation process can function as soft templates and promote the formation of porous structure [33,34]. The external surface area is very important in photocatalytic action as the reactions take place mainly on the external surface that is exposed to light irradiation. The surface area and porous structure of the prepared samples were further studied on the basis of nitrogen gas adsorption-desorption isotherms and pore size distribution curves. As shown in Figure 3, all the samples exhibit a classical type IV isotherms, which is the characteristic of the typical mesoporous materials. The BET surface areas and pore volume of the as-prepared samples were summarized in Table 1. It can be seen that U-g-C3N4 has the largest specific surface area (SBET = 81.060 m 2 g −1 ) and pore volume (Vmeso = 0.164 cc g −1 ), while the BET surface areas and pore volume of D-CN (SBET = 8.779 m 2 g −1 , Vmeso = 0.025 cc g −1 ), M-CN (SBET = 8.363 m 2 g −1 , Vmeso = 0.021 cc g −1 ), and T-CN (SBET = 7.262 m 2 g −1 , Vmeso = 0.019 cc g −1 ) have little difference. These results are in good agreement with the morphologies of the samples. The presence of the oxygen heteroatom in urea might also play an important role in increasing the BET surface area and pore volume of the U-CN sample. The emission of pyrolysis-generated gases during the thermal condensation process can function as soft templates and promote the formation of porous structure [33,34]. The external surface area is very important in photocatalytic action as the reactions take place mainly on the external surface that is exposed to light irradiation.  The XPS spectra of the samples are presented in Figure 4. The survey scan XPS spectra shown in Figure 4a illustrates that the obtained g-C3N4 samples are composed of C, N, and O elements. It can also be seen from Figure 4 that C1s, N1s, and O1s have no obvious energy shifts while core electrons occur, indicating the chemical states of the three elements are the same in D-CN, M-CN, T-CN, and U-CN. As shown in Figure 4b, the highresolution XPS spectra of the C 1s for the samples can be deconvoluted into three peaks with binding energies of 284.67 eV, 285.45 eV, and 288.22 eV, which are characteristic of  The XPS spectra of the samples are presented in Figure 4. The survey scan XPS spectra shown in Figure 4a illustrates that the obtained g-C 3 N 4 samples are composed of C, N, and O elements. It can also be seen from Figure 4 that C1s, N1s, and O1s have no obvious energy shifts while core electrons occur, indicating the chemical states of the three elements are the same in D-CN, M-CN, T-CN, and U-CN. As shown in Figure 4b, the high-resolution XPS spectra of the C 1s for the samples can be deconvoluted into three peaks with binding energies of 284.67 eV, 285.45 eV, and 288.22 eV, which are characteristic of the sp 2 C-C bond, C-O bond, and N-C=N bond, respectively [35]. The high-definition N 1s spectra of the samples can be fitted into three distinct peaks at 398.74 eV, 399.59 eV, and 401.12 eV, corresponding to C-N=C, tertiary nitrogen N-(C) 3 and -NH 2 ( Figure 4c) [36]. In addition, a weak energy peak at 532.54 eV can also be observed in the high resolution XPS spectra of O1s, which can be attributed to adsorbed H 2 O on the sample surface [37] (Figure 4d). The XPS spectra of the samples are presented in Figure 4. The survey scan XPS spectra shown in Figure 4a illustrates that the obtained g-C3N4 samples are composed of C, N, and O elements. It can also be seen from Figure 4 that C1s, N1s, and O1s have no obvious energy shifts while core electrons occur, indicating the chemical states of the three elements are the same in D-CN, M-CN, T-CN, and U-CN. As shown in Figure 4b, the highresolution XPS spectra of the C 1s for the samples can be deconvoluted into three peaks with binding energies of 284.67 eV, 285.45 eV, and 288.22 eV, which are characteristic of the sp 2 C-C bond, C-O bond, and N-C=N bond, respectively [35]. The high-definition N 1s spectra of the samples can be fitted into three distinct peaks at 398.74 eV, 399.59 eV, and 401.12 eV, corresponding to C-N=C, tertiary nitrogen N-(C)3 and -NH2 (Figure 4c) [36]. In addition, a weak energy peak at 532.54 eV can also be observed in the high resolution XPS spectra of O1s, which can be attributed to adsorbed H2O on the sample surface [37] (Figure  4d).  The surface elements content of the samples can be analyzed from the integrated peak areas under C1s, N 1s, and O 1s ( Table 1). The atomic ratios of C/N of D-CN, M-CN, T-CN and UCN are determined to be 0.82, 0.81, 0.86, and 0.72, respectively. The C/N ratio of U-CN is the lowest, implying a more defective structure and lower polymerization. The zeta potentials of the samples were measured to investigate the surface charges of the samples. In the suspension with the initial pH, the zeta potentials of D-CN, MCN, TCN, The surface elements content of the samples can be analyzed from the integrated peak areas under C1s, N 1s, and O 1s ( Table 1). The atomic ratios of C/N of D-CN, M-CN, T-CN and UCN are determined to be 0.82, 0.81, 0.86, and 0.72, respectively. The C/N ratio of U-CN is the lowest, implying a more defective structure and lower polymerization. The zeta potentials of the samples were measured to investigate the surface charges of the samples. In the suspension with the initial pH, the zeta potentials of D-CN, MCN, TCN, and U-CN are −21.45 eV, −21.75 eV, −23.31 eV, and −15.13 eV, respectively. g-C 3 N 4 contains abundant Lewis acid and base sites, which are derived from the terminal and bridging NH-groups and lone pairs of N in triazine/heptazine rings, respectively. Amine groups can act as proton acceptors and acquire positive surface charges. Additionally, hydroxyl ions can react with primary and secondary amine groups to produce negative charges on the surface of g-C 3 N 4 . The surface charges and zeta potentials of g-C 3 N 4 are determined by the number of amine groups on the carbon surface and the pH value of the suspension [38]. The amine groups on the U-CN surface are substantially higher than those on the D-CN, M-CN, and T-CN because of the huge specific surface area and much higher surface N element concentration, resulting in a much stronger ability to adsorb hydrogen ions in the solution with the same pH value. Furthermore, Lewis acid and base sites on the surface of g-C 3 N 4 are potential anchoring sites for cocatalysts [39]. A higher occupation of surface amine group may provide more active sites for the photocatalytic reaction.

Photocatalytic Performance
The photocatalytic reduction performance of Cr(VI) over the as-prepared photocatalysts were evaluated under white light irradiation. As shown in Figure 5a Combined with the previous morphological structure and surface characteristics analyses, it was found that the photocatalytic reduction effect of the catalyst has a certain correspondence with its specific surface area. As shown in Figure 5c, a close correlation was found between the C/C 0 and S BET with a correlation coefficient 0.7967. U-CN with nanosheet morphology, large specific surface area has the best photocatalytic reduction performance. The large surface area facilitates the separation and transfer of photoinduced charges in U-CN, as well as provide more active sites for Cr(VI) reduction. In addition, it is worth noting that, despite having a similar surface area as D-CN and M-CN, T-CN had the lowest photocatalytic reduction activity. By analyzing the surface characteristics of the three samples, it is found that the surface amine group content of the samples has a significant impact on the photocatalytic activity. The surface amine groups can act as exciton dissociation traps which are conducive to the rapid splitting of photogenerated excitons and promote the formation of hydrogen adsorption atomic centers, thus facilitating the photocatalytic reduction of Cr(VI) [39]. The effect of S BET is excluded while investigating the possible role of surface amino groups in the photocatalytic reduction process. Figure 4d shows the association between normalized C/C 0 ((C/C 0 )/S BET ) and zeta potential. As illustrated in Figure 4d, the value of normalized C/C 0 is closely correlation with the zeta potential which is determined by the varied specific area and the occupation of the surface amine groups [38]. The correlation coefficient is conculcated to be 0.8686. The results reveal that the photocatalytic reduction activity of Cr(VI) is affected by the specific surface area and the amount of surface amine groups, and the amount of surface amine groups has a greater impact. tion traps which are conducive to the rapid splitting of photogenerated excitons and promote the formation of hydrogen adsorption atomic centers, thus facilitating the photocatalytic reduction of Cr(VI) [39]. The effect of SBET is excluded while investigating the possible role of surface amino groups in the photocatalytic reduction process. Figure 4d shows the association between normalized C/C0 ((C/C0)/SBET) and zeta potential. As illustrated in Figure 4d, the value of normalized C/C0 is closely correlation with the zeta potential which is determined by the varied specific area and the occupation of the surface amine groups [38]. The correlation coefficient is conculcated to be 0.8686. The results reveal that the photocatalytic reduction activity of Cr(VI) is affected by the specific surface area and the amount of surface amine groups, and the amount of surface amine groups has a greater impact.   [40]. The surface of the catalyst becomes highly protonated at low pH value, which makes the surface of the catalytic more conducive to the accumulation of HCrO 4 − . While the surface of g-C 3 N 4 is negatively charged at alkaline solution, which tends to repel the Cr 2 O 7 2− . The Cr (VI) photoreduction was achieved following Equations (1) and (2) under acidic solution, and the hydrogen ion was beneficial to the reduction reaction. In contrast, the Cr(VI) reduction under alkaline solution was accomplished following Equation (3) [41]. In addition, Cr(OH) 3 may be formed under high pH value and covers the active sites of the photocatalyst, leading to the declining Cr(VI) reduction performance [42].
reduction rate of Cr(VI) increases dramatically as the mount of U-CN catalyst is increased. The improvement in the photocatalytic degradation rate of Cr(VI) declined as the catalyst dosage was increased further. The reason might be that the effect of photocatalytic reaction is related to the catalyst surface's reaction sites [26]. The catalyst's reaction sites can be effectively increased by increasing the catalyst dosage. While the surface active sites meet the need for Cr(VI) reduction as the dose is increased, the influence of the catalyst dosage on the total reaction rate of Cr(VI) reduction is lowered. Furthermore, the Influencing factors on Cr(VI) photocatalytic reduction including light sources, hole scavengers, initial Cr(VI) concentration are presented in Supplementary Material (Figures S2-S4).  The effect of U-CN dosage was also tested. As shown in Figure 6b, the photocatalytic reduction rate of Cr(VI) increases dramatically as the mount of U-CN catalyst is increased. The improvement in the photocatalytic degradation rate of Cr(VI) declined as the catalyst dosage was increased further. The reason might be that the effect of photocatalytic reaction is related to the catalyst surface's reaction sites [26]. The catalyst's reaction sites can be effectively increased by increasing the catalyst dosage. While the surface active sites meet the need for Cr(VI) reduction as the dose is increased, the influence of the catalyst dosage on the total reaction rate of Cr(VI) reduction is lowered.
Furthermore, the Influencing factors on Cr(VI) photocatalytic reduction including light sources, hole scavengers, initial Cr(VI) concentration are presented in Supplementary  Material (Figures S2-S4). Figure 7a shows the UV-visible absorption spectra of the samples. T-CN exhibits stronger light absorption in the range of 400 to 700 nm than that of the other three samples. While the absorption edge of U-CN showed a significant blue shift with respect to the other three samples [16,21]. The band-gap energies of the samples are calculated by plots of (αhv)1/2 versus photo energy. As depicted in Figure 7b, the bandgaps of T-CN, M-CN, D-CN and U-CN are estimated to be 2.68, 2.70, 2.72 and 2.84 eV, correspondingly. U-CN has a larger band gap might be attributed to the quantum size effect caused by smaller and disordered crystalline domains [33,34]. The band structures of the samples were further characterized by valence band XPS(VB-XPS). As shown in Figure 7c [43]. Ac cording to the results of UV-DRS spectra and CB XPS, the VB potential of D-CN, M-CN T-CN and U-CN are conculcated to be −0.50, −0.48, −0.56, −0.36 eV, respectively. Due t the most negative CB potential in UCN, the photogenerated electrons produced by U-CN have the most reducing ability for Cr(VI) reduction in comparison to those in the othe three samples. The transfer behavior and separation efficiency of photogenerated charge carriers i the samples can be reflected by the photoluminescence (PL) spectra. As shown in Figur 7d, the emission peak centers of the four samples are around 440 nm, which represente the irradiative recombination of e − and h + [21]. The emission intensity is lower for the U CN as compared to the other three samples, indicating the recombination rate of electron and holes under white light irradiation is lower in U-CN. To further investigate the sepa ration efficiency of photogenerated charges during the photoreactions, photoelectrochem ical measurements were performed. As displayed in Figure 8a, fast photocurrent re The transfer behavior and separation efficiency of photogenerated charge carriers in the samples can be reflected by the photoluminescence (PL) spectra. As shown in Figure 7d, the emission peak centers of the four samples are around 440 nm, which represented the irradiative recombination of e − and h + [21]. The emission intensity is lower for the U-CN as compared to the other three samples, indicating the recombination rate of electrons and holes under white light irradiation is lower in U-CN. To further investigate the separation efficiency of photogenerated charges during the photoreactions, photoelectrochemical measurements were performed. As displayed in Figure 8a, fast photocurrent responses via on-off cycles were observed for all the samples, while the photocurrent intensity of U-CN was obviously higher than that of the other three samples. In addition, the arc radium of U-CN in the EIS Nyquist plot shown in Figure 8b was also smaller than that of the other three samples. It was clear to see that the diameter of arc radius followed in the order of D-CN < M-CN < T-CN < U-CN, which agrees well with the PL spectra.

Photoelectrochemical Properties
As mentioned above, the photoelectric properties have little effect on the photocatalytic reduction of Cr(VI). Although having the widest bandgap, U-CN displays the best photocatalytic reduction performance. T-CN has the highest white light absorption and much higher charge separation efficiency, but it displays the lowest photocatalytic reduction activity.
sponses via on-off cycles were observed for all the samples, while the photocurrent in sity of U-CN was obviously higher than that of the other three samples. In addition arc radium of U-CN in the EIS Nyquist plot shown in Figure 8b was also smaller than of the other three samples. It was clear to see that the diameter of arc radius follow the order of D-CN < M-CN < T-CN < U-CN, which agrees well with the PL spectra. As mentioned above, the photoelectric properties have little effect on the photo lytic reduction of Cr(VI). Although having the widest bandgap, U-CN displays the photocatalytic reduction performance. T-CN has the highest white light absorption much higher charge separation efficiency, but it displays the lowest photocatalytic re tion activity.

Possible Reaction Mechanism
Previous studies have shown that the photocatalytic reduction of Cr (VI) on the face of g-C3N4 occurs via a direct or indirect reduction of photogenerated electrons trapping of photogenerated electrons by O2 is critical for Cr (VI) reduction in the redu process [3,29,44,45]. To investigate the role of O2 on the photocatalytic reduction of Cr over U-CN in the presence of citric acid, the comparison experiments were carried o different gas atmospheres. It is found from Figure 9a that the photocatalytic reductio Cr(VI) over U-CN in the N2 atmosphere was obviously enhanced, whereas the photo lytic reduction of Cr(VI) over U-CN in O2 atmosphere was depressed to some extent results indicate that O2 is involved in the photocatalytic reduction of Cr (VI) over U ESR technique was further employed to measure the reactive species generated du photocatalysis. As shown in Figure 9b, signals of DMPO-•O2 − could be detected in m anolic suspension of U-CN under white light irradiation, reflecting that •O2 − is gener via electron transfer from conduction band of U-CN to the dissolved molecular ox under white light illumination. Thus, it can be concluded that in the photocatalytic re tion system, Cr (VI) is direct reduced by photogenerated electrons over U-CN, wh O2 in the solution competes with Cr(VI) for the photogenerated electrons, inhib Cr(VI) reduction. While citric acid acts as sacrificial agent of reactive oxygen species i system, avoiding the Cr(III) re-oxidation, while the surplus electrons will participa the reaction of Cr(VI) reduction or trapped by O2 to form •O2 − (Figure 10) [46,47].

Possible Reaction Mechanism
Previous studies have shown that the photocatalytic reduction of Cr (VI) on the surface of g-C 3 N 4 occurs via a direct or indirect reduction of photogenerated electrons. The trapping of photogenerated electrons by O 2 is critical for Cr (VI) reduction in the reduction process [3,29,44,45]. To investigate the role of O 2 on the photocatalytic reduction of Cr (VI) over U-CN in the presence of citric acid, the comparison experiments were carried out in different gas atmospheres. It is found from Figure 9a that the photocatalytic reduction of Cr(VI) over U-CN in the N 2 atmosphere was obviously enhanced, whereas the photocatalytic reduction of Cr(VI) over U-CN in O 2 atmosphere was depressed to some extent. The results indicate that O 2 is involved in the photocatalytic reduction of Cr (VI) over U-CN. ESR technique was further employed to measure the reactive species generated during photocatalysis. As shown in Figure 9b

Preparation of Graphitic Carbon Nitride (g-C3N4)
The g-C3N4 samples were synthesized by a thermal polymerization method using dicyandiamide, melamine, thiourea, and urea as precursors. Typically, 10 g of the precursor powder was put into an alumina crucible and heated at 5 °C/min up to 550 °C for 3 h in a covered muffle furnace. After the crucible cooling to room temperature, four products were collected and ground into powder. The obtained products were named as D-CN, M-CN, T-CN, and U-CN, respectively ( Figure S1, Supplementary Materials).

Photocatalytic Experiments
The photocatalytic activity of the synthesized samples for the removal of Cr(VI) was evaluated under white light provided by a Xe light with the power of 300 W (Perfectlight). At room temperature a certain amount of g-C3N4 was suspended into 150 mL aqueous solution of Cr(VI) (50 mg/L) with the addition of citric acid (0.9 mM). The solution pH was adjusted to 3 by H2SO4 (1 M) or NaOH (1 M). Prior to irradiation, the suspension was

Preparation of Graphitic Carbon Nitride (g-C 3 N 4 )
The g-C 3 N 4 samples were synthesized by a thermal polymerization method using dicyandiamide, melamine, thiourea, and urea as precursors. Typically, 10 g of the precursor powder was put into an alumina crucible and heated at 5 • C/min up to 550 • C for 3 h in a covered muffle furnace. After the crucible cooling to room temperature, four products were collected and ground into powder. The obtained products were named as D-CN, M-CN, T-CN, and U-CN, respectively ( Figure S1, Supplementary Materials).

Photocatalytic Experiments
The photocatalytic activity of the synthesized samples for the removal of Cr(VI) was evaluated under white light provided by a Xe light with the power of 300 W (Perfectlight). At room temperature a certain amount of g-C 3 N 4 was suspended into 150 mL aqueous solution of Cr(VI) (50 mg/L) with the addition of citric acid (0.9 mM). The solution pH was adjusted to 3 by H 2 SO 4 (1 M) or NaOH (1 M). Prior to irradiation, the suspension was ultrasonicated for 10 min and stirred in the dark for 30 min to establish the adsorptiondesorption equilibrium. Subsequently, the light source was switched on. During the illumination process, a certain amount of the solution was taken at a predetermined time interval and filtered through a 0.45 µm filter. The Cr(VI) concentration in the supernatant was determined at 540 nm using the diphenylcarbazide (DPC) method by a spectrophotometer at the characteristic wavelength of 540 nm.

Characterization
X-ray diffraction (XRD) patterns were examined using X-ray diffractometer (Panalytical X'Pert-pro MPD, Almelo, The Netherlands) with Cu Kα radiation source (λ = 1.54056 Å) in the range of 10 • -60 • . The Fourier transform infrared spectra were measured using an infrared spectrometer (Nicolet iS 10, Madison, WI, USA) with KBr pallets. The morphology of the sample was studied using a scanning electron microscope (Hitachi S-4700, Tokyo, Japan). Thermogravimetry (TG) of the samples was analyzed by TA synchronous thermal analyzer ( TA Q600, New Castle, USA). The surface area was calculated using the multi-point BET (Quantachrome NOVA 2000e, Boynton Beach, USA) method. X-ray photo-electron spectroscopy (XPS) was performed using an X-ray photoelectron (ThermFischer ESCALAB 250Xi, Waltham, MA, USA) spectrometer with Al monochrome (hv = 1486.6 Ev) as the X-ray source. The zeta potential of the prepared sample was measured at 293 K using a zeta potential analyzer (Malvern, Zetasizer Nano ZS90, Worcestershire, UK). The UV-vis diffuse reflectance spectra (UV-DRS) were measured on a UV-vis spectrometer (Shimadzu UV-2600, Kyoto, Japan). The photoluminescence (PL) spectra of the samples were recorded by using a fluorescence spectrophotometer (Hitachi F-7100, Tokyo, Japan). The ESR spectra were recorded on a ESR spectrometer (Bruker EMX plus X-band CW, Rheinstetten, Germany) using DMPO as a spin trap agent at room temperature.
The photoelectrochemical measurements were applied in a standard three-electrode cell with an electrochemical workstation (CHI770E, CHN). A Pt and Ag/AgCl (saturated KCl) electrode were chosen as counter and reference electrode, respectively. ITO coated with the prepared catalyst served as the working electrode. The working electrodes were prepared as follows: 5 mg of the as-obtained photocatalyst was suspended in a mixed solution (10 µL of 5% nafion and 2 mL of ethanol) with ultrasound. After that, 100 µL of the obtained suspension was drop-coated on the ITO glass (10 mm × 10 mm), and dried in the air to completely eliminate water. Additionally, 0.1 M Na 2 SO 4 was used as electrolyte.

Conclusions
In this study, g-C 3 N 4 composites are synthesized via a facile polymerized method with four different precursors (i.e., melamine, dicyandiamide, thiourea, and urea). It was found that the type of precursors has a significant impact on the morphology and structure of g-C 3 N 4 and further affects the performance of photocatalytic reduction of Cr(VI). Urea-derived U-CN with nanosheet morphology, large specific surface area, and high occupancies of surface amine groups exhibit superior photocatalytic activity. These results demonstrate that large surface area and high surface amine groups can provide more catalytically active sites. This work confirms the effect of surface properties on the photocatalytic activity of g-C 3 N 4 and provides a theoretical and technical foundation for the construction of practical and high-efficiency photocatalysts based on g-C 3 N 4 .