Enhancing Light-Driven Production of Hydrogen Peroxide by Anchoring Au onto C3N4 Catalysts

Light-driven production of hydrogen peroxide (H2O2) is a green and sustainable way to achieve solar-to-chemical energy conversion. During such a conversion, both the high activity and the stability of catalysts were critical. We prepared an Au-supported C3N4 catalyst—i.e., Au/C3N4-500(N2)—by strongly anchoring Au nanoparticles (~5 nm) onto a C3N4 matrix—which simultaneously enhanced the activity towards the photosynthesis of H2O2 and the stability when it was reused. The yield of H2O2 reached 1320 μmol L−1 on Au/C3N4-500(N2) after 4 h of light irradiation in an acidic solution (pH 3), which was higher than that (1067 μmol L−1) of the control sample Au/C3N4-500(Air) and 2.3 times higher than that of the pristine C3N4. Particularly, the catalyst Au/C3N4-500(N2) retained a much higher stability. The yield of H2O2 had a marginal decrease on the spent catalyst—i.e., 98% yield was kept. In comparison, only 70% yield was obtained from the spent control catalyst. The robust anchoring of Au onto C3N4 improved their interaction, which remarkably decreased the Au leaching when it was used and avoided the aggregation and aging of Au particles. Minimal Au leaching was detected on the spent catalyst. The kinetic analyses indicated that the highest formation rate of H2O2 was achieved on the Au/C3N4-500(N2) catalyst. The decomposition tests and kinetic behaviors of H2O2 were also carried out. These findings suggested that the formation rate of H2O2 could be a determining factor for efficient production of H2O2.


Introduction
Hydrogen peroxide (H 2 O 2 )-a mild and environment-friendly oxidant-has been used as a green chemical and applied in a variety of industrial and living activities-such as papermaking, water treatment, chemical synthesis, disinfection, and sterilization [1]. In addition, H 2 O 2 is a promising liquid fuel for H 2 O 2 -based fuel cells, because only water is emitted [2]. Compared with the widely known fuel H 2 , H 2 O 2 is easier to store and is more conveniently transported with a low risk of wastage. The current industrial production of H 2 O 2 is based on an anthraquinone method. The advantages of this are a scale-up and high productivity of H 2 O 2 . However, huge energy consumption, complicated by-products, and safety issues are still challenges being faced. It is highly desirable to pursue an efficient, mild, and sustained method as an alternative for the synthesis of H 2 O 2 .
The light-driven formation of H 2 O 2 has received increased attention in recent decades [3][4][5][6][7][8]. The process involves light capture on the photocatalyst, charge separation under light irradiation, charge transfer to the surface of the photocatalyst, water/alcohol oxidation by photo-generated holes, and the reduction of dioxygen by photo-generated electrons [9][10][11]. In such a process, a highly efficient photocatalyst could be a core factor in order to direct the synthesis of H 2 O 2 . Titanium dioxide (TiO 2 ), a widely studied photocatalyst, has been found to enable the formation of H 2 O 2 [12][13][14][15][16]. Several strategies have been developed to enhance the H 2 O 2 yield on TiO 2 -based photocatalysts. For example, the surface modification of TiO 2 by fluorine ions was proposed [13]. The loading of noble metal (Au) and alloys (AuAg) on TiO 2 were studied [14,16]. However, the robust hole oxidative capability of TiO 2 accelerated the decomposition of H 2 O 2 on the surface of TiO 2 , and thus lowered the H 2 O 2 yield [17]. Additionally, the ultraviolet (UV) responsive characteristic of TiO 2 limited its light absorption and utilization. The graphitic phase carbon nitride (g-C 3 N 4 )-a polymeric layered material-was explored as a photocatalyst for pollutant decomposition and water splitting, because of its relatively narrow bandgap (~2.7 eV), and its suitable valence band and conduction band levels [18,19]. Recently, C 3 N 4 -based materials have been utilized in the light synthesis of H 2 O 2 [20][21][22][23][24][25]. For instance, Shiraishi et al. reported the selective production of H 2 O 2 on g-C 3 N 4 photocatalyst [20]. The H 2 O 2 production from water and molecular oxygen was improved on a pyromellitic diimide (PDI)-modified C 3 N 4 -by the same authors [21]. The modification caused the conduction and valence band levels of C 3 N 4 to positively shift, which was more favorable to the two-electron reduction of dioxygen needed to produce H 2 O 2 . The mesoporous g-C 3 N 4 was also prepared by using silica sol as a soft template [22]. The effects of surface defects on photocatalytic H 2 O 2 production were also investigated. The results indicated that the suitably large surface area favored an increased H 2 O 2 yield. Recently, a carbon nitride-aromatic diimide-graphene nanohybrid was synthesized which demonstrated extremely high activity towards H 2 O 2 generation-possessing a solar-to-hydrogen peroxide conversion efficiency of 0.2% [24]. Furthermore, Moon et al. reported a photochemical production of H 2 O 2 over carbon nitride frameworks incorporated with multiple heteroelements (K, P, and O) [25].
In our previous studies, the g-C 3 N 4 exhibited enhanced activity when combined with mixed-metal oxide (MMO) [26]. The reason for this is that the positive shift of the conduction band of C 3 N 4 improved the two-electron reduction of dioxygen to H 2 O 2 . Additionally, the MMO plays the role of the water oxidation catalyst (WOC), which has been recognized to be active in oxygen-evolving reactions [27][28][29]. Furthermore, the H 2 O 2 production was increased on the carbon-modified g-C 3 N 4 , which regulates the energy levels and is more suitable for the selective two-electron reduction of oxygen by photo-generated electrons [30]. Consequently, the efficient separation of photo-excited electrons and holes, and appropriate band positions are advantageous to the production of H 2 O 2 . The noble metals (e.g., Au, Ag, and Pt) revealed the reductive ability in supported catalysts by collecting electrons, similar to a reservoir. The activity is closely related to the size of the noble metal nanoparticles. Generally, when the size of the metal particles is greater than 7 nm, the activity drops sharply [31]. However, the smaller-sized metal particles possess a low Tammann temperature and high surface energy, leading to low thermal stability and the sintering phenomenon when heated [32]. In this article, we investigated Au-supported C 3 N 4 prepared by a stabilized method. The Au/C 3 N 4 catalyst exhibited high activity towards H 2 O 2 production and superior stability when reused. The strong anchoring of Au nanoparticles (around 5 nm) on C 3 N 4 inhibited the leaching and prevented the particles from agglomerating in the sintering process of the catalyst. The kinetics were also studied in order to understand the determining pathway for H 2 O 2 production.

Catalyst Characterization
The Au loading and distribution on the C 3 N 4 matrix were observed by transmission electron microscopy TEM (Figure 1). The average size of Au particles was 5.0 nm (Figure 1a,b) and 7.5 nm (Figure 1d,e) in the Au/C 3 N 4 -500(N 2 ) and Au/C 3 N 4 -500(Air) catalysts, respectively. The Au particles had a narrower size distribution on the Au/C 3 N 4 -500(N 2 ) than on the Au/C 3 N 4 -500(Air). The high-resolution transmission electron microscopy photographs showed that the distance of the neighboring planes was 0.235 nm in both catalysts, corresponding to the d-spacing of the plane (111) in Au [32]. Au particles had more crystal defects on the Au/C 3 N 4 -500(N 2 ) (Figure 1c) than on the Au/C 3 N 4 -500(Air) (Figure 1f). The smaller-sized Au possessed more edges, corners, and steps, which were favorable for the adsorption of reactive species and further catalytic reactions. The energy dispersive spectrometer EDS mapping of the Au/C 3 N 4 -500(N 2 ) is presented in Figure 1g. The results suggested that Au, C, and N elements had a homogeneous distribution in the catalyst. The phase structure of the catalysts was determined by X-ray diffractometer XRD ( Figure 2). The diffractions at 2θ = 13.3 • and 27.4 • were indexed to the planes-namely, 100 and 002-of g-C 3 N 4 (JCPDS 87-1526) in all samples, which originated from the in-plane repetition of tri-s-triazine and the stacking of the conjugated aromatic units, respectively [33]. The formation of Au was verified in the XRD patterns of the supported catalysts, in which the diffractions were indexed to the planes-namely, 111, 200, and 220 (JCPDS 04-0784) [34]. The loading of Au was estimated by Thermogravimetric analysis (TGA). The weight of Au was 1.05% and 1.35% in Au/C 3 N 4 -500(N 2 ) and Au/C 3 N 4 -500(Air), respectively ( Figure S1). This indicated that the loading of Au in the two catalysts had minimal difference. The elemental valences and surface species were analyzed by X-ray photoelectron spectroscopy XPS (Figure 3), because of its surface-sensitive technique with a detection depth of less than ten nanometers. Two deconvoluted peaks, at 83.7 and 87.4 eV, were observed in the core level spectra of Au4f in Au/C 3 N 4 -500(N 2 ), which were assigned to zerovalent Au [35,36]. The peaks showed a slight shift (0.2 eV) towards decreased binding energy in that of Au/C 3 N 4 -500(Air) ( Table 1). In both catalysts, the N1s core level spectra could be deconvoluted into three peaks. In the g-C 3 N 4 phase, the peaks at 400.6, 399.2, and 398.5 eV corresponded to N-(C) 3 , =N-, and C 2 NH species, respectively [37,38]. The binding energy of the three peaks had a slight shift (<0.2 eV) between the Au/C 3 N 4 -500(N 2 ) and Au/C 3 N 4 -500(Air), which suggested that the nitrogen species had minimal change in both catalysts.  The optical absorption properties of the supported catalysts were characterized by UV-visible diffuse reflectance spectra (Figure 4a). These spectra were obtained by converting the reflection data that was measured using the Kubelka-Munk equation: F(R) = (1 − R) 2 /2R, where R is the reflectance [39]. The three Au/C 3 N 4 samples showed enhanced absorption in both the UV and visible regions when compared with the bare C 3 N 4 . The absorption band below 400 nm in the UV region corresponded to the π-π* electron transition of the aromatic 1-, 3-, and 5-triazine compounds [40][41][42]. The localized surface plasmon resonance (LSPR) absorption of Au was apparent, peaking at~550 nm in the Au/C 3 N 4 -250 and Au/C 3 N 4 -500(Air) samples. The Au/C 3 N 4 -500(N 2 ) sample presented more intensive absorption throughout the spectral region measured. This suggested that the metal-support interaction differed between the Au/C 3 N 4 -500(N 2 ) and Au/C 3 N 4 -500(Air), which could have affected the photocatalytic properties. The fluorescence spectra were measured because the fluorescence emission properties were associated with the carrier recombination ( Figure 4b). The C 3 N 4 itself exhibited an emission band that peaked at 462 nm, which was attributed to the transition between the lone pair valence band and the π* conduction band in the framework of graphitic C 3 N 4 [43]. After the loading of Au nanoparticles, the intensity of the emission band decreased significantly in the Au/C 3 N 4 -250 and Au/C 3 N 4 -500(Air). Notably, the emission in the Au/C 3 N 4 -500(N 2 ) was almost quenched. The quenching of fluorescence emission indicated the suppression of carrier recombination under photo-excitation. The efficient blockage of the recombination channel could have favored the carrier separation and thus stimulated the electrons and holes to involve the photocatalytic reactions on the surface of catalysts. The reduction of emission intensity was more significant in the Au/C 3 N 4 -500(N 2 ) than in the Au/C 3 N 4 -500(Air). It could have indicated that the metal-support interaction was more intensive in the former, which was beneficial for the feasible electron transfer of photo-excited C 3 N 4 to Au.

Photosynthesis of H 2 O 2
The photocatalytic generation of H 2 O 2 on the catalysts was conducted in an O 2 -equilibrated acidic aqueous solution (pH 3) under light irradiation. The H 2 O 2 yield is shown (Figure 5a) as a function of time. The yield increased with time on all catalysts. The higher yield was observed on the Au/C 3 N 4 -500(N 2 ) from the beginning of the reactions. After 240 min of irradiation, the H 2 O 2 yield reached 1320 and 1067 µmol L −1 on the Au/C 3 N 4 -500(N 2 ) and Au/C 3 N 4 -500(Air), respectively-both of which were higher than that on the C 3 N 4 (582 µmol L −1 ) ( Table 2). The yield was higher for the Au/C 3 N 4 -500(N 2 ) than for the Au/C 3 N 4 -500(Air), although the latter had a larger Au loading-1.05% wt in Au/C 3 N 4 -500(N 2 ) and 1.35% wt in Au/C 3 N 4 -500(Air) catalyst. We concluded that the catalytic activity of Au/C 3 N 4 -500(N 2 ) was higher than that of Au/C 3 N 4 -500(Air). This was potentially ascribed to the smaller size of Au nanoparticles in the Au/C 3 N 4 -500(N 2 ) (Figure 1). Similar findings were also observed in the Au-and AuAg-loaded TiO 2 photocatalytsts [14,16]. The apparent quantum yield (Φ AQY ) for H 2 O 2 formation over the Au/C 3 N 4 -500(N 2 ) was calculated using the equation Φ AQY (%) = (2 × [H 2 O 2 ] formed)/(photon number entered into the reaction vessel) × 100. The Φ AQY value measured at a monochromic light of 400 nm was 3.63% (see Supplementary Materials for details).
In order to understand the effect of the light wavelength irradiated, we measured the H 2 O 2 generation under light irradiation with a cut-off filter of 420 nm ( Figure S2). After 4 h of irradiation, the H 2 O 2 yield reached 694 µmol L −1 under the cut-off filter of 420 nm, lower than that measured without the cut-off filter on the Au/C 3 N 4 -500(N 2 ) (1320 µmol L −1 ). This suggested that the irradiation of UV light contributed to the formation of H 2 O 2 when no cut-off filter was used. This indicated that the UV irradiation possessed high energy and could stimulate the C 3 N 4 (with a bandgap of~2.8 eV) to generate photo-carriers. The Au nanoparticles showed the LSPR absorption band at around 550 nm. As a result, the Au/C 3 N 4 -500(N 2 ) had intensive absorption above 500 nm, while the C 3 N 4 showed minimal absorption in the same spectral region (>500 nm) (Figure 4a). To understand the effect of LSPR on the production of H 2 O 2 , we used a cut-off filter of 510 nm. The H 2 O 2 yield was 10 times higher for the Au/C 3 N 4 -500(N 2 ) (293 µmol L −1 ) than for the C 3 N 4 (25 µmol L −1 ) after 4 h of irradiation ( Figure S3). The LSPR of Au enhanced light absorption around the plasmon resonance band and contributed considerably to H 2 O 2 production. To compare the effect of Au loading on the activity, we prepared another two samples of Au/C 3 N 4 -500(N 2 ), with an Au loading of 1% and 3% (a nominal weight). The H 2 O 2 production was carried out on the two catalysts under the same reaction conditions. The H 2 O 2 yield after 4 h of irradiation was 743, 1320, and 917 µmol L −1 , on 1%, 2%, and 3% Au-supported catalysts, respectively ( Figure S4). This finding indicated that an appropriate Au loading led to a higher H 2 O 2 yield. Higher Au loading might have caused larger-sized Au particles and the reduction in activity [44,45]. The Au/C 3 N 4 -500(N 2 ) catalyst was prepared by a carbon-layer-stabilized method, where the dopamine molecule was used as a precursor of the carbon layer. In order to show the effect of dopamine, we prepared a control sample without a treatment of dopamine and named it Au/C 3 N 4 -500(N 2 )-without dopa. The H 2 O 2 yield after 4 h of irradiation was 1320 and 1065 µmol L −1 on the Au/C 3 N 4 -500(N 2 ) and Au/C 3 N 4 -500(N 2 )-without dopa, respectively ( Figure S5). This finding verified the effect of dopamine on the activity of the catalyst.
The H 2 O 2 concentration-time data (Figure 5a) was fitted using the equation [H 2 O 2 ] = (K f /K d ) {1-exp(-K d t)}, where K f and K d were the formation and decomposition rate constants of H 2 O 2 , respectively, and t was the irradiation time [4,26]. The solid lines (Figure 5a) were the fitted curves. Based on the fitted results ( Figure S6), the K f and K d values were listed ( Table 2) and plotted ( Figure 5b). Both K f and K d values were the smallest on the C 3 N 4 . The K f value was the largest on the Au/C 3 N 4 -500(N 2 ) (7.893 µmol L −1 min −1 ) and the K d value was smaller on the Au/C 3 N 4 -500(N 2 ) than on the Au/C 3 N 4 -500(Air) (0.00354 vs. 0.00446 min −1 ). This indicated that a larger K f value and smaller K d value led to a higher yield of H 2 O 2 . The formation rate of H 2 O 2 was dominant in the process of H 2 O 2 production. This was consistent with our previous studies on the metal oxide@C 3 N 4 and carbon-doped C 3 N 4 catalysts [26,30].   (Figure 6). The decomposition data was fitted using a pseudo-first-order equation: −dC(t)/dt = kt, where C(t) is the H 2 O 2 concentration at certain a time, k is the decomposition rate constant, and t is the time. The fitted results showed that the decomposition rate constant followed the order of Au/C 3 N 4 -500(N 2 ) (0.00406 min −1 ) > Au/C 3 N 4 -500(Air) (0.00251 min −1 ) > Au/C 3 N 4 -250 (0.00181 min −1 ) > C 3 N 4 (0.00097 min −1 ) ( Figure S7). The k values had the same order of magnitude as those in the production of H 2 O 2 ( Table 2). The decomposition of H 2 O 2 on the Au-loaded C 3 N 4 catalysts was less dependent on the concentration of H 2 O 2 . The sustained production of H 2 O 2 was mainly dominated by its formation rate.

Stability of the Photocatalysts
The stable production of H 2 O 2 was required on the spent catalysts. After being used for the photosynthesis of H 2 O 2 , the catalysts were separated from the solution and recycled for reuse. We studied the repeated use of the Au/C 3 N 4 catalysts to confirm their stability. The H 2 O 2 yield was almost the same on the spent Au/C 3 N 4 -500(N 2 ) as on the original Au/C 3 N 4 -500(N 2 ), having a negligible decrease of less than 2% (Figure 7). However, the H 2 O 2 yield showed a significant reduction on the reused Au/C 3 N 4 -500(Air) and only 70% of the original yield was retained. This suggested that Au/C 3 N 4 -500(N 2 ) had a much higher stability than Au/C 3 N 4 -500(Air) for the photosynthesis of H 2 O 2 . This was verified by the structural characterizations of the spent catalysts. The spent catalysts were characterized by TEM, inductively-coupled plasma spectrum ICP, and XPS. The average size of Au particles increased from 5 to 7 nm in the Au/C 3 N 4 -500(N 2 ) after reactions (Figure 8a,b). The Au particles had less agglomeration and were still highly dispersed in the C 3 N 4 matrix. In contrast, the Au particles showed an increase in size from 7.5 to 10 nm in the Au/C 3 N 4 -500(Air) (Figure 8c,d). The Au particles exhibited obvious aggregates. In addition, the elemental analyses showed that the Au loading had a slight loss of 5.7% in the spent Au/C 3 N 4 -500(N 2 ), while the Au remarkably lost 17.8% in the spent Au/C 3 N 4 -500(Air) (Figure S1). Such a significant leaching of Au could be a critical factor for the sharp decrease of H 2 O 2 yield, as well as the increased size of Au particles in Au/C 3 N 4 -500(Air) (Figure 7). The XPS results indicated that the valence of Au had no change and Au still existed in the form of zero-valence. The fitted peaks in the N1s spectra had minimal shift compared to the original catalysts ( Figure S8). The flatband potentials (E fb ) of the catalysts were estimated from the electrochemical Mott-Schottky plots, measured at varied frequencies. The Mott-Schottky plots exhibited positive slopes, characteristic of n-type semiconductors ( Figure 9) [26]. The E fb values were derived by extrapolating the plots to the x-axis, which were −0.86 V, −0.68 V, and −0.70 V (vs. RHE) for C 3 N 4 , Au/C 3 N 4 -500(N 2 ), and Au/C 3 N 4 -500(Air), respectively. The E fb values of Au-supported C 3 N 4 catalysts had a minimal positive shift (~0.16 V) compared with that of C 3 N 4 , and had minimal difference between Au/C 3 N 4 -500(N 2 ) and Au/C 3 N 4 -500(Air). Generally, the conduction band potential was approximately equal to the E fb [30]. Consequently, the conduction band levels hardly changed in the Au-supported catalysts, regardless of the stability. The bandgap of the catalysts was also determined from the Tauc plots. The Tauc plots of Au/C 3 N 4 -500(N 2 ) and Au/C 3 N 4 -500(Air) were derived according to the equation (F(R)·hυ) 2 = A(hυ-E g ), where h was Plank constant, υ was the frequency, A was a constant, and E g was the band gap ( Figure S9) [39,46,47]. The estimated bandgap was 2.83 eV and 2.85 eV for the Au/C 3 N 4 -500(N 2 ) and Au/C 3 N 4 -500(Air), respectively. Additionally, the bandgap values had no change in the spent catalysts. From the energy level analyses, we concluded that the different activity and stability of the Au-supported C 3 N 4 catalysts were mainly affected by the Au sizes and metal-support interactions rather than the energy band structure.

Preparation of g-C 3 N 4
The g-C 3 N 4 was prepared by heating of a melamine precursor [33]. The melamine (10 g) was added to a porcelain cup and calcined for 4 h at 520 • C (heating rate of 4 • C min −1 ). The product was a yellow powder.

Au/C 3 N 4 -250 Catalyst
The Au/C 3 N 4 -250 catalyst was synthesized by a method similar to the literature [32]. Briefly, the pH value of an aqueous solution (0.02 g HAuCl 4 ·3H 2 O in 60 mL water) was adjusted to 10.0 by adding 1 M NaOH. 1 g of C 3 N 4 powder was added to the solution and heated to 70 • C for 2 h under stirring. The suspension was washed three times using deionized water by centrifugation. Finally, the product was dried for 2 h at 40 • C and subsequently calcined in air for 2 h at 250 • C. Before further use, the product was sealed in vacuum desiccator at room temperature. Au/C 3 N 4 -500(N 2 ) Catalyst Firstly, the Au/C 3 N 4 -250 sample was modified with dopamine according to the previous method [48]. The 0.5 g of Au/C 3 N 4 -250 powder was added to 1 mg mL −1 dopamine-containing solution, which was pre-adjusted to pH 8.5 by a tris-buffer solution (10 mM). The suspension was stirred for 24 h, allowing for the formation of polydopamine (PDA) on the surface of solid. Subsequently, the suspension was separated by centrifugation, washed three times using deionized water and once using ethanol. The product was dried for 2 h at 40 • C. Subsequently, the dried product was annealed in a gas flow of nitrogen for 2 h (heating rate of 5 • C min −1 ) at 500 • C. The resulting powder was further annealed in air for 2 h at 500 • C to obtain the Au/C 3 N 4 -500(N 2 ) catalyst.

Au/C 3 N 4 -500(Air) Catalyst
The 0.5 g of Au/C 3 N 4 -250 powder was directly annealed in air for 2 h at 500 • C to obtain the Au/C 3 N 4 -500(Air) catalyst.

Conclusions
The Au nanoparticles were strongly anchored onto the C 3 N 4 matrix by a carbon-layer-stabilized method. The as-prepared Au/C 3 N 4 -500(N 2 ) catalyst showed higher activity towards light-driven synthesis of H 2 O 2 than the control Au/C 3 N 4 -500(Air). This was ascribed to the smaller size of Au nanoparticles (~5 nm) and the stronger metal-support interaction on the Au/C 3 N 4 -500(N 2 ). The H 2 O 2 yield reached 1320 µmol L −1 after 240 min of light irradiation in an acidic solution (pH 3), 2.3 times higher than that of the pristine C 3 N 4 . The spent Au/C 3 N 4 -500(N 2 ) catalyst exhibited a marginal loss of activity and the H 2 O 2 yield remained >98% of the original yield. However, the H 2 O 2 yield decreased to 70% of the original yield on the Au/C 3 N 4 -500(Air) catalyst when reused. There were primarily two reasons for this. One was the significant leaching of Au (17.8%) from the Au/C 3 N 4 -500(Air) catalyst after reactions, much greater than that (5.7%) from the Au/C 3 N 4 -500(N 2 ) catalyst. The other was the increased sizes of Au (~10 nm) on Au/C 3 N 4 -500(Air), which led to the decrease in activity. The energy band structures of the Au/C 3 N 4 -500(N 2 ) and the Au/C 3 N 4 -500(Air) showed minimal difference before or after the reactions. Therefore, the highly stable Au-supported C 3 N 4 catalyst could be favorable for the sustained production of H 2 O 2 via a two-electron reduction of dioxygen driven by solar light.