Identification and Analysis of Multiple Factors Controlling Solar-Driven H 2O 2 Synthesis Using Engineered Polymeric Carbon Nitride

SummarySolar-driven hydrogen peroxide (H<sub>2</sub>O<sub>2</sub>) production presents unique merits of sustainability and environmental friendliness. Herein, highly efficient solar-driven H<sub>2</sub>O<sub>2</sub> production through the reduction of dioxygen coupled with the oxidation of biomass-derived glycerol as a sustainable reductant was achieved by employing an unusual p-type polymeric carbon nitride (PCN) framework with sodium cyanaminate moiety (PCN-NaCA), which exhibited a highly enhanced production rate of 15.0 mmol H<sub>2</sub>O<sub>2</sub> h<sup>−1</sup> g<sup>−1</sup><sub>(catal.)</sub> and a notable apparent quantum yield of 27.6% at 380 nm. The overall photocatalytic transformation process was systematically analyzed using various static/time-resolved spectroscopic and computational methods. The presence of sodium cyanaminate moiety in PCN-NaCA induced the following multiple effects: enhancing photon absorption, switching to p-type character with accumulating more electrons in the surface region, retarding radiative charge recombination, enhancing surface adsorption of O<sub>2</sub>, and favoring highly selective 2e<sup>−</sup> ORR. In particular, it was found that the adsorption of O<sub>2</sub> (an electron acceptor) on PCN-NaCA actually enhances the population and lifetime of trapped electrons in the ps-ns time regime, which should have a notable synergic effect on oxygen reduction process. All of these unique properties of PCN-NaCA positively contribute to the extraordinary photoactivity of H<sub>2</sub>O<sub>2</sub> production.


3
The pendant amino group on PCN is proven to introduce energetically deeper trapping sites which may negatively influence the photocatalytic activity. 29 Various methods for modifying the structure and composition of PCN have been investigated to improve its photocatalytic activity. 30,31 For example, substitution of the pendant amino groups by cyanamide units improves the solar hydrogen evolution reaction (HER) activity of the PCN photocatalysts. 26,27,32 However, a comprehensive mechanistic understanding on how the specific structural features influence each step in the photocatalytic 2e − oxygen reduction reaction (ORR) is challenging and relatively unexplored while such information is critical for the rational design of a highly efficient solar H 2 O 2 production system. Herein, superior solar-driven 2e − ORR performance is achieved on a p-type PCN framework. The key mechanistic features of the overall photocatalytic process are analyzed by tracing the consecutive electrons transfer steps involved in the photoinduced processes and the subsequent surface reactions (Scheme 1). Introduction of sodium cyanaminate moiety switches the n-type character of PCN to p-type. The p-type photocatalyst exhibits enhanced photon absorption capacity and retarded emissive charge-recombination by trapping a significant fraction of charge carriers. It also shows enhanced accumulation of surface charge, stronger surface affinity for dioxygen, and more catalytic active sites for selective 2e − ORR.

Results and Discussion
PCN was synthesized by condensation reaction of the melamine-cyanuric acid complex under high temperature ( Figure   1a). The resulting PCN was further treated with sodium thiocyanate (NaSCN) molten salt to tailor the conjugated electronic structure and the surface properties. Further condensation reaction occurs in the molten salt and leads to two favorable structural features: (1) improved polymerization degree and expanded conjugated electronic structure; [33][34][35][36] and (2) conversion of the amino group to the sodium cyanaminate moiety (Figures 1a, S1, and S2). The PCN frameworks with sodium cyanaminate moiety is denoted by PCN-NaCA-n (n : 1, 2, and 3 refer to the sample with the salt/PCN weight ratio of 0.5, 1, and 2, respectively). The structure of PCN for the simulation is the linear melon with infinite repeating units. For PCN-NaCA-2, sodium ion interacts with four nitrogen atoms from two adjacent heptazine units in the optimized structure (Figures 1b, 1c).
The photocatalytic selective reduction of O 2 to H 2 O 2 , in an ideal scenario, should employ H 2 O as the proton/electron donor, so that there is no additional CO 2 emission from this process. [37][38][39][40] However, the electron and proton extraction through water oxidation process (2H 2 O → 4H + + 4e  + O 2 ) is inefficient as the hole transfer kinetics toward water oxidation is sluggish, causing severe charge recombination, which results in low solar conversion efficiency. 41 In natural photosynthesis, electrons/protons are extracted from water via complicated bio-enzymatic reactions and subsequently used Electronic copy available at: https://ssrn.com/abstract=3606779 4 in transforming CO 2 to biomass to achieve the energy-uphill reaction. 42,43 An alternative solution is to utilize more reactive organic substances (as an electron/proton donor instead of water) that are abundant and cheap. In the rapidly rising biodiesel industry, glycerol is the byproduct and its yield accounts for 10 wt.% of the biodiesel production, but the limited consumption of glycerol makes it surplus. 44 Developing a proper process of consuming and valorizing glycerol well matches the market need. The crude glycerol is cheap with price of 10 -15 c/lb. (80 wt.%, Oleoline), while H 2 O 2 is a moderately valuable chemical with price of around 6.0 USD/lb. (35 wt. %, Supleco). Moreover, glycerol is non-toxic and bio-degradable, rendering it an ideal practical electron/proton donor for the solar fuel production. [37][38][39][40] Therefore, solar production of H 2 O 2 with using glycerol as the electron/proton donor can be proposed as an environmentally benign and cost-effective solution.    Figure 1d). The AQY is 27.6% and 11.8% at 380 nm and 420 nm, respectively, and rapidly decreases with further increasing the wavelength, which matches well with the absorption spectral profile of the photocatalyst. The fact that the action spectrum of H 2 O 2 production is closely correlated with the optical absorption spectrum supports the photocatalytic mechanism based on ORR.
To comprehensively understand the rationale for the above-mentioned superior photoactivity of the cyanaminatemodified PCN, we carried out systematic mechanistic investigations on the following aspects: 1) excitation and emissive decay process; 2) non-emissive states, focusing on population and decay kinetics of the trapped electrons; and 3) surface processes that include surface electron trapping, dioxygen adsorption, and ORR activity and selectivity (Scheme 1).
Compared to PCN, the photon absorption spectra of PCN-NaCA-n samples show significantly improved absorbance at 350 nm -380 nm which is commonly observed in the conjugated aromatic systems with π − π* transition ( Figure 3a). 46,47 The absorbance at 450 -500 nm is attributed to n − π* transitions involving lone pair electrons on the N atoms of the amino group and the secondary amine unit in the framework. The n − π* transition is forbidden for perfectly symmetric and planar s-triazine/heptazine units, but they become allowed as the structures develop distortions. 46,47 For the PCN-NaCA-n samples, the improved polymerization degree increases layer buckling, and the interruption from the sodium cyanaminate moiety also leads to distortion of the conjugated heptazine structure. [48][49][50][51] As a result, PCN-NaCA-n samples show increased absorbance not only at π − π* transition but also at n − π* transition at 450 -500 nm.
In the Tauc plots from Kubelka-Munk function transformation, the optical band gap is determined to be 2. demonstrating that 2e − ORR by conduction band electrons is thermodynamically feasible (Figure 3d). The enhanced photon absorption is a primary prerequisite for the high activity of H 2 O 2 production, as this step provides the initial driving force for the whole solar energy conversion process.
Electronic copy available at: https://ssrn.com/abstract=3606779 Upon photon absorption, the exited photocatalyst will either relax to the ground state via photoluminescence (PL) or transit to non-emissive state through charge trapping wherein some of the trapped charges will participate in the expected surface chemical reaction steps. 52 Steady-state and transient PL spectroscopy is thus employed as an indirect method for analyzing the situation of the trapped charges. 53 As shown in Figure 3e, PCN shows strong PL emission peak at 488 nm, while the emission intensity of PCN-NaCA-2 is significantly attenuated. Moreover, considering the stronger photon absorption of PCN-NaCA-2 than PCN, much larger proportion of the excited states should transit to the non-emissive states on PCN-NaCA-2 than PCN at this stage. 50 It is also interesting to note a blue shift of 18 nm for PCN-NaCA-2 as compared to that of PCN, which is attributed to the quantum confinement caused by the decreased thickness of the layer stacking. 48,54−57 Electronic copy available at: https://ssrn.com/abstract=3606779 To further understand the variation of the electronic structure of the conjugated system with sodium cyanaminate moiety, the decay kinetics of the emissive state is thereafter analyzed by time resolved photoluminescence spectroscopy.
As shown in Figure 3f, PCN-NaCA-2 shows faster PL decay kinetics than that of PCN, showing average lifetime of 3.46 and 8.27 ns for PCN-NaCA-2 and PCN, respectively (Table S2). The shorter lifetime and weaker PL emission intensity of PCN-NaCA-2 compared to PCN indicates the fast quenching of luminescence, which might be attributed to the enhanced charge separation by improved polymerization degree. 58,59 We Under vacuum condition, after excitation by laser pulse with photon fluence of 79.6 μJ cm −2 , PCN-NaCA-2 presents characteristic broad absorption peak at 640 nm, which is identified as trapped electrons (Figures 4a and 4b). 60 Figure 4c shows the decay kinetics profiles of the photo-induced electrons with various photon fluence, and the initial absorption Electronic copy available at: https://ssrn.com/abstract=3606779  Electronic copy available at: https://ssrn.com/abstract=3606779  Figure 6c, PCN-NaCA-2 exhibits significantly higher dioxygen adsorption capacity (by mass), which is around 3 times larger than that of the pristine PCN. Moreover, PCN-NaCA-2 has a much lower BET surface area than that of PCN, i.e., 11.9 cm 2 g −1 for PCN-NaCA-2 versus 83.2 cm 2 g −1 for PCN. This implies that the density of the surface adsorbed dioxygen molecules on PCN-NaCA-2 is around 20 times larger than that on PCN. More importantly, it is noted that PCN-NaCA-2 has much higher surface binding affinity for O 2 , as the dioxygen desorption peak on PCN-NaCA-2 appears at 160 o C, much higher than that of PCN at 104 o C. The stronger surface binding affinity for O 2 as well as the high density of adsorbed O 2 should contribute synergically to the highly enhanced ORR activity.
Finally, for an efficient H 2 O 2 production, the selectivity towards 2e − ORR is of critical importance. The performance of 2e − ORR to H 2 O 2 is thus evaluated by analyzing its electrochemical selectivity on a rotating ring disc electrode (RRDE) wherein the disc current comes from the dioxygen reduction reactions (including 1e − , 2e − , and 4e − transfer pathways) and the ring current comes from the 2e − oxidation of H 2 O 2 produced from the disc. For further understanding the fundamental mechanism of the above-mentioned superior surface-catalytic 2e − ORR performance, we carried out first-principal calculations based on proposed 2e − ORR steps in alkaline solution. Figures 7a and 7b show, respectively, the optimized configurations of the surface dioxygen adsorption on PCN and PCN-NaCA-2. The adsorption of dioxygen on PCN is very weak, presenting an adsorption energy of −0.017 eV, which is mainly Van der Waals interaction.
On the other hand, PCN-NaCA-2 exhibits much higher adsorption energy of −0.446 eV, which explains why the strong surface dioxygen affinity was observed in the O 2 -TPD characterization (Figure 7g). This high adsorption energy may result from the interaction between surface adsorbed dioxygen with conjugated π-electrons as well as the sodium cation. The free energy of the intermediate *OOH on PCN-NaCA-2 drops further to −0.784 eV, which is much lower than that on PCN (Figures 7c, 7d, and   7g). After electron/ proton transfer, the surface adsorbed H 2 O 2 (*H 2 O 2 ) forms and its free energy on PCN-NaCA-2 reaches Electronic copy available at: https://ssrn.com/abstract=3606779 13 −2.345 eV, which is still lower than that on PCN with free energy of −1.714 eV. From this analysis, it is evident that 2e − ORR pathway is energetically more favorable on PCN-NaCA-2 surface than on PCN surface.

Conclusion
In the solar-driven selective 2e − ORR with using the biomass-derived glycerol as the electron/proton donor, the p-type carbon nitride framework with cyanaminate sodium salt moiety exhibits superior photoactivity for H 2 O 2 production, which is 21 times higher than that of n-type PCN. The critical factors and steps in the overall photoconversion process, which include the band energy levels, photon absorption, charge recombination/separation/trapping, dioxygen adsorption, and interfacial electron

Declaration of interests
The authors declare no competing interests.

Synthesis of the Photocatalysts:
PCN was synthesized by the polymerization reaction of cynauric acid-melamine super-molecular assembly under high temperature. 1 In a typical synthesis, 5 gram cynauric acid and 5 gram melamine was mixed with 180 mL water and 20 mL isopropyl alcohol, and magnetically stirred for 24 hours. The white precursor was then collected by centrifugation and dried in vacuum overnight. Dry precursor was calcined at 600 o C for 3 hours with temperature ramping rate of 5 o C/min in nitrogen flow. The as prepared yellow sample was washed with water, ethanol, and dried in vacuum oven over night, and then pulverized by ball milling.
Grafting of the sodium cyanaminate moiety on the polymeric carbon nitride framework was realized by molten salt treatment. In a typical synthesis, 1 g PCN and 2 g NaSCN was mixed with a small amount of water for improving the contact of the salt and PCN. The mixture was then dried in vacuum overnight before subject to 400 o C heating in a tube furnace with nitrogen flow. PCN-NaCA-1, PCN-NaCA-2, and PCN-NaCA-3 was, respectively, prepared by mixing 0.5, 1, and 2 gram of NaSCN with 1 gram PCN under the otherwise same reaction condition.

General Procedure of the Photocatalytic Reactions:
The photocatalytic hydrogen peroxide production reactions were conducted in photoreactor filled with 35 mL glycerol aqueous solution and catalyst. 5 mg of photocatalyst is used for catalyst performance comparison and pH impact evaluation; and 10 mg photocatalyst is used for the long-term running experiments. The pH of the mixture was adjusted by KOH or HClO 4 to the expected value. The photoreactor was irradiated by monochromatic light from a Xeon lamp with a series of band-pass filters. The reaction mixture was sampled at specific time intervals.
The concentration of the hydrogen peroxide was determined with DPD/POD method reported elsewhere, which is based on the horsedish peroxides (POD)-catalyzed oxidation of N, N-diethyl-p-phenylenediamine (DPD) by hydrogen peroxide. 2 The calibration curve is prepared with a series of H 2 O 2 solutions with known concentrations. The concentration of the standard H 2 O 2 solution was determined by KMnO 4 titration, and Na 2 C 2 O 4 standard solution was employed for the analysis of the concentration of the titrant (KMnO 4 solution).

Apparent Quantum Yield (AQY) Measurement:
Reaction mixture was prepared by ultra-sonication dispersion of 10 mg PCN-NaCA-2 in 35 mL glycerol aqueous solution with concentration of 0.38 mol/L. 2 mL of the oxygen saturated reaction mixture was added into a quantz cuvette, which was capped by septum and wrapped by aluminum foil with a 1 cm 2 window. The cuvette was then subject to monochromatic light irridiation for a short time duration, e.g. λ = 380 nm, 420 nm light irradiation for 0.5 min, 450 nm light irradiation for 3 min, 475 nm, 500 nm, and 550 nm light irradiation for 30 min.
The amount of photons (Mp) absorbed by the photocatalyst was calculated by the equation: P λ is the power density on the surface of the cuvette; a is the area of the window on the cuvette (0.0001 m 2 ); t is the irradiation time (s); E λ is the energy per photon; N A is Avogadro's constant ( ); A is the 6.02 × 10 23 -1 absorbance of the cuvette with reaction mixture at λ nm on UV-Vis spectrometer with the water as the background.
The apparent quantum yield (AQY) was calculated by the equation:

Characterizations:
FT-IR spectra was collected on Thermo Scientific Nicolet iS 50 with diamond attenuate total reflection (ATR) unit. The morphology of the catalyst samples were determined by scanning electron microscope (SEM) and high resolution transmission electron microscope (HRTEM) on Hitachi S4300 and JEOL 2100, respectively. X-ray diffraction patterns of the catalysts were obtained on a Rigaku D/MAX 2500 diffractometer with Cu radiation (Cu Kα=0.15406 nm). X-ray photoelectron spectroscopy (XPS) patterns was recorded on an ESCA laboratory 220i-XL spectrometer with an Al Kα (1486.6 eV) X-ray source and a charge neutralizer; all the binding energy were calibrated to C 1s peak at 284.6 eV. BET surface area was measured via nitrogen sorption at 77 K on a surface area analyzer (QuadraSorb SI); the samples were degassed at 200 o C before nitrogen adsorption. Contents of nitrogen and carbon in the sample was analyzed with vario El cube (Elementar, Germany). The UV-Vis absorbance property the samples was recorded on Shimazu UV-2600.

SPV and TPV measurement:
The SPV spectra measurement was conducted based on the lock-in amplifier. 3,4 The measurement system consists of a 500 W xenon lamp with monochromator (SBP500, Zolix) as light sourve, a lock-in amplifier (SR830, Stanford Research Systems, Inc.) with a light chopper (SR540, Stanford Research Systems, Inc.), and a sample chamber. The monochromatic light was chopped with a frequency of 23 Hz. The monochromator and the lock-in amplifier were controlled by a computer. The input resistance of the lock-in amplifier is 10 MΩ. SPV spectra were carried out by scanning from low photon energy to high. The TPV measurement system consists of a Nd:YAG laser source (Polaris II, New Wave Research, Inc.) providing 355 nm laser pulse radiation (pulse width 5 ns), a 500 MHz digital phosphor oscilloscope (TDS 5054, Tektronix) with a preamplifier, and a sample chamber. The oscilloscope was triggered by a synchronous signal provided by a photomultiplier. The input resistance of the preamplifier is 100 MΩ.

Electrochemical oxygen reduction reaction (ORR) performance measurement:
The electrochemical ORR performance is evaluated on a CHI 760E workstation equipped with PINE rotating ring disk electrode (RRDE) unit. The electrochemical system setup is assembled on the basis of three-electrode configuration consisting of catalyst-coated RRDE working electrode, Ag/AgCl reference electrode, Pt/C counter electrode, and O 2 -saturated 0.1 M KOH electrolyte. The working electrode is fabricated by a traditional "dropcasting" method. Specifically, the ink was prepared by mixing 5 mg of catalyst, 750 μL of isopropyl alcohol, 250 μL of water, and 10 μL of Nafion solution (5%) under ultrasonication. Next, 10 μL ink was dropped onto the surface of electrode and dried in air under room temperature. The loading of the catalyst on RRDE electrode is 0.202 mg cm −2 .

Density functional theory (DFT) calculation method:
The explored model is linear melon structure with infinite repeating unit with edge nitrogen saturated by hydrogen (for PCN) or replaced by cyanaminate sodium salt unit (for PCN-NaCA). The calculations were based on spinpolarized density functional theory (DFT) using projector augmented wave (PAW) methods, as implemented in the Vienna ab initial simulation package (VASP 5.2). A plane-wave basis set with 400 eV kinetic-energy cut-off was used to expand the wave function of valence electrons. The generalized gradient approximation (GGA) with the Perdew-Burke-Ernzerhof (PBE) functional was used for describing the exchange-correlation interactions.
In this work, we construct 1-D nano-line to module the PCN and PCN-NaCA. 20 Å vacuum space was set to prevent the interaction between adjacent nanolines. The Brillouin-zone integration was sampled with 5 × 1 × 1 k-points Monkhorst-Pack mesh. The structural relaxations were performed by computing the Hellmann-Feynman forces within the total energy and force convergences of 10 -4 eV and 10 -2 eV/Å, respectively. The adsorption energy E ad of interaction between the carbon nitride and oxygen molecule is calculated based on the following equation: where E (CN+O2) and E (CN) is, respectively, the total free energy of carbon nitride with and without dioxygen molecule, E (O2) is the free energy of dioxygen molecule.
The electro-catalytic active sites and process of oxygen reduction reaction (ORR) on the surface of carbon nitride were examined by first-principle calculations. Only intermediate states in the ORR process were considered, and energy barriers might exist but were not considered in here. The calculation is based on the following reaction steps:

Interpretation of the characterizations:
FT-IR spectra: The absorption peak at 808 cm −1 is due to the out-of-plane bending of the tri-s-triazine unit. For PCN, strong and broad absorption peak at 3000 cm −1 to 3500 cm −1 from the N-H stretching vibration is observed, stating the presence of rich amino groups. 5 With the increase of the salt/PCN ratio (from PCN-NaCA-1 to PCN-NaCA-3), the asymmetric vibration peak of cyano group at 2175 cm −1 grows, and the N-H vibration peak decreases simultaneously, demonstrating the conversion of the amino group to the cyanaminate moiety.

XRD patterns:
The PCN presents a strong and sharp X-ray diffraction peak at 27.6 o , demonstrating the layer stacking structure on (002) direction with interlayer spacing of 0.33 nm. All PCN-NaCA-n samples present the same diffraction angle, i.e., the introduction of sodium cyanaminate moiety does not alter the interlayer spacing of the stackings. However, further polymerization reaction in molten salt changes the intensity and width of the diffraction peak. As compared to PCN, PCN-NaCA-2 shows much stronger and wider diffraction peak. This states that further polyerization reaction in the molten salt produces new carbon nitride layer stackings with much thinner thickness than that of PCN.

XPS profiles:
For the sample PCN, C 1s peaks can be deconvoluted into three peaks at 288.1 eV, 286.3 eV, and 284.7 eV, which are respectively assigned to the N−C=N in the heptazine unit, the carbon atoms connected to −NH 2/1 or cyano groups, and the adventitious carbon (C−C or C=C). 6 In C 1s profile of PCN-NaCA-2, the percentage of the deconvoluted peaks at 286.3 eV increased. Based on the analysis on the FT-IR and 13 C spectra, this intensified peak is repeatedly stating the formation of the cyanaminate moiety on the edge of the carbon nitride framework as well. The deconvoluted peak at 398.6 eV is assigned to the C-N=C in the heptazine unit; and peak at 400.1 eV is attributed to the tertiary N atoms (C-N 3 ), the N atoms in the amino group (-NH 2 ) at the edge conjugated plane, as well as the secondary amine (−NH−) moiety connecting two heptazine units. Comparing to sample PCN, the percentage of the 400.1 eV peak decreases, indicating the depletion of the amino group on the edge of the heptazine unit, which is consistent to the analysis in FTIR spectra, and should be the combined results of further polymerization reaction in the molten salt and the formation of the sodium cyanaminate moiety. 7

C NMR spectra:
In the 13 C NMR spectra of PCN and PCN-NaCA-2, the peaks at 157 ppm and 164 ppm are assigned to the C(3N) and C(2N, NH x ), respectively. For the sample PCN-NaCA-2, there is an additional peak appearing at 171 ppm is assigned to the carbon connecting with the sodium cyanaminate moiety. 8

STEM-EDS:
Uniform distribution of sodium in the matrix is observed.
Elemental Analysis: The C/N ratio of the framework is, respectively, 0.685 and 0.719 for PCN and PCN-NaCA-2.  Notes: The samples were characterized by the time resolved photoluminescence spectroscope with a single photon picosecond laser at a wavelength of 355 nm. The PL decay curves was fitted exponentially using the following equation: where, A1 and A2 represent the normalized amplitudes of each decay component, and