Introduction

With rising concern about environmental degradation, there has been increasing interest in the cost-effective production of hydrogen peroxide (H 2 O 2 ), an environmentally friendly oxidizing agent. 1 H 2 O 2 is an important chemical in a variety of industries, including textile manufacturing, disinfectants, semiconductor cleaning, and oilfield sludge and sulfide treatment. [2][3][4][5][6] In addition, H 2 O 2 can be a potential energy carrier to generate electricity in fuel cells as an alternative to hydrogen. 7,8 The global H 2 O 2 market demand was 4.5 million metric tons in 2020 and the market demand is projected to increase to 5.7 million metric tons by 2027. 9 However, industrial production of H 2 O 2 is dependent on the energy-intensive anthraquinone oxidation process (AOP), which requires large infrastructure, generates chemical waste, and makes on-site H 2 O 2 production difficult. 10 Direct synthesis of H 2 O 2 through the atom economy method provides a facile route to replace the conventional AOP, but it could result in possible explosions from the hydrogen and oxygen gas mixture and has limited efficiency. 11,12 Recently, direct H 2 O 2 production through electrochemical reduction of oxygen has attracted attention due to its high efficiency, sustainability, inhibition of hazardous transportation, and environmentally benign on-site H 2 O 2 production capability under ambient conditions. [13][14][15][16] In principle, the electrochemical reduction of oxygen molecules produces H 2 O 2 or H 2 O via the two-electron (2e À ) oxygen reduction reaction (ORR, eqn (1)) or four-electron (4e À ) ORR pathways (eqn (2)), respectively. 5,17 O 2 + 2H + + 2e À -H 2 O 2 E1 = 0.70 V vs. RHE

E1 is the standard equilibrium reaction potential for each pathway and RHE is the reversible hydrogen electrode. 18 The 2e À ORR pathway includes only OOH* intermediate, while the 4e À ORR pathway involves additional O* and OH* intermediates after reduction of OOH*. 19,20 Therefore, efficient electrochemical H 2 O 2 production requires cathode catalysts with high selectivity to prevent further reduction water. An ideal catalyst also should exhibit high activity, and stability to enable industrial adaption. Some noble metals such as Au, Pd, Pt, Pd-Hg, and Au-Pd have shown high activity and selectivity for the 2e À ORR with a low overpotential. 13,[21][22][23][24] However, the high cost and embodied carbon footprint of noble metals hinders their large-scale application. Introducing a very small amount of metal in carbon materials can generate active metal-nitrogen-carbon (M-N-C) sites for the ORR. [25][26][27] The activity and selectivity of M-N-C catalysts can be controlled by tuning the metal composition and the surrounding atomic structure of the metal centers. 25,28 Co-N-C catalysts show higher activity and selectivity for 2e À ORR than other M-N-C catalysts because they have weak binding between Co and O 2 . 25,28,29 Nevertheless, finetuning the coordination structures in M-N-C catalysts is challenging as their synthesis method relies heavily on simple impregnation and reduction processes with toxic NH 3 gas. 25,26 Recently, metal-free carbon-based catalysts have shown great potential as alternative catalysts for H 2 O 2 production due to their low cost, high activity, tunability, and excellent stability. 17,[30][31][32][33][34][35] In particular, carbon catalysts containing certain surface oxygen and nitrogen groups have shown highly selective H 2 O 2 electrosynthesis in alkaline media. 17,35,36 For example, oxidized carbon nanotubes and reduced graphene oxides with oxygen functional groups on their surface showed high selectivity and catalytic activity for 2e À ORR. 17,35 Additionally, the pyrrolic-N group promotes H 2 O 2 production in nitrogen-rich graphene with few layers. 36 However, most carbon-based catalysts showed a narrow overpotential window (r0.4 V vs. RHE) for maintaining high peroxide selectivity. 17,33,34 Furthermore, the structure-mechanismperformance relationship for metal-free carbon catalysts, which is related to their specific functional groups, morphology, and defect structure in carbon, has not been established.

In this study, we prepare a series of oxidized crumpled graphene (OCG) catalysts with adjustable morphologies, oxygen functional groups, and surface defects to establish the structure-mechanism-performance relationship for functionalized carbon catalysts. Various particle sizes, oxygen functional groups, and defect structures of the 3D-structured OCG catalysts are systematically controlled in situ using the aerosol spray drying process at different annealing temperatures. 2D graphene-based catalysts can reduce the surface area and block the diffusion pathway for O 2 due to the restacking of graphene sheets. However, this unique 3D structure can increase the electrochemically active surface area as well as decrease the mass transport resistance for efficient ORR. OCG synthesized at 800 1C (denoted as OCG-800) has the optimized oxygen functional groups and defect structure, exhibiting the highest H 2 O 2 selectivity (492%) over a wide range of potential (0.05-0.7 V vs. RHE) in alkaline media. In addition, the OCG-800 catalyst delivers a high H 2 O 2 productivity of 473.9 mmol gcat À1 h À1 and excellent stability over 46 h during the chronoamperometry test at 0.4 V vs. RHE. Furthermore, density functional theory (DFT) computations reveal the ORR mechanism of OCG electrocatalysts and the role of functional groups and defect sites in the 2e À transfer reaction pathway. The results shed light on the structure-mechanism-performance relationship of functionalized carbon catalysts for the 2e À ORR and provide important insights for the design of metal-free carbon catalysts with controlled functional groups and defect structures for efficient production of H 2 O 2 .

Results and discussion

Characterization of OCG catalysts OCG was synthesized using a one-step aerosol spray drying process (Fig. S1, ESI †). 37 The prepared graphene oxide (GO) solution was diluted in deionized water (B1 mg mL À1 ) and sprayed with an ultrasonic nebulizer to form GO droplets. The resulting GO droplets were passed through a preheated tubular furnace with argon gas at a flow rate of B1 L min À1 . The preheated temperature was set to a controlled reduction temperature from 250 to 1000 1C. The GO sheets accumulated on the surface of the droplets during the annealing process and rapidly shrunk into 3D-shaped OCG owing to the capillary compression force during rapid water evaporation. 38 The physical morphologies of OCGs synthesized at different temperatures (250, 500, 800, and 1000 1C) were investigated by scanning electron microscopy (SEM), transmission electron microscopy (TEM), and high-resolution TEM (HRTEM) analysis of the dspacing profiles. The resultant OCG particles showed a 3D crumpled ball-like shape consisting of smooth surfaces and sharp ridges (Fig. 1a, b and Fig. S2,ESI †). The SEM images of the OCGs showed particle sizes of 720 AE 132 nm for OCG-250, 578 AE 123 nm for OCG-500, 502 AE 82 nm for OCG-800, and 422 AE 98 nm for OCG-1000, with more ridges at higher annealing temperatures. Narrow ridge regions were formed by the restacking of graphene sheets, as shown in the HRTEM

View Article Online

View Journal | View Issue image of OCG-800 with an interlayer lattice spacing of B0.39 nm (Fig. 1c). These narrow ridge regions are important for maintaining the 3D structure of OCG particles during solution processing. 39,40 The stacking structure of OCG particles was further analyzed using X-ray diffraction (XRD). OCG-250 showed a relatively sharp (002) diffraction peak at 2y = 11.61 with an interplanar spacing of 0.76 nm (Fig. S3, ESI †). Compared to OCG-250, the (002) peaks of OCG-500, OCG-800, and OCG-1000 gradually shifted towards higher diffraction angles, and their interplanar spacings were reduced to 0.40, 0.36, and 0.34 nm, respectively. The reduced interlayer distance can be attributed to the removal of oxygen functional groups at higher temperatures. The surface defect structure of the OCGs was studied using Raman spectroscopy (Fig. 1d). All OCGs showed a broad D peak at B1360 cm À1 , in addition to a sharp G peak at B1580 cm À1 , revealing the incorporation of structural defects in the sp 2 hybridized carbon systems. The integrated ratio of D to G peak (A G /A D ) was calculated to be 0.74, 0.71, 0.67, and 0.65 for OCG-250, OCG-500, OCG-800, and OCG-1000, respectively. This indicates that higher reduction temperatures form more defects on the graphene surface. The thermal reduction of oxygen functional groups is accompanied by the generation of O 2 , CO, and CO 2 products, 41 causing the formation of more defect sites. This intensity ratio was further used to define the amount of defects in the radar plot.

The chemical structure and composition of OCG particles were characterized by X-ray photoelectron spectroscopy (XPS). The XPS survey scans showed reduced oxygen to carbon ratios (O/C) with increasing heating temperatures: 0.49 for OCG-250, 0.22 for OCG-500, 0.16 for OCG-800, and 0.11 for OCG-1000 (Fig. 1e). The XPS high-resolution C1s spectra of OCG catalysts can be fitted by 284.5 eV AE 0.1 eV for sp 2 -C (CQC), 285.2 AE 0.2 eV for sp 3 -C (C-C), 286.5 AE 0.2 eV for C-O, 288.4 AE 0.1 eV for CQO, and 290.2 AE 0.2 eV for COOH (Fig. 1f). A comparison of the C1s spectra illustrates that the graphene sp 2 -C system gradually recovers through the removal of oxygen functional groups at higher temperatures. Interestingly, the dominant epoxide group in OCG-250 was effectively removed in the OCGs synthesized at higher temperatures (Z500 1C). Fourier transform infrared spectroscopy (FTIR) analysis further supported this decreasing trend in the amount of oxygen functional groups with higher reduction temperatures (Fig. S4,ESI †). In addition, O1s XPS spectra of OCG catalysts were deconvoluted by 531. 1g). OCG-250 exhibited a predominant O II species over other oxygen functional groups, but the O II species progressively decreased with increasing annealing temperatures (Fig. 1h and Table S1, ESI †). These results indicate that the physical and chemical structures of OCG, including surface defects and oxygen functional groups, can be effectively controlled by adjusting the reduction temperature (Fig. 1i), which provides an important model system for studying the structure-mechanism-performance relationship of graphene catalysts for the 2e À ORR.

Electrocatalytic ORR performance of OCG catalysts

The electrochemical active surface area (ECSA) of the OCG catalysts was compared by measuring their double-layer capacitance (C dl ) using rate-dependent cyclic voltammetry (CV) scans (Fig. S5 andS6, ESI †). The C dl values of OCG catalysts increased with increasing annealing temperature-from 1.88 mF cm À2 for OCG-250 to 2.54 mF cm À2 for OCG-1000. The increase in ECSA at higher temperatures may be attributed to the increased number of defect sites on the surface of graphene. The ORR performance of the OCG catalysts was evaluated using a 3-electrode cell with a rotating ring-disk electrode (RRDE) at 1600 rpm in O 2 -saturated 0.1 M KOH. The collection efficiency (N = 0.25) was pre-calibrated using the reversible [Fe(CN) 6 ] 4À/3À redox couple (Fig. S7,ESI †). The ORR polarization curves for the OCG catalysts with the oxygen reduction current density at the disk electrode (solid line), and the H 2 O 2 oxidation current density at the ring electrode (dashed line) are shown in Fig. 2a. The corresponding H 2 O 2 selectivity and transference number (n) were calculated over a wide potential window from 0.05 to 0.7 V vs. RHE. OCG-800 maintained a high H 2 O 2 selectivity of 492% over the entire potential, while OCG-500 and OCG-1000 exhibited a reduced H 2 O 2 selectivity of 82-90%, and OCG-250 showed the lowest H 2 O 2 selectivity of 60-78% (Fig. 2b). Notably, OCG-800 showed the excellent H 2 O 2 selectivity of 100% at 0.6 and 0.65 V vs. RHE. It should be noted that OCG-800 exhibited the highest selectivity in the wide range of 0.05 to 0.7 V vs. RHE among previously reported carbon-based and metal-embedded electrocatalysts (Fig. 2c). 25,32,33,35,36,42,43 The calculated n value of OCG-800 from the disk and ring current was the closest to 2.0, further supporting its higher suitability for the 2e À reaction compared to other OCG catalysts (Fig. 2d). These results were in good agreement with the values of n obtained according to the Koutechy-Levich equation using a rotating disk electrode (RDE) (Fig. S8, ESI †). The calculated Tafel slope of OCG-800 (49 mV dec À1 ) was also much lower than those of the other OCG catalysts, indicating the faster reaction kinetics of OCG-800 (Fig. 2e). This outstanding electrochemical performance of OCG-800 can be attributed to the facilitated mass transfer of the reactants to the active sites on the 3D-structured graphene surface containing optimized oxygen functional groups and defect sites. It should be noted that ORR performance is more closely related to oxygen functional group content and defect structure than ECSA (Fig. S9, ESI †). OCG-1000 with the highest ECSA exhibited a similar activity to OCG-500, supporting the importance of the combination of functional groups and defect structures. In addition, the amount of catalyst loaded on the glass carbon (GC) disk electrode can be an important factor for the ORR performance. Therefore, the optimal loading was determined by varying the loading amount of OCG-800 from 5 to 50 mg cm À2 , and it was found that a loading amount of 10 mg cm À2 delivered optimal H 2 O 2 activity and selectivity (Fig. S10,ESI †). Furthermore, the stability of the OCG-800 catalyst was demonstrated by a chronoamperometry (CA) test for 10 h with an average H 2 O 2 selectivity of 98% (Fig. 2f). In addition to the excellent performance of OCG-800 in alkaline media, a high selectivity of B88% was achieved in a neutral solution (0.5 M NaCl and Fig. S11,ESI †). The 0.5 M NaCl electrolyte represents seawater that can be utilized in various energy fields. [44][45][46][47] This result suggests that seawater can be a promising resource as an electrolyte for large-scale, costeffective electrochemical production of H 2 O 2 . 29,[48][49][50][51] The kinetic current density of OCG-800 for H 2 O 2 production was obtained according to eqn (S2) and (S3) (ESI †) in the Experimental section to compensate for the mass-transport limitation. OCG-800 exhibited a kinetic current density of 3.2 mA cm À2 RDE (at 0.65 V vs. RHE) and a mass activity of 158 A g À1 (at 0.65 V vs. RHE), surpassing the state-of-the-art catalysts (Fig. 3a). 17,25,33,35,36,42,43 In addition, the H 2 O 2 production yield of OCG catalysts was measured in the H-cell configuration, and the yield rate was normalized by the catalyst loading amount (B0.78 mg cm À2 , Fig. 3b). The generated H 2 O 2 concentrations were estimated by calibrating the titration method via UV-visible spectroscopy (Fig. S12, ESI †). OCG-800 showed the highest H 2 O 2 productivity of 473.9 (AE29) mmol g cat À1 h À1 at 0.4 V vs. RHE. This H 2 O 2 production rate is higher than any previously reported metal-free catalyst (Table S2, ESI †). Furthermore, the OCG-800 electrode showed excellent stability over 46 h at 0.4 V vs. RHE, and the production rate in the alkaline solution was reduced by only 2.8% (460.6 (AE35) mmol g cat À1 h À1 ).

After the stability test, OCG-800 retained its original 3D morphology and defect structure, further confirming its structural stability (Fig. 3c and Fig. S13, ESI †). In addition, the current density-dependent chronopotentiometry test showed the stable operation of OCG-800 even at 60 mA cm À2 in H-cell, suggesting the potential for stable H 2 O 2 production via the electrochemical synthesis route (Fig. S14, ESI †).

The catalytic active sites for H 2 O 2 synthesis

The experimental results demonstrate that the annealing process of OCG catalysts can optimize the structures of the catalytic active sites of graphene by controlling oxygen functional groups and defect sites. To identify the catalytic active sites for H 2 O 2 production in terms of the functional groups and defects in the OCG, DFT calculations were systematically performed. As shown in Fig. 4a and Fig. S15 (ESI †), various oxygen functional groups of the graphene model system, such as epoxide, carbonyl (CQO), ether (C-O-C), carboxyl (COOH), and hydroxyl (C-OH) in the basal planes and edge sites, were considered. 35,52,53 volcano plot. 14,25 The limiting potential (U L ), which is the lowest potential at which all reaction steps are downhill in free energy, was calculated as a function of DG OOH* (color map in Fig. 4b). The black and red circles in Fig. 4b indicate the ether groups and other oxygen functional groups respectively. In the ether group, the carbon atom adjacent to oxygen is the active site (black circle in Fig. 4a), showing moderate OOH adsorption strength and a small overpotential for H 2 O 2 production. In addition, OOH was more strongly adsorbed as the local defect level near the active site increased, and C-O-C basal 1 (3V C ) showed optimal OOH adsorption for the 2e À ORR (Fig. 4b). we first suggested that the selectivity of functionalized carbon catalysts can be predicted based on the scaling relation between DG OOH* and DG H2O2* -DG O* (Fig. 4b). All functional groups except ether (red circle in Fig. 4b) had a weak oxygen adsorption tendency and showed good performance in terms of selectivity.

In particular, the carbon atoms adjacent to the CQO functional groups, such as CQO edge 1 and CQO basal 2 and 3, were found to advantageous in terms of selectivity for 2e À ORR due to the weak oxygen adsorption energy. In addition, these activity sites maintained their initial structure after oxygen adsorption (Fig. S16 and S17, ESI †). In contrast, the O* state of the ether group was excessively stable by breaking the C-O-C bond to form a stable CQO bond (Fig. 4c and Fig. S17, ESI †). This additional CQO bond formation serves as a driving force for strong oxygen adsorption to the ether. All other functional groups retained their initial configuration, even in the O* state. Furthermore, the free energy diagram showed that both CQO and C-O-C have suitable OOH adsorption energies for H 2 O 2 production, whereas C-O-C tends to proceed to 4e À ORR due to strong oxygen adsorption (Fig. S18 and Table S3, ESI †). Therefore, considering both the activity (OOH adsorption strength) and selectivity (O adsorption strength), the best performance of OCG-800 can be attributed to the combination of optimal defect structures and CQO functional groups. Moreover, the results demonstrate that the systematic control of surface defects and functional groups can be effectively applied to the design of metal-free carbon catalysts, maximizing both activity and selectivity.

Conclusions

3D graphene catalysts with controlled oxygen functional groups and defects were successfully synthesized using the one-step aerosol spray drying process. The systematic correlation between the structure and electrochemical performance of the graphene catalysts revealed the important role of oxygen functional groups and defects in the effective production of H 2 O 2 . The graphene catalyst with optimized oxygen functional groups and defect sites, OCG-800, exhibited outstanding H 2 O 2 selectivity (92-100%) over a wide potential range with excellent stability. In addition, OCG-800 achieved a high H 2 O 2 production rate of 473.9 (AE29) mmol g cat À1 h À1 at 0.4 V vs. RHE. Furthermore, DFT calculations revealed the contribution of various oxygen functional groups and defect sites to the 2e À ORR pathway through the scaling relation between OOH and O adsorption strengths. The established structure-mechanism-performance relationship of the nanostructured carbon system for 2e À ORR provides important insights for designing highly active and selective metal-free carbon electrocatalysts for the production of H 2 O 2 by finetuning oxygen functional groups and defect structures.