Electrochemical oxygen reduction to hydrogen peroxide at practical rates in strong acidic media

Electrochemical oxygen reduction to hydrogen peroxide (H2O2) in acidic media, especially in proton exchange membrane (PEM) electrode assembly reactors, suffers from low selectivity and the lack of low-cost catalysts. Here we present a cation-regulated interfacial engineering approach to promote the H2O2 selectivity (over 80%) under industrial-relevant generation rates (over 400 mA cm−2) in strong acidic media using just carbon black catalyst and a small number of alkali metal cations, representing a 25-fold improvement compared to that without cation additives. Our density functional theory simulation suggests a “shielding effect” of alkali metal cations which squeeze away the catalyst/electrolyte interfacial protons and thus prevent further reduction of generated H2O2 to water. A double-PEM solid electrolyte reactor was further developed to realize a continuous, selective (∼90%) and stable (over 500 hours) generation of H2O2 via implementing this cation effect for practical applications.

H ydrogen peroxide (H 2 O 2 ) is ranked as one of the top 10 most energy-intensive chemicals in the chemical manufacturing bandwidth study by the Advanced Manufacturing Office in Department of Energy 1 . It is currently manufactured industrially by the energy-and waste-intensive anthraquinone cycling process [2][3][4] , which consumes a primary current typical energy of~13,000 Btu/lb (8.1 kWh/kg) without taking into account the H 2 /O 2 feedstocks 1 . Electrochemical synthesis of H 2 O 2 via the 2e − oxygen reduction reaction (2e − -ORR), where the O 2 molecule is electrochemically reduced to H 2 O 2 via a two-electron (2e − ) pathway, provides a promising energy-efficient and low-waste alternative [2][3][4][5][6][7] . Recent efforts have been mostly focused on developing catalysts in alkaline solutions, in which small overpotential and high selectivity of 2e − -ORR toward H 2 O 2 have been comparatively easy to achieve on lowcost materials such as carbon [8][9][10][11][12][13][14][15][16][17][18] . However, in alkaline solutions, H 2 O 2 is deprotonated (pK a > 11) and easily degraded 19 . Moreover, for practical electrolyzers such as membrane electrode assembly (MEA), catalysts developed in alkaline solutions need to be applied on an anion exchange membrane (AEM), which is typically not as stable as its counterpart of proton exchange membrane (PEM), e.g., Nafion 4 , especially operated under air. In addition, with stronger oxidation ability in acid, the acidic H 2 O 2 solution shows a wider range of applications and greater demand 6,20 , which strongly motivates studies in highperformance electrochemical generation of H 2 O 2 in acidic media.
Till so far, there are only a few known noble metal catalysts, including Pt-and Pd-based catalysts, demonstrated to be selective and stable for 2e − -ORR in strong acids [21][22][23][24] , but their high cost and toxicity of heavy metals (in the case of PtHg alloy 22,23 ) could limit their applications in large-scale H 2 O 2 generation. Some lowcost catalysts such as carbon materials may also show good H 2 O 2 selectivity in acids within small overpotential and small current density regions (typically less than 10 mA cm -2 ) [25][26][27][28][29] , but their H 2 O 2 selectivity and stability were dramatically dropped when an industrial-relevant current was reached 5,27,30 . In acidic media, carbon catalysts present sluggish ORR kinetics and typically require a large overpotential (>300 mV) to initiate the ORR reaction, and consequently, a large negative cathodic overpotential is required to deliver a high current density [26][27][28][29] . While the carbon surface might intrinsically prefer a 2e − -ORR pathway due to their relatively weak binding with oxygen intermediates as demonstrated by several previous studies 26,27,30,31 , such negative overpotential could further push the ORR reaction all the way down to H 2 O with significantly decreased H 2 O 2 selectivity and production rate especially in acids. This is because under negative potentials in acids, the catalyst surface accumulates concentrated protons that are prone to further reduce the locally generated H 2 O 2 molecules to H 2 O (H 2 O 2 + 2e − + 2H + = 2H 2 O). Therefore, diluting the local proton concentration and minimizing the electrochemical dissociation of asproduced H 2 O 2 to H 2 O could be one promising strategy for resolving this H 2 O 2 selectivity-activity dilemma, and delivering industrial-relevant H 2 O 2 production rate in acidic solution while maintaining good H 2 O 2 selectivity 32,33 (Fig. 1).

Results
The alkali metal cation effect on acidic H 2 O 2 generation in flow cell. We first employed a standard three-electrode flow cell reactor to investigate the cation effect towards H 2 O 2 generation in acid (Fig. S1), allowing to evaluate our hypothesis in a more practical environment and produce H 2 O 2 under higher current densities compared to the traditional RRDE setup (Supplementary Note 1, Fig. 2a). The commercially available carbon black catalyst (BP2000) with a high surface area was used as a model ORR catalyst in this study (Fig. S2). Its intrinsic H 2 O 2 activity and selectivity were first evaluated in 0.1 M H 2 SO 4 electrolyte (pH = 0.96) (Fig. 2b, c). We observed that while the H 2 O 2 FE of carbon black catalyst in acid remained relatively good (~70%) under small current regions, it started to decrease dramatically once the current density is over 100 mA cm −2 with very negative applied potentials (Fig. 2c) Fig. 3g). After introducing a trace amount of Na + additive (5 mM Na 2 SO 4 ) into the acidic electrolyte, while the ORR activity did not show that much difference (Fig. 2b), the H 2 O 2 FE was significantly improved especially under high current densities. As shown in Fig. 2c, in the potential range of the ORR onset, the impact of Na + (5 mM Na 2 SO 4 ) toward H 2 O 2 is negligible; while in the range of large overpotentials at high current densities, the small amount of Na + helped the carbon black catalyst to hold a high H 2 O 2 FE plateau of over 80% until 200 mA cm −2 , suggesting a more than doubled FE compared to that in pure acid (Fig. 2c).
The measured trends of H 2 O 2 FE show that the promotion effect of Na + is more pronounced at high current density toward H 2 O 2 formation. We understand that Na + and protons are both positively charged ions and will be attracted towards the ORR  catalyst surface to form the electrochemical double layer under negative potentials 34 . We would assume that, Na + ions could be more competitive than protons to be aligned along the electrochemical double layer, which dramatically reduces the local proton concentration and thus protects the generated H 2 O 2 from further reductions coupling protons and electrons (discussed in the following sections in detail). To further amplify the cation effect and drive the O 2 -to-H 2 O 2 production at even higher current densities in acid (>200 mA cm −2 ), we gradually increased the Na + cation concentration in the electrolyte. As shown in Fig. 3a Table S1). b Effect of 0.005 M Na 2 SO 4 on the total activity over carbon catalysts BP2000 in a flow cell. The cell voltages were 85% iR compensated. The inset is the cross-section SEM image of the electrode. The scale bar is 200 µm. c The comparison of H 2 O 2 FE and production rate over carbon catalyst BP2000 with/without 0.005 M Na 2 SO 4 in a flow cell. The Na + serves as a promoter for the production of H 2 O 2 through 2e − -ORR. The H 2 O 2 FE and production rate are both increased, especially at the high current range.
The low threshold of the alkali metal cation concentration towards promoting H 2 O 2 generation puts forward new requirements for the purity of the electrolyte during ORR to H 2 O 2 tests in acids. For the traditional electrolytic ORR process, the Na 2 SO 4 is widely used as the anolyte to balance the electrochemical reaction. However, even far away from the cathode side and separated by ion exchange membranes, we found that the Na + in the anolyte can still penetrate the PEM and move to the cathode chamber, and thus significantly improve the H 2 O 2 FE of ORR at the cathode (Fig. S8). Therefore, it is highly recommended to perform the ORR reaction using the same acidic electrolyte to avoid any cross-over cation contaminations which could significantly improve H 2 O 2 performance in acids.
The promotion effect is not only limited to the Na + . The H 2 O 2 production rate can also be enhanced by using other alkali metal ions. Figure 3d shows the I-V curves for four different alkali metal cations in each sulfuric acid electrolyte with a concentration of 0.01 M (0.005 M X 2 SO 4 , X = Li, Na, K, Cs). As compared with the pure H 2 SO 4 electrolyte, while the ORR activities were slightly improved, significant improvements in H 2 O 2 FE were observed for all the cations (Fig. 3d-f). The FE and production rates of H 2 O 2 are relatively unaffected by the size of the alkali metal cations in the electrolyte. As differences in total current density exist for electrolytes containing different cations, the H 2 O 2 production rate provides a better representation of trends in product formation rates than FEs. As shown in Fig. 3f, with 0.005 M X 2 SO 4 as the additive, all the alkali metal cations are able to drive the O 2 -to-H 2 O 2 reaction efficiently with high production rates, and the differences induced by different cations are relatively marginal. Nevertheless, we find that the promotion effect is only limited to IA alkali metal cations (such as Li + , Na + , K + , Cs + ), while the other cations (including the IIA cations such as Mg 2+ , Ca 2+ and IIIA cations such as Al 3+ ) decrease the H 2 O 2 FE dramatically. This might due to the local environment change from acid to alkaline induced by the cation additives during ORR (will discuss the details in the simulation part). The alkaline local environment could induce the formation of solid metal hydroxide on the catalyst surface and block the ORR reaction, decreasing the H 2 O 2 FE and production rate (Fig. S9).
It was also exciting to find out that the cation additives not only promote the H 2 O 2 selectivity but also improve the long-term operation stability, which is another important target for practical In pure acidic electrolyte, the H 2 O 2 FE rapidly dropped to less than 10% within 2 h. As a sharp contrast, with the presence of Na + cations, the potential and H 2 O 2 FE showed negligible changes for over 20 h. We suppose the improved activity and stability are induced by the alkalization of the local environment induced by the cation additives during ORR. During the ORR process, the solvated alkali metal cations, compared to concentrated protons in acids, could preferentially be attracted to the catalyst/electrolyte interface, which gives rise to a local alkaline environment (will discuss the details in the simulation part). At alkaline conditions, the carbon-based catalyst typically shows higher activity and better stability compared to acidic conditions [8][9][10] . We then found this cation promotion effect has broad applicability to different catalysts and acidic electrolytes. With a small amount of Na 2 SO 4 as the additive in 0.1 M H 2 SO 4 ( Fig. 4), high H 2 O 2 FE under high production rates can be achieved on various carbon catalysts including carbon nanotube (CNT), reduced graphene oxide (rGO), XC-72, and the carbonbased single-atom catalysts (e.g., Zn-N-C). This promotion effect is independent of the catalyst structures, surface morphology or surface area (Figs. S10-S12). In addition, this cation impact on 2e − -ORR also applies to other acidic electrolytes such as HClO 4 , and the Na + sources can also come from other salts such as NaHSO 4 (Figs. S13, S14).
Mechanistic understanding of the cation effect on H 2 O 2 selectivity. To further explore our aforementioned hypothesis and understand the cation effects on 2e − -ORR selectivity, we carried out constant-potential ab-initio molecular dynamics (AIMD) simulations 35 based on DFT. In the simulations, we use a typical potential V RHE (potential relative to RHE) = −1 V at which the cation effect is most prominent. The hydrogen evolution reaction was not considered because only a trace amount of H 2 byproduct (from H 2 evolution at large overpotentials) was detected from the cathode side (Fig. S15). We consider the acid condition and use pH = 0 in our simulations. Since alkali metal cations are not likely to directly participate in ORR, we explore two aspects of the cation effects: (1) how alkali metal cations affect the distribution of the protons, (2) how the redistribution of protons will influence the selectivity of O 2 reduction to H 2 O 2 . For the first question, we adopted a model with relatively thick water layers (equivalent to 6 ice layers) on (6 × 6) periodic graphene (Fig. S16), and then put a certain number of cations/protons to represent the electrode environment under the lower potential like V RHE = −1 V. For the second question, we use a thinner model (~4 ice layers) with single-vacancy to represent the reaction site, and then use slow-growth method 36 to evaluate the reaction barrier under different conditions). The details of the simulations can be found in the experiment section in supporting information (SI). As shown in Fig. 5a, b, both Na + cations drift towards the surface in molecular dynamics. This is not unexpected considering that under the low potential V RHE = − 1 V, the surface is charged by~3e − . Such a fast drift may have two consequences: firstly, the local concentration of Na + can be much higher than that in the bulk; secondly, as a charge carrier that compensates the net charge of the substrate, Na + may compete with protons which is the major charge carrier in the acid electrolyte when Na + cations are not added. Indeed, as shown in Fig. S17, when Na + cations and protons co-exist near the interface, the cations compete with the protons by repelling the proton away from the surface. After 3 picoseconds of AIMD, both cations stay~4.5 Å from the surface, while the protons end at~8 Å from the surface. These results clearly suggest that the cations, which can be enriched by the attraction of the negatively charged surface, can strongly repel the local protons, and thus dramatically reduce the local proton concentrations.
Then we consider how the selectivity of the 2e − -ORR is affected by the presence of Na + cations and the reduced local proton concentrations. We understand that carbon catalysts typically bind oxygen intermediates weakly and thus present an intrinsic selectivity towards H 2 O 2 , which can be seen from our above experimental results (high H 2 O 2 selectivity under low currents in Fig. 2c) as well as   Fig. 5c). As shown in Fig. 5c, the H 2 O 2 reduction activity was greatly suppressed, especially under large overpotentials, when Na + was added, suggesting that the introduction of cations can greatly inhibit the H 2 O 2 dissociation to H 2 O under reductive potential environments. Please be noted here that the only possible side reaction, the hydrogen evolution reaction, was also taken into consideration when measuring the H 2 O 2 reduction currents (Fig. S15). This suppression effect can be further validated from our simulation results. Clearly, as shown in Fig. 5d Another possible factor for improved H 2 O 2 selectivity is that the cations could suppress the dissociation process of peroxide intermediate (*OOH) during ORR due to local proton depletion. To explore this hypothesis, we further used a slow-growth approach based on AIMD to evaluate the reaction barriers of both 2e − and 4e − paths under different proton concentration conditions. As displayed in Fig. 5d, when a proton is present, both the 2e − path and 4e − paths have an extremely low barrier and take place spontaneously at 300 K. In contrast, the absence of protons increases the *-O breaking (2e − path) barrier to 0.136 eV, in comparison to 0.167 eV for O-OH breaking (4e − path), suggesting 2e − pathway being favorable by exp ((0.167-0.136)k B T) = 3.36 times than the 4e − pathway and thus enhanced H 2 O 2 selectivity. Therefore, the local absence of protons, a result of cation accumulation near the surface, can strongly enhance the H 2 O 2 selectivity in acids.
We also evaluate the effect of Mg 2+ on the 2e − -ORR by replacing two Na + by one Mg 2+ , and running the AIMD simulations similar to the case of Na + (Fig. 5b). As shown in Fig. S27, Mg 2+ does not show the same effect as Na + . This is likely due to two reasons: (1) Mg 2+ is more efficiently screened as the Mg 2+ bonds stronger with O of H 2 O than the case of Na + , which is evidenced by the significantly shorter distance between Mg and O (averaged distance d Mg-O = 2.05 Å vs. d Na-O = 2.55 Å); (2) the mole concentration of Mg 2+ is only half of that of Na + in the electrical double layer and the electrostatic repelling decays in the form of 1/r, so there is more "screened" space for H + in the electrical double layer.
Practical generation of H 2 O 2 using cation exchange membrane solid electrolyte reactor. Obtaining good H 2 O 2 selectivity and activity in acidic ORR is a prerequisite for practical implementations of membrane electrode assembly (MEA) reactor using reliable and well-established PEM such as Nafion (sulfonated tetrafluoroethylene based fluoropolymer-copolymer membrane). However, till so far, only noble metal catalysts such as PtHg, PdHg or PtP 2 nanocrystals could deliver reasonable H 2 O 2 selectivity and stability in proton exchange MEA device 22,23,37 . This observed notable promotion effect of cations on low-cost and non-toxic carbon catalysts in acidic H 2 O 2 generation therefore provides us with a great opportunity to deliver practical H 2 O 2 activity, selectivity, and stability. Our basic assumption is that, as the alkali metal cations can move across the PEM, they may help to regulate the local environment of the catalyst/membrane interface for better H 2 O 2 activity and selectivity. To explore how to successfully make use of this cation tuning effect, we first evaluated its applicability in a traditional PEM-MEA cell configuration (Figs. S22-S24). First, it is well within our expectation that the traditional PEM-MEA cell using commercial carbon black catalyst with 0.1 M H 2 SO 4 as the anolyte, presented negligible H 2 O 2 selectivity, due to the high proton flux at the catalyst/ membrane interface (Fig. S22). However, we found out that even with the addition of cations in acids (Fig. S23), or directly using Na 2 SO 4 solution as the anolyte (Fig. S24), no obvious improvements were observed. This is because on the anode side, a significant number of protons will be generated locally at the catalyst/membrane interface during the oxygen evolution reaction and then immediately transported across the membrane to the cathode side, suppressing the possibility of Na + ion transportation from the bulk anolyte towards the cathode to regulate the interfacial environment of improved 2e − -ORR. The low H 2 O 2 selectivity in PEM-MEA drove us to design a new cell configuration to employ this cation effect for continuous production of H 2 O 2 in a practical way. Instead of the traditional PEM-MEA cell design, here we developed a SE reactor with three chambers separated by two PEMs to fully implement this cation effect for high-performance H 2 O 2 generation (Fig. 6a and Fig. S25). Specifically, the cathode (carbon black) and anode (IrO 2 ) of our device are catalyst-coated GDL electrodes, which were separated by a thin SE layer sandwiched by two identical PEMs (Nafion-117). The cathode side was continuously supplied with a mixture of O 2 stream and water flow for 2e − -ORR, while the anode side was circulated with H 2 O for water oxidation. In the middle chamber, a SE layer consisting of porous polymer ion conductors was used to minimize the iR-drop between cathode and anode 38,39 . A dilute cation solution flows through this SE layer to introduce the cation effects on the cathode side 2e − -ORR. Please be noted here that without this SE layer, the cell voltage was significantly increased due to the increased cell resistance between the cathode and anode (Fig. S26). Under a negative reduction potential, the cations in the middle SE chamber are driven by the electrical field to penetrate the PEM toward the cathode surface and thus regulate the local environment at the catalyst/PEM interface to promote H 2 O 2 generation (Supplementary Note 2 and Fig. S27). The H 2 O 2 molecules formed at the cathode side are then efficiently brought out via the oxygen and DI water flow stream. Meanwhile, protons generated from water oxidation at the anode penetrate the righthand side PEM and move into the middle chamber to compensate for the charge.
The I-V curve of our 4-cm 2 three-chamber PEM SE cell flowing 0.03 M Na 2 SO 4 solution in the middle layer is plotted in Fig. 6b. Of note that the concentration of Na 2 SO 4 can be varied to higher values. Our target is to realize the high production rate of H 2 O 2 while minimizing cations' consumption for practical demonstrations. Therefore, a 0.03 M Na 2 SO 4 solution is adopted to regulate the interfacial environment of improved 2e − -ORR in our SE cell. A mixture of O 2 gas flow (180 sccm) and DI water flow (1.8 mL·min −1 ) was supplied to the cathode. The flow of DI water is to efficiently bring out the generated H 2 O 2 . By ramping up the overall current density, the cell voltage of the SE reactor gradually increased. The H 2 O 2 FE remained over 85% across the entire cell voltage range, with a maximum of 96% at 5 and 20 mA cm −2 (Fig. 6b), much higher than the traditional MEA cell configuration. In comparison, the reactor without alkali metal cations requires a higher potential to deliver and shows much lower H 2 O 2 FE, practically at high current densities (Fig. S28). The crosse-over Na + from the middle SE layer to the cathode plays a key role in determining the H 2 O 2 FE (Fig. S27).
The electrolysis stability is always one of the most important but challenging parts in practical applications. Benefiting from the stable material properties in carbon black and PEM as well as reliable cation effects, our SE cell with the double-PEM configuration presents excellent long-term stability in producing H 2 O 2 . The SE reactor stability was evaluated by holding a 50 mA·cm −2 cell current density (200 mA total current). As shown in Fig. 6c and d, when supplying dilute Na + ions in the SE layer , a continuous generation of H 2 O 2 solution on the cathode side can be stably operated for over 500 h with no degradations in product FE (~90%). As a sharp contrast, in the absence of cations, the H 2 O 2 FE started with less than 60% and rapidly dropped to less than 10% in 6 h, which can even be recovered to~90% when cations were later introduced (Fig. S29), clearly suggesting this prominent cation effect. The observation indicates that the carbon catalyst still works well, and this is not the reason for FE degradation. We suppose the FE degradation is because of the accumulation of local H + to high concentrations with the extension of reaction time, which accelerates the further reduction of H 2 O 2 to H 2 O and decrease the H 2 O 2 selectivity.
A continuous supply of cation solutions in the SE layer could limit the device's real applications due to the following two reasons. First, it could result in a significant consumption of cations as most of them are flowing out of the SE layer and only part of them crossed to the cathode chamber. Second, the generated H 2 O 2 solution in the cathode would be slightly alkaline due to the Na + ion crossover from the SE layer (resulting in slight acidity in the SE layer downstream flow). To further explore the high potential of our SE cell for practical uses, we operated the cell by circulating the cation solution from the SE layer into the cathode side, and then back to the SE layer for H 2 O 2 accumulation. The outlet solution in the middle chamber was mixed with oxygen flow and supplied to the cathode to produce H 2 O 2 . Then the as-produced solution containing H 2 O 2 and cations was cycled back into the middle chamber (Fig. 6e). By doing so we could continuously reuse the cations by circulating them back to the SE layer for a closed system without the need for a continuous cation stream supply. Also, the excessive OH − groups generated from the cathode would be neutralized by exactly the same number of excessive protons in the SE layer. Our target is to accumulate the H 2 O 2 concentration to~5000 ppm in a 250 mL solution (containing only 60 mM of Na + ) for each operation cycle via maintaining a 50 mA cm −2 cell current. As shown in Fig. 6f, the H 2 O 2 concentration continuously increased to~5000 ppm in about 13 h. The cell can be operated for more than 200 h with negligible degradations. During this stability test, a total of 3.7 L 5000 ppm H 2 O 2 solution was obtained. We also observed that, in each operation cycle the H 2 O 2 FE slightly decreases with increased H 2 O 2 concentration, which may be due to the H 2 O 2 self-decomposition, further reductions on the cathode, and/or the crossover oxidation on the anode side. For each operation cycle, the H 2 O 2 concentration can reach up tõ 0.15 M, which is 5 times of the Na 2 SO 4 additive in the final product. A higher concentration of H 2 O 2 solutions can also be produced by extending the operation time while maintaining the current and FE. As a result, a high concentration of 20,000 ppm H 2 O 2 solution was achieved in 17 h by circulation of the 50 mL water at the cathode side (Fig. S30). Based on the above promising H 2 O 2 activity, selectivity and especially the durability, and since all the reactor components, including catalysts, membrane, and the polymer SE are all commercially available, our PEM-based H 2 O 2 SE cell with cation promotion effect has a great potential for future's practical applications.
To conclude, we presented a cation-regulated interfacial engineering approach to improve the catalytic performance of O 2 reduction to H 2 O 2 at industrial-relevant rates in strong acidic media. By adding a small number of alkali metal cations in acid solutions, the selectivity and stability of H 2 O 2 generation using commercial carbon black catalyst can be dramatically improved, especially under large ORR current densities (over 400 mA cm −2 ). Modeling of reaction and local environment suggest that the cations could preferentially be attracted to the catalyst/electrolyte interface, showing a "shielding effect" to squeeze out the catalyst/ electrolyte interfacial protons and thus prevent further reduction of generated H 2 O 2 to water. A double-PEM-based reactor was further developed for continuous production of H 2 O 2 solution. By using only 0.03 M Na 2 SO 4 as the cation source, a promoted H 2 O 2 FE (~90%) and stability (>500 h) were achieved. In light of this performance, this would be a promising demonstration of the use of renewable electricity for the continuous generation of H 2 O 2 through O 2 reduction at a more practical scale. This cation "shielding effect" could also be used in other electrocatalytic reactions such as selective CO 2 reduction into fuels and chemicals or N 2 reduction into ammonia.   Preparation of electrodes. Typically, 40 mg conductive carbon black (BP2000) and 80 μL of Nafion (527084-25 mL) binder solution was mixed with 4 mL of 2-proponal (Sigma-Aldrich) and 1 ml methanol. After sonication in ice water for 30 min, the obtained homogeneous ink was air-brushed onto a 5 × 5-cm 2 gas diffusion layer (GDL, Sigracet 28 BC, Fuel Cell Store) electrode at room temperature. Then the prepared electrode was dried in a vacuum at room temperature for 24 h before use. The procedure for preparation electrodes with other catalysts is same as that of carbon black BP2000. The reduced graphene oxide (rGO) catalyst was pretreated using HCl and acetone to remove impurities before making the catalyst ink.

Experiment
Activation of the Nafion-117 membrane.  Fig. 6a and Fig. S25. The cathode side was supplied with an oxygen/water mixture of 180 sccm of O 2 gas and 10.8 mL min −1 of DI water. The gas flow rate was controlled by a mass flow meter (MFC) and the water flow rate was controlled by a syringe pump.
Electrochemical H 2 O 2 dissociation. The electrochemical H 2 O 2 dissociation was conducted in a customized gas-tight H-type glass cell at 25°C. Before the experiment, the glass cell was carefully cleaned by boiling the cell in a mixture of H 2 SO 4 : H 2 O 2 (3:1) for 1 h. After being thoroughly cleaned by DI at room temperature, the cell was further boiled in DI water for another 1 h to totally remove H 2 O 2 residual.
The electrochemical H 2 O 2 dissociation was conducted with a BioLogic VMP3 workstation. The cathode electrode was prepared by spray coating carbon black (BP2000) on a GDL (Sigracet 28 BC, Fuel Cell Store), and the anode electrode was a carbon rod. The cathode electrode was fixed using a gold-coated clip and the exposed geometric surface area of each electrode was 1 cm 2 . Since the traditional clip made of iron can be easily dissolved to Fe 2+ /Fe 3+ by acid and may contribute to the H 2 O 2 dissociation, the gold-coated clip is necessary to avoid the dissolution of the iron clip during the process. The working and counter electrodes were parallel and separated by a clean PEM. The mixture of 0.2 M H 2 O 2 + 0.1 M H 2 SO 4 was used as catholyte, 0.1 M H 2 SO 4 was used as anolyte, and each volume of the electrolyte was 25 mL. A gas dispersion frit was used at the cathode chamber to provide vigorous electrolyte mixing. The cathode chamber was supplied with Ar gas (99.999% Praxair) at a rate of 20 sccm for 30 min before the electrochemical measurements. During electrolysis, continuous Ar flow was supplied throughout the experiment and the gas outlet was connected to a gas chromatograph (GC, Shimadzu GC-2014 GC) for detection of the H 2 gas. The H 2 amount was quantified by a thermal conductivity detector. After the electrochemical decomposition, the amount of H 2 O 2 remaining was determined by using the standard potassium permanganate titration process.
Theoretical simulation. The Vienna Ab initio Simulation Package (VASP) 40,41 , together with the VASPsol patch 42 , was employed to perform slow-growth calculation. The constant potential along the MD track is realized by adjusting the number of electrons onthe-fly, as described in our previous work 35 . Perdew-Burke-Ernzerhof (PBE) functional 43 together with D3 van der Waals correction 44 were employed in most of the calculations. The cutoff energy of the plane-wave basis is 400 eV in the relaxation while 300 eV was used in the MD simulations. Gamma-only MD calculations were done for the thick model in Fig. 5a, b (water layers equivalent to 6 ice layers), while 3 × 3 × 1 Gamma-centered k-mesh was used in MD simulations using the thinner defectgraphene model shown in Fig. S17. One proton is added into the 45 H 2 O molecules to simulate the pH = 0 condition. We choose the bond length (*-O or O-OH) to be the reaction coordinate (ε). Time step in MD was set to be 0.5 femtoseconds, and the ∂ε in slow-growth method 35 was set to be 0.0004 Å. A Nose-Hoover thermostat 45 was used to keep temperature (statistically) constant at 300 K. The proton is determined as the H that is farthest from the central O among two other H atoms of a H 3 O + for snapshots that are evenly distributed along the AIMD track.