Understanding activity trends in electrochemical water oxidation to form hydrogen peroxide

Electrochemical production of hydrogen peroxide (H2O2) from water oxidation could provide a very attractive route to locally produce a chemically valuable product from an abundant resource. Herein using density functional theory calculations, we predict trends in activity for water oxidation towards H2O2 evolution on four different metal oxides, i.e., WO3, SnO2, TiO2 and BiVO4. The density functional theory predicted trend for H2O2 evolution is further confirmed by our experimental measurements. Moreover, we identify that BiVO4 has the best H2O2 generation amount of those oxides and can achieve a Faraday efficiency of about 98% for H2O2 production.

This paper reports the DFT predictions predict trends in activity for water oxidation towards production of H2O2 on four different metal oxides i.e., WO3, SnO2, TiO2 and BiVO4. The trend for production of H2O2 was confirmed by experiments. BiVO4 has provided the faraday efficiency about 95% for H2O2 production. The results are interesting enough to warrant publication in Nat. Commun. However, the following points should be clarified prior to publication.
(1) The recent reports on photocatalytic production of H2O2 (Mase et al., Nat. Commun. 7, 11470 (2016); ACS Energy Lett. 1, 913-919 (2016)) should be cited. (2) The formation of H2O2 and the yield should be confirmed by using different methods. The yield of O2 should also be determined. (3) The labeling experiments using H218O are also required to confirm that oxygen in H2O2 comes from O2.
Reviewer #3 (Remarks to the Author): The manuscript entitled "Understanding Activity Trends in Electrochemical Water Oxidation to Form Hydrogen Peroxide" demonstrates a theoretical study using DFT calculation to predict the activity trend of four representative metal oxides on photo-electrochemically hydrogen peroxide generation. Experimental investigation also confirmed the trend predicted by the DFT calculation. I believe the results are important for advancing our understanding of photo-electrochemical reactions. This topic is also expected to attract a broader readership as hydrogen peroxide generation is a important process in almost every water-related catalytic activity. The paper is suitable for publishing in Nature Communications after some technical issues in the manuscript should be addressed and clarified. 1. The introduction part is confusing. Information given in this part seems contradictory. For example, why does Eq.(2) represent a two-electron water oxidation process (page 2, line 49) but it is clearly a reducing process? Why is Eq. (5) an oxygen generation reaction while the product is not oxygen gas? In addition, no background was provided for the participation of OH*, O* and OOH* in this section, albeit they were frequently mentioned in the calculation section. The introduction part needs to be revised. 2. The rationale of choosing some specific crystal facets for the DFT calculation is unclear. If it is based on experimental observation, XRD and/or high-resolution TEM should be provided. 3. Information on how to evaluate the Faraday efficiency is missing. 4. The authors mentioned that the concentration of obtained hydrogen peroxide was experimentally determined by a titration process using potassium permanganate. Details (e.g., mechanism, accuracy, systematic errors, consistency comparing to the strip measurement etc.) should be provided. Please also comment on whether hydrogen peroxide can decompose during titration due to its chemical instability. 5. It is well-known that BiVO4 is photo-electrochemically unstable. The authors should comment on the stability of BiVO4 on hydrogen peroxide generation. 6. The rationale of the step-wise annealing process for material synthesis should be explained.

Point-by-Point response to the reviewers' comments (in blue)
Reviewer #1:

Comment 1
Fuku, et al. recently reported the oxidative H 2 O 2 production on eleven oxide/FTO anodes including WO 3 , SnO 2 , TiO 2 , BiVO 4 in the dark in KHCO 3 , and concluded that BiVO 4 is the best material among them (Chemistry Select, 2016, 1, 5721). They also reported BiVO 4 /FTO photoanode is very efficient for the oxidative H 2 O 2 production in KHCO3 (Ref. 12). These results are completely same as the main experimental conclusion in this manuscript.

Response:
We agree that Fuku et al. have published the pioneering work of using common metal oxides for oxidative H 2 O 2 production, and our work falls into the same general area. However, we built upon their excellent work and went beyond their experimental scopes to achieve significantly higher performance for H 2 O 2 production with complimentary theoretical efforts.
Specifically, Fuku's work (Chemistry Select, 2016) studied 11 oxides in a two-electrode system under a fixed bias of 3V. They studied BiVO 4 generation under a very large applied bias (3V-7V, vs. Pt mesh) with 1V interval, which is away from optimized condition for H 2 O 2 production. We studied 4 oxides in a three-electrode system with a refined applied bias range and intervals for measurement points, to catch the optimal condition for all these four common used metal oxide. The variation of bias may appear trivial, but it is critical for investigating the competition among one-, two-and fourelectron pathways under different bias. The elaborated study with different bias allows us not only to identify the optimal bias range for H 2 O 2 production experimental, but also to compare with theoretical results on how these competing reactions vary with bias.
Because of our more systematic efforts on the bias dependence, as well as the different surface density controlling for dark and illumination, we have achieved much higher faraday efficiencies and H 2 O 2 production rates under both conditions (see table above), in comparison to two papers by Fuku et al (which have been cited as the key references in the revised manuscript).

Comment 2
2. WO 3 , SnO 2 , TiO 2 and BiVO 4 are wide bandgap semiconductors and they act as insulator in the dark generally unless some defect level is present in the forbidden band. Therefore, the electron transfer may occur mainly on the FTO surface under the porous oxide layer or on the interface of oxide/FTO. The onset potential is defined by many factors such as coverage of insulator oxide, electro-catalytic activity at the interface of oxide/FTO, adsorption of cation/anion in the electrolyte solution. The onset potential is very changeable by the preparation methods of oxide/FTO. Both O 2 and H 2 O 2 productions occur around the onset potential. The theoretical limiting potentials of oxides in this manuscript are calculated on the pure oxide surface. The electrolyte solution for WO 3 /FTO was different from those for the others. The effect of carbonate anion is very important for the H 2 O 2 production, but this effect is not considered in the DFT calculation. In conclusion, the fundamental idea of the comparison between theoretical limiting potential and the onset potential of oxide/FTO electrode is wrong.

Response:
We agree with the reviewer that our theory cannot duplicate the exact experimental conditions and experimental results are sensitive to the experimental details, a situation that is true whenever we compare any theories with experiments. This is exactly the reason that although our experiments appear to the same as Fuku's work in studying oxides for H 2 O 2 production, our efforts on varying the applied bias are important. In addition, we strongly disagree with reviewer's view that comparing theory with experiment is not meaningful. Theory has played a vital role by providing insights on the nature of active sites and guiding the design and optimization of various catalysts (see the following references for example, Nørskov, J. K., Bligaard, T. We disagree with the reviewer's comment regarding similar onset potential for both O 2 and H 2 O 2 because thermodynamically the O 2 is evolved at 1.23 V while H 2 O 2 is evolved at 1.76 V as can be seen in Figure 1 (in the revised manuscript). This difference in potential is reflected in the onset potentials and thus, production of O 2 and H 2 O 2 occurs at different onset potential. Therefore, H 2 O 2 evolution doesn't necessarily occur at the same bias when O 2 evolves. In fact, this is the main message of our study which we tried to highlight the principle for better H 2 O 2 generation over O 2 generation. More information about this is covered in the comment Four.
We agree that the J-V curve onset is sensitive the interface of oxide/FTO, but this holds only if a large portion of FTO is exposed. In our experiments, we have optimized the coating of each oxide to minimize the exposure of FTO ( Figure S4). Under those conditions, the onset potential is mainly affected by the surface properties of those oxides for O 2 /H 2 O 2 evolution. Our dark J-V curve onsets are very similar to those previous studies on BiVO 4 [1,2] and WO 3 [3,4] for water oxidation reactions, regardless of their precursor and/or fabrication method used. Moreover, we would like to clarify that we have studied the most stable facet of the oxides theoretically, however, we do not claim a one to one correspondence between the theoretical values and the experimental J-V onset for the studied oxides. Instead, we have focused on the trends (the hollow symbols in Figure 2b in the manuscript) in this work. Furthermore, we used not only J-V onset potential, but also the actual faraday efficiency of H 2 O 2 production under different bias, to compare with our theoretical analysis (the solid symbols in Figure 2b in the manuscript), which is more meaningful since it reflects the real starting point for H 2 O 2 evolution.
We want to clarify that the reason for choosing different electrolyte solution for WO 3 /FTO is due to the poor stability of WO 3 in alkaline electrolyte (the as-prepared 1M NaHCO 3 solution has a pH 8.3). For WO 3 , we still used the NaHCO 3 solution but adjusted its pH using an acid solution to solve the stability problem. We have considered the pH effect and normalized all the potentials to the ones vs. RHE.
Finally, we agree with the reviewer on the promotional effect of carbonate anion on H 2 O 2 production, as Fuku's work suggested. This work focuses on the mechanism of H 2 O 2 generation dependency on the surface properties for different materials as well as the applied bias, while the study of the anion effect is out of the scope of this work.

Comment 3
3. High Faraday efficiency is interesting, but it is only the initial reaction in 10 min. The H2O2 will be re-oxidized or photo-or thermal-decomposed to O2 easily. The accumulation data of H2O2 and Faraday efficiency after a long time are essential to utilize the H2O2 solution. High Faraday efficiency (92% -99%) is already reported using Ge-oxyl complex photoanode, though the photocurrent properties and the accumulated amount of H2O2 was not high (T. Shiragami et al., Journal of Photochemistry and Photobiology A: Chemistry 313 (2015) 131).

Response:
We agree that the accumulation data of H 2 O 2 is important for practical H 2 O 2 production, which is a system level optimization. This work aims to address the first step: identify suitable H 2 O 2 production bias over metal oxides and understand the fundamental thermodynamic properties that affect the generation of H 2 O 2 . With those understandings, we achieved a high faraday efficiency for H 2 O 2 production over BiVO 4 . We have not studied the stability issue of H 2 O 2 and are currently working on the accumulation and long-term storage of H 2 O 2 obtained from the approach in this work.

Comment 4
4. The onset potential at 1.1 V vs RHE under solar light is not efficient compared to other previous BiVO4 photoanode. Moreover, there is no photocurrent density-potential data. Why? It is probably very poor.

Response:
We didn't include the J-V curves not because they are poor but rather the current value includes both O 2 production and H 2 O 2 production; therefore instead, we've used faraday efficiency of H 2 O 2 that is more meaningful for H 2 O 2 production. Per reviewer's request, we have added the J-V curve and the amount of H 2 O 2 produced here. The onset potential of our BiVO 4 is well in the normal range as reported previously (0.47V-0.8V vs. RHE) [1,2], and since this is in bicarbonate, the performance is even higher than the previous reports [1,2,5].
To address reviewer's point the following figure has been added as the new Figure S5 in the revised SI, and the following contents were added in the section "H 2 O 2 production on BiVO 4 under different conditions" in the main text: "The J-V curve and the measured H 2 O 2 generation rate under illumination for this 9-layer BiVO 4 are shown as Figure S5".

Comment 5
5. In Fig. 4, the Faraday efficiency was significantly changed by the potential. The author should explain the reason. Because the hole on the valence band in photoanode is hardly affected by the potential. The Faraday efficiency was also changed by the potential under the dark condition in Fig. 3, and the optimum potential for the maximum Faraday efficiency was different in each anodes. Why?

Response:
The reviewer's comment is exactly the key point for this work: H 2 O 2 production efficiency is bias dependent. This is a major difference between our work (varying bias) and Fuku's work (mainly focus on fixed bias). Having holes on the valence band is only a necessary, not sufficient, condition for H 2 O 2 production. Optimizing H 2 O 2 production needs to consider the competing O 2 generation at low bias and OH* formation at high bias, which leads to the bell shape faraday efficiency for H 2 O 2 vs. applied bias. If only a rough and broad range is considered, the optimal condition may be missed since faraday efficiency under both higher or lower potential is low. Therefore, a refined potential range should be applied guided by the theoretical study. The competition between three reactions also depends on the materials. Our theoretical results ( Figure 1) help us to understand the activity of different oxides in their relevant bias range, which is a very valuable contribution.

Response:
We thank the reviewer for the correction and corrected the text accordingly.

Reviewer #2
This paper reports the DFT predictions predict trends in activity for water oxidation towards production of H 2 O 2 on four different metal oxides i.e., WO 3 , SnO 2 , TiO 2 and BiVO 4 . The trend for production of H 2 O 2 was confirmed by experiments. BiVO 4 has provided the optimal faraday efficiency for H 2 O 2 production. The results are interesting enough to warrant publication in Nat. Commun. However, the following points should be clarified prior to publication.

Response:
We thank the reviewer for this comment and have cited the mentioned literatures. These two papers have given great contribution on the use of cobalt (II) chlorin complex as an effective catalyst for H 2 O 2 production from O 2 reduction.

Comment 2
(2) The formation of H 2 O 2 and the yield should be confirmed by using different methods. The yield of O 2 should also be determined.

Response:
We thank the reviewer for this comment and to address this point we have used KMnO 4 solution to further confirm the H 2 O 2 yield. The details about the mechanism and accuracy on it are described in the "Methods" section as follows: "The generated H 2 O 2 concentration was further confirmed with a titration process by using potassium permanganate (KMnO 4 , ≥99.0%, Aldrich). The permanganate ion has a dark purple color, and the color disappears during titration as the MnO 4 being consumed based on the following equation: (6) In this work the sulfuric acid (H 2 SO 4 , Acros Organics) was used as the H+ source. We measured five H 2 O 2 solution samples with different degree of dilution from a same initial concentration 100ppm, by both the standard strips and the permanganate titration are shown as Figure S7 (also shown below), which shows that two methods basically agree with each other, confirming the actuary of the H 2 O 2 concentration measurement." The faraday efficiency for the yield of O 2 is also determined and added in the supplementary information as Figure S6 (shown in the Comment Three). In addition, the following contents are added in the sentence "H 2 O 2 production on BiVO 4 under different conditions" in the main text: "In addition, we have measured the evolved gaseous O 2 and H 2 and the measured FEs are shows as Figure S6. Figure S6 shows

Response:
We thank the reviewer for this comment. To answer reviewer's point we would like to mention that in this work we've focused on the study of H 2 O 2 generation from H 2 O oxidation at the working electrode, other than from O 2 reduction at the counter electrode. We used a Nafion membrane to separate the working and counter and found that the yield of H 2 O 2 at the counter is zero. In addition, the measured faraday efficiency for hydrogen is approximately 100% in a broad potential range, as shown in the Figure S6 (also shown below), which indicates all the electrons have been used to reduce H + to generate H 2 , instead of reducing O 2 .

Reviewer #3
The manuscript entitled "Understanding Activity Trends in Electrochemical Water Oxidation to Form Hydrogen Peroxide" demonstrates a theoretical study using DFT calculation to predict the activity trend of four representative metal oxides on photo-electrochemically hydrogen peroxide generation. Experimental investigation also confirmed the trend predicted by the DFT calculation. I believe the results are important for advancing our understanding of photo-electrochemical reactions. This topic is also expected to attract a broader readership as hydrogen peroxide generation is a important process in almost every water-related catalytic activity. The paper is suitable for publishing in Nature Communications after some technical issues in the manuscript should be addressed and clarified.

Comment 1
The introduction part is confusing. Information given in this part seems contradictory. For example, why does Eq.(2) represent a two-electron water oxidation process (page 2, line 49) but it is clearly a reducing process? Why is Eq. (5) an oxygen generation reaction while the product is not oxygen gas? In addition, no background was provided for the participation of OH*, O* and OOH* in this section, albeit they were frequently mentioned in the calculation section. The introduction part needs to be revised.

Response:
We thank the reviewer for this comment and pointing to typo error in the numbering of the equations in the text. To address this point, we have corrected the numbering of the equations in the text. We have also added the following sentence in the last line of the page 2 and the top of the page 3 to address reviewer's point regarding the background for OH*, O* and OOH*.

Comment 2
The rationale of choosing some specific crystal facets for the DFT calculation is unclear. If it is based on experimental observation, XRD and/or high-resolution TEM should be provided.

Response:
We thank the reviewer for this comment. To answer reviewer's point we would like to mention that we have theoretically investigated the activity of all the low-index surfaces ((001), (101), (011), (111), (110), (100) and (010)) of fm-BiVO 4 (" Figure I" below). Among all the examined surfaces we found (111) and (110) as the most active surfaces for two-electron water oxidation (" Figure II" below). In fact (111) facet is among the most stable facets in terms of formation energies and it is highly likely to be an exposed facet in the crystal structure of the BiVO 4 . The details of this analysis is out of the scope of this study and therefore we removed it from the SI in the first place. These results will be published separately. However, to address reviewer's point we have added the following sentence in the theoretical section in page 3.
"We only focus on the (111) surface, which has been shown theoretically and experimentally to be stable and exposed in the BiVO 4 crystal structure. 33 "

Comment 3
Information on how to evaluate the Faraday efficiency is missing.

Response:
We thank the reviewer for this comment. To address this point we have added the following explanation in the revised "Methods" section in the main text: "The FE for H 2 O 2 production (%) is calculated by (7) Where the theoretical generated H 2 O 2 is equal to the total number of electrons divided by two (in mol);"

Comment 4
4. The authors mentioned that the concentration of obtained hydrogen peroxide was experimentally determined by a titration process using potassium permanganate. Details (e.g., mechanism, accuracy, consistency comparing to the strip measurement etc.) should be provided. Please also comment on whether hydrogen peroxide can decompose during titration due to its chemical instability.

Response:
We thank the reviewer for this comment. To address this point, we added the requested information as well as the related discussion in the revised "Methods" section: "The generated H 2 O 2 concentration was further confirmed with a titration process by using potassium permanganate (KMnO 4 , ≥99.0%, Aldrich). The permanganate ion has a dark purple color, and the color disappears during titration as the MnO 4 being consumed based on the following equation: (6) In this work the sulfuric acid (H 2 SO 4 , Acros Organics) was used as the H+ source. We measured five H 2 O 2 solution samples with different degree of dilution from a same initial concentration 100ppm, by both the standard strips and the permanganate titration are shown as Figure S7 (also shown below), which shows that two methods basically agree with each other, confirming the actuary of the H 2 O 2 concentration measurement." In addition, from the consistency between two methods, as well as the linear relationship shown for the five points it can be seen within the 0-100ppm range (the range used in this work) the permanganate titration doesn't affect much on the H 2 O 2 stability.

Comment 5
5. It is well-known that BiVO 4 is photo-electrochemically unstable. The authors should comment on the stability of BiVO 4 on hydrogen peroxide generation.

Response:
We agree that the photoelectrochemical stability of BiVO 4 is an issue due to the V 5+ dissolution into solution [6,7], when pH is far from neutral conditions. However, as reported recently [8], BiVO 4 is quite stable in the near neutral region and our bicarbonate solution has a near neutral pH value (8.3 as measured). In fact, Fuku's work (Chem.Commun., 2016, 52 ,5406) shows that the WO 3 /BiVO 4 has a very good stability for H 2 O 2 production when also measured in bicarbonate solution. Based on this we've added some comments on the stability of BiVO 4 for H 2 O 2 production in the text right before the "Conclusion" section: "…the photoelectrochemical stability of BiVO 4 is known to be an issue when the electrolyte is far from neutral conditions because the V 5+ tends to dissolve into solution. 42,43 However, we used the bicarbonate electrolyte with a measured pH value of 8.3, so BiVO 4 is relatively stable in this near neutral region. 44 "

Comment 6
The rationale of the step-wise annealing process for material synthesis should be explained.

Response:
The step-wise annealing process is commonly used when fabricating metal oxide films from the solgel process. A rapid temperature rise during annealing will cause the rapid boiling of the solvent, leading to poor film morphology. To address this point, we have added the following information in the "Methods" section.
"Similar step-wise annealing process was commonly used for metal oxide fabrication, and the purpose is to slowly evaporate the solvent to achieve a better film morphology."