Colloidal zinc oxide-copper(I) oxide nanocatalysts for selective aqueous photocatalytic carbon dioxide conversion into methane

Developing catalytic systems with high efficiency and selectivity is a fundamental issue for photochemical carbon dioxide conversion. In particular, rigorous control of the structure and morphology of photocatalysts is decisive for catalytic performance. Here, we report the synthesis of zinc oxide-copper(I) oxide hybrid nanoparticles as colloidal forms bearing copper(I) oxide nanocubes bound to zinc oxide spherical cores. The zinc oxide-copper(I) oxide nanoparticles behave as photocatalysts for the direct conversion of carbon dioxide to methane in an aqueous medium, under ambient pressure and temperature. The catalysts produce methane with an activity of 1080 μmol gcat −1 h−1, a quantum yield of 1.5% and a selectivity for methane of >99%. The catalytic ability of the zinc oxide-copper(I) oxide hybrid catalyst is attributed to excellent band alignment of the zinc-oxide and copper(I) oxide domains, few surface defects which reduce defect-induced charge recombination and enhance electron transfer to the reagents, and a high-surface area colloidal morphology.


8.
Line 257-258, Why do the authors believe that O2 is oxidizing the Cu2O phase instead of this phase being oxidized by the photogenerated holes?
Once these issues have been addressed, I believe this paper will be publishable.

For Reviewer #1:
The manuscript by Bae et al reports an efficient photocatalytic CO2 reduction to methane using a heterostructure composed of a central aggregate of ZnO nanocrystals surrounded by cubic Cu2O crystals. The authors claim high selectivity towards the formation of CH4 over the formation of CO. I find the topic to be interesting and suitable for the journal. The reported efficiency is certainly of note. The characterization is properly written, but the discussion of the results is unfortunately of lesser standard. In addition, I have serious doubts about the proposed mechanism and the accuracy of some of the measurements.
Author response: We deeply appreciate the reviewer's evaluation that our work is interesting and our results are certainly of note. We agree the reviewer's concern on the proposed mechanism and its explanation. We have checked our measurements to enhance the accuracy, and have completely re-written the section of band alignment and electron transfer mechanism related to Fig. 3. 1. The authors claim that the CO2 reduction step proceeds on ZnO crystals due to Type II alignment between ZnO and Cu2O. This would be very unusual as ZnO is not known for good photocatalytic properties towards CO2 reduction (without the use of noble metal cocatalysts). On the other hand Cu2O is considered to be one of the best candidates. In fact the measurements taken by the authors for isolated ZnO and Cu2O confirm this trend (15micromol vs 180micromol). The authors would need very good arguments for the reaction to proceed on ZnO, which are not present in the current manuscript.
Author response: We totally respect the reviewer's opinion for the description of mechanistic aspects. In fact, there have been two opposite pathways proposed for metal oxide hybrid catalytic systems; one is the electron transfer from the conduction band of one material at a higher energy level to the conduction band of the other material at a lower energy level, which was what we proposed in our manuscript. The ZnO islands on CuO nanowires (Biswas et al. ACS Appl. Mater. Interfaces 7, 5685-5692 (2015)) and Cu 2 O/TiO 2 porous materials (Ye et al. Nanotechnology 25, 165402 (2014)) were the systems explained using this model of bandgap alignment. On the contrary, most of other literatures used Zscheme-like reaction pathways where the electron is transferred from the conduction band of one material at a low energy level to the valence band of another material (Grimes et al. ACS Nano 4, 1259-1278(2010; Li et al. ACS Appl. Mater. Interfaces 7, 8631-8639 (2015)). Although there remains some ambiguity between the opposite reaction pathways, we think that the latter mechanism is more appropriate for our catalytic system, because of the reasons that the reviewer commented. Figure 3c clearly shows that the Cu 2 O sites are more active than the ZnO sites as a pure form. The sufficient energy of the excited electrons on the conduction band of the Cu 2 O sites leads to the effective reduction of CO 2 into other compounds. On the other hand, the water oxidation generally requires a large overpotential, therefore, the hole transfer from the valence band of the ZnO domains is more reasonable to understand the high activity of this catalytic system. We have changed Fig. 3e, and have rewritten the section of "Mechanistic aspects of CO 2 conversion reactions", by the proposition of the latter pathway for CO 2 reduction.

Revision made:
Page 10: Fig. 3e was changed into the proper diagram depicting bandgap alignment and reaction pathways. data presented are in the millimol/g•h with 1.5 % quantum efficiency that is a remarkable activity. Nevertheless, publication in Nature Communications of this submission is not recommended, because the work has not addressed the main drawbacks of the two metal oxide, i.e., their instability and their tendency to undergo photocorrosion and deactivation upon extended use. Just a simple comment that the material is "stable enough" is not convincing in view of the extensive data in the literature that, both component ZnO and Cu2O, even though quite photoactive at short times, eventually deactivate. Figure 5 that is key in this issue shows a significant decrease in the production rate from o to 2 h compared to 10-14 h in which the reaction does not progress. Why?
Author response: At first, we appreciate the reviewer's evaluation of the importance of our work. We also agree the reviewer's main concern on the catalyst stability. As the reviewer commented, both components of ZnO and Cu 2 O are known to be unstable under the low pH conditions. But, under the nearly neutral conditions, the particles maintained their original structure when we carried out the reactions for more than 24 h. For Fig. 5, we reported some saturation behaviors of the reactions by prolonged irradiation, which were explained by the reaction equilibrium between the reactants and products. However, as the other reviewer claimed, we were totally incorrect, because neither CO 2 reduction into methane nor water oxidation are reversible reactions. We think that the decrease of the CH 4 production at 10-14 h is mainly due to the depletion of CO 2 . To prove it, we have carried out more stability tests as follows: Experiment 1: Additional 10 h reaction after the reaction for 14 h We have carried out the CO 2 reduction experiment for 14 h under the present reaction conditions. Then, the catalyst particles were re-dispersed in a fresh reaction medium with 0.2 M Na 2 CO 3 , and the reaction was carried out again under the identical conditions. The catalysts were still active after the 14 h reaction, and showed a similar and even better activity than that of the first reaction.
Experiment 2: repeating 4 h reactions three times using fresh reaction media In this experiment, the catalyst particles were re-dispersed in the fresh reaction medium with 0.2 M Na 2 CO 3 after the first 4 h reaction, and the reaction was carried out. This process was repeated again to show the catalyst stability lasting for 12 h. As shown in the figure, the catalytic activity was regenerated after the use of the fresh reaction medium.
Based on these tests, we concluded that the catalyst stability maintains during the reaction for more than 12 h, when the fresh reaction medium is supplied. The present reaction proceeded inside the closed chamber, resulting in the CO 2 depletion by the prolonged reaction time. This would be a main cause for the significant decrease of the activity at the late stage of the reaction. The CO 2 depletion may change the pH of the reaction medium and slow down the CO 2 reduction reaction, which leads to the oxidation of Cu 2 O surface into CuO by photogenerated holes. We have added the data of Experiment 2 in Fig. 5b, and have rewritten the section of "Stability of the ZnO-Cu 2 O catalysts".

Revision made:
Pages 14-15: The section of "Stability of the ZnO-Cu 2 O catalysts" was also completely rewritten. Fig. 5b was changed as follows. 2. Other points that the authors may consider are the following: The proportion between ZnO and Cu2O has not been optimized. Only a material with a fixed percentage has been studied.
Author response: We have already carried out the optimization of the ratio between ZnO and Cu 2 O domains for the catalyst purpose. The CH 4 production rate was clearly dependent upon the Zn/Cu ratio as follows. However, in the cases of other catalysts with the Zn/Cu ratios distinct from 1 : 1, either the Cu 2 O domain was not fully grown or the ZnO-Cu 2 O catalyst structure was not sufficiently defined well. Our purpose for this manuscript is the importance of well-defined nanostructures for CO 2 conversion reactions, therefore, we did not include this optimization process in the main text.
3. The text comments twice that assignment of the particles in HRTEM is based on the difference between 0.247 (ZnO) and 0.246 nm (Cu2O). What is the resolution of the instrument? In contrast to the extremely high resolution of the instrument, the claimed epitaxial growth of Cu2O on top of ZnO has not been supported with the corresponding HRTEM images of this region.
Author response: As the reviewer commented, we have checked the HRTEM images of different samples. In fact, it is very hard to clearly see the junction area, because the ZnO domain is an aggregate of small single-crystalline domains. See below for a representative HRTEM image. We still think that the low lattice mismatch between ZnO and Cu 2 O domains enables the direct attachment of the Cu 2 O seeds on the surface of ZnO, but we agree that the term of "epitaxial growth" is misused, because it is generally used for good crystallographic alignment between the different domains in a large area during the thin film growth process. We have changed the descriptions related to "epitaxial".

Revision made:
Pages 6-7: We modified the sentence as "This low lattice mismatch may lead to the direct growth of Cu 2 O on the ZnO surface forming good junctions.".
Page 12: We modified the sentence as "As shown in Fig. 1c, the average distance of Cu 2 O(111) planes matches that of ZnO(101) planes, which forms good contact between the different domains.".
Page 16: We modified the sentence as "ZnO-Cu 2 O hybrid nanoparticles were synthesized through the direct surface growth of Cu 2 O on ZnO spheres.". 4. 13CO2 has to be used as starting material and detect the corresponding 13CH4 in the required 13C percentage.
Author response: We have conducted an isotope experiment using 13 CO 2 and Na 2 13 CO 3 . The gas product was analyzed by GC/MS. A signal at m/e = 17 increased a lot when the 13 C reagents were used for the reaction, indicative of the production of 13 CH 4 . This result implies that CH 4 was directly produced from CO 2 , which is relevant to the control experiment of no CH 4 formation under a N 2 flow.

Revision made:
Supplementary Figure  13 CO 2 as carbon sources for CO 2 reduction.
Pages 8-9: We inserted sentences as "The reaction in the presence of 13 CO 2 was also carried out ( Supplementary Fig. 2). A signal at m/e = 17, assignable to the 13 CH 4 peak, increased a lot in the gas chromatography-mass spectrometry (GC-MS) chromatogram when 13 CO 2 and Na 2 13 CO 3 were used, which were another indication of the direct CO 2 reduction into CH 4 .".
5. It is known that the binding energy value in XPS cannot differentiate between Cu(0) and Cu(I). However, the authors attribute the XPS Cu2p peak exclusively to Cu(I).
Author response: As the reviewer commented, Cu(0) and Cu(I) cannot be distinguished in the XPS spectrum. However, XPS data can indicate the absence of Cu(II) species. By the combination of XRD data, it was informed that Cu 2 O was exclusively generated from the reaction. We have modified the descriptions related to the XPS data.

Revision made:
Page 6: We changed the sentence as "In particular, the spectrum in the Cu 2p 3/2 region shows a single symmetric peak at 932.1 eV, indicating that there were no formation of Cu(II) during the synthesis (Fig. 1i).".
Page 14: We changed the sentence as "The XPS data in the region of Cu 2p 3/2 also indicates the presence of Cu species on the surface (Fig. 4b).".
6. Since Cu2O is active under visible light, the photoresponse of the material has to be presented.
Author response: As the reviewer recommended, we have checked the photoresponse of our catalyst, ZnO-Cu 2 O, by the irradiation of UV-vis and visible light (> 425 nm). The catalyst deposited on a FTO electrode generated cathodic photocurrents at an applied potential of -0.45 V vs. Ag/AgCl in a phosphate buffer, implying the p-type characteristics of the Cu 2 O domains. The electrode generated almost identical photocurrents under UV-Vis and visible light irradiation, indicating that the Cu 2 O is highly effective in the visible light region.

Revision made:
Page 10: We inserted the sentences as, "The photoresponse of the ZnO-Cu 2 O catalysts was measured by the irradiation of UV-Vis and visible light (> 425 nm). The catalyst deposited on a FTO electrode generated cathodic photocurrents at an applied potential of -0.45 V vs. Ag/AgCl in a phosphate buffer, implying the p-type characteristics of the Cu 2 O domains. The electrode generated almost identical photocurrents under UV-Vis and visible light irradiation, indicating that the Cu 2 O is highly effective in the visible light region ( Supplementary Fig.  4).".
Supplementary Figure  7. Figure 3e regarding the mechanism implies that both ZnO and Cu2O can absorb photons, leading to equal efficiency? ZnO absorbs in the UV, while Cu2O absorbs in the visible. This issue of which component is most efficient should be addressed.
8. According to Figure 3 e, irradiation of ZnO under UV light should also produce the same reaction without the need of Cu2O. This has to be confirmed.
Author response: We agree that our explanation of the mechanistic aspect was incorrect. Figure 3c clearly shows that the Cu 2 O sites are more efficient than the ZnO sites as a pure form for CO 2 reduction. Apparently, the Cu 2 O sites should be the active surface where the CO 2 reduction occurs. In fact, there have been two opposite pathways proposed for metal oxide hybrid catalytic systems; one is the electron transfer from the conduction band of one material at a higher energy level to the conduction band of the other material at a lower energy level, which was what we proposed in our manuscript. The ZnO islands on CuO nanowires ( (2015)).
Although there remains some ambiguity between these opposite reaction pathways, we think that the latter mechanism is more appropriate for our catalytic system, because of the reasons that the reviewer commented well. The sufficient energy of the excited electrons on the conduction band of the Cu 2 O sites leads to the effective reduction of CO 2 into other compounds. On the other hand, the water oxidation generally requires a large overpotential, therefore, the hole transfer from the valence band of the ZnO domains is more reasonable to understand the high activity of this catalytic system. We have changed Fig. 3e, and have rewritten the section of "Mechanistic aspects of CO 2 conversion reactions", by the proposition of the latter pathway for CO 2 reduction.

Revision made:
Page 10: Fig. 3e was changed into the proper diagram depicting bandgap alignment and reaction pathways. Pages 10-13: The section of "Mechanistic aspects of CO 2 conversion reactions" was completely re-written.

For Reviewer #3:
This manuscript reports an impressive result in terms of photocatalytic efficiency for the conversion of CO2 to methane. The authors claim they have achieved this result by synthesizing an improved p-n junction between nanoparticles of ZnO and Cu2O, and they provide strong evidence supporting this claim. However, in the manuscript they make several statements that are either incorrect or speculative (when proof could be obtained in a direct manner). These items need to be considered prior to publication. Once these issues have been addressed, I believe this paper will be publishable.
Author response: We deeply appreciate the reviewer's evaluation of our achievement and its strong evidences. We agree that the original manuscript has several incorrect statements; therefore, we have carefully addressed all issues that the reviewer kindly commented.
1. Lines 130-132, The authors make a CO2 saturated electrolyte by starting with a 0.2M Na2CO3 solution which is acidified with perchloric acid to pH=7.4 under a pressure of CO2(g). But, at this pH carbonate is converted to CO2. So, this procedure makes little sense. Is there something going on here that is not being described?
Author response: We have used the right condition of pH = 7.4 as we described in the manuscript. As the reviewer commented, most of the CO 2 reduction reactions were carried out under the basic conditions dissolving more CO 2 in water as a carbonate form. However, multiple protons are particularly needed for CH 4 production; therefore, there should be a middle pH to satisfy these two opposite requirements. As a matter of fact, there are several results using neutral conditions for CO 2 reduction (Chen et al. J. Phys. Chem. Solids 73, 661-669 (2012); Chen et al. Catal. Commun. 8, 1546-1549(2007). In the case of AgBr/TiO 2 nanocomposites, the catalyst exhibited maximum activities for the methane and methanol production in the range of pH 6.0 -8.5 (He et al. Catal. Today 175, 256-263 (2011)). In the Nafion/Pd-TiO 2 catalyst, even pH 1 was the best condition of the CH 4 and C 2 H 6 generation compared to the conditions at pH = 3 and 11. In our experimental conditions, we think that the high proton concentration at neutral pH may help the effective formation of CH 4 . We described this factor in the section for the selectivity issue.

Revision made:
Pages 12-13: We inserted the description of the proton concentration as "In the present reaction conditions, the reaction medium, water, with a high proton concentration at neutral pH behaves a rich hydrogen source, and leads to the CH 4 production more effectively. In addition, it is reported that the intermediates such as CO are particularly stabilized on the Cu 2 O(100) surface, which may allow more efficient coupling with adsorbed protons during the reaction.".
2. Lines 138-141, The authors carry out a run under N2, which only produces a trace of CH4 to "prove" that the observed CH4 derives from CO2. This is a necessary but insufficient control. One can image a number of reasons that methane is not produced under nitrogen, but it does not come from CO2. This is especially true because of the comment made in point 1 about the authors "hidden" source of CO2. The essential experiment to carry out is a run using 13CO2 showing that 13CH4 is formed. This manuscript should not be published without this critical control experiment.
Author response: As the reviewer recommended, we have conducted an isotope experiment using 13 CO 2 and Na 2 13 CO 3 . The gas product was analyzed by GC/MS. A signal at m/e = 17 increased a lot when the 13 C reagents were used for the reaction, indicative of the production of 13 CH 4 . This result implies that CH 4 was directly produced from CO 2 , which is relevant to the control experiment of no CH 4 formation under a N 2 flow. 13 CO 2 as carbon sources for CO 2 reduction.
Pages 8-9: We inserted sentences as "The reaction in the presence of 13 CO 2 was also carried out ( Supplementary Fig. 2). A signal at m/e = 17, assignable to the 13 CH 4 peak, increased a lot in the gas chromatography-mass spectrometry (GC-MS) chromatogram when 13 CO 2 and Na 2 13 CO 3 were used, which were another indication of the direct CO 2 reduction into CH 4 .".
3. Lines 152-155, This is not the normal definition of quantum yield. The QE for a product is normally the moles of product formed per moles of photons incident. By inserting a factor of 8, the authors have converted the quantum yield from CO production to the quantum yield for electron production (leading to CO formation). This becomes important when they compare their results to other works.
Author response: We understand that the quantum efficiency is defined by following equation,

QE /
where  (2011)). We have also followed the same definition, and obtained the quantum efficiency of 1.5% for our catalyst.

Revision made:
Page 9: We added a sentence, "It is noted that 8 electrons are required for the production of one CH 4 molecule from CO 2 .".
4. Line 196 indicates that water is oxidized in this reaction. However, no data is provided showing the formation of O2. This needs to be demonstrated if one wants to claim water oxidation.
Author response: Using the TCD detector in GC, we have detected the oxygen amount during the reaction. The amount of oxygen increased along the reaction progress for 3 h. However, the oxygen evolution was slower than what we expected in the ideal reaction. We think that photogenerated holes were partly consumed by the oxidation of the Cu 2 O domains. The OH radicals, generated by the holes with water, may also react with the active surfaces. The actual role of the OH radicals in the possible reactions remains unclear, as some literatures described (Grimes et al. ACS Nano 2010, 4, 1259-1278(2010).

Revision made:
Page 14: We omitted the description of oxygen formation, because the discussion of the stability issue was completely changed (see below).
5. Lines 201-204, The selectivity of the process for methane is impressive, but it is not due to the indicated conduction band position.
Author response: For the issue of selectivity, we agree the reviewer's opinion, in which the product selectivity is limited by the energies of intermediates with kinetic factors, but is not determined by thermodynamic conduction band position. We have begun to study gas-phase reaction conditions using the present ZnO-Cu 2 O catalysts, and as a preliminary result, the CH 4 generation was significantly diminished with the large production of CO when the small amount of water was used. This result indicates that water as a reaction medium behaves as a rich hydrogen source, leading to nearly quantitative production of CH 4 over other products. It is also reported that adsorbed CO are particularly stabilized on the Cu 2 O(100) surface, which may allow more efficient coupling with adsorbed protons during the reaction. We have inserted the discussion of selectivity in the main manuscript.

Revision made:
Page 10: Figure 3e was changed into the proper diagram depicting bandgap alignment and reaction pathways. Pages 10-12: The section of "Mechanistic aspects of CO 2 conversion reactions" was completely re-written.
Pages 12-13: We inserted a paragraph as "For the issue of selectivity, this photocatalytic system provides sufficient energy based on the Z-scheme with photoexcited electrons at a high energy level to CO 2 reduction. It is known that the products are highly dependent upon relative energy levels of intermediates. In the present reaction system, the reaction medium, water, with a high proton concentration behaves a rich hydrogen source, and leads to the production of CH 4 more effectively. In addition, it is reported that the intermediates such as CO are particularly stabilized on the Cu 2 O(100) surface, which may allow more efficient coupling with adsorbed protons during the reaction.".
6. Line 237-8, the authors state, "Regarding direct CO2 conversion, the ZnO-Cu2O catalysts are definitely superior to TiO2-Cu2O for both reaction activity and selectivity." But, in fact, what the data proves is that their catalyst is an improvement over P25-Cu2O only.
7. Line 252-256, This is a false statement. Neither the reduction of CO2 to methane or the oxidation of water to O2 is reversible by any stretch of the imagination, under any set of conditions present in the experiment under consideration. In fact, these reactions are often used to illustrate the prototypically irreversible reaction.
8. Line 257-258, Why do the authors believe that O2 is oxidizing the Cu2O phase instead of this phase being oxidized by the photogenerated holes?
Author response: We deeply thank the reviewer for these valuable comments. In Fig. 5, we reported some saturation behaviors of the reactions by prolonged irradiation, which were explained by the reaction equilibrium between the reactants and products. However, as the reviewer claimed, we were totally incorrect, because neither CO 2 reduction into methane nor water oxidation are reversible reactions. We think that the decrease of the CH 4 production at 10-14 h is mainly due to the depletion of CO 2 . To prove it, we have carried out more stability tests. In this experiment, the catalyst particles were re-dispersed in the fresh reaction medium with 0.2 M Na 2 CO 3 after the first 4 h reaction, and the reaction was carried out. This process was repeated again to show the catalyst stability. As shown in the figure, the catalytic activity was regenerated after the use of the fresh reaction medium.
Based on these tests, we concluded that the catalyst stability maintains during the reaction for more than 12 h, when the fresh reaction medium is supplied. The present reaction proceeded inside the closed chamber, resulting in the CO 2 depletion along the prolonged reaction time. This would be a main cause for the significant decrease of the activity at the late stage of the reaction. The CO 2 depletion may change the pH of the reaction medium and slow down the CO 2 reduction reaction, which leads to the Cu 2 O oxidation into CuO by photogenerated holes. We have added the data in the main text, and rewritten the section of "Stability of the ZnO-Cu 2 O catalysts".

Revision made:
Pages 14-15: The section of "Stability of the ZnO-Cu 2 O catalysts" was completely re-written. Fig. 5b was changed as follows. It was a big challenge for us to address every issue that the reviewers raised, but we think we successfully reached the final goal. During this revision, we have learned many important things in photocatalytic CO 2 conversion, and we deeply respect the reviewers for their valuable comments and recommendations. We hope that we correctly represent our responses to the reviewers' concerns. Please kindly consider our revision that has been rationally prepared to provide a better manuscript suitable for the journal. Thank you for your cooperation on this matter.

Reviewer #1 (Remarks to the Author):
I have read the revised version of the manuscript "ZnO-Cu2O Colloidal Nanocatalysts for Highly Selective Photocatalytic CO2 Conversion into Methane by Water" by Bae et al together with its associated reply letter. In my opinion this version offers a substantial improvement over its processor, with most points addressed correctly.
My main concern remains the mechanism. The authors have accepted my criticism to the original version that the CO2 reduction cannot happen on ZnO, but rather on Cu2O. However, I am very skeptical about the Z-scheme proposed by the authors. In a Z-Scheme (reviewed e.g. by Maeda, ACS Catalysis 2013, 3, 1486-1503) the two absorbers are usually separated. A mediator redox pair is then employed to transfer the electrons from the oxidation component to the reduction component and to prevent the reverse process. In the system proposed by the authors, with the very good contact between the ZnO and Cu2O parts, I do not see why the electron would not transfer from Cu2O to ZnO (assuming the alignment as in Fig 3e). This would be expected to be much quicker than the CO2 electrocatalytic reaction at the surface.
In addition the time-resolved PL measurements also suggest that the recombination in a single component is faster than in the composite which would also suggest that the proposed recombination (Fig.3e) between CB of ZnO and VB of ZnO is not that fast.
In effect, I find the proposed mechanism to be very unlikely. The authors would need to provide considerable more evidence that the other electron transfer processes are not happening.
That said, I find the results -in particular the efficiency -to be very interesting and worthy of publication at high level. I suggest that the authors try to perform the photocatalytic experiments using only visible light (to exclude absorption by ZnO) to verify whether absorption by ZnO is required or whether absorption by CuO is sufficient (adjusting the light intensity for comparable photon flux). Additionally, light intensity could also help to determine whether the reaction is a twoor single-photon process. The former would be expected in a Z-scheme.
The authors characterized the samples by XPS measurements prior to the irradiation. Could such measurement be performed on a sample after the irradiation? This could help establish whether a Cu(I)/Cu(II) junction is formed, perhaps facilitated by the ZnO phase.
I have two more comments on the manuscript: 1. The time-resolved Pl decay curves should be presented on a scale of several nanoseconds if the lifetimes are in hundreds of ps (cf. Fig. 3d). It does not make sense to present the time until 1000ns in such case.
2. The arguments for the selectivity towards CH4 over CO are rather weak in the paper. The additional experiments would probably be outside the scope, but I would suggest extending the discussion based on available literature on CO2 reduction on copper(I) surfaces.
In summary, the manuscript is improved and the reported efficiency is notable, but the mechanism appears to be wrong. I suggest that the paper is not accepted at this time, but a further revision based on the experiments suggested above (and additional ones, if necessary to prove the point) could bring the paper to the level suitable for Nature Communications.

Reviewer #2 (Remarks to the Author):
In my previous report, I did not recommend publication in Nature Communications of this submission, since reports in the literature clearly indicate the instability of the material over the time, due to oxidation of Cu2O and corrosion of ZnO. Now the authors have addressed this issue and, particularly corrected Figure 5, by performing a reuse and making other tests. I still have some concerns, but I have to recognize that the authors have presented now data in support of photocatalyst stability. In view of this as well as the answers to the other reports, now I recommend publication with minor changes. The authors still have to address the following comments: • The reasons why the photocatalytic activity upon reuse increases have to be explained.
• The authors claim that the decrease in photocatalytic activity over 14 h is due to substrate depletion. Considering that CO2 is in a very large excess, this reason seems unlike. The authors have to provide CO2 conversion at 14 h to convince the reader of CO2 depletion.
• Figure showing the influence of ZnO/Cu2O ratio on the photocatalytic activity that is provided in the author's answer has to be included in the manuscript, at least in the supplementary information and mentioned in the text.
• 13-labelled CO2 experiment needs to be presented better. Instead of stating that the m/z peak at 17 increases a lot, the percentage of 13CH4 vs. 12CH4 has to be quantified and indicated in the text. Also Figure S2 has to be expanded showing the mass spectra of the 13CO2 labelled experiment.
• Visible light photoresponse has to be completed by adding the CH4 production under these conditions.
Once these changes are made, publication would be recommended.
Reviewer #3 (Remarks to the Author): The authors have carefully addressed all of the concerns of the referees. Additional experiments have been incorporated that address key open issues in the original manuscript, and conclusions have been adjusted to include the new information. Based on this analysis and my original conclusion that the manuscript reported material that was publishable, I now recommend publication of the paper. I do note that there are several small grammatical errors that should be corrected to improve the readability of the paper.

For Reviewer #1:
I have read the revised version of the manuscript "ZnO-Cu2O Colloidal Nanocatalysts for Highly Selective Photocatalytic CO2 Conversion into Methane by Water" by Bae et al together with its associated reply letter. In my opinion this version offers a substantial improvement over its processor, with most points addressed correctly.
Author response: We deeply appreciate the reviewer's opinion that the previous revision addressed most of the issues properly.
1. My main concern remains the mechanism. The authors have accepted my criticism to the original version that the CO2 reduction cannot happen on ZnO, but rather on Cu2O. However, I am very skeptical about the Z-scheme proposed by the authors. In a Z-Scheme (reviewed e.g. by Maeda, ACS Catalysis 2013, 3, 1486-1503) the two absorbers are usually separated. A mediator redox pair is then employed to transfer the electrons from the oxidation component to the reduction component and to prevent the reverse process. In the system proposed by the authors, with the very good contact between the ZnO and Cu2O parts, I do not see why the electron would not transfer from Cu2O to ZnO (assuming the alignment as in Fig 3e). This would be expected to be much quicker than the CO2 electrocatalytic reaction at the surface.
Author response: As we modified our previous manuscript, we totally agreed the reviewer's opinion describing the fact that Cu 2 O is an active surface for CO 2 reduction, and changed our explanation into Z-scheme-like mechanism. In fact, the Z-scheme mechanisms have been generally used for the systems with separated absorbers in the presence of mediators, however, the Z-schemes having intimately contacted semiconductors without mediators have also been reported.  (2017)). Grimes et al. suggested the Z-scheme for the CO 2 reduction into CO in a TiO 2 -CuO system in their review paper (ACS Nano 3, 1259-1278 (2010)). Recently, Li et al. compared between two opposite schemes of double-charge transfer and Z-scheme mechanisms on the Fe 2 O 3 /Cu 2 O heterostructures, and concluded that the direct Z-scheme mechanism was more reasonable in their system (ACS Appl. Mater. Interfaces 7, 8631-8639 (2015)). It is noted that when a photocatalyst is immersed in water, charge transfer occurs at the semiconductor-solution interface due to the equilibration of electron density between two phases. The net result is the formation of an electrical field at the semiconductor surface, which leads to the hole transfer to the surface oxidizing water in n-type semiconductors (ZnO) when photogenerated electron-hole pairs forms in the space charge region. Similarly, photogenerated electrons move to the surface reducing CO 2 in p-type semiconductors (Cu 2 O). (Nozik et al. Appl. Phys. Lett. 30, 567-569 (1977); Walter et al. Chem. Rev. 10, 6446-6473 (2010)).
For the facile electron transfer from ZnO to Cu 2 O, we think that the excited electrons in the ZnO domains are trapped on the interface states, which may reduce the charge recombination rate to the valence band of ZnO and facilitate the tunneling to the valence band of Cu 2 O (see below for responses 2 and 3). We have added the comments for the comparison between double charge transfer and Z-scheme mechanisms in the manuscript.

Revision made:
Page 12, Lines 8-15: Some sentences were modified as "Second, the formation of uniform domain structures facilitates highly efficient electron and hole transfers to the reagents. When a photocatalyst is immersed in water, charge transfer occurs at the semiconductor-solution interface due to the equilibration of electron density between two phases. The net result is the formation of an electrical field at the semiconductor surface, which leads to the hole transfer to the surface oxidizing water in n-type semiconductors (ZnO), when phtotogenerated electron-hole pairs forms in the space charge region. Similarly, photogenerated electrons move to the surface reducing CO 2 in p-type semiconductors (Cu 2 O)." with proper citations.
Page 13, Lines 3-13: A paragraph was added to explain the reaction mechanism as "The other mechanism, double charge transfer, which includes electron transfer from the conduction band of Cu 2 O to ZnO domains and hole transfer from the valance band of ZnO to Cu 2 O, has also been proposed in several photoreduction systems. However, in our catalysts, the CH 4 production of the pure ZnO aggregates was negligible, while the pure Cu 2 O nanoparticles showed a significant activity (Figure 3c), indicating that the Cu 2 O domains are main active sites for CO 2 reduction. In the aspect of band edge energies, CO 2 reduction needs sufficient energy over its electrochemical potential, and water oxidation also requires a large overpotential in general. Therefore, the Z-scheme mechanism in Figure 3e is more reasonable, where electron transfer occurs from the high-lying conduction band of the Cu 2 O domains to CO 2 molecules. The low-lying valance band of the ZnO domains is also able to provide a sufficient overpotential for water oxidation reactions." with proper citations.
2. In addition the time-resolved PL measurements also suggest that the recombination in a single component is faster than in the composite which would also suggest that the proposed recombination (Fig.3e) between CB of ZnO and VB of ZnO is not that fast.
Author response: Indeed, a similar time-resolved PL behavior was also observed in ZnO islands on CuO nanowires (Biswas et al. ACS Appl. Mater. Interfaces 7, 5685-5692 (2015)). In this literature, the authors explained the long photoexcited electron lifetime by the electron transfer from CuO to ZnO domains, which is the mechanism that we have originally proposed in our previous manuscript. However, as you commented, this double charge transfer mechanism has many problems unable to explain the negligible activity of ZnO and insufficient band edge energies to supply overpotentials of the redox reactions. Instead, we agree your and other reviewers' suggestions of the Z-scheme mechanism, which is more reasonable for our catalytic system (also see below for response 3). For the long lifetime of photoexcited electrons, it is known that interface states are generated between the Cu 2 O and ZnO domains, which trap charge carriers. In fact, the peak centered at 620 nm used for the PL measurement appears only in the spectrum of the ZnO-Cu 2 O nanoparticles, but not in those of the pure ZnO and Cu 2 O nanoparticles. This indicates that the 620 nm peak may root from the interfacial energy band transition, which was already observed in the Cu 2 O/ZnO heterojunction nanostructure (Yu et al. Nanoscale 4, 7817-7824 (2012); Schmidt-Mende et al. Adv. Funct. Mater. 21, 573-582 (2011)). These trap levels may reduce the charge recombination rate from the conduction band to the valence band of ZnO, and facilitate the tunneling to the valence band of Cu 2 O. To understand the photophysical mechanism in detail, further study is required.

Revision made:
Page 9, Lines 17-21: Some sentences were modified as, "On the other hand, the pure ZnO aggregates exhibit only weak signals in this region, and the Cu 2 O nanoparticles have a distinct peak at 570 nm. Therefore, the peak centred at 620 nm roots from the interfacial energy band transition, as observed in the Cu 2 O/ZnO heterojunction nanostructure. The decay of transient absorption was measured at this wavelength.".
Page 10, Lines 1-3: A sentence, "This may be attributed to the interface states trapping electrons, which reduce the charge recombination rate to from the conduction band to the valence band of ZnO and facilitate tunnelling to the valence band of Cu 2 O." was added with the citation of references 31 and 32.
3. In effect, I find the proposed mechanism to be very unlikely. The authors would need to provide considerable more evidence that the other electron transfer processes are not happening. That said, I find the results -in particular the efficiency -to be very interesting and worthy of publication at high level. I suggest that the authors try to perform the photocatalytic experiments using only visible light (to exclude absorption by ZnO) to verify whether absorption by ZnO is required or whether absorption by CuO is sufficient (adjusting the light intensity for comparable photon flux). Additionally, light intensity could also help to determine whether the reaction is a two-or single-photon process. The former would be expected in a Z-scheme. The authors characterized the samples by XPS measurements prior to the irradiation. Could such measurement be performed on a sample after the irradiation? This could help establish whether a Cu(I)/Cu(II) junction is formed, perhaps facilitated by the ZnO phase.
Author response: We sincerely thank to the reviewer for this valuable comment. The reviewer's suggestion would be a direct evidence to prove the mechanism. When we irradiated visible light with a UV cut off filter (λ > 420 nm), CH 4 production was nearly negligible, and after the removal of the filter with a fixed light intensity (0.59 Wcm -2 ), CH 4 was generated with the activity similar to the previous experiment. There was no change of the surface states in the XPS spectrum after the visible light irradiation. This result indicates that the excitation of electrons in the ZnO domain is critical to activate the catalyst, and demonstrates that the Z-scheme is a reliable photochemical reaction mechanism. We have inserted this experimental result in the main text and supplementary information.

Supplementary Figure 6 | Photocatalytic reactions by the irradiation of visible light.
Amount of CH 4 production under the irradiation of visible light with a UV cut-off filter (λ > 420 nm). After 6.5 h irradiation, the filter was removed. The light intensity before and after the removal of the filter was fixed at 0.59 Wcm -2 by adjusting the distance between the light source and the reactor.

Revision made:
Page 13, Lines 13-21: The sentences were added as "To prove the proper photophysical mechanism, the reaction was carried out by the irradiation of visible light under the present condition. The CH 4 production was almost negligible by the light irradiation with a UV cutoff filter (λ > 420 nm), and the surface state of the catalyst was unchanged after the reaction. However, CH 4 was generated with the activity similar to that of the original experiment after the removal of the cut-off filter at a fixed light intensity of 0.59 Wcm -2 (Supplementary Figure 6). This result demonstrates that the excitation of electrons in the ZnO domain is critical to activate the catalyst, and consequently, Z-scheme is a more reliable reaction mechanism in our catalytic system.".
Supplementary Figure 6 was inserted in Supplementary Information. 4. I have two more comments on the manuscript: The time-resolved Pl decay curves should be presented on a scale of several nanoseconds if the lifetimes are in hundreds of ps (cf. Fig.  3d). It does not make sense to present the time until 1000ns in such case.
Author response: As the reviewer recommended, we have inserted expanded spectra in the region of 0 -4 ns as an inset in Figure 3d. Figure 3d was changed with the inclusion of spectra in the region of 0 -4 ns.

Revision made:
5. The arguments for the selectivity towards CH4 over CO are rather weak in the paper. The additional experiments would probably be outside the scope, but I would suggest extending the discussion based on available literature on CO2 reduction on copper(I) surfaces.
Author response: For the issue of selectivity, there are a few proposed mechanisms, such as formaldehyde and carbine pathways. In our experiment, the selectivity of CH 4 is remarkably over 99%; therefore, it is hard to distinguish any of the mechanisms. The more important point is that the intermediates are strongly bound to the Cu 2 O(100) surface, which induce the efficient reduction of CO into CH 4 with a rich hydrogen source of aqueous solutions at neutral pH. As the reviewer recommended, we extend the discussion of CO 2 reduction based on commonly acceptable mechanisms with adding the comments of the uniqueness of the Cu 2 O(100) surface.

Revision made:
Page 14, Lines 1-12: The description of the reaction mechanism was inserted as "Gattrell and many other researchers suggested that the radical anion of CO 2 is adsorbed on the metal surface and forms a carboxylic radical, which converts to CO by the interaction with surface hydrogen radical. According to the calculations, the rate determining step of the process is the hydrogenation of CO into the formyl radical, which majorly decides the product distribution. Cu has a strong binding strength for adsorbed intermediates and facilitate the hydrogenation. More specifically, it is reported that the intermediates are particularly stabilized on the Cu 2 O(100) surface, which prevents the desorption of CO and allow efficient coupling with protons during the reaction. In the present reaction conditions, the reaction medium, water, with a high proton concentration at neutral pH behaves a rich hydrogen source and directly supplies protons. The resulting intermediates, such as formyl radicals or carbenes, are further hydrogenized to produce CH 4 eventually. To understand the reaction mechanism in detail, however, further study is demanded." with proper citations of references 41-44.

For Reviewer #2:
In my previous report, I did not recommend publication in Nature Communications of this submission, since reports in the literature clearly indicate the instability of the material over the time, due to oxidation of Cu2O and corrosion of ZnO. Now the authors have addressed this issue and, particularly corrected Figure 5, by performing a reuse and making other tests. I still have some concerns, but I have to recognize that the authors have presented now data in support of photocatalyst stability. In view of this as well as the answers to the other reports, now I recommend publication with minor changes. The authors still have to address the following comments: Author response: We deeply appreciate the reviewer's opinion that the previous revision was effective to prove the photocatalyst stability.
1. The reasons why the photocatalytic activity upon reuse increases have to be explained.
Author response: We think that the photocatalytic activity partly depends on the reaction environment. In the case of recycle at each 4 h reaction time as shown in Fig. 5b, the activity did not change a lot between the cycles. In the continuous 14 h reaction periods, we expect that some surfactants were presumably detached from the catalysts, generating naked active Cu 2 O surface to generate more CH 4 .
2. The authors claim that the decrease in photocatalytic activity over 14 h is due to substrate depletion. Considering that CO2 is in a very large excess, this reason seems unlike. The authors have to provide CO2 conversion at 14 h to convince the reader of CO2 depletion.
Author response: To address the reviewer's concern, we have calculated CO 2 conversion based on the amount of dissolved CO 2 in water using the following equations (Ridgwell et al. Nature Education Knowledge 3, 21 (2012) As a result, the CO 2 conversion is estimated as 47% after the 14 h irradiation of light. The CO 2 depletion may be compensated from CO 2 gas inside the chamber, but in the real experiment, the CO 2 concentration in water may decrease further due to the temperature rising by prolonged irradiation. Severe decrease of the reactivity due to the transfer of dissolved CO 2 to the gas phase was also reported in the literature (Choi et al. Energy Environ. Sci. 5, 6066-6070 (2012)).

Revision made:
Page 16, Lines 1-3: A sentence was modified as "The reaction is carried out in a closed chamber; therefore, the CO 2 depletion in the reaction medium may be the main reason for this activity decrease (See Supplementary Information).".
Supplementary Information: The section of "Calculation of CO 2 Conversion" was added with the proper citations.
3. Figure showing the influence of ZnO/Cu2O ratio on the photocatalytic activity that is provided in the author's answer has to be included in the manuscript, at least in the supplementary information and mentioned in the text.
Author response: As the reviewer recommended, we have inserted a paragraph describing the ZnO-Cu 2 O catalysts having different Zn/Cu ratios, and added the CH 4 production graph in Supplementary Information. Figure 5 | CH 4 production using the catalysts with different Zn/Cu ratios. Amount of CH 4 production using the ZnO-Cu 2 O catalysts synthesized from the Zn and Cu precursor ratios of 2:1 (red), 1:1 (blue, optimized structure), and 2:3 (green), and using pure ZnO aggregates (black).

Revision made:
Page 10, Lines 10-14: A paragraph was inserted as "The CH 4 production rates were also measured using ZnO-Cu 2 O catalysts synthesized from the various ratio of the Zn/Cu precursors, but the activities were inferior to that of the optimized catalyst ( Supplementary  Fig. 5). It is because either the Cu 2 O domains were not fully grown on the ZnO surface, or the resulting catalyst was not uniform in its morphology. This indicates that the catalyst structure is an essential factor to maximize the catalytic performances." Supplementary Figure 5 was added in Supplementary Information. 4. 13-labelled CO2 experiment needs to be presented better. Instead of stating that the m/z peak at 17 increases a lot, the percentage of 13CH4 vs. 12CH4 has to be quantified and indicated in the text. Also Figure S2 has to be expanded showing the mass spectra of the 13CO2 labelled experiment.
Author response: In the isotope experiment, mostly 13 CH 4 but also a small amount of 13 CO 2 were produced. It was very difficult to isolate the only CH 4 signals in our GC-MS instrument, because the fragmentation peaks from small molecules such as CH 4 and H 2 O usually appeared together in the early region of the spectrum. Instead, the peak at m/e = 17 significantly increased when the 13 CO was used as shown in Supplementary Fig. 2. In addition, although CO was produced in a small amount in the reaction, the 13 CO peak (m/z = 29) uniquely appeared when 13 CO 2 was used for the reaction. Based on the control experiment with N 2 and the large amount generation of CH 4 by prolonged irradiation, we ensure that CH 4 was actually generated from CO 2 reduction. 5. Visible light photoresponse has to be completed by adding the CH4 production under these conditions.
Author response: We thank to the reviewer for this valuable comment. As the reviewer recommended, we have carried out the visible light experiment. When we irradiated visible light with a UV cut off filter (λ > 420 nm), CH 4 production was nearly negligible, and after the removal of the filter with a fixed light intensity (0.59 Wcm -2 ), CH 4 was generated with the activity similar to the previous experiment. There was no change of the surface states in the XPS spectrum after the irradiation. This result indicates that the excitation of electrons in the ZnO domain is critical to activate the catalyst, and demonstrates that the Z-scheme is a reliable photochemical reaction mechanism. We have inserted this experimental result in the main text and supplementary information.

Revision made:
Page 13, Lines 13-21: The sentences were added as "To prove the proper photophysical mechanism, the reaction was carried out by the irradiation of visible light under the present condition. The CH 4 production was almost negligible by the light irradiation with a UV cutoff filter (λ > 420 nm), and the surface state of the catalyst was unchanged after the reaction. However, CH 4 was generated with the activity similar to that of the original experiment after the removal of the cut-off filter at a fixed light intensity of 0.59 Wcm -2 (Supplementary Figure 6). This result demonstrates that the excitation of electrons in the ZnO domain is critical to activate the catalyst, and consequently, Z-scheme is a more reliable reaction mechanism in our catalytic system.".
Supplementary Figure 6 was inserted in Supplementary Information.

Supplementary Figure 6 | Photocatalytic reactions by the irradiation of visible light.
Amount of CH 4 production under the irradiation of visible light with a UV cut-off filter (λ > 420 nm). After 6.5 h irradiation, the filter was removed. The light intensity before and after the removal of the filter was fixed at 0.59 Wcm -2 by adjusting the distance between the light source and the reactor.