Copper-on-nitride enhances the stable electrosynthesis of multi-carbon products from CO2

Copper-based materials are promising electrocatalysts for CO2 reduction. Prior studies show that the mixture of copper (I) and copper (0) at the catalyst surface enhances multi-carbon products from CO2 reduction; however, the stable presence of copper (I) remains the subject of debate. Here we report a copper on copper (I) composite that stabilizes copper (I) during CO2 reduction through the use of copper nitride as an underlying copper (I) species. We synthesize a copper-on-nitride catalyst that exhibits a Faradaic efficiency of 64 ± 2% for C2+ products. We achieve a 40-fold enhancement in the ratio of C2+ to the competing CH4 compared to the case of pure copper. We further show that the copper-on-nitride catalyst performs stable CO2 reduction over 30 h. Mechanistic studies suggest that the use of copper nitride contributes to reducing the CO dimerization energy barrier—a rate-limiting step in CO2 reduction to multi-carbon products.

• The geometric arrangement of Cu atoms within the metallic overlayers on Cu2O and Cu3N appear to be quite different. Previous work in single crystal electrochemistry has shown that the C-C coupling activity of Cu electrodes are quite sensitive to the exposed facet. Are these the most thermodynamically stable facet terminations for the metallic overlayer? Otherwise, why were these particular facets chosen? There doesn't seem to be a strong correlation between the CO adsorption energies and the kinetic barriers for CO dimerization. How much of the decrease in the kinetic barrier for CO dimerization is due to a geometric effect vs. a ligand or strain effect from the support?
• In the best case scenarios of having the metallic overlayers on Cu2O and Cu3N, the improvements in CO dimerization barrier height appear to be quite small in comparison to metallic Cu (~0.1 eV). In particular, the differences in barrier height between the metallic overlayers on Cu2O and Cu3N appear to be very small. Are these improvements within the error of the DFT calculations?
• For the Cu2O or Cu3N support to affect the binding of reaction intermediates through either ligand or strain effects, the metallic Cu overlayer must be very thin and the authors suggest a thickness of ~1-2 monolayers. However, the combination of physical characterization techniques utilized by the authors does not have a high enough spatial resolution to determine how thick this metallic Cu overlayer is on the Cu2O or Cu3N support. Experimentally demonstrating either the thickness of the overlayer or that its electronic structure is affected by the Cu2O or Cu3N support is critical to substantiate the enhancement mechanism that the authors have proposed. Have the authors considered valence band spectroscopy measurements to examine the differences in electronic structure between the metallic Cu nanoparticles and composite structures?
• Instead of a strain or ligand effect from the Cu2O or Cu3N support, have the authors considered a mechanism of enhancement from a difference in exposed Cu facets? Single-crystal studies have shown similar trends in C2+/C1 ratios through improvements in C-C coupling activity (and CH4 suppression) from introducing more undercoordinated sites on the surface (DOI: 10.1007/978-0-387-49489-0_3).
• How does the stoichiometry of the Cu overlayer on Cu2O or Cu3N composite change after steady state electrocatalysis measurements? Are there any bulk structural changes in the catalyst after these longer-term measurements? Is there any evidence of O2-or N3-leaching over time?
• Why is the Cu overlayer on Cu3N or Cu2O more stable than pure Cu? The transmission electron micrographs in the Supporting Information don't show much change in the particle morphology after the reduction treatment. Have the authors used ICP measurements to determine whether there is any dissolution or delamination of Cu from the electrodes over time?
• Please check the fitting of the X-ray photoelectron spectra. In particular, the fitting for the Cu Auger LMM spectra needs more attention to be published.
• It is difficult to deconvolute the effect of the ligand from the catalytic behavior of Cu3N for the measurements shown in Figure S13 and Table S6.
Reviewer #3 (Remarks to the Author): This study explores the design of Cu-based catalysts where Cu metal (Cu0) is deposited on a Cu+ support to enhance the selectivity of CO2 reduction to C2+ products. The idea is partly supported by literature work on Cu oxide derived catalysts and DFT calculations within the paper evaluating CO adsorption energy versus CO dimerization barrier. The hypothesis is that the selectivity step between C1 and C2 products is hydrogenation of CO versus CO dimerization. Based on the finding that Cu on Cu3N shows that the CO dimerization barrier is lowered in comparison to Cu metal and Cu on Cu2O, they synthesis Cu-on-Cu3N catalysts. These catalysts were tested along with Cu and Cu-on-Cu2O catalysts. They see that the FE to C2 is enhanced substantially on the Cu-on-Cu3N catalysts. Very interestingly (and this is rare in the CO2 electroreduction work in the literature) they report the stability of the CO2 electroreduction over 32 hours and see much more stability than Cu-on-Cu2O and Cu metal. This stability result is promising since one of the issues with much of the CO2 electroreduction work is that the performance of the catalysts degrades rapidly.
Overall, this paper reports a very interesting new finding for CO2 electroreduction. This is a hot area with a lot of results coming quickly in the past 2-3 years but this paper is promising since they not only see a dramatic increase in selectivity but also study stability and have a hypothesis-driven approach to the catalyst design. I am not entirely convinced their hypothesis is correct for the reasons they give -i.e enhance CO dimerization. The main suggestion and question I have are related to the DFT calculations, which I found to be relatively weak component of this paper.
(1) They should report more detail on the CO dimerization calculation. There has been some controversy about this pathway -the initial proposal was that a CO + CO-reaction occurs. Not two neutral CO dimers, which was found to have a very large barrier. The more recent work from the Norskov group examines this reaction and showed that cations and protons could be used to facilitate this reaction. There are no references to the work of Koper and Norskov (and also Goddard on the CO dimerization on Cu(100)) in this paper. I realize that references are limited but the authors should provide this context in the SI and explain what was involved in their CO dimerization calculation in reference to this work. If two neutral CO molecules were coupled to form a CO dimer then this would be a non-electrochemical step and would not explain any of the potential dependence seen in the experiments.
(2) If the hypothesis is that C1 versus C2 is a competition between CO hydrogenation and CO dimerization, does it not make sense to look at the barrier to CO hydrogenation on these various surfaces? i.e. CO* + H+ + e-◊ COH or CHO. The potential dependent barrier for this step has been examined by several groups. It would be helpful if this barrier or free energy of reaction was calculated and added to the discussion.
(3) They note in the SI that they examined CO dimerization on both (111) and (100) facets for the Cu-on-Cu2O and Cu-on-Cu3N. It would be helpful to report all the various surfaces that were examined and the results in Table S1. If only one facet is showing preference for CO dimerization it would be useful to know this since presumably one could explore how that correlates to the morphology of the experimental catalyst nanoparticles.
Because of the promising experimental results and the difficulty of resolving a clear mechanism for this complex reaction/catalyst, I do not believe the DFT has to provide a irrefutable explanation of the results (there is ongoing debate of these questions in the literature and it is fair to say that this is outside of the scope of this paper) but it would be helpful if they provide context to what has been done in the literature and are clear on their calculations. All of these additions could be made with a more detailed description, with references, of the DFT calculations in the SI.

Response and list of changes implemented
"Copper-on-nitride enhances the stable electrosynthesis of multi-carbon products from CO 2 " Reviewer #1 (Remarks to the Author): As the reviewer recommends, we now provide in-situ X-ray absorption spectroscopy (XAS). We investigate thereby the effect of the Cu + support on the structure and chemical state of the catalyst, all as a function of time under operation. To gain more insight into the role of the Cu + support, we recorded in-situ Cu K-edge spectra of Cu-on-Cu 3 N and Cu-on-Cu 2 O catalysts following 30 min under CO 2 reduction (Fig. R1c). We found that the absorption edges of the two catalysts are between Cu + and Cu 0 , indicating the presence of a mixture during the reaction. However, the absorption edge of Cu-on-Cu 2 O was at a lower energy than that of Cuon-Cu 3 N, with energies at 8979.4 eV and 8979.8 eV, respectively. We also calculated the ratio of Cu oxidation states with respect to the reaction time (Fig. R1d). The Cu-on-Cu 3 N catalyst shows that the structure becomes stable with Cu 3 N and Cu after the initial 60 min, while Cu-on-Cu 2 O presents only the Cu component after one hour. This observation indicates suppression of the partial reduction of the catalyst when we use the Cu 3 N support.
Contents of manuscript: Authors prepared Cu on Cu3N, which they believe to be an effective catalyst for producing C2 products from CO2. The premise for this move is that the Cu3N core helps to keep the outer layer of Cu oxidized. The highest CO2R activities found are at -0.95 V RHE: C2H4 (39% FE), C2H5OH (19% FE) and propanol (6% FE).
There are some issues with this manuscript: 1. No operando or in situ characterization techniques were done to ensure that the Cu outer layer stays oxidized during CO2 reduction. Hence, the foundational evidence to support the basic claims of this manuscript is not there.

Fig. R1.
In-situ characterization of the structure and chemical state for the catalysts during CO 2 reduction. a, In-situ Cu K-edge XAS spectra of Cu-on-Cu 3 N catalyst as function of reaction time at -0.95 V vs RHE. b, The first derivatives of the XAS spectra in a. c, In-situ Cu K-edge XAS spectra during the first 30 min on the catalysts: Cu-on-Cu 3 N (green) and Cu-on-Cu 2 O (orange). Spectra of Cu (wine red) and Cu 3 N (earthy yellow) are also listed as references. d, Calculated ratio of Cu + species with respect to the reaction time at -0.95 V vs RHE.
In this work, we aimed at improving CO 2 reduction stability while achieving a high selectivity and activity comparable to reported Cu 2 O-derived catalysts.
Prior studies have found that the mixture of Cu + /Cu 0 active species synergistically promotes CO 2 reduction in favour of C 2+ selectivity. However, whether there exists a stable presence of active Cu + during CO 2 reduction has remained the subject of debate.
2. The CO2R activities reported above are supposed to be the best. But really, they are not any better than many previous works using the so-called Cu2O-derived Cu.
Previous results suggest that metal nitrides offer a stable compound and can be employed not only as the catalytic active species but also as support 1 . In our report, we present a Cu-on-Cu 3 N catalyst in which the surface Cu protects the Cu 3 N support from further reduction over the course of CO 2 reduction, a fact we now confirm in detail using in-situ XAS results ( Fig. 3a-b in the revised manuscript). This composite leads to a modified Cu electronic structure affected by the Cu 3 N support. Together with the suppressed reduction of Cu 3 N support, the new strategy results in stable selectivity for at least 30 hours with only 10% of the loss in CO 2 reduction selectivity.
We summarize the results and compare the present advance to the best prior reports in Table R1. Inspired by the reviewer's suggestion, we reviewed the proposed reaction mechanism in light of the experimental results. We now provide an improved explanation in the manuscript (pp. 9-11).
We first evaluated the effect of surface defects on C 2+ selectivity. For Cu-on-Cu 2 O catalyst, we believe that surface defects -such as grain boundaries -may play a role in the selectivity toward C 2+ in the case of the oxide-derived process. For the Cu-on-Cu 3 N catalyst, surface defects can also affect C 2+ selectivity. However, compared with the Cu-on-Cu 2 O, which was quickly reduced to Cu (Figure 3d in the revised manuscript), the Cu-on-Cu 3 N catalyst retained a higher C 2+ selectivity under CO 2 reduction. We offer that suppressed reduction of the Cu 3 N support thus play a significant role in the high selectivity over the course of CO 2 reduction. Together with the prolonged presence of Cu + over time, it allows for increasedstability C 2+ electrosynthesis under CO 2 reduction.
We also considered local pH as another possible contributor. It is expected that local pH will rise on rough surfaces, suppressing pH-dependent CO protonation toward methane formation; while the pHindependent CO dimerization is unaffected, resulting in increased C 2+ selectivity 2-6 .
Compared to the pure Cu catalyst, Cu-on-Cu + catalysts display a suppression in methane selectivity, which can be attributed to increased local pH. We also performed CO 2 reduction in buffer solution (0.  (Fig. R2), which confirms the effect of local pH. Comparing Cu-on-Cu 2 O and Cu-on-Cu 3 N catalysts, the geometric current densities are similar (Fig. S15), which indicates a nearly identical consumption rate of local protons during CO 2 reduction. We propose therefore that differences in local pH do not account for the higher C 2+ selectivity for Cu-on-Cu 3 N relative to Cu-on-Cu 2 O.
As demonstrated by the in-situ XAS and STEM-EELS results ( Fig. 2-3 in the revised manuscript and Fig.  S6), both Cu and Cu 3 N features are observed during CO 2 reduction. EELS mapping revealed that the Cu surface is no more uniformly-shaped following the electroreduction (Fig. S6). We conclude that Cu + species are supplied to subsurface layer during the reduction, resulting in the selectivity for C 2+ . DFT calculations reveal that Cu 3 N support modifies the electronic structure of the surface Cu, decreasing the energy barrier of CO dimerization (Fig. S23). Together with the suppressed reduction of Cu + using Cu 3 N as a support, we achieve the highest C 2+ production under CO 2 reduction among the three catalysts. We performed each of the experiments proposed by the reviewer, including in-situ XAS and STEM-EELS to examine the catalyst evolution during CO 2 reduction, and the relevant electrochemical measurement in buffer solution to exclude the effect of local pH. We also carried out DFT calculations to investigate the relationship between the catalyst structure and the catalytic performance. Based on these results, we now provide improved substantiation of the proposed mechanism within the revised manuscript.
Because of # 1, #2 and #4, I am not able to recommend acceptance of this manuscript, in its present form. Much more work needs to be done, especially to address #1.
In light of the reviewer's suggestion, we built different numbers of Cu layers on top of the Cu 3 N and Cu 2 O supports ( Fig. S18-21) and calculated the diffusion free energy barriers of the oxygen and nitrogen atoms from core to the surface. Our calculations show that there is a large diffusion barrier (more than 2 eV) for both O and N to diffuse from their original positions within Cu 2 O and Cu 3 N, respectively, to one layer above their original position (diffusion from level 0 to level 1, Fig. S24-26). This is in contrast to the report of Garza et al. 7 ; however, in the Garza model, the oxygen in the sublayer is a defect within pure copper. In contrast, in our model, nitrogen/oxygen are within the pristine crystalline structure of Cu 3 N or Cu 2 O, and their diffusion to the top Cu layers will not only will leave a vacancy defect within Cu 3 N and Cu 2 O structures, but will also create another defect in the top pure Cu layers. We now explain that larger energy barriers for oxygen and nitrogen diffusion are reasonable in our catalyst model compared to the sublayer oxygen model studied by Garza et al.
We also studied the diffusion energy barriers from the first layer to the upper layers toward the surface (up to level 5). Interestingly, as in Garza's study, we observed no energy barrier for oxygen diffusion at these stages. However, there was another energy barrier for nitrogen diffusion from subsurface layers (level 3 and 4) to the surface (level 5) of around 1 eV. This energy barrier leads nitrogen to be trapped in the sublayers near the surface (all initial and relaxed configurations are in Fig. S24-26 and all relevant energy barriers are summarized in Table S6.).
Therefore, based on our computational results, we posit that, oxygen diffusion is easier than nitrogen diffusion. As a result, the Cu-on-Cu 2 O will convert to pure Cu much faster than Cu-on-Cu 3 N. For Cu-on-Cu 3 N, we expect to have nitrogen on sublayers close to the surface, and to be able to observe its effect on the surface copper's electronic structure. and Cu-on-Cu 3 N will introduce a large barrier for both oxygen and nitrogen to diffuse from the pristine core to the pure copper shell, and this delays reduction of these compounds to pure copper.
In addition, experimentally, EELS mapping shows a Cu 3 N signal after the reaction (Fig. S6). Due to the change of the surface structure, the Cu outlayer is no longer uniformly shaped, and some of Cu + species may be supplied to the subsurface layer (Fig. S6f), resulting in a selectivity for C 2+ during the reduction. It is likely that these Cu + species remain relatively stable in the sublayer since the activity and selectivity remain constant within the reaction time. Furthermore, the in-situ XAS data also show the presence of Cu and Cu 3 N during the 120 min and a relatively stable ratio of Cu 3 N/Cu after one-hour of reaction (Fig. 3 in the revised manuscript).
To investigate geometric effects, we compared different catalysts with the same facets. We considered the two widely-studied (111) and (100) facets as the most stable closed pack surface and active surface for C-C coupling, respectively, for the Cu-on-Cu 2 O and Cu-on-Cu 3 N catalysts as well as Cu, Cu 2 O and Cu 3 N.
Our calculations show that, as in previous studies, (100) shows a stronger catalytic activity toward CO dimerization compared to the (111) facet. However, comparing different models with the same facet, we observe the ligand effect and the effect of the altered electronic structure due to the sublayer nitrogen atoms. These models are present in Fig. S17-21, and the calculated energy barriers for CO dimerization are listed in Table S5.
We agree that the difference in thermodynamic energy barriers is quite small. However, the activation energy barrier calculated from the transition state is larger (0.412 eV). We added these calculations to the Supplementary Computational Details portion (pp. [4][5][6][7][8] and the calculated activation energy barriers are listed in Fig. S23. We considered Cu, Cu-on-Cu 2 O (with 1 Cu top layer), and Cu-on-Cu 3 N (with 1 Cu top layer) with (100) facet. We expect that the same trend will be observed for the other models studied in this paper. As recommended by the reviewer, we re-examined the HRTEM-EELS mapping of Cu-on-Cu 3 N catalyst before (Fig. R1a-c) and after two-hour reaction (Fig. R1d-f). The catalyst surface exhibited indications of surface reconstruction following operation under reduction conditions. 8 As shown in Fig. R1c and R1f, our analysis indicates a maximum 3 nm of the surface Cu layer, while in some areas, the thickness is less than 1 nm. It is likely that some Cu + species will reside in the subsurface layer during reduction, resulting in selectivity for C 2+ . We also measured the valence band spectra (VBS) to examine the difference in the valence electronic structure between the Cu and the composite (Fig. R2). Compared with pure Cu, we found that the Fermilevel (E F ) shifted toward VB m by 0.33 eV for Cu-on-Cu 3 N and 0.08 eV for Cu-on-Cu 2 O, respectively, indicating that the core-level Cu 3 N or Cu 2 O support affect the electronic structure of the surface Cu.

For the Cu2O or Cu3N support to affect the binding of reaction intermediates through either ligand or strain effects, the metallic Cu overlayer must be very thin and the authors suggest a thickness of ~1-2 monolayers. However, the combination of physical characterization techniques utilized by the authors does not have a high enough spatial resolution to determine how thick this metallic Cu
overlayer is on the Cu2O or Cu3N support. Experimentally demonstrating either the thickness of the overlayer or that its electronic structure is affected by the Cu2O or Cu3N support is critical to substantiate the enhancement mechanism that the authors have proposed. Have the authors considered valence band spectroscopy measurements to examine the differences in electronic structure between the metallic Cu nanoparticles and composite structures? We also considered systems having different Cu monolayers on top of Cu 2 O and Cu 3 N. As the reviewer mentioned, the effect of sublayer nitrogen and oxygen on the surface electronic structure dramatically decreases when one increases the number of copper overlayers (the electron localization potential of these models with different Cu layers are shown in Fig. S18-21, and the calculated charge density by the Bader charge analysis are tabulated in Table S5). However, our diffusion barrier calculations show that nitrogen atoms can exist on sublayers very close to the surface, and that it still has a significant effect on the surface electronic structure and consequently on the CO dimerization.
We now add data and related explanations to the revised manuscript and the Supplementary Materials (Fig. S6, S9, S18-21, Table S5).
The previous results show that single-crystal Cu (111) facets favor methane formation, while (100) facets help to increase the C-C coupling toward ethylene production 9-12 .
In this work, we synthesized three catalysts using an initial electroreduction of the surface oxidation layer using a negative cyclic voltammetry (CV) scan, yielding the final catalyst. During this process, the Cu species possesses a polycrystalline structure (Fig. S16). These structures do not exhibit the specific facet 5. Instead of a strain or ligand effect from the Cu2O or Cu3N support, have the authors considered a mechanism of enhancement from a difference in exposed Cu facets? Single-crystal studies have shown similar trends in C2+/C1 ratios through improvements in C-C coupling activity (and CH4 suppression) from introducing more undercoordinated sites on the surface (DOI: 10.1007/978-0-387-49489-0_3).
orientation. Therefore, we would not expect that they contribute in a quantitatively significant way to increase C 2+ selectivity.
We found that, following the steady-state electrocatalysis measurements, the catalyst began to lose activity and saw a loss in performance. We acquired TEM images of the Cu-on-Cu 3 N catalyst following the long-term tests, and the results showed an aggregated morphology (Fig. R3). We were not able to determine the change of Cu stoichiometry of Cu the overlayer on this Cu 3 N composite by this measurement.

Fig. R3. TEM characterization of the Cu-on-Cu 3 N catalyst after the long-term electrocatalysis test.
We refer therefore to in-situ XAS results which show that the ratio of Cu + /(Cu + +Cu 0 ) decreased and then gradually became stable after the initial 60 min (Fig. 3d in the revised manuscript). Since the sample is stable after 60 min, we characterized structural changes following two hours. TEM images show that the bulk Cu 3 N is still present, but is decreased compared with the case before reaction (Fig. R1). This is consistent with the in-situ XAS results.
Overall, we conclude, based on in-situ XAS and TEM mapping results, that the stoichiometry of the Cu overlayer on Cu 3 N composite increases, the bulk structure is still present but decreased, and that N 3leaches out as a result of the reaction.
For the Cu-on-Cu 2 O catalyst, in-situ XAS results show a prominent metallic Cu feature after 60 min (Fig.  S10). Therefore, we also believe that the Cu overlayer increases, bulk structure changes to derived Cu, and O 2leaches out after the reaction compared with that before reaction.
We point out that our diffusion free energy barrier calculations support the view that the oxygen in sublayer is less stable than is the nitrogen one, consisting with our finding that Cu-on-Cu 3 N exhibits increased stability compared to Cu 2 O.
6. How does the stoichiometry of the Cu overlayer on Cu2O or Cu3N composite change after steady state electrocatalysis measurements? Are there any bulk structural changes in the catalyst after these longer-term measurements? Is there any evidence of O2-or N3-leaching over time?
As recommended by the reviewer, we obtained ICP of the electrolyte at various reaction times of each catalyst (Fig. R4).
For pure Cu, we found that a certain content of Cu was delaminated into the electrolyte, which might cause the poor stability of CO 2 reduction. For the Cu-on-Cu 3 N and Cu-on-Cu 2 O catalysts, the concentrations of Cu dissolution are very low with time evolution. We conclude that Cu dissolution is expected to have a negligible effect on the stability.
Based on the VBS (Fig. S9), in-situ XAS (Fig. 3 in the revised manuscript) and STEM-EELS mapping (Fig. S6) results, we propose that the Cu 3 N support affects the electronic structure of the surface Cu and thereby the performance of CO 2 reduction. Together with the suppressed reduction of Cu + over time, which prolongs the effect of the Cu 3 N support on the surface Cu, it allows for an increased C 2+ electrosynthesis stability during CO 2 reduction. We have corrected the fitting curve of Cu Auger LMM spectra in Fig. 1 in the revised manuscript. We agree with reviewer that the ligand affects the reactivity of Cu 3 N catalyst. CO 2 reduction shows a low activity because of the poor conductivity of Cu 3 N capped with surface long-chain organic ligand. We also confirm that all reduction products come from CO 2 rather than the ligands -a fact we validated using N 2saturated electrolyte in control experiments (Fig. S13).
To avoid confusing the reviewer and readers, we have removed this dataset from the revised manuscript since it secondary to the main narrative.
9. It is difficult to deconvolute the effect of the ligand from the catalytic behavior of Cu3N for the measurements shown in Figure S13 and Table S6.
We appreciate the reviewer's feedback. We have acted on each suggestion as documented below, with a major focus on improving the DFT studies in the manuscript. We add detailed information to the revised manuscript (pp. [11][12] and the Supplementary Materials (pp. 4-8, Fig. S17-26, Table S5-6).
In the revised Supplementary Materials, we now discuss prior mechanistic studies on CO dimerization by the Koper, Norskov, Goddard, Bell, Head-Gordon, Janik and Asthagiri groups. Our methodology is based on the non-electrochemical CO dimerization to evaluate potential-independent catalytic activity of three systems: Cu, Cu-on-Cu 3 N and Cu-on-Cu 2 O. We recognize that considering the parameters involved in the reaction such as anions and cations, applied potential and also surface coverage of the reactants will improve the quality of the DFT computations and will provide a more realistic model. Therefore, as in 1. They should report more detail on the CO dimerization calculation. There has been some controversy about this pathway -the initial proposal was that a CO + CO-reaction occurs. Not two neutral CO dimers, which was found to have a very large barrier. The more recent work from the Norskov group examines this reaction and showed that cations and protons could be used to facilitate this reaction. There are no references to the work of Koper and Norskov (and also Goddard on the CO dimerization on Cu(100)) in this paper. I realize that references are limited but the authors should provide this context in the SI and explain what was involved in their CO dimerization calculation in reference to this work. If two neutral CO molecules were coupled to form a CO dimer then this would be a non-electrochemical step and would not explain any of the potential dependence seen in the experiments.
previous works from the above-mentioned groups, we considered the CO dimerization on different models to see the effect of the structure and consequently the oxidation state on the catalytic activity.
In light of the reviewer's suggestion, we calculated the reaction free energy barrier for CO hydrogenation (both COH and CHO) on both Cu-on-Cu 3 N and Cu-on-Cu 2 O catalysts with different number of Cu layers on top (Table S5). Our computations show that the energy barrier for CO hydrogenation on Cu-on-Cu 3 N (0.933 eV) is higher than that on Cu-on-Cu 2 O (0.749 eV) and pure Cu (0.721 eV) on (100) facets, indicating improved catalytic activity toward C 2+ production on Cu-on-Cu 3 N catalyst.
We have calculated the CO adsorption energy and CO dimerization energy barrier for two different facets of (111) and (100) for both Cu-on-Cu 3 N and Cu-on-Cu 2 O catalysts with different number of Cu layers on top (Fig. S17-21 and Table S5). The relevant discussion of geometrical effects and oxidation state effects are also added to the Supplementary computational details (pp. 4-8).
2. If the hypothesis is that C1 versus C2 is a competition between CO hydrogenation and CO dimerization, does it not make sense to look at the barrier to CO hydrogenation on these various surfaces? i.e. CO* + H+ + e-à COH or CHO. The potential dependent barrier for this step has been examined by several groups. It would be helpful if this barrier or free energy of reaction was calculated and added to the discussion.
3. They note in the SI that they examined CO dimerization on both (111) and (100) facets for the Cuon-Cu2O and Cu-on-Cu3N. It would be helpful to report all the various surfaces that were examined and the results in Table S1. If only one facet is showing preference for CO dimerization it would be useful to know this since presumably one could explore how that correlates to the morphology of the experimental catalyst nanoparticles.