Isolated copper single sites for high-performance electroreduction of carbon monoxide to multicarbon products

Electrochemical carbon monoxide reduction is a promising strategy for the production of value-added multicarbon compounds, albeit yielding diverse products with low selectivities and Faradaic efficiencies. Here, copper single atoms anchored to Ti3C2Tx MXene nanosheets are firstly demonstrated as effective and robust catalysts for electrochemical carbon monoxide reduction, achieving an ultrahigh selectivity of 98% for the formation of multicarbon products. Particularly, it exhibits a high Faradaic efficiency of 71% towards ethylene at −0.7 V versus the reversible hydrogen electrode, superior to the previously reported copper-based catalysts. Besides, it shows a stable activity during the 68-h electrolysis. Theoretical simulations reveal that atomically dispersed Cu–O3 sites favor the C–C coupling of carbon monoxide molecules to generate the key *CO-CHO species, and then induce the decreased free energy barrier of the potential-determining step, thus accounting for the high activity and selectivity of copper single atoms for carbon monoxide reduction.

(4) in the reference section, there are some problems in references citations. in ref. 16 After the authors corrected all the above issues with a minor revision, this paper can be published after a further round review.
Reviewer #3: Remarks to the Author: This manuscript by Bao et al. reports a record-level selectivity of > 90 % for the formation of C2 products (ethylene, ethanol, Acetate) In CO electrochemical reduction. The catalyst is described as Cu single atoms (Cu SAs) anchored to Ti3C2Tx MXene nanosheets. Although the C2 production selectivity reported here is very high, the present study and the conclusions made about the active sites are not sufficiently supported by the data presented. The data analysis of several of the experimental methods discussed is also not rigorous enough. Therefore, the reviewer cannot recommend the acceptance of this work in Nature communications at current state. However, if all these comments are addressed, I would like to reconsider it. The detailed comments could be considered are given below: (1) The claim that Cu species in Cu-SA/Ti3C2Tx are atomically dispersed is questionable. Firstly, EXAFS is an average result of bulk phase and there is a bulge at radical distance of ~2.5A, which is possible from Cu-Cu interaction. Moreover, fitting parameters of Figure 1g should be provided. Without fitting parameters, the fitting is meaningless and unconvincing. In addition, dimers cannot be ruled out in HAADF-STEM of Supplementary Figure 5. The author just marked spots with larger distance, how about the ones with short distance? CO FT-IR or DRIFT is suggested to probe the dispersion state of Cu in Cu-SA/Ti3C2Tx.
(2) In Electrochemical COR performances test, the control experiments of pristine Ti3C2Tx and NaBH4 reduced Ti3C2Tx without Cu species need to be conducted and compared. In addition, considering both Ti3C2Tx and carbon paper contain carbon species, C18O Isotope labeling experiment must be conducted to confirm the products is from CO reduction other than from the reduction of carbon species in Ti3C2Tx and carbon paper. Moreover, the authors used ethanol to prepare the catalyst ink, thus lead the selectivity of ethanol questionable. Can the authors exclude that ethanol measured in solution after electrochemistry is not partly due to 'leaching' of residual ethanol on catalysts?
(3) Can the authors provide an example NMR spectrum to demonstrate and illustrate the analysis of the liquid? What was the concentration of ethanol and acetate in the liquid, and can the authors express the formation of ethanol and acetate in terms of a rate? What's the detection limits of NMR test? The authors did not give information about the reaction time of each potential, so the data of selectivity give here is unconvincing. The authors show error bars in Figure 2 but did not give explanation how they are obtained. The details of the COR tests and other characterizations also need improvement.
(4) The authors compared the XRD patterns of fresh and used Cu-SA/Ti3C2Tx (Figure 1b and Supplementary Figure 18) states that an amorphous broad peak around 24° were formed After the stability test, and was caused by the exfoliation of nanosheets resulting from the sonication process, as well as the intercalation of electrolyte ions during the electrolytic process. Thus, it is very likely the surface structure of the catalysts has changed and some carbon species in Ti3C2Tx is leached during COR test. The authors need to address this problem. (5) In DFT calculations, as it is questionable to claim that Cu species in Cu-SA/Ti3C2Tx are atomically dispersed (see above comment 1), the model used here will be unconvincing. The case of Cu dimers model is suggested included in calculation for comparison.
Reviewer #4: Remarks to the Author: The authors report a combined theory/experimental study revealing a single atom copper catalyst supported on MXene with high CO reduction performance. The Faradaic efficiency and selectivity to overall C2 products appears to compare favorably with the literature and to Cu nanoparticles on the same support. The paper is fairly well written and clear in its presentation. This is an important work showcasing a successful application of using single atom catalysts for CORR with good stability and activity. This manuscript should be publishable pending revisions.
Comments regarding the computational portion of the work: The justification for the computational model used here is very sparse. A model system with a single Cu atom on a pristine MXene is used; however the surface may become hydroxylated or contain impurities such as Cl and F from the synthesis. This will strongly impact the chemical bonding and activity of the single atom. The authors should provide justification the pure O termination under the reaction conditions.
What is the binding strength of the single atom on the surface, relative to bulk Cu? What is the kinetic stability of this site, in terms of surface mobility or agglomeration from single atom to dimer? This should match experimental observations of stability.
What is the charge state of the Cu single atom in the model? How does this compare with a 0<x<+1 charge assignment from XPS? One way would be to compare computed charges with charges in Cu2O as a reference, for example.
The higher activity of the single atom over the nanoparticles is explained by the difference in the rate-limiting formation of the "COCOH" intermediate, though the difference in free energy is very small (0.73 vs 0.78 eV). This is not entirely convincing and the authors should supplement this with at least the calculation of the activation barriers for the formation of this COCOH intermediate (and ideally some other important intermediates) from transition state search. This may tell more information about the difference between the single atom and the bulk surface, if the barrier for C-C coupling is improved on the single atom (for either steric or electronic reasons). The barriers may also explain the difference in selectivity between C2H4 and EtOH.
Another experimental observation is the suppressed or decreased favorability of competing HER in the single atom versus the nanoparticle. Do the free energy of adsorption of hydrogen (GH) on the computational models follow the same experimental trends? For example, the single atom may overbind H relative to the surface and suppress HER. This is easy to test and would provide more support for the model.
Another consideration into the different performance between the single atom and nanoparticles could be due to solvation. Explicit solvation is commonly included in computational studies for CO2RR or approximated with energetic corrections. Due to the different local geometry of the single atom site the solvation shell and stabilization energies will likely differ. This should at least be addressed in text or via calculations as a possible explanation.
For free energy calculations (G), were ZPE and entropy contributions (at reaction temperature) included? They should be, and if so, should be mentioned in the computational method. the authors truly had achieved a major advance in FE to desirable products, and understood why, this could be of interest.

Response:
Thank you very much indeed for reviewing our manuscript. We greatly appreciate your inspiring, useful, helpful and constructive comments. Details of the corresponding answer and changes made are described below point by point. For your convenience, all changes have been highlighted in yellow in the revised main text and the revised supplementary information (SI) files. Figure 2: It's a little surprising that Cu NP catalyst 2a have such vastly lower total current densities. Was the Cu NP loading incredibly low? That would explain all the hydrogen Figure 2c. Tom Jaramillo has documented near-100% FE to C2+ in CORR based on Cu so something is very badly wrong in the controls.

Response:
Thanks for your valuable comments here. The Cu loading content in the control sample (that is, Cu nanoparticles supported on Ti 3 C 2 T x nanosheets (Cu-NP/Ti 3 C 2 T x )) was 5.2 wt% measured by inductively coupled plasma-optical emission spectrometer 2 / 30 analysis. According to your suggestions, we have also carefully synthesized two other Cu nanoparticles control samples (denoted as Cu-NP/Ti 3 C 2 T x -x; x is the actual Cu loading in wt%, and it is 9.8 or 20.3 here) with a higher Cu NP content by adding more Cu precursor. The corresponding characterization results and synthesis details of Cu-NP/Ti 3 C 2 T x -x have been added as Supplementary Figures 28 and 29 of the revised SI file and in Pages 16 and 17 of the revised main text file, respectively. We have also examined the CO reduction (COR) activity of Cu-NP/Ti 3 C 2 T x -x under the same conditions as those of Cu single atoms (SAs) supported by Ti 3 C 2 T x nanosheets (namely Cu-SA/Ti 3 C 2 T x , Figure 2a Cu-NP/Ti 3 C 2 T x -9.8 and Cu-NP/Ti 3 C 2 T x -20.3 afford improved reduction current densities of -21.97 and -34.0 mA cm −2 at -1.0 V versus the reversible hydrogen electrode (vs RHE) in comparison with that of Cu-NP/Ti 3 C 2 T x (-16.2 mA cm −2 ).
Meantime, Cu-NP/Ti 3 C 2 T x -9.8 and Cu-NP/Ti 3 C 2 T x -20.3 exhibited the maximum Faradaic efficiency (FE) of 54.0% and 59.5% at -0.7 V vs RHE for C 2 products, respectively. However, the obtained two indicators are still inferior to those of Cu-SA/Ti 3 C 2 T x (-52.2 mA cm −2 vs 98%), suggesting that the Cu NP content is not the major contribution to the COR activity. This comparison further confirmed that Cu SAs in Cu-SA/Ti 3 C 2 T x can effectively catalyze CO electrolysis, leading to its high COR activity. We reasoned that the less electrochemical accessible surface and lower intrinsic catalytic activity of active sites induced the inferior COR performance of Cu-NP/Ti 3 C 2 T x , which was verified by measuring the electrochemically active surface area (ECSA) and ECSA-corrected COR polarization curves of Cu-NP/Ti 3 C 2 T x and Cu-SA/Ti 3 C 2 T x (Supplementary Figure 15 in our initially submitted version, it has been renumbered as Supplementary Figure 36 in Page 37 of the revised SI file because some new figures have been added). Furthermore, the theoretical calculations confirmed that the introduction of Cu SAs will induce a relatively low barrier energy for the potential-determining step (Figure 3a,b), which indicates the CO electrolysis is more likely to occur over Cu-SA/Ti 3 C 2 T x .

/ 30
Indeed, Tom Jaramillo et al. (Nat. Catal. 2019, 2, 702) reported Cu nanoflowers, which exhibited a maximum FE of near-100% for C 2+ products at -0.23 V vs RHE. This work is a significant advancement for Cu-based COR electrocatalysts. Following it, our Cu-SA/Ti 3 C 2 T x has shown an impressive COR performance, such as a nearly 100% FE for C 2 products at -0.70 V vs RHE and remarkable stability over 65 h. Moreover, we have read through not only the above Tom Jaramillo's paper but also all other papers about Cu-based materials, in all of which no Cu SAs have been reported for COR. In addition, comparison of Tom Jaramillo's work and our designed control sample (Cu-NP/Ti 3 C 2 T x ) reveals that the low roughness factor of our designed Cu-NP/Ti 3 C 2 T x sample (ca. 10.3) resulted in an enhanced hydrogen evolution reaction (HER) activity, and thus leading to an inferior activity for the formation of C 2 products.
In addition to Supplementary Figures 28-30, the description of the above content has been added in Page 11 of the revised main text file, and Page 31 of the revised SI file. The paper (Nat. Catal. 2019, 2, 702) has been cited as ref. 42 in the reference list of the revised main text file.
2. And, as noted above, the CORR FE to ethylene, and the CORR FE to C2+, are good but not better than prior reports. One explanation for their findings: in their notionally Cu SA catalysts, they loaded up fairly well with Cu (nanoparticles, Cu~ polycrystalline metal, or whatever). In their NP controls, they didn't.

Response:
Thank you very much for your helpful comments. Indeed, our designed Cu-SA/Ti 3 C 2 T x achieved a total C 2 FE of ~98%, which is comparable to that of the best-reported results (please see details in Supplementary Cu-based COR catalysts. This enhanced COR activity for Cu-SA/Ti 3 C 2 T x can be attributed to the abundant exposed active sites (Supplementary Figure 34), the high intrinsic activity (Supplementary Figure 36) and the decreased free energy barrier of the potential-determining step (Figure 3a,b). While for Cu-NP/Ti 3 C 2 T x , as discussed above in comment #1, the inferior COR activity mainly resulted from the low roughness factor, which suppresses the intrinsic activity for the COR. This viewpoint is consistent with the experimental results ( Figure 2c), which showed a maximum FE of 81% for H 2 production. Furthermore, this finding is also confirmed by Tom Jaramillo's work (Nat. Catal. 2019, 2, 702).
The description of the above content has been added in Page 9 of the revised main Thanks for your valuable comments and suggestion. The atomic resolution high-angle annular dark-field scanning transmission electron microscopy (HAADF-STEM) imaging is acquired by an aberration-corrected electron microscopy with a sub-Å-resolution.
In these HAADF-STEM images, the image intensity is approximately proportional to the square of the atomic number (Z). Therefore, the isolated heavier Cu SAs can be discerned in the Ti 3 C 2 T x support because of a different Z-contrast between Cu and Ti.
In this regard, the aberration-corrected HAADF-STEM imaging is a powerful and When more Cu precursor was added in the synthetic process, the co-existence of Cu SAs and abundant Cu nanoclusters could be observed in the Ti 3 C 2 T x support (namely Cu-NC/Ti 3 C 2 T x ), which was confirmed by the newly added Supplementary Figure 8a 0.27 nm. By comparing the statistical result and geometrically optimized Cu dimers model, we can conclude that above 96% of the Cu atoms in Cu-SA/Ti 3 C 2 T x are atomically anchored to the Ti 3 C 2 T x support. Meantime, because of the large distance between the single Cu atoms (average: 0.61 nm), it is unlikely to form a patchy Cu monolayer in Cu-SA/Ti 3 C 2 T x .
To further confirm the atomic dispersion of Cu on the Ti 3 C 2 T x support, the XAS measurements provide deep insights on the macro-scale. Based on the Fourier transform extended X-ray absorption fine structure (FT-EXAFS) analysis (please see details in Figure 1f in Page 6 of the revised main text file), it clearly indicated that the Cu atoms are atomically dispersed; Cu-Cu paths at 2.2 Å are absent. All these results demonstrated that no Cu NPs can be found, and Cu SAs are uniformly dispersed on the Ti 3 C 2 T x support in the Cu-SA/Ti 3 C 2 T x sample.
In addition to Supplementary Figures 1, 4-6 and 8, the description of the above content has been added in Page 5 of the revised main text file and Pages 5-7 and 9 of the revised SI file. 7 / 30 4. Presumably a deep XAS expert will review this paper. I am skeptical of the evidence overall in this work for single Cu atom catalysts, but someone who does

XAS for a living can provide expert review on this topic. A key question is whether
XAS truly is capable of resolving this point i.e. can it distinguish materials that are mostly SA, mostly double-atom, or a blend of various coordination numbers of Cu.
Meanwhile, HAADF-STEM imaging is also used to give a direct observation of the distribution of single atoms. Compared with HAADF-STEM images which reflected regional information, XAS technologies give average information on the atomic structure and electronic structure, including the distance between adjacent atoms, the number and type of coordinated atoms, and the oxidation state of atoms. The successful synthesis of SAs can be directly determined by analyzing the coordination information of metal atoms through the R space data in EXAFS. Of note, some pioneering works (Science 2016, 352, 797; Science 2019, 364, 1091) have also used these two methods for the determination of SA-based catalysts.
To confirm the structure of Cu atoms in Cu-SA/Ti 3 C 2 T x , we performed a fitting for the R region of FT-EXAFS spectra, and the fitting parameters were added in Supplementary Table 1 in Page 43 of the revised SI file. According to the fitting results, the main peak at ~1.6 Å is associated with Cu-O single scattering path which represents the first coordination shell of the Cu-SA catalyst (phase uncorrected). The 8 / 30 smaller peak at ~2.7 Å is associated with the average Cu-Ti single scattering path (second shell) in agreement with the theoretical optimized model. According to the previous reports (refs. 21-23 in our initially submitted version; Science 2016, 352, 797; Science 2019, 364, 1091), obvious metal-metal bonds can be distinguished in R region of FT-EXAFS data if the metal atom aggregates were formed. However, no Cu-Cu bonds (2.2 Å) can be observed in our work, indicative of the extremely small amount of metal dimers in Cu-SA/Ti 3 C 2 T x . Furthermore, we made a statistical analysis of HAADF-STEM images as discussed in comment #3, which indicated that above 96% of the Cu atoms are atomically dispersed. Therefore, according to the XAS and HAADF-STEM results, it is concluded that almost all of the Cu sites in Cu-SA/Ti 3 C 2 T x are single atoms.
As for the coordination numbers of Cu, it is extremely difficult and impractical to make a statistic analysis for the exact distribution of the coordination number of Cu SAs. Alternatively, from the perfect R space fitting results of Cu-SA/Ti 3 C 2 T x model, it indicated an average coordination number of Cu-O of 3.2 ± 0.4, which is also consistent with our Cu-SA/Ti 3 C 2 T x models (see inset of Figure  Thank you for your helpful comments and suggestion. We are so sorry about the ambiguous illustration of Figure 3d, which may be caused by the absence of figure legends representing different atoms. In the revised main text file, a clearer model has been given; hopefully, it gives a clear description of Cu SAs.
As you suggested, a computational chemist has been invited to review our work, and the more detailed discussion can be found in the responses to the comments of Reviewer #4.

/ 30
Responses to the comments of Reviewer #2:

Comment:
The authors performed an interesting study on the Isolated Cu single sites for high-performance electroreduction of CO to multicarbon products. After carefully go through the whole manuscript, some problems can be found as follows:

Response:
Thank you very much indeed for reviewing our manuscript. We greatly appreciate your inspiring, useful, helpful and constructive comments. Details of the corresponding answer and changes made are described below point by point. For your convenience, all changes have been highlighted in yellow in the revised main text and the revised supplementary information (SI) files.
1. in the second paragraph of introduction section, the authors mentioned that the single atom catalysts for electrochemical CO 2 reduction (CO 2 R). The authors can make further discussion in depth on the single atom catalysts for CO 2 R and many two dimensional materials used as single atom catalysts due to the large specific surface area and precisely defined active center. The authors can add one sentence, two dimensional materials is a powerful platform to design the single atom catalysts for CO 2 R. and cite the following references:

The authors should make a comparison of the catalytic activity of the current
Cu-containing single-atom catalyst with the reported Cu-containing CO 2 R catalysts, such as, Cu single crystals, Cu nanoclusters, etc in literature, and highlight the features of present Cu-containing single atom catalyst. This can make the present work more solid and interesting.

Response:
Thank you very much for pointing this out. According to your suggestion, the currently reported Cu-containing catalysts for COR have been added in Supplementary Thank you for this helpful comment. Compared with other substrates, Ti 3 C 2 T x MXene shows great promise in electrochemical applications due to its excellent electronic conductivity, catalytically active basal planes, and graphene-like unique layered

Response:
Thank you for these constructive comments. We fully agree that the EXAFS reflects an average result of the bulk phase. However, up till now, the XAS technology is considered to be a reliable measurement for the determination of single-atom catalysts, as widely reported by the pioneering works (refs. 21, 23, 30  atoms in Cu-SA/Ti 3 C 2 T x are atomically anchored to the Ti 3 C 2 T x support. Only a small fraction of Cu pairs (less than 4%) are formed; however, it is important to keep in mind that such a small portion of Cu pairs does very little contribution to the high COR activity of Cu-SA/Ti 3 C 2 T x .
In addition to the statistical results, according to the comments of Reviewer #1, the HAADF-STEM images of two control samples (i.e. Ti 3 C 2 T x support and Cu nanoclusters on Ti 3 C 2 T x (denoted as Cu-NC/Ti 3 C 2 T x )), have also been added as Supplementary Figures 5 and 8 in Pages 6 and 9 of the revised SI file, respectively. By comparing the HAADF-STEM images of Cu-SA/Ti 3 C 2 T x and Ti 3 C 2 T x , we can also conclude that the Cu SAs uniformly anchored to the Ti 3 C 2 T x support. Meanwhile, with the appearance of Cu aggregated sites in Cu-NC/Ti 3 C 2 T x (new control sample), a prominent peak representing the Cu-Cu bond emerged in FT-EXAFS curves, which is added as Supplementary Figure 8e in Page 9 of the revised SI file. Accordingly, the XAS, HAADF-STEM images and the corresponding statistical results all confirmed the atomically dispersed Cu atoms in Cu-SA/Ti 3 C 2 T x .
As you suggested, we also have performed the CO-DRIFTS measurements, and the corresponding results are added as Supplementary Figure 12 in Page 13 of the revised SI file. Unfortunately, there is no characteristic peak corresponds to linearly bonded CO on Cu SAs in Cu-SA/Ti 3 C 2 T x . This finding may be ascribed to the weakening of CO adsorption on the surface of Cu-SA/Ti 3 C 2 T x . It is widely accepted that CO-DRIFTS characterization is an effective method to confirm a few of noble metal SAs, such as Pt, Pd and Rh SAs (Nat. Commun. 2018, 9, 4454;ACS Catal. 2014, 4, 1546Nat. Commun. 2020, 11, 954).
In addition to Table S1 and Supplementary Figures 5, 6, 8, 11 and 12, the description of the above content has been added in Pages 5 and 7 of the revised main text file, as well as Pages 6, 7, 9 and 13 of the revised SI file.

In Electrochemical COR performances test, the control experiments of pristine
Ti 3 C 2 T x and NaBH 4 reduced Ti 3 C 2 T x without Cu species need to be conducted and compared. In addition, considering both Ti 3 C 2 T x and carbon paper contain carbon species, C 18 O Isotope labeling experiment must be conducted to confirm the products is from CO reduction other than from the reduction of carbon species in Ti 3 C 2 T x and carbon paper. Moreover, the authors used ethanol to prepare the catalyst ink, thus lead the selectivity of ethanol questionable. Can the authors exclude that ethanol measured in solution after electrochemistry is not partly due to 'leaching' of residual ethanol on catalysts?

Response:
Thank you for your inspiring comments and helpful suggestion, according to which the COR performances of the pristine Ti 3 C 2 T x and NaBH 4 reduced Ti 3 C 2 T x (namely R-Ti 3 C 2 T x ) without Cu species have been prepared, their structural characterizations and 34 of the revised SI file, respectively. As seen, the two control samples (that is, Ti 3 C 2 T x and R-Ti 3 C 2 T x ) showed inferior catalytic performance compared to that of Cu-SA/Ti 3 C 2 T x . In detail, the maximum FEs of C 2 H 4 formation for Ti 3 C 2 T x and R-Ti 3 C 2 T x were 7.5% (-0.8 V vs RHE) and 5.2 % (-0.7 V vs RHE), respectively, both of them are significantly lower than that of Cu-SA/Ti 3 C 2 T x (71% at -0.7 V vs RHE). This finding demonstrates that the Ti 3 C 2 T x support help to capture and stabilize the Cu SAs and the impact on the selectivity of COR is negligible.
As you suggested, isotopic labeling experiments were performed to confirm that the carbon in the COR products was derived from CO. 13 CO was used as the feeding gas.
The COR products were analyzed by a quadrupole-type mass spectrometer (MS Based on these results, the carbon source of the evolved gas and liquid reduction products are confirmed to be originated from the CO electrolysis. Note that, the COR activities of the catalysts were evaluated using controlled potential electrolysis with CO-saturated electrolyte for 2 h.
In this work, the error bars correspond to the standard deviations of measurements over three separately prepared samples under the same testing conditions.
In addition to Supplementary Figures 13 and 14

structure of the catalysts has changed and some carbon species in Ti 3 C 2 T x is leached
during COR test. The authors need to address this problem.

Response:
Thank you for this helpful comment. According to the responses to your comment #2, the 13 CO isotopic labeling experiment strongly confirmed that the carbon in the evolved products (including gas and liquid products) originates from the gaseous CO supplied. Further, the elemental analyzer (Vario EL Cube) was used to detect the content of carbon in Cu-SA/Ti 3 C 2 T x before and after the long-term stability test. As expected, the carbon content of the fresh Cu-SA/Ti 3 C 2 T x and the used one is measured to be 35.7 and 36.2 wt%, respectively, indicative of the good stability of Cu-SA/Ti 3 C 2 T x during the COR test.
The description of the above content has been added in Page 17 of the revised SI file.
In addition, the details of the element analysis have also been added in Page 17 of the revised main text file.

In DFT calculations, as it is questionable to claim that Cu species in Cu-SA/Ti 3 C 2 T x
are atomically dispersed (see above comment 1), the model used here will be unconvincing. The case of Cu dimers model is suggested included in calculation for comparison.

Response:
Thank you very much for your valuable comments. According to the responses to your comment #1 and the comments #3 and #4 from Reviewer #1 (please see details in the above responses), we can conclude that almost all of the single Cu atoms are isolatedly dispersed on the Ti 3 C 2 T x support. Therefore, the case of Cu dimers model is not taken into consideration in this work.

/ 30
Responses to the comments of Reviewer #4:

Comment:
The authors report a combined theory/experimental study revealing a single atom copper catalyst supported on MXene with high CO reduction performance. The Faradaic efficiency and selectivity to overall C2 products appears to compare favorably with the literature and to Cu nanoparticles on the same support. The paper is fairly well written and clear in its presentation. This is an important work showcasing a successful application of using single atom catalysts for CORR with good stability and activity. This manuscript should be publishable pending revisions.

Response:
Thank you very much indeed for reviewing our manuscript. We greatly appreciate

Response:
Thank you for your valuable comments and suggestion. Indeed, the surface of

/ 30
MXene is always functionalized with T x groups, including -O, -OH and -F. Besides, a slight amount of Cl element also can be detected (ca. 3 at % by XPS measurement in this work), however, most of the residual Cl can be washed off during the synthesis of Ti 3 C 2 T x MXene nanosheets (as described in the Methods). In addition, the composition of Ti 3 C 2 T x surface functional groups are influenced by the etching method of Ti 3 AlC 2 . In this work, we used LiF and HCl as etching agent, therefore, the amount of -O terminations is supposed to be higher than that of -F and -OH (Phys. Chem. Chem. Phys. 2016, 18, 5099). Moreover, EELS analysis also used to determine the elements present in the Ti 3 C 2 T x flakes, and the result has been added as Therefore, in our initially submitted version, we use the -O terminated one as the model of Cu-SA/Ti 3 C 2 T x .
As you suggested, the -O termination may become hydroxylated in an aqueous solution. Therefore, we carefully checked the thermodynamical stability of Cu-SA/Ti 3 C 2 T x model under the reaction condition. According to the surface pourbaix diagrams of Ti 3 C 2 (ACS Catal. 2017, 7, 494), the MXenes will be oxidized by H 2 O and turn into -OH terminated ones under a relatively negative applied potential.
Therefore, in this revised version, we changed the O-terminated model into the OH terminated ones, and the relative DFT analysis has been recalculated based on this new model. The newly calculated results replaced the original version of Figure 3. In brief, the rate-limiting step (RLS) for Cu-SA/Ti 3 C 2 T x was determined to be *CHO + CO → *CHO-CO. In contrast, the RLS of Cu (111) was 2*CO → *COCO; however, the free energy barrier of the RLS is 0.94 eV, much higher than that of Cu-SA/Ti 3 C 2 T x (0.32 eV).
The theoretical calculation results agree well with the high COR activity of Cu-SAs from the experimental observation (Fig. 2b,c).

/ 30
In addition to Supplementary Figure 38, the description of the above content has been added in Page 12 of the revised main text file. The papers (ACS Nano 2016, 10, 9193;Sci. Bull. 2018, 63, 1397ACS Catal. 2017, 7, 494) have been cited as ref. S7, and refs. 48-49 in the reference list of the SI file and revised main text file, respectively.

What is the binding strength of the single atom on the surface, relative to bulk
Cu? What is the kinetic stability of this site, in terms of surface mobility or agglomeration from single atom to dimer? This should match experimental observations of stability.

Response:
Thank you very much for proposing this vital comments. To estimate the stability of Cu single atom on Ti 3 C 2 T x support, the binding energy of Cu-SA/Ti 3 C 2 T x (E Cu-SA/Ti 3 C 2 T x bind ) was determined. It is calculated to be +2.16 eV, smaller than the cohesive energy of +3.48 eV for bulk Cu (E Cu bulk coh ); however, the binding strength is still quite strong and agrees well with previous reports for SA catalysts (Nanoscale 2018, 10, 17893;Nat. Commun. 2016, 7, 1080. Besides, the kinetic stability that affecting the stability in practical electrocatalysis should be taken into consideration as well. Thus, to estimate the kinetic stability of Cu-SA/Ti 3 C 2 T x , the surface mobility of Cu SAs was discussed by simulating the mobilization of a single Cu atom on Ti 3 C 2 T x surface using  (1(7)) site to an adjacent oxygen top site (1(7)→11), which gives a larger energy barrier of 1.63 eV. This energy barrier is comparable to those reported for Pd SAs (1.67 eV) (Chem. Mater. 2017, 29, 9456) and Cu SAs (0.95 eV) (Nanoscale 2018, 10, 17893) on CeO 2 (111), which indicates that single Cu atom on the Ti 3 C 2 T x surface is difficult to move away from the triangle oxygen traps, suggesting the high kinetic stability of Cu-SA/Ti 3 C 2 T x .
Another possible instability formation for single Cu atoms is the appearance of Cu-dimer species, which are potential CO adsorption sites that would affect and change the total COR mechanism. Therefore, it is necessary to investigate, in the theoretical aspect, whether dimers can be formed in our highly dispersed Cu SAs during the electrolysis. The possible surface agglomeration from single Cu atom to Cu dimer paths and corresponding energy barriers have been calculated using CI-NEB method. The corresponding results have been added as Supplementary   Figure 19 in Page 20 of the revised SI file. It illustrates that two adjacent Cu single atoms need a relatively high energy barrier of 1.23 eV to form a Cu dimer through atom migration, indicating that Cu dimer is difficult to be formed.
The good stability of Cu-SA/Ti 3 C 2 T x can also be evidenced by the atomic-resolution HADDF-STEM images in Supplementary Figure 15, in which no nanoparticle or clusters were observed after the long-term stability test. Therefore, the consistency of the theoretical calculation and experimental results confirmed the good stability of Cu-SA/Ti 3 C 2 T x in the electrocatalysis.
In addition to Supplementary Figures 18 and 19

Response:
We greatly appreciate your constructive and important comments. According to your 27 / 30 comment, we recalculated the free energy profile of to EtOH and C 2 H 4 pathways on Cu-SA/Ti 3 C 2 T x and Cu (111)

Response:
Thank you very much for this helpful suggestion, according to which we compared the reaction free energy profiles of the HER on Cu-SA/Ti 3 C 2 T x and on the Cu (111) surface. The calculated free-energy diagrams have been added as Supplementary   Figure 41 in Page 42 of the revised SI file. As seen, the difference of free energy (∆G H* ) of Cu-SA/Ti 3 C 2 T x is -0.35 eV, much more negative than that on the Cu-NP/Ti 3 C 2 T x (+0.18 eV). Therefore, the greater adsorption energy absolute value of Cu-SA/Ti 3 C 2 T x led to a stable absorption regime, which accounts for the suppression of the HER activity in Cu-SA/Ti 3 C 2 T x in comparison with the Cu-SA/Ti 3 C 2 T x counterpart. This finding is also consistent with the experimental results (Figure 2b,c). In addition to Supplementary Figure  method gives a more intuitively realistic picture of the practical reaction that occurred in the aqueous phase.
In this work, given that our Ti 3 C 2 T x substrate model surface is OH-terminated, and at the same time, the reaction adsorbates are different in each elemental reaction step, therefore, rather different conformations will be produced by reacting with a large number of solvent molecules after consideration of the explicit solvation effect.
Therefore, the full analysis of the energetics of the complex pathways towards C 2+ products after explicit consideration solvent layers would need a lot of time and cost.
However, restricted by our shoestring computational budget, the number and configurations of water molecules included in the explicit solvation model are quite limited, leading to results lack of generality.
Here, the implicit solvation model is a feasible compromise for this issue, which represents solvent as a continuous medium instead of individual "explicit" solvent molecules. It need not consider the complex real solvent molecules, showing an advantage of fast convergence. Previous studies have shown that the implicit model can provide some insights into CO 2 reduction (J. Am. Chem. Soc. 2017, 139, 130;J. Am. Chem. Soc. 2016, 138, 483;Chem 2017, 3, 652;Nat. Commun. 2018, 9, 1320. Therefore, the implicit solvation model is used here using the Poisson−Boltzmann model implemented in VASPsol (Phys. Rev. B 2012, 86, 075140), and the calculated results have been added in newly drawn Figure 3a 8. Pg 13 Line 210 "mechanical" should be mechanistic.

Response:
Thank you for pointing this out. We are so sorry for the mistake we made. The typo error has been corrected. We carefully checked and confirmed that no similar mistakes exist in the revised main text and SI files.

Reviewers' Comments:
Reviewer #2: Remarks to the Author: After carefully check the revised manuscript, I found that the authors have made significant improvements and corrected all the problems the referee proposed. So, I recommend to publish the paper as it is.
Reviewer #3: Remarks to the Author: Although the authors made a major revision based on the reviewers' comments, the structure of catalyst and the identification of the catalytic active sites still remain ambiguous. Therefore, the reviewer cannot recommend the acceptance of this work in Nature Communications.  Figure 3d, the Cu atom is connected to 2 O atoms, and all of the DFT calculations were conducted based on this model; it is hard to connect the structure characterization with the theoretical study, making the theoretical study not meaningful. 4: Something wrong in caption of Figure 3, "d, e Charge density difference of the *COCO-adsorbed and *COCHO-adsorbed configuration in e Cu (111) and f Cu-SA/Ti3C2Tx, respectively." 5: The authors were asked to provide an example NMR spectrum to demonstrate and illustrate the analysis of the liquid products? In the revised version of the manuscript, one NMR spectrum was provided as SI Fig. 23c, however, this NMR spectrum cannot provide any quantitative information of the product concentration. Specifically, they have provided ample evidence for the model system they use, and provided quantification(s) of their stability (formation energy, diffusion barriers, charge state, etc.). The reaction profile is also much more complete and supplemented by reaction barriers. The lower C-C coupling barrier on the single atom system is also enlightening to see. The improved catalytic performance of the modeled single atom system over the Cu (111) surface appears to be convincing.
I don't believe additional revisions are necessary from the theoretical side.

Responses to the Comments
Responses to the comments of Reviewer #2:

Comments:
After carefully check the revised manuscript, I found that the authors have made significant improvements and corrected all the problems the referee proposed. So, I recommend to publish the paper as it is.

Response:
Thank you very much for reviewing our revised manuscript. We greatly appreciate your inspiring and constructive comments.

/ 12
Responses to the comments of Reviewer #3:

Comments:
Although the authors made a major revision based on the reviewers' comments, the structure of catalyst and the identification of the catalytic active sites still remain ambiguous. Therefore, the reviewer cannot recommend the acceptance of this work in Nature Communications.

Response:
We truly thank you for reviewing the revised version of our manuscript and greatly appreciate your helpful and constructive comments. Details of the corresponding answers and changes made are described below point by point. For your convenience, all changes have been highlighted in yellow in the revised main text and the revised Supplementary Information (SI) files.

1: The coordination environment of Cu-SA is unclear, Cu atom coordinated with 3 O atoms or 3 O and 3 Ti atoms? The authors claimed the Bader charge of Cu in
Cu-SA/Ti 3 C 2 T x to be about +0.42, which model is used to calculate the Bader charge?

Response:
Thank you very much for pointing this out. Actually, Cu atom was coordinated with 3 O atoms in the first coordination shell and 3 Ti atoms in the second coordination shell as illustrated by the EXAFS fittings. That is to say, the Cu atom was chemically bonded with 3 nearest O atoms, while the farther 3 Ti atoms also influenced the electronic structure of Cu atoms, which contribute to the peak at 2.7 Å in FT-EXAFS spectrum ( Figure R1a). The distances of Cu atoms to both shells are consistent with our fitting models (Figure 1g). To better illustrate this, Figure R1b  Regarding your second point, the Bader charge of Cu in Cu-SA/Ti 3 C 2 T x is calculated using the optimized DFT model (the results are added as Figure R2), which is 4 / 12 consistent with the model used in EXAFS fitting and theoretical calculations.
The description of the above content has been added in Page 7 of the revised main text file. Figures R1 and R2 have been added as Supplementary Figures 11 and 12 in Pages 12 and 13 of the revised SI file, respectively.
2: Experimental data is required to show that the isolated Cu atom is the real active site for catalytic CO reduction.

Response:
Thank you for the constructive comments. In the last submitted version, the comprehensive characterizations of Cu-SA/Ti 3 C 2 T x , including atomic-resolution high-angle annular dark-field scanning transmission electron microscopy (HAADF-STEM) images and X-ray absorption fine structure (XAFS), have proved that the stable presence of Cu single atoms on Ti 3 C 2 T x . In addition, to validate the high reactivity of Cu-SA/Ti 3 C 2 T x , a series of control samples (i.e. Cu-NP/Ti 3 C 2 T x , Ti 3 C 2 T x and R-Ti 3 C 2 T x ) have been conducted. Cu-NP/Ti 3 C 2 T x shows a very low current density, and the highest Faradic efficiencies (FEs) of C 2 H 4 and EtOH are about 3.4-and 2.1-fold lower than that on Cu-SA/Ti 3 C 2 T x , respectively ( Figure 2b). This circumstance was not improved even increasing the Cu content in Cu-NP/Ti 3 C 2 T x from 5.2 to 20.3 wt%. For pristine Ti 3 C 2 T x and reduced Ti 3 C 2 T x (i.e. R-Ti 3 C 2 T x ), the maximum C 2 H 4 FEs of Ti 3 C 2 T x and R-Ti 3 C 2 T x are significantly lower than that of Cu-SA/Ti 3 C 2 T x , demonstrating that the Ti 3 C 2 T x support helps to capture and stabilize Cu species and the impact on the selectivity of CO reduction is negligible. All the control results indicated that the high CO reduction activity of Cu-SA/Ti 3 C 2 T x comes from the Cu single atoms on Ti 3 C 2 T x .
Those experiment results are also in line with the DFT calculation results, in which Cu-SA/Ti 3 C 2 T x showed much lower free energy barriers for CO reduction than that of Cu (111).

/ 12
According to your comments, additional control experiments have been conducted to further prove the reactivity of Cu single-atom sites for catalytic CO reduction.
Since that the SCNanions can coordinate with Cu atoms and poison the Cu single sites during the catalysis (Small 2019, 15, 1902410;iScience 2019, 22, 97). We therefore have examined CO reduction activity of Cu-SA/Ti 3 C 2 T x in 1 M KOH electrolyte containing 0.1 mM KSCN. The corresponding CO electroreduction results are added as Figure R3. In the presence of SCN -, Cu-SA/Ti 3 C 2 T x exhibits a noticeable reduction in current density, meanwhile, the obtained highest C 2+ Faradic efficiency of Cu-SA/Ti 3 C 2 T x is as low as 26% (-0.7 V vs. RHE), much smaller than that of electrolyte without SCN - (Figure 2a,b). These results demonstrate that Cu single atoms act as the CO reduction sites. In addition, the COR pathway on pure Ti 3 C 2 T x was also calculated by the DFT method and the results are added as Figure R4. Note that the 2*CO → *COCO pathway is infeasible on Cu-SA/Ti 3 C 2 T x model, and the *CHO + CO → *CHO-CO pathway is adopted here. The results show that a large energy barrier of 1.29 eV is required for the activation of CO on Ti 3 C 2 T x , much higher than that on Cu-SA/Ti 3 C 2 T x . This agrees 6 / 12 well with the poor COR performance of Ti 3 C 2 T x . It further confirmed the high COR activity of Cu-SA/Ti 3 C 2 T x comes from the Cu single atoms on Ti 3 C 2 T x . Figure R4. The optimized COR pathway on pure Ti 3 C 2 T x surface.
The description of the above content has been added in Pages 11 and 13 of the Cu-SA/Ti 3 C 2 T x , respectively."

Response:
Thank you very much for pointing this out. We are so sorry for the mistakes we made.
These errors have been corrected. Meantime, we have carefully checked and confirmed that no similar mistakes exist in the revised main text and SI files. Thanks for your helpful comments. As you stated, the 1 H NMR spectrum in Supplementary Figure 23c was conducted under an open-circuit voltage, which was given as evidence for that: the EtOH was generated from the electroreduction of dissolved CO by Cu-SA/Ti 3 C 2 T x rather than the residual EtOH. In the last revised version, we have added the standard curves ( Figure R4a  Cu-SA/Ti 3 C 2 T x at -0.7 V (RHE). Note that Figure R4a,b is the same one of Supplementary Figure 14a,b in the previously revised version.
According to your suggestion, the representative 1 H NMR pattern recorded at -0.7 V vs. RHE for 2 h is added as Figure R4c. The ratios of the areas of the produced acetate 9 / 12 and EtOH (peak 1) to the DMSO peak area were calculated to be 27.98 and 0.05164, respectively. The obtained ratios were then compared to the standard curves ( Figure   R4a,b) to quantify the concentrations of the reaction products. Accordingly, the concentrations of EtOH and acetate (the blue balls in Figure R4a,b) were measured to be 5.4 and 0.88 mM, respectively. In this case, the corresponding formation rates of EtOH and acetate were calculated to be 2.79 and 0.44 mM h −1 at -0.7 V vs. RHE, respectively.
The description of the above content has been added in Page 15 of the revised SI file. Figure R4c has been added as Supplementary Figure 14c in the revised SI file.
6: In the DFT calculations, why is the *CHO + CO → *CHO-CO pathway energetically more favorable for Cu-SA/Ti 3 C 2 T x , as compared to the case on Cu (111) surface? or 2*CO → *COCO on the Cu-SA/Ti 3 C 2 T x ?

Response:
Thank you very much for the very helpful instructions. Actually, both *CHO + CO → *CHO-CO and 2*CO → *COCO coupling processes probably existed in CO reductions (ACS Catal. 2018, 8, 1490. The 2*CO → *COCO process is always reported on Cu bulks, which needs two nearest Cu adsorption sites. According to your suggestion, the 2*CO → *COCO process was also calculated on Cu-SA/Ti 3 C 2 T x . A model of 2Cu/Ti 3 C 2 T x was used, in which two nearest isolated Cu atoms (a distance of 3.2 Å) were constructed on Ti 3 C 2 T x substrate ( Figure R5). Note that, the distance is nearer than the actual average Cu-Cu interatomic distance (0.61 Å) in Cu-SA/Ti 3 C 2 T x (Supplementary Figure 6), and this 2Cu/Ti 3 C 2 T x model was just used for investigating the possibility of 2*CO coupling on Cu-SA/Ti 3 C 2 T x . Accordingly, the geometric optimization was proceeded on *COCO-2Cu/Ti 3 C 2 T x model to investigate the existence of *COCO intermediate. Unfortunately, the energy minimum value point for *COCO intermediate was not found in the local region. Instead, *COCO was 10 / 12 separated to form two isolated *CO ( Figure R5). So the 2*CO → *COCO pathway is infeasible on the 2Cu/Ti 3 C 2 T x surface. Therefore, the *CHO + CO → *CHO-CO pathway is adopted for the calculation of CO reduction pathways of Cu-SA/Ti 3 C 2 T x . Figure R5. The optimization process of *COCO on Cu-SA/Ti 3 C 2 T x .
In addition, the *CHO + CO → *CHO-CO pathway is also calculated on Cu (111). As illustrated in Figure R6, the *CO →*COH becomes the rate-limiting step with a free energy barrier of 1.18 eV, a little higher than that of 2*CO → *COCO step (0.94 eV) on Cu (111). Therefore, the two pathways co-existed on Cu (111), however, they are both energetically more unfavorable that the *CHO + CO → *CHO-CO pathway on Cu-SA/Ti 3 C 2 T x , suggesting the inferior CO reduction activity of Cu (111). Figure R6. The reaction mechanism of the COR on Cu (111) through *CHO pathway.