Electro-reduction of carbon dioxide at low over-potential at a metal–organic framework decorated cathode

Electrochemical reduction of carbon dioxide is a clean and highly attractive strategy for the production of organic products. However, this is hindered severely by the high negative potential required to activate carbon dioxide. Here, we report the preparation of a copper-electrode onto which the porous metal–organic framework [Cu2(L)] [H4L = 4,4′,4″,4′′′-(1,4-phenylenebis(pyridine-4,2,6-triyl))tetrabenzoic acid] can be deposited by electro-synthesis templated by an ionic liquid. This decorated electrode shows a remarkable onset potential for reduction of carbon dioxide to formic acid at −1.45 V vs. Ag/Ag+, representing a low value for electro-reduction of carbon dioxide in an organic electrolyte. A current density of 65.8 mA·cm−2 at −1.8 V vs. Ag/Ag+ is observed with a Faradaic efficiency to formic acid of 90.5%. Electron paramagnetic resonance spectroscopy confirms that the templated electro-synthesis affords structural defects in the metal–organic framework film comprising uncoupled Cu(II) centres homogenously distributed throughout. These active sites promote catalytic performance as confirmed by computational modelling.

a remarkable potential to start the reduction of CO2 to formic acid at -1.45 V compared to Ag/Ag+, which represents a record value for CO2 electro-reduction in an organic electrolyte compared to other studies in literature. These rather remarkable results in terms of conversion of CO2 to formic acid have been the subject of work in electrochemistry but also in EPR in order to understand the active sites that support the reactivity of the material. The authors thus showed unambiguously that this electroreductive activity is supported by monomeric copper defects while the binuclear sites are only slightly affected. These results are supported by a multi-frequency EPR approach (X-band and Qband). I find it a pity that the extremely representative Figure S8 is not incorporated in the body of the paper. One of the very innovative aspects of this paper is to have combined a multi-technique experimental and theoretical approach leading to relevant and very convincing results. From my point of view this paper is very innovative in the field of CO2 remediation by electrochemical techniques.

Détail comments
Page 3: The authors describe the binuclear system as strongly coupled by a strong antiferromagnetic exchange. 1-Did the authors measure this value experimentally by varying the temperature? 2-have the authors been able to observe the forbidden transition at half field as it is often the case for copper binuclear complexes? 3-In the simulation of the spectra, did the authors introduce this J value? If this is the case, the authors could add this parameter in the S2 tables. Figure S19 : 1 -was this experiment carried out in operando where these measurements were made ex-situ. If they were carried out in opérando it would be interesting to mention it in the manuscript. 2-it is interesting to see that the EPR measurements are fully consistent with the progressive loss after 75 min of mononuclear Cu(II) sites and the XPS is interpreted as the formation of metallic copper. Did the authors try to detect the presence of this species by lowering the temperature for example, or by working at very low amplitude modulation and low microwave power. Is it conceivable that the reduction of mononuclear sites is not done at 2 electrons but at 1 electron which would lead to the formation of Cu(I). page7 : on the EPR analyses section, should provide the amplitude modulation used but also the code they have used for spectra simulation.

Conclusion
In conclusion, this work brings new results regarding CO2 electro-reduction and in particular the fact that only the mono nuclear species of copper are active. Moreover, these electro-synthetized materials have a remarkably low potential compared to what is described in the current literature. For all these reasons I think that this article has its place to be published in Nature Communications which will allow a wide dissemination of these results and should stimulate research on CO2 electro-reduction based on such electrode materials.
Reviewer #3 (Remarks to the Author): The authors report a presumably successful synthesis for creating a defect-rich MOF catalyst for CO2 reduction to formic acid with low overpotential. The authors argue the high activity is driven by coordinatively unsaturated copper centers in the defect-rich catalyst. However, a convincing causal link between higher concentration of unsaturated Cu sites and activity is lacking; the obvious induction period of the catalyst and the co-evolution of Cu(0) along with uncoupled Cu(II) points to at least several alternative hypotheses for the active site besides the uncoupled Cu(II). Furthermore, the DFT calculations are incomplete and fail to support the argument of CO2 reduction on the uncoupled Cu(II) sites. In light of these concerns regarding the conclusions of the paper I would not recommend publishing in Nature Communications. Some specific comments below: 1. Figures showing the DFT geometries of the intermediates from CO2 (adsorbed) to formic acid should be supplied in the SI.
2. The hydrogen binding free energies must be explicitly calculated on the same site to verify the the similar suppression of HER experimentally.
3. Generally the path to produce formate is via a HCOO intermediate, not a COOH intermediate, as shown by many theoretical studies on copper (COOH leads to the formation of CO instead). At the very least, both HCOO and COOH pathways should be calculated on the pristine and defect sites. In addition the path to form CO should also be considered as a comparison. There are many references regarding this, to name a few: ACS Catal. 2017, 7, 7, 4822-4827;J. Phys. Chem. Lett. 2015, 6, 20, 4073-4082 4. The binding free energy of COOH is hard to rationalize in terms of the experimental performance. The binding energy appears to be very positive at 2-2.5 eV. This seems unreasonably high (and would point to a very high overpotential) and suggests these sites are not actually active for CO2 reduction. A comparison with the pure Cu surface should also be provided using the same computational method (e.g. in a recent DFT work here: J. Phys. Chem. C 2018, 122, 21, 11392-11398. the values are much lower in energy on metallic copper.) 5. Solvation effects are completely omitted in either the calculation or discussion.

We thank the reviewers for their constructive comments and our responses are given below in bold italics. We attach files for the revised manuscript and SI with all changes highlighted in yellow for clarity and have also submitted clean versions of both.
Reviewer #1 (Remarks to the Author): Comments: In this manuscript, the author fabricated a metal-organic framework decorated copper cathode for CO 2 electroreduction, namely the Cu 2 (L)-e/Cu, which exhibited remarkable performance. The utilization of metal-organic frameworks as electrocatalyst is a less explored research area. However, the technical quality and discussion of mechanism need improvements. Therefore, I suggest major revision before publication in Nature Communications. The comments are as follows: 1. Cu catalysts are well known to produce unique C2+ products with low selectivity. So many researchers struggled to raise the FE of C2+ for Cu catalyst. The significance to produce formate with Cu catalysts should be rationalized and justified in the introduction part.

Revised. Additional discussions have been added to the introduction on page 1.
2. There are many literature reports on CO 2 RR to formate (for example, Sn-based and Bi-based catalysts). The authors are suggested to compare their catalyst with the state-of-the-art and justify their results. Besides, the description of the catalytic performance section compared with the state-of-the-art should be given as the more relevant data in charts or tables. Supplementary Table 3. 3. The article made some mistakes and errors. For example, the description of legend does not correspond to the figure. In figure 1, the scale bars of b, c and d are 300 um, 300 nm and 100 nm, respectively. Corrected. We apologise for this typo.

Revised. A comprehensive comparison with literature examples is now given in
4. The catalyst was formed by electro-synthesis of the MOF on copper foam. However, the Cu foam itself already has been explored as electrocatalyst (ACS Catal. 2016, 6, 6, 3804.) and could produce a variety of products. In this case, the polycrystalline MOF coating could probably not cover the copper foam entirely. But why was formic acid the only carbon-containing product? This counterintuitive phenomenon and its mechanism should be throughout discussed, and the in-situ technology is probably needed to explore the "real" catalysts. The free Cu-foam electrode with FE HCOOH of 20.5% should be discussed in the context of Cu 2 (L)-e/Cu and Cu 2 (L)-t/CP. Fig. 1. We have added this reference and revised the manuscript accordingly to clarify this point on page 5. 4. Furthermore, from the aspect of MOFs, Cu-carboxylate MOFs, such as HKUST-1 (J. Am. Chem. Soc. 2018, 140, 36, 11378) and 2-D Cu-MOF (Chem. Sci., 2019, 10, 2199) underwent the calcination or in-situ transformation to be the "catalyst" and enhance the stability in water or ionic liquid electrolytes. The real catalyst is not intact Cu-carboxylate MOF anymore in the previous two cases. In this paper, the active site was described as the copper center with uncoupled electron in Cu 2 (L), which was structurally unclear to the readers. Is it still MOF? Furthermore, why the decoration method could generate the highly active uncoupled electrons in MOFs? Please try to tune the number of uncoupled Cu (II) active sites in the Cu 2 (L)-e/Cu system and test the catalytic activity.

As an aside, turning a strongly antiferromagnetic system with an exchange coupling ~ 300 cm -1 that contains four carboxylate bridging ligands into a ferromagnetic species (thus mimicking an uncoupled Cu(II) site) that would have the exact spectral parameters of the excited S = 1 spin state of the [Cu 2 (OOCR) 4 ] systems is highly unlikely.
5. Is Cu 2 (L)-e/Cu still catalytically active in the water? Or other electrolyte systems.

Unfortunately, Cu 2 (L)-e/Cu is not stable in water. This is consistent with reported results (ref. 15) that Cu-MOFs generally have poor stability in aqueous electrolyte for CO 2 reduction and decompose to copper oxides rapidly. This has been added to page 5.
6. The EPR measurement of the organic ligand, pure Cu foam and ionic liquid should be performed to check the background of EPR signal and give a fair comparison. Figs. 5, 6, 19).

Revised. These spectra have been added to the revised manuscript (Supplementary
7. The quantities of uncoupled Cu (II) sites slightly decreased after a peak maximum (Figure 4a), which may be indicated the decomposition of catalysts. The long-term stability of the materials should be evaluated. Supplementary Figs. 21, 22 and 23. 8. "The onset potential for reduction of CO 2 to formic acid at −1.45 V vs Ag/Ag+, representing a record low value for electro-reduction of CO 2 in an organic electrolyte." should be revised. See Chem. Sci. 2019, 10, 2199.

Revised. This reference has been cited as ref 17 and text modified on page 5.
9. Please provide the details about the simulation, especially the different structural models for Cu 2 (L)-e and Cu 2 (L)-t. The authors must consider the potential in-situ structural changes during electrocatalysis in ILs.

Revised. The structural model and atomic coordinates for all intermediates used in the DFT calculations have been added to the SI (page 21 -52). We agree entirely with the reviewer that there may well be potential in situ structural changes during electrocatalysis. However, the exact nature of this can hardly be predicted or modelled accurately by DFT. The potential impact of structural change of electrodes has been added in the revised manuscript on pages 6 and 7, and the new ref 42 highlights reduction of Cu(II) to Cu(I) in a paddlewheel MOF with concomitant distortion of the coordination sphere.
Reviewer #2 (Remarks to the Author): The manuscript entitled "Electro-reduction of CO 2 at low over-potential at a metal-organic framework decorated cathode" report the synthesis of a Cu-elcetrode. This decorated copper-based electrode has a remarkable potential to start the reduction of CO 2 to formic acid at -1.45 V compared to Ag/Ag+, which represents a record value for CO 2 electro-reduction in an organic electrolyte compared to other studies in literature. These rather remarkable results in terms of conversion of CO 2 to formic acid have been the subject of work in electrochemistry but also in EPR in order to understand the active sites that support the reactivity of the material. The authors thus showed unambiguously that this electroreductive activity is supported by monomeric copper defects while the binuclear sites are only slightly affected. These results are supported by a multi-frequency EPR approach (X-band and Q-band). I find it a pity that the extremely representative Figure S8 is not incorporated in the body of the paper. One of the very innovative aspects of this paper is to have combined a multi-technique experimental and theoretical approach leading to relevant and very convincing results. From my point of view this paper is very innovative in the field of CO 2 remediation by electrochemical techniques. Supplementary Fig. 8 has been moved to the main text (Figs. 2h, 2i)

. This makes Figure 2 rather large and complex but given the on line nature of the journal this is acceptable in our view.
Page 3: The authors describe the binuclear system as strongly coupled by a strong antiferromagnetic exchange. 1-Did the authors measure this value experimentally by varying the temperature? (298 cm -1 ; H Güdel, Inorg. Chem.  1979, 18, 1021), and magnetic susceptibility studies (315 cm -1 ; ref. 34). The observed EPR spectra at room temperature are due to the excited triplet (S = 1) spin state, and have very specific g, D, E characteristics: 2.35 ± 0.03, g x,y = 2.07 ± 0.03, D = 0.34 ± 0.03 cm -1 , and E~0, and display a 'half-field' forbidden   transition (ms =± 2). Our recorded spectra are typical for such a Cu (II) paddlewheel structure (ref. 50). In this work, we conducted electro-reduction of CO 2 at room temperature, and conducted EPR studies at the same temperature aiming to monitor changes in MOF structure during the electrolysis.

The magnetic behavior of copper(II) acetate and related Cu(II) paddlewheel systems is well known, and is marked by strong antiferromagnetic coupling between Cu(II) ions leading to a well isolated singlet (S = 0) ground state. The singlet-triplet energy gap has been measured by INS
2-have the authors been able to observe the forbidden transition at half field as it is often the case for copper binuclear complexes? We observed the forbidden transition at ~5300 G in Fig. 2g and Supplementary Figs. 8a, 8b, confirming  the S = 1 spin state of a binuclear Cu 2 entity. We have added this to the text on page 3. 3-In the simulation of the spectra, did the authors introduce this J value? If this is the case, the authors could add this parameter in the S2 tables.

Revised. Simulation of the EPR spectra of S = 1 species involved a spin Hamiltonian that included the axial and rhombic zero-field splitting terms (D and E) in order to compare the results to known paddlewheel systems. We have added the equation of the Spin Hamiltonian in the main text and provided additional information on the modelling of EPR data on page 8 of the manuscript.
Figure S19 : 1 -was this experiment carried out in operando where these measurements were made ex-situ. If they were carried out in operando it would be interesting to mention it in the manuscript. Supplementary Fig. 19 2-it is interesting to see that the EPR measurements are fully consistent with the progressive loss after 75 min of mononuclear Cu(II) sites and the XPS is interpreted as the formation of metallic copper. Did the authors try to detect the presence of this species by lowering the temperature for example, or by working at very low amplitude modulation and low microwave power. Is it conceivable that the reduction of mononuclear sites is not done at 2 electrons but at 1 electron which would lead to the formation of Cu(I). page7 : on the EPR analyses section, should provide the amplitude modulation used but also the code they have used for spectra simulation. Supplementary Fig. 20

), with additional ref 42 supporting Cu(I) formation. Amplitude modulation and code used for EPR simulation have been added to the revised manuscript on page 8 of the manuscript.
In conclusion, this work brings new results regarding CO 2 electro-reduction and in particular the fact that only the mono nuclear species of copper are active. Moreover, these electro-synthetized materials have a remarkably low potential compared to what is described in the current literature. For all these reasons I think that this article has its place to be published in Nature Communications which will allow a wide dissemination of these results and should stimulate research on CO 2 electro-reduction based on such electrode materials.
The authors report a presumably successful synthesis for creating a defect-rich MOF catalyst for CO 2 reduction to formic acid with low overpotential. The authors argue the high activity is driven by coordinatively unsaturated copper centers in the defect-rich catalyst. However, a convincing causal link between higher concentration of unsaturated Cu sites and activity is lacking; the obvious induction period of the catalyst and the co-evolution of Cu(0) along with uncoupled Cu(II) points to at least several alternative hypotheses for the active site besides the uncoupled Cu(II). Furthermore, the DFT calculations are incomplete and fail to support the argument of CO 2 reduction on the uncoupled Cu(II) sites. In light of these concerns regarding the conclusions of the paper I would not recommend publishing in Nature Communications.

Additional EPR experiments and DFT calculations have been conducted to strengthen the link between active Cu sites and catalytic activity. Please see below.
Some specific comments below: 1. Figures showing the DFT geometries of the intermediates from CO 2 (adsorbed) to formic acid should be supplied in the SI.

Revised. These have been added in SI.
2. The hydrogen binding free energies must be explicitly calculated on the same site to verify the the similar suppression of HER experimentally. Fig. 28. 3. Generally the path to produce formate is via a HCOO intermediate, not a COOH intermediate, as shown by many theoretical studies on copper (COOH leads to the formation of CO instead). At the very least, both HCOO and COOH pathways should be calculated on the pristine and defect sites. In addition the path to form CO should also be considered as a comparison. There are many references regarding this, to name a few: ACS Catal. 2017, 7, 7, 4822-4827;J. Phys. Chem. Lett. 2015, 6, 20, 4073-4082 Fig. 25-27) Fig. 25). This discussion has been added to the revised manuscript (pages 6-7) and the two references cited above have been added as refs 46 and 47.

. The formation of O-bound HCOO is more facile over both pristine and defect Cu 2 (L) than formation of C-bound COOH. Particularly, defect Cu 2 (L) has a much lower energy barrier for the formation of O-bound HCOO, resulting in the production of formic acid. This also excludes the possibility of CO or other C-containing products being formed due to the pathway that requires C-bound COOH being less favoured (Supplementary
4. The binding free energy of COOH is hard to rationalize in terms of the experimental performance. The binding energy appears to be very positive at 2-2.5 eV. This seems unreasonably high (and would point to a very high overpotential) and suggests these sites are not actually active for CO 2 reduction. A comparison with the pure Cu surface should also be provided using the same computational method (e.g. in a recent DFT work here: J. Phys. Chem. C 2018, 122, 21, 11392-11398. the values are much lower in energy on metallic copper.) Both O-bound HCOO and C-bound COOH are plausible intermediates. We have calculated these two pathways over pristine and defect Cu 2 (L) electrodes. Pristine Cu 2 (L) shows higher binding free energy over these two pathways, which suggests indeed a low activity for CO 2

reduction. Defect Cu 2 (L) has an evidently lower free energy toward the O-bound HCOO pathway, which is comparable with other efficient electrodes. Pure copper has a low binding energy and is considered as an effective electrode for CO 2 reduction. The DFT analysis reported in the above reference (ref. 48 in the revised manuscript) has confirmed that pure copper is prone to generate both C-bound COOH and O-bound HCOO intermediates,
yielding various C-containing products. In this work, formic acid was the only product from CO 2 , and therefore we believe Cu 2+ is the active sites in Cu 2 (L). Additional DFT analysis and discussions have been added to the revised manuscript.
5. Solvation effects are completely omitted in either the calculation or discussion. Supplementary Fig. 27). For hydrogen evolution solvation effects are less dramatic (new Supplementary Fig. 28). This has been added to the revised manuscript.