Low coordination number copper catalysts for electrochemical CO2 methanation in a membrane electrode assembly

The electrochemical conversion of CO2 to methane provides a means to store intermittent renewable electricity in the form of a carbon-neutral hydrocarbon fuel that benefits from an established global distribution network. The stability and selectivity of reported approaches reside below technoeconomic-related requirements. Membrane electrode assembly-based reactors offer a known path to stability; however, highly alkaline conditions on the cathode favour C-C coupling and multi-carbon products. In computational studies herein, we find that copper in a low coordination number favours methane even under highly alkaline conditions. Experimentally, we develop a carbon nanoparticle moderator strategy that confines a copper-complex catalyst when employed in a membrane electrode assembly. In-situ XAS measurements confirm that increased carbon nanoparticle loadings can reduce the metallic copper coordination number. At a copper coordination number of 4.2 we demonstrate a CO2-to-methane selectivity of 62%, a methane partial current density of 136 mA cm−2, and > 110 hours of stable operation.

6. Similarly, I found the experimental methods section to be significantly lacking detail regarding XAS data. A better description of the XAS experiments is absolutely needed since these are such an integral part of this study. The fitting method for the EXAFS data needs to be described: Which software was used, what fitting parameters were used, which parameters were varied or held constant, etc. All EXAFS fitting results need to be included and presented in a table (coordination number, bond length, amplitude reduction factor, debye waller factor, E0, r-factor, etc.). Over what range (k and R) were the data fit? Which scattering paths were used to fit the data? At the moment there is nothing to really indicate the quality of the EXAFS fitting. Ideally, an example of one of the fits in k-space and r-space (with the real or imaginary component) should be shown.
7. The second main point that currently appears unclear from the article is the timeframe of the transformation of the CuPC to Cu nanoparticles. Could the author develop on that point? In particular, how long do CNP:CuPC samples need to be reduced before they reach optimal selectivity for CH4? Is there a break in period or some preconditioning applied where the CO2RR is increasing as the catalyst is formed? Or are the data in Figure 2c/d collected immediately after assembling the cell?
8. In the same line, is there a reason that the XAS results are shown for OCC (i.e. 0 min), 3, 6, and 10 min for the in-situ characterization of CuPC while for the preparation of 4:1 CNP:CuPC the first data point is after 25 minutes? Is there anything interesting that happens within the first few minutes that shows the formation of the catalyst? From the XANES it seems that all the Cu(II) is reduced on a shorter time scale for the 4:1 sample. 9. Page 8, line 10-12 the authors state "The XAS characterisation employed here yields an average coordination number within the first ~100 nm of the bulk, and thus provides a conservative (e.g. high) estimate of the coordination number at the surface.
I am not sure what is meant by this statement. How big do the authors expect the Cu particles to be? From the post-electrolysis TEM image (supp. Figure 9) it appears they should be around 2-5 nm, so XAS should be very representative of what is happening at the surface. Anything much larger than 20 nm in diameter would only yield information related to the bulk.
10. The reported electrolyte concentration seems rather low (0.05 M KHCO3). Could the authors comment on this choice? 11. The XPS figures in the SI (Supp. Figure 5) are very difficult to read because of the color choice. The light grey color used to plot the measured signal is nearly impossible to read against the white background. This needs to be updated to a darker color that is easily visible. The total fit envelope of the data also needs to be shown and compared to the measured signal, not just the fits of the deconvoluted peak components.
12. From these XPS data, I would argue that a significant amount of CuPC remains even after electrolysis, considering that the CuPC peaks in the Cu 2p and N-Cu peaks in the N 1s are clearly visible in all samples after electrolysis, and also in the XRD. A quantitative analysis should be applied to the XPS data if possible in order to identify how much CuPC remains after electrolysis. The incomplete conversion of CuPC to CNP:CuPC also seems to be visible in the EXAFS shown in Supp. Figure 3b and 3c (i.e. the peak remaining around 1.5 A), which may also contribute to the lower overall CNs that are derived. Do the authors have a hypothesis on this point? 13. The authors state that "the higher magnification TEM images taken pre-and post-electrolysis also support the claim that no sizable metallic Cu nanoparticles were formed during the reaction while also suggesting that the metallic Cu was in an amorphous state during electrolysis." What do the author consider sizeable metallic Cu nanoparticles? Based on the comparison of the TEM images, it seems that some particles in the range of 2-5 nm could be suspected. I recommend adding STEM and EDX mapping to the paper to help with identifying the presence or absence of small Cu nanoparticles. Being able to map the distribution of Cu before and after electrolysis would be highly desirable.
Some additional minor issues: 14. In the SI Figure 2 caption, "(b) The Faradaic Efficiency (FE) toward hydrogen with different CNP to CuPc ratios." is written twice.
15. Supp. Figure 11 is not referenced in the article despite its interesting content and should be mentioned in the main text. The study focuses on the ultra-low coordination number copper catalysts for electrochemical CO2 reduction, which has good methane selectivity and stability. However, the key argument of this paper, XAS analysis of ultra-low coordinated copper catalysts, looks invalid and incomplete, which needs more explanation and transparency. The ultra-low coordination number obtained from this paper actually indicate the Cu-metal with nano-size, which can be quantitatively estimated with high quality data fitting (A. I. Frenkel, et al. Annu. Rev. Anal. Chem. 4, 23-39, 2011; Z. Weng, et al., Nature Communications, 9, 415, 2018). Previous report has shown the effective of nano-cluster Cu metal converted from CuPc as the active site for CO2 reduction (Z. Weng, et al., Nature Communications, 9, 415, 2018). I don't see any significant new things that warrant the publishing in Nat Comm.

Are the XANES and EXAFS shown in
In addition, I have the following comments for authors to address for the improvement and possible submission to other journals.
1. The author argued that the mixture of CuPc with CNP can help form the metallic Cu with nanosize to catalyze the reaction. It is possible the pure CuPc catalysts authors prepared have low surface-tobulk ratio. The surface CuPc convert to Cu metals due to the surface reactions. XAS is a bulk technique that cannot detect the surface changes, thus no formation of Cu is observed. CuPc to CNP with 1:7 ratio was used for electrochemical performance but unfortunately not used for XAS analysis. This part is critical to create the turning point if authors conclusion can be held. I will suggest to add another data point, say CuPc to CNP with 1:10 ratio for both electrochemical and XAS analysis, in addition to 1:7 ratio one for XAS analysis. This is to guarantee the effective of the volcano shape for both electrochemical performance and XAS analysis.
2. Only XAS measurements with controlled current are performed. It will be interesting and more important to perform XAS under voltage-controlled mode, namely at different applied voltage to see the reversible or irreversible changes of CuPc to Cu, similar to what report in previous study (Z. Weng, et al., Nature Communications, 9, 415, 2018). This is to confirm the metallic Cu as the true catalytic site for this reaction.
3. It seems the Method section has been written in a rush, at least it does not allow experimentally reproducing the results presented in this study. Please provide a detailed Method section for XAS experiment and data analysis, such as in-situ cell design and XAS analysis package.
4. Please provide the detailed method to determine all sample compositions by XAS (using XANES or EXAFS? Linear combination fit or linear function analysis?) and cited related reference.
5. The author argued that 4:1 ratio sample stabilized after 50 minutes; however, the XANES and first derivative of XANES do not show any difference between 25 mins and 50 mins. The author should provide the enlarged view of XANES, full XAS spectra, and EXAFS k-space spectra to prove the stable after 50 minutes.
6. The author should provide the method of how to get the Cu-Cu coordination number such as 4.2. The percent of composition cannot determine the Cu-Cu coordination number. EXAFS fitting needs to be done since the author did not mention any analysis method of EXAFS. Suppose the author did run the EXAFS fitting, the fitting table with all necessary fitting parameters (such as amplitude reduction factor, scattering path length, K-range, R-factor, Debye-Waller factor, and the corresponding error bars). The fitting of EXAFS k-space and comparison of EXAFS k-space also need to be provided to see the actual differences.

REVIEWER COMMENTS
Reviewer #1 (Remarks to the Author): In this article, the authors investigate electrocatalysts for the conversion of CO2 to CH4.
Specifically, the authors find via computational studies that Cu in a low coordination environment favors the conversion of CO2 to CH4 in alkaline media as opposed to multi-C products. The authors then develop a strategy to use carbon nanoparticles to confine Cu to an ultra-low coordination environment in their prepared MEAs. By increasing the amount of C nanoparticles, the authors show that the Cu coordination number can be reduced to 4.2, which hinders the formation of C-C products and promotes CH4 formation. Such a study of the restructuring of Cu-based site isolated catalysts and its tuning to modulate product distribution is of prime importance for the field, and fully justifies its publication in this journal.
However, this article would require some corrections and additions before the work could be published, mainly regarding the processing and analysis of XAS data. A list of comments and questions can be found below:

Response from authors:
We appreciate the reviewer's suggestions, and we have revised the manuscript and SI based on the specific comments below.
1. Page 6, the authors state "The ratio of 7:1 is the exception to this trend, as all CO2RR product FEs, including CH4 and C2H4, decreased while the hydrogen evolution increased ( Supplementary Fig. 2b). We attribute this selectivity shift to a shortage of Cu reaction sites, suggesting a trade-off between methanation and reduced CO2RR activity as the ratio of CNP to CuPc is increased" This formulation is not very clear. Are the authors claiming a shift in selectivity originating from a mass-transport limitation of CO2 to the active sites? This statement may need to be reformulated.

Response from authors:
Increasing the CNP-to-CuPc ratio decreases the Cu active sites and increases the C active sites. The C sites produce predominantly H 2 (Supplementary Fig. 1), whereas the Cu sites are the only ones capable of reducing CO 2 to methane. As the Cu sites become less prevalent, we expect more H 2 to be generated. We have modified the text to better reflect this: "As the CNP active sites increased relative to that of the Cu, more H 2 production is expected because CNP active sites cannot perform CO 2 RR and instead produce H 2 ( Supplementary Fig. 1). A lower density of Cu sites thus lowers CO 2 RR activity. This suggested a trade-off between methanation and reduced CO 2 RR activity as the ratio of CNP to CuPc was increased." 2. The XANES and EXAFS for 4:1 CNP:Cu do not seem to agree so well. The XANES indicate that all of the Cu(II) is reduced to Cu(0) within the first 25 minutes and that edge position remains stable afterwards. This is further highlighted by the first derivative ( Figure 3c). The EXAFS, on the other hand, show that the Cu(II) is gradually reduced and a significant amount remains after 25 minutes and there is even still some present after 50 minutes (Supp. Figure 3c). In comparison, the XANES for 1:1 CNP:CuPc also seem to be in disagreement. The XANES (Figure 3b) seem to show a gradual shift from Cu(II) to Cu(0) with time while the EXAFS show that most of the Cu(II) has been converted to Cu(0) within the first 20 min and changes only slightly afterwards (SI Figure   3b).
Could the authors explain these discrepancies? Is it possible that the spectra in Figures 3b and 3c were accidently switched?

Response from authors:
The EXAFS spectra of the 4:1 sample were incorrectly labelled and had since been corrected with the proper data ( Fig. 3 and Supplementary Fig. 6); we have also repeated the XANES measurements on the 1:1 sample ( Fig. 3 and Supplementary Fig. 4). These 4. Page 8, lines 1-2, the authors state "The fitted metallic Cu-Cu bond scattering 1 path spectra in Fig. 3d were plotted based on fitting parameters shown in Supplementary Table   1." However, this table does not really provide a summary of the fit. What do the CuPC and Cu ratios refer to? This is not explained or described in the text at all. A table of the fitting parameters including the CN, bond length, debye waller factor, e0, r-factor, etc. used or calculated for each scattering path that was included in the fit should be provided for each sample at each potential.
In addition, it appears from that table that the EXAFS fits were carried out assuming only Cu-Cu coordination. Wouldn't the author expect some Cu-C bonds once CNPs are added?
If very small Cu particles are considered, then some evidence of Cu-C coordination from the CNPs would be likely, at least from where they are anchored to the CNP surface; could the author comment on that point?

Response from authors:
Supplementary of X-ray absorption near edge structure (XANES) using copper foil and pure CuPc as the standards in the Athena software 1 . The fitting range is from -20 to 30 eV." We have provided more details about our EXAFS analysis and curve fitting to the SI. For example, the fitting factors are now described in more detail in the Supplementary Table 2. We also clarify in the SI that we used multi-path fitting to consider Cu-C bonds (although none were detected): "Multi-path fitting was considered for Cu-Cu paths, Cu-N paths (within CuPc molecules), and Cu-C paths (interface between the metallic Cu cluster and CNP). The contribution from Cu-C paths was negligible and the fitting using two paths (Cu-Cu and Cu-N) matched well with the experimental spectra ( Supplementary Fig. 5). This result confirmed that there were no detectable Cu-C paths formed once CNP was added. The fitting results were listed (shown in Supplementary Table 2) using only two-path fitting (Cu-Cu and Cu-N) to clarify these shells in the EXAFS spectra. These fitting results were also consistent with the XANES fitting using linear combination results.
. 5. Similarly, I found the experimental methods section to be significantly lacking detail regarding XAS data. A better description of the XAS experiments is absolutely needed since these are such an integral part of this study. The fitting method for the EXAFS data needs to be described: Which software was used, what fitting parameters were used, which parameters were varied or held constant, etc. All EXAFS fitting results need to be included and presented in a table (coordination number, bond length, amplitude reduction factor, debye waller factor, E0, r-factor, etc.). Over what range (k and R) were the data fit?
Which scattering paths were used to fit the data? At the moment there is nothing to really indicate the quality of the EXAFS fitting. Ideally, an example of one of the fits in k-space and r-space (with the real or imaginary component) should be shown.

Response from authors:
We have added these details, and a table, to the SI. Regarding our XAS experimental measurements and the fitting parameters:   applied where the CO2RR is increasing as the catalyst is formed? Or are the data in Figure 2c/d collected immediately after assembling the cell?

Response from authors:
Our approach was to compare the performance of the different CNP:CuPc samples at steady state conditions. We waited 2 hours prior to taking gas product samples to accommodate any restructuring. Our XANES data (Fig. 3a-c)

Response from authors:
We repeated the XANES measurements of the 1:1 sample (see updated Fig. 3 and Supplemental Fig. 4) and the new results are more consistent with that of the 4:1 sample (which was previously mislabelled as explained in our response to concern #2): specifically, they both exhibit significant change within the first 20 minutes and minimal change after. It was very challenging to make the in-situ measurements for 1:1 and 4:1 samples because the low copper concentrations gave weak fly scan results. To compensate for the faint signals in these scans, we used longer scan times at the expense of some temporal resolution. Moreover, we were not concerned to analyse the 1:1 and 4:1 at the shorter time scales because we knew, from our prior experiments, that they were relatively stable over the span of hours and we wanted to showcase the coordination stability at longer time scales. We have added some clarification to the text:

Response from authors:
We agree with the reviewer that our XAS measurements should be representative of the surface since our cluster size is only 2-5 nm (Supplementary Fig. 15 & Fig. 16). We have deleted this discussion from the text and mentioned the size discussion in MS and SI: "STEM/EDX and TEM images taken pre-and post-electrolysis also support the claim that Cu nanoclusters (~2-5 nm) were formed during the reaction (Supplementary Fig. 15 & Fig. 16).  Figure 5) are very difficult to read because of the color choice. The light grey color used to plot the measured signal is nearly impossible to read against the white background. This needs to be updated to a darker color that is easily visible. The total fit envelope of the data also needs to be shown and compared to the measured signal, not just the fits of the deconvoluted peak components.

Response from authors:
We have replotted the figures accordingly: "

Response from authors:
We agree that there is some CuPc left after reaction as supported by our in-situ XANES and ex-situ XPS measurements. We have attempted to quantify the amount of CuPc remaining by normalising the N-Cu peak by the N=C peak, as we now describe in the SI: Supplementary "The deconvolved N 1s peaks demonstrate the N-Cu bond in the CuPc molecular structure was decomposed irrevocably 40 (Fig. 4a, Supplementary Fig. 10,   Supplementary Table 3)." We hypothesise that as the CNP ratio is increased, the catalyst ink becomes more conductive and more CuPc sites are believed to be electronically connected, thereby facilitating their conversion during CO 2 RR conditions. The EXAFS peak remaining at 1.4 Å is found in metallic copper as well as CuPc and cannot be differentiated easily. We have added an EXAFS spectrum of metallic Cu to the Supplementary Fig. 6. to clarify: 12. The authors state that "the higher magnification TEM images taken pre-and postelectrolysis also support the claim that no sizable metallic Cu nanoparticles were formed during the reaction while also suggesting that the metallic Cu was in an amorphous state during electrolysis." What do the author consider sizeable metallic Cu nanoparticles? Based on the comparison of the TEM images, it seems that some particles in the range of 2-5 nm could be suspected. I recommend adding STEM and EDX mapping to the paper to help with identifying the presence or absence of small Cu nanoparticles. Being able to map the distribution of Cu before and after electrolysis would be highly desirable.

Response from authors:
We have now performed both SEM with EDX mapping, and STEM with EDX mapping on a sample after electrolysis: Cu elemental mapping. 25 We mention these results in the manuscript: "Energy dispersive X-ray (EDX) mapping and spectroscopy results proved the Cu element is evenly distributed on the GDE (Supplementary Fig. 13 & Fig. 15 & Supplementary Table 4).
STEM/EDX and TEM images taken pre-and post-electrolysis also support the claim that Cu nanoclusters (~2-5 nm) were formed during the reaction (Supplementary Fig. 15 & Fig. 16).
" Some additional minor issues: 13. In the SI Figure 2 caption, "(b) The Faradaic Efficiency (FE) toward hydrogen with different CNP to CuPc ratios." is written twice.

Response from authors:
Thank you for pointing this out, we have now removed the duplicate reference:  Figure 11 is not referenced in the article despite its interesting content and should be mentioned in the main text.

Response from authors:
We now call out this figure in the main text: "The highest FE toward CH 4 , 62%, was exhibited by the 4:1 sample at -4.00 V and 220 mA cm -2 (liquid product analysis shown in Supplementary Fig. 2)." Figure 3 performed in a saturated electrolyte (i.e.

Are the XANES and EXAFS shown in
under constant CO2 bubbling)? This should be indicated in the figure caption.

Response from authors:
As noted in response #5, we provide full details on the XANES and EXAFS measurements, and describe in the SI that we performed these measurements in a flow cell with a gas diffusion electrode, and as such CO 2 bubbling into electrolyte was not required.

Response from authors:
We have added these details to the experimental section and directed the reader to this section in the main text: The study focuses on the ultra-low coordination number copper catalysts for electrochemical CO2 reduction, which has good methane selectivity and stability. However, the key argument of this paper, XAS analysis of ultra-low coordinated copper catalysts, looks invalid and incomplete, which needs more explanation and transparency. The ultra-low coordination number obtained from this paper actually indicate the Cu-metal with nano-size, which can be quantitatively In addition, I have the following comments for authors to address for the improvement and possible submission to other journals.

Response from authors:
We appreciate the reviewer's perspective and follow-up on the characterisation concern and the novelty point, with new measurements and clarifications in the text as described in-line below.
1. The author argued that the mixture of CuPc with CNP can help form the metallic Cu with nanosize to catalyze the reaction. It is possible the pure CuPc catalysts authors prepared have low surface-to-bulk ratio. The surface CuPc convert to Cu metals due to the surface reactions. XAS is a bulk technique that cannot detect the surface changes, thus no formation of Cu is observed. CuPc to CNP with 1:7 ratio was used for electrochemical performance but unfortunately not used for XAS analysis. This part is critical to create the turning point if authors conclusion can be held. I will suggest to add another data point, say CuPc to CNP with 1:10 ratio for both electrochemical and XAS analysis, in addition to 1:7 ratio one for XAS analysis. This is to guarantee the effective of the volcano shape for both electrochemical performance and XAS analysis.

Response from authors:
As the reviewer requested, we attempted to characterise the 7:1 sample but the very low copper concentration in this sample made measurements incredibly challenging. We used 29 the silicon drift detector, the most sensitive detector available to us, but we could not get reliable signals. This is to confirm the metallic Cu as the true catalytic site for this reaction.

Response from authors:
We thank the reviewer for highlighting this prior work which studied, but did not try to control, Cu agglomeration from CuPc. This paper does not present any experiments or characterisations taken at different times, so the stability of their synthesis approach is not known. Our work builds on this by developing a means to stabilise these low coordination numbers for methane production.
As the reviewer requested, we have now performed a set of XANES and EXAFS measurements with the 4:1 sample at different constant voltages as requested ( Supplementary Fig. 8). Unlike the referenced paper, we observe only small changes in CuPc transformation for the more negative applied voltages, at all voltages most of the CuPc is transformed into Cu, emphasising that the CNP moderator strategy can effectively limit Cu agglomeration. We now mention these results in the manuscript: "Potentiostatic XAS measurements suggested that CuPc agglomeration was not influenced significantly by the applied potential when the CNP moderator strategy was employed (Supplementary Fig. 8.).

Response from authors:
We now provide a comprehensive XAS analysis method section in SI, including the cell design and the XAS analysis approach. "

In-situ X-ray absorption spectroscopy (XAS) Characterization
In-situ XAS was carried out under the same conditions as electrochemical testing using a REX2000 software using ab initio-calculated phases and amplitudes from the program FEFF 8.2 was used for the EXAFS fitting. The theoretical properties of the Cu-Cu and Cu-N paths, from the crystal information file of metallic Cu and CuPc, were incorporated to fit the experimental results. The ab initio phases and amplitudes were used in the EXAFS equation: The neighbouring atoms with different distances were divided into j shells. N j represented the coordination number of shell j at a distance of R j relative to the central atom. f effj (π,k,R j ) was the ab initio amplitude function for shell j, while the Debye−Waller factor accounted for the damping that resulted from static and thermal disorder in absorber−backscatterer distances. The mean free path term ( ) reflected losses due to inelastic scattering, where λj(k) was the electron mean free path. The sinusoidal term sin(2kR j + ij (k)), where ij (k) was the ab initio phase function for shell j, reflected the oscillations in the EXAFS. Shake-up/shake-off processes at the central atom(s) affected the amplitude reduction factor, . CN, R, ΔE, and the EXAFS Debye−Waller factor (DW; σ 2 ) were variable parameters of the EXAFS equation for fitting the experimental result. The R range for fitting ranged from 1.20 to 2.80 and the k fell between 3.70 and 9.55. Multi-path fitting was considered for Cu-Cu paths, Cu-N paths (within CuPc molecules), and

Supplementary
Cu-C paths (interface between the metallic Cu cluster and CNP). The contribution from Cu-C paths was negligible and the fitting using two paths (Cu-Cu and Cu-N) matched well with the experimental spectra (Supplementary Fig. 5). This result confirmed that there were no detectable Cu-C paths formed once CNP was added. The fitting results were listed (shown in Supplementary Table 2) using only two-path fitting (Cu-Cu and Cu-N) to clarify these shells in the EXAFS spectra. These fitting results were also consistent with the XANES fitting using linear combination results.
4. Please provide the detailed method to determine all sample compositions by XAS (using XANES or EXAFS? Linear combination fit or linear function analysis?) and cited related reference.

Response from authors:
We have added this description to the caption of Supplemental Table 1  The author should provide the enlarged view of XANES, full XAS spectra, and EXAFS k-space spectra to prove the stable after 50 minutes.

Response from authors:
The original manuscript was not sufficiently clear here. We have now clarified that the sample was stable by 50 minutes of operation (and not after): "The EXAFS spectra of the 4:1 ratio sample stabilised within 50 minutes of operation with little change between the spectra at 50 and 80 minutes, demonstrating that the structure was stable once agglomerated (Supplementary Fig. 3c)" We have corrected the EXAFS measurements of the 4:1 sample (which were originally mislabelled in error). The XANES and EXAFS measurements of the 4:1 sample now show similar spectra for all times during CO 2 RR, suggesting a stable structure.
" 6. The author should provide the method of how to get the Cu-Cu coordination number such as 4.2. The percent of composition cannot determine the Cu-Cu coordination number.
EXAFS fitting needs to be done since the author did not mention any analysis method of EXAFS. Suppose the author did run the EXAFS fitting, the fitting table with all necessary fitting parameters (such as amplitude reduction factor, scattering path length, K-range, R-factor, Debye-Waller factor, and the corresponding error bars). The fitting of EXAFS k-space and comparison of EXAFS k-space also need to be provided to see the actual differences.

Response from authors:
As described above in response #3, we now provide the full details on our XANES and EXAFS measurements and associated fittings in the manuscript and SI.