A metal-supported single-atom catalytic site enables carbon dioxide hydrogenation

Nitrogen-doped graphene-supported single atoms convert CO2 to CO, but fail to provide further hydrogenation to methane – a finding attributable to the weak adsorption of CO intermediates. To regulate the adsorption energy, here we investigate the metal-supported single atoms to enable CO2 hydrogenation. We find a copper-supported iron-single-atom catalyst producing a high-rate methane. Density functional theory calculations and in-situ Raman spectroscopy show that the iron atoms attract surrounding intermediates and carry out hydrogenation to generate methane. The catalyst is realized by assembling iron phthalocyanine on the copper surface, followed by in-situ formation of single iron atoms during electrocatalysis, identified using operando X-ray absorption spectroscopy. The copper-supported iron-single-atom catalyst exhibits a CO2-to-methane Faradaic efficiency of 64% and a partial current density of 128 mA cm−2, while the nitrogen-doped graphene-supported one produces only CO. The activity is 32 times higher than a pristine copper under the same conditions of electrolyte and bias.


Reviewer #1 (Remarks to the Author):
This is a clever writing way to report the results. The copper-supported iron-single-atom catalyst, Cu-FeSA is essentially the same with the explosively reported copper based electroctalysts. In the materials aspects, the Cu-FeSA was achieved by trapping iron phthalocyanine on copper surface and subsequent in situ reduction by applying electricpotential. This, in other way, revealed the unstable characteristic of the metal phthalocyanine. I agree this would be a way to prepare single atom catalyst, however few advantages of this method can be expected. In the aspect of electroreduction performance, the CO2-to-methane Faradaic efficiency of 64% at a partial current density of 128 mA cm-2 in GDE electrode is reached. This is also not impressive. It is straightforward to understand the different performance as compared to nitrogen-doped graphene-supported Fe as well as the 32 times enhanced activity than pristine copper. I find no major flaws in experiments and data analysis, however, there is lack of novelty for this journal. Thus, I cannot recommend its publication.

Response:
We have revised the manuscript to be clear that we achieved one of the best CO2-tomethane Faradaic efficiency, equal to FE=64%. Given the referee's comment on nitrogen-doped graphene-supported Fe, we study this reference catalyst too, and obtain CO alone. The coppersupported iron-single-atom catalyst, Cu-FeSA, majorly produces CH4 rather than CO. It shows that the metal-supported iron-single-atom catalyst is a new catalyst system, broadening the understanding of single-site catalysts in CO2 reduction.

Reviewer #2 (Remarks to the Author):
The electrochemical hydrogenation of CO2 to methane (CH4) represents a promising mean to store renewable electricity in the form of chemical fuels. In this work, the author reported that the Cu support Fe single atom is an excellent electrocatalyst toward CO2 reduction to methane with a high Faradaic efficiency of 64% and a partial current density of 128 mA cm-2. Through a series of experimental characterization combined with theoretical calculations, they explored the origin of the high activity of the catalyst. The results are impressive and this work is well-organized. However, several issues should be addressed before I can recommend publication.
1. It is known that Cu and Fe are vulnerable to the oxidation and leaching in harsh chemical environments. Several works reveal that in aqueous electrolyte Cu may spontaneous oxide and alter the product profile of CO2RR (J. Am. Chem. Soc. 2020, 142, 28, 12119-12132;ACS Cent. Sci. 2019, 5 (12), 1998-2009. Moreover, the nanoparticles with large surface area often suffer from the oxidation or aggregation problems. Therefore, it is necessary to emphasize the dynamic phase transformation and/or surface reconstruction during the cathodic CO2RR process.

Response:
We now report in-situ time-resolved X-ray absorption spectroscopy (XAS) with the goal of investigating the dynamic reoxidation/reduction behavior and the phase transformation. This is now provided in Revised Supplementary Fig. S36. Time-resolved Cu K-edge XAS allows us to collect the spectrum every 2 min rather than every 30 min. We report that, at 200 mA cm -2 , sputtered Cu in Cu-FeSA exhibits a metallic state within 2 min, and remains in this state over continuous operation of 30 min. We better explain that we do not observe the dynamic reoxidation/reduction behavior and the phase transformation seen in prior reports based instead on H-cell systems. Using in-situ Raman spectroscopy (Fig. 3f), we report that the peak for surface copper oxide decreases progressively with increasing applied potential, and we do not see evidence of reoxidation. We now discuss our findings and explain that, in a flow-cell system, the catalyst undergoes a strong driving force of reduction at 200 mA cm -2 , leading to the lack of detected reoxidation/reduction.

These considerations inform this new main manuscript text:
"According to literature, 58,59 Cu can undergo a dynamic phase transformation and/or surface reconstruction during CO2RR. We performed in-situ time-resolved X-ray absorption spectroscopy (XAS) to investigate this possibility, shown in Supplementary Fig. 40. We found that the sputtered Cu in Cu-FeSA exhibits a metallic state in 2 min, and remains in this state following 30 minutes of continuous operation. We did not observe dynamic reoxidation/reduction behavior and phase transformation reported in prior H-cell system studies. In-situ Raman spectroscopy (Fig. 3f) indicated that the peak for surface copper oxide decreases progressively with increasing applied potential, without reoxidation. In flow-cell system, the catalyst undergoes an intensive driving force of reduction at 200 mA cm -2 , accounting for the lack of detectable reoxidation/reduction." "In-situ time-resolved XAS was conducted at the 44A beamline of the National Synchrotron Radiation Research Center (NSRRC, Hsinchu, Taiwan)." 2. The selectivity of CO2 reduction is of great importance. Previous studies have shown that local * CO coverage affects the distribution of products in CO2RR (ACS Catal. 2017, 7, 1749Nat. Catal. 2019, 2, 1124. Therefore, the relationship between local CO surface coverage and methane selectivity over Cu-supported Fe monatomic catalysts should be analyzed.

Response:
We have now provided DFT calculations that seek to investigate the coverage effect in the calculation of the CO hydrogenation step (to methane) and coupling step (to C2 products), shown below and in Supplementary Fig. 5. We find that both hydrogenation and coupling reactions are promoted with increased CO coverage, consistent with the previous studies, (ACS Catal. 2017, 7, 1749and Nat. Catal. 2019, 2, 1124 but that the latter exhibits a more acute decline in reaction energies as a function of increasing coverage, agreeing with preferred coupling in the case of high CO coverage. We further compared energy difference between coupling step and hydrogenation step to identify the preference of CO coupling ( Supplementary Fig. 5b). The reaction energy of CO coupling gradually decreases and becomes close to equaling the hydrogenation energy, indicating an improved priority of C2 products at high coverage.
The full revised context is shown below and in Main text: With CO coverage increases from 2/9 (2 *CO on surface with 9 metal atoms) to 5/9, the reaction energy of CO coupling gradually decreases and close to hydrogenation energy, which suggests an improved priority of C2 products at high coverage.
3. Previous studies have pointed out that the CH4 generation of copper-based catalysts depends on the pH value. If the pH value of the electrolyte is high, the early CC substitution can dynamically inhibit the C1 pathway (Phys. Chem. Chem. Phys. Phys, 2015, 17, 17) 18924). J. Am. Chem. Soc. 2016, 138, 483). Therefore, the variation of the local pH near the electrode surface may affect the product distribution. The authors need to comment on this issue.

Response:
We now better discuss the relationship to previous work (Nature 2020, 557, 509, figure shown below): we discuss that, in that work, the local pH in KHCO3 at 200 mA cm -2 was 12.1. Thus, as seen in J. Am. Chem. Soc. 2016, 138, 483, at high pH, C-C coupling through adsorbed CO dimerization dominates, suppressing C1 pathways, thereby increasing selectivity to multi-carbon products. This matches with the catalytic result seen using pristine Cu, whose major product is ethylene (FE > 50%). Under the same catalytic conditions, Cu-FeSA catalyzes CO2 into methane (C1 product). The Raman spectrum in Fig. 3f showed that Fe sites attract and convert CO2 to *CO on Fe sites, and suppress adsorbed *CO on Cu sites. The more active single-atom Fe is present at the surface of the Cu and increases methane production, even at high pH.
The full revised context is shown below and in Main text: "In comparison, the major product on bare Cu is ethylene (maximum 54% FE at a current density of 200 mA cm -2 ) and the methane FE is only 2% FE. The local pH in KHCO3 at 200 mA cm -2 is 12. 51 At high pH, C-C coupling dominates, and the C1 pathways are suppressed, facilitating the selectivity of C2 products, in line with the catalytic result of bare Cu. 52 " however, more elaborated scientific presentation should be provided before being accepted on Nat

Supplementary
Communication. This manuscript is recommended for a major revision subject to further review, and the following comments need to be addressed in the revision.
Comments to general presentation: 1. In page 4, The authors claimed "we calculated the post-adsorption steps on Cu-FeSA: CO hydrogenation and C-C coupling, which determine branching of C1 (e.g. CH 4 , CH 3 OH) vs. C 2 (e.g. C 2 H 4 , C 2 H 5 OH) products. 21 2. In page 4, the authors claimed "it is unlikely for *CO adsorbed on atop Fe sites to bond with neighboring *CO on Cu sites to generate C2 products, consistent with 17 prior theoretical studies.
". Since ref 17 was done on Au or Ag support, the so called "consistency" need to be re-elaborated.

Response:
We have rewritten: "We found that C-C coupling is energetically unfavorable on Cu-FeSA (both on Cu-Cu and Cu-Fe sites) compared to on Cu (Fig. 2e) Taking the CO hydrogenation reaction *CO + H2O + e -→ *COH + OHas example, the reaction energy is equals to where, G(*COH) and G(*CO) is the total free energy of *COH and *CO adsorption configuration; G(OH -) is the free energy of hydroxyl ion, to calculate this value, we assume the equilibrium Which relates the chemical potentials as Then, Here, + + − can be calculated using computational hydrogen electrode (CHE) model developed by Nørskov and co-workers.  (111) is not provided in the manuscript.

Response:
We now discuss more clearly the calculated surface energy results seen in Supplementary   Table S1. The (111) facets have the lowest surface energy among all low-index facets of Cu, with an fcc crystal structure, indicating that (111) is the most stable facet. For experimental evidence, the XRD pattern in Fig. 3a shows that the preferred orientation of the sputtered Cu is (111) facet before and after the molecular assembly. We now write: "We began by investigating, using DFT calculations, whether CO2 methanation is feasible on atomically-dispersed elements on Cu (111), the preferred orientation in polycrystalline copper exhibiting the lowest surface energy in all low-index facets of Cu with an fcc crystal structure

in-situ EXAFS
whether there is a contribution of solvation ( Fig. 2f and Supplementary Figs. 6 and 7), while *CHO is preferred on the pristine Cu ( Supplementary Fig. 8)."

Supplementary Fig. 8 | Hydrogenation energy of the intermediates for methane production on
the pristine Cu. Even though pure Cu has a lower (~0.3 eV) hydrogenation energy through *CHO reaction pathway, an energy difference ~1 eV of CO coupling between pure Cu and CuFe suggests a higher preference for C2 products on pure Cu compared to CuFe SAA. 5. Mechanism of C2 channel: a common critical C-C bond coupling step, *CO + *CO → *OCCO, was not taken into account in this study? Is any evidence to exclude this process?

Response:
We now better explain the atomic force microscope images of sputtered copper on flat Si wafer shown in Supplementary Fig. 12. The images display the surface of sputtered copper, whose roughness is approximately 9 nm. The rough surface provides abundant adsorption sites for iron phthalocyanine. The catalyst surface becomes smoother on the iron-phthalocyanine-modified copper, as observed using atomic force microscope ( Supplementary Fig. 13), consistent with SEM in Supplementary Fig. 14. We thus correlate the presence of iron phthalocyanine with the observation of a slightly less rough copper surface: "The rough surface of sputtered copper provides ample adsorption sites for iron phthalocyanine, and exhibits an intensified monolayer XRD signal. We also conduct atomic force microscope studies; the roughness of approximately 9 nm on the sputtered copper ( Supplementary Fig. 12), while a 9 Angstrom step height, corresponding to the predicted distance Fe⋯3-MPA⋯Cu, is observed on the iron-phthalocyanine-modified copper ( Supplementary Fig. 13). The Cu surface becomes smoother following molecular assembly ( Supplementary Fig. 14) but keeps (111)  The revised manuscript pretty much addresses all of my concerns. This manuscript could be publishable if the following comment can be further clarified.
The metallic Fe resulted from the negative-potential induced demetallation process out of phthalocyanine (Pc) moiety, and subsequently diffuses to Cu(111) surface to form Cu-FeSA.
Since the 3-MPA bonded FePc layer exists before the demetallation step, Fe is presumptively forced to populate on the binding sites not occupied by the thiol groups of 3-MPA. Therefore, the formation of Cu-FeSA catalytic sites may not be comparable with the scenario of simply positioning a Fe atom on Cu (111)  (1) Strong 650 cm-1 signal is shown in Figure 3f but weak corresponding feature is shown in Figure S32. How is the intensity scale defined in Figure 3f and Figure S32 becomes very important for these interpretations. It should be clearly noted in the captions.
(2) If the 740 cm-1 signal correctly represents the presence of 3-MPA, the color evolution in respect to the voltage seems to be inconsistent. Noticeable red spots only appear at -0.7 ~ -1.0 V (vs RHE). Shouldn't it be gradually decayed as the voltage goes to more negative, meaning more red spots should be seen at V > -0.7?
If 3-MPA-Pc only plays the role as the material template for synthesizing Cu-FeSA (and no role to the catalytic mechanism as noted in the manuscript), it would also show the same superior catalytic activity with removing the 3-MPA layer before starting the catalytic measurement. With that, I would like to request the evidence of the control experiment.

Manuscript ID: NCOMMS-21-10097-A "A Metal-Supported Single-Atom Catalytic Site Enables Carbon Dioxide
Hydrogenation" Reviewer #3 (Remarks to the Author): The revised manuscript pretty much addresses all of my concerns. This manuscript could be publishable if the following comment can be further clarified.
The metallic Fe resulted from the negative-potential induced demetallation process out of phthalocyanine (Pc) moiety, and subsequently diffuses to Cu (111)  (1) Strong 650 cm-1 signal is shown in Figure 3f but weak corresponding feature is shown in Figure S32. How is the intensity scale defined in Figure 3f and Figure S32 becomes very important for these interpretations. It should be clearly noted in the captions.

Response:
We are now more clear that the applied potential in the previous Figure   RS32 was -1.2 V vs RHE, at which the signal is weak at 650 cm -1 . We now are explicit in providing the applied potential in the revised Figure RS32 and we add a new in-situ Raman result for the case of Cu-FeSA at -0.6 V vs. RHE to make it clearer. We are now explicit that the intensity scale of revised Figure S32 is the same as that of Figure R3f. We also state explicitly the intensity scale in the caption of Figure R3f.
The Raman spectra of the Cu and Cu-3-MPA were obtained in a dry condition. The Raman signal was not reduced by the electrolyte, so that the intensity was similar to the in-situ spectrum of the Fe-FeSA.
We now separate the Raman results for the case of in-situ and dry conditions, and these are provided in revised Figure RS32. (2) If the 740 cm-1 signal correctly represents the presence of 3-MPA, the color evolution in respect to the voltage seems to be inconsistent. Noticeable red spots only appear at -0.7 ~ -1.0 V (vs RHE). Shouldn't it be gradually decayed as the voltage goes to more negative, meaning more red spots should be seen at V > -0.7?

Response:
We assign the 740 cm -1 signal to the presence of 3-MPA and FePc (JACS, 2019, 141, 5684) at the Cu surface. We consider the 740 cm -1 signal at -0.6 V vs RHE to arise from 3-MPA, since it is close to the Cu surface. At -0.7 ~ -1.0 V vs RHE, the noticeable red spots we assign to FePc, which moves toward the copper surface and leads to an increase in the Raman intensity as the 3-MPA detaches from the Cu surface. We also observe that the intensity at 660 cm -1 and 935 cm -1 increases from -0.6 V to -0.7 V vs. RHE and then decreases at a more negative voltage, suggesting that the phthalocyanine ring of FePc is detached from the Cu surface.
The full revised context is shown below and in the Main text: "We associate the 740 cm -1 signal at -0.6 V vs. RHE with 3-MPA ( Supplementary Fig.   R32) since it is close to the Cu surface. At -0.7 ~ -1.0 V vs. RHE, the noticeable red spots can be attributed to FePc, which moves toward the copper surface and increases the Raman intensity as the 3-MPA detaches from the Cu surface. The intensity at 660 cm -1 and 935 cm -1 increases from -0.6 V to -0.7 V vs. RHE and then decreases at a more negative voltage, suggesting that the phthalocyanine ring of FePc is detached from the Cu surface." (3) If 3-MPA-Pc only plays the role as the material template for synthesizing Cu-FeSA (and no role to the catalytic mechanism as noted in the manuscript), it would also show the same superior catalytic activity with removing the 3-MPA layer before starting the catalytic measurement. With that, I would like to request the evidence of the control experiment.

Response:
We disassembled the flow cell and washed the gas-diffusion electrode as the iron-single-atom formed, attempting to remove the 3-MPA and H 2 Pc layer.
Unfortunately, we observed mainly hydrogen evolution. We propose that iron is oxidized when we disrupt the reductive current, and that the copper-supported iron-decorated surface is thus disrupted.
We sought another route to see whether the remaining 3-MPA and H 2 Pc on the copper surface influence the catalytic activity. We obtained an in-situ time-evolution Raman spectrum of Cu-FeSA. Initially, we observed a signal associated with phthalocyanine ligand and 3-MPA remaining on the catalytic surface. The phthalocyanine ligand and 3-MPA then progressively depart the catalytic surface. We correlate this to the activation process in the stability measurement of Figure 4c. We conclude that the remaining 3-MPA and H 2 Pc affect activity, but can be removed through the electrolyte flowing during CO 2 RR.
The full revised text is shown below and is now in the Main text: "We observe a signal associated with the phthalocyanine ligand that remains following CO 2 RR, consistent with the view that iron atoms are attracted to the copper surface, and that some phthalocyanine rings remain near the copper surface but depart from the catalytic surface after approximately one hour. (Supplementary Fig. 33)" "We offer that the activation process correlates with the fact that the phthalocyanine rings remain and then leave the copper surface, shown in Supplementary Fig. 33."