Enhanced multi-carbon alcohol electroproduction from CO via modulated hydrogen adsorption

Multi-carbon alcohols such as ethanol are valued as fuels in view of their high energy density and ready transport. Unfortunately, the selectivity toward alcohols in CO2/CO electroreduction is diminished by ethylene production, especially when operating at high current densities (>100 mA cm−2). Here we report a metal doping approach to tune the adsorption of hydrogen at the copper surface and thereby promote alcohol production. Using density functional theory calculations, we screen a suite of transition metal dopants and find that incorporating Pd in Cu moderates hydrogen adsorption and assists the hydrogenation of C2 intermediates, providing a means to favour alcohol production and suppress ethylene. We synthesize a Pd-doped Cu catalyst that achieves a Faradaic efficiency of 40% toward alcohols and a partial current density of 277 mA cm−2 from CO electroreduction. The activity exceeds that of prior reports by a factor of 2.


Reviewer 1
We thank Review #1 for this feedback.
We performed inductively coupled plasma optical emission spectrometry (ICP-OES) for all CuPd samples as suggested by the reviewer. We also renamed all samples according to the Cu:Pd ratios applied in the precursor solutions which correlate well with the Cu:Pd ratios determined by ICP-OES shown in Supplementary Table 5 (i.e. CuPd 0.010 , CuPd 0.015 and CuPd 0.022 have now renamed as CuPd 0.005 , CuPd 0.010 and CuPd 0.015 , respectively).
The Pd contents of CuPd 0.015 catalysts (now termed CuPd 0.010 since a Pd:Cu ratio of 0.010 in the precursor solution was used during synthesis) in both atomic and aggregate forms were determined by inductively coupled plasma optical emission spectrometry (ICP-OES). A similar Cu:Pd bulk ratio was found in both CuPd 0.010 samples.
The work by Li and Xu et. al. presents a system with a record selectivity for the production of oxygenate species through carbon monoxide reduction using a Pd-doped Cu surface. The experiments are well thought out and well realised and lead to a system whose activity warrants publication in Nature Communications. Nevertheless, the theoretical interpretation of the data has a few inconsistencies to be rectified before publication.
1. While it is clear that addition of Pd and Pt increase the selectivity for oxygenates, the emphasis on single-atom Pd sites being key to the high selectivity is not sufficiently proven for oxygenates over ethylene. In Figure 4e it appears that the aggregated Pd catalyst still shows high selectivity for oxygenate vs. ethylene formation, but the aggregates have reduced overall selectivity for CO reduction due to competing proton reduction, which is expected given the propensity of Pd for H2 evolution. To prove that aggregates are worse at oxygenate vs. ethylene selectivity, an experiment comparing CuPd0.015 with single atom and aggregated Pd is shown, however it is not clear if these two electrodes have the same Pd content. How was Pd content of the aggregated CuPd0.015 analysed? This must be added to the text. If it is the same galvanic exchange solution (as used for the single atom CuPd0.015) but with a different preparation procedure (as is implied by the heading of Figure S15), it should not be assumed that the electrode content is still CuPd0.015. This comment stems from the fact that to create an electrode containing the exact same amount of Pd but in a different morphology would be difficult without optimisation. That is not to say the data are not very interesting; doping with a small number of hydrogenation sites to increase oxygenate selectivity is an important finding. I recommend the authors concentrate on the use of single-atom Pd sites for creating a surface with high oxygenate/low H2 selectivity. The use of single atoms vs. aggregated catalysts for superior oxygenate formation over ethylene is only proven in the DFT data. We performed inductively coupled plasma optical emission spectrometry (ICP-OES) to determine the bulk ratio of Cu:Pd in all of our catalysts. The results are tabulated above and in the revised SI (Supplementary Table 5).

Supplementary
We further clarify this aspect in the text: Page 7, Line 128: "Using a Pd:Cu ratio of 0.01 in the precursor solution, a similar Pd:Cu ratio of 0.007 was achieved, as measured by inductively coupled plasma atomic emission spectroscopy (ICP-OES, Supplementary 3. It is unusual to plot 'peak selectivity' across a range of surfaces without indication of the potential at which the 'peak' was achieved, as in Figure 4e. It would also be useful to see error bars on these graphs. We now extend the COR performance of pure Cu to large negative potentials in our revised  We added the following in the revised manuscript (Page 8, Line 167): "Selectivities (FE, %) for individual products from COR using different Cu catalysts are provided in Supplementary Fig. 15 and Supplementary Table 7, and the corresponding product activities (current density, mA cm -2 ) are shown in Fig. 4 and We have added the ohmic drop correction details in the revised manuscript (Page 14, Line

308):
"Potentials reported in this work were calculated to the reversible hydrogen electrode (RHE) reference scale using E RHE = (E Ag/AgCl ) iR + 0.235 V + 0.059 × pH. The ohmic drop correction (i.e. iR compensation: i is the applied current and R is the cell resistance) was conducted using (E Ag/AgCl ) iR = E Ag/AgCl − 0.85 × i × R, in which E Ag/AgCl was the applied potential before iR compensation and a R value of 3.4 Ω was determined by performing an electrochemical impedance spectroscopy measurement using an Autolab PGSTAT302N electrochemical workstation coupled with a FRA32M module. A factor of 0.85 is applied in iR compensation during flow cell operation due to a low resistivity of 1 M KOH electrolyte which holds a relatively low voltage drop over the electrolyte. 18 " All corrections have been made to the revised files.
We thank Reviewer #1 for these helpful comments.
6. The authors should specify the nature of the GDL more specifically in the methods, as 'Sigracet' GDLs are available in a variety of compositions.

Indication of the ohmic drop correction (if any) should be described in the methodology.
8. Line 187 -oxygenate is misspelled 9. Line 262 -area is misspelled 10. Figure S23b -roughness is misspelled If these comments are addressed I can recommend this work be published as soon as possible.

Reviewer #2
We thank Reviewer #2 and detail below our responses that seek to address the reviewer's comments.
We clarify in the revision that the mechanism we proposed, hydrogenation vs. dehydroxylation, does not conflict with the CO dimer mechanism. In this work we focus on tuning the selectivity of alcohol vs. hydrocarbon, with the focus on steering the post-C-C coupling reaction steps. In particular, Ref 2 (Nat. Energy 4, 732-745 (2019)) indicates "… these C 2 species are formed through common intermediates… the hydrogenation of which leads to acetaldehyde and subsequently ethanol, and the hydrogenolysis of which leads to ethylene", which aligns with our proposed mechanism for favouring alcohols over ethylene, i.e. hydrogenation for alcohol vs. hydrogenolysis (dehydroxylation) for ethylene. We also note that the production of ethylene and ethanol from CO require the same hydrogen stoichiometry: We clarified these aspects in the revised manuscript (Page 4, Line 70): "CO dimerization has been suggested as the rate-determining step for CO-to-C 2+ conversion, generally 2, 21, 22 . Recent works by Goddard and co-workers 19, 20 have shown that the reaction of the intermediate HOCCH* through either a hydrogenation pathway (i.e. * + * → * ) or a dehydroxylation pathway (i.e. * + → * + ) ( Fig. 1a and Supplementary Fig. 1) determines the selectivity towards alcohol vs. ethylene, which is similar to the mechanism proposed by Koper and co-workers 2, 22 , showing that hydrogenation of C 2 intermediates leads to acetaldehyde and subsequently ethanol generation. We reasoned that controlling the catalytic hydrogenation of HOCCH* intermediates could steer high-rate COR selectivity from ethylene to alcohols." We performed atomic resolution HAADF-STEM and included the results in revised Fig. 2. Individual Pd atoms are now resolved on the Cu crystal surface via the difference in Z-contrast.
2. While single atom Pd catalyst is used in DFT, the authors do not successfully synthesize single atom Pd to modify Cu. At least, the evidence of single atom Pd is not very strong since there is a lack of atomic resolution STEM. This is very misleading and may confuse the readers. Nanocluster, instead of single atom, is a more appropriate term and should be used in DFT calculations. We now include these in the revised manuscript (Page 6, Line 121): "The Pd dopants were determined to be evenly distributed in the Cu structure using the aberration-correction high-angle annular dark-field scanning transmission electron microscopy (HAADF-STEM), coupled with electron energy loss spectroscopy (EELS) mapping ( Fig. 2b and 2c), as well as energy-dispersive X-ray (EDX) mapping (Supplementary Fig. 10). Individual Pd atoms were discerned on the crystal surface of Cu in the Pd-doped Cu catalysts based on the difference in Z-contrast."

XPS is used to determine the surface ratio of Cu to Pd and the authors admitted that this method may underestimate the amount of Pd due to the probing depth of XPS. Since this information is very important for the readers and those who want to reproduce the results, have the authors considered electrochemical methods such as Pb UPD, which is more surface-sensitive, to determine the amount of Pd?
We have performed Pb UPD of the CuPd 0.010 catalysts and found that no Pb deposition on Pd is observed, consistent with literature (J. Electrochem. Soc. 165, J3074-J3082 (2018)). To further support these measurements, we quantified the Cu:Pd bulk ratio of various Cu catalysts using inductively coupled plasma optical emission spectrometry. The resulting Cu:Pd ratios are tabulated below and included in the revised SI. We are now clear that we focus on alcohol, including only ethanol and 1-propanol, throughout the revised files and updated Fig. 4 shown below: Figure 4 since 'total FE of oxygenates' is a tricky term. It is well known that CO reduction process generates a lot of acetate, which cannot be directly used as fuel additive (only ethanol and n-propanol can be directly used). More importantly, the mechanism of acetate formation is likely to be significantly different from the formation of ethanol/n-propanol and simply adding them together is not acceptable (JACS 141,(4191)(4192)(4193)  We now include detailed product distributions, by catalyst, in Supplementary Table 7 We also updated the performance comparison to focus on the generation of alcohols, as tabulated in Supplementary Table 8.  We included the information in the revised manuscript (Page 8, Line 167):

A detailed breakdown of the oxygenates is necessary in
"Selectivities (FE, %) for individual products from COR using different Cu catalysts are provided in Supplementary Fig. 15 and Supplementary Table 7, and the corresponding product activities (current density, mA cm -2 ) are shown in Fig. 4 and Supplementary Fig. 16." We have now plotted the partial current densities for each and every product at each potential in Supplementary Fig. 16. b. They should plot the partial current densities for each and every product at each potential.
We further quantify the missing current as below: c. They should also plot the unbalanced charge for at each potential We have calibrated our Ag/AgCl/1 mg/kg KCl reference electrode as detailed below: Figure R3. Calibration of Ag/AgCl/1 mg/kg KCl reference electrode (E 0 = 0.235 V at 25 C).
Test was performed in ultrapure H 2 -saturated 1 M HClO 4 solution (pH = 0), the onset potential of hydrogen evolution reaction on polished Pt foil was determined to be 0.240 -0.235 = 0.005 V ≈ 0 V, indicating the accuracy of Ag/AgCl/1 mg/kg KCl reference electrode used in this work.

2)
The authors should calibrate the reference electrode experimentally using a real RHE.
We added the ohmic loss correction information into the revised manuscript (Page 14, Line

308):
"Potentials reported in this work were calculated to the reversible hydrogen electrode (RHE) reference scale using E RHE = (E Ag/AgCl ) iR + 0.235 V + 0.059 × pH. The ohmic drop correction (i.e. iR compensation: i is the applied current and R is the cell resistance) was conducted using (E Ag/AgCl ) iR = E Ag/AgCl − 0.85 × i × R, in which E Ag/AgCl was the applied potential before iR compensation and a R value of 3.4 Ω was determined by performing an electrochemical impedance spectroscopy measurement using an Autolab PGSTAT302N electrochemical workstation coupled with a FRA32M module. A factor of 0.85 is applied in iR compensation during flow cell operation due to a low resistivity of  (2019)).
We clarified these aspects in the revised Methods section (Page 12, Line 251): "The energetically favourable galvanic replacement of Cu with Pd/Pt proceeds by Cu + Pd 2+  Cu 2+ + Pd or 2Cu + [PtCl 6 ] 2- 2Cu 2+ + Pt + 6Cl -. In the case of adding a small amount of Pd/Pt precursor in the bulk Cu solutions, ultrasound waves consisting of compression and rarefaction cycles not only improve the reaction rate of galvanic replacement but also inhibit the aggregation of Pd/Pt atoms, in agreement with prior works 27,42 ."

3) They should also correct for Ohmic losses and state exactly how they measured Ohmic losses.
4) The authors should provide more convincing evidence to explain why their synthesis method should favour single atom catalysts.
All Cu K-edge and Pt L 3 -edge XAS spectra in this work were indeed performed in operando, at the 9BM beamline of the Advanced Photon Source (9BM@APS, https://www.aps.anl.gov/Spectroscopy/Beamlines/9-BM). We clarify this aspect in the manuscript and reference details of the operando XAS cell design and testing in our previous work (Ref 16: Nat. Commun. 9, 4614 (2018)).
Since the absorption energy of Pd K-edge (~24350 eV) exceeds the detection energy range of 9BM beamline (2.1 -24 keV), we conducted the Pd K-edge measurements at 20BM beamline of APS with a detection energy range of 2.7 -30 keV (https://www.aps.anl.gov/Spectroscopy/Beamlines/20-BM). However, the sample testing stage at 20BM was in an open system and could not accommodate CO gas for operando COR testing. In this case we performed instead ex-situ Pd K-edge testing. Samples after COR were washed with DI water, dried using N 2 gas and quickly sealed with Kapton tape, no oxidation of Pd was observed in our XAS spectra, suggesting the presence of metallic active Pd sites during COR.
We further assessed the Pd structure by carrying out a long-term COR test using the Pd-doped Cu catalysts (Supplementary Fig. 19). After a 7-hour COR operation, some Pd atoms had aggregated, reducing alcohol selectivity and increasing H 2 production. These results indicate importance of atomic-level Pd dopants in assisting alcohol production from COR, consistent with our DFT calculations.