Remarkable active-site dependent H2O promoting effect in CO oxidation

The interfacial sites of supported metal catalysts are often critical in determining their performance. Single-atom catalysts (SACs), with every atom contacted to the support, can maximize the number of interfacial sites. However, it is still an open question whether the single-atom sites possess similar catalytic properties to those of the interfacial sites of nanocatalysts. Herein, we report an active-site dependent catalytic performance on supported gold single atoms and nanoparticles (NPs), where CO oxidation on the single-atom sites is dramatically promoted by the presence of H2O whereas on NPs’ interfacial sites the promoting effect is much weaker. The remarkable H2O promoting effect makes the Au SAC two orders of magnitude more active than the commercial three-way catalyst. Theoretical studies reveal that the dramatic promoting effect of water on SACs originates from their unique local atomic structure and electronic properties that facilitate an efficient reaction channel of CO + OH.

shows the CO conversion as a function of the reaction time. In this, a rapid deactivation of the catalyst when the reaction is carried out at 200ºC in absence of water is observed. The authors think that carbonate species are not responsible for this deactivation because the reaction temperature is high. Are you sure that the catalysts surface is free of carbonates in these conditions?? Some types of cerium carbonates show a very high thermal stability. The absence of carbonates on the surface can be easily demonstrated by inspection of DRIFT spectra. Moreover, in figure S3-A, the reaction is carried out in dry conditions, but no deactivation is observed even at very high conversions at 120-160ºC. Why if the temperature is lower than 200ºC? Figure S3-B shows the CO conversion as a function of the time for two cycles of reaction in similar conditions (with water) and it is observed the improvement of the catalytic activity in the second one. In my opinion, a modification of the catalytic behaviour of a solid in consecutive reaction cycles, means that the surface (species, electronic density, …) has been modified during the first one. What happens in the case of Au1/CeO2? For the Au/CeO2-NP, a second cycle of reaction, in dry conditions (figure S3-C) also provokes an improvement of the catalytic activity. I think that a detailed discussion of these aspects is needed in the paper.
In addition, the paper is focused to demonstrate that water influences the catalytic activity of SAC in a different way than the interfacial sites in gold nanoparticles. However, the catalytic reactions have been carried out in a mixture CO+O2, what is the role of oxygen? The authors attribute the Au1/CeO2 deactivation in dry conditions to the consumption of OH species. This idea suggests that oxygen has not effect on the CO oxidation reaction, but it is not the case. This point has to be discussed in order to better understand the reaction. The oxygen activation in the CO oxidation reaction has been a very discussed point in the literature, and therefore, I suggest the discussion of this aspect.
An operando DRIFTS analysis has also been carried out in order to detect the modification during the reaction. Figure 4 shows difference spectra recorded during the reaction without and with water. In this, bands of CO in gas phase are overlapping other possible species such as the typical band due to an electronic transition in Ce3+, around 2130 cm-1. The authors try to demonstrate that Ce4+ is reduced during the "dry" reaction while after the water introduction, the Ce3+ is oxidized to Ce4+. In this analysis, a band at 2089 cm-1 (hardly observed due to the presence of CO in gas phase) has been attributed to CO adsorbed on Ce3+. I don't agree with this attribution, because a carbonyl bands appearing at wavenumbers lowers than 2143 cm-1, reflects the presence of sites with a very high backdonation ability (high electronic density), and this is not the case for Ce3+ species. The attribution of this band has to be considered. Moreover, the attribution of OH species should be revised.
1. The authors do not address at all the issue of sintering of Au species. I think that they should provide a more solid evidence that sintering can be excluded. 2. The nomenclature and charge of the Au1/CeO2 system is puzzling. As far as I understand the Au1 atom is substitutional for a surface Ce atom. The system is therefore more similar to a solid solution than to a supported catalyst. This could explain why sintering is not observed, although -again -the authors should clarify it (see point 1). What is puzzling is the charge of this Au1 atom, reported to be "positively charged, non-zero valentine's Au^{\delta +} atom. In the literature, this notation implicitly suggest a small positive charge. Instead, for a solid solution, the ionic nature of the Au1 species should be prominent, with oxidation states close to 3+ or 4+, as previously reported on the basis of photoemission spectroscopy. This issue should in my opinion made clear, and some experimental data assessing the charge should be presented. 3. The computational study model the Au1/CeO2 system on the basis of a substitutional Au atom on a flat (111) surface. The authors should provide more solid evidence that this highly ideal model does capture the physical features of the new catalyst. For example, the Au atom in this model will be highly ionic (see point 2).

Reviewer #1 (Remarks to the Author):
The subject treated in this paper is very interesting in the catalysis field: the role of water in oxidation reactions. The authors analyse the catalytic activity of SACs (Au1/CeO2) and Au/CeO2 (gold nanoparticles) in the CO oxidation in the presence of water. Results evidence that water has a promoting effect in SAC more important than in the case of interfacial sites of gold nanoparticles catalysts. Moreover, computational studies were carried out and permitted to attribute the special behaviour of SAC to the electronic structure and local geometry of gold atoms in this system. Finally, the authors propose that the reaction proceeds with the participation of OH groups generated in the presence of water.
Despite the great interest of this work, I have some comments and suggestions. Figure 1 shows the CO conversion as a function of the reaction time. In this, a rapid deactivation of the catalyst when the reaction is carried out at 200ºC in absence of water is observed.

Question (1):
The authors think that carbonate species are not responsible for this deactivation because the reaction temperature is high. Are you sure that the catalysts surface is free of carbonates in these conditions??
Some types of cerium carbonates show a very high thermal stability. The absence of carbonates on the surface can be easily demonstrated by inspection of DRIFT spectra. Moreover, in figure S3-A, the reaction is carried out in dry conditions, but no deactivation is observed even at very high conversions at 120-160ºC. Why if the temperature is lower than 200ºC?

Response:
We thank your nice comments on the value of our research work and we also appreciate your insightful questions that help us to clarify our discussions in the revised manuscript.
We agree with you that during the CO oxidation reaction the formation of carbonate species on CeO 2 surfaces can occur, especially at low reaction temperatures. In order to unambiguously identify the sources of the observed deactivation we have now checked the CO adsorption band region of 1200-1800 cm -1 (carbonates region) by conducting operando FT-IR with and without the presence of H 2 O in the reactant gas mixture. As shown below in Figure R1A, for CO oxidation without the presence of H 2 O, three bands appeared immediately after the reaction gas being introduced and the intensity of these bands gradually increased with the reaction time, unambiguously demonstrating the formation of carbonates on CeO 2 surfaces. However, the intensity of these bands kept almost unchanged when H 2 O was introduced, in contrast to the significant increase of intensity of the OH-group with time. These results proved that stable carbonates were indeed formed during CO oxidation but they did not change appreciably when H 2 O was introduced or eliminated. Therefore, we propose that the presence of the carbonates on the CeO 2 surfaces did not cause the observed deactivation.
In order to further confirm this hypothesis we have performed in situ heat-treatment of the deactivated catalyst by He and O 2 at 200 o C for 1 h. Such treatment has been proved effective (at least partially) to remove the carbonates and recover the catalyst activity for CO oxidation (Nat Chem 2011, 3 (8), 634, ACS Catal. 2014, 4 (7), 2113). Figure R2 shows clearly that both the He and O 2 treatments have no effect on the recovery of the CO conversion rate. The CO oxidation activity, however, increased dramatically after the introduction of a small amount of water. These new experimental results further corroborate our conclusion that stable carbonate species indeed formed during the CO oxidation reaction but the presence of these carbonate species do not seem to be responsible for the experimentally observed rapid deactivation and reactivation of our Au 1 /CeO 2 SAC catalyst.
As to the catalyst behavior in Figure S3a, we believe that the deactivation occurred during the CO oxidation process. The rate of deactivation is not obvious because of i) the catalyst may deactivate slower at temperatures lower than that at 200 o C (The consumption of the OH groups is higher at higher temperatures) and ii) the fact that the CO conversion rate increases with reaction temperature offsets the deactivation with reaction time. Therefore, it is difficult to evaluate the deactivation rate by simply examining the "conversion vs. reaction temperature" plot. To confirm this hypothesis, we have conducted a stability test at 160 o C. As shown in Figure R3, the catalyst certainly deactivated but indeed experienced a slower deactivation process compared with that conducted at 200 o C. In addition, the introduction of water immediately recovers the activity, similar to that at 200 o C. We have added the new data and discussions in pages 4-6 of the revised manuscript and pages 5-9 of the updated supporting information ( Figure S4, S6 and S7).  Figure S3

Response:
We appreciate the reviewer's comments and question. First of all, although this peak is difficult to detect, we are pretty sure that the observed peak DOES exist because in several repeated measurements the phenomenon of appearance and disappearance of this small peak with removal and introduction of water were all observed. Secondly, we can rule out the possibility of assigning the peak to CO adsorption on Au atoms because the Au loading level is too low to provide detectable CO adsorption peaks. We had tried CO adsorption on 0.05 wt% Au 1 /CeO 2 at room temperature but did not detect reasonable signal. Furthermore, CO adsorption on Au 0 usually yields a band at 2100-2110 cm -1 but on positively charged Au δ+ the band should be at even higher frequencies (2120-2150 cm -1 ), which is similar to or higher than that on Ce 3+ . Based on these reasons, we believe that ascribing this band to CO adsorption on Ce 3+ instead of Au δ+ is a more reasonable assignment. To further support this assignment we performed a DFT calculation of CO adsorption on our hydroxylated CeO 2 surface and found that the adsorption band exists at a frequency of 2072 cm -1 , very close to the experimentally observed band at 2089 cm -1 when considering that the used GGA functional usually underestimate frequencies. The DFT calculation suggests that such a frequency of CO adsorption on Ce is theoretically possible. All the above analyses indicate that our assignment is not unreasonable.
However, in the revised manuscript we have changed the text into "tentatively ascribed this band at 2089 cm -1 to CO adsorption on Ce 3+ on hydroxylated CeO 2 surfaces.".
As to the assignment of the OH group (3667 cm -1 ), we do not seem to see any issues. There is a systemic study on the IR spectrum of CeO 2 (J. Chem. Soc., Faraday Trans. 1989, 85 (6)

Reviewer #2 (Remarks to the Author):
The authors report on the results of an experimental and theoretical study on the promoting effect of water on single-atom Pt-CeO2 catalysts. Response: We thank the reviewer for the nice comments on our work and the critical questions that help us to improve our manuscript. First of all, the possibility (or risk) of testing surface models one by one until a reasonable correspondence could be ruled out because it is not the case and would be extremely time-consuming even if one likes to. The model selection is generally on the basis of the experimental results as well as the stability of the surface. For example, if we have a given structure or a specific exposed surface, it is easy to make the model. However, in our case, unfortunately the CeO 2 was synthesized by a co-precipitation method and existed in a small nanoparticle form thus no preferentially exposed surface can be identified.  Figure S3 in the main text and not only in the SI.

Response:
We appreciate the reviewer's question and comments. We apologize for not making this point clearer in the original manuscript. All specific reaction rates were measured in a differential mode operation at higher GHSV (up to ~3 125 000 ml g cat -1 h -1 ) and low CO conversions (≤30 %) to eliminate possible mass-and heat-transfer limitations. For each run at a specified reaction temperature, the CO conversions at 20 min, 40 min, and 60 min were averaged and used for calculations of the specific rate. The turnover frequency (TOF) was then calculated based on the specific rate and the dispersion, where the dispersion of SACs was assumed to be 100% and that of the Au/CeO 2 NP catalysts was calculated according to the equation D = 0.9/d where D represents the dispersion and d is the average diameter of Au NPs (in nm). We have added these details into the Methods section in the revised manuscript.
3. In Figure S3, the light-off curves are given for all three catalysts, but the effect of water is only shown for the Pt-CeO2-SAC. For completeness the effect of water should also be shown for the other two catalysts.

Response:
We thank the reviewer's excellent suggestion. The effect of water on the other two Au NP catalysts is now presented as "Run 3-W" on Figure S3C and Figure S3D, respectively. To make it clearer to the readers, we have added a note in the legend of Figure S3 in the revised manuscript.

L42/43: "…catalysts because the interfacial perimeter sites usually serve as the dominant active
sites." This is certainly not true in general. In fact, there are many cases, in which the boundary sites is not the most active.

Response:
We thank the reviewer for identifying this problem in our original manuscript. We have replaced the word "usually" with "in many cases", thus the sentence has been changed into "…catalysts because the interfacial perimeter sites in many cases serve as the dominant active sites." 5. L. 31/32: "…due to the quantum size and ensemble effects of metal nanoparticles…". Talking about heterogeneous catalysis in general, there are other effects that may play a role, for example the electronic and support effects discussed in this work.

Response:
We appreciate the reviewer's comments. In order to better express our intention we have added the "and the differences between SACs and nanocatalysts in electronic and support effects as well as metal-support interaction" into the revision (L.32/33 and L.48/49).
6. Figure 3: the arrow for CO2 has the wrong orientation (against the catalytic cycle).

Response:
We thank the reviewer for kindly pointing this error. We have corrected the arrow in Figure   3 of the revised manuscript. 7. Line 267: "which makes the reaction exothermic", change to "which makes the reaction step exothermic" (the overall thermodynamics is not changed).
Response: Thank you so much for kindly pointing this out. We have changed the sentence in the revised manuscript as "The presence of the highly positively charged Au 1 single atoms is crucial to the CO + OH reaction channel due to their flexibility in variation of oxidation state; the Au 1 atom can be reduced from Au(III) to Au(II) or Au(I), with Bader charge decreased from +1.10 to +0.69 eV during the CO oxidation reaction ( Figure S17 and Table S2), which makes the reaction step exothermic.".

Reviewer #3 (Remarks to the Author):
This work analyses the catalytic performances of a promising catalyst Au1/CeO2, that was already proposed by a subset of authors (see Ref. 19, ACS Catalysis 2015), as evidenced by ac-STEM-HAADF images ( Figure S1 and images in our previous work (ACS Catal. 2015, 5, 6249-6254)). In this case, the Au atoms should exist in a high oxidation state (e.g. Au(III) or Au +3 , which is often written as Au 3+ charge states). Although oxidation state of Au can be measured by XPS or XANES, experimentally proving the Au oxidation state in our case is very challenge due to the extremely low loading of the Au atoms. In our previous report, we have examined the oxidation state of Au in Au 1 /CeO 2 with an Au loading of 0.05 wt% (ACS Catal. 2015, 5, 6249-6254). As shown in Figure R7, although the Au4f signal is very weak we can still deduce that the Au atom likely exists mainly as Au + and Au 3+ oxidation state, as is expected. Given the non-zero oxidation state (Au in Response: This is a good suggestion. Indeed, as shown experimentally Au 1 substitutes the Ce atom in ceria, thus existing in a positive oxidation state. To estimate the possible oxidation states of Au atoms in the catalysts, we have now calculated the Bader charges of Au atoms in the Au 1 /CeO 2 as well as in the Au(OH) x (x = 1-3) molecules as a comparison. As shown in Table R1, Au atom in the Au 1 /CeO 2 -OH is indeed positively charged (atomic charge +1.10 |e -|) and the calculated atomic charge is close to that in Au(OH) 3 (+0.988 |e -|), indicating that the oxidation state of Au 1 can be assigned as Au(III) and the charge state is thus Au 3+ , which is consistent with previous studies [Nano Res. 2015, 8(9): 2913-2924]. As presented in the projected DOS, Au 5d-orbitals indeed have two peaks above the Fermi level, suggesting the formation of an Au 3+ cation. The additional results also confirm the high oxidation states and positively charged Au atoms in Au 1 /CeO 2 SAC. We have added the new data into Table S2 and Figure S17 in the revised support information.