Molecular-level insights into the electronic effects in platinum-catalyzed carbon monoxide oxidation

A molecular-level understanding of how the electronic structure of metal center tunes the catalytic behaviors remains a grand challenge in heterogeneous catalysis. Herein, we report an unconventional kinetics strategy for bridging the microscopic metal electronic structure and the macroscopic steady-state rate for CO oxidation over Pt catalysts. X-ray absorption and photoelectron spectroscopy as well as electron paramagnetic resonance investigations unambiguously reveal the tunable Pt electronic structures with well-designed carbon support surface chemistry. Diminishing the electron density of Pt consolidates the CO-assisted O2 dissociation pathway via the O*-O-C*-O intermediate directly observed by isotopic labeling studies and rationalized by density-functional theory calculations. A combined steady-state isotopic transient kinetic and in situ electronic analyses identifies Pt charge as the kinetics indicators by being closely related to the frequency factor, site coverage, and activation energy. Further incorporation of catalyst structural parameters yields a novel model for quantifying the electronic effects and predicting the catalytic performance. These could serve as a benchmark of catalyst design by a comprehensive kinetics study at the molecular level.


The caption of
is unclear (XPS). The caption said "ambient pressure" while the pressure in the experiment chamber shows 20 torr. The ambient pressure stands for atmospheric pressure conditions and temperature (~1 bar, RT). In general, the information on the in situ XPS experiments should be more precise! They should state clearly the conditions of XPS experiments (P and T).
2. The binding energy shift of Pt 4f7/2 was attributed to "shift is mainly ascribed to the variation in Pt electron density due to the electron transfer between Pt and CNT". What is the driving force for this effect? It would have been interesting to investigate this phenomenon through the examination of the molecular orbitales involved. 3. Although I understand the challenge of recording XPS spectra in in situ conditions, using a laboratory setup (not a synchrotron-based one where high photon brilliance exists), a main critic concerns the in situ XPS spectra especially Pt 4f (more specifically Fig. S9). They were acquired with low counts. During the experiment, they should have been accumulated with more spectra in order to improve the resolution. Figure S9 is not good! The convolution of the pics is quite arbitrary! I am surprised to see such poorly conducted XPS experiment. 4. The figure 4a is irrelevant and does not bring any new information. The technical details shown are obvious and rather standard. It can be removed or displaced to the Supplementary document. 5. Regarding the minimization of the Pt-CO bonding strength, the author stated that "…from the point view of theoretical calculation, the positively charged Pt weakens the normally strong CO adsorption…". This statement is probably not obvious if we consider that a strong Pt-CO bonding may result from the electron retro-donation from Pt d states to CO molecular orbitals. A Pt positively charged Pt may on the contrary promote CO adsorption. Since this statement is important in the author's discussion, they should emphasize this effect. 6. The author stated that "Similarly, the employment of Pt B.E. could also help derive the expression of TOF' in good agreement with the experimental TOF' as shown in Fig. S27". It is not obvious how the BE was used to derive the TOF. The authors should provide more details regarding this statement.
Reviewer #2 (Remarks to the Author): In this nice study, Chen and colleagues vary the charge on Carbon-supported Pt nanoparticles and measure the effect on the CO oxidation activity. Similar demonstrations of the effect of charge transfer on the catalytic activity have been reported over the past decade, but the detailed characterization of the Pt electronic structure is the key strength of this paper. The paper however has many shortcomings that need to be fixed before publication.
A kinetic model links activity to conditions (T, pressures). This paper does not contain a kinetic study. There is no "breakthrough in microkinetic modeling" or a "nanokinetics model" The use of the term volcano is misleading in this context. In catalysis, volcano curves link activity to a thermodynamic property (adsorption energy, charge, d-band center,…). Here, activity if plotted as a function of the preparation temperature.
It would be nice if the electronic properties of the various carbon supports could be characterized as well, e.g., via C XANES, C XPS, C NMR, or conductivity measurements.
What is the ratio between the number of Pt particles and the number of O defects? Does every particle nucleate at a defect?
In addition to a CO isotopic switch to determine the number of active sites, H2 pulse chemisorption and H2-D2 exchange should be used. Since the CO adsorption energy is sensitive to the Pt charge, the CO isotopic switch might behave differently on the 6 catalysts.
It is unfortunate that no kinetic evaluation has been performed. Light-off curves are generally poor measures of reaction kinetic and are affected by heat and mass transfer. A few measurements around 100C, at limited conversion, and for a range of CO and O2 partial pressures would provide much more information than the data reported here. The orders in CO and O2 would be particularly interesting.
The coverage in the DFT calculations does not match the SSITKA coverages, and the analysis assumes a model that is first order in CO and O2. This is not correct. The correlation between the CO + O2 adsorption energy and the Bader charge is misleading, as the CO and O2 adsorption energy respond very differently to charge (see SSITKA data). CO adsorption generally weakens with charge, O2 adsorption strengthens. From Figure 2f, it seems the effect is largest on O2.
The absence of O2 isotope scrambling (p 11) does not prove non-dissociative adsorption. It proves that reaction of O* with CO* is faster than reaction of O* with O*.
The relation between the CO and the oxygen coverage and the Pt charge measured by SSITKA results from reaction kinetics, not only from the adsorption energy. Moreover, based on theory, one would expect a linear relation between charge and adsorption energy, and hence an exponential relation between coverage and charge.
The manuscript is at times difficult to follow because of poor grammar.
Many thanks for the valuable comments and suggestions from the two reviewers. We have revised our paper by fully taking into account all the comments and suggestions.
Reviewer #1 (Remarks to the Author): The effect of the electronic properties on the adsorption and activation of reactants is an important phenomenon for understanding catalytic reactions mechanistic. Establishing a clear relationship between electronic properties of a catalyst and its activity is a challenging and a long-standing issue in the heterogenous catalysis. In this regard, this study is timely and valuable. The authors conducted a systematic and a solid investigation using a variety of in situ tools and theoretical modeling. They experimentally elucidate the influences of Pt electronic properties on (CO, O2) reactants adsorption properties, especially in operando conditions. Indeed, the activation of these reactants was correlated with Pt electronic properties (Pt charge). I believe that one of the strong assets of this study is the investigation of this phenomenon in situ under realistic catalytic condition of the CO oxidation reaction. Therefore, this manuscript brings sound results on the effect of electronic properties on CO oxidation reaction over Pt catalysts at the molecular-level. In my opinion it deserves to be published in Nature Communications. Figure 4 is unclear (XPS). The caption said "ambient pressure" while the pressure in the experiment chamber shows 20 torr. The ambient pressure stands for atmospheric pressure conditions and temperature (~1 bar, RT). In general, the information on the in situ XPS experiments should be more precise! They should state clearly the conditions of XPS experiments (P and T).

Response:
We are very sorry for this mistake about the pressure of XPS measurement. Indeed, the quasi in situ XPS equipment is made up of two chamber: the reaction chamber is working at ambient pressure, while the analysis chamber working under 10 -9 torr pressure. To make it more clearly, we have revised the relevant description as following: "The catalysts were treated under the real reaction conditions with elevated temperature at a homemade reaction chamber under ambient pressure, and then transferred to the XPS analysis chamber for XPS measurement under 10 -9 torr pressure and room temperature through a load-lock gate without exposure to air." 2. The binding energy shift of Pt 4f7/2 was attributed to "shift is mainly ascribed to the variation in Pt electron density due to the electron transfer between Pt and CNT". What is the driving force for this effect? It would have been interesting to investigate this phenomenon through the examination of the molecular orbitales involved.

Response:
Thanks for this good question. The electronic effects of the oxygen-containing groups (OCGs) of CNT on the supported metal particles could be categorized into two types: i) the inductive effects, involving the polarization of bonds owing to the differences in the electronegativity; ii) the resonance effects, involving the actual movement of electrons through a π-bond system (Organic Pharmaceutical Chemistry. In Remington: The Science and Practice of Pharmacy, 21st ed.; Troy, D. B., Ed.;Lippincott Williams & Wilkins: Philadelphia, PA, 2005;Chapter 25, pp 386−409). Typically, an electron-donating group (EDG) is a functional group that donates some of its electron density to a conjugated π system via the resonance (+R) or inductive effect (+I), whereas an electron-withdrawing group (EWG) removes electron density via the resonance (-R) or inductive effect (-I). Hence, the driving force for this electronic effect is mainly arising from the inductive/resonance effects (Chem. Mater., 2015, 27, 7362-7369).
To make it more clearly, we have conducted atomic orbital analysis based on the suggestions of the reviewer, and added the relevant description in the revised version as following: "Considering the high electron conductivity of CNT (   3. Although I understand the challenge of recording XPS spectra in in situ conditions, using a laboratory setup (not a synchrotron-based one where high photon brilliance exists), a main critic concerns the in situ XPS spectra especially Pt 4f (more specifically Fig. S9). They were acquired with low counts. During the experiment, they should have been accumulated with more spectra in order to improve the resolution. Figure S9 is not good! The convolution of the pics is quite arbitrary! I am surprised to see such poorly conducted XPS experiment.

Response:
Thanks very much for kindly reminding us on this issue. According to your valuable suggestion, we have accumulated with more spectra to improve the resolution of Figure S9, and the results are still in good consistence with the current analyses on the Pt electronic properties. Hence, we have replaced Figure S9 with the new one in the revised version as following: shown are obvious and rather standard. It can be removed or displaced to the Supplementary document.

Response:
We fully agree with the reviewer's good suggestion. Hence, we have removed Figure 4a in the revised version, and the new Figure 4 has been shown as following: 5. Regarding the minimization of the Pt-CO bonding strength, the author stated that "…from the point view of theoretical calculation, the positively charged Pt weakens the normally strong CO adsorption…". This statement is probably not obvious if we consider that a strong Pt-CO bonding may result from the electron retro-donation from Pt d states to CO molecular orbitals. A Pt positively charged Pt may on the contrary promote CO adsorption. Since this statement is important in the author's discussion, they should emphasize this effect.
We are sorry for not emphasizing this effect of electron retro-donation on the adsorption of CO. As the reviewer suggested, a strong Pt-CO bonding may result from the electron retro-donation from Pt d states to CO molecular orbitals. Typically, a decrease of the electron density of Pt results in a decrease of the back-donation of the metal electrons into 2π* antibonding orbitals of the CO molecule and a strengthening of the C-O bond, but not necessarily a weakening of the Pt-CO bond (J. Phys. Chem. B, 2005, 109, 23430-23443;Surf. Sci. 1998, 396, 156-175). Therefore, there is no clear relationship between the charge of Pt and the adsorption of CO. Based on the above analysis, we have revised the relevant discussion as following: "Notably, the above result is still speculative because of the lack of direct evidence on the activation and adsorption of reaction species under operando conditions. Moreover, it has been suggested in previous study that the changes of the electronic properties of metal may result in an increase of electron back-donation, but not necessarily a strengthening of the M-CO bond 12 .
Therefore, in situ kinetics information to bridge the microscopic electronic properties of the Pt active site with the macroscopic catalytic performance of CO oxidation were obtained as follows." 6. The author stated that "Similarly, the employment of Pt B.E. could also help derive the expression of TOF' in good agreement with the experimental TOF' as shown in Fig. S27". It is not obvious how the BE was used to derive the TOF. The authors should provide more details regarding this statement.

Response:
Thanks for the reviewer's kind suggestion. In the revised version, we have added the details on the derivation of TOF based on Pt B.E. in the Supporting Information as following:

The derivation of TOF based on Pt B.E.
As shown in Fig. S31a in which a, b, and c is determined in  Fig. 31f. Reviewer #2 (Remarks to the Author): In this nice study, Chen and colleagues vary the charge on Carbon-supported Pt nanoparticles and measure the effect on the CO oxidation activity. Similar demonstrations of the effect of charge transfer on the catalytic activity have been reported over the past decade, but the detailed characterization of the Pt electronic structure is the key strength of this paper.
The paper however has many shortcomings that need to be fixed before publication.
1. A kinetic model links activity to conditions (T, pressures). This paper does not contain a kinetic study. There is no "breakthrough in microkinetic modeling" or a "nanokinetics model"

Response:
We are very sorry to use these exaggerated words, and we have revised them as following:  Fig. 1b." 3. It would be nice if the electronic properties of the various carbon supports could be characterized as well, e.g., via C XANES, C XPS, C NMR, or conductivity measurements.

Response:
Thanks very much for the reviewer's kind suggestion. We have conducted C 1s XPS and electronic conductivity measurements, and added the relevant description in the revised version as following: "Considering the high electron conductivity of CNT (Table S4)   5. In addition to a CO isotopic switch to determine the number of active sites, H2 pulse chemisorption and H2-D2 exchange should be used. Since the CO adsorption energy is sensitive to the Pt charge, the CO isotopic switch might behave differently on the 6 catalysts.

Response:
According to the reviewer's suggestion, we have conducted H2 pulse chemisorption measurement, and the results are shown in Table S2. It can be seen that the amount of H2 adsorption is higher than that of Pt atoms for each catalyst, which could be attributed to H spillover to carbon support (J. Catal. 1979, 58, 287-295), H diffusion into the bulk (Surf. Sci. 1985, 160, 37-45), or the ability of under-coordinated metal atoms present at the edges and corners of supported particles to bind more than one H (J. Catal. 1972, 24, 367-384). These results are also consistent with the high dispersion of Pt particles over carbon support, and we have added the relevant description in the revised version as following: "The average Pt particle size (dPt) for these catalysts was determined to be 1.2-1.3 nm by averaging 200 random particles (Fig. 1b), and the highly dispersed Pt particles were confirmed by the H2-chemisorption results in Table S2." b based on the Pt particle size from HAADF-STEM measurement. c determined from the 12 CO-13 CO isotopic switches at 100 o C. d based on the reversible adsorption of CO from the 12 CO-13 CO isotopic switches at 100 o C.
*Note: It can be seen that the amount of H2 adsorption is higher than that of Pt atoms for each catalyst, which could be attributed to H spillover to carbon support (J. Catal. 1979, 58, 287-295), H diffusion into the bulk (Surf. Sci. 1985, 160, 37-45), or the ability of under-coordinated metal atoms present at the edges and corners of supported particles to bind more than one H (J. Catal. 1972, 24, 367-384).
These results are also consistent with the high dispersion of Pt particles over carbon support.
6. It is unfortunate that no kinetic evaluation has been performed. Light-off curves are generally poor measures of reaction kinetic and are affected by heat and mass transfer. A few measurements around 100 o C, at limited conversion, and for a range of CO and O2 partial pressures would provide much more information than the data reported here. The orders in CO and O2 would be particularly interesting.

Response:
Thanks for this kind suggestion. We have measured the dependence of reaction rate on the partial pressures of CO and O2, and added the relevant description in the revised version: "Moreover, the reaction orders of CO and O2 for the most positively charged Pt/CNT-600 and negatively charged Pt/CNT-0 were measured as shown in Fig. S30. It is obvious that Pt/CNT-0 exhibits much lower CO reaction order of -0.58 compared with that of -0.10 for Pt/CNT-600, consistent with its higher CO site coverage. On the other hand, the almost same O2 reaction orders around 0.9 further evidence the much lower O2 site coverages for these catalysts." Figure S30. The kinetic reaction orders of (a) CO and (b) O2 for Pt/CNT-0 and Pt/CNT-600.
7. The coverage in the DFT calculations does not match the SSITKA coverages, and the analysis assumes a model that is first order in CO and O2. This is not correct. The correlation between the CO + O2 adsorption energy and the Bader charge is misleading, as the CO and O2 adsorption energy respond very differently to charge (see SSITKA data). CO adsorption generally weakens with charge, O2 adsorption strengthens. From Figure 2f, it seems the effect is largest on O2.

Response:
Thanks for the reviewer's kindly reminding us on the coverage in the DFT calculation mismatching the SSITKA coverage. We have carried out more DFT calculations to study the effects of CO coverage on the reaction energy, in which the CO-preadsorbed models were constructed in Fig. S24. As a result, the energy barrier for the formation of OOCO complex from adsorbed CO and O2 could be calculated as shown in Fig. S25a. It is found that the energy barrier still demonstrate an almost linear decline with Pt Bader charge in Fig. S25b.
Accordingly, we have added the relevant description in the revised version as following: "Moreover, the effects of CO coverage on the reaction energy were investigated, in which the corresponding CO-preadsorbed configurations were constructed in Fig. S24. Accordingly, the energy barrier for the formation of OOCO complex from adsorbed CO and O2 could be calculated as shown in Fig. S25a, which still demonstrates an almost linear decline with Pt Bader charge in Fig. S25b."  CO-preadsorbed Pt-basal, Pt-hydroxyl, Pt-carboxyl, Pt-carbonyl, and Pt-ester. (b) The relationships between the energy barrier (ΔEi) and Pt Bader charge.
Moreover, based on the above analysis, it is found that the rate-determining step remains to be the formation of OOCO complex from adsorbed O2 and CO. Because the reaction order for each reactant in an elementary step is equal to its stoichiometric coefficient, the reaction orders for CO and O2 in this rate-determining step are determined to be 1. To make it more clearly, we have revised the description as following: "According to the above DFT calculations, the rate-determining step for CO oxidation for these Pt catalysts involves the formation of OOCO species from adsorbed CO and O2 (Fig.   2d). Because the reaction order for each reactant in an elementary step is equal to its stoichiometric coefficient, the reaction orders of CO and O2 in this rate-determining step (CO*+O2*→OOCO*+*) are determined to be 1." Lastly, the reviewer is right that the correlation between the CO + O2 adsorption energy and the Bader charge is misleading, because the CO and O2 adsorption energy respond very differently to charge as reflected by SSITKA. Hence, we have removed the correlation in Fig.   2f, and changed the relevant description in the revised version as following: "Considering that the adsorption of reactants, specifically CO, has been suggested as the key factor for this reaction, the effects of Pt charge on their adsorption were further compared in Fig. S26 "Similarly, inspired by the linear dependence of adsorption energy (Eads) on Pt Bader charge (Fig.   2f), the adsorption behaviors of the reaction species (θCO and θoxygen) were further correlated with the Pt charge by exponential functions (Fig. 3e)."  10. The manuscript is at times difficult to follow because of poor grammar.

Response:
In the revised version, we have obtained helps from Editage (www.editage.com) for English language editing, and improved the language as possible as we can. We hope it meet the standard of the publication.

<b>REVIEWERS' COMMENTS</b>
Reviewer #1 (Remarks to the Author): The authors have taken my comments into account and made the recommended changes. The manuscript has been considerably improved as far as I am concerned. I think that this manuscript can be published.
Reviewer #2 (Remarks to the Author): The authors have addressed my questions and somewhat improved the analysis, but some inconsistencies remain. Since the data are interesting, I recommend publication.
Additional suggestions: Question 3: It is counterintuitive that the conductivity of the support (e.g., for CNT-1000 vs CNT-600) does not correlate with the XPS shifts and the derived Pt charges. One would expect that a higher conductivity results from a higher free carrier concentration in the support and hence results in an increased charge transfer to the Pt particles, as is the case with typical semiconductor supports. This is not the case here and the connection between support electronic properties and the catalyst activity is somewhat indirect.
Question 6. The reduction in the CO order is consistent with weaker CO adsorption on Pt/CNT-600. The reduction in the O2 order for Pt/CNT-600 is consistent with stronger O2 adsorption on Pt/CNT-600. The relation between charge and adsorption energy is hence opposite for both reactants, in line with the SSITKA results ( Figure 3e). The DFT analysis misses this point, since only CO and O2 co-adsorption is considered (Figure 2d), and the kinetic analysis compares the energy of the TS with the energy of coadsorbed CO and O2 (energy barrier in Figure 2f). The kinetically relevant energy difference is however between the TS, and adsorbed CO (CO*, zero to negative order) and gas phase O2 (nearly first order). It is unfortunate that the authors do not use this kinetic information to improve the DFT calculations. Pt/CNT-600 is more active because (largely) the reaction is less hindered by strong CO adsorption. The weaker CO adsorption also explains the lower measured activation energy.
The authors have taken my comments into account and made the recommended changes. The manuscript has been considerably improved as far as I am concerned. I think that this manuscript can be published.

Response:
We appreciate the reviewer for the positive comments and the previous valuable suggestions.
Reviewer #2 (Remarks to the Author): The authors have addressed my questions and somewhat improved the analysis, but some inconsistencies remain. Since the data are interesting, I recommend publication.
Additional suggestions: Question 3: It is counterintuitive that the conductivity of the support (e.g., for CNT-1000 vs CNT-600) does not correlate with the XPS shifts and the derived Pt charges. One would expect that a higher conductivity results from a higher free carrier concentration in the support and hence results in an increased charge transfer to the Pt particles, as is the case with typical semiconductor supports. This is not the case here and the connection between support electronic properties and the catalyst activity is somewhat indirect.

Response:
Thanks for this valuable question. Indeed, the electric conductivity of carbon materials has been widely observed to increase with the annealing temperature ascribed to the elimination of oxygen-containing groups (Carbon, 2013, 59, 2-32;Carbon, 2016, 96, 174-183;Carbon, 2019, 147, 27-34), which is consistent with the trend in Table S4.
Moreover, for semiconductor supports, the impurities (doping) changes the number of charge carriers (electrons or holes) and therefore changes the Fermi level, resulting in increased mobile negative charge carriers in the conduction band or positive charge carriers in the valence band (Cowper, P., 2017. Azulene-eee en el ell.). Similarly, in this work, the thermal treatment of oxidized support changed the relative concentrations of electron-withdrawing groups (EWG) and electron-donating groups (EDG), which would decrease and increase the electron density of Pt, respectively. In this regard, we also made correlations between the Pt B.E., Pt charge and the molar ratio of EWG to EDG (nEWG/nEDG) as depicted in Fig. 1f, and further the catalytic activity in Fig. 2c. To make it more clearly, we have revised the relevant description as following: Table S4.  Question 6. The reduction in the CO order is consistent with weaker CO adsorption on Pt/CNT-600. The reduction in the O2 order for Pt/CNT-600 is consistent with stronger O2 adsorption on Pt/CNT-600. The relation between charge and adsorption energy is hence opposite for both reactants, in line with the SSITKA results ( Figure 3e). The DFT analysis misses this point, since only CO and O2 co-adsorption is considered (Figure 2d), and the kinetic analysis compares the energy of the TS with the energy of co-adsorbed CO and O2 (energy barrier in Figure 2f). The kinetically relevant energy difference is however between the TS, and adsorbed CO (CO*, zero to negative order) and gas phase O2 (nearly first order). It is unfortunate that the authors do not use this kinetic information to improve the DFT calculations. Pt/CNT-600 is more active because (largely) the reaction is less hindered by strong CO adsorption. The weaker CO adsorption also explains the lower measured activation energy.

Response:
Thanks for this constructive comment. We fully agree with the referee that the reduction in the CO order is consistent with the weak adsorption of CO, and the increase in the O2 order is also consistent with the strong adsorption of O2. In this regard, we have conducted DFT calculations to gain a fundamental understanding of the effects of Pt charge on the adsorption and activation of reactants. Generally, the gas molecule adsorption ability determines the reaction pathways on the catalyst. Due to the much larger adsorption energy of CO compared with O2, the Pt surface could be dominantly covered by CO. In this case, the Pt reactive site would be blocked to hinder the ER reaction, in which the coadsorption of CO and O2 on Pt and the LH reaction should be favored, which has been verified by previous DFT calculations (Phys. Chem. Chem. Phys. 2012, 14, 16566-16572). Hence, the LH reaction pathway was chosen to study the energy change across the reaction coordinate for these models, and the results were also in line with the SSITKA results as the referee said.
Moreover, according to the referee's suggestion, it is still interesting to compare the energy change between the LH and ER reaction to incorporate the electronic effects of oxygencontaining groups on reaction pathway. Considering the large computational cost to obtain reliable results, it will be studied in our future work.