In situ identification of the metallic state of Ag nanoclusters in oxidative dispersion

Oxidative dispersion has been widely used in regeneration of sintered metal catalysts and fabrication of single atom catalysts, which is attributed to an oxidation-induced dispersion mechanism. However, the interplay of gas-metal-support interaction in the dispersion processes, especially the gas-metal interaction has not been well illustrated. Here, we show dynamic dispersion of silver nanostructures on silicon nitride surface under reducing/oxidizing conditions and during carbon monoxide oxidation reaction. Utilizing environmental scanning (transmission) electron microscopy and near-ambient pressure photoelectron spectroscopy/photoemission electron microscopy, we unravel a new adsorption-induced dispersion mechanism in such a typical oxidative dispersion process. The strong gas-metal interaction achieved by chemisorption of oxygen on nearly-metallic silver nanoclusters is the internal driving force for dispersion. In situ observations show that the dispersed nearly-metallic silver nanoclusters are oxidized upon cooling in oxygen atmosphere, which could mislead to the understanding of oxidation-induced dispersion. We further understand the oxidative dispersion mechanism from the view of dynamic equilibrium taking temperature and gas pressure into account, which should be applied to many other metals such as gold, copper, palladium, etc. and other reaction conditions.

This work reported an in situ characterizations of the oxidative dispersion of Ag nanostructures on Si3N4 surface during CO oxiation. A typical oxidative dispersion process Ag remains in the metallic state due to the chemisorption of O2. It is interesting to see that the Ag nanoclusters could be oxidized upon cooling in the O2 atmosphere. The mechanism was explained by dynamic equilibrium considering temperature and pressure and the authors proposed that such mechanism could be employed to further demonstrate the oxidative dispersion of many other metals. The authors claimed the importance of the new data. Specific comments are listed below.
(1) A brief description of the synthesis process of Ag nanowires should be given in SM.
(2) What about the desorption energies and reaction rates?
(3) How to consider the surface defects as the defects may affect the reaction mechanism dramatically? (4) Although the authors tried explaining the reaction mechanism, the reactivity of CO2 production is not so attractive compared to other catalysts. Also, the durability and reproducibility of the obtained results were not provided.
Reviewer #2 (Remarks to the Author): In the manuscript "In situ identification of the metallic state of Ag nanoclusters in oxidative dispersion" the Dalian groups with correspondign authours Gao, Yang, Fu and Bao present an interesting work on the dispersion of silver-nanowires on a Si3N4-support under oxidizing conditions. Such sintering and (re)-dispersion studies are extremely important in catalysis and need further attention to obtain stable catalysts. The Dalian groups are well known in the field of single site catalysis, high-end characterization and heterogeneous catalysis in general. For this study, state of the art and highly attractive methods like NAP-XPS and ESEM were used to in situ follow the dispersion process of Ag. As model reaction CO-oxidation was used. The main outcome of the study is that the dispersion of silver is observed under oxidizing conditions whereas aggregation of the Ag nanostructures is observed under more reducing conditions. I have the following comments: 1) This dispersion/redispersion is a concept that has been reported in literature as also outlined by the authors, e.g. the intelligent catalysts by Toyota (ref. 8). It has also been the topic of several review articles (e.g. ref. 3 and 4). In addition to the studies given in the manuscript, also a few further may be cited (e.g. L. Kibis, J. Phys Chem. C, 10.1021/acs.jpcc.7b09983) The authors claim that the understanding has been "established mainly based on ex situ spectroscopic characterization…" (first page of "main") and in the abstract they write "Lacking in situ characterizations, however, the interplay …" has not been well illustrated. I disagree with these statements, as there are various studies that report "in situ" studies, e.g. Yasutaka Nagai et al. report in Ang. Chem. (doi.org/10.1002/anie.200803126, 2008 oxidative redispersion of Pt nanoparticles supported on ceria-based oxide in situ XANES analysis. An atomic migration model accounts for the observed redispersion through the trapping of Pt species at sites on the Ce support that exhibit strong interactions between the Pt oxide and the support. There are various further in situ studies like in situ TEM (ref. 9) that show redispersion in oxygen and aggregation in reducing conditions, which was further extended e.g. by G. Ferre et al. recently (Catal. Sci Technol. , doi: 10.1039/d0cy00732c). Also, on Pt/TiO2 (L. DeRita, et al., Nature Materials volume 18, pages746-751(2019)) in situ studies have been reported. A good additional study is on Rh-based ceria catalysts by Ikemoto et al., also studying the reversible behavior by in situ spectroscopies (PCCP, 2019, doi.org/10.1039/C9CP04625A). Turning to silver-based catalysts as outlined in this study, also in this case the dynamics has been reported before, also by in situ TEM (e.g., Y. Gao, Catal. Sci. Technol., 10.1039/c7cy00831g, 2017, soot oxidation e.g. D. Gardini et al., Appl. Catal. B, 10.1016/j.apcatb.2015.10.029, 2016. Hence, the novelty of the manuscript needs to be sharpened and it is therefore not acceptable in its present form. 2) Also other groups like the Schlögl group have been very active in the area of silver catalysed oxidation and need attention. Maybe I overlooked a corresponding reference as mostly the first authour is only given.
3) The dispersion of Ag and its dispersion is described to occur on reduced Ag species with "adsorbed" oxygen rather than oxidized Ag. However, this claim is not clear for me and also not sufficiently convincing for me, as no quantification of the oxidation state of the Ag-species are performed with XPS (only the position of the peak maximum is presented). Note also that in the abstract the authors write "…we unravel that in such a typical oxidative dispersion process Ag actually remains in its metallic state and the dispersion is caused by the chemisorption of O2." A chemisorption -so the formation of a chemical bond -of O2 is an oxidation if considering cluster/particles. The authors should overthink their wording. They state that Ag is not oxidized, but it is still called a "oxidative dispersion". This is contradictory. 4) When reading the manuscript, it becomes not clear that 3 different type of samples were used for the characterization and not only 1 sample. Especially for deducting mechanistic insights, this is to my opinion misleading, since due to the different surface concentrations, the samples can behave as well differently. In the experimental part it is written: "prepared three kind of samples with different Ag density (high, medium, and low). To acquire strong Auger spectra, samples with high NW density were used and shown in Fig. 4. Samples with medium AgNW density were used to get obvious change of Ag/Si ratio during various treatments, also to make XPS peak intensity strong enough, shown in Figs. 1-3." 5) Comment to the conclusion: "We provide a solid evidence for the fact that oxide formation is not necessary in oxidative dispersion, but a result during cooling in O2." This is contradictory to the DFT results, where it was stated that "On the contrary, associative O2 and dissociative O atoms adsorb stably at the Ag3-Si3N4 ( = -2.24 eV) and Ag8-Si3N4 ( = -3.97 eV) interfaces". The provided data does suggest, that a surface oxidation (nothing else is a O2 chemisorption of "dissociative O atoms") is needed for the dispersion and not a bulk oxidation. As mentioned before, the wording has to be improved and the terminology should be defined to enhance the clearness of the manuscript.
In summary, the manuscript is not recommended for publication in Nature Communications in its present version because the novelty is not clear enough yet, the process of redispersion (chemisorption / oxidation) and also some experimental details/choice of the experimental study (systematics) are not clear.

Reviewer #3 (Remarks to the Author):
It is an interesting and solid piece of study where in situ XPS technique has been applied to study the "oxidative dispersing" of Ag particles. It was observed that metallic Ag, instead of Ag oxide, was the main species being dispersed. The authors further confirmed that the chemisorption of O2 is the driving force for the re-dispersion of metallic Ag clusters. While the study is thorough, and selfconsistent, its wide applicability and general validity remains a question. What is reported here may represent a special case rather than a general phenomenon.
First of all, Si3N4 is not a typical support. Ag/Si3N4 is not a widely used commercial catalyst. Ag/Si3N4 is not a representative system. There is no guarantee that the dispersion behaviour of metal on non-oxides to be the same as on oxides. Further, oxidative re-dispersion is normally conducted at

Reviewer #1:
This work reported an in situ characterizations of the oxidative dispersion of Ag nanostructures on Si 3 N 4 surface during CO oxidation. A typical oxidative dispersion process Ag remains in the metallic state due to the chemisorption of O 2 . It is interesting to see that the Ag nanoclusters could be oxidized upon cooling in the O 2 atmosphere. The mechanism was explained by dynamic equilibrium considering temperature and pressure and the authors proposed that such mechanism could be employed to further demonstrate the oxidative dispersion of many other metals. The authors claimed the importance of the new data.

Response:
We thank the referee for the positive evaluation of our work.
Specific comments are listed below.
1) A brief description of the synthesis process of Ag nanowires should be given in SM.

Response:
We thank the referee for the suggestion. A brief description of the synthesis process of Ag nanowires has been added in the revised SI on Page 3. Also see as below: "In a typical synthesis, PVP (K-90) was dissolved in ethylene glycol (160 mL, 12.5 mg /mL) at 150 °C with constant stirring. Then, the sodium chloride in ethylene glycol (16 mL, 0.42 mg/mL) and silver nitrate in ethylene glycol (40 mL, 50 mg/mL) were added when the solution was cooled to room temperature. The mixture was heated in an oven at 110 °C for 12 h after vigorous magnetic stirring for 5 min, to obtain AgNWs." 2) What about the desorption energies and reaction rates?

Response:
We thank the referee for this question. We have carried out additional DFT calculations for the desorption energy and reaction rate using the model of Ag 8 cluster.
The energy pathway of the reaction is shown in Fig. R1. The energy barrier during the reaction is 0.11 eV, in the elementary step of forming adsorbed OCO* from adsorbed O* and CO*. According to the transition state theory, the reaction rate constant of this step is calculated as 2.10 × 10 /s at 673 K.
The details for this calculation are shown below and added to the revised SI on Page 14 (supplementary Fig. 16). We also found that the formed CO 2 is readily to desorb from the cluster. The energy of the system becomes lower after the desorption.
The reaction rate constant K 1 can be calculated based on the transition state theory: is the partition function of the difference species, including rotational, transitional and vibrational parts. Here, it was assumed 1. Finally, we can obtain = 2.10 × 10 /s (∆ = 0.11 , = 673 ) 3) How to consider the surface defects as the defects may affect the reaction mechanism dramatically?

Response:
We fully agree with the referee on this point that surface defects may dramatically affect the reaction mechanism, particularly on reducible oxides. However, Si 3 N 4 is a very inert material containing rare surface defects and can remain stable under harsh conditions [Ref. 55]. This is also confirmed by our XPS results (see 4) Although the authors tried explaining the reaction mechanism, the reactivity of CO 2 production is not so attractive compared to other catalysts.

Response:
We thank the referee for pointing this out. We are aware that Ag is not usually considered as an excellent metal catalyst for low temperature CO oxidation. In case of our Ag/Si 3 N 4 model catalyst of this work, the low temperature activity (CO conversion) is not so efficient largely due to the ultra-small metal loading of only 0.001mg per sample. (see Fig. R3a Tod. 2019,332,[189][190][191][192][193][194]. To address this point, we have replaced the CO 2 production rate with CO conversion in Fig. 5b and the corresponding description in the revised manuscript on Page 13 as "The highly dispersed Ag nanoclusters formed during reaction exhibit an enhanced activity especially at low temperature, higher than the initial state of AgNWs upon heating." We have also modified the supplementary Fig. 12 to show mass spectroscopy results of both CO and CO 2 in the revised SI on Page 12. 22]. However, it was unable to fully understand the correlation between the metal dispersion/redispersion and the corresponding oxidation state during the dynamic process of dispersion/redispersion. As we present in this work, the oxygen adsorption (lower oxidation degree), likes a ligand in organic chemistry, is able to protect the Ag clusters from sintering. It opens a new way to utilize the common gas (O 2 ) as the protector. To our best knowledge, our study is the very first work to perform in situ microscopy and spectroscopy under exactly the same conditions which leads us to a solid evidence for the identification of metallic state (to be exactly, nearly metallic state with low surface oxidation degree) during Ag dispersion in O 2 .
To be accurate on this point, we have removed all the phrases: "lacking in situ characterizations" in the revised manuscript. Furthermore, we have revised the abstract in the revised manuscript on Page 2 as "……However, the interplay of gas-metal-support interaction in the dispersion processes, especially the gas-metal interaction has not been well illustrated. …… we unravel a new adsorption-induced dispersion mechanism in such a typical oxidative dispersion process. The strong

gas-metal interaction achieved by chemisorption of O 2 on metallic Ag clusters is the internal driving force for dispersion. ……"
We have also revised the main in the revised manuscript on Pages 3-4 as "……The common consensus on the oxidative dispersion includes the formation of mobile metal oxide species from large metal particles and capture of these species on the support, which is regarded as the most crucial step in the so-called oxidation-induced dispersion mechanism revealed by in-situ TEM and in-situ spectroscopy including near ambient pressure X-ray photoelectron spectroscopy (NAP-XPS) and in-situ X-ray absorption spectroscopy (XAS). Nagai et al. ascribed the redispersion of Pt nanoparticles to the formation of Pt oxide species observed in O 2 at 873 K by in-situ XANES which is due to the strong PtO-support interaction. The underlying mechanism needs to be fully depicted.……" We have cited these papers as the referee suggested above when discussing the dispersion mechanism (see Refs. 21-28).
2) Also other groups like the Schlögl group have been very active in the area of silver catalysed oxidation and need attention. Maybe I overlooked a corresponding reference as mostly the first authour is only given.

Response:
We thank the referee for pointing this out. Schlögl's and Ertl's groups have made great contributions in the fundamental studies of silver-based catalysis including both model-and real-catalysts, and we have cited a few of their papers as refs. 45 and 46 in our manuscript.
To further address their contribution in this field, we have cited one more relevant paper by Schlögl's and Ertl's group in the discussion of oxygen adsorption and silver-catalyzed oxidation reactions. See Refs. 46,47,53,54 in the revised manuscript.
3) The dispersion of Ag and its dispersion is described to occur on reduced Ag species with "adsorbed" oxygen rather than oxidized Ag. However, this claim is not clear for me and also not sufficiently convincing for me, as no quantification of the oxidation state of the Ag-species are performed with XPS (only the position of the peak maximum is presented).

Response:
For the quantification of Ag oxidation state, we actually use Auger parameters i.e. the sums of Auger kinetic energy (KE) and photoelectron BE, because XPS core-level of Ag is unable to distinguish its oxide or metallic states. As shown in Figure R5 (adopted from Handbook of X-ray Photoelectron Spectroscopy), the Auger parameters of metallic Ag is around 726 eV, whereas the value for silver oxide is usually lower than 725 eV. The values of the Auger parameters can represent the degree of oxidation in Ag quantitatively, which has been widely adopted in literatures [Ref. 11,12 in SI].
To make clear this point, we have revised the manuscript on Page 6: "(see the description of Auger parameter in the supplementary discussion)", and added description to supplementary discussion in the revised SI on Page 5, as "For the quantification of Ag oxidation state, we employed Auger parameters i.e. the sums of Auger kinetic energy (KE) and photoelectron BE, because XPS core-level of Ag is unable to distinguish its oxide or metallic states. As shown in supplementary Fig. 17 (adopted from Handbook of X-ray Photoelectron Spectroscopy), 10 the Auger parameters of metallic Ag is around 726 eV, whereas the value for silver oxide is usually lower than 725 eV. The values of the Auger parameters can represent the degree of oxidation in Ag quantitatively which has been widely adopted in many other literatures. 11, 12 "

Figure R5. Wagner plots for Ag.
Note also that in the abstract the authors write "…we unravel that in such a typical oxidative dispersion process Ag actually remains in its metallic state and the dispersion is caused by the chemisorption of O 2 ." A chemisorption -so the formation of a chemical bond -of O 2 is an oxidation if considering cluster/particles. The authors should overthink their wording. They state that Ag is not oxidized, but it is still called a "oxidative dispersion". This is contradictory.

Response:
We agree with the reviewer that the chemisorption of O 2 would lead to some charge transfer between Ag clusters and O 2 molecules. However, the difference in the degree of oxidation makes our findings different from the conventional understanding. In the conventional understanding, the dispersed metal atoms/clusters normally have a high oxidation degree (formation of bulk oxides) because they form strong chemical bonds not only to O from the atmosphere but also to the lattice oxygen of the oxide support surface. The later one is considered as a key point to stabilize the dispersed atoms/clusters [Refs. 12,21,and 22]. Our work shows that besides the strong metal-support interaction, the dispersion could occur with the assistance of O 2 in the atmosphere. This is also shown in our theoretical model that the dispersion is energetically favored when only 1 out of 3 and 4 out of 8 Ag atoms are bonded to O. Particularly, in the case of Ag 3 cluster the dispersion was caused by the non-dissociative adsorption of molecular oxygen rather than dissociative adsorption of atomic oxygen, which verifies a preferred metallic state under in situ conditions (1 mbar, 673 K).
To make our point clearer, we used the wording of "high" and "low" oxidation degree in the revised manuscript following the reviewer's suggestion. Also see the changes in revised manuscript, "bulk oxidation with a high oxidation degree" on Page 9, "nearly metallic clusters with a low surface oxidation degree" on Page 9, etc.

4) When reading the manuscript, it becomes not clear that 3 different type of samples
were used for the characterization and not only 1 sample. Especially for deducting mechanistic insights, this is to my opinion misleading, since due to the different surface concentrations, the samples can behave as well differently. In the experimental part it is written: "prepared three kind of samples with different Ag density (high, medium, and low). To acquire strong Auger spectra, samples with high NW density were used and shown in Fig. 4. Samples with medium AgNW density were used to get obvious change of Ag/Si ratio during various treatments, also to make XPS peak intensity strong enough, shown in Figs. 1-3."

Response:
We thank the referee for pointing this out. Here, the purpose of using three types of samples with different Ag density is only to improve the signal-to-noise ratio of XPS data for presentation, as the intensity drops dramatically under different environmental pressures. In our experiments, the three samples show the same trend of behavior/dynamics during dispersion, despite only slight difference in the dispersion rate (duration to reach equilibrium) and dispersion degree, as verified by XPS in Fig. 1e, Fig. 3b, and Fig. 4c.
To further eliminate the effect of Ag density, we have performed additional ESEM experiments on the dispersion of Ag nanowires with medium and high density. As shown in Figure R5, both samples are dispersed into small clusters/particles after the O 2 treatment and become invisible under the SEM imaging condition. We can thus safely conclude that the surface Ag density has minor effect on the dispersion dynamics under our experimental conditions.
To address this point, we have put an additional supplementary Fig. 16 (Fig. R6) and supplementary discussion in the SI on Pages 6 and 15: "Effect of Ag density. The effect of surface Ag nanowire density can't be ignored when we discuss the dispersion process. As shown by XPS (Fig. 1e, Fig. 3b and Fig.   4c) and SEM ( Fig. 1c and supplementary Fig. 14) results, the three samples showed the same trend of behavior/dynamics during dispersion, despite only slight difference in the dispersion rate (duration to reach equilibrium) and dispersion degree. All samples were dispersed into small clusters/particles after the O 2 treatment and become invisible under the SEM imaging condition. We can thus safely conclude that the surface Ag density has minor effect on the dispersion dynamics under our experimental conditions." 5) Comment to the conclusion: "We provide a solid evidence for the fact that oxide formation is not necessary in oxidative dispersion, but a result during cooling in O 2 ." This is contradictory to the DFT results, where it was stated that "On the contrary, associative O 2 and dissociative O atoms adsorb stably at the Ag 3 -Si 3 N 4 ( = −2.24 eV) and Ag 8 -Si 3 N 4 ( = −3.97 eV) interfaces". The provided data does suggest, that a surface oxidation (nothing else is a O 2 chemisorption of "dissociative O atoms") is needed for the dispersion and not a bulk oxidation. As mentioned before, the wording has to be improved and the terminology should be defined to enhance the clearness of the manuscript.

Response:
We thank for the referee's comment. Following the referee's suggestion, here we distinguish a surface oxidation with a low oxidation degree from a bulk oxidation with a high oxidation degree. We have changed the text of "oxide formation" to be "a bulk oxide with a high oxidation degree". See the changes in the revised manuscript as "bulk oxide with a high oxidation degree" on Page 9, "AgO x cluster with high oxidation degree" on Page 10, etc.
In summary, the manuscript is not recommended for publication in Nature Communications in its present version because the novelty is not clear enough yet, the process of redispersion (chemisorption/oxidation) and also some experimental details/choice of the experimental study (systematics) are not clear.

Reviewer #3:
It is an interesting and solid piece of study where in situ XPS technique has been applied to study the "oxidative dispersing" of Ag particles. It was observed that metallic Ag, instead of Ag oxide, was the main species being dispersed. The authors further confirmed that the chemisorption of O 2 is the driving force for the re-dispersion of metallic Ag clusters. While the study is thorough, and self-consistent, its wide applicability and general validity remains a question. What is reported here may represent a special case rather than a general phenomenon.

Response:
We thank the referee for the positive evaluation of our paper as a thorough and self-consistent study. 1) First of all, Si 3 N 4 is not a typical support. Ag/Si 3 N 4 is not a widely used commercial catalyst. Ag/Si 3 N 4 is not a representative system. There is no guarantee that the dispersion behaviour of metal on non-oxides to be the same as on oxides.

Response:
We agree with the referee that Si 3 N 4 is not often used as catalyst support.
The choice of this substrate in the work is to make use of the inert property of the and also reducible oxide: SrTiO 3 (100) (Fig. R7c). The same dispersion behavior of Ag (upshift of Ag 3d BE induced by size effect and increasing Ag 3d peak intensity) was clearly observed under oxygen-rich environment as shown in Figure R7, verifying that our findings on Si 3 N 4 surfaces is a general phenomenon which can be applied to other oxides. The nearly-metallic state of Ag during dynamic dispersion process was also observed on silica substrate (SiO 2 /Si(111)) ( Fig. R8).
To address this point, we have added an extra supplementary Fig. 3, 7, and supplementary  (110) substrates. The upshift of Ag 3d BE and the increase in the peak intensity were observed on all the substrates (supplementary Fig. 3), similar to that on the Si 3 N 4 surface." "Similar metallic state of dispersed Ag species was also observed on SiO 2 /Si(100) in O 2 at elevated temperatures, verifying a universal phenomenon validated for SiO 2 and other oxides (see Supplementary Fig. 7)."   2) Further, oxidative re-dispersion is normally conducted at 1 bar in air. The concentration of O 2 is orders of magnitude higher than the O 2 used in in situ XPS.
This will change hugely the oxidation behaviour of Ag. Unless the same observation is obtained on a more typical metal oxide support under conditions closer to general practice, the implication of the study is compromised. Consider doing this requires effort beyond a major revision, I would suggest the authors to go for a more specialized journal. It is a very well conducted piece, and I feel it can be published in any surface chemistry journals with essentially no revision.

Response:
We thank the referee for the important comment. Firstly, the universality of a more typical metal oxide support has been fully addressed by performing the same investigation over the Ag nanowires on silica, sapphire and SrTiO 3 surfaces, as shown in the response to the former question, as well as in the revised manuscript on Pages 6 and 10 and the revised SI on Pages 8 and 10.
Regarding the O 2 pressure effect, our quasi in-situ XPS experimental results for the 1 bar pressure have indicated the same dispersion behavior (see supplementary Fig. 11).
Though XPS experiment for 1 bar pressure is still technically limited by the current NAP-XPS technique, the previous works have confirmed that increasing the O 2 pressure from 1 mbar to 1 bar at 673 K will lead to higher coverage of adsorbed oxygen but not the oxidation (Phys. Rev. Lett. 2003, 90, 256102) and Ag 2 O decomposes at the temperature of 673 K in air (Phys. Chem. Chem. Phys. 2001, 3, 3838-3845). Therefore, from the thermodynamic point of view the same mechanism should be valid in 1 bar O 2 as that in 1 mbar O 2 .
To address this point, we have added a paragraph to supplementary discussion in the revised SI on Page 6, as "Though XPS experiment for 1 bar pressure is still technically limited by the current NAP-XPS technique, the previous works have confirmed that increasing the O 2 pressure from 1 mbar to 1 bar at 673 K will lead to higher coverage of adsorbed oxygen but not the oxidation 18 and Ag 2 O decomposes at the temperatures of 673 K in air. 19 Therefore, from the thermodynamic point of view the same mechanism should be valid in 1 bar O 2 as that in 1 mbar O 2 ."