In situ observations of an active MoS2 model hydrodesulfurization catalyst

The hydrodesulfurization process is one of the cornerstones of the chemical industry, removing harmful sulfur from oil to produce clean hydrocarbons. The reaction is catalyzed by the edges of MoS2 nanoislands and is operated in hydrogen-oil mixtures at 5–160 bar and 260–380 °C. Until now, it has remained unclear how these harsh conditions affect the structure of the catalyst. Using a special-purpose high-pressure scanning tunneling microscope, we provide direct observations of an active MoS2 model catalyst under reaction conditions. We show that the active edge sites adapt their sulfur, hydrogen, and hydrocarbon coverages depending on the gas environment. By comparing these observations to density functional theory calculations, we propose that the dominant edge structure during the desulfurization of CH3SH contains a mixture of adsorbed sulfur and CH3SH.

The application of the reactor STM for this kind of work in hydrodesulfurization is a very strong achievement on a very high technical level, and is a valuable next step in model studies of this important catalyst. Unfortunately the STM data set produced is not very extensive, which is probably due to the significant challenges involved, and this also influences the overall strength of the conclusions (see below). It may be of general interest to the readers of a multidisciplinary journal such as Nature Communications to observe that in-situ scanning probe microscopy has improved to the level where sulfidic systems can be studied. This is a spectacular achievement, but in terms of new findings in the particular field of hydrodesulfurization catalysis, the paper is less significant at this point. Furthermore the analysis related to comparison between STM experiment and modelling has some problems that need to be addressed.
-One of the difficulties in the comparison between experiment and STM simulations is that basically all fine structure predicted in simulations is argued to be smeared out due to fast diffusion of species along the edge at the imaging temperature. While these assignments are quite reasonable given the temperature, the overall determination of the exact edge structure using this imaging method becomes somewhat speculative since such a scenario will fit with any suggested edge structure. Here a more extensive set of experiments could have been helpful in supporting the conclusions, i.e. temperature series, or imaging quenched samples in vacuum afterwards.
-The comparison between STM images to determine the changes in position of edge protrusions between Figure 2a-c needs to be quantitative in order to become convincing. The image quality makes it hard to judge from the illustrations, so further quantitative measures such as line scans and distances need to be included and compared with the simulations. It is hard to discern exactly what the authors mean by edge protrusions. The authors use the terms "in" and out of "registry", but a 2D grid placed on the particle is needed to determine this in detail instead of just a single guide line. Since so much of the paper scope relies on the observation that something happens between figure 2a, b and c when the authors claim to see an effect of the CH3SH, this important aspect needs to be stronger presented.
-STM image quality needs to be discussed in general, in particular in relation to the resolution on the edges, as the images presented here appear show different types of contrast within the same particles ( Figure 1, and large images in figure 2). Only small averaged sections are discussed, but how robust are the conclusions when applied to the other edge of the same particles within a dataset? -Theory and also recent IR and TEM results (ACS Catal. 6, 1081(2016) indicate that the MoS2 shape itself should be a function of the conditions, i.e. that the somewhat regular triangular shape in Fig 1 should change to a more hexagon like shape when exposed to hydrogen. Was there any indication in the experiment to address this effect? -Previous STM images (e.g. Phys. Rev. Lett. 87, 196803 (2001)) have resolved bright stripes adjacent to the edge (claimed to be due to metallic states), but these seem much differently presented here or even absent. Is this an effect of the STM imaging conditions or due to H2 at elevated pressure? -The calculations use a Au surface which is stretched to match the MoS2 lattice. Is the Au underneath the MoS2 also stretched in the experiment? This is a significant elongation of the bonds and correspondingly this would make the Au surface in the model more reactive. What influence does this have on the conclusions and the equilibrium coverages? -It is mentioned that a (1x1) Au(111) surface is imaged during exposure (figure 1b). Does this mean that the herringbone reconstruction on the Au(111) surface is lifted due to the thiols at these pressures? In addition, I am puzzled about the long term build-up of S on the Au? How does this structure compare with previously proposed S-adlayer structures on Au? And was an experiment without MoS2 on the surface done to correlate if it is entirely due to the CH3SH? -I am missing information on what was done to ensure purity of the gases supplied? Since H2S is corrosive, there is a need to select materials in the gas systems to prevent transport of impurities. Also to which extent does the CH3SH decompose before it reaches the sample? Was some kind of spectroscopy (e.g. XPS) used to confirm sample purity after the experiment? -It is highlighted in the paper that a fractional coverage of 63% S and 50% H is surprisingly found to be stable under imaging conditions, which contrasts previous findings. Most previous studies looked at less wide unit cells, and the emergence of the claimed new structures here is most likely due to the better resolution of a 4 unit cell wide model. The trends are in reality the same, and I suppose if one used a 5 unit cell wide model, one would see and even more gradual transition between high S and low S coverages and a more complicated phase diagram. In addition, it is likely that corners of a particle may play a role here (as also discussed in the paper), which could change the picture altogether.
-How does the claimed presence of CH3-S species adsorbed on the edges compare with literature? There are very extensive operando IR studies on the catalysts system, so it would be interesting if additional support exists in the literature for such stable hydrocarbon species. Some studies also report on Mo-carbide formation under working conditions, but is the structure proposed here compatible with this idea?
Reviewer #2 (Remarks to the Author): Using advanced STM techniques combined with DFT calculations, the authors have performed a detailed investigation toward detection of the in-situ variation of active edge site of MoS2 under CH3SH hydrodesulfurization conditions. After careful evaluation of the manuscript, I think the developed STM techniques are state-of-the-art, the computational methods are reasonable. Moreover, the paper is novel and the conclusions are original which will be of interest to others in the broad community. Thus, I suggest the paper should be published in Nature communications eventually. However, I have three comments and the authors should be cautiously considered at this stage. The first one is, the phase diagram of MoS2 edge structures in H2/H2S mixtures have been extensively explored in previous studies. Although the authors listed the result offsets with those of published results in the supporting information, a detailed comparison should be given. For examples, what is the reference of sulfur structures in these literatures, how do they get these offsets and the relative energies, and what is the difference with the phase diagrams reported previously, i.e., the emergence of 63% and 38% S. Secondly, the authors use CH3SH as a model compound to simulate the S-contained species in oil. However, the compositions of oil are rather complicated, depending on many factors, such as place of origin, etc. The authors should comment whether CH3SH has a representativeness for Scontained oil. I am wondering what the situations are for other typical model compounds, such as thiophene, tetrahydrothiophene and other refractory S-contained species. Similar question is, the authors claimed that they use a dedicated high-pressure STM, but the pressure is still largely lower than the realistic condition. I don't know whether it can represent the real situation. Finally, the rates calculated by formula (12) seem questionable. If two gases involved in the elementary reaction, two coverages of theta should be used. Kn and Kn-1 should be explained in the text. I suggest the kinetic parts should be fully described.
Reviewer 1 -One of the difficulties in the comparison between experiment and STM simulations is that basically all fine structure predicted in simulations is argued to be smeared out due to fast diffusion of species along the edge at the imaging temperature. While these assignments are quite reasonable given the temperature, the overall determination of the exact edge structure using this imaging method becomes somewhat speculative since such a scenario will fit with any suggested edge structure. Here a more extensive set of experiments could have been helpful in supporting the conclusions, i.e. temperature series, or imaging quenched samples in vacuum afterwards.
We thank the reviewer for this comment, which requires some additional evidence. This can, however, not be obtained through a temperature series or quenching for two reasons: first, changing the temperature will result in different thermodynamic and kinetic conditions, which may well change the edge state that is obtained. To give an extreme example: the 100%S Mo edge cannot be reduced using pure H 2 in ultrahigh vacuum due to kinetic limitations. Higher temperatures in our ReactorSTM setup are not an option, as we are already close to the limit of the chemical and thermal stability of our Kalrez reactor seal. The second reason is that CH 3 SH adsorption and (rapid) decomposition on Au(111) occurs at temperatures (slightly) lower than probed here. This will result in the formation of a sulfur overlayer on the Au(111) substrate. The rate at which this occurs is much higher at intermediate temperatures than at our reaction temperature due to the increased CH 3 SH coverage on the surface. The edge sites of MoS 2 particles surrounded by a sulfur overlayer are exceedingly difficult to probe, and may be altered due to the presence of the overlayer.
That said, we would like to strengthen our argumentation by explicit simulation of the temperatureaveraged STM images. In the main text, we have added a temperature-averaged simulation of the 38%S-25% CH 3 SH structure to Figure 4. This simulation clearly shows that the edge protrusions do not follow the registry of the basal plane atoms, similar to our experimental observations. We also performed temperature averaging for the 50%S-50%H structure in Figure S2. Again, this simulation corroborates our conclusion that the 50%S and 50%S-50%H structures look very similar to eachother, yet distinctly different from the 38%S-25% CH 3 SH structure.
-The comparison between STM images to determine the changes in position of edge protrusions between Figure 2a-c needs to be quantitative in order to become convincing. The image quality makes it hard to judge from the illustrations, so further quantitative measures such as line scans and distances need to be included and compared with the simulations. It is hard to discern exactly what the authors mean by edge protrusions. The authors use the terms "in" and out of "registry", but a 2D grid placed on the particle is needed to determine this in detail instead of just a single guide line. Since so much of the paper scope relies on the observation that something happens between figure 2a, b and c when the authors claim to see an effect of the CH 3 SH, this important aspect needs to be stronger presented.
Following the opportunities for stronger presentation suggested by the reviewer, we have implemented a 2D grid in Figures 2, 4, and S1-4. To clarify that terms "in registry" and "out of registry" refer to the position of the edge protrusions with respect to the basal plane atomic registry we added the following sentence in the last paragraph of page 3: "Indeed, Figure 2a shows that the apparent location of the edge atoms does not follow the registry of the basal plane atoms." To ensure clarity of the term "edge protrusions" we added the sentence: "To clearly separate the nomenclature for apparent and actual edge atom positions, we refer to the apparent positions of the edge atoms as edge protrusions hereafter." We should point out that further quantitative analysis of the data will likely lead to overinterpretation, as the noise level in the tunneling current due to fast adsorption/desorption processes on the tip apex induce inhibitively high noise levels in the tunneling current. However, as also discussed in the next point, we have added additional analyses of edge structures in the supporting information to provide a more confident assignment.
-STM image quality needs to be discussed in general, in particular in relation to the resolution on the edges, as the images presented here appear show different types of contrast within the same particles (Figure 1, and large images in figure 2). Only small averaged sections are discussed, but how robust are the conclusions when applied to the other edge of the same particles within a dataset?
We thank the reviewer for raising this point, which we should further clarify. In regular STM experiments, the tip is often conditioned until the tip apex is symmetrical. Once a symmetrical apex is obtained, it will usually remain in this good state for a prolonged amount of time. Under our highpressure, high-temperature conditions however, frequent changes in the tip apex are unavoidable. Therefore, obtaining an extremely sharp and symmetric apex for a period long enough to acquire an image is virtually impossible, even with a PtIr tip. Hence, the images show some asymmetry, often showing sufficient resolution on only part of the edges. In the manuscript, this point is addressed at the top of page 4: "Note that some tip asymmetry could not be avoided under high-pressure, hightemperature conditions, resulting in somewhat asymmetrical image sharpness. To prevent misinterpretation, the conclusions from Figure 2b were corroborated using additional data (see Figure S3)." To ensure correct interpretation of the edge structure under HDS conditions, we followed the reviewer's suggestion and analyzed the alternative edge of Figure 2c. The result is shown in Figure S4 and discussed in the neighboring paragraph.
-Theory and also recent IR and TEM results (ACS Catal. 6, 1081 (2016)) indicate that the MoS 2 shape itself should be a function of the conditions, i.e. that the somewhat regular triangular shape in Fig 1 should change to a more hexagon like shape when exposed to hydrogen. Was there any indication in the experiment to address this effect?
Upon the start of our experiments, we had the same expectation as the reviewer. To test for shape changes, we exposed as-prepared particles to hydrogen for several hours. We counted the fraction of pure triangular particles before and after this exposure, but did not find a convincing indication for any shape change. We attribute this to kinetic limitations, that were also observed in somewhat similar experiments by Lauritsen et al. Journal of Catalysis 221 (2004) 510-522. In the meantime, the same group has shown that shape change can be induced at higher temperatures. As this point is clearly of interest, we have added it to the first paragraph of page 4.

-Previous STM images (e.g. Phys. Rev. Lett. 87, 196803 (2001)) have resolved bright stripes adjacent to the edge (claimed to be due to metallic states), but these seem much differently presented here or even absent. Is this an effect of the STM imaging conditions or due to H 2 at elevated pressure?
The bright stripes at the particle edges were usually observed, independent of the gas environment, although their intensity appears to depend on the tip structure (adsorbates on the apex) and likely also on the edge state. However, we have shown mostly differentiated images in the manuscript, in which the bright stripes appear much weaker. We have now stressed this in the last paragraph of page 3: "We should also point out that the STM images in Figure 2 were differentiated to highlight the atomic contrast. In this display method, the metallic "bright" edge states that are usually observed at the edges of the MoS 2 particles are less apparent (see also Figure S5 in the Supporting Information)." Furthermore, we note in the second paragraph of page 4 that: "In agreement with the observations on reduced MoS 2 edge structures in the literature, we note that the "bright" metallic edge states are maintained (see Figure S5 in the Supporting Information)". Figure S5 shows the image in Figure 2b as a regular line-by-line background subtracted image rather than in differentiated display.

-The calculations use a Au surface which is stretched to match the MoS 2 lattice. Is the Au underneath the MoS 2 also stretched in the experiment? This is a significant elongation of the bonds and correspondingly this would make the Au surface in the model more reactive. What influence does this have on the conclusions and the equilibrium coverages?
The reviewer raises a just point here. We noted that we had not referred to Section 2 in the SI in the main text, which discusses the effect of the Au(111) support. We have added this reference in the discussion of Figure 3 and addressed the overestimation of the support effect due to the stretching of the Au(111) substrate in section S2 of the Supporting Information. We have also extended this discussion to the adsorption of CH 3 SH in Supporting Information Section S4.

-It is mentioned that a (1x1) Au(111) surface is imaged during exposure (figure 1b). Does this mean that the herringbone reconstruction on the Au(111) surface is lifted due to the thiols at these pressures? In addition, I am puzzled about the long term build-up of S on the Au? How does this structure compare with previously proposed S-adlayer structures on Au? And was an experiment without MoS 2 on the surface done to correlate if it is entirely due to the CH 3 SH?
We have intensively studied the interaction of CH 3 SH and Au(111) (more publications will follow), and added more details based on the importance that the reviewer points out here. Page 3, 3 rd paragraph: "However, the absence of the herringbone reconstruction observed on clean Au(111) [Barth et al., PRB 42, 1990] indicates that some (dissociated) CH 3 S is present on the surface. Decomposition of CH 3 S or H 2 S leads to the formation of a sulfur layer over time (see Figure 1c), with a structure resembling that observed by Lay et al. [Lay et al., Langmuir 19, 2003]. The formation of the sulfur overlayer occurred independent on whether MoS 2 particles were present on the Au(111) substrate or not."

-I am missing information on what was done to ensure purity of the gases supplied? Since H 2 S is corrosive, there is a need to select materials in the gas systems to prevent transport of impurities. Also to which extent does the CH 3 SH decompose before it reaches the sample? Was some kind of spectroscopy (e.g. XPS) used to confirm sample purity after the experiment?
The reviewer raises an important point here, which we addressed experimentally, yet did not discuss in the manuscript. We have added the following in the methods section: "The employed gases (Westfalen AG), Ar N5.0, H 2 N5.0 and CH 3 SH N2.8 (main impurities dimethylsulfide and dimethyldisulfide) were fed through particle filters before use. Their purity was confirmed using mass spectroscopy. To prevent corrosion, the gas lines were made of Hastelloy C alloy, whereas the reactor consists of PEEK, Zerodur glass and Kalrez. Measurements on the bare Au(111) surface under HDS conditions confirmed the absence of impurity deposition other than the slow formation of a sulfur overlayer".
-It is highlighted in the paper that a fractional coverage of 63% S and 50% H is surprisingly found to be stable under imaging conditions, which contrasts previous findings. Most previous studies looked at less wide unit cells, and the emergence of the claimed new structures here is most likely due to the better resolution of a 4 unit cell wide model. The trends are in reality the same, and I suppose if one used a 5 unit cell wide model, one would see and even more gradual transition between high S and low S coverages and a more complicated phase diagram. In addition, it is likely that corners of a particle may play a role here (as also discussed in the paper), which could change the picture altogether.
We agree in part with the reviewer on this point. On the one hand, a larger unit cell may indeed yield an even larger variety of stable edge structures (as stressed in the first paragraph on page 5: "However, we should point out that for a larger unit cell size, an even larger variety of structures may appear in the phase diagram."). On the other hand, the transition from 0%S to 100%S is not gradual over the whole range of chemical potentials. As pointed out in the second paragraph of page 5, many phases are "skipped" in the phase diagram, because the intermediate coverage is never more stable than either the lower or the higher coverage. For instance, 50%S, 50%H is missing in the phase diagram for Au-supported MoS 2 . The effect is even stronger at high S coverage, where the 75% S and 87% S are not present. Hence, we do feel that the stability of the 63%S and 50%H phase, in particular with respect to the 50%S phases, is surprising and meaningful. To make it more clear for the reader why this phase has not been found in the literature, we extended our literature analysis in section S3 in the ESI: "We should note that there were also differences in the employed unit cells and the range of investigated structures. Where  Obviously, larger unit cells lead to more flexibility in the observed structures. The low-symmetry 38%S-x%H and 63%S-x%H structures which we have found to be stable over a wide range of conditions were not accessable to Bollinger et al., Lauritsen et al. and Cristol et al.. Prodhomme et al. did use the same unit cell size as in the present work, but did not consider structures with a coverage higher than 50%S. On the other hand, they did consider the 38%S structures and found their stability somewhat lower than in our work: the transition from 38%S to 50%S occurs already at Δμ S =-0.76 eV, whereas our calculations (without Au support) put the transition at -0.51 eV. This difference could be related to the differences in used methodology (VASP/projector augmented waves for Prodhomme et al. versus the orbital-based basis set used in the BAND package employed here). The orbital-based basis set used in our case is particularly suited for the simulation of situations where the electron density shows strong gradients, as is the case with the irregular 38%S-x%H and 63%S-x%H structures. This could lead to a higher stabilization of these structures." -How does the claimed presence of CH3-S species adsorbed on the edges compare with literature? There are very extensive operando IR studies on the catalysts system, so it would be interesting if additional support exists in the literature for such stable hydrocarbon species. Some studies also report on Mo-carbide formation under working conditions, but is the structure proposed here compatible with this idea?
We thank the reviewer for this suggestion, which allows us to further substantiate our claims. While no direct comparison to DFT or IR studies for CH 3 SH on Au-supported MoS 2 is available, some relevant studies on unsupported catalysts have been conducted. We have added the CH 3 SH adsorption energies for unsupported MoS 2 in section S4 in the ESI and show good agreement with literature DFT. Furthermore, we have provided references to in situ and ex situ IR spectroscopy showing that adsorption of hydrocarbons in H 2 -hydrocarbon mixtures is a general phenomenon that can be expected to occur over a wide range of HDS conditions, and that adsorption is promoted by having a reducing atmosphere (see the first paragraph on page 8).
We agree with the reviewer that the possibility of Mo-carbide formation on industrial HDS catalysts should not be excluded, given the clear evidence that MoS 2 is prone to form Mo 2 C under certain conditions (Zhang et al., Int. J. of Min. Met and Mat. 25, 2018, Kelty et al., App. Cat. A 322, 2007, Jeon et al., ACS Nano 12, 2018. However, it seems that carbon-rich, sulfur-poor conditions are necessary and the reaction is usually carried out at high temperatures. Hence, Mo 2 C formation under our sulfur-rich, low-temperature conditions seems unlikely. This point is now made in the first paragraph on page 5: "Mo carbide formation can be excluded, since our low-temperature, sulfur-rich HDS environment is far away from the conditions for which Mo 2 C or MoS x C y formation were observed [30][31][32] ."

Reviewer 2
The first one is, the phase diagram of MoS2 edge structures in H2/H2S mixtures have been extensively explored in previous studies. Although the authors listed the result offsets with those of published results in the supporting information, a detailed comparison should be given. For examples, what is the reference of sulfur structures in these literatures, how do they get these offsets and the relative energies, and what is the difference with the phase diagrams reported previously, i.e., the emergence of 63% and 38% S.
To improve the completeness of the literature comparison, we have listed the reference and unit cell size used in the cited papers in section 3 of the supporting information. A detailed discussion on the emergence of the 38% and 63% structures is also provided.
Secondly, the authors use CH3SH as a model compound to simulate the S-contained species in oil. However, the compositions of oil are rather complicated, depending on many factors, such as place of origin, etc. The authors should comment whether CH3SH has a representativeness for Scontained oil. I am wondering what the situations are for other typical model compounds, such as thiophene, tetrahydrothiophene and other refractory S-contained species.Similar question is, the authors claimed that they use a dedicated high-pressure STM, but the pressure is still largely lower than the realistic condition. I don't know whether it can represent the real situation.
We thank the reviewer for raising these questions, as it appears we have not sufficiently stressed the important discussion of representativeness. We have addressed this in the second paragraph on page 3: "Naturally, a single organosulfur compound can never fully represent the complex mixture in oil. However, mercaptans such as CH 3 SH constitute a major component in crude oil 23 . Our applied temperature and pressure (250 °C, 1 bar) are close to the typical hydrodesulfurization conditions applied for the light naphtha fraction (260-380 °C, 5-10 bar) 2 and are sufficient to achieve catalytic turnover 24 ." Additionally we discuss in the first paragraph on page 8: "In our experiments, this leads to the counterintuitive observation that the edge S content is reduced due to CH 3 SH adsorption: the 63%S-50%H structure is favored in the absence of CH 3 SH adsorption, whereas the 38%S-25%CH 3 SH structure is the most stable one when we do take CH 3 SH adsorption into account. This effect is likely also present for other industrially important reaction intermediates such as reduced thiophenes, which adsorb even stronger than CH 3 SH 33 . Indeed, infrared spectroscopy has shown that thiophene adsorption is much more pronounced in a reducing atmosphere 34 . Other compounds were also found to adsorb under HDS conditions 35 . However, weakly adsorbing and/or sterically hindered (e.g. dimethyldibenzothiophene) organosulfur molecules may not have sufficient interaction with the MoS 2 edge to significantly alter the catalyst's resting state." Finally, the rates calculated by formula (12) seem questionable. If two gases involved in the elementary reaction, two coverages of theta should be used. Kn and Kn-1 should be explained in the text. I suggest the kinetic parts should be fully described.
Based on the remarks of the reviewer, it is clear that we have insufficiently described the kinetic model. To aid the reader, an extensive description of the model has now been added to the supporting information (section 5).