Model catalyst synthesisSingle-layer Co-promoted MoS2 slabs were grown on an Au(111) substrate by simultaneous physical vapor deposition of Co and Mo followed by a crystallization step, all under an H2S background. This synthesis procedure is adapted from the work of Kibsgaard et al.20. and is described in detail in the experimental methods section. Figure 1a shows an STM image of the single-layer Co-promoted MoS2 synthesized on Au(111). All Co-promoted MoS2 slabs are observed to have a hexagonal shape. This shape is attributed to the substitution of the S-edge Mo atoms by Co which renders this edge thermodynamically stable under the sulfur-rich synthesis conditions17,20. Pristine MoS2 slabs synthesized under the same conditions exclusively have a triangular shape displaying only the Mo-terminated edges due to the S-edge being unstable [see Helveg et al.25. and the SI, S3]. Additionally, the Co-promoted MoS2 slabs in Fig. 1a have a bright periphery along their edges. These bright edge sites are further resolved in Fig. 1b, which show that the bright edge features are present along the penultimate atomic rows. The bright 1D edge features are attributed to the buildup of electronic states in order to compensate for the polarization due to the lattice termination at the edge26,27. These 1D metallic edge states are also called brim sites in literature17. The intensity of the brim sites varies between adjacent edges exposed to vacuum, as illustrated in Fig. 1b. This variation arises from the diverse edge terminations—whether Mo or S termination—and their stability, along with their interactions with the Au substrate and gases. Different types of dangling bonds may exist on the edges as a result of these interactions. This leads to varying degrees of polarity at the Co- and Mo-terminated edges, resulting in differing levels of edge metallicity that contribute to the brightness of the brim sites. Figure 1a also shows the formation of 2D CoS2 islands which are discussed in our prior work28. As the Co sulfides are not known to be active for HDS reactions, we focus on the Co-promoted MoS2 slabs in this work.Fig. 1: As-synthesized CoMoS model catalyst in UHV.a Large-scale STM image of Co-promoted MoS2 slabs on an Au(111) substrate; b Zoomed-in STM images showing the Co- and Mo-terminated edges and the brim sites (STM images acquired at sample voltage of −1 V and tunneling current of 150 pA). A lattice is drawn in b for the ease of visualizing the edge registry. In b, blue circles indicate the expected positions of the protrusions and green circles mark the actual positions.In Fig. 1b, the edges of a CoMoS slab are shown with atomic resolution. In studies of MoS2 and Co-promoted MoS2 slabs, the alignment of edge protrusions in STM images with respect to the line of atoms in the basal plane—referred to as edge registry—strongly correlates with the atomic structure of the edge17. Past STM investigations have often revealed that the observed locations of bright features along the edges may not precisely correspond to the expected atomic positions. Instead, they might be slightly translated along one or more of the crystallographic directions, or along a line perpendicular to the line of atoms on the edge25. Moreover, recent STM research by Grønborg17 and Mom18 has highlighted that both the edge registry and the brim site contrasts are highly sensitive to the interaction of edge S atoms with the Au substrate. By considering all of these effects in the STM images, it becomes possible to deduce the atomic structure of the edge.The bright edge protrusions along the edge marked (2) in Fig. 1b of the Co-promoted MoS2 slab are observed to be in registry with the line of atoms in the basal plane of the CoMoS slab while on the adjacent edge marked (1) in Fig. 1b, they are imaged as being out-of-registry with the line of atoms in the basal plane. We define the bright features on the edges as edge protrusions in this work, for the sake of clarity. Grønborg et al.17. have suggested, based on ab initio thermodynamic modeling, that the 100% S saturated Co and Mo edges are the most stable edges under the physical vapor deposition synthesis conditions. Additionally, for the Mo edge, combined experiments and DFT modeling in the past have confirmed that a thermodynamically stable 100% S edge results in an out-of-registry edge structure in the STM images17,18,29,30. This type of edge has two sulfur atoms for every Mo atom. Therefore, the edge (1) of the Co-promoted MoS2 slabs in Fig. 1 which has an out-of-registry edge structure is identified as the Mo-terminated edge.It is, therefore, expected that the adjacent edge marked (2) with ‘in-registry’ edge protrusions, is the Co-substituted S edge based on the 2D Wulff construction of the Co-promoted MoS2 slabs20. For a discussion on the STM image simulation of the 100% S Co-substituted S edge, we refer to the SI S4. Furthermore, the possibility of 2D CoS2 sheets influencing the edges (both Mo and Co-substituted S) was also considered for the analysis presented in this work. We found that only those step edges of the Co-promoted MoS2 nanoclusters which directly interact with 2D CoS2 (or even Au step edges) have a different atomic structure, which is expected considering the chemical bonding between the two phases. Only very small nanoclusters (<1 nm) not considered in this work are strongly influenced by neighboring phases. In our study, most of the nanocluster edges are too long (7 to 15 sites) to be influenced by this effect. For Fig. 1 (and later for Fig. 2 in this work), only the edges exposed to the vacuum are considered for analysis. For atom-resolved STM images of Co-promoted MoS2 nanoclusters not interacting with 2D CoS2 slabs, we refer to the additional STM images provided in SI, Fig. S3B.Fig. 2: CoMoS model catalyst under H2 and CH3SH.a Large-scale STM image of Co-promoted MoS2 slabs on Au(111) measured in situ under 1 bar H2 at 510 K (sample voltage = −1 V, tunneling current = 200 pA); b Zoomed-in STM image of activated Co-promoted MoS2 slabs on Au(111) measured in situ under 1 bar H2 at 510 K (sample voltage = −0.3 V, tunneling current = 500 pA); c STM image of activated Co-promoted MoS2 slabs on Au(111) measured in situ under 0.1 bar CH3SH at 510 K (sample voltage = −0.25 V, tunneling current = 600 pA). The inset shows the out-of-registry structure on the Mo edge marked (1) in c; d Zoom of the region marked by a dotted rectangle in c, showing the complex edge structure on the activated Co edge marked (3) in c. Lattices are drawn in b–d for the ease of visualizing the edge registry. Blue circles indicate the expected positions of the protrusions and green circles mark the actual positions of the edge protrusions.In situ observation under reaction gasesHaving synthesized the single-layer Co-promoted MoS2 slabs, we investigated the effects of the individual HDS reaction gases, namely H2 and CH3SH on the Mo and Co-terminated edges as a next step towards understanding the HDS of CH3SH. CH3SH is chosen as a model organosulfur molecule to represent the aliphatic sulfur-containing organic contaminants present in fossil fuels. CH3SH itself is a very common contaminant in most industrial feeds which are desulfurized commercially. Of course, commercial feedstock also contain aromatic sulfides like substituted thiophenes and benzothiophenes, where the S atom is not only involved in the aromatic bonding, but also is sterically hindered. In such molecules, more complex HDS reaction pathways involving one or more hydrogenation steps before the desulfurization (C-S bond breaking) steps are expected. To understand the multi-pathway desulfurization of complex aromatic molecules like substituted thiophenes, it is important to separate the hydrogenation and the C-S scission steps. Additionally, the desulfurization of substituted thiophenes typically involves ring opening after hydrogenation steps31 and therefore, a fundamental understanding of the desulfurization of the C1-C6 thiols with in situ experimentation is necessary and currently, this experimental information is lacking. The usage of a simple gaseous thiol like CH3SH meets many of these challenges, as the C-S bond breaking occurs without the need of prior hydrogenation steps, thereby allowing the investigation of many fundamental questions such as the effect of incorporating the Co promoter on the C-S bond breaking process, by directly imaging the active sites in situ using the STM. Furthermore, usage of C2-C6 thiols (liquids) in a set up like the ReactorSTM requires significant modifications for handling liquid thiols which are beyond the scope of this experimental work. Nevertheless, later in this work, we also discuss the applicability of our results to thiophene-like molecules, with an aim of using complex HDS precursors in our future explorations.The as-synthesized Co-promoted MoS2 slabs are first imaged in situ under pure H2 (see Fig. 2). We expect that the hydrogen exposure will reduce and activate the Co and Mo-terminated edges by forming unsaturated edge sites. The activated CoMoS slabs are also imaged in situ under pure CH3SH to understand the interactions between an organosulfur molecule like CH3SH and the activated edges, which are expected to be sulfided again upon exposure to a sulfiding gas.Figure 2a, b shows the large-scale and zoomed-in in situ STM images of the reduced Co-promoted MoS2 slabs obtained during exposure to 1 bar hydrogen gas. Figure 2a shows that the truncated hexagonal shape and morphology of the Co-promoted MoS2 slabs remain largely unchanged after 180 min of reduction in 1 bar of H2 (510 K) which is very well in agreement with ab initio thermodynamics calculations and the hydrogen reduction experiments of Grønborg and coworkers (see also, SI, S3)17. The zoom-in of one of the CoMoS slabs in Fig. 2a is shown in Fig. 2b. In Fig. 2b, we observe that all edges have bright edge protrusions which are in registry with the line of atoms in the basal plane. This change of the edge structure compared to the as-prepared system under UHV indicates that the edges of the CoMoS slabs have been reduced due to the reaction with H2. However, the set of edges marked 1 and 3 in Fig. 2b have a much brighter brim compared to the ones marked 2, 4 and 6. Grønborg et al.17. and Mom et al.18. have extensively studied this reduction process using STM and DFT. Particularly, Grønborg et al.17. have used adsorption of molecules like pyridine to confirm the formation of dissociated H on the CoMoS S edges. Based on their work, upon reduction, a 100% S saturated Mo edge with double S edge atoms is expected to reduce into a 50% S edge with single S edge atoms. This reduced edge may also be populated with additional hydrogen atoms depending on the H2 partial pressure. The possibility of additional H occupancy was not considered extensively in the atomic models of Mom and coworkers18. Nevertheless, a 50% S Mo edge is expected to show an in-registry-edge-apparent structure in the STM images with a bright brim17,18. Mom et al.18. have experimentally observed this transformation of the Mo edge on pristine MoS2 slabs using the ReactorSTM during atmospheric-pressure exposure to H2 and temperature conditions similar to the ones used in our experiment. For the Co edge, Grønborg and co-workers17 have shown that at low hydrogen partial pressures, reduction may occur with and without partial hydrogenation of the Co-substituted edge. The S edges without hydrogenation have a 50% S structure where the single sulfur atoms are laying low to interact with gold and hence are not imaged well by the STM tip, while the hydrogenated S edges also have a 50% S structure but with some sulfur atoms hydrogenated. The formation of -SH causes the sulfur atom to interact less with the Au substrate due to the electron donating effects of the H and thus are imaged well by the STM tip. The strong edge S–Au interaction in the case of the 50% S edge without hydrogenation also causes brim site quenching. However, a partial hydrogenation even on a single site was found to repopulate the edge states causing the local brim sites on the Co edges to appear very bright in contrast to those on the Mo edges in the STM images. Having attained a ~1000 times higher hydrogen pressure than in the experiments of Grønborg and coworkers17, we expect to favor additional kinetic processes which will allow for a very high occupancy of hydrogen on the Co edge. In our experiment we find that all edges of the Co-promoted MoS2 slabs are imaged clearly with bright brim site contrasts, which supports this expectation further. Nevertheless, based on the work of Mom and Grønborg17,18, we identify the even numbered edges of Fig. 2b with the darker brim as the Mo-terminated edge and the odd numbered edges in Fig. 2b with a very bright brim as the Co-substituted S edge, both of which may have additional H occupancy. We further verify this assignment with our own DFT modeling which is discussed in the subsequent paragraphs.Having observed the activated CoMoS model catalyst under hydrogen, we next observe the model catalyst under 0.1 bar 100% CH3SH which is close to the typical partial pressures of organosulfides during industrial HDS reactions. Figure 2c, d shows the STM images of reduced Co-promoted MoS2 slabs acquired in situ under 0.1 bar CH3SH at 510 K. On the edges marked 1, 2 and 6 in Fig. 2c (also, see inset), a regular arrangement of bright protrusions which are out-of-registry with the line of atoms in the basal plane is observed. We identify this type of edge observed in a sulfiding gas environment as the Mo-terminated edge. It is well-known that as long as the sulfur saturation conditions are met, pristine MoS2 slabs grown with different types of sulfiding agents form triangular slabs with 100% S Mo edge terminations which show out-of-registry edge protrusions in the STM images identical to the one observed in Fig. 2c32,33. The Mo edges in our experiment are expected to behave in a similar fashion as there are only traces of hydrogen in the gas stream, if any, for any significant reducing effects on the Mo edge. On the edges of the CoMoS slab marked 3, 4 and 5 in Fig. 2c (see also, Fig. 2d), a more complex edge structure consisting of repeating units of a pair of diffuse bright protrusions in registry with the basal line of atoms and a darker out-of-registry protrusion occurring at an immediately adjacent edge site is observed.To further understand the changes occurring on each of the edges under H2 and CH3SH, we make use of DFT calculations. We first try to understand the reduction process of the CoMoS slabs. A 12-atom long 1D periodic CoMoS stripe model discussed earlier, but with a 50% S saturation on the Co and Mo edge is considered as the basis, as we expect that the hydrogen gas will partially reduce both edges. It is also very likely that the edges will be equilibrated to the H2 gas given the long waiting time before acquiring the STM images. Therefore, the H occupancy of both the Co and Mo edges to find the most stable configuration is considered. The free energies of the Mo and Co edges for various H occupancies are shown in Fig. 3a. We find that a 50% S Mo edge with 50% H occupancy and a 50%S Co edge with 100% H are the most stable among the 7 possible edge configurations considered. A CoMoS stripe model with these stable edge configurations is relaxed on a two-atom thick gold slab and the resulting STM image is calculated. Because of the high temperature used and the ease of H hopping on the Mo and the Co edge (discussed later in this work), we expect the STM images of both edges to be time-averaged in the laboratory time frame. Figure 3b shows the time-averaged simulated STM image of the Co and Mo edges with 50%S 100%H and 50%S 50%H structures, respectively. Figure 3b shows that both time-averaged edges have an in-registry configuration of the edge protrusions with respect to the basal plane matching excellently with the experimental STM image in Fig. 2b. Furthermore, the 50%S 100%H Co edge has a brighter brim compared to the 50%S 50%H Mo edge which again agrees well with our interpretation of the experimentally observed Co and Mo edges in Fig. 2b.Fig. 3: DFT modeling of the CoMoS slab edges.a Graph showing the ΔG vs edge H occupancy for both Co and Mo edges with 50% S occupancy; b simulated STM image of a 50% S 100% H Co edge; and c simulated STM image of a 50% S Co edge with a single dissociatively adsorbed CH3SH molecule. The overlay of the atomic model is also shown. Blue atoms are those of Co, grey atoms are of H, lavender atoms are those of Mo, yellow are those of S, and orange are of Au, and black atoms are those of C.For the CH3SH adsorption, the interaction of a single CH3SH molecule with the Co edge is considered. We find that the complex paired edge structure observed in Fig. 2d on the Co edge can best be explained by the dissociative chemisorption of CH3SH on the Co edge with 50% S coverage, as shown in Fig. 3c. This is not surprising as the dissociative chemisorption of CH3SH forming adsorbed methanethiolate (CH3S-) and an adjacent –SH on the edges of supported and unsupported MoS2 slabs have been detected through infrared (IR) spectroscopy in many previous experiments34,35,36,37. Furthermore, many molecular adsorption studies of organosulfur compounds on the edges of MoS2 and CoMoS model catalysts in the past have shown a strong influence of the adsorbates (such as atomic H-) on the local density of states on the edges and the nearby brim sites38,39,40,41. A significant change in the edge structure is therefore expected in the STM images upon CH3SH adsorption. The Co edge structure in Fig. 3c also involves an additional thiol transfer to the nearby site after the adsorption of CH3SH, the mechanism of which is discussed later in this work. Adsorption experiments of a variety of organothiols at vacuum pressures on the edges of CoMoS slabs have shown that their dissociative chemisorption leads to the formation of very similar intensely bright and dark paired lobes on the edge39,41,42. Additionally, one may also argue that when a 50%S–100%H hydrogenated Co edge is exposed to CH3SH, the dissociative chemisorption can also lead to self-desulfurization, since it generates adsorbed CH3S- and -SH units of which the CH3 and the H may interact with each other and form methane. We show later in this paper, that this reaction pathway has a very high activation barrier at 510 K to be of any industrial significance. Nevertheless, the observation of such a CH3S- and SH- structure only on the Co edge, shows that the complete sulfidation of the Co edge is hindered because of the much stronger binding of CH3SH to the Co edge compared to the Mo edge at 510 K. Additionally, in our experiment, the bright (or dark) edge features on the re-sulfided Co edges were always separated by at least one atomic site, with the paired feature having a preferred orientation in all of the re-sulfided CoMoS slabs imaged under CH3SH. The potential origins of the preferred orientation of the paired CH3S- and -SSH units on the Co edge are discussed later in this work.Having observed the effects of the individual gases on the edges of Co-promoted MoS2 slabs, a fresh identical sample containing Co-promoted MoS2 slabs supported on Au(111) is prepared to study the HDS reaction of CH3SH. After reducing the CoMoS slabs under 1 bar H2, a desulfurizing gas feed containing 9:1 H2:CH3SH is let into the STM reactor to initiate the HDS reaction, while maintaining the total pressure and temperature. The experimental procedure for attaining the HDS reaction conditions is detailed in the experimental methods section. Using a temperature close to the industrial HDS temperature, we expect some HDS chemical reaction to occur. We investigate the Co-promoted MoS2 slabs by acquiring their STM images every 10–15 s. Figure 4a–h shows the STM images acquired 14 s apart under the HDS reaction gas. The STM images in Fig. 4 show that one set of the edges (marked 2,4 and 6, Fig. 4a) of the hexagonal slabs have the brim contrast quenched while the adjacent edges (marked 1,3 and 5, Fig. 4a) have a relatively much higher brim contrast. The later also shows a static out-of-registry structure with respect to the basal plane which is identified in Fig. 4h. The features of this edge match well with those of Mo-terminated edges of pristine MoS2 slabs from the work of Mom et al.18, where it was shown that a 38% S-25% CH3SH edge structure over the Mo edge is most stable at the reaction conditions used. In their work, an assumption was made that the HDS reactions are too slow to disturb the equilibrium of the Mo edge with H2 and CH3SH. At an imaging temperature of 510 K, the structure undergoes time averaging, resulting in an out-of-registry edge structure with respect to the basal plane atoms. This phenomenon is expected to occur on the Mo edge of a CoMoS slab as well. Additionally, we assume that any spillovers along the corner sites from the adjacent Co edge do not affect the chemistry of the Mo edge. We designate the edges exhibiting the brightest brim sites and out-of-registry edge protrusions (marked 1,3 and 5, Fig. 4a) as the Mo-terminated edges of the CoMoS slab. Consistently, we observe that the adjacent set of edges (marked 2,4 and 6, Fig. 4a) of the CoMoS slabs exhibit high activity under the reaction conditions and display a darker brim compared to the Mo edges. Following the Wulff construction, we attribute this active edge as the Co-substituted S edge. Furthermore, we observe significant variations in the lengths of the Mo and Co edges among the nanoclusters. The size of the edges of the slabs was not found to affect the assignment made here.Fig. 4: CoMoS model catalyst under HDS conditions.a–h Succesive STM images of a CoMoS slab supported on Au(111) imaged in situ during the desulfurization of CH3SH. The images were acquired 14 s apart with a sample voltage of −0.3 V and a tunneling current of 700 pA under 1 bar of 9:1 H2:CH3SH at 510 K. The blue arrows indicate the location of the transient bright features observed on the Co edge. The arrows in a–g indicate the positions of the bright edge feature; h STM image of a CoMoS slab supported on Au(111) showing the out-of-registry structure on the Mo edge. A lattice is drawn in h for the ease of visualizing the edge registry. In h, blue circles indicate the expected positions of the protrusions and green circles mark the measured positions of the edge protrusions.In the successive STM images in Fig. 4a–g, a time-varying edge structure which consists of the appearance and disappearance of bright features can be seen on the Co edge. The process was observed to occur over all Co-promoted MoS2 slabs imaged over a continuous period of 16 h. The time-varying edge structure was found to have no relation to the bias voltage, tunneling current, scanning speed and direction or the duration of the scanning, indicating that the process was not influenced by the STM tip but rather a result of exposing the CoMoS slabs to the HDS reaction gases. Scanning at speeds slower than 30 s/image resulted in poor resolution of the edges suggesting a turn-over frequency of the order of ~10−2 s−1 for the edge structure fluctuations. This kind of turn-over frequency is expected for the HDS reaction conditions used in our experiment. In fact, highly-optimized industrial catalysts have reported turn-over frequencies of 10−2 to 100 s−1 43,44. The observation of a time-varying edge structure also strongly suggests that the Co edges of the CoMoS slabs are not at equilibrium and that their atomic structure is reaction-rate governed. Unfortunately, the HDS reaction gases (especially hydrogen) are also found to strongly interact with the Pt-Ir tip and cause many spontaneous tip state changes, making STM imaging to better resolve the time-varying edge structure over long durations of time very challenging.To resolve the time-varying edge structure better, it is necessary to successfully image the CoMoS slabs with good atomic resolution over long periods of time. For this purpose, a fresh CoMoS model catalyst was synthesized by the physical vapor deposition (PVD) process and reduced in the ReactorSTM with 1 bar H2 for 60 min, followed by the transition to lower-pressure HDS conditions with a total pressure of 0.3 bar. Furthermore, a milder HDS gas mixture containing 1:1 H2:CH3SH was used. By lowering the partial pressure of hydrogen, we expect that the HDS reactions on the Co edge will slow down and better resolution of the time-varying edge structure can be attained. Figure 5a–k shows the successive STM images of Co-promoted MoS2 slabs imaged in situ under 0.3 bar, 510 K, with a gas composition of 1:1 H2:CH3SH. We observe that the use of lower pressure and milder HDS conditions significantly improves the imaging quality. In Fig. 5, the edge identification is carried out in a similar fashion to that for Fig. 4 as we are still within HDS conditions. We make an assumption that the slightly lower hydrogen partial pressure does not significantly affect the edge chemistries. The edges with bright brim sites (marked 1,3 and 5, Fig. 5e) are observed to have an out-of-registry structure with respect to the basal plane atoms. These edges are designated as the Mo terminated edges with a 38%S–25% CH3SH time-averaged structure18. The adjacent edges (marked 2,4 and 6, Fig. 5e) with the darker brim sites are identified as the Co-substituted S edges. The successive STM images in Fig. 5a–k very clearly show that Co edges of the Co-promoted MoS2 slabs have a time-varying edge structure consisting of the appearance and disappearance of bright and diffuse edge features similar to those observed on the Co edge after exposing a reduced CoMoS slab to CH3SH. The process is observed for up to 44 h over all CoMoS slabs imaged.Fig. 5: CoMoS model catalyst under low-pressure HDS conditions.a–k Succesive STM images of a CoMoS slab supported on Au(111) imaged in situ during the desulfurization of CH3SH under mild HDs conditions. The images were acquired over a continuous 154 s period (14 s time resolution) with a sample voltage of −0.2 V and a tunneling current of 850 pA under 0.3 bar of 1:1 H2:CH3SH at 510 K. The blue arrows indicate the location of the brightest features on the Co edges. In e, blue circles indicate the expected positions of the protrusions and green circles mark the measured positions of edge protrusions. STM images acquired in the ‘irregular’ tip-state are marked with a * in the lower left corner.The measured height profiles along the Co edge for two successive frames in Fig. 5k, h are shown in Fig. 6a, b, respectively. The height profiles in Fig. 6 show that there are three distinct edge features. First one that consists of a single bright feature that is imaged brighter than the basal plane with a measured height of 3–5 Å (marked I); one that has a measured height of 2–2.5 Å and has about the same contrast as the basal plane (marked II); and one that consists of I and II occurring on immediately adjacent sites (marked as III). In all CoMoS slabs imaged, feature III was always observed to form at least one atomic site away from features I and II. Additionally, features I and II were more commonly observed than III. We note that feature III appears to be very similar to the one observed when reduced Co edges of the CoMoS slabs were exposed to CH3SH (see Fig. 2d), wherein we attributed the formation of the paired bright and dark features to dissociatively chemisorbed CH3SH on the Co edge.Fig. 6: Analysis of the slab edges and reaction gas stream.a, b Zoomed-in STM images of the Co edge from Fig. 5 with the measured heights along the blue and red lines drawn on the time-variant edge structures of the Co edge. The three types of recurring edge structures are marked as I, II and III in the height lines and the STM images. c Distribution of structures I, II, and III on Co edges at various H2:CH3SH ratios for CoMoS slabs of different sizes. The slab sizes are shown in the legend. d Co edge occupancy under various H2:CH3SH ratios for CoMoS slabs of different sizes. The slab sizes are shown in the legend. e Species detected in the RGA on the exit stream for the reaction condition H2:CH3SH = 9:1 at room temperature and 510 K. The error bars in c and d indicate the standard deviation. The statistics in c and d are based on 2-3 min of in situ observation of 50–55 Co edges (18–20 slabs) each for H2:CH3SH up to 6:1 and 20 Co edges (6 slabs) for H2:CH3SH = 9:1. STM images acquired in the ‘irregular’ tip-state are marked with a * in the lower left corner.In order to gain more insight into the distribution of structures I, II, and III, statistical analysis of the Co-substituted S edges of the CoMoS slabs is performed. For this purpose, the HDS experiments were also repeated for intermediate pressure ratios (H2:CH3SH) of 3:1 and 6:1 on identically prepared model catalysts, to gain additional insights into the effects of hydrogen partial pressure, while keeping the partial pressure of CH3SH the same for all the experiments. STM images of several CoMoS slabs of sizes between 3–7 nm were analyzed for the reaction gas mixtures containing H2:CH3SH ratios of 1:1, 3:1, 6:1 and 9:1. Figure 6c shows the distribution of structures I, II, and III among all of the occupied sites on the Co edge. The total occupancy of the Co edges by I, II, and III is shown in Fig. 6d for slabs of various sizes. Generally, the slab sizes are found to not have a significant effect on the occupancies or the distribution of structures I, II, and III. Figure 6c shows that upon increasing the hydrogen partial pressure, the occurrence of structures II and III decreases. Additionally, there is also a corresponding slight decrease in the number of occupied sites on the Co edges suggesting that hydrogen likely plays a role in the disappearance of II and III.To confirm the HDS activity in our experiments, the experimental detection of product gases is also necessary. The downstream mass spectrometer was used to analyze the gases leaving the reactor. Given that only the STM tip and the sample are exposed to the reaction gases, we expect that any observable activity can be attributed to the model catalyst surface as no other parts of the system have any catalytic activity. Figure 6e shows the product analysis for a H2:CH3SH ratio of 9:1 measured at room temperature where we do not expect significant HDS activity and at 510 K where the time-variant Co edge structures were observed in the STM images. The CH4 (m/z = 16) and H2S (m/z = 34) detected at room temperature are attributed to the trace contaminants present in the H2 and CH3SH gas bottles. If there is no catalytic activity, we expect the background CH4 and H2S signals to remain unchanged irrespective of the model catalyst temperature. In comparison to this baseline, the product streams at 510 K show the formation of additional CH4 and H2S by the same order of magnitude. Given the mild HDS conditions used, small amount of catalyst present, and the design of the ReactorSTM, we expect low yields of methane if any, such as in Fig. 6e.We note that the STM images presented in this work show several tip-state fluctuations which may result in the CoMoS slabs being imaged with a darker contrast. Prior literature works on CoMoS slabs involving vacuum and low-pressure studies do not report such electronic effects. While scanning under the reaction gases (especially under H2), the STM tip undergoes spontaneous tip-state changes every few 101 s because of the gas-tip interactions and likely chemical reactions on the tip, making the process of maintaining atomic resolution, scanning stability, and tip stability at the same time very challenging. In one of the less likely of these states, the CoMoS nanoclusters are imaged darker. This spontaneous switch in the tip states is visible between Fig. 5d, e and 5h, i. When imaged in this state, all of the edge states including brim contrasts, appear lower by about 1 Å but their overall positions and registry with respect to the basal plane atomic rows remain unchanged. This effect is also visible in the height profiles of the Co edges in Fig. 6a, b. For e.g., structure I in Fig. 6a (slabs imaged bright) and Fig. 6b (slab imaged dark) appear to be 4.2 Å and 3 Å high respectively, without any significant registry shift in the features. Additionally, the contrast difference between the brim sites on Mo and Co edges are not significantly influenced by this effect. For the sake of consistency in this work and ease of comparison with prior works, we consider the STM images obtained with the tip states where the basal planes and edges are imaged bright, whenever possible, for the edge identification.DFT modeling of the reaction pathwayIn order to better understand the time-varying edge structures on the Co edge of the CoMoS slabs, we make use of DFT calculations to construct a reaction pathway based on equilibrium structures and transition states. We particularly focus on modeling of the Co edge to understand the dynamics of this edge under HDS reaction gases. Furthermore, the Mo edges have already been modeled in detail in the work of Mom et al. under similar conditions18. Given that the structural changes are observed in a gas mixture containing H2 and CH3SH, both the hydrogen activation and C-S bond breaking steps are critical to understand the edge structure. In the earlier work of Mom18, it was shown that hydrogen can be activated through the interaction with a 50% S Mo-terminated edge of an MoS2 slab because the Mo atoms in such an edge behave like a noble metal atom towards hydrogen dissociation. We first consider the question whether the Co-substituted edge may also behave similarly. As a 50%S 100%H Co edge is experimentally observed in our experiment upon exposure to 1 bar H2, we consider a 50% S Co-substituted edge in our model as the basis of the calculations. A possible reaction pathway is modeled by which a molecule of hydrogen may interact with the edge Co, dissociate, and hop to two adjacent sulfur sites, as shown in Fig. 7a. For the complete energetics, we refer to SI, S8. The activation energies and the reaction steps are calculated and plotted in Fig. 7a, structure 1 to 8. The activation energy for the overall reaction, considered as the difference between the states (2) (di-hydrogen pre-adsorption) and (5) (dissociated H2 transition state), is 0.69 eV/atom. This low activation barrier shows that hydrogen dissociation and activation for hydrogenation readily occur on the Co-substituted S edge as well. The hydrogen activation calculations for the Mo-terminated edge of the 1D periodic model slab are shown in the SI, Fig. S5.Fig. 7: DFT modeling of H2 adsorption.a DFT-modeled reaction pathway for the H2 dissociation on a 50% S Co edge; b DFT modeled reaction pathway for atomic H hopping on a 50% S Co edge. Blue atoms are those of Co, lavender atoms are those of Mo, yellow are those of S, and white-grey are those of H. The red arrows indicate the hopping of the H atom.The ease of hydrogen transport on the Co edge with 50% S saturation is also considered in Fig. 7b. An initial state where two hydrogen atoms are adsorbed to two different S atoms, four sites apart is considered and the pathway to the two H’s being five sites apart is computed. The reaction pathway is shown in Fig. 7b starting with structure (1) (two H atoms, 4 edge sites apart), evolving into structure (5) (two H atoms, 5 edge sites apart). The overall activation energy for the reaction, taken as a difference between the states (1) (two H atoms, 4 edge sites apart) and (4) (two H atoms, transition state to 5 edge sites apart), is found to be 0.65 eV/ atom. Similarly, for the Mo edge, an activation energy of 0.43 eV/atom is found for the diffusion of an adsorbed H atom, 5 sites away (see SI, Fig. S5). While these numbers depend on the consideration of the initial state including the number of H atoms and their locations on the S atoms for the respective edges, the low activation barriers for H diffusion show that, under the HDS reaction conditions in which the STM images are acquired, the H atoms are thermally averaged over both Mo- and Co-terminated edges and one may safely assume that an H atom is always present at the site adjacent to which the C-S bond breaking may potentially occur.To get a better picture of the CH3SH interaction with the Co edge, the chemisorption of a single CH3SH molecule on a 50% S Co-substituted edge is considered. For the complete energetics, we refer to SI, S8. Based on the work of Šarić and coworkers31, the dissociative chemisorption of CH3SH on a 50% S Co-edge is unfavorable, very much in disagreement with our experimental findings from the STM where we have clearly seen that the activated Co edge reacts with CH3SH (Fig. 2d). Our calculations also find that the energy of the Co-substituted S edge containing a dissociatively adsorbed CH3SH molecule is 0.31 eV/atom lower than the sum of the energies of the edge and the non-dissociated molecule, strongly suggesting that the chemisorption of CH3SH is a thermodynamically favorable process on the Co edge. Based on this information, the potential self-desulfurization of the adsorbed CH3SH is considered, given that the chemisorption of CH3SH without any H2 can still result in the formation of an active H on an adjacent site due to the S-H bond breaking. An example of this is shown in structure (4) (CH3SH adsorbed, H transferred) of Fig. 8a, the formation of which is very favorable (state (2) (CH3SH adsorbed) to state (4) (CH3SH adsorbed, H transferred)). We consider the possibility that the H atom in state (4) (CH3SH adsorbed, H transferred) reacts with the thiolate by an H transfer. Overall, we find that the process has a very high activation barrier of 2.00 eV/atom, indicating that the CH3SH self-desulfurization is a very unfavorable process. This is very much in agreement with our experimental findings where we find that the Co-substituted edge is saturated with dissociatively adsorbed CH3SH entities at 510 K, but no further reaction occurs on this edge.Fig. 8: DFT modeling of the HDS reaction on the Co edge.a DFT-modeled reaction pathway for self-desulfurization of chemisorbed CH3SH on a 50% S Co edge; b DFT-modeled reaction pathway for hydrodesulfurization of chemisorbed CH3SH on a 50% S Co edge. The grey, red and green pathways correspond to direct desulfurization, hydrogen migration, and ‘methyl transfer’ pathways for methane formation. The key steps of the ‘methyl transfer’ pathway are enlarged in the insets for the ease of viewing; and c DFT-modeled reaction pathway for hydrodesulfurization of chemisorbed CH3SH via thiol transfer on a 50% S Co edge. This reaction pathway is feasible when edge H occupancy (and hence, the H partial pressure) is lower leading to slow H2S removal. In all of the structures: Blue atoms are those of Co, lavender atoms are those of Mo, yellow atoms are those of S, brown atoms are of C and white-grey are those of H. The S atoms associated with adsorbed CH3SH molecule are marked dark yellow. The red and blue arrows indicate the hopping of the H- and CH3-, respectively.We consider the additional presence of hydrogen in the system which is readily expected to form edge SH groups. CH3SH adsorption on such an edge would result in a structure similar to state (4) of Fig. 8a, but with the addition of edge H atoms, as shown in structure (1) (CH3SH adsorbed, H transferred) in Fig. 8b. On such an edge structure, the removal of H2S by the transfer of the new H on the adjacent site may also occur since the self-desulfurization pathway was found to be unfavorable. With an activation energy of 1.08 eV/atom and a reaction energy of 0.66 eV/atom, the proton next to the reactive site can be transferred to SH to form H2S (state (1) to (3) (CH3S adsorbed, H2S formation), Fig. 8b). With an energy cost of 0.45 eV/atom, this molecule can desorb leading to state (4) (CH3S adsorbed) in Fig. 8b. A pathway in which H moves first to Co and from there to SH to form H2S is not likely, given that the target Co is already coordinated to three S atoms. Nevertheless, with an activation energy of 1.08 eV/atom, H2S formation is feasible at the reaction conditions used with low turn-over frequencies of ~10−2, similar to the turn-over frequencies of the S edges seen in our experiments. This finding also agrees well with the detection of H2S formation in Fig. 6e at 510 K. This step leaves a CH3S- on the active edge site. Given that the H hopping and H2 dissociation on the Co edge is a very fast process, we expect that the nearby sites are rapidly populated by hydrogen atoms without any significant barriers resulting in structure (5) of Fig. 8b. An effect of having a neighboring –SH and –[SSH] group is that the –CH3 points away from the edge (see inset, Fig. 8c). The directional ordering of the adsorbed -SCH3 units at high edge coverage such as that seen in Fig. 2c is attributed to this nearest-neighbor interaction between the adsorbed species. DFT calculations were not carried out to further optimize this structure as it is a special case that occurs when CH3SH is abundant in the system, a condition not typically encountered during HDS reactions. Nevertheless, from this point, the HDS reaction may occur via three hypothetical pathways which are individually considered in this work.First, a direct transfer of the proton from the adjacent -SH, could result in C-S bond breaking and subsequent methane formation. We consider this pathway and find that the activation barriers for this step are too high to be of any catalytic relevance, which is in agreement with the findings of Šarić and coworkers31. This reaction pathway is called ‘direct desulfurization’ denoted by grey arrows in Fig. 8b. We consider a second hypothesis wherein the Co atom also takes part in the HDS reaction. This could happen in two distinct ways. One, where the adjacent H hops to the Co atom to form a hydride, which then hydrogenates the CH3S- to form methane and two, where the methyl transfer occurs to the Co, followed by a H- hop and hydrogenation on the Co atom to form methane. Typically, addition of Co into the R-S bond is known in general chemistry experiments when either the C-S bond is polarized or the organic group has either conjugation or aromaticity, such as with thiophenes45. Studies in the past have shown that the charge transfer from Co to CH3S- can induce additional polarity in the, otherwise weakly, polar C-S bond46. Additionally, CH3SH adsorption experiments on Co-covered Mo low-index surfaces have reported the formation of methyl radicals and methane on the surface at elevated temperatures, suggesting methyl group transfer to the metal atom47. We use these previous findings as an inspiration to investigate this alternative methyl-hopping pathway via the Co site for desulfurization of CH3SH. Furthermore, these two possibilities have not been considered by Šarić and coworkers31, nor, to our knowledge, by any other theoretical works on the Co-substituted S edge. The hydride formation pathway from structure (5), Fig. 8b is shown in red arrows and is called ‘hydrogen migration’. We find that the activation energy for transferring H from S to Co is rather low (as we might expect based on the previous sections), but the activation energy to create CH4 by hydrogenation is again very high (1.71 eV/atom). The “methyl transfer” pathway is shown in green arrows from structure (5) in Fig. 8b. We investigated a two-step pathway in which the C-S bond is broken first by metal insertion, followed by a reaction between the proton and the CH3 group on Co. We find that the Co-CH3 intermediate (state (9) (CH3S adsorbed, CH3 transferred to Co), Fig. 8b) is surprisingly stable. The subsequent hydrogenation of Co-CH3 by hydrogen transfer can occur with a low activation barrier of 0.73 eV/atom. Based on these possibilities, it is clear that the easiest route to HDS is via the Co-CH3 intermediate compared to the other two possible pathways considered. It is quite remarkable that, while S-CH3 and S-H are both formally positive and might therefore be difficult to bring together, Co-H is formally negatively charged and should react much easier with S-CH3, but it does not. In order for CH3 to react with H, the C-S bond has to be broken first and based on our reaction models, the Co atom has to participate and break the C-S bond. This reaction pathway highlights the importance of the C-S bond breaking step for the HDS reaction to occur in aliphatic thiols. For aromatic thiols like thiophenes, one or more preceding hydrogenation steps can make the C-S bond more accessible for the ‘methyl transfer’ to occur. Our models for the HDS of a hydrogenated thiophene molecule show that ‘methyl transfer’ is still the most favorable reaction pathway (see SI, S7). The additional pre-hydrogenation steps may provide additional activation barriers for the overall HDS reaction depending on the type of substitution and the accessibility of the alpha carbon atom.A similar possibility is also considered for the Mo-terminated edge as this was not investigated in the work of Mom et al.18. The results of our modeling show that Mo can also add across the C-S bond in a similar fashion with comparable activation energies. The details of the model for the Mo edge are presented in the SI, Fig. S5. While the Mo edge can also perform C-S bond breaking via Mo addition, we also find that the CH3SH adsorption on the Mo edge is weaker than on the Co edge, which decreases the lifetime of CH3S- on the Mo edge. Thus, there is a lower probability for this reaction to occur on a pristine Mo edge, which is in agreement with the earlier work of Mom et al.18.We note that for the ‘methyl transfer’ reaction pathway to occur, the initial step of the H2S removal remains essential in structure (5) (CH3S adsorbed, active H nearby) of Fig. 8b. This, however, need not be true always. It may also be possible that the -SH simply migrates to another adjacent site so that the H2S removal occurs as an independent side reaction. In order to investigate this possibility, we also calculate an alternative reaction pathway wherein the desulfurization is preceded by an SH unit hopping along the edge, as shown in Fig. 8c, state (1) to (4) (CH3SH adsorbed, alpha site blocked by SH transfer). We note that upon adsorption of CH3SH such as in structure (1) of Fig. 8b, ‘methyl transfer’ is not possible as the adjacent Co atoms of the CH3S unit are triply coordinated. Thus, the SH hopping entails breaking of one Co-S bond at a site. We find that the most favorable pathway for this hopping is via transition state (2) (CH3SH adsorbed, transition state SH transfer) in Fig. 8c, with an activation energy of 0.87 eV/atom and a reaction energy of 0.65 eV/atom. Structure (3) in Fig. 8c contains an [SSH]- ion. This structure demonstrates the immense flexibility of the Co edge with rearranging the sulfur atoms, and thus transfers or accepts electrons readily to any adsorbed species. With a medium activation barrier of 0.54 eV/atom, structure (3) may convert to (5) (CH3S adsorbed, active H nearby, alpha site blocked by SH transfer) via a transition state (4), where the S-S bond is broken to form single S and SH. It is also very likely that the SH group hops further away from the CH3S- to minimize steric hindrance but this is not considered in our model. Nevertheless, we consider structure (5) of Fig. 8c as the starting point of the desulfurization step in this reaction pathway. Having established that the hydrogen activation occurs readily on the Co edge, we expect that the neighboring sulfur atoms to the CH3S moiety are hydrogenated. We note that, if the SH group hops away further, then this reaction pathway is identical to the ‘methyl transfer’ pathway (green) considered in Fig. 8b. Having fixed the SH group at the adjacent active site, we model a situation where the SH migration may be temporarily blocked on one side, as a special case. This type of a situation can be easily achieved by using a lower H2:CH3SH ratio and lower H2 partial pressure, such as in our experiments in Fig. 5, where a H2:CH3SH ratio of 1:1 at a pressure of 0.3 bar was used instead of 9:1 at 1 bar, in Fig. 4. This would in turn increase the edge occupancy of the CH3S- moieties which can block -SH hopping until they are desulfurized. In this special case, ‘methyl transfer’ into the C-S bond is found to occur with a higher activation barrier of 1.08 eV/atom and a reaction energy of 0.86 eV/atom through a transition state (6), into the final structure (7) (CH4 formation) in Fig. 8c. We note that structure (6) is a ‘late transition state’ of (7), having only a slightly higher energy. The methyl group on Co subsequently reacts with the proton next to it with an activation energy of 0.83 eV/atom through transition state (8) and a reaction energy of −1.01 eV/atom into the final state (9), where a methane molecule is formed. The two activation energies are somewhat higher than the corresponding ones computed with prior H2S removal as in the green pathway of Fig. 8b and we also note that the H2S removal for active site regeneration itself has an activation barrier of 1.08 eV/atom. Thus, it seems that these two reaction pathways are equally likely and are competing especially under mild HDS conditions. It may be that one is favored over the other, depending on the edge H saturation, which in turn depends on the relative hydrogen partial pressure in the system. Overall, the two favorable reaction pathways for HDS indicate that having higher H2 partial pressures favors the path with lower activation barriers for the C-S bond breaking. The ability of the Co edge to increase the lifetime of the adsorbed aliphatic thiol species enough for their HDS reaction to occur, can be one of the reasons for the enhanced selectivity of the MoS2 catalyst towards C-S bond breaking, when a Co promoter is used in the industrial catalysts.In general, all the adsorption steps discussed in the reaction pathways so far for the Co edge, are considered for single sulfur vacancy sites on the edge corresponding to a 50% S saturation. In the study by Grønborg et al.17, they explore the stability of edges with varying degrees of sulfur vacancies. For the conditions utilized in our experiments, we do not anticipate the formation of double sulfur vacancies at thermodynamic equilibrium. Nevertheless, it is worth noting that the adsorption of CH3SH may alter this outcome. In our previous work (Mom et al.).18, we observed that CH3SH adsorption leads to the creation of additional sulfur vacancies on a Mo edge. While we have not explicitly investigated the possibility of structures with lower sulfur vacancies for the Co edges in this study, as we have not experimentally observed any such structures, it may be worthwhile to explore this reaction pathway in future investigations. It is conceivable that at very high partial pressures of H2, such as those encountered in industrial settings, double sulfur vacancies may occur more frequently on the Co edge during the HDS reaction. This could be an effect of bridging another 2 orders of magnitude of the pressure gap from our experiments.The models presented in our work describe the simplest mechanism involving C-S bond scission with assistance of single Co sites. This is done to demonstrate the feasibility of this pathway in comparison to C-S bond breaking by hydrogen transfer, for example. There is no evidence in our work that suggests that multiple Co metal atoms are involved in this reaction. Furthermore, given the low hydrogen pressures used, the S edges have many different structures. Therefore, it is highly unlikely that a process involving multiple metal atom sites occurs under the reaction conditions used in our work. Multiple-site reaction pathways may be more feasible at higher hydrogen pressures. Additionally, one of the goals of the modeling carried out in this work is to investigate an alternative reaction pathway for the HDS of CH3SH that works without an Au support, as the typical DDS pathway is not feasible on the S edge. This general model should then be adapted further to specific systems involving complex model industrial supports in future studies. Therefore, the effect of the Au support on the structures was not considered. Geometry optimization showed that the effect of the Au slab on the intermediate structures was minor, mostly in terms of edge sulfur atoms interacting with the support.Having considered many reaction pathways, we relate these structures to the three types of time-varying edge structures observed under HDS conditions (Fig. 6a, b). Since the STM acquires images every tens of seconds, we expect that the structures observed are most likely those which are succeeded by a large activation barrier, and hence the slowest steps of the feasible reaction pathways in Fig. 8b, c are considered for simulating STM images. Consequently, we select structure (5) of Fig. 8b, structure (5) of Fig. 8c, and structure (9) of Fig. 8c, (9) being the final state for the HDS reaction. We simulate their STM images after relaxing these structures on a gold slab as detailed in the theoretical methods. The simulated STM images are shown in Fig. 9a–c. We find that structure (5) of Fig. 8b, that consists of a dissociatively-chemisorbed CH3SH along with an H atom on a nearby S site (Fig. 9a), forms a very bright and diffuse feature, with a brighter contrast than the basal plane S atoms, and has a very good match with experimentally observed feature (I) of the time-variant edge structures in Fig. 6a, b. Structure (5) of Fig. 8c, which involves a thiol transfer, forms two bright lobes on adjacent sites on the Co edge, and matches very well with feature III in Fig. 6b. The final state (9) of Fig. 8c after HDS leaves behind a single S and SH on the edge and their combined contribution to the LDOS causes a relatively less bright lobe on the Co edge, as in Fig. 9c, but matches very well with feature II of Fig. 6a, b. Thus, we identify structures I and III, as those related to the pre-HDS states, namely after CH3SH adsorption with and without thiol transfer, respectively. Structure II is an indication of the completion of the HDS reaction, the disappearance of which marks the end of the HDS catalytic cycle through the regeneration of the edge site. Observation of the thiol transferred variant III is not surprising as the time-variant structures of Fig. 6 were obtained under mild HDS conditions, which can favor two competing pathways as discussed earlier. We note that despite the significantly lower barrier for H2S removal (1.08 eV, as shown in Fig. 8b), we still observe a number of post-hydrodesulfurization (HDS) structures on the Co edges. This phenomenon is attributed to the rather crowded nature of the edges under the relatively mild HDS conditions employed. The statistical analysis presented in Fig. 6c, d shows a lower occurrence of the post-HDS state (structure II) with increasing partial pressure of hydrogen to 0.9 bar, indicating the role of H2 in its removal. Overall, we find that the structures from our reaction model result in simulated STM images that corroborate nicely with our experimentally observed transient edge structures on the Co edges of CoMoS slabs during the HDS of CH3SH. If the partial pressure of H2 is raised to 99.9 bar as in many industrial systems, keeping the CH3SH partial pressure 0.1 bar, we expect the Co edge to have a slightly different thermodynamically stable edge structure with lower sulfur saturation than 50% S, likely with double sulfur vacancies. We expect these types of sites to be more reactive and hence, resulting in different barriers for the rate determining steps than in our models.Fig. 9: Understanding the HDS reaction cycle.a The DFT-relaxed atomic model and the corresponding STM image of structure (5), Fig. 8b; b The DFT-relaxed atomic model and the corresponding STM image of structure (5), Fig. 8c; and c The DFT-relaxed atomic model and the corresponding STM image of structure (9), Fig. 8c. A two-atom thick gold slab (not shown in the atomic models) has been considered for simulating the STM images. In the atomic models, brown atoms are of C, grey of H, blue atoms are those of Co, lavender atoms are those of Mo, and yellow are those of S; d Rationalization of the reaction pathway for a complete HDS catalytic cycle of CH3SH on the Co edge of a CoMoS slab based on the reaction models in Fig. 8 and the experimentally observed structures in Figs. 4, 5 and 6.Overall, we rationalize the time-variant edge structures observed in Figs. 4 and 5 and their simulated STM images in Fig. 9a–c as a complete HDS catalytic cycle as shown in Fig. 9d. Starting with a hydrogenated 50% S Co edge, we have observed time-variant structures on the CoMoS slabs in situ that correspond to various stages of the HDS reaction. The formation of a single bright lobe on an active site of the Co edge denotes the dissociative chemisorption of CH3SH. The splitting of a single bright lobe into two adjacent bright lobes corresponds to the transfer of the thiol moiety to an adjacent site. The appearance of a smaller single bright lobe with a darker contrast denotes the formation of methane. The formation of this feature is possible when the hydrogen occupancy on the edge is lower allowing for the competing pathway with thiol transfer. The disappearance of this feature indicates the regeneration of the active site through H2S removal. In Fig. 9d, we also take into account that the HDS may occur without thiol migration, especially when higher H2 partial pressures are used. Additionally, Fig. 9d also suggests that higher partial pressures of hydrogen could favor the removal of H2S. This effect is observed in Fig. 6c, d, where increasing the hydrogen content in the gas leads to a decrease in the number of observed structures II and III on the Co edges of the CoMoS slabs and also a slight decrease in the overall occupancy of all the structures on the Co edges. Thus, we find an excellent agreement between our rationalization of the HDS reaction of CH3SH and our experimental observations using the operando ReactorSTM at near-industrial HDS conditions.For the first time in the history of HDS catalysis, we have directly probed the active Co and Mo edges of the Co-promoted MoS2 model catalyst at work and observed the HDS catalytic reaction of CH3SH. We used the state-of-art ReactorSTM set up combined with DFT modeling to resolve the atomic structures of the edges and the reaction pathways through which the HDS reaction can occur. Our reaction models show that the transfer of the methyl group to the coordinatively-unsaturated Co site is an essential step for the CH3SH HDS reaction to occur. We also find that the atomic structures corresponding to the rate determining steps in our modeled pathways match very well with the transient atomic structures observed in the STM images of the Co edges of the CoMoS slabs. Ultimately, we have rationalized that the transient atomic structures are an indication of the starting and completion of an HDS catalytic cycle. Our results also show that addition of a promoter like Co increases the lifetime of adsorbed thiol species sufficiently to enhance the probability of their HDS to occur, thus explaining the enhanced selectivity upon using Co as a promoter in industrial MoS2 catalysts. The results presented in this work demonstrate the importance of observing the active sites in situ under relevant reaction conditions and the efficacy of combining these results with DFT calculations to obtain experimentally-verified reaction models for complex heterogeneous catalytic reactions.
