Rational design of novel catalytic materials used to synthesize storable fuels via the CO hydrogenation reaction has recently received considerable attention. In this work, defect poor and defect rich 2D-MoS2 as well as 2D-MoS2 decorated with Mo clusters are employed as catalysts for the generation of acetylene (C2H2) via the CO hydrogenation reaction. Temperature programmed desorption is used to study the interaction of CO and H2 molecules with the MoS2 surface as well as the formation of reaction products. The experiments indicate the presence of four CO adsorption sites below room temperature and a competitive adsorption between the CO and H2 molecules. The investigations show that CO hydrogenation is not possible on defect poor MoS2 at low temperatures. However, on defect rich 2D-MoS2, small amounts of C2H2 are produced, which desorb from the surface at temperatures between 170 K and 250 K. A similar C2H2 signal is detected from defect poor 2D-MoS2 decorated with Mo clusters, which indicates that low coordinated Mo atoms on 2D-MoS2 are responsible for the formation of C2H2. Density functional theory investigations are performed to explore possible adsorption sites of CO and understand the formation mechanism of C2H2 on MoS2 and Mo7/MoS2. The theoretical investigation indicates a strong binding of C2H2 on the Mo sites of MoS2 preventing the direct desorption of C2H2 at low temperatures as observed experimentally. Instead, the theoretical results suggest that the experimental data are consistent with a mechanism in which CHO radical dimers lead to the formation of C2H2 that presents an exothermic desorption.

A new chapter in the field of nanotechnologies has been opened with the advance in the fabrication of large, highly crystalline two dimensional transition metal dichalcogenides (2D-MX2, where M is a transition metal such as Mo, Ta, W, Nb, Re, Co, and X is a chalcogen, such as S, Se, and Te), which are promising alternatives to the existing technologies due to their unique thickness dependent optical and electronic properties. Due to its prevalence in nature as molybdenite, molybdenum disulfide (MoS2) is one of the most investigated 2D-MX2. 2D-MoS2 is a prototypical 2D layered transition metal dichalcogenide material consisting of three atomic layers, one molybdenum and two sulfur atomic layers, arranged to a sandwich structure by covalent bonds in a sequence of S—Mo—S. A MoS2 monolayer is a semiconductor with a direct bandgap of about 1.8–1.9 eV1–3 that has typically a thickness of about 0.65 nm, while the monolayers are held together by weak van der Waals forces.

Due to its unique properties, 2D-MoS2 was successfully used in applications such as low-power transistors,4,5 phototransistors,6,7 complex electronic circuits,8,9 hydrogen evolution reaction, hydro sulfurization, optoelectronics, and energy storage applications.10 Recently, 2D-MoS2 has also attracted a lot of attention due to its catalytic properties.11–16 Highly-crystalline, pure 2D-MoS2 has been found to have poor intrinsic activity, considerably reducing its applications in heterogeneous catalysis. New strategies to stimulate the intrinsic activity of 2D-MoS2 have been developed that include self-structure changing (layer thickness, plane size, and defects of MoS2) and modulation of hybrid structure (heteroatom doping, metal loading, and heterostructure).

CO hydrogenation reaction on MoS2 has been studied toward the production of alcohol and hydrocarbon fuels. Anderson and Yu employed molecular orbital theory to investigate the conversion of CO and methane to ethane and ethanol on MoS2, and they studied the CO adsorption energy and the energy barriers for CH2 + CH2, CH2 + CH3, CH3 + CH3, CH3 + CO, and H + CH2CH3 coupling reactions.17 Lee et al. studied the effect of various potassium promoters on the activity and selectivity of the CO hydrogenation reactions.18 Their results revealed that unpromoted MoS2 facilitates the formation of almost exclusively C1—C5 hydrocarbons. A similar analysis performed by Zhang et al. revealed that the absence of promoters lead to no alcohols formation on MgO—SiO2-supported MoS2 catalyst, while CH4 was found to be the dominant product.19 Additional studies (by Dow and Union Carbide companies) found that alkali doped MoS2 shifts the selectivity of the CO hydrogenation reaction toward the formation of alcohols.20 

The density functional theory (DFT) method has been employed by Huang and Cho to study the CO hydrogenation reaction on MoS2 surface.21 Methane and CO2 have been found to be the main products of the CO hydrogenation reaction on pure MoS2. Recently, Fariduddin and co-workers performed DFT calculation and microkinetics simulation to explore the mechanism of the CO hydrogenation reaction at the bare and partially sulfurized Mo-edges of MoS2(100).22 Their results indicated that at high temperature (>800 K), methane is the major product, while only small amounts of ethane, ethylene, and methanol are formed.

Relevant to the CO hydrogenation reaction, MonSm clusters of various sizes such as Mo6S8,23 Mo7S14,24 Mo16S32,25 and Mo28S84,26 which resemble the type of edges and corners found on the MoS2 surface or coordinatively unsaturated MoS2 surface, have also been used as model systems in DFT calculations. CH4 and CO2 were predicted to be the dominant products on the unpromoted MoS2 surfaces. DFT calculations performed to investigate the catalytic activity of the basal plane of MoS2 modified with rows and patches of sulfur-vacancy found the energetics for alcohol synthesis from syngas to be more favorable for a MoS2 layer with a sulfur-vacancy patch.12 This analysis revealed that the alcohol synthesis from syngas is possible through the manipulation of the sulfur-vacancy geometry.

The role of the metal nanoparticles on MoS2 surfaces in the CO hydrogenation reaction has been investigated as well. Theoretical investigations revealed that Au particles act as active sites for the generation of methanol molecules via CO hydrogenation.27 Experiments conducted by Li et al. have shown similar alcohol selectivity when Ni particles are loaded on alkali promoted MoS2.28 An alternative pathway toward ethanol as a primary product with a preference toward low linear carbon chain alcohols (C1—C5) was observed. The addition of Co promoters on the alkali-doped MoS2 has also been observed to increase the selectivity of reaction toward the formation of ethanol and higher alcohols.29–31 

Despite of the large number of investigations, the mechanism of CO hydrogenation on MoS2 is still a subject of debate. Therefore, in this work, the initial steps involved in the CO hydrogenation reaction are investigated under ultrahigh vacuum (UHV) conditions on large area, crystalline 2D-MoS2 films as well as on 2D-MoS2 films modified by defects or decorated with Mo clusters. These investigations reveal that CO hydrogenation is not possible on defect poor MoS2 at low temperatures and pressures. However, on defect rich MoS2 films or MoS2 films decorated with Mo clusters, the CO hydrogenation reaction leads to the formation of C2H2 indicating that low coordinated Mo atoms on 2D-MoS2 are the active sites responsible for the catalytic reaction. Even though C2H2 can be formed on the Mo active sites, the desorbed C2H2 detected in the present experiment is found, based on DFT calculations, to originate from CHO radical dimers, which decompose to form C2H2.

For these investigations, an experimental apparatus that consists of a UHV surface science chamber equipped with standard tools for surface preparation and investigation is employed. An Ar+ sputtering gun is used for substrate cleaning, a twin pocket electron beam evaporator, and three thermal evaporators are used to evaporate metals in order to synthesize thin films and grow metal particles on surfaces. A gas feedthrough connected to a specialized gas manifold in conjunction with leak valves is used to introduce high purity reactive gases such as CO, CO2, H2, and H2S in the UHV chamber. The gas partial pressure inside the UHV chamber is measured with an ion gauge detector calibrated for N2. The UHV chamber also hosts a low electron energy diffraction (LEED) device to investigate the structure of crystalline surfaces, and an Auger electron spectrometer (AES) to investigate the cleanliness of the substrates as well as the composition and thickness of the synthesized 2D materials. A quadrupole mass spectrometer (QMS) mounted into a differentially pumped enclosure is used for residual gas analysis as well as temperature programmed desorption (TPD) and temperature programmed reaction (TPR).

The substrate, consisting of a 10 × 10 × 0.5 mm3 Cu(111) single crystal is attached to a liquid nitrogen cryostat and it is mounted in the center of the UHV chamber. The surface sample-cryostat assembly can be vertically translated and 360° horizontally rotated by means of a mechanical manipulator to place the surface in front of each preparation and characterization tool. Furthermore, a xy-horizontal translation stage allows a movement of the assembly by ±12 mm from the midpoint. The crystal position can be reproduced to 0.02 mm and 0.5° accuracy.

A type K thermocouple wire attached on the back of the Cu(111) substrate is used to measure the temperature of the sample. The sample can be resistively heated from the lowest temperature of 100 K–1000 K.

Defect poor, highly crystalline 2D-MoS2 is synthesized on a Cu(111) substrate via a multi-step process based on physical vapor deposition using recipes available in the literature.32–34 LEED and AES are employed after every preparation step to characterize the crystalline structure of the surface as well as the composition and the thickness of the material. Initially, the Cu(111) substrate is cleaned via Ar+ sputtering for about 40 min (IAr=2μA,Ekin(Ar+)=600 eV). Subsequently, the substrate is annealed at 850 K for 5 min to reestablish the crystallinity of the surface.

Figure 1(a) (black curve) shows an Auger electron spectrum recorded from the Cu(111) after Ar+ sputtering and annealing. The spectrum displays a single peak at 60 eV, which corresponds to the CuMVV Auger transition. No peaks corresponding to carbon (273 eV) or oxygen (510 eV), characteristic to contaminated surfaces are detected, indicating that the substrate is clean.

FIG. 1.

(a) Auger electron spectra recorded from a clean Cu(111) substrate (black curve), a sulfurized Cu (blue curve), and 2.1 ML MoS2 film grown on Cu(111) (red curve), and (b) Auger electron intensity ratio of Mo to Cu peaks as a function of Mo evaporation time in H2S environment on a sulfurized Cu substrate. The break point in the AES ratio plot corresponds very closely to the initial appearance of the MoS2 second layer.

FIG. 1.

(a) Auger electron spectra recorded from a clean Cu(111) substrate (black curve), a sulfurized Cu (blue curve), and 2.1 ML MoS2 film grown on Cu(111) (red curve), and (b) Auger electron intensity ratio of Mo to Cu peaks as a function of Mo evaporation time in H2S environment on a sulfurized Cu substrate. The break point in the AES ratio plot corresponds very closely to the initial appearance of the MoS2 second layer.

Close modal

In the next step, the cleaned Cu(111) substrate is annealed for 15 min up to 850 K in an H2S atmosphere (PH2S = 5 × 10−7 Torr), leading to the surface sulfurization. An Auger electron spectrum recorded from the sulfurized Cu(111) surface [cf. blue curve in Fig. 1(a)] shows that the CuMVV Auger transition decreases while, a new transition specific for sulfur (SLVV) appears at 152 eV. The SLVV transition intensity does not change if the H2S partial pressure is increased above 5 × 10−7 Torr or if the sulfurization time is longer than 15 min, indicating that the sulfurized copper layer is saturated.

In the last step, 0.25 ML Mo is evaporated on the sulfurized Cu(111) at room temperature, followed by 10 min annealing at 850 K in 5 × 10−7 Torr H2S. To grow thicker MoS2 films, multiple cycles of 0.25 ML Mo evaporation in conjunction with annealing in H2S atmosphere are repeated. An Auger electron spectrum recorded from a 2.1 ML MoS2 on Cu(111) is displayed in Fig. 1(a), which displays a diminished CuMVV transition peak, an increase of the SLVV transition intensity, and new peaks at 186 eV and 221 eV corresponding to the MoMNN Auger transition.

Auger electron spectroscopy is also employed to calibrate the MoS2 deposition rate and to determine the MoS2 film thickness, as mentioned above. Figure 1(b) shows the Mo/Cu AES ratio as a function of the Mo evaporation time on the sulfurized Cu(111). During the Mo evaporation, the partial pressure of H2S is kept at 5 × 10−7 Torr. A gradual increase of the Mo evaporation time on the S—Cu(111) surface leads to a break-point in the intensity ratio of the Mo/Cu AES, which corresponds to the amount of Mo needed for the completion of the first MoS2 layer on a sulfurized Cu(111) surface. The resulting Mo evaporation rate is determined to be 0.055 ML/min.

Figure 2 displays LEED diffraction patterns recorded from a clean Cu(111) substrate, a sulfurized Cu(111), as well as from defect poor 0.4 ML and 2.1 ML MoS2 on Cu(111). The hexagonal LEED pattern typical for the 111 orientation is obtained in the case of the clean Cu substrate (red marked diffraction spots in Fig. 2). After annealing the clean Cu(111) in H2S atmosphere, the LEED pattern exhibit a 7×7 R19° diffraction pattern characteristic to a sulfurized Cu(111) overlayer.32 The evaporation of Mo in H2S atmosphere on the sulfurized Cu(111) overlayer leads to a new hexagonal diffraction pattern corresponding to crystalline MoS2 (green marked diffraction spots in Fig. 2). In Fig. 2, the ratio of the distance between the central diffraction spot (00) and any red marked spots, i.e., diffraction from Cu(111) [cf. RCu in Fig. 2(a)], and the distance between the central spot (00) and any green spots, i.e., diffraction from MoS2 [cf. RMoS2 in Fig. 2(c)] is about 0.8, indicating the epitaxial growth of (4 × 4) unit cells of MoS2 on (5 × 5) atoms of the Cu(111) substrate as previously reported by Sun et al.32Figure 2 indicates that the MoS2 film aligns with the crystallographic axes of the Cu(111) substrate and present the Moiré pattern due to the epitaxial growth [blue highlighted spots in Fig. 2(c)].

FIG. 2.

LEED images of (a) Cu(111), (b) √7 × √7 R19° Cu—S overlayer obtained via H2S sulfurization of Cu(111), (c) 0.4 ML MoS2 on the Cu—S overlayer, and (d) 2.1 ML MoS2 on Cu(111). On the right-hand side of each figure, circles are used to highlight specific diffraction spots. The red circles in (a)–(c) highlight the diffraction from the Cu(111), while the green circles in (c) and (d) highlight the diffraction from MoS2. The red and green arrows show the distance between the central diffraction spot (00) and diffraction spots of Cu and MoS2, respectively. The blue circles highlight the Moiré pattern due to the epitaxial growth of (4 × 4) unit cells of MoS2 on (5 × 5) atoms of the Cu(111) substrate (see text for details). The LEED images are recorded with an e-beam energy of 79 eV.

FIG. 2.

LEED images of (a) Cu(111), (b) √7 × √7 R19° Cu—S overlayer obtained via H2S sulfurization of Cu(111), (c) 0.4 ML MoS2 on the Cu—S overlayer, and (d) 2.1 ML MoS2 on Cu(111). On the right-hand side of each figure, circles are used to highlight specific diffraction spots. The red circles in (a)–(c) highlight the diffraction from the Cu(111), while the green circles in (c) and (d) highlight the diffraction from MoS2. The red and green arrows show the distance between the central diffraction spot (00) and diffraction spots of Cu and MoS2, respectively. The blue circles highlight the Moiré pattern due to the epitaxial growth of (4 × 4) unit cells of MoS2 on (5 × 5) atoms of the Cu(111) substrate (see text for details). The LEED images are recorded with an e-beam energy of 79 eV.

Close modal

At submonolayer coverages of MoS2 on Cu(111), both the diffraction patterns of sulfurized Cu(111) and MoS2 are visible in LEED as displayed in Fig. 2(c). The diffraction pattern of the sulfurized Cu(111) vanishes at the same MoS2 coverage at which the break point in Fig. 1(b) appears, which marks the completion of the first MoS2 layer on Cu(111). Identical Auger spectra and LEED images are recorded anywhere on the sample, indicating that the MoS2 films are homogeneous and have the same surface area as the Cu(111) substrate, i.e., 1 cm2.

To synthesize defect poor MoS2 films, it is imperative to keep the Mo evaporation rate low, i.e., 0.055 ML/min, while the H2S pressure has to be higher than 1 × 10−7 Torr. Therefore, the procedure described above leads to defect poor, highly crystalline MoS2 films. To synthesize on purpose defect rich MoS2 films with reproducible characteristics, which are expected to have enhanced catalytic properties, various procedure have been employed in this laboratory, such as (i) synthesis of a defect poor MoS2 followed by gentle Ar+ sputtering to produce sulfur vacancies, (ii) Mo evaporation at a high rate, or Mo evaporation at low H2S partial pressures. The first technique leads to highly defective films with a large number of sulfur vacancies, which become less defective through thermal annealing during the TPD measurements. Therefore, these defect rich MoS2 films produced via Ar+ sputtering cannot be used for more than a single TPD or TPR experiment. In this study, we employed the second procedure, which consists of Mo evaporation at a rate of 0.4 ML/min in 2 × 10−7 Torr H2S, to produce defect rich MoS2 films, which do not change their morphology during the annealing process. AES analysis of these films shows that the content of sulfur is 30% lower than in the case of defect poor films. A LEED image of a defect rich 2 ML MoS2 film synthesized on Cu(111) at a Mo evaporation rate of 0.4 ML/min in 2 × 10−7 Torr H2S is displayed in Fig. S1.

To investigate the interaction of CO molecules with the surface, as well as to study the mechanism of CO hydrogenation reaction on 2D-MoS2 film, TPD and TPR experiments are performed. For these experiments, the sample initially cooled down to 100 K is positioned in the middle of the UHV chamber, far from the gas feedthrough, to measure with high accuracy, the gas partial pressure in the vicinity of the sample. The gas doses are measured in units of Langmuir (1 L = 1 × 10−6 Torr · 1 s) using a LabVIEW software, which accurately reads and integrates the ion gauge partial pressure as a function of time. Subsequently, the QMS is turned on, and after a short break of about 2–4 min, in which unabsorbed gas-phase molecules are pumped down, the sample is positioned in front of the QMS skimmer entrance, with the surface perpendicular to the longitudinal axis of the spectrometer. The distance between the QMS skimmer entrance and the surface is below 2 mm to ensure that just the molecules desorbing from the surface are analyzed.

During the TPD measurement, the temperature of the sample is ramped-up uniformly by a proportional-integral-differential (PID) controller implemented into a LabView program. The same LabView program reads the surface temperature as well as accomplishes the mass spectrometer signal acquisition. The surface temperature is increased at a rate of 1 K/s until 750 K.

Plane-wave DFT calculations are carried out using the Vienna Ab initio Simulation Package (VASP) code.35,36 The Perdew-Burke-Ernzerhof (PBE) functional,37 within the generalized-gradient approximation is employed, together with the DFT-D338 correction to describe the exchange-correlation of electrons and projector augmented wave method (PAW) pseudopotentials39 for the interactions between the core and valence electrons. A cutoff energy of 500 eV for plane-wave expansion is used. As described above and shown in prior work,33,40 the Moiré unit cell of MoS2 on the Cu(111) surface consists of (4 × 4) MoS2 and (5 × 5) Cu(111). Thus, the basic simulation supercell, used in our first set of calculations, is built using a five-layer (5 × 5) Cu(111) slab, a (4 × 4) MoS2 adsorbed on one side of the Cu slab, and a vacuum of 20 Å to eliminate the artificial interaction along the normal direction as the result of periodic boundary condition. The Cu slab and MoS2 layer are constructed using their optimized lattice parameters, which are 3.57 Å and 3.16 Å, respectively. From the basic simulation supercell, systems with S vacancies and edges are constructed as described below. For the calculations of Mo cluster supported by MoS2, the calculation supercell consists of a 7 atom Mo cluster (Mo7) that is adsorbed on one side of a (6 × 6) MoS2. All structures are relaxed until all force components acting on each atom reach 0.01 eV/Å or lower, except for those of the bottom two Cu layers that are held fixed. The total energy for electronic iterations converged to 10−5 eV. The Brillouin zone with (3 × 3 × 1) zone-centered mesh and with one point at the zone center are sampled for the relaxation of these structures of the two set of calculations, respectively. Reaction barriers are calculated using the climbing-image nudged-elastic-band (CI-NEB).41,42

To study the CO hydrogenation reaction, first a set of experiments is performed to understand the interaction between the CO molecules and MoS2 surface. Figure 3 shows a series of TPD spectra recorded from various amounts of CO dosed on defect poor MoS2. The highest coverage is obtained for a dose of 0.6 L CO, which corresponds to MoS2 saturation with CO molecules at a surface temperature of 100 K. In this experiment, a thickness of 2 ML is chosen for the MoS2 film, in order to avoid any contributions of CO desorption from the Cu substrate that can appear at lower coverages.

FIG. 3.

TPD spectra of various amounts of CO dosed onto a defect poor 2.0 ML MoS2/Cu(111). The thin curves represent the measured data, while the thick curves are smooths of the measured data obtained using 10 points adjacent averaging function. All spectra were obtained with a heating rate of 1 K/s.

FIG. 3.

TPD spectra of various amounts of CO dosed onto a defect poor 2.0 ML MoS2/Cu(111). The thin curves represent the measured data, while the thick curves are smooths of the measured data obtained using 10 points adjacent averaging function. All spectra were obtained with a heating rate of 1 K/s.

Close modal

At the lowest CO dose of 0.003 L, four desorption features are observed at 110 K (α), 140 K (β), 190 K (γ), and 280 K (δ), respectively. As the CO dose is gradually increased, the intensity of those desorption features increases. The feature α at 110 K does not shift as the CO dose is increased, which reflects a first order desorption process. In contrast, the desorption feature β initially at 140 K shifts to lower temperatures, i.e., 135 K as the CO dose is increased to 0.6 L. The desorption feature γ and δ follow the same trend as feature β, shifting to lower temperatures as the CO coverage is increased from 0.007 L to 0.6 L, i.e., 200–150 K and from 290 K to 270 K, respectively. The shift of the β, γ, and δ features to lower temperatures as the molecular coverage is increased can be attributed to a repulsive desorption of the CO molecules from the MoS2 surface. The features α, β, and γ in Fig. 3 merge into a single peak structure at the highest molecular coverage obtained here, while δ stays as a distinct desorption feature. In addition, the CO desorption from defect poor MoS2/Cu(111) (cf. Fig. 3) does not occur sequentially, i.e., saturation of high temperature desorption peaks before the appearance of the low temperature ones. The simultaneous appearance of the low and high temperature desorption peaks is attributed to CO desorption from sites that are on different edges of the MoS2 domains. The adsorption and desorption of molecules from sites that are not on the same edge are independent.

TPD spectra of CO from the Cu(111) substrate are also recorded (not shown here) to ensure that the TPD spectra of CO from MoS2/Cu are not affected by CO desorption from the Cu(111) substrate. The CO desorption features from the bare Cu(111) substrate, which accurately resembles the TPD spectra of CO from a Cu(111) surface obtained by Kirstein et al.,43 have a different shape and desorption temperatures than the spectra obtained from MoS2/Cu(111), confirming that MoS2 is the only contributor to the observed spectra in Fig. 3.

Computational calculations are performed to explore possible adsorption sites of CO on defect poor MoS2 grown on the Cu(111) surface by studying the adsorption of CO at the edges and small S vacancy defects on the basal plane of MoS2. The edges of MoS2 on Cu(111) are modeled by an infinite strip of MoS2, whose supercell is constructed by double the size of Moiré unit cell of MoS2 on the Cu(111), aka (4 × 4) MoS2 and (5 × 5) Cu(111), along one direction and remove 4 Mo rows and 3 S rows, as shown in Fig. 4(a). In this model, the Mo edge, aka (1¯010) edge, and S edge, aka (101¯0) edge, are fully covered by S (100% S coverage), representing the S rich condition on the Cu(111) surface. The bare Mo edge, representing the Mo rich condition, is also modeled.

FIG. 4.

(a) Surface supercell (black parallelogram) used in modeling S and Mo edge of MoS2 on Cu(111). Configurations of CO binding on (b) S edge with 100% S coverage, on [(c) and (d)] Mo edge with 100% S coverage at two binding sites, and on (e) bare Mo edge. Binding energy is listed below each configuration. Pink, blue, yellow, gray, and red spheres represent Cu, Mo, S, C, and O atoms, respectively.

FIG. 4.

(a) Surface supercell (black parallelogram) used in modeling S and Mo edge of MoS2 on Cu(111). Configurations of CO binding on (b) S edge with 100% S coverage, on [(c) and (d)] Mo edge with 100% S coverage at two binding sites, and on (e) bare Mo edge. Binding energy is listed below each configuration. Pink, blue, yellow, gray, and red spheres represent Cu, Mo, S, C, and O atoms, respectively.

Close modal

Figure 4(b) shows the adsorption configuration of CO at the S edge with 100% S coverage. The binding energy of CO is found to be 43.07 kJ/mol. In this configuration, CO binds to the edge via C—Mo bond. On Mo edge with 100% S coverage, the CO molecule binds at S—S bridge in two configurations shown in Figs. 4(c) and 4(d) with binding energy of 110.21 kJ/mol and 111.66 kJ/mol, respectively. Although the calculated values of binding energy are higher than the corresponding temperatures measured in the TPD experiments, they agree well with the two major TPD peaks measured in this experiment, i.e., peak β and δ in Fig. 3. The other sites that correspond to peak α and γ in experimental TPD spectra could not be identified. However, they could be attributed to CO desorption from different structure of edges that were not considered in this work.

Figure 4(e) shows the adsorption configuration of CO on bare Mo edge. The binding energy of CO is found to be 214.59 kJ/mol. The corresponding TPD desorption peak would be around 800 K, which is out of the temperature range of the TPD spectra shown in Fig. 3. The adsorption of CO at sulfur vacancies of the basal plane of MoS2 is also considered. The binding energies of CO at these sulfur vacancies sites are extremely high (in the range from 150 kJ/mol to 210 kJ/mol for CO binding onto single, double, or triple S vacancies), as compared to the TPD data in Fig. 3, indicating that α, β, γ, and δ features cannot originate from these defect sites.

Additional calculations have been performed to study the interaction of CO with a 7-atom Mo cluster (Mo7) supported by MoS2. It is worth noticing that the adsorption energy of CO on Mo cluster is much stronger than that on the edge of MoS2. The calculated DFT adsorption energy of CO on Mo7 cluster on MoS2 is at least −219.58 kJ/mol (Fig. S2). This value would correspond to a TPD peak of about 900 K, which is a far higher temperature than those observed in the defect-poor case, suggesting that the peaks detected in TPD spectra in Fig. 3 are not originating from Mo cluster.

Previous theoretical investigations predicted that molecular hydrogen interaction with MoS2 leads to hydrogen atoms adsorbed mostly on Mo sites, as a result of dissociative adsorption of H2.12,21,44 In this study, no H or H2 desorption peaks have been observed in the 100–750 K surface temperature range, in the TPD measurements in which various amounts of H2 are dosed on the MoS2 surface. The weakest binding energy of atomic hydrogen on the Mo edges of MoS2 obtained from DFT calculations performed in this study is −265.56 kJ/mol, which would correspond to a TPD desorption peak around 1000 K.

Although no atomic or molecular hydrogen is observed in TPD, this experiment does not rule out that hydrogen can be adsorbed on the MoS2 surface, because as will be discussed in Sec. III C, a hydrogenated product is only observed when the surface is exposed to H2. Moreover, recent theoretical investigations of Fariduddin and co-workers predicted that CO would compete with H for the Mo sites at temperatures below 575 K during the CO hydrogenation reaction on MoS2.22 Therefore, to get insights into the hydrogen interaction with MoS2, TPD spectra of CO are recorded from a MoS2 surface pre-dosed with various amounts of H2 [cf. Fig. 5(a)]. As can be seen in Fig. 5(a), by increasing the pre-exposure of MoS2 to H2, the δ desorption feature, i.e., desorption peak between 250 K and 300 K does not shift, but decreases in intensity. A similar behavior is observed in the case of desorption features β and γ. The desorption feature α, below 120 K, is not influenced by the MoS2 pre-exposure to H2.

FIG. 5.

(a) TPD spectra of 0.6 L CO from a 2 ML MoS2/Cu(111) pre-exposed to various amounts of H2. (b) TPD spectra of 0.6 L CO from a 2 ML MoS2 film pre- and post-exposed to 500 L of H2. All spectra are obtained with a heating rate of 1 K/s.

FIG. 5.

(a) TPD spectra of 0.6 L CO from a 2 ML MoS2/Cu(111) pre-exposed to various amounts of H2. (b) TPD spectra of 0.6 L CO from a 2 ML MoS2 film pre- and post-exposed to 500 L of H2. All spectra are obtained with a heating rate of 1 K/s.

Close modal

Figure 5(b) compares TPD spectra of 0.6 L CO recorded from a 2 ML MoS2 surface pre- and post-dosed with 500 L of H2. This experiment indicates that the desorption feature β is only slightly affected, while γ and δ are apparently influenced if CO is dosed first and H2 later. The desorption feature α below 120 K is not influenced by the order in which gases are dosed on the surface [cf. Figs. 5(a) and 5(b)].

The TPD spectra shown in Fig. 5 clearly indicate that hydrogen is present on the MoS2 surface. Moreover, these spectra show a competitive adsorption between CO and H2 on the β, γ, and δ sites, which is in agreement with the theoretical predictions of Fariduddin and co-workers.22 

To study the CO hydrogenation reaction, the MoS2 surface was first exposed to 10 L H2 and subsequently to 0.6 L CO, which corresponds to the CO saturation limit of MoS2 at 100 K, as mentioned above. We chose to expose the sample to 10 L of H2, because at this dose the δ desorption feature of CO decreases considerably as can be seen by comparing the blue-color TPD spectra in Figs. 3 and 5(a), which clearly indicates that hydrogen is absorbed on the surface. No chemical species such as formyl radical (CHO), or formaldehyde (CH2O), which are the first intermediates that are expected to form during CO hydrogenation reaction, have been detected in TPR experiments on both defect poor and rich MoS2 films. Moreover, no methylidene radical (CH2), methyl radical (CH3), or methane (CH4) species have been detected either. A low intensity peak of C2H2 is detected between 160 K and 250 K only on defect rich MoS2 [cf. Figs. 6(a) and 6(b)]. An even lower intensity signal of CH is also detected (not shown here). Because the CH signal resembles the features of C2H2 signal, it is attributed to the fragmentation of C2H2 into the ionizer of the QMS used to record TPD and TPR spectra. No C2H2 signal is detected if either H2 or CO are not dossed on the MoS2 surface.

FIG. 6.

TPR spectra of C2H2 obtained by dosing 10 L H2 followed by 0.6 L CO on (a) defect poor 2 ML MoS2/Cu(111), (b) defect rich 2 ML MoS2/Cu(111) as well as on defect poor 2 ML MoS2/Cu(111) decorated with (c) 0.03 ML Mo and (d) 0.08 ML Mo. The spectra are obtained with a heating rate of 1 K/s.

FIG. 6.

TPR spectra of C2H2 obtained by dosing 10 L H2 followed by 0.6 L CO on (a) defect poor 2 ML MoS2/Cu(111), (b) defect rich 2 ML MoS2/Cu(111) as well as on defect poor 2 ML MoS2/Cu(111) decorated with (c) 0.03 ML Mo and (d) 0.08 ML Mo. The spectra are obtained with a heating rate of 1 K/s.

Close modal

The defect rich MoS2 films on which C2H2 is formed, is expected to have a large number of low coordinated molybdenum atoms, which presumably are the catalytic sites accountable for the CO hydrogenation reaction. To demonstrate that the low coordinated Mo atoms of MoS2 are the catalytic centers that lead to the formation of C2H2, different amounts of Mo, i.e., 0.03 ML and 0.08 ML, are evaporated on a defect poor MoS2. Recently, Batzill45 demonstrated that evaporation of Mo on a 2D-MoS2 surface leads to a MoS2 surface decorated with Mo clusters. Previous STM investigations also demonstrate that evaporation of very small amounts of metals below 0.04 ML on surfaces held at ≤100 K leads to surfaces decorated with mostly atoms because at this low temperature, the metal atoms are immobile even on defect free inert surfaces.46,47 Moreover, photoemission investigations confirm that the evaporation of small amounts of metals below 0.04 ML on surfaces held at 100 K results in the formation of very small clusters while evaporation of metals below 0.12 ML results in the formation of slightly larger clusters, which do not have a metallic character.48 Therefore, it is expected that in this experiment, evaporation of 0.03 ML Mo on MoS2 at 100 K leads to a MoS2 surface decorated mostly with Mo monomers, while the evaporation of 0.08 ML Mo leads to the formation of small Mo clusters composed of a few atoms. Since a very small amount of extra Mo was added to the surface, the LEED images of defect poor MoS2 films decorated with 0.03 ML and 0.08 ML Mo resemble the LEED image of defect poor 2.1 ML MoS2/Cu(111) displayed in Fig. 2(d) and hence are not shown separately.

Figures 6(c) and 6(d) display TPR spectra of C2H2 recorded from a defect poor 2 ML MoS2 film decorated with 0.03 ML and 0.08 ML Mo. The C2H2 spectrum recorded from 0.03 ML Mo on defect poor 2 ML MoS2 display a peak structure between 160 K and 250 K with a maximum at 205 K, which basically resembles the C2H2 spectrum obtained from defect rich MoS2 [cf. Fig. 6(b)]. The C2H2 spectrum recorded from 0.08 ML Mo/MoS2 also displays a peak that has a higher intensity then the signals obtained from defect rich MoS2 and 0.03 ML Mo/MoS2. This peak has the onset and maximum intensity at slightly lower temperatures than the signal recorded from defect rich MoS2 and 0.03 Mo/MoS2, i.e., 150 K, and 195 K, respectively. In addition, this signal presents a shoulder that extends up to 390 K. Furthermore, a small peak at temperatures below 150 K is detected in both C2H2 TPR spectra recorded from defect poor MoS2 decorated with Mo clusters. The C2H2 peak below 150 K as well as the signal detected from 0.08 ML Mo/MoS2, which extends to higher temperatures than the signal obtained from defect rich MoS2 and 0.03 ML Mo/MoS2 [cf. Figs. 6(b) and 6(c)], could be attributed to catalytic centers consisting of Mo clusters, composed of a few Mo atoms, which might have various shapes and sizes and hence various electronic and catalytic properties. The spectra displayed in Figs. 6(c) and 6(d) clearly demonstrates that C2H2 detected from defect rich MoS2 originates from low coordinated Mo atoms.

To get insights into the formation mechanism of C2H2, by-products such as H2O, O2, and CO2 that can be formed during the CO hydrogenation reaction are monitored via TPR. No H2O or O2 are detected on any surfaces investigated here. A negligible CO2 signal is detected from defect poor MoS2. However, if the defect poor MoS2 surface is decorated with 0.04 ML Mo, a CO2 signal is detected. The CO2 TPR signal from 0.04 ML Mo/MoS2 displays a peak structure with a maximum at 135 K followed by a shoulder that extends up to 370 K (cf. blue curve in Fig. 7). Interestingly, the CO2 TPR signal from 0.04 ML Mo/MoS2 is detected when only CO is dosed on the surface, without pre-dosing the surface with H2, which indicates that CO is the only precursor needed for the formation of CO2.

FIG. 7.

TPR spectra obtained by dosing 0.6 L CO on samples consisting of 0.04 ML and 0.08 ML Mo evaporated on a defect poor 2 ML MoS2/Cu(111) (blue and red curves). A TPD spectrum of CO2 dosed on 0.08 ML Mo evaporated on a defect poor 2 ML MoS2/Cu(111) is shown as a reference (filled gray curve). The CO2 dose was chosen to match the TPR desorption intensity of the red curve.

FIG. 7.

TPR spectra obtained by dosing 0.6 L CO on samples consisting of 0.04 ML and 0.08 ML Mo evaporated on a defect poor 2 ML MoS2/Cu(111) (blue and red curves). A TPD spectrum of CO2 dosed on 0.08 ML Mo evaporated on a defect poor 2 ML MoS2/Cu(111) is shown as a reference (filled gray curve). The CO2 dose was chosen to match the TPR desorption intensity of the red curve.

Close modal

When 0.08 ML Mo is evaporated on MoS2, the intensity of CO2 increases, the desorption maximum of the peak structure shifts to 140 K, while the overall intensity of the shoulder that extends to 370 K dramatically increases (cf. red curve in Fig. 7). The detection of CO2 when only CO is dosed on the MoS2 surface decorated with Mo clusters indicates that the CO molecule splits on the Mo clusters grown on MoS2 to form C and O atoms. Subsequently, a portion of the CO molecules that did not split can be oxidized to form CO2. As a reference, CO2 has also been dosed on a MoS2 surface decorated with 0.08 ML Mo (cf. gray filled curve in Fig. 7). The TPD spectrum of CO2 displays a peak structure, with a maximum at 125 K with a low intensity shoulder that extends up to 250 K, which clearly demonstrate the CO2 molecules produced via TPR, i.e., via CO splitting and oxidation have a different origin than the CO2 molecules released during TPD. In Fig. 7, the broad CO2 signal resulting from CO adsorption on 2D-MoS2 decorated with Mo clusters can be attributed to (i) a desorption-controlled feature, i.e., the sharp peak at low temperatures and (ii) a reaction-controlled feature, i.e., the shoulder that extends to high temperatures, consisting of CO2 molecules formed as the surface temperature is ramped-up.

To our knowledge, no comprehensive studies regarding CO splitting on mass selected Mo clusters have been published so far, that report how the Mo cluster size influence the CO activation and fragmentation. However, CO splitting has been reported on both Mo(100),49–55 and Mo(111)56,57 surfaces to occur at very low temperatures. Therefore, the CO splitting on small Mo clusters might be even more efficient than on Mo single crystals and most probably depends on the size of Mo clusters. The C atoms left after the CO splitting on Mo clusters could react with H atoms to form CH radicals. The formation of CH radicals can also be facilitated by the hydrogen assisted CO dissociation (HCO* → CH* + O*).

Theoretical investigations are performed to verify the possibility of C2H2 formation on Mo edges of MoS2 and on MoS2 supported Mo7 cluster (cf. Fig. 8). The calculations indicate that C2H2 binds very strongly to the bare Mo edge of MoS2 and the Mo7, with a binding energy of 238.85 kJ/mol and 339.97 kJ/mol, respectively. Because such high binding energies require a much higher desorption temperature (about 1000 K) than the one measured in TPR spectra shown in Fig. 6, the results suggest that while the formation of C2H2 on the systems is possible, the desorption of the molecule is not seen in the presented TPR spectra.

FIG. 8.

Reaction pathways of CHO* radical formation (a), CHO* dimer formation and subsequent C2H2 desorption (b). Green bars show transition states while blue bars represent intermediate states (initial and final states for each reaction). Cyan, yellow, gray, red, and green balls represent Mo, S, C, O, and H atoms, respectively. Eb and ΔE are the activation barrier and reaction energy, respectively.

FIG. 8.

Reaction pathways of CHO* radical formation (a), CHO* dimer formation and subsequent C2H2 desorption (b). Green bars show transition states while blue bars represent intermediate states (initial and final states for each reaction). Cyan, yellow, gray, red, and green balls represent Mo, S, C, O, and H atoms, respectively. Eb and ΔE are the activation barrier and reaction energy, respectively.

Close modal

Note that in this experiment, a high dose of H2 is used. Therefore, the atomic hydrogen occupies the majority of the exposed Mo sites. The dissociation of CO is thus limited due to lack of available sites. Instead, we expect the formation of CHO* radical group (through exothermic reaction CO* + H* → CHO*) to be more favorable.22 Our calculations on MoS2 supported Mo7 cluster show that the barrier for such formation is 77.83 kJ/mol. Subsequently, the formation of CHO dimers (OCH=CHO) is found to be exothermic, with a reaction enthalpy of −50.76 kJ/mol and an activation barrier of 12.83 kJ/mol. From the dimer, C2H2 exothermically desorbs leaving two O* on the Mo edge, which is an endothermic desorption with a reaction enthalpy of 48.09 kJ/mol and an activation barrier of 77.16 kJ/mol. While it is not possible to have a direct correspondence of the DFT values with exact reaction temperatures, in the absence of kinetic effects, the low reaction barriers suggest that this series of reactions (Fig. 8) occur at low temperatures. Thus, while we are not eliminating the possibility of C2H2 formation on Mo sites, we believe that the mechanism that leads to the detection of C2H2 in the present TPR experiments can be described by the following reactions:

CO* + H* → CHO* → OCH=CHO* (CHO dimer) → 2O* + C2H2(g), where “g” denotes gas phase species, while “*” denoted adsorbed species. As mention above, no molecular oxygen has been detected in the TPR experiments. Therefore, O* could contribute to the formation of CO2, which subsequently can be released from the surface (cf. Fig. 7).

In this work, TPD and TPR experiments are performed to investigate the initial steps involved in the CO hydrogenation reaction on defect poor and defect rich 2D-MoS2 as well as on defect poor 2D-MoS2 decorated with Mo clusters obtained by evaporating well defined amounts of Mo on the surface. The TPD experiments revealed four CO adsorption sites below room temperature. Moreover, a competitive adsorption behavior between CO and H2 was observed. The TPR experiments show that highly crystalline 2D-MoS2 has poor catalytic properties. However, defect rich 2D-MoS2 as well as defect poor 2D-MoS2 decorated with small Mo clusters facilitate the formation of acetylene through the CO hydrogenation reaction. Our DFT calculations suggest that the detected C2H2 in TPR originates from OCH=CHO (CHO dimer) adsorbed on Mo atoms or Mo7 clusters. Although C2H2 is the only hydrocarbon detected in this experiment performed under UHV conditions, we cannot exclude the formation of other hydrocarbon molecules that can be formed by increasing the gas pressure and thus the number of the collisions between the reagent molecules at the surface. The results presented in this work suggest that it would be worth investigating the CO hydrogenation reaction on mass selected Mo clusters supported on MoS2 as well as other substrates to understand how Mo cluster size and its support influence the catalytic process.

See the supplementary material for a LEED image recorded from a defect rich MoS2 layer grown on Cu(111) (Fig. S1) and the optimized geometries of CO adsorbed on Mo7 cluster on MoS2 (Fig. S2).

B.T.Y. and M.A.K.P. have contributed equally to this work.

M.E.V. gratefully acknowledges financial support provided by UCF through the “VPR Advancement of Early Career Researchers” program. T.J., D.L., and T.S.R. contributed to the DFT investigations, which was funded by the U.S. Department of Energy Grant No. DE-FG02-07ER15842. Computational resources were provided by the UCF Advanced Research Computing Center and the National Energy Research Scientific Computing Center. N.M., T.N., and C.E.J. acknowledge financial support provided by UCF Office of Undergraduate Research. T.N. gratefully acknowledges financial support provided by UCF through the “Burnett Research Scholars” program.

1.
K. F.
Mak
,
C.
Lee
,
J.
Hone
,
J.
Shan
, and
T. F.
Heinz
,
Phys. Rev. Lett.
105
,
136805
(
2010
).
2.
A.
Splendiani
,
L.
Sun
,
Y.
Zhang
,
T.
Li
,
J.
Kim
,
C.-Y.
Chim
,
G.
Galli
, and
F.
Wang
,
Nano Lett.
10
,
1271
(
2010
).
3.
R.
Ganatra
and
Q.
Zhang
,
ACS Nano
8
,
4074
(
2014
).
4.
B.
Radisavljevic
,
A.
Radenovic
,
J.
Brivio
,
V.
Giacometti
, and
A.
Kis
,
Nat. Nanotechnol.
6
,
147
(
2011
).
5.
Y.
Zhang
,
J.
Ye
,
Y.
Matsuhashi
, and
Y.
Iwasa
,
Nano Lett.
12
,
1136
(
2012
).
6.
Z.
Yin
,
H.
Li
,
H.
Li
,
L.
Jiang
,
Y.
Shi
,
Y.
Sun
,
G.
Lu
,
Q.
Zhang
,
X.
Chen
, and
H.
Zhang
,
ACS Nano
6
,
74
(
2012
).
7.
O.
Lopez-Sanchez
,
D.
Lembke
,
M.
Kayci
,
A.
Radenovic
, and
A.
Kis
,
Nat. Nanotechnol.
8
,
497
(
2013
).
8.
B.
Radisavljevic
,
M. B.
Whitwick
, and
A.
Kis
,
ACS Nano
5
,
9934
(
2011
).
9.
H.
Wang
,
L.
Yu
,
Y.-H.
Lee
,
Y.
Shi
,
A.
Hsu
,
M. L.
Chin
,
L.-J.
Li
,
M.
Dubey
,
J.
Kong
, and
T.
Palacios
,
Nano Lett.
12
,
4674
(
2012
).
10.
M.
Chhowalla
,
H. S.
Shin
,
G.
Eda
,
L.-J.
Li
,
K. P.
Loh
, and
H.
Zhang
,
Nat. Chem.
5
,
263
(
2013
).
11.
T. F.
Jaramillo
,
K. P.
Jørgensen
,
J.
Bonde
,
J. H.
Nielsen
,
S.
Horch
, and
I.
Chorkendorff
,
Science
317
,
100
(
2007
).
12.
D.
Le
,
T. B.
Rawal
, and
T. S.
Rahman
,
J. Phys. Chem. C
118
,
5346
(
2014
).
13.
J.
Mao
,
Y.
Wang
,
Z.
Zheng
, and
D.
Deng
,
Front. Phys.
13
,
138118
(
2018
).
14.
V. P.
Santos
,
B.
van der Linden
,
A.
Chojecki
,
G.
Budroni
,
S.
Corthals
,
H.
Shibata
,
G. R.
Meima
,
F.
Kapteijn
,
M.
Makkee
, and
J.
Gascon
,
ACS Catal.
3
,
1634
(
2013
).
15.
Y. Q.
Yang
,
C. T.
Tye
, and
K. J.
Smith
,
Catal. Commun.
9
,
1364
(
2008
).
16.
B.
Yoosuk
,
D.
Tumnantong
, and
P.
Prasassarakich
,
Chem. Eng. Sci.
79
,
1
(
2012
).
17.
A. B.
Anderson
and
J.
Yu
,
J. Catal.
119
,
135
(
1989
).
18.
J. S.
Lee
,
S.
Kim
,
K. H.
Lee
,
I.-S.
Nam
,
J. S.
Chung
,
Y. G.
Kim
, and
H. C.
Woo
,
Appl. Catal., A
110
,
11
(
1994
).
19.
J.
Zhang
,
Y.
Wang
, and
L.
Chang
,
Appl. Catal., A
126
,
L205
(
1995
).
20.
R.
Murchison
,
M.
Conway
,
R.
Steven
, and
G. J.
Quarderer
, in
Proceedings of Ninth International Congress on Catalysis
(
Chemical Institute of Canada
,
Ottawa, Calgary
,
1988
).
21.
M.
Huang
and
K.
Cho
,
J. Phys. Chem. C
113
,
5238
(
2009
).
22.
F.
Fariduddin
,
I. A. W.
Filot
,
B.
Zijlstra
, and
E. J. M.
Hensen
,
Chem. Eng. Sci.
198
,
166
(
2019
).
23.
P.
Liu
,
Y.
Choi
,
Y.
Yang
, and
M. G.
White
,
J. Phys. Chem. A
114
,
3888
(
2010
).
24.
M. L.
Neiman
,
J. A.
Spirko
, and
K.
Klier
, “
Modelling of syngas reactions and hydrogen generation over sulfide
,” Final Technical Progress Report,
Department of Chemistry, Lehigh University Bethlehem
,
2004
.
25.
T.
Zeng
,
X.-D.
Wen
,
G.-S.
Wu
,
Y.-W.
Li
, and
H.
Jiao
,
J. Phys. Chem. B
109
,
2846
(
2005
).
26.
T.
Zeng
,
X.-D.
Wen
,
Y.-W.
Li
, and
H.
Jiao
,
J. Phys. Chem. B
109
,
13704
(
2005
).
27.
T. B.
Rawal
,
D.
Le
, and
T. S.
Rahman
,
J. Phys.: Condens. Matter
29
,
415201
(
2017
).
28.
D.
Li
,
C.
Yang
,
W.
Li
,
Y.
Sun
, and
B.
Zhong
,
Top. Catal.
32
,
233
(
2005
).
29.
Z.
Li
,
Y.
Fu
,
J.
Bao
,
M.
Jiang
,
T.
Hu
,
T.
Liu
, and
Y.-n.
Xie
,
Appl. Catal., A
220
,
21
(
2001
).
30.
J.
Iranmahboob
and
D. O.
Hill
,
Catal. Lett.
78
,
49
(
2002
).
31.
J.
Iranmahboob
,
H.
Toghiani
, and
D. O.
Hill
,
Appl. Catal., A
247
,
207
(
2003
).
32.
D.
Sun
,
W.
Lu
,
D.
Le
,
Q.
Ma
,
M.
Aminpour
,
M.
Alcántara Ortigoza
,
S.
Bobek
,
J.
Mann
,
J.
Wyrick
,
T. S.
Rahman
, and
L.
Bartels
,
Angew. Chem., Int. Ed.
51
,
10284
(
2012
).
33.
D.
Kim
,
D.
Sun
,
W.
Lu
,
Z.
Cheng
,
Y.
Zhu
,
D.
Le
,
T. S.
Rahman
, and
L.
Bartels
,
Langmuir
27
,
11650
(
2011
).
34.
Y.
Lee
,
J.
Lee
,
H.
Bark
,
I.-K.
Oh
,
G. H.
Ryu
,
Z.
Lee
,
H.
Kim
,
J. H.
Cho
,
J.-H.
Ahn
, and
C.
Lee
,
Nanoscale
6
,
2821
(
2014
).
35.
G.
Kresse
and
J.
Furthmüller
,
Phys. Rev. B
54
,
11169
(
1996
).
36.
G.
Kresse
and
J.
Furthmüller
,
Comput. Mater. Sci.
6
,
15
(
1996
).
37.
J. P.
Perdew
,
K.
Burke
, and
M.
Ernzerhof
,
Phys. Rev. Lett.
77
,
3865
(
1996
).
38.
S.
Grimme
,
J.
Antony
,
S.
Ehrlich
, and
H.
Krieg
,
J. Chem. Phys.
132
,
154104
(
2010
).
39.
P. E.
Blöchl
,
Phys. Rev. B
50
,
17953
(
1994
).
40.
D.
Le
,
D. Z.
Sun
,
W. H.
Lu
,
L.
Bartels
, and
T. S.
Rahman
,
Phys. Rev. B
85
,
075429
(
2012
).
41.
G.
Henkelman
and
H.
Jonsson
,
J. Chem. Phys.
113
,
9978
(
2000
).
42.
G.
Henkelman
,
B. P.
Uberuaga
, and
H.
Jonsson
,
J. Chem. Phys.
113
,
9901
(
2000
).
43.
W.
Kirstein
,
B.
Krüger
, and
F.
Thieme
,
Surf. Sci.
176
,
505
(
1986
).
44.
S.
Cristol
,
J. F.
Paul
,
E.
Payen
,
D.
Bougeard
,
S.
Clémendot
, and
F.
Hutschka
,
J. Phys. Chem. B
106
,
5659
(
2002
).
45.
M.
Batzill
,
J. Phys.: Condens. Matter
30
,
493001
(
2018
).
46.
M.
Sterrer
,
T.
Risse
,
M.
Heyde
,
H.-P.
Rust
, and
H.-J.
Freund
,
Phys. Rev. Lett.
98
,
206103
(
2007
).
47.
M.
Yulikov
,
M.
Sterrer
,
T.
Risse
, and
H. J.
Freund
,
Surf. Sci.
603
,
1622
(
2009
).
48.
M. E.
Vaida
,
B. M.
Marsh
, and
S. R.
Leone
,
Nano Lett.
18
,
4107
(
2018
).
49.
J.
Lecante
,
R.
Riwan
, and
C.
Guillot
,
Surf. Sci.
35
,
271
(
1973
).
50.
C.
Guillot
,
R.
Riwan
, and
J.
Lecante
,
Surf. Sci.
59
,
581
(
1976
).
51.
T. E.
Felter
and
P. J.
Estrup
,
Surf. Sci.
76
,
464
(
1978
).
52.
E. I.
Ko
and
R. J.
Madix
,
Surf. Sci.
100
,
L505
(
1980
).
53.
F. J. E.
Scheijen
,
J. W.
Niemantsverdriet
, and
D. C.
Ferré
,
J. Phys. Chem. C
111
,
13473
(
2007
).
54.
F. J. E.
Scheijen
,
D. C.
Ferré
, and
J. W.
Niemantsverdriet
,
J. Phys. Chem. C
113
,
11041
(
2009
).
55.
X.
Tian
,
T.
Wang
, and
H.
Jiao
,
Phys. Chem. Chem. Phys.
19
,
2186
(
2017
).
56.
J. G.
Chen
,
M. L.
Colaianni
,
W. H.
Weinberg
, and
J. T.
Yates
,
Chem. Phys. Lett.
177
,
113
(
1991
).
57.
A. J.
Jaworowski
,
M.
Smedh
,
M.
Borg
,
A.
Sandell
,
A.
Beutler
,
S. L.
Sorensen
,
E.
Lundgren
, and
J. N.
Andersen
,
Surf. Sci.
492
,
185
(
2001
).

Supplementary Material