Di- and Tetrameric Molybdenum Sulfide Clusters Activate and Stabilize Dihydrogen as Hydrides.

NaY zeolite-encapsulated dimeric (Mo2S4) and tetrameric (Mo4S4) molybdenum sulfide clusters stabilize hydrogen as hydride binding to Mo atoms. Density functional theory (DFT) calculations and adsorption measurements suggest that stabilization of hydrogen as sulfhydryl (SH) groups, as typical for layered MoS2, is thermodynamically disfavored. Competitive adsorption of H2 and ethene on Mo was probed by quantifying adsorbed CO on partly hydrogen and/or ethene covered samples with IR spectroscopy. During hydrogenation, experiment and theory suggest that Mo is covered predominately with ethene and sparsely with hydride. DFT calculations further predict that, under reaction conditions, each Mo x S y cluster can activate only one H2, suggesting that the entire cluster (irrespective of its nuclearity) acts as one active site for hydrogenation. The nearly identical turnover frequencies (24.7 ± 3.3 molethane·h-1·molcluster -1), apparent activation energies (31-32 kJ·mol-1), and reaction orders (∼0.5 in ethene and ∼1.0 in H2) show that the active sites in both clusters are catalytically indistinguishable.

Introduction

Two-dimensional, nanostructured transition metal sulfides (TMS) have a wide variety of applications in energy conversion.[1] Owing to their robust performance in the presence of heteroatoms, layered TMS materials, primarily based on Mo(W)S2 promoted by Co/Ni, have long been used as hydrodefunctionalization catalysts in hydroprocessing.[2−7] In addition, TMS catalysts show promise for a multitude of processes related to the production of fuels and chemicals,[1] including the upgrading of bioderived feedstocks (e.g., hydrodeoxygenation),[8−10] (reverse) water gas shift (WGS/rWGS),[11,12] and other synthetic approaches including syntheses from CO/CO2.[13−16] More recently, TMS have been studied as electrocatalysts.[17−21]

Most reactions studied on these catalysts involve H2 as a reactant on sites that have been characterized by microscopy,[22,23] spectroscopy,[24,25] and adsorption of molecular probes,[26,27] guided and aided by theory.[28−30] The efforts have led to significantly improved catalytic properties, as well as to better atomistic understanding of local structures, and detailed mechanisms for hydrogenation/hydrogenolysis reactions. While the link between these catalysts and enzymes with sulfide based active sites seems intuitive, clear analogies have not been substantiated.[2,6,31]

The impressive progress in the understanding of naturally occurring enzymes featuring metal–sulfur moieties in their active sites allows us now to draw stronger analogies to inorganic materials containing atomistically defined sulfide clusters. The active sites in enzymes are the key to redox and hydrogenation catalysis, such as H+/H2 conversion by hydrogenases and nitrogen fixation by nitrogenases.[32−34] To translate this chemistry to nonenzyme systems, several supported (multinuclear) metal–sulfur clusters have been reported, some of which are known to mimic key structural motifs of sulfur-based enzyme cofactors.[35] To make them better accessible for catalytic hydrogenation, inorganic scaffolds are used for stabilization.[14,36]

Activation of H2 on sulfide materials occurs by dissociative adsorption, resulting in distinct final states for adsorbed hydrogen, e.g., SH groups on sulfide slabs or hydride species on metal sites.[31,37−40] Interestingly, the nitrogenase enzyme’s FeMo-cofactor was recently shown to stabilize hydrogen in the form of hydride species; in this case a fraction of adsorbed hydrogen is present as μ-bridging Hδ- on the Fe centers.[34,41−43]

We recently reported MoS clusters encapsulated in faujasite-type NaY zeolite with precisely defined nuclearity, geometry, and atomic connectivity.[44−46] Using thermal treatment in sulfiding/reducing atmosphere, two different molecular cluster sizes were stabilized, viz., dimeric Mo2S4 and tetrameric Mo4S4. The latter cluster structurally and electronically resembles the cubane motifs in the nitrogenase enzyme, i.e., the FeMo-cofactor.[32,33] These catalysts exhibited remarkable stability for ethene hydrogenation in the absence of continuous sulfur supply to the reaction feed, while the classic layered MoS2 catalyst deactivated significantly under the same conditions.[46] This deactivation behavior of conventional MoS2 catalysts in the absence of sulfur in the feed has been reported extensively in literature.[47−49]

These promising results motivated us to address how these MoS clusters activate hydrogen and how their structures dynamically adapt to the reaction environment. Combining IR spectroscopy of adsorbed probe molecules with kinetic measurements and DFT calculations, employing ethene hydrogenation as a model reaction, we develop here the most plausible configurations of MoS clusters and the changes in their geometric and electronic properties upon interactions with reactive gases (ethene, H2, and their mixtures) at low temperature (i.e., 173 K; relevant for CO adsorption measurements) and high temperature (i.e., 473 K; typical hydrogenation reaction temperature). The analyses provide insights into the similarities and differences in H2 activation and the identity of surface hydrogen species among MoS clusters, layered MoS2, and enzymes containing TMS-based structural motifs, and explains why it is justified to treat the entire cluster, rather than individual Mo atoms, as an active site for hydrogenation catalysis.

Results and Discussion

Active Sites for CO Adsorption and Hydrogenation

Direct spectroscopic evidence for the presence of hydrides on Mo atoms of MoS phases has not been achieved so far and appears at present elusive. Thus, we turn to an indirect method to characterize the location and concentration of adsorbed hydrogen, using CO as a probe molecule. CO has been successfully used as a probe for Lewis acid sites (LAS) on TMS.[26,50−52] In this case, CO appears to be an ideal probe. First, as long as CO molecules are not aligned, the wavenumbers of IR bands of CO adsorbed on LAS can be directly related to the electronic properties of the metal sites. Therefore, the observed shift in the band of adsorbed CO allows us to characterize the electronic state of Mo sites in the MoS clusters compared to bulk MoS2. Second, a decrease in the intensity of bands after pre-exposing the catalyst to H2 and/or ethene indicates blockage of Mo sites by adsorbed hydrogen/ethene. Third, the relative change in the intensity of IR bands (after pre-equilibrating with H2/ethene) allows to quantitatively estimate the fraction of Mo covered with hydrogen/ethene.

Figure shows the IR spectra of CO adsorbed on Mo2S4/NaY, Mo4S4/NaY, and MoS2/γ-Al2O3. In addition to the bands attributed to CO adsorbed on the acid sites of the support (i.e., ∼2175 and ∼2125 cm–1 for NaY, ∼2195 and ∼2150 cm–1 for γ-Al2O3), all catalysts showed characteristic broad bands assigned to CO adsorbed on the Mo sites of MoS at ∼2035, ∼2075, and 2085 cm–1 for the MoS/NaY, and at ∼2055, ∼2095, and ∼2105 cm–1 for MoS2/γ-Al2O3.[50,53]

Figure 1

IR spectra of adsorbed CO on Mo2S4/NaY (top), Mo4S4/NaY (middle), and MoS2/γ-Al2O3 (bottom) after quenching to 173 K in vacuum (left) or H2 (right). All spectra are normalized to catalyst wafer thickness and mass.

In comparison to MoS2/γ-Al2O3, the bands of CO adsorbed on Mo in MoS/NaY were red-shifted by ∼20 cm–1, indicating an enhanced electron back-donation from Mo to CO in the case of cluster catalysts. This red-shift suggests a higher electron density in the Mo d-orbitals of MoS clusters in comparison to the Mo atoms in MoS2 slabs. It is interesting to note that the observed shift in the wavenumber was identical, suggesting the local electronic environment of Mo to be similar in both dimeric and tetrameric clusters. We also noticed that the amount of adsorbed CO, normalized to Mo content, on the two MoS/NaY catalysts (after quenching in vacuum) was comparable: 1455–1708 au on Mo2S4/NaY and 1730–1760 au on Mo4S4/NaY (Tables S2.1 and S2.2), suggesting that both dimeric and tetrameric clusters can accommodate similar number of CO molecules per Mo and that all Mo sites are accessible to CO adsorption.

We simulated the adsorption of CO on Mo2S4 and Mo4S4 clusters using DFT. Adsorption of three CO molecules per Mo atom resulted in an octahedral-like coordination environment around the Mo centers (Figure S3.1). This configuration was found to be thermodynamically stable on both clusters (Figure S3.2), thereby confirming the experimental observation that each Mo atom, irrespective of the nuclearity of the cluster it belongs to, can accommodate the same number of CO molecules.

CO binding to the Mo LAS on these NaY-encapsulated MoS clusters shows pronounced similarities to CO binding to the nitrogenase enzyme’s FeMo-cofactor.[54,55] In both cases, CO is adsorbed on a metal atom (Mo in our case; Fe in FeMo-cofactor) that itself is coordinated to three nonmetallic atoms (S only in our case; S and C in the FeMo-cofactor) in its first coordination shell. The similarity between these systems is especially intriguing, considering that the activity of nitrogenase enzyme for catalytic hydrogenation of acetylene was shown to be almost completely lost in the presence of CO.[54] Thus, we hypothesize that the Mo centers are at least part of the active site for ethene hydrogenation on these cluster catalysts.

For ethene hydrogenation, representative Mo2S4/NaY and Mo4S4/NaY catalysts (with ∼3.2 wt % Mo) showed similar apparent activation energies (31 ± 1 kJ·mol–1 on Mo2S4/NaY and 32 ± 3 kJ·mol–1 on Mo4S4/NaY), similar reaction orders of ∼1.0 in H2 and ∼0.5 in C2H4 (Figure ), and virtually identical turnover frequencies (discussed later) indicating that the active sites in both catalysts are catalytically indistinguishable despite different geometries and compositions of the two clusters. It is worth mentioning that X-ray absorption spectroscopy (XAS) measurements verify that the nuclearity of both clusters remains intact under reaction conditions (section S11 in the Supporting Information).

Figure 2

(a) Arrhenius-type plots (T = 463–493 K, pH ≈ 49 kPa, pC ≈ 2.5 kPa) and (b,c) steady-state ethane formation rates as a function of H2 pressure (T ≈ 473 K, pH = 25–74 kPa, pC ≈ 2.5 kPa) and C2H4 pressure (T ≈ 473 K, pH ≈ 49 kPa, and pC = 1.5–5 kPa) on representative Mo2S4/NaY and Mo4S4/NaY catalysts with ∼3.2 wt % Mo.

Hydrogen Adsorption

The amount of CO adsorbed on MoS clusters, measured as the relative peak area of corresponding IR bands, decreased when either catalyst was exposed to H2 (Figure ). This suggests that hydrogen is adsorbed on the same site as CO, i.e., the Mo atoms. In contrast, the intensity of the bands associated with CO adsorbed on MoS2/γ-Al2O3 were unaffected by exposure to H2 (decrease by only ∼5%). This confirms that hydrogen does not bind to Mo but to sulfur atoms at the perimeter, forming SH groups, on the bulk MoS2 catalysts.[31,37,38,52]

Figure 3

Relative amount of adsorbed CO (normalized to catalyst wafer thickness and mass), measured with IR spectroscopy, on Mo2S4/NaY, Mo4S4/NaY, and MoS2/γ-Al2O3 after quenching in either vacuum or H2.

Using 4,6-dimethylpyridine (DMP) to probe weakly Brønsted acidic SH groups via IR bands at ∼1650 and ∼1625 cm–1 (attributed to protonated DMP),[56,57] we had shown that, for Al2O3-supported MoS2, the concentration of SH groups increased after exposure to H2.[58] However, for Mo2S4/NaY (Figure ), we observed only bands of very low intensity and these bands did not increase after exposure to H2. Therefore, we conclude that these small bands result from residual Brønsted acid sites (BAS) of the zeolite support and that the MoS nanoclusters do not form Brønsted acidic SH groups. In consequence, we conclude that hydrogen is indeed adsorbed on the Mo atoms of MoS clusters.

Figure 4

IR spectra of adsorbed 4,6-dimethylpyridine (DMP) on Mo2S4/NaY (before and after admission of 1 bar H2) and parent NaY zeolite. The bands at ∼1650 and ∼1625 cm–1 are assigned to ring vibrations of protonated DMP, the band at ∼1602 cm–1 to weakly physisorbed DMP, and the band at ∼1580 cm–1 to DMP physisorbed/adsorbed on zeolite Lewis acid sites.[56−58]

To investigate the adsorption structure further, we optimized (with DFT) the geometries of Mo2S4 and Mo4S4 clusters with hydrogen dissociatively adsorbed at five distinct locations: a single Mo atom, two different Mo atoms, one Mo and one bridging S atom, two bridging S atoms, and a single bridging S atom. The thermodynamically most stable configuration for both clusters was found to be the one with H2 dissociatively adsorbed on two different Mo atoms (Figure e,f). This configuration was thermodynamically more stable (Tables S4.1 and S5.1) than the configuration with hydrides stabilized on the same Mo atom (Figure c,d). However, the transition from the configuration with hydrides stabilized on a single Mo atom to separate Mo atoms involved a high free energy barrier on both dimeric and tetrameric MoS clusters (Figures S4.2 and S5.2).

Figure 5

DFT/B3LYP/def2-TZVP-optimized geometries of bare Mo2S4 and Mo4S4 clusters (a,b) and configurations involving dissociatively adsorbed H2 on the same Mo atom (c,d), separate Mo atoms (e,f) or one Mo atom and one bridging S atom (g,h). Optimized geometries obtained using PBE0/def2-TZVP were similar. S: yellow; Mo: blue; H: pink.

These high free energy barriers suggest that even though the thermodynamically most stable configuration is the one with hydrides stabilized on separate Mo atoms, these states are kinetically not accessible at reaction conditions. Therefore, under typical reaction conditions, H2 is likely dissociatively adsorbed on the same Mo atom for both Mo2S4 and Mo4S4. This adsorbed hydrogen resembles the homolytic H2 splitting on (noble) metal catalysts[39,40] and points to the similarity to the FeMo cofactor, which similarly is able to stabilize hydrogen as hydride species.[34,42,43,59] We speculate that electronic and structural similarities cause this analogous mode of interaction.

Interestingly, in contrast to the remarkable stability of structures that involved the formation of hydride species, configurations with hydrogen stabilized as SH groups (e.g., Figure g,h) were much less stable (Tables S4.1 and S5.1). For the tetrameric cluster, it was not even possible to stabilize hydrogen as SH groups without destroying the structural integrity of the cluster.

The difference between hydrogen activation on MoS clusters and bulk MoS2 however cannot be directly linked to the dimension or the nuclearity of the transition metal sulfide phase as hydrogen was proven to be stabilized as SH groups on metal–organic S-bridged Mo dimers.[60] Therefore, the key difference has to lie within the electronic properties, and more specifically the electron density (or reducibility) of the involved Mo centers. Stabilization of hydrogen in the form of protons is formally an oxidation process and, therefore, requires reduction of the Mo centers.[61] DFT calculations suggested that stabilization of hydrogen as SH groups resulted in electron density transfer from the H to the MoS cluster. In contrast, stabilization of hydrogen as hydride species on Mo resulted in electron density transfer from the cluster to the H adatoms (Tables S4.3 and S5.3).

Our DFT models predict that a higher electron density on Mo should favor hydride formation, while a lower electron density on Mo atoms should favor sulfhydryl formation. In the case of extended MoS2 slabs, upon formation of SH groups, the resulting electron density transfer from hydrogen is likely compensated by multiple Mo atoms and, thus, SH groups as a final state are thermodynamically feasible. DFT calculations additionally predict a barrierless transfer of adsorbed hydrogen from Mo centers to the neighboring S atoms.[31,37,38,62] In the case of zeolite-encapsulated MoS nanoclusters, on the other hand, we hypothesize that the more covalent character of Mo–S bonds prevents reduction of these Mo centers perhaps due to a higher local electron density on the Mo atoms.[46,63] For this reason, stabilization of hydrogen as hydride species on these molecular clusters is thermodynamically favored over sulfhydryl group formation. It must be noted in passing that these findings are related to the difference between molybdenum and ruthenium sulfide phases for hydrogen binding observed using inelastic neutron scattering. Sulfhydryl groups were the single hydrogen species on MoS2 slabs, while a mixture of SH groups and hydride species on the metal atoms of a more metallic RuS2 phase were observed.[64]

Adsorption of Multiple H2

As both dimeric and tetrameric MoS clusters comprise of multiple Mo atoms, which could potentially adsorb more than one H2, we simulated the adsorption of multiple H2 on both dimeric and tetrameric clusters (section S6 in the Supporting Information).

For Mo2S4, stable geometries for cases with two dissociatively adsorbed H2 molecules could not be achieved. The second H2 could only be stabilized as physisorbed species. For Mo4S4, on the other hand, stable geometries could be achieved for clusters accommodating more than one H2 molecule (Figure S6.1). At 473 K, adsorption of more than one H2, however, was thermodynamically unfavorable (Figure and Table S6.1). We conclude, therefore, that, at typical reaction temperature, both clusters can only stabilize a single dissociatively adsorbed H2, thereby acting as one hydrogenation site. Consequentially, the number of MoS clusters must be the basis for calculating the turnover frequency for ethene hydrogenation. Applying this for a series of MoXS/NaY catalysts with varying Mo loading, we observed, as predicted, a constant turnover frequency of 24.7 ± 3.3 molethane·h–1·molcluster–1 (Figure ). It must be noted that a Mo loading of ∼9.7 wt % corresponds to ∼1.6 Mo atoms or ∼0.8 Mo2S4 clusters per NaY zeolite supercage.

Figure 6

Differential free energy change per additionally adsorbed H2 on Mo2S4 and Mo4S4 clusters, computed at T = 473 K and T = 173 K. Error bars correspond to ±1/2D gas-phase translational entropy of H2.

Figure 7

Steady-state ethane formation rates, normalized per cluster, on Mo2S4/NaY and Mo4S4/NaY catalysts with increasing Mo loading. Reaction conditions: T ≈ 473 K, pH ≈ 96 kPa, and pC ≈ 5 kPa.

Interestingly, the spin-multiplicity of the (thermodynamically most stable) Mo4S4 structure decreased from ω = 5 (i.e., four unpaired electrons) for the bare Mo4S4 cluster to ω = 3 (i.e., two unpaired electrons) for the Mo4S4 cluster with one adsorbed H2 and finally to ω = 1 (i.e., no unpaired electrons) for the Mo4S4 cluster with two adsorbed H2 (Tables S6.1 and S6.2). Similarly, the spin-multiplicity of the dimeric cluster also decreased from ω = 3 (i.e., two unpaired electrons) for the bare Mo2S4 cluster to ω = 1 (i.e., no unpaired electrons) for the Mo2S4 cluster with one adsorbed H2 molecule (Table S4.1). Based on these observations, we hypothesize that the unpaired electrons on Mo atoms likely interact with hydrogen species upon adsorption.

Ethene Adsorption

Figure shows the IR spectra of adsorbed CO on MoS/NaY catalysts after quenching to 173 K in ethene atmosphere. Quantitative analysis of the IR spectra shows that the concentration of CO adsorbed on Mo Lewis acid sites was reduced to ∼14% on Mo2S4/NaY and to ∼11% on Mo4S4/NaY after equilibration with ethene at 173 K (Figure and Table S2.2). This suggests that in the presence of ethene a significantly higher fraction of Mo (as compared to the experiments with H2 pretreatment) becomes inaccessible to CO, thereby suggesting that the coverage of ethene must be higher than that of hydrogen on Mo under the tested conditions.

Figure 8

IR spectra of adsorbed CO on Mo2S4/NaY (left) and Mo4S4/NaY (right) after quenching to 173 K in vacuum (top), ethene (middle), or a mixture of hydrogen and ethene (bottom). All spectra are normalized to catalyst wafer thickness and mass.

Figure 9

Relative amount of adsorbed CO (normalized to catalyst wafer thickness and mass), measured with IR spectroscopy, on Mo2S4/NaY and Mo4S4/NaY after quenching in vacuum, ethene, or a mixture of ethene and hydrogen.

DFT calculations of the adsorption of ethene on Mo2S4 and Mo4S4 (section S7 and S8 in the Supporting Information) showed that ethene could not be stabilized on the S atoms of MoS clusters but only on Mo atoms. However, in contrast to hydrogen adsorption, adsorption of multiple ethene molecules was thermodynamically favored on both Mo2S4 and Mo4S4 (Figure ).

Figure 10

Differential free energy change per additionally adsorbed ethene molecule on Mo2S4 and Mo4S4 clusters, computed at T = 473 K and T = 173 K. Error bars correspond to ±1/2D gas-phase translational entropy of ethene.

Based on free energy calculations at 173 and 473 K, the configurations with two chemisorbed ethene molecules on Mo2S4 (Figure a) and four chemisorbed ethene molecules on Mo4S4 (Figure c) were thermodynamically most favored. The differential standard free energy change, however, decreased with the adsorption of each additional ethene molecule. DFT therefore predicts that, for CO adsorption experiments, a significant fraction of Mo atoms is expected to be covered by ethene. This prediction agrees with IR spectroscopy measurements that showed that ∼86% and ∼89% of Mo atoms were inaccessible to CO once the cluster had been in contact with ethene (Figure and Table S2.2).

Figure 11

DFT/B3LYP/def2-TZVP-optimized geometries of a Mo2S4 cluster with (a) two adsorbed ethene molecules and (b) one ethene and one hydrogen molecule, and a Mo4S4 cluster with (c) four adsorbed ethene molecules and (d) one hydrogen and three ethene molecules. S: yellow; Mo: blue; C: gray; H: pink.

Competitive Adsorption between Ethene and Hydrogen

Using IR spectroscopy and DFT calculations, we showed above that both H2 and ethene are adsorbed on the Mo atoms of the MoS clusters. As both ethene and H2 are simultaneously present during ethene hydrogenation and compete for the same sites, we turned to the IR spectra of adsorbed CO after quenching the samples to 173 K in a typical reaction mixture containing ∼960 mbar bar H2 and ∼50 mbar ethene, resulting in blocking of ∼81% Mo sites for Mo2S4/NaY and ∼72% Mo sites for Mo4S4/NaY (Figure and Table S2.2).

DFT (sections S9 and S10 in the Supporting Information) showed that the Mo2S4 cluster with one ethene molecule and one H2 molecule adsorbed on different Mo atoms (Figure b) was thermodynamically less stable than the thermodynamically most stable configuration with two adsorbed ethene molecules (Figure a).

Similarly, for the tetrameric cluster, the configuration with one hydrogen and three ethene molecules adsorbed on separate Mo atoms (Figure d) was higher in free energy in comparison to the thermodynamically most stable configuration with four adsorbed ethene molecules (Figure c). As this holds true at 173 K as well as at 473 K, we conclude that the clusters with at least one dissociatively adsorbed H2 (i.e., the ones likely functioning as starting point for hydrogenation) are a minority species under reaction conditions. Note that this agrees well with the fact that ethene hydrogenation had a first order pressure dependence in H2 on both Mo2S4/NaY and Mo4S4/NaY catalysts (Figure b).

Given the large distance between Mo atoms on MoS clusters, and consequentially between the reactants adsorbed on these sites (Figure b,d), we hypothesize that the reaction between adsorbed ethene and adsorbed H2 species to form ethane, i.e., Langmuir–Hinshelwood (LH) type mechanism, is unlikely. We instead propose that the reaction proceeds via an Eley–Rideal (ER) type mechanism, wherein ethene weakly adsorbed in the zeolite pores reacts with an adsorbed hydrogen. The fractional reaction order in ethene (instead of unity as expected for ER type reactions) is attributed to the negative influence of the ethene partial pressure on the hydrogen coverage with both ethene and hydrogen competing for the same sites.

DFT simulations for LH-type (involving an adsorbed H2 and an adsorbed ethene) and ER-type (involving as adsorbed H2 and a gas-phase ethene) reaction pathways on the Mo2S4 cluster are presented in section S12 in the Supporting Information. The standard enthalpic barriers for the LH-type reaction pathway were computed to be higher than that for the ER-type reaction pathway (Figures S12.1 and S12.2). The DFT calculations, therefore, support our hypothesis that the ER-type reaction mechanism (as opposed to LH-type reaction mechanism) likely proceeds on these zeolite pore stabilized molybdenum sulfide clusters.

Conclusions

NaY zeolite-encapsulated MoS clusters adsorb H2 dissociatively, binding both H atoms to a single Mo atom. DFT calculations suggest that this adsorption structure is caused by a high barrier to distribute the H atoms evenly among all Mo atoms of the cluster. Adsorption of hydrogen as hydrides contrasts the stabilization of hydrogen as sulfhydryl groups on the edge of conventional MoS2 slabs. The difference is attributed to the easier reducibility of Mo in the larger MoS2 slabs.

Both dimeric and tetrameric MoS cluster catalysts show stable rates of hydrogenation scaling with the concentration of clusters in the catalyst, independent of the cluster nuclearity. This is also reflected by DFT calculations indicating that only one hydrogen can be dissociatively adsorbed per cluster under reaction conditions. The nature of the active site is concluded to be identical for both clusters, i.e., a single Mo center, as demonstrated by constant activation energies and reaction orders in H2 and ethene on both Mo2S4/NaY and Mo4S4/NaY catalysts.

Thus, experiments and theory suggest jointly that active sites in both dimeric and tetrameric catalysts are catalytically indistinguishable for ethene hydrogenation. Theory and experiment also show that ethene, H2, and CO, competitively adsorb on Mo similarly to the situation reported for the FeMo sulfide cluster in nitrogenase. As ethene adsorbs more strongly than H2 on both cluster catalysts, it is predicted to be the most abundant surface species under reaction conditions. Ethene hydrogenation is postulated to proceed via Eley–Rideal type mechanism, with a weakly adsorbed ethene in the zeolite pores reacting with an adsorbed hydrogen. The reaction between adsorbed ethene and adsorbed hydrogen, i.e., the Langmuir–Hinshelwood type pathway, was found to have a high enthalpic barrier owing to large distance between the Mo atoms in these clusters.

Experimental and Computational Methods

Catalyst Precursor Preparation

MoS2/γ-Al2O3 catalyst precursors were prepared by incipient wetness impregnation of γ-Al2O3 (provided by the Chevron company) with an ammonium heptamolybdate (99.98% purity; Sigma-Aldrich) solution of appropriate concentration. The impregnated sample was dried at 383 K overnight followed by calcination in 100 mL·min–1 synthetic air (temperature ramp: 5 K·min–1 to 673 K, hold for 2 h).

Carbonyl-based catalyst precursors were prepared by chemical vapor deposition (CVD). Approximately 200 mg of NaY (Zeolyst CBV100; Si/Al ∼ 2.5; pelletized and sieved to 250–355 μm) was treated under reduced pressure (10–2 mbar) at elevated temperatures (temperature ramp: 5 K·min–1 to 408 K, held for 2 h; 5 K·min–1 to 503 K, held for 2 h; 5 K·min–1 to 653 K, held for 1 h) to carefully remove adsorbed water. Molybdenum hexacarbonyl (>99.9% purity; Sigma-Aldrich) was loaded on the dried zeolite at room temperature under static conditions for a defined amount of time. In the final step, the catalyst precursors were treated under reduced pressure (10–2 mbar) for 10 min to remove physisorbed Mo(CO)6. All carbonyl-based precursors were stored in a glovebox to avoid exposure to air/moisture at any time.

Catalyst Preparation

MoS2/γ-Al2O3 and NaY-encapsulated cluster catalysts were prepared in a lab-scale plug flow reactor (quartz glass tube; 4 mm i.d.). Bronkhorst mass flow controllers were used to regulate gas flow rates. To avoid formation of hotspots, all precursors were diluted 1/10 in SiC (sieved to 500–1000 μm) and placed in the quartz tube supported with quartz wool on both sides. The precursors prepared via incipient wetness impregnation were sulfided in a stream of 20 mL·min–1 H2S (10% v/v in H2) at ambient pressure (temperature ramp: 5 K·min–1 to 673 K, held for 16 h). Precursors prepared via CVD were also sulfided in a flow of 20 mL·min–1 H2S (10% v/v in H2) at ambient pressure (temperature ramp: 5 K·min–1 to 673 K, held for 2 h) to form sulfided MoS/NaY catalyst (previously shown to be primarily composed of dimeric Mo2S4 clusters and denoted as Mo2S4/NaY).[46] The Mo2S4/NaY catalyst was then treated in a stream of pure H2 (temperature ramp: 5 K·min–1 to 673 K, held for 2 h) to obtain the reduced MoS/NaY catalyst (previously shown to be primarily comprised of tetrameric Mo4S4 clusters and denoted as Mo4S4/NaY).[46] After thermal treatments, all catalysts were purged with N2 for 30 min prior to any catalytic reactions.

Catalytic Reactions

All catalytic reactions were also performed in the lab-scale plug flow reactor. Ethene hydrogenation was studied at ∼473 K and ambient pressure with a H2/ethene volumetric ratio of ∼20. The product stream composition was analyzed by online gas chromatography using an Agilent 7890B GC. Ethane formation rates were determined using space–time yields under differential conditions after the catalysts reached a stable steady state (after ∼24 h). External mass transport limitations have been excluded for the applied reaction conditions by varying the amount of loaded catalyst and its particle size. Limitations by internal mass transport were excluded owing to the small size of MoS clusters (∼5 Å for Mo4S4) and reactants/products in comparison to the diameter of the NaY zeolite supercages (∼12 Å) and the pore openings (∼7 Å).

Infrared (IR) Spectroscopy of Adsorbed Probe Molecules

IR spectroscopy of adsorbed probe molecules was performed using a Nicolet 6700 IR spectrometer with a resolution of 4 cm–1. The catalyst materials were ground and pressed into self-supporting wafers (∼5 mg·cm–2). The catalyst wafers were first sulfided/reduced in a stream of 20 mL·min–1 H2S (10% v/v in H2) or 20 mL·min–1 H2 at ambient pressure (temperature ramp: 5 K·min–1 to 673 K, held for 2 h). Then, for activation, a treatment in H2 (4 cycles, total of 24 h) at 473 K each followed by evacuation at 10–6 mbar for 30 min was applied to all samples.

CO adsorption was performed on activated samples after cooling to 173 K using liquid nitrogen. The samples were examined after cooling in either H2, ethene, a mixture of H2 and ethene, or high vacuum. In a first run, after cooling to 173 K (by skipping the last evacuation cycle) in either H2, ethene, or a mixture of H2 and ethene, CO adsorption isotherms were obtained by applying controlled doses of CO ranging from 0.01 to 2 mbar. After this, CO and other adsorbed gases were desorbed at room temperature under high vacuum (10–7 mbar). Subsequently, the samples were thermally treated at 473 K for 1 h and cooled down to 173 K again under high vacuum (10–7 mbar) before the second run of CO adsorption. Possible errors caused by different thicknesses of the catalyst wafers are prevented in this procedure as multiple series of measurements are conducted on the same wafer. In lieu of molar extinction coefficients for bands of CO adsorbed on cluster catalysts, we analyzed the areas normalized to wafer thickness for all experiments.

IR spectroscopy of adsorbed 4,6-dimethylpyridine (DMP) was performed on the activated samples at 323 K by applying small doses of DMP (up to 0.5 mbar) into the IR cell and equilibrating for 0.5 h. A second spectrum was taken after exposing the catalyst to an additional ∼1 bar of H2 and equilibrating for another 0.5 h.

All IR spectra were background corrected using the OMNIC software package, and they are presented as difference spectra against the reference spectra at 10–7 mbar.

Computational Details

Unrestricted Kohn–Sham (UKS) DFT calculations were performed on gas-phase MoS clusters using the Orca quantum chemistry package version 4.2.[65−67] The calculations were performed using two hybrid exchange-correlational functionals: B3LYP and PBE0. Relativistic effects were taken into account by zeroth-order regular relativistic approximations (ZORA), and Grimme’s atom-pairwise dispersion correction with the Becke–Johnson damping scheme (D3BJ) was used for dispersion corrections.[68−70] Relativistically recontracted versions of the all-electron Ahlrichs def2 basis sets with triple-ζ polarization functions, ZORA-def2-TZVP, were employed for geometry optimization and vibrational frequency calculations, while ZORA-def2-TZVPP basis-sets were employed for single-point energy (SPE) calculations.[71] The hybrid functionals were employed with RIJCOSX approximation to speed up the calculations, and general auxiliary basis sets SARC/J were used for this purpose.[72−77] The RIJCOSX approach incorporates the resolution-of-identity (RI) approximation for the evaluation of the Coulomb matrices and the chain-of-spheres algorithm for the formation of the exchange-type matrices.[78−80] The core electrons (1s22s22p63s23p63d10 for Mo, 1s22s22p6 for S, 1s2 for C, and none for H) were kept frozen during geometry optimization and vibrational frequency calculations. The Hirshfeld charge populations and Mayer bond orders were computed for the optimized geometries. Standard thermodynamics equations were used for computing the free energy and free enthalpy of different structures, and the methodology is presented in detail in section S1 in the Supporting Information.