Methanol Synthesis from H2 and CO2 on a Mo6S8 Cluster: A Density

Oct 30, 2009 - Department of Chemistry, Brookhaven National Laboratory, Upton, New York 11973, and Department of Chemistry, State University of New Yo...
3 downloads 13 Views 3MB Size
Download PDF
3888

J. Phys. Chem. A 2010, 114, 3888–3895

Methanol Synthesis from H2 and CO2 on a Mo6S8 Cluster: A Density Functional Study† Ping Liu,*,‡ YongMan Choi,‡ Yixiong Yang,§ and Michael G. White‡,§ Department of Chemistry, BrookhaVen National Laboratory, Upton, New York 11973, and Department of Chemistry, State UniVersity of New York (SUNY) Stony Brook, Stony Brook, New York 11794 ReceiVed: July 17, 2009; ReVised Manuscript ReceiVed: September 14, 2009

Catalytic CO2 hydrogenation to methanol has received considerable attention as an effective way to utilize CO2. In this paper, density functional theory was employed to investigate the methanol synthesis from CO2 and H2 on a Mo6S8 cluster. The Mo6S8 cluster is the structural building block of the Chevrel phase of molybdenum sulfide, and has a cagelike structure with an octahedral Mo6 metallic core. Our calculations indicate that the preferred catalytic pathway for methanol synthesis on the Mo6S8 cluster is very different from that of bulklike MoS2. MoS2 promotes the C-O scission of HxCO intermediates, and therefore, only hydrocarbons are produced. The lower S/Mo ratio for the cluster compared to stoichiometric MoS2 might be expected to lead to higher activity because more low-coordinated Mo sites are available for reaction. However, our results show that the Mo6S8 cluster is not as reactive as bulk MoS2 because it is unable to break the C-O bond of HxCO intermediates and therefore cannot produce hydrocarbons. Yet, the Mo6S8 cluster is predicted to have moderate activity for converting CO2 and H2 to methanol. The overall reaction pathway involves the reverse water-gas shift reaction (CO2 + H2 f CO + H2O), followed by CO hydrogenation via HCO (CO + 2H2 f CH3OH) to form methanol. The rate-limiting step is CO hydrogenation to the HCO with a calculated barrier of +1 eV. This barrier is much lower than that calculated for a comparably sized Cu nanoparticle, which is the prototypical metal catalyst used for methanol synthesis from syngas (CO + H2). Both the Mo and S sites participate in the reaction with CO2, CO, and CHxO preferentially binding to the Mo sites, whereas S atoms facilitate H-H bond cleavage by forming relatively strong S-H bonds. Our study reveals that the unexpected activity of the Mo6S8 cluster is the result of the interplay between shifts in the Mo d-band and S p-band and its unique cagelike geometry. 1. Introduction The increasing consumption of oil and the environmental crisis resulting from fossil fuel combustion (CO2 emissions) demands attempts to meet global energy needs without further impact on the environment. Compared to other alternative sources of energy (solar, hydro, wind, or nuclear), developing renewable technologies to produce simple liquid fuels such as methanol or ethanol is an attractive option due to their high energy density and ease of storage and distribution.1,2 Using renewable agricultural products or biomass conversion for the direct production of alcohols suffers from relatively low yields and thus does not have the potential to meet global needs.3 The industrial production of liquid hydrocarbon fuels from “syngas”, a mixture of CO and H2, is based on well-developed technology (Fischer-Tropsch synthesis) that in principle can be scaled to meet the world’s needs for liquid fuels such as diesel.4 Syngas is currently derived from steam-reforming of methane, a fossilderived feedstock, so that current practices are not sustainable. A more attractive option of recycling CO2 into alcohols via hydrogenation is sustainable, where the hydrogen is produced from a variety of renewable technologies, such as solar water splitting. Despite the favorable thermodynamics of the CO2 conversion to methanol or ethanol, kinetic limitations for many of the elementary reaction steps (e.g., CO2 dissociation) require † Part of the special issue “Green Chemistry in Energy Production Symposium”. * Corresponding author. E-mail: [email protected]. ‡ Brookhaven National Laboratory. § SUNY.

the use of catalysts, which so far suffer from poor activity and selectivity.5-9 Molybdenum sulfide catalysts, specifically MoS2, have also been evaluated for alcohol synthesis from syngas.10 Both supported (e.g., Al2O3 and carbon) or doped (e.g., Rh, Ni, and Co) MoS2 show promising catalytic activity for the conversion of syngas to alcohols.8 In particular, MoS2-based catalysts exhibit higher sulfur- and coke-resistance and better selectivity to higher alcohols than other catalysts for alcohol synthesis (e.g. Rh-based and Cu-based catalysts8). On the other hand, bare or unsupported MoS2 particles convert syngas only to hydrocarbons.11,12 The promoters, such as alkali and transition metals, are necessary for alcohol synthesis,8 even though the absolute yield and selectivity is still too low to be viable as a commercial process. Moreover, the design and optimization of Mo sulfide catalysts is hindered by continuing uncertainties regarding the nature of the active sites and the reaction mechanism.8 A recent theoretical study showed that CO hydrogenation takes place at Mo-edge sites of bulk MoS2 and proceeds via HxCO formation, followed by C-O bond cleavage to produce methane and CO2.12 More realistic catalysts consist of small MoS2 particles dispersed on a high-surface-area support,8 and particle-support interactions could modify the active sites or edge termination structures and thereby alter the catalytic mechanism and activity.13,14 In the present paper, density functional theory (DFT) was employed to investigate methanol synthesis from CO2 and H2 (CO2 + 3H2 f CH3OH + H2O) on a Mo6S8 cluster. The Mo6S8 cluster has been previously identified as the building block of the well-known Chevrel phase of molybdenum sulfide in AxMo6S8 or ABMo6S8 solid state compounds.15,16 Crystalline

10.1021/jp906780a  2010 American Chemical Society Published on Web 10/30/2009

Methanol Synthesis from H2, CO2 on a Mo6S8 Cluster Chevrel phases and amorphous ternary molybdenum sulfides have been found to be active catalysts for methanethiol synthesis and hydrodesulfurization.17 In the gas phase, the highly symmetric cagelike structure of the Mo6S8 cluster (Oh symmetry) is especially stable with a relatively large HOMO-LUMO gap of 0.8-0.9 eV.18-26 Density functional calculations on the free MoxSy clusters (M ) Mo, W; x/y ) 2/5, 3/7, 4/6, 5/7, 6/8) show that that the M6S8 clusters have the highest CO binding energy,18 and this trend is also observed experimentally for the MoxSy clusters (x/y ) 3/7, 4/6, 5/7, 6/8, 7/10, 8/12) deposited on a Au(111) surface.27 Because the hydrogenation of CO is expected to play a role in the methanol synthesis mechanism, a stronger Mo-CO bond could facilitate the reaction by weakening the C-O bond and/or stabilizing the HxCO intermediates. On the basis of the expectation of high catalytic activity and high structural stability, the Mo6S8 cluster was chosen for study in this work as a potentially novel Mo sulfide nanocatalyst. Compared to stoichiometric MoS2, the Mo6S8 cluster has a lower S/Mo ratio, which should lead to less Mo-to-S electron transfer and more low-coordinated Mo sites available for the reaction.12-14,28-31 On the basis of these considerations, the Mo6S8 cluster is expected to be more reactive than the Mo-edge sites of bulk MoS2. The results presented in this work show that this is not the case. Specifically, the Mo6S8 cluster is not as active as MoS2 in promoting C-O bond rupture of HxCO intermediates that lead to hydrocarbon products, yet following Sabatier’s principle, the cluster’s more moderate bonding of the reaction intermediates facilitates the selective conversion of CO2 and H2 to methanol. Moreover, the Mo6S8 cluster stays structurally intact throughout the reaction process. The methanol synthesis reaction proceeds via the reverse water-gas shift reaction (RWGS, CO2 + H2 f CO + H2O), which is followed by CO hydrogenation to methanol via HCO (CO + 2H2 f CH3OH). Both the Mo and S sites participate in the reaction with CO2, CO and CHxO preferentially binding to the Mo sites, whereas S atoms facilitate H-H bond cleavage by forming relatively strong S-H bonds. 2. Computational Methods DFT calculations for the interaction of CO with unsupported Mo6S8 and Mo6S8/Au(111) were performed with the projectoraugmented wave32 method using the Vienna ab initio simulation package.33,34 The generalized gradient approximation (GGA) with the Perdew-Burke-Ernzerhof (PBE) exchange-correlation functional35 was used. The kinetic energy cutoff for a plane wave basis set was 400 eV. Only Γ-points was considered for the calculations of the unsupported Mo6S8 cluster, whereas a Monkhorst-Pack mesh36 was applied with 3 × 3 × 1 in the case of Mo6S8/Au(111), allowing convergence to 0.01 eV of the total electronic energy. The Au(111) surface was modeled by a three-layer slab with a (4 × 4) unit cell, which was separated by enough vacuum space. The Mo and S edges of bulk MoS2 (Figure 1a) were modeled by cleaving the bulk parallel to the (100) or (010) plane. For surfaces, the top metal or S-Mo-S layer was allowed to relax with the adsorbates or the supported Mo6S8 cluster while the other layers of atoms were fixed in their optimized bulk positions. The calculations for CO2 hydrogenation to methanol on a Mo6S8 cluster were conducted using DMol3,37,38 which utilizes the effective core potentials, double-numerical basis set with polarization functions and GGA-PBE35 for the exchange and correlation functional. A global orbital cutoff of 5.5 Å was used. The transition states were identified by synchronous transit methods,39 which yield

J. Phys. Chem. A, Vol. 114, No. 11, 2010 3889

Figure 1. Optimized edge surface of MoS2 and Mo6S8 cluster (Mo, big cyan; S, small yellow).

TABLE 1: Optimized Geometric Parameters for the Mo Edge of MoS2 and Mo6S8a d d q q n n a

(Mo-S) (Å) (Mo-Mo) (Å) (Mo) (e) (S) (e) (Mo) (S)

Mo edge of MoS2

Mo6S8

2.37 3.20 0.19 -0.07 4 3

2.46 2.66 0.27 -0.20 8 3

d: bond length; q: partial charge; n: coordination number.

results close to those obtainable by eigenvector following methods. The Mo6S8 cluster was allowed to fully relax with no constraints. 3. Results and Discussions 3.1. CO Adsorption on Free and Au(111)-Supported Mo6S8. Before investigating the methanol synthesis on Mo6S8, we first tried to identify the active sites and the stability of the cluster using CO as a probe molecule. The structure of the Mo6S8 cluster found in the Chevrel phases is highly symmetric with a Mo6 octahedral core and the sulfur atoms located in the faces of the octahedron.40 Previous DFT calculations show that the symmetric “Chevrel” structure is also the lowest-energy structure for the free Mo6S8 cluster (cation and neutral).20,41,42 Our calculated lowest-energy structure for the Mo6S8 cluster, shown in Figure 1b, is similar to these earlier results.43 The calculated Mo-Mo bond length of Mo6S8 is ∼2.66 Å, and the Mo-S bond length is ∼2.46 Å (see Table 1). Compared with the Mo edge of MoS2 (Figure 1a), which is active for syngas conversion,12 the Mo-S bond of Mo6S8 is slightly longer, whereas the Mo-Mo bond is much shorter. With a bond length of 3.20 Å, the Mo-Mo interaction in MoS2 is negligible. Therefore, rather than generating more low-coordinated sites on moving from bulk to nanoparticles,28,29,44 the coordination number of Mo in the small Mo6S8 cluster is twice as large as that on the Mo edge of MoS2 (Figure 1). As a result, the Mo atoms in Mo6S8 are more positively charged (Table 1), and the Mo d-band shifts away from the Fermi level (Figure 2). In contrast, due to the lower S/Mo ratio, the S atoms of Mo6S8 are more negatively charged (Table 1), and the S p-band lies closer to the Fermi level compared to that of MoS2 (Figure 2). As shown below, the differences in electronic structure result in a less active Mo site and a more active S site in going from MoS2 to Mo6S8. Moreover, the S atoms in Mo6S8 are not spectators, but directly participate in the reaction to stabilize the dissociated H atoms. Figure 3a plots the adsorption energies [∆Eads(n)] of CO on Mo6S8 as a function of the number of CO molecules (n), where ∆Eads(n) ) E(nCO/Mo6S8) - E(Mo6S8) - nE(CO). CO adsorbs

3890

J. Phys. Chem. A, Vol. 114, No. 11, 2010

Liu et al.

Figure 2. Partial density of states of the Mo edge of MoS2 and the Mo6S8 cluster.

Figure 3. Calculated adsorption energies of CO on (a) Mo6S8 and (b) Mo6S8/Au(111) as a function of the number of CO. The numbers labeled in the figure represent the sequential adsorption energy of CO.

strongly at the Mo sites with a slightly tilted structure, and the cluster stays intact with slight distortion. The calculated adsorption energy of the first CO molecule is around -1.3 eV, which is much weaker than that on the Mo edge of MoS2 (-2.24 eV).12 This is due to the downward shifted Mo d-band of Mo6S8, as shown in Figure 2, which moderates the activity of Mo. Our results are consistent with other studies,45,46 which indicate the essential role of the d-band position in determining metaladsorbate (e.g., CO, O, OH, S, ...) bond strengths. We also calculated the sequential adsorption energies using ∆Eseq_ads(n) ) ∆Eads(n) - ∆Eads(n - 1). As shown in Figure 3, the adsorption of up to six CO molecules is predicted to be highly exothermic, and ∆Eseq_ads(n) for adding each ligand is almost identical. The adsorption of additional CO (n > 6) results in two CO molecules sharing one Mo site, which levels off the adsorption energy. Specifically, the binding energy of the seventh CO is small [∆Eseq_ads (n ) 7) ) -0.62 eV] and is likely to be unbound. This result is similar to that observed experimentally for CO adduct formation on the free Mo6S8+ cation cluster for which adducts with n > 6 have very low probability.20 To evaluate catalytic activity, it is important to understand how the structure, stability, and activity change when depositing the cluster on a support material. Similar to the previous studies,47,48 a flat Au(111) surface was used here as a support for the Mo6S8 cluster. To obtain the most plausible surface model, we carried out ab initio molecular dynamics (MD) simulations (T ) 300 K) based on the structure proposed by Popov and co-workers.48 In agreement with previous results,48 our calculations show that Mo6S8 is very stable on Au(111), with a binding energy of -1.79 eV. The corresponding conformation is shown in Figure 4, where the cluster anchors on the surface through one Mo atom and four S atoms and the shortest distance between adsorbed Mo6S8 and the Au surface is 2.323 Å. To compare with unsupported Mo6S8, we calculated

Figure 4. Side view of optimized geometries for CO adsorption on Mo6S8/Au(111) as a function of the number of CO (n ) 1-6). For clarity, only the topmost Au layer is shown (Mo, big cyan; S, small yellow; C, small gray; O, small red).

the sequential adsorption energies of CO on Mo6S8/Au(111) (Figure 3b). The corresponding optimized geometries are shown in Figure 4. One can see that the effect of the Au support on CO adsorption is negligible in terms of adsorption energies and geometries. Similar to the unsupported case, CO prefers the Mo top sites, and the corresponding binding energy changes only slightly (Figure 3). According to our charge density calculations, the charge transfer from Au(111) to Mo6S8 is negligible (0.03e). The only significant effect is to decrease the number of active Mo sites from six in Mo6S8 to five in Mo6S8/Au(111). As a result, Mo6S8/Au(111) is able to strongly adsorb up to five CO molecules, and the addition of the sixth CO levels off the energy with ∆Eseq_ads (n ) 6) ) -0.35 eV (Figure 3b), since it has to share one Mo atom with another CO (Figure 4). Therefore, Au(111) seems to function only as a support to stabilize the cluster and does not affect the chemical activity of the supported Mo6S8. In addition, one can also see that the structure of the Mo6S8 cluster stays intact after the adsorption of multiple CO molecules, not only in the gas phase but also supported on Au(111). Therefore, a study of the unsupported Mo6S8 cluster is meaningful for simulating its catalytic activity of the cluster supported on materials such as Au, which are not active enough to participate in the reaction directly; however, a detailed kinetic study is necessary when using active supports, such as Al2O3. It was reported that the Mo edge of MoS2 is very active, being able to break the C-O bond of CHxO and form methane, whereas the production of alcohols is very unlikely.49,12 Ac-

Methanol Synthesis from H2, CO2 on a Mo6S8 Cluster

J. Phys. Chem. A, Vol. 114, No. 11, 2010 3891

Figure 6. Possible reaction paths for CO2 hydrogenation on Mo6S8, where the reaction energies (top) and barriers (bottom in parentheses) are expressed in eV (Mo, big cyan; S, small yellow; C, small gray; O, small red; H, small white). Figure 5. Optimized geometries for coadsorption of the dissociated H and other important intermediates, CO2 (a), HOCO (b), CO (c), HCO (d), H2CO (e), and H3CO (f), in CO2 hydrogenation to methanol on Mo6S8 (Mo, big cyan; S, small yellow; C, small gray; O, small red; H, small white).

cording to our calculations, the Mo atoms of Mo6S8 are less active than the Mo edge of MoS2. The question is whether this lower activity will suppress the C-O bond cleavage and therefore facilitate alcohol formation. In the next section, we investigate the first step in methanol synthesis, which involves the hydrogenation of CO2 adsorbed on the Mo6S8 cluster. 3.2. CO2 Hydrogenation on Mo6S8. 3.2.1. H2, CO2 Adsorption and Dissociation. Facile dissociation of H2 was predicted on the Mo edge of MoS2, where atomic H prefers the Mo-bridge sites.12 According to our calculations, a H2 molecule will dissociate spontaneously on Mo6S8, although the H2 adsorption is 1 eV weaker than on the edge of MoS2. The dissociated H atom sits at a Mo-S bridge site, which is accompanied by electron transfer from the H atom to the Mo and S atoms. We have shown in Section 3.1 that for CO adsorption, only the Mo sites of Mo6S8 are active. However, in the case of hydrogen, the S atoms of Mo6S8 are not spectators, but participate in binding of the adsorbed H atoms together with Mo. In the case of MoS2, no H-S interaction was observed.12 This is attributed to the up-shifted S p-band of Mo6S8 (Figure 2), which allows a stronger S-H interaction than that of MoS2. In a hydrogenation process, our calculations show that the H atoms also prefer to occupy the Mo-S bridge sites close to the other adsorbed intermediate to make a facile hydrogenation (Figure 5). In addition, the effect of coadsorption on the binding energy is very small, which is less than 0.2 eV, as compared to the case without coadsorbed H atoms. Therefore, we did not include the coadsorbed species in the overall reaction pathway. CO2 was found to strongly adsorb at the Mo edge of MoS2 (Figure 1a) through the C atom. The adsorption energy is -1.60 eV, and the O-C-O bond is bent.12 On the Mo6S8 cluster, the binding of CO2 is much weaker (-0.29 eV) via one of the O atoms, and the O-C-O bond stays linear (Figure 6). Similar to the case of CO, the weaker CO2-Mo interaction in the case of Mo6S8 is associated with the downward shift of the Mo

d-band (Figure 2), which should result in negligible activity toward CO2 dissociation. As shown in Figure 6, both the dissociated CO and atomic O adsorb on top of the Mo atom and the reaction is highly endothermic, with a reaction energy, ∆Er, of +2.55 eV, which is unlikely to occur under the methanol synthesis conditions. By contrast, the dissociation of CO2 on MoS2 is an exothermic reaction (∆Er ) -0.67 eV).12 3.2.2. CO Production Wia ReWerse Water-Gas-Shift Reaction. In contrast to CO2 dissociation, CO2 hydrogenation is energetically preferred on the Mo6S8 cluster. Two possible steps were considered here: CO2 + H f HCOO and CO2 + H f HOCO. One can see in Figure 6 that both are less endothermic than the CO2 dissociation. Furthermore, the barrier (∆Ea) for carboxyl (HOCO) formation is +0.82 eV, which is 0.77 eV lower than that for the formate (HCOO; see Figure 6). Therefore, the first step to CO2 hydrogenation is likely to produce HOCO. Similarly, two pathways for the sequential reaction of HOCO were included. One is direct dissociation (HOCO f HO + CO) and the other is H-guided dissociation (HOCO + H f H2O + CO). Both reactions involve C-O bond cleavage; however, our calculations (Figure 6) show that the presence of adsorbed H atoms helps lower the barrier for C-O bond cleavage. The H-guided dissociation is highly exothermic with ∆Er ) -1.78 eV and ∆Ea ) +0.69 eV, whereas the direct dissociation is endothermic (∆Er ) 0.73 eV). That is, the HO + CO intermediate on Mo6S8 is less stable than the transition state for HOCO + H f H2O + CO, and the corresponding barrier for HOCO f HO + CO should be higher than that for HOCO + H f H2O + CO. Therefore, the reaction between H2 and CO2 on Mo6S8 starts with the RWGS reaction, converting CO2 into CO and H2O via the HOCO intermediate. The water product desorbs, and the CO remains on the cluster for further hydrogenation (Figure 6). The highest barrier during this process is +0.82 eV, corresponding to CO2 hydrogenation to HOCO. This is similar to MoS2, which was found to be able to catalyze the RWGS reaction with high CO selectivity.50 3.2.3. Methanol Production Wia CO Hydrogenation. The next question to be addressed is whether Mo6S8 behaves like MoS2 to produce hydrocarbons from CO hydrogenation.11,12 To answer this question, we optimized the reaction pathway for CO hydrogenation to methanol on the Mo6S8 cluster. There are

3892

J. Phys. Chem. A, Vol. 114, No. 11, 2010

Figure 7. Possible reaction paths for CO hydrogenation on Mo6S8, where the reaction energies (top) and barriers (bottom in parentheses) are expressed in eV (Mo, big cyan; S, small yellow; C, small gray; O, small red; H, small white). The optimal path is labeled by red arrows.

four possible products for the first CO hydrogenation step: COH, HCO, OH + C, and HC + O. Note that direct CO dissociation was not considered here. As demonstrated above, Mo6S8 can not break the CdO double bond of CO2, and therefore, cleavage of the CtO triple bond of CO should not be feasible. In fact, CO dissociation is also highly activated at the Mo edge of MoS2,12 which bonds the adsorbates stronger than Mo6S8. One can see in Figure 7 that the H-guided C-O bond cleavage resulting in OH + C or HC + O is endothermic, with ∆Er ≈ +2 eV. In the hydrogenation processes, the formation of COH is also energetically uphill (∆Er ) +1.75 eV). In contrast, the production of formyl (HCO) is much more favorable, which corresponds to the lowest energy path with ∆Er ) +0.96 eV and ∆Ea ) +1.00 eV. This is attributed to different binding configurations. As seen in Figure 7, both COH and HCO bind to the Mo top site; however, COH binds only through the C atom (η1-C), and both the C and O atoms interact with Mo in the case of HCO (η2-C, O). The negatively charged dangling O atom of HCO strongly interacts with the positively charged Mo atom (Mo-O bond length of 2.44 Å), which helps stabilize the HCO species on Mo6S8. Accordingly, the CO hydrogenation on Mo6S8 starts with HCO formation. Following the same idea, we also calculated the energies for the sequential reaction of HCO: direct dissociation (HCO f HC + O), H-guided dissociation (HCO + H f H2C + O or HC + OH), and hydrogenation (HCO + H f HCOH or H2CO) (Figure 8). The energetically preferred path is the production of formaldehyde (H2CO) with ∆Er ) -0.60 eV and ∆Ea ) +0.61 eV. The hydrogenation of H2CO produces methoxy (H3CO), where the C-Mo bond is broken and H3CO binds with Mo6S8 via a Mo-O bond (Figure 9). This reaction is downhill (∆Er ) -0.61 eV), and the corresponding ∆Ea is as low as

Liu et al.

Figure 8. Possible reaction paths for CHO hydrogenation on Mo6S8, where the reaction energies (top) and barriers (bottom in parentheses) are expressed in eV (Mo, big cyan; S, small yellow; C, small gray; O, small red; H, small white). The optimal path is labeled by red arrows.

0.43 eV. By comparison, H2CO dissociation (H2CO f H2C + O, H2CO + H f H2C + OH) and hydrogenation at the O atom to produce H2COH are energetically less favorable. Finally, the H3CO intermediate does not dissociate (H3CO f H3C + O, H3CO + H f H3C + OH) because the barriers for both processes are very high (+1.53 and +2.19 eV, respectively). The only plausible path for H3CO is the formation of methanol, which is exothermic (∆Er ) -0.91 eV), and the corresponding barrier is very low (∆Ea ) +0.34 eV). Figure 10 displays the potential energy diagram of the optimal pathway for methanol synthesis from CO2 and H2 on Mo6S8. One can see that the reaction starts with the RWGS reaction, in which CO2 is converted into CO and H2O via the HOCO intermediate. Sequential hydrogenation of CO is initiated with HCO formation, followed by H2CO, H3CO, and finally, methanol productions. The overall reaction is exothermic (-0.56 eV). The rate-limiting step is CO hydrogenation to HCO, and the corresponding barrier is +1 eV. Using the same code and a setup similar to the present calculation, we also calculated the methanol synthesis on a Cu nanoparticle, which is considered to represent the catalytic activity of the commercial catalysts for methanol synthesis (Cu-ZnO/Al2O3).51,52 The results show that the barrier for the rate-limiting C-O bond-breaking in dioxomethylene (H2COO) is +1.77 eV. Given that, Mo6S8 could also be a good catalyst for conversion of CO2 to methanol. We also calculated the overall barrier, which is defined as the highest energy to overcome along the reaction pathway. There are two highest transition states for methanol synthesis on Mo6S8sCO formation and H2CO formationswhich are 1.2 eV higher than the energy of the gas-phase reactants. The overall barrier for CO formation starting from CO2 adsorbed is +1.48 eV, and that for H2CO formation starting from adsorbed CO is +1.57 eV. Therefore, the formation of methanol from CO2 and H will have to overcome a barrier of ∼1.6 eV. Although a barrier of 1.6 eV may look high, it should be possible to overcome this

Methanol Synthesis from H2, CO2 on a Mo6S8 Cluster

J. Phys. Chem. A, Vol. 114, No. 11, 2010 3893

Figure 9. Possible reaction paths for CH2O and CH3O hydrogenation on Mo6; the reaction energies (top) and barriers (bottom in parentheses) are expressed in eV (Mo, big cyan; S, small yellow; C, small gray; O, small red; H, small white). The optimal path is labeled by red arrows.

Figure 10. Optimized potential energy diagram for methanol synthesis from CO2 and H2 on a Mo6S8 cluster. The energies in the figure are expressed with respect to free Mo6S8, CO2, and H2 in gas phase. Thin bars represent the reactants, products, and intermediates; the thick bars stand for the transition states. The geometries of the transition states involved in the reaction were also included (Mo, big cyan; S, small yellow; C, small gray; O, small red; H, small white).

barrier under experimental conditions. To estimate the reaction rate, we use the Arrhenius equation, rate ≈ νe-∆Ea/kT. Assuming the standard value for the preexponential factor of ν ≈ 1013 and setting ∆Ea ) +1.6 eV, we obtain a reaction rate of ∼0.1 s-1 at T ) 573 K. Significantly higher rates are possible with modest increases in reaction temperature (e.g., rate ≈ 102 s-1 at T ) 773 K) or a higher pre-exponential factor (rate ≈ 103

s-1 at T ) 573 K with ν ≈ 1017). In addition, higher CO2 and H2 pressures should also help to increase the rate. Experimentally, it is known that the RWGS rate for MoS2 increases with increasing CO2 and H2 pressures.50 For the RWGS reaction on Mo6S8 (Figure 10), high pressures of CO2 and H2 will shift the equilibrium, CO2 + H2 f CO + H2O, toward more CO, which would then be available for methanol synthesis. Furthermore,

3894

J. Phys. Chem. A, Vol. 114, No. 11, 2010

as shown in Section 3.1, a high CO coverage will also force several CO molecules to share one Mo atom, which makes the CO-Mo bond weaker (Figures 3 and 4) and thus enhance the rate-limiting CO hydrogenation to HCO according to the Brøndsted-Evans-Polanyi relation.53 As shown in Figure 10, once the HCO intermediate is formed, the subsequent hydrogenation steps are exothermic all the way to methanol. Finally, we also notice that the Mo6S8 cluster is quite stable and only slightly distorted during the reaction. In the next step, molecular dynamics (MD) simulations will be employed to explore the stability of Mo6S8 under the experimental temperature and pressure. 3.2.4. Implication for Catalysis on Nanoparticle. Nanoparticles can have unique chemical activities with respect to those of bulk species and, thus, have found fascinating uses in catalysis.28,44,54 In this respect, most of the attention has been focused on understanding the behavior of noble-metal nanoparticles.28,29,44 For example, when compared to Au bulk, the exceptional catalytic activity of a Au nanoparticle can be a consequence of the presence of active edge sites that have a particular structure.55,56 Much less is known about the physics responsible for the chemical and catalytic properties of metal compound nanoparticles. For these systems, one must consider ligand (electronic perturbations induced by compound formation) and ensemble effects (reduction in the number of exposed metal sites), together with possible variations in stoichiometry and geometry with respect to the bulk materials. In general, it has been found that the activity of the metal sites is lower in metal compounds and increases with an increasing nonmetal/ metal ratio.12-14 The present study shows that the Mo6S8 cluster behaves differently. Despite its smaller size and lower S/Mo ratio, the Mo6S8 cluster has lower activity as compared to the Mo edge of MoS2. MoS2 promotes the C-O scission in HxCO intermediates, and therefore, only hydrocarbons are produced. As shown in Figures 6-9, the C-O cleavage of HxCO on Mo6S8 via either direct or H-guided dissociation is always more difficult than the hydrogenation. This can be attributed to the lower-lying d-band of Mo and, therefore, a weaker binding strength as compared to the Mo atoms of MoS2 (Figure 2). In contrast, with an up-shifted p-band, the S atoms of Mo6S8 are more active, being able to facilitate the H2 dissociation by forming S-H bonds, whereas in the case of MoS2 the S atoms do not interact with any intermediate. This combination of electronic factors promotes the selective production of methanol on the Mo6S8 cluster by facilitating the hydrogenation and suppressing the C-O cleavage of HxCO. IV. Conclusion Catalytic CO2 hydrogenation to methanol has received considerable attention as an effective way to utilize CO2. DFT calculations were employed to investigate the methanol synthesis from CO2 and H2 on a model catalyst of molybdenum sulfide, a Mo6S8 cluster. The present study shows that a Mo6S8 cluster behaves differently from the Mo edge of MoS2, which is active for syngas conversion. MoS2 promotes C-O scission in HxCO intermediates, which leads to the formation of methane (or higher hydrocarbons). The CO2 hydrogenation to methanol is most likely to occur on the Mo6S8 cluster. The reaction undergoes the RWGS reaction by converting CO2 and H2 to CO and H2O via the HOCO intermediate. The formed CO is then hydrogenated to HCO radical and therefore methanol. The rate-limiting step for the overall conversion is CO hydrogenation to HCO. The corresponding barrier is 1 eV, which is much lower

Liu et al. than that of a Cu nanoparticle. Both the Mo and S sites participate in the reaction directly. Mo adsorbs CO2, CO, and CHxO, whereas S facilitates the H-H bond cleavage by stabilizing the dissociated hydrogen. According to our DFT results, Mo6S8 exhibits a behavior more complex than that of pure metal nanoparticles. The substoichiometric Mo6S8 cluster has a lower intrinsic activity than MoS2, despite the smaller size and the lower S/Mo ratio. Due to the overly strong interactions with adsorbates, MoS2 has poor selectivity to methanol synthesis from syngas mixtures. The unique atomic and electronic structure of the Mo6S8 cluster results in generally moderate interactions with the reactant and intermediates involved in methanol synthesis, that is, CO2, CO, H, and CHxO. As a consequence, the C-O bond cleavage, which leads to hydrocarbon formation, is suppressed during the reaction sequence. Hence, the Mo6S8 cluster is predicted to be a more selective, albeit less active, catalyst for methanol formation than MoS2. Given that, decreasing the nonmetal concentration or the size of the metal compounds does not necessarily lead to an increased activity. Depending on the actual structures of the metal compounds and the nonmetal/metal ratio, the effect may lead to the different catalytic activities from the bulk materials. Acknowledgment. This research was carried out at Brookhaven National Laboratory under Contract DE-AC02-98CH10886 with the U.S. Department of Energy, Division of Chemical Sciences. The DFT calculations were carried out at Centers for Functional Nanomaterials at Brookhaven National Laboratory and the National Energy Research Scientific Computing (NERSC) Center, which is supported by the Office of Science of the U.S. Department of Energy under Contract no. DE-AC02-05CH11231. References and Notes (1) Wasmus, S.; Kuver, A. J. Electroanal. Chem. 1999, 14. (2) Kowal, A.; Li, M.; Shao, M.; Sasaki, K.; Vukmirovic, M. B.; Zhang, J.; Marinkovic, N. S.; Liu, P.; Frenkel, A.; Adzic, R. R. Nat. Mater. 2009, 8, 325. (3) Song, C. Catal. Today 2006, 115, 2. (4) Fisher, F. The coVersion of coal into oils; Ernest Benn Ltd.: London, 1925. (5) Xu, M.; Gines, M. J. L.; Hilmen, A.-M.; Stephens, B. L.; Iglesia, E. J. Catal. 1997, 171, 130. (6) Spivey, J. J.; Egbebi, A. Chem. Soc. ReV. 2007, 36, 1514. (7) Liu, X.; Lu, G. Q.; Yan, Z.; Beltramini, J. Ind. Eng. Chem. Res. 2003, 42, 6518. (8) Subramani, V.; Gangwal, S. K. Energy Fuels 2008, 22, 814. (9) He, J.; Zhang, W.-n. J. Zhejiang UniV. Sci. 2008, 9, 714. (10) Herman, R. G. Catal. Today 2000, 55, 233. (11) Lee, J. S.; Kim, S.; Lee, K. H.; Nam, I.; Chung, J. S.; Kim, Y. G.; Woo, H. C. Appl. Catal., A 1994, 110, 11. (12) Huang, M.; Cho, K. J. Phys. Chem. C 2009, 113, 5238. (13) Liu, P.; Rodriguez, J. A.; Muckerman, J. T. J. Chem. Phys. 2004, 121, 10321. (14) Vin˜es, F.; Rodriguez, J. A.; Liu, P.; Illas, F. J. Catal. 2008, 260, 103. (15) Umarji, A. M.; Rao, G. V. S.; Janawadkar, M. P.; Radhakrishnan, T. S. J. Phys. Chem. Solids 1980, 41, 421. (16) Mancour-Billah, A.; Chevrel, R. J. Solid State Chem. 2003, 170, 281. (17) Paskach, T. J.; Schrader, G. L.; McCarley, R. E. J. Catal. 2002, 211, 285. (18) Bertram, N.; Kim, Y. D.; Gantefor, G.; Sun, Q.; Jena, P.; Tamuliene, J.; Seifert, G. Chem. Phys. Lett. 2004, 396, 341. (19) Lightstone, J. M.; Patterson, M. J.; White, M. G. Chem. Phys. Lett. 2005, 413, 429. (20) Patterson, M. J.; Lightstone, J. M.; White, M. G. J. Phys. Chem. A 2008, 112, 12011. (21) Lightstone, J. M.; Patterson, M. J.; Liu, P.; Lofaro, J. C.; White, M. G. J. Phys. Chem. C 2008, 112, 11495. (22) Enyashin, A. N.; Gemming, S.; Seifert, G. Eur. Phys. J. Special Top. 2007, 149, 103. (23) Lightstone, J. M.; Mann, H. A.; Wu, M.; Johnson, P. M.; White, M. G. J. Phys. Chem. B 2003, 107, 10359.

Methanol Synthesis from H2, CO2 on a Mo6S8 Cluster (24) Enyashin, A. N.; Gemming, S.; Bar-Sandan, M.; Popovits-Biro, R.; Hong, S. Y.; Prior, Y.; Tenne, R.; Seifert, G. Angew. Chem., Int. Ed. 2007, 46, 623. (25) Zeng, T.; Wen, X. D.; Li, Y. W.; Jiao, H. J. Chem. Phys. B 2005, 109, 13704. (26) Seifert, G.; Tamuliene, J.; Gemming, S. Comput. Mater. Sci. 2006, 35, 316. (27) Patterson, M. J.; White, M. G. to be published. (28) Valden, M.; Lai, X.; Goodman, D. W. Science 1998, 281, 1647. (29) Falsig, H.; Hvolbæk, B.; Kristensen, I. S.; Tao, J.; Bligaard, T.; Christensen, C. H.; Nørskov, J. K. Angew. Chem., Int. Ed. 2008, 47, 4835. (30) Lee, S.; Molina, L. M.; Lopez, M. J.; Alonso, J. A.; Hammer, B.; Lee, B.; Seifert, S.; Winans, R. E.; Elam, J. W.; Pellin, M. J.; Vajda, S. Angew. Chem., Int. Ed. 2009, 48, 1467. (31) Rodriguez, J. A.; Liu, P.; Hrbek, J.; Evans, J.; Perez, M. Angew. Chem., Int. Ed. 2007, 46, 1329. (32) Blo¨chl, P. E. Phys. ReV. B 1994, 50, 17953. (33) Kresse, G.; Hafner, J. Phys. ReV. B 1993, 47, 558. (34) Kresse, G.; Furthmu¨ller, J. Phys. ReV. B 1996, 54, 11169. (35) Perdew, J. P.; Burke, K.; Ernzerhof, M. Phys. ReV. Lett. 1996, 77, 3865. (36) Monkhorst, H. J.; Pack, J. D. Phys. ReV. B 1976, 13, 5188. (37) Delley, B. J. Chem. Phys. 1990, 92, 508. (38) Delley, B. J. Chem. Phys. 2000, 113, 7756. (39) Halgren, T. A.; Lipscomb, W. N. Chem. Phys. Lett. 1977, 49, 225. (40) Yvon, K. Current Topics in Materials Science; North-Holland Publishing,: Amsterdam, 1979; Vol. 3.

J. Phys. Chem. A, Vol. 114, No. 11, 2010 3895 (41) Seifert, G.; Tamuliene, J.; Gemming, S. Comput. Mater. Sci. 2006, 35, 316. (42) Murugan, P.; Kumar, V.; Kawazoe, Y.; Ota, N. J. Phys. Chem. A 2007, 111, 2778. (43) Murugan, P.; Kumar, V.; Kawazoe, Y.; Ota, N. Phys. ReV. B 2005, 71, 063203. (44) Bell, A. T. Science 2003, 229, 1688. (45) Hammer, B.; Nørskov, J. K. AdV. Catal. 2000, 45, 71. (46) Greeley, J.; Nørskov, J. K.; Mavrikakis, M. Annu. ReV. Phys. Chem. 2002, 53, 319. (47) Popov, I.; Kunze, T.; Semming, S.; Seifert, G. Eur. Phys. J. D 2007, 45, 439. (48) Popov, I.; Semming, S.; Seifert, G. Phys. ReV. B 2007, 75, 245436. (49) Subramani, V.; Gangwal, S. K. Energy Fuels 2008, 22, 814. (50) Osaki, T.; Narita, N.; Horiuchi, T.; Sugiyama, T.; Masuda, H.; Suzuki, K. J. Mol. Catal. A: Chem. 1997, 125, 63. (51) Waugh, K. C. Catal. Today 1992, 15, 51. (52) Weigel, J.; Koeppel, R. A.; Baiker, A.; Wokaun, A. Langmuir 1996, 12, 5319. (53) Nørskov, J. K.; Bligaard, T.; Logadottir, A.; Bahn, S.; Hansen, L. B.; Bollinger, M.; Bengaard, H.; Hammer, B.; Sljivancanin, Z.; Mavrikakis, M.; Xu, Y.; Dahi, S.; Jacobsen, C. J. H. J. Catal. 2002, 209, 275. (54) Rohmer, M.-M.; Be´nard, M.; Poblet, J.-M. Chem. ReV. 2000, 100, 495. (55) Liu, P.; Rodriguez, J. A. J. Chem. Phys. 2007, 126, 164705. (56) Barrio, L.; Liu, P.; Rodrı´guez, J. A.; Campos-Martı´n, J. M.; Fierro, J. L. G. J. Chem. Phys. 2006, 125, 164715.

JP906780A