Surface Science 605 (2011) 513–519

Contents lists available at ScienceDirect

Surface Science

j ourna l homepage: www.e lsev ie r.com/ locate /susc

Chemisorption on cobalt surfaces: The effect of subsurface rhenium atoms fromquantum chemical cluster model calculations

Vebjørn Bakken 1,2, Erling Rytter 3, Ole Swang ⁎,2

SINTEF Materials and Chemistry, Department of Process Chemistry, P. O. Box 124 Blindern, N-0314 Oslo, Norway; Statoil R&D, N-7465 Trondheim, Norway

⁎ Corresponding author. Tel.: +47 98243934; fax: +E-mail address: ole.swang@sintef.no (O. Swang).

1 Present address: Faculty administration, Faculty ofSciences, University of Oslo, P. O. Box 1032 Blindern, N-

2 SINTEF.3 Statoil.

0039-6028/$ – see front matter © 2010 Elsevier B.V. Aldoi:10.1016/j.susc.2010.12.008

a b s t r a c t

a r t i c l e i n f o

Article history:Received 6 September 2010Accepted 7 December 2010Available online 17 December 2010

Keywords:Ab initio quantum chemical methods andcalculationsCatalysisChemisorptionSurface chemical reactionCobaltRhenium

We have calculated chemisorption energies for different sorbates on cluster models for a number of sites onpure and subsurface rhenium-doped cobalt surfaces. Bonding energies follow the trend waterbCObpro-pylbmethylbhydrogenbhydroxyl, and are in good agreement with experimental results where available. Theresults indicate that for single-bond radicals (hydrogen, alkyl, hydroxyl), rhenium inclusion stabilizes thechemisorbed species. Further, the stabilization leads to a greater number of sites being energetically close tothe most stable ones, possibly enhancing surface mobility of chemisorbed species. Hydrogen is less stabilizedby rhenium substitution compared to propyl, indicating a possible mechanism for the greater yield of long-chained hydrocarbons afforded by rhenium-doped catalysts. For carbon monoxide, the results are lessconclusive as rhenium substitution does not influence chemisorption energies so strongly.

47 22067350.

Mathematics and the Natural0315 Oslo, Norway.

l rights reserved.

© 2010 Elsevier B.V. All rights reserved.

1. Introduction

Metal clusters have proven to be useful models for surface and bulkphenomena both in theoretical [1] and experimental [2–4] investiga-tions. In particular smaller metal clusters are well suited for quantummechanical studies, as indicated by the amount of existing literature onthe subject [5]. It is worth noting that the quantum chemical literatureon cluster studies mostly deals with copper [6] and nickel [7], and thatrelatively little has been published on cobalt [8,9].

Some of the most important heterogeneous catalysts are based ontransition metals. They may consist of a single metal or an alloy. In thelatter case, one speaks of a bimetallic or promoted catalyst. Statoil [10]has developed a bimetallic catalyst for Fischer–Tropsch (F–T) synthesis,in which cobalt promoted with rhenium, presumably an alloy, issupported ona stabilizedγ-alumina. The F–Tprocesswasfirst describedmore than 80 years ago [11] and implemented on a full industrial scalemore than 70 years ago [12]. F–T synthesis has seen renewed interestlately, at least partially due to high crude oil prices. Nevertheless, theprocess is still not fully understood; the role of rheniumor other dopantsbeing one of the points still under debate.

Elsewhere we have reported the interaction between rhenium andcobalt through quantum chemical cluster model calculations [13,14]

and Low Energy Ionic Scattering Spectroscopy (LEISS) [14]. Ourresults, which corroborate earlier computational semiempiricalresults published by Ruban et al. [15], strongly indicate that rheniumpreferentially occupies subsurface sites. An interesting question is,then, how subsurface rhenium influences the reactivity of the cobaltsurface. In the present contribution, we describe quantum chemicalcluster model studies of chemisorption of a number of species knownto be involved in elementary steps of F–T synthesis. The focus of thisinvestigation is not the specific clusters as such, but rather to uncovertrends in chemisorption energies for different sorption sites assubsurface rhenium atoms are introduced.

Hydrogen, one of the reactants, has been shown to chemisorbdissociatively on cobalt surfaces forming atoms; methyl is a postulatedintermediate in the reaction; propyl represents the growing chain and isa direct precursor for propene, a by-product; CO is a reactant and H2O isa primary product. Chemisorption energies for these species on purecobalt and bimetallic cobalt/rhenium clusters have been computed toelucidate the effect of rhenium doping upon chemisorption energies.

2. Computational details

All calculations were carried out using the PW91 [16] densityfunctional as implemented in the Amsterdam Density Functionalpackage (ADF) [17–19]. The inner core orbitals (up to and including3p for Co and 4f for Re), were kept frozen, while the valence orbitalswere described by the TZP Slater-type basis sets built into the ADFpackage. Relativistic effects were taken into account using the ZORAmethod [20], and spin-orbit coupling has thus not been included in

Co20 Co19Re Co18Re2

Fig. 1. 20-atom cluster models. Cobalt atoms are white, rhenium black.

4 Substitution energies calculated relatively to the relevant clusters and single metalatoms in vacuum.

514 V. Bakken et al. / Surface Science 605 (2011) 513–519

the description of the electronic structures. Reported partial atomiccharges are computed using the Hirshfeld formalism [21].

20-atom clusters have been used to model the pure and rhenium-doped cobalt surfaces; see Fig. 1. Ground-state spin multiplicities ofthe clusters were taken as reference states. For calculating chemi-sorption energies of sorbates that may be called “one-electronligands” in organometallic parlance, the total spin polarization vasreduced by 2 compared to the sum of spin polarizations for the twoseparated systems (cluster and sorbate, the latter in these cases beinga doublet radical). In the calculations involving closed-shell sorbates,spin polarization was kept at the ground-state value for thecorresponding bare cluster. Literature results [22] suggest thatsorption of small ligands on metal surfaces may be described aslocal phenomena. For clusters described by all-electron GGA DFTcalculations, absolute binding energies for organic fragments areusually converged to within some 30 kJ/mol for clusters of evensmaller size than the one used here. The primary focus in the presentwork is differences (trends) in chemisorption energies, and these willbe more reliable than the absolute values due to cancellation of error.

3. Results and discussion

3.1. The clusters

Three 20-atom clusters, Co20, Co19Re, and Co18Re2, shown in Fig. 1,have been used to model different surface sites. Between them, theyoffer models for different sites of varying coordination and proximityof subsurface rhenium. Themetal moieties of the FT catalyst are highlyirregular, showing entwined oxide and metal regions and possiblyeven mixed phases (cobalt aluminates) [23]. While the metalagglomerates of the working catalyst contain much more than 20atoms, it has been shown that single chemisorption events may beregarded as quite local phenomenae. Chemisorption energies andgeometries of chemisorbed species on metal surfaces have beensuccessfully determined using computational methods and quitesmall cluster models [7]. Reboredo and Galli [24] found thatchemisorption energies for CO on cobalt clusters vary monotonicallyand predictably with cluster size down to 13 atoms. We conclude thatthe 20-atom clusters used should be entirely adequate for describingthe trends that are the subject of the present contribution.

The cluster geometries were constructed based on our previousfindings for smaller cobalt/rhenium clusters [25]. For reasons ofcomputational feasibility, it proved necessary to use clusters thatmaintain a modicum of symmetry (i.e. at least Cs-symmetry) evenafter adsorption. The chosen clusters have D3h-, C3v- and D3h-symmetry, respectively, yielding C3v- and Cs-symmetry upon adsorp-tion. Prior to the sorption calculations, each cluster was fullyoptimized with regards to both spin polarization and geometrywithin the given symmetry. The selected cluster geometries likelyconstitute low-energy structures, though they are not necessarily

global minima for their stoichiometries. For example for Co20, atwisted configuration having D3-symmetry was found to be approx-imately 100 kJ/mol lower in energy than the D3h-structure. However,the former structure was deemed unsuitable for the currentinvestigation, due to the fact that the calculations of adsorbed specieswould have to be run without any symmetry. Our subject here isanyway not the clusters per se, but rather the qualitative changes insurface chemistry induced by rhenium doping.

The total partial Hirshfeld charge on rhenium is the same forCo19Re and Co18Re2: 0.2 e−/atom in the former case and 0.1 e−/atomin the latter. This should be compared to the partial charges on thetwo central atoms in the pure cobalt cluster which are essentially zero(−0.03 e−). The relatively small change correlates well with the factthat the two elements have quite similar electronegativities.

The spin polarization (defined as the difference between thenumbers of electrons of majority and minority spin, N↑–N↓) of themost stable electronic state decreases by 6 units upon substitution ofone cobalt atom with rhenium, and additionally 2 when the secondsubstitution takes place. This downcoupling of spins is in line withresults from earlier cluster studies [8,13,14]. Additionally, thesubstitution of cobalt by rhenium stabilizes the cluster by about150 kJ/mol per rhenium atom.4 The fact that the stabilization is almostidentical for the first and second Re-atom indicates that the 20-atomclusters are large enough to model the energetics of incorporation ofup to two rhenium atoms. This as opposed to the 14-atom clusterspreviously investigated [13].

Rhenium has a larger metallic radius than cobalt, and thereplacement of cobalt with rhenium at the two central positions inthe cluster unsurprisingly expands the cluster. Looking at the distancebetween the two interior positions, Co20 yields a Co–Co distance of2.38 Å, whereas the Re–Re distance in Co18Re2 is 2.49 Å, although thetrend is broken by Co19Re having a central Co–Re distance of 2.36 Å.The effect on the three cobalt atomsmaking up our chosen adsorptionsite is that the Co–Co distance is increased slightly from 2.39 Å in thepure cobalt cluster, to 2.40 Å/2.41 Å and 2.45 Å in Co19Re and Co18Re2,respectively. Experimentally, the closest bulk interatomic distances inthe pure metals are 2.5 and 2.75 Å for cobalt and rhenium,respectively.

As the different sorbates are investigated, the clusters are keptfrozen at their optimized geometries. Relaxing the cluster geometriesupon chemisorption would lead to greater changes compared to thereal catalyst, where the metal aggregates are much larger andsupported by the oxide. Hence, we hold that relaxation of the clusterswould waste computational resources without improving the predic-tive power of the calculations. One referee has raised the question ofsorbate-induced segregation, i.e. the possibility that sorption couldlead to rhenium being pulled out to the surface. Such a situation could

Table 1Adsorption of hydrogen to the different sites on Co20, Co19Re, and Co18Re2; adsorptionenergies, bond distance, partial charges and spin polarization are listed. All clusters arekept frozen at their respective optimized geometries (without sorbate).

Hydrogen

Cluster Site Eads (kJ/mol) rCo–H (Å) qH (e) μtot

Co20 3-fold top 216 1.79 −0.07 37Co20 2-fold top 217 1.67 −0.07 37Co20 1-fold top 189 1.54 −0.10 37Co20 2-fold side 253 1.70 −0.08 37Co20 3-fold side 251 1.81 −0.07 37Co19Reα 3-fold top 244 1.81 −0.08 31Co19Reα 2-fold side 259 1.66 −0.07 31Co19Reα 3-fold side 256 1.79 −0.07 31Co19Reβ 3-fold top 245 1.83 −0.09 31Co19Reβ 2-fold side 241 1.68 −0.07 31Co19Reβ 3-fold side 247 1.79 −0.07 31Co18Re2 3-fold top 260 1.83 −0.09 29Co18Re2 2-fold side 250 1.67 −0.07 29Co18Re2 3-fold side 254 1.80 −0.07 29

515V. Bakken et al. / Surface Science 605 (2011) 513–519

clearly not be described by the reported calculations. However, it ishighly unlikely in the case of rhenium in cobalt: CO, the sorbent mostlikely to cause segregation, has very similar sorption energies on thetwo pure metals [9,26], and the preference for rhenium to avoid thesurface appears to be very strong [14,15].

3.2. Chemisorption

In the following, results are analyzed first and foremost in terms ofenergies and geometries. Population analyses gave no conclusiveresults, so we have chosen to restrict ourselves to observables.

3.2.1. HydrogenFig. 2 shows hydrogen chemisorption on the 3-fold site. Co19Re,

being the least symmetric of the clusters, may be a model for twodifferent 3-fold end sites, depending on the placement of rhenium. Inthe following, the sorption position closest to rhenium will bedenoted α, and the opposite side β. From the data in Table 1, wenote that sorption energies increase with the number of subsurfacerhenium atoms. For Co19Re, the energy differences between thesorption on α and β positions are small.

The Co–H distances increase slightly upon introduction ofrhenium, despite the increase in bonding strength. The partialelectronic charge on hydrogen is negligible and does not changeappreciably upon rhenium substitution, indicating that the bondbetween cluster and sorbate is mainly covalent.

Greeley and Mavrikakis [27] calculated the sorption energy ofhydogen on a pure Co(0001) surface using a slab model, a 0.25monolayer coverage, and the same density functional (PW91) as usedhere, and report a chemisorption energy of 277 kJ/mol. Klinke andBroadbelt [1], using periodic calculations and a different functional,reported 279 kJ/mol for coverages of both 0.5 and 1.0 monolayers.These values are close to the 251 kJ/mol found on the most stable3-fold site on the presently used cluster. Incidentally, the experimen-tally measured energy on a pure surface is exactly 251 kJ/mole [28].Different adsorbates show different preferences as to the most stablesites of chemisorption. For hydrogen, the most stable sites on thedifferent clusters are shown in Fig. 3. As can be seen in Table 1, only thepure cobalt cluster gives rise to 2-fold and 1-fold top-site chemisorp-tion. Attempts to locate these sites for Co19Re resulted in the hydrogencrawling over the edge and down to either a 2-fold or a 3-fold site onthe side of the cluster. For Co18Re2, the hydrogen instead moved in tothe 3-fold top-site. As discussed earlier, chemisorption at the 3-fold topsite is markedly stabilized by the introduction of rhenium. The sidesites, on the other hand, are found to yield very similar adsorptionenergies across the clusters investigated. On Co(0001) single crystalsurfaces, hydrogen and alkyls chemisorb at 3-fold hollow sites [29]. Thisalso appears to be the case on the areas of the present cluster modelwith the least surface curvature. While our approach describes asituation with low surface coverage, we note that literature results [30]

3-fold 2-fol

Fig. 2. Adsorption of hydrogen to the three different top-sites of Co20. Cobalt ato

indicate that there is one hydrogen atom per cobalt surface atom atsaturation.

If hydrogen is allowed to adsorb on the preferred sites, themaximum sorption energies on the three different clusters differ by7 kJ/mol only. A prominent effect of rhenium substitution is that itchanges the preferred adsorption site. Additionally, the number ofstable adsorption sites for hydrogen seems to increase for the clusterscontaining rhenium; allowing higher mobility.

3.2.2. Methyl and propylMethyl and propyl adsorbates show behaviour rather similar to

that of hydrogen, see Tables 2 and 3. An important difference is thatpropyl adsorbed in a three-fold site is not stable. It was necessary topartially freeze the carbon atom coordinating the metal cluster tokeep it at the three-fold site. The vertical distance between the alpha-carbon and the metal cluster was allowed to vary along with thepositions of all the other atoms in the ligand. The main trend formethyl and propyl is the same as for hydrogen: sorption energiesincrease upon rhenium substitution. This is to be expected, as theligands share similar frontier orbital structures. Top-sites and side-sites are in competition, and again the former seem to be stabilized bythe addition of rhenium, while the latter is less influenced as can beseen from Table 2. There is a somewhat arbitrary distinction betweenthe 1-fold top and 1-fold side sites: The latter has methyl bonded to asingle top-site cobalt atom, but the carbon atom of methyl lies prettymuch in the same plane as the three top-site cobalt atoms, thuspointing “sideways”. Fig. 4 shows the most stable adsorption site foreach cluster, but as can be seen from Table 2 all clusters havealternative sites with very similar energies. For most of the adsorptionsites, the methyl group may be turned 180° to yield another stable

d 1-fold

ms are shown as large white spheres and hydrogen as a small white sphere.

Co20-H Co19Reα-H Co19Reβ-H Co18Re2-H

Fig. 3. The preferred adsorption sites for hydrogen on the different clusters. Cobalt atoms are shown as large white spheres, rhenium atoms as black spheres and hydrogen atoms assmall white spheres.

516 V. Bakken et al. / Surface Science 605 (2011) 513–519

adsorption mode. The differences in energy between these conforma-tions are generally of the order 1–3 kJ/mol. Even without havinginvestigated rotational barriers, we conclude that the methyl groupswill rotate freely at F–T reaction temperatures. Our most stablemethyl species bind with 185 kJ/mol, a value that may be comparedwith the 193 kJ/mol found by Zuo et al. [31] in their periodic DFT studyfor methyl at a hollow site of a Co(111) surface.

Despite significant changes in adsorption energy as rhenium isintroduced, the Co–C bond distances are almost constant, changing byno more than a few hundredths of an Ångström. The key parametergoverning the bond length appears to be the number of coordinatingcobalt atoms, rather than the proximity of rhenium. As expected, thefewer cobalt atoms the methyl (carbon atom) coordinates to, theshorter the carbon–cobalt bond length becomes. Furthermore, thecarbon atom is found to have a partial atomic charge of between−0.15 and −0.20 e−, but this is partially cancelled by small negativecharges on the hydrogen atoms. Overall, the methyl group has a smallpartial negative charge between−0.05 and−0.10 e−.

The adsorption energies of methyl are typically 50–60 kJ/mollower than for hydrogen. Moving on to propyl, the energies decreaseby another 20 kJ/mol or so, see Table 3. Propyl is larger and muchmore flexible than the other sorbates considered here, yielding arather flat potential energy surface that is less well-behavednumerically. Only top-sites were examined, though for some of theclusters the 1-fold top is very close to being a 1-fold side (as discussedfor methyl above). The 2-fold top site is significantly stabilized by theintroduction of rhenium, making this the preferred adsorption site in

Table 2Adsorption of methyl to the different sites on Co20, Co19Re, and Co18Re2; adsorptionenergies, bond distance, partial charges and spin polarization are listed. All clusters arekept frozen at their respective optimized geometries (without sorbate).

Methyl

Cluster Site Eads (kJ/mol) rCo–C (Å) qC (e) μtot

Co20 3-fold top 154 2.23 −0.16 37Co20 2-fold top 166 2.12 −0.17 37Co20 1-fold top 185 1.96 −0.20 37Co20 3-fold side 185 2.26 −0.16 37Co19Reα 3-fold top 188 2.25 −0.17 31Co19Reα 2-fold top 195 2.13 −0.17 31Co19Reα 3-fold side 195 2.23 −0.16 31Co19Reα 1-fold side 180 1.99 −0.18 31Co19Reβ 3-fold top 184 2.28 −0.18 31Co19Reβ 2-fold top 195 2.14 −0.18 31Co19Reβ 1-fold top 188 1.97 −0.21 31Co19Reβ 3-fold side 189 2.24 −0.16 31Co18Re2 3-fold top 200 2.27 −0.17 29Co18Re2 2-fold top 203 2.14 −0.18 29Co18Re2 3-fold side 195 2.23 −0.16 29Co18Re2 1-fold side 181 1.99 −0.19 29

all cases except the pure cobalt cluster, see Fig. 5. Again, it seemsreasonable to assume that propyl will twist and turn quite freely atrelevant temperatures. We note subtle signs of agostic interactions astheα carbon–hydrogen bonds are elongated by about 2 pm comparedto the situation in the free propyl radical, disregarding clusterstoichiometry. Such agostic interactions may point to a propensityfor termination to the olefin, which is a known by-product of the FTsynthesis.

The strongest sorption energies for methyl and propyl increase bysome 10–15 kJ/mol upon rhenium substitution. This could account forthe observation that rhenium-doped catalysts give a greater selectiv-ity to long-chained hydrocarbons, as chain termination becomes lessthermodynamically favourable — this would be the case for both theα-hydride addition and the β-hydride abstraction mechanisms, as thehydrogen sorption energy is less influenced by rhenium substitution.

3.2.3. HydroxylAs far as the three-fold site chemisorption is concerned, hydroxyl

breaks the trend of the previous sorbates by yielding reduced stabilityas the number of present rhenium atoms increases. However, asshown in Table 4, this changes as symmetry constraints are lifted, andwe will return to this shortly. Hydroxyl chemisorption at the 3-foldsite actually showed a decrease in adsorption energy for the rhenium-containing clusters. However, by lifting the C3v symmetry (reducing itto Cs), the by now prominent pattern of stabilization is restored, as canbe seen in Table 4. In particular, the 2-fold top-site benefits from thepresence of rhenium, and with the exception of Co20 it is the moststable site. For Co20 the 1-fold top site is more stable than the 2-fold,however, the highest adsorption energy is found for a 2-fold side-site.For each of the four cluster configurations, the most stable adsorption

Table 3Adsorption of propyl to the different sites on Co20, Co19Re, and Co18Re2; adsorptionenergies, bond distance, partial charges and spin polarization are listed. All clusters arekept frozen at their respective optimized geometries (without sorbate).

Cluster Site Eads (kJ/mol) rCo–C(α) (Å) qC(α) (e) μtot

Co20 3-fold topa 130 2.30 −0.10 37Co20 2-fold top 147 2.12 −0.10 37Co20 1-fold top 165 1.97 −0.13 37Co19Reα 3-fold topa 170 2.25 −0.11 31Co19Reα 2-fold top 174 2.13 −0.11 31Co19Reα 1-fold top 162 1.98 −0.12 31Co19Reβ 3-fold topa 171 2.27 −0.11 31Co19Reβ 2-fold top 174 2.13 −0.12 31Co19Reβ 1-fold top 165 1.98 −0.14 31Co18Re2 3-fold topa 176 2.26 −0.11 29Co18Re2 2-fold top 180 2.14 −0.12 29Co18Re2 1-fold top 163 1.97 −0.13 29

a The three-fold adsorption was achieved by partially freezing C(α).

Co20-CH3 Co19Reα-CH3 Co19Reβ-CH3 Co18Re2-CH3

Fig. 4. The preferred adsorption sites for methyl on the different clusters. Cobalt atoms are shown as large white spheres, rhenium atoms as black spheres, carbon as small black andhydrogen atoms as small white spheres.

Co20-C3H7 Co19Reα-C3H7 Co19Reβ-C3H7 Co18Re2-C3H7

Fig. 5. The preferred adsorption sites for propyl on the different clusters. Cobalt atoms are shown as large white spheres, rhenium atoms as black spheres, carbon as small black andhydrogen atoms as small white spheres.

Table 4Adsorption of hydroxyl to the different sites on Co20, Co19Re, and Co18Re2; adsorptionenergies, bond distance, partial charges and spin polarization are listed. All clusters arekept frozen at their respective optimized geometries (without sorbate).

Hydroxyl

Cluster Site Eads (kJ/mol) rCo–O (Å) qO (e) μtot

Co20 3-fold top 329 2.10 −0.26 37Co20 2-fold top 334 2.03 −0.29 37Co20 1-fold top 339 1.81 −0.31 37Co20 2-fold side 349 2.01 −0.26 37Co19Reα 3-fold top 303 2.10 −0.25 31Co19Reα 2-fold top 365 2.00 −0.26 31Co19Reα 1-fold top 332 1.82 −0.32 31Co19Reα 2-fold side 363 1.97 −0.24 31Co19Reβ 3-fold top 291 2.10 −0.25 31Co19Reβ 2-fold top 370 2.01 −0.27 31Co19Reβ 1-fold top 352 1.81 −0.32 31Co19Reβ 2-fold side 339 2.01 −0.26 31Co18Re2 3-fold top 267 2.09 −0.24 29Co18Re2 2-fold top 381 2.02 −0.27 29Co18Re2 1-fold top 349 1.82 −0.33 29Co18Re2 2-fold side 357 2.00 −0.26 29

517V. Bakken et al. / Surface Science 605 (2011) 513–519

complex is depicted in Fig. 6. Hydroxyl yields the highest sorptionenergies of all the sorbates considered in this study, as well as thelargest difference between clusters (32 kJ/mol).

3.2.4. Carbon monoxideCarbon monoxide is a qualitatively different sorbate, with its well

known σ-donor/π-acceptor properties. For the three-fold top-site,chemisorption energies generally seem to decrease as the number ofrhenium atoms increase, see Table 5. This trend, however, is notmonotonous, as the adsorption energy of CO is larger for Co18Re2 thanfor Co19Reβ, but less than for Co19Reα. In other words, there is asignificant difference between the two configurations of Co19Re whenabsorbing carbon monoxide, once more setting this sorbate apartfrom the others. For all four clusters, the most stable adsorption modeis found to be coordination to a single top-site cobalt atom, see Table 5and Fig. 7. This is in line with literature results [24]. We note that theCo–C distance found for the most stable structures (1.78 Å) is inexcellent agreement with experimental LEED data for a single crystalsurface at high coverage [32]. There is no unambiguous trend in thechemisorption energy as more rhenium is introduced. Themost stableadsorption is found for Co19Reβ and the least stable for Co18Re2, the

Co20-OH Co19Reα-OH Co19Reβ-OH Co18Re2-OH

Fig. 6. The preferred adsorption sites for hydroxyl on the different clusters. Cobalt atoms are shown as large white spheres, rhenium atoms as black spheres, oxygen atoms as smallshaded spheres and hydrogen as small white spheres.

518 V. Bakken et al. / Surface Science 605 (2011) 513–519

two being separated by 20 kJ/mol. Investigating the partial atomicspin densities of the coordinating cobalt atoms, it is found thatchemisorption of CO gives a significantly larger decrease than theother sorbates considered here. This observation is also in line with aσ-donor picture.

3.2.5. WaterThe final sorbate examined is water, and the results are given in

Table 6. We have only found stable structures involving oxygencoordinating to a single cobalt atom. At around 50 kJ/mol, the sorptionenergies of water are small enough that nondissociative sorption ofwater may be regarded as uninteresting at the temperaturesencountered in Fischer–Tropsch synthesis. Water formed in thereaction will spontaneously leave the surface. This result is in linewith experiment [33]. Sorption on the “backside” of the Co19Re clustergives some 7 kJ/mol higher sorption energy than for the other threeclusters, which yield energies equal to within 1 kJ/mol. Populationanalysis reveals that the water molecule has an overall positive partialcharge of around 0.2 e−, not surprisingly indicating that electrondensity in this case is donated from the sorbate to the metal cluster.

4. Conclusions

Upon substitution of cobalt with rhenium, the ground statemultiplicities of the clusters decrease significantly, in line with earlier

Table 5Adsorption of carbon monoxide to the different sites on Co20, Co19Re, and Co18Re2;adsorption energies, bond distance, partial charges and spin polarization are listed. Allclusters are kept frozen at their respective optimized geometries (without sorbate).

Carbon monoxide

Cluster Site Eads (kJ/mol) rCo–C (Å) qC (e) μtot

Co20 3-fold top 116 2.04 0.00 38Co20 2-fold top 134 1.95 0.02 38Co20 1-fold top 163 1.78 0.04 38Co19Reα 3-fold top 110 2.02 −0.01 32Co19Reα 2-fold top 132 1.93 0.02 32Co19Reα 1-fold top 161 1.80 0.04 32Co19Reβ 3-fold top 92 2.01 −0.01 32Co19Reβ 2-fold top 141 1.92 0.01 32Co19Reβ 1-fold top 173 1.78 0.04 32Co18Re2 3-fold top 103 1.99 −0.02 30Co18Re2 2-fold top 138 1.91 0.01 30Co18Re2 1-fold top 153 1.80 0.04 30

results. The present cluster calculations indicate that the addition ofsubsurface rhenium increases the chemisorption energy of hydrogen,methyl, propyl and hydroxyl on the terminal three-fold sites.Rhenium substitution has a smaller effect on the sorption energieson other sites. This results in a situation where there exist a number ofsites with sorption energies very close to the ground state. One mightspeculate that subsurface rhenium may result in greater mobility ofthese species on corrugated surfaces. A further indication of mobilityfor hydrogen, methyl, and propyl is the close similarity between theenergies for 1, 2, and-3-coordinated sorption sites.

Turning to the overall maximum sorption energies as a function ofrhenium substitution, we note that rhenium stabilizes the alkylsorbents more strongly compared to hydrogen. This could explainexperimental observation that rhenium-doped catalysts give a higheryield of long-chained hydrocarbons.

The sorption energies of carbon monoxide are much lessinfluenced by the cluster stoichiometry. It shows a strong tendencyto chemisorb on top of single metal atoms of low coordination, in linewith earlier findings. The adsorption energy of water is small for allsites and clusters. It always prefers a 1-fold coordination.

As only individual adsorbed species have been investigated, wemight speculate that coordination numbers for sorbates on a crowdedsurface will be smaller than what is observed here. Assuming thatsorbate–sorbate interactions are moderate, the present results indicatethat all sorbates may assume 1-fold coordination at high coverage.

Fig. 7. The preferred adsorption site for carbon monoxide on all clusters. Metal atomsare shown as large white spheres, carbon atoms as small black spheres and oxygenatoms as small shaded spheres.

Table 6Adsorption of water to Co20, Co19Re, and Co18Re2; adsorption energies, bond distance,partial charges and spin polarization are listed. All clusters are kept frozen at theirrespective optimized geometries (without sorbate).

Water

Cluster Site Eads (kJ/mol) rCo–O (Å) qO (e) μtot

Co20 1-fold top 48 2.17 −0.17 38Co19Reα 1-fold top 49 2.16 −0.18 32Co19Reβ 1-fold top 57 2.14 −0.17 32Co18Re2 1-fold top 50 2.18 −0.18 30

519V. Bakken et al. / Surface Science 605 (2011) 513–519

Acknowledgements

The authors wish to thank Professor Unni Olsbye for valuablediscussions. A generous grant of computer time from the NorwegianResearch Council (NOTUR project, account No. NN2923K) and aresearch grant from Statoil are both gratefully acknowledged. V.B.thanks the VISTA foundation for a postdoctoral fellowship.

References

[1] G. Pacchioni, Heterog. Chem. Rev. 2 (1995) 213;R. Hoffmann, Angew. Chem. Int. Ed. Engl. 26 (1987) 846;L.J. Broadbelt, R.Q. Snurr, Appl. Catal. A 200 (2000) 23.

[2] M.B. Knickelbein, Ann. Rev. Phys. Chem. 50 (1999) 79.[3] K.O. Fossan, E. Uggerud, Dalton Trans. (2004) 892;

M.S. Ford, M.L. Anderson, M.P. Barrow, D.P. Woodruff, T. Drewello, P.J. Derrick, S.R.Mackenzie, Phys. Chem. Chem. Phys. 7 (2005) 975;P.B. Armentrout, Ann. Rev. Phys. Chem. 52 (2001) 423.

[4] G.B. Sergeev, J. Nanoparticle Res. 5 (2003) 529;W.E. Klotzbucher, M.A. Petrukhina, A.V. Nemukhin, A.Y. Ermilov, B.L. Grigorenko,Spectrochim. Acta A 57 (2001) 1093.

[5] A.M. Joshi, M.H. Tucker, W.N. Delgass, K.T. Thomson, J. Chem. Phys. 125 (2006)194707;T. Futschek, J. Hafner, M. Marsman, J. Phys. Condens. Matter 18 (2006) 9703;

M.P. Jigato, K. Somasundram, V. Termath, N.C. Handy, D.A. King, Surf. Sci. 380 (1997)83.

[6] A. Mattsson, I. Panas, P. Siegbahn, U. Wahlgren, H. Åkeby, Phys. Rev. B 36 (1987)7389.

[7] O. Swang, K. Faegri Jr., O. Gropen, U.Wahlgren, P. Siegbahn, Chem. Phys. 156 (1991)3798 and references therein.

[8] O. Swang, K. Faegri Jr., O. Gropen, U. Wahlgren, Int. J. Quantum Chem. 57 (1996)105.

[9] S.H. Ma, X.T. Zu, Z.Y. Jiao, X.Z. Zhang, J. Mol. Struct. THEOCHEM 864 (2008) 68.[10] D. Schanke, E. Rytter, F.O. Jaer, Stud. Surf. Sci. Catal. 147 (2004) 43.[11] V.F. Fischer, H. Tropsch, Brennst. Chem. 7 (1926) 97.[12] M.E. Dry, Catal. Today 71 (2002) 227.[13] V. Bakken, O. Swang, J. Chem. Phys. 128 (2008) 084712.[14] V. Bakken, E. Bergene, E. Rytter, O. Swang, Catal. Lett. 135 (2010) 21.[15] A.V. Ruban, H.L. Skriver, J.K. Norskov, Phys Rev. B 59 (1999) 15990.[16] J.P. Perdew, J.A. Chevary, S.H. Vosko, K.A. Jackson, M.R. Pederson, D.J. Singh, C.

Fiolhais, Phys. Rev. B 46 (1992) 6671.[17] G. te Velde, F.M. Bickelhaupt, S.J.A. van Gisbergen, C. Fonseca Guerra, E.J. Baerends,

J.G. Snijders, T. Ziegler, J. Comput. Chem. 22 (2001) 931.[18] C. FonsecaGuerra, J.G. Snijders, G. teVelde, E.J. Baerends, Theor. Chem. Acc. 99 (1998)

391.[19] ADF2005.01, SCM, Theoretical Chemistry, Vrije Universiteit, Amsterdam, The

Netherlands.[20] E. van Lenthe, A.E. Ehlers, E.J. Baerends, J. Chem. Phys. 110 (1999) 8943.[21] F.L. Hirshfeld, Theor. Chim. Acta 44 (1977) 129.[22] O. Swang, K. Faegri Jr., O. Gropen, U. Wahlgren, Int. J. Quantum Chem. 57 (1996)

105;I.V. Yudanov, R. Sahnoun, K.M. Neyman, N. Rösch, J. Chem. Phys. 117 (2002) 9887;M.B. Jensen, U. Olsbye, O. Swang, Chem. Phys. Lett. 432 (2006) 94.

[23] H. Schultz, Top. Catal. 26 (2003) 73.[24] F. Reboredo, G. Galli, J. Phys. Chem. B 110 (2006) 7979.[25] V. Bakken, O. Swang, J. Chem. Phys. 128 (2008) 084712.[26] C. Pauls, D. Przyrembel, K. Christmann, J. Phys. Chem. B 108 (2004) 14749.[27] J. Greeley, M. Mavrikakis, J. Phys. Chem. B 109 (2005) 3460.[28] K. Christmann, Surf. Sci. Rep. 9 (1988) 1.[29] X.-Q. Gong, R. Raval, P. Hu, J. Chem. Phys. 122 (2005) 024711.[30] A.Y. Khodakov, W. Chu, P. Fongarland, Chem. Rev. 107 (2007) 1692;

K.H. Ernst, E. Schwarz, K. Christmann, Chem. Phys. 101 (1994) 5388.[31] Z. Zuo, W. Huang, P. Han, Z. Li, Appl. Surf. Sci. 256 (2010) 5929.[32] J. Lahtinen, J. Vaari, K. Kauraala, E.A. Soares, M.A. van Hove, Surf. Sci. 448 (2000)

269.[33] L.S. Xu, Y.S. Ma, Y.L. Zhang, B.H. Chen, Z.F. Wu, Z.Q. Jiang, W.X. Huang, J. Phys.

Chem. C 114 (2010) 17023.