13310
2006, 110, 13310-13313 Published on Web 06/16/2006
On the Performance of Au(111) for Ethylene Epoxidation: A Density Functional Study Daniel Torres and Francesc Illas* Departament de Quı´mica Fı´sica & CeRQT, UniVersitat de Barcelona & Parc Cientı´ficde Barcelona, C/ Martı´ i Franque` s 1, E08028 Barcelona, Spain ReceiVed: April 27, 2006; In Final Form: June 6, 2006
This work presents a periodic density functional study of the epoxidation mechanism of ethylene on Au(111). It is found that, once atomic oxygen is adsorbed on the surface, partial oxidation to ethylene oxide becomes possible. Calculated transition state theory rate constants for the elementary steps involved in the reaction predict that the selectivity of Au(111) toward epoxide is of ∼40% in good agreement with recent experimental findings for styrene epoxidation on Au(111).
I. Introduction Ethylene partial oxidation to ethylene oxide (EO) is one of the most important catalytic processes in the chemical industry, because EO is a chemical intermediate in many synthetic routes in organic chemistry. In practice, the reaction is carried out on catalysts consisting of silver particles supported on alumina with added Cl to enhance the selectivity. Because of the relevance of this synthetic process, the literature is extensive, and we will just refer the reader to several review papers,1,2 to recent relevant experimental studies using model catalysts mainly based on the use of Ag(111) single-crystal surfaces and ultrahigh-vacuum conditions,3-5 or to pertinent theoretical studies.6-8 From the UHV model studies where styrene is employed as a model molecule for ethylene, there is compelling evidence that this reaction is kinetically controlled9 and that the mechanism involves an oxametallacycle intermediate (OMME) leading either to the epoxide or to the corresponding aldehyde, the former being the kinetically favored product, the latter the thermodynamically favored one leading finally to CO2 and H2O as combustion products.10,11 This mechanism has also been confirmed by density functional calculations carried out for ethylene epoxidation on slab models of Ag(111).6,7 The UHV studies also have proven that partial oxidation to epoxide is entirely due to chemisorbed oxygen atoms, and therefore, the first step in the mechanism involves molecular oxygen dissociation.12 In the absence of promoters, the selectivity of the Ag(111) surface toward the epoxide is in the 40-50% range. Nevertheless, the question regarding the rather unique properties of silver as a specific catalyst for the ethylene epoxidation reactions remains open and hot. In fact, recent experiments also using styrene indicate that, at low oxygen coverages, copper is also active in epoxidizing the olefinic bond, but indeed with ∼100% selectivity.13 Density functional calculations for the ethylene epoxidation on Ag(111) and Cu(111) are fully consistent with the experiments using styrene and in addition show that the origin of the superior selectivity of Cu lies in the different character of the transition state leading to EO, which is of early type for Ag and of late type for Cu.14 Indeed, this can be related to the relative stability of the OMME intermediate 10.1021/jp0625917 CCC: $33.50
with respect to the products. Following the suggestion that Au particles catalyze epoxidation reactions,15 Deng and Friend have very recently studied the possible activity of Au(111) in styrene epoxidation.16 These authors find that Au(111) is also capable of promoting olefin oxidation, provided that atomic oxygen is dosed onto the surface. The experiments were carried out using relatively low oxygen coverage (∼0.2 ML) and evidence that selectivity toward the desired product is of ∼50% and, hence, very similar to that exhibited by the Ag(111) surface. To further check the validity of the proposed molecular mechanism for ethylene epoxidation on these metal surfaces, we have carried out density functional calculations for the reaction on Au(111) between adsorbed atomic oxygen and adsorbed ethylene. It is found that the overall reaction energy profile for the molecular mechanism corresponding to ethylene epoxidation on Au(111) is very similar to that reported for the reaction on Ag(111) with a concomitant similar selectivity. II. Computational Approach Periodic density functional (DF) calculations have been carried out for the adsorption of atomic oxygen and ethylene on Au(111) and their reaction to produce either EO or acetaldehyde. The Au(111) surface was modeled by slabs containing four atomic layers with a full relaxation of the two outermost layers, and with the repeated slabs being separated by a vacuum width corresponding to five atomic layers. A p(2 × 2) supercell was employed to describe a 0.25 coverage of both reactants, which is similar to the experimental conditions. Several initial geometries for the reactants were considered the most stable one being described in Figure 1; this has been selected as the starting point for the partial oxidations reaction. The DF calculations were carried out within the usual KohnSham (KS) implementation and using the PW91 version of the generalized gradient approximation (GGA) for the exchangecorrelation potential.17 The inner electrons have been described by the projector augmented wave (PAW) method,18 and the KS valence states were expanded in a plane-waves basis set with a cutoff at 315 eV for the kinetic energy. The reciprocal space of the p(2 × 2) cell was described with a Monkhorst-Pack19 mesh © 2006 American Chemical Society
Letters
J. Phys. Chem. B, Vol. 110, No. 27, 2006 13311
Figure 1. Molecular arrangement of ethylene and oxygen coadsorbed on Au(111). Carbon atoms in are labeled C1 (top) and C2 (fcc).
with 7 × 7 × 1 k-points. The spurious energy contribution due to the dipole arising from the adsorption of reactants in one of the surfaces was removed using the standard dipole correction.21 The present settings ensure that the calculated energies are converged within 0.01 eV. All calculations were carried out using the VASP code,20-22 and transition-state (TS) structures were located through the climbing image-nudged elastic band method (cNEB),23-25 using four interpolated images along the minimal-energy pathway. All minima on the potential energy surface were relaxed until self-consistent forces are lower than 0.01 eV/Å and TS structures were fully characterized with a pertinent vibrational analysis through diagonalization of the Hessian matrix obtained by numerical difference of analytical gradients and making sure TS structures show a single normal mode associated with an imaginary frequency. Rate constants at 550 K for the reaction to EO or acetaldehyde were estimated from conventional transition-state theory assuming minimal loss of degrees of freedom of reactants, products, and TSssi.e., reducing the partition function to the vibrational components and taking into account thermal effects. Hence, for each reaction from OMME, one has
k)
( )( ) kBT h
qTS
qOMME
e(-Eact/kBT)
(1)
where kB is the Boltzmann constant, T the temperature, qTS and qOMME the vibrational partition function of TS leading either to the epoxide or acetaldehyde and oxametallacycle, respectively, and Eact the activation energy including the correction for the zero-point vibrational energies. III. Results and Discussion In the following, we will describe each one of the elementary steps involved in the ethylene partial oxidation. For completeness, the problem of molecular oxygen dissociation and of adsorbed atomic recombination on Au(111) is also discussed. The full energy profile including kinetic barrier is shown in Figure 2. Having atomic oxygen adsorbed on the metal surface is the main problem to using Au as the ethylene epoxidation agent because of the low activity of extended gold surfaces toward dissociation of molecular species.26 Consequently, molecular oxygen does not adsorb on Au(111) over a wide range of temperatures. This is consistent with a calculated endothermic binding energy of 0.47 eV. However, atomic oxygen is known to be readily adsorbed on Au(111),16 as also shown by different theoretical papers.27,28 This explains the strategy used in the
Figure 2. Reaction profile and structures for ethylene partial to epoxide and acetaldehyde from ethylene and atomic oxygen adsorbed on Au(111). The zero energy is taken as the sum of the energies of the surface slab model, gas-phase ethylene, and one-half gas-phase molecular oxygen. ET+O stands for ethylene at infinite distance from the surface having adsorbed atomic oxygen, ET-O the situation where ethylene and atomic oxygen are coadsorbed, and OMME for the oxametallacycle intermediate.
experimental work of Deng and Friend who used one of the several successful preparation methods to dose atomic oxygen on Au(111).29,30 However, one has to keep in mind that the energy of the Au(111) surface with two adsorbed oxygen atoms on the fcc threefold sites lies ∼1.1 eV above the energy of separated Au(111) and O2. Therefore, recombination of adsorbed oxygen atoms into molecular oxygen, which will easily desorb from the surface, is the thermodynamically favored route. This route is partially hindered by an energy barrier, which the present calculations estimates as ∼0.4 eV. The O-O distance in the TS for recombination is of 2.188 Å, ∼0.8 Å larger than the gas-phase O2. In the absence of adsorbed atomic oxygen, the adsorption energy of ethylene on Au(111) is very weak (Eads ≈ 0.08 eV), and the geometry of adsorbed ethylene is nearly that calculated for the gas-phase molecule with minor changes in geometry due to adsorption on different adsorption sites. The small lifetime of ethylene on the surface corresponding to such a weak interaction is enhanced by the presence of preadsorbed oxygen. A configuration with ethylene adsorbed with the C atoms in atop and fcc sites (top-fcc) and a threefold hollow site for oxygen was selected to study the coadsorption and the following reactivity toward epoxidation (Figure 1). The geometry of this initial arrangement has been fully optimized, including relaxation of the two outermost metal layers, and the final geometry properly characterized as a minimum on the potential energy surface by pertinent frequency analysis. The presence of adsorbed atomic oxygen enhances the adsorption energy of ethylene by 0.13 eV, as previously reported by different metals.14,31 However, a close inspection to the geometry of the adsorbed molecule indicates that it remains nonactivated. The adsorption energies and geometries of ethylene on clean and oxygen-precovered Au(111) are shown in Table 1. Adsorbed species (ethylene and oxygen) react to form the OMME intermediate as shown in the energy profile in Figure 2. The minimum-energy path corresponds to a rotation of the ethylene molecule through an axis perpendicular to the surface placed on the C1 atom (see Figure 1 for atomic labels). The formation of the OMME intermediate implies two new bonds: one between C2 adsorbed O and a second one between C1 atom and a metal surface atom. The C-C distance elongates from
13312 J. Phys. Chem. B, Vol. 110, No. 27, 2006
Letters
TABLE 1: Adsorption Energy (Eads in eV) and Relevant Distances (d in Å) of Ethylene Adsorbed on Clean and Oxygen-Precovered (0.25 ML) Au(111) Surfacesa O coverage (ML) 0 0.25
Eads
d(O-Au) d(C1-H) d(C1-C2) d(O-C2) d(C1-Au)
0.08 -0.05
2.160
free C2H4
1.045 1.094
1.356 1.357
1.094
1.334
3.272
2.584 2.599
a
Adsorption energies are given with respect to gas-phase ethylene and the clean or oxygen-precovered surface, respectively. The calculated geometry of the gas-phase ethylene is given for comparison.
TABLE 2: Structural Parameters for Reactants (coadsorbed ethylene and atomic oxygen), OMME Intermediate, and Products (epoxide or acetaldehyde) on the Clean Au(111) Surfacesa Eads reactants OMME epoxide acetaldehyde
-0.12 -0.81 -1.24 -2.39
d(O-Au) d(C1-Au) d(C-C) d(C2-O) d(C1-H) 2.182 2.341 3.087 3.843
2.560 2.155 3.718 3.657
1.356 1.522 1.472 1.500
3.271 1.432 1.448 1.223
1.092 1.117 1.096 1.131
a Adsorption energies are in eV and with respect to gas-phase O2 and ethylene, and distances are in Å.
TABLE 3: Structure and Activation Energy (Ea) for the Transition States on Au(111) Leading to the OMME Intermediate and from It to the Epoxide or Acetaldehydea TS
Ea d(C1-Au) d(C1-C2) d(C1-O) d(C2-O) d(C2-H)
to OMME 0.50 to epoxide 0.88 to acetaldehyde 0.73
2.273 2.724 3.047
1.462 1.485 1.475
2.880 2.053 2.452
1.457 1.444 1.343
1.097 1.109 1.172
a
Adsorption energies are in eV and with respect to gas-phase O2 and ethylene, and distances are in Å.
1.357 to 1.522 Å with an intermediate value of 1.462 Å at the TS. The other relevant change involves the C1-Au distance, which changes from 2.599 Å in the intial state to 2.273 Å in the TS state and to 2.155 Å in the OMME. The energy barrier for this reaction is 0.50 eV, and the OMME lies -0.69 eV below the energy of adsorbed reactants. Adsorption energies and geometrical parameters for the OMME intermediate are shown in Table 2 where energies and geometrical properties of the TS involved in the pathway are presented in Table 3. Figure 2 shows that, from the OMME intermediate, two competitive paths leading to epoxide and acetaldehyde product are possible, as found previously for the reaction on Ag6,7 or Cu.14 The cyclization of OMME to form epoxide implies the formation of a new C-O bond and can be monitored by the change in the C1-O distance, which changes from 2.528 Å in the OMME to 2.053 Å in the TS (Figure 3) and 1.448 Å in the epoxide. The energy barrier for this reaction is 0.88 eV, and the product lies 0.43 eV below the OMME. The formation of acetaldehyde can be understood as a 1,2-hydrogen shift, and the C2-H distance provides a good approach to the reaction coordinate; this changes from 1.117 to 1.172 Å in the TS (Figure 3). The energy barrier for this reaction is 0.73 eV, corresponding to a change in energy of -1.58 eV, between OMME and the adsorbed acetaldehyde. This difference in stability of the products with respect to the OMME is very similar to previously reported values for Ag(111).7 Both products are weakly bound to Au(111) with adsorption energies of 0.10 and -0.10 eV for epoxide and acetaldehyde, respectively. It is worth comparing the geometry of the TS involved in the formation of the OMME intermediate in Cu, Ag, and Au. For Cu and Ag, the C2-O distance is almost the same and indeed close to that of the OMME intermediate, while in the
Figure 3. Schematic representation of the three transition-state structures involved in the epoxidation mechanism.
case of Au, this distance is reminiscent of the coadsorbed structure. The fact that OMME formation is easier in Au results from the difference in the character of the TS leading to OMME in Cu and Ag (late type) and in Au (early type) as it would be expected from the Hammond’s postulate.32 However, in the cyclization step leading to EO, the C1-O distance for the TS in Au is similar to the corresponding structure in Ag and similar to reactants (early TS), while in copper, this distance is significantly shorter (late TS). For the elementary step leading to the acetaldehyde, the TSs for the three metal surfaces are very similar. From this qualitative analysis, one can expect a similar selectivity of Au and Ag toward the epoxide as well as toward acetaldehyde. In the following, we will provide more quantitative arguments. The analysis of the thermodynamic profile of the full molecular mechanism for ethylene partial oxidation (Figure 2) shows that, once atomic oxygen is adsorbed on the surface, the formation of epoxide and also acetaldehyde is straighforward at typical catalytic temperatures with the elementary step from the OMME to the product being the rate-determining step. In this respect, the molecular mechanism for ethylene partial oxidation on Au(111) is very close to that reported for Ag(111) by different authors.6,7,14 Conventional transition-state theory was used in order to estimate the selectivity for the epoxide formation. The estimated selectivity toward epoxide at 550 K is ∼40%, which indeed is in rather good agreement with experiment findings for styrene. Here, a caveat is necessary, since it has been shown recently that small variations in the energy barriers caused may result in noticeable difference in the predicted selectivity.8 The difference in Eact in the energy profile in Figure 2 is sufficiently large so that the qualitative conclusions will remain unchanged upon increasing the accuracy of the present DFT calculations. In fact, from Figure 2 it appears that the similar selectivity found for gold and silver is directly related to energy difference between both TS structures leading to epoxide and acetaldehyde, respectively, which is ∼0.1 eV for either Ag(111) or Au(111) but significantly larger for Cu(111).14 The similarity between the two metals is also evident in the complete energy profile. IV. Conclusions The present work evidences that, as far as ethylene epoxidation is concerned, the catalytic activity of Au(111) is very similar to that of Ag(111). Therefore, for this particular reaction, there is no special feature in the catalytic activity of Au(111), which is at variance with exciting findings concerning the catalytic activity of Au nanoparticles.33,34 This claim is based
Letters on the comparison of present density functional model calculations for the complete energy profile for the reaction on Au(111) to earlier results published for Ag(111). This comparison clearly proves that the energy profile for the molecular mechanism for ethylene oxidation to the epoxide on Au(111) is similar to that of Ag(111), with the main difference being in the initial step concerning dissociation of molecular oxygen which is required for Au(111) and not for Ag(111). Moreover, the present calculations predict that the selectivity of Au(111) toward ethylene epoxidation (∼40%) is also very similar to that of Ag(111). Indeed, this is in good agreement with the experimental findings of Deng and Friend for styrene epoxidation.16 Acknowledgment. The authors are indebted to Prof. R. M. Lambert and Dr. N. Lopez for stimulating discussions. Financial support has been provided by the Spanish Ministry of Education and Science (projects CTQ2005-08459-CO2-01 and UNBA0533-001) and, in part, by Generalitat de Catalunya (projects 2005SGR-00697, 2005 PEIR 0051/69 and Distincio´ per a la Promocio´ de la Recerca Universita`ria de la Generalitat de Catalunya granted to F.I.). Part of the computer time was provided by the Centre de Supercomputacio´ de Catalunya, CESCA and Barcelona Supercomputing Centre, BSC. D. Torres thanks the Universitat de Barcelona for a predoctoral fellowship. References and Notes (1) van Santen, R. A.; Kuipers, H. P. C. E. AdV. Catal. 1987, 35, 265. (2) Serafin, J. G.; Liu, A. C.; Seyedmonir, S. R. J. Mol. Catal., A 1998, 131, 157. (3) Klust A.; Madix, R. J. J. Am. Chem. Soc. 2006, 128, 1034. (4) Williams, F. J.; Bird, D. P. C.; Palermo, A.; Santra A. K.; Lambert, R. M. J. Am. Chem. Soc. 2004, 126, 8509. (5) Monnier, J. R.; Stavinoha, J. L., Jr.; Hartley, G. W. J. Catal. 2004, 226, 321.
J. Phys. Chem. B, Vol. 110, No. 27, 2006 13313 (6) Linic S.; Barteau, M. A. J. Am. Chem. Soc 2003, 125, 4034. (7) Bocquet, M.-L.; Michaelides, A.; Loffreda, D.; Sautet, P.; Alavi, A.; King, D. A. J. Am. Chem. Soc. 2003, 125, 5620. (8) Bocquet, M.-L.; Loffreda, D. J. Am. Chem. Soc. 2005, 127, 17207. (9) Stegelmann, C.; Stoltze, P. J. Catal. 2004, 226, 129. (10) Jones, G. S.; Mavrikakis, M.; Barteau, M. A.; Vohs, J. M. J. Am. Chem. Soc. 1998, 120, 3196. (11) Linic, S.; Piao, H.; Adib, K.; Barteau, M. A. Angew. Chem., Int. Ed. 2004, 43, 2918. (12) Grant, R. B.; Lambert, R. M. J. Catal. 1985, 92, 364. (13) Cowell, J. J.; Santra, A. K.; Lambert, R. M. J. Am. Chem. Soc. 2000, 122, 2381. (14) Torres, D.; Lopez, N.; Illas, F.; Lambert, R. M. J. Am. Chem. Soc. 2005, 127, 10774. (15) Hayashi, T.; Tanaka, K.; Haruta, M. J. Catal. 1998, 178, 566. (16) Deng, X.; Friend, C. M. J. Am. Chem. Soc. 2005, 127, 17178. (17) Perdew, J. P.; Wang, Y. Phys. ReV. B 1992, 45, 13244. (18) Blochl, P. E. Phys. ReV. B 1994, 50, 17953. (19) Monkhorst, H. J.; Pack, J. D. Phys. ReV. B 1976, 13, 5188. (20) Kresse, G.; Hafner, J. Phys. ReV. B 1993, 47, 558. (21) Kresse, G.; Furthmuller, J. Comput. Mater. Sci. 1996, 6, 15. (22) Kresse, G.; Furthmuller, J. Phys. ReV. B 1996, 54, 11169. (23) Mills, G.; Jonsson, H. Phys. ReV. Lett. 1994, 72, 1124. (24) Mills, G.; Jonsson, H.; Schenter, G. K. Surf. Sci. 1995, 324, 305. (25) Henkelman, G.; Uberuaga, B. P.; Jonsson, H. J. Chem. Phys. 2000, 113, 9901. (26) Meyer, R.; Lemire, C.; Shaikhutdinov, Sk.; Freund, H.-J. Gold Bull. 2004, 37, 72. (27) Torres, D.; Gonza´lez, S.; Neyman, K. M.; Illas, F. Chem. Phys. Lett. 2006, 422, 412. (28) Xu, Y.; Mavrikakis, M. J. Phys. Chem. 2003, 107, 9298. (29) Deng, X.; Min, B. K.; Guloy, A.; Friend, C. M. J. Am. Chem. Soc. 2005, 127, 9267. (30) Saliba, N.; Parker, D. H.; Koel, B. E. Surf. Sci. 1998, 410, 270. (31) Kokalj, A.; Dal Corso, A.; de Gironcoli, S.; Baroni, S. J. Phys. Chem. B 2006, 110, 367. (32) Hammond, G. S. J. Am. Chem. Soc. 1955, 77, 334. (33) Haruta, M.; Yamada, N.; Kobayashi, T.; Iijima, S. J. Catal. 1989, 115, 301. (34) Valden, M.; Lai, X.; Goodman, D. W. Science 1998, 281, 1647.