7992
J. Phys. Chem. C 2007, 111, 7992-7999
On the Role of Subsurface Oxygen and Ethylenedioxy in Ethylene Epoxidation on Silver Jeff Greeley† and Manos Mavrikakis* Department of Chemical and Biological Engineering, UniVersity of Wisconsin-Madison, Madison, Wisconsin 53706 ReceiVed: January 19, 2007; In Final Form: March 24, 2007
The thermochemical stability of various three-component phases containing oxygen, ethylene, and Ag(111) was determined as a function of oxygen and ethylene chemical potential using periodic, self-consistent density functional theory calculations. Ethylenedioxy is stable over a wide range of conditions, although its formation may be kinetically hindered in some cases. Ethylene and ethylene-containing oxametallacycles are also found to be stable over a reasonably large range of chemical potentials, particularly if ethylenedioxy formation is neglected. Furthermore, subsurface oxygen (Osb) is seen to be present in the three-component systems at a variety of conditions; minimum energy path calculations performed at a coverage of 1/2 ML Osb suggest that this species may actually increase the reaction barrier for ring closure leading to ethylene oxide elimination from Ag(111).
Introduction Ethylene epoxidation, the production of ethylene oxide (EO) via the partial oxidation of ethylene, is among the most important processes in the modern chemical industry. With an annual production value of billions of dollars, EO is used in the synthesis of ethylene glycol (an important component of antifreeze agents), polyester fabrics, surfactants, and detergents. The principal challenge associated with the epoxidation process, which employs an alkali- and chlorine-promoted silver catalyst, is the prevention of the total oxidation of ethylene; the competition between this undesired reaction and the desired epoxidation reaction determines the selectivity of EO production.1-3 The need to increase the selectivity of EO production has driven a significant effort to understand the fundamental reaction mechanism, and a variety of possible reactive intermediates has been discussed in the literature. For example, Campbell4 proposed the existence of a common intermediate for the epoxidation and total oxidation pathways. Linic and Barteau5-8 have presented convincing evidence identifying this common intermediate as a surface oxametallacycle; in this case, the total oxidation pathway is reported to proceed through acetaldehyde.9 Stegalmann et al.10 have proposed the existence of CH2CHO and CH2CHOH species in the epoxidation pathway. Still more intermediates have been proposed for closely related reactions; Bulushev et al.11 identified glycol-like species during EO reactions on a Ag/R-Al2O3/hydroxylated support, and Capote and Madix12 observed an ethylenedioxy (OCH2CH2O) species during ethylene glycol decomposition on Ag(110). Another active area of study in ethylene epoxidation concerns the chemical state of oxygen on the catalyst surface under typical epoxidation conditions. It has been suggested that oxygen may be present in either a nucleophilic or an electrophilic form; the former is active for total oxidation, while the latter leads to epoxidation.13-19 A variety of structures for the EO active, * To whom correspondence should be addressed. E-mail: manos@ engr.wisc.edu. Phone: (608) 262-9053. Fax: (608) 262-5434. † Present address: Center for Nanoscale Materials, Argonne National Laboratory, Argonne, IL 60439.
electrophilic phase has been postulated: surface (atomic) oxygen,18 surface oxygen with chemical properties strongly modified by surface defects,15 surface oxygen with properties modified by subsurface oxygen,13,20 and perhaps even a surface oxide phase.21-27 The above considerations suggest that many important questions regarding the details of the ethylene epoxidation reaction mechanism remain unresolved. For example, the effect of subsurface oxygen on epoxidation kinetics is not yet fully understood,28,29 nor has the effect of ethylene itself on the structure of the binary O/Ag system been fully analyzed. In this paper, we begin to address these questions by offering some preliminary insights into the complex, three-way relationship between ethylene, oxygen, and silver on the Ag(111) surface. We use a first principles-based thermodynamic analysis to investigate the stability of the Ag(111) surface phases in mixed oxygen and ethylene atmospheres, and we identify a possible intermediate in the ethylene epoxidation reaction, ethylenedioxy (EDO), that has hitherto received less attention in the EO literature. We also determine whether or not phases containing both ethylene and subsurface O may be stable under typical ethylene epoxidation conditions, and we report the effect of subsurface O on the kinetics of ring closure in oxametallacycles to form EO. Materials and Methods DACAPO, a density functional theory (DFT) based total energy calculation code,30 was used for all calculations in this study.31 A four-layer slab representing Ag(111), periodically repeated in a super cell geometry with six equivalent layers of vacuum between any two successive metal slabs, was used; given the lower surface energy of the Ag(111) facet as compared to other low-index crystal facets, we expect this surface to be very relevant, if not completely dominant, at catalytically relevant temperatures.32 A 2 × 2 surface unit cell was employed, and the top two layers of the slab were allowed to relax. Adsorption was allowed on only one of the two surfaces exposed, and the electrostatic potential was adjusted accordingly.33 Ionic cores were described by ultrasoft pseudopotentials,34 and the Kohn-Sham one-electron valence states
10.1021/jp070490i CCC: $37.00 © 2007 American Chemical Society Published on Web 05/12/2007
Oxygen and Ethylenedioxy in Ethylene Epoxidation on Ag
J. Phys. Chem. C, Vol. 111, No. 22, 2007 7993
Figure 1. Phase diagrams for the O-C2H4-Ag(111) system with (a) EDO explicitly incorporated into the thermochemical analysis and (b) EDO omitted from the analysis. The maximum values of µO and µEt correspond to the calculated total gas-phase energies of 1/2 O 2 and C2H4, respectively. The dark circles show an estimate of the industrial conditions for ethylene epoxidation (∼500 K, 15 bar O2 and C2H4). OMME: oxametallacycle. Osu: surface oxygen. Osb: subsurface oxygen. In both diagrams, a region of stability for a 1/4 ML Osu/1/4 ML Osb phase exists between 1/4 ML Osu and 1/4 ML Osu/1/2 ML Osb phases; this region is too narrow to appear at the resolution of the diagrams. No ZPE corrections are included in the analysis. (c) Selected phases for various coverages of O and C2H4. For each coverage, the most stable phases are located on the x-axis. Not all indicated phases are present in the phase diagrams of Figure 1a,b. Coverages can be obtained by dividing the number of atoms of the respective species by 4. Pure C2H4 has been omitted from the diagram because all configurations are essentially degenerate. tbt: top-bridge-top configuration for C2H4. See also captions of Tables 1 and 3 for additional descriptions of site nomenclature. An asterisk indicates that the species in question is accompanied by one surface and two subsurface oxygen atoms.
were expanded in a basis of plane waves with kinetic energy up to 25 Ry. The surface Brillouin zone was sampled at 18 special k-points. The exchange correlation energy and potential were described by the generalized gradient approximation (GGA-PW91).35,36 The self-consistent PW91 density was determined by iterative diagonalization of the Kohn-Sham Hamiltonian, Fermi population of the Kohn-Sham states (kBT ) 0.1 eV), and Pulay mixing of the resulting electronic density.37 All total energies were extrapolated to kBT ) 0 eV. No zero-
point energy corrections were included in the reported results. The lattice constant for bulk Ag was calculated to be 4.14 Å, in good agreement with the experimental value of 4.08 Å.38 The minimum energy reaction paths of various elementary steps were studied using the climbing image nudged elastic band (CI-NEB) method.39-42 Vibrational frequencies were calculated by numerical differentiation of the forces using a second-order finite difference approach with a step size of 0.015 Å.43 The resulting Hessian matrix was then mass-weighted and diago-
7994 J. Phys. Chem. C, Vol. 111, No. 22, 2007
Greeley and Mavrikakis
Figure 2. (a) Potential energy surface (PES) for the reaction of an oxametallacycle (OMME) to form (i) EDO and (ii) EO. O* represents an adsorbed atomic oxygen. The energetics for all states involving simultaneous adsorption of more than one species on the surface is reported with the assumption that the species are at infinite separation on Ag(111) from one another; hence, it is assumed, for example, that the extra adsorbed O atom in the leftmost panel has no effect on the OMME formation process. The oxygen atom in the adsorbed EDO on the right-hand side of the figure is far (approximately 3.5 Å) from the Ag(111) surface. Dashed horizontal line: energy of gas-phase ethylene plus two coadsorbed oxygen atoms at infinite separation from one another. (b) Simplified schematic of possible reaction pathways from ethylene to EO. Sketches of both OMME and EDO are also shown.
nalized to yield the frequencies and normal modes of the system. Specular HREELS intensities were estimated from the spatial derivatives of the dipole moment of the system; these derivatives were calculated with the same numerical scheme as was used for the force derivatives. Phase diagrams were constructed by determining the most stable phase in the grand canonical ensemble as a function of the chemical potential of oxygen and ethylene. Details of the formalism have been provided by Greeley and Mavrikakis.43 The procedure to evaluate the approximate oxygen and ethylene chemical potentials associated with industrial ethylene epoxidation processes (∼500 K, 15 bar oxygen and ethylene) was also similar to the procedure described in the indicated reference. In the present analysis, however, we obtained the ground state total energy of O2 by subtracting the experimentally determined value of the oxygen bond energy44 from twice the DFT calculated total energy of atomic oxygen. This procedure avoids a calculation of the gas-phase O2 total energy; this bond is known to be poorly described by DFT.45,46 Results and Discussion We begin by studying the thermochemistry of the three-way interaction between ethylene, silver, and oxygen on Ag(111). In doing so, we strive to determine both the important ethylene-
related intermediates that could be present under typical ethylene epoxidation conditions and the effect that these intermediates have on the ad(b)sorption properties of surface and subsurface oxygen. To this end, the lowest energy configurations for a variety of surface phases involving these three species have been analyzed. Structures for three classes of ethylene-related moieties, including ethylene itself, surface oxametallacycles, and EDO, have been determined on clean Ag(111) and on Ag(111) with coadsorbed surface and subsurface oxygen (Table 1). Total coverages of oxygen (including both surface and subsurface O species) between 0 and 1 ML have been studied, and coverages of 0 and 0.25 ML have been analyzed for ethylene-related species. We do not consider EO itself in the phase diagrams, as this species does not interact strongly with the Ag(111) surface and most likely desorbs as soon as it is formed. In the absence of adsorbed ethylene-related species, pure surface oxygen is the most stable phase only up to a total oxygen coverage of 0.25 ML. At higher total oxygen coverages, the population of subsurface oxygen sites begins (Table 1). As the coverage approaches 1 ML, oxygen can partition in either a 1:3 or a 1:1 ratio between surface and subsurface sites (these two configurations are nearly degenerate). We note that we do not consider the population of sites below the first subsurface layer in this analysis and that we do not consider other, perhaps
Oxygen and Ethylenedioxy in Ethylene Epoxidation on Ag
J. Phys. Chem. C, Vol. 111, No. 22, 2007 7995
TABLE 1: Most Stable Configurations for Oxygen-Ethylene-Silver(111) System at Various Coverages of O and Ethylenea
configuration
O coverage (ML)
ethylene coverage (ML)
total energy (eV)
C2H4 binding energy (eV)
clean Osu (fcc) Osu (fcc)/Osb (tetr I) Osu (hcp)/2Osb (oct) Osu (hcp)/3Osb (oct)b C2H4 (button) OMME EDO EDO/Osb (oct) EDO/2Osb (oct)
0 0.25 0.5 0.75 1.0 0 0.25 0.5 0.75 1.0
0 0 0 0 0 0.25 0.25 0.25 0.25 0.25
-20704.06 -21139.73 -21575.22 -22010.67 -22445.46 -21078.71 -21514.88 -21952.20 -22387.53 -22822.87
-0.10 -0.61 -2.43 -2.31 -2.88
a Configurations involving ethylenedioxy (EDO) are explicitly included in this analysis, but configurations with ethylene oxide (EO: very weakly adsorbed species) are omitted. OMME: oxametallacycle (oxygen and carbon are bound to different silver atoms). Button: both carbon atoms are bound to the same silver atom. Oct: octahedral (under a surface fcc site). Tetr I: tetrahedral (under a surface hcp site). C2H4 binding energies (B.E.) are calculated with respect to the most stable configuration of oxygen (in the absence of C2H4) at indicated oxygen coverage. b Configuration with 2Osu(hcp)/2Osb(oct) is nearly degenerate.
larger, surface unit cells, which might provide a more detailed phase diagram. Nevertheless, we believe that the insights derived with this setup should be a useful initial step toward understanding this complex system. When 0.25 ML ethylene is added to the O/Ag system, a significant increase in the complexity of the phase diagram is observed. At low total oxygen coverages (0.25 ML), surface O atoms are bound up in surface oxametallacycles, consistent with the work of Barteau and Linic5 (Table 1). As the oxygen coverage is increased to 0.5 ML, however, all O atoms are incorporated into EDO; the high stability of this species can be seen by the energy change of EDO formation from gas-phase ethylene and surface O/Ag, -2.43 eV, amounting to roughly 4 times the oxametallacycle formation energy from C2H4(g) and surface oxygen (Table 1). In contrast to the chemistry observed for the pure O/Ag system, it is only at higher oxygen coverages, after EDO has formed, that subsurface oxygen can be favorably created; the total oxygen coverage is 0.75 ML in this case. As the O coverage is further increased to 1 ML, additional subsurface O is formed. We note, in passing, that an additional effect of ethylene adsorption on these O/Ag slabs is to induce significant surface rearrangements. This effect is consistent with the recent experiments of Guo et al.47-49 that demonstrate a pronounced tendency for Ag(110) to reconstruct in the presence of oxygen and other adsorbates, and it is also reminiscent of the substantial surface rearrangements that have been recently shown to occur in the presence of Ag adatoms.26,27 The magnitude of this effect seems to increase as the coverage of subsurface oxygen increases; to illustrate the maximum possible size of the effect within the range of oxygen coverages studied, we have analyzed ethylene adsorption on a metastable O/Ag slab with 1 ML of subsurface oxygen. In that case, the surface is completely reconstructed to yield an Ag adatom to which ethylene is bound.28 To better understand the thermochemical characteristics of the phases described above, we have constructed diagrams showing the most stable surface phases at given values of the chemical potentials of oxygen and ethylene. In Figure 1a, the maximum chemical potentials of both oxygen and ethylene correspond to the total energies of these species in their stable,
Figure 3. Calculated HREELS spectra for (a) EDO on clean Ag(111), (b) EDO on Ag(111) with 1/2 ML of subsurface oxygen, and (c) an oxametallacycle (OMME) on clean Ag(111). More detailed information on these vibrational frequencies is provided in Table 2.
gas-phase forms (1/2 O2 and C2H4, respectively). These energies represent physically realistic upper bounds to the values that the corresponding chemical potentials may take, while chemical
7996 J. Phys. Chem. C, Vol. 111, No. 22, 2007
Greeley and Mavrikakis
TABLE 2: Calculated Vibrational Frequencies of Ethylenedioxy (EDO) and Oxametallacycle (OMME) on Clean Ag(111)a species EDO on clean Ag(111)
OMME on clean Ag(111)
frequency scaled (cm-1) intensity 2951
0.00
2934 2909 2896 1425 1422 1319 1306 1206 1202 1070 1044 1020 866 841 558 380 288 246 213
0.83 0.68 0.04 0.10 0.00 0.02 0.00 0.05 0.00 0.84 1.00 0.21 0.26 0.00 0.01 0.00 0.33 0.04 0.19
3087 3019 2949 2917 1435 1416 1293 1212 1111 1017 949 872 868 544 514 288 257 246
0.00 0.25 0.00 1.00 0.15 0.01 0.33 0.13 0.02 0.09 0.05 0.00 0.88 0.05 0.01 0.01 0.56 0.28
mode description
species EDO on Ag(111) with 1/2 ML subsurface oxygen
C-H stretch C-H stretch
C-C twist C-O stretch
O-Ag stretch
frequency scaled (cm-1) intensity 2969
0.08
2955 2921 2915 1420 1419 1331 1313 1237 1201 1078 1066 1028 860 848 553 419 410 371 357 339 329 297 277 255 248 215
0.68 0.30 0.18 0.02 0.01 0.13 0.01 0.01 0.00 1.00 0.47 0.00 0.60 0.19 0.04 0.02 0.02 0.05 0.05 0.02 0.19 0.00 0.05 0.05 0.04 0.00
mode description
C-H stretch C-H stretch
C-O stretch C-C twist C-C stretch
C-H stretch CH2 wag
O-C-C scissor + CH2 wag C-Ag stretch CO-Ag stretch
a
Frequencies of EDO on Ag(111) with 1/2 ML of subsurface oxygen present are also given. Only frequencies with magnitudes greater than 200 cm-1 are shown. The intensity of the highest intensity mode in both spectra is set to unity, and the other intensities are scaled accordingly. The spectrum with subsurface oxygen present explicitly includes the vibrational modes of the subsurface O. Some qualitative descriptions are given for the most intense modes only.
potentials under industrial ethylene epoxidation conditions will lie somewhere in the upper middle of the phase diagrams. For a broad range of conditions, EDO (either with or without subsurface oxygen) is seen to be the most stable species on the Ag(111) surface. In fact, the stability of this species is such that it permits the presence of surface oxygen (in the form of EDO) at much lower oxygen chemical potentials than would be possible in the absence of ethylene. In contrast, the oxygen chemical potential at which subsurface O forms does not seem to be significantly decreased by EDO. Nonetheless, the phase diagram suggests that subsurface O could be present under typical ethylene epoxidation conditions. The phase diagrams indicate that EDO dominates the thermochemical behavior of the ethylene/oxygen/silver system for a broad range of oxygen and ethylene chemical potentials. This species (and its single carbon analogue, dioxymethylene) has been proposed and observed in some chemistries on silver and copper surfaces.11,12,50 However, to our knowledge, the literature
on ethylene epoxidation makes no mention of EDO. This absence could be attributed to a combination of two factors. First, EDO formation is likely to be kinetically unfavorable under some conditions. Figure 2a shows the calculated energetics for the competition between oxametallacycle (OMME) ring closure toward EO elimination and the addition of a surface O atom to the C-Ag bond in OMME to form EDO. Figure 2b provides a simplified schematic representation of these processes, together with sketches of the OMME and EDO molecules. The latter process has a reaction barrier ∼0.3 eV higher than the former, suggesting that EDO formation will not be particularly competitive at low temperatures on clean Ag(111). Second, it is possible that any EDO that does form on silver surfaces would be very difficult to identify experimentally. Spectroscopic identification of EDO might, for example, require techniques and the design of experiments at appropriate temperature and pressure windows, constituting an approach analogous to the one used by Barteau and co-workers
Oxygen and Ethylenedioxy in Ethylene Epoxidation on Ag to identify oxametallacycles on Ag(111) surfaces.6,51 In Figure 3 and Table 2, we present calculated HREELS spectra for EDO, both in the presence and in the absence of subsurface O, and for an oxametallacycle without subsurface O. If a suitable method for preparing EDO on Ag(111) could be found (perhaps involving dosing EO onto a surface with preadsorbed O, as Figure 2 suggests might be energetically feasible), spectra in Figure 3 might be helpful in identifying EDO and in distinguishing it from oxametallacycles. The two spectra (Figure 3a,c) are qualitatively similar, but EDO seems to have an intense, asymmetric C-O stretch mode at 1044 cm-1, while a comparably intense mode for the oxametallacycle is found at a lower frequency, 868 cm-1 (this mode is a combination O-C-C inplane scissor and a CH2 wag). Although we caution that our calculated HREELS intensities and peak positions are only approximate, the difference in frequencies between these intense modes might be enough to experimentally distinguish one species from the other. Furthermore, substitution of both C atoms with 13C yields a calculated decrease of ∼25 cm-1 in the most intense EDO mode; a decrease of ∼12 cm-1 is predicted for the most intense oxametallacycle mode. These characteristics might permit differentiation of the indicated species by isotopic labeling of ethylene. The above kinetic considerations notwithstanding, it is clear that EDO is significantly more stable on Ag(111) than are other species in the commonly accepted reaction network for ethylene epoxidation. Furthermore, the possibility that pathways, other than the addition of O to OMME, exist for EDO formation cannot be ruled out,11,12 and it is also likely that subsurface oxygen may alter the relative kinetics of the pathways described above. We also note that ∼80% of the EDO formation barrier comes from repulsion between coadsorbed OMME and atomic oxygen; this repulsion might be reduced at certain coverage regimes, thereby facilitating EDO formation. In addition, the barriers to EDO decomposition, at least through the pathway analyzed in Figure 2, are quite high, indicating that any EDO that forms will likely remain on the surface. Thus, some EDO may well be present, probably in the form of a spectator species, on the Ag(111) surface during ethylene epoxidation. Given that EDO may act primarily as a spectator species in ethylene epoxidation reactions, we have reanalyzed the thermochemical behavior of the ethylene-oxygen-silver system in the absence of EDO. The most stable configurations are listed in Table 3; it is seen that the resulting system is dominated by oxametallacycles at ethylene coverages of 0.25 ML. The data are presented in the form of a phase diagram in Figure 1b, whereas the full dataset for the phase diagrams presented in both Figure 1a,b is presented schematically in Figure 1c. The diagram in Figure 1b shows the presence of oxametallacycles relatively close to the chemical potentials that might be expected under industrial ethylene epoxidation conditions. In situations where EDO formation is kinetically limited, therefore, oxametallacycles should be present on the surface, consistent with the results of Barteau and co-workers.6,7,51 We note, in passing, that the ethylene chemical potentials required to yield stable oxametallacycles on the surface are considerably higher than are the corresponding potentials required to produce EDO (compare Figure 1a,b); this result simply reflects the very high thermochemical stability of EDO on Ag(111) surfaces. We further note that stable phases containing both oxametallacycles and subsurface oxygen do exist near industrial ethylene epoxidation conditions; the presence of oxametallacycles does not, however, seem to significantly alter the oxygen chemical
J. Phys. Chem. C, Vol. 111, No. 22, 2007 7997 TABLE 3: Most Stable Configurations for Oxygen-Ethylene-Silver(111) System at Various Coverages of O and Ethylenea
configuration Clean Osu (fcc) Osu (fcc)/Osb (tetr I) Osu (hcp)/2Osb (oct) Osu (hcp)/3Osb (oct)b C2H4 (button) OMME C2H4 (button)/Osu (fcc)/ Osb (tetr I)c OMME/Osb (tetr I)/ Osb (tetr II)d OMME/Osb (tetr I)/ Osb (tetr II)/Osb (oct)
O ethylene coverage coverage (ML) (ML)
total energy (eV)
C2H4 binding energy (eV)
0 0.25 0.5 0.75 1.0 0 0.25 0.5
0 0 0 0 0 0.25 0.25 0.25
-20704.06 -21139.73 -21575.22 -22010.67 -22445.46 -21078.71 -0.10 -21514.88 -0.61 -21950.01 -0.24
0.75
0.25
-22385.89 -0.67e
1.0
0.25
-22821.04 -1.03e
a Configurations involving ethylenedioxy (EDO) and ethylene oxide (EO) are not included in this analysis (EO is weakly adsorbed and is assumed to desorb immediately upon formation). OMME: oxametallacycle (oxygen and carbon are bound to different silver atoms). Oct: octahedral (under a surface fcc site). Tetr I: tetrahedral (under a surface hcp site). Tetr II: tetrahedral (directly under a surface Ag atom). Button: both carbon atoms are bound to the same silver atom. C2H4 binding energies (B.E.) are calculated with respect to the most stable configuration of oxygen (in the absence of C2H4) at the indicated oxygen coverage. b Configuration with 2Osu (hcp)/2Osb (oct) is nearly degenerate. c OMME configuration exists but is 0.15 eV less stable. d Buttonlike configuration is nearly degenerate with the OMME. e Change in the preferred subsurface O positions as compared to the C2H4-free slabs is induced by C2H4 adsorption.
potential needed to produce subsurface oxygen, as compared to the oxygen chemical potential needed in the absence of ethylene. These results indicate that subsurface oxygen may be thermochemically stable under typical ethylene epoxidation conditions. This fact, combined with the extensive speculation in the literature about the influence of subsurface O on epoxidation kinetics, may suggest that an analysis of the relevant kinetics on a subsurface O covered slab might be interesting. We have focused on a single, kinetically significant step in the reaction, that of ring closure in oxametallacycles to form EO.8,23 We analyze this reaction on a surface with 1/2 ML of subsurface oxygen. This phase is stable near industrially relevant conditions (see dark circle in Figure 1), assuming that EDO formation is kinetically inhibited (see above discussion). We note that the small difference between the calculated chemical potentials at industrial conditions and the ethylene-containing phases in Figure 1 is likely because of inaccuracies in the chosen DFT functional (e.g., the lack of dispersion forces52). The calculated barrier for ring closure toward EO elimination in the presence of 1/2 ML of subsurface O, 1.46 eV (Figure 4), is significantly higher than is the corresponding barrier, 0.86 eV, on clean Ag(111). The reaction coordinate for ring closure both with and without subsurface oxygen involves initial movement of the surface oxygen parallel to the silver surface; C-O bond formation occurs subsequent to the oxygen diffusion. This quasidiffusive initiation of the reaction coordinate suggests that the overall barrier for ring closure could be related to the diffusive properties of atomic oxygen on the respective surfaces. It is known, in turn, that subsurface oxygen can alter the properties of adsorbed surface oxygen in a complex manner28,53 and can lead to a significant enhancement of the surface O binding energy and of its diffusion barrier (we estimate an increase of
7998 J. Phys. Chem. C, Vol. 111, No. 22, 2007
Greeley and Mavrikakis
Figure 4. Potential energy surfaces for ring closure in oxametallacycles (OMME) on (a) clean Ag(111) and (b) Ag(111) with 1/2 ML of subsurface oxygen. O* and Osb represent surface and subsurface oxygen, respectively. All energies are calculated with respect to surface coadsorbates at infinite separation from one another (in Table 1, the reference states involve coadsorbed oxygen on Ag(111)). The oxametallacycle configuration shown in panel a is degenerate to the configuration shown in Figure 3c. The indicated EO* species are far from the Ag(111) surface (see also Figure 2 caption). In panel b, all energetics are referenced to an Ag(111) slab with 1/4 ML surface oxygen and 1/2 ML subsurface oxygen.
a few tenths of an electronvolt in the barrier for O diffusion on the Ag surface in the model system with 1 ML of subsurface oxygen, as compared to a clean Ag(111) slab). This increase might be partially responsible for the increased reaction barrier leading to EO elimination via oxametallacycle ring closure in the presence of subsurface oxygen. The above analysis suggests that subsurface oxygen does not facilitate, and may even inhibit, ethylene epoxidation kinetics. We stress, however, that significantly more work needs to be done before the role of subsurface oxygen in this reaction can be determined in greater detail. For example, it would, in general, be useful to determine if the conclusions reached using our relatively simple surface model would be affected by the use of a more detailed model, for example, that of the p(4 × 4) surface oxide,26,27 or by the use of a more elaborate description of the active catalytic phase, under realistic ethylene epoxidation conditions, whenever that becomes available. More specifically, it would be interesting to understand the effect of subsurface O on other steps in the ethylene epoxidation reaction network, including EDO chemistry. It would also be desirable to analyze the OMME ring closure kinetics as a function of both subsurface and surface oxygen coverage; the effect might be strongly coverage dependent. We note that, even if the ring closure barrier is found to be higher at all coverages of subsurface oxygen, ethylene epoxidation under industrial conditions may still be facilitated by subsurface O. It has been recently shown, for example, that the barrier for O2 dissociation on Ag(111) (another kinetically significant step in the ethylene epoxidation reaction mechanism8) is lowered in the presence of subsurface O.28 This effect might offset increases in the ring closure barrier under industrial ethylene epoxidation conditions. Furthermore, under those conditions, subsurface oxygen may act synergistically with promoters (e.g., alkaline, chlorine) to alter the ethylene epoxidation kinetics. Conclusion Three-component phase diagrams for systems with oxygen, ethylene, and Ag(111) present have been determined as a
function of oxygen and ethylene chemical potentials. EDO is found to dominate these phase diagrams, but its formation may be kinetically hindered in some cases. If the formation of EDO is not considered, then ethylene-containing oxametallacycles are also found to be present in several phases. Furthermore, subsurface oxygen (Osb) is found to be stable at a variety of conditions; minimum energy path calculations performed on Ag(111) in the presence of 1/2 ML Osb suggest that Osb may actually increase the reaction barrier for EO elimination from surface oxametallacycles. Acknowledgment. DOE-BES, Office of Chemical Sciences supported this work. The majority of the calculations was performed with National Energy Research Scientific Computing Center (NERSC) and Molecular Science Computing Facility (MSCF-PNNL) resources. Additional calculations were performed with computing resources at ORNL. References and Notes (1) Satterfield, C. N. Heterogeneous Catalysis in Industrial Practice, 2nd ed.; Krieger Pulishing Company: Malabar, 1996. (2) Chorkendorff, I.; Niemantsverdriet, H. Concepts of Modern Catalysis and Kinetics; Wiley-VCH: Weinheim, 2003. (3) Somorjai, G. A. Introduction to Surface Chemistry and Catalysis; John Wiley: New York, 1994. (4) Campbell, C. T. J. Catal. 1985, 94, 436. (5) Linic, S.; Barteau, M. A. J. Am. Chem. Soc. 2003, 125, 4034. (6) Linic, S.; Barteau, M. A. J. Am. Chem. Soc. 2002, 124, 310. (7) Linic, S.; Piao, H.; Adib, K.; Barteau, M. A. Angew. Chem., Intl. Ed. 2004, 43, 2918. (8) Linic, S.; Barteau, M. A. J. Catal. 2003, 214, 200. (9) Torres, D.; Lopez, N.; Illas, F.; Lambert, R. M. J. Am. Chem. Soc. 2005, 127, 10774. (10) Stegelmann, C.; Schiødt, N. C.; Campbell, C. T.; Stoltze, P. J. Catal. 2004, 221, 630. (11) Bulushev, D. A.; Paukshtis, E. A.; Nogin, Y. N.; Bal’zhinimaev, B. S. Appl. Catal., A 1995, 123, 301. (12) Capote, A. J.; Madix, R. J. J. Am. Chem. Soc. 1989, 111, 3570. (13) Grant, R. B.; Lambert, R. M. J. Catal. 1985, 92, 364. (14) van Santen, R. A.; Kuipers, H. P. C. E. AdV. Catal. 1987, 35, 265. (15) Bal’zhinimaev, B. S. Kinet. Catal. 1999, 40, 795. (16) Carter, E. A.; Goddard, W. A., III. Surf. Sci. 1989, 209, 243.
Oxygen and Ethylenedioxy in Ethylene Epoxidation on Ag (17) Bukhtiyarov, V. I.; Prosvirin, I. P.; Kvon, R. I. Surf. Sci. Lett. 1994, 320, 47. (18) Bukhtiyarov, V. I.; Ha¨vecker, M.; Kaichev, V. V.; Knop-Gericke, A.; Mayer, R. W.; Schlo¨gl, R. Phys. ReV. B 2003, 67, 235422. (19) Bukhtiyarov, V. I.; Nozovskii, A. I.; Bluhm, H.; Havecker, M.; Kleimenov, E.; Knop-Gericke, A.; Schlogl, R. J. Catal. 2006, 238, 260. (20) van Santen, R. A.; de Groot, C. P. M. J. Catal. 1986, 98, 530. (21) Temkin, M. I. AdV. Catal. 1979, 28, 173. (22) Carlisle, C. I.; King, D. A. Phys. ReV. Lett. 2000, 84, 3899. (23) Bocquet, M.-L.; Michaelides, A.; Loffreda, D.; Sautet, P.; Alavi, A.; King, D. A. J. Am. Chem. Soc. 2003, 125, 5620. (24) Michaelides, A.; Bocquet, M.-L.; Sautet, P.; Alavi, A.; King, D. A. Chem. Phys. Lett. 2003, 367, 344. (25) Li, W.-X.; Stampfl, C.; Scheffler, M. Phys. ReV. Lett. 2003, 90, 256102. (26) Schnadt, J.; Michaelides, A.; Knudsen, J.; Vang, R. T.; Reuter, K.; Laegsgaard, E.; Scheffler, M.; Besenbacher, F. Phys. ReV. Lett. 2006, 96, 146101. (27) Schmid, M.; Reicho, A.; Stierle, A.; Costina, I.; Klikovits, J.; Kostelnik, P.; Dubay, O.; Kresse, G.; Gustafson, J.; Lundgren, E.; Andersen, J. N.; Dosch, H.; Varga, P. Phys. ReV. Lett. 2006, 96, 146102. (28) Xu, Y.; Greeley, J.; Mavrikakis, M. J. Am. Chem. Soc. 2005, 127, 12823. (29) Bocquet, M. L.; Loffreda, D. J. Am. Chem. Soc. 2005, 127, 17207. (30) Hammer, B.; Hansen, L. B.; Nørskov, J. K. Phys. ReV. B 1999, 59, 7413. (31) Greeley, J.; Nørskov, J. K.; Mavrikakis, M. Annu. ReV. Phys. Chem. 2002, 53, 319. (32) Vitos, L.; Ruban, A. V.; Skriver, H. L.; Kollar, J. Surf. Sci. 1998, 411, 186.
J. Phys. Chem. C, Vol. 111, No. 22, 2007 7999 (33) Bengtsson, L. Phys. ReV. B 1999, 59, 12301. (34) Vanderbilt, D. Phys. ReV. B 1990, 41, 7892. (35) Perdew, J. P.; Chevary, J. A.; Vosko, S. H.; Jackson, K. A.; Pederson, M. R.; Singh, D. J.; Fiolhais, C. Phys. ReV. B 1992, 46, 6671. (36) White, J. A.; Bird, D. M. Phys. ReV. B 1994, 50, 4954. (37) Kresse, G.; Furthmuller, J. Comput. Mater. Sci. 1996, 6, 15. (38) CRC Handbook of Chemistry and Physics, 76th ed.; CRC Press: New York, 1996. (39) Ulitsky, A.; Elber, R. J. Chem. Phys. 1990, 92, 1510. (40) Mills, G.; Jonsson, H.; Schenter, G. K. Surf. Sci. 1995, 324, 305. (41) Henkelman, G.; Jo´nsson, H. J. Chem. Phys. 2000, 113, 9978. (42) Henkelman, G.; Uberuaga, B. P.; Jo´nsson, H. J. Chem. Phys. 2000, 113, 9901. (43) Greeley, J.; Mavrikakis, M. Surf. Sci. 2003, 540, 215. (44) Brix, P.; Herzberg, G. J. Chem. Phys. 1953, 21, 2240. (45) Jones, R. O.; Gunnarsson, O. ReV. Modern Phys. 1989, 61, 689. (46) Kurth, S.; Perdew, J. P.; Blaha, P. Int. J. Quantum Chem. 1999, 75, 889. (47) Guo, X. C.; Madix, R. J. Surf. Sci. 2004, 564, 21. (48) Guo, X. C.; Madix, R. J. Surf. Sci. 2002, 496, 39. (49) Guo, X. C.; Madix, R. J. Surf. Sci. 2004, 550, 81. (50) Askgaard, T. S.; Nørskov, J. K.; Ovesen, C. V.; Stoltze, P. J. Catal. 1995, 156, 229. (51) Jones, G. S.; Mavrikakis, M.; Barteau, M. A.; Vohs, J. M. J. Am. Chem. Soc. 1998, 120, 3196. (52) Rydberg, H.; Dion, M.; Jacobson, N.; Schroder, E.; Hyldgaard, P.; Simak, S. I.; Langreth, D. C.; Lundqvist, B. I. Phys. ReV. Lett. 2003, 91, 126402. (53) Li, W. X.; Stampfl, C.; Scheffler, M. Phys. ReV. B 2003, 67, 45408.