Quantum Chemical Study of the Mechanism of Partial Oxidation

1084

J. Phys. Chem. B 1999, 103, 1084-1095

Quantum Chemical Study of the Mechanism of Partial Oxidation Reactivity in Titanosilicate Catalysts: Active Site Formation, Oxygen Transfer, and Catalyst Deactivation Phillip E. Sinclair and C. Richard A. Catlow* The DaVy Faraday Research Laboratory, The Royal Institution of Great Britain, 21 Albemarle Street, London, W1X 4BS ReceiVed: May 7, 1998; In Final Form: July 29, 1998

Density functional theory calculations are presented on the oxidation of ethene over Ti-silicate catalysts within the cluster approximation. Results using the BP86 functional and a DZVP basis suggest that, on interaction with hydrogen peroxide, a Ti(η2-OOH) species is formed, which then donates an oxygen to a weakly bound alkene molecule. A number of different models for the TiIV site were used, and a number of different pathways were considered, e.g., different routes for the formation of the η2 complex, a route via formation of a TidO species and different possibilities for the interaction of the alkene with the TiIV site. Our calculated estimate of the activation barrier for oxidation, in good agreement with experiment, is around 70 kJ mol-1, depending on the route followed. A mechanism for alkene oxidation over TiIV-silica catalysts in the presence of hydroperoxides is proposed, consistent with available experimental and theoretical data and including the effects on the reaction of different solvents, peroxides, additives, and bases.

Introduction The success of aluminosilicate zeolites in numerous industrial catalytic processes1 has stimulated the search for catalysts containing elements other than Al3+ embedded in a silica framework; TiIV is one important example. TiIV-silica partial oxidation catalysts have played a major role for many years in the production of organic feedstocks, e.g., propylene oxide, hydroquinone, and cyclohexanone oxime, but it was not until as late as 1983 with the discovery of titanosilicate-1 (TS-1)2 that the field was transformed. These crystalline, frameworksubstituted TiIV-silicas are active for many oxidation processes under mild conditions ( CH3CH2OH > t-ButOH and have postulated that such strong solvent effects can only be explained by the direct involvement of the solvent in an elementary reaction step.26-28 They suggest that, in the presence of ROOH and a coordinating solvent, the reaction shown in Figure 2 occurs with formation of an active titanium hydroperoxo complex coordinated, within a fivemembered ring, to the solvent (species I). The scheme in Figure 2 has also been used to rationalize the observed acidity of titanosilicates in the presence of H2O2 but not alkyl or aryl hydroperoxides, i.e., when R ) H.26,28,29 Unfortunately, no titanium hydroperoxo compounds have been prepared to date, and no direct evidence for their presence has been recorded. It should also be noted that TS-1 is active for alkene epoxidation in the absence of solvents (although with reduced activity),29 and species I is therefore not necessary for this mode of reaction. An alternative rationale for the observed solvent effects is discussed in detail by Arends et al.4 They consider that the molecular sieve acts as a second solvent, extracting the substrate molecules from the bulk solvent. The exact behavior of the molecular sieve is thus dependent on the size and hydrophobicity

J. Phys. Chem. B, Vol. 103, No. 7, 1999 1085

Figure 2. Formation of solvated titanium peroxo species (I). Clerici and co-workers26-28,56 have advocated this species as being crucial to explaining the behavior of TiIV-silicas oxidation catalysts.

of the pores, e.g., hydrophilic materials such as Ti-Al-β, TiO2SiO2, or Ti-MCM-41 will selectively absorb polar molecules. The question of hydrophobic/hydrophilic effects has been recently addressed by several workers.30,31 Effect of the Sacrificial Oxidant. In addition to host shape selectivity effects, Sheldon et al.25 and Ledon and Varescon32 have demonstrated that the structure of the hydroperoxide, ROOH, can play an important role in determining the product selectivities and the rates of the epoxidation process. For example, electron-withdrawing substituents in the hydroperoxide were shown to increase the rate of epoxidation by enhancing the electrophilic nature of the TiIV-hydroperoxide complex, e.g., species I in Figure 2. These results have been interpreted as indicating that the entire alkyl peroxide moeity (-OOR) is part of the active oxygen-donating complex.32 Effect of Additives. A number of authors have studied the effect on catalysis of adding bases, acids, fluorides, etc. to the reaction mixture.4,26,29 Generally, small amounts of alkali metal hydroxides or acetates improve the yields of epoxide product. Increased concentrations, however, inhibit activity. Metal salts of strong acids, e.g., LiCl and NaNO3, were generally found to have little effect on the rate of epoxidation. Addition of mineral acids improved rates, whereas fluorides strongly reduced them. These effects have been rationalized in terms of species I in Figure 2 being the active oxygen-donating species;26 small amounts of base improve the selectivity by reducing acidity (see Figure 2) and therefore secondary acid-catalyzed reactions. At a higher pH, however, the formation of stable, charged peroxides, species III in Figure 2 which are inert toward alkenes due to electrostatic repulsion has been postulated. Strong acids shift the equilibrium back toward species I, re-establishing catalytic activity. Fluorides were considered to titrate available TiIV sites, preventing active TiIV-peroxide complexes from forming. TiIV Peroxo Complex. The five-membered, solvent-containing ring structure, species I, has been extensively discussed with respect to oxidation activity, solvent effects, and acidic properties, of TiIV-silica catalysts; it has, in fact, been proposed as the active species for oxygen transfer during epoxidation.26-28 It is already known from liquid-phase studies that H2O2 will displace other ligands to form very stable η2-O2 complexes.3 Alkyl hydroperoxides are known to form both η1- and η2-OOR

1086 J. Phys. Chem. B, Vol. 103, No. 7, 1999

Sinclair and Catlow

Figure 4. Suggested pathway for epoxide formation via a bidentate TiIV(η2-OOH) complex as suggested by Karlsen and Scho¨ffel.9 Figure 3. Pathway for epoxide formation from a TiIV-OOH‚‚‚ROH complex as suggested by Clerici and co-workers.26-28

complexes, depending on the binding strength of the other ligands, e.g., solvents, with the stability of the η2 species being enhanced by back-donation to the transition metal center.3 Density functional theory calculations have corroborated this result by showing that for small Ti(OH)4 clusters and in the absence of solvent, TiIV(η2-OOH) is ∼33 kJ mol-1 more stable than the η1 case.9 The formation of η2-peroxo on treatment of H2O2 and TiIVsilicas is well explained; a UV-vis band at 26 000 cm-1 ( i.e., the sample turns yellow) is comparable to that from [TiF5(O2)]3-, attributed to an O22- f Ti4+ charge transfer process.22 These data do not, however, indicate unambiguosly whether the peroxo species in TiIV-silicas is a η2-O2, η2-O2-, hydro-, or alkylperoxide ligand. Neither do the data indicate that this η2-O2 complex is active for partial oxidation processes, a suggestion made by several authors.13,33 Clerici and Ingallina argue that titanium η2-O2 species cannot explain the observed solvent effects, the acidic properties of TS-1/H2O2 mixtures, or the observed behavior toward additives (see above). Further, Ledon and Varescon32 have shown that TiIV(η2-O2) complexes are the poisoned form of active TidO porphyrin catalysts.32 Their studies suggest that a cis-TiIV(OH)(OOR) complex is likely to be the crucial intermediate. Notari3 has also argued that formation of this TiIV-OOR complex (including bidentate binding of -OOR) would require less bond breaking in the TiIV site than formation of η2-O2. In agreement, Karlsen and Scho¨ffel9 have reported density functional theory calculations on small clusters which suggest that formation of TiIV(η2-O2) is unfavorable. Relevant Mechanistic Proposals. Heterolytic mechanisms have been advocated on the basis of liquid-phase analogues and are consistent with experimental observations that the reactions involve electrophilic attack of the peroxo compound on the organic substrate.24,26,34 Clerici and Ingallina,26 for example, report that the reactivity of similarly hindered alkenes in epoxidation reactions over TS-1 follows the trend allyl alcohol < allyl chloride < 1-butene and 3-methyl-1-butene < 2-methyl1-butene < 2-methyl-2-butene, reflecting the change in electron density at the double bond. However, in microporous molecular sieves, shape selectivity can often mask these electronic effects.26 High epoxide yields, high epoxide stereospecificity,26,29 and recent radical clock experiments also indicate a nonradical mechanism.35 Clerici and co-workers26-28 have proposed the pathway for epoxide formation shown in Figure 3. This mechanism (with slight modification) has been studied in detail using DFT cluster methods by Neurock and Manzer.36 They reported calculations on ethene epoxidation with methanol as the coordinating solvent. Ethene was found preferentially to attack the TiOOH complex at the oxygen attached to TiIV; interaction with the other oxygen was more repulsive. Clerici and Ingallina26 have argued that attack at the peroxide oxygen furthest away from TiIV was most

Figure 5. Pathway for epoxide formation as suggested by the study of Adam et al.,38 based on the measurement of the diastereoselectivities of oxidation of various allylic alcohols.

likely on the basis of steric arguments; others, however, disagree.29 In accord with the suggestions of Clerici and coworkers,26-28 the DFT studies show that hydrogen bonding in a five-membered ring stabilizes both the hydrogen peroxide complex and the transition state for epoxide formation; they calculated an overall reaction energy of -220 kJ mol-1 for ethene epoxidation with H2O2 and an activation barrier of 210 kJ mol-1.37 Karlsen and Scho¨ffel have reported a DFT study of the oxidation of ethene via the pathway shown in Figure 4.9 Using small Ti(OH)4 clusters to model the TiIV site, they found that O1 in the TiIV(η2-OOH) complex is highly activated, with an O1-O2 bond distance of 1.52 Å relative to that in H2O2 of 1.49 Å. The η2-OOH complex was found to be 33 kJ mol-1 more stable than the η1-OOH complex, and formation of TiIV(η2-O2) was found to be unfavorable by 90 kJ mol-1. The dominant orbital overlap in the transition state was found to be between the HOMO of ethene and the LUMO of the η2-OOH complex, reflecting the electrophilic nature of the reaction. They calculated an activation barrier of 96 kJ mol-1 for the oxygentransfer step to ethene and an overall energy change of -105 kJ mol-1 for epoxide formation. Adam et al.38 have suggested the mechanism shown in Figure 5 for epoxidation of various allylic alcohols. Their study was based on the comparison of diastereoselectivities of epoxidation over TS-1 and Ti-β with those from established systems. Although this detailed study is very convincing, we consider the results to be inconclusive. For example, their studies were based on the measurement of diastereoselectivities of systems in different solvents. Although they note this problem and use solvents with similar polarities, this process has a marked solvent dependence, especially for coordinating solvents; they compare results from systems in (CH3)2CdO, CH3N, CH2Cl2, and C6H6. Further, Adam et al.38 compared results from the use of a number of different peroxides, TBHP, H2O2, and the urea adduct of H2O2 (UHP); the dependence of this reaction on peroxide has already been noted. The mechanism shown in Figure 5 is

Reactivity in Titanosilicate Catalysts


Figure 6. Schematic representation of a TiIV-octasilsesquioxane showing only the tetrahedral T atoms. The T centers are connected through oxygen. These soluble systems are used to model the active sites in TiIV-silicas.

also incompatible with the use of organic peroxides as oxygen donors (which were, in fact, used in their study). Not only would the process involve transfer of an organic group in place of H* but both UHP and TBHP contain bulky organic groups that would severely hinder approach of the substrate to O*. More details and additional mechanistic proposals are discussed in ref 3. TiIV-Silsesquioxanes. Soluble, molecular titanosilicates or Ti-silsesquioxanes have recently been employed in order to model chemically the active sites of heterogeneous TiIV-silica catalysts.39-42 A schematic picture of one of these molecular systems is shown in Figure 6. The varying behavior in catalytic alkene epoxidation of silsesquioxanes with different “R” groups has been interpreted as indicating that Ti-R bonds in tripodally anchored materials, e.g., hydrolyzed crystalline Ti-silicas or surface-grafted Ti-MCM-41, are stable under the reaction conditions toward hydrolysis.39 Abbenhuis et al.40 reported similar results, indicating that as far as these model systems can be compared to heterogeneous catalysts, the Ti-R bond of tripodally anchored TiR groups appears to be stable toward hydrolysis while for catalytic oxidation to occur a Ti-OSi bond must be cleaved. In summary, evidence is overwhelmingly in favor of TiIV being in tetrahedral coordination within the framework or grafted onto the internal surface of the silica host. The TiIV sites can reversibly expand their coordination number up to 6 in the presence of water or other adsorbable molecules. The unique chemical and catalytic properties of TiIV-silicas are likely to be due to the atomic, and probably random, dispersion of the TiIV centers within these silicas, and the importance of Ti-OSi linkages in titanosilicate-catalyzed oxidation cannot be overstated. Of particular importance to mechanistic studies of alkene oxidation are the marked solvent effects, the behavior in the presence of additives, the effect of varying the peroxide structure, and the convincing arguments for nonradical, electrophilic reactivity of the active oxygen-donating species. A number of important questions, however, still remain. Is alkene binding to TiIV an important requirement for oxidation? What is the active oxygen-donating species? What, if any, is the role of the titanyl, TidO, species? What is the rate-determining step for substrate oxidation? We now present DFT cluster calculations aimed at answering some of the questions posed above. We show that formation of an active η2-OOR complex is as energetically costly as the oxygen-transfer step itself. We present calculations of catalyst deactivation by η2-O2 formation and stress the likelihood of competitive pathways for epoxidation and deactivation. The relevance of the TidO species in epoxidation catalysis will also be discussed. Finally, we will present a general mechanistic scheme rationalizing all of the major observations concerning the behavior of TiIV-silicas in epoxidation reactions.

Figure 7. BP86/DZVP-optimized geometries along the pathway for reaction of TiOH with ethene. The TiOH complex is modeled by (H3SiO)3TiOH. All Si positions were fixed throughout; see text. Energy changes are in kJ mol-1; “ts” indicates that the structure is a transition state.

Methodological Details The nonlocal spin density formulation of density functional theory (DFT),43 as implemented in the code DGauss 3.0,44 has been used throughout. We employed the cluster approximation for all DFT calculations, with dangling bonds being terminated by hydrogens. The Becke45 nonlocal exchange and the Perdew46 nonlocal correlation corrections to the local density functional (Vosko, Wilk, and Nusair local correlation functional47 were applied self-consistently to calculate energies and optimize geometries. All calculations used a valence double-ζ basis set with polarization functions on heavy atoms only (DZVP) and employed an “A1” triple-ζ auxiliary basis on all centers. These basis sets were specifically optimized for DFT calculations.48 Geometry searches were carried out without symmetry constraints and in Cartesian coordinates. However, as in previous studies,6,7 to prevent unrealistic relaxation of terminal SiH3 groups during geometry optimization, only the TiOH cluster (see, for example, Figures 1 and 7) was fully optimized; all other calculations had the Si coordinates fixed at these positions. Smaller clusters, e.g., the Ti(OH)2 structure were built from the TiOH fragment, with the resulting Si coordinates fixed. All other parameters were optimized. We will explicitly note where a different protocol is used. Thus, calculated energy differences are not reaction energies for the equilibrium geometries but reflect more accurately the constrained processes occurring at surface sites. For the purpose of reproducibility, the Cartesian coordinates of the defining cluster models are listed in Table 1. As a result of the large number of species involved, confirmations of minima and transition state structures from independent Hessian calculations were not carried out. Instead, information on curvature was gauged on the basis of the eigenvalues of the approximate Hessian matrix built up during the quasi-Newton optimization procedure. We recognize that Hessian updating schemes can bias the update to maintain desirable Hessian characteristics and intend the use of the approximate Hessian as a rough guide only. All transition state structures had a single, dominant, negative (approximate) Hessian eigenvalue, while all minimum energy structures had zero (approximate) Hessian eigenvalues. The ability of this type of calculation to predict accurately the local geometries of the TiIV site in titanosilicate materials has been noted previously6,9,36 and calculated geometries will not be discussed extensively. All effective partial charges were calculated using the natural bond analysis, as implemented in Gaussian 94;49 basis sets for the

1088 J. Phys. Chem. B, Vol. 103, No. 7, 1999 TABLE 1: Cartesian Coordinates (in Å) of Representative Cluster Models Ti O O O O H Si H H H Si H H H Si H H H Ti O O O O Si H H O Si H H H Si H H H O O H H Si O H H Ti O O O O H Si H H H Si H H H H

4T cluster, TiOH (see, e.g., Figure 7) -1.112354 -0.141602 -0.004085 -2.918477 -0.378197 0.062445 -0.661114 1.516158 0.573352 -0.560180 -0.272627 -1.730312 -0.276710 -1.401510 0.991423 -3.561699 0.354593 -0.026301 0.241451 2.550487 1.540218 1.648914 2.042466 1.645161 -0.384538 2.615120 2.901688 0.241590 3.915033 0.918549 0.863084 -2.566499 1.376962 2.015274 -2.497628 0.418131 0.224565 -3.920912 1.311240 1.354025 -2.307467 2.769838 0.627593 0.039951 -2.869038 0.338970 -0.763531 -4.099375 1.973976 -0.355839 -2.331542 0.641426 1.502295 -3.201672 5T cluster, TiOH (see e.g., Figure 12) -0.646480 -1.094441 0.219303 -2.413353 -1.447912 0.212619 -0.502645 0.554456 1.047343 -0.000958 -0.964894 -1.475125 0.256146 -2.379810 1.113620 0.429559 1.946006 0.922277 1.802664 1.656139 1.444349 -0.250912 3.023338 1.692802 0.494164 2.410121 -0.666032 1.497521 -3.202684 1.880741 2.724285 -2.338761 1.933175 1.811771 -4.446990 1.107404 1.063395 -3.549631 3.272496 1.151200 -0.632792 -2.632450 0.799811 -1.324164 -3.903600 2.488743 -1.063673 -2.108058 -3.044694 -0.671357 0.236346 -3.137730 1.317866 1.934935 -3.839298 0.954345 0.683179 -2.218172 1.010171 1.699591 -4.730055 0.767144 1.057720 1.379557 2.406482 -2.075228 1.179882 1.003124 -2.947519 0.874729 3.519121 -2.922352 2.830871 2.548791 -1.727535 3T cluster, Ti(OH)2 (see e.g., Figure 10) 0.000000 0.000000 0.000000 -0.570290 -0.951834 -1.441726 1.819313 0.000000 0.000000 -0.687662 -0.722836 1.520219 -0.604995 1.718049 0.000000 -1.475444 -1.302184 -1.565500 -1.795053 2.596882 0.793308 -1.740736 2.341093 2.272342 -3.152794 2.206926 0.283605 -1.572719 4.056366 0.531848 -0.726400 -1.235059 3.112257 -2.105136 -1.742680 3.413293 -0.408338 -0.085775 4.024101 0.272518 -2.335365 3.320375 2.349880 0.748587 0.343930

analysis were identical to those used for calculating energies and optimizing geomtries. We emphasize that all energies are internal energies resulting from static electronic structure calculations at stationary points on the adiabatic potential energy surface. Thus, as with all such calculations, thermal and entropic effects are neglected. Results and Discussion We now present the results of our calculations concerning the reaction of an alkene (CH2CH2) and a peroxide (H2O2), both separately and then together, with various model TiIV sites. Calculations concerning catalyst acidity and deactivation will

Sinclair and Catlow then be followed by a general discussion of the mechanism of alkene epoxidation. Interaction with Ethene. From the literature discussed above and our previous studies,6,7 it is clear that, in materials in which TiIV is embedded into the silica framework, at least three forms of dehydrated TiIV sites are possible, (SiO)4Ti, TiOH, and Ti(OH)2; see Figure 1. On a dehydrated TiIV-silica generated by a grafting procedure, only TiOH and Ti(OH)2 sites are likely (it is unlikely that nearby silanols exists in the correct orientations for condensation to (SiO)4Ti). Furthermore, in the presence of ROH adsorbates, TidO species are also energetically accessible.7 We therefore need to consider the interaction of ethene with models representative of these species; we have chosen TiOH and TidO. Figure 7 illustrates the interaction and reaction of a model TiOH species with ethene. A number of attempts were made to locate a stable minimum energy structure for the TiIV(η2CH2CH2) complex; all resulted in the ethene substrate being expelled out of the TiIV coordination sphere. The configuration space search was not exhaustive, but the results suggest that strong bidentate binding is unlikely (although van der Waals interactions, which are not included explicitly, would provide additional weak binding). We consider that the lack of η2 binding probably reflects the energetic cost of distorting the TiIV site geometry relative to the energy gained on binding; we therefore expect a similar result at the (SiO)IVTi and Ti(OH)2 sites. Physisorption of the substrate to TiOH is also unnatractive, with a negligibly weak sorption energy of 2 kJ mol-1 (which would probably vanish on correction for basis set super position errors). Activation of the alkene to form a surface SiOCH2CH3 moiety, datively covalently bonded to a neighboring TidO species, proceeds as shown in Figure 7. In the process shown in Figure 7, cleavage of a Ti-O-Si bond will lead to geometric strain, since the calculations were carried out with the Si positions fixed. In a real system, some relaxation can occur, even within the constraints imposed by a host lattice. This strain will be discussed in more detail below, where it is estimated to contribute around -8 kJ mol-1 to the activation barrier. With this is mind, the calculated absolute activation barrier for formation of TidO from TiOH in the presence of ethene is 114 kJ mol-1; comparison should be made with the calculated activation barrier for its formation in the presence of H2O of 67 kJ mol-1 (59 kJ mol-1 after a similar relaxation correction).7 It is clear, therefore, that CH2CH2 is probably inert under the mild reaction conditions experimentally employed toward silica-embedded titanium (IV) hydroxides. Figure 8 shows the interaction of CH2CH2 with a TidO model. Adsorption of CH2CH2 with the site is two-fold; there is one hydrogen bond to SiOH and one to TidO. The activation barrier of 43 kJ mol-1 for alkene activation suggests that if Tid O species are present, either naturally or from the conversion of titanium hydroxides in the presence of ROH adsorbates, TiOR groups will probably form, depending on the substrate. It is known that the R group can affect the catalytic activity of the TiOR species in TiIV-silsesquioxanes39,42 and that, at temperatures of 100 °C, TS-1 is active for double bond isomerization;50 the formation of TiOR is consistent with this observation. In conclusion, the most important result from this section, the lack of TiIV(η2-CH2CH2) binding, sheds doubt on a number of proposed mechanisms that require a strong substrate interaction.3,51-53 Interaction with H2O2. This section describes calculations on the interaction of our TiIV sites with H2O2. Figure 9 shows the reaction profile for formation of a Ti(η2-OOH) complex from


Figure 8. BP86/DZVP-optimized geometries along the pathway for reaction of TidO with ethene. The TidO complex is modeled by (H3SiO)2Ti(dO)‚‚‚O(H)SiH3. All Si positions were fixed throughout; see text. Energy changes are in kJ mol-1; “ts” indicates that the structure is a transition state.

Figure 9. BP86/DZVP-optimized geometries along the pathway for hydrolysis of the hydroxy function of TiOH with H2O2. The TiOH complex is modeled by (H3SiO)3TiOH. All Si positions were fixed throughout; see text. The energy change marked with an asterix was estimated on the basis of the process using a Ti(OH)2 model; see Figure 10. Energy changes are in kJ mol-1; “ts” indicates that the structure is a transition state.

TiOH and H2O2. Figure 10 shows the same reaction involving a Ti(OH)2 cluster. The pathway involves cleavage of a hydroxy function bonded to TiIV and results in formation of a Ti(η2OOH) complex and water. It is clear from the energetic profiles that there is little difference in reactivity between the TiOH and Ti(OH)2 sites, with the major difference being initial H2O2 binding. The similarity in reactivity is echoed in the effective charges on the O centers in the bond being broken, -0.97 electrons in both cases. The Ti(OH)(η2-OOH) cis-hydroxo(hydroperoxide) complex formed on the Ti(OH)2 site is similar to the one suggested by Ledon and Varescon32 for the active oxygen-donating species in TidO porphyrin catalysts. In the


Figure 10. BP86/DZVP-optimized geometries along the pathway for hydrolysis of the hydroxy function of Ti(OH)2 with H2O2. The Ti(OH)2 complex is modeled by (H3SiO)2Ti(OH)2. All Si positions were fixed throughout; see text. Energy changes are in kJ mol-1; “ts” indicates that the structure is a transition state.

present work, the Ti(η2-OOH) complexes arose naturally during geometry optimization. Given the fact that the η2 complex was calculated to be 33 kJ mol-1 more stable than the η1 complex,9 and that the calculated activation barrier for epoxidation via a fivemembered, solvent-coordinated ring complex26-28 as shown in Figure 2, is very high37 (210 kJ mol-1), we have chosen to proceed, as did Karlsen and Scho¨ffel,9 with the hypothesis that the Ti(η2-OOH) complex is the active oxygen-donating species. Karlsen and Scho¨ffel noted from their calculations that the Ti(η2-OOH) complex is highly activated for oxygen donation, with the O-O distance being 1.52 Å. The current calculations predict a smaller O-O bond distance of 1.48 Å relative to the value in hydrogen peroxide of 1.49 Å. The cause of the difference is unclear. Karlsen and Scho¨ffel used a smaller 1T cluster to model the reaction path, but we consider that this is unlikely to result in such a discrepancy. The effective partial charges on the peroxidic oxygens of -0.488 and -0.462 electrons compared to those in H2O2 of -0.480 does, however, indicate the electrophilic activation of the oxygen directly bonded to titanium. The Ti(η2-OOH) complex satisfies the experimental observation that use of R′OOH oxidants leads to reactivity that is dependent on the R′ group, i.e., R′ is part of the active oxygen-donating species; here R′ ) H. Last, the hypothesis that an η2, rather than an η1, complex is the active oxygen-donating species can explain the retarding effect of strongly coordinating solvents on the reaction, as the latter may block the extra coordination site required for η2-OOH formation. We have previously noted the R dependence on the activity of TiOR-substituted octasilsesquioxanes39,42 and the fact that a Ti-O bond (which in the experiment reported was a Ti-OSi bond) cleaves during epoxidation.40 These two results imply that, in epoxidation reactions with TiOR derivatives, a Ti-OSi bond, rather than a Ti-OR bond, is being broken. To test this hypothesis, we have calculated the reaction path shown in Figure 11 for formation of the active oxygen-donating Ti(η2-OOH) species via fracture of a siloxy bond of the TiOH site. Unlike the previous case, cleavage of a Ti-OSi bond leads to a SiOH group, dative covalently bonded to TiIV. Fixing the Si positions during geometry optimization will therefore lead to structural


Figure 11. BP86/DZVP-optimized geometries along the pathway for hydrolysis of the siloxy function of TiOH with H2O2. The TiOH complex is modeled by (H3SiO)3TiOH. All Si positions were fixed throughout; see text. Energy changes are in kJ mol-1; “ts” indicates that the structure is a transition state.

Figure 12. BP86/DZVP-optimized geometries along the pathway for hydrolysis of the siloxy function of a modified TiOH complex with H2O2. The model contains an extra Si-O-Si bridge in comparison with previous TiOH models. The transition-state structure shown at the top was optimized with all Si fixed at their position in the reactant. The transition state shown at the bottom was optimized with the highlighted Si center relaxed. This procedure allows us to estimate the contribution of the strain to the activation barrier for hydolysis of the Ti-O-Si bond when using the smaller 4T model, where all Si positions are kept fixed; see Figure 11 and the text. Energy changes are in kJ mol-1; “ts” indicates that the structure is a transition state.

strain, since the SiOH unit cannot relax. To estimate the magnitude of this effect, while retaining some constraints to mimic those of the host lattice, the structures shown in Figure 12 were studied. All of the geometrical degrees of freedom in the reactant were optimized. One transition state was optimized with the Si fixed at the positions in the reactant, while the other was optimized with the Si* center shown in Figure 12 relaxed. The contribution of this strain to the activation barrier was thus calculated to be -8 kJ mol-1. We therefore estimate that Ti(η2-OOH) formation at a TiOH or Ti(OH)2 site occurs with an absolute barrier of 48-56 kJ mol-1 or 46 kJ mol-1 for hydroxy or siloxy bond cleavage, respectively, i.e., with little energetic preference for either pathway. Differences in the nature of the TiIV site, e.g., TiOH versus Ti(OH)2, and in its environment (resulting in changes in strain energy) appear to contribute only a small fraction to the barrier.

Sinclair and Catlow

Figure 13. BP86/DZVP-optimized geometries along the pathway for formation of a Ti(η2-OOH) complex from TidO with H2O2. The Tid O complex is modeled by (H3SiO)2Ti(dO)‚‚‚O(H)SiH3. All Si positions were fixed throughout; see text. Energy changes are in kJ mol-1; “ts” indicates that the structure is a transition state.

Last, Figure 13 shows the energetic profile for formation of Ti(η2-OOH) from a TidO complex. It is clear that in the presence of H2O2, TidO will rapidly convert to Ti(η2-OOH). The initial step of Figure 13 is merely adsorption of water, a prerequisite for TidO formation7 (adsorption of a related ROH molecule would also probably suffice. Note that the calculated barrier for formation of TidO from a TiOH complex and water is 67 kJ mol-1,7 minus 8 kJ mol-1 strain energy, to give an estimate of 59 kJ mol-1). In summary, formation of the active oxygen-donating species, Ti(η2-OOH), can proceed with an absolute activation barrier of 48-56 kJ mol-1, by cleavage of a hydroxy function or with a barrier of 46 kJ mol-1, by cleavage of a siloxy function of a titanium (IV) hydroxide complex. Little preference was found for either pathway. We have also shown that TidO species (formed in the presence of protic solvents or occurring naturally) are very reactive toward H2O2, with Ti(η2-OOH) formation being dependent, essentially, on the availability of TidO species. Formation of Ethene Epoxide. For the processes described above, we have shown that the Ti(OH)2 model gives a reaction profile that is very similar to that from the TiOH model. We will therefore employ the computationally cheaper Ti(OH)2 cluster from this point forward. Figure 14 shows the reaction of the active oxygen-donating species with ethene. As noted by Neurock and Manzer,36 interaction of ethene with a Ti(η2-OOH) complex is repulsive. Our calculations are in agreement, and although the [Ti(η2OOH)‚‚‚ethene] complex is a converged minimum energy structure, the positive adsorption energy suggests that a very flat potential energy surface has led to failure of the optimization algorithm. Also in agreement with Neurock and Manzer,36 the peroxidic oxygen center attached to TiIV is more electrophilic than the other (less negative effective partial charge), and it is this oxygen that is donated in the pathway shown. The mechanism is appealing in that the OH of the peroxide that is not donated to the alkene substrate is strongly coordinated to the TiIV center prior to the reaction (Ti‚‚‚O distance of 2.25 Å) and little atomic motion is required in order to “extract” the active oxygen. The activation barrier of 43 kJ mol-1 is comparable to the barrier for formation of the Ti(η2-OOH) complex itself. However, the barrier reported by Karlsen and Scho¨ffel9 for the same step over a Ti(OH)2(η2-OOH) fragment of 96 kJ mol-1 is substantially different. Given the difference in the peroxidic


Figure 14. BP86/DZVP-optimized geometries along the pathway for ethene oxide formation from Ti(OH)(η2-OOH) and CH2CH2. The Ti complex is modeled by (H3SiO)2Ti(OH)OOH. All Si positions were fixed throughout; see text. Energy changes are in kJ mol-1; “ts” indicates that the structure is a transition state.

O-O distance between their work and the current calculations discussed above, we consider that their quantitative results should be treated with care. Using the value of 43 kJ mol-1 for the oxygen-donation step the overall barrier for the catalytic formation of ethene oxide by the reaction cat

H2O2 + CH2CH2 98 H2O + CH2(O)CH2 at a Ti(OH)2 site is calculated to be 66 kJ mol-1 relative to the [Ti(OH)2‚‚‚H2O2] adsorption complex and 28 kJ mol-1 relative to the separate reactants (the estimate is based on bond breaking in a hydroxy function). Using our previously calculated reaction path for formation of the TidO species from TiOH groups in the presence of water,7 and including the contribution of -8 kJ mol-1 for strain energy in the transition state, we estimate that catalysis proceeding via a TidO species involves an overall barrier of 70 kJ mol-1 relative to a [TiOH‚‚‚H2O] adsorption complex and one of 43 kJ mol-1 relative to separate reactants. We note that initial formation of the titanyl species is expected to be strongly solvent dependent. Larger quantities and different protic solvents could result in a much larger role for this pathway. The overall energy of reaction is -204 kJ mol-1 via a Ti(OH)2 site and -189 kJ mol-1 via a TidO (and a TiOH) site. To understand this step further, the diagram shown in Figure 15 was constructed from the Kohn-Sham molecular orbitals (KS MOs) of the isolated fragments and the oxygen-transfer transition state. Although the formal relationship between these single-particle functions and the properties of the real, interacting system is unclear, they have been used by a number of authors to rationalize chemical phenomena,54 and we will proceed without further comment except to note that the KS MOs are complicated in these unsymmetrical cases. Thus, for example, the fact that both ethene π to catalyst σ* combinations are stabilized is simply due to mixing of these orbitals with others of the active site, that is, the labels merely denote that the MO contains a large contribution from the idealized atomic valence combinations. In agreement with Karlsen and Scho¨ffel,9 the


Figure 15. BP86/DZVP Kohn-Sham MO diagram describing the oxygen-transfer step from a TiIV(η2-OOH) active site to ethene. Also included are the MO diagrams for the isolated TiIV(η2-OCH2CH3) and allyl chloride fragments. See text for more details.

results indicate that the ethene HOMO (π) to catalyst LUMO (σ*) is the major interaction that leads to stabilization of the transition state and therefore to oxygen transfer; the catalyst HOMO (π*) to ethene LUMO (π*) separation is larger (-6.03 eV compared to -5.07 eV). This result is consistent with the observed electrophilic nature of the epoxidation step. All other relevant frontier orbitals combine in an approximately nonbonding manner. With this in mind, we have considered the effect of changing both the alkene and active complex; the results are shown on the far right and left of Figure 15. Changing the active complex from TiIV(η2-OOH) to TiIV(η2OOCH2CH3) raises the KS MO energies by 0.3-0.4 eV. Since the dominant interaction leading to oxygen transfer and transition state stabilization involves the catalyst σ* orbital, it is expected that the interaction of TiIV(η2-OOCH2CH3) will be less than that with the TiIV(η2-OOH) complex. This conclusion is supported by the effective charges on the oxygen being donated of -0.462 and -0.474 electrons for the TiIV(η2-OOH) and TiIV(η2-COCH2CH3) complexes, respectively. However, oxygen donation also involves formation of a new Ti-OH bond. The formally nonbonded Ti‚‚‚O distance in TiIV(η2-OOH) is 2.21 Å, whereas that in the TiIV(η2-OOCH2CH3) complex is only 2.19 Å, and we expect that the latter system will be the preferred one for the Ti-O bond forming component of this step. Changing the alkene from ethene to allyl chloride, the latter being less reactive for epoxidation, leads to a small stabilization of the π HOMO and a larger stabilization of the π* LUMO (∆E ) -0.73 eV). Given that the HOMO contributes primarily to the formation of nonbonding orbitals, a small stabilization of this fragment KS MO is unlikely to have a marked effect on the stability of the transition state. However, the substantially more stable π* LUMO results in very similar HOMO-LUMO, catalyst-alkene, and HOMO-LUMO, alkene-catalyst, gaps of 5.30 and 5.18 eV, respectively, raising

1092 J. Phys. Chem. B, Vol. 103, No. 7, 1999 the possibility of a competing channel of reaction. Analysis of the form of the KS MOs suggests that this alternative pathway could involve chlorination of the peroxidic part of the active site with formation of a stable allyl cation, CH2CHCH2+. It is clear that if this alternative channel were important, it would be observable in the product distribution resulting from oxidation of allyl chloride. However, we were unable to find any experimental evidence indicative of such behavior, and the reduced reactivity of allyl chloride relative to unsubstituted alkenes probably derives from other sources; note, however, that Sheldon et al.25 have observed the heterolytic decomposition of the sacrificial oxidant during allyl chloride epoxidation over TiO2-SiO2. It is clear, however, that the less reactive nature of allyl chloride is unlikely to be directly related to changes in the oxygen-donation step. In conclusion, the calculations suggest that epoxidation by a TiIV(η2-OOH) complex will proceed at the mild reaction temperatures experimentally employed. More encouraging is comparison with experimental activation barriers. Although we were unable to aquire data on expoxidation reactions, van der Pol et al.55 have published a barrier of 71 kJ mol-1 for oxidation of straight chain alcohols. Assuming a similar pathway and noting that alcohols and alkenes have different intrinsic activities, agreement with our calculated barriers of 62-71 and 70 kJ mol-1 for reaction via a titanium hydroxide and titanyl species, respectively, is favorable (the former range includes reaction at both model titanium (IV) hydroxides and via cleavage of either hydroxy or siloxy functions). The current calculations also indicate that titanyl groups are as likely to be involved in epoxidations over TiIV-silicas as not, provided that protic solvents are present. Analysis of the molecular orbitals of the reactants and transition states indicate that the major interaction leading to oxygen transfer is between the π HOMO of the alkene and the σ* LUMO of the catalyst. ROOH oxidants, leading to TiIV(η2-OOR) active sites will cause the activity for epoxidation to vary according to the nature of the R group. Last, frontier orbital analysis suggests that allyl chloride could participate in a competing reaction during the epoxidation processes. However, the less reactive nature of this alkene to oxidation is unlikely to be directly related to changes in the oxygen-donation step. We will discuss mechanistic implications in more detail below. Catalyst Deactivation. It has been observed that catalyst deactivation occurs when the TiIV-silica host is preloaded with hydroperoxide.3,35 The catalyst changes color from white to yellow and shows an UV-vis band at around 260 000 cm-1 assigned to formation of Ti(η2-O2) complexes.3 Figure 16 shows the formation of such an η2-O2 complex from Ti(OH)(η2-OOH) by the transfer of hydrogen and the loss of water (note the similarity to the Clerici proposal, Figure 2). In agreement with Karlsen and Scho¨ffel,9 we found that formation of an isolated η2-O2 complex is highly unfavorable. The absolute activation barrier for formation of this η2-O2 function of 63 kJ mol-1 is higher than that reported in the last section for the oxygen-transfer step. This implies, of course, that in the presence of an alkene, epoxidation will proceed, while in its absence a slower, proton-transfer process is operative that deactivates the catalyst. This rationale also explains the effect of base (which will aid deprotonation of the Ti(η2-OOH) complex) and the effect of acid (which will assist reformation of the Ti(η2-OOH) complex) on catalytic activity. Basic species, could, however, also deactivate the catalyst by competitive adsorption at TiIV centers.

Sinclair and Catlow

Figure 16. BP86/DZVP-optimized geometries along the pathway for deactivation of the active oxygen-donating species, Ti(OH)(η2-OOH), in the absence of an oxidizable substrate. The Ti complex is modeled by (H3SiO)2Ti(OH)OOH. All Si positions were fixed throughout; see text. Energy changes are in kJ mol-1; “ts” indicates that the structure is a transition state.

Deactivation of the catalyst on preloading of alkyl hydroperoxides cannot be as easily explained by the process shown in Figure 16, since, for example, transfer of a CH2CH3 group from the -OOR to the -OH ligand has a calculated barrier of 222 kJ mol-1 (for which detailed results are not shown). However, it must be noted that the alkyl groups of the peroxides that have been used to study this effect, e.g., tert-butyl, form carbenium ions that are much more stable, and it is expected that the barrier will then be considerably lower. Acidity Properties. We have already noted the acidic behavior of titanosilicate catalysts in the presence of H2O2. In this respect, our active (η2-OOH) complexes could also be important. The calculated proton affinities of the Ti(OH)(η2OOH) and Ti(OH)2 complexes of the 3T models, which have only two siloxy bridges attached to the TiIV center, are 1333 and 1380 kJ mol-1, respectively. Thus, η2 binding of the -OOH ligand can result in increased acidity relative to the parent titanium (IV) hydroxide complexes and can do so in the absence of a protic solvent. How the acidity of the Ti(η2-OOH) complex compares with that of Clerici’s model (species I in Figure 2) is currently being studied. Epoxidation of Alkenes Summary and Discussion. To summarize, the extensive range of results on the interaction of TiIV sites with ethene and hydrogen peroxide, presented above, have indicated that ethene probably does not bind to the TiIV sites in an η2 fashion. A large absolute barrier of 114 kJ mol-1 for ethene activation (to form a TidO and SiOCH2CH3 complex) suggests that, under the mild conditions used for epoxidation reactions, ethene is probably inert to this type of site. Reaction of ethene with Tid O groups, however, will readily form a TiOCH2CH3 complex (calculated barrier of 43 kJ mol-1), implying that for TidO in, for example, porphyrins32 or TiIV-silicas,7 formation of organic derivatives, TiOR, could be an important catalyst-modifying process. The calculations have also indicated that on interaction of a titanium (IV) hydroxyl with hydrogen peroxide and in the



Figure 17. Possible catalytic cycles for alkene epoxidation at Ti-OR sites based on the individual steps studied in the current work; an alternative active oxygen-donating species would, however, lead to similar paths. See text for more details.

absence of any solvent, a Ti(η2-OOH) species can form. We have shown that formation of this complex via fracture of either a Ti-OH or a Ti-OSi bond can occur with approximately the same activation barrier, 48-56 and 46 kJ mol-1, respectively. In addition, we have shown that reaction at a TiOH site with three siloxy functions to the TiIV center is similar to reaction at a Ti(OH)2 site, which only has two siloxy functions. Small differences in the absolute activation barrier of the latter two TiIV sites as well as the observation of geometrical strain in the system when a Ti-OSi bond is broken suggests that the actual barrier will be somewhat dependent on the nature and environment of the TiIV site. Finally, we showed that reaction of hydrogen peroxide with a TidO function is very favorable but that formation of the Ti(η2-OOH) species is still dominated by generation of the TidO group itself. Following Karlsen and Scho¨ffel9 and Ledon and Varescon,32 the Ti(η2-OOH) complex was suggested as the active oxygen-donating species. Once formation of the active Ti(η2-OOH) species has occurred, formation of the epoxide on interaction of ethene is rapid, with the active peroxidic oxygen being the one attached to the TiIV center. The activation barrier of 43 kJ mol-1 leads to an overall barrier relative to the [hydroxide‚‚‚water] adsorption complex of 62-71 or 70 kJ mol-1 for epoxidation at a titanium (IV) hydroxide or TidO species, respectively (the former includes reaction at both model titanium (IV) hydroxides and via fracture of either a hydroxy or a siloxy bond); that is, the calculations clearly indicate that TidO groups are viable intermediates on the pathway for alkene epoxidation. Unlike the Clerici mechanism, coordination of a protic solvent is not required for activity. Solvent effects are presumed to arise from competitive adsorption at TiIV coordination sites that are required for η2-OOH binding or from the hydrophobic/hydrophilic and shape selectivity properties of the host. Analysis of the molecular orbitals of the reactants and transition state indicate that the major interaction leading to oxygen transfer is that between the π HOMO of the alkene and the σ* LUMO of the

catalyst. ROOH oxidants, leading to TiIV-OOR) active sites will cause the activity for epoxidation to vary according to the nature of the R group. Last, frontier orbital analysis suggests that allyl chloride could participate in a competing reaction during epoxidation processes. However, the reduced activity of this alkene in oxidation reactions is more likely to be due to other factors, e.g., electronic repulsion preventing initial approach of the substrate to the active site and/or formation of stable adsorption complexes of the reactant; it is unlikely that the reduced activity is directly related to changes in the oxygendonation step. The Ti(η2-OOH) complex was shown to be more acidic than the initial titanium (IV) hydroxides and was shown to deactivate to a Ti(η2-O2) complex with a barrier of 63 kJ mol-1. This barrier is significantly higher than that for oxygen transfer and suggests that deactivation of the catalyst via this pathway (leading to a yellow colouring of the catalyst) occurs more slowly than epoxidation and would therefore only occur in the absence of an oxidizable substrate. Catalytic Cycles. To present a complete catalytic mechanism for titanosilicate-catalyzed alkene epoxidation, we must still address the results based on silsesquioxanes mentioned in the Introduction. We recall that these indicate that a Ti-OSi bond is cleaved during epoxidation catalysis40 but that Ti-R bonds, or indeed Ti-OR bonds, in organic derivatives of the silsesquioxanes are not.39,40,42 Figure 17 shows the possible catalytic cycles for alkene epoxidation at Ti-OR sites based on the individual steps studied in the current work. Path 1 assumes cleavage of a Ti-OR function on interaction with R′OOH, while path 2 involves inital cleavage of a siloxy function. Over many cycles, paths 1 and 2a are unlikely to lead to variations in catalytic activity with different initial R groups attached to the TiIV. In addition, path 2b is likely to lead to TiIV leaching. Path 2c however, in which the R group from the original catalyst is retained, would lead to the observed effects. Further, initial calculations of the catalyst “self-repair” step using clusters


Sinclair and Catlow

Figure 18. The proposed mechanism for alkene epoxidation at Ti-OR sites based on the current work and that in the open literature. R represents either tSiO-, H, or an organic function. Step 1 is likely to be faster than step 2 and probably represents the path to the active oxygen-donating species; its formation will be dependent on the concentration of hydroperoxide at the TiIV site and on competitive adsorption with solvent or base for the two coordination sites. Step 3 gives rise to acidity in the presence of hydrogen but not alkyl hydroperoxides. Steps 4 and 5 describe alkene epoxidation and catalyst deactivation, respectively. These two processes are likely to compete, but since the activation barrier is less for step 5 than for step 4, deactivation will only be significant when the concentration of alkene becomes smaller than that of the active oxygen-donating species. Finally, step 6, self-repair of the catalyst and diffusion of the products, closes the catalytic cycle.

similar to those in Figure 12 and with the Si* center relaxed, for the case when R ) R′ ) H, gave an activation barrier of 74 kJ mol-1 (for which detailed results have not been presented), suggesting that this step could be a major contributor to the observed rate. We propose, therefore, the mechanism shown in Figure 18 (based on path 2c) for alkene epoxidation over TiIV-silicas (and octasilsesquioxanes). The mechanism is consistent with the current calculations and the open literature, including solvent effects, acidity, behavior toward base, and catalyst deactivation. The fact that the current calculations do not distinguish markedly between reactions involving breakage of a Ti-OH or a Ti-OSi bond indicates that the dependence on the process of the R group in TiOR derivatives is primarily steric. Finally, we are now in a position to speculate on ways to improve the catalytic effectiveness of TiIV-silicas in the alkene epoxidation processes. Obviously, the host material needs to be chosen with the reactants and products in mind, such that reactant and product shape selectivities as well as hydrophobic considerations are optimal. The first chemical step could, for example, be optimized by the choice of the best protic solvent, which will favor the pathway via a TidO species but that will not compete too effectively with other steps, for example, by strongly coordinating to the TiIV center. Oxygen transfer to the alkene will be accelerated by careful choice of ROOH such that the strength of the nonbonded Ti‚‚‚OR interaction in the η2-OOR complex is maximized without excessively reducing

the electrophilicity of the activated oxygen. This end may also be achieved by use of electron-withdrawing ligands at the TiIV center. The final chemical step, self-repair of the active site, may be rate determining, and optimization of this step is therefore likely to be the most effective. We suggest that electron-withdrawing ligands on the TiIV center, to increase its electrophilicity toward the neighboring SiOH species as well as structural constraints on the potentially bulky product of step 5, may aid this step. Finally, we note that increasing activity via this heterolytic pathway may not have much effect on the epoxidation of alkenes such as allyl chloride. If the reduced activity is due to formation of stable adsorption complexes that reduce the alkene concentration at the active site, then careful catalyst engineering to remove these adsorption sites may help. If, however, the reduced reactivity is due to electronic repulsion between the active site and the alkene, then solving of the problem is more difficult. One possibility may be to include a hydrogen-bonding functionality into the R group of TiOR derivatives in order to aid the approach of the alkene to the active site through hydrogen bond formation. Indeed, it may be profitable in a more general sense to include such a “reactiondirecting” functionality into the active site, for example, to have more regioselective control or even to map chirality into the products. Our studies into these novel approaches continue. Acknowledgment. We thank Dr. M. Neurock (University of Virginia) for discussions and access to unpublished work.

Reactivity in Titanosilicate Catalysts We also thank Prof. R. A. van Santen (TUE, The Netherlands), Prof. Sir J. M. Thomas, Dr. G. Sankar, and Dr. R. Oldroyd (The Royal Institution of Great Britain, UK) for useful discussions. We are grateful to EPSRC for financial support and computer facilities. References and Notes (1) Thomas, J. M.; Bell, R. G.; Catlow, C. R. A. A synoptic guide to zeolites and their catalytic properties. In HandBook of Heterogeneous Catalysis; Ertl, G., Kno¨zinger, H. A., Eds.; VCH: Weinheim, 1997. (2) Taramasso, M.; Perego, G.; Notari, B. US Patent 4410501, 1983. (3) Notari, B. AdV. Catal. 1996, 41, 253. (4) Arends, I. W. C. E.; Sheldon, R. A.; Wallau, M.; Schuchardt, U. Angew. Chem. 1997, 36, 1144. (5) Notari, B. Catal. Today 1993, 18, 163. (6) Sinclair, P. E.; Sankar, G.; Catlow, C. R. A.; Thomas, J. M.; Maschmeyer, T. J. Chem. Phys. B 1997, 101, 4232. (7) Sinclair, P. E.; Catlow, C. R. A. J. Chem. Soc., Chem. Comm. 1997, 1881. (8) Jentys, A.; Catlow, C. R. A. Catal. Lett. 1993, 22, 251. (9) Karlsen, E.; Scho¨ffel, K. Catal. Today 1996, 32, 107. (10) de Man, A. J. M.; Sauer, J. J. Phys. Chem. 1996, 100, 5025. (11) Zicovich-Wilson, C. M.; Dovesi, R. J. Mol. Catal. A 1997, 119, 449. (12) Perego, G.; Bellussi, G.; Corno, C.; Taramasso, M.; Buonomo, F.; Esposito, A. Stud. Surf. Sci. Catal. 1986, 28, 129. (13) Notari, B. Stud. Surf. Sci. Catal. 1988, 37, 413. (14) Bordiga, S. D.; Coluccia, S.; Lamberti, C.; Marchese, L.; Zecchina, A.; Boscherini, F.; Buffa, F.; Genoni, F.; Leofanti, G.; Petrini, G.; Vlaic, G. J. Phys. Chem. 1994, 98, 4125. (15) Blasco, T.; Camblor, M. A.; Corma, A.; Perez-Periente, J. J. Am. Chem. Soc. 1993, 115, 11806. (16) Alba, M. D.; Luan, Z.; Klinowski, J. J. Phys. Chem. 1996, 100, 2178. (17) Sankar, G.; Rey, F.; Thomas, J. M.; Greaves, G. N.; Corma, A.; Dobson, B. R.; Dent, A. J. Chem. Soc., Chem. Comm. 1994, 2279. (18) Blasco, T.; Corma, A.; Navarro, M. T.; Perez-Periente, J. J. Catal. 1995, 156, 65. (19) Lopez, A.; Tuiler, M. H.; Kessler, H.; Guth, J. L.; Delmotte, L.; Popa, J. M. J. Solid State Chem. 1993, 102, 480. (20) Markgraf, S. A.; Halliyal, A.; Bhalla, A. S.; Newnham, R. E.; Prewitt, C. T. Ferroelectrics 1985, 62, 17. (21) Boccuti, M.; Rao, K.; Zeccina, A.; Leofanti, G.; Petrini, G. Stud. Surf. Sci. Catal. 1989, 48, 133. (22) Geobaldo, F.; Bordiga, S.; Zecchina, A.; Giamello, E.; Leofanti, G.; Petrini, G. Cat. Lett. 1992, 16, 109. (23) Maschmeyer, T.; Rey, F.; Sankar, G.; Thomas, J. M. Nature 1995, 378, 159. (24) Sheldon, R. A.; von Doorn, J. A. J. Catal. 1973, 31, 427. (25) Sheldon, R. A.; van Doorn, J. A.; Schram, C. W. A.; de Jong, A. J. J. Catal. 1973, 31, 438. (26) Clerici, M. G.; Ingallina, P. J. Catal. 1993, 140, 71. (27) Bellussi, G.; Carati, A.; Clerici, M.; Maddinelli, G.; Millini, R. J. Catal. 1992, 133, 220.

J. Phys. Chem. B, Vol. 103, No. 7, 1999 1095 (28) Clerici, M. G.; Bellussi, G.; Romano, U. J. Catal. 1991, 129, 159. (29) Khouw, C. B.; Dartt, C. B.; Labinger, J. A.; Davis, M. E. J. Catal. 1994, 149, 195. (30) Ingold, K. U.; Snelgrove, D. W.; MacFaulo, P. A.; Oldroyd, R. D.; Thomas, J. M. Catal. Lett. 1997, 48, 21. (31) Camblor, M.; Corma, A.; Esteve, P.; Martinez, A.; Valencia, S. Chem. Comm. 1997, 795. (32) Ledon, H. J.; Varescon, F. Inorg. Chem. 1984, 23, 2735. (33) Huybrechts, D. R. C.; de Bruycker, L.; Jacobs, P. A. Nature 1990, 345, 240. (34) Jorgensen, K. A. Chem. ReV. 1989, 89, 431. (35) Oldroyd, R. D.; Thomas, J. M.; Maschmeyer, T.; MacFaul, P. A.; Snelgrove, D. W.; Ingold, K. U.; Wayner, D. D. M. Angew. Chem., Int. Ed. Engl. 1996, 35, 2787. (36) Neurock, M.; Manzer, L. E. J. Chem. Soc., Chem. Comm. 1996, 1133. (37) Neurock, M. Personal communication. (38) Adam, W.; Corma, A.; Reddy, T. I.; Renz, M. J. Org. Chem. 1997, 62, 3631. (39) Maschmeyer, T.; Klunduk, M. C.; Martin, C. M.; Shephard, D. S.; Thomas, J. M.; Johnson, B. F. G. J. Chem. Soc., Chem. Commun., in press. (40) Abbenhuis, H. C. L.; Krijnen, S.; van Santen, R. A. J. Chem. Soc., Chem. Comm. 1997, 331. (41) Voigt, A.; Murugavel, R.; Montero, M.; Wessel, H.; Liu, F. Q.; Roesky, H.; Uson, I.; Albers, T.; Parisini, E. Angew. Chem., Int. Ed. Engl. 1997, 36, 1001. (42) Crocker, M.; Herold, B. H. M.; Orpen, A. G. J. Chem. Soc., Chem. Commun. 1997, 2411. (43) Andzelm, J.; Wimmer, E. J. Chem. Phys. 1992, 96, 1280. (44) DGauss 3.0 is a pure DFT code originally developed by Cray Research Inc. Oxford Molecular now develops and distributes DGauss 4.0. (45) Becke, A. D. Phys. ReV. A 1988, 38, 3098. (46) Perdew, J.; Wang, Y. Phys. ReV. B 1986, 33, 8800. (47) Vosko, S. H.; Wilk, L.; Nusair, W. Can. J. Phys. 1980, 58, 1200. (48) Godbout, N.; Andzelm, J.; Wimmer, E.; Salahub, D. R. Can. J. Chem. 1992, 70, 560. (49) Frisch, M.; Trucks, G.; Schlegel, H.; Gill, P.; Cheeseman, M.; Robb, J.; Keith, T.; Petersson, G.; Montgomery, J.; Raghavachari, K.; Al-Laham, M.; Zakrzewski, V.; Ortiz, J.; Foresman, J.; Cioslowski, J.; Stefanov, B.; Nanayakkara, N.; Challacombe, M.; Peng, C.; Ayala, P.; Chen, W.; Wong, M.; Andres, J.; Replogle, E.; Gomperts, R.; Martin, R.; Fox, D.; Binkley, J.; Defrees, D.; Baker, J.; Stewart, J.; Head-Gordon, M.; Gonzalez, C.; Pople, J. Gaussian 94; Gaussian, Inc.: Pittsburgh, PA, 1994. Gaussian 94 is a suite of codes including amongst others HF, MP2, and DFT modules. (50) Robert, R.; Rajamohanan, P. R.; Hegde, S. G.; Chandwaldkar, A. J.; Ratnasamy, P. J. Catal. 1995, 155, 345. (51) Huybrechts, D. R. C.; Buskens, P. L.; Jacobs, P. A. J. Mol. Catal. 1992, 71, 129. (52) Mimoun, H. Angew. Chem., Int. Ed. Engl. 1982, 21, 734. (53) Mimoun, H. Catal. Today 1987, 1, 281. (54) Baerends, E.; Gritsenko, O. J. Chem. Phys. 1997, 101, 5383. (55) van der Pol, A.; van Hooff, J. H. C. Appl. Catal. A 1993, 106, 97. (56) Clerici, M. G. Appl. Catal. 1991, 68, 249.

Quantum Chemical Study of the Mechanism of Partial Oxidation

Recommend Documents