TMCnet News
Catalytic Performance of Vanadium MIL-47 and Linker-Substituted Variants in the Oxidation of Cyclohexene: A Combined Theoretical and Experimental Approach [ChemPlusChem](ChemPlusChem Via Acquire Media NewsEdge) The epoxidation of cyclohexene has been investigated on a metal-organic framework MIL-47 containing saturated V+IV sites linked with functionalized terephthalate linkers (MIL-47-X, X=OH, F, Cl, Br, CH3, NH2). Experimental catalytic tests have been performed on the MIL-47-X materials to elucidate the effect of linker substitution on the conversion. Notwithstanding the fact that these substituted materials are prone to leaching in the performed catalytic tests, the initial catalytic activity of these materials correlates with the Hammett substituent constants. In general, substituents led to an increased activity relative to the parent MIL-47. To rationalize the experimental findings, first-principles kinetic calculations were performed on periodic models of MIL-47 to determine the most important active sites by creating defect structures in the interior of the crystalline material. In a next step these defect structures were used to propose extended cluster models, which are able to reproduce in an adequate way the direct environment of the active metal site. An alkylperoxo species V+V O-(OOtBu) was identified as the most abundant and therefore the most active epoxidation site. The structure of the most active site was a starting basis for the construction of extended cluster models including substituents. They were used for quantifying the effect of functionalization of the linkers on the catalytic performance of the heterogeneous catalyst MIL-47-X. Electron-withdrawing as well as electron-donating groups have been considered. The epoxidation activity of the functionalized models has been compared with the measured experimental conversion of cyclohexene. The agreement is fairly good. This combined experimental-theoretical study makes it possible to elucidate the structure of the most active site and to quantify the electronic modulating effects of linker substituents on the catalytic activity. Keywords : ab initio calculations · epoxidation · heterogeneous catalysis · metal-organic frameworks · substituent effects (ProQuest: ... denotes formulae omitted.) Introduction Over the past decade, metal-organic frameworks (MOFs) have emerged as an important new class of nanoporous materials and can be considered as the most recent development and challenge in the field of ordered porous materials.[1] This new class of porous materials has a truly hybrid character, since both inorganic and organic moieties are necessary to build up the framework. In this way, MOFs can provide a bridge be- tween organic polymers and inorganic crystalline zeolites and are sometimes also referred to as a crystalline subclass within the porous coordination polymers. Since the 1990s, this area of chemistry has experienced an almost unparalleled growth, wit- nessed not only by the large number of research and review papers published but also by the ever-expanding scope of the research.[2] MOFs have been reviewed both from a general per- spective[1, 3] and as more application-related. For example, MOFs have been studied extensively for adsorption and gas storage,[4] separation,[5] catalysis,[6] optics and sensing,[7] and electrical/magnetic properties.[8] Recently, MOFs have also been examined for biomedical applications such as controlled drug delivery.[9] In view of the broad variety of applications, more and more attempts have been made to tune the properties of MOFs at the nanoscale level. Various studies have been published on the functionalization of the organic linkers without altering the original topology of the material. A class of MOFs that can be regarded as a prototype of nanoporous materials, in which pore size and functionality on the organic linker can be varied systematically without changing the topology, is MOF-5. It is constructed from octahedral Zn^O^C clusters and benzene linkers, and its three-dimensional pore system can be function- alized with the organic groups Br, NH2,OC3H7,OC5H11, and C2H4.[10] Another example is presented in the work of Serra- Crespo et al. , in which a pure phase amine-functionalized MIL- 101(Al) was synthesized and found to be an excellent candi- date for the selective adsorption of CO2 from methane and N2.[11] Various studies on the substituted versions of MIL-53(M) with M = Cr,[12] Fe,[13] Al,[14] and Sc[15] are also highly interesting, as they have shown different adsorption isotherms and breath- ing behavior. The influence of functionalization was particularly explored for sorption and separation studies of molecules such as CO2,CH4, and so forth. Recently, the UiO-66-X series was in- vestigated for both CO2/CH4 separation and catalysis. Grand canonical Monte Carlo calculations showed that SO3H- and CO2H-functionalized materials exhibit the highest selectivity for CO2/CH4 separation.[16] With respect to catalysis, some of the present authors have shown a remarkable influence of elec- tronic linker modulations of the UiO-66 (UiO-66-X) material on the catalytic activity for the citronellal cyclization.[17] Functional- ized linkers could greatly enhance the catalytic performance of the coordinatively unsaturated Lewis acid sites. Nitro-group substitution resulted in a 56-fold increase in rate. This effect was confirmed by theoretical modeling on a well-chosen defect site, and resulted in a sub- stantial lowering of both the ap- parent and intrinsic free-energy barrier. MOFs have also been explored as heterogeneous catalysts for oxidation reactions using perox- ides or molecular oxygen as oxi- dant. An excellent review on this item was recently presented by Garcia and co-workers.[6b] Hetero- geneous catalytic oxidation reac- tions were also investigated in Co-MOFs (MFU-1 and MFU-2).[18] The liquid-phase oxidation of cy- clohexene, employing tert-butyl hydroperoxide (TBHP) as oxi- dant, has been investigated and for both MOFs large conversions of cyclohexene were noticed. Al- though MFU-1 performed as a truly heterogeneous and robust catalyst, MFU-2 suffered from metal leaching into solu- tion. In a previous study of some of the authors,[19] the same oxi- dation reaction has been studied over a coordinatively saturated vanadium-MOF : V-MIL-47. Apart from cyclohexene oxide 2, the product distribution consisted of the consecutively formed cyclo- hexane-1,2-diol 3 and (radical) side products tert-butyl-2-cyclo- hexenyl-1-peroxide 4 and cyclo- hexen-2-one 5 (Figure 1 a). The catalyst is believed to act partly heterogeneously, which was confirmed by hot filtration experi- ments.[19] After hot filtration of the catalyst, the conversion to- wards the epoxide almost stops while the radical products are still formed. This could be a result of 1) the presence of radical pathways that are also able to produce epoxide[20] and 2) leached V sites in solution.[21] In a recent theoretical study, we performed a complete mechanistic investigation of the possible epoxidation pathways for the homogeneous VO- (acac)2 (acac =acetylacetonate) system.[21] The concerted Sharp- less mechanism turned out to be preferred and the alkylper- oxo species V+VO(L1)(L2)(OOtBu) (with L1,L2 = OH, acac, acetate (OAc), and tert-butoxide anion (OtBu)) were proposed as the most active sites in the epoxidation of cyclohexene. The aim of this study is twofold. First, the effects of linker substitution on the epoxidation performance of the vanadium catalyst MIL-47 are studied in a combined experimental-com- putational effort. Modifications with both electron-withdrawing (OH, F, Cl, Br) and electron-donating (CH3,NH2) groups are in- vestigated. To the best of our knowledge the functionalized MIL-47-X series has not been studied before for this purpose. Second, special attention is devoted to the methodology (clus- ter versus periodic approaches) to describe the structure of the most active catalytic sites in MIL-47. The MIL-47 structure con- sists of a three-dimensional framework, in which each V + IV center is coordinated to four oxygen atoms from four different carboxylate groups and two oxygen atoms from the vanadyl V=O^V=O chain. As such, vanadium is fully saturated in an oc- tahedral coordination mode. The main catalytic pathways that have been modeled in our previous study[20] are shown in Figure 1 b. The reaction cycle starts with an activation step, in which the oxidant TBHP coor- dinates with the vanadium center to form an alkylperoxo spe- cies 7. First, a direct epoxidation pathway may lead to the for- mation of cyclohexene oxide 2 and brings the catalyst to a less active complex 8. Second, a radical mechanism is plausible (7!9 in Figure 1 b). In this route V+V vanadium complexes 9 are formed by homolytic cleavage of the peroxo linkage, which can then be further activated with TBHP. The generated V + V activated complexes 10 and 12 can again epoxidize cyclo- hexene. Both direct epoxidations (A and C) and radical (B) routes are completed by a regeneration step, which closes the cycle by reducing the oxidation state + V back to + IV (Fig- ure 1 b). Many more reactive pathways can occur through the catalytic cycle but the considered pathways here are selected as a proof of principle to determine an accurate catalyst model and investigate the influence of linker substitution. A complete overview of all possible epoxidation pathways can be found in our earlier study on VO(acac)2.[21] The epoxidation reactions (7!8, 10 !11, 12 !11) and radical generation reactions (7! 9, 12 !6) were described previously with the help of extended cluster models[20] (Figure 1 b). In this study, the reaction 10!11 was omitted, as it requires a higher free-energy barrier. It is more likely that complex 10 would react further to give com- plex 12, thereby forming an active V + V epoxidation species. Herein, the creation of defect structures-required for making a vanadium site catalytically active-is investigated taking into account adequately the molecular environment. For this purpose, the methodology for a theoretical investiga- tion of epoxidation reactions has been extended. In the search for the most appropriate active site, the reactions have been modeled in their most natural molecular environment. Un- doubtedly, such a description is best provided by periodic models. They include a larger part of the environment of the active site than cluster models, and hence describe much better geometrical and electronic effects. Furthermore, periodic models give a significant added value to the discussion, as they provide extra insight into the epoxidation behavior of the MIL-47 catalyst, which is of primary importance to identify the exact nature of the active site. Once the structure of the most active site is determined by the periodic calculations, a cluster model can be constructed, which exhibits the same features of the active metal site as in the periodic model. Extended cluster calculations are computationally less demanding and therefore of high utility as a viable alternative for periodic calculations. The extracted cluster structure can be further used as starting basis for the construction of substituted MIL-47-X cluster models. As experimental data on cyclohexene conversion in functionalized MIL-47-X as catalysts are still missing in the liter- ature, these materials were also synthesized in the frame of the goal of this study. The catalytic performance of these mate- rials was investigated and compared carefully with the catalytic activity of the parent MIL-47. The combined theoretical and ex- perimental approach will give us insight into the electronic modulation and its effect on the epoxidation mechanism of substituted MIL-47 materials. Results and Discussion Defects and active sites in MIL-47 materials : theoretical in- sight Periodic model : creation of defects In an undamaged fully saturated MIL-47 framework, the coor- dination sphere around vanadyl [V=O]2 + is completely blocked by terephthalate linkers, which leaves no free position for sub- strate chemisorption or reactions. The catalytic activity of the material in the cyclohexene epoxidation with TBHP can only be started up by the presence of structural defects, thus making the vanadium sites available for reaction with TBHP and cyclohexene. These defects can be present at the surface of the material, but they can also be generated in the interior of the material by removal of at least two terephthalic acid (TA) molecules that are both connected to the same vanadium center. To study defect formation in a realistic molecular envi- ronment, it is mandatory to use periodic calculations that take into account the full crystalline environment of the defect. Ad- ditionally, such calculations enable us to evaluate the energy required for the creation of the various defects. The procedure followed to induce defects in the interior of the material and thus create active metal sites in the pores is outlined in Figure 2 and is as follows. First, the terephthalate linker is re- moved by interacting with water. Terephthalic acid (TA) is re- pelled and in the crystal replaced by terminating hydroxyl groups (this step is easily visualized in the top of Figure 2). Then, a second linker, connected to the same metal site, is also cut. Actually, there are two TA linkers that are candidates to be removed : they are highlighted with red tubes in the second row of Figure 2 (note that Figure 2 displays unit cells) and lead to defects of type A or B. The removal of a second terephtha- late linker requires a substantially lower energy than the 227 kJ mol^1 free reaction energy needed to break up the first TA linker. The defect of type A appears to be energetically fa- vored owing to a beneficial entropic contribution. In principle, there is now sufficient space to allow the TBHP oxidant and the cyclohexene substrate to come close to the vanadium site : a first active site has been created in the interior of the materi- al. In reality it is not excluded that a vanadium site can be leached in solution and driven away from the crystal. This leaching process is indeed observed experimentally.[20,21] It is highly realistic that such defect structures actually occur in the interior of the material. The diagrams at the bottom of Figure 2 visualize two distinct vanadium centers that are re- moved, thus leaving the material in a disjointed structure. The energy cost for the creation of defect structures is given in Figure 2. The required free energies are reasonable. At 323 K the cost is limited to 156 kJ mol^1 for the removal of the second TA linker and 149 kJ mol^1 for the removal of a vanadi- um site VO(OH)2. A more realistic theoretical model would be the consideration of a 2^ 2^ 2 supercell consisting of at least eight unit cells. In this way the defect structure would be better embedded in the material and well isolated from the other defects. With the above cutting proce- dure, large parts of the material can be leached creating large holes in the interior of the crys- tal. An estimate of the requested energy cost to achieve such a defect model is also given. A description of these defects re- quests large supercells, which make their use in further appli- cations (e.g. , simulation of chemical reactions in the pores) almost unfeasible owing to large computational costs. More realis- tic is the introduction of a period- ic slab model. It mimics as well the active sites present at the outer surface of the material as the inner surface of the holes in the material. The proposed peri- odic slab model is created by cutting two TA linkers and two vanadyl groups from the opti- mized 2 ^ 1 ^1 MIL-47 supercell (Figure 3). The defects were taken parallel with the 100 crys- tal plane according to the crys- tallographic information file pro- vided by Barthelet et al.[22] The way defects can be created is not unique, and we can distin- guish between two different structures of open metal sites depending on how the linkers and vanadyl groups have been removed : removal of two TA linkers and two vanadyl groups enclosed in the blue rectangle (see Figure 3) leads to an active site of type I, whereas an active site of type II is characterized by removing the linkers and vanadi- um enclosed in the red rectan- gle. Water is added to satisfy the stoichiometry of the leaching re- action : the two unsaturated va- nadium atoms at the end of the vanadyl chain are terminated with OH groups, thus maintain- ing the + IV oxidation state for the vanadium atoms. Addition- ally, to act as possible active metal sites occurring at the sur- face of the material or at the surface of large holes in the ma- terial, we incorporate a vacuum layer in the model to prevent artificial interactions between the terminal hydroxyl groups and the interrupted vanadyl chain. In this way we dispose of two structurally different periodic slab models. The periodic slab of type II was found to be around 66 kJ mol^1 more stable than the slab of type I. As two defect sites are present in each periodic slab, this difference amounts to 33 kJ mol^1 per active site. This can probably be ascribed to some extra interaction between the terminating vanadyl oxygen and the saturating hydrogen terminating the carboxyl group in the case of MIL-47 slab type II (Figure 3). Periodic model : reactivity for cyclohexene oxidation In this section the activity of the constructed defect sites for cyclohexene oxidation is investigated. The reaction kinetics of the three main reactions 7!8, 7!9, and 12 !11 in respective catalytic cycles A, B, and C (Figure 1) have been calculated in the periodic model. Computational details are given in the Ex- perimental Section (see below). Only apparent energy and free-energy barriers (at 323 K) are presented in Table 1; for completeness all the other results are given in the Supporting Information (Tables S4 and S5). In these particular cases we un- derstand under apparent kinetics that cyclohexene is handled in the gas phase as well as the product after the epoxidation. Following the results reported in Table 1 the direct epoxidation reaction 7!8 on type I defect sites is unlikely to compete with the faster radical generation 7!9 on the basis of the huge dif- ferences in free-energy barrier (101.9 versus 32.8 kJ mol^1). Note also that the radical generation reaction 7!9 is unimolecular and that the lack of entropic contributions makes that energy and the Gibbs free-energy barri- er close to each other (Table 1). This reaction is by far the most favorable reaction path starting from an active V + IV alkylperoxo species. Usage of the PBE func- tional confirms the same qualita- tive trend (Table S5), but the ab- solute values deviate a lot. PBE predictions for apparent barriers are mostly underestimated,[18, 23] which is typical for generalized gradient approximation-type functionals[24] even with inclu- sion of the semiempirical correc- tion for dispersion. PBE0 is much more reliable for chemical kinet- ics, and can best be compared with values predicted by other hybrid functionals, such as B3LYP. For this reason we only retain the PBE0-D2 results for further assessment and compari- son with the extended cluster calculations. We do not have the intention to repeat here the mechanism of the whole cyclohexene oxidation process with TBHP in MIL-47; it has been outlined in the Introduction and discussed extensively in reference [21] . We only stress the im- portance of the three main reactions we take into consider- ation, as they govern the whole catalytic cycle : the epoxidation reactions starting from V + IV vanadium complexes, those start- ing from V + V vanadium complexes, and the radical generation reactions transforming V+IV to V+V. The regeneration step 12! 6 bringing a V+V complex back to oxidation state + IV requires a reaction free energy of some 70 kJ mol^1 [21] and closes the catalytic cycle. Taking into account that all reaction steps are ir- reversible, the rate-determining step is the epoxidation step from an active V + V complex. Regardless of the level of theory or type of active site, our results on the periodic slab model show that the radical generation reaction 7!9 is the fastest, followed by the epoxidation on a vanadium + IV site 7!8, and that 12 !11 is the slowest reaction. Periodic calculations seem to prefer type II active sites for the epoxidation reactions, but the energy differences of a few kJ mol^1 are such that we may not exclude active catalytic centers of type I in an overall assessment. Simulations on active sites in the interior of the material (type A and B, Figure 2) are also instructive. The low apparent barrier of 14.7 kJ mol^1, observed for the 7!8 epoxidation re- action at a vanadium center of type B, largely deviates from the higher barriers predicted in the other models. Similar fea- tures are observed at a vanadium center of type A. This finding can be ascribed to a somewhat constrained confinement effect, as a result of the limited space around the active vana- dium center, and may be regarded as unrealistic with the cy- clohexene molecule described in the gas phase. More specifi- cally, there is a strong stabilizing hydrogen bond between an underlying OH group and the alkylperoxo group in the transi- tion state. At the active site B in the interior of the material, the epoxidation reaction from an active V+V complex (e.g. , 12 !11) can even not take place owing to space limitations. On the other hand, the radical decomposition reaction 7!9 does not suffer from space limitation and needs a free-energy barrier of 23.1 kJ mol^1. In conclusion, reaction sites of the kind encountered in interior defect models of type A and B are too confined in space to be realistic. The cavity in the interior can further be enlarged by continuing the leaching process follow- ing the protocol leading to the construction of active sites of the type illustrated by the last two unit cells displayed in Figure 2. Such a description needs an enlargement of the su- percell, and is scarcely feasible on the computational level. For further assessment, the periodic slab model will be used. Extended cluster model : reactivity of cyclohexene oxidation The various defect structures constructed in the periodic models discussed above enable the construction of extended clusters as shown in Figure 3. In the following we will assess how far such models describe the correct reactivity for cyclo- hexene oxidation pathways, compared with the periodic calcu- lation. This information is interesting as the construction of re- action pathways with extended cluster models is far less de- manding than with periodic calculations. In the latter, localiza- tion of transition states is not straightforward and the determi- nation of normal modes is computationally very demanding. As the defect structures of type A and B are too limited in space, the predicted kinetics are too largely affected by the space confinement. It is unrealistic to expect that their results can be reproduced by cluster predictions. Therefore, we regard the reaction kinetics simulated in the periodic slab as the refer- ence data for the cluster calculations. Results are also tabulat- ed in Table 1. Cluster predictions using the same functional as in the periodic PBE0-D2 calculations are better suited to be submitted to a comparative study. The most significant devia- tions are observed in the barrier of the radical decomposition reaction 7!9 switching the oxidation state for vanadium from + IV to + V. Periodic calculations predict barriers of circa 30 kJ mol^1, independent of the structure of the active site, but clearly much smaller than the 54 kJ mol^1 predicted in the clus- ter model using the same density functional (see Table 1 b and c). This is remarkable as the epoxidation barriers do not vary much. To elucidate this, we display the transition-state struc- tures for both active sites in the periodic slab in Figure 4 (tran- sition-state structures in the extended cluster model are taken up in Figure S3), from which essential features can be extract- ed. The epoxidation reaction described in this study follows a Sharpless-type mechanism. In this scheme the distance be- tween the cyclohexene and the oxygen of the peroxo ligand of the complex plays a prominent role. In the transition state 12!11 this distance is the shortest (^ 2.20 ^) and the O^O distance in the peroxo ligand also turns out to be the largest. These distances are ideal to promote the Sharpless mechanism as the main reaction mechanism for the cyclohexene epoxida- tion. They are regarded as reference values for the transition- state structures resulting from cluster calculations. These dis- tances, reported in Figure S3, do indeed agree fairly well with the periodic predictions. As a result, the free-energy barriers for the 12 !11 epoxidation reaction are also very close to each other (87 versus 89 kJ mol^1) in the case of a reaction site of type II. The cluster predictions obtained with another level of theory (B3LYP/6-311 + g(d,p)-D3) are in line with previous re- sults.[21] To elucidate the discrepancy observed in the barrier of the radical decomposition reaction 7!9 in periodic versus cluster predictions, we notice a significantly larger value for the break- ing O^O distance in the cluster model (1.80 versus 1.63-1.65 ^; Figures S3 and S4). This could probably be ascribed to the dif- ferent functional used for optimization (PBE versus B3LYP) and to the different environment employed in both approaches. However, consistent use of the same DFT method (B3LYP/6- 311 + g(d,p)-D3) in the extended cluster model for geometry optimization and energy calculation yields free-energy barriers of 31.4 and 39.7 kJ mol^1 for the radical reaction 7!9 at the two types I and II of active sites. These numbers are now much more in agreement with the periodic predictions of 32.8 and 28.0 kJ mol^1. The new theoretical data confirm what has been reported before.[20] There are two competitive cyclohexene epoxidation pathways : the first starting from an active V +IVO(OOtBu) com- plex (7!8) and the second through an active V + VO(OOtBu) complex (12 !11). The first complex prefers kinetically to follow a radical generation reaction 7!9 switching the oxida- tion state for vanadium from +IV to +V. A V+V-V+IV recycling step 12 !6 is not excluded, but this step is energetically less favored. Although all pathways are very competitive, we expect the epoxidation route via the V + VO(OOtBu) species 12 to be dominant because of the higher number of active V +V species,[20] and decided to consider the 12 !11 epoxidation re- action as the model reaction in the further investigation of the effect of linker substitution. It is also instructive to discuss the similarities found in the cyclohexene epoxidation for the homogeneous catalytic system VO(acac)2 with TBHP as oxidant. In a recent paper[21] the authors performed an extensive theoretical study involving all possible reaction paths leading to the epoxidation mecha- nism, taking into account various ligand-exchange reactions taking place among the active and inactive complexes in which vanadium may either have oxidation state + IV or + V. Radical decomposition reactions transforming V + IV to V + V spe- cies and vice versa have also been taken into consideration in the extended model space of active and inactive complexes with ligands varying from hydroxyl (OH) to acetylacetonate (acac), acetate (OAc), or tert-butoxide anion (OtBu). A compari- son between both studies is instructive as the similarities are large : only the ligand L differs substantially. In the current study the ligand is bidentate and contains two terephthalate linkers, which simulates the topology in some way, whereas in reference [21] the ligand is essentially less bulky. We can assume that the acac and OAc ligands are best comparable with the TA ligand in the current study. The energy barriers of these two selected active species are also reported in Table 1. The barriers agree to a large extent with those predicted in the heterogeneous catalyst. With the acetate ligand the agree- ment is even more striking. These results confirm our earlier hypothesis that the homogeneous catalytic system VO(acac)2 is a good model system for the heterogeneous MIL-47 catalyst. This was also observed in reference [19] from an experimental point of view. Summarizing, the proposed extended cluster of type II fits the requirements we imposed on the cluster model to mimic at best the periodic results. In addition, type II sites are ther- modynamically more stable than type I sites. In the remainder of this article we will further elaborate on this cluster model. Experimental results on the MIL-47-X series To investigate the influence of substituents on the catalytic performance, a set of catalytic experiments was performed on the parent MIL-47 and MIL-47-X series (X = OH, Br, Cl, F, CH3, NH2). The time-conversion plot for the formation of cyclohex- ene oxide in the first catalytic runs is shown in Figure 5 where- as the values of turnover number (TON), turnover frequency (TOF), and vanadium leaching for the catalysts are presented in Table 2. Owing to the high amounts of vanadium leaching obtained after 6 hours of catalysis for all the compounds except MIL-47 and MIL-47-NH2, only the first hour of the cata- lytic process can be assumed to be mainly heterogeneous. During this period, the TOF values follow a trend: MIL-47-OH > MIL-47-Br>MIL-47-Cl>MIL-47-F>MIL-47-CH3>MIL-47-NH2> MIL-47 (Table 2). Thus, all the substituted MIL-47-X catalysts show higher activity towards the formation of cyclohexene oxide relative to the parent MIL-47, thus indicating a significant effect of the functionalization on the catalytic performance. We tried to correlate this trend of catalytic activity with the electronic effect of the grafted functional groups. In essence, a Hammett-type structure-activity linear free-energy relation- ship (LFER) can be established using the experimental rate constants (kX) and Hammett's s (substituent constant)[25] values of the substituents. The plot of log(kX/kH) versus s should result in a Hammett-Taft equation[26] of the type: log(kX/kH) = 1s + Es, in which 1 and Es represent the reaction constant and steric substituent constant, respectively. As the substituents can never reside in a para position to a carboxyl group in the case of terephthalate substitution, we are limited in examining the LFER only as a function of the meta substituent constant sm. A Hammett plot is displayed in Figure 6. The linear fit through the points belonging to mon- oatomic substituents goes through the origin of the plot, and gives evidence that no steric effects are playing a role (Es = 0). On the contrary, in the case of multiatomic substituents, such as OH, NH2, and CH3, a steric substituent coefficient of Es = 0.71 is observed. The OH substituent constant sm(OH) is determined by deprotonation of a carboxylic acid with OH in the meta po- sition, which indicates that there is no extra hydrogen bond with the carboxyl group connected to the active site. This ex- plains why the OH group follows nicely the trend of the mul- tiatomic substituents. In a realistic situation we can, however, have OH groups both in either a meta or ortho position-or even more likely in a meta-ortho combination-relative to the carboxyl group connected to the active site. One of the com- putational tasks will be to carefully address some possible sub- stituent orientations. Recently, the UiO-66-X series was tested for the cyclization of citronellal. It was found that the defect linkers were respon- sible for the catalytic activity. For the UiO-66-X series, a Ham- mett equation as a function of sm[17] could also be found, in which electron-withdrawing substituents led to an increased rate and electron-donating substituents to a decreased rate. The electronic modulation effect on the conversion was in the same line as in the case of the epoxidation with MIL-47-X : a re- action constant 1 = 2.35 was found compared with 2.09 in the latter case. Furthermore, the Hammett-Taft equation did not lead to any steric substituent effect (Es = 0). Those results were not surprising, since the substituents in a meta position with respect to the carboxyl group connected to the active Zr site cannot have much steric influence on the transition state. After longer periods some catalysts become more active but this may be ascribed to the formation of more defects in the material and larger amounts of leached vanadium centers. At this point, homogeneous pathways may also contribute to the higher activity. Although MIL-47-NH2 showed much less leach- ing than the other substituted MIL-47-X after 6 hours of con- version, this did not lead to a decreased initial activity for MIL- 47-NH2. The initial number of active sites in the MIL-47-X mate- rials should thus be similar. From an experimental point of view it is hard to differentiate between reactions taking place on heterogeneously active sites or on homogeneously working active sites owing to leaching. To illustrate the difference in catalytic activity of both active sites, the epoxidation reaction of cyclohexene has also been performed in a reaction mixture of vanadyl acetylacetonate VO(acac)2 and OH-functionalized terephthalate linkers, thereby creating in this way some "ho- mogeneous" variant of the heterogeneous MIL-47-OH catalyst. By respecting equal amounts of vanadium species and linkers, we create a mixture that proportionally resembles the reaction solution that is found after leaching of fractions of the crystal- line material. We may expect that their catalytic activity will behave in a similar way. The conversion plots are shown in Fig- ure S5 and S6, and they reveal that the homogeneous vanadi- um centers do indeed act as a very efficient homogeneous cat- alyst, which leads to a higher conversion of cyclohexene, espe- cially during the initial stages of the reaction. Figure S6 shows that after 5 hours the homogeneous and heterogeneous plots coincide with each other. The conversion plots are very clear : the formation of the epoxide product is much faster for the homogeneous variant and the convergence pattern is reached after 1.5 hours, thus indicating that there are more active V species in the homogeneous catalyst relative to the MIL-47- OH. Theoretical results on the MIL-47-X series Based on the conclusions drawn above with regard to the structure of the active site and the rate-determining step in the cyclohexene oxidation with TBHP in MIL-47, we investigat- ed the effect of linker substitutions on the epoxidation rate of V + VO(OOtBu)(OH) species 12 on an active site of type II. The basic cluster upon which substitutions can be performed is dis- played in Figure 7 a, in which the transition state is visualized for the 12 !11 epoxidation reaction. In principle the substituents may be placed at various posi- tions as indicated in Figure 7 b. If we assume that each organic linker is functionalized there are in total 16 distinct combina- tions. Each combination corresponds with a distinct cluster, which can be defined by the notation X-i/j in which i and j refer to the position of the substituent X on the left/right linker. By convention the right linker is closest to the alkylperoxo group. Not all positions are equally interesting ; one can expect that the global electronic modulation of the functionalized linkers is larger for ortho/meta positions than for ortho'/meta' posi- tions, with the convention that ortho'/meta' is further away from the active V center (see Figure 7 b). Owing to the high computational cost in considering all possible pairs i/j, we limit our selection to ortho/ortho, meta/meta, ortho/meta, meta/ ortho, ortho'/meta', and meta'/ortho'. The last two combina- tions have been added to verify if its electronic modulation has indeed a negligible effect on the rate constant for epoxida- tion, as intuitively expected. As averaging procedure we intro- duce a Boltzmann weight factor exp ... for each variant i/ j of active site with DGi0=j the Gibbs free energy of formation relative to the most stable isomer of the MIL-47-X series. The average rate constant, which could be associated with the en- semble of active sites in MIL-47-X and which in principle should match at best the experimentally measured value, is then elaborated as [Eq. (1)]: ... (1) For each structural isomer, characterized by the pairwise in- dices i/j, of the MIL-47-X series with functional group X = OH, F, Cl, or NH2, the geometry of the cluster representing the active LV + VO(OOtBu)(OH) site has been optimized in an uncon- strained fashion, that is, all atoms were allowed to relax. In a previous paper on the parent MIL-47 material, we have shown that a description of the reaction in the unconstrained or constrained cluster model leads to similar kinetic and ther- modynamic results.[21] Owing to the existence of many transi- tion states, it is not always trivial to deduce the correct state for the reactants, that is, LV + VO(OOtBu)(OH) and the cyclohex- ene molecule in the gas phase. We therefore applied consis- tently the following procedure. First, the structure of the de- sired transition state was determined. In the second step the cyclohexene molecule was removed and the remaining struc- ture reoptimized. Depending on the specific position of the functional group substituted on the linker, the energies of for- mation will vary among the different clusters under consider- ation. Table 3 reports the relative free energies between the various active LV + VO(OOtBu)(OH) metal sites in each MIL-47-X series. Note that the free-energy differences are enthalpy driven (Table S6). In all cases the position ortho/meta is energetically preferred, and is more stable than the configuration with the two sub- stituents both in the ortho position. Many features lie on the basis of this finding. A first aspect is the position of the sub- stituent with regard to the vanadium of the active site (see Figure 7 b), which determines to a large extent the electrostatic interaction between the two atoms. The second aspect is the presence of hydrogen bonds, which can be formed depending on the spatial orientation of the multiatomic substituent (in the case of OH and NH2) and which stabilize the conformation. Resonance effects take place in the molecular system not only in the phenyl ring of the terephthalate linker, but also in rings that are occasionally constructed by the presence of the hy- drogen bonds. It is a complex interplay of all these features that ultimately determines the final electronic modulation in- duced by the substituent. Not surprisingly, it can vary strongly per case. In an attempt to elucidate the different electronic modula- tion patterns induced by the substituents, we investigated the charge shift occurring in the linkers as a result of the function- alization of the linkers. Several charge population schemes can be used (Mulliken, Becke,[27] Iterative Stockholder,[28] Hirsh- feld,[29] Hirshfeld-i,[30] Hirshfeld-e[31] partitioning). One of the most recent partition schemes is offered by Hirshfeld-e combin- ing electrostatic potential accuracy with transferability.[31] Their charge predictions for the TA linkers (including the two carbox- ylate oxygen atoms ; Figure 7) and for the benzoic linkers (ex- cluding the two carboxylate oxygen atoms) are given in Table 4. We only take up the results for the energetically most favorable positions of the substituents. Different positions may cause some slight variations of the charges on the linkers, but the main trend is given by the type of substituent rather than its position (the whole list is given in the Supporting Informa- tion). With respect to the parent MIL-47, which we regard as the reference, the linkers with electron-withdrawing substitu- ents attract more electronic charge than NH2. The hydroxyl group is the most electronegative by far, enforced by the for- mation of electric dipoles that are formed in the various hydro- gen bonds present in MIL-47-OH (see Figure 8 b). Results de- rived from other charge population schemes are taken up in the Supporting Information. During the reaction there is some slight electronic charge transfer from the catalytic site to the benzoic linkers (also in the case of the parent MIL-47). Even the (positive) charge of the active vanadium center is submitted to some charge drop. In the transition state, electronic charge is transferred from the alkylperoxo complex to the catalyst. The more electron density is present on the benzoic linker, the higher the reaction rate for the epoxidation reaction. The total charge for the benzoic linkers in the parent catalyst amounts to ^0.07 and becomes more negative for all substituents considered in this study. All substituted MIL-47-X catalysts show higher activity both exper- imentally and theoretically towards epoxidation than the parent MIL-47, so there is a clear correlation between the charge concentrated on the linker and the reaction rate con- stant. The transition states for the epoxidation reaction 12 !11 in the parent MIL-47 cluster and some of the substituted clusters are shown in Figures 7 a and 8. The tert-butoxy group is coordi- nated with one vanadium center whereas the cyclohexene is located in a free spot in front of the two linkers. This particular transition state, which involves quite bulky groups from all re- actants, stresses once more that structural defects need to be introduced into the periodic structure to allow enough space for the reaction to take place. In the case of the hydroxo clus- ters (Figure 8), the hydrogen atoms of the OH groups provide extra stabilization for the transition state through hydrogen bonds, which are formed in the epoxidation transition states. The obtained kinetic data are collected in Table 5, together with the enthalpy and entropy contributions to the Gibbs free- energy barriers. All energies are referred to the separated reac- tants and thus "apparent" reaction kinetics are obtained. We notice some significant differences in the free-energy barriers depending on the particular substituent position. They are all energetically driven ; the entropic contributions are fluc- tuating around the value of 58.7 kJ mol^1 corresponding with the unsubstituted species and are not decisive in causing an increase of the catalytic activity. Some substituent positions are manifestly more favorable than others generating a de- crease of the free-energy barrier of more than 10 kJ mol^1. They all prefer the substituent in an ortho/ortho or a meta/ ortho position. The catalytic performance of these functional- ized materials is largely determined by those configurations with substituents on preferential positions. They determine to a large extent the weighted average < kX/kH > stat and lie at the origin of the significant enhancement of the epoxidation rate constant as reported in Table 3. The overall agreement with the experimental TOFX/TOFH ratios is quite satisfactory (last column in Table 3). The largest enhancement is noticed for MIL-47-OH. Theory overestimates the enhancement and this is mainly owing to the two dominant contributions arising from the ortho/ortho and ortho/meta conformations. Probably the orientation of the electron-withdrawing hydroxyl group in our cluster model lies on the origin of a strong hydrogen bond at- tracting too much electron density from the carboxyl group. This is illustrated in the bottom-left structure of Figure 8 dis- playing the transition state with the OH group in the ortho po- sition. There are also other features that require attention : the apparent difference by three orders of magnitude in the pre- diction of the epoxidation rate constant between the hydroxyl substituent in the meta/meta and ortho/meta position on the one hand and in the ortho/ortho and meta/ortho position on the other hand. The mean reaction rate is determined com- pletely by one or two conformations with substituents on pref- erential positions. We will briefly elaborate on the explanation for the mismatch in rates by analyzing the charges of the link- ers for some preferential positions of the functional groups (Table 4). We refer to the Supporting Information for a more complete overview of the charge distributions in the prereac- tive complex and transition state. The charge of the V center does not change on functionalizing the linker with a substitu- ent. The presence of a functional group results in a charge transfer from the alkylperoxo complex to the linkers, and this feature appears to have a significant influence on the epoxida- tion rate constant. The appearance or disappearance of hydrogen bonds in the transition state compared with the prereactive complex may affect the reaction barrier to a large extent. The transition state in the ortho/ortho conformation of MIL-47-OH is clearly stabi- lized by the H bond between the hydrogen of the OH substitu- ent in the ortho position and one of the peroxo oxygen atoms (bottom-left structure in Figure 8), whereas in the meta posi- tion this H bond does not occur (bottom-right structure in Figure 8). The situation is even more dramatic in the case of NH2 as substituent. In the prereactive complex with an amino hydrogen in the ortho position with respect to the vanadium alkylperoxo group, a strong hydrogen bond is formed with the oxygen of the hydroxyl group bound with the metal. This bond disappears completely in the transition state inducing a serious increase of the free-energy barrier by more than 16 kJ mol^1 compared with the meta/ortho conformer. This lies at the origin of the fact that these MIL-47-NH2 species are sig- nificantly better stabilized than the other amino-functionalized MIL-47 variants. Combined with a larger epoxidation rate con- stant these two species (ortho/ortho and meta/ortho) deter- mine the average enhancement with respect to the reference (unsubstituted active site), and this largely explains the ob- served overestimation with respect to experiment (8.8 versus 1.7). Similarly for MIL-47-OH, we have an overestimation with respect to experiment (79.6 versus 10.5). These features are in- herent to the usage of a static approach, in which the energet- ically most stable geometrical configuration is that with the strongest hydrogen bonds. In reality the orientation of the amino group fluctuates and can best be reproduced in molec- ular dynamics simulations. In such a model, we expect that the weighted < kX > /kH factor will decrease, and might come closer to the experimental value. In the case of a halogen-functionalized MIL-47 variant, the issue of a preferential spatial orientation of the substituent is completely missing. No hydrogen bonds are formed, and all conformations show more or less the same stability and cata- lytic activity, independent of the position of the substituent (F or Cl). Theory predicts an enhancement by a factor of 2.6, close to the experimental value of 5.0. Summarizing for a multiatomic substituent, the variations in catalytic activity can largely be affected by the substituent po- sition. The slightly overestimated theoretical enhancement no- ticed at the catalyst MIL-47-NH2 is largely a result of the strong hydrogen bonding of the amino hydrogen atoms with the car- boxylate oxygen atoms. As experiment is unable to disentan- gle the different positions, the theoretically observed optimal positions for the substituents cannot be verified. The experi- mentally measured enhancement in the cyclohexene epoxida- tion performance is a statistical average. In all cases the pres- ence of substituents on the linkers does increase the epoxida- tion rate constant, and this trend is in line with theory. Conclusion In this study, we have tested both experimentally and theoreti- cally the effect of linker modifications on the metal-organic framework MIL-47-containing coordinatively saturated V + IV sites linked by terephthalate linkers-for the epoxidation of cy- clohexene. The use of periodic models is indispensable to take the environment of the active site fully into account, but the high computational cost and the intensive efforts needed to perform correct frequency analyses make it a tool for occasion- al assessment rather than for routine kinetic studies. On the other hand, extended cluster models require a substantially lower computational cost and are better suited to locate tran- sition states and to determine the normal modes, which are in- dispensable for a proper computation of the entropy contribu- tions. The effect of linker substitution was therefore investigat- ed on a cluster model. The size and structure of the extended cluster are determined on the basis of a comparison of the ki- netics predicted by the periodic model and the cluster model. We do indeed find that all substituents, such as F, Cl, Br, CH3, and NH2, give higher rates of epoxidation as was observed ex- perimentally. The largest increase was found for the OH sub- stituent, in agreement with experiment. Theoretically various positions of the substituents are possible and we calculated relative rates compared with the unmodified material based on a statistical method that incorporates the relative probabili- ties of the various clusters. The governing factors for the in- crease may be ascribed to specific interactions in the transition state, such as a hydrogen bond between an electron-donating or -withdrawing functional group in the ortho position with a neighboring carboxylate group connected to the vanadium atom. They generate charge transfers to the linkers. Overall, a good qualitative agreement is found between the electronic modulation of a selection of substituents (OH, F, Cl, NH2) and the experimentally observed catalytic activity for MIL-47-type materials for cyclohexene oxidation. From an experimental point of view, we can conclude that the substituted MIL-47 materials are much more prone to leaching during the epoxi- dation catalysis than the parent MIL-47, which makes them less interesting for industrial applications, in which the stability of the materials is mandatory. Yet, with the recent develop- ment of mixed-metal MOFs with vanadium (e.g. V, Fe), one can expect that the stability of these V-MIL-47 frameworks can still be increased with metal doping, which makes this study a val- uable intermediate report towards more stable heterogeneous epoxidation catalysts. Experimental Section Computational methods Periodic calculations Periodic DFT calculations were performed with the help of the Vienna Ab Initio Simulation Package (VASP).[32] For all periodic sim- ulations, the Brillouin zone sampling was restricted to the G point. Furthermore, the PBE exchange-correlation functional of Perdew, Burke, and Ernzerhof was applied[33] for optimization together with van der Waals corrections of type D2[34] as implemented in VASP 5.2.12. The projector-augmented wave (PAW) approximation[35] to- gether with a plane-wave basis set (kinetic energy cutoff of 600 eV) was used. The RMM-DIIS algorithm was chosen as optimiz- er to converge the atomic forces below 0.01 eV ^^1. A Gaussian smearing[32b] of 0.01 eV was also applied. First, the cell vectors of MIL-47 were optimized starting from XRD data.[22] Second, the periodic slabs were constructed (see Results and Discussion and Figure 3), and the unit cell parameters were maintained throughout the whole computational study. The atoms within each periodic slab were allowed to move freely upon opti- mization of the slab. For the reaction pathways only the active site and the neighboring terephthalic groups were relaxed. During this partial optimization approach, the atomic forces were converged below 0.003 eV ^^1. Defects in the interior of the crystal were modeled with a slightly different procedure ; all atoms were allowed to move freely apply- ing a Gaussian smearing of 0.05 eV and self-consistent field (SCF) criterion of 10^6 eV while the forces were converged below 0.01 eV ^^1. Next, a partial vibrational analysis was performed including only the active site, which consisted of the central vanadium atom and nearest neighbors together with the reactive area (cyclohexene + tBuOO complexes). Displacements in the a, b, and c axes of ^ 0.015 ^ were used in the subsequent partial Hessian calculation. The nature of the minima and transition states was very carefully checked and if residual imaginary frequencies were still present, the following strategies were used to get rid of them : 1) the con- vergence criterion for the electronic SCF problem was increased from 10^6 up to 10^10 eV;[36] and 2) a small displacement along the trajectory of the imaginary frequency was followed by a reoptimiza- tion of the found minimum. We refer to the Supporting Informa- tion for all structures subjected to the partial Hessian calculation. All systems contained V +IV atoms, which were modeled in their high-spin state fixing the value Na^Nb to the number of V + IV atoms in the unit cell. We anticipated that this had a negligible effect on the energy barriers of the material. The optimized geo- metries were used for single-point calculations at the PBE0-D2 level, employing a standard kinetic energy cutoff (400 eV) and SCF criterion (10^5 eV). For the dispersion part of the energy, D2 correc- tions according to Grimme were applied.[34, 37] Both PBE-D2 as PBE0-D2 results are taken up in the Supporting In- formation. Extended cluster calculations on the various reactions within the catalytic cycle Previously, a cluster model[21] had already been investigated by some of the present authors ; it contained two active sites, of which only one was investigated. The constrained character of the MOF could be preserved by fixing the outer carboxyl oxygen atoms. For the optimization of the geometries, an ONIOM (our own n-layered integrated molecular orbital and molecular mechan- ics) approach was used to accurately describe the active site (VO(COO))2 and the reacting molecules on a high level of theory and the surrounding linkers at a lower level of theory (Figure 2). Geometry optimizations were performed with the Gaussian 09 package[31] using the B3LYP hybrid functional.[38] Within the high- level part, the double-zeta Pople basis set 6-31+ G(d) was used for all high-level atoms for which the LANL2DZ effective core potential and basis set was applied.[39] For the sake of clarity, this combined basis will be abbreviated as BS1. Furthermore, the lower-level part was described at the B3LYP/3-21g level. The general optimization procedure could thus be abbreviated as ONIOM(B3LYP/BS1:B3LYP/ 3-21g). Afterwards, energy refinements, including the third version of dispersion corrections according to Grimme,[34] were executed at three different levels of theory (B3LYP/6-311+ g(d,p)-D3, PBE/6- 311+ g(d,p)-D3, and PBE0/6-311 + g(d,p)-D3) to allow for a good comparison with the energies of the periodic simulations at the PBE-D2 and PBE0-D2 level of theory. Note that variations in the structures of the transition state obtained in the two calculations are marginal (compare Figures S3 and S4). The procedure for the extended cluster calculations on the substi- tuted MIL-47-X is similar. Only the application of an ONIOM ap- proach was omitted to have a more reliable optimization of the substituent groups. Using the standard notation "LOT-E"//"LOT-G" (LOT-E and LOT-G are the electronic levels of theory used for the energy and structure optimizations, respectively), all results dis- cussed herein were obtained with the method denoted as "B3LYP/ 6-311g(d,p)-D3//B3LYP/BS1". The calculation of all kinetic and ther- modynamic parameters was performed using the in-house-devel- oped software module TAMkin.[40] In both cluster and periodic approaches, the partial Hessian vibra- tional analysis (PHVA) method,[41] as implemented in the in-house- developed post-processing toolkit TAMkin,[40] was then used to construct the partition functions, which were required to deter- mine the kinetics of the studied reactions. The PHVA method was already applied successfully in previous first-principles studies on kinetics.[21, 23, 42] Use of PHVA is recommended in the constrained cluster calculations, as the outer carboxyl oxygen atoms have been fixed at their crystallographic position. In the case of the periodic calculations, only a partial Hessian was constructed with finite dif- ferences to save computational time. The program ZEOBUILDER was used for the construction of initial geometries for the periodic calculations.[43] Experimental procedures Synthesis and activation of MIL-47-X materials MIL-47 was synthesized according to a route described in the liter- ature.[22] MIL-47-X compounds were prepared in a rapid micro- wave-assisted method as described below. The acronym "as" stands for as synthesized. Synthesis of MIL-47-NH2-as : A mixture of VCl3 (267 mg, 1.74 mmol) and 2-aminoterephthalic acid (308 mg, 1.74 mmol) in water (3 mL) was placed in a Pyrex tube (10 mL). The tube was sealed and heated in a microwave synthesizer (CEM, Discover S) to 150 8Cat 150 W, held under these conditions for 20 min with stirring, then cooled to room temperature. The greenish yellow precipitate was collected by filtration and dried in air. Synthesis of MIL-47-F-as : A mixture of VCl3 (100 mg, 0.64 mmol) and 2-fluoroterephthalic acid (117 mg, 0.64 mmol) in water (2 mL) was placed in a Pyrex tube (10 mL). The tube was sealed and heated in a microwave synthesizer (CEM, Discover S) to 170 8Cat 150 W, held under these conditions for 30 min with stirring, then cooled to room temperature. The greenish yellow precipitate was collected by filtration and dried in air. Synthesis of MIL-47-Cl-as : This compound was obtained as a green- ish yellow powder by the procedure described for MIL-47-F, except the linker used was 2-chloroterephthalic acid (156 mg, 0.64 mmol) instead of 2-fluoroterephthalic acid. Synthesis of MIL-47-Br-as : This compound was obtained as a green- ish yellow powder by the procedure described for MIL-47-F, except the linker used was 2-bromoterephthalic acid (156 mg, 0.64 mmol) instead of 2-fluoroterephthalic acid. Synthesis of MIL-47-CH3-as : This compound was obtained as a greenish yellow powder by the procedure described for MIL-47-F, except the linker used was 2-methylterephthalic acid (115 mg, 0.64 mmol) instead of 2-fluoroterephthalic acid and the tempera- ture used was 1508C instead of 170 8C. Synthesis of MIL-47-OH-as : This compound was obtained as a greenish yellow powder by the procedure described for MIL-47-F, except the linker used was 2-hydroxyterephthalic acid (116 mg, 0.64 mmol) instead of 2-fluoroterephthalic acid. Activation of MIL-47-X-as : A suspension of each of MIL-47-NH2-as and MIL-47-Cl-as in dimethylformamide was heated (125 8C, 1.5 h, MIL-47-NH2-as; 150 8C, 5 h, MIL-47-Cl-as) in an oil bath. The solids were isolated by filtration and heated (80 8C, 2 h, MIL-47-NH2-as; 320 8C, 24 h, MIL-47-Cl-as) under dynamic vacuum to give the acti- vated forms of the compounds. A slow heating ramp (ca. 0.338C min^1)to808C, before being held at this temperature for 2 h, was crucial to activate the MIL-47-NH2-as sample. Each of MIL-47-Br-as, MIL-47-CH3-as, and MIL-47-OH-as was heated directly (300, 300, and 280 8C, respectively) under dynamic vacuum for 24 h to obtain the evacuated forms of the compounds. The MIL-47 sample, synthesized by using conventional electric heating, was calcined at 300 8C for 21.5 h in air to give the thermal- ly activated compound. Catalytic tests During a typical catalytic test, a 100 mL round-bottom flask was charged with chloroform (30 mL, anhydrous) as solvent, cyclohex- ene (5 mL), and 1,2,4-trichlorobenzene (6.2 mL, used as internal standard for the GC analysis). The oxidant used was tert-butyl hy- droperoxide (TBHP) dissolved as a 5.5 m solution in decane. The latter was chosen as oxidant because of the instability of MIL-47 in TBHP in water.[19] The molar ratio of cyclohexene/oxidant was 1:2. All the catalytic tests were performed at a temperature of 50 8C and with an Ar-containing balloon on top of the condenser. Blank reactions at this temperature showed no catalytic conversion of cy- clohexene. The reaction mixture was stirred under an argon atmosphere. Ali- quots were gradually taken out of the mixture, diluted with ethyl acetate (500 mL), and subsequently analyzed by GC with flame ioni- zation detection. After a catalytic run, the catalyst was recovered by filtration on a combined nylon-membrane filter, washed several times with acetone, and vacuum dried overnight. The filtrate was analyzed by X-ray fluorescence to determine the leached vanadi- um. To eliminate experimental noise and GC errors, a polynomial fit (of order 4) through the cyclohexene conversion was performed to determine the turnover frequency (TOF) and turnover number (TON), which were calculated after 30 min and 6 h of reaction time, respectively. In a similar fashion, the amount of cyclohexene oxide produced per catalytic site per hour was also calculated (TONox, h^1) after 30 min to determine the average epoxidation rate per catalyst MIL-47. Acknowledgements M.V. acknowledges funding from both Scientific Research-Flan- ders (FWO) and the research board of Ghent University (BOF). V.V.S. , P.V.D.V., and S.B. acknowledge funding obtained from Ghent University by the GOA grant no. 01G00710. K.L. acknowl- edges financial support from the BOF of Ghent University (post- doctoral grant). V.V.S. and M.W. acknowledge BELSPO in the frame of IAP-PAI P7/05. V.V.S. acknowledges funding from the Eu- ropean Research Council under the European Community's Sev- enth Framework Programme [FP7(2007-2013) ERC grant agree- ment number 240483] . Computational resources and services were provided by Ghent University (Stevin Supercomputer Infra- structure). Grants for computer time at the HLRN cluster of the North-German Supercomputing Alliance and the JUROPA cluster/ FZ Julich are acknowledged. [1] a) G. F^rey, Chem. Soc. Rev. 2008, 37, 191 - 214 ; b) S. Kitagawa, R. Ki- taura, S. Noro, Angew. Chem. 2004, 116, 2388 - 2430; Angew. Chem. Int. Ed. 2004, 43, 2334 - 2375 ; c) J. L. C. Rowsell, O. M. Yaghi, Microporous Mesoporous Mater. 2004, 73, 3 - 14. [2] H.-C. Zhou, J. R. Long, O. M. Yaghi, Chem. Rev. 2012, 112, 673- 674. [3] a) S. T. Meek, J. A. Greathouse, M. D. Allendorf, Adv. Mater. 2011, 23, 249 - 267; b) R. J. Kuppler, D. J. Timmons, Q. R. Fang, J. R. Li, T. A. Makal, M. D. Young, D. Q. Yuan, D. Zhao, W. J. Zhuang, H. C. Zhou, Coord. Chem. Rev. 2009, 253, 3042 - 3066 ; c) S. Horike, S. Shimomura, S. Kitaga- wa, Nat. Chem. 2009, 1, 695 - 704 ; d) U. Mueller, M. Schubert, F. Teich, H. Puetter, K. Schierle-Arndt, J. Pastre, J. Mater. Chem. 2006, 16, 626 - 636. [4] a) K. Sumida, D. L. Rogow, J. A. Mason, T. M. McDonald, E. D. Bloch, Z. R. Herm, T.-H. Bae, J.R. Long, Chem. Rev. 2012, 112, 724-781; b)M.P. Suh, H. J. Park, T. K. Prasad, D.-W. Lim, Chem. Rev. 2012, 112, 782-835; c) H. Wu, Q. Gong, D. H. Olson, J. Li, Chem. Rev. 2012, 112, 836-868; d) G. F^rey, C. Serre, T. Devic, G. Maurin, H. Jobic, P. L. Llewellyn, G. De Weir- eld, A. Vimont, M. Daturi, J.-S. Chang, Chem. Soc. Rev. 2011, 40, 550 - 562; e) L. J. Murray, M. Dinca, J. R. Long, Chem. Soc. Rev. 2009, 38, 1294 - 1314 ; f) S. S. Han, J. L. Mendoza-Cortes, W. A. Goddard IIII, Chem. Soc. Rev. 2009, 38, 1460 - 1476. [5] a) J.-R. Li, J. Sculley, H.-C. Zhou, Chem. Rev. 2012, 112, 869-932; b) J.-R. Li, R. J. Kuppler, H.-C. Zhou, Chem. Soc. Rev. 2009, 38, 1477 - 1504. [6] a) M. Yoon, R. Srirambalaji, K. Kim, Chem. Rev. 2012, 112, 1196 - 1231; b) A. Dhakshinamoorthy, M. Alvaro, H. Garcia, Catal. Sci. Technol. 2011, 1, 856 - 867; c) J. Lee, O. K. Farha, J. Roberts, K. A. Scheidt, S. T. Nguyen, J. T. Hupp, Chem. Soc. Rev. 2009, 38, 1450-1459; d) D. Farrusseng, S. Aguado, C. Pinel, Angew. Chem. 2009, 121, 7638 - 7649 ; Angew. Chem. Int. Ed. 2009, 48, 7502 -7513. [7] a) Y. Cui, Y. Yue, G. Qian, B. Chen, Chem. Rev. 2012, 112, 1126 - 1162; b) L. E. Kreno, K. Leong, O. K. Farha, M. Allendorf, R. P. Van Duyne, J. T. Hupp, Chem. Rev. 2012, 112, 1105 - 1125; c) C. Wang, T. Zhang, W. Lin, Chem. Rev. 2012, 112, 1084-1104; d)J. Rocha, L. D. Carlos, F. A. A. Paz, D. Ananias, Chem. Soc. Rev. 2011, 40, 926 - 940 ; e) M. D. Allendorf, C. A. Bauer, R. K. Bhakta, R. J. T. Houk, Chem. Soc. Rev. 2009, 38, 1330 -1352. [8] a) W. Zhang, R.-G. Xiong, Chem. Rev. 2012, 112, 1163-1195; b) M. Clem- ente-Le^n, E. Coronado, C. Mart^-Gastaldo, F. M. Romero, Chem. Soc. Rev. 2011, 40, 473-497;c)M. Kurmoo,Chem.Soc. Rev. 2009, 38, 1353-1379. [9] P. Horcajada, R. Gref, T. Baati, P. K. Allan, G. Maurin, P. Couvreur, G. F^rey, R. E. Morris, C. Serre, Chem. Rev. 2012, 112, 1232 - 1268. [10] M. Eddaoudi, J. Kim, N. Rosi, D. Vodak, J. Wachter, M. O'Keeffe, O. M. Yaghi, Science 2002, 295, 469 - 472. [11] P. Serra-Crespo, E. V. Ramos-Fernandez, J. Gascon, F. Kapteijn, Chem. Mater. 2011, 23, 2565 - 2572. [12] F. X. Coudert, C. Mellot-Draznieks, A. H. Fuchs, A. Boutin, J. Am. Chem. Soc. 2009, 131, 3442 - 3443. [13] T. Devic, P. Horcajada, C. Serre, F. Salles, G. Maurin, B. Moulin, D. Heur- taux, G. Clet, A. Vimont, J. M. Greneche, B. Le Ouay, F. Moreau, E. Magni- er, Y. Filinchuk, J. Marrot, J. C. Lavalley, M. Daturi, G. F^rey, J. Am. Chem. Soc. 2010, 132, 1127 - 1136. [14] S. Biswas, T. Ahnfeldt, N. Stock, Inorg. Chem. 2011, 50, 9518 - 9526. [15] J. P. S. Mowat, V. R. Seymour, J. M. Griffin, S. P. Thompson, A. M. Z. Slawin, D. Fairen-Jimenez, T. Duren, S. E. Ashbrook, P. A. Wright, Dalton Trans. 2012, 41, 3937 - 3941. [16] Q. Yang, A. D. Wiersum, P. L. Llewellyn, V. Guillerm, C. Serre, G. Maurin, Chem. Commun. 2011, 47, 9603 - 9605. [17] F. Vermoortele, M. Vandichel, B. Van de Voorde, R. Ameloot, M. Waroqui- er, V. Van Speybroeck, D. E. De Vos, Angew. Chem. 2012, 51, 4887 - 4890. [18] M. Tonigold, Y. Lu, A. Mavrandonakis, A. Puls, R. Staudt, J. Moellmer, J. Sauer, D. Volkmer, Chem. Eur. J. 2011, 17, 8671 - 8695. [19] K. Leus, I. Muylaert, M. Vandichel, G. B. Marin, M. Waroquier, V. Van Spey- broeck, P. Van der Voort, Chem. Commun. 2010, 46, 5085 - 5087. [20] K. Leus, M. Vandichel, Y. Y. Liu, I. Muylaert, J. Musschoot, S. Pyl, H. Vrie- linck, G. B. Marin, C. Detavernier, P. V. Wiper, Y. Z. Khimyak, M. Waroquier, V. Van Speybroeck, P. Van der Voort, J. Catal. 2012, 285, 196 - 207. [21] M. Vandichel, K. Leus, P. Van der Voort, M. Waroquier, V. Van Speybroeck, J. Catal. 2012, 294, 1 - 18. [22] K. Barthelet, J. Marrot, D. Riou, G. Ferey, Angew. Chem. 2002, 114, 291 - 294; Angew. Chem. Int. Ed. 2002, 41, 281 - 284. [23] V. Van Speybroeck, J. Van der Mynsbrugge, M. Vandichel, K. Hemelsoet, D. Lesthaeghe, A. Ghysels, G. B. Marin, M. Waroquier, J. Am. Chem. Soc. 2011, 133, 888 - 899. [24] Y. Zhang, W. Pan, W. Yang, J. Chem. Phys. 1997, 107, 7921 - 7925. [25] C. Hansch, A. Leo, R. W. Taft, Chem. Rev. 1991, 91, 165 - 195. [26] R. W. Taft, J. Am. Chem. Soc. 1952, 74, 3120 -3128. [27] A. D. Becke, J. Chem. Phys. 1988, 88, 2547 - 2553. [28] T. C. Lillestolen, R. J. Wheatley, Chem. Commun. 2008, 5909 - 5911. [29] F. L. Hirshfeld, Theor. Chim. Acta 1977, 44, 129 -138. [30] P. Bultinck, C. VanAlsenoy, P. W. Ayers, R. Carb^-Dorca, J. Chem. Phys. 2007, 126, 144111. [31] T. Verstraelen, P. W. Ayers, V. Van Speybroeck, M. Waroquier, J. Chem. Theory Comput. 2013, 9, 2221 - 2225. [32] a) G. Kresse, J. Furthm^ller, Phys. Rev. B 1996, 54, 11169 - 11186 ; b) G. Kresse, J. Furthm^ller, Comput. Mater. Sci. 1996, 6, 15 - 50 ; c) G. Kresse, J. Hafner, Phys. Rev. B 1993, 47, 558-561; d) G. Kresse, J. Hafner, Phys. Rev. B 1994, 49, 14251 - 14269. [33] a) J. P. Perdew, K. Burke, M. Ernzerhof, Phys. Rev. Lett. 1996 , 77, 3865 - 3868 ; b) J. P. Perdew, K. Burke, M. Ernzerhof, Phys. Rev. Lett. 1997, 78, 1396-1396. [34] S. Grimme, J. Comput. Chem. 2006, 27, 1787 - 1799. [35] P. E. Blçchl, Phys. Rev. B 1994, 50, 17953. [36] B. A. De Moor, A. Ghysels, M. F. Reyniers, V. Van Speybroeck, M. Waroqui- er, G. B. Marin, J. Chem. Theory Comput. 2011, 7, 1090 - 1101. [37] T. Kerber, M. Sierka, J. Sauer, J. Comput. Chem. 2008, 29, 2088 - 2097. [38] a)A. D. Becke, J. Chem. Phys. 1993, 98, 5648-5652; b)C.T. Lee, W. T. Yang, R. G. Parr, Phys. Rev. B 1988, 37, 785 - 789. [39] P. J. Hay, W. R. Wadt, J. Chem. Phys. 1985, 82, 270 - 283. [40] A. Ghysels, T. Verstraelen, K. Hemelsoet, M. Waroquier, V. Van Spey- broeck, J. Chem. Inf. Model. 2010, 50, 1736 - 1750. [41] a) A. Ghysels, D. Van Neck, M. Waroquier, J. Chem. Phys. 2007, 127, 164108 ; b) A. Ghysels, V. Van Speybroeck, T. Verstraelen, D. Van Neck, M. Waroquier, J. Chem. Theory Comput. 2008, 4, 614 - 625 ; c) A. Ghysels, V. Van Speybroeck, E. Pauwels, D. Van Neck, B. R. Brooks, M. Waroquier, J. Chem. Theory Comput. 2009, 5, 1203 - 1215. [42] M. Vandichel, D. Lesthaeghe, J. Van der Mynsbrugge, M. Waroquier, V. Van Speybroeck, J. Catal. 2010, 271, 67 - 78. [43] T. Verstraelen, V. Van Speybroeck, M. Waroquier, J. Chem. Inf. Model. 2008, 48, 1530 - 1541. Received : March 26, 2014 Published online on June 17, 2014 Matthias Vandichel,*[a] Shyam Biswas,[b] Karen Leus,[b] Joachim Paier,[c] Joachim Sauer,[c] Toon Verstraelen,[a] Pascal Van Der Voort,*[b] Michel Waroquier,*[a] and Veronique Van Speybroeck*[a] [a] Dr. M. Vandichel, Prof. T. Verstraelen, Prof. M. Waroquier, Prof. V. Van Speybroeck Center for Molecular Modeling Universiteit Gent, Technologiepark 903, 9052 Zwijnaarde (Belgium) E-mail : [email protected] [email protected] [email protected] [b] Dr. S. Biswas, Dr. K. Leus, Prof. P. Van Der Voort COMOC-Centre for Ordered Materials, Organometallics and Catalysis Universiteit Gent, Krijgslaan 281-S3, 9000 Gent (Belgium) E-mail : [email protected] [c] Dr. J. Paier, Prof. J. Sauer Institut f^r Chemie, Humboldt-Universit^t zu Berlin Unter den Linden 6, 10099 Berlin (Germany) * Supporting information for this article is available on the WWW under http://dx.doi.org/10.1002/cplu.201402007. (c) 2014 Blackwell Publishing Ltd. |