MOF Encapsulation of Ru Oleﬁn Metathesis Catalysts to Block Catalyst Decomposition

: In the present work, a catalyst variation of the second-generation Hoveyda–Grubbs catalyst, particularly the ammonium-tagged Ru-alkylidene metathesis catalyst AquaMet TM , is under study, not simply to increase the e ﬃ ciency in oleﬁn metathesis but also the solubility in polar solvents. Moreover, this ionic catalyst was combined with the metal organic framework (MOF) (Cr)MIL-101-SO 3 − (Na · 15-crown-5) + . We started from the experimental results by Grela et al., who increased the performance when the ruthenium catalyst was conﬁned inside the cavities of the MOF, achieving non-covalent interactions between both moieties. Here, using density functional theory (DFT) calculations, the role of the ammonium N-heterocyclic carbene (NHC) tagged and the conﬁnement e ﬀ ects are checked. The kinetics are used to compare reaction proﬁles, whereas SambVca steric maps and NCI plots are used to characterize the role of the MOF structurally and electronically.


Introduction
Olefin metathesis by Ru-based catalysts certainly has a promising position for finding new applications for industry [1][2][3]. The basic transformations of raw materials in the oil refinery, polymer chemistry, as well as the fine chemical synthesis in the pharmaceutical industry, are the main examples of the industrial-level potential of this transition metal-catalysed reaction [4][5][6][7]. Basically, the goal is to obtain double C-C bonds from other existing ones. Although it may seem like an easy redistribution of C-C double bonds [8,9], a thorough understanding of its mechanism, as well as any unwanted parallel processes that could decrease its efficiency are necessary [10][11][12][13]. If that were not enough, apart from the activity, then an additional effort is needed to control the chemo-, regio-and stereoselectivity of the metathesis [14][15][16].
To improve the performance of olefin metathesis catalysts, attempts to anchor them by means of their ionization led to ammonium-tagged Ru-alkylidene metathesis catalysts [17][18][19]. This represents the addition of the Brönsted acid nature in the framework of olefin metathesis [20][21][22]. Actually, depending on the generation of olefin metathesis catalysts, results were significantly different. The first generation of catalysts has the ammonium group installed in the benzylidene ligand, giving relatively pure metathesis products. Moreover, they are used in polar solvents including water [23] or immobilised on various supports. On the other hand, for the second generation, catalysts tagged in the N-heterocyclic carbene (NHC) ligand became more stable, and consequently the metal contamination levels decreased. Promising for future industrial purposes, the non-dissociating ligand tagged systems were successfully immobilised on zeolites and metal organic frameworks (MOFs). This allows their use in batch and in continuous flow conditions. including water [23] or immobilised on various supports. On the other hand, for the second generation, catalysts tagged in the N-heterocyclic carbene (NHC) ligand became more stable, and consequently the metal contamination levels decreased. Promising for future industrial purposes, the non-dissociating ligand tagged systems were successfully immobilised on zeolites and metal organic frameworks (MOFs). This allows their use in batch and in continuous flow conditions. Furthermore, to increase selectivity, a combination of olefin metathesis catalysts with MOFs by Grela and co-workers [24] imposed a special confinement [25]. Particularly, the MOF (Cr)MIL-101-SO3 − (Na·15-crown-5) + with the catalyst AquaMet TM created a non-covalent immobilization in the MOF [26][27][28].
This study, using density functional theory (DFT) calculations, aims to unveil the role of the NHC-tagged catalyst AquaMet, and type of the confinement of the ruthenium catalyst inside a cavity of the MOF (Cr)MIL-101-SO3 -(Na·15-crown-5) + . The X-ray structure of the MOF included in Figure 1a was obtained from Grela and Chmielewski [26], and the model of the MOF (Figure 1b) consists of a Cr-trimer linked to a six 1,4-benzenedicarboxylic acid (bdc) ligand.

Results and Discussion
The study started with the reaction profile with ethylene of the initiation for the neutral Hoveyda-type catalyst (HOV) displayed in Scheme 1a, for the sake of comparison with the ammonium-tagged ones (Scheme 1b,c). The first step of the reaction profile can be dissociative or concerted [71], especially considering the small nature of the olefin chosen. Figure 2 confirms that the 18e species is too sterically demanding and the system kinetically prefers 2.1 kcal/mol to go first via the 14e species. We applied the method of Martin and co-workers [72] to delicately deal with the overestimation of the entropy when joining several chemical moieties, and proved olefin metathesis for the activation of Ru-based olefin metathesis catalysts [73]. Otherwise, this energy difference

Results and Discussion
The study started with the reaction profile with ethylene of the initiation for the neutral Hoveyda-type catalyst (HOV) displayed in Scheme 1a, for the sake of comparison with the ammonium-tagged ones (Scheme 1b,c). The first step of the reaction profile can be dissociative or concerted [71], especially considering the small nature of the olefin chosen. Figure 2 confirms that the 18e species is too sterically demanding and the system kinetically prefers 2.1 kcal/mol to go first via the 14e species. We applied the method of Martin and co-workers [72] to delicately deal with the overestimation of the entropy when joining several chemical moieties, and proved olefin metathesis for the activation of Ru-based olefin metathesis catalysts [73]. Otherwise, this energy difference would enlarge up to 5.5 kcal/mol without this correction, confirming the dissociative nature of the first step for HOV. Next, from the latter 14e species, the entering olefin bonds to ruthenium overcoming an energy barrier of 3.4 kcal/mol. The corresponding coordination intermediate Ci1 is rather unstable and by overcoming an energy barrier of just only 2.5 kcal/mol the Mcy is reached. This latter metallacycle is interestingly rather unstable, placed 16.7 kcal/mol above the initial catalyst. The opening of the metallacycle is 4.3 kcal/mol more expensive than the previous closure, and leads to Catalysts 2020, 10, 687 3 of 15 a second coordination intermediate Ci2, also less stable, by 7.3 kcal/mol than the first. Finally, via a barrierless process the olefin is released. However, there is probably a more energetic step in-between that consists of a nearly 90 • rotation of the product olefin. Even though the rate determining step (rds) were supposed to be the olefin release according to Solans-Monfort and co-workers [74], the opening of the metallacycle would be here. would enlarge up to 5.5 kcal/mol without this correction, confirming the dissociative nature of the first step for HOV. Next, from the latter 14e species, the entering olefin bonds to ruthenium overcoming an energy barrier of 3.4 kcal/mol. The corresponding coordination intermediate Ci1 is rather unstable and by overcoming an energy barrier of just only 2.5 kcal/mol the Mcy is reached. This latter metallacycle is interestingly rather unstable, placed 16.7 kcal/mol above the initial catalyst. The opening of the metallacycle is 4.3 kcal/mol more expensive than the previous closure, and leads to a second coordination intermediate Ci2, also less stable, by 7.3 kcal/mol than the first. Finally, via a barrierless process the olefin is released. However, there is probably a more energetic step inbetween that consists of a nearly 90° rotation of the product olefin. Even though the rate determining step (rds) were supposed to be the olefin release according to Solans-Monfort and co-workers [74], the opening of the metallacycle would be here.  As shown in Scheme 1b, the HOV was combined with a chloride counteranion, since the substitution of one hydrogen of the backbone of the NHC ligand by a cationic chain led to the neutral ammonium-tagged AquaMet TM . For the sake of consistency, the cationic AquaMet TM+ was also studied (i.e., an ammonium NHC-tagged olefin metathesis catalyst). Moreover, it is its cationic part that deals with the MOF. The comparison of the results displayed in Figure 2, with respect to the homologous ammonium NHC-tagged olefin metathesis catalyst, unveil minimal energy differences [75]. Table 1 confirms that there are insignificant differences not only between the neutral HOV and AquaMet TM , but also with respect to the charged AquaMet TM+ . To point out, the rds goes down by 2.0 would enlarge up to 5.5 kcal/mol without this correction, confirming the dissociative nature of the first step for HOV. Next, from the latter 14e species, the entering olefin bonds to ruthenium overcoming an energy barrier of 3.4 kcal/mol. The corresponding coordination intermediate Ci1 is rather unstable and by overcoming an energy barrier of just only 2.5 kcal/mol the Mcy is reached. This latter metallacycle is interestingly rather unstable, placed 16.7 kcal/mol above the initial catalyst. The opening of the metallacycle is 4.3 kcal/mol more expensive than the previous closure, and leads to a second coordination intermediate Ci2, also less stable, by 7.3 kcal/mol than the first. Finally, via a barrierless process the olefin is released. However, there is probably a more energetic step inbetween that consists of a nearly 90° rotation of the product olefin. Even though the rate determining step (rds) were supposed to be the olefin release according to Solans-Monfort and co-workers [74], the opening of the metallacycle would be here.  As shown in Scheme 1b, the HOV was combined with a chloride counteranion, since the substitution of one hydrogen of the backbone of the NHC ligand by a cationic chain led to the neutral ammonium-tagged AquaMet TM . For the sake of consistency, the cationic AquaMet TM+ was also studied (i.e., an ammonium NHC-tagged olefin metathesis catalyst). Moreover, it is its cationic part that deals with the MOF. The comparison of the results displayed in Figure 2, with respect to the homologous ammonium NHC-tagged olefin metathesis catalyst, unveil minimal energy differences [75]. Table 1 confirms that there are insignificant differences not only between the neutral HOV and AquaMet TM , but also with respect to the charged AquaMet TM+ . To point out, the rds goes down by 2.0 As shown in Scheme 1b, the HOV was combined with a chloride counteranion, since the substitution of one hydrogen of the backbone of the NHC ligand by a cationic chain led to the neutral ammonium-tagged AquaMet TM . For the sake of consistency, the cationic AquaMet TM+ was also studied (i.e., an ammonium NHC-tagged olefin metathesis catalyst). Moreover, it is its cationic part that deals with the MOF. The comparison of the results displayed in Figure 2, with respect to the homologous ammonium NHC-tagged olefin metathesis catalyst, unveil minimal energy differences [75]. Table 1 confirms that there are insignificant differences not only between the neutral HOV and AquaMet TM , but also with respect to the charged AquaMet TM+ . To point out, the rds goes down by 2.0 and 1.5 kcal/mol for AquaMet TM+ and AquaMet TM , respectively (see Figure 3). Thus, kinetically speaking, the ammonium-tagged catalysts should perform slightly better, whereas the thermodynamics are quite similar. On the other hand, the concerted transition state is located even further, and the energy difference with respect to the initial Ru-O bond cleavage rises up from 1.7 kcal/mol for HOV to 8.3 and 9.0 kcal/mol for AquaMet TM+ and AquaMet TM , respectively.  Figure 3). Thus, kinetically speaking, the ammonium-tagged catalysts should perform slightly better, whereas the thermodynamics are quite similar. On the other hand, the concerted transition state is located even further, and the energy difference with respect to the initial Ru-O bond cleavage rises up from 1.7 kcal/mol for HOV to 8.3 and 9.0 kcal/mol for AquaMet TM+ and AquaMet TM , respectively.    Table 2 demonstrates that the key bonds are not that different holding the ammonium-tagged ligand or not. The ruthenium with the ylidene group bonds similarly. In addition, the Ru-O bond is slightly weaker (elongated by 0.007 Å) and translates into a decrease of the energy barrier of the Ru-O bond cleavage by 4.1 and 4.5 kcal/mol for AquaMet TM+ and AquaMet TM , respectively. Thus, the presence of the ammonium-tagged ligand facilitates this bond cleavage. Moving to electronics, conceptual DFT was considered to find out if the nature of the studied olefin metathesis catalysts allows any differentiation. Among definitions, electrophilicity and chemical hardness are the parameters that could fit here. The electrophilicity of the catalysts was evaluated by means of the Parr electrophilicity index, using Equation (1) [76], where µ and η are the chemical potential and the molecular hardness, respectively. Using DFT [77], µ and η for an N-electron system with total electronic energy E and subject to an external potential are defined as the first and second derivatives of the energy with respect to N at a fixed external potential. By Koopmans' theorem [78], µ and η can be approximated with the finite difference formulas of Equation (2), where ε H and ε L are the energies of the highest occupied molecular orbital (HOMO) and the lowest unoccupied molecular orbital (LUMO), respectively.
Regarding the results in Table 3, and focusing on chemical hardness, the addition of the ammonium moiety is demonstrated as sterile, whereas electrophilicity shows subtle differentiating effects. Although electrophilicity is similar for AquaMet TM , it is worth noting the almost doubly positive value for the cationic part of it. Since with the MOF the interaction occurs from this fragment, although it is partially stabilized with the negative charge of a sulfonated group, here the electrophilicity is more incipient for this cationic fragment. This trend is maintained for both 14e species, Act and I14e. The positive charge of the cationic ammonium species remains on the ammonium-tagged ending group, whereas the first sphere containing the atoms around the metal centre is not significantly affected. Table 4 gathers all the information related to natural bond orbital (NBO) charges. The charge on the ruthenium is the same for the first 14e species Act, but interestingly for the initial precatalytic Catalysts 2020, 10, 687 6 of 15 structure Precat, HOV presents a less positive charge; thus, it is less prone to react with potential olefins, but the difference is just 0.003 e − . To check if the MOF has any role in terms of energetics or nature of the reaction intermediates (see Figure 4 to see how the olefin metathesis catalyst interacts with the MOF), CP2K calculations were performed. To point out that the MOF was truncated (see Figure 1b), and the anionic MOF moiety was coupled with the ammonium NHC-tagged olefin metathesis catalyst MOF-AquaMet TM (displayed in Scheme 1c) without the chloride. The binding energy for the Precat was 63.8 kcal/mol between both ionic moieties. In particular, the sulfonate of the MOF and the ammonium group are linked by a series of H-bonds. Table 5 collects the relative energies, collected in gas phase since calculations by CP2K were performed without explicit solvent molecules. The different nature of the calculations with both computational packages (CP2K and Gaussian), despite significant absolute energies, shows results that qualitatively agree, and are very close in terms of electronic energies. Particularly, among the results with CP2K, the introduction of the MOF model, instead of the chloride counteranion, involves a flattening of the potential energy surface. Interestingly, both coordination intermediates (Ci1 and Ci2) are especially stabilized, while the metallacycle is relatively less stable, which would favor olefin metathesis [79].
calculations with both computational packages (CP2K and Gaussian), despite significant absolute energies, shows results that qualitatively agree, and are very close in terms of electronic energies. Particularly, among the results with CP2K, the introduction of the MOF model, instead of the chloride counteranion, involves a flattening of the potential energy surface. Interestingly, both coordination intermediates (Ci1 and Ci2) are especially stabilized, while the metallacycle is relatively less stable, which would favor olefin metathesis [79].

Model Catalyst
Precat Act Ci1 MCy Ci2 I14e   To see if the ammonium-tagged catalysts change their catalytic properties for structural reasons in presence of the MOF around, or if the entry of a substrate is prevented, especially when the metal catalyst is confined inside the MOF, steric maps were made around the metal [80]. In the first sphere, at 3.5 Å, this is where reactivity takes place [81,82]. If a ligand marks the reactivity by ruthenium complexes, this reaction is the effect caused by the NHC ligand. Consequently, the study was performed with all the catalysts studied here. To determine this steric hindrance around the metal [83], topographical steric maps of NHC ligands were obtained by SambVca 2.1 [84], developed by Cavallo and co-workers. The radius of the sphere around the metal centre was set to 3.5 Å, whereas for the atoms we adopted the Bondi radii scaled by 1.17, and a mesh of 0.1 Å was used to scan the sphere for buried voxels [85]. As reported for the interaction of small molecules with carbo-benzenes [86], the study was extended to higher ranks (i.e., 5.0, 8.0, 10.0, 12.0 and 15.0 Å). The elucidation of the steric maps, together with the total and quadrant %V Bur values, give quantitative and qualitative data to predict the reactivity of the metal catalysts. These two-dimensional isocontours represent the interaction surface as topographic maps. Even though the NHC ligand affects up to 12-13 Å, its interaction is basically in the first 3.5 Å length around the metal (see Figure 5b) [87]. However, here we had to check how the MOF could sterically take part in the region around the metal where the olefin enters. From Table 6 it is clear that the MOF alone has no significant participation till a range of  Figure 5c,d), which confirms that the catalysis inside the MOF is the same as outside the MOF for the Ru-based olefin metathesis catalyst. The %V Bur and steric maps including all the atoms, with or without the MOF, are quite similar around the metal (see Supplementary Information for further details). On the other hand, the ammonium-tagged catalyst AquaMet TM is as sterically demanding as HOV, with almost null differences. with a radius of 10 Å. The linking C atom of the NHC is on the z axis, and the metal atom is 2 Å below the plane described by the metal and both chloride atoms. The isocontour curves of the steric maps are given in Å. Due to the low covalent character presented by the interactions between the metal catalyst and the MOF, we computed the NCI plots using the NCIPLOT package of Contreras-Garcia and coworkers [88,89]. The NCI plots allow to observe and qualitatively evaluate the strength of the noncovalent interactions between different moieties, pointing out that they are not available for pseudopotential. This did not represent an issue since the nature of the metal, ruthenium, is not that affected by relativistic effects [90,91]. Figure 6 shows the NCI plot obtained for Precat. Qualitatively we did not notice any significant difference with respect to the other NCI plots (see Table S1).   Due to the low covalent character presented by the interactions between the metal catalyst and the MOF, we computed the NCI plots using the NCIPLOT package of Contreras-Garcia and co-workers [88,89]. The NCI plots allow to observe and qualitatively evaluate the strength of the non-covalent interactions between different moieties, pointing out that they are not available for pseudopotential. This did not represent an issue since the nature of the metal, ruthenium, is not that Catalysts 2020, 10, 687 9 of 15 affected by relativistic effects [90,91]. Figure 6 shows the NCI plot obtained for Precat. Qualitatively we did not notice any significant difference with respect to the other NCI plots (see Table S1). we only observed a rather strong interaction between the model of the MOF and the cationic moiety on the backbone of the NHC ligand, defined by a clear H-bond between an oxygen of the anionic sulfonate and the ending H atom of the cationic chain on the backbone of the NHC ligand (O···H = 2.178 Å). Other relevant interactions came from the hydroxyl of a carboxylic chemical group and the closest isopropyl of the NHC ligand. However, the intensity is rather low in agreement with the weak H-bond among them (O···H = 4.389 Å).

Conclusions
When starting a computational project, the aim is always to achieve challenging objectives, that at an experimental level were not possible with the second-generation Hoveyda-Grubbs catalyst studied here, or that could not be explained as a whole or part. The research group of Grela studied catalytic activity on olefin metathesis [26] confined within a MOF and managed to find an explanation that is both humble and simple, and in line with the results on capsules by Reek and collaborators [92]. Anchoring the catalyst within the MOF is intended to be almost like a heterogeneous catalyst, and to prevent undesired decomposition reactions. Computationally, the results are somewhat similar, with a slight destabilization of the metallacycle that helps to get better catalytic performance [79]. The cationic moiety on the backbone of the NHC is unveiled to positively affect the kinetics, with roughly 2 kcal/mol of stabilization. Despite not finding big differences, we observed that the confinement of the catalysts enhances their catalytic capacity [26]; this could be a result of avoiding the decomposition of olefin metathesis by means of the interaction of two metal moieties [71]. Thus, either we find a vaccine (i.e., a modification of the catalysts to minimize or remove the decomposition reactions) or the MOFs here are a perfect vaccine for this issue. Overall this study confirms that the NHC-tagged catalysts perform better for the Brönsted acid nature, but the enhancement of the reactivity is also due to the confinement of the ruthenium catalyst inside a cavity of the MOF, blocking the potential decomposition of the metal catalyst.

Computational Details
All DFT calculations were performed using the Gaussian09 set of programs [93]. In these calculations, the BP86 of Becke and Perdew was employed [94,95]. The electronic configuration of the In the representations, we plotted the isocontour obtained for a value of 0.5 on the reduced density gradient; and for the colour scale, we used the interval from −0.5 to 0.5 of the second density Hessian eigenvalue, going from blue (attractive) to red (repulsive). From a qualitative point of view, we only observed a rather strong interaction between the model of the MOF and the cationic moiety on the backbone of the NHC ligand, defined by a clear H-bond between an oxygen of the anionic sulfonate and the ending H atom of the cationic chain on the backbone of the NHC ligand (O···H = 2.178 Å). Other relevant interactions came from the hydroxyl of a carboxylic chemical group and the closest isopropyl of the NHC ligand. However, the intensity is rather low in agreement with the weak H-bond among them (O···H = 4.389 Å).

Conclusions
When starting a computational project, the aim is always to achieve challenging objectives, that at an experimental level were not possible with the second-generation Hoveyda-Grubbs catalyst studied here, or that could not be explained as a whole or part. The research group of Grela studied catalytic activity on olefin metathesis [26] confined within a MOF and managed to find an explanation that is both humble and simple, and in line with the results on capsules by Reek and collaborators [92]. Anchoring the catalyst within the MOF is intended to be almost like a heterogeneous catalyst, and to prevent undesired decomposition reactions. Computationally, the results are somewhat similar, with a slight destabilization of the metallacycle that helps to get better catalytic performance [79]. The cationic moiety on the backbone of the NHC is unveiled to positively affect the kinetics, with roughly 2 kcal/mol of stabilization. Despite not finding big differences, we observed that the confinement of the catalysts enhances their catalytic capacity [26]; this could be a result of avoiding the decomposition of olefin metathesis by means of the interaction of two metal moieties [71]. Thus, either we find a vaccine (i.e., a modification of the catalysts to minimize or remove the decomposition reactions) or the MOFs here are a perfect vaccine for this issue. Overall this study confirms that the NHC-tagged catalysts perform better for the Brönsted acid nature, but the enhancement of the reactivity is also due to the confinement of the ruthenium catalyst inside a cavity of the MOF, blocking the potential decomposition of the metal catalyst.

Computational Details
All DFT calculations were performed using the Gaussian09 set of programs [93]. In these calculations, the BP86 of Becke and Perdew was employed [94,95]. The electronic configuration of the studied molecules was described with the standard split valence basis set with a polarization function for H, C, Cl, N and O (Def2SVP keyword in Gaussian) of Ahlrichs and co-workers [96]. The quasi relativistic, small-core, effective core potential of Stuttgart/Dresden, with an associated valence basis set (SDD keyword in Gaussian) was used for Ru atom [97].
Solvent effects on the potential energy surfaces of the oligomerization cycle were estimated based on the polarizable continuum solvation model (PCM) using dimethyl carbonate (DMC) as solvent [98,99], the B3LYP, hybrid GGA functional of Becke-Lee, Parr, and Yang [100] and triple-ζ basis set (cc-pVTZ keyword in Gaussian) [101], together with the Grimme D3 correction term for the electronic energy [102]. Thus, the free energies discussed throughout the manuscript include the electronic energies in solvent that are corrected by the thermal corrections calculated in the gas phase at T = 323.15 K and P = 1354 atm [72].
The unit cell of Mil-101 is excessively large for periodic DFT calculations, therefore, we performed the calculations on a fragmented cluster of the MOF (Figure 1). All the simulations of the cluster were performed with CP2K [103] at density functional level of theory. The semi-local PBE functional of Perdew, Burke and Ernzerhof was adopted [104] using the DZVP-MOLOPT-SR-GTH Gaussian basis set for all the atom types [105], and a cutoff of 450 Ry for the plane wave auxiliary basis set. Atom positions were optimised converging the force up to 5 × 10 −3 a.u. and the electronic structure up to 1× 10 −3 a.u. The cubic simulation box size was set to 25 × 25 × 25 Å 3 ensuring isolated molecule simulation.