A computational perspective of olefins metathesis catalyzed by N-heterocyclic carbene ruthenium (pre)catalysts

Abstract

In recent years olefin metathesis catalyzed by N-heterocyclic carbene ruthenium complexes has attracted remarkable attention as a versatile tool to form new C=C bonds. The last developed (pre)catalysts show excellent performances, and this achievement has been possible because of continuous experimental and computational efforts to understand the laws controlling the behavior of these systems. This perspective rapidly traces the ideas and discoveries that computational chemistry contributed to the development of these catalysts, with particular emphasis on catalysts presenting a N-heterocyclic carbene ligand.

---

Raffaele Credendino

Raffaele Credendino earned his PhD in Chemistry in 2009, working under the supervision of Vincenzo Busico at University of Naples, Italy, defending a computational thesis on heterogeneous Ziegler–Natta catalyst systems. During his PhD he spent six months in the University of Manitoba, Canada, in the group of Peter Budzelaar. In 2010, he joined the group of Luigi Cavallo at the University of Salerno as a postdoctoral fellow. Here he follows projects on asymmetric and ring closing metathesis and heterogeneous Ziegler–Natta catalysis.

Albert Poater

Born in Banyoles, Catalonia, Spain, in 1979. He received his PhD in computational and theoretical chemistry at the University of Girona in 2006, under the supervision of Miquel Solà and Miquel Duran. His thesis was based on DFT calculations of bioinorganic complexes. Later he joined the group of Luigi Cavallo at the University of Salerno, working on olefin metathesis. In 2010 he was awarded a Ramón y Cajal fellowship in Institut Català de Recerca de l'Aigua (ICRA) in Girona, Spain. His current research interests are olefin metathesis, water oxidation catalysts and gold catalysis.

Francesco Ragone

Francesco Ragone was born in 1982 in Bergamo, Italy. He took his PhD in chemistry in 2011 at the University of Salerno, studying the reactivity of NHCs with a computational approach, under the supervision of Luigi Cavallo. During his PhD he has spent nine months in the Universität Zürich in the group of Jürg Hutter under the supervision of Marcella Iannuzzi. His main research interests are homogeneous catalysis, gold nanoparticles and programming.

Luigi Cavallo

Luigi Cavallo earned his PhD in Chemistry in 1991 at the University of Naples, Italy, under the supervision of Paolo Corradini working on the modelling of stereoselective olefins polymerization. After a postdoc experience with Tom Ziegler at the University of Calgary, Canada, he was appointed as Assistant Professor at the University of Naples. In 2002 he moved as Associate Professor at the University of Salerno. His main work interest is to unravel the mechanism of industrially relevant reactions with a computer.

---

1. Introduction

1.1. The relevance of Ru-catalyzed olefins metathesis

Olefin metathesis is at the heart of chemistry in its traditional meaning, since it is a powerful and versatile synthetic tool for the formation of new C=C bonds. The basic transformation involved in olefin metathesis is the scrambling of the alkylidene groups around two C=C bonds, see Fig. 1. Since the same type of bonds is present in reactants and products, olefins metathesis is often an almost thermoneutral reaction. The mechanism operative in olefins metathesis is that proposed by Chauvin, also displayed in Fig. 1.¹ The key step is formation of the metallacycle intermediate, whose opening in the backward or the forward direction connects reactants and products. To date, there are two families of well working and well developed catalysts, based on Mo or Ru as the catalytic metal. Their experimental performances have been extensively reviewed.^2–8 The general idea is that Mo-based catalysts are more active but more delicate, in the sense that they are less tolerant to other functional groups in the substrate, while Ru-based catalysts are less active but tolerate the presence of almost any other functional group in the substrate. Research activity is dedicated to solve the weaknesses of both families of catalysts, and progresses in the field are continuous.

Fig. 1 Scope of olefins metathesis and the Chauvin mechanism.

Ru-catalyzed metathesis can span a variety of applications, the most typical are shown in Fig. 2. Ring closing metathesis (RCM) is a powerful tool to close rings of various dimensions, particularly medium size rings, and is often relevant in pharmaceutical applications.^9,10 Ring opening metathesis polymerization (ROMP) represents one of the first industrial applications of Ru-catalyzed metathesis, through the synthesis of poly-dicyclopentadiene, poly-DCPD.⁶ Cross metathesis (CM) of terminal olefins might seem the simplest metathesis reaction, but controlling the products distribution, as well as achieving quantitative formation of tetrasubstituted or Z-olefins, still is a challenge.^11–13 CM has the additional challenge that it does not benefit the entropic driving force shifting the equilibrium to the right in RCM (increased molecularity), or of the enthalpic driving force shifting the equilibrium to the right in ROMP (release of ring strain). Enyne metathesis is a growing area, since the products bear a conjugated diene functionality that can be the handle for further transformations.^6,14,15

Fig. 2 Typical scopes of Ru-catalyzed olefins metathesis.

The historical development of Ru-catalysts for olefin metathesis is summarized in Fig. 3. The bis-phosphine complex 1, usually defined as a 1st generation (pre)catalyst, is the first example of a Ru-complex very effective in olefin metathesis.¹⁶ The discovery of stable N-heterocyclic carbenes (NHC)¹⁷ opened the route to catalysts in which one phosphine is replaced by a NHC ligand, resulting in complexes, such as 2 in Fig. 3, defined as 2nd generation (pre)catalysts.^18–20 Other improvements in catalytic performances were the substitution of the second phosphine by a chelating alkoxy-alkylidene ligand, such as in 3,^21,22 followed by the introduction of an electron withdrawing NO₂ group para to the chelating isopropyloxy group, as in 4.^23,24 Nevertheless, despite the already excellent performances of the last generation catalysts there is still large space to improve activity, stability and selectivity, which justifies the great efforts in the field by both academic and industrial groups.

Fig. 3 Historical development of Ru olefin metathesis (pre)catalysts.

The beneficial effect of NHC ligands has been correlated to their higher σ-basicity, which has been suggested to stabilize the catalyst and to improve the key activation step. Indeed, NHC as ligands have been demonstrated to induce novel or better performances in a broad series of catalysis promoted by transition metals.^25–27 NHC are rather flexible and versatile ligands whose electronic and steric properties can be tuned to a large extent.^27–29 The most relevant NHC ligands used in olefins metathesis, with the commonly used nomenclature, are reported in Fig. 4.

Fig. 4 Typical NHC ligands used in Ru-catalyzed olefins metathesis.

In the following we will present the main achievements obtained computationally in the field of Ru-catalyzed olefins metathesis. Following the natural life of the catalyst, we will start with a summary of our computational knowledge of the events associated with (pre)catalyst activation, Section 2.1, we will move then inside the catalytic cycle, Sections 2.2 to 2.4, and we will finish with the death of the catalyst through deactivation reactions, Section 2.5. At the end of each section we will briefly summarize in which direction computational action could move to contribute to a more rapid development of these catalysts. Since the last improvements in the field focus on catalysts bearing a NHC ligand, we will focus this perspective on this catalyst family.

2. The fundamental steps of Ru-catalyzed olefins metathesis

2.1. Activation

It is commonly accepted that the catalytically active species in Ru-catalyzed olefins metathesis is the 14e species formed by dissociation of one labile ligand from the (pre)catalyst, step A in Fig. 5. The so formed 14e species then coordinates the C=C bond of the substrate, step B in Fig. 5, an event that is followed by metallacycle formation, step C in Fig. 5, whose opening in the forward direction, step D in Fig. 5, followed by product release, step E in Fig. 5, closes the catalytic cycle. Controlling dissociation of the labile ligand L is fundamental to achieve controlled activation, and (pre)catalysts that are activated more rapidly are usually associated to more labile ligands.

Fig. 5 Schematic catalytic cycle in the presence of NHC Ru-complexes.

According to calculations, dissociation of the labile L ligand, step A in Fig. 5, is frequently rate-limiting. This step could follow both a dissociative or an associative pathway. The former pathway does not require coordination of the olefin before dissociation of the L ligand from the (pre)catalyst. Rather, L dissociation leads to the 14e species that can coordinate the olefin. The associative pathway, instead, starts with olefin coordination to form an octahedral intermediate, followed by dissociation of the L ligand. In the case of L = PR₃, Grubbs and coworkers showed that the mechanism of dissociative displacement is favored for phosphine exchange with most of the 1st and 2nd generation (pre)catalysts, and that phosphines bind more strongly to Ru in the 2nd rather than in the 1st generation systems.³⁰ On the other hand, recent experimental results by Plenio and coworkers indicated that an associative activation mechanism could be operative with chelating alkoxy (pre)catalysts.³¹

Calculations performed by Chen and coworkers indicated that for a 1st generation (pre)catalyst the associative mechanism is disfavored by large energy barriers,³² while in the case of the dissociative mechanism there is no enthalpic barrier to phosphine dissociation, since it proceeds without considerable rearrangement of the complex. Thiel and coworkers extended these calculations to 2nd generation catalysts, including entropic effects. Working on small model systems the dissociative pathway is favored by more than 10 kcal mol⁻¹ in terms of free energy.³³ Moving to systems more representative of those used experimentally, calculations on 2nd generation (pre)catalysts confirmed that dissociation of the phosphines is more feasible than dissociation of the NHC ligand, consistently with the experimental evidence that in these systems the phosphine rather than the NHC ligand is lost during olefin metathesis.^34,35 Dissociation of a typical NHC, such as the SIMes in 2, costs 5 kcal mol⁻¹ more than dissociation of a typical phosphine, such as PCy₃.³² Calculations also showed that the energy required to dissociate the phosphine is reduced in solution, which indicates that care must be taken when quantitatively comparing calculations and experiments. Using the experimental estimate of 36.9 ± 2.3 kcal mol⁻¹ for PCy₃ dissociation from a 2nd generation (pre)catalyst in the gas-phase,³⁶ Chen and coworkers achieved a remarkable agreement with the experimental estimate only using last generation density functionals, such as M06-L, whereas more mature density functionals, such as BP86 or B3LYP, clearly underestimate the experimental value.³⁶ However, the same functionals are less able to reproduce correctly the experimental dissociative PCy₃ exchange activation energy in solution, ΔH^≠ = 27 kcal mol⁻¹, if the simple dissociation of PCy₃ is calculated.³⁷ Considering only electrostatic solvation effects , a too large binding energy is predicted. To solve this discrepancy Goddard and coworkers suggested that a benzene solvent molecule could coordinate to the Ru center. If the global process is calculated as displacement of PCy₃ by a benzene molecule, a reasonable agreement with the experimental value is obtained.³⁷

Focusing on (pre)catalysts with an alkoxybenzylidene chelating ligand, and using the unsubstituted parent (pre)catalyst 3 in Fig. 3 as a reference, experiments indicated that an electron withdrawing nitro group para to the alkoxy position lowers the reaction times and yields full conversion,³⁸ while an electron donor methoxy group in the same position results in longer reaction times.³⁹ Blechert and coworkers suggested that the inclusion of electron withdrawing groups decreases the electron density either of the Ru–alkoxy bond or of the Ru–alkylidene bond, thereby enhancing catalytic activity.³⁹ Solans-Monfort and coauthors rationalized these data by investigating alkoxide dissociation in the series of chelating alkoxy (pre)catalysts as shown in Fig. 6.⁴⁰ They demonstrated that substitution on the aromatic ring of the chelating alkoxybenzylidene group produces significant variations in the Ru⋯O and C1–C2 distances. Focusing on the alkoxy dissociation energy barriers, they found only small differences, within 2 kcal mol⁻¹, between the various systems, but the trend follows the experimentally observed catalytic activities.^23,38,39 Interestingly, calculations showed no direct relationship between the energy barrier and the Ru⋯O interaction strength. Actually, the experimental catalytic efficiencies correlate better with the C1–C2 bond length, that is, its double bond character, rather than with the Ru⋯O distance. The C1–C2 double-bond character was taken to be indicative of π-electron density delocalization, which is almost completely lost during alkoxide dissociation.

Fig. 6 Alkoxybenzylidene (pre)catalysts investigated computationally.

A correlation between aromaticity in the chelating alkoxyarylidene group and catalytic activity was already proposed by Grela and coworkers that, based on Clar's rules of aromaticity,⁴³ explained the lack of activity of the alkoxynaphthylidene (pre)catalysts with the naphthyl bond in the C3–C4 or C5–C6 positions, Fig. 7, to an increased conjugation (aromaticity) on the chelating Ru–C1–C2–C7–O ring.⁴² Still according to Clar's rules, they correlated the preserved catalytic activity of the (pre)catalyst with the naphthyl bond in the C4–C5 position to the reduced conjugation on the chelating ring.⁴³

Fig. 7 Chelating alkoxynaphthylidene NHC–Ru (pre)catalysts.

Despite of intense computational activity, our comprehension of the activation step is still limited, and a clear relationship between (pre)catalysts structure and experimental activity is missing. On the other hand, rationalizing activity remains one of the biggest challenges for computational chemistry, since activity often depends on too many parameters. This is an area where computational chemistry could better contribute through synergic studies with experimental chemists on a series of well characterized and well representative systems.

2.2. Inside the catalytic cycle

In the framework that the active species is the 14e Ru-species formed by dissociation of one labile ligand, see Fig. 5, some early theoretical studies on simplified models confirmed the Chauvin mechanism.^44,45 Starting from the olefin coordination intermediate, the reacting atoms assume an almost planar four-centers geometry at the transition state, see Fig. 8.

Fig. 8 DFT geometry of the transition state for trans-butene metathesis reaction with the Ru–methylidene bond. Distances in Å.

The Ru–alkylidene bond bends toward the olefin, whereas the overall position of the two olefin C atoms is substantially unmodified. The transition state then collapses into the metallacycle intermediate. These conclusions were also supported by the first computational study in which real size catalysts were modeled.³⁵ This study suggests that only a small energy barrier prevents the coordinated olefin to react with the Ru–alkylidene bond, thus precipitating into the metallacycle, which is the most stable structure along the reaction pathway. This study also suggests that steric pressure of the N-substituents on the alkylidene moiety stabilizes the metallacycle, since the formed ruthenacyclobutane develops away from the NHC ligand more than the alkylidene moiety. Incidentally, in a synergic experimental and theoretical study Piers, Woo and coworkers showed that the C_α–C_β bonds of the metallacycle are engaged in an agostic interaction with the Ru center, and that the C_α–C_β bond within the formed 5-membered ring in RCM is remarkably longer than the C_α–C_β bond outside the cycle (Fig. 9). In other words, the metallacycle intermediate is distorted towards a more stable ring opening metathesis product, rather than towards a less stable RCM product.⁴⁶

Fig. 9 Representation of the destiny of the metallacycle intermediate presenting a 5-membered ring. Distance in Å, energies in kcal mol⁻¹.

Other well accepted point coming from calculations is that coordination of a small olefin, such as ethene, presents no or negligible enthalpic barrier, since no severe structural rearrangement of the 14e active species occurs upon olefin coordination. Of course, unfavorable entropic effects could give rise to a free energy barrier to olefin coordination.⁴⁷ An unfavorable enthalpic term could be present in the case of the folding of the substrate prior to a RCM step.

Ethene coordination energy to Ru in a series of complexes was shown to follow the same trend calculated for phosphines, and inclusion of solvent effects decreases both phosphine and olefin coordination energy. Reduction is larger in the case of phosphines, which results in a smaller preference for phosphine coordination and, consequently, higher initiation rates.³⁵ As anticipated, quite low energy barrier has been calculated for metallacycle formation independently of the system considered.^32,35,40,48 Barriers of this height support the idea “that all of the steps after olefin coordination (particularly metallacyclobutane formation) are fast”, proposed by Grubbs to develop his mechanistic scheme.³⁰ Compared to 1st generation catalysts, the NHC-based systems have slower initiation rates because of the higher energy required to dissociate the phosphine. As argued by Grubbs they have a higher activity because olefin coordination is more competitive with respect to re-binding of the phosphine. Nonetheless, the higher activity of the NHC-based systems can be also connected to their substantially lower energy barrier for the metathesis reaction. This helps the active species of NHC-based catalysts to perform more metathesis steps than that of 1st generation catalysts, before being trapped back from free phosphine.

Although the mechanism depicted above is generally well accepted, there are some modifications which are worth mentioning.^50,51 For example, to rationalize low temperature NMR experiments on ethene metathesis exchange,⁴⁹ Webster proposed that a second olefin could coordinate trans to the Ru–alkylidene bond to form an octahedral complex, see Fig. 10.⁵⁰ The second olefins would assist metallacycle opening and dissociation of the formed ethene molecule. It is difficult to extrapolate these results to substrates and temperatures used in productive metathesis.

Fig. 10 Mechanism of associative ethene exchange.

Differently, a long standing question tackled from the experimental and the theoretical side has been whether the exact mechanism in the case of the Ru-based catalysts involves an isomerization from the initial trans-Cl to a cis-Cl geometry leading to the so called “cis” pathway, see Fig. 11.^12,52

Fig. 11 Representation of the cis and trans metathesis pathways.

The cis pathway was hypothesized by Grubbs et al.^12,52 on the basis of the experimental evidence that some Ru-complexes that could be considered as models of Ru-catalysts active in metathesis presented a chelating C=C bond or N group cis to the NHC ligand.^53,54 Further, a cis-Cl geometry was hypothesized to explain stereoselectivity in asymmetric metathesis (vide infra).^52,55 However, a series of computational studies by Chen and Adlhart,³² Cavallo and Correa⁵⁶ and Goddard et al.⁴⁸ showed that the preferred mechanism follows the trans pathway. Calculations suggest that olefin coordination can actually occur both trans and cis to the NHC ligand, and that solvent effects were of paramount relevance to the high stability of complexes with a cis-Cl geometry.⁵⁷ However, calculations also indicated that at the transition state for metallacycle formation the cis-Cl geometry is of much higher energy. This destabilization was explained by the different orientation of the olefin in the cis-Cl geometry on going from the coordination intermediate to the transition state. In the coordination intermediate the olefin C=C double bond is almost perpendicular to the Ru–alkylidene bond to minimize steric interaction with the NHC ligand. At the cis-Cl transition state, instead, the olefin is forced to be parallel to the Ru–alkylidene bond. This increases steric interaction between the olefin and the NHC ligand. Considering that these effects are stronger with bulky olefins, and that have been primarily calculated for NHC based catalysts, this conclusion cannot be generalized to the 1st generation catalysts, although there are experimental indications that the trans-Cl pathway could be favored also in this case.^58,59

While most of the computational studies have used ethene as representative olefin, a few valuable investigations focused on more complex substrates. Solans-Monfort and coauthors reported an exhaustive computational analysis of the RCM of a prototype diene with a series of chelating alkoxy (pre)catalysts, see Fig. 12.⁴⁰ Calculations showed that all the energy barriers within the metathesis cycle, the ring-closing step included, are rather small and definitely smaller than the initial dissociation of the chelating alkoxy group. Thus, they concluded that differences in activity exhibited by (pre)catalysts bearing different alkoxy groups are related to the ability of the (pre)catalyst to generate the 14e active species, rather than to differences in the catalytic cycle.

Fig. 12 Schematic representation of diene RCM.

Chen and coworkers modeled the ROMP of norbornene with 1st and 2nd generation catalysts. Results indicated that metathesis of norbornene is clearly exergonic so that the reaction, at variance from standard CM reactions, is irreversible. Incidentally, their calculations were in remarkable agreement with gas-phase energy values from mass spectroscopy.^32,36 Termination of norbornene ROMP by ethene or 1,4-dichloro-2-butene, see Fig. 13, was modeled by Fomine and coauthors.⁶⁰

Fig. 13 Chain termination by ethene in ROMP of norbornene.

Regarding the various selectivities that could be operative during Ru-catalyzed metathesis, it is surprising the paucity of computational studies. One recent report by Cavallo and coworkers investigated the origin of E/Z selectivity in the case of the model CM of propene with a prototype NHC based catalyst, see Fig. 14.⁶¹ Controlling this selectivity is one of the biggest challenges in Ru-catalyzed metathesis, and it is usually biased towards the formation of the E-isomer. Unexpectedly, calculations indicated that the preferential formation of E olefins, also at low conversions,^62,63 is not determined either at metallacycle formation or opening, and that the key step determining the experimentally observed preferential formation of E-olefins is instead product release.⁶¹

Fig. 14 Schematic representation of the CM of propene.

Another selectivity issue investigated computationally is the competition between the ene and yne functionalities in the case of enynes and dienynes metathesis, with a focus to understand if an ene–then–yne or a yne–then–ene pathway is followed.^15,64,65 In the first study, Lippstreu and Straub investigated the competitive intermolecular metathesis of olefins and alkynes, as well as intramolecular metathesis of a model 1,6-enyne, see Fig. 15.⁶⁶ Their calculations indicated that alkynes coordinate stronger than olefins to the 14e species, but the following metathesis event between the coordinated alkyne and the Ru–alkylidene bond presents a quite high energy barrier. However, alkynes metathesis is irreversible. The same yne–then–ene pathway was suggested to be operative in the case of intramolecular enyne metathesis.

Fig. 15 Schematic representation of ene + yne metathesis.

More complex substrates, such as tosylamine-containing dienynes, and commonly used catalysts were compared by Nolan, Cavallo and coauthors in a synergic experimental and theoretical study.⁶⁷ Using a series of different 1st and 2nd generation (pre)catalysts they showed that the products distribution was (pre)catalyst dependent, with formation of mono and bicyclic products, as well as of 5- and 6-membered rings, see Fig. 16.

Fig. 16 Schematic representation of dienyne metathesis.

The synergic experimental and computational approach permitted us to conclude that for 1st generation PCy₃ containing catalysts the ene–then–yne pathway is adopted, since the bicyclic product 4 is exclusively formed, and no traces of the monocyclized products 1 or 3 are observed. This suggested that 1st generation catalysts are less reactive towards the alkyne moiety, and calculations confirmed their minor ability, relative to the NHC based catalysts, to capture both the ene or yne moieties of the substrate. On the other hand, for catalysts bearing NHC ligands, the combined approach indicated that both the ene–then–yne and the yne–then–ene pathways are possible. In line with the results of Lippstreu and Straub the yne functionality coordinates better than the ene functionality, but presents a higher energy barrier for the metathesis reaction, which results in a competition between the two pathways. Steric hindrance of the NHC ligand as well as around the C=C and C≡C bonds appeared as a key factor governing the mechanistic pathway adopted (hindrance around the C=C bonds push towards the yne–then–ene pathway) and consequently affects the nature of the formed products.

As this overview has indicated, comprehension of the basic events occurring in the active cycle is quite extensive. Nevertheless, there are several points which have been scarcely considered. For example, very little is known of fundamental points such as the preferential regiochemistry of the metathesis event, as well as of the factors controlling the formation of Z-olefins, and what really prevents formation of highly substituted olefins. Similarly, almost nothing has been done to rationalize chemoselectivity. These are key points where computational chemistry could excel and could help experimentalists to design better performing catalysts. The scarcity of computational studies is instead surprising.

2.3. Dynamic behavior of Ru–NHC catalysts

The concept of NHC based Ru-catalysts as static entities whose behavior can be largely derived from inspection of crystallographic structures is being replaced by the more complex picture of highly dynamic structures. This dynamic behavior, if controlled and rationalized, can open the route to new possibilities. Along this line, recently there has been an increasing interest on understanding better the isomerization between cis-Cl and trans-Cl geometries of Ru-precatalysts, see Fig. 17. Cis-Cl (pre)catalysts are of practical interest because they are seen as (pre)catalysts whose activation can be controlled via their isomerization to the trans-Cl isomer.^53,68–70 Controlled activation is particularly relevant in ROMP, since latent catalysts allow for a proper mixing of the monomer and the catalyst before polymerization starts.

Fig. 17 Example of cis/trans equilibrium in selected Ru-complexes.

Goddard and Benitez rationalized the high stability of cis-Cl complexes.⁵⁷ Their calculations indicated that the trans-Cl geometry of a Ru complex bearing a chelating alkylidene-pyridine ligand is quite more stable than the cis-Cl geometry in the gas-phase, but inclusion of solvent effects results in the cis-Cl geometry slightly favored. This solvent effect originates from the strong difference in the dipole moment of the two geometries, which translates into very different solvation energies.⁵⁷

This first computational effort was followed by Lemcoff and coworkers that focused on the σ donors and π acid capability of the chelating atom in a series of structurally related Ru-complexes, see Fig. 18.⁷¹ Stronger σ-donors were shown to destabilize the trans-Cl geometry due to a destabilizing trans interaction with the strong σ-donor NHC ligand. Also π acid coordinating atoms, such as phosphines, destabilize the trans-Cl geometry, or better stabilize the cis-Cl geometry, since they prefer to be coordinated trans to a π donor, such as Cl. Conversely, weak σ and π donors, such as ethers, stabilize the trans-Cl geometry, since they prefer to coordinate trans to a strong σ donor and π acceptor, such as the NHC ligand.⁷¹

Fig. 18 Ru-complexes investigated in the context of the cis-Cl/trans-Cl equilibrium.

Regarding the mechanism of isomerization between cis-Cl and trans-Cl geometries, preliminary calculations by Goddard and Benitez indicated the viability of the dissociative pathway shown in Fig. 19 for the chelating pyridine complex of Fig. 17.⁵⁷

Fig. 19 Schematic representation of the proposed pathways for the isomerization between the cis-Cl and trans-Cl geometries.

Extensive analysis of the two pathways was performed by Grela, Cavallo and coworkers in a synergic experimental and theoretical study of the isomerization in the two Ru-complexes shown in Fig. 17. Calculations suggested a sizeable preference for the concerted pathway in the case of the quinoline complex, whereas the associative pathway is only slightly preferred in the case of the pyridine complex. The negligible influence of cyclohexene or free quinoline on the ¹H NMR isomerization kinetics of the complex containing a chelated quinoline moiety supported the preference for the concerted pathway in this specific case.⁷²

Still on the flexible behavior of NHC based Ru-catalysts, NHC ligands have been shown rather flexible architectures whose stereoelectronics can be modified to a large extent, giving the chance of an accurate tuning of catalytic activity. In the case of olefins metathesis, even catalysts based on structurally similar NHC ligands show rather different stabilities and activities. For example, catalysts bearing the SIMes NHC ligand exhibit low efficiency in the RCM or CM of sterically encumbered olefins. These limitations can be circumvented, while preserving catalysts robustness, by reducing the steric bulk of the NHC ligand.¹¹ Along this line, Grubbs and coworkers developed catalysts bearing NHC ligands with o-tolyl N-substituent, such as that in Fig. 20, which proved rather successful in this regard.¹²

Fig. 20 ortho-Tolyl based (pre)catalyst.

Rationalization of this result was achieved in a synergic experimental and theoretical study by Grubbs, Goddard and co-workers. Computations clearly indicated that the o-tolyl based NHCs can respond actively to the steric requirements of an incoming ligand by adopting conformations that minimize their steric bulkiness in the first coordination sphere of the metal.⁷³ To minimize steric congestion the o-tolyl rings are rotated away from the bound olefin. In short, the substituted side of the N-tolyl rings appears “smaller” than the unsubstituted side, due to rotation of these rings.⁷³ This conclusion is in agreement with previous computational analysis on asymmetric metathesis catalysts by Cavallo and Costabile.⁷⁴

To achieve a more comprehensive understanding of this point, Cavallo and coworkers performed an ab initio molecular dynamics study of a series of 17 NHC based Ru (pre)catalysts.⁷⁵ The main results are summarized in Fig. 21, where the distribution of the ϕ₁ angle around the N-substituent bonds on the Ru–alkylidene side, as obtained by the statistical analysis of the molecular dynamics trajectories, is reported. System 6 is used as a reference, and its ϕ₁ distribution shows a peak centered at 90°. Conversely, the ϕ₁ distribution of the SIPr based system 8 shows a flatter and somewhat broader and jagged peak, while the N-phenyl system 5 presents a rather broad ϕ₁ distribution with long tails. The o-tolyl system 7 also presents a quite peaked ϕ₁ distribution, which means that a single Me group in the ortho position of the N-substituent is enough to reduce remarkably flexibility. The distribution of 7 presents a larger shoulder at low ϕ₁ values, indicating that the unsubstituted side of the o-tolyl ring is folded towards the Cl–Ru–Cl plane. This result relates with the good catalytic performances of 7, since it is less hindered than 6, but the limited flexibility relative to 5 prevents it from suffering C–H deactivation reactions.⁷⁶

Fig. 21 Distribution of the ϕ₁ angle in Ru (pre)catalysts. [Ru] stands for the Ru(Cl₂)(PMe₃) moiety. ϕ₁ = ±90° corresponds to geometries with the ring of the N-substituent perpendicular to the plane of the NHC ring.

On quantitative grounds, the steric pressure introduced by different NHC ligands can be measured by the percent of buried volume, %V_Bur, defined as the percent of the total volume of the first coordination sphere around the metal occupied by a given ligand, developed by Nolan and Cavallo et al.^29,77,78 Analysis of a series of NHC ligands, see Fig. 22, indicated that saturated NHCs consistently show a %V_Bur slightly greater than that of the unsaturated analogues. NHC presenting ortho isopropyl groups on the N-aryl ring is quite bulkier than the NHC ligand presenting N-mesityl substituents, and is comparable in bulkiness with bulky phosphines, such as PCy₃.⁷⁸ In this context, Cavallo and coworkers analyzed the optimized structure of a series of (pre)catalysts using topographic steric maps to highlight the real impact of the NHC ligands in the first coordination sphere of the metal, which is the place where catalysis occurs. Topographic steric maps can be considered a chemical analogy to geographical physical maps, which display the natural features of the earth, like the location of mountains and valleys.^75,79 Comparison of the steric maps of (pre)catalysts 6 and 7, see Fig. 23, indicates that they shape quite different reactive pockets. Rather flat with a constant pressure on the halide–Ru–halide plane in the case of 6, while in the case of 7 the steric map clearly highlights the higher impact of the unsubstituted side of the ligand in the first coordination sphere of the metal and evidences an overall C_S-symmetric reactive pocket.

Fig. 22 Schematic representation of the buried volume, and %V_Bur value for representative ligands.

Fig. 23 Topographic steric map of the NHC ligand of (pre)catalysts 6 and 7. The colouring scale of the isocontour levels, in Å, is also reported.

However, new experimental results are already putting new challenges. Phosphites have been shown to replace effectively phosphines to generate very active catalysts with a peculiar tendency to give cis-Cl geometries,⁸⁰ while the flexibility and the isomerization between different geometries have been also related to the possible dissociation of one of the halide ligands.⁸¹ Considering the insight that could derive from a better understanding of these phenomena, future computational work in this area is highly desirable.

2.4. Asymmetric metathesis

A promising application of Ru-catalyzed olefins metathesis is asymmetric metathesis.^{11,52,55,82–84} Several catalysts have been developed to achieve this fundamental transformation that can increase the value of starting achiral reactants by transforming them into chiral products, see Fig. 24 for an example.

Fig. 24 Ru-catalyzed desymmetrization of achiral trienes.

A prototype catalyst able to effectively desymmetrize achiral trienes is that shown in Fig. 24, which presents two phenyl substituents on the C4 and C5 atoms of the NHC skeleton.^52,55 Mechanistically, this catalyst offers a remarkable challenge since the stereogenic centers on the C4 and C5 positions of the NHC bridge are spatially well away from the coordination position that will be occupied by the incoming substrate.⁸⁵ Computations were fundamental to illuminate the origin of stereoselectivity in the ring closing step of a model substrate. The competing transition states for ring closure are shown in Fig. 25. The chiral NHC ligand assumes a chiral folding around the ϕ₁ and ϕ₂ torsional angles, imposed by the Ph substituents on the C4 and C5 atoms of the NHC ring, so that the unsubstituted side of the N-aryl ring is pushed down, toward the Cl–Ru–Cl plane. As a consequence of this folding, the unsubstituted side of the N-aryl ring appears more bulky to the substrate than the ortho-isopropyl substituted side, and it is the steric pressure of the unsubstituted side of the N-substituent that shapes a chiral reactive pocket resulting in an energy difference between the two competing transition states, thus generating stereoselectivity.⁷⁴ Synergic NMR experiments and DFT calculations performed by Grubbs, Goddard and co-workers confirmed that in a strictly related Ru-complex, bearing a NHC ligand with o-tolyl N-substituents, the unsubstituted side of the o-tolyl ring is folded down towards the Ru–alkylidene moiety, and thus appears bulkier to the substrate in solution.⁷³

Fig. 25 Competing transition states for stereoselective RCM.

The best evidence of the asymmetry of the chiral pocket shaped by the chiral NHC ligand in Fig. 24 is through analysis of the steric map around the metal center, see Fig. 26.^75,79 The steric map is asymmetric both in the tail (left) and head (right) quadrants. In both cases the steric map confirms that the less encumbered quadrants are those occupied by the i-Pr substituted side of the N-substituents.⁷⁵

Fig. 26 Steric map of the chiral NHC ligand is shown. The quadrants occupied by the ortho i-Pr groups are indicated by a black dot. The colouring scale of the isocontour levels, in Å, is also reported.

Asymmetric metathesis is probably the field where large improvements are still possible, and computational chemistry could represent a very powerful tool to help experimentalists to design new and better performing catalysts. Despite effective stereoselectivity requires an energy difference as little as 2–3 kcal mol⁻¹ between the competing transition states, this often is one of the most reliable values that computational chemistry can provide. In fact, all the errors in the computational recipe tend to cancel out because two extremely similar transition states are compared. Once the origin of stereoselectivity has been well clarified and the computational recipe has been tested using an experimentally characterized catalyst as benchmark, computational chemistry could be extremely useful to screen a large number of ligands, and could indicate to the experimental chemists which are the most promising catalysts. Of course, designing the new ligands in collaboration with the experimentalists would be more productive, since screening would be focused on ligands whose synthesis is feasible.

2.5. Deactivation

To rationally design new and more effective catalysts it is crucial to understand the decomposition pathways that reduce the performance of the existing catalysts. Nevertheless, despite the importance of the issue, only a few computational studies have been performed on this topic.^86–91 One of the reasons for the paucity of computational studies is in the often obscure nature of the decomposition products, which makes almost impossible for computational chemists to know which reaction pathway to model. In these cases, even morcels of experimental information are fundamental to guide the computations. Along this direction, van Rensburg and coauthors performed a synergic experimental/theoretical study of the decomposition of 1st and 2nd generation catalysts in ethene exchange metathesis induced by the substrate.⁸⁶ The key step of this decomposition pathway is the β-H transfer from the metallacycle intermediate to the Ru center to form a hydride Ru(allyl) species, from which higher olefins are formed, see Fig. 27. Calculations indicated that 1st generation catalysts are more prone than 2nd generation catalysts to this deactivation route.

Fig. 27 Allylic deactivation pathway. L = PR₃ or NHC.

Suresh and coworkers,⁸⁸ and immediately after Cavallo and coworkers,^89,90 reported on deactivation pathways involving the activation of an ortho C–H bond of the N-aryl ring of the NHC ligand of 2nd generation catalysts. As correctly suggested by Grubbs et al.,⁷⁶ this deactivation pathway involves a key Ru-hydride intermediate, as well as all the single steps composing the whole deactivation pathway. However, calculations were fundamental to place the single steps in the proper sequential order. As indicated in Fig. 28, the initial C–H activation transfers the H atom to the benzylidene moiety to form a Ru-benzyl moiety, followed by the formation of the Ru-hydride species with regeneration of the benzylidene group. At this point the benzylidene groups are transferred to the activated N-aryl ring, see again Fig. 28. This leads to the first product experimentally observed. Activation of an ortho C–H bond of the other N-aryl ring, promoted by free PCy₃, explains the formation of the second product experimentally observed, see again Fig. 28. Both the experimental products are stabilized by η⁶-coordination of the aromatic ring of the former benzylidene group to the Ru center.

Fig. 28 Decomposition pathway involving C–H activation of the NHC ligand.

Another decomposition reaction of the Ru (pre)catalyst was investigated by Diver and coworkers, see Fig. 29.^92,93 (Pre)catalyst decomposition is promoted by coordination of a CO molecule trans to the Ru–alkylidene bond, and proceeds through the insertion of the alkylidene moiety into the nearby N-aryl ring via a Buchner type ring expansion.

Fig. 29 Decomposition of Ru (pre)catalysts induced by CO.

Calculations of Cavallo and coworkers suggested a continuous energy decay along the pathway, with a cascade of reactions with meaningless barriers, including the Buchner ring expansion step. This suggests that from a kinetic point of view the systems would roll down to the products with no kinetically stable intermediate. Ab initio molecular dynamics calculations confirmed this hypothesis.⁹¹ The activation role of the CO was related to its ability to weaken the π-back-bonding between the Ru atom and the ylidene ligand, making the latter more electrophilic and disengaging it from the metal center, see Fig. 30.

Fig. 30 Relevant molecular orbitals involved in the CO–Ru–alkylidene interaction.

As clear from this summary, understanding catalysts deactivation is a real challenge both for experimental and computational chemists, since deactivation pathways can proceed along unexpected and not completely clear reaction pathways. The contribution of computational chemistry can be thus extremely useful since it could allow us to characterize properly the deactivation mechanisms by connecting the starting species to the deactivation products. After the weak point of the catalyst has been understood, computational chemistry could be used to explore those structural modifications that prevent catalyst deactivation while preserving good catalytic activity.

Acknowledgements

The research leading to these results received funding from the European Community's Seventh Framework Programme (FP7/2007-2013) under grant agreement no. CP-FP 211468-2 EUMET. AP thanks the Spanish MICINN for a Ramón y Cajal contract.

Notes and references

1. J.-L. Herisson and Y. Chauvin, Makromol. Chem., 1971, 141, 161.
2. A. Fürstner, Angew. Chem., Int. Ed., 2000, 39, 3012.
3. R. R. Schrock and A. H. Hoveyda, Angew. Chem., Int. Ed., 2003, 42, 4592.
4. A. H. Hoveyda and R. R. Schrock, Chem.–Eur. J., 2001, 7, 945.
5. T. M. Trnka and R. H. Grubbs, Acc. Chem. Res., 2000, 34, 18.
6. R. H. Grubbs, Handbook of Olefin Metathesis, Wiley-VCH, Weinheim, Germany, 2003.
7. G. C. Vougioukalakis and R. H. Grubbs, Chem. Rev., 2009, 110, 1746.
8. C. Samojłowicz, M. Bieniek and K. Grela, Chem. Rev., 2009, 109, 3708.
9. T. Nicola, M. Brenner, K. Donsbach and P. Kreye, Org. Process Res. Dev., 2005, 9, 513.
10. Y. S. Tsantrizos, J.-M. Ferland, A. McClory, M. Poirier, V. Farina, N. K. Yee, X.-j. Wang, N. Haddad, X. Wei, J. Xu and L. Zhang, J. Organomet. Chem., 2006, 691, 5163.
11. J. Berlin, S. D. Goldberg and R. H. Grubbs, Angew. Chem., Int. Ed., 2006, 45, 7591.
12. I. C. Stewart, C. J. Douglas and R. H. Grubbs, Org. Lett., 2008, 10, 441.
13. A. B. Smith, III, C. M. Adams and S. A. Kozmin, J. Am. Chem. Soc., 2001, 123, 990.
14. M. Mori, Adv. Synth. Catal., 2007, 349, 121.
15. S. T. Diver, Coord. Chem. Rev., 2007, 251, 671.
16. S. T. Nguyen, R. H. Grubbs and J. W. Ziller, J. Am. Chem. Soc., 1993, 115, 9858.
17. A. J. Arduengo, R. L. Harlow and M. Kline, J. Am. Chem. Soc., 1991, 113, 361.
18. M. Scholl, S. Ding, C. W. Lee and R. H. Grubbs, Org. Lett., 1999, 1, 953.
19. T. Weskamp, F. J. Kohl, W. Hieringer, D. Gleich and W. A. Herrmann, Angew. Chem., Int. Ed., 1999, 38, 2416.
20. J. Huang, E. D. Stevens, S. P. Nolan and J. L. Petersen, J. Am. Chem. Soc., 1999, 121, 2674.
21. J. S. Kingsbury, J. P. A. Harrity, P. J. Bonitatebus and A. H. Hoveyda, J. Am. Chem. Soc., 1999, 121, 791.
22. S. Gessler, S. Randl and S. Blechert, Tetrahedron Lett., 2000, 41, 9973.
23. A. Michrowska, R. Bujok, S. Harutyunyan, V. Sashuk, G. Dolgonos and K. Grela, J. Am. Chem. Soc., 2004, 126, 9318.
24. M. Bieniek, R. Bujok, M. Cabaj, N. Lugan, G. Lavigne, D. Arlt and K. Grela, J. Am. Chem. Soc., 2006, 128, 13652.
25. N. Marion and S. P. Nolan, Acc. Chem. Res., 2008, 41, 1440.
26. S. Würtz and F. Glorius, Acc. Chem. Res., 2008, 41, 1523.
27. S. Diez-Gonzalez, N. Marion and S. P. Nolan, Chem. Rev., 2009, 109, 3612.
28. W. A. Herrmann, Angew. Chem., Int. Ed., 2002, 41, 1291.
29. L. Cavallo, A. Correa, C. Costabile and H. Jacobsen, J. Organomet. Chem., 2005, 690, 5407.
30. M. S. Sanford, J. A. Love and R. H. Grubbs, J. Am. Chem. Soc., 2001, 123, 6543.
31. T. Vorfalt, K. J. Wannowius and H. Plenio, Angew. Chem., Int. Ed., 2010, 49, 5533.
32. C. Adlhart and P. Chen, J. Am. Chem. Soc., 2004, 126, 3496.
33. S. F. Vyboishchikov, M. Bühl and W. Thiel, Chem.–Eur. J., 2002, 8, 3962.
34. T. Weskamp, W. C. Schattenmann, M. Spiegler and W. A. Herrmann, Angew. Chem., Int. Ed., 1998, 37, 2490.
35. L. Cavallo, J. Am. Chem. Soc., 2002, 124, 8965.
36. S. Torker, D. Merki and P. Chen, J. Am. Chem. Soc., 2008, 130, 4808.
37. D. Benitez, E. Tkatchouk and W. A. Goddard, Organometallics, 2009, 28, 2643.
38. K. Grela, S. Harutyunyan and A. Michrowska, Angew. Chem., Int. Ed., 2002, 41, 4038.
39. M. Zaja, S. J. Connon, A. M. Dunne, M. Rivard, N. Buschmann, J. Jiricek and S. Blechert, Tetrahedron, 2003, 59, 6545.
40. X. Solans-Monfort, R. Pleixats and M. Sodupe, Chem.–Eur. J., 2010, 16, 7331.
41. G. Portella, J. Poater and M. Solà, J. Phys. Org. Chem., 2005, 18, 785.
42. H. Masui, Coord. Chem. Rev., 2001, 219–221, 957.
43. M. Barbasiewicz, A. Szadkowska, A. Makal, K. Jarzembska, K. Wozacuteniak and K. Grela, Chem.–Eur. J., 2008, 14, 9330.
44. O. M. Aagaard, R. J. Meier and F. Buda, J. Am. Chem. Soc., 1998, 120, 7174.
45. C. Adlhart, C. Hinderling, H. Baumann and P. Chen, J. Am. Chem. Soc., 2000, 122, 8204.
46. C. N. Rowley, E. F. van der Eide, W. E. Piers and T. K. Woo, Organometallics, 2008, 27, 6043.
47. T. K. Woo, P. E. Blöchl and T. Ziegler, J. Phys. Chem. A, 2000, 104, 121.
48. D. Benitez, E. Tkatchouk and W. A. I. Goddard, Chem. Commun., 2008, 6194.
49. P. E. Romero and W. E. Piers, J. Am. Chem. Soc., 2007, 129, 1698.
50. C. E. Webster, J. Am. Chem. Soc., 2007, 129, 7490.
51. B. F. Straub, Adv. Synth. Catal., 2007, 349, 204.
52. T. J. Seiders, D. W. Ward and R. H. Grubbs, Org. Lett., 2001, 3, 3225.
53. T. Ung, A. Hejl, R. H. Grubbs and Y. Schrodi, Organometallics, 2004, 23, 5399.
54. D. R. Anderson, D. D. Hickstein, D. J. O'Leary and R. H. Grubbs, J. Am. Chem. Soc., 2006, 128, 8386.
55. T. W. Funk, J. M. Berlin and R. H. Grubbs, J. Am. Chem. Soc., 2006, 128, 1840.
56. A. Correa and L. Cavallo, J. Am. Chem. Soc., 2006, 128, 13352.
57. D. Benitez and W. A. Goddard, III, J. Am. Chem. Soc., 2005, 127, 12218.
58. J. A. Tallarico, P. J. Bonitatebus Jr. and M. L. Snapper, J. Am. Chem. Soc., 1997, 119, 7157.
59. E. F. van der Eide, P. E. Romero and W. E. Piers, J. Am. Chem. Soc., 2008, 130, 4485.
60. S. Fomine, J. V. Ortega and M. A. Tlenkopatchev, J. Mol. Catal. A: Chem., 2005, 236, 156.
61. N. Bahri-Laleh, R. Credendino and L. Cavallo, Beilstein J. Org. Chem., 2011, 7, 40.
62. C. Copéret, Beilstein J. Org. Chem., 2011, 7, 13.
63. D. R. Anderson, T. Ung, G. Mkrtumyan, G. Bertrand, R. H. Grubbs and Y. Schrodi, Organometallics, 2008, 27, 563.
64. N. Dieltiens, K. Moonen and C. V. Stevens, Chem.–Eur. J., 2007, 13, 203.
65. G. C. Lloyd-Jones, R. G. Margue and J. G. de Vries, Angew. Chem., 2005, 117, 7608.
66. J. J. Lippstreu and B. F. Straub, J. Am. Chem. Soc., 2005, 127, 7444.
67. H. Clavier, A. Correa, E. C. Escudero-Adán, J. Benet-Buchholz, L. Cavallo and S. P. Nolan, Chem.–Eur. J., 2009, 15, 10244.
68. S. Pruhs, C. W. Lehmann and A. Furstner, Organometallics, 2003, 23, 280.
69. M. Barbasiewicz, A. Szadkowska, R. Bujok and K. Grela, Organometallics, 2006, 25, 3599.
70. C. Slugovc, B. Perner, F. Stelzer and K. Mereiter, Organometallics, 2004, 23, 3622.
71. C. E. Diesendruck, E. Tzur, A. Ben-Asuly, I. Goldberg, B. F. Straub and N. G. Lemcoff, Inorg. Chem., 2009, 48, 10819.
72. A. Poater, F. Ragone, A. Correa, A. Szadkowska, M. Barbasiewicz, K. Grela and L. Cavallo, Chem.–Eur. J., 2010, 16, 14354.
73. I. C. Stewart, D. Benitez, D. J. O'Leary, E. Tkatchouk, M. W. Day, W. A. Goddard and R. H. Grubbs, J. Am. Chem. Soc., 2009, 131, 1931.
74. C. Costabile and L. Cavallo, J. Am. Chem. Soc., 2004, 126, 9592.
75. F. Ragone, A. Poater and L. Cavallo, J. Am. Chem. Soc., 2010, 132, 4249.
76. S. H. Hong, A. Chlenov, M. W. Day and R. H. Grubbs, Angew. Chem., Int. Ed., 2007, 46, 5148.
77. A. C. Hillier, W. J. Sommer, B. S. Yong, J. L. Petersen, L. Cavallo and S. P. Nolan, Organometallics, 2003, 22, 4322.
78. A. Poater, B. Cosenza, A. Correa, S. Giudice, F. Ragone, V. Scarano and L. Cavallo, Eur. J. Inorg. Chem., 2009, 1759.
79. A. Poater, F. Ragone, R. Mariz, R. Dorta and L. Cavallo, Chem.–Eur. J., 2010, 16, 14348.
80. X. Bantreil, T. E. Schmid, R. A. M. Randall, A. M. Z. Slawin and C. S. J. Cazin, Chem. Commun., 2010, 46, 7115.
81. M. Zirngast, E. Pump, A. Leitgeb, J. H. Albering and C. Slugovc, Chem. Commun., 2011, 47, 2261.
82. R. E. Giudici and A. H. Hoveyda, J. Am. Chem. Soc., 2007, 129, 3824.
83. P.-A. Fournier, J. Savoie, B. Stenne, M. Bédard, A. Grandbois and S. K. Collins, Chem.–Eur. J., 2008, 14, 8690.
84. F. Grisi, A. Mariconda, C. Costabile, V. Bertolasi and P. Longo, Organometallics, 2009, 28, 4988.
85. K. B. Lipkowitz, C. A. D'Hue, T. Sakamoto and J. N. Stack, J. Am. Chem. Soc., 2002, 124, 14255.
86. W. J. van Rensburg, P. J. Steynberg, W. H. Meyer, M. M. Kirk and G. S. Forman, J. Am. Chem. Soc., 2004, 126, 14332.
87. W. J. van Rensburg, P. J. Steynberg, M. M. Kirk, W. H. Meyer and G. S. Forman, J. Organomet. Chem., 2006, 691, 5312.
88. J. Mathew, N. Koga and C. H. Suresh, Organometallics, 2008, 27, 4666.
89. Deactivation, ed. A. Poater and L. Cavallo, Kluwer, 2009.
90. A. Poater and L. Cavallo, J. Mol. Catal. A: Chem., 2010, 324, 75.
91. A. Poater, F. Ragone, A. Correa and L. Cavallo, J. Am. Chem. Soc., 2009, 131, 9000.
92. B. R. Galan, M. Gembicky, P. M. Dominiak, J. B. Keister and S. T. Diver, J. Am. Chem. Soc., 2005, 127, 15702.
93. B. R. Galan, M. Pitak, M. Gembicky, J. B. Keister and S. T. Diver, J. Am. Chem. Soc., 2009, 131, 6822.

---