The divergent effects of strong NHC donation in catalysis

The inverse relationship between NHC donicity and catalyst initiation.


Introduction
The remarkable impact of N-heterocyclic carbene (NHC) ligands on transition-metal catalysis [1][2][3][4] is due largely to their strong sdonor character, a feature highlighted in even the earliest reviews. [5][6][7] Strong NHC binding is believed to inhibit decomposition of molecular catalysts, 1,8 and to stabilize the higher oxidation states essential in multiple catalytic contexts, including olen metathesis and cross-coupling reactions. [1][2][3] As well, however, emerging work points toward the potential for NHC donation to inuence bonding interactions with other ligands present, both ancillary ligands and bound substrate. [9][10][11] In a leading recent example, the Neidig group reported evidence for ground-state weakening of the Fe-Cl bond by sdonation from the NHC ligand in tetrahedral FeX 2 (NHC) 2 complexes. 9 The implied potential labilization of p-donor ligands by NHC ligands is of keen interest. The potentially broad implications of such behaviour in catalysis prompted us to explore the impact of NHC donicity on neutral, dative donor ligands, particularly in geometries that reinforce inter-ligand electronic communication. Here we demonstrate the impact of the NHC ligand on trans-ligand binding, in an important example drawn from olen metathesis.
The breakthrough activity of the second-generation Grubbs catalysts, 12,13 which greatly expanded the scope of the reaction relative to the parent system GI (Fig. 1), was originally attributed to labilization of the s-donor PCy 3 ligand by the strongly donating trans-NHC ligand. 14 In a seminal kinetics study, however, Grubbs and co-workers demonstrated that PCy 3 loss is in fact slower for GII than the rst-generation catalyst GI. 14 A leading explanation for this "inverse trans effect" highlights alkylidene rotation as a trigger for PCy 3 dissociation, pointing out higher torsional barriers to such rotation in the NHC complexes. 15 An alternative view emerges from Kennepohl's discovery, based on groundbreaking X-ray absorbance spectroscopy (XAS) studies, that the Ru center in s-GII is more electropositive than that in GI. 16 This implies that the NHC ligand is a poor net charge donor, relative to PCy 3 . An increased electrostatic attraction between the more electron-decient Ru center in GII and the strongly-donating PCy 3 ligand was proposed to account for the reduced phosphine lability.
Adopting the majority view of NHC ligands as strong sdonors, we speculated that NHC donation might itself be a factor: that strong s-donation could in fact strengthen the trans Ru-PCy 3 bond, by increasing Ru / PCy 3 backbonding. In exploring this possibility, we focused on the methylidene species GIIm (Fig. 2), to eliminate steric or p-stacking effects associated with the benzylidene moiety, and electronic perturbation arising from benzylidene p-acidity. GIIm is, moreover, a key player in catalysis, as the resting-state species in most ring-closing and cross-metathesis reactions promoted by GII. That is, because GIIm is thermodynamically stable relative to both the benzylidene precatalyst GII, and other ruthenium species present in the catalytic cycle, its concentration builds up during metathesis. Recently-developed 17 routes to the secondgeneration methylidene complexes enable their direct study.
Here we quantify the differences in PCy 3 lability in GIIm; we demonstrate that strong s-donation from the H 2 IMes ligand is indeed tempered by p-backbonding onto the NHC, as evidenced by restricted rotation about the Ru-H 2 IMes bond, and that PCy 3 loss is dramatically slower for the IMes system, in which NHC s-donation is unrelieved by NHC p-acidity (as conrmed by room-temperature rotation about the Ru-IMes bond). Based on these observations, we propose that enhanced backbonding onto the PCy 3 ligand is a key, overlooked contributor to the low phosphine lability characteristic of the second-generation Grubbs catalysts. Such Ru / PCy 3 backbonding relieves the heightened electron density at Ru that would otherwise result from strong NHC s-donation, and consequently strengthens the Ru-P bond. The broader implications for catalysis are discussed.

Results and discussion
Assaying PCy 3 lability for GIIm Direct assessment of PCy 3 lability for the second-generation methylidene complexes is hampered by a combination of strong phosphine binding and thermal instability. Even for the more labile benzylidene pre-catalysts, PCy 3 loss from the IMes derivative u-GII was 640 times slower than from the rst-generation complex GI. 14 Qualitative evidence indicated drastically lower lability for the methylidene complexes GIIm, but attempts to measure rate constants were thwarted by decomposition at the temperatures required to induce PCy 3 exchange (ca. 85 C). 14 This underscores the point that the thermodynamic stability of GIIm relative to other catalytically relevant species does not equate to stability against decomposition. Indeed, the methylidene complexes are signicantly more vulnerable than their benzylidene precursors, owing to their susceptibility to nucleophilic attack at the Ru]CH 2 site. [30][31][32] We considered that this vulnerability, which constituted a problem in the original exchange experiments, could offer a disguised opportunity to assess phosphine lability. Specically, if decomposition of GIIm proceeds via rate-limiting loss of PCy 3 , 30 then the rate of decomposition reports on the rate of PCy 3 loss. To conrm that this reaction proceeds only via fourcoordinate Ru-1, we examined the impact of added PCy 3 on the reaction kinetics. If phosphine attack occurs on Ru-1 (Scheme 1a), the rate of decomposition should be unaffected, for the reasons discussed below. If, however, GIIm can react directly with PCy 3 (Scheme 1b), decomposition should be accelerated.
As seen from Fig. 3, the rate of decomposition is unaffected by added PCy 3 , indicating reaction via the dissociative pathway (Scheme 1a). The preference is unsurprising, given steric restrictions on the approach of PCy 3 to the methylidene carbon in ve-coordinate GIIm. The absence of an inverse dependence on [PCy 3 ] may at rst seem inconsistent with rate-determining loss of PCy 3 . This reects the participation of PCy 3 in the k 2 step (i.e. the Ru-1 / Ru-2 transformation), as well as the k À1 step  (the Ru-1 / GIIm back-reaction). If nucleophilic attack on Ru-1 is much faster than phosphine re-binding (i.e. k 2 [ k À1 ), the rate expression reduces to k 1 [GIIm] (see ESI †).
(For completeness, it may be noted that even if k 2 and k À1 were of comparable magnitudeor indeed if k 2 ( k À1no phosphine inhibition would result. Because the rate of the k À1 step is k À1 [Ru-1][PCy 3 ], and that of the k 2 step is k 2 [Ru-1][PCy 3 ], any change in [PCy 3 ] alters both rates equivalently, and the phosphine concentration term cancels out. Thus the rate of reaction is independent of [PCy 3 ], irrespective of the relative magnitudes of k 2 and k À1 ).
To assess the rates of PCy 3 loss from s-GIIm and u-GIIm, in the present case, where k 2 [ k À1 , we measured the rates of decomposition of these complexes in C 6 D 6 . Decreases in the proportion of GIIm over time were established by 1 H NMR analysis. The integrated intensity of the methylidene singlet was measured relative to 1,3,5-trimethoxybenzene (TMB; d CH 6.26 ppm) as internal standard. Decomposition was nearly eight times faster for s-GIIm than u-GIIm, as shown by the rate curves in Fig. 4. The relative rates show little change from 40-80 C: in each case, loss of PCy 3 from the IMes derivative was 7-8 times slower. DFT studies by the Jensen group reported an identical trend for the parent benzylidene catalysts, with k 1 for u-GII being seven-fold lower than for s-GII. 34 The lower phosphine lability of u-GIIm relative to s-GIIm was maintained in other solvents (Fig. 5). In these experiments, the proportion of GIIm remaining aer 6 h at 60 C was measured. Decomposition was marginally faster in chlorinated media than in aromatic solvents, and dramatically faster in the coordinating solvent THF. The solvent-dependence of PCy 3 dissociation thus follows the trend C 7 H 8 $ C 6 H 6 < CH 2 Cl 2 $ CHCl 3 ( THF, for both the IMes and H 2 IMes methylidene complexes. This agrees with the trend previously established for initiation of the benzylidene precatalyst s-GII, for which the rate-determining step is likewise PCy 3 loss. 14 The consistency in these reactivity patterns, as well as the excellent agreement with the relative rate constants computed by Jensen (see above), validate the use of decomposition rates to quantify rates of PCy 3 loss from GIIm. Also noteworthy is the close correlation between relative rates of initiation of GII in different solvents, and relative rates of decomposition of GIIm. This correlation accounts for the observation that increasing the rate of initiation does not improve reaction rates for the Grubbs catalysts. 35 Instead, because productive metathesis generates an unprotected methylidene moiety, faster initiation is offset by faster methylidene abstraction by free PCy 3 .

Crystallographic analysis: comparison of u-GIIm with s-GIIm
In the hope of gaining insight into the bonding interactions that distinguish the IMes and H 2 IMes analogues, we undertook    4 Assessing rates of PCy 3 loss from the decomposition of s-GIIm and u-GIIm in C 6 D 6 . Left: Rate curves at 60 C. Right: Initial rate constants and k rel (normalized to u-GIIm) at 40 C, 60 C, and 80 C. For half-lives and rate plots at other temperatures, see the ESI. † Fig. 5 Assessing the relative stability of u-GIIm and s-GIIm in common solvents, as a proxy for PCy 3 lability (6 h, 60 C oil-bath; 1 H NMR integration vs. TMB). Key chemical shift data for GII and GIIm in these solvents are tabulated in the ESI. † a crystallographic study of u-GIIm, for comparison with the reported structure of s-GIIm. 36 The instability of these complexes in solution can be minimized by low-temperature handling, and X-ray quality crystals of u-GIIm deposited from concentrated solutions in toluene over days at À35 C. The ORTEP plot is shown in Fig. 6; key bond lengths and angles are compared with those for s-GIIm in Table 1.
The geometry at Ru is square pyramidal in both cases, as indicated by the s values of 0.19 (cf. s ¼ 0 for a perfect square pyramid, and s ¼ 1 for a perfect trigonal bipyramid). 37 While the P-Ru-C NHC angle shows some distortion from the 180 ideal (ca. 166 in both u-GIIm and s-GIIm), excellent orbital communication is expected between the trans-disposed phosphine and NHC ligands. Importantly, however, the Ru-P bond distances in s-GIIm and u-GIIm are statistically indistinguishable, despite the nearly tenfold difference in phosphine lability. The absence of a correlation between Ru-PCy 3 bond length and bond strength was pointed out for the parent benzylidene complexes, 14 but has gone widely unnoticed. Frenking has pointed out that metal-ligand bond lengths are not reliable indicators of bond strength, where the ligand can function as an acceptor as well as a donor. 38 The p-acceptor properties of the phosphine ligand in the NHC complexes are discussed below.

Molecular dynamics study: Ru]C NHC rotation and bond order
More direct insight emerged from a molecular dynamics study, in which 2D NOESY-NMR was used to assess rotational exchange between the mesityl rings above and below the basal plane (Fig. 7, top). Exchange cross-peaks were observed for all four unique mesityl methyl signals in u-GIIm and u-GII, indicating rotation about the Ru-IMes bond at room temperature (Fig. 7a). No such cross-peaks were evident for s-GIIm and s-GII (Fig. 7b), even for the well-resolved p-Me singlets (the o-Me singlets are less well resolved, perhaps due to [Ru]]CHPh swiveling). Slower rotation of the H 2 IMes ligand in both the methylidene complex s-GIIm and its benzylidene parent s-GII is important in indicating that restricted rotation is unrelated to the steric demand of the [Ru]]CHR substituent.
Restricted rotation about the Ru-H 2 IMes bond implies increased Ru-C NHC double-bond character, arising from pback-donation from the metal onto the vacant p-orbital on the NHC carbon. Free rotation of the IMes ligand, in contrast, indicates a high proportion of single-bond character in the Ru-C NHC bond. This accords with the experimental and computational ndings described above, showing stronger p-acceptor character for the H 2 IMes ligand than IMes. Bertrand and coworkers drew a similar conclusion in a comparative study of H 2 IPr-PPh and IPr-PPh adducts, also on the basis of a solution dynamics study (IPr ¼ 1,3-bis(2,6-diisopropylphenyl)imidazol-2ylidene). 23 Thus, the saturated H 2 IPr derivative was classied as a phosphaalkene species, and the unsaturated IPr adduct as a phosphinidene.

Origin of the inverse trans effect
As noted in the Introduction, the origin of the dramatically reduced phosphine lability in the second-generation Grubbs catalysts is a puzzle of long standing. Straub suggested that faster PCy 3 loss from GI is due to repulsive interactions between the chloride ligands and the b-hydrogen atoms of the cyclohexyl rings. 39 More recently, Yang, Truhlar and co-workers reported DFT evidence showing that alkylidene rotation functions as a toggle to trigger PCy 3 dissociation, and that the torsional barriers to rotation are higher for s-GII. 15 Kennepohl's XAS study stands out, however, for the unexpected revelation that s-GII exhibits a higher 1s ionization potential for Ruthat is, a more electron-decient metal center than does the rst-generation parent GI. We suggest that this is due to enhanced p-donation from Ru onto the NHC and PCy 3 ligands. It should be noted that the Kennepohl study examined   (8) this possibility for s-GIIm. It was rejected, as calculations at the level of theory then available indicated limited Ru / PCy 3 backbonding (in consequence of which, stronger PCy 3 binding was attributed to an enhanced Ru/PCy 3 electrostatic attraction). Importantly, however, consideration of dispersion forces has since emerged as critical to quantitative evaluation of the PCy 3 dissociation step. 40 The limited role heretofor assigned to Ru-PCy 3 p-acceptor interactions in this system is perhaps unsurprising, given the widespread perception of alkylphosphines as strong s-donors and weak p-acceptors (a situation also encountered in the context of NHC donicity; see above). Here too, however, a reevaluation is in progress. In an analysis of electron density and structural effects, Leyssens, Harvey and co-workers demonstrated that p-backbonding from the metal atom onto the P-R s*-antibonding orbitals can represent a signicant component of metal-phosphine bonding, including for trialkylphosphine complexes. 41 A recent leading review of computational approaches to the understanding of metalphosphorus bonding likewise emphasizes that calculated ligand descriptors for phosphine ligands must consider their p-acceptor character. 42 In light of these developments, we suggest that p-backdonation onto the phosphine is a signicant, overlooked contribution to the low PCy 3 lability in the second-generation Grubbs catalysts. The potent s-donor properties of the NHC ligand constrain back-donation onto any p-acceptor ligands present. For precatalyst s-GII, three ligands can participate in pbackbonding: H 2 IMes, PCy 3 , and benzylidene. 39 In the case of u-GIIm, the poor p-acceptor character of the IMes and methylidene ligands leaves the PCy 3 ligand as the sole entity that can ameliorate the buildup of charge on the metal. We propose that this buildup is offset for u-GIIm by greater Ru / PCy 3 backdonation (Fig. 8), and for s-GIIm, by greater Ru / H 2 IMes backdonation, accompanied by a lesser amount of Ru / PCy 3 backdonation. This would account for the poor net charge donation from the saturated NHC ligand observed in the Kennepohl study. Also relevant in this context is an energy decomposition analysis by Poblet and co-workers, which suggested that the pacceptor capacity of H 2 IMes reduces total charge donation to the metal for s-GIIm, relative to its IMes analogue. 21 Several consequences can be envisaged, which have a profound impact on catalytic behaviour. Most obviously, stronger Ru-P backbonding would account for the reduced lability of the PCy 3 ligand in the IMes complexes, relative to their H 2 IMes analogues. Slower loss of PCy 3 would in turn account for the 7-8-fold longer lifetime shown above for u-GIIm, relative to s-GIIm. Because phosphine dissociation is required for entry into the active catalytic cycle, however, the advantage of longer lifetime is offset by slower initiation for the precatalyst u-GII, and slower re-entry for the resting-state species u-GIIm. This proposal claries the greatly enhanced initiation efficiency of phosphine-free, Hoveyda-class metathesis catalysts, 43 in which the p-accepting PCy 3 ligand is replaced by a p-donating ether ligand, as well as the high latency of the Cazin catalysts, in which a much more strongly p-acidic phosphite ligand is present. 44 In the Neidig study cited in the Introduction, 9 the NHC ligands were shown to signicantly reduce the binding strength of a chloride ligand in tetrahedral Fe-NHC complexes. The strengthening of the trans-PCy 3 bond observed herein is a striking further manifestation of the impact of NHC donicity on M-L binding. Beyond the specic context of olen metathesis, similar inhibition of uptake into catalysis may be expected whenever a p-acceptor ligand must be released in order to bind substrate, particularly where this ligand is trans to an NHC. Such effects are enhanced for systems in which the strong sdonor character of the NHC ligand is undiminished by NHC pacceptor capacity, as illustrated here for the IMes system.

Conclusions
Strong NHC donation is arguably the dening feature of the second-generation Grubbs catalysts, as the parameter that enables their high inherent reactivity. The foregoing reveals that such strong donation wears a Janus face. Enhancing the electron density at the metal center activates the Ru-olen intermediate, and stabilizes the Ru(IV) metallacyclobutane intermediate. However, it also greatly amplies Ru / PCy 3 backbonding: Ru-P bond strengths are thereby increased, and loss of phosphine is severely inhibited. This inverse trans effect is manifested in retarded initiation of the benzylidene precatalysts GII, and very slow re-entry into the catalytic cycle from the resting-state methylidene complexes GIIm.
Notwithstanding the central importance of the Grubbs catalysts and their descendents in olen metathesis, the implications are considerably broader. The transformative impact of NHC ligands on homogeneous catalysis has long been assigned to their capacity to enhance the electron density at the metal. The inuence of NHC donicity on the ancillary ligands, however, is now beginning to be examined more closely. The ndings above contribute to emerging understanding of the profound impact of NHC donicity on M-L binding, and hence on catalytic behaviour. Specically, inhibited initiation is predicted to be a general feature for M-NHC catalysts in which a pacidic ancillary ligand occupies a latent substrate binding site, particularly where such ligands are trans to the NHC. The potential for activation of a p-accepting substrate located in this site is an obvious corollory. These ndings complement recent work highlighting the labilizing effect of the NHC ligand on pdonor ligands in tetrahedral iron complexes. Differences in NHC p-acceptor capacity can thus either mitigate or reinforce trans-type M-L bonding interactions, with major consequences for catalyst conscription and activity.

General procedures
Reactions were carried out under N 2 using standard glovebox techniques, at ambient temperature (RT; 25-27 C, unless otherwise noted). Dry, oxygen-free toluene was obtained using a Glass Contour solvent purication system. All NMR solvents (Cambridge Isotopes) were stored under N 2 over Linde 4Å molecular sieves for at least 6 h prior to use. Dimethyl terephthalate (DMT, >99%), 1,3,5-trimethoxybenzene (TMB, >99%), used as internal integration standards to support quantication in 1 H NMR experiments, were obtained from Sigma-Aldrich. The methylidene complexes u-GIIm and s-GIIm were prepared by literature methods. 17,45 X-ray quality crystals of u-GIIm were grown from toluene at À35 C over 48 h.
NMR spectra were recorded on Bruker Avance 300 and 500 spectrometers at 23 C (unless otherwise noted), and referenced to the residual proton of the solvent. Signals are reported in ppm, relative to TMS ( 1 H) or 85% H 3 PO 4 ( 31 P) at 0 ppm.

Representative procedure for measuring decomposition rates
In the glovebox, a J. Young NMR tube was charged with GIIm (10 mg, 0.013 mmol), TMB (ca. 0.5 mg), and C 6 D 6 (660 mL). The sample was removed from the glovebox and a 1 H NMR spectrum was measured to establish the initial ratio of s-GIIm to TMB. The NMR tube was then transferred to a 40 C oil bath (thermocouple-equipped; AE1.5 C). The rate was determined by collecting 1 H NMR spectra at regular intervals. Rate proles for u-GIIm and s-GIIm at 40 C and 80 C are given in the ESI. † To examine the [PCy 3 ]-dependence of decomposition, a corresponding experiment was carried out with s-GIIm (9.2 mg, 0.0127 mmol), TMB (ca. 0.5 mg), and PCy 3 (35.7 mg, 0.127 mmol, 10 equiv.) in C 6 D 6 (635 mL) at 60 C. Time-points were taken at regular intervals until decomposition was complete.

Exploring the impact of solvent on decomposition of GIIm
These experiments were carried out as above at a bath temperature of 60 C, with NMR analysis at a single time-point (6 h). Thermolysis experiments in CD 2 Cl 2 (b.p. 40 C) were carried out in thick-walled J. Young NMR tubes.