Toward E-selective Olefin Metathesis: Computational Design and Experimental Realization of Ruthenium Thio-Indolate Catalysts

The selective transformation of 1-alkenes into E-olefins is a long-standing challenge in olefin metathesis. Density functional theory (DFT) calculations predict high E-selectivity for catalysts incorporating a bidentate, dianionic thio-indolate ligand within a RuXX’(NHC)(py)(= CHR) platform (NHC = N-heterocyclic carbene; py = pyridine). Such complexes are predicted to yield E-olefins by favoring anti-disposed substituents in the transition state expected to be rate-determining: specifically, that for cycloreversion of the metallacyclobutane intermediate. Three pyridine-stabilized catalysts Ru21a-c were synthesized, in which the thio-indolate ligand bears a H, Me, or Ph substituent at the C2 position, and the NHC ligand is the unsaturated imidazoline-2-ylidene Me2IMes (which bears N-mesityl groups and methyl groups on the C4,5 backbone). Single-crystal X-ray diffraction analysis of Ru21c confirms the ligand orientation required for E-selective metathesis, with the thio-indolate sulfur atom binding cis to the NHC, and the indolate nitrogen atom trans to the NHC. However, whereas the new complexes mediated metathetic exchange of their 2-thienylmethylidene ligand in the presence of the common metathesis substrates styrene and allylbenzene, no corresponding self-metathesis products were obtained. Only small amounts of 2-butene (73% (Z)-2-butene) were obtained in self-metathesis of propene using Ru21a. Detailed DFT analysis of this process revealed that product release is surprisingly slow, limiting the reaction rate and explaining the low metathesis activity. With the barrier to dissociation of (Z)-2-butene being lower than that of (E)-2-butene, the calculations also account for the observed Z-selectivity of Ru21a. These findings provide guidelines for catalyst redesign in pursuit of the ambitious goal of E-selective 1-alkene metathesis.

Even more elusive are catalysts for E-selective olefin metathesis. Despite more than 20 years of effort, no catalyst for E-selective metathesis of 1-alkenes has yet been achieved. To date, metathetical access to E-olefin products can be achieved only via "stereoretentive" catalysts (Chart 1b), which can transform stereochemically defined E-olefin substrates into E-configured products (Scheme 1) [36,[52][53][54][55]. The utility of stereoretentive metathesis is limited by the cost and accessibility of the isomerically pure starting materials required. Production of E-olefinic products from 1-alkenes represents an intellectually and economically attractive alternative.
Here we describe work toward the design of catalysts for E-selective 1-alkene metathesis. Building on insights obtained in earlier modifications of stereoretentive catalysts [56], we explore a new family of thio-indolate catalysts for which E-selectivity is predicted on the basis of density functional theory (DFT) calculations.

Initial Considerations
Stereoselective metathesis is achieved by controlling the catalyst stereochemistry in of the rate-determining step of the Chauvin mechanism [57], typically cycloreversion to release the product olefin from the metallacyclobutane (MCB) intermediate [53,[58][59][60]. A syn relationship between the substituents in the MCB results in a Z-configured product; antidisposed MCB substituents yield an E-configured product.
Seminal computational mechanistic work showed that the MCB intermediates and the associated transition states of the preferred dissociative reaction pathways adopt trigonal bipyramidal (TBP) geometries, in which the η 2 -C α -C β -C α ring occupies the equatorial plane [61]. Such a geometry is likewise observed for stereoretentive catalysts Ru13-16 (Scheme 1b) [36,[53][54][55][62][63][64]. In the MCB intermediate, the NHC ligand occupies one of the two axial sites: that is, it is cis to the MCB ring. This positions the NHC N-aryl substituents in close proximity to the MCB α-carbon atoms. Steric pressure from the N-Ar groups hence forces orientation of the C α -substituents away from the NHC.
If stereochemically-defined internal olefins are used as substrates, this also sets the orientation of the β-substituent. The substrate stereochemistry thus controls the stereochemistry of the metathesis product. That is, use of an E-olefin dictates formation of an E-configured product; a Z-olefin substrate yields a Z-configured product (Scheme 1) [60]. These stereoretentive catalysts offer the only current metathetical route to E-olefins. They are therefore an attractive starting point for the next logical step: developing catalysts that selectively convert 1-alkenes to E-configured alkene products, and circumventing the need for isomerically pure starting materials.
Discussed above is the means by which steric pressure on the α-substituent of the MCB enables formation of Z-(Scheme 1b) or E-configured products (Scheme 1c) via retention of the substrate stereochemistry. E-selective metathesis of 1-alkenes, in contrast, requires steric pressure on the β-substituent. Grubbs and co-workers pursued this objective by introducing a phenanthrene-dithiolate ligand ( Fig. 1) [56]. In the key MCB intermediate, however, the favored isomer is the undesired Ru17_22, in which the phenanthrene ring system is oriented away from the MCB β-position. Moreover, even for the target isomer Ru17_22', DFT calculations on a model, unsubstituted MCB predicted a 5 Å separation between H β and the phenanthrene ring. Larger substituents would reduce this distance, but the catecholthiolate ligand appears too distant to influence the orientation of the β-substituent to any great extent. This prompted us to pursue design of alternative dianionic ligands with greater influence on the β-site of the MCB, using DFT calculations as a guide.

Computational Ligand Design
To reduce the distance between the selectivity-inducing group and the MCB β-position, we envisaged replacing one S-donor with a trivalent, anionic donor, thereby retaining a neutral Ru complex. Specifically, we considered introduction of a nitrogen center bearing a substituent that would reduce the distance to the MCB ring. We further stipulated a planar, rigid, and bicyclic κ 2 -S,N ligand, to maximize steric pressure on the MCB β-position.
These requirements led us to the thio-indolate scaffold shown in the model unsubstituted MCB (Ru19a_22) in Chart 2. The DFT-optimized geometry of Ru19a_22 revealed a much shorter distance between H β and the indole ring than between H β and the thiocatecholate ring in Ru13_22 (2.35 Å vs 5.95 Å, respectively: Chart 2). Substitution at position 2 of the indolate ring should reduce the distance further, increasing the steric pressure. Using propene as a computationally efficient model 1-alkene, we investigated the impact of different substituents at the indole 2-position using DFT calculations.
Thio-indolate catalyst Ru19a, with only a hydrogen atom at position 2 of the indole ring, is predicted to be E-selective, as judged from the difference in free energy between the transition states leading to (Z)-or (E)-2-butene (ΔΔG ‡ (E/Z) = ΔG ‡ Z -ΔG ‡ E = 3.4 kcal mol −1 : see Table 1). In contrast, the state-of-the-art catecholthiolate catalyst Ru13 [53] (Chart 1) is predicted to be Z-selective (ΔΔG ‡ (E/Z) = -1.2 kcal mol −1 ), in agreement with experiment: using Ru13, 83% Z-selectivity is obtained in self-metathesis of propene. 1 Also of note, lower barriers to metathesis are predicted for thio-indolate catalysts Ru19a and Ru19b (the latter bearing a 2-Me substituent) vs Ru13 (∆G ‡ Z = 21.0 kcal mol −1 ) as well vs a corresponding (hypothetical) catalyst precursor Ru18 bearing the same methoxybenzylidene as Ru19a-c. These data reinforce the potential of the thio-indolate ruthenium alkylidenes as olefin metathesis catalysts.
2-Substitution destabilizes the Z-transition state more than the corresponding E-isomer, leading to increasing E-selectivity with increasing substituent size (Table 1). With a methyl substituent, for example (Ru19b), ΔΔG ‡ (E/Z) is 0.6 kcal mol −1 higher than that of Ru19a. A phenyl substituent (Ru19c) is predicted to increase the selectivity substantially (ΔΔG ‡ (E/Z) = 11.5 kcal mol −1 ), but the significantly higher barrier to metathesis is expected to limit the catalytic activity of Ru19c relative to Ru19a and Ru19b.
The increased selectivity predicted for Ru19a is a result of the steric pressure on the MCB β-substituent, as discussed above. This pressure is reflected in the increased Ru-C β -CH 3 angle (119° in Ru13_TS4,5 Z , vs 126° in Ru19a_TS4,5 Z : Fig. 2), and the decreased C NHC -Ru-C β angle (which declines from 105° to 92°).

Experimental Realization
In light of the promising effect of the thio-indolate ligands on metathesis stereoselectivity, we proceeded to synthesize the first versions of ruthenium alkylidenes bearing such ligands. The required 7-bromoindoles L1a-c (Scheme 2) are commercially available.
A thiol group was installed in place of the bromine atom by Pd-catalyzed exchange with a silylthiolester [65], following which the thiol was deprotected with HCl (R = Me, Ph) or N n Bu 4 F (R = H). This two-step protocol delivered the corresponding thio-indoles L3a-c in yields of 68-72%. Ensuing treatment with KH in THF afforded the dianions as potassium salts (L4a-c), which were extracted with hexane to remove PPh 3 residues introduced in the prior steps.
Salt metathesis of L4a-c with the second-generation Hoveyda catalyst HII to give Ru19a-c failed, despite the success of the corresponding reactions with unsubstituted [53, Fig. 1 Phenanthrene-dithiolate complex Ru17 and its TBP-like MCB intermediates [56]. The steric bulk of the extended dithiolate ligand is located far from the critical β -position in the favored isomer Ru17_22 Chart 2 Thiocatecholate vs Thio-indolate Ligands: Amplifying Steric Pressure at the MCB β-Position. DFT-calculated interatomic distances from the ratio of (Z)-and (E)-2-butene (86:14) in the first 1 H NMR spectrum (t = 5 min), to reduce the effect of isomerization (see Table S1). 62] or sterically demanding [56] catecholthiolates. Reaction of L4c with the third-generation Grubbs catalyst GIII proceeded, but the formed complex was unstable and could not be purified. Greater success was achieved in salt metathesis with the Evonik catalyst Ru20, perhaps because of the reduced steric demand of the 2-thienylmethylidene and the unsaturated NHC ligand. It may be noted that Ru20 has been successfully used as a precursor to other Z-selective catalysts bearing sterically demanding thiolates [49]. The target thio-indolate alkylidene complexes Ru21a-c (a: R = H; b: R = 2-Me; c: R = 2-Ph; Scheme 2) were obtained in 60-65% yield. These are, to our knowledge, the first transition-metal complexes bearing thio-indolate chelate ligands. The new complexes were characterized by NMR and MS analysis, and, in the case of Ru21a and Ru21c, single-crystal X-ray diffraction (Fig. 3). The X-ray crystal structure of Ru21a confirms that the atom connectivity is analogous to that of Ru21c, but the diffraction quality is too low for detailed structural analysis. The X-ray structure confirms binding of the thio-indolate fragment as a S,N-chelate. Crucially, and as predicted by the DFT calculations, the dianionic ligand adopts the orientation required for E-selective metathesis, with the thiolate sulfur cis, and the indolate nitrogen trans, to the NHC (Table S6). However, DFT calculations predict that the unintended isomer (Ru21a'), with the thiolate sulfur trans and the indolate nitrogen cis to the NHC, is only 1.9 kcal mol −1 less stable and may thus also be present (see Table S7). Indeed, 1 H NMR spectra of catalyst Ru21a consistently exhibit a minor alkylidene singlet (5%) at δ = 15.4 ppm, with the main alkylidene signal located at δ = 16.2. A NOESY experiment confirms the existence of an exchange equilibrium between these two alkylidene singlets, with the minor alkylidene species being 1.8 kcal mol −1 less stable than the dominating species. From the agreement between the NMR experiments and the DFT calculations, the minor species is presumed to be Ru21a'. Such minor alkylidene species are not observed for the larger complexes Ru21b-c, and calculations also indicate that the isomers Ru21b' and Ru21c' with rotated S,N ligands are high in energy (Table S6).
The metathesis activity of Ru21a-b was initially assessed by reaction with styrene at room temperature (Scheme 3). Unexpectedly, 1 H NMR analysis showed no evidence of the stilbene self-metathesis product. However, a new alkylidene singlet was observed, along with vinylthiophene, in the experiments involving Ru21a and Ru21b (δ 17.2 and 16.9 ppm, respectively, in C 6 D 6 : Fig. S2, S3). The new alkylidene species, identified as the benzylidene analogues of Ru21a and Ru21b (labeled Ru21a_29 and Ru21b_29, respectively) result from unproductive metathesis of styrene, via Ru21a_28 (see Scheme 3, Fig. S2-S4, and Scheme S1). 2 Complex Ru21a reaches equilibrium within 15 min, vs nearly an hour for catalyst Ru21b. The slower reaction of Ru21b is consistent with the higher barrier to metathesis Fig. 2 Optimized geometries of the Z-isomeric transition states for cycloreversion in propene self-metathesis by complexes Ru13 and Ru19a. Ruthenium is shown in violet, sulfur in yellow, nitrogen in blue, chlorine in green, and carbon in gray. Distances in [Å], angles in [°] calculated for the bulkier thio-indolate ligands of the H 2 IMes analogues Ru19 (Table 1). Increasing the reaction temperature or time resulted in loss of the alkylidene signals as well as a black precipitate indicating catalyst decomposition and formation of Ru nanoparticles [69].
In light of the low reactivity documented above, especially for Ru21b, we speculated that unfavorable steric interactions may hamper productive metathesis. Formation of stilbene would necessitate an MCB structure in which the β-phenyl substituent approaches the two N-mesityl groups (for the (E)-stilbene), or the thio-indolate ligand (for the (Z)-stilbene). The impact on the barriers to metathesis is explored computationally below.
To test whether reduced steric bulk at C β would enable productive metathesis, we examined the reaction with allylbenzene (Fig. S5). Again, however, no metathesis products were detected. In the case of Ru21b, 35% vinylthiophene was detected by 1 H NMR analysis (Fig. S5, S6), but 17% Ru21b remained even after 12 h, confirming slow initiation and a relatively stable precatalyst. Isomerization of allylbenzene was also observed, presumably catalyzed by decomposed Ru species [69,70]. In sum, attempts at selfmetathesis of allylbenzene led to catalyst decomposition and substrate isomerization, rather than productive metathesis.
Next, to reduce the steric pressure as much as possible, with the goal of facilitating productive metathesis, we attempted self-metathesis of propene in NMR experiments with Ru21a-c. 3 Ru21a gave the expected butene product in low yield (11 mol% vs catalyst loading), and the proportion of the Z-isomer was slightly lower than that obtained with catalyst Ru13 (73% vs 83%). The more sterically demanding catalysts Ru21b-c afforded no butene product. For catalyst Ru21b, this is clearly due in part to low metathesis activity, as unreacted Ru21b remained even after 96 h at 50 °C. In contrast, the alkylidene signal of catalyst Ru21c disappeared within 12 h of reaction time.
The small amount of butene obtained using Ru21a in propene self-metathesis may be due to either low catalytic activity, perhaps caused by the steric hindrance of the thioindolate ligand, or to catalyst decomposition. To probe its susceptibility to β-hydride elimination from the unsubstituted MCB (a key decomposition pathway in 1-alkene metathesis for a range of Ru-NHC catalysts [71]), Ru21a Scheme 2 Ligand synthesis and installation was reacted with ethylene. Both vinylthiophene and propene 4 were detected, evidence for alkylidene exchange (catalyst initiation), and β-hydride elimination (Fig. S7, S8). The proportion of propene is consistent with decomposition of ca. 40% of the catalyst via β-hydride elimination. (Bimolecular decomposition of the 4-coordinate methylidene species may also occur [72], but the ethylene product is indistinguishable from ethylene formed via metathesis). The competition between metathesis and decomposition is further explored in the mechanistic computational analysis below.

Mechanistic Calculations
The calculated barrier to cycloreversion -the presumed rate-determining step -for Ru19a (the H 2 IMes analogue of Ru21a) is similar to that of the known metathesis catalyst Ru13 [53,[58][59][60]. Nevertheless, Ru21a is inactive in metathesis of styrene or allylbenzene, and produced only small proportions of 2-butene in self-metathesis of propene. Moreover, the major stereoisomer produced was (Z)-2-butene, despite the predicted E-selectivity of Ru19a (Table 1).
To uncover the factors underlying the discrepancy between the catalytic properties predicted for Ru19a and those observed for Ru21a, the latter was subjected to detailed computational analysis.
The calculated barrier to cycloreversion in self-metathesis of styrene and allylbenzene to E-configured products by Ru21a is 36.9 and 37.5 kcal mol −1 (Scheme S12, Table S7), respectively, consistent with the absence of product 5 observed experimentally. In contrast, for propene, the cycloreversion barrier (via Ru21a_TS4,5 E , 23.8 kcal mol −1 vs the precursor Ru21a) leading to formation of (E)-2-butene, is only 2.8 kcal mol −1 higher than that of the state-of-the-art catecholthiolate catalyst Ru13 (Table 1), indicating that the critical bond rupture and formation of propene metathesis should be within reach for Ru21a. The catalytic potential of Ru21a seems even clearer when eliminating the effect of the precursor (the pyridine-coordinated Ru21a vs the isopropoxybenzylidene-coordinated Ru13), by calculating free energies relative to the two active ethylidene complexes derived from Ru21a (Ru21a_2) and Ru13 (Ru13_2), respectively (Scheme 4). In fact, with initiation completed, Ru21a_2 should mediate both cycloaddition and cycloreversion faster than Ru13_2. However, Ru21a is observed to be less active in metathesis than Ru13. Because Ru21a is unlikely to be limited by slow initiation (initiation was observed even for styrene and allylbenzene; see above), the most likely explanation for its slow metathesis lies in a reaction step other than cycloreversion.
In searching for an alternative rate-limiting step, we did not initially consider 2-butene dissociation. However, Cavallo and co-workers, in an early computational study of the stereoselectivity of propene self-metathesis using a RuCl 2 (H 2 IMes)-ethylidene catalyst [73], found the methylidene complex and 2-butene to be of higher energy than any other minimum or transition state in the catalytic cycle. Even if this is not the case for Ru21a, the relatively high energies of complexes toward the end of the pathway indicate that product release could be slow. For example, the 2-butene π-complexes Ru21a_5 are significantly less stable than the propene counterparts Ru21a_3. We therefore considered whether product release from Ru21a_5 might be rate determining. Indeed, product release is surprisingly costly (10-13 kcal mol −1 vs Ru21a_5, see Scheme 4, or 29-32 kcal mol −1 vs Ru21a), and is the undisputed bottleneck for the thio-indolate catalyst. This step, which is difficult to follow computationally, 6 was not investigated for the corresponding thio-catecholate catalyst Ru13. However, Indolate-induced deprotonation of the MCB [71] was ruled out as a source of propene, since calculations with stepwise reduced N-H Cβ distance invariably led, instead, to β-hydride elimination. Also, upon completion of the reaction, no NH signals were observed in the 1 H-15 N-HSQC-NMR spectrum. product release is not expected to be a bottleneck for the latter, which is less bulky than Ru21a. Previous computational studies of this catalyst do not suggest rate-limiting product dissociation [53,55,60]. Instead, cycloreversion has been suggested to be rate limiting for this and other stereoretentive catalysts [60].
In summary, with cycloreversion assumed to be rate determining for Ru13 and product release being identified as the bottleneck for Ru21a, the calculations are consistent with the much lower catalytic activity of Ru21a relative to the closely related thio-catecholate catalyst Ru13. The calculations are also consistent with the observed Z-selectivity of both Ru21a and Ru13, given the lower barrier to dissociation of (Z)-2-butene than (E)-2-butene from Ru21a_5, and the lower barrier to cycloreversion via Ru13_TS4,5 Z than Ru13_TS4,5 E .
The calculations indicate that product release is a twostep process. Surprisingly high barriers for the thio-indolate catalyst are located in the first step, involving rearrangement from η 2 -to η 1 -coordinated 2-butene (Fig. 4). This rearrangement requires considerable activation, as the Ru-butene π-bond is lost at the same time as the steric repulsion between the ligands (the NHC and the thio-indolate) and the leaving, but still largely η 2 -coordinated, 2-butene, is large. The steric repulsion is lower for (Z)-2-butene than for (E)-2-butene, resulting in lower barriers to rearrangement to the agostic complex. The reduced steric hindrance results, at least in part, from (Z)-2-butene being more compact than its  . 4 The optimized transition states for rearrangement of η 2 -bound to η 1 -bound 2-butene in π-complexes Ru21a_5E and Ru21a_5Z to give the corresponding agostic complexes Ru21a_6E and Ru21a_6Z, respectively. This rearrangement is the rate-determining step of Ru21a-mediated propene metathesis and initiates product release. The subsequent (E)or (Z)-2-butene dissociation to give methylidene complex Ru21a_7 requires less geometric adaption and less activation. E-counterpart, with a smaller molecular volume and surface area, (Fig. 4). Faster release of the Z-configured product is likely to be a challenge extending far beyond the current thio-indolate catalyst design: regardless which of the existing catalyst frameworks is chosen as a starting point for design of E-selective catalysts, substitution is likely to be essential to disfavor formation of Z-configured products. This substitution will increase the overall steric pressure and tend to make product release the kinetic bottleneck.
The slowest step of the product release, the η 2 -to-η 1bound 2-butene rearrangement, leads to sterically less encumbered complexes Ru21a_6, in which 2-butene is bound to Ru via an agostic methyl C-H bond. From Ru21a_6, the continued 2-butene dissociation to methylidene Ru21a_7 and free 2-butene requires much less geometric adaption (see Fig. 4) and is thus expected to require little activation. Constrained geometry optimizations at increasing R-H agostic distances and failed attempts at locating the corresponding transition states, 6 confirm that this, the final part of the product release, requires only negligible enthalpic activation from Ru21a_6.
To shed further light on the factors underlying the low observed metathesis activity, we also investigated a range of decomposition modes for Ru21a and its isomer Ru21a'. Specifically, we considered β-H elimination [74] and nucleophilic attack of the thio-indolate ligand on the alkylidene [53] during productive (Scheme S5), non-productive (Scheme S6) and regenerative propene metathesis (i.e., regeneration of the ethylidene Ru21a_2: Scheme S7). In addition, nucleophilic attack and β-hydride elimination, occurring during reaction of the Ru-methylidene with ethylene, were examined (Scheme S8).
Consistent with the observed formation of propene on reaction of Ru21a with ethylene (see above), the calculations indicate a relatively low barrier to β-H elimination of the unsubstituted MCB (ΔG ‡ = 21.7 kcal mol −1 vs Ru21a). However, the barrier to β-hydride elimination is consistently higher (by 4.5-6.2 kcal mol −1 ) than that of cycloreversion during metathesis itself (Scheme S5-S8). Thus, thio-indolate catalyst Ru21a does not appear to be particularly vulnerable to β-H elimination.
The calculations suggest that nucleophilic attack is more likely (Scheme S5-S8). Indeed, consistent with the previously reported nucleophilic attacks of both catechothiolates [53] and amines [75,76] on alkylidenes, indolate attack on the ethylidene during regenerative metathesis emerges as the dominant decomposition pathway (ΔG ‡ = 19.7 kcal mol −1 vs Ru21a, Scheme S7). This reaction releases ethylene, which, in turn, may react with a second ruthenium ethylidene or methylidene complex, thereby accelerating decomposition. However, the calculations do not suggest that this vulnerability to nucleophilic attack is inherent to the design of Ru21a, with the indolate nitrogen atom trans to the NHC.
In short, the calculations reveal no decomposition modes intrinsic to the thio-indolate catalyst design that should make these catalysts more vulnerable than, e.g., the related dithiolate catalyst Ru13. Instead, the low catalytic activity and the decomposition observed for these catalysts appear to be the result of unusually high barriers to releasing the internal-olefin product at the end of the metathesis reaction. The exceptional height of these barriers originates from the added steric pressure of the thio-indolate ligand required to achieve E-selectivity.

Conclusion
Based on considerations of the geometries of stereoretentive metathesis catalysts [77] and on catecholthiolate modifications aimed at increasing the share of E-isomeric product [56], a thio-indolate ligand scaffold was designed to exert steric pressure on the β-substituent of the MCB, and the MCB-like transition states for cycloaddition and cycloreversion. DFT calculations predicted that the S,N-thio-indolate chelate should bind to ruthenium with an orientation suitable to exert the desired steric pressure. Furthermore, DFT-calculated energy differences (ΔΔG ‡ (E/Z ) between the cycloreversion transition states for propene self-metathesis leading to (Z)-2-butene and (E)-2-butene suggested that ruthenium-alkylidene thio-indolate complexes would favor E-isomer products.
To follow up the computational predictions, the first metal complexes bearing bidentate thio-indolate ligands were synthesized, isolated, and characterized (Ru21a-c). The thio-indolate chelates are compatible with the 2-thienylmethylidene and Me 2 IMes ligands present in known, active metathesis catalysts, and the general robustness of complexes Ru21a-c is comparable to that of other ruthenium catalysts for olefin metathesis.
However, whereas the new complexes participated in metathetic exchange with styrene and allylbenzene, liberating the 2-thienylmethylidene ligand, no self-metathesis products were obtained. Even self-metathesis of propene using Ru21a yielded only small proportions of 2-butene (73% (Z)).
Detailed mechanistic DFT calculations of propene selfmetathesis by Ru21a and its isomer Ru21a' revealed barriers to product release from the Ru-2-butene π -complex much higher than those of cycloreversion of the MCB, the step repeatedly identified as rate limiting in computational studies 1 3 of ruthenium-catalyzed olefin metathesis [53,[58][59][60]. The barriers to product release are also much higher than those of common decomposition reactions, and are caused by the additional steric bulk of the thio-indolate. This steric bulk leaves little room for the escaping, bulky disubstituted olefin, which experiences steric repulsion from the thio-indolate ligand, in particular, and the mesityl methyl groups of the Me 2 IMes ligand. This repulsion adds to the cost of losing the Ru-olefin π-bond, resulting in unusually high barriers to product release. The negative effect of steric bulk predicted by the calculations is consistent with the metathesis inactivity of the bulkier complexes Ru21b and Ru21c.
E-selective 1-alkene metathesis catalysts have been sought in vain for more than two decades. The present study reveals the dual challenge of this molecular-design goal: (1) Steric pressure must be exerted in opposite directions to closelyspaced substituents of the nascent disubstituted olefin. (2) The net steric congestion must be sufficiently low to permit productive metathesis and, in particular, for the product olefin to dissociate from the metal. The thio-indolate complexes described herein are, to our knowledge, the first catalysts for which calculations suggest that Challenge 1 can be met. However, these findings also underline the difficulty in achieving E-selectivity without incurring excessive steric congestion.

Acknowledgements
The authors gratefully acknowledge financial support from the Research Council of Norway (RCN) via the FRIPRO program (grant number 262370) and via the Norwegian NMR Platform, NNP (226244). The RCN is also thanked for CPU (NN2506K) and storage resources (NS2506K). Bjarte Holmelid is thanked for assistance with the HRMS ESI + analyses and Inger Johanne Fjellanger for assistance with the elemental analysis.
Funding Open access funding provided by University of Bergen (incl Haukeland University Hospital). The authors gratefully acknowledge financial support from the Research Council of Norway (RCN) via the FRIPRO program (grant number 262370) and via the Norwegian NMR Platform, NNP (226244). The RCN is also thanked for CPU (NN2506K) and storage resources (NS2506K.

Availability of data and material
The crystal structure of compound Ru21c is available in the Cambridge Structural Database (CSD, as compound number CCDC-2,086,885). The following is available in an additional supporting file (PDF): A description of the experimental procedures and of additional experimental results, NMR, MS, and X-ray data for new compounds, a description of the computational details and of additional computational results, and calculated electronic and free energies of all computationally investigated minima and transition states. The corresponding DFT-optimized molecular Cartesian coordinates are available in an additional supporting file, in a format convenient for visualization (XYZ).

Conflict of interest
The authors declare no competing financial interest.
Open Access This article is licensed under a Creative Commons Attribution 4.0 International License, which permits use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons licence, and indicate if changes were made. The images or other third party material in this article are included in the article's Creative Commons licence, unless indicated otherwise in a credit line to the material. If material is not included in the article's Creative Commons licence and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this licence, visit http:// creat iveco mmons. org/ licen ses/ by/4. 0/.