An Initiation Kinetics Prediction Model Enables Rational Design of Ruthenium Olefin Metathesis Catalysts Bearing Modified Chelating Benzylidenes

Abstract

Rational design of second-generation ruthenium olefin metathesis catalysts with desired initiation rates can be enabled by a computational model that depends on a single thermodynamic parameter. Using a computational model with no assumption about the specific initiation mechanism, the initiation kinetics of a spectrum of second-generation ruthenium olefin metathesis catalysts bearing modified chelating ortho-alkoxy benzylidenes were predicted in this work. Experimental tests of the validity of the computational model were achieved by the synthesis of a series of ruthenium olefin metathesis catalysts and investigation of initiation rates by UV/Vis kinetics, NMR spectroscopy, and structural characterization by X-ray crystallography. Included in this series of catalysts were thirteen catalysts bearing alkoxy groups with varied steric bulk on the chelating benzylidene, ranging from ethoxy to dicyclohexylmethoxy groups. The experimentally observed initiation kinetics of the synthesized catalysts were in good accordance with computational predictions. Notably, the fast initiation rate of the dicyclohexylmethoxy catalyst was successfully predicted by the model, and this complex is believed to be among the fastest initiating Hoveyda–Grubbs-type catalysts reported to date. The compatibility of the predictive model with other catalyst families, including those bearing alternative NHC ligands or disubstituted alkoxy benzylidenes, was also examined.

Graphical Abstract

INTRODUCTION

Olefin metathesis is a powerful method to form carbon–carbon double bonds¹ with control of regio-² and stereoselectivity.³ There are extensive applications of cross metathesis (CM), ring closing metathesis (RCM) and ring-opening metathesis polymerization (ROMP) for a variety of purposes, ranging from natural product synthesis⁴ to chemical feedstock conversion⁵ to advanced materials synthesis.⁶ One of the major milestones in the development of olefin metathesis as a synthetic method was the discovery of well-defined ruthenium–alkylidene catalysts,⁷ which demonstrated superior tolerance towards moisture, air and functional groups compared to early transition metal catalysts.

Figure 1 shows a collection of commonly used ruthenium-based olefin metathesis catalysts. Each type of catalyst is unique in its reactivity and selectivity profile and hence has unique olefin metathesis applications. For example, catalysts 6 and 7 are especially fast-initiating and suitable for catalyzing ROMP to give polymers with high molecular weight and low dispersity.⁸ Catalyst 5, discovered by Hoveyda and Blechert,⁹ is well known for its stability and tolerance to air, moisture, and Lewis basic functional groups. The chelation of the ortho-alkoxy benzylidene to the ruthenium center at the axial site results in the exceptional stability of 5 and its derivatives, making them valuable tools for organic synthesis.^{1a, 1b, 10} Another advantage of catalyst 5 is the absence of the labile phosphine ligands which are responsible for certain phosphine-mediated catalyst decomposition pathways.¹¹

Figure 1.

Commonly used ruthenium-based olefin metathesis catalysts.

The initiation kinetics of an olefin metathesis catalyst have significant impact on its reactivity and hence its applications.^{1a, 1b, 10a, 12} Since the initiation mechanism for Hoveyda–Grubbs-type catalysts is an ongoing topic of investigation,¹³ it is valuable to establish a model to correlate initiation kinetics with thermodynamic parameters in a manner that avoids making assumptions of the initiation mechanism. In a previous report, we systematically investigated a series of ruthenium-based olefin metathesis catalysts bearing modified chelating ortho-alkoxy/aryloxybenzylidenes (Figure 2).¹⁴ It was shown that by tuning the structure of the alkoxy/aryloxy group, catalysts with a wide range of initiation rates could be accessed. More importantly, a computational model that makes no specific assumptions regarding the initiation mechanism was proposed and was tested versus the initiation rates of the catalysts studied. This model established a good correlation between the computed reaction energy of initiation with butyl vinyl ether (BVE) to form a common 14-electron Fischer carbene complex C (ΔG_r(A → C)) and the experimental initiation rate of the catalyst (Figure 3). The correlation is given by Equation 1. In the present work, this model is employed to predict the initiation kinetics of new catalysts, and these predictions are then tested experimentally.

Figure 2.

Catalysts investigated previously with Hoveyda-type chelating benzylidenes (8–15).

Figure 3.

Computational model for catalyst initiation kinetics.

RESULTS AND DISCUSSION

Catalyst Design.

Having previously established the aforementioned computational model with an initial training set, we next sought to use this model to extrapolate and predict the initiation rates of other ruthenium-based olefin metathesis catalysts bearing modified chelating ortho-alkoxy benzylidenes. To this end, a wide variety of Hoveyda–Grubbs type olefin metathesis catalysts were considered (16–38). Catalyst 20, 23 and 25 are known catalysts,¹⁵ yet their initiation rates have not been previously measured. As analogues to the standard i-Pr control, catalysts 16–19 were proposed to examine the effect of increasing steric bulk from ethoxy to tert-butoxy. Catalysts 20 and 21 were considered to evaluate the effect of the distance of the phenyl ring on the initiation rates as compared to catalyst 11 discovered by Plenio and coworkers.¹⁶ This series of alkoxy-modified Hoveyda–Grubbs variants also includes cycloalkyl groups of various ring sizes, from cyclobutyl to cyclooctyl groups (22–26). They provide a direct comparison with the cyclopropyl catalyst (12), which was included in the original training set. Dicyclohexylmethoxy (27) catalyst was intended to mimic the parent i-Pr catalyst, but with substantially greater steric bulk.¹⁷ Similar to the rationale in proposing catalysts 20 and 21, catalyst 28 was included to probe the impact of additional methylene spacer at the 1-adamantyl position compared with catalyst 14. As catalyst 15 was discovered to be fast-initiating in our previous report,¹⁴ we hoped to enhance the initiation rate even further by introducing an additional methyl group at the 2-adamantyl position. It is also interesting to examine how well the model is able to predict the initiation properties of catalysts bearing electron-withdrawing groups attached to oxygen,^{12g, 15a} such as a trifluoromethyl group (30–31). To further evaluate the scope of the computational model, we also examined the known 2,6-dimethoxy analogue (32)^{12e, 18} of catalyst 10. In particular, we were interested to investigate the relationship between OMe-exchange and initiation in this catalyst. Lastly, we sought to test the scope of the computational model by examining catalysts with modified N-heterocyclic carbene (NHC) ligands. As common variants of the standard SIMes NHC ligand, SIPr¹⁹ and SITol²⁰ NHC ligands were used, and a series of catalysts bearing these two ligands and chelating benzylidenes with different steric bulk at the alkoxy position was prepared (33 – 38).

Computed Ru–O Bond Energies and Predicted Initiation Rates.

The computational method used for the proposed catalysts is consistent with previous computational studies.^14,21 The strengths of Ru–O bonding interactions were evaluated in two ways (Figure 3): (1) by calculating the energy difference between the nonchelated 14-electron complex B formed by dechelation of the Ru–O bond and rotation of the o-alkoxyphenyl group and the ground-state chelated conformation A (ΔG_r(A → B)) and (2) by calculating the reaction energy to form a common 14-electron Fischer carbene complex C from the ground-state chelated conformation A and BVE (ΔG_r(A → C)). The geometries were optimized in the gas phase using B3LYP with a mixed basis set of LANL2DZ for ruthenium and 6–31G(d) other atoms. Energies were calculated with M06 single-point calculations with a mixed basis set of SDD for ruthenium and 6–311+G(d,p) for other atoms using the SMD solvation model in toluene. B3LYP, BP86, M06, M06-L and ωB97x-D have been previously employed to optimize the geometries; B3LYP and BP86 were found to produce better correlation between the computed Ru–O bond strengths and the observed initiation rate (ln(k_obs)) than the others.¹⁴

Computed Ru–O bond strengths derived from both methods (ΔG_r(A → B) and ΔG_r(A → C)) are summarized in Table 1. The initiation rates of the proposed catalysts (16–38) were predicted by Equation 1 with the calculated ΔG_r(A → C) values. In order to clearly compare and contrast the computational results, the calculated ΔG_r(A → C) values for catalysts bearing SIMes NHC ligand are presented in Figure 4, and the results with catalysts bearing non-SIMes NHC ligands are shown in Figure 5. It should be noted that, except for the ethoxy catalyst 16, all proposed catalysts have a ΔG_r(A → C) value lower than the i-Pr control (entry 1, Table 1), hence they are predicted to initiate faster than 5. In particular, the calculated ΔG_r(A → C) values for 27 and 29–31 are significantly lower than others, which is attributed to steric clashing between the bulky alkoxy group (dicylohexylmethoxy or 2-methyl-2-adamantyloxy) and the SIMes-dichlororuthenium fragment and to the strong electron-withdrawing effect of the trifluoromethyl group in weakening Ru–O coordination and destabilizing the catalysts. The benzyl catalyst (20) is predicted to initiate faster than the phenethyl catalyst (21), as the phenyl ring is closer to the ruthenium center in 20. Moreover, ΔG_r(A → C) values for catalysts bearing larger cycloalkyl rings (24–26) are not significantly smaller than those of catalysts with smaller ring sizes (22–23) due to the structural flexibility of the rings. Consistent with the results from a recent computational study by Trzaskowski et al.,²² catalysts 22 and 23 are predicted to initiate with similar rates as the isopropyl control 5. It is also surprising to see that the calculated ΔG_r(A → C) values for catalysts bearing non-SIMes NHC ligands (33–38) remain relatively constant across the different alkoxy groups (Figure 5). The introduction of the 2-Ada group is expected to significantly decrease the ΔG_r(A → C) value, yet this is not borne out in the computational data for complexes 33–38. This data further motivated us to experimentally investigate the initiation kinetics of 33–38 in order to shed light on to the scope and limitations of the current model.

Table 1.

Computed Ru–O Bond Strengths and Predicted Initiation Rates.

					Ru–O Bond Strength (kcal•mol−1)		Predicted Observed Initiation Rate ln(10−4 s−1)
entry	cat.	Ar =	R1 =	R2 =	ΔGr(A→B)	ΔGr(A→C)	ln(kobs)
1	5	Mes	i-Pr	H	12.8	6.3	−0.92 (expt.)
2	16	Mes	Et	H	13.0	6.9	−2.07
3	17	Mes	n-Pr	H	12.3	5.2	0.18
4	18	Mes	t-Bu	H	9.6	5.5	−0.28
5	19	Mes	i-Bu	H	11.4	5.3	0.02
6	20	Mes	Bn	H	8.9	4.1	1.57
7	21	Mes	PhEt	H	11.5	6.2	−1.15
8	22	Mes	c-Bu	H	12.9	5.9	−0.78
9	23	Mes	c-Pentyl	H	12.8	6.0	−0.91
10	24	Mes	Cy	H	10.2	5.5	−0.17
11	25	Mes	c-Heptyl	H	12.7	6.3	−1.25
12	26	Mes	c-Octyl	H	12.4	5.7	−0.44
13	27	Mes	CHCy2	H	9.9	2.4	3.85
14	28	Mes	CH2-1-Ada	H	10.7	3.4	2.51
15	29	Mes	2-Me-2-Ada	H	7.2	−0.1	7.07
16	30	Mes	CF3	H	6.8	−0.1	7.06
17	31	Mes	CH2CF3	H	9.3	0.7	6.07
18	32	Mes	Me	OMe	–a	5.8	−0.60
19	33	Dipp	i-Pr	H	11.6	5.6	−0.31
20	34	Dipp	1-Ada	H	11.6	4.6	0.92
21	35	Dipp	2-Ada	H	11.3	5.6	−0.31
22	36	o-Tol	i-Pr	H	11.5	5.9	−0.79
23	37	o-Tol	1-Ada	H	9.3	4.9	0.53
24	38	o-Tol	2-Ada	H	9.3	4.7	0.80

^aCalculation did not converge to a stable non-chelated 14 e⁻ complex. See Supporting Information for energy scan plot.

Figure 4.

Computed Ru–O Bond Strengths (ΔG_r(A → C)) for Representative SIMes Catalysts.

Figure 5.

Computed Ru–O Bond Strengths (ΔG_r(A → C)) for Non-SIMes Catalysts.

Synthesis of catalysts.

The synthesis of the proposed catalysts and their precursors followed reported procedures.^{14, 23} 2-Alkoxy-benzaldehyde intermediates were synthesized from 2-fluorobenzaldehyde and the respective alcohols via a three-step S_NAr procedure (Scheme 1). The resulting benzaldehydes were converted to the respective styrenes by Wittig olefination.

Scheme 1.

Synthesis of benzaldehyde intermediates and styrene chelates. ^aFor detailed procedures, see Supporting Information.

Three different synthetic procedures (Methods A–C) were employed for the final step to introduce the chelating benzylidene group onto the complex. A summary of methods and yields for the synthesis of 16–38 (except for commercially available complexes 33 and 36) is shown in Table 2. Most of the proposed catalysts were synthesized in moderate to high yields. Consistent with the observation in our previous study,¹⁴ Method C was found to be more effective in synthesizing the more sterically demanding catalyst 27 (entry 12, Table 2). Notably, attempts to synthesize catalysts 29–31 were unsuccessful, possibly due to weak Ru–O bonding from electronic (30–31) or steric (29) perturbation on the alkoxy group, as predicted by the DFT calculations. This observation is consistent with the unsuccessful attempts by Barbasiewicz and coworkers to obtain 30 by lowering the temperature or terminating the reaction at lower conversion.^15a For catalyst 32, Method B under extended reaction time was effective, whereas Method C was found to be incompatible (entry 17, Table 2).

Table 2.

Complexation attempts to prepare 16–32, 34–35 and 37–38.

entry	Ar =	R1 =	R2 =	cat.	Method	Yield
1	Mes	Et	H	16	A	51%
2	Mes	n-Pr	H	17	A	78%
3	Mes	t-Bu	H	18	A	91%
4	Mes	i-Bu	H	19	A	63%
5	Mes	Bn	H	20	A	58%
6	Mes	PhEt	H	21	A	85%
7	Mes	c-Bu	H	22	A	96%
8	Mes	c-Pentyl	H	23	A	95%
9	Mes	Cy	H	24	A	94%
10	Mes	c-Heptyl	H	25	A	98%
11	Mes	c-Octyl	H	26	A	77%
12	Mes	CHCy2	H	27	A C	8% 47%
13	Mes	CH2-1-Ada	H	28	B	54%
14	Mes	2-Me-2-Ada	H	29	-	0%
15	Mes	CF3	H	30	-	0%
16	Mes	CH2-CF3	H	31	-	0%
17	Mes	Me	OMe	32	A Bb	42% 55%
18	Dipp	1-Ada	H	34	B	64%
19	Dipp	2-Ada	H	35	B	42%
20	o-Tol	1-Ada	H	37	B	40%
21	o-Tol	2-Ada	H	38	B	33%

^aMethod A: 2 (0.2 mmol), styrene (0.2 mmol), and CuCl (0.2 mmol); Method B: 2/3/4 (0.2 mmol), styrene (0.4 mmol), and Amberlyst-15 (0.8 mmol); Method C: 39 (0.2 mmol), styrene (0.2 mmol), and Amberlyst-15 (0.8 mmol)

^bExtended reaction time (6 h). For detailed procedures, see Supporting Information.

X-ray Crystal Structures.

Single crystals suitable for X-ray diffraction were grown and analyzed for all newly synthesized catalysts.²⁴ Graphical representations of the single crystal structures are summarized in Figure 6, confirming the expected connectivity of the catalysts. The series includes commercially available catalyst 33 (the structure of which had not been previously reported). The X-ray structures of 32 and 35 have been previously reported.^{18a, 20} Catalysts 17–18, 21, 24–25, 26 and 32 crystallized in the space group P2₁/n, while catalysts 20, 28 and 34 crystallized in the space group P2₁/c. Catalysts 22, 33 and 35 crystallized in the space group $\stackrel{-}{\text{P}1}$, and catalysts 16 and 23 crystallized in the space group C2/c. Catalyst 19 crystallized in the space group Pbcn. Catalysts 27 and 38 crystallized in the space group Pbca. Catalyst 37 crystallized in the space group P2₁2₁2₁. Two crystallographically inequivalent conformers were found in the unit cell for catalysts 18, 22, 26 and 33, and in these cases the quoted bond lengths represent the average of the values of the two conformers. In general, there is no apparent correlation between the Ru–O bond length and the size of the alkoxy substituent; the Ru–O and Ru=C bond lengths (Table 3) are observed to be fairly constant for most of the catalysts in the series, with the exception of 27 (entry 12, Table 3), 35 (entry 17, Table 3), and 38 (entry 20, Table 3), whose Ru–O bonds are exceptionally long. Similar to the X-ray structure for catalyst 15,¹⁴ elongation of the Ru–O bonds serves to accommodate the energetically unfavorable steric clash between the SIMes-dichlororuthenium fragment and a methylene unit on the 2-Ada group (35 and 38) or the dicyclohexyl group (27).

Figure 6.

Illustration of X-ray structures of the synthesized catalysts.

Table 3.

Summary of experimental Ru–O and Ru=C bond distances for catalysts 16–28 and 32–38.

					Expt. (X-ray, Å)a
entry	cat.	R1 =	R2 =	R3 =	Ru–O	Ru=C
1	16	Mes	Et	H	2.254	1.818
2	17	Mes	n-Pr	H	2.226	1.832
3b	18	Mes	t-Bu	H	2.277	1.828
4	19	Mes	i-Bu	H	2.239	1.832
5	20	Mes	Bn	H	2.264	1.831
6	21	Mes	PhEt	H	2.229	1.828
7b	22	Mes	c-Bu	H	2.262	1.836
8	23	Mes	c-Pentyl	H	2.238	1.833
9	24	Mes	Cy	H	2.268	1.836
10	25	Mes	c-Heptyl	H	2.254	1.830
11b	26	Mes	c-Octyl	H	2.262	1.842
12	27	Mes	CHCy2	H	2.394	1.821
13	28	Mes	CH2-1-Ada	H	2.265	1.828
14	32	Mes	Methyl	MeO	2.218	1.839
15b	33	Dipp	i-Propyl	H	2.242	1.832
16	34	Dipp	1-Ada	H	2.249	1.828
17	35	Dipp	2-Ada	H	2.323	1.829
18c	36	o-Tol	i-Propyl	H	2.298	1.833
19	37	o-Tol	1-Ada	H	2.268	1.827
20	38	o-Tol	2-Ada	H	2.338	1.828

^aExperimental values are shown to three decimal places without estimated standard deviations for clarity. Estimated standard errors are typically in the range of 0.001–0.005 Å. See Supporting Information for additional details.

^bExperimental value represents the average of the two crystallographically inequivalent molecules found in the unit cell.

^cRef. 20.

NMR Spectra.

The series of newly synthesized catalysts was next studied by ¹H and ¹³C NMR spectroscopy in CD₂Cl₂. The characteristic benzylidene peaks of the catalysts were observed to be in the 16–17 ppm (¹H) and 280–290 ppm (¹³C) ranges. Detailed results are summarized in Table 4. Chemical shifts of the carbenes on the NHC ligands, which are typically observed at 210 ppm (¹³C), are also compiled in Table 4. There seems to be no definitive trend between the NMR shifts of the benzylidene moiety and the size of the alkoxy group.

Table 4.

Diagnostic NMR peaks of the catalysts.

					Ru=CHAr		NHC
entry	cat	Ar =	R1 =	R2 =	1H	13C	13C
1	16	Mes	Et	H	16.52	294.71	211.58
2	17	Mes	n-Pr	H	16.52	295.23	211.53
3	18	Mes	t-Bu	H	16.50	298.91	211.77
4	19	Mes	i-Bu	H	16.57	296.08c	210.81
5b	20	Mes	Bn	H	16.55	290.80	209.83
6	21	Mes	PhEt	H	16.45	294.77	211.04
7	22	Mes	c-Bu	H	16.48	294.46c	211.55
8	23	Mes	c-Pentyl	H	16.52	296.20c	211.73
9	24	Mes	Cy	H	16.51	296.76c	211.49
10	25	Mes	c-Heptyl	H	16.50	296.63c	211.74
11	26	Mes	c-Octyl	H	16.51	296.44	211.21
12	27	Mes	CHCy2	H	16.72	296.60c	208.83
13	28	Mes	CH2-1-Ada	H	16.59	295.58c	210.22
14b	32	Mes	Methyl	OMe	17.25d	286.02c	211.18
15b	33	Dipp	i-Pr	H	16.37	289.60c	213.60
16	34	Dipp	1-Ada	H	16.43	294.83c	214.67
17	35	Dipp	2-Ada	H	16.62	293.18c	212.23
18b	36	o-Tol	i-Propyl	H	16.47	295.21	210.89
19	37	o-Tol	1-Ada	H	16.39	297.91	211.54
20	38	o-Tol	2-Ada	H	16.57	296.26	209.09

^aSpectra in CD₂Cl₂

^bThese values were independently measured and corresponded closely with the published values in the literature: entry 5 (Ref. 15a), entry 14 (Ref. 18a), entry 15 (Ref. 19) and entry 18 (Ref. 20).

^cAverage of two or more peaks corresponding to the benzylidene NMR peak.

^dBased on the major trans isomer.

Initiation Rates by UV/Vis Kinetics Studies.

To test the validity of the predictions made by the computational model, the initiation rates of the synthesized catalysts were measured experimentally by UV/Vis spectrometry. We have previously found that results from this assay are in good agreement with initiation rates of a representative catalytic RCM reaction.¹⁴ Following identical procedures as in our previous report, initiation rates were determined by reacting the catalysts (10⁻⁴ M) with a large excess of butyl vinyl ether (BVE) (30 equiv) at 10 °C under pseudo-first-order conditions. The observed initiation rates were calculated based on the decay of the λ_max peak by UV/Vis spectrometry as the catalysts reacted with BVE. Detailed results are summarized in Table 5, and the relative rate with respect to catalyst 5 (k_rel) was also calculated. A large range of initiation rates were observed throughout the series of catalysts. Taking the i-Pr catalyst as control (entry 1, Table 5), k_rel values of the catalysts span four orders of magnitude, ranging from 0.69 (22, entry 3, Table 5) to 340 (27, entry 15, Table 5). The increase in initiation rate tracks closely with increasing steric bulk at the alkoxy position. In particular, dicyclohexylmethoxy catalyst 27 was found to initiate the fastest in the entire series, consistent with the prediction from the computational model. In fact, 27 initiates faster than the well-known fast-initiating Blechert (8) and Plenio (11) catalysts with BVE at 10 °C.²⁵ The single-crystal X-ray structure of 27 (Figure 6) shows that one of the cyclohexyl rings is positioned in an energetically unfavorable pseudo-axial conformation, thus lowering the energy barrier to access the catalytically active non-chelated 14-electron complex. The unfavorable energy profile of the chelated form of catalyst 27 is also reflected in the low yield of its preparation using method C (47%, entry 12, Table 2), which was typically found to be effective for the chelation of more challenging styrenes.¹⁴ The fast-initiating properties of 27 are also consistent with the exceptionally long Ru–O bond length found in its X-ray structure (2.394 Å, entry 12, Table 3). Consistent with the computed results, the catalysts with larger cycloalkyl groups such as cyclooctyl (26, entry 7, Table 5) and cycloheptyl (25, entry 5, Table 5) rings do not initiate significantly faster than the catalysts with smaller cycloalkyl groups like cyclobutyl (22, entry 3, Table 5) or cyclopentyl (23, entry 9, Table 5). Again, this observation is attributed to the flexible ring structures that have limited contribution to the steric bulk of the alkoxy group and hence insignificant impact on initiation rates. Benzyl catalyst (20, entry 13, Table 5) was found to initiate 10 times faster than the phenethyl analogue (21, entry 6, Table 5), an effect that could be steric or electronic in nature. In comparing the crystal structures of 20 and 21 (Figure 6), one can see that the benzyl group of 20 is situated perpendicular to the mesityl ring on the NHC ligand, whereas the phenyl ring in 21 is parallel to the mesityl ring and pointing away from the SIMes-dichlororuthenium fragment, suggesting potential edge-to-face and face-to-face π-stacking respectively. It should also be noted that the 2,6-diOMe catalyst 32 was observed to initiate only at elevated temperatures (55 °C),¹⁸ possibly due to rapid exchange of the methoxy groups, a phenomenon that merited further investigation.

Table 5.

Initiation rates of catalysts, displayed in increasing order for catalysts bearing the same NHC ligand.^a,b

entry	cat.	Ar =	R1 =	R2 =	λmax	kobs (10−4 s−1)	krel
1	5	Mes	i-Pr	H	378	0.40 ± 0.04	1.0
2	32	Mes	Me	OMe	376	-c	-
3	22	Mes	c-Butyl	H	378	0.28 ± 0.02	0.7
4	16	Mes	Et	H	376	0.40 ± 0.04	1.0
5	25	Mes	c-Heptyl	H	380	0.41 ± 0.03	1.0
6	21	Mes	PhEt	H	376	0.46 ± 0.07	1.1
7	26	Mes	c-Octyl	H	380	0.46 ± 0.01	1.2
8	24	Mes	Cy	H	380	0.48 ± 0.06	1.2
9	23	Mes	c-Pentyl	H	380	0.58 ± 0.04	1.4
10	17	Mes	n-Pr	H	376	0.71 ± 0.02	1.8
11	18	Mes	t-Bu	H	380	1.24 ± 0.04	3.1
12	19	Mes	i-Bu	H	378	1.85 ± 0.06	4.6
13	20	Mes	Bn	H	370	5.07 ± 0.62	13
14	28	Mes	CH2-1-Ada	H	374	5.50 ± 0.37	14
15	27	Mes	CHCy2	H	376	135 ± 4.24	340
16	33	Dipp	i-Pr	H	374	0.04d	0.1
17	34	Dipp	1-Ada	H	378	0.10d	0.3
18	35	Dipp	2-Ada	H	378	2.91 ± 0.09	7.3
19	36	o-Tol	i-Pr	H	374	0.20 ± 0.06	0.5
20	37	o-Tol	1-Ada	H	378	2.48 ± 0.31	6.2
21	38	o-Tol	2-Ada	H	376	75.5 ± 8.94	190

^aThe k_obs values are reported as averages (with 95% confidence intervals) determined from three independent trials.

^bThe relative rate (k_rel) was calculated by dividing the k_obs value of the catalyst of interest by the k_obs value for catalyst 5.

^cCatalyst 32 was found not to initiate at 10 °C.

^dDue to the slow initiation kinetics, the k_obs values were determined from a single trial.

Variation of the NHC ligand also had a significant impact on the initiation rates of the catalysts. Catalysts bearing the SIPr NHC ligand (33–35, entries 16–18, Table 5) were found to initiate much slower than their SIMes counterparts, consistent with the observations reported by Percy and coworkers,²⁶ whereas catalysts with the SITol NHC ligand (36–38, entries 19–21, Table 5) were found to initiate with comparable rates. Notably, across the NHC ligand series, there was a consistent trend of faster initiation rate with increasing steric bulk on the alkoxy group from i-Pr (5, 33, 36) to 1-Ada (14, 34, 37) to 2-Ada groups (15, 35, 38) (Table 6). In particular, fast-initiating catalysts such as 38 and 27 initiate about 2000 and 3600 times faster than 33, respectively. This observation is another demonstration of the synergistic power of tuning the alkoxy group and the NHC ligand together to access catalysts with a broad range of initiation rates.

Table 6.

Effect of NHC ligand variations on initiation rates.^a

Alkoxy╲NHC	SIPr	SIMes	SITol
i-Pr	krel = 1.0 (33)	krel = 11 (5)	krel = 5.5 (36)
1-Ada	krel = 2.6 (34)	krel = 94 (14)	krel = 67 (37)
2-Ada	krel = 79 (35)	krel = 1500 (15)	krel = 2000 (38)

^aThe relative rate (k_rel) was calculated by dividing the k_obs value of the catalyst of interest by the k_obs value for catalyst 33.

Validity of Computational Model in Predicting Experimental Initiation Rates.

Predicted initiation kinetics of the synthesized catalysts are tabulated and compared against the experimental values from the UV/Vis kinetics experiments (Table 7). To better illustrate the comparison and evaluate the accuracy of the initiation-prediction model, scatter plots of the SIMes catalysts series (Figure 7) and SIPr and SITol series (Figure 8) are shown. The experimental results are illustrated as colored points on the scatterplot (SIMes: green stars; SIPr and SITol: red circles), whereas the solid black line represents the predicted trend given by Equation 1. To gauge the deviation of the experimental initiation rates from the predicted trend line, dashed blue lines are plotted around the predicted trend line to show the 95% confidence band derived and extrapolated from the training set (5 and 8–15) in the prediction model.

Table 7.

Comparison between predicted and experimental initiation rates

					Predicted	Experimental
entry	cat.	Ar =	R1 =	R2 =	ln(kobs) ln(10−4 s−1)	ln(kobs) ln(10−4 s−1)
1	22	Mes	c-Butyl	H	−0.78	−1.28
2	16	Mes	Et	H	−2.07	−0.93
3	25	Mes	c-Heptyl	H	−1.25	−0.90
4	21	Mes	PhEt	H	−1.15	−0.78
5	26	Mes	c-Octyl	H	−0.44	−0.77
6	24	Mes	Cy	H	−0.17	−0.73
7	23	Mes	c-Pentyl	H	−0.91	−0.55
8	17	Mes	n-Pr	H	0.14	−0.35
9	18	Mes	t-Bu	H	−0.28	0.22
10	19	Mes	i-Bu	H	0.02	0.62
11	20	Mes	Bn	H	1.57	1.62
12	28	Mes	CH2-1-Ada	H	2.51	1.70
13	27	Mes	CHCy2	H	3.85	4.91
14	33	Dipp	i-Pr	H	−0.31	−3.30
15	34	Dipp	1-Ada	H	0.92	−2.32
16	35	Dipp	2-Ada	H	−0.31	1.07
17	36	o-Tol	i-Pr	H	−0.79	−1.59
18	37	o-Tol	1-Ada	H	0.53	0.91
19	38	o-Tol	2-Ada	H	0.80	4.32

Figure 7.

Validation of the initiation-prediction model results (SIMes catalysts).

Figure 8.

Validation of the initiation-prediction model results (SIPr and SITol catalysts).

As shown in Figure 7, all of the experimental initiation rate values fall within the 95% confidence prediction band of the prediction model. This indicates that the model provides a reasonably accurate prediction of the initiation kinetics for the newly synthesized catalysts. Notably, catalysts 21 (phenethyl, entry 4, Table 7) and 25 (c-heptyl, entry 3, Table 7) are predicted to have very similar initiation rates despite their significantly different steric and electronic properties, and this prediction is borne out experimentally. Considering the fact that the initiation-prediction model is derived from a single-variable linear regression with an R–squared value of 0.79 (Equation 1), the new experimental results (R² = 0.86 with respect to the solid prediction trendline) are in well accordance with the prediction model.

However, for the catalysts bearing different NHC ligands, as illustrated in Figure 8, the initiation-prediction model shows poor accuracy as illustrated by the large deviation of the experimental values from the predicted trend line (R² = 0.16). Many contradictory results are observed (Table 7). In particular, catalysts 33 and 35 are predicted to have essentially the same initiation rates, yet the experimental results show that 35 initiates about 80 times faster. Similarly, catalyst 34 is predicted to initiate faster than 38, but the experimental results show that 38 is actually about 760 times faster. This suggests that the current initiation-prediction model, which was developed based on catalysts bearing the SIMes NHC ligand, cannot be extended to catalyst bearing the SIPr and SITol NHC ligands. Future directions include developing a new prediction model (or collection of models) to accommodate different NHC ligands. Moreover, although catalysts 29–31 are too unstable to be synthesized, their computed Ru–O bond strengths can be valuable in identifying the limits of Ru–O chelation in the current initiation-prediction model. This boundary condition can serve as a convenient indication of structural stability in the future design of fast-initiating catalysts.

In our previous report,¹⁴ the computed Ru–O bond strength from the dechelation process (ΔG_r(A → B)) was found to be a poor predictor for catalyst initiation kinetics. This relationship was reevaluated with the new data from the catalysts synthesized in this study. Again, the computed Ru–O bond strength from ΔG_r(A → B) shows weak correlation with the observed initiation rates (R² = 0.50, Figure 9), recapitulating that the use of ΔG_r(A → C) is superior to using ΔG_r(A → B) in Ru–O bond strength estimation for initiation kinetics studies.²⁷

Figure 9.

ln(k_obs) versus Ru–O bond strength ΔG_r(A → B))

Chelate Dissociation Rate versus Initiation Rate.

It is especially interesting to investigate the initiation kinetics of the 2,6-diOMe catalyst 32 because the 2,6-dimethoxy group on the benzylidene chelate provides an opportunity to study the chelate dissociation rate of Grubbs–Hoveyda olefin metathesis catalysts.

Even though catalyst 32 is predicted to have similar initiation kinetics as catalyst 22 or 26, it did not initiate at 10 °C like the rest of the catalysts in the series (entry 2, Table 5). Nevertheless, it was previously shown to catalyze ring-closing metathesis (RCM) reaction at elevated temperature.¹⁸ We found that even at higher temperature (50–100 °C), the initiation rate of catalyst 32 is too low to obtain an accurate measurement at the concentration typically used for UV/Vis experiments. As such, reaction kinetics were measured by ¹H NMR spectroscopy ([Ru]₀ = 5 mM) by subjecting 32 to a large excess of BVE (30 equiv) at 70 °C in d₈-toluene.^{12a, 12b} The initiation rate profile of 32 (Figure 10) is obtained by measuring the decay of the benzylidene ¹H NMR peak (17.27 ppm in toluene-d₈), and the results are in good accordance with the assumption of pseudo-first-order decay. As the methoxy group in 32 presents less steric bulk and the incoming BVE is relatively electron-rich and sterically non-demanding, the initiation mode of 32 is expected to lean towards the interchange mechanism where the initiation kinetics is dominated by the interchange rate k_I, which is approximated by k_obs/[BVE].^13b

Figure 10.

Initiation kinetics of catalyst 32 approximately over the course of two half-lives of the initiation reaction.

To gauge the dissociation rate of the Ru–O bond, a series of ¹H–¹H EXSY NMR experiments were performed with 32 using pulse sequences with different mixing times to observe the exchange of the two methoxy groups.²⁸ A representative spectrum is shown in Figure 11. The EXSY spectra allowed for direct observation of methoxy group exchange in 32 at 70 °C in deuterated toluene, and the quantitative rate data for the exchange process was thus obtained.

Figure 11.

Representative ¹H–¹H EXSY spectrum for OMe exchange in 32 (t_mix = 500 ms)

It is instructive to compare the rate of methoxy group exchange in 32 against its initiation kinetics. At 70 °C in deuterated toluene, the methoxy groups exchange rate in 32 is about 3300 times faster than the catalyst’s interchange rate with the incoming olefin during the initiation process (Table 8).²⁹ The rate of Ru–O dissociation is expected to be even higher since the methoxy groups do not necessarily exchange every time the Ru–O bond dissociates. These results shed light on the initiation mechanism of 32. In particular, the observation that Ru–O dissociation is rapid compared to initiation is inconsistent with a purely dissociative initiation pathway with 32. Rather, interchange or associative mechanisms are more likely in this case.

Table 8.

Comparison of OMe exchange and interchange rate of 32

process	k (10−3 s−1)	T (°C)	rel. rate
OMe exchange	450	70	3300
Interchange with BVE	0.14	70	1

CONCLUSION

The initiation rates of a series of ruthenium-based olefin metathesis catalysts bearing modified ortho-alkoxy benzylidenes have been predicted by a previously developed computational model. This series of catalysts was investigated using a combination of organometallic synthesis, NMR spectroscopy, X-ray crystallography and initiation kinetics. The computational model satisfactorily predicts the initiation kinetics of the catalysts bearing the SIMes NHC ligand. A total of fourteen new second-generation Grubbs–Hoveyda catalysts have been synthesized and characterized. Among them, the dicyclohexylmethoxy catalyst 27 was found to be one of the fastest-initiating second-generation Hoveyda–Grubbs olefin metathesis catalysts reported to date. The enhanced initiation rate of 27 was successfully predicted by the computational model. A notable aspect of this model is that it does not make any assumptions about the mechanism of catalyst initiation and relies on a single, easily computed thermodynamic metric. This metric gives simple and generally reliable predictions of olefin metathesis catalyst properties in silico. Further improvements of the computational model, such as extensions to other NHC ligands, will ultimately provide a more powerful toolkit for the design of customized catalysts with the desired initiation kinetics.