Iron-Catalyzed Vinylsilane Dimerization and Cross-Cycloadditions with 1,3-Dienes: Probing the Origins of Chemo- and Regioselectivity

Abstract

The selective, intermolecular, homodimerization and cross-cycloaddition of vinylsilanes with unbiased 1,3-dienes, catalyzed by a pyridine-2,6-diimine (PDI) iron complex is described. In the absence of a diene coupling partner, vinylsilane hydroalkenylation products were obtained chemoselectively with unusual head-to-head regioselectivity (up to >98% purity, 98:2 E/Z). In the presence of a 4- or 2-substituted diene coupling partner, under otherwise identical reaction conditions, formation of value-added [2+2]- and [4+2]-cycloadducts, respectively, was observed. The chemoselectivity profile was distinct from that observed for analogous α-olefin dimerization and cross-reactions with 1,3-dienes. Mechanistic studies conducted with well-defined, single-component precatalysts (^MePDI)Fe(L₂) (where ^MePDI = 2,6-(2,6-Me₂-C₆H₃N═CMe)₂C₅H₃N; L₂ = butadiene or 2(N₂)) provided insights into the kinetic and thermodynamic factors contributing to the substrate-controlled regioselectivity for both the homodimerization and cross cycloadditions. Diamagnetic iron diene and paramagnetic iron olefin complexes were identified as catalyst resting states, were characterized by in situ NMR and Mössbauer spectroscopic studies, and were corroborated with DFT calculations. Stoichiometric reactions and computational models provided evidence for a common mechanistic regime where competing steric and orbital-symmetry requirements dictate the regioselectivity of oxidative cyclization. Although distinct chemoselectivity profiles were observed in cross-cycloadditions with the vinylsilane congeners of α-olefins, these products arose from metallacycles with the same connectivity. The silyl substituents ultimately governed the relative rates of β-H elimination and C–C reductive elimination to dictate final product formation.

Graphical Abstract

Introduction

Metallacycle formation has emerged as a versatile strategy for controlling chemo- and site-selectivity in metal-mediated and metal-catalyzed transformations.¹ Oxidative cyclization of multiple unsaturated coupling partners at a metal center enables access to cyclic and/or linear products following reductive elimination, β-elimination, or cycloreversion.² For example, the exceptional chemoselectivity in the chromium-catalyzed trimerization of ethylene to 1-hexene, which is practiced industrially on kilo tons per annum (kta) scales,³ arises from selective β-elimination from a chromacycloheptane.⁴ The utility of this atom-economical approach for generating structural complexity in a single synthetic step has also motivated its application with electronically biased coupling partners to generate hetero- and carbocycles of interest for small-molecule synthesis (e.g. pyrones, pyridones, pyridines, benzene derivatives).^1b,5 These formal cycloaddition products are generally distinct from those that can be accessed through thermal, photochemical, or Lewis acid catalyzed processes alone. Additionally, polarized substrates offer a direct handle for tuning regioselectivity.^1b,5 However, electronically neutral or unbiased substrates such as alkene, diene, and alkyne feedstocks pose challenges for control of chemo- and especially regioselectivity.⁶

Our group and others have recently introduced iron,^6e,7 cobalt,⁸ and nickel⁹-catalyzed processes for the [2+2]-cyclodimerization of unsaturated hydrocarbons to generate cyclobutane or cyclobutene products. These rigid carbocycles are scaffolds of interest for the synthesis of small-molecule targets ranging from pharmaceuticals and fragrances to fuels and materials.¹⁰ Mechanistic studies of both intra- and intermolecular olefin cyclodimerization have provided support for a mechanism involving oxidative cyclization to form a substituted metallacyclopentane prior to C–C bond-forming reductive elimination (Scheme 1A).¹¹ With related catalysts derived from these Earth-abundant metals, metallacyclic intermediates have been observed and/or implicated in cross-[2+2] reactions between alkenes and alkynes,^6d,12 enynes,^6c,6d,13 allenes,¹⁴ and conjugated dienes¹⁵ (Scheme 1B) as well as in hydroalkenylations of alkenes and conjugated dienes.^16–17

Scheme 1. Contrasting Reactivity of α-olefins and Vinylsilanes in Iron- and Cobalt-Catalyzed Cycloaddition Reactions.

^a Ref. 6e; Ar¹ = 2,6-(c-C₅H₉)₂-C₆H₃ ^b Ref. 6e; Ar² = 2-^tBu-C₆H₄ ^c Refs. 6d, 13 ^d Refs. 6e, 15. [Si] = silyl

As part of an ongoing program toward the development of non-canonical cycloaddition reactions to upgrade abundant hydrocarbons, we have sought to manipulate the reactivity of iron metallacycles to achieve bond constructions with unique regioselectivity patterns using unactivated substrates.^6e,6g,15,17e As part of these efforts, exposure of vinyltrimethylsilane to standard conditions for iron-catalyzed α-olefin cyclodimerization resulted in the formation of the linear head-to-head dimer rather than the corresponding cyclobutane or more typically observed tail-to-tail products (Scheme 1A). This anomalous product is consistent with β-H elimination and C–H reductive elimination from a metallacyclopentane intermediate with both SiMe₃ groups proximal (α) to the iron center. Oxidative cyclizations and migratory insertions of alkynyl silanes have previously been demonstrated to favor placement of the silyl substituent α to early transition metals in stoichiometric and catalytic reactions.¹⁸ This “α-[Si] effect” has, in turn, been attributed to orbital symmetry requirements, electrostatic polarization, and the possibility of C–Si agostic interactions (Scheme S1, see Supporting Information).^18–21 While similar preferences have not been typically observed with mid to late transition metals, this result suggested that pyridine-2,6-diimine (PDI) supported iron complexes may exhibit a kinetic and/or thermodynamic preference for α-silicon incorporation. Given the availability of varied vinylsilanes, which are employed as (co)monomers in α-olefin polymerizations,²² as well as the versatility of organosilicon groups toward further synthetic manipulation, evaluation of their potential to access unique regioselectivity patterns through iron-catalyzed cycloaddition reactions was explored.

Here we describe the development of three [(PDI)Fe]-catalyzed methods for the chemo- and regioselective elaboration of vinylsilanes: (i) head-to-head vinylsilane dimerization, (ii) cross-[2+2]-cycloaddition with 4-substituted 1,3-dienes, and (iii) cross-[4+2]-cycloaddition with 2-substituted 1,3-dienes (Scheme 1). Each case enabled access to unique scaffolds, including relatively unactivated Diels–Alder products that are inaccessible with conventional methods,²³ with potential for strategic applications in the silicones industry as well as in small-molecule synthesis. Mechanistic studies including stoichiometric reactions, kinetic measurements, in situ Mössbauer spectroscopy, and DFT analyses provide support for pathways in which competing steric and orbital-symmetry requirements between the vinylsilane and 1,3-diene coupling partners coupled with β-[Si]-suppressed β-H elimination give rise to the unusual selectivity profiles.

RESULTS AND DISCUSSION

Head-to-Head Vinylsilane Dimerization.

Our study began with re-examination of the iron-catalyzed homocoupling of vinylsilanes using [(^MePDI)Fe(N₂)]₂(μ₂-N₂) (where ^MePDI = 2,6-(2,6-Me₂-C₆H₃N═CMe)₂C₅H₃N) as the precatalyst. As reported previously, exposure of a benzene-d₆ solution of the iron precatalyst (2.5 mol% dimer, 5 mol% [Fe]) to an excess of propylene (1a) resulted in near complete (96%) consumption of 1a within 48 hours at ambient temperature (~23 °C) and formation of a mixture of dimerization products including [2+2]-cycloadduct 2a and the tail-to-tail dimer 3a (Scheme 2A).^6e Neither the corresponding head-to-head dimer (4a) nor head-to-tail dimer 5a was detected by NMR spectroscopy. By contrast, exposure of vinyltrimethylsilane (6a) to the same reaction conditions resulted in negligible conversion. Conducting the reaction in neat vinylsilane produced (32% yield after 48 h) but clean formation of the head-to-head dimer 9a; [2+2]-cycloadduct 7a, tail-to-tail dimer 8a, and head-to-tail dimer 10a were not observed (Scheme 2B). Control reactions conducted without iron or the PDI ligand resulted in no conversion or the formation of an intractable product mixture, respectively. It is useful at this point to explain the labeling scheme used throughout the manuscript. Each of the products arising either from hydrovinylation or cycloaddition is labeled with a number referring to the carbon skeleton followed by a letter signaling the vinylsilane from which it was derived.

Scheme 2. Inverted Regioselectivity for Vinylsilane Dimerization.

^a Ref. 6e. ^b rsm = remaining starting material ^c Neither 7a nor any other regio- or stereoisomeric [2+2] cycloaddition products were observed.

Homocoupling of various vinylsilanes and vinylsiloxanes was examined, and clean conversion to the head-to-head dimer (9) was observed in all cases (Scheme 3). The chemoselectivity was insensitive to both steric and electronic variation of the silyl group and preferential formation products with (E)-olefin geometry was uniform. The low yield (32%) for 9c was a likely result of the dilute reaction conditions. Notably, 1,3-divinyltetramethyldisiloxane (6g), a commodity chemical used widely as a crosslinking agent in the silicones industry,^22a,24 formed C–C bonds from both termini to afford macrocyclic product 9g, which exhibits the same head-to-head orientation within each repeat unit. The macrocycle was assigned as the 21-membered cyclic trimer on the basis of the parent mass observed by atmospheric pressure chemical ionization (APCI) mass-spectrometric analysis under low fragmentation conditions and corroborated by diffusion ordered NMR spectroscopy (DOSY) in chloroform-d. This viscous, slippery oil bears resemblance to the silicone surfactants in consumer products such as soaps and cosmetics.²⁴

Scheme 3. Substrate Scope for Iron-Catalyzed Vinylsilane Dimerization.^a.

^a Reactions conducted on 0.5 mmol scale. Isolated yields reported (with selectivity for the head-to-head hydroalkenylation product in parentheses). ^b Reaction conducted in toluene.

Despite the general efficacy of the vinylsilane coupling, the homologous allyltrimethylsilane (1b) underwent isomerization to trimethyl(prop-1-en-1-yl)silane, rather than homocoupling, upon exposure to [(^MePDI)Fe(N₂)]₂(μ₂-N₂). Furthermore, t-butyl-ethylene (1c) and 3,3,3-trifluoropropylene (1d), which were examined as steric mimics of 6a, failed to undergo coupling reactions upon exposure to [(^MePDI)Fe(N₂)]₂(μ₂-N₂), even at elevated temperature (50 °C), at which point decomposition of the iron precatalyst and elemental iron deposition was observed. Given that these substrate-based probes did not prove mechanistically informative, additional experiments were conducted to elucidate the basis for the chemo- and regioselectivity of the (PDI)Fe-catalyzed dimerization of vinylsilanes.

A series of stoichiometric reactions was undertaken. Addition of trimethylvinylsilane (6a), dimethylphenylvinylsilane (6b), or diethoxymethylvinylsilane (6d) to a solution of [(^MePDI)Fe(N₂)]₂(μ₂-N₂) (2 equiv of vinylsilane relative to Fe) in benzene-d₆ resulted in an immediate color change from red-brown to yellow-brown and afforded the corresponding iron bis(vinylsilane) complex in >98% conversion within 15 minutes at ambient temperature (Scheme 4). In each case, the ¹H NMR spectrum was consistent with that expected for a paramagnetic, effectively C_2v-symmetric complex with paramagnetically shifted resonances ranging in chemical shift from δ 170 to −110 ppm. A representative, solution-state magnetic moment of 2.8 μ_B was measured for (^MePDI)Fe(η²-6a)₂ at 23 °C (benzene-d₆, Evans method), consistent with the spin-only value for two unpaired electrons and an S = 1 ground state. However, the bis(vinylsilane) complexes proved unstable to vacuum, and attempts to isolate or crystallize them, even in the presence of excess vinylsilane, resulted in re-isolation of [(^MePDI)Fe(N₂)]₂(μ₂-N₂). Based on these results and prior studies with α-olefins,¹¹ it is likely that (^MePDI)Fe(η²-6a)₂, (^MePDI)Fe(η²-6a), and [(^MePDI)Fe(N₂)]₂(μ₂-N₂) are in equilibrium under catalytic conditions. A zero-field ⁵⁷Fe Mössbauer spectrum (80 K, toluene glass) was collected following exposure of [(^MePDI)Fe(N₂)]₂(μ₂-N₂) (0.5 equiv) to excess trimethylvinylsilane (6a, 10 equiv) in toluene at ambient temperature to generate (^MePDI)Fe(6a)₂ in situ (Figure 1). The spectrum was fit with two quadrupole doublets, where the major component exhibited parameters (92%: δ = 0.66 mm/s, |ΔE| = 1.04 mm/s) in accord with those reported previously for isolable (κ²-PDI)Fe olefin complexes assigned as the Fe(I) oxidation state.¹¹ These parameters also agreed with those predicted for (κ²-^MePDI)Fe(η²-6a)₂ using full-molecule density functional theory (δ = 0.73 mm/s, |ΔE| = 1.72 mm/s; Figure 2).^{25, 26}

Scheme 4.

Synthesis and Hydrogenolysis of (^MePDI)Fe(vinylsilane)₂ complexes.

Figure 1.

Solution, zero-field ⁵⁷Fe Mössbauer spectrum (toluene glass, 80 K) of (^MePDI)Fe(6a)₂ generated in situ. Simulated spectra plotted for the major (92%, blue) and minor (12%, red) components with the listed parameters, along with the net spectrum (grey).

Figure 2.

(A) Lowest-energy optimized structure and (B) spin-density plot for (^MePDI)Fe(6a)₂, computed at the B3LYP level of density functional theory with the ZORA-def2-TZVP(-f) basis set augmented by SARC/J terms for Fe, N, and C atoms in the primary coordination sphere, and the ZORA-def2-SVP basis set augmented with SARC/J terms for all other atoms. Hydrogen atoms omitted for clarity. H = white, C = gray, N = blue, Fe = tan, Fe = red-orange, positive spin density = red, negative spin density = yellow.

Given the challenges associated with isolation of the iron vinylsilane complexes, alternative strategies were pursued to establish their connectivity and reactivity. Hydrogenolysis of (^MePDI)Fe(η²-6b)₂ furnished the head-to-head coupled alkane 12b as the exclusive organic product (Scheme 4B). However, exposure of (^MePDI)Fe(η²-6a)₂ and (^MePDI)Fe(η²-6d)₂ to 4 atm of H₂ instead resulted in direct hydrogenation to yield alkylsilanes 11a and 11d, respectively. While both (a) direct hydrogenolysis of a transient metallacycle (Scheme 4B) and (b) hydrogen-induced alkene dimerization are precedented, the divergent outcomes observed with (^MePDI)Fe(η²-6b)₂ versus (^MePDI)Fe(η²-6a)₂ and (^MePDI)Fe(η²-6d)₂ suggested that a metal hydride was most likely not responsible for the observed catalytic dimerization.²⁷ Regardless, the combined results indicated that the (PDI)-supported iron complex exhibited a pronounced and consistent α-Si preference in the C–C bond-forming step.

Cross-[2+2]-Cycloadditions with 4-Substituted-1,3-Dienes.

Given the evidence for an α-[Si] preference in metallocycle formation and ultimately vinylsilane dimerization, we were motivated to explore whether a similar effect could be applied to alter the regioselectivity of other (PDI)Fe-catalyzed coupling reactions. [(^MePDI)Fe(N₂)]₂(μ₂-N₂) serves as an active and selective precatalyst for the cross-[2+2]-cycloaddition of (E)-piperylene (13a) with ethylene and α-olefins, yielding substituted propenylcyclobutanes (14; Scheme 5A).^15b The 1,3-cyclobutane substitution pattern has been attributed to the combination of maximizing orbital overlap (d_z²→π*)¹⁹ and avoiding steric encumbrance from geminal substituents during selectivity-determining olefin coordination or oxidative cyclization. Although silylalkynes have been reported to exhibit inverted π* polarization relative to their aliphatic counterparts (vide supra), NBO analyses of the orbital coefficients for representative vinylsilanes (6a–d,f) revealed only negligible polarization, comparable in magnitude and sign to that of α-olefins (see Supporting Information).²⁸ Accordingly, vinylsilanes proved to be excellent substrates for (PDI)Fe-catalyzed cross-[2+2]-cycloadditions with (E)-piperylene (Scheme 5B). No change in regioselectivity was observed relative to the pure hydrocarbon case. While the diastereoselectivity proved modest, high chemoselectivity for the [2+2]-products (17aa–ga) and exclusive reaction with the (E)-diene isomers was observed, even when using (E/Z) substrate mixtures across a variety of vinylsilane and 4-substituted diene coupling partners (Scheme 6). For clarity, the cross-cycloaddition products are each labeled with a number referring to the carbon skeleton and two letters indicating the vinylsilane (first letter) and diene (second letter) from which they were derived.

Scheme 5. Conserved Chemoselectivity for Cross-Cycloadditions with 4-Substituted-1,3-Dienes.

^a Conducted on 1.0 mmol scale with 1.25 mol% [(^Me(Et)PDI)Fe(N₂)]₂(μ₂-N₂). Yield (and selectivity for [2+2]-cycloadduct 14aa) determined by integration of diagnostic ¹H NMR signals relative to an internal standard. See Ref. 15b. ^b Conducted on 1.0 mmol scale with 2.5 mol% [(^MePDI)Fe(N₂)]₂(μ₂-N₂). Isolated yield reported (with selectivity for the [2+2]-cycloadduct 17aa product in parentheses).

Scheme 6. Substrate Scope for Iron-Catalyzed Cross-[2+2]-Cycloaddition of Vinylsilanes with 4-Substituted-1,3-Dienes.^a.

^a Reactions conducted on 1.0 mmol scale unless indicated otherwise. Isolated yields reported (with selectivity for the [2+2]-cycloadduct 17aa product in parentheses). ^b Reaction conducted in toluene. ^c With 5 mol% [Fe] and 2.4 equiv 13a relative to 6g. ^d Mass balance (14%) comprised of the [2+2] cycloadduct where only one of the two vinyl groups of 6g reacted (60:40 d.r.) ^e With 1 mol% (^MePDI)Fe(butadiene) ^f Mass balance comprised of the adduct resulting from thermal [4+2]-cyclodimerization of 13b ^f With 2.5 mol% (^MePDI)Fe(butadiene)

Cross-[4+2]-Cycloadditions with 2-Substituted-1,3-Dienes.

Despite the high chemoselectivity observed in (PDI)Fe-catalyzed cross-[2+2]-cycloadditions with piperylene and other 4-substituted dienes, the method has previously been shown to be highly sensitive to the position of the diene substituent. For example, exposing precatalyst [(^MePDI)Fe(N₂)]₂(μ₂-N₂) to an equimolar mixture of isoprene (13i) and ethylene (1e) afforded the hydrovinylation product 16ie (instead of [2+2] adduct 14ie) as the exclusive organic product.^15a Subjecting representative α-olefin/myrcene (13j) combinations to analogous conditions generated mixtures of hydroalkenylation products 16aj–dg along with [2+2]-cycloadducts 14aj–dg (Scheme 7A). A more specialized ligand (^iPr(TB)PDI) has also been identified to achieve high [2+2]-chemoselectivity in these cases.^6e

Scheme 7. Inverted Chemoselectivity for Cross-Cycloadditions of Vinylsilanes with 2-Substituted-1,3-Dienes.^a.

^a Reactions conducted on 0.25 or 0.5 mmol scale. Combined isolated yields reported (with selectivity for each product listed in parentheses). ^b Reaction conducted in C₆D₆ using isoprene (13i) in place of myrcene (13j) in C₆D₆. Yield (and selectivity for hydrovinylation product 16ei) determined by integration of diagnostic ¹H NMR signals relative to an internal standard. ^c With 5 mol% [Fe] ^d Reactions conducted on 0.25 mmol scale. Conversion and product ratio determined after 24 h by gas chromatographic analysis relative to mesitylene, which was added as an internal standard at the end of the reaction. ^e 5 mol% Mg(butadiene)•2THF added as an in situ reductant. ^f Reaction performed on a 1.9 mmol scale at 200 °C for 24 h with 5 mg BHT. ^g Complex mixture of decomposition products. ^h Reactions conducted on 1.0 mmol scale. Combined isolated yields reported (with selectivity for each product listed in parentheses).

Like its α-olefin counterparts, vinyltrimethylsilane proved to be a competent (albeit sluggish) substrate for (^iPr(TB)PDI)Fe-catalyzed cross-[2+2]-cycloaddition with myrcene (Scheme 7B, entry 1). However, we were intrigued as to whether vinylsilane–myrcene coupling would exhibit an analogous selectivity profile using the more readily accessed (^MePDI) ligand. Remarkably, exposure of an equimolar mixture of vinyltrimethylsilane (6a) and myrcene (13j) to [(^MePDI)Fe(N₂)]₂(μ₂-N₂) yielded [4+2]-cycloadduct 18aj rather than corresponding [2+2]-product 17aj (Scheme 7B, entry 2). Ligands with larger 2,6-aryl substituents, e.g. (^iPrPDI)Fe(N₂)]₂(μ₂-N₂), resulted in lower levels of chemoselectivty. Butadiene complex (^MePDI)Fe(η⁴-butadiene) afforded 18aj with effectively identical selectivity to [(^MePDI)Fe(N₂)]₂(μ₂-N₂) (Scheme 7B, entry 3) and exhibited improved shelf-stability. The two precatalysts were employed interchangeably in subsequent studies to extend this finding. With either precatalyst, the preference for [4+2]-cycloaddition with vinyltrimethylsilane was consistent with a variety of 2-aryl- and 2-alkyl-substituted 1,3-dienes (Scheme 7C). However, neither polar nor disubstituted dienes (13q, 20–23) afforded any cross products under analogous conditions (even upon heating to 50 °C); trace amounts of vinylsilane dimerization product 9a were instead detected after 24 hours. These results are remarkable in that they produce formal Diels–Alder products with unactivated substrates that are not kinetically competent for thermal cycloaddition by a pericyclic mechanism (Scheme 7B, entry 8).

Similar reactivity patterns were observed with varied vinylsilane and vinylsiloxane coupling partners, where the [4+2]-cycloadducts with 13j were the major products in all cases examined (Scheme 7C). Exclusive [4+2]-chemoselectivity was observed with alkyl- and aryl-substituted vinylsilanes 6a–c; however, competitive formation of the analogous 1,4-hydroalkenylation products (19dj–gj) was observed with vinylsiloxanes 6d–g. The decreased chemoselectivity observed with vinylsiloxane substrates was hypothesized to arise from increased flexibility of the resulting metallacycle (vide infra). However, in control experiments, hydroalkenylation product 16aj also formed cleanly upon exposure to a combination of FeCl₂ and Mg(butadiene)•2THF, even in the absence of added ligand (Scheme 7B, entry 6).²⁹ Although all substrates were dried over CaH₂, degassed, and filtered through alumina prior to use, the presence of trace impurities in the vinylsiloxanes that could degrade the (^MePDI)Fe(L)₂ precatalyst and result in a competitive “ligand-free” background reaction cannot be excluded.

Potential Mechanistic Bases for [4+2] vs. [2+2] Selectivity.

To inform the development of new synthetic methods and strategies for catalyst-control, the basis for the unusual chemoselectivity observed for the [(PDI)Fe]-catalyzed cycloaddition with the specific combination of vinylsilane and 2-substituted-1,3-diene substrates was of interest. In comparison to the mechanisms described previously to account for [(PDI)Fe]-catalyzed cross-reactions with α-olefins (Scheme 8, Mechanistic Hypothesis 0), three distinct possibilities were considered (Scheme 8), as follows:

Scheme 8.

Mechanistic Hypotheses Accounting for Divergent Chemoselectivity of Vinylsilane/Diene Cycloadditions.

Mechanistic Hypothesis A:

As outlined for the formation of 14 from α-olefin/diene cycloaddition, the [(^MePDI)Fe]-catalyzed vinylsilane/diene cycloaddition proceeds through metallacyclopentane A-iv to afford [2+2]-product 17 regardless of diene substitution pattern. However, only the [2+2]-cycloadducts arising from 2-substitued dienes and vinylsilanes undergo rapid thermal or Lewis-acid promoted rearrangement to 18 on the timescale of the cycloaddition.

Mechanistic Hypothesis B:

As outlined for the formation of 14 and 16 from α-olefin/diene coupling, the [(^MePDI)Fe]-catalyzed vinylsilane/diene cycloadditions proceed through metallacycles B-iv and B-vii with [Si] substituents distal (β) to [Fe]. Oxidative cyclization from the s-trans or s-cis diene isomer is dictated predominantly by the diene substitution pattern. 4-Subsituted dienes are biased to react through an s-trans conformation (B-iii), giving rise to metallacyclopentane B-iv. By contrast, 2-substituted dienes may react through an s-cis conformation (B-vi) giving rise to metallacycloheptene B-vii. The flexibility of either resulting metallacycle (with or without β-silyl substituents) then determines the relative rates of C–C reductive elimination giving rise to 17 and 18 versus β-H elimination and C–H reductive elimination giving rise to 19 (where the latter case is not depicted). This mechanistic pathway would require that the α-[Si] effect is negligible in the presence of 1,3-dienes.

Mechanistic Hypothesis C:

If the α-[Si] effect is competitive with the directing effects of 1,3-dienes, oxidative cyclization from vinylsilane complex C-x occurs selectively with [Si] proximal (α) to [Fe]. This inverted regioselectivity is accessible only with 2-substituted dienes and requires reaction through the 1,2-coordianted s-cis diene complexes C-ix and C-x, giving rise to metallacycloheptene C-xi, which is poised for C–C reductive elimination to afford 18. However, reaction through this s-cis conformation is not energetically feasible with 4-substituted dienes, obviating the α-Si effect and forcing reaction through the metallacyclopentane intermediate C-iv en route to 17.

Mechanistic hypotheses B and C may also be modified to describe related scenarios in which olefin coordination or reductive elimination, rather than oxidative cyclization, are selectivity-determining.

Stability of Independently Synthesized 17bi.

To probe the viability of mechanistic hypothesis A, a representative example of the posited intermediate, a [2+2]-cycloadduct, was independently prepared (see Supporting Information). This isopropenylcyclobutane (17bi) was exposed to a battery of conditions to probe for kinetically accessible paths for thermal or Lewis acid-promoted rearrangement (Scheme 9). Notably the cyclobutane was stable to prolonged heating in benzene-d₆ (up to 80 °C), exposure to [(^MePDI)Fe(N₂)]₂(μ₂-N₂) at ambient temperature or 50 °C, or treatment with a variety of Lewis acids (including deliberately decomposed [(^MePDI)Fe(N₂)]₂(μ₂-N₂)). While 17bi decomposed to an intractable mixture of products upon exposure to AlCl₃, formation of the putative rearrangement product, 18bi, was not observed in any case. Taken together, these results eliminate mechanistic hypothesis A.

Scheme 9.

Probing the Viability of Mechanistic Hypothesis A.

Observation of Catalytically Relevant Iron Complexes.

To distinguish between mechanistic hypotheses B and C, the identity of the catalyst resting state and speciation in the context of the iron-catalyzed [4+2] cycloaddition reaction was studied. As reported previously, (^MePDI)Fe(η⁴-isoprene) is formed readily in solution from [(^MePDI)Fe(N₂)]₂(μ₂-N₂) upon addition of isoprene (13i).^15b Similarly, a diamagnetic, C₁ symmetric complex assigned as (^MePDI)Fe(η⁴-13j) was observed within two hours upon treatment of a benzene-d₆ solution of [(^MePDI)Fe(N₂)]₂(μ₂-N₂) with myrcene (13j, 2 equiv per dimer, 1 equiv per [Fe]) at ambient temperature (~23 °C; Scheme 10). This complex proved unstable to vacuum or crystallization under an N₂ atmosphere.

Scheme 10.

Synthesis of (^MePDI)Fe(diene) Complexes.

The zero-field ⁵⁷Fe Mössbauer spectrum (80 K, toluene glass) was recorded following exposure of [(^MePDI)Fe(N₂)]₂(μ₂-N₂) (0.5 equiv) to excess myrcene (13j, 10 equiv) in toluene at ambient temperature. Two partially overlapping signals were observed and fit to quadrupole doublets (Figure 3). The major component (63%) exhibited parameters (δ_major = 0.24 mm/s, |ΔE|_major = 1.32 mm/s) in accord with those reported previously for isolable (PDI)Fe(s-trans-η⁴-diene) complexes in the solid state¹⁵ and was assigned as (^MePDI)Fe(s-trans-η⁴-13j). The minor component (37%) exhibited parameters (δ_minor = 0.52 mm/s, |ΔE|_minor = 0.82 mm/s) similar to κ²-pyridine(diimine)¹¹ and iminopyridine iron(I) s-cis-η⁴ complexes^6g as well as α-diimine and iminopyridine iron(I) bis(olefin) complexes.^6f,6g,15b The minor species was thus tentatively assigned as either (^MePDI)Fe(η²-13j)₂ or (^MePDI)Fe(s-cis-η⁴-13j). Both assignments were corroborated with full-molecule DFT predictions (see Supporting Information); however, alternative possibilities cannot be excluded based on these data alone.

Figure 3.

Solution-state, zero-field ⁵⁷Fe Mössbauer spectrum (toluene glass, 80 K) of (^MePDI)Fe(13j)_n (n = 1,2) generated in situ. Simulated spectra plotted for the major (63%, blue) and minor (37%, red) components with the listed parameters, along with the net spectrum (grey).

With diagnostic spectroscopic data in hand for both (^MePDI)Fe(η⁴-13) and (^MePDI)Fe(η²-6)₂ complexes, iron speciation in the presence of both coupling partners was examined. A mixture of [(^MePDI)Fe(N₂)]₂(μ₂-N₂), dimethylphenylvinylsilane (6b), and isoprene (13i) was prepared in benzene-d₆ in a 1:4:2 ratio and examined by ¹H NMR spectroscopy (Scheme 11). An equilibrium mixture of both complexes as well as free 13i and 6b was observed, with an equilibrium constant favoring isoprene binding over the vinylsilane (K_eq = 10.8 at 23 °C). Additionally, both complexes were detected under catalytic conditions (e.g. 5 mol% [Fe]). The proposed metallacycle was not observed spectroscopically, which was consistent with previous findings from our group.¹⁵ Taken together, these data suggest that multiple diene coordination modes and electronic structures of pyridine(diimine) iron 1,3-diene and vinylsilane species are accessible under catalytically relevant conditions but that (^MePDI)Fe(η⁴-13) predominates.³⁰

Scheme 11.

Relative Stability of Representative (^MePDI)Fe(vinylsilane)₂ and (^MePDI)Fe(isoprene) Complexes.

Kinetic Analysis.

The rate-determining steps in the [(PDI)Fe]-catalyzed intra- and intermolecular olefin [2+2]-cyclodimerization and cross-[2+2]-cycloaddition are highly system dependent.^11,13b In an effort to identify the possible rate- and selectivity-determining step(s) in cross-[4+2]-cycloaddition of vinylsilanes and 1,3-dienes, reaction progress kinetic analysis was performed for the cycloaddition of vinyldimethphenylsilane (6b) with myrcene (13j) using [(^MePDI)Fe(N₂)]₂(μ₂-N₂) as the precatalyst.³¹ Although preparative reactions were generally conducted in neat substrate, concentrated stock solutions of catalyst and substrate were employed for the kinetic measurements while maintaining a constant total volume. Under these conditions, the chemoselectivity for [4+2] cycloaddition remained high, and the disappearance of vinylsilane 6b and formation of [4+2]-cycloadduct 16bi were monitored readily over the entire course of the reaction by gas chromatographic analysis of aliquots removed from the reaction mixture at regular timepoints. Select cases were validated by ¹H NMR spectroscopic measurements. Runs initiated at different concentrations of vinylsilane 6b but with the same excess concentration of 13i afforded concentration versus time data that overlaid graphically, indicating that negligible catalyst decomposition or product inhibition occurred over the course of the reactions (Figure 4A).

Figure 4.

Kinetic analysis of cross-vinylsilane/diene-[4+2]-cycloaddition. (A.) Rate profiles for trials conducted with a ‘same excess’ of diene (dark open markers vs. light filled markers) but a different initial concentration. (B.) Rate dependence on diene 13j concentration. (C.) Rate dependence on vinylsilane 6b. (D.) Rate dependence on [Fe]_tot. See supporting information for details.

For all of the kinetic measurements performed, the profiles of rate versus substrate concentration were linear, signaling a net first-order dependence on the vinylsilane substrate concentration. Correspondingly, effective rate constants were readily obtained from fitting the concentration versus time data to an exponential, first-order rate law. Given the well-behaved kinetic regime and the long reaction times necessary to reach full conversion with the most dilute reaction concentrations, the orders in [5b], [13j], and [Fe]_tot were determined using the method of initial rates, varying the initial concentration of each component while maintaining that of all others. This established no rate dependence on diene [13i] (Figure 4B) and a saturation-dependence in vinylsilane [6b] (Figure 4C) with typical conditions falling in the first-order regime. A strictly first-order dependence on [Fe]_tot was observed (Figure 4D). The rate law was determined to be rate = k[Fe][vinylsilane]. Taken in combination with the stoichiometric studies, these kinetic data are consistent with a mechanism in which either vinylsilane coordination to a resting state iron diene complex (at low concentration) or oxidative cyclization (at high concentration) is rate-determining. Given that these are the first effectively irreversible steps involving both substrates, this indicates that either vinylsilane coordination or oxidative cyclization is also regioselectivity-determining. Although C–C reductive elimination cannot be ruled out as rate- and regioselectivity-determining based on kinetic data alone, this possibility is disfavored by the stoichiometric experiments.

Selectivity Model and Analogy to α-Olefins.

Given the evidence implicating olefin coordination and/or oxidative cyclization as selectivity determining, the relative energies of intermediates B-vi/B-vii (β-[Si]) vs. C-x/C-xi (α-[Si]) (Scheme 8) were assessed using full-molecule density functional theory.³² Although the complicated electronic structures of the relevant iron complexes would necessitate in-depth multiconfigurational analysis for rigorous analysis of the full reaction coordinate,³³ the ground-state DFT structures were evaluated to provide qualitative understanding of the regioisomeric candidate metallacycles. In the lowest energy conformers identified for each pair of intermediates, the β-[Si] disposition was preferred. These ground state calculations suggest that there is no intrinsic preference for α-[Si] incorporation. To the extent that the transition state for oxidative cyclization resembles the incipient metallacycle, these computational models support mechanistic hypothesis B.

Further experimental support for mechanistic hypothesis B follows from reexamination of the product mixtures formed from (^MePDI)Fe-catalyzed α-olefin/myrcene coupling (Scheme 7A). In these cases, the formation of hydroalkenylation products (16) reports on the intermediacy of metallacycloheptene vii (Scheme 8). Furthermore, the connectivity of 16 was analogous to that of the hydroalkenylation side-products (19dj–gj) generated using vinylsiloxane substrates, which likely formed from the analogous metallacycloheptene B-vii.³⁴ The hydroalkenylation isomers that would arise via metallacycloheptene C-xi were not observed. Although this metallacyclic intermediate cannot be conclusively ruled out, our findings support that it is not part of the mechanistic pathway. The parallel regioselectivity trends are suggestive of a shared mechanistic pathway in which any α-[Si] effect is negligible. In the presence of a conjugated diene, the diene substitution pattern alone controls metallacycle formation. In this case, the difference in [4+2]-cycloaddition versus hydroalkenylation chemoselectivity arises from the facility of β-H elimination and C–H reductive elimination relative to C–C reductive elimination. We hypothesize that the highly substituted β-SiR₃ groups impart enhanced conformational rigidity in the intermediate metallacycle, thereby preventing the orbital overlap necessary for β-H elimination and instead favoring C–C reductive elimination.

CONCLUSIONS

In summary, three types of [(PDI)Fe]-catalyzed carbon-carbon bond forming reactions for the chemo- and regioselective elaboratio of vinylsilanes that are distinct from analogous processes with pure hydrocarbons. Specifically: (i) head-to-head vinylsilane dimerization, (ii) cross-[2+2]-cycloaddition with 4-substituted 1,3-dienes, and (iii) cross-[4+2]-cycloaddition with 2-substituted 1,3-dienes were achieved with selectivity derived from substrate control and with commodity reagents yield products with potential applications in silicones and small-molecule synthesis. Stoichiometric reactions, kinetic measurements, in situ Mössbauer spectroscopy, and DFT analyses provided insights into the origin of the observed selectivities. In contrast to α-olefin coupling partners, vinylsilanes exhibit distinct stereoelectronic preferences in some cases and specifically, suppress β-H elimination from key metallacyclic intermediates. Nonetheless, addition of a silyl group was insufficient for perturbing the regioselectivity of metallacycle formation in the presence of 1,3-dienes. These insights shed light on the interplay of ligand and substrate control in the development of selective chemistry that leverages metallacyclic intermediates to upgrade unsaturated coupling partners.