Low-valent tungsten redox catalysis enables controlled isomerization and carbonylative functionalization of alkenes

Abstract

The controlled isomerization and functionalization of alkenes is a cornerstone achievement in organometallic catalysis that is now widely used industrially to produce detergents, fragrances, and pharmaceuticals. In particular, the addition of carbon monoxide (CO) and hydrogen (H₂) to an alkene, also known as the oxo-process, is used in a constantly expanding industry that produces linear aldehydes from crude alkene feedstocks. In these catalytic reactions, isomerization is governed by thermodynamics, giving rise to functionalization at the most stable alkylmetal species. Despite the ubiquitous industrial applications of tandem alkene isomerization/functionalization reactions, selective functionalization at internal positions has remained largely unexplored. Here we report that the simple W(0) precatalyst, W(CO)₆, catalyzes the isomerization of alkenes to unactivated internal positions and subsequent hydrocarbonylation with CO. The 6- to 7-coordinate geometry changes that are characteristic of the W(0)/W(II) redox cycle and the conformationally flexible directing group are key factors in allowing isomerization to take place over multiple positions and stop at a defined unactivated internal site that is primed for in situ functionalization.

Tungsten is widely employed in commercial applications of central importance to modern society, including hard materials (e.g., tungsten carbide and metal alloys) and petroleum upgrading processes¹ (e.g., the Shell Higher Olefin Process and the Neohexene Process). Most notable in this context are catalytic processes that employ high-valent tungsten (e.g., +6) in a fixed oxidation state for alkene and alkyne metathesis² and polymerization reactions³ (Figure 1a). Low-valent tungsten (e.g., +0), on the other hand, has a strikingly different reactivity profile, in that it possesses the potential for redox reactions, such as oxidative addition and reductive elimination. Despite these promising features and the intriguing reactivity that has been demonstrated with stoichiometric low-valent tungsten, such as dearomatization^4–5 and alkene hydrofunctionalization^6–7, low-valent tungsten redox catalysis has been limited to a few specific types of transformations, including alkene isomerization, hydrogenation, and allylic substitution^8–10 (Figure 1b). We were drawn to the W(0)/W(II) redox cycle because of the ability of the metal to naturally fluctuate between 6-coordinate octahedral and noncanonical 7-coordinate geometries (Figure 1c). In theoretical^11–12 and experimental^13–14 studies it has been shown that the facile nature of migratory insertion and reductive elimination at 7-coordinate W(II) or Mo(II) is due to the spatial proximity of the ligands and attractive non-bonding interactions. Additionally, 6-coordinate alkyltungsten(II) species are known to undergo rapid β-H elimination to form 18-electron 7-coordinate W(II)–H species^15–16.

Fig. 1. Overview of tungsten reactivity and isomerization-carbonylation reactions.

(a) Examples of widely-used molecular tungsten (pre)catalysts and the differences in electronic properties of high valent and low valent catalysts. I W(Cl)₆, is a precatalyst used for alkyne polymerization. II shows the general structure of alkyne metathesis catalysts. III shows the general structure for imido tungsten olefin metathesis catalysts. IV W(CO)₆, the catalyst used in this study. (b) An early example of non-selective alkene isomerization using W(CO)₆ as a catalyst (top). Seminal example of a stoichiometric alkene hydrofunctionalization reaction using low-valent tungsten (bottom). (c) Model redox reaction of W(0)/(II) cycle showing the key changes in coordination chemistry upon changing the oxidation state. (d) Hydrocarbonylation of terminal alkenes with linear or branched selectivity (top). Isomerization-hydrocarbonylation of internal alkenes resulting in formation of the linear product (middle). Isomerization-hydrocarbonylation of internal alkenes resulting in formation of the α-functionalized product, where [G] = aryl, carbonyl, ether, etc. (middle). (e) The strategy of utilizing a directing to achieve internal selectivity for isomerization-carbonylation utilized in this work.

With these findings in mind, we hypothesized that the interplay between 6- and 7-coordinate species would be particularly well-suited to address a synthetically important problem of tandem isomerization and hydrocarbonylation of alkenes at a classically disfavored position. Alkene isomerization/hydroformylation is one of the highest volume homogenous catalytic reactions in the chemical industry^17–19, making it a useful reaction for benchmarking reactivity and selectivity patterns. Typically, in this family of reactions, the regioselectivity is dictated by thermodynamics, with functionalization occurring at the most stable alkylmetal intermediate^20–21. In particular, migration of the metal to the terminal position or to a carbon next to a π-system can reliably give linear or α-functionalized products, respectively, in high regioselectivity. Nevertheless, for a given molecule of interest, alternative positions cannot generally be functionalized with current methodology (Figure 1d). More broadly speaking, controlled alkene isomerization to classically inaccessible internal positions has been limited to a few examples where the isomerization is coupled to an intramolecular cyclization or rearrangement, including 5-endo cyclizations by Kochi²² and [3,3]-sigmatropic rearrangements by Marek²³. Capturing these intermediates via intermolecular functionalization, namely coupling with CO, would offer different site-selectivity in regiocontrolled alkene functionalization (Figure 1e).

We reasoned that the isomerization and tandem carbonylation process could be guided to a traditionally unfavorable position via metallacycle contraction²⁴ or “metallacycle zipping,” with site-selectivity governed by the nature of the directing group employed²⁵. We envisioned the geometric flexibility of 7-coordinate W(II) would facilitate otherwise challenging endocyclic β-H eliminations required for the directed alkene isomerization. Stabilizing an alkylmetal species of a defined ring size through coordination of the directing group (Fig. 1e) would then allow carbonylative functionalization²⁶ to take place with high regioselectivity²⁷. The development of a W(0)-catalyzed directed migratory carbonylation would complement other substrate-directed tandem alkene isomerization/functionalization reactions, such as oxidative arylation and hydroalkylation catalyzed by nickel^28–29.

Results and discussion

Identification of an effective directing group and catalyst.

To initiate the study, we tested a library of known directing groups (DGs) for their ability to facilitate isomerization and carbonylation of an alkenyl carboxylic acid derivative with stoichiometric W(CO)₆ at 150 °C (table 1, entries 1–6). Chatani’s NH–Pic (Pic = 2-picolinyl) directing group (DG4), which was originally developed for ruthenium-catalyzed C(sp²)–H carbonylation³⁰, gave quantitative yield of the desired product as a single regioisomer. All other directing groups resulted in mixtures of internal alkenes and failed to give any carbonylated products. Given that W(CO)₆ readily dissociates CO upon UV irradiation^{6–7, 9, 15–16, 31} we questioned whether the reaction would take place at lower temperature (25 °C) under photocatalytic conditions. Indeed, the desired product was formed in good yield when using stoichiometric W(CO)₆ under an argon atmosphere with irradiation from a 450 Watt UV lamp (entry 7). With 20 mol% W(CO)₆, there was evidence of turnover (entry 8), but all reactivity shut down with catalytic tungsten under an atmosphere of CO (table 1, entries 9–10). One explanation is that the relatively high concentration of CO in solution at room temperature leads to coordinative saturation of the catalyst and inhibits substrate association. Under thermal conditions (150 °C), 93% of product 2a was obtained using 20 mol% W(CO)₆ under 1 atm CO (entry 12). Other homoleptic metal carbonyls were less effective (Mo, entry 14) or ineffective all together (Cr and Fe, entries 15–16). Although molybdenum, and to some extent chromium, are known to exhibit similar reactivity and coordination chemistry to tungsten^{4, 9–10,13–14}, the former metals resulted in catalyst death, as evidenced by rapid (<1 h) precipitation of black solids. We attribute the higher yield of tungsten in this case to its greater stability under the reaction conditions^31,32, and the more stable M–H bond of tungsten (see supplementary Figure S26 for DFT calculations). Additionally, a representative panel of Rh, Ru, Co, and Ir precatalysts failed to promote the reaction or did so with low efficiency (See supplementary Table S2).

Table 1. Reaction Optimization.

Cond. 1 describe a photochemical procedure which utilizes UV light to liberate CO ligands from the tungsten center to free coordination sites and allow for substrate coordination. Cond. 2 describes a thermal procedure which liberates CO ligands by heating. All reactions were carried out on 0.2 mmol scale unless otherwise stated.

^aCO pressure at ambient temperature before heating (see Supplementary Information “General Procedure A” for full details).

^bYield determined by ¹H NMR with CH₂Br₂ as an internal standard.

^c1 atm of argon was used with no added CO. DG, directing group; atm, atmospheres of pressure; AQ, 8-aminoquinoline; PicNH, 2-aminomethylpyridine.

Reaction scope.

Having optimized the reaction conditions, we next investigated the reaction scope (table 2). Notably, γ,δ-unsaturated amides bearing various substitution patterns are readily accessible via numerous robust organic reactions³³ including the Diels–Alder cycloaddition, Ireland- or Johnson–Claissen rearrangement, and Wittig olefination. γ,δ-Unsaturated amides containing a terminal alkene and α-monosubstitution afforded the products with high yields and moderate d.r. (2b–2c), while an α,α-disubstituted alkenyl amide (2d) gave lower yield. A derivative of (S)-allylglycine underwent the transformation (2e) and retained high e.e. (98%) at the α-stereocenter in the case of the major diastereomer. The minor diastereomer showed erosion to 52% e.e., suggesting epimerization of the kinetically favored cis product to the thermodynamically preferred trans product. The reaction was effective with 1,1- and 1,2-disubstituted alkenes (2f–2h). The presence of a primary alkyl chloride (2i) or a second, more distal alkene (2j) was well tolerated.

Table 2. Substrate scope of isomerization-hydrocarbonylation.

Unless otherwise stated, all reactions were carried out in a standard Schlenk tube on 0.2 mmol scale using W(CO)₆ (20 mol%), toluene (2.0 mL) with 1 atm of CO and run for 24–72 hours with the isolated yields shown.

^aCO pressure (1 atm) at ambient temperature before heating; see Supplementary Information “General Procedure A” for full details.

^b1 equiv. of W(CO)₆ under argon; see Supplementary Information “General Procedure B” for full details.

^cYield determined by ¹H NMR using CH₂Br₂ as an internal standard. For further data on functional group robustness test, see Supplementary Information Table S3.

Next, we investigated substrates that require thermodynamically uphill alkene isomerization, which we reasoned could be compensated for in the highly exergonic carbonylation step (vide infra). Indeed, trisubstituted alkenes were isomerized to disubstituted alkenes, which then afforded the desired carbonylated products (2k–2n, 2w) in moderate to high yields. Endocyclic alkenes represent another class of challenging substrates for selective alkene isomerization/functionalization reactions. Pleasingly, Diels–Alder-derived substrates bearing di- (2u–2v) or trisubstituted (2w) alkenes gave moderate yield with exclusively cis stereochemistry at the ring juncture. A cyclopentenyl substrate (2t) also reacted in high yield. Remarkably, exocyclic methylidene cyclobutane was efficiently isomerized and carbonylated via the intermediacy of a highly strained endocyclic cyclobutene, providing 2o in modest yield and >20:1 d.r. Various conjugated alkenes were next evaluated. An α,β-unsaturated ester (2q) provided the product in high yield and with high chemoselectivity. Styrenes with versatile (and potentially sensitive) functional handles, including aryl iodides, bromides, and nitriles (2q–2s) proceeded in high yields. A series of additive inhibition experiments revealed the reaction to be compatible with ketone, alcohol, and free amine functional groups and incompatible with thiol and carboxylic acid groups (see Supplementary Information Table S3).

Lastly, we probed the distance dependence of this isomerization/carbonylation process. A substrate requiring no initial isomerization (2x) or isomerization over one position (2a) gave the corresponding products in greater than 90% yield. Moving the alkene farther away led to a gradual decrease in yield from 71% (2g) to 6% (2h). In cases with incomplete conversion to carbonylated product, we observed formation of a mixture of internal alkene isomers accompanied by gradual catalyst death, as evidenced by formation of a black precipitate. We attribute the decrease in product yield over distance to increased activation entropy in the directed isomerization pathway and the sluggish nature of the alternative non-directed pathway⁹. Synthesis of 2a on 5 mmol scale provided the product in 89% yield (951 mg), which could then be hydrolyzed to provide the corresponding 1,4-dicarboxylic acid in 92% yield (See Supplementary Scheme S2).

We were then curious if chiral branched variants of DG4 could control relative stereochemistry in the CO insertion step to provide diastereoselective carbonylated products. Introduction of a methyl group at the benzylic carbon (DG7) led to 5:1 d.r. and 82% yield, while progressively more bulky alkyl groups (DG8–DG10) did not lead to further improvement in diastereoselectivity and eventually led to a drastic drop in yield. Use of commercially available, enantiopure (S)-DG7 under catalytic conditions provided the product 2aa in 75% yield and 5:1 d.r. (>99% e.e. for both diastereomers). Overall, this result establishes a platform for further development of chiral directing auxiliaries for stereocontrolled isomerization/hydrofunctionalization reactions.

Reactivity of model organometallic intermediates.

The unique features of this catalytic isomerization/carbonylation process motivated us to interrogate the reaction mechanism and the underlying coordination chemistry of the intermediates involved. To examine the potential involvement of W(0)/W(II) redox processes, we first attempted direct synthesis of organometallic intermediates from an alkenyl amide starting material (1xa) by treatment with W(MeCN)₃(CO)₃ (1 equiv) in CDCl₃ at temperatures ranging from 3–60 °C, but these experiments only showed formation of the organic product 2x and unreacted starting material 1xa by ¹H NMR (see supplementary Figure S4). Thus, we next turned our attention to independent synthesis of isolable model complexes W-1–W-3 in an effort to establish the feasibility of these W(II) intermediates to undergo CO migratory insertion and reductive elimination (Figure 2a). Oxidative addition of alkyl iodide SM-1 to W(MeCN)₃(CO)₃ furnished model complex W-1 which is stable in solution for several days. After full conversion of SM-1, AgOTf and KOt-Bu/18-crown-6 were added to W-1 at room temperature, triggering migratory insertion and reductive elimination to afford the desired product (P-1) in high yield. Alternatively, migratory insertion and reductive elimination can be thermally promoted. Heating W-1 to 150 °C under 1 atm CO in the presence of excess 1a yields both P-1 (30%) and 2a (12%) (for full details see Supplementary Scheme S4). Due to our inability to isolate this alkyltungsten(II) intermediate in crystalline form, further studies focused on more thermally stable aryltungsten(II) species³⁴ to facilitate analysis by X-ray crystallography (Figure 2b).

Fig. 2. Synthesis and reactivity of model W(II) intermediates.

(a) The reactions were conducted in argon-filled glovebox. Reaction conditions: (Top) SM-1 (0.1 mmol), W(MeCN)₃(CO)₃ (0.1 mmol), THF (3.0mL), 2 h, 23 °C. Then, AgOTf (0.1 mmol), KOt-Bu (0.1 mmol), and 18-crown-6 (0.1 mmol), 48 h, 23 °C. (Bottom) SM-2/3 (0.3 mmol), W(MeCN)₃(CO)₃ (0.3 mmol), THF (4.0mL), 2 h, 45 °C. Then, AgOTf (0.3 mmol), 30 min, 45 °C. Then, PMe₃ (6.5 mmol), KOt-Bu (0.3 mmol), and 18-crown-6 (0.1 mmol), 2 h, 23 °C. (b) X-ray structures of complexes W-3 (CCDC 2020047) and W-3’ (CCDC 204639) with methyl group hydrogens and solvent omitted for clarity.

Using an analogous synthetic sequence, the complexes W-2 and W-3 were prepared by oxidative addition of the corresponding aryl iodides SM-2 and SM-3 to W(0)³⁵. W-3 was then treated with AgOTf to abstract the iodide, and further treatment with KOt-Bu/18-crown-6 deprotonated the amide. PMe₃ (2.5 equiv) was added to prevent migratory insertion and reductive elimination, as was observed with W-1. Without addition of PMe₃, we were unable to isolate any organotungsten species at this stage, and other ligands (PPh₃, PhCN, tert-butyl isocyanide, or pyridine) failed to form isolable complexes. While W-2’ was found to be unisolable, W-3’ is air- and moisture-stable and can be purified by column chromatography on neutral alumina. The X-ray structure of W-3’ revealed the directing group to be bound in the amidate-form, demonstrating the feasibility of this coordination mode.

Mechanistic studies.

Having established plausible oxidation states and coordination structures for many of the key proposed intermediates, we next investigated the mechanism of alkene isomerization. Earlier observations that the amino acid substrate (2e) maintained its stereochemistry at the α-position and that the α,α-disubstituted substrate (2d) gave product indicated that the carbonylation reaction likely proceeds through the β,γ-unsaturated isomer (as drawn in table 2). To further exclude the alternative hypothesis that the carbonylation proceeds through the conjugated α,β-unsaturated amide, we subjected the corresponding crotonyl amide (1xb) to the standard reaction conditions, and <10% product 23 was observed by ¹H NMR of the crude reaction mixture, with the remainder being unreacted starting material (see Supplementary Scheme S6).

After determining the positional aspects of isomerization, we then considered possible mechanisms, namely 1,2- and 1,3-hydride shift pathways^36–37, through a series of deuterium labeling experiments and density functional theory (DFT) calculations. To test the feasibility of tungsten to undergo N–H oxidative addition and W(II)–H insertion into alkenes, we conducted a series of experiments in which the amide N–H was exchanged to N–D in situ via treatment with excess CH₃OD under the reaction conditions (see Supplementary Figure S7). Evaluation of substrates with varying distances between the amide and the alkene as well as different substitution patterns (1xa, 1gb, and 1ga) gave evidence for this mechanism, with 82–96% deuterium incorporation into the distal position. An experiment with 1g-d₃ under standard conditions revealed that isomerization is unidirectional. In an effort to test for potential contributions from a 1,3-H shift pathway, substrate 1a-d₂ was subjected to the reaction conditions. Analysis of the product showed that 0.50 D was incorporated at the γ-methylene, while 0.30 D was also found in the δ-methyl. Careful inspection of HRMS data revealed considerable erosion and intermolecular scrambling of the deuterium labels. This could arise from a mechanism of substrate association, N–H oxidative addition, W(II)–H migratory insertion, β-D elimination, and then D⁺/H⁺ exchange of the resulting W(N–D) species (see supplementary Scheme S7 for a graphical depiction). The observation of intermolecular isotope exchange raises the possibility that D found at the terminal (δ) position does not exclusively arise from a 1,3-shift pathway, as one might anticipate a priori. The overall product distribution from this experiment suggests that both the 1,3- and 1,2-shift pathways may occur in tandem, with the 1,2-H shift as the dominant mechanism.

DFT calculations were performed to identify the catalytically active W species³⁸ and its reactivity in promoting N–H oxidative addition and disambiguate the mechanism of alkene isomerization (Figure 4). The six-coordinate W(0) complex I-1 was found to be the most likely species in N–H oxidative addition (see supplementary Figure S18 for details), leading to a seven-coordinate W(II)–H (I-2). Subsequent metal–hydride insertion (TS2), endocyclic β-hydride elimination (TS3), and metal–hydride reinsertion (TS4) are all facile, leading to a more stable five-membered metallacycle (I-5). An agostic interaction with the endocyclic β-C–H bond in I-3 promotes the otherwise challenging endocyclic β-hydride elimination (TS3). The seven-coordinate geometry of TS3 and the flexibility of the DG are both critical to alleviate strain in the transition state with a bidentate DG. After CO insertion (TS5) and binding with another CO molecule, the C–N reductive elimination from the seven-coordinate W(II) (I-7) requires a very low barrier (TS6) and is highly exergonic. It was found that 1,3-hydride shift pathways are kinetically less favorable (see supplementary Figure S19) and likely arise from a non-directed pathway^9,37. These DFT results in combination with the mechanistic experiments above show that this 1,2-hydride shift pathway is the prevailing mode of isomerization, while minor contributions from a 1,3-hydride shift may also occur in tandem.

Fig. 4. Computed reaction energy profile of the isomerization–carbonylation of 1a.

M06-D3/SDD-6–311+G(d,p)/SMD(toluene)//B3LYP-D3/LANL2DZ-6–31G(d) level of theory. After the N–H oxidative addition (TS1), seven-coordinated W(II)–H undergoes facile alkene isomerization via TS2 and TS3.

During the course of the isomerization process, buildup of intermediate internal alkenes is detected for reactions halted prior to completion (see supplementary Scheme S4), suggesting substrate dissociation and re-association are likely occurring, which is further supported by incorporation of multiple deuterium equivalents at different chain positions. DFT also predicts that ligand exchange of the coordinated alkene with free CO is energetically feasible (see Supplementary Figure S21), which can explain why intermediate internal alkenes can be observed when stopping the reaction early. In chain-walking reactions, premature dissociation of the alkene from the substrate is typically deleterious and leads to side products or intermediates that cannot feed back into the catalytic cycle³⁹; however, in this reaction it appears to have little effect on the overall yield and regioselectivity.

The results with deuterium-labeled substrate 1g-d₃ revealed an unusual phenomenon³⁹ that isomerization proceeds in a uni-directional fashion exclusively towards the directing group. Running the complementary experiment with internal alkene 1gb further corroborated the uni-directional nature of isomerization, as no deuterium was detected in the terminal position (Figure 3). The uni-directional alkene isomerization is consistent with the greater stability of the π-complex with internal alkene (I-4) than the terminal alkene complex (I-2).

Fig. 3. Mechanistic experiments and deuterium labeling studies.

D incorporation determined by ¹H NMR analysis of purified products; D distribution determined by HRMS of purified starting material and product. a, Incorporation of deuterium via in situ N–H/D exchange and N–D oxidative addition and D-insertion into the alkene. b, D scrambling studies with deuterium incorporated in various positions of starting material. ^a CO pressure (1 atm) at ambient temperature before heating (see Supplementary Information “General Procedure A” for details). Pic, 2-picoyl.

Conclusion

When compared to relevant industrial carbonylation processes, this method still has drawbacks that will be addressed in future work, namely the requirement for a stoichiometric auxiliary and relatively high catalyst loadings. Along these lines, directed alkene isomerization/functionalization employing native functional groups holds great promise owing to the ability of W(0) to form stable adducts with weakly Lewis basic groups⁴⁰. Nevertheless, these findings provide an experimental and computational framework for exploring W(0)/W(II) catalytic processes and illustrate how this catalytic cycle can enable exceptional regioselectivity with a diverse range of alkenes in a challenging tandem isomerization/functionalization transformation.

Methods

Caution:

Carbon monoxide (CO) is a colorless, odorless gas that is extremely toxic. Filling or opening reaction tubes that contain CO needs to be done in a well-ventilated fumehood with a CO monitor. W(CO)₆ and other M(CO)_x complexes used in this study thermally decompose to give off free CO, so all reactions run with M(CO)_x complexes should be handled with the same precautions as those using gaseous CO, while taking into account any potential hazards introduced by the metal carbonyl species themselves.

General procedure for catalytic carbonylation:

A 25-mL Schlenk tube with a Teflon-coated magnetic stir bar containing W(CO)₆ (14 mg, 0.4 mmol) and the appropriate alkene (0.2 mmol) was introduced into an argon filled glovebox. Toluene (2.0 mL) was added via micropipette, with care taken to ensure that solids on the wall were all washed into the bottom of the tube. The flask was closed and removed from the glovebox then connected to a Schlenk line. A carbon monoxide tank was directly connected to a Schlenk line, and the line was purged with CO. The reaction tube was submerged in a cooling bath of dry ice and acetone (−78 °C), and the reaction solution was then carefully exposed to vacuum. CO was then back-filled into the Schlenk tube and sealed. Alternatively, standard freeze-pump-thaw technique can be used to introduce CO gas. The reaction tube was allowed to warm room temperature and was then submerged into a preheated oil bath at 150 °C and heated for 24–72 h.

Data availability statement

All data generated or analyzed during this study are included in this Article (and its Supplementary Information). The structures of W-2, W-3, W-3’, and 2u in the solid state were determined by single-crystal X-ray diffraction and the crystallographic data have been deposited with the Cambridge Crystallographic Data Centre: CCDC 2020047 (W-2), 2050458 (W-3), 2045639 (W-3’), 2008992 (2u). Copies of the data can be obtained free of charge via https://www.ccdc.cam.ac.uk/structures/.