Experimental and Computational Studies on Cobalt(I)-Catalyzed Regioselective Allylic Alkylation Reactions

Abstract

Here, we report the development of cobalt(I)-catalyzed regioselective allylic alkylation reactions of tertiary allyl carbonates with 1,3-dicarbonyl compounds. A family of well-defined tetrahedral cobalt(I) complexes bearing commercially available bidentate bis(phosphine) ligands [(P,P)Co(PPh₃)Cl] are synthesized and explored as catalysts in allylic alkylation reactions. The catalyst [(dppp)Co(PPh₃)Cl] (dppp = 1,3-Bis(diphenylphosphino)propane) enables the alkylation of a large variety of tertiary allyl carbonates with high yields and excellent regioselectivity for the branched product. Remarkably, this methodology is selective for the activation of tertiary allyl carbonates even in the presence of secondary allyl carbonates. This contrasts with the selectivity observed in cobalt-catalyzed allylic alkylations enabled by visible light photocatalysis. Mechanistic insights by means of experimental and computational investigations support a Co(I)/Co(III) catalytic cycle.

Graphical Abstract

A cobalt(I)-catalyzed regioselective allylic alkylation of tertiary allylic carbonates with 1,3-dicarbonyl compounds is described. The alkylated products are obtained in high yields and with excellent selectivity towards the kinetic branched regioisomer. Mechanistic insights by combining experimental and computational investigations support a Co(I)/Co(III) catalytic cycle.

Introduction

Transition metal-catalyzed nucleophilic allylic substitutions are well-established methods for the construction of carbon–carbon and carbon–heteroatom bonds in organic synthesis.^[1] The introduction of soft carbon-based nucleophiles, such as malonates, have been traditionally dominated by palladium (as in the Tsuji-Trost reaction),^[1] rhodium^[2] or iridium,^[3] whereas reactions using hard carbon-based nucleophiles (RMgX, RZnX, R₂Zn, RLi) are usually catalyzed by copper.^[4] Recent advances in cobalt-catalyzed nucleophilic allylic substitutions illustrate that high levels of regio- and enantioselective control have also been achieved with this metal.^[5] Selectivity towards the branched product is desirable when seeking for methods to synthesize quaternary stereogenic centers employing achiral nucleophiles.^[6] In this regard, cobalt enables high regioselectivity towards branched addition products with stabilized enolates,^{[5a, 5c]} similar to the heavier group 9 metals, rhodium and iridium.^[7] This contrasts with cobalt-catalyzed reactions of hard carbon-based nucleophiles with allylic electrophiles since, in these cases, the linear product is usually the major regioisomer.^{[5d, 5e, 8]} Although cobalt-catalyzed asymmetric allylic substitution reactions are underdeveloped, these precedents prompt the interest for exploring Co as a more abundant and environmentally benign alternative for such transformations, with specific aims at finding complementary reactivity to the well-established methods.

The mechanistic scenario depicted for cobalt-catalyzed allylic substitutions is largely extrapolated from the knowledge obtained with other metal catalysts. Thus, it has been proposed that cobalt-catalyzed allylic substitutions with soft carbon-nucleophiles proceed via Co(I)/Co(III) catalytic cycles (Figure 1). Most mechanistic scenarios converge on the suggestion that a cobalt(I) species is likely the active catalytic intermediate that enables the activation of the electrophile, which is commonly an allyl carbonate. The catalytic cycle is closed by the nucleophilic attack to the putative π-allylcobalt(III) intermediate, which regenerates the Co(I) and delivers the product that contains a new functional group at the allylic position. Thus, most methods for allylic substitution reactions require reducing conditions to form in situ low-valent cobalt species as the active catalytic intermediates from Co(II) precursors.^{[5b, 5c, 9]} However, mechanistic studies in allylic alkylations catalyzed by well-defined Co(I) complexes are still limited.^{[5a, 5b, 10]} It is worth noting that cobalt(II) complexes have been shown to catalyze the allylation of 1,3-dicarbonyl compounds with allyl acetates, though with limited efficiency in comparison with the reductive methods.^[11]

Figure 1.

Previous work on cobalt-catalyzed allylic substitution reactions.

As part of our interest in the field of low-valent base-metal catalysis (Fe, Co, Ni),^[12] we aimed at the development of well-defined cobalt(I) catalysts bearing commercially available bidentate bis(phosphine) complexes and to investigate their catalytic activity and the mechanism in allylic alkylations reactions with 1,3-dicarbonyl compounds. The results obtained in this work shed light on the reaction mechanism of cobalt(I)-catalyzed allylic substitution reactions and provide a blueprint for rational catalyst design.

Results and Discussion

The combination of cobalt salts and bidentate bis(phosphine) ligands has been shown to give access to highly efficient catalysts for a variety of hydrogenations, alkene isomerizations, hydroborations, cycloadditions, Alder-Ene and hydroacylation reactions, among others, under reducing conditions.^[13] On the other hand, the isolation of Co(I) complexes bearing bidentate bis(phosphine) ligands showcases that cobalt can adopt different geometries depending on the ligand, the stoichiometry and reduction conditions.^{[5a, 13i, 14]} In this regard, in order to control the geometry of the Co(I) catalysts bearing bidentate bis(phosphine) ligands we sought to prepare tetracoordinate Co(I) complexes by ligand substitution reaction at CoCl(PPh₃)₃.^[15] CoCl(PPh₃)₃ is a convenient Co(I) precursor since it is commercially available or, alternatively, can be readily prepared in gram-scale in the laboratory. This strategy enables the synthesis of well-defined tetrahedral Co(I) complexes and a systematic evaluation of their catalytic activity.

Our study began with the synthesis of tetracoordinate Co(I) chloride complexes bearing bidentate phosphine ligands that present a variety of natural bite angles (from 83° for dppbz to 111° for Xantphos).^[16] As shown in Figure 2a, the direct coordination of several bidentate phosphine ligands (and concomitant dissociation of two PPh₃) to CoCl(PPh₃)₃ enabled the efficient synthesis of distorted tetrahedral Co(I) complexes [(P,P)Co(PPh₃)Cl] in yields ranging from 75% to 96%. These cobalt(I) complexes are reasonably stable to air and moisture in solid state, although they get oxidized rapidly in solution, similarly to their Co(I) precursor. Gratifyingly, we obtained the solid state structure for six of these complexes (1^Co-7^Co, Figure 3).^[17] A closer look at the structures of 6^Co and 7^Co that bear bis(phosphine) ligands with large bite angles (dppf and Xantphos respectively)^[18] show slightly distorted tetrahedral geometries. On the other hand, bidentate ligands with smaller bite angles led to highly distorted tetrahedral geometries in complexes 1^Co, 2^Co and 3^Co. Similarly, complex 4^Co, which contains the more electron-donating bis(phosphine) ligand dcpe, also exhibits a distorted tetrahedral geometry, featuring a slightly larger bite angle and larger Co–P bond lengths in comparison with 3^Co bearing dppe ligand.^[19]

Figure 2.

a) Synthesis of well-defined tetrahedral Co(I) chloride catalysts by ligand substitution reaction. b) Synthesis of cationic Co(I) complexes by anion exchange with Na[BAr^F₄] from Co(I) chloride precursors.

Figure 3.

Solid-state structures of complexes 1^Co, 2^Co, 3^Co, 4^Co, 6^Co, 7^Co, 11^Co and 12^Co and selected bond lengths and angles. For 2^Co, 3^Co and 11^Co the bond lengths are an average of 2, 3 and 3 independent crystallographic molecules, respectively. Displacement ellipsoids are shown at 50% probability; hydrogen atoms are omitted for clarity.

The Co(I) chloride complexes 1^Co-7^Co are convenient precursors to synthesize cationic Co(I) complexes bearing bidentate bis(phosphine) ligands [(P,P)Co(η⁶-C₇H₈)][BAr^F₄] (BAr^F₄ = B[(3,5-(CF₃)₂)C₆H₃]₄)). Thus, chloride exchange with Na[BAr^F₄] in aromatic solvents at complexes 1^Co, 2^Co and 5^Co afforded complexes 8^Co-10^Co in good yields (Figure 2b). Our two step synthesis from CoCl(PPh₃)₃ precursor is a synthetic alternative to previously reported procedures towards [(P,P)Co(η⁶-C₇H₈)]⁺ complexes.^{[13i, 13l, 14b, 20]} Recrystallization in benzene:pentane afforded suitable single crystals for X-ray diffraction (XRD) analysis for complexes [(dppbz)Co(η⁶-C₆H₆)][BAr^F₄] (11^Co) and [(dppe)Co(η⁶-C₆H₆)][BAr^F₄] 12^Co, which are shown in Figure 3.

Next, we evaluated the catalytic activity of the isolated cobalt(I) complexes in allylic alkylation reactions with soft carbon-based nucleophiles. Particularly, we explored the reaction of allylic carbonate 1a^OCO2Me and diethyl malonate 2 using 10 mol% of Co(I) precatalyst and 20 mol% of NaBF₄ as additive in MeCN, at 60 °C for 24 h (Figure 4). We observed a strong influence of the bite angle of the bis(phosphine) ligand on the reaction yield.^[21] While Co(I) precatalysts bearing dppf and Xantphos gave no conversion, the reaction proceeded with good yields (49–77%) when using Co(I) complexes featuring bis(phosphines) with smaller bite angles such as dppe, dppen and dppbz. A similar observation was reported for the enantioselective reverse prenylation of β-ketoesters.^[5a] The coordination of the more electron-donating dcpe ligand to the Co(I) center in [(dcpe)Co(PPh₃)Cl] (4^Co) catalyst ceases the catalytic activity. A control experiment with catalyst (dppp)CoCl₂, where the metal is in oxidation state +2, gave no conversion (see Table SI-1 in the Supporting Information (SI)).

Figure 4.

Screening of the catalytic activity of isolated (P,P)Co(PPh₃)Cl complexes. Conditions: 1.5 equiv of diethyl malonate, 1 equiv of allyl carbonate (0.2 M), (P,P)Co(PPh₃)Cl (10 mol%), NaBF₄ (20 mol%), MeCN, 60 °C, 24 h. The conversion (in parenthesis) and the yield were determined by GC using undecane as internal standard. n.d. = not detected.

Among the catalysts tested, complex 5^Co bearing dppp is the most efficient one under our standard reaction conditions, giving the product with 77% yield. In the absence of the additive NaBF₄ the reaction yield drops to 21%. Other salts such NaBAr^F₄, LiOTf or La(OTf)₃ were tested as additives but lower yields were obtained (Figure 4). Our synthetic method enables the addition of 1,3-dicarbonyl compounds without the use of external base, in contrast with the common reaction conditions reported for nucleophilic allylic substitutions catalyzed by Pd, Rh, Ir.^{[1c, 1d, 2–3, 22]} The performance of the cationic Co(I) complexes 8^Co-10^Co in catalysis was also tested. Similarly to the experiments with the Co(I) chloride complex 5^Co, the addition of NaBF₄ also had a positive impact in the yield (Figure 4). However, overall the highest yield was obtained when using complex 5^Co, which was selected for further exploring the scope of the reaction.

As shown in Table 1, our method is applicable for a wide range of tertiary allyl carbonates 1 containing two aliphatic substituents to afford the branched product 3 with high regioselectivity (the linear product is not detected by ¹H-NMR spectroscopy of the reaction crude). Allyl carbonates containing alkene (3h), ether (3c), silyl ether (3i), unprotected alcohol (3j), aryl and alkyl chlorides (3g and 3k), aryl fluorides (3f), amides (3i) and ester (3m, 3n) are suitable substrates. The method tolerates several heteroaromatics in the substrate such as pyrrole (3o), indole (3p), furan (3q), benzofuran (3r), thiophene (3s) and benzothiophene (3t). However, only secondary allyl carbonates containing an electron-deficient aryl group (3u, 3v, 3w) afford the branched alkylated product, and in low yields (27–30%). Other 1,3-diesters as dibenzyl malonate and diisopropyl malonate were suitable nucleophiles and they furnished the alkylated products 3x and 3y in 62% and 59% yields, respectively. The reaction using β-ketoesters such as ethyl 3-oxobutanoate and phenyl 3-oxobutanoate gave the corresponding branched product 3z and 3aa in 82% and 77% respectively. The reverse prenylation of β-ketoester 1ab containing the tetralone scaffold gave product 3ab in 69% yield. However, the ester group was required in the nucleophile since no conversion was observed when using β-diketones, as pentane-2,4-dione and 1,3-diphenylpropane-1,3-dione, and malononitrile derivatives (see SI for further details).

Table 1.

Screening of substrate scope.^[a]

^[a]Conditions: 81.5 equiv of diethyl malonate, 1 equiv of allyl carbonate (0.2 M), (dppp)Co(PPh₃)Cl (10 mol%), NaBF₄ (20 mol%), MeCN, 60°C, 24 h. [b] with 20 mol% of (dppp)Co(PPh₃)Cl. [c] Using the substrate containing the methyl carbonate instead of tert-butyl carbonate (Boc).

The cobalt(I)-catalyzed allylic alkylation protocol can be further simplified by preparing in-situ the (dppp)Co(PPh₃)Cl (5^Co) catalyst. Thus, a mixture of CoCl(PPh₃)₃ and dppp ligand in MeCN added directly to the reaction vessel that contains the allylic carbonate 1h, diethyl malonate and NaBF₄ afforded the branched alkylated product 3h in 56% yield (70% conversion). This yield is lower in comparison with the yield obtained when using isolated 5^Co as precatalyst (73% yield). We considered that PPh₃ could be an inhibitor for the reaction. Indeed, addition of increasing amounts of PPh₃ (from 0.5 to 5 equiv) to the standard reaction conditions using catalyst 5^Co caused a decrease in the yield of 3h to 30% already after addition of 1 equiv (see Table SI-5), which corroborates this hypothesis. Similarly, the addition of PPh₃ has also an inhibitory effect in the reactions catalyzed by the cationic Co(I) complexes [(P,P)Co(η⁶-C₇H₈)][BAr^F₄] (see Table SI-6).

The selectivity of the cobalt(I)-catalyzed allylic alkylation reaction reported here is showcased by testing substrate 1ac, that contains both secondary and tertiary allyl carbonates (Figure 5). Under the standard reaction conditions, product 3ac is isolated in 76% yield, which indicates that the malonate is only incorporated at the tertiary allyl carbonate, whereas the secondary allyl carbonate remains unreacted. The observed selectivity pattern is complementary to the selectivity commonly reported for palladium-catalyzed Tsuji-Trost reaction.^{[1c, 1d, 22–23]} In this regard, the reaction of diethyl malonate and substrate 1ac catalyzed by Pd(PPh₃)₄ afforded product 4 in 30%. Palladium(0) reacts with the secondary allyl carbonate to give the linear alkylated product but undergoes β-hydride elimination to form the 1,3-diene upon cleavage of the tertiary allyl carbonate.^[24]

Figure 5.

Divergent reactivity between Pd(PPh₃)₄ and (dppp)Co(PPh₃)Cl with allyl carbonates. Conditions: a) 1.5 equiv of diethyl malonate, 1 equiv of allyl carbonate (0.2 M), (dppp)Co(PPh₃)Cl (10 mol%), NaBF₄ (20 mol%), MeCN, 60°C, 24 h. b) 1.5 equiv of diethyl malonate, 1 equiv of allyl carbonate (0.1 M), Pd(PPh₃)₄ (5 mol%), THF, 30°C, 20 min.

The selectivity observed with isolated cobalt(I) precatalyst (dppp)Co(PPh₃)Cl (5^Co) in allylic alkylation reactions is also complementary to the selectivity achieved by the reported cobalt-catalyzed allylic alkylation under visible-light photoredox conditions (Figure 6).^[5c] Reaction of diethyl malonate 2 with tertiary allylic carbonates 1d and 1ad catalyzed by CoBr₂/dppp (5 mol%) and 1,2,3,5-Tetrakis(carbazol-9-yl)-4,6-dicyanobenzene (4Cz-IPN, 0.5 mol%) in the presence of ⁱPr₂NEt₂ (1 equiv) in THF, 4Å MS, rt, 16 h under visible-light irradiation did not give conversion. In contrast, as previously reported, the visible light photocatalytic reaction of nucleophile 2 with secondary allyl carbonates 1ae and 1af afforded the corresponding alkylated products 3ae and 3af in 78% and 81% yield, respectively. Similar results were obtained when CoCl₂ was used as Co(II) salt instead of CoBr₂. Our protocol enables the activation of more sterically hindered tertiary allylic carbonates, differing from photocatalytic methods that are limited to secondary allylic carbonates. This likely indicates that the thermal activation with well-defined Co(I) precatalysts and the photocatalytic activation with the CoX₂/dppp and 4Cz-IPN dual catalytic system proceed via different reaction mechanisms.

Figure 6.

Comparison of thermal conditions (method A) vs. visible-light photocatalysis (method B) catalyzed by cobalt. 4Cz-IPN = 1,2,3,5-Tetrakis(carbazol-9-yl)-4,6-dicyanobenzene.

Next, we focused on shedding light on the mechanism of the developed cobalt(I)-catalyzed allylic alkylation reaction by means of mechanistic probes in combination with theoretical calculations. We carried out a catalysis using substrate 1ag, which contains a cyclopropyl ring in α-position to the tertiary allyl carbonate. The corresponding branched alkylated product 3ag was isolated in 40% yield (Figure 7a). Importantly, the ring opening product was not detected, which rules out the intermediacy of long-lived allyl radicals. In addition, no conversion was observed in the reaction of Co(I) precatalyst 5^Co with the allyl carbonate 1d without the addition of diethyl malonate in MeCN at 60°C (Figure 7a). These results contrast those recently reported by Meng and co-workers on the reductive cobalt-catalyzed enantioselective addition of allyl carbonates to aldehydes.^[25] In that precedent, the use of phosphorous-oxazoline ligand in combination with CoI₂ and Mn as a reductant was key for the success of the reaction. The formation of radical intermediates upon activation of the allyl carbonate was proposed based on the observation of the ring opening product of an allyl carbonate containing a cyclopropyl ring as radical probe. In addition, it was reported the isolation of the homodimerization product of the allyl carbonate in the absence of aldehyde.

Figure 7.

a) Mechanistic probes. b) Proposed Co-catalysed mechanism for regioselective allylic alkylation reaction. The energies shown were computed at the uB3LYP-D3/def2tzvpp-CPCM(acetonitrile)//uB3LYP-D3/def2SVP-CPCM(acetonitrile) level of theory. c) Relevant transition state structures. All energies in the scheme are in kcal/mol.

The reaction of the enantiopure allyl carbonate (R)-1h with dibenzyl benzoate catalyzed by (dppp)Co(PPh₃)Cl afforded the alkylated product 3ah in 75% yield as a racemate (Figure 7a). The loss of stereochemical information is also observed when using the chiral Co(I) catalyst formed in situ by combining (PPh₃)₃CoCl and (R,R)-Ph-BPE ligand. This loss of stereochemical information during catalysis could indicate a faster oxidative addition of Co(I) into the allyl carbonate and isomerization of the π-allylcobalt(III) intermediate in comparison with the next steps that lead to the product formation, which is in agreement with the theoretical calculations (vide infra). Finally, analysis of the headspace of the reaction vessel after catalysis confirmed the quantitative formation of CO₂ (Figure SI–4).

In order to further understand the mechanism and the origin of the observed selectivity, we carried out dispersion-corrected density functional theory (DFT) calculations (see the SI for the details on the computational methods). For simplicity, we only show and discuss in detail the lowest energy pathway (see the SI for alternative mechanisms considered). As shown in Figure 7b, firstly an oxidative addition reaction of the cationic (dppp)Co(I)⁺ in the presence of the alkene 1a^OCO2Me proceeds rapidly and irreversibly via a ³TS-A-B (barrier, ~7.4 kcal/mol) to form the corresponding Co(III) intermediate ³B. Alternative (dppp)Co(I) species (i.e., (dppp)Co(I)-Cl, (dppp)Co(I)-MeCN and (dppp)Co(I)-PPh₃; Figure SI–41) were considered for the oxidative addition step, but the transition states were found to be much higher >12.0 kcal/mol). Then, the release of CO₂ promoted by cobalt was considered from intermediate ³B, as CO₂ was experimentally detected. Specifically, our calculations show that species ³B could undergo facile decarboxylation through the ³TS-B-B´ (barrier, ~16.4 kcal/mol) to form the corresponding Co(III) intermediate ³B’ and release of CO₂. From ³B’, the calculations show that the coordination (and deprotonation) of dimethyl malonate to afford the π-allylcobalt(III) intermediate ³C and MeOH is thermodynamically favourable. In turn, intermediate ³C is prone to undergo a branched-selective reductive elimination (via ³TS-C-D_branched) to furnish the desired product and restart the catalytic cycle. Notably, in agreement with experiments, the linear-selective reductive elimination (via ³TS-C-D_linear) is significantly higher than branched-selective reductive elimination (ΔΔG^‡ = 15.6 kcal mol⁻¹ versus 9.2 kcal mol⁻¹, Figure 7c). Noncovalent interaction (NCI) analysis revealed greater steric repulsions between the alkene substituents and the phenyl groups in the dppp ligand in ³TS-C-D_linear in comparison to ³TS-C-D_branched leading to exclusive regioselectivity (Figure SI–42). In addition, Natural Bond Orbital (NBO) analysis revealed that ³TS-C-D_branched benefits from favorable donor–acceptor interaction energy from the lone pair (LP) on the cobalt atom to the neighboring antibond (BD*) of the C1–C2 bond (Figure SI–43). Finally, higher barriers were observed for secondary allylic carbonate systems, consistent with the experimental data (See SI for further details and discussion).

Conclusion

In conclusion, we have developed regioselective allylic alkylation reactions with soft carbon-nucleophiles catalyzed by well-defined tetrahedral bis(phosphine) cobalt(I) complexes. The reaction of 1,3-diesters and β-ketoesters compounds with tertiary allyl carbonates catalyzed by cobalt(I) proceeds under mild conditions and with excellent regioselectivity towards the branched alkylated product. Remarkably, this method enables the alkylation of tertiary allyl carbonates in the presence of secondary allyl carbonates. This unusual selectivity is complementary to the previously observed selectivity achieved in cobalt-catalyzed allylic alkylation reactions under visible light photoredox catalysis. Radical probe experiments in combination with DFT computational studies were employed to elucidate the Co(I)-catalyzed mechanistic pathway for this transformation. All the collected data support a Co(I)/Co(III) catalytic cycle. Theoretical studies provide a better understanding of the non-covalent interactions that induce regioselectivity in this reaction.

Data Availability Statement

The data that support the findings of this study are available in the supplementary material of this article.