Cobalt-Catalyzed C(sp²)–C(sp³) Suzuki–Miyaura Cross-Coupling Enabled by Well-Defined Precatalysts with L,X-Type Ligands

Abstract

Cobalt(II) halides in combination with phenoxy-imine (FI) ligands generated efficient precatalysts in situ for the C(sp²)–C(sp³) Suzuki–Miyaura cross coupling between alkyl bromides and neopentylglycol (hetero)arylboronic esters. The protocol enabled efficient C–C bond formation with a host of nucleophiles and electrophiles (36 examples, 34–95%) with precatalyst loadings of 5 mol%. Studies with alkyl halide electrophiles that function as radical clocks support the intermediacy of alkyl radicals during the course of the catalytic reaction. The improved performance of the FI-cobalt catalyst was correlated with decreased lifetimes of cage-escaped radicals as compared to diamine-type ligands. Studies of the phenoxy(imine)-cobalt coordination chemistry validate the L,X interaction leading to the discovery of an optimal, well defined, air-stable mono-FI cobalt(II) precatalyst structure.

Graphical Abstract

INTRODUCTION

Transition metal–catalyzed cross coupling is an enabling and widely applied technology for the synthesis of pharmaceuticals, agrochemicals and fine chemicals.¹ Among the many variants used for C(sp²)–C(sp²) bond formation, the Suzuki–Miyaura reaction is the most widely used in medicinal chemistry,^2,3 and is by far the most common type of cross-coupling employed on scale for the synthesis of drug candidates.¹ The bench-stability and availability of the coupling partners, particularly neutral organoboron nucleophiles; the reliability and scalability with advanced intermediates; and the mild reaction conditions, distinguishes the Suzuki–Miyaura reaction from other catalytic cross-coupling methods.^1,4

Evolution of the metal-catalyzed Suzuki-Miyaura cross coupling to include C(sp³)-based partners is of interest in drug discovery because increasing three-dimensionality that has been linked to the performance and success of drug candidates.^5,6 First-row transition metals offer considerable promise for C(sp²)–C(sp³) cross coupling owing to a lower propensity for competitive β-hydrogen elimination when compared to palladium catalysts.^7,8 Examples with copper have been demonstrated but are principally limited to primary alkyl electrophiles.^9,10 Nickel catalysts have been widely explored for this transformation,^11,12,13,14 however widespread adoption has thus far been limited.¹⁵

Cobalt is appealing as a catalyst for this transformation because it is Earth-abundant, relatively inexpensive, and available in quantities that are amenable to large-scale pharmaceutical applications.^16,17,18 Cobalt catalysis would also provide an alternative in the synthetic toolbox for the installation of C(sp²)–C(sp³) bonds. Cobalt catalysts have demonstrated promosing reactivity with typically “unactivated” alkyl (pseudo)halide electrophiles in related couplings employing more reactive organomagnesium and organozinc nucleophiles.^19,20,21,22 However, unlike other first-row metals like nickel, cobalt-catalyzed Suzuki–Miyaura coupling remains in its infancy.

Examples of cobalt-catalyzed C(sp²)–C(sp²) cross-coupling with organoboron nucleophiles have been reported using select neutral ligands, including bis(phosphino)pyridine pincers,²³ terpyridine chelates,²⁴ and aryl-substituted N-heterocyclic carbenes (Scheme 1A).^25,26 The understanding of ligand-to-metal complexation has significant implications on the reproducibility of a cross-coupling reaction, and demonstrates the inherent difficulty in cross-coupling reaction development. More broadly, understanding catalytically relevant oxidation states, the nature of the active species, and redox cycles in cobalt-catalyzed cross-coupling lags behind iron- and nickel-catalyzed variants.^27,28

Scheme 1.

Cobalt-catalyzed C(sp²)–C(sp³) Suzuki–Miyaura cross-coupling

Our laboratory recently reported that the combination of cobalt(II) salts with trans-N,N’-dimethylcyclohexane-1,2-diamine in the presence of KOMe was catalytically active for the C(sp²)–C(sp³) Suzuki–Miyaura cross-coupling between neopentylglycol (hetero)arylboronic esters and alkyl bromides (Scheme 1B).²⁹ While the scope of the transformation included sterically and electronically diverse aryl boronate nucleophiles, the catalyst loadings were high (15 mol%), and little is known about metal-ligand interactions, the origin of the effects of the specific diamine substituents, or the mechanism of the C–C bond-forming event.⁴ For the development of new catalytic reactions with improved performance, an understanding of metal–ligand interactions and reactivity is a key aspect of successful reaction design.^30,31,32

Here we describe the application of readily prepared and modular phenoxyimine (FI) ligands to next-generation cobalt catalysts for C(sp²)–C(sp³) Suzuki–Miyaura cross-coupling (Scheme 1C). These compounds operate with lower cobalt loadings (5 mol%) and enable broader substrate scope, with lower amounts of side-products arising from alkyl bromide reduction. Substrates with known radical clocks support the intermediacy of radicals derived from the alkyl electrophile, which are stabilized by the [(FI)Co] complexes to a higher degree compared to complexes with diamine ligands. X-ray crystallographic and EPR spectroscopic studies provided insight into the metal–ligand interaction in cobalt precatalysts as well as to identify deactivation pathways of cobalt(II) sources.

RESULTS AND DISCUSSION

Part I. Discovery and Evaluation of the Methodology.

The discovery of next generation cobalt catalysts for C(sp²)–C(sp³) Suzuki–Miyaura reaction began with high-throughput experimentation (HTE)³³ to evaluate classes of ligands (Table 1). A representative catalytic reaction was selected using N-Cbz-4-bromopiperidine 1a (1 equiv) as a prototypical electrophile with pharmaceutical relevance,³⁴ neopentylglycol phenylboronic ester 2a (PhB(neo)) (1.5 equiv) as the nucleophilic partner, and potassium methoxide (1.25 equiv, according to the previously method) as the base.²⁹ Various solvent classes including amide, nitrile, ester, aromatic, ether, and protic/aprotic, cobalt(II) sources such as CoCl₂•6H₂O or CoBr₂, and ligand classes were evaluated. Reaction success was measured according to relative area percent of C(sp²)–C(sp³) cross-coupled product 3a as determined by ultra-performance liquid chromatography–mass spectrometry (UPLC-MS).³⁵ A key design feature in the screening procedure was the use of 5 mol% of the cobalt(II) source, significantly lower than the 15 mol% loading required for the previously reported cobalt–diamine catalyst.²⁹ Among the evaluated ligand classes, ligands previously reported for cobalt-catalyzed cross-coupling such as bipyridine-/terpyridine-type ligands^24,26a,36 or phosphines (see SI) were generally inadequate (Table 1).¹⁶ Consistent with previous studies,²⁹ the use of diamine-type ligands generally resulted in modest to good yield of 3a (up to 48% UPLCMS area) (Table 1).^20,23,37 Another class of ligand that exhibited promising performance was 8-hydroxyquinolines (QNOL), which gave the desired product ( 3a ) in up to 42% UPLCMS area (Table 1).³⁸ Notably, this class of ligands likely forms “L,X-type” interactions where deprotonation of the phenol generates an anionic donor (vide infra). Related N,O ligands such as amino alcohols have also been successful in nickel-catalyzed Suzuki–Miyaura arylation reactions of unactivated alkyl electrophiles.³⁹

Table 1.

Evaluation of ligands by high-throughput experimentation (HTE)^a

^aReactions performed on 0.010-mmol scale using CoX₂ (5 mol %) and various polar aprotic solvents at 60 °C. Yields represent relative UPLC-MS area percent. See SI for details.

With this preliminary ligand data in hand, a more comprehensive ligand comparison was performed in batch on 0.25-mmol scale (Table 2A). Reactions were conducted using N-Boc-4-bromopiperidine (1b), 3-methoxyphenylboronic acid neopentylglycol ester (2b), CoCl₂ (5 mol %), and KOMe (1.5 equiv) in DMA at 60 °C, and analyzed by GC to determine the yield of desired product (3b) as well as the yields of alkyl bromide-derived side-products, namely reduction (4a) and elimination (5a). The head-to-head comparison of the preferred diamine ligand from the first-generation catalyst, DMCyDA (L1) and the potential “L,X-type” 2,5,7-tri-Me-QNOL (L2) demonstrated that both ligands delivered the desired product (3b) in comparable yield, within GC-FID error (57% and 61%, respectively) (Table 2B). Notably, the distribution of side products arising from catalysts derived from L1 or L2 differed, with L1 forming 4a and 5a in 18% and 13% yield, respectively, and L2 forming these side-products in 4% and 29%, respectively. These data suggest that the QNOL-based ligand is more effective in suppressing formation of the reduced side-product (4a) and that formation of 4a is likely catalyst-dependent (i.e., dependent on (i) the rate of cobalt-catalyzed product formation and (ii) the ability of the cobalt catalyst to “stabilize” cage-escaped radicals; vide infra). By contrast, formation of 5a is likely a background reaction arising from interaction of the base with the electrophile that is independent of the cobalt catalyst (vide infra). These results suggested that L,X-type ligands may be able to maintain high selectivity for 3b versus 4a, and that in an optimized case, could accelerate the reaction rate to outpace background elimination (5a).

Table 2.

Optimization of ligand structure^a

^aReactions performed on 0.10-mmol scale using 5 mol% CoCl₂ and 5 mol% ligand; yields determined by GC-FID using n-dodecane as an internal standard. Mass balances >100% reflect error of GC yields. DMA = N,N-dimethylacetamide; neo = neo-pentylglycol.

Based on these findings, other structurally similar and potential L,X-type ligands were explored and evaluated for the yield of desired product (3b) relative to elimination byproduct (5a) (Table 2C). Within the series of QNOL ligands, minimal formation of the reduced side product (4a) was consistently observed (typically <5%; in all cases <10%, see SI) with the remaining mass balance usually being unreacted starting material. QNOL ligands without substituents at the 2- and 7-positions performed poorly (L3, L4) while 2,7-diMe-QNOL (L5) displayed good reactivity (56%). 2,4,5,7-TetraMe-QNOL (L6) performed better than 2,5,7-triMe-QNOL (L2) (74% vs. 61%). Methylation of 5,7-diMe-QNOL (L7) resulted in comparable yield (8%) and selectivity to the background reaction in the absence of ligand (Table 2B), supporting the role of phenol deprotonation for improved catalytic performance.

Evaluation of other potential L,X-type ligands was conducted to determine the general role of this structural type on catalytic performance and to explore if ligand more readily synthesized and modular than QNOL were effective. A benzylidene-ortho-aminophenol (L8) produced no cross coupled product (0%) and a pyridine-2-carboxylate (L9) furnished the desired product in modest yield (32%). While ortho-2-imidazolylphenol (L10) and ortho-pyrazolylphenol (L11) did not outperform QNOL ligands, ortho-2-pyridinylphenol (L12) gave the desired product in an improved 79% yield by GC-FID. Inspired by reports in the polymerization literature that phenoxy-imine-based (FI) ligands performed similarly to QNOLs,⁴⁰ the former class of ligands was explored for cobalt-catalyzed C(sp²)–C(sp³) Suzuki-Miyuara cross coupling. This ligand class has been previously studied in the context of cobalt-catalyzed C(sp²)–C(sp²) Suzuki–Miyaura cross-coupling,⁴¹ and modest reactivity with yields up to 47% as judged by GC-MS using activated lithium borate nucleophiles were reported.⁴²

FI ligands derived from strongly electron-donating (L13) or electron-withdrawing (L18) anilines performed poorly for C(sp²)–C(sp³) coupling; however a range of FI ligands with electron-neutral or moderately electron-withdrawing N-aryl substituents (L14–L17)⁴³ yielded product 3b with high yield (81–90%) and selectivity. Under these same conditions, a control reaction in the absence of CoCl₂ resulted in full conversion to side-product 5a, indicating that formation of 5a arises from background elimination rather than β-hydride elimination from a cobalt complex with an FI ligand (see SI). Notably, extension of the FI-type ligand to the related tetradentate salphen⁴⁴ ligand class (L19) resulted in no formation of cross coupled product. Variations to the FI were also explored on the phenol ring; however no change to steric or electronic structure of the phenol resulted in a discernable improvement in performance (see SI).

To gain insight into the reactivity of diamine and FI ligands under optimized conditions, experiments were performed to probe alkyl radical behavior (Scheme 2). Radical cyclization was examined (Scheme 2A). Namely, for a cage-escaped radical derived from the activation of 6-bromo-1-hexene (1c), the proportion of linear (L) (6) to cyclized (C) (7) cross-coupled product will reflect the lifetime of the free radical in solution, because the radical can undergo 5-exo-trig cyclization (k_cyclization ≃ 1 × 10⁶ s⁻¹)^45,46 to yield the cyclized intermediate. In related experiments, Biswas and Weix⁴⁷ and Shenvi and coworkers⁴⁸ demonstrated that, in the presence of a nickel or cobalt catalyst that can recombine with an alkyl radical, the proportion of radical cyclization will change as a function of catalyst loading; higher concentrations of catalyst tend to result in more recombination and less 5-exo-trig radical cyclization, according to the recombination-propensity of the catalyst. Employing either 4-CF₃-Ph-FI (L17) or DMCyDA (L1), L/C ratios were measured with varying catalyst concentrations (2–20 mol%) and a 1:1 CoCl₂:ligand ratio (Scheme 2A). In both cases, a positive slope was obtained, indicating that both L17- and L1-supported cobalt catalysts undergo recombination with cage-escaped alkyl radicals as a function of catalyst loading. However, the L17 data set ratios were shifted upwards compared to those for L1 across the same catalyst concentrations (b = 0.112 and 0.049, respectively), indicating that radical capture is more efficient with L17 than with L1 under otherwise identical conditions. It should be noted that a coordination–insertion mechanism cannot be excluded based on these observations.

Scheme 2. Experiments to probe the intermediacy of alkyl radicals.

^aSee SI for details; ^bDetermined by GC-FID; ^cDetermined by ¹H NMR spectroscopy.

These data suggest that an important catalytic feature of L,X-type FI ligands versus diamine ligands is the ability to productively capture alkyl radicals, an advantageous property for reactions involving C(sp³) coupling partners.⁴⁹ In other chemistry, FI ligands are common for reactions involving alkyl partners, e.g. nickel-catalyzed ethylene polymerization.⁵⁰ Related chelating ligands with X and L donors for cobalt, such as salen or cobaloxime, are known to participate in recombination with alkyl radicals to act as stabilizing “reservoirs” of alkyl fragments in catalytic reactions.^48,51 The reduced side product (4) likely forms from a cage-escaped radical (R•) that undergoes HAT from solvent (SH) to yield RH and S•. Yield of 4a (Table 2) is expected to be dependent on radical lifetime and likely the origin as to why FI ligands (L14–L17) tend to produce lower quantities of the reduced side-product than DMCyDA.

The formation of alkyl radicals was corroborated with a cyclopropylcarbinyl radical probe (Scheme 2B). When bromide 1d was used as the substrate, the corresponding arylated product (8) was formed in moderate yield with exclusive selectivity for ring-opening.

With an optimized metal-ligand combination and reaction conditions, the scope of the cobalt-catalyzed C(sp²)–C(sp³) Suzuki-Miyaura cross coupling was explored (Table 3). Standard conditions employed 5 mol% of each of the FI ligand and CoCl₂ along with 1.5 equivalents of KOMe in 0.25 M DMA at 60 °C. Notably, FI ligands of interest were routinely prepared from commodity chemicals on decagram scale by condensation and collection by filtration, leading to large quantities of bench-stable materials (Table 3, photos). For alkyl bromide coupling partners, a variety of secondary substrates were tolerated, including hydrocarbon (3e–3f, 3h), ether-containing (3d, 3j, 3l, 3n), and carbamate-containing (3c, 3g, 3i, 3m) substrates. A TBS-protected alcohol was compatible (3k). Primary electrophiles were also compatible (3o–3q), including an acyclic acetal (3p), albeit with higher catalyst loadings (10 mol% each of CoCl₂ and L14), which resulted in less formation of the elimination side-product compared to the standard 5 mol% loading. Substrates where elimination was facile (3r) yielded lower amounts of product. Among compatible alkyl substituents, a few are also of note due to their beneficial physiochemical properties,⁵² namely oxetane (3l),⁵³ azetadine (3m),⁵⁴ and an oxaspirocycle (3n).⁵⁵

Table 3.

Scope of the cobalt-catalyzed C(sp²)–C(sp³) Suzuki-Miyaura cross coupling reaction^a

^aReactions performed on 0.125–0.25 mmol scale; yields are isolated. For incompatible and supplementary substrates, see SI.

^bDetermined by ¹H NMR analysis of the crude reaction mixture.

^cUsing 10 mol% CoCl₂ and 10 mol% 4-OMePh-FI L14.

^dSee SI for details.

With respect to arylboronic esters, the reaction was compatible with electron-donating (3s, 3t), electron-neutral (3u–x, 3z, 3ab) and electron-withdrawing aryl substituents (3c, 3g, 3k, 3q, 3aa). A ketone-containing substrate furnished the coupled product in 34% yield (3y). Substrates with one ortho substituent (3ac–3ad) were compatible, including an ortho-chloro-substituted arylboronic acid (3ad). A cyclic-acetal-substituted aryl nucleophile gave the desired product in good yield (3af). Extended aromatic substrates (3ae, 3ag) also worked. Some heteroarylboronic esters were also viable, including 3-pyridyl (3ah), 2-furanyl (2ai), 2-benzofuranyl (2aj), and N-Me 2-indolyl (3ak). Substrates incompatible with the current method and gave low yields (<10%) of product and include those bearing functional groups such as benzaldehyde, primary alcohol, and phthalimide, among others (see SI for specific examples).

The method was also viable for gram-scale synthesis (Scheme 3). With 37.5 mmol of arylboronic ester 2b and 25 mmol of bromocyclohexane 1d, standard reaction conditions were employed with the modifications that instead of CoCl₂, 5 mol% of less hygroscopic CoBr₂ was used to allow weighing on benchtop, and that the reaction was stirred in an 80 °C oil bath to ensure adequate heat-transfer to the round-bottom flask (cf. Table S9). These conditions yielded 3.37 g of the desired cross-coupled product 3al, representing 71% yield.

Scheme 3.

Application of the method on 25-mmol scale

Part II. Synthesis of Phenoxyimine Cobalt Complexes and Relevance to Catalysis.

The improved catalytic performance of the phenoxy-imine based cobalt catalysts prompted more detailed studies into the coordination chemistry of this ligand class, elucidation of the importance of the “L,X” interaction, and understanding the oxidation state of the cobalt precursor. These studies were of importance as results initially obtained when translating reactions with FI ligands from HTE to batch were often variable. Specifically, the productivity and yield of the reactions depended on the order of addition of the reagents (Table 4). Optimized performance was observed when the aryl boronate (2) was premixed with the FI ligand (L) and KOMe, followed by mixing with CoCl₂ for 1–2 minutes, then addition of alkyl bromide (1) and immediate heating of the reaction mixture. Using this procedure, the cross-coupled product (3) was obtained in 83% yield as judged by GC after 16 hours (Entry 1).

Table 4.

Impact of the order of addition of reagents on yield and selectivity for cobalt-catalyzed cross coupling in batch and HTE^a

Entry	Setup	Order of addition to DMA solution	1 (%)b	3 (%)b	4 (%)b	5 (%)b
1	batch	(i) 2 + L + KOMe; (ii) Cocl2; (iii) 1	0	83	2	20
2		(i) L + CoCl2; (ii) 2 + KOMe; (iii) 1	8	2	0	87
3		(i) 2 + L + KOMe + CoCl2; (ii) 1	27	41	2	44
4		(i) KOMe + L;c (ii) 2; (iii) CoCl2; (iv) 1	8	83	2	11
5d	HTE	(i) L + CoCl2; (ii) 1 + 2; (iii) KOMee	4	43	1	23

^aSee SI for details

^bDetermined by GC-FID

^cUsing preformed FI-phenoxide

^dPerformed by HTE, yields represent relative area percent as determined by UPLC-MS

^eUsing KOMe (1.25 equiv) prepared from MeOH (2 equiv) and KHMDS (1.25 equiv).

This order of addition was optimal as premixing 2 and KOMe establishes an equilibrium with the borate species (eq. 1, above), minimizing competing elimination upon addition of the electrophile. If L and CoCl₂ were premixed prior to the addition of 2 and KOMe, minimal yield (~2%) of 3 was observed with the elimination product 5 accounting for most of the mass balance (entry 2). Combination of all of the reagents prior to addition of 1 also resulted in a reduced yield of 41% (Entry 3). Stirring KOMe and L to promote deprotonation also resulted in an 83% yield, identical to the optimized conditions (Entries 4 and 1). It should be noted that in HTE procedures, the KOMe base was added last and a lower yield of 43% was obtained (Entry 5). Given the acidity of the FI ligand (pK_a of MeOH = 29.0 in DMSO⁵⁶; pK_a of ArOH ≃ 13–19 in DMSO⁵⁷), it was expected that KOMe would promote rapid deprotonation. Experimentally, mixing 4-F-Ph-FI (L16) with KOMe DMSO-d₆ resulted in >95% conversion to the anion as judged by ¹⁹F NMR spectroscopy.⁵⁸

With a reproducible set-up for the reaction, the impact of the L:CoCl₂ ratio on the performance and reproducibility of the catalytic cross coupling reaction was studied (Table 5). Namely, reactions with FI ligands (e.g., 4-H-PhFI, L15) were sensitive to higher ligand loadings. For example, an optimal reaction using L15 (5 mol %, 1:1 L:CoCl₂) gave the desired product in 83% yield (Entry 1), whereas a reaction using double the amount of L15 (10 mol %, 2:1 L:CoCl₂) resulted in an inefficient reaction (10% yield), with the major product being elimination side-product 5a (70%) (Entry 2).

Table 5.

Conversion and selectivity according to ligand loading^a

Entry	mol% L15	1b (%)b	3b (%)b	4a (%)b	5a (%)b
1	5	0	83	2	20
2	10	20	10	0	70

^aReactions performed on 0.10-mmol scale.

^bDetermined by GC-FID.

To better understand the effects of the L:CoCl₂ ratio, the interactions of FI ligands with cobalt precursors were investigated. With the sterically hindered 2,6-di-i-PrPh^di-t-BuFI (9) ligand, deprotonation with one equivalent of NaH in DMA followed by addition to a DMA slurry containing one equivalent of CoCl₂ produced a green solution (Scheme 4A). Stirring for 16 hours followed by treatment with Et₂O and recrystallization from a DMA-Et₂O mixture at −30 °C produced red crystals suitable for X-ray diffraction. A representation of the solid-state structure is shown in Scheme 4B, which identified the product of the reaction as the four-coordinate, idealized tetrahedral cobalt(II) complex (τ₄ = 0.91), (2,6-di-i-PrPh^di-t-BuFI)CoCl(DMA) 10, containing a deprotonated FI ligand coordinated in an “L,X” fashion along with a molecule of DMA to complete the coordination sphere. The ¹H NMR spectrum in benzene-d₆ of paramagnetic 10 exhibits broad signals with chemical shifts ranging from 451 to −34 ppm.⁴² This is a rare example of a monomeric cobalt complex supported by an FI ligand.^59,60 The structure of 10 provides insight to at least one role of the DMA solvent; namely, to serve as a ligand to stabilize the resulting [(FI)Co] precatalyst.

Scheme 4. Synthesis of (FI)Co(Cl)(DMA)^a.

^aSee SI for details; ^bRepresentation of the molecular structure of (2,6-di-i-PrPh^di-t-BuFI)CoCl(DMA) at 30% probability ellipsoids; hydrogen atoms omitted for clarity. Selected bond distances (Å) and angles (°): Co1–O1 1.9025(12), Co1–O2 1.9882(13), Co1–N1 1.9864(14), Co1–Cl1 2.2258(5); O1–Co1–N1 96.12(5), Cl1–Co1–O2 118.34(4).

Compound 10 was stable in DMA solution for months as determined by ¹H NMR spectroscopy but decomposed upon attempted isolation. Removal of the solvent by high-vacuum resulted in decomposition of 10 after 6 hours. Dilution of the DMA stock solution with MeCN, MeOH, CDCl₃ or benzene-d₆ also produced decomposition of 10 within 1 hour as determined by ¹H NMR spectroscopy, likely a result of DMA dissociation to form an unstable three-coordinate cobalt compound. Attempted synthesis of 10 using stoichiometric DMA in THF as the solvent resulted in isolation of the corresponding bis(chelate) compound, (2,6-di-i-Pr-Ph^di-t-BuFI)₂Co, which was characterized by independent synthesis (see SI). This bis(chelate) species was also identified as a side-product formed during the synthesis of 10 that was obtained in variable amounts (0–20%).

The synthesis of related compounds with catalytically optimal ligands L14–L17 was targeted (Scheme 5). Deprotonation of ligands L14–L17 was accomplished by addition of NaH, followed by treatment with CoCl₂ and a stoichiometric quantity of DMA.⁶¹ The resulting reaction mixtures were stirred for 16 hours, and green solids identified as 11a–d were isolated. Single crystals were obtained by vapor diffusion recrystallization of 11c (R = F) from DMF–Et₂O and established the formation of a dimeric cobalt complex, [(4-F-Ph-FI)CoCl(DMF)]₂ (12c) (Figure 1). Each subunit contained an idealized trigonal bipyramidal cobalt (τ₅ = 0.55) ligated by an L,X FI ligand, a chloride and a molecule of DMF. A bridging oxygen from an adjacent FI ligand and the imine nitrogen of the FI define the axial positions.⁶² The presence of DMF in the crystal structure also established the lability of the coordinating solvent.

Scheme 5. Preparation of [(FI)CoCl(DMA)]₂.

^aInseparable from bisligand chelate (4-F-PhFI)₂Co (15c). Sample represents an 80:20 molar ratio of 11c:15c; yield is with respect to desired compound.

Figure 1.

Representation of the molecular structure of [(4-F-Ph-FI)CoCl(DMF)]₂ 12 with 30% probability ellipsoids; hydrogen atoms omitted for clarity. Selected bond distances (Å) and angles (°): Co1–O1 1.9609(12), Co1–O2 2.0473(11), Co1–O1’ 2.1537(12), Co1–N1 2.0651(13), Co1–Cl1 2.3069(9); O1–Co1–N1 92.31(4), Cl1–Co1–O2 113.77(3).

The corresponding dimers with different aniline para-substituents (OMe, H, F, CF₃, 11a–d respectively) were obtained in comparable yield (34–38%) (Scheme 5). These compounds were stored on benchtop and were stable to air and moisture for months. The remaining mass balance for these reactions was attributed to formation of bis-FI chelates, namely (FI)₂Co complexes (vide infra), which were removed from the desired dimer by washing with Et₂O or PhMe. The formation of (FI)₂Co under these conditions may be partially attributed to the poor solubility of CoCl₂ in THF. A coordinating solvent (e.g., DMSO-d₆) could be used instead of THF to improve the yield of dimeric mono-FI compounds 11a–11d as determined by ¹H and ¹⁹F NMR (vide infra), highlighting the importance of coordinating solvent for the synthesis and stability of mono-FI cobalt(II) compounds.

The X-band EPR spectra of the dimeric cobalt compounds 11a–d were recorded in DMA glass at 8 K and exhibited rhombic signals (Figure 2). The simulation parameters were indistinguishable for each entry in the series (Table 6) with notably large g₁ values (5.07–5.29).⁶³ In the solid state, magnetic moments of 6.6–7.1 μ_B were obtained for the dimers (Table 6).⁶⁴ Together, the X-ray, EPR and magnetic moment data all support high-spin (S = 3/2) cobalt(II) for 11a–d.

Figure 2.

X-band EPR spectra of compounds 11a, 11b, and 11d recorded in DMA glass at 8 K. Black = experimental; red = simulated. Collection parameters for 11a: microwave frequency = 9.363 GHz, power = 2.0 mW, modulation amplitude = 4 G. Collection parameters for 11b: microwave frequency = 9.364 GHz, power = 0.06 mW, modulation amplitude = 4 G. Collection parameters for 11d: microwave frequency = 9.367 GHz, power = 0.6 mW, modulation amplitude = 4 G.

Table 6.

EPR simulation parameters and solid-state magnetic moment data for compounds 11a–d

		EPR Simulation Parameters
Compound	FI 4-Ph-N Substituent	S	g 1	g 2	g 3	g strain	μeff (μB)
11a	OMe	3/2	5.07	1.71	3.39	(1.50, 0.71,2.58)	7.02
11b	H	3/2	5.13	1.63	3.29	(1.38, 0.58, 2.60)	7.13
11d	CF3	3/2	5.29	1.76	3.06	(1.45, 0.79, 2.22)	6.59

¹H and ¹⁹F NMR spectroscopy were used to correlate isolated dimer 11d (R = CF₃) with the cobalt compounds generated in situ during precatalyst formation (Scheme 6). Deprotonation of L17 (R = CF₃) with one equivalent of KOMe in DMSO-d₆ followed by addition of one equivalent of CoCl₂ and monitoring the reaction by ¹H and ¹⁹F NMR spectroscopies established dimeric cobalt compound, 13d as the major product (Scheme 6). Repeating this experiment with L16 (R = F) using C₆H₅F as internal standard, also resulted in formation of the corresponding dimer in 80% ¹⁹F NMR yield (see SI). Together, these data suggest [(FI)CoCl(DMA)]₂ is likely the relevant precatalyst for the cobalt-catalyzed C(sp²)–C(sp³) Suzuki–Miyaura cross-coupling procedure.

Scheme 6.

In situ formation of FI–cobalt(II) compound

The synthesis of bis(FI) cobalt(II) complexes was also explored. Stirring two equivalents of neutral ligands L14–L16 with CoCl₂ yielded deep green solids assigned as neutral FI coordination complexes, namely (H-FI)₂CoCl₂ compounds (14a–14c) (Table 7). For compounds 14a–14c (R = OMe, H, F, respectively), moderate to excellent yields of the desired deep green compounds were obtained (50–95%). For 14d (R = CF₃), free ligand (L17) was obtained as the major product, potentially due to the weaker donor ability of the electron-deficient salicylaldimine. Employing L14 and CoCl₂•6H₂O and stirring in refluxing EtOH yielded the same product (4-OMe-H-FI)₂CoCl₂, 14a (80%) (see SI).^41,42 The solid-state structure of 14a was confirmed by X-ray crystallography (Figure 3), establishing an idealized tetrahedral geometry at the metal (τ₄ = 0.89). Solid-state magnetic data were collected for the isolated compounds and values of μ_eff between 4.1 and 4.3 μ_B were obtained (Table 7), consistent with high spin (S = 3/2) cobalt(II) center and the observed solid-state structure of 14a. Attempts to prepare similar neutral bis(chelate) cobalt compounds using FI ligands bearing substituents ortho to the phenol resulted in recovery of free ligand, likely due to steric interference preventing O–Co coordination.

Table 7.

Bis(chelate) Cobalt Compounds, (H-FI)₂CoCl₂ Compounds 14a–d with Protonated Ligands.

Compound	H-FI 4-Ph-N Substituent	Yield (%)a	μeff (μB)b
15a	OMe	95	4.33
15b	H	95	4.32
15c	F	50	4.14
15d	CF3	<20	n.d.

^aIsolated yields.

^bMeasured in the solid state.

Figure 3.

Representation of the molecular structure of (4-OMe-Ph-H-FI)₂CoCl₂ 14a at 30% probability ellipsoids; hydrogen atoms omitted for clarity. [H1] and [H2] have been located. Selected bond distances (Å) and angles (°): Co1–O1 1.9596(12), Co1–O2 1.9432(12), Co1–Cl1 2.273(6); O1–Co1–O2 117.79(5), Cl1–Co1–Cl2 112.43(14).

The ¹H NMR spectra of (H-FI)₂CoCl₂ compounds 14a–c in DMSO-d₆ did not exhibit signals in the paramagnetic region of the spectrum. Instead, slightly broadened signals nearly identical to the free ligand were observed (see SI). These observations support that in the presence of coordinating solvent, (H-FI)₂CoCl₂ compounds 14a–c are in equilibrium with free ligand and solvated [CoCl₂]_x(S)_y. These bis(ligated) cobalt compounds (14a–14c) were insoluble in noncoordinating solvents, prohibiting further characterization in solution.

The synthesis of bis(chelate) cobalt(II) derivatives containing anionic FI ligands, (FI)₂Co (15a–d), was accomplished by deprotonating ligands L14–L17 with NaH, followed by addition of CoCl₂ (Scheme 7A). (FI)₂Co compounds 15a–d were paramagnetic orange-red solids, consistent with known (FI)₂Co compounds.^62,63,65 A bulkier derivative derived from ligand 9 was also synthesized using the same procedure (see SI).

Scheme 7. Synthesis of (FI)₂Co Compounds (15a-d).

^aSolid-state structure of (4-OMePhFI)₂Co 15a at 30% probability ellipsoids; hydrogen atoms omitted for clarity. Selected bond distances (Å) and angles (°): Co1–O2 1.8998(11), Co1–N1 2.0039(12); O1–Co1–N2 96.96(5). ^bX-Band EPR spectrum of compound 15a at 10 K in DMA glass. Collection parameters: microwave frequency = 9.363 GHz, power = 0.02 mW, modulation amplitude = 4 G. Simulation parameters: S = 3/2, g₁ = 4.21, g₂ = 1.99, g₃ = 4.75, g_strain = (1.20, 0.42, 1.58).

A representation of the molecular structure of 15a (R = OMe) determined by X-ray diffraction is presented in Scheme 7B. As expected from other known (FI)₂Co complexes,⁶⁶ an idealized tetrahedral geometry was observed (τ₄ = 0.82). The series of (FI)₂Co compounds (15a–d) were further characterized by low-temperature X-band EPR spectroscopy (e.g., 15a, Scheme 7C; see SI). In each case, rhombic spectra were obtained, consistent with metal-centered radical character. Together, these data support these complexes being high-spin (S = 3/2) cobalt(II).⁶³

The role of base was evaluated with respect to the protonation state of the FI chelate as premixing the ligand and CoCl₂ before addition of KOMe was detrimental for catalysis (cf. Table 4, Entry 2). The 4-F-Ph-FI ligand (L16) was added to CoCl₂•6H₂O and stirred in refluxing EtOH for 12 h (Scheme 8). This produced a heterogeneous mixture composed of a green EtOH solution and green precipitate, indicating formation of 14c, consistent with observations made during the independent synthesis of this class of compounds (eq. 2). Subsequently, two equivalents of KOH were added, and the reaction was stirred at 80 °C for 6 hours. This resulted in the formation of 15c, obtained in 47% isolated yield, confirming that 14c was converted to 15c in the presence of base.

Scheme 8.

Conversion of (H-FI)₂CoCl₂ to (FI)₂Co

The impact of FI:CoCl₂ stoichiometry was evaluated with respect to precatalyst formation and thermodynamics in solution, as 2:1 FI:CoCl₂ stoichiometry was detrimental to catalysis compared to 1:1 stoichiometry (cf. Table 5, entries 3–4). Namely, the series of (FI)₂Co compounds 15a–d was separately added to one equivalent of CoCl₂ in DMSO-d₆ and stirred at room temperature for one hour (eq. 1). Analysis of the reaction mixtures by ¹H NMR spectroscopy revealed the mono-FI dimers 13a–d to be the major components in all cases, and (FI)₂Co chelates 15a–d to be minor components. Analysis by ¹⁹F NMR spectroscopy for the reactions employing 15c (R = F) and 15d (R = CF₃) revealed 63% and 88% conversion to the mono-FI compounds 13c and 13d, respectively. These data indicate that FI:CoCl₂ stoichiometry in a coordinating solvent like DMSO-d₆ affects the formation of mono- versus bis-FI cobalt(II) compounds. Empirically, the results in DMSO-d₆ contrast with the protocol used to synthesize [(FI)CoCl(DMA)]₂ compounds 11a–d that utilizes THF as solvent, for which a significant product is undesired (FI)₂Co (Scheme 6). These data imply that coordinating solvent like DMA may help favor mono-FI cobalt intermediates, especially in catalysis.

(1)

The competence of the isolated [(FI)Co] coordination complexes in the C(sp²)–C(sp³) Suzuki–Miyaura cross-coupling reaction of 1b and 2b was evaluated using 5 mol% of each isolated cobalt compound (Table 8). These reactions were set up and performed on the benchtop using standard Schlenk techniques; besides 10, all complexes were air-stable over a period of months. Modest cross coupling reactivity was observed for reactions conducted with a stock solution of model complex 10 in DMA under otherwise standard catalytic conditions. The 12% yield of 3b using 10 was the same as when the in situ catalyst generation method was used, highlighting the detrimental effect of bulky substituents ortho to the FI phenol moiety, with respect to catalytic performance; the remaining mass balance in this reaction was predominantly elimination (5b). Dimers [(FI)CoCl(DMA)]₂, 11a–d, were all competent in catalysis (78–95% yield), consistent with the yields for in situ precatalyst generation. These data support that the isolated cobalt compounds lead to on-cycle species upon exposure to the components of the cross-coupling reaction. In contrast, the bisligated (H-FI)CoCl₂ compounds with neutral H-FI ligands, 14a–c, performed poorly in catalysis. Finally, with chelate complexes bearing two deprotonated FI ligands (15a–d), no turnover was observed.

Table 8.

Performance of isolated phenoxyimine cobalt complexes in catalytic C(sp²)–C(sp³) cross coupling^a

^aYields determined by GC-FID using n-dodecane as internal standard. The designation of “5 mol%” is with respect to equiv of [Co] (5 mol% compound loading for monomers, 2.5 mol% compound loading for dimers 11a, 11c, and 11d)

^bUsing 5 mol% of the corresponding FI ligand and 5 mol% CoCl₂ instead of the isolated cobalt complex.

From the experimental observations and synthesis and evaluation of isolated compounds, a unified picture for pathways leading to active and inactive precatalysts is presented in Scheme 9. Neutral FI ligands (L14–L17) undergo deprotonation with KOMe to yield potassium phenoxides 16a–d. Treatment of phenoxides 16a–d with one equivalent of CoCl₂ generates [(FI)CoCl(DMA)]₂ 11a–d, which are active precatalysts for C(sp²)–C(sp³) Suzuki–Miyaura cross-coupling. Treatment of ligands L14–L17 with half an equivalent of CoCl₂ prior to deprotonation results in an equilibrium with neutral (H-FI)₂CoCl₂ compounds 14a–d in coordinating solvent such as DMSO-d₆. Compounds 14a–d do not produce active catalysts as treatment with KOMe resulted in rapid formation of (FI)₂Co complexes 15a–d that are not catalytically competent. The bis(chelate) cobalt complexes 15a–d were also formed if excess phenoxide (16a–d) is added to CoCl₂ (middle arrow) or to dimers [(FI)CoCl(DMA)]₂ 11a–d. However, the treatment of (FI)₂Co chelates 15a–d with stoichiometric CoCl₂ in a coordinating solvent yields the dimers [(FI)CoCl(DMA)]₂ 11a–d, albeit with lower conversion than the direct treatment of CoCl₂ to phenoxides 16a–d. Together, this picture highlights the importance of coordinating solvent and of the FI:Co stoichiometry to access precatalysts for cobalt-catalyzed C(sp²)–C(sp³) Suzuki–Miyaura cross-coupling.

Scheme 9.

Formation of catalytically active and inactive phenoxy(imine) cobalt complexes.

CONCLUDING REMARKS

A readily synthesized and modular class of “L,X”-type ligands for cobalt-catalyzed C(sp²)–C(sp³) Suzuki–Miyaura cross-coupling has been found to enable efficient C–C bond formation and a host of products were isolated in synthetically useful yields. A Studies into the coordination chemistry established a well-defined mono-FI cobalt(II) precatalyst and demonstrate the importance of FI:CoCl₂ stoichiometry, order of addition, and coordinating solvent (DMA), which solvent acts as L-type ligand to stabilize catalytically relevant cobalt compounds.