Palladium-Catalyzed Carbonylative Difunctionalization of Unactivated Alkenes Initiated by Unstabilized Enolates

Abstract

This report describes the first example of palladium-catalyzed carbonylative difunctionalization of unactivated alkenes initiated by enolate nucleophiles. The approach involves initiation by an unstabilized enolate nucleophile under an atmospheric pressure of CO and termination with a carbon electrophile. This process is compatible with a diverse range of electrophiles, including aryl, heteroaryl, and vinyl iodides to yield synthetically useful 1,5-diketone products, which were demonstrated to be precursors for multi-substituted pyridines. A Pd(I)-dimer complex with two bridging CO units was observed although its role in catalysis is not yet understood.

Graphical Abstract

Pengpeng Zhang and Prof. Timothy R. Newhouse*

A palladium-catalyzed carbonylative difunctionalization of unactivated alkenes is described that is initiated by enolate nucleophiles. Synthetically useful 1,5-diketone products were smoothly transformed into highly substituted pyridines. The method tolerates a range of electrophiles and proceeds under an atmospheric pressure of CO.

Carbonylation,^[1,2] employing carbon monoxide (CO) as an inexpensive and abundant C1 source,^[3] represents one of the most powerful and straightforward strategies for the construction of carbonyl compounds.^[4] A particularly attractive type of carbonylation reaction is the carbonylative 1,2-difunctionalization of alkenes^[5] wherein three bonds are simultaneously formed to generate high value carbonyl compounds from simple starting materials. The mechanistic complexity of this type of multicomponent coupling reaction presents a significant synthetic challenge due to the possibility of undesired reactivity with multiple organometallic intermediates.

A wide variety of reports have demonstrated the development and application of 1,2-difunctionalization of alkenes that install a carbon atom and a carbonyl group across a C=C bond.^[5,6] The majority of these transformations are initiated by electrophiles that react with alkenes and terminate by acyl-metal complexes reacting with alcohols and amines to yield the corresponding esters and amides as products (via head-to-tail cyclization, Scheme 1A). In addition to oxygen- and nitrogen-based nucleophiles, carbon nucleophiles have been employed for carbonylative 1,2-difunctionalization of alkenes to generate ketone products, such as with aryl-boronic acids.^[7] Additionally, different radical-based mechanistic pathways allow for the use of alkyl-iodides as initiating electrophiles with termination by boron-based or acetylene carbon nucleophiles.^[8–10] Recently alternative types of coupling partners, such as B₂Pin₂ (bis(pinacolato)-diboron), aryl-sulfonyl chlorides, and alkyne-appended nucleophiles, have been utilized in carbonylative 1,2-difunctionalization of alkenes.^[11]

Scheme 1.

Transition metal catalyzed carbonylative difunctionalization of unactivated alkenes and potential applications.

In contrast, transformations that employ unstabilized enolates as nucleophilic partners for carbonylative reactions,^[12] especially for carbonylative 1,2-difunctionalization of alkenes, remains a significant challenge. This would be a beneficial transformation to develop, as ketones are abundantly available and ubiquitous, as well as useful for a wide range of applications. Early precedent comes from studies by Hegedus and co-workers,^[13] who reported a few examples of carbo-acylation of olefins with carbanions and CO using a stoichiometric amount of Pd. Those reactions proceed through a stepwise process: the nucleophilic attack of carbanions on Pd-complexed olefins, followed by 1,1-insertion with CO, and finally trapping with methanol or organostannanes. The challenge of rendering this type of mechanistic paradigm catalytic for non-latent nucleophiles rests with the general incompatibility of carbanionic nucleophiles with typical oxidants.^[14]

By inverting well-established head-to-tail carbonylative difunctionalization pathways (Scheme 1A), we designed a tail-to-head pathway, wherein the order of bond formation is reversed through carbo-zincation^[15] of the tethered olefin and followed by a palladium-catalyzed carbonylative coupling with external electrophiles (Scheme 1B). This tail-to-head pathway allows for carbonylative reactions with abundantly available ketones as pro-nucleophiles without the need for pre-activation in a separate step. Furthermore, this pathway produces synthetically useful 1,5-diketones as products that we demonstrate could be transformed into highly substituted pyridines, which are commonly found in bioactive small molecules (Scheme 1C).^[16] The excellent compatibility of a diverse range of aryl, heteroaryl, and vinyl electrophiles, such as iodo-pyridine, -quinoline, -benzoxazole, -benzothiazole, -indazole, -benzofuran, and -thiophene, suggests this method’s utility for various applications.^[17]

Given our previously reported vicinal difunctionalization of alkenes^[18b] employing intramolecular enolate nucleophiles and aryl electrophiles, we initiated the present study by using 1a as a model substrate and iodobenzene as an electrophile in the presence of a Pd catalyst and a CO balloon. Zn(TMP)₂ (bis(2,2,6,6-tetramethylpiperidinyl)zinc) was used as the base for deprotonation of the ketone’s α-proton, and Zn(OTf)₂ was used as an additive to promote the zinc-mediated anionic cyclization in DME (dimethoxyethane).^[18,19]

Using 10 mol% of Pd(OAc)₂ as a pre-catalyst, we screened various monodentate phosphine ligands, and bi-dentate P,P-, P,N-, N,N-ligands. As shown in Table 1, when monodentate phosphine ligands were employed such as PPh₃, the desired product 2a was observed in low yield (13%, along with various byproducts). The use of TMTU (tetramethylthiourea), which was used as an effective ligand for a Pd-catalyzed Pauson-Khand reaction,^[20] gave a cleaner reaction with 46% yield of 2a (entry 2). Bidentate phosphine ligands with increasing bite angle (entries 3–5) improved the yield of the desired product 2a and suppressed undesired β-hydride elimination product 2a’.^[18a] When XantPhos was used as a ligand, the yield of 2a was improved to 81% (entry 5). Unfortunately, this did not prove to be a general ligand, as it was unsuccessful for substrate 1b (32%, see SI). The use of P,N-ligands provided moderate yields (entry 6). Curiously, when Pd(OAc)₂ and Pd/C were used as pre-catalysts without an added ligand, the product could be formed in modest yield (52%, 46%, respectively).

Table 1.

Optimization of reaction conditions.^[a]

^[a]1.5 equiv Zn(OTf)₂, 1.5 equiv Zn(TMP)₂, 10 mol% Pd(OAc)₂, 30 mol% L for entry 1 to entry 2, or 15 mol% L for entry 3 to entry 14, 1 atm CO, 1.5 equiv PhI; Yield of 2a and 2a’ was determined by ¹H-NMR analysis using 1,3,5-trimethoxybenzene as an internal standard.

Ultimately, a breakthrough was realized through the use of N,N-type ligands, which dramatically suppressed the formation of 2a’. We found that L2, with electron-donating substituents on the bi-pyridine, improved the yield to 82% (entry 9). BnBox (benzyl- bisoxazoline) ligand L5 gave the most optimal result, forming the desired product 2a in excellent yields without detection of 2a’ (entry 12). For this substrate, both enantiomers of the ligand provided similar yields (91% for (R,R)-L5 and 89% for (S,S)-L5). Other types of Box ligands or terpyridine (entries 13–14) were also competent for this transformation but proceeded with lower efficiency (see SI for additional optimization with substrate 1b, Table S1). Notably, with L5 as the optimal ligand, replacing Zn(OTf)₂ with other additives, such as ZnCl₂ or LiOTf, resulted in reduced yields (53% or 55%, respectively). Likewise, the use of Zn(TMP)₂∙MgCl₂∙LiCl as the base led to a depreciation in conversion (60%, Table S2), which is likely due to the detrimental effects of both lithium and chloride on cyclization.^[18] This result demonstrates the importance of using salt-free Zn(TMP)₂ for this transformation. Reducing the amount of Zn(OTf)₂ to 20 mol% resulted in full conversion of 1a to 2a, albeit in a somewhat diminished yield (72%).

With the optimized reaction conditions in hand, we began to explore the substrate scope to evaluate the generality of this transformation. Notably within minutes, the reaction to form 2a is nearly complete (>80%, Table S3), but longer times were used to evaluate the scope to ensure complete conversion was uniformly achieved. As shown in Table 2, employing iodobenzene as an electrophile, a panel of ketones with pendant alkenes was subjected to the reaction conditions with reduced catalyst loading (5 mol% Pd(OAc)₂, 7.5 mol% (S,S)-L5). Both 5-membered and 6-membered cyclic ketones could be employed as substrates, and both 5-membered (2a-2j) and 6-membered rings (2k-2m) could be constructed, giving access to various cis-fused bicyclic ring systems. Replacement of the β-methyl group with a 4-OMe-phenyl group gave 2j in 49% yield, while the major byproduct arose from protonation of the anionic cyclization intermediate (33%). We also interrogated the effect of substitution on the alkene. A 2,2-disubstituted alkene worked well in this transformation thereby forming a quaternary carbon center (2h, 59%). Notably, an acyclic methyl ketone was also suitable for this transformation and provided a functionalized cyclohexanone (2n) in 49% yield, 2.5 equiv of Zn(TMP)₂ was used to reduce undesired protonation of the anionic cyclization intermediate. Enone substrates (2g, 2o) were tolerated under these conditions. More excitingly, the bicyclo[3.3.1]nonane (2o) could be constructed albeit in a modest yield (32%), indicating its potential application in the synthesis of complex small molecules.

Table 2.

Substrate scope of carbonylative alkene difunctionalization.^[a]

^[a]Isolated yields.

^[b]2.5 equiv of Zn(TMP)₂ was used.

^[c]XantPhos was used as ligand.

We explored the scope with respect to the electrophile and investigated the electronic effects on the aryl coupling partner using various 4-substituted iodobenzenes for the carbonylative difunctionalization reaction. As shown in Table 2, electron-withdrawing groups, such as –CO₂Me (2aa), –CN (2ab), –NO₂ (2ac) and even –OTf (2ag) at the para-position provided slightly depreciated yields (50–67%) relative to the unsubstituted analog (2a). In addition, iodobenzenes bearing electron-donating groups, such as –OMe (2ad) and –NMe₂ (2ae), were also viable electrophiles, providing good yields (71–75%). Employing meta-substituted 1,3-bis(trifluoromethyl)-5-iodobenzene (2ai) gave 71% yield, and using an ortho-substituted iodothioanisole (2aj) was also successful; a moderate yield (51%) was observed as a result of the steric effect or a possible catalyst poisoning by sulphidation.^[21] Additionally, a variety of heteroaryl electrophiles were explored. Both electron-deficient and electron-rich heteroarenes could be employed, including 3-iodo-9-phenyl-carbazole (2ak), iodoquinolines (2al, 2as), 2-iodo-pyridine (2am), 5-iodo-1-methyl-indazole (2an), 5-iodo-2-methylbenzo[d]-thiazole (2ao), 2-iodobenzofuran (2ap) and 2-iodothiophene (2at).

It should be emphasized that the unique selectivity^[22] of iodides as electrophiles enabled good compatibility with other halogens (2m, 2ah), pseudo-halogens (2ag) and boronic esters (2af), which could be used as pre-installed oxidation states for subsequent transition-metal catalyzed coupling reactions. In addition to arenes and heteroarenes, synthetically versatile vinyl-electrophile (2aq) could also be employed. The diminished yield (35%) may be due to the instability of the electrophile (1-iodovinyl-benzene), the decomposition of which was observed after 2 hours at room temperature. Surprisingly, when 2-iodopyridine (2am) was used as the electrophile under the optimized conditions with L5, the arylation product 2am’ was obtained wherein CO insertion did not occur. However, this side reaction was mitigated to some extent by using Xantphos as a ligand (48% of 2am, 28% of 2am’). The challenge of using 2-iodopyridine as an electrophile may be related to an increased propensity for the electron-deficient aryl group to undergo reductive elimination^[23] or the ability of the pyridine nitrogen to coordinate to palladium.^[24]

Under our standard conditions, a suspicious black suspension is formed upon premixing Pd(OAc)₂ with L5 in DME under an atmosphere of CO. The IR (Infrared Rays) spectrum of the precipitate showed a similar broad absorption at 1911 cm^-1. We attribute it to υ_CO,^[25] and the wavenumber in our spectrum is more consistent with the Pd(I)-dimer complex reported by Ragaini and co-workers^[25b,c] than a Pd(II)-CO complex. They identified several Pd(I)-dimers with two bridging CO units (υ_CO = 1875 cm⁻¹ for [Pd(Me₂Bipy)I]₂(μ-CO)_2, Scheme 2A) by treating neutral Pd(II) complexes in the presence of chelating nitrogen ligands under a high CO pressure (30 bar). Therefore, we tentatively conclude that a Pd(I) dimer complex, [Pd(BnBox)OAc]₂(μ-CO)₂ was formed under our standard conditions (Scheme 2B). Unfortunately, further characterization proved challenging due to its limited stability and the associated challenge purifying it from metallic palladium.

Scheme 2.

Discovery of Pd(I)-dimer complexes and their performance in the reaction.

To provide additional support for this structure, we synthesized [Pd(BnBox)I]₂(μ-CO)₂ by a comproportionation reaction of Pd(0)-BnBox with Pd(II)-BnBox (Scheme 2C),^[25b] which instantaneously resulted in formation of a black precipitate (υ_CO = 1871 cm⁻¹, with iodide rather than acetate as the counterion). The freshly prepared complex was applied as a precatalyst (3 mol%) for the carbonylation reaction under otherwise identical conditions, which gave a comparable yield (98%) of 2a in only 10 mins (Scheme 2C). It is apparent from these considerations that a correlation exists between the putitive Pd(I)-dimer and the formation of the desired product, but its role remains unclear.^[26]

To provide additional insight into the mechanism, a series of straightforward trapping experiments were attempted. In our related study on Ni-catalyzed alkene difunctionalization, cross-coupling was inhibited by the addition of TEMPO ((2,2,6,6-tetramethylpiperidin-1-yl)oxyl).^[18b] However, this Pd-catalyzed carbonylative reaction with added TEMPO in stoichiometric amounts did not impact the yield of the desired product (2a, 89%). The addition of a stoichiometric amount of the nucleophile KOMe under otherwise standard conditions was expected to form PhCO₂Me by intercepting an acyl-Pd species.^[27] However, surprisingly methyl benzoate was not observed and the desired product 2a was formed in typical yield (94%). These two findings argue against the presence of long-lived persistent radicals and a long-lived acyl-Pd(II) species.

Further mechanistic study is warranted to identify the role of the Pd(I)-dimer.^[28] There are several compelling possibilities to consider in future investigations: (1) the novel Pd(I)-dimer may be a resevoir for a Pd(0)/Pd(II) cycle that prevents aggregation of Pd(0) into metallic Pd;^[29] (2) the Pd(I)-dimer may be an on-cycle catalytically active species;^[22,30] or (3) the Pd(I)-dimer may be produced, as an intermediate on route to the formation of catalytically active Pd nanoparticles,^[31] that are differentiated as a function of added N,N-chelating ligands.

The 1,5-diketone products provided by this methodology are broadly useful motifs, including for synthesis of multi-substituted pyridines. Nitrogen-rich heterocycles, especially pyridines, are widely prevalent structures in FDA-approved (Food and Drug Administration) pharmaceuticals and natural products.^[16] Additionally, N-containing ligands play important roles in transition-metal catalysis.^[17] The efficient synthesis of novel skeletons and multi-substituted pyridines remains an important challenge.^[32]

As shown in Table 3, 1,5-diketone products obtained from this methodology can be readily converted to pyridines with NH₄OAc in the presence of FeCl₃∙6H₂O in AcOH. Products were generally formed in good yields (67–80%), including pinocarvone derivatives (3c, 3e) and an androsterone derivative (3d).^[33] The α-position of starting material ketone 2c was epimerized under the acidic conditions, thus giving 3b as a mixture of diastereomers. The TBS (tert-butyldimethylsilyl) group of 2i was deprotected and formed an acetate 3d in situ. Novel skeletons, such as chiral pyridines (3b, 3c), 2-(thiophen-2-yl)pyridine (3g), and 8-(pyridin-2-yl)quinoline (3h) provide entry to a scaffold potentially significant as chiral ligands.^[34] The combination of the carbonylative difunctionalization disclosed herein with condensation to form pyridines represents a new strategy for pyridine synthesis which may find utility for a wide range of disciplines, including drug development.

Table 3.

Application for the synthesis of multi-substituted pyridines.^[a]

^[a]Isolated yield. d.r. was determined by ¹H NMR after aqueous work up.

In conclusion, we developed the first palladium-catalyzed carbonylative difunctionalization of unactivated alkenes initiated by unstabilized ketone enolates. This method enables efficient access to various bicyclic 1,5-diketone architectures with concomitant incorporation of a diverse range of synthetically useful electrophiles. In addition, 1,5-diketone products were further transformed into various multi-substituted pyridines. We expect this strategy to enable significant diversification of existing libraries of complex molecules. Further mechanistic studies are needed to understand the role of the Pd(I)-dimer that is formed, as these studies could be informative for future developments in catalysis at unusual oxidation states.