Generation of Axially Chiral Fluoroallenes through a Copper-Catalyzed Enantioselective β-Fluoride Elimination

Abstract

Herein we report the copper-catalyzed silylation of propargylic difluorides to generate axially chiral, tetrasubstituted monofluoroallenes in both good yields (27 examples >80%) and enantioselectivities (82–98% ee). Compared to previously reported synthetic routes to axially chiral allenes (ACAs) from prochiral substrates, a mechanistically distinct reaction has been developed: the enantiodiscrimination between enantiotopic fluorides to set an axial stereocenter. DFT calculations and vibrational circular dichroism (VCD) suggest that β-fluoride elimination from an alkenyl copper intermediate likely proceeds through a syn-β-fluoride elimination pathway rather than an anti-elimination pathway. The effects of the C1-symmetric Josiphos-derived ligand on reactivity and enantioselectivity were investigated. Not only does this report showcase that alkenyl copper species (like their alkyl counterparts) can undergo β-fluoride elimination, but this elimination can be achieved in an enantioselective fashion.

Graphical Abstract

INTRODUCTION

The pursuit of synthetic methods to access structurally diverse allenes stems from their applications in both medicinal and materials chemistry as well as their ability to serve as a reactive functional group for further synthetic manipulations.^1,2 These cumulenes can exhibit axial chirality, and most of the naturally occurring allenic compounds that have been isolated are nonracemic.^1a-c Although enantiopure bromoallenes have been discovered in nature (Scheme 1a) and axially chiral bromo-, chloro-, and iodoallenes have been synthesized, the corresponding axially chiral fluoroallenes are largely unknown.^3,4 Specifically, to the best of our knowledge, there is a single report where an enantioenriched tetrasubstituted, axially chiral monofluoroallene has been prepared, albeit in modest enantioselectivity.^4c Due to a lack of general synthetic routes toward this chiral, fluorinated motif, their potential applications have remained unexplored.

Scheme 1.

Allenyl Halide Natural Products and Synthesis of Monofluoro ACAs via Enantioselective β-Fluoride Elimination

An emerging route toward the catalytic synthesis of ACAs has been from prochiral substrates.⁵ This strategy overcomes the requirement for a stoichiometric amount of a chiral auxiliary and/or enantioenriched substrates. Prochiral substrates that have been transformed into ACAs include propargylic electrophiles, 1,3-enynes, terminal alkynes, 1,3-dienes, racemic allenes, and vinyl triflates.⁶ In the case of vinyl triflates, it was demonstrated that β-hydride elimination occurred in an enantioselective fashion, revealing a new mechanistic route for the synthesis of ACAs.^6f Of the classes of ACAs (1,3-di-, tri-, and tetrasubstituted), tetrasubstituted ACAs remain difficult to synthesize in high enantiopurity.⁷

Moreover, the incorporation of a functional group directly attached to ACAs that permits further transformations has gained popularity as evidenced by recent reports of boryl and silyl substituted ACAs.^8,9 Access to tetrasubstituted, boryl, or silyl monofluoro ACAs would permit an array of further transformations that could generate quaternary, fluorine-containing stereocenters by way of axial-to-point chirality transfer.^3a,10

The development of defluorination methods of (poly)-fluorinated compounds has emerged as a complementary strategy to access complex, fluorine-functionalized motifs that have typically been accessed from nonfluorinated substrates.¹¹ Of the metals that catalyze such defluorination reactions, copper has been shown to be exceptionally competent, as evidenced by the numerous reports of catalytic hydrodefluorinations, defluoroborylations, and defluorosilylations of fluoroarenes and fluoroolefins.¹² An emerging trend in this field is the enantioselective defluorination of allylic CF₃ or CF₂R groups, forging stereocenters adjacent to mono- and difluoroolefins (Scheme 1b).¹³ Recently, a rare example of enantioselective defluorination via oxidative addition to a gem-difluoride has been reported as a method to generate products with a fluorine-containing stereogenic unit.¹⁴ We envisioned that a mechanistically different approach to the desymmetrization of difluoromethylene groups, proceeding through an enantioselective β-fluoride elimination reaction, might be employed in the enantioselective synthesis of monofluoro ACAs (Scheme 1c).

To this end, we hypothesized that reaction of propargylic difluorides with a suitable chiral copper nucleophile would form an alkenyl copper species that could undergo an enantioselective β-fluoride elimination to generate tetrasubstituted monofluoro ACAs (Scheme 1c). Potential obstacles toward achieving this transformation included controlling both the regioselectivity¹⁵ and enantioselectivity of the process, avoiding undesired reactivity of the alkenyl copper intermediate,¹⁶ and preventing further silylation of the product.¹⁷ Herein we demonstrate that β-fluoride elimination from an alkenyl copper species is possible and the discrimination of enantiotopic fluorides is a viable elementary process to achieve asymmetric synthesis of ACAs. DFT studies predict that this elimination proceeds through a syn-elimination pathway, which is in contrast to some studies of alkyl copper species that undergo β-fluoride elimination.^{12b,g-i,13b,c,f,g}

RESULTS AND DISCUSSION

We began our investigation by determining if β-fluoride elimination was feasible from a vinyl copper intermediate. In the presence of a suitable base, the borylation of 1a with B₂pin₂ generated the desired boryl monofluoroallene in 46% NMR yield (eq 1); however, attempts at isolating this product were

(1)

unsuccessful.^8b We hypothesized that the corresponding silyl monofluoroallene of 1a would be isolable and gratifyingly discovered that the silylation of 1a with PhMe₂SiBpin (2) led to 1b. After a brief optimization of conditions, over 40 chiral ligands were examined for this transformation. Of the chiral ligands employed, only four gave the desired allene in greater than 20% ee.

Fortunately (R,S)-Josiphos afforded 1b in a modest yield (71%) and promising enantioselectivity (20% ee). A series of Josiphos ligands were generated to evaluate their steric and electronic effects (see Supporting Information Tables S1 and S2) on the reaction. Josiphos ligands containing aryl groups on both phosphorus atoms demonstrated good reactivity and enantioselectivity, specifically when the alkyldiarylphosphorus moiety possessed bulky 3,5-substituted arenes (Table 1, compare entries 2 and 3; Supporting Information Tables S2 and S3). However, once the 3,5-substituents became sterically too demanding, for example, with the TTB derivative, the reactivity dropped significantly (entry 5). Moreover, it appeared that steric rather than electronic factors played a decisive role in determining the enantioselectivity of the transformation (compare entries 3 and 8).

Table 1.

Structural Effect of Josiphos on Transformation^a

entry	R; R'	yield (SM %)b	ee (%)c
1	Me; H	59% (23)	45
2	Me; OMe	31% (14)	44
3	tBu; OMe	41% (<1)	70
4	Mes; H	58% (38)	75
5	TTB; H	3% (79)	nd
6	TMS; H	73% (<1)	75
7	TES; H	66% (<1)	83
8	CF3; H	43% (<2)	71
9	iPr(F7); H	44% (43)	86

^aStandard conditions: 1a (0.10 mmol, 1.0 equiv), 2 (0.135 mmol, 1.35 equiv), CuCl (5 mol %), L (6 mol %), THF (1.0 mL), 65 °C, 24 h.

^bYield was determined by ¹⁹F NMR of crude reaction, using PhF as an internal standard.

^cDetermined by HPLC with a chiral stationary phase.

Further optimization led to simplified reaction conditions (Table 2). By lowering the temperature and changing the solvent, the enantiomeric excess of 1b improved to 90%, but the overall conversion of 1a dropped (entry 1). Increasing the amount of phenoxide base resulted in decoordination of the ligand, determined by ³¹P NMR spectroscopy, and erosion of the ee of 1b.¹⁸ By use of an insoluble source of fluoride (CsF), an increase in the chemical yield of 1b was achieved (83%, entry 2). Possible roles of CsF could be either trapping FBpin¹⁹ and/or releasing CsOAr from ArOBpin.²⁰ Fortunately, the transformation proceeded with CsF as the sole base, removing the chance of phosphine decoordination due to excess phenoxide.²¹ The reaction proceeded well in nonpolar, low coordinating solvents (entries 4–7), and the removal of MeCN led to in an increase in catalytic activity (entries 8, 9). Switching to CuOTf·¹/₂C₆H₆ as the copper source afforded the desired allene in almost quantitative yield (entry 10).

Table 2.

Reaction Optimization^a

entry	Cu	base (mol %)	solvent	yield (SM %)b	ee (%)c
1	Cu(MeCN)4BF4	A (20)	PhMe	50% (49)	90
2	“	A (20), CsF (100)	“	83% (13)	90
3	“	CsF (150)	“	75% (11)	91
4	Cu(MeCN)4BF4	“	THF	39% (48)	74
5	“	“	dioxane	65% (20)	91
6	“	“	cyclohexane	50% (31)	87
7	“	“	MTBE	75% (10)	92
8	“d	“	“	74% (<1)	91
9	Cu(MeCN)4OTfd	“	“	82% (4)	90
10	CuOTf·1/2C6H6	CsF (160)		98% (–)	90

^aStandard conditions: 1a (0.10 mmol, 1.0 equiv), 2 (0.135 mmol, 1.35 equiv), [Cu] (5 mol %), L (6 mol %), solvent (1.5 mL), 32 °C, 24 h. A = NaO(2-OMeC₆H₄).

^bYield was determined by ¹⁹F NMR of crude reaction, using PhF as an internal standard.

^cDetermined by HPLC with a chiral stationary phase.

^dMeCN removed before reaction.

With the optimal conditions established, an alternative synthetic route employing readily available material to prepare propargylic difluorides was developed (Scheme 2). The decarboxylative bromination of difluorocarboxylic acids and the copper-catalyzed Sonogashira cross-coupling of terminal alkynes with difluorobenzyl bromides afforded difluoroalkynes, which were subjected to defluorosilylation under the optimized conditions. A range of functional groups were tolerated in the copper catalyzed transformation, affording the desired allenes 1–24b in high yields (83–98%) and in good enantioselectivities (82–98%) after isolation (Table 3). Notably, alkynes (19b), alkenes (20b), enynes (24b), aldehydes (8b), ketones (9b), propargylic acetates (23b) as well as alkyl and aryl halides (3b and 6b) were tolerated. Coordinating heterocycles (14–16b, 18b, and 25b), amides (7b, 11–13b), and nitriles (5b and 10b) also did not hamper catalysis. Although changing the electronics of the aryl ring slightly decreased the enantiomeric excess of the reaction (1–6b), increasing the steric bulk of the aryl group was well tolerated (25b). Although many functional groups were tolerated, the reaction proved more sensitive to alterations of the substituents directly attached to the allene, which could affect the barrier of silylation of the alkyne (26–27b) or impact the C─F bond strength (28–29b).

Scheme 2.

Improved Synthesis of Propargylic Difluorides

Table 3.

Scope of Asymmetric β-Fluoride Elimination^a

^aStandard conditions: 1–25a (0.20 mmol, 1.0 equiv), 2 (0.27 mmol, 1.35 equiv), CuOTf·½C₆H₆ (6 mol %), L (7 mol %), CsF (1.6–1.8 equiv), 9:1 PhMe:MTBE or MTBE (3.0 mL), 35–45 °C, 24 h. Reported yields are of isolated allene. Enantiomeric excess was determined by HPLC with a chiral stationary phase.

^b1a (6.0 mmol), CsF(25%)–CaF₂ (1.6 equiv), 30 h.

^c1a (2.5–3.0 mmol), R₃SiBpin (1.35–1.45 equiv), Cu (8–9 mol %), 3,5-TripJosiphos (9–10 mol %), CsF(25%)–CaF₂ (1.8–2.5 equiv), MTBE, 28–45°C, 30–48 h.

^dNaO(2-OMeC₆H₄) (30 mol %), CsF (1.0 equiv), PhMe, 27 °C.

To explore the scalability of this method, 1b was synthesized on a 6 mmol scale without a significant loss in enantioselectivity or chemical yield. It was discovered that other silylboranes could be utilized for this transformation when (R,S)-3,5-Trip-Josiphos was employed as the ligand (1b-BnMe₂Si, 1b-CyMe₂Si, and 1b-Et₃Si). These allenyl silanes were synthesized on gram scales with comparable yields and enantioselectivities to 1b-PhMe₂Si.

Determination of the absolute configuration of allene 10b was achieved using vibrational circular dichroism (VCD).²² After a careful conformational search (see Supporting Information for details) at four levels of theory, B3LYP/6-31G(d), B3PW91/6-31G(d), B3LYP/cc-pVTZ, and B3PW91/cc-pVTZ,²³ the resulting conformers were Boltzmann averaged and plotted with a line width of 5 cm⁻¹ to produce the final theoretical spectra. The IR and VCD spectra were then frequency scaled²³ for comparison to the experimental data. Calculations at all four levels of theory matched well, proving the absolute configuration of 10b to be S. Of the four methods employed, the best agreement with experimental data was from the B3PW91/cc-pVTZ level. The comparison of experimental and theoretical spectra was quantified²⁴ using BioTools (Jupiter, FL) CompareVOA software, with high neighborhood similarity for IR (90.4) and VCD (69.6), ESI (enantiomeric similarity index) for VCD (62.9), and a confidence level of 99%. Of particular note was the asymmetric allene C─C─C stretch observed at 1933 cm⁻¹, which was one of many closely correlated bands between experiment and theory. In addition to allene 10b, the absolute configuration of 1b-Et₃Si, 12b, 17b, and 25b were also determined to be S by VCD analysis (see Supporting Information).

On the basis of previous reports regarding CuF^{12c,e,16a,21,25} and copper silyl species,^15b,f,17d,26 we propose the following mechanism for the copper-catalyzed reaction (Figure 1). First, a complex between Josiphos and [CuOTf] undergoes salt metathesis with CsF to generate JosiphosCuF (Cu1).^16a,21,25d σ-Bond metathesis with PhMe₂SiBpin (2) generates JosiphosCuSiMe₂Ph (Cu2) and releases FBpin.^12c-e Subsequent coordination and silylation of the triple bond generates an alkenyl Cu species (Cu3).^15a,e,27 A β-fluoride elimination^12g regenerates Cu1, which is trapped by FBpin,^19,28 2^15c or decomposes the formed allenylsilane (c). As FBpin is more Lewis acidic than B₂pin₂,²⁸ the same is likely true with 2. By using a judicious amount of CsF and a nonpolar solvent, we propose that the precipitation of Cs[F₂Bpin]^19,20 drives this reaction forward. ¹H, ¹⁹F, and ¹¹B NMR studies have identified Cu1, Cu2, LCuF₂Bpin, as well as LCuOH and confirmed the generation of Cu2 from both Cu1 and LCuF₂Bpin (see Supporting Information section 7). Cu2 was also shown to react with alkyne 1a, generating both allene 1b and FBpin. It appears that LCuF₂Bpin acts as a reservoir of Cu^IF, and under catalytic conditions, a monomeric or dimeric CuF was not observed. Over the course of the reaction only Cu2, LCuF₂Bpin, and LCuOH were observed, which converged to Cu2 after the alkyne has been consumed (see Supporting Information section 7). Although the exact structure of Cu1 is unknown, the speciation of Cu1 appears to be both solvent and temperature dependent, in which the former has been observed for other Josiphos copper halide complexes (see Supporting Information section 7).²⁹ On the basis of our experiments, we propose that the silylation of the alkyne and the β-fluoride elimination reactions are the rate- and selectivity-determining steps, respectively. Using (R,S)-3,5-TES-JosiphosCuF₂Bpin as a catalyst, the desired allene 1b was obtained in a similar yield and enantiomeric excess, demonstrating its catalytic competence (see Supporting Information section 7).

Figure 1.

Proposed catalytic cycle.

Density functional theory (DFT) calculations were performed to investigate the reaction mechanism and origin of enantioselectivity of this Cu-catalyzed asymmetric silylation of propargylic difluorides. The DFT calculations were performed at the M06/SDD(Cu,Fe,Cs)-6-311+G(d,p)/SMD-(toluene)//B3LYP-D3(zero)/SDD(Cu,Fe,Cs)-6-31G(d) level of theory using difluoroalkyne 27a and PhMe₂SiBpin (2) as model substrates. The (R,S)-3,5-TMS-Josiphos ligand was used in the DFT calculations for simplicity because the use of this ligand in the ligand screening provided only slightly lower ee than using (R,S)-3,5-TES-Josiphos (Table 1, entries 6 and 7). On the basis of the proposed catalytic cycle, the computed reaction energy profile is shown in Figure 2. The association of FBpin to monomeric LCuF (26) to form a heterodimer (29) is exergonic by 4.0 kcal/mol, suggesting that the more stable complex 29 can be an off-cycle reservoir of Cu^IF. Although the dimerization of LCuF is exergonic by 7.3 kcal/mol, its formation is expected to be less favorable than forming 29 due to the low concentrations of LCuF under catalytic conditions (see Figure S1 for detailed discussions about the equilibrium of 26, 29, and the dimer of LCuF).

Figure 2.

Computed reaction energy profiles of the Cu-catalyzed silylation and asymmetric β-fluoride elimination. Gibbs free energies and enthalpies (in kcal/mol) are with respect to the monomeric copper fluoride 26.

The σ-bond metathesis between monomeric LCuF (26) and PhMe₂SiBpin³⁰ (2) takes place via a four-membered cyclic transition state (TS-1) to form silyl copper intermediate 30 and FBpin. This step requires a low activation barrier of 6.7 kcal/mol with respect to 26 and is exergonic by 24.7 kcal/mol. Migratory insertion of alkyne 27a into the silyl copper (TS-2) gives alkenyl copper species 31. This migratory insertion is highly regioselective for the formation of Cu─C bond at the alkyne terminus adjacent to the difluoromethylene. The transition state leading to the other regioisomer, TS-2′, is 10.9 kcal/mol higher in energy than TS-2. The high level of regioselectivity is due to steric repulsions between the silyl and the more hindered alkyne terminus (C1) in TS-2′ as well as inductive effects of the difluoro substituents that stabilize the building of negative charge at C1 in TS-2. From 31, both syn^12g-i and anti^13b β-fluoride elimination pathways were calculated. The syn-elimination of either of the two diastereotopic β-F in 31 (via TS-3 and TS-4) involves a four-membered cyclic transition state, while the anti-β-fluoride elimination is facilitated by CsF as a Lewis acid (via TS-5 and TS-6). The FBPin-facilitated anti-elimination was also computed and is also less favorable than the syn-elimination (see Figure S3). The syn-elimination pathways require much lower barriers than the anti-elimination, which is in contrast to a computational study by Hoveyda and Torker that suggested the β-fluoride elimination from alkyl copper species favors the anti-pathway due to Lewis acid (i.e., Na⁺) coordination to the F⁻ leaving group and the Bpin group on the substrate.^13b In the present study, the lack of such chelating Lewis-acid coordination in the anti-pathway, the weaker Lewis acidity of CsF, and the strain release effect that alleviates steric repulsions between the SiMe₂Ph group and the Cu in the syn-elimination transition state changed the reaction mechanism to favor the syn-elimination.³¹

Among the three key elementary steps in the catalytic cycle, the alkyne migratory insertion (TS-2) has the highest activation free energy (ΔG^‡ = 19.9 kcal/mol with respect to 30). This finding is consistent with our experimental results that suggest this step being the rate-determining step (vide supra). The enantioselectivity-determining step is the syn-β-fluoride elimination. TS-3, which leads to the (S)-enantiomer of the monofluoroallene product, is 1.9 kcal/mol more stable than TS-4 that leads to the (R)-enantiomer. The predicted enantioselectivity is consistent with the absolute configuration of the product identified by the VCD analysis.

Next, we performed a detailed analysis to investigate the effects of the Josiphos ligand on the reactivity and enantioselectivity of the β-fluoride elimination. Because of the conformational flexibility of the (R,S)-3,5-TMS-Josiphos ligand,³² a careful conformational search was performed for all intermediates and transition states in the catalytic cycle. These calculations revealed at least four different conformers of the 3,5-TMS-Josiphos-supported copper complexes. The two most stable and catalytically active ligand conformations A and B are shown in Table 4 (see Figures S4-S6 for all possible ligand conformations). Ligand conformation A involves a twist-boat-type six-membered ring and is more favorable in the copper fluoride (26), the σ-bond metathesis transition state (TS-1), and the silyl copper intermediate (30). In more sterically encumbered structures, including the migratory insertion transition state (TS-2), alkenyl copper (31), and the β-fluoride elimination transition states (TS-3 and TS-4), ligand conformation B becomes more favorable. This ligand conformation involves a half-chair type six-membered ring, which points the Ar and Ph groups in quadrants I and II away from the Cu center. As such, the bulky SiMe₂Ph group is placed between these unoccupied quadrants to minimize steric repulsions between the ligand and the SiMe₂Ph group on the substrate.

Table 4.

Relative Free Energies of Conformers of (R,S)-3,5-TMS-Josiphos-Supported Copper Complexes and Transition States^a

						syn-β-fluoride elimination
ligand conformation	LCuF (26)	σ-bond metathesis (TS-1)	silyl copper (30)	migratory insertion (TS-2)	alkenyl copper (31)	TS-3	TS-4
A	0.0	6.7	−24.7	2.9	−42.8	−26.7	−28.4
B	4.8	10.8	−23.5	−4.8	−47.1	−32.5	−30.6

^aAll Gibbs free energies are in kcal/mol with respect to the monomeric copper fluoride 26. Bold numbers indicate the favorable ligand conformation.

Ligand conformation B not only stabilizes the silylalkenyl copper species but also plays a significant role in controlling the enantioselectivity of the β-fluoride elimination. The P-phenyl group in quadrant III and the P-3,5-TMS-phenyl group in quadrant IV point toward the Cu center and thus occupy these quadrants. The larger size of the 3,5-TMS-phenyl compared to phenyl indicates that the ligand–substrate repulsions in quadrant IV would be more pronounced than those in quadrant III. Indeed, quadrant diagrams of the β-fluoride elimination transition states (Figure 3) support this hypothesis. In the less favorable transition state TS-4, the phenyl group on the substrate is located in the more occupied quadrant IV, leading to steric repulsion with a TMS group on the ligand. By contrast, in the more favorable β-fluoride elimination transition state TS-3, the much smaller fluoro group is located in quadrant IV, and thus the ligand–substrate steric repulsions are diminished. Next, we performed energy decomposition analysis (EDA)³³ calculations to quantitatively analyze the ligand–substrate noncovalent interactions in TS-3 and TS-4 (see Supporting Information for computational details). The EDA calculations revealed that the dominant factor controlling the enantioselectivity is the Pauli repulsion (i.e., steric repulsion) between the (R,S)-3,5-TMS-Josiphos ligand and the substrate. The Pauli repulsion energy (ΔE_Pauli) in TS-4 is 1.7 kcal/mol higher than that in TS-3 and thus destabilizes the former transition state.

Figure 3.

Origin of enantioselectivity in syn-β-fluoride elimination. Most hydrogen atoms are omitted for clarity. Gibbs free energies and enthalpies are with respect to 31. ΔE_Pauli is the Pauli repulsion energy between the substrate and the Josiphos ligand from EDA calculations.

Finally, we calculated the enantioselectivity-determining syn-β-F elimination transition states of the reaction of the same substrate (27a) catalyzed by a SegPhos-supported Cu complex. The computed enantioselectivity is diminished (ΔΔG^‡ = 0.2 kcal/mol, Figure S7), indicating the C2-symmetric SegPhos ligand is not effective for asymmetric induction. This prediction is consistent with the low ee of 12% obtained experimentally at 65°C with 5 mol % CuCl, 40 mol % sodium phenoxide, and 6 mol % SegPhos ligand. Taken together, these ligand effect analyses revealed the unique roles of the conformationally flexible C1-symmetric Josiphos ligand, where it lowers the activation barrier for the rate-determining alkyne migratory insertion step and improves the enantioselectivity of the β-F elimination.

CONCLUSION

The first copper-catalyzed, enantioselective β-fluoride elimination has been achieved. The resulting monofluoro ACAs represent the first examples of fluorine-containing, chiral tetrasubtituted allenes.³⁴ It is expected that such a motif will find value in pharmaceutical and agrochemical chemistry in addition to being a valuable building block for the generation of more elaborate fluorine-containing stereocenters. DFT calculations of the reaction mechanisms predicted that this elimination occurs in a syn-fashion, which is promoted by strain release of the Z-β-silylalkenyl copper intermediate. The unique roles of the C1-symmetric Josiphos-derived ligand in promoting the reactivity and enantioselectivity were investigated. It is believed that lessons learned from this desymmetrization could be leveraged for the creation of other fluorine-containing stereocenters via defluorination pathways.