Introduction

Optically active allene-containing organic compounds have found widespread application in organic synthesis and materials chemistry1,2,3. A synthetic method that can rapidly lead to chiral substituted allenes from simple and readily available starting materials is highly desirable4,5,6,7,8. Concerning efficiency and versatility, transition metal catalysis has emerged as an attractive method to transform propargylic alcohol derivatives into propargylated9,10,11,12,13,14,15,16,17,18,19, 1,3-dienyl20,21, or allenyl products22,23,24,25,26,27,28,29,30,31 through the use of various nucleophiles32,33,34. However, control of regioselectivity in catalytic propargylic substitution reactions has rarely been demonstrated35,36. We hypothesized that it might be possible to tune the desired selectivities through careful choice of phosphine ligand, leading to the formation of phosphoryl derivatives with functional diversity.

Allenylphosphine oxides and phosphinoyl 1,3-butadienes are widely recognized as synthetic intermediates37,38,39, chiral ligands40,41, and biologically active reagents42. Therefore, the development of enantioselective methods to access diverse allenylphosphoryl derivatives rapidly and efficiently is highly desirable. Classical synthetic avenues toward allenylphosphine oxide synthesis usually rely on [2,3]-sigmatropic rearrangement43,44. The process hitherto developed still suffers from severe limitations, such as multistep synthetic sequences, the use of relatively unstable and hazardous phosphorus chlorides, and poor tolerance of functional groups. Enantioselective couplings of propargylic alcohol derivatives with various nucleophiles have been reported to afford chiral allenes (Fig. 1A)45,46,47,48,49,50. Despite extensive efforts, there is no extant catalytic protocol for constructing enantiomerically enriched allenylphosphoryl derivatives from two readily available fragments. On the other hand, transition metal-catalyzed propargylic substitution reactions have been established for carbon-phosphorus bond formation in racemic versions51,52,53,54,55,56. Recently, Sakata, Nishibayashi, and coworkers described the chiral ruthenium complex-catalyzed enantioselective phosphinylation of terminal propargylic alcohols delivering propargylic derivatives with high regio- and enantioselectivity (Fig. 1B)57. The Murakami group reported a seminal nickel-catalyzed reaction of propargylic carbonates with phenols for the synthesis of aryloxy-1,3-dienes21. Therefore, the development of a simplified and stereospecific strategy to access various phosphoryl derivatives would be warranted to further widen the synthetic potential of these scaffolds and build up molecular complexity. However, several challenges have to be overcome, such as (1) the selective generation of different products from identical substrates; (2) catalytic regioselective protocols utilizing ligands rather than substrate control; and (3) the identification of reaction conditions that afford high enantioselectivities.

Fig. 1: Nickel-catalyzed regioselective strategy for 1,3-dienylation and enantioselective allenylation.
figure 1

A Pd-catalyzed enantioselective allenylation. B Ru-catalyzed enantioselective propargylic phosphinylation. C This work: Ni-catalyzed 1,3-dienylation and enantioselective allenylation.

The allenylnickel complex, as revealed by recent elegant reports, was proposed as the key intermediate in nickel-catalyzed asymmetric propargylic substitution reactions58,59,60,61,62. Therefore, we speculated that the nickel-catalytic system with the rational design and development of chiral ligands would meet the challenges above to achieve carbon-phosphorus bond formation in an asymmetric manner. In this work, we report the successful introduction of nickel-catalyzed 1,3-dienylation and enantioselective allenylation, providing phosphinoyl 1,3-butadienes and allenylphosphine oxides with excellent chemo- and regioselectivity (Fig. 1C).

Results

Reaction optimization

In an initial experiment, the coupling reaction of racemic propargylic carbonate 1a and diphenylphosphine oxide 2a was examined (Table 1). The use of (S,S)-Ph-BPE (L1) afforded trisubstituted allenylphosphine oxide 3a and phosphinoyl 1,3-butadiene 4a as a mixture in a ratio of 1:2 with 26% yield, and 3a was obtained with 12% e.e. (entry 1). We next evaluated chiral phosphine ligands that were crucial in terms of both reactivity and selectivity (entries 2–8). Gratifyingly, using phosphine ligand iPr-DuPhos (L3), 1,3-dienyl isomer 4a was isolated in 45% yield as the sole product (entry 3). We next evaluated a series of chiral bidentate phosphine ligands (L4-L8), and a change from iPr-DuPhos (L3) to (S,S)-BDPP (L8) was accompanied by a switch in selectivity from the 1,3-dienyl isomer to the allenyl isomer under otherwise identical reaction conditions (entries 3 versus 8). Further studies focused on the effect of phosphine ligand on the regioselective reaction (entries 9 and 10), and the use of an achiral dcypbz ligand (L10) afforded 3a and 4a as a mixture in a ratio of 1:3. Notably, the yield of 1,3-dienyl isomer 4a increased to 65% with excellent regioselectivity when DMF was employed as the solvent with L3 as the ligand (entry 11). Interestingly, 4a can also be accessed as the main product by using diphenylphosphinic acid as the additive with the dcypbz ligand L10 (entry 12, see Supplementary Information for details). Using L8 as the chiral ligand, 3a was delivered in 34% yield and 48% e.e. when the reaction was performed at 25 °C with the use of quinuclidine as an additive (entry 13). The stereoselectivity was further enhanced by using DCM as the solvent (entry 14, 35% yield, 69% e.e.).

Table 1 Condition evaluation

Subsequently, we designed and synthesized a series of BDPP analogs (L11-L26) to alter the electronic and steric properties and further investigate their influence on the reaction outcomes (Fig. 2)63. It appears that substituents at the para position of the benzene ring linked to the phosphorus atom had some effect on the catalytic properties (L11-L14), and the stereoselectivity was enhanced to 74% e.e. by using L13 with the tert-butyl group. The steric properties of different substitutions at meta positions (L15-L18) showed no improvement in reaction efficiency. Other multiple substituents were then studied (L19-L22), and (S,S)-3,5-(Me)2BDPP (L19) afforded 3a with 76% e.e., albeit with low conversion. Significantly, the desired product 3a could be obtained with 80% e.e. by using the newly synthesized (S,S)-(2-Naph)BDPP (L23). More importantly, BDPP-type ligand L23 could be easily prepared on a gram scale from commercially available starting materials (see Supplementary Information for details). The molecular structure and absolute configuration of L23 were identified by single-crystal X-ray diffraction analysis. No improvement in yield or enantioselectivity was observed with the use of chiral ligand L24. Further optimization studies revealed that the chiral backbone of the phosphine ligands (L25 and L26) showed remarkable effects on the outcome of the reaction, and finally, L23 was identified as the optimal choice.

Fig. 2: Investigation of BDPP-type ligands in enantioselective allenylation.
figure 2

Reactions were conducted by using Ni(COD)2 (10 mol%), L* (12 mol%), (rac)-1a (0.3 mmol), 2a (0.1 mmol), and quinuclidine (1.0 equiv) in DCM (1 mL) at 25 °C for 24 h. e.e. values were determined by high-performance liquid chromatography analysis. The isolated yields of after chromatography are shown. NR = no reaction.

Notably, the use of HCO2Li as an additive can dramatically enhance the activity of the reaction to afford 3a with 89% yield and 82% e.e. (Table 2, entry 2 vs 1). Among all the additives examined (entries 3–6), 1,3-DCP was found to give the optimum results with desired product 3a obtained in 81% yield with 88% e.e. (entry 6). An increase in stereoselectivity was observed by fine-tuning the reaction conditions (entry 7).

Table 2 Further optimization

Substrate scope

Having optimized the reaction conditions for the regioisomeric allenylation, we next investigated the generality of the reaction substrates (Fig. 3A). A diverse array of propargylic carbonates 1 with various alkyl substituents performed well in the presence of nickel catalyst, affording the desired products in high yields and excellent enantioselectivities (3a-3f). Propargylic carbonates with electron-donating or electron-withdrawing substituents on the benzene ring also proved to be suitable nucleophiles in the coupling reactions, thus furnishing the corresponding allenylphosphine oxides in 70–86% yield and 87–90% e.e. (3g-3j). The meta substituted propargylic carbonates also survived our catalytic conditions (3k and 3l). This method was compatible with furan and thiophene heterocycle-substituents (3m and 3n). Moreover, various dialkyl substituted propargylic carbonates were well tolerated, leading to the corresponding products in good yields with excellent enantioselectivity (3o-3s).

Fig. 3: Nickel-catalyzed enantioselective allenylation.
figure 3

Reactions were performed by using Ni(COD)2 (10 mol%), L23 (12 mol%), (rac)-1 (0.3 mmol), 2 (0.1 mmol), HCO2Li (0.2 equiv), quinuclidine (3.0 equiv), and 1,3-DCP (30 μL) in DCM (1 mL) at 10 °C. Isolated yields after chromatography are shown. e.e. values were determined by HPLC analysis. aQuinuclidine (5.0 equiv), without 1,3-DCP, 0 °C.

The generality of the reaction with respect to the substituents of the phosphine oxides was also investigated (Fig. 3B). Both electron-donating and electron-withdrawing substituents were accommodated on the phenyl ring, and excellent levels of enantioselectivity were obtained (3t-3ab). In addition, the extension of the protocol to heterocyclic substituted phosphine oxide was also successful in affording the desired product in 67% yield and 87% e.e. (3ac).

Having established the stereoselective allenylation of phosphine oxides with (S,S)-(2-Naph)BDPP (L23) as the ligand, we further turned our attention to complementary regioselective 1,3-dienylation with the phosphine ligand dcypbz (L10). Under the optimized reaction conditions (Table 1, entry 12), a broad range of propargylic carbonates and phosphine oxides were investigated (Fig. 4). Various substituted propargylic carbonates smoothly underwent this transformation, delivering phosphinoyl 1,3-butadienes 4 in high yields (4a-4h). The structure of 4b was confirmed by X-ray single-crystal diffraction. Phosphine oxides with different substituents on the aromatic ring also afforded good results (4i-4o).

Fig. 4: Nickel-catalyzed 1,3-dienylation of phosphine oxides.
figure 4

Reactions were conducted by using Ni(COD)2 (10 mol%), L10 (12 mol%), (rac)-1 (0.12 mmol), 2 (0.1 mmol), and Ph2P(O)OH (50 mol%) in DMF at 100 °C. Isolated yields after chromatography are shown.

Synthetic applications

To further demonstrate the practical utility of our protocol, a scale-up reaction was carried out under standard reaction conditions, furnishing 3a in 71% yield with 90% e.e. (Fig. 5A). In addition, the treatment of allenylphosphine oxide 3a with PhSeCl gave rise to desired selenohydroxylation product 5 in 86% yield with a maintained e.e.64, and the absolute configuration of 5 was assigned by single-crystal X-ray diffraction analysis. Exposure of allenylphosphine oxide 3a to iodine in MeCN/H2O afforded 6, which was subsequently treated with acetic anhydride to generate ester 7 in high yield without loss of enantiopurity. Subsequently, the Suzuki coupling of 7 with phenylboronic acid in MeCN afforded 8 with 90% e.e. in 86% yield. 7 could also be readily reduced and then oxidized in the presence of elemental selenium and sulfur to obtain the corresponding phosphine selenide 10 and phosphine sulfide 11, respectively. In addition, propargylic carbonate 1b was subjected to 1,3-dienylation on a large scale, and the corresponding adduct 4b was obtained in 65% yield (Fig. 5B). Hydrogenation of 4b in the presence of a catalytic amount of Pd/C generated the corresponding phosphine oxide 12 in 87% yield. Furthermore, the treatment of 4b with trichlorosilane furnished the corresponding phosphine 13, which was then oxidized in the presence of elemental selenium to obtain phosphine selenide 14 in a high yield.

Fig. 5: Synthetic versatility of the catalytic system.
figure 5

A Derivatization of product 3a. B Derivatization of product 4b.

Mechanistic studies

A series of experiments were conducted to gain a better understanding of the mechanistic details of this process (Fig. 6). The reactions carried out with the racemate and both enantiomers of propargylic carbonate 1p produce good graphical overlay in the kinetic data (Fig. 6A), and under all conditions, (S)-3p was obtained in essentially identical e.e. (92% e.e.) monitored over different reaction times. Furthermore, reactions of (rac)−1p, (R)-1p (96% e.e.), and (S)-1p (93% e.e.) were separately conducted under the standard reaction conditions (Fig. 6B), and e.e. of the recovered 1p was essentially unchanged throughout the reaction process, suggesting that the irreversible oxidative addition process occurred under the reaction conditions. All these results established that the chiral ligand effectively controls the absolute configuration of the product, regardless of the stereochemistry of the starting electrophilic substrates. The relationship between the e.e. value of L23 and that of 3p was then investigated using a ligand with different levels of enantiopurity (Fig. 6C, racemic, 20%, 40%, 50%, 60%, 80%, and >99% e.e.). Indeed, the nonlinear effect study revealed a linear relationship between the e.e. of 3p and the enantiopurity of phosphine ligand L23, suggesting that one phosphine ligand is most likely involved in the active nickel species.

Fig. 6: Mechanistic studies.
figure 6

A Time course of the kinetic resolution of 1p. B e.e. of recovered 1p at partial conversion. C Nonlinear relationship between the enantiopurity of phosphine ligand L23 and adduct 3p.

We achieved control over the reaction outcome in Ni-catalyzed reactions by slightly changing the reaction conditions. The selectivity between 3a and 4a could be controlled by the choice of the phosphine ligand (Table 1, entries 3 vs 8) and the acid additive (Table 1, entries 10 vs 12). These unusual results encouraged us to investigate the effect of acid or base effect on the regioselectivity (see Supplementary Information for details). As shown in Table 3, there were clear trends that allenylation pathway leading to 3a with different phosphine ligands is favored with base additives (entries 1,4 vs 2,5), while the acid additive resulted in an improved formation of 1,3-dienylation product 4a (entries 3,6 vs 2,5).

Table 3 Additive effect on the outcome of the reaction

A proposed mechanism for the nickel-catalyzed coupling reactions of propargylic carbonates and phosphine oxides is illustrated in Fig. 7. The selectivity between the competing 1,3-dienylation and allenylation pathways was determined by the structure of the phosphine ligands and the additives65,66,67,68. As shown in Fig. 7, the proposed catalytic cycle for the generation of allenylphosphine oxide 3 begins with the Ni-mediated decarboxylation of propargylic carbonate 1, and further delivers the allenylnickel complex I or IV58,59,60,61,62. Under basic conditions with L23 as the ligand (left cycle), the in situ generated tert-butoxy anion could deprotonate hydroxydiphenylphosphane to generate the diphenylphosphinite anion. Therefore, the nucleophilicity of oxygen is higher than that of phosphine in 2’, which might favor nickel-catalyzed O-propargylation62. Therefore, nucleophilic addition of hydroxyphosphine anion 2’ onto phosphine L23-bound allenylnickel intermediate I provides a stereoselective route to access chiral propargyl phosphinate III62, and the following [2,3]-sigmatropic rearrangement furnishes the final allenylphosphine oxide 369. Alternatively, under acidic conditions with L10 as the ligand (right cycle), the nucleophilicity of phosphine is higher than that of oxygen in 2”57. A possible diphenylphosphinic acid assisted proton shift process70 is operated to generate an allyl nickel intermediate V71, which is also consistent with the deuteration experiments (see Supplementary Information for details). The ligand exchange of intermediate V with 2” and subsequent reductive elimination of nickel complex VI71 regenerates the nickel catalyst and furnishes final 1,3-dienyl isomer 4.

Fig. 7: Proposed catalytic cycle.
figure 7

Pathways are shown for the catalytically afforded reactive species from propargylic carbonate with nickel catalysts. Under basic condition, the phosphine L23-bound allenylnickel intermediate I can react with hydroxyphosphine anion 2’ to produce 3 (left cycle). Under the acidic condition, an allyl nickel intermediate V undergoes the reaction with 2” to furnishes the final 1,3-dienyl isomer 4 (right cycle).

Discussion

In summary, a nickel-catalyzed 1,3-dienylation and enantioselective allenylation of phosphine oxides have been developed. The employment of different phosphine ligands allows for the highly regio- and stereoselective formation of allenylphosphine oxides and phosphinoyl 1,3-butadienes from identical starting materials. Furthermore, this reaction represents the successful example of catalytic enantioselective construction of allenylphosphoryl frameworks with excellent enantioselectivity (up to 94% e.e.) using a highly efficient type of chiral BDPP-type ligand. The high efficiency, stereoselectivity, and operational simplicity of this transformation, coupled with the rational design of chiral ligands, are expected to render this method a valuable tool in asymmetric synthesis.

Methods

General procedure for nickel-catalyzed enantioselective allenylation

In a 10 mL Schlenk tube, Ni(COD)2 (2.8 mg, 0.01 mmol, 10 mol%), L23 (7.6 mg, 0.012 mmol, 12 mol%), and 1,3-DCP (30 μL) were stirred in 1 mL anhydrous DCM under argon at room temperature for 20 min. Propargylic carbonate 1 (0.3 mmol), phosphine oxide 2 (0.1 mmol, 1.0 equiv), quinuclidine (33.3 mg, 0.3 mmol), and HCO2Li (1.4 mg, 0.02 mmol) were then added successively. The reaction mixture was stirred at 10 °C until the reaction was complete (monitored by TLC). The reaction mixture was concentrated, and the residue was purified by column chromatography to afford the corresponding product 3.

General procedure for nickel-catalyzed 1,3-dienylation

In a 10 mL Schlenk tube, Ni(COD)2 (2.8 mg, 0.01 mmol, 10 mol%) and dcypbz L10 (5.6 mg, 0.012 mmol, 12 mol%) were stirred in 1 mL anhydrous DMF under argon at 80 °C for 10 min. Propargylic carbonate 1 (0.12 mmol), phosphine oxide 2 (0.1 mmol, 1.0 equiv) and HOP(O)Ph2 (10.9 mg, 0.05 mmol) were then added successively. The reaction mixture was stirred at 100 °C until the reaction was complete (monitored by TLC). The reaction mixture was concentrated, and the residue was purified by flash column chromatography on silica gel to give the desired product 4.