Optimization of the reaction conditions. We initiated our exploration by evaluating the transfer hydrogenation of the model substrate 1a with simple nickel sources and 2,2'-bipyridine ligands (Table 1). The first obstacle to overcome is the activation of the inert H2O molecule in our nickel catalyst system.28-30 Gratifyingly, boron reagents showed unique effect, and the alkenes were obtained in high yield and selectivity using Na2CO3 as base. B2pin2 turned out to be more efficient than other diboron compounds such as B2(OH)4, B2cat2 and B2neop2 (Supplementary Table 1).31-33 Although diboron were found to be capable of activating water in Pd-catalyzed systems,31-35 including hydrogenation of unsaturated C-C bonds to saturated alkanes,31 it is, as far as we know, the first case for such activation effect in Ni catalyst systems. Notably, E-alkene 3a was formed as the major isomer, and over-reduced alkane product was not observed. Solvents turned out to exert a profound influence on the reactivity (Supplementary Table 1), 72% yield of alkenes were obtained with 11/89 isomeric ratio in DMF (entry 1). Decorating the bipyridine ligand with electron-withdrawing ester groups totally suppressed the reactivity (entry 2). Subsequent screening of other bipyridine derivatives as well as phenanthroline ligands L3-L6 provided comparatively inferior results to 2,2'-bipyridine (entries 3-6). Systematic screening of nickel catalyst, ligand, base, boron and water (Supplementary Table 2 and 3) showed that base exerted an unexpected, yet decisive role in the control of selectivity. As shown in Table 1, the reaction was evidently inclined to E-selectivity by K2CO3, NaOH and CF3CO2Na, with the later showing the best result, affording 3a in 84% isolated yield and 6/94 Z/E ratio (entries 7-9). Interestingly, a slant to Z-selectivity was shown with CH3CO2Na, providing 2a with 69/31 Z/E ratio (entry 10). Organic bases such as DABCO and Et3N were also tested, and E-alkene 3a was delivered as the major product (entries 11 and 12). The catalyst loading could be lowered to 5 mol% with no erosion of the yield or selectivity (entry 13). The reactivity was almost totally shut down at a lower temperature of 60 oC (entry 14), which might due to the insufficient energy for the isomerization of alkenyl nickel intermediate (Figure 1i). In contrast, comparable results were observed at higher temperatures (entries 15 and16).
The above results inspired us to further proceed with other bases aiming at the optimization for Z-selective semihydrogenation of 1a. As shown in Table 2, CH3CO2K and CH3CO2Cs acted similarly as CH3CO2Na, indicating that metal ions are not responsible for the selectivity reversal (entries 1 and 2). Only moderate selectivity was achieved when HCO2Na was added (entry 3). To our delight, PhCO2Na gave a promising result, providing the final olefins in 80/20 selectivity (entry 4). Again, dicarboxylate ligand L2 showed dramatically decreased reactivity (entry5). In contrast, 4,4'-dimethoxy-2,2'-bipyridine L3 improved the selectivity to 93/7 (entry 6). Ligands L4 and L5 bearing methyl groups at 3,3’- or 4,4’-positions both gave slightly reduced selectivity than L3 (entries 7 and 8). When the loading of the catalyst and base were reduced, alkenes were retrieved in slightly improved yield and selectivity (entries 9 and 10). Contrary to E-selective system (Table 1, entry 14), the reaction could still proceed smoothly at a lower temperature, albeit 1a was partially recovered (entry 11). Performing the reaction at higher temperatures resulted in poorer selectivities (entries 12 and 13).
Mechanistic investigations. Several questions deserve exploration to better understand this unprecedented system: (a) is water in the system indeed the hydrogen donor? (b) are alkenes generated from hydrometallation of in situ formed Ni-H species or hydrolysis of vinyl boron compounds? (c) does isomerization of Z-olefins take effect similarly as most precedents to afford E-olefins? (d) what are the roles of the bases in modulation of the reaction outcomes? To answer these questions, a series of mechanistic studies were carried out. Firstly, deuterium-labeled experiments were conducted (Figure 1a). The deuterium was incorporated into both the 1,2-olefinic positions of 2a' and 3a' with D2O instead of H2O under both standard conditions (equations (1) and (2)). Similar results were also observed for unsymmetric alkynes 1z and 1g, with the former leading to even higher deuterations (equations (7) and (8)). In contrast, there was no sign of deuteration on the products using DMF-d7 as solvent (equations (3) and (4)). When the reactions of 1a using D2O were placed in hydrogen atmosphere, comparative deuterium isotopic contents as in argon were observed (equations (5) and (6)), proving that releasing of H2 and consequent hydrogenation was not involved in the catalytic pathway. Control experimental studies of vinylboron reagents 4 and 5 were performed under the standard reaction conditions.36,37 Olefin products 2a and 3a were not detected, excluding the possibility of hydrolysis of vinylboron derivatives (Figure 1b, equations (9) and (10)). This, together with the reactions under H2 atmosphere, indicated that Ni-H species were formed between the nickel pre-catalyst and H2O assisted by B2pin2, which would deliver alkenyl nickel intermediates to accomplish the catalytic cycle.
To deeper understand the process of selective semi-reduction, the kinetic behavior of the reaction system was monitored (Fig.1c). The kinetic profile of Z-selective semihydrogenation showed that 2a was generated by degrees in the initial 5 hours, and the yield stayed closely aligned with the conversion. After this period, 3a began to show up and gradually increased to 6% yield, alongside with a sharp decline of the yield growth rate of 2a (Fig.1c, left). We postulate that the small amount of E-alkene in this system derives from isomerization of the Z-isomer, which was suppressed in the initial 5 hours since competitive coordination of alkyne 1a with the metal center. Consumption of most 1a after 5 hours left space for the coordination of 2a for the subsequent isomerization process, which still need 1a as auxiliary since the selectivity remained unchanged after disappearance of 1a. The E-selective reaction profile with CF3CO2Na as base clearly indicated the nonexistence of Z/E isomerization (Fig.1c, right). Approximately 6% of Z-alkene was already formed at the early stage of the reaction, which maintained in this level until 1a was completely converted. The concentration of 3a increased gradually, which was independent with the amount of 2a.
To further verify the above inferences, a series of control experiments were conducted (Fig.1d). When Z-alkene 2a was put in both standard conditions, only less than 5% of E-alkene was detected (equations (11) and (12)), demonstrating the reluctance of the Z/E isomerization in these conditions. Elevating the reaction temperature showed a beneficial effect for the isomerization, which was promoted to 13% by heating 2a at 120 oC under the Z-selective condition (equation (13)). Consistently, the reaction of 1a at 120 oC under this condition afforded the corresponding olefinic products in 86/14 selectivity (equation (14)), compared with 93/7 at 80 oC.
The color changes between the two reaction systems were significantly different. As shown in Figure 1e, the Z-selective system seemed turbid and beige at the very beginning, which turned to light brown after several minutes and got darker later. The color changed to tan-yellow gradually in about one hour and became lighter to milk-white after another one hour, which remained till the end. A completely different visual appearance mutation was observed for the E-selective system, which looked transparent black and got darker quickly at the very early stage. Interestingly, as soon as the reaction was over as monitored of the crude mixture, the color changed to bright yellow immediately, which could be regarded as a simple hint for the complete of the reaction. We postulate that the dark color ascribes to the coordination of the triple bond to the metal center, which was terminated promptly once alkynes were exhausted.24 The distinction in colors of the two systems indicates that different nickel species might be involved, leading to the corresponding olefinic products in totally unrelated pathways. The color variation of the control experiments on base was quite similar to the above observation (Fig. 1e, bottom): the initial pale green color changed to tint of turbidity yellow and clarify black color separately after addition of PhCO2Na and CF3CO2Na, respectively, indicating the formation of different nickel species was modulated with the choice of base.
Competitive control experiments of the bases were conducted to further illustrate their functions (Figure 1f). After the standard Z-selective mixture using PhCO2Na was stirred for 1 h, another 2.0 equivalent of CF3CO2Na was added, and no apparent influence on the reaction outcomes was observed (equation (15)). By contrary, a worse selectivity was caused by addition of PhCO2Na into the E-selective system (28/72 vs 4/96) (equation (16)).
All the mechanistic insights and the visual phenomenon pointed to distinct catalytic pathways for the two reaction systems, inspiring us to further inquire whether different metal species were taking effect inherently. To detect whether nanoparticles were involved in our Ni-B-H2O system, general mercury drop experiments were performed32,34 (Figure 1g). The yield or selectivity was not affected in either systems (equations (17) and (18)), excluding heterogeneous catalytic pathways. Despite the failure in capture of metallic intermediates, electron paramagnetic resonance (EPR) analyses provided clues on the active nickel species and the base effect. As shown in Figure 1h (2), strong EPR signals were observed in the E-selective mixture, indicating the formation of Ni(I) or Ni(III) species.38-41 The signals of such Ni species could not be found at ambient temperature, which is in accordance with our experimental observations that semihydrogenations of 1a were not permitted at rt (Supplementary Table 3, entry 25). In contrast, EPR active species was not observed in Z-selective system (Figure 1h (1)), featuring a Ni(0)/Ni(II) catalytic cycle. In agreement with the competitive experiments of bases (equation (16)), the EPR signals for the reactions using CF3CO2Na as base were markedly weakened after the addition of PhCO2Na (Figure 1h (3)). In line with the fact that use of HCO2Na as base gave an almost 1:1 ratio of the Z- and E-alkenes (Table 2, entry 3), the EPR signal of the system with HCO2Na was less significant than that with CF3CO2Na (Figure 1h (4)), but much more significant than that of PhCO2Na system (Figure 1h (1)).
Although more experimental supports are awaited to uncover the detailed mechanism, a general scenario could be delineated based on the above results and relative literatures31,32, 42-50 (Figure 1i): NiBr2 would interact with the bases firstly, delivering carboxylates carrying different counter anions. The difference in electronic properties between the benzoate and the trifluoroacetate endows them with distinct reactivities towards B2pin2. Consequently, Ni(II) species C is generated directly from the benzoate B and B2pin2. Activation of H2O molecule delivers Ni(II)-H species D, which undergo syn-addition to the triple bond to afford alkenyl Ni(II) intermediate E. Participation of another H2O molecule release the cis-olefin and regenerate C with the assistance of B2pin2. Based on the kinetic experiments, coordination and insertion of the Z-alkene to the Ni-H species assisted by alkyne precursor would occur in the late stage of the reaction, followed by isomerization process resulting in slight stereo-impurity. We propose that isomerization of a vinyl Ni(I) species is responsible for the E-selectivity observed in this approach, the specific oxidation state at Ni could provide an opportunity for isomerization42,43. At the beginning of the cycle, Ni(0) species H might be generated firstly from nickel trifluoroacetate G and B2pin2. Oxidative addition of another molecule of B2pin2 furnishes Ni(II) species C, comproportionation between C and H occurs instantly, forging Ni(I) species I to initiate the catalytic cycle. Activation of H2O molecule would deliver Ni(I)-H species J, followed by insertion of alkyne leading to vinyl Ni(I) intermediate K, which may undergo isomerization43 to E-alkenyl nickel intermediate L. Thermodynamically more stable product 3 is generated by hydrolysis of L, and the acquired nickel hydroxide M could be transformed back to Ni(I) species I in the aid of B2pin2.
Substrate scope. The synthetic practicability of this system was sufficiently embodied in the functional group compatibility investigations. In Table 3a, the Z-selective semi-reduction of various alkynes 1 using PhCO2Na as base is summarized. This reaction proceeded successfully toward substituted diarylethynes bearing a diverse set of substituents. Specifically, substrates bearing methyl or tert-butyl groups at p- or m-positions all worked smoothly under the standard conditions (2a-2d), as well as hindered isopropyl (2e) or phenyl (2f) groups located in the ortho-position of the aryl terminus, suggesting the insensitivity of the system to steric effect. Electron-donating methoxy substituent was well accommodated, and the diaryl alkenes were generated in high yields and selectivity (2h, 2i and 2j). Amino functional group 2k was no exception, well tolerated in this catalytic semihydrogenation process. Z-olefins with electron-withdrawing trifluoromethyl (1l), cyano (1m, 1n), ester (1o) and acyl (1p) groups could also be achieved uneventfully. Fluoro- and chloro-containing products (2q-2t) were furnished from the corresponding alkynes, leaving space for further functionalization. The generality of the system was further showcased by the tolerance of naphthyl (2u) and heterocycles including thienyl (2v), benzofuryl (2w) and pyridyl motifs (2x), particularly the latter, considering pyridinyl ligands were used in our catalytic system. Moreover, running in a longer reaction time or higher temperature, alkynes carrying both naphthenic and linear alkyl terminuses could be reduced to the corresponding olefinic products efficiently (2z-2ff). Notably, only Z-alkenes were formed specifically from the alkyl substrates, supporting our previous deduction that the E-alkenes in the Z-selective conditions might derive from the isomerization process, which was sluggish for alkyl alkenes due to their weak coordinating ability to the metallic species. The compatibility of the system was further underlined by successful involvement of unprotected primary OH group (2dd), which was unaffected under the catalytic conditions. Natural product derived alkyne with estrone skeleton proceeded smoothly, and the desired product 2ee was furnished in excellent yield and selectivity. Finally, internal alkyne 1ff bearing 1,2-dialkyl substituents also gave high yield and perfect stereoselectivity.
A survey on the substrate scope was performed next to demonstrate the robustness of the E-selective TH process using CF3CO2Na as base (Table 3b). Similar as the former system, diaromatic internal alkynes with a wide range of functional groups such as methyl (1a-1c), tert-butyl (1d), isopropyl (1e), methoxyl (1i, 1j), amino (1k), trifluoromethyl (1l), cyano (1m), ester (1o), acyl (1p) and halogen substituents (1q-1t) were all hydrogenated to the desired trans-alkenes uneventfully. Heteroaromatic rings including thienyl (1v), benzofuryl (1w) and pyridyl (1x, 1y) substituents were compatible again, delivering the alkenyl heterocycles selectively. The reaction of alkyl acetylene was more challenging, affording alkene 3gg in moderate yield and inferior selectivity. Propargylic esters were transformed to E-olefins (3hh-3jj) as single isomers in moderate yields and excellent selectivity. Consistent with the previous observation, for all the E-selective experiments, a mutation of color from black to bright yellow was observed as soon as the reaction finished.
Finally, we tested the reactivity of terminal alkynes, which are more inclined to over-reduction. As shown in Table 3c, alkene 6 was obtained in high yield in Z-selective conditions from 1kk, and saturated ethyl product was not observed. The condition could also be extended to diynes 1ll and 1mm, with both triple bonds being hydrogenated in high selectivity. Interestingly, the reaction of conjugated enyne 1nn in Z-selective conditions afforded diene 9 with E-configuration as the major product. On the contrary, Z-enyne 10 was obtained in high selectivity when 1,3-diyne 1oo was loaded in E-selective conditions.