Bismuth radical catalysis in the activation and coupling of redox-active electrophiles

Radical cross-coupling reactions represent a revolutionary tool to make C(sp3)–C and C(sp3)–heteroatom bonds by means of transition metals and photoredox or electrochemical approaches. However, the use of main-group elements to harness this type of reactivity has been little explored. Here we show how a low-valency bismuth complex is able to undergo one-electron oxidative addition with redox-active alkyl-radical precursors, mimicking the behaviour of first-row transition metals. This reactivity paradigm for bismuth gives rise to well-defined oxidative addition complexes, which could be fully characterized in solution and in the solid state. The resulting Bi(III)–C(sp3) intermediates display divergent reactivity patterns depending on the α-substituents of the alkyl fragment. Mechanistic investigations of this reactivity led to the development of a bismuth-catalysed C(sp3)–N cross-coupling reaction that operates under mild conditions and accommodates synthetically relevant NH-heterocycles as coupling partners.

oxidation states (Fig. 1a, right). This particular chemical behavior leads to the facile generation of alkylradical fragments through single-electron-transfer (SET) oxidative addition from precursors such as redoxactive esters (RAE) or Katritzky salts (KS). The recent years have witnessed significant efforts towards mimicking the redox behavior of transition metals by main-group elements. 14 For instance, pnictogens can take part in SN2-type polar oxidative additions resulting in two-electron maneuvering throughout (n)/(n+2) redox catalytic cycles, emulating those of late transition metals (Fig. 1a, left). 15 However, the radicalactivation of redox-active electrophiles is often restricted to first-row transition metals, and examples by a well-defined main group-element complex are rare.
Very recently, bismuth redox catalysis has been established as an emerging platform for organic synthesis. 16 In this context, our group has shown how polar Bi(III/V) or Bi(I/III) catalytic cycles can lead to the development of C-F, 17 C-O, 18 or C-H 19 bond-forming reactions, among others. 20 Nevertheless, even though persistent and stable radicals of heavier main-group elements are known, 21 the studies for bismuth are much more limited, and bismuth radical catalysis has been significantly underexplored. 22 Bi(II/III) catalytic cycles have been postulated for the living radical polymerization of alkenes, 23 or the cycloisomerization of 4iodoalkenes. 24 This, together with further reports probing the existence of bismuth(II)-centered radicals, 25 prompted us to explore the behavior of the Bi(I/II) pair in SET-based oxidative additions of redox-active alkyl electrophiles. Here, we disclose how a well-defined bismuthinidene (1) reacts with alkyl phthalimide esters and alkyl Katritzky salts to give alkyl-bismuth(III) adducts, which were found to behave as Bi-C radical-equilibrium complexes (Fig. 1b, bottom). Additionally, we discovered that α-amino alkyl-radical fragments resulting from this process can be easily oxidized by Bi(II), giving rise to iminium ions 26,27 that can be trapped by N-nucleophiles. This observation led to the development of a Bi-catalyzed radical C-N cross-coupling reaction with a wide scope of both coupling partners (Fig. 1b, top). In spite of the vast number of alkyl-radical couplings developed during the past decade, seldom examples of C(sp 3 )-N bond formation from redox-active radical precursors have been reported, 28 mainly relying on photoredox setups, 29,30,31,32 electrochemical synthesis, 33,34 or the use of an excess of chemical oxidant. 35 In this report, we demonstrate that catalytic amounts of a Bi(I) complex can promote this type of transformation in an autonomous manner, without the need of a photoredox system, a chemical oxidant, an external base, or an electrochemical setup.

Results and discussion
As a result of the high nucleophilicity of the 6p 2 lone pair on the Bi(I) center, bismuthinidene 1 has recently been shown to engage in polar SN2-type reactions with alkyl halides and triflates. 36 Similarly, 1 reacted quantitatively with a range of benzyl (pseudo)halides (Cl, Br, I, OMs) to give benzyl bismuth(III) complexes 5-8 (Fig. 2b). Cyclic voltammetry analysis of 1 (E1/2 = -0.85 vs Fc 0/+ ) evidences that such C-X cleavages should proceed through a classical SN2 pathway (Ep/2 < -2.0 V vs Fc 0/+ ). On the other hand, the electrochemical behavior observed suggested that 1 could potentially engage in SET oxidative addition processes with alkyl redox-active electrophiles (Fig. 2a). Accordingly, reaction of 1 with 1 equivalent of tetrachlorophthalimide ester 2 (Ep/2 = -1.1 V vs Fc 0/+ ) cleanly afforded benzyl bismuth(III) complex 9, after SET and fragmentation upon release of CO2, followed by radical recombination. 37 The resulting alkyl bismuth(III) adduct could be fully characterized by NMR, HRMS and single-crystal X-ray diffraction.
Furthermore, Katritzky salt 4 (Ep/2 = -1.2 V vs Fc 0/+ ) also underwent radical oxidative addition with 1 to give 10. As expected, non-chlorinated phthalimide ester 3 (Ep/2 = -2.0 V vs Fc 0/+ ) remained unreacted when mixed with 1. Besides benzyl groups, the same process occurs with primary (11), or secondary (13) RAEs, leading to stable alkyl-bismuth(III) complexes. 38 Interestingly, the process is orthogonal to classical polar transition-metal oxidative additions, as it could be performed in the presence of an aryl bromide, giving 12 as the sole product in 93% yield (Fig 2b). This reactivity is a rare example where bismuth, besides emulating the redox behavior of first-row transition metals during oxidative addition, allows the isolation and characterization of the corresponding alkyl-metal species resulting from radical recombination. Furthermore, when 13 was analyzed by EPR spectroscopy a weak but steady signal of the carbon-centered radical could be observed (see Supplementary Materials), which points to 13 being a radical-equilibrium complex. 39,40 Based on this, the reaction of bismuthinidene 1 with cyclopropylmethyl iodide was monitored by NMR at low temperature in the dark (Fig. 2c). Complete conversion to cyclopropylmethyl adduct 14 was observed within one hour at -20 °C. When the mixture was warmed to 50 °C, a slow but steady conversion to ring-opening compound 16 was observed (35% after 12 h). This indicates that homolysis of the Bi-C bond in 14 takes place, and the complex is in equilibrium with its corresponding in-cage radical pair (18). Conversely, when cyclopropylmethyl redox-active ester 15 was reacted with bismuthinidene 1, a complex analogous to 14 was not observed. Instead, radical ring-opening product 17 was immediately obtained, even in the dark. This is consistent with the two distinct mechanistic scenarios for the oxidative addition. On one hand, polar SN2-type reaction of cyclopropylmethyl iodide with 1 initially leads to 14, which eventually ring-opens via radical equilibrium. On the other hand, SET and fragmentation of RAE 15 leads to an incage bismuth(II)/alkyl radical pair (18), for which cyclopropane ring-opening is faster than radical recombination, resulting in the formation of 17 (Fig. 2b). The radical-equilibrium hypothesis is consistent with the reactivity displayed by these complexes: the secondary alkyl radical derived from 13 reacts with Michael acceptors such as phenylvinylsulfone giving Giese addition product 21, either in the dark (57%) or under blue-light irradiation (85%) (Fig. 3a, left). The alkyl-radical fragment of several complexes was successfully trapped with TEMPO leading to C(sp 3 )-OTEMP adducts. Moreover, we observed that catalytic amounts of 1 can promote Giese-type reactions, among others, upon blue light irradiation (for these catalytic examples, and for TEMPO trapping experiments, see Supplementary Information). 42 Interestingly, when attempting the isolation of α-amino alkyl-bismuth(III) adduct 23 derived from proline, we observed the exclusive formation of the product of decarboxylative amination 24, with recovery of bismuthinidene 1 (Fig.   3a, right). 29,32 It was speculated that product 24 would arise from the oxidation of the corresponding α-amino alkyl radical by a highly reactive bismuth(II) species. This would lead to the formation of an electrophilic iminium ion, 27 which ultimately reacts with the tetrachlorophthalimide (TCNPhth) anion to forge the C-N bond (for details, see Fig. 4b and Supplementary Information). At this point, it was envisaged that this reactivity could be exploited to engage the corresponding iminium with external N-nucleophiles, leading to a formal C-N cross-coupling reaction. After optimization of the conditions, we found that the reaction of RAE 22 with 3 equivalents of benzimidazole (25) in the presence of 10 mol% of 1 in DMA at 25 °C afforded the product of C-N cross coupling (26) in 88% yield within 2 h (Fig. 3b). Under these conditions, the only observed side products were 24 (nucleophilic competition by TCPhth) and the expected amide bond-formation product 27, which could be minimized by controlling the stoichiometry and selecting the appropriate solvent (Fig. 3b, entry 1 vs 3), respectively (see Supplementary   Information for optimization details). Control experiments without Bi catalyst led exclusively to acyltransfer product (entry 2). The high efficiency of the optimized reaction relies on the faster kinetics of the Bi-radical process compared to the background amide formation. For example, the reaction could be carried out at -30 °C in DMF (entry 4), or at room temperature in DMA in only 10 min (entry 5), giving the desired product in 56% and 85% yield, respectively. To exclude completely the requirement of photoexcitation for any of the steps of the transformation to proceed, the reaction was carried out under exclusion of ambient light, giving comparative results (entry 6). Finally, we found that addition of TEMPO completely inhibits the reaction, presumably due to the inertness of the formed bismuth-TEMPO adduct (entry 7). 39 The scope of the transformation is presented in Fig. 4a. RAEs of readily available natural and non-natural α-amino acids were interrogated (either fully protected, or with free N-H bonds) as electrophilic partners.
C-N coupling products derived from proline (26), phenylalanine (28), valine (29) or pipecolic acid (30,32) were successfully obtained in good to excellent yields. In order to compete favorably with TCPhth, synthetically relevant N-heterocycles bearing free N-H bonds were evaluated. Using proline-derived RAE 22, the corresponding C-N coupling products with a variety of benzimidazoles (26,40), imidazoles (33,39,41), and pyrazoles (31, 32, 35-38) were obtained. Non-symmetrical heterocycles such as benzotriazole (34) could also be accommodated, providing the product in a 5:1 N1/N2 ratio of regioisomers. A range of functional groups with different electronic properties were also tolerated (38,41,45). Since the radical process is orthogonal to classical transition metal-catalyzed cross-coupling reactions, different heteroaryl halides (35)(36)(37) and heteroaryl boronic esters (31, 32) could be well tolerated. The strategy was successfully applied in the modification of bioactive molecules, such as theophylline (42, 57%, single regioisomer). The successful coupling using thiabendazole (43, 61%) provides another illustrative example of the orthogonal reactivity to transition metals, as the Lewis-basic sites on both starting material and product could inhibit catalysis by binding to a metal center. In the absence of external nucleophiles, the product of decarboxylative amination via formal CO2 extrusion was obtained. For this process, both α-amino RAEs and α-oxo RAEs reacted, giving hemiaminal-type structures such as 44 and 45 in good yields. Overall, this strategy is complementary to the photochemical protocol reported by Fu and co-workers, allowing the use of α-heteroatom RAEs, instead of unbiased alkyl substrates. 29 To shed light into the mechanism, we monitored the catalytic reaction of α-amino RAE 22 by NMR with 10 mol% of bismuthinidene 1 at -40 °C using DMF-d7 as solvent. In this scenario, the pair of rotamers of α-amino alkyl-bismuth(III) intermediate 23 accumulated upon consumption of the RAE, coexisting with bismuthinidene 1. It is important to mention that complex 23 was characterized by reaction of 1 with 22 in a separate stoichiometric experiment (see Supplementary Information for details). The accumulated 23 decays into 1 after 1 h at -20 °C (Fig. 4b, slow Bi(I/II/III) pathway). However, we observed that the consumption of RAE 22 to give decarboxylative-amination product 24 occurs at a faster rate than the former process (for details on kinetic analysis, see Supplementary Information). Thus, an alternative pathway should be considered in which in-cage radical pair 23' reacts through SET, leading to iminium cation 46 upon regeneration of Bi(I) (Fig. 4b, fast Bi(I/II) pathway). 43 Competing radical recombination of 23' leads to some accumulation of 23, which eventually collapses into the reaction product (24) and 1. Overall, the radical oxidative addition appears to be the rate-limiting step of the dominant pathway, as suggested by the continuous presence of 1 throughout the entire course of the reaction.

Conclusions
In summary, we have developed a radical oxidative addition of redox-active carbon electrophiles to lowvalent bismuth, based on the SET from a well-defined Bi(I) complex to alkyl RAEs and KS. This process led to a family of alkyl-bismuth(III) complexes, which were found to be in equilibrium with the corresponding in-cage radical pair formed by bismuth(II) and a free alkyl radical. Unbiased alkylbismuth(III) complexes are stable and can be characterized both in solution and in solid state. On the other hand, α-amino alkyl-bismuth(III) intermediates collapse back to bismuth(I) upon releasing iminium cations, which can be trapped by external N-nucleophiles. This led to the development of a bismuth-catalyzed C-N cross-coupling reaction, using complex N-heterocyclic compounds. This new type of radical catalysis is promoted by bismuth in an autonomous manner, through a radical Bi(I/II) or Bi(I/II/III) redox cycle, without the need of a photoredox system, a chemical oxidant, an external base, or an electrochemical set-up. Overall, these findings open up a field of radical couplings by a main-group element and pave the way for the rational design of synthetically relevant transformations based on Bi radical catalysis.

Data availability
The Supplementary Information contains all experimental procedures and analytical data ( 1 H, 19 F, 13