Given the fundamental significance of organophosphorus compounds ranging from synthetic, pharmaceutical, and agricultural chemistries to materials and life sciences, the construction of C–P bonds has long been an important task for generations of synthetic chemists [1-7]. Since the pioneering work from Hirao et al., who developed the palladium-catalyzed phosphonation of aryl iodides or bromides in the 1980s [8-10], transition-metal-catalyzed C–P couplings have been among the most reliable tools for synthesizing organophosphorus molecules. In recent decades, many elegant protocols using aryl halides, boronic acids, triflates, sulfonates, and other precursors as coupling partners have been developed for this purpose [11-16]. However, the requirement of expensive and air-sensitive ligands and oxidants and the harsh reaction conditions in these transition metal catalysis systems usually resulted in a limited substrate scope and poor functional group tolerance, as well as high synthetic costs. Consequently, the development of efficient and economical methods for C–P bond formation is still highly desirable.

Recently, a number of impressive approaches have been reported for constructing C–P bonds following the advent of visible photoredox catalysis as a powerful and ecofriendly synthetic tool [17-32]. Particularly, the strategy of synergistic catalysis by merging visible-light photoredox catalysis [33-35] with organometallic catalysis has provided several attractive accesses to organophosphorus molecules [29-32]. For example, in 2015, Toste et al. [29] disclosed an elegant C–P coupling reaction of aryldiazonium salts and H-phosphonates in the presence of gold catalysts and ruthenium-based photocatalysts. In the same year, a dual catalytic C(sp²)−P formation reaction of diphenylphosphine oxides with aryl iodides by using nickel catalysts and ruthenium-based photosensitizers was reported [30]. In 2017, Lei et al. [31] reported a visible-light-induced oxidant-free oxidative phosphonylation of a C(sp²)–H bond of arenes via a dual photoredox and cobalt catalysis. Despite these important advances in the field of photocatalytic C–P bond formation, regardless of whether single photoredox catalysis or the synergistic photoredox and organometallic catalysis is used, homogeneous photocatalysts, including ruthenium or iridium complexes and organic dyes, are widely applied. This leads to the problem of the inability to separate, recover, and reuse compared with heterogeneous photocatalysts (Fig. 1(a)) [36-44]. In this work, because of continuing interest in visible-light photocatalysis [45-50], this C–P formation reaction is updated by the joint use of Ni catalysts and heterogeneous cadmium sulfide (CdS, 2.4-eV bandgap) photocatalysts (Fig. 1(b)) [51]. Compared with previous endeavors, this protocol should attract wide interest from the synthetic community because of the recyclable photocatalysts, a less expensive catalyst system, and a broad substrate scope.

The study began with ethyl 4-bromobenzoate and diphenylphosphine oxide as substrates. Exposure of these two chemicals in CH₃CN to blue light-emitting diodes (LEDs) in the presence of 20-mol% CdS, 5-mol% NiCl₂•6H₂O, 5-mol% dtbbpy, and 2.0 equiv. DBU under an Ar atmosphere at 55 ℃ resulted in the desired product 3aa in 21% yield after 24 h. Because of this result, different parameters were evaluated to improve the reaction efficiency. First, it was found that organic base Et₃N gave the best yield of the product (Table 1, entry 4, 80% yield) among all the screened bases. Second, after other reaction media were screened, DMA stood out as the optimal one (Table 1, entry 6), and the best substrate ratio of 1a to 2a was 1:2 (Table S2). Third, the reaction, when performed at room temperature, gave a low reaction efficiency (Table 1, entry 10). Finally, control experiments revealed that Ni, CdS, base, and light were all essential components for obtaining high efficiency (Table S3).

After the optimized conditions were established, the generality of this catalyst system for the C–P coupling was then investigated. Different aryl bromides bearing a variety of functional groups can be utilized in this protocol to deliver the coupled product with a good-to-excellent yield (Scheme 1). For instance, aryl bromides containing electron-withdrawing groups, such as ester, ketone, nitrile, trifluoromethyl, and fluoride, at the para- or meta-position were adequate coupling partners (3aa-3da, 3ka, and 3la: 70%‒95% yields). With the existence of a competing chlorine group, the reaction exhibits good chemoselectivity, giving the C–P coupled product a good yield (3ja: 82% yield). Furthermore, aryl bromides bearing electron-donating groups, such as methyl, methoxyl, tert-butyl, protected amine, and phenyl, as well as 3, 5-dimethylphenyl and fused rings, can readily participate in this coupling reaction (3ea-3ia, 3ma, and 3na: 53%‒86% yields). Moreover, heteroaryl bromides were also compatible in this coupling system and converted to the corresponding products in good yields (3oa: 73% yield; 3pa: 65% yield).

Next, a representative set of aryl chlorides and triflates was examined as possible partners for the C–P coupling. As exemplified in Scheme 2, because of the lower reactivity of aryl chlorides compared with aryl bromides, long reaction times were required for aryl chlorides, and, generally, decreased yields were observed for aryl chlorides and triflates (from chlorides 4: 46%‒69% yields; from triflates 5: 49%‒65% yields). However, the C–P coupling reactions of 2-chloropyridines 4q and 4r can produce the corresponding products in high yields (3qa: 75% yield; 3ra: 81% yield).

Then, the scope of the phosphine oxide substrate was evaluated. As shown in Scheme 3, a series of P–H compounds, including H-phosphinates, H-phosphonates, and secondary phosphine oxides, with varied electronic and steric properties, can be well applied as efficient substrates, affording the corresponding organophosphorus products with excellent levels of efficiency (3ab-3ae: 90%‒98% yields). However, aliphatic secondary phosphine oxides failed to participate in this transformation at the current stage for unknown reasons.

To demonstrate the practicality of this protocol, the coupling reaction of 1a and 2a was performed on a gram scale (10 mmol) under the standard reaction conditions, finally affording product 3aa (3.01 g) in an excellent yield (Scheme 4, Eq. (1)). Moreover, considering the importance of vinylphosphine motifs in polymer science and in the synthesis of heterocycles [52-56], whether vinyl bromides could be used as suitable coupling partners for this coupling was considered. Alkyl, aryl, and diaryl-substituted vinyl bromides were successfully converted into alkenyl phosphine oxides in good yields under the standard dual catalysis conditions (Scheme 4, Eq. (2): 7aa-7ca, 67%-93% yields). In addition, it was found that the CdS photocatalyst can be easily recovered through a simple filtration and washing with DMA after each run. Subsequently, substrates 1a and 2a, together with a Ni salt, a ligand, Et₃N, and DMA, were added for a new run. As shown in Fig. 2, the catalyst could be recycled at least five times without any significant loss of efficiency (note: a control experiment revealed that no desired product is observed in the absence of a Ni catalyst and the ligand after each run).

Furthermore, an attempt was made to gain insight into the reaction mechanism. When 2.0 equiv. of the radical scavenger 2, 2, 6, 6-tetramethylpiperidine-1-oxyl (TEMPO) was added to the reaction under standard conditions, the formation of 3aa was completely inhibited, and the phosphine-TEMPO adduct was detected by HRMS, which indicated the involvement of radicals in this reaction. Based on this result and previous reports [30, 32-44, 51], a plausible reaction mechanism was proposed for the formation of the desired C–P bond (Fig. 3). Upon exposure to visible light, isomer A from substrate 2a was oxidized by the excited CdS photocatalyst via injection of an electron into the conduction band of CdS. The formed radical-cation intermediate was subsequently deprotonated by a base to afford the P-centered radical B. Concurrently, the oxidative additive of the aryl halide to Ni(0) catalyst C would deliver the electrophilic Ni(Ⅱ) intermediate D, which could rapidly capture the P-centered radical B to form the Ni(Ⅲ) complex E, thereby setting the stage for C–P bond-forming reductive elimination to afford the desired product 3aa. Finally, the interface of the two catalytic cycles (F ↔ CdS^•-) via single-electron transfer would regenerate the base state of CdS photocatalyst and the Ni(0) catalyst C. However, it was recognized that an alternative mechanism, where the generated P-centered radical is trapped by Ni⁰ species before oxidative addition, cannot be ruled out at this current stage.

In conclusion, a visible-light-induced heterogeneous CdS/Ni synergistic catalysis was developed for efficiently constructing C–P bonds. This method employs readily available aryl and vinyl halides, as well as aryl triflates, to produce organophosphorus compounds with good-to-excellent yields. Notably, the recyclable photocatalyst and inexpensive catalyst system, together with a broad substrate scope and high functional group tolerance, make the updated C–P bond formation protocol practical and attractive for synthetic chemists.

The authors declare no competing financial interest.

We are grateful to the National Science Foundation of China (21822103, 21820102003, 21772052, 21772053, 21572074, 21472057), the Program of Introducing Talents of Discipline to Universities of China (111 Program, B17019), the Natural Science Foundation of Hubei Province (2017AHB047), and the International Joint Research Center for Intelligent Biosensing Technology and Health for support of this research.