To optimize reaction conditions, the cross-coupling of N-methylaniline (1) and methyl 4-bromobenzonate (2) is picked as the model reaction. Catalytic tests with different conditions were carried out in an undivided PEC cell with a three-electrode configuration (Figure S1). The BiVO4 photoanode was prepared using previously reported work36. Concurrently, a Ni foam electrode and a Ag/AgNO3 electrode were employed as the cathode and the reference electrode, respectively. In a typical test, the experiment was performed in N2 atmosphere under the irradiation of a 465 nm blue light emitting diode (LED) with a constant external bias of − 0.4 V vs. Ag/AgNO3 (Fig. 1b).
Initially, the reaction was conducted using acetone as the solvent and LiBr as the electrolyte. NiBr2 glyme and the ligand 4,4′-Di-tert-butyl-2,2′-dipyridyl (dtbppy) were added to the reaction system simultaneously for the formation of Ni catalyst. 1,4-Diazabicyclo[2.2.2]octane (DABCO), a hydrogen-atom-transfer (HAT) mediator, was introduced to the PEC cell to facilitate the electron transfer between the photoanode and amine (Fig. 1c, entry 1). Such recipe led to a high yield of 82% towards the target product methyl 4-(methyl(phenyl)amino)benzoate (3). To investigate the solvent effect, acetone was in turn replaced by acetonitrile (MeCN), dimethylacetamide (DMA) and dichloroethane (DCE), but poor results were observed (Fig. 1c, entry 2–4). As for the influence of electrolyte, when LiBr was changed to nBu4NBr or nBu4NBF4, the yield dropped below 60% (Fig. 1c, entry 5–6). Additionally, trace product was obtained when Pt plate was employed as the cathode (Fig. 1c, entry 7). Changing Ni precatalyst or reducing the ligand loading amount exhibited limited impact on the reaction, with yields consistently exceeding 70% (Fig. 1c, entry 8–9). However, replacing dtbppy with 2,2'-bipyridine (bpy) almost stopped the reaction from proceeding (Fig. 1c, entry 10).
The role of each component in the reaction was explored. As expected, no reaction happened when Ni precatalyst or ligand was omitted (Fig. 1c, entry 11). Similarly, only trace product was found in the absence of DABCO (Fig. 1c, entry 12) or electricity (Fig. 1d, left column). These results demonstrate that Ni catalyst, DABCO and electricity all play crucial roles in the catalysis. Electrocatalysis under dark condition was also carried out (Fig. 1d, middle column), and no product was detected until the potential was raised up to + 1.5 V vs. Ag/AgNO3. Furthermore, due to the high overpotential of electrocatalysis, the yield and selectivity of product 3 was merely 16% and 62% at + 1.5 V vs. Ag/AgNO3, respectively (Table S1). Notably, introducing light to the electrocatalysis dramatically lowered the potential to -0.4 V vs. Ag/AgNO3, and consequently improved the yield and selectivity of product 3 to 82% and 89% (Fig. 1d, right column).
Subsequently, linear sweep voltammetry (LSV) and cyclic voltammetry (CV) were applied to investigate the PEC properties of the reaction systems (See Figure S2-S4 for more details). As shown in Fig. 2a, the LSV diagram of BiVO4 anode shows significant onset potential decrease when light is introduced to the system, which highlights the superiority of PEC cell. Besides, adding DABCO also contributes to the onset potential reduction (Fig. 2a, pink curve vs. blue curve), indicating that DABCO serves as a mediator to promote the charge transfer between the anode and the substrate. The function of DABCO is further revealed by CV curves (Fig. 2b-c). The considerably lower oxidation potential of DABCO (EOx,DABCO = + 0.41 V vs. Ag/AgNO3) than that of the substrate N-methylaniline (EOx,N−methylaniline = + 0.53 V vs. Ag/AgNO3) demonstrates that it is easier to be oxidized by the anode, which well accords with LSV results. Moreover, in the presence of N-methylaniline (Fig. 2b, pink curve), the intensity of DABCO oxidation peak obviously exceeds the original one (Fig. 2b, orange curve), suggesting the interactions between oxidized DABCO and N-methylaniline. Furthermore, the EC properties of Ni catalyst were studied via CV test. As illustrated in the right diagram of Fig. 2c, the CV curve of Ni catalyst displays a reduction peak at Ered,1 = − 1.73 V vs. Ag/AgNO3, which is attributed to the Ni(II/I) redox couple (orange curve). After the inclusion of aryl bromide, the intensity of reduction peak increases significantly (pink curve). The change in reduction signal reveals the oxidation addition of aryl bromide on Ni(I) catalyst. Meanwhile, the concurrent observation of enhanced Ni(I/0) redox signals (Ered,2 = − 2.24 V vs. Ag/AgNO3) and the emergence of a new pair of reversible redox peaks at − 2.0 V vs. Ag/AgNO3 fortifies the existence of Ni(I)-participated oxidation addition15, 47.
To dig deep into the reaction pathway, several control experiments and radical trapping experiment were designed and carried out. When the control experiment was conducted in a divided cell where all the reagents except for LiBr were placed in the cathode chamber (Fig. 3a, top), only trace 3 was obtained, displaying the importance of anode oxidation reaction. The addition of butylated hydroxytoluene (BHT), a radical-scavenging agent, greatly suppressed the generation of 3 (Fig. 3a, middle). Instead, BHT adduct 4 was detected by high-resolution mass spectrometry (HR-MS), demonstrating that the nitrogen-centered radical was an important intermediate in the PEC process. Control experiment in the absence of Ni catalyst led to 1,2-dimethyl-1,2-diphenylhydrazine (5) from coupling of 1 (Fig. 3a, bottom), offering another evidence about the existence of radical intermediate. This result indicated that the Ni catalyst played a critical role in the cross-coupling of 1 and 2. Further getting rid of DABCO from the reaction system resulted in the vanishment of 5, showing that DABCO took part in the generation of amine radical.
Based on the experimental results above, a plausible reaction mechanism of PEC C–N coupling was proposed (Fig. 3b). Under light irradiation, the electron in the valence band (VB) of BiVO4 photoanode is excited to the conduction band (CB) and leaves a hole behind. After migrating to the surface, the photogenerated hole oxidizes DABCO (6) to DABCO cation radical (7) via a single-electron transfer reaction. The HAT between DABCO amine radical and the aniline substrate gives rise to the protonated DABCO (8) and the aniline radical (9). The deprotonation of the former accomplishes the organocatalytic cycle and the latter participates in the Ni-catalytic cycle. Meanwhile, the Ni-catalytic cycle starts from the reduction of NiII catalyst 10 at the Ni foam cathode. As-generated NiI intermediate 11 involves the oxidative addition of aryl bromide to form NiIII intermediate 12, which then receives an electron from the cathode to generate NiII intermediate 13. Afterwards, the NiII intermediate is added by the aniline radical to produce NiIII intermediate 14. Finally, a reductive elimination occurs and the coupling product is formed.
Density functional theory (DFT) simulations were performed (See Supporting Information for details) to verify the rationality of the proposed reaction mechanism. The HAT from aniline to DABCO cation radical has a low activation energy of ∆G‡HAT−TS = + 9.4 kcal mol–1 (Figure S5), indicating the fast kinetics of this process. As shown in Fig. 3c, the oxidative addition step of aryl bromide on 11 is proven to be a fast step (∆G‡11→TS1 = + 9.4 kcal mol–1) with a Gibbs free energy change of ∆G11→12 = − 4.4 kcal mol–1. The reduction step for the intermediate 12 is calculated to be endothermic (∆G12→13 = + 22.9 kcal mol–1). Once NiII intermediate 13 is formed, it is possible to be attacked by N-methylaniline or its radical. Noteworthily, DFT simulation elucidates that compared with the traditional amine attacking pathway (13→TS2’→15) in electrocatalysis, the radical attacking pathway (13→TS2→14) has a much lower energy barrier (∆G‡13→TS2 = + 5.0 kcal mol–1 vs. ∆G‡13→TS2’ = +9.9 kcal mol–1) and is more thermodynamically favorable (∆G13→14 = − 9.2 kcal mol–1 vs. ∆G13→15 = − 2.8 kcal mol–1), revealing the intrinsic advantage of PEC approach. Finally, the reductive elimination is a thermodynamically favorable step (∆G14→3 = − 23.5 kcal mol–1) with an energy barrier of ∆G‡14→TS3 = + 11.3 kcal mol–1.
Armed with mechanistic insights, we set out to explore the generality of our PEC C–N cross-coupling method. The scope of aryl bromide substrate was first investigated. As displayed in Fig. 4, the para-position exhibits excellent group tolerance under the standard condition. When the ester group is replaced by ketocarbonyl, phenyl and trifluoromethyl groups, good yields are obtained (16–18). Employing 1-bromo-4-chlorobenzene as substrate results in the selective production of 19, implying that the Ni-catalytic cycle has a high selectivity on activating halide groups. Thio- and cyano-containing substrates are also successfully coupled with N-methylaniline with slightly lower yields (20, 21). Even for bulky substituent such as tert-butyl or carbazole, the reaction takes place with moderate yields (22, 23). Similar tolerance is observed for the meta- and multi-substituted substrates (24–31). Bromobenzene, bromonaphthalene and some more complicated bromated (hetero) aromatic compounds were further tested and fair to good yields were observed (32–38). In regard to the scope of amine substrate, aniline derivatives, including para- and meta-substituted N-methylaniline, N-Methyl-2-naphthylamine and some cyclic amines, were tried (39–50). The catalytic performance varies greatly among different substrates, likely because the stability of amine radical is sensitive to the electronic effect and the steric hinderance.
Last but not least, we showcased the synthetic applicability of this synergistically catalytic approach with a few examples of late-stage functionalization. Impressively, the analogue of phytol and L-menthol, two types of important natural product, are readily synthesized using this method (51, 52). Bromides derived from pharmaceutical chemicals (canagliflozin, bedaquiline and celecoxib precursor) also incorporate well into this protocol (53–55). Additionally, fenofibrate, indomethacin and etoricoxib, three famous drugs with aryl chloride moiety, are found to be compatible with this system, evincing that it is promising to apply the PEC mode to other halides (56–58).