Introduction

The carbon–oxygen (C–O) bond-forming reactions are of particular interest in the field of organic synthesis, as C–O bonds are prevalent in many natural products, pharmaceuticals, and agrochemicals1,2,3. Although Pd-based C–O cross-coupling catalysts are well established4,5,6,7, there has been a great interest in developing sustainable catalytic systems based on earth-abundant nickel. In this regard, photoredox/nickel dual catalysis and light-promoted Ni catalysis have emerged as powerful strategies for challenging C–O bond formations8,9,10,11. Accordingly, an in-depth mechanistic understanding of Ni-mediated photocatalysis is essential for rational reaction design and optimization. From a mechanistic viewpoint, while the photocatalytic cycle is well-understood, the Ni-mediated redox cycle remains elusive as intermediates in various oxidation states (from Ni0 to NiIV) have been proposed12. For example, the involvement of paramagnetic NiI and NiIII species has been commonly implicated, yet such intermediates have not been thoroughly investigated, and the key catalytic steps of oxidative addition, transmetalation, and reductive elimination have rarely been observed directly10,13,14.

Our group has employed tetradentate azamacrocycle N,N’-dialkyl-2,11-diaza[3.3](2,6)pyridinophanes (RN4) ligands to isolate and characterize NiIII species capable of C–C15,16,17,18 and C–heteroatom15,19,20 bond formation reactions, and to probe the involvement of high-valent Ni species in these reactions. Since the most ubiquitous ligands in Ni cross-coupling and photocatalytic reactions are the bidentate bipyridine ligands8,9,10,11,21, the RN4 ligands were thought to be less effective ligands for catalytic applications due to the crowded environment around the Ni center, which is also expected to disfavor the formation of NiI species. By contrast, a bidentate ligand structure is not suitable for stabilizing high-valent Ni species, thus hampering the investigation of such Ni intermediates22. Considering these aspects, we sought to identify an optimal ligand framework positioned in between bipyridine and RN4 ligands in terms of denticity, molecular structure, and functionality, and thus the pyridinophane tridentate RN3 ligands were targeted (Fig. 1a). These RN3 ligands are structurally analogous to the RN4 ligands, possessing a rigid aromatic pyridinophane framework and containing only one flexible chelating arm that allows for either a κ2 or κ3 coordination. Surprisingly, such RN3 ligands have never been synthesized to date, likely due to the lack of an efficient synthetic route. Several other well-known fac-capping tridentate N-donor ligands such as tris(pyrazoyl)borate (Tp–)22, tris(pyrazolyl)methane (Tpm)23, or 1,4,7-trimethyl-1,4,7-triazacyclononane (Me3TACN, Fig. 1b)24 have been reported to stabilize high-valent organometallic Ni complexes, or even an organometallic NiI complex25, yet these ligands do not seem to stabilize mononuclear LNiX2 (X = halide) complexes, dinuclear halide-bridged or homoleptic 2:1 L:Ni complexes being generated instead26,27,28.

Fig. 1: Comparison of different N-donor ligands.
figure 1

a Tridentate ligand framework design. b Representative tridentate N-donor ligands commonly used for stabilizing uncommon Ni oxidation states and the RN3 tridentate ligands developed in this work.

Full size image

In this work, we report the development of N-alkyl-2-aza[3.2](2,6)pyridinophanes (RN3, R = Me or iPr) ligands, which embody our design criteria for photocatalytic applications and mechanistic studies. The corresponding (RN3)Ni complexes are active C–O cross-coupling catalysts upon light excitation and without an added precious-metal photocatalyst. Importantly, we show that the utilization of the tridentate RN3 ligands allows for a comprehensive examination of the Ni intermediates proposed in the individual steps of the cross-coupling catalytic cycle: oxidative addition, transmetalation/ligand exchange, and reductive elimination, all these steps involving paramagnetic NiI or NiIII species.

Results and discussion

Ligand development

In 1958, Overberger et al. reported the reduction of N-nitrosodibenzylamines by Na2S2O4 to generate the hydrocarbon product with evolution of N229. However, this reaction has received little attention since then, and only one report by Takemura et al. in 1988 exploited the N-nitroso reduction reaction to obtain [2.2]cyclophane derivatives30. We speculated that this N-extrusion reaction would provide an efficient synthetic pathway for tridentate pyridinophane ligands. Our synthesis starts with the unsymmetrical pryridinophane precursor N-tosyl-2,11-diaza[3.3](2,6)pyridionphane (TsHN4, Fig. 2a)31. For the nitrosylation step, TsHN4 was treated with HNO2 to generate the N-nitroso intermediate TsNON4 in 85% yield. Subsequent reductive elimination of nitrogen from the N-nitrosoamine was performed with Na2S2O4 under basic conditions, converting TsNON4 to TsN3 in 75% yield. Additional steps of detosylation and methylation or isopropylation afforded the alkylated ligand variants MeN3 or iPrN3 in ~45% overall yields, confirming that further functionalization can be easily achieved. The N-nitroso reduction conditions are mild, suggesting a broad functional group tolerance. Since all reaction steps do not require extensive purification and their yields are high, the development of this synthetic route provides an opportunity to access a variety of ligand structures by further functionalization or starting from different secondary amine precursors. Notably, Levin et al. have recently reported the development of an anomeric amide reagent for the oxidative extrusion of nitrogen from dibenzylamines via a isodiazene intermediate to yield C–C coupled products32. By comparison, our approach employs a reduction of the N-nitrosoamine intermediate and only uses simple and commonly available reagents, and thus allowing for bis-2-picolylamines to be used as substrates.

Fig. 2: Synthesis and structural characterization.
figure 2

a Synthesis of RN3 pyridinophane ligands. b Synthesis of (RN3)NiCl2 complexes (1a and 1b) and c ORTEP representation of the 1a obtained by X-ray diffraction (ellipsoids are shown at 50% probability and the hydrogen atoms omitted for clarity). Selected bond lengths (Å): Ni–Npy 2.058, Ni–Namine 2.056, Ni–Cl 2.322. d Space-filling model of 1a highlighting the steric protection of the Ni center by the H atoms of the ethylene bridge.

Full size image

Synthesis and characterization of (RN3)NiCl2 complexes

The reaction of MeN3 or iPrN3 with Ni(DME)Cl2 in CH2Cl2 generated the green NiII-dichloride complexes (MeN3)NiCl2 (1a) and (iPrN3)NiCl2 (1b), respectively (Fig. 2b). This is in contrast to other tridentate N-donor ligands such as Tp–, Tpm, or Me3TACN, which tend to form dinuclear halide-bridged or homoleptic 2:1 L:Ni complexes instead of LNiCl2 complexes26,27,28. Both complexes 1a and 1b are paramagnetic, as shown by the paramagnetic NMR spectrum of 1b (Supplementary Fig. 11), and the effective magnetic moment μeff of 3.03 μb determined using the Evans method is consistent with a triplet ground state (S = 1), as expected for a high-spin NiII center33,34. The obtained magnetic moment value is similar to those of triplet ground state Ni(bpy)X2 (X = halide) complexes35, although these complexes adopt a tetrahedral geometry36. The molecular structures of 1a and 1b were determined by single crystal X-ray diffraction (Fig. 2c), to reveal similar coordination environments around the Ni center, except that the structure of 1b exhibits slight distortions due to the bulkier iPr substituent (Supplementary Fig. 55). The structure of 1a reveals a five-coordinate NiII center in a square pyramidal geometry with a calculated trigonality index parameter τ of 037, with an average Ni–Npy bond length of 2.058 Å and an axial Ni–Namine bond length of 2.056 Å, which are similar to those of the related complex (tBuN4)NiCl238. A closer look at the space-filling representation of 1a, as shown from the bottom view and along the Ni–Namine axial axis (Fig. 2d), reveals that the Ni center is sterically protected by the H atoms of the ethylene bridge and thus could lead to an increased stability of the high- or low-valent Ni intermediates (see below).

The electrochemical properties of (RN3)NiCl2 complexes were investigated by cyclic voltammetry (CV, Fig. 3c, Supplementary Figs. 13 and 14). The CVs of 1a and 1b in MeCN reveal a pseudoreversible oxidation wave at 0.71 V vs Fc+/0, which is assigned to the NiIII/II couple, as well as an irreversible reduction wave at −1.20 V vs Fc+/0 for 1a and −1.22 V for 1b, followed by a pseudoreversible reduction wave at −1.60 V vs Fc+/0 for 1a and −1.72 V for 1b, respectively. The first reduction wave is proposed to correspond to the conformation of the Ni complex in which the RN3 ligand adopts a κ2 binding mode, in line with the flexibility of the pyridinophane ligands observed previously24,39. The electrochemical irreversibility of the second reduction event is likely due to the loss of a Cl− ligand and the change in coordination geometry from square pyramidal to tetrahedral upon formation of the NiI species (i.e., an EC mechanism: an electron transfer followed by a chemical reaction)40. It is important to note that the oxidation and reduction potentials for 1a and 1b are similar to those of (dtbbpy)NiCl2, which exhibits an oxidation wave at 0.56 V vs Fc+/0 and an irreversible reduction wave at −1.37 V vs Fc+/0, followed by a pseudoreversible reduction wave at −1.74 V vs Fc+/0 (Supplementary Fig. 16). By comparison, while for (MeN4)NiCl2 the NiIII/II oxidation event was observed at 0.52 V vs Fc+/0, no reduction event was observed up to −2.00 V vs Fc+/041, supporting our hypothesis that tetradantate ligands cannot stabilize NiI species to a large extent. The accessible reduction events observed for (RN3)NiCl2 suggest that for these complexes the NiI oxidation state is accessible via one-electron reduction with a chemical reductant and thus it can be exploited in reactivity and mechanistic studies. When compared to (dtbbpy)NiCl2, the (RN3)NiCl2 complexes exhibit similar oxidation and reduction potentials, which is essential for photocatalysis, yet they possess a structural advantage by containing the flexible axial amine arm that is beneficial for characterization of reaction intermediates (see below).

Fig. 3: Characterization of 1a.
figure 3

a Experimental (black line) and simulated (red line, using Gaussian peak shapes with a full width at half maximum FWHM of 0.33 eV) UV–vis spectrum of 1a with overlaid relevant oscillator strengths (solid blue bars) from TD–DFT calculations. b Natural transition orbitals (NTOs, 0.05 isocontour value) associated with the visible absorption bands of 1a. c Cyclic voltammograms (CVs) of 1a in 0.1 M TBAPF6/MeCN (scan rate 0.1 V/s). d A simplified excited-state diagram for 1a, showing the population of the lower energy excited state, a 3MC d-d state.

Full size image

The UV-visible absorption spectra of 1a and 1b were then obtained and also analyzed computationally. The UV-vis absorption spectrum of 1a shows an intense band around 380 nm and a less intense band around 640 nm (Fig. 3a). To probe the nature of the electronic transitions, time-dependent density functional theory (TD-DFT) calculations were performed. As the ground state of 1a has triplet spin multiplicity, the higher energy absorption feature (380 nm) is attributed to a combination of a spin-allowed Cl → RN3 ligand-to-ligand charge transfer (3LLCT) and a Ni → RN3 metal-to-ligand charge transfer (3MLCT), while the lower energy absorption band (640 nm) is attributed to a metal-centered (3MC) d-d electronic transition (Supplementary Figs. 63–66). The energy of the 3MC d-d transition (640 nm ≈ 1.93 eV) is consistent with the electrochemical gap observed by CV (0.71 V − (−1.20 V) = 1.91 eV), further supporting that the lowest excited state (T1) has primarily MC d–d character. For a qualitative representation of the specified transitions between the ground and excited states, we performed a natural transition orbital (NTO) analysis that provides a localized picture of the transition density matrix (Fig. 3b)42. The NTO analysis supports the assignment of the 380 nm transition to a Tn excited state with a mixed LLCT/MLCT character, while the 633 nm transition involves an electron excitation between predominantly metal-based orbitals, thus indicating a MC (d–d) transition. For 3d transition metal complexes the MC (d–d) state usually lies lower in energy than the CT state(s), due to the intrinsically low ligand field strength, and thus resulting in the relaxation of the CT state to the lower-lying MC (d–d) state43,44. Therefore, we propose that on initial population of the MLCT/LLCT state is followed by fast relaxation to the lower MC (d–d) state with an appreciable Ni–Cl σ* orbital character (Fig. 3d), which then promotes a metal–ligand homolytic bond cleavage45,46, as observed recently by Doyle et al. for a similar Ni system47. Since 1a shows a prominent absorption band at 380 nm, a purple LED lamp (λmax = 390 nm) was chosen as a light source for the direct light-promoted C–O coupling reaction (see below). Finally, the UV–vis absorption spectrum and TD-DFT calculation results for 1b are similar to those for 1a, indicating a minimal effect of the N-substituent on the electronic structure (Supplementary Figs. 67–72).

Photocatalytic C–O coupling reactions mediated by (RN3)NiCl2 complexes

To probe the relevance of the (RN3)NiCl2 complexes in photocatalysis, we evaluated their light-promoted catalytic activity in C–O cross-coupling without any additional photocatalyst. Our initial efforts focused on optimizing the reaction conditions using methanol and 4-bromoacetophenone (Table 1 and Supplementary Tables 2–8). In the presence of 2 mol% (MeN3)NiCl2 (1a) and quinuclidine in THF, the desired C–O coupled product 4-methoxyacetophenone was obtained in 95% yield under purple LED (λmax = 390 nm) irradiation (Table 1, entry 1). Control experiments revealed that the reaction did not proceed without Ni catalyst, light, or base (Table 1, entries 2–4). The desired product was not observed when the reaction was irradiated with blue LED, suggesting the crucial role of the excitation energy (Table 1, entry 5). While inorganic bases such as K2CO3 were not effective, possibly due to the precipitation of halides required to stabilize the NiI intermediate14 or the unproductive decomposition of the NiIII intermediates (Supplementary Figs. 46–47), DABCO was effective but proceeded at a slightly slower rate and thus requiring a longer reaction times (Table 1, entries 6–8). Switching from 1a to 1b or (MeN3)NiBr2, lowering of catalyst loading to 0.2 mol%, or the addition of MeN3 and Ni(DME)Cl2 separately instead of 1a gave comparable yields (Table 1, entries 9–12). Notably, a 0.2 mol% catalyst loading is significantly less (~25–50 fold decrease) than the loadings used in other Ni-photocatalyzed C–O cross-couplings10,11. Moreover, other alcohols such as 1-hexanol and benzyl alcohol could also be employed (Supplementary Table 7, 86–90% product yield). The combination of 1a and heterogenous Zn0 or Mn0 without light was ineffective in our conditions, likely due to the slow electron transfer kinetics or insufficient generation of the active species during catalysis48,49. Interestingly, the use of a Ni complex supported by another tridentate N-donor ligand 1,4,7-triosopropyl-1,4,7-triazacyclononane, (iPr3TACN)NiIICl2, generated the desired product only in a moderate yield (51%, Supplementary Table 8), while the (RN4)NiIICl2 complex supported by the tetradentate ligand MeN4 yielded a negligible amount of product, validating our ligand design principle (Table 1, entry 13).

Table 1 Light-promoted Ni catalytic reaction development and optimizationa.
Full size table

The same C–O coupling reaction was also probed under common dual Ni/photoredox catalytic conditions using various Ir or Ru photocatalysts and blue LED light (λmax = 456 nm) (Table 2). The photocatalyst screening experiments revealed that the Ir photocatalysts tested here gave the desired C–O coupled product in 80–98% yield, while the Ru catalysts showed no reactivity.

Table 2 Dual Ni/photoredox catalytic reaction optimizationa.
Full size table

EPR and reactivity studies

Several reports have proposed the oxidative addition of an organic halide to NiI to generate a NiIII species is a key step in the C–O cross-coupling catalytic cycle13,14,47, yet minimal evidence was provided and such a step has been underexplored experimentally50,51. Our Ni catalysts have accessible NiII/I redox potentials, thus we aimed to probe the reduction of NiII to NiI and the subsequent oxidative addition of an aryl halide to generate a NiIII intermediate by using X-band electron paramagnetic resonance (EPR) spectroscopy. For the EPR studies, 1b was used due to its better solubility vs 1a. Upon the reduction of 1b by 1 equiv decamethylcobaltocene (CoCp*2) at −95 °C (Fig. 4a), a NiI species 2 was formed that exhibits a rhombic EPR spectrum with gx = 2.241, gy = 2.113, and gz = 2.049 (Fig. 4d), and suggesting the presence of a metal-based radical. The detection of an EPR-active mononuclear NiI species is important since such an intermediate is essential for an efficient oxidative addition step. By comparison, in the catalytic systems involving bidentate bipyridine ligands relatively stable dinuclear NiI species are formed as side products that hinder catalysis, especially at higher substrate concentrations and catalyst loadings13,47. In our case, the NiI species 2 decays rapidly to EPR inactive species in the absence of substrate, yet no dinuclear NiI species are formed, suggesting that the tridentate RN3 ligand might prevent the formation of such catalytically inactive species due to the steric hindrance provided by the axial amine arm of the RN3 ligand (see Supplementary Figs. 58 and 59 for the space filling models of 1a and 1b). This behavior might explain why higher concentrations of substrate and catalyst can be employed herein, which is in contrast to the conditions utilized in other reported C–O cross-coupling reactions10,11,14. When 20 equiv 4-bromoacetophenone were added to a thawing solution of 2 (Fig. 4b), the resulting EPR spectrum could be simulated with two sets of parameters corresponding to two different species in a 0.6:1 ratio and that were tentatively assigned to the initial NiI species and a new NiIII species (Fig. 4e and Supplementary Fig. 32). We tentatively propose that the NiIII species is a transient 6-coordinate (iPrN3)NiIII(PhAc)BrCl (3) complex generated through the oxidative addition of 4-bromoacetophenone to 2. Moreover, when 40 equiv 4-bromoacetophenone were added to 2 (Fig. 4c), only a rhombic EPR signal with gx = 2.298, gy = 2.197, and gz = 2.037 formed (Fig. 4f), which was assigned to the proposed structure 3 and thus strongly supporting a direct conversion from NiI (2) to NiIII (3) that is promoted by the excess aryl halide. The observed EPR spectrum for 3 decays quickly upon warming up the solution for seconds at −80 °C (Supplementary Fig. 33), supporting the increased reactivity of 3 and its proposed involvement in catalysis. Finally, while the exact structure of this species 3 is not unambiguously known, it is important to note that to the best of our knowledge the direction observation of the oxidation addition of an aryl halide to a NiI species to generate a NiIII species has not been reported to date.

Fig. 4: EPR studies of the NiI species.
figure 4

a–c (left) Experimental setup for EPR detection of proposed NiI and NiIII species upon reduction of 1b and subsequent oxidative addition of aryl bromide. (right) Experimental (black) and simulated (red) EPR spectra of the reaction mixtures recorded at 77 K in 1:5 MeOH:PrCN frozen glass: d 1b + CoCp*2, e 1b + CoCp*2 + 20 equiv 4-bromoacetophenone, f 1b + CoCp*2 + 40 equiv 4-bromoacetophenone. The following parameters were used for simulation of d gx = 2.241, gy = 2.113, gz = 2.049; f gx = 2.298, gy = 2.197, gz = 2.037; e the simulations in d and f were added in a 0.6:1 ratio. The structure proposed for species 3 is tentative.

Full size image

To further probe the role of various NiIII species in the C–O coupling reaction, we independently synthesized the (iPrN3)NiII(PhAc)Br complex 4 via oxidative addition of 4-bromoacetophenone to Ni(COD)2 in the presence of iPrN3 (Fig. 5a). While 4 is unstable in solution at room temperature, rapidly decomposing to the green (iPrN3)NiIIBr2 complex, it is stable at −40 °C for several hours. The decomposition of a related nickel complex, (bpy)NiIIPhBr, to (bpy)NiIIBr2 has been observed previously and proposed to proceed via a bimolecular pathway52,53. The instability of 4 in solution prevented us from obtaining X-ray quality crystals, nonetheless since 4 is diamagnetic its structural evidence was provided by 1H NMR obtained at −40 °C, which shows a slight peak broadening due to the flexible amine arm (Supplementary Fig. 2)24. Oxidation of 4 with acetylferrocenium hexafluorophosphate (AcFcPF6) at −40 °C generates the NiIII [(iPrN3)NiIII(PhAc)(Br)]+ complex 5 (Fig. 5a). The effective magnetic moment μeff of 1.98 μb, determined using the Evans method, is consistent with a S = 1/2 ground state for 5, as expected for a NiIII center33,34. The rhombic EPR spectrum of 5 with gx = 2.356, gy = 2.158, gz = 2.059 is significantly different than that of 3 (Fig. 5e vs. Fig. 4f), solidifying the proposed 6-coordinate structure for 3 in Fig. 4c. Complex 5 was stable in solution at −35 °C and can be isolated as a red solid, yet it was recalcitrant to form single crystals. Therefore, the presence of a NiIII center was further confirmed by X-ray photoelectron spectroscopy (XPS), which shows an increase of the Ni 2p3/2 and 2p1/2 binding energies of ~1.5 eV for 5 versus 4, demonstrating the presence of a more oxidized Ni center (Supplementary Fig. 60)54.

Fig. 5: EPR studies of NiIII species.
figure 5

(left) Experimental setup for the synthesis of (iPrN3)NiII(PhAc)Br (4) and EPR detection of a [(iPrN3)NiIII(PhAc)(Br)]+ (5), b [(iPrN3)NiIII(PhAc)(Cl)]+ (6), and c and d [(iPrN3)NiIII(PhAc)(OMe)]+ (7). (right) Experimental (black) and simulated (red) EPR spectra of the reaction mixtures recorded at 77 K in 1:1:3 MeOH:CH2Cl2:Toluene frozen glass: e [(iPrN3)NiIII(PhAc)(Br)]+ (5), f [(iPrN3)NiIII(PhAc)(Cl)]+ (6), and g and h [(iPrN3)NiIII(PhAc)(OMe)]+ (7). The following parameters were used for simulation of e gx = 2.356, gy = 2.158, gz = 2.059; f gx = 2.363, gy = 2.144, gz = 2.047; g gx = 2.326, gy = 2.151, gz = 2.027 (Az(N) = 19.4 G); h gx = 2.326, gy = 2.152, gz = 2.027 (Az(N) = 19.5 G).

Full size image

The appreciable stability of 5 allowed us to investigate its reactivity. When 5 was reacted with 1 equiv tetraethylammonium chloride (TEACl, Fig. 5b), a rhombic EPR spectrum with gx = 2.363, gy = 2.144, gz = 2.047 and sharper signals than the spectrum of 5 was obtained, indicating the likely formation of the [(iPrN3)NiIII(PhAc)(Cl)]+ species 6 (Fig. 5f vs. Fig. 5e). A similar EPR signal broadening due to superhyperfine coupling to the Br atom (I = 3/2) vs the Cl atom (I = 3/2) was observed in a previous study by our group, where the structural analysis of [(tBuN4)NiIIIAr(X)]+ complexes (X = Br or Cl) confirmed the effect of the halide ligand onto the corresponding EPR spectra15. Importantly, an EPR spectrum resembling the proposed 6-coordinate species (iPrN3)NiIII(PhAc)BrCl (3) was not observed, supporting the instability of this sterically hindered species. GC-MS analysis of the solution of 6 after warming up to room temperature in the absence of MeOH reveals Ar–Cl product formation (Supplementary Fig. 37), while the reaction between (iPrN3)NiII(PhAc)Br (4) and PhICl2 generates an EPR spectrum identical to that of 6 (Supplementary Fig. 39), providing additional experimental evidence for the halide exchange step. Next, we probed the ligand exchange step with an alkoxide. The addition at −80 °C of quinuclidine in MeOH to 5 generates a species assigned as the NiIII-aryl alkoxide complex [(iPrN3)NiIII(PhAc)(OMe)]+ (7), which exhibits a rhombic EPR signal with gx = 2.326, gy = 2.151, gz = 2.027 and superhyperfine coupling to the one axial N (I = 1) donor in the gz direction of Az(N) = 19.4 G (Fig. 5g). We posit that replacement of the halides that exhibit superhyperfine interactions (Br or Cl, I = 3/2) by the methoxide ligand resolves the superhyperfine coupling to the N atom along the gz direction, while the weaker π-donor methoxide leads to a stronger Ni–Naxial interaction. The identity of 7 was further confirmed by the treatment of 5 with 1 equiv NaOMe, which generated an identical EPR spectrum (Fig. 5h and Supplementary Fig. 45). Importantly, to the best of our knowledge this is the first experimental observation of a ligand exchange reaction (either by a halide or alkoxide) occurring at a NiIII metal center, which is typically proposed as a key step in Ni-mediated C–C and C–heteroatom bond formation reactions55,56. Finally, we also note that 5 is a catalytically competent species (Supplementary Table 8) and affords the C–O cross-coupled product in 31% yield, suggesting that 5 can successfully enter the catalytic cycle. In this case, the lower yield is likely due to the slight decomposition of 5 in RT during the catalytic reaction.

Upon warming up to room temperature, species 6 and 7 undergo C–O bond-forming reductive elimination, affording 4-methoxyacetophenone together with different amounts of acetophenone (Fig. 6 and Supplementary Figs. 49–54), and consistent with a favorable reductive elimination from a NiIII species57,58. Moreover, the addition of excess anionic ligands (Cl− or OMe−) leads to increased yields of the C–O coupled product (Fig. 6) and faster decomposition of 6 and 7, as seen in the fast decay of their EPR signals (Supplementary Figs. 42 and 44). We propose that coordination of an additional anionic ligand in the presence of excess base leads to the formation of a congested 6-coordinate NiIII center due to the steric hindrance of the ethylene bridge that blocks the second axial coordination site (Fig. 2d), and consequently results in accelerated C–O reductive elimination. This is also in line with the observed rapid decay of the 6-coordinate species (iPrN3)NiIII(PhAc)BrCl (3, Fig. 4c and Supplementary Fig. 33). In addition, the rapid reductive elimination from a 6-coordinate NiIII center may preclude β-hydride elimination (Fig. 6, entry 6), which is likely the source of the formation of b. Overall, these EPR and reactivity studies provide a complete picture of the NiI/NiIII catalytic cycle for the C–O cross-coupling reaction, with experimental evidence for each essential step in this process: oxidative addition, transmetalation/ligand exchange, and reductive elimination.

Fig. 6: Stoichiometric reactivity studies.
figure 6

Stoichiometric reactivity of 5 with TEACl, quinuclidine, or NaOMe to yield the C‒O coupled product through reductive elimination from NiIII intermediates.

Full size image

Mechanistic considerations

Finally, based on the photoredox/Ni dual catalytic studies under blue LED light (Table 2), we evaluated the thermodynamic feasibility of the photocatalytic process in the C–O coupling reaction. A comparison of the redox potentials of the Ni complexes and photocatalysts implies that a reductive quenching pathway is not operating, as the observed production yields contradicts the trend of the oxidation strength of excited photocatalysts (see Supplementary Discussion). Instead, an oxidative quenching pathway is proposed to be operative, in which the initial step is the single-electron reduction of NiII by the excited photocatalyst, and this is in line with our results in which all photocatalysts with E(M+/*M) < −0.87 V afford the C–O coupling product. By combining the insights obtained from EPR and reactivity studies and the photoredox/Ni dual catalysis, two photocatalytic mechanisms are proposed (Fig. 7). The (RN3)NiII complex is reduced to a NiI species either by irradiation with purple LED light (Fig. 7a) or via a single electron transfer (SET) from the excited photocatalyst (Fig. 7b), followed by an oxidative addition of the aryl halide to form a transient (RN3)NiIIIAr(Br)X species. For both of the direct light-promoted Ni catalysis and dual Ni/photoredox catalysis, the (RN3)NiIII species can undergo a ligand exchange with the alcohol substrate in presence of excess base to generate the (RN3)NiIIIAr(OR)X species, which is then proposed to undergo a subsequent reductive elimination to yield the C–O coupled product and regenerate a (RN3)NiI species, completing the catalytic cycle. As the proposed catalytic cycle in Fig. 7a suggests a NiI/NiIII dark catalytic cycle, then the theoretical quantum yield for the light-promoted Ni catalysis should be above 1 (ΦNi > 1). The measured quantum yield was determined experimentally to be below 1 (ΦNi < 0.26, Supplementary information page 32), which is likely due to the less efficient absorption of light in the heterogeneous reaction mixture that contains insoluble material and thus prevents an accurate determination of the number of photons absorbed by Ni under catalytic conditions. In addition, the quantum yield (ΦNi < 0.26) obtained by actinometry could be significantly overestimated. Given that there are several possible side reactions that would generate off-cycle NiII species, if ΦNi ≪ 1, then the dark NiI/NiIII dark catalytic cycle may not be able to compensate for the deleterious side reactions. While values of Φ > 1 were reported previously for similar photoredox/Ni dual catalysis C–O cross-coupling reactions13, a dark NiI/NiIII cycle cannot be unambiguously ruled out. To provide further evidence for a dark catalytic cycle, we performed a light on/off experiment and employed ReactIR to monitor in real time the consumption of the starting material 4-bromoacetophenone, which exhibits a characteristic IR C=O stretch at 1690 cm−1 (Supplementary information page 34). When the reaction was performed with the light on 25% of the time and the light off 75% of the time, the reaction yield was only marginally lower than when the reaction was performed under continuous illumination (Supplementary Table 9), while the ReactIR monitoring clearly showed the continuous consumption of the substrate during the light-off periods (Supplementary Fig. 25). Overall, these results strongly support a self-sustained NiI/NiIII cycle, although continuous irradiation during catalysis seems to be beneficial to regenerate any depleted (RN3)NiI species due to a NiI/NiIII comproportionation process.

Fig. 7: Proposed photocatalytic mechanisms.
figure 7

Proposed mechanisms for the C–O cross coupling reaction for a direct light-promoted Ni catalysis with purple LED light (proposed catalyst inactivation pathway in absence of light is shown in dashed arrow), and b dual Ni/photoredox catalysis with blue LED light. X = Cl or Br, L = MeN3 or iPrN3.

Full size image

To further probe the proposed NiI/NiIII comproportionation reaction that can generate off-cycle NiII species, we investigated this process via EPR and NMR spectroscopy. The reaction between 5 and in situ generated 2 (i.e., 1b + CoCp*2) leads to the rapid decay of the EPR signal, indicating the decomposition of 5 and conversion into an EPR inactive species (Supplementary Fig. 26). In addition, NMR analysis of the reaction confirmed the formation of (iPrN3)NiIICl2 complex 1b as the terminal product of comproportionation, along with several organic products (Supplementary Figs. 27–30). These EPR and NMR results thus support the deleterious NiI/NiIII comproportionation leading to off-cycle NiII-dihalide species, which can re-enter the catalytic cycle upon photoexcitation and is also consistent with previously reported computational13 and experimental reports59. For the proposed dual Ni/photoredox catalysis (Fig. 7b), the (RN3)NiI species can directly re-enter the catalytic cycle, or it can be oxidized to the (RN3)NiII complex via an SET process with the oxidized photocatalyst IrIV, thus regenerating the IrIII photocatalyst. Overall, both mechanisms are proposed to involve both NiI and NiIII intermediates, which are involved in the oxidation addition step, and the ligand exchange/transmetalation and reductive elimination steps, respectively. The main difference between the two pathways is the generation of the catalytically active (RN3)NiI species, either through light-promoted reduction of the NiII precursor (Fig. 7a), or by the excited photocatalyst (Fig. 7b).

In conclusion, herein we report the tridentate pyridinophane RN3 ligands and their Ni complexes, which were employed in mechanistic studies of the photocatalytic C–O cross-coupling reaction. The (RN3)NiCl2 complexes are active photocatalysts for the C–O cross-coupling in the absence of a precious metal photocatalyst. Employing these developed RN3 ligands allowed us to directly probe the key steps of the C–O cross-coupling catalytic cycle involving paramagnetic Ni species: the oxidative addition of an aryl halide to a NiI species to generate a NiIII species, the ligand exchange/transmetalation step at a NiIII center, and the C–O bond forming reductive elimination from a NiIII species. Overall, the present work suggests that the RN3 ligands can lead to the development of improved Ni catalysts, and are also a practical platform for detailed mechanistic studies of related Ni-catalyzed reactions.

Methods

General procedure for the C‒O cross-coupling reaction

In a N2-filled glovebox, 4-bromoacetophenone (0.4 mmol, 1.0 equiv), quinuclidine (0.44 mmol, 1.1 equiv), MeOH (1.6 mmol, 4 equiv), (MeN3)NiCl2 (2 mol%), and a magnetic stir bar were added into 0.4 mL THF in a vial. The vial was irradiated with one Kessil purple LED lamp (λmax = 390 nm, max 52 W) under fan cooling. After 24 h, 1,3-benzodioxole (0.4 mmol) was added to the reaction mixture at a standard, and the reaction mixture was analyzed by 1H NMR in CDCl3 and GC-FID to determine the yield of C-O coupled product. Authentic 4-methoxyacetophenone was purchased from AK scientific and used to determine the retention time and response factor for GC-FID quantification.

General procedure for EPR detection of 2 and 3

In a N2-filled glovebox, a MeOH solution of 1b was layered on top of a frozen PrCN solution of CoCp*2 and 4-bromoacetophenone in the EPR tube. The mixture was quickly frozen and taken outside of the glove box. The reaction mixture was mixed at −95 °C, quickly frozen at 77 K, and then the sample was warmed up for ~10 s to −95 °C to allow for a complete reaction.

General procedure for EPR detection of 5, 6, and 7

In a N2-filled glovebox, an EPR tube was charged with 1:3 CH2Cl2:Toluene or CH2Cl2:PrCN solution of 5 and frozen at 77 K. A MeOH solution of TEACl, quinuclidine, or NaOMe was added over the frozen solution of 5. After taking out the EPR tube from the glovebox, an initial EPR spectrum was taken at 77 K (for 5). After a quick shake of the tube for 10 s at −95 °C (for 6 or 7), the sample was frozen at 77 K and the EPR spectrum was collected.