Photocatalytic C–C cross-coupling reactionsA series of two-coordinated Au(I) complexes having N-heterocyclic carbene and carbazolide ligands with various alkyl substituents were selected as potent candidates for the visible light-absorbing photoredox catalysts. These complexes exhibit visible absorption due to the LLCT transition (Fig. 3 and Supplementary Fig. 1). For example, the ultraviolet–visible (UV–Vis) absorption spectrum of Au(BZI)(TMCz) has an onset wavelength of 470 nm (Fig. 3 and Supplementary Table 3). The excited state of Au(BZI)(TMCz) is long-lived, with a τobs as long as 210 ns in deaerated toluene, without any short-lived components (Supplementary Fig. 2). Au(BZI)(TMCz) exhibits ground-state oxidation (Eox) and reduction (Ered) potentials of 0.56 and −2.26 V vs SCE, respectively (Supplementary Fig. 3). The corresponding E*ox value is calculated to be −2.16 V vs SCE, being more cathodic than those of widely used photoredox catalysts such as Ptppy (−2.07 V vs SCE)36, fac-Ir(ppy)3 (−1.73 V vs SCE)37, 4CzIPN (−1.18 V vs SCE)4, and Mes–Acr+ (−0.57 V vs SCE)38. It should be emphasized that, as compared in Fig. 2, Au(BZI)(TMCz) exhibits a balance between τobs and E*ox, which is a prerequisite for a potent photoreducing catalyst.Fig. 3: Photon absorption.UV–Vis absorption spectrum of Au(BZI)(TMCz) recorded in toluene at 298 K. Inset figures denote the hole and electron distributions calculated at the CAM-B3LYP level of theory for the singlet transition of the triplet geometry of Au(BZI)(TMCz) with exclusive ligand-to-ligand charge-transfer (LLCT) transition character. The LANL2DZ and 6−311+G(d) basis sets were used for the Au atom and the other atoms, respectively.The photoredox catalytic efficacy of these Au(I) complexes was evaluated through comparisons with several established photoredox catalysts for an intermolecular C–C cross-coupling reaction between model (hetero)aryl chlorides (specifically, methyl-4-chlorobenzoate (1a) and 2-chloroquinoline (1b)) and N-methylpyrrole (2a) (Table 1). The reactions were carried out with a 0.25 mol% or a 0.1 mol% photocatalyst for 1a or 1b, respectively, and 2 equivalents of N,N-di(isopropyl)ethylamine (DIPEA) in Ar-saturated dimethyl sulfoxide (DMSO, 0.50 M) under blue LEDs (405 nm) irradiation. Notably, Au complexes without alkyl substituents in the carbene ligands, Au(BZI)(Cz) (Cz = carbazolide) and Au(BZI)(TMCz), exhibited superior reactivity, successfully yielding the desired C–C cross-coupled product 3 in 71–91% yields. The distinctly superior catalytic performance of Au(BZI)(Cz) and Au(BZI)(TMCz) compared with that of the other Au(I) complexes is attributable to their interactions with 2a, being less sterically encumbered. The added 2a results in an increase in τobs, suggesting that 2a is not engaged in photoinduced electron transfer but suppresses nonradiative relaxation of the Au(I) complexes (Supplementary Fig. 2).By contrast, the commonly used fac-Ir(ppy)3 demonstrated lower reactivity, and 4CzIPN, fluorescein, and Ptppy36 were inactive in this transformation. Optimization of the reaction parameters enabled the synthesis of the heteroaryl–heteroaryl coupled product (3ba) in 84% yield in the presence of only 0.1 mol% Au(BZI)(TMCz). The optimization results are compiled in Supplementary Table 4. Notably, this catalytic loading represents a significant advance, rearely reported in the literature. The quantum yield for the reaction of 1a with 2a, which was determined using the standard ferrioxalate actinometry, is as large as 34%. Finally, control experiments revealed that the reaction requires both the photocatalyst and irradiation with visible light (Supplementary Table 4).With the reaction parameters in hand, we explored the versatility of the Au(I) complex-catalyzed C–C coupling protocol in synthesizing various (hetero)aryl– heteroaryl products (3). Au(BZI)(TMCz), which has an E*ox as negative as –2.16 V vs SCE, enables the incorporation of a range of redox-resistant (hetero)aryl chlorides (1a, 1b, 1c, 1k, 1m, 1n, and 1q) with reduction potentials (Ereds) in the range –1.61 to –2.12 V vs SCE (see Supplementary Fig. 4 for the corresponding voltammograms), as well as bromides and iodides, in the C–C coupled products (Table 2). We observed a distinct variation in the reactivity among different (hetero)aryl halides, leading to a chemoselective process. Specifically, in the case of dihalogenated substrates 1r and 1s, the iodo group demonstrated selective reactivity over the bromo substituent, resulting in 3ra and 3sa as the major products, respectively. For 1m and 1n, which contain two chloro substituents, regioselective reactivity was noted at the chloride positioned para to the aldehyde and CF3 groups, yielding 3ma and 3na, respectively. In general, N-heteroaryl halides exhibited higher reactivities than aryl halides, even with a reduced catalyst loading (0.1 mol% vs 0.25 mol%). Furthermore, the substitution patterns did not affect reactivity: ortho-, meta-, and para-substituted aryl halides were all suitable. Notably, our photoredox catalysis protocol tolerated the presence of various functional groups, including aldehyde (3ja and 3ma), ketone (3ia), ester (3aa, 3sa, 3ab, and 3ac), and nitrile (3ka) groups, as well as medicinally significant fluoride (3pa) and CF3 (3da, 3la, 3na, 3oa) groups. In general, dehalogenation side products were also observed, albeit in negligible amounts (<5%) for bromide and iodide substrates.Table 2 Substrate scopea,bMechanistic investigationsHaving demonstrated the photoredox catalytic ability of the Au(I) complex, we sought to elucidate the role of the catalyst in the C–C cross-coupling reaction. The E*ox value of Au(BZI)(TMCz) (−2.16 V vs SCE) is more negative than the Ered value of 1a (−1.81 V vs SCE), implying that the excited state of Au(BZI)(TMCz) (denoted as [Au(BZI)(TMCz)]* hereafter) can be oxidatively quenched by 1a with the driving force for heterobimolecular one-electron transfer (−ΔGeT, −ΔGeT = e·[E*ox(Au(BZI)(TMCz)) − Ered(1a)], where e is the elementary charge and we ignore the Coulomb term because of our use of the polar solvent DMSO of 0.35 eV). By contrast, reductive quenching of [Au(BZI)(TMCz)]* (E*red = 0.46 V vs SCE) by DIPEA (Eox = 0.64 V vs SCE) is predicted to be disfavored due to the negative −ΔGeT of −0.18 eV.We validated the aforementioned thermodynamic considerations for the initial electron transfer by using photoluminescence titration experiments. As shown in Fig. 4a, the increased concentrations of 1a (0–100 mM) elicit a concentration-dependent decrease in the photoluminescence intensity of 50 μM Au(BZI)(TMCz). An analogous decrease is also observed for τobs (Fig. 4b). The quenching rate computed through the relationship, quenching rate = 1/τobs(1a) − 1/τobs(0), where τobs(1a) and τobs(0) are the τobs of Au(BZI)(TMCz) in the presence and absence of 1a, respectively, increases with the concentration of 1a (Fig. 4c). In all cases, the decays of photoluminescence intensities obey the first-order kinetics to 19–45 ns. A plot of the pseudo-first-order quenching rate as a function of the concentration of 1a is linear, with an apparent heterobimolecular quenching rate constant (kq) of 7.2 ± 0.1 × 108 M−1 s−1. Our Stern–Volmer analysis for the steady-state photoluminescence intensity data yields a similar kq of 8.4 ± 0.3 × 108 M−1 s−1 (Supplementary Fig. 5). Deconvolution of the diffusion rate constant (kdiff, 3.3 × 109 M−1 s−1 for DMSO at 298 K) from kq yields a second-order quenching rate constant (kQ) of 9.1 ± 0.1 × 108 M−1 s−1. The quantum yield for quenching (ΦQ) is as large as 95% in the presence of 0.50 M 1a, which is estimated according to the relationship ΦQ = kQ·[1a] / (kQ·[1a] + kd), where [1a] is expressed in molarity (0.50 M) and kd is the intrinsic decay rate of [Au(BZI)(TMCz)]* (2.2 × 107 s−1). In stark contrast, DIPEA does not quench the photoluminescence of [Au(BZI)(TMCz)]* even at a concentration as high as 500 mM (Fig. 4d). The rapid deactivation of [Au(BZI)(TMCz)]* by 1a can be ascribed to electron or energy transfer. We exclude the energy-transfer pathway because Au(BZI)(TMCz) not only exhibits negligible spectral overlap with 1a, 2a, and DIPEA but also exhibits a T1 energy (2.79 eV) lower than that of 1a (3.16 eV) (refer to Supplementary Fig. 6 for details). Collectively, our electrochemical and photoluminescence results strongly suggest rapid and exclusive electron transfer from [Au(BZI)(TMCz)]* to 1a. Nanosecond laser flash photolysis investigations provide direct spectroscopic evidence for electron transfer (vide infra).Fig. 4: Oxidative quenching.a Photoluminescence (λex = 380 nm) spectra of Ar-saturated DMSO containing 50 μM Au(BZI)(TMCz), recorded with increasing concentration of 1a (0–100 mM). The peak marked with an asterisk (*) is the Raman signal of the solvent. b Photoluminescence decay traces of Ar-saturated DMSO containing 50 μM Au(BZI)(TMCz), recorded with increasing concentration of 1a (0–40 mM) at a wavelength of 540 nm after picosecond pulsed laser photoexcitation at 377 nm (pulse duration = 25 ps). c Corresponding pseudo-first-order kinetics analysis of the quenching rate as a function of added 1a. The quenching rate was calculated according to the relationship rate = 1/τobs(1a) − 1/τobs(0), where τobs(1a) and τobs(0) are the observed photoluminescence lifetime of 50 μM Au(BZI)(TMCz) in the presence and absence, respectively, of 1a. d Photoluminescence (λex = 380 nm) spectra of Ar-saturated DMSO containing 50 μM Au(BZI)(TMCz) recorded with increasing concentration of DIPEA (0–500 mM).The photoinduced electron transfer produces a radical-ion pair (RIP) consisting of [Au(BZI)(TMCz)]•+ and 1a•−. We used nanosecond visible–near-infrared (Vis–NIR) transient absorption spectroscopy to directly monitor the genesis of the radical-ion species. As shown in Fig. 5a, the heat map of the photoinduced Vis–NIR absorption difference spectra of 100 μM Au(BZI)(TMCz) after nanosecond pulsed laser photoexcitation at a wavelength of 355 nm contains negative signals in the visible region because of the stimulated emission (see also the top- and bottom-most panels in Fig. 5c). In sharp contrast, positive signals emerge at a peak wavelength of 870 nm in the presence of 200 mM 1a (Fig. 5b and the second panel in Fig. 5c). The NIR signals are attributable to [Au(BZI)(TMCz)]•+ because [Au(BZI)(TMCz)]•+ electrochemically generated at an anodic potential of 0.45 V vs Ag+/0 exhibits a broad absorption band in this region (third panel in Fig. 5c). This spectral assignment is further corroborated by the close match with the [Au(BZI)(TMCz)]•+ electronic transition spectrum quantum chemically simulated at the CAM-B3LYP level of theory with a conductor-like polarizable continuum model parameterized to DMSO (fourth panel in Fig. 5c). The transient spectroscopy and calculation investigations provide direct evidence for photoinduced electron transfer from [Au(BZI)(TMCz)]* to 1a.Fig. 5: Electron transfer.a, b Heat maps showing nanosecond photoinduced transient Vis–NIR absorption difference signals of Ar-saturated DMSO containing 100 μM Au(BZI)(TMCz) recorded in the absence (a) and presence (b) of 200 mM 1a, recorded after 355 nm pulsed laser photoexcitation. The legend shows ΔAbsorbance. c Top-most panel, selected photoinduced transient Vis–NIR absorption difference spectra of Ar-saturated DMSO containing 100 μM Au(BZI)(TMCz); second panel, selected photoinduced transient Vis–NIR absorption difference spectra of Ar-saturated DMSO containing 100 μM Au(BZI)(TMCz) recorded in the presence of 200 mM 1a; third panel, Vis–NIR absorption difference spectra of 2.0 mM Au(BZI)(TMCz) recorded under an anodic potential of 0.45 V vs Ag+/0 (conditions: Pt mesh working electrode, Pt coil counter electrode, Ag/AgNO3 pseudo-reference electrode, and Ar-saturated DMSO containing 0.10 M Bu4NPF6 and the Au(I) complex); fourth panel, the absorption spectrum simulated for [Au(BZI)(TMCz)]•+ (CAM-B3LYP and LANL2DZ basis sets for Au and 6–311+g(d,p) basis set for the other atoms), where the vertical bars indicate oscillator strengths; bottom-most panel, photoluminescence spectrum of Ar-saturated DMSO containing 10 μM Au(BZI)(TMCz).Once formed, the RIP is rapidly annihilated by charge recombination through back electron transfer from 1a•− to [Au(BZI)(TMCz)]•+ because the driving force for charge recombination (−ΔGCR, −ΔGCR = e·[Ered(1a) − Eox(Au(BZI)(TMCz))]) is as large as 2.37 eV. Notably, in many photoredox catalysis processes, the charge recombination is ultrafast and detrimental to the catalysis cycle, limiting the overall photocatalytic performance36,39. The charge recombination process in [Au(BZI)(TMCz)]•+ can be monitored at a wavelength of 870 nm. Surprisingly, [Au(BZI)(TMCz)]•+ is long-lived, with an apparent lifetime of 18 μs (Fig. 6a). Second-order kinetics analyses based on the molar absorbance of [Au(BZI)(TMCz)]•+ at 870 nm (96 M−1 cm−1) indicate that the rate constant for charge recombination (kCR) with 1a•− is 3.3 ± 0.1 × 108 M−1 s−1 (Fig. 6b). Notably, the yield for the liberation of free-radical-ion species from the geminate RIP (i.e., [[Au(BZI)(TMCz)]•+···1a•−] → [Au(BZI)(TMCz)]•+ and 1a•−) can be computed according to the relationship k − diff / (k − diff + kCR) to be as large as 91%. This value is greater than the yield for charge recombination of cyclometalated Pt(II) complexes with CF3I•− (47–77%), which are highly reducing photocatalysts established by us36. The kCR values with other aryl halide substrates were also determined to be in the range (0.7 ± 0.04 to 2.9 ± 0.7) × 109 M−1 s−1 (Supplementary Fig. 7). Finally, we found that the kCR values adhere to the Jortner curves for electron transfer with large reorganization energies, which suggested strong interactions with solvents (Supplementary Fig. 8).Fig. 6: Electron transfer kinetics.a Temporal changes of the 870 nm [Au(BZI)(TMCz)]•+ traces. b Second-order kinetics analysis for charge recombination between [Au(BZI)(TMCz)]•+ and 1a•−. See Supplementary Fig. 7 for the results for the other substrates. c Decay traces recorded at 870 nm in the presence of 50 mM 1a and increased concentrations of 2a (0–300 mM). d Pseudo-first-order kinetics analysis for the catalyst recovery through electron transfer to [Au(BZI)(TMCz)]•+. Error bars indicate the standard deviations of exponential fits of the data points in (c). See Supplementary Fig. 9 for the results for the other substrates.The liberated 1a•− is cleaved into an aryl radical species and Cl−. The aryl radical species subsequently reacts with 2a to form a C–C cross-coupled adduct (3aaH• in Fig. 7). We hypothesized that the 3aaH• species plays a key role in completing the photoredox catalytic cycle. Specifically, we hypothesized that the [Au(BZI)(TMCz)]•+ is neutralized to the original Au(BZI)(TMCz) state through one-electron withdrawal from 3aaH• and that DIPEA subsequently deprotonates the resultant one-electron-oxidized 3aaH+ intermediate, producing the final product (3aa).Fig. 7: Catalysis cycle.Plausible mechanism of the photoredox catalytic C–C cross-coupling reaction. Refer to the main text for the definitions of the symbols.To validate this hypothesis, we performed nanosecond laser flash photolysis for the mixture of 100 μM Au(BZI)(TMCz) and 50 mM 1a with the addition of 2a in various concentrations (0–300 mM). We expected that the effective concentration of 3aaH• would increase with the increased 2a concentration. Gratifyingly, we found that the increase in concentration of 2a shortens the lifetime of the 870 nm signal of [Au(BZI)(TMCz)]•+ (Fig. 6c). The concentration-dependent decays can be best interpreted as the recovery of Au(BZI)(TMCz) through electron transfer from 3aaH• to [Au(BZI)(TMCz)]•+. Our pseudo-first-order kinetics analysis of the 870 nm signals as a function of [2a] yields a second-order rate constant of 3.8 ± 0.4 × 105 M−1 s−1 for bimolecular electron transfer (keT) from 3aaH• to [Au(BZI)(TMCz)]•+ (Fig. 6d). Results obtained with the other substrates are summarized in Supplementary Fig. 9. Although the determined value should serve as a lower limit due to the pre-steps for the generation of 3aaH•, this keT value is three orders of magnitude smaller than the rate constants for the other electron-transfer steps, kQ (9.1 ± 0.1 × 108 M−1 s−1) and kCR (3.3 ± 0.1 × 108 M−1 s−1). This comparison indicates that the catalyst recovery is the rate-determining step in the overall photoredox catalysis cycle. The relatively slow electron transfer is presumably attributable to the multiple steps, including C–Cl bond cleavage and the radical addition to 2a, required to form 3aaH•.
