Introduction

Photoinduced transition-metal catalysis has established a powerful paradigm for achieving challenging organic transformations1,2,3. Recently, photoexcited gold catalysts have attracted significant attention owing to their unique catalytic performance by integration of exceptional carbophilicity, tunable redox behavior, and rich photophysical properties4,5,6,7,8,9. In particular, dinuclear gold complexes have fundamentally expanded catalytic possibilities beyond mononuclear systems. Recent studies have revealed that the synergistic Au-Au interactions can dramatically reduce activation barriers through cooperative bond cleavage and formation processes10,11,12,13,14. However, the limited synthetic accessibility of structurally diverse dinuclear gold complexes has severely hindered the advances in their photocatalytic applications, as most reported systems are based on a single class of symmetric ligands (Fig. 1a)15,16,17,18,19,20,21.

Fig. 1: Design and catalytic applications of dinuclear tricoordinate gold complexes.
Fig. 1: Design and catalytic applications of dinuclear tricoordinate gold complexes.
Full size image

a Photoinduced dinuclear gold catalysis. b The combination of bidentate P^P and N^N ligands. c Ligand to ligand charge transfer (LLCT) promoted C-Br bond activation for R-Br. d A kind of phtoexcited dinuclear gold-catalyzed three component carbonylative amidation reaction.

As our continual interests in bimetallic chemistry10,11,12,13,14,22,23,24,25,26,27,28,29,30, we sought to employ two different kinds of bidentate ligands (P^P and N^N frameworks) to precisely modulate the coordination microenvironments and structural diversity of dinuclear gold complexes, thereby broadening their synthetic landscape (Fig. 1b). By pairing σ-donating bisphosphine and π-conjugated bidentate nitrogen ligands, the resulting complexes enable ligand-to-ligand charge transfer31,32 (LLCT) to channel photoexcitation energy toward substrate activation. In the case of organic halides, the Au-Au interaction serves as both a structural anchor and electronic conduit. This synergistic interaction, combined with LLCT, stabilizes a polarized excited state and lowers the C-X bond activation barrier through back-donation (Fig. 1c), ultimately producing open-shell radicals via inner-sphere single-electron transfer (ISET)33,34,35. Consequently, these alkyl radicals can further capture CO to undergo mild carbonylative amidation36,37,38,39,40 with amines, enabled by dinuclear gold photoredox catalysis (Fig. 1d). Notably, homogeneous gold-catalyzed carbonylation reactions are generally challenging due to the poor compatibility of CO within redox catalytic systems for the formation of Au-carbonyl complex41,42,43,44,45. A limitation that has recently been addressed by Patil and co-workers applies only in the CO insertion of aryl/vinyl iodides assisted by chelated ligands46.

Here, we report a method for the synthesis of dinuclear gold complexes through a mixed-ligand strategy incorporating both P^P- and N^N-bidentate ligands. A comprehensive investigation of their photophysical properties reveals the critical role of LLCT and Au-Au interactions in modulating catalytic performance. The LLCT process in the photoexcited state acts as a key driving force for ISET, effectively cleaving unactivated C-Br bonds through back-donation. These complexes exhibit excellent catalytic efficiency in carbonylative amidation of alkyl halides, with broad functional group tolerance and a wide substrate scope. Mechanistic studies uncover a dual ISET/outer-sphere electron transfer (OSET) pathway involving both radical initiation and 4-dimethylaminopyridine (DMAP)-assisted acyl pyridinium formation.

Results and discussion

Design of mixed ligand dinuclear gold complexes

To synthesize dinuclear gold complexes based on a mixed-ligand strategy, we employed complementary P^P- and N^N-type bidentate ligands. Nitrogen-based bidentate ligands, such as bipyridine (bpy) and 1,10-phenanthroline (phen), with their distinctive π-accepting character and rigid chelating geometry47,48,49,50, are well-suited to modulate the electronic excitation properties of dinuclear gold complexes, and allow fine-tuning of catalytic performance through functional group modification51,52. Coordination of P^P-(AuCl)2 complexes with bipyridines bearing either electron-withdrawing or electron-donating groups afforded mixed-ligand dinuclear gold-complexes (4–12) in good yields (64–92%) (Fig. 2). Replacing bis(diphenylphosphino)methane (dppm) with P^N^P-type phosphines or 1,3-bis(diphenylphosphino)propane (dppp) gave complexes 13 and 14 in slightly reduced yields (62% and 68%, respectively). A series of additional gold complexes (15–19) incorporating various counterions were also prepared in consistently good yields. Single-crystal X-ray diffraction analyses of representative complexes (4, 7, 9 and 12) revealed Au-Au distances ranging from 3.0 to 3.2 Å, suggesting the presence of significant aurophilic interactions. It is noted that the distance between gold and nitrogen is unequal, and this unsymmetric coordination of the N,N-chelates to gold reflects a clear manifestation of the antichelate effect53,54.

Fig. 2: Mixed ligands in dinuclear gold complexes.
Fig. 2: Mixed ligands in dinuclear gold complexes.
Full size image

Reaction conditions: 1 (1.0 equiv.), 2 (2.0 equiv.), 3 (2.0 equiv.), ambient temperature, 12 h. The trifluoromethanesulfonate anion is omitted in the single-crystal structure diagrams.

Photophysical properties study

Subsequent investigation of the photophysical properties of gold complex 4 revealed significant solvent-dependent modulation. UV-vis absorption spectroscopy demonstrated a pronounced bathochromic shift in both absorption and emission spectra when the solvent was changed from dichloromethane (DCM) to dimethyl sulfoxide (DMSO) (Fig. 3a, b). This strong solvatochromic effect indicates the influence of solvent environment on the excited-state properties of the gold complexes. Time-dependent density functional theory (TD-DFT) simulations successfully reproduced the experimental absorption spectrum of complex 4 in DMSO. Notably, a prolonged fluorescence lifetime of 1.8 μs was observed in DMSO, providing valuable insight into the long-lived excited state of the complex and underscoring its potential applications in light-driven applications.

Fig. 3: Photophysical characterization of gold complexes.
Fig. 3: Photophysical characterization of gold complexes.
Full size image

a UV-Vis absorption spectra of 4 in different solvents. b Photoluminescence and fluorescence lifetime of 4. c Hole (blue)-electron (green) analysis of gold complexes 4, 5, 6, 7 at the level of TDDFT: PBE0(D3)/DKH2/DKH-def2-TZVP (nonmetals)/SARC-DKH-TZVP (metals). Isosurfaces for hole-electron of selected structures are shown (isovalue = 0.002).

To elucidate the charge transfer mechanisms of these dinuclear gold complexes, we performed electron-hole analyses55,56,57 on representative examples, as illustrated in Fig. 3c. For clarity and comparability, each complex was partitioned into four structural fragments: the dinuclear gold core (#1), the phosphine ligand (#2), the counterion (#3), and the nitrogen ligand (#4) (see Supplementary Information for details). Our objective was to quantify the contribution of each fragment to the excitation process and thereby elucidate the underlying charge transfer modes. The computational results demonstrated a high degree of complexity and diversity in the electron distribution across the excited states. In the singlet excited state (S1), all examined complexes consistently exhibited a predominant charge transfer character from the metal center (M) and phosphine ligand (P) to the nitrogen ligand (N), namely MLP(LN)CT. This consistency suggests specific electronic arrangements and interaction patterns between the phosphine and nitrogen ligands in the complexes, shaped by their alignment with the relative frontier orbital energies of the metal center.

However, upon transition to the triplet excited state (T1), the diversity in electron distribution becomes more diversified. Taking the T1 state of complex 4 as an example, it primarily exhibits LLCT characteristics. This indicates that in the triplet state, excitation-induced electron redistribution is no longer restricted to a single mode. Such changes are likely attributed to alterations in the electron spin configuration and orbital reordering in the triplet state, which significantly modulate the electron distribution and excited-state electronic structure and in turn have a significant impact on the electron transfer modes. Collectively, these findings highlight that the photophysical properties of these mixed-ligand dinuclear gold complexes are governed not only by their ground-state electronic structures, but also critically depend on the excited-state charge transfer nature.

Photochemical reaction discovery

Building upon the fundamental understanding of the photophysical properties of the synthesized dinuclear gold complexes, we next examined their photocatalytic performance. Specifically, we explored their application in direct dehalogenative three-component carbonylative amidation reactions under photochemical conditions. Notably, when employing [Au2dppm(dOMebpy)2](OTf)2 (4) as the photocatalyst irradiated by 370 nm LEDs, the reaction between bromocyclohexane (20) and 1-(1-naphthyl)ethylamine (21) in DMSO under a CO atmosphere (balloon), with DMAP as an additive, afforded the target product (22) in 86% yield (Table 1, entry 1). Other dinuclear gold catalysts exhibited decreased reaction efficiency. The mononuclear gold complex58 (25) was tested but proved catalytically inactive, underscoring the essential role of the dinuclear architecture in this transformation. Additionally, other commercially available photocatalysts, such as the strongly reducing fac-Ir(ppy)3 (26), the strongly oxidizing Ir[dF(CF3)ppy]2(dtbpy)]PF6 (27), and the organic photocatalyst 4-CzIPN (28), failed to promote the reaction. Control experiments further confirmed that the photocatalyst, additive, and light irradiation are all crucial for this three-component carbonylative amidation reaction (Table 1, entries 2-5).

Table 1 Optimization of the reaction conditionsa
Full size table

Reaction mechanism study

A series of control experiments were carried out to gain insight into the mechanism of this gold-catalyzed three-component carbonylative amidation reaction. Screening different additives under standard conditions revealed that only DMAP was an efficient additive to promote this transformation (Fig. 4a). We speculated that an acyl pyridinium cation intermediate might be formed during the reaction59. To validate this hypothesis, the corresponding acyl pyridinium salt (29) was synthesized and reacted with primary amine (21) under standard conditions, which cleanly afforded the corresponding target product (22). This result supports the possible generation of acyl pyridinium intermediate in the catalytic cycle. Additional evidence for a radical pathway was obtained from radical inhibition experiments. The addition of TEMPO completely suppressed formation of product (22), and the corresponding cyclohexyl-TEMPO adduct was detected by HRMS, confirming a radical process (see Supplementary Information for details). Furthermore, addition of the spin-trap reagent N-tert-butyl-α-phenylnitrone (PBN) to the model reaction led to a strong radical signal in electron paramagnetic resonance (EPR) spectroscopy (Fig. 4b). Comparison with simulated spectra confirmed the signal originated from the PBN-trapped alkyl radical, with specific fitting parameters of g = 2.0062, aN = 14.6 G, and aH = 3.2 G. Complementary evidence was provided by a radical clock experiment using 6-bromo-1-hexene (30) as the substrate, which exclusively afforded the cyclized product (32) under the standard conditions (Fig. 4c), further corroborating the presence of free radical intermediates.

Fig. 4: Mechanism studies.
Fig. 4: Mechanism studies.
Full size image

a Effect of additives. b EPR experimental result. c Radical clock experimental result. d Proposed mechanism.

Based on the experimental results, two plausible pathways were proposed for the three-component carbonylative amidation reaction (Fig. 4d). In path A, gold catalyst (4) undergoes photoexcitation to form its excited-state species 4*. Subsequent dynamic coordination of the alkyl bromide to photoexcited 4* via the lone pair on the halogen atom leads to the formation of transition state TS1. This interaction facilitates an ISET process, resulting in the C-Br bond cleavage and formation of the cyclohexyl radical (33), along with the dinuclear gold species ([Au2Br2+][OTf]2) (34). The resulting radical (33) then rapidly adds to CO, forming the acyl radical intermediate (35). In the presence of DMAP, the acyl radical  adds to DMAP, yielding the zwitterionic radical intermediate (36). Subsequent SET oxidation of (36) by gold complex (34) produces the electrophilic alkylacyl-DMAP salt (37). This electrophilic intermediate then undergoes nucleophilic substitution with aniline (21) to afford amide product (22), concomitant with the regeneration of the ground-state gold catalyst. In path B, the acyl radical intermediate (35) is instead oxidized by the gold species (34) to form a highly reactive acylium cation (38). This intermediate subsequently is captured by the primary amine nucleophile (21) under DMAP assistance to afford the target product (22), regenerating the active gold catalyst and completing the catalytic cycle.

To gain a better insight into the reaction mechanism, density functional theory (DFT) calculations were conducted for the model reaction with 20 and 21 as substrates (Fig. 5, see details in Supplementary Information and Supplementary Data 1). Initially, the excited state of dinuclear gold complex [Au2dppm(dOMebpy)2(OTf)2] (4*) undergoes an SET process with 20, which results in the cleavage of C-Br bond and simultaneous generation of a cyclohexyl radical. Comparative analysis of different mechanistic pathways revealed that the ISET mechanism proceeds with a significantly lower activation barrier (ΔG‡ = 19.0 kcal mol⁻¹) than the outer-sphere SET (OSET) pathway, indicating a clear energetic preference for ISET60,61,62. The generated cyclohexyl radical subsequently undergoes carbonylation with CO and affords an acyl radical intermediate (35), which then reacts with DMAP to form a zwitterionic radical intermediate (36). The transition state energy barrier for this step is computed to be 14.3 kcal mol-1. In contrast, direct generation of an acyl carbonium ion was found to require a free energy gain of 59.5 kcal mol-1, rendering this pathway energetically prohibitive under the reaction conditions. Following the formation of intermediate (36), a subsequent SET oxidation by gold species (34) generates the acyl pyridinium cation (37) with a low activation barrier (ΔG‡ = 2.6 kcal mol⁻¹). During this step, the bromine-containing dinuclear gold species (34) is reduced back to the original dinuclear gold catalyst. Ultimately, 37 undergoes nucleophilic substitution with 21, affording the target product (22) and completing the catalytic cycle. To further investigate the distribution of unpaired electrons in the excited-state dinuclear gold complex during the reaction, spin density analyses were performed on the key intermediates (4*, INT1, TS1, INT2 and 34). These results reveal a gradual electron transfer from the bidentate nitrogen ligands fragment to the metal center and alkyl halide substrate (20), which further confirms that the activation of alkyl halides would be promoted by the LLCT process of the dinuclear gold complex.

Fig. 5: Computational studies about reaction mechanism.
Fig. 5: Computational studies about reaction mechanism.
Full size image

Free energy profiles calculated at the PBE(D3)/def2-SVP/def2-TZVP/CPCM(DMSO)//ωB97M-V/def2-TZVP/SMD(DMSO) level of theory. Isosurfaces for spin densities of selected structures are shown below (isovalue = 0.005).

Reaction substrate scope

Under the optimized reaction conditions, bromocyclohexane was selected as the starting material to explore the scope of amines in the three-component carbonylative amidation reaction (Fig. 6). The results indicated broad amine compatibility, with both aliphatic (primary and secondary) and aromatic amines proving to be competent coupling partners, affording the desired products in yields up to 96%. Notably, the transformation exhibited excellent functional group tolerance, as evidenced by successful incorporation of diverse functional groups such as ethers (41–42), cyano group (43), difluoromethyl (54), fluorine (63), and trifluoromethyl group (67). Furthermore, amines containing an internal olefin or terminal alkyne (44–46), as well as those bearing heterocyclic moieties (50–52) or steric hindered group (56), were well tolerated. This approach also exhibited good compatibility with sulfonamide substrate (63) and 2,2,2-trifluoroethylamine molecule (64). The late-stage functionalization of pharmaceutical molecules, including paroxetine (65), vortioxetine (66), desloratadine (67), dehydroabiethylamine (68), and fluoxetine (69), was successfully performed, affording the desired products with good yields ranging from 72% to 88%. We further explored the scope of arylamine substrates and found that a wide range of arylamines, bearing either electron-withdrawing or electron-donating groups, were well tolerated under the optimized reaction conditions. The corresponding amide products (70–80) were obtained in 41% to 82% yields.

Fig. 6: The substrate scope of amines.
Fig. 6: The substrate scope of amines.
Full size image

Reaction conditions: 4 (3 mol%), bromoalkanes (0.6 mmol, 3.0 equiv.), amines (0.2 mmol, 1.0 equiv), DMAP (0.6 mmol, 3.0 equiv.), CO balloon, DMSO (1.0 mL), 370 nm LEDs, 12 h.

As shown in Fig. 7, the protocol also exhibits broad applicability across primary, secondary, and tertiary bromoalkanes, delivering products (81–98) in moderate to good yields. It is worth mentioning that various active functional groups, such as silyl (83), fluoro (84), trifluoromethyl (85), chloro (87), ester (88), and boron compounds (89), were well tolerated. Additionally, cyclic secondary bromides bearing four to seven-membered rings proved to be competent substrates, giving rise to the desired products (91-95) in synthetically useful yields. For alkyl bromides containing heterocyclic structures (e.g., tetrahydropyran and piperidine), the reaction proceeded smoothly as well, providing products (92–94) with 58%-92% yields. Remarkably, even highly sterically hindered substrates afforded products (96 and 97) in 56% and 55% yields, respectively. Our investigation was further extended to hydroxyl molecules in this transformation. Notably, with 2-hydroxybenzyl alcohol (99) as substrate, the reaction tended to occur in the aliphatic hydroxyl group rather than the phenolic hydroxyl group. The phenolic hydroxyl group, despite its relatively lower nucleophilicity, participated smoothly in the transformation, producing the corresponding ester product (102) in 47% yield. Furthermore, simple 3-bromopropionic acid (103) was also tested, enabling the successful synthesis of sacubitril (105), an important therapeutic for chronic-heart-failure treatment63.

Fig. 7: The substrate scope of alkyl bromides and other nucleophiles.
Fig. 7: The substrate scope of alkyl bromides and other nucleophiles.
Full size image

Reaction conditions: 4 (3 mol%), bromoalkanes (0.6 mmol, 3.0 equiv.), nucleophiles (0.2 mmol, 1.0 equiv.), DMAP (0.6 mmol, 3.0 equiv.), CO balloon, DMSO (1.0 mL), 370 nm LEDs, 12 h.

Discussion

In summary, we have developed a bidentate mixed-ligand strategy for the construction of a class of dinuclear gold complexes, in which P^P- and N^N-type ligands coordinate to the gold centers. Interestingly, photophysical studies revealed that LLCT and Au–Au interactions play key roles in governing the excited-state behavior and catalytic performance of these complexes. Moreover, under mild conditions, these complexes have demonstrated excellent photoredox catalytic performance in selective three-component carbonylative amidation of unactivated alkyl halides with amines and CO (balloon). A wide range of structurally diverse amides are produced in synthetically useful yields (64 examples, up to 96% yield). Mechanistic investigations uncovered a hybrid ISET/OSET pathway, in which the dinuclear gold catalyst facilitates both substrate activation and DMAP-assisted acyl pyridinium formation. These findings not only advance our fundamental understanding of gold photoredox catalysis, but also provide an avenue for the rational design of next-generation photocatalysts.

Methods

General procedure for synthesis of dinuclear gold-complex

To an oven-dried 10 mL sealed tube equipped with a magnetic stirring bar, 1 (1.0 equiv.) is added, followed by dry DCM. The tube is backfilled with argon. To this solution, the bipyridine ligand 2 (2.0 equiv.) and silver salt 3 (2.0 equiv.) are added, and the reaction mixture is stirred for 10 h. Afterward, celite is added, and the precipitate is filtered off. The solvent is concentrated, and the complex is precipitated with diethyl ether. Filtration and washing with diethyl ether afford dinuclear gold complexes 4–19 as a solid powder.

General procedure for three-component carbonylative amidation

To an oven-dried 10 mL sealed tube equipped with a magnetic stirring bar, 4 (3 mol%), bromoalkanes (0.6 mmol, 3.0 equiv.), amines (0.2 mmol, 1.0 equiv.), DMAP (0.6 mmol, 3.0 equiv., 73.2 mg) and DMSO (1.0 mL) are added. The reaction mixture is degassed by three times backed with CO cycles and thenis vigorously stirred under the irradiation of 370 nm LEDs (distance ~4.0 cm from the bulb) for 12 h. After completion, the reaction mixture is removed from the light, diluted with water (10 mL), and extracted with ethyl acetate (5 mL) three times. The combined organic layers are washed with brine three times and dried over Na2SO4 and concentrated under reduced pressure. The residue is purified by silica gel chromatography (petroleum ether/ethyl acetate) to afford the corresponding desired product.