Exceptionally fast carbon-carbon bond reductive elimination from gold(III).

Reductive elimination of carbon-carbon bonds occurs in numerous metal-catalysed reactions. This process is well documented for a variety of transition metal complexes. However, carbon-carbon bond reductive elimination from a limited number of Au(III) complexes has been shown to be a slow and prohibitive process that generally requires elevated temperatures. Herein we show that oxidation of a series of mono- and bimetallic Au(I) aryl complexes at low temperature generates observable Au(III) and Au(II) intermediates. We also show that aryl-aryl bond reductive elimination from these oxidized species is not only among the fastest observed for any transition metal, but is also mechanistically distinct from previously studied alkyl-alkyl and aryl-alkyl reductive eliminations from Au(III).

Reductive elimination of C-H, C-C, and C-X bonds is a key step in many metal-catalyzed reactions.[1] These processes have been extensively studied from Ni[2-6], Pd[7-15], and Pt[16-22], yet relatively little is known about the reductive elimination from Au, which is markedly more stable toward air and water. In fact, since Kochi[23-25] and Tobias’s[26] seminal investigations of high-temperature dialkylgold(III) reductive eliminations over 35 years ago, fundamental studies of C-C bond reductive coupling from oxidized Au centers remain exceedingly rare. Previous studies have shown that cis-EtMe2Au(PPh3)[24] and [cis-(CH3)2Au(PPh3)2](PF6)[26] are stable at room temperature, undergoing alkyl-alkyl bond reductive elimination at 70 °C (kobs ~ 10−5 − 10−3 s−1), while cis-(CH3)2AuCl(PPh3)[25] reductively eliminates ethane slowly at 40 °C (kobs ~ 10−7 s−1). These processes are inverse first-order in PPh3 and are severely inhibited by even small amounts of free ligand. Kinetic analysis suggests that reductive elimination does not occur directly from the 4-coordinate complex, but rather from a high-energy T-shaped intermediate, formed by slow, reversible phosphine dissociation[23]. Since these reports, the field has not only witnessed the advent and maturation of now-ubiquitous palladium-catalyzed cross-coupling methods, but the development of homogeneous redox-neutral gold catalysis as well. The shortage of mechanistic insight of processes at Au(III) may be due in part to synthetic challenges in accessing appropriate Au(III) models, and to the high oxidation potential of Au(I) that traditionally precludes Au(I)/Au(III) redox cycling[27].

Electronic cooperation between metal centers is well established in redox processes. Hence, bimetallic Au complexes may be useful systems for studying the fundamental chemistry of oxidized organogold species. In the bimetallic core, each Au(I) can formally undergo a one electron oxidation, resulting in two d[9] metal centers with unpaired electrons capable of making a Au(II)-Au(II) σ-bond. Indeed, Fackler[28-34] and Laguna[35-38] have shown that dihalides and alkyl halides oxidize A-frame bimetallic Au(I) complexes to access Au(II) species, yet reductive elimination processes from Au(II) have not been rigorously investigated. In fact, well-behaved reductive eliminations of carbon-carbon bonds from bimetallic Au(II) species are incredibly rare[39]. Due to the preference for linearity by Au(I), A-frame bimetallic Au(I) complexes lack available coordination sites and may only coordinate nucleophiles upon oxidation. Therefore, while non-A-frame systems may be more appropriate for mechanistic studies that ultimately inform catalysis, only one non-A-frame bis(aryl) bimetallic Au(II) complex has been prepared[38]; due to the strongly electron-withdrawing perfluorinated aryl ligands, biaryl reductive elimination does not occur at room temperature, and reductive elimination at elevated temperatures was not investigated.

Current studies of the fundamental chemistry at oxidized gold centers are essential to establishing new modes of gold reactivity – particularly those involving redox cycling[40]. Herein we report our preparation of several monometallic and non-A-frame bimetallic bis(aryl) gold(I) complexes that allow kinetic analysis of C-C bond-forming reductive eliminations from oxidized species. Contrary to Kochi’s pioneering report, we show that C-C bond reductive elimination at Au(III) is not necessarily a “disfavored”[41,42] process, but is among the fastest C-C bond-forming reductive eliminations reported for any transition metal complex between −50 and −10 °C. In one case, we calculate the fastest C-C bond reductive elimination recorded to date. We show that biaryl reductive elimination at monometallic Au(III) proceeds via an unexpected mechanism, and report the mechanism of biaryl reductive elimination from non-A-frame bimetallic gold complexes.

Results and discussion

Biaryl reductive elimination from a mononuclear gold complex

Ph3PAu(4-F-C6H4)2Cl (3) was synthesized upon partial oxidation of Ph3PAu(4-F-C6H4) (1) with PhICl2 at −78 °C, followed by fast transmetalation from remaining 1 to Ph3PAu(4-F-C6H4)Cl2 (2) (Figure 1). The reaction is unaffected by equimolar or excess amounts of oxidant, indicating that transmetalation from 1 to 2 is faster than oxidation of 1. While 3 cannot be isolated, the cis-aryl relationship is supported by 19F (two 1:1 singlets at −118.4, −119.5 ppm) and 31P NMR (one singlet at 27.9 ppm).

Figure 1

Oxidation of monometallic Au(I) and biaryl reductive elimination from Au(III). A CD2Cl2 solution of Ph3PAu(4-F-C6H4) (1) undergoes fast partial oxidation by PhICl2 upon thawing at low temperature to afford 2, which reacts rapidly with remaining 1 to generate a cis-diaryl Au(III) species (3). Reductive elimination from 3 affords 4,4’-difluorobiphenyl. Associative ligand exchange of 3 with excess PPh3 generates cationic 4, which undergoes very fast biaryl reductive elimination.

In light of Kochi’s findings, we expected biaryl reductive elimination from 3 to require heating via a mechanism involving phosphine dissociation. However, to our surprise, 3 was immediately consumed at room temperature (20 °C), quantitatively generating 4,4’-difluorobiphenyl and (Ph3P)AuCl. At −52 °C, aryl-aryl reductive elimination from 3 could be monitored by 19F NMR (kobs = (1.5 ± 0.1)×10−4 s−1) (Figure 2); at −23 °C, reductive elimination is 100 times faster (kobs = 0.015 ± 0.001 s−1). The observed rate constant remains unchanged over a range of concentrations, as well as in the presence of 10 equivalents of either Bu4NCl or Ph3PAuCl, indicating a unimolecular process that is first-order in 3.

Figure 2

Kinetics of decay of cis-(Ph3P)Au(4-F-C6H4)2Cl (3). a, First-order decay of 3 monitored by 19F NMR (470 MHz, CD2Cl2) at −32 °C over a period of 35 minutes. b, A natural log plot of [3]/[3]t=0 indicative of a first order process. Error bars represent the inherent error in quantifying 19F NMR signal intensity.

That reductive elimination is substantially more facile from 3 than from alkylgold(III) complexes may not simply be attributed to faster coupling of aryl-aryl relative to alkyl-alkyl bonds[9], but also suggests fundamentally different rate-determining steps and mechanisms. Contrary to what is established for alkyl reductive elimination of Au(III), excess PPh3 increases the observed rate of reductive elimination (Figure 3). In the presence of 10 equivalents of PPh3 at −52 °C, the rate of reductive elimination from 3 increases by an order of magnitude (kobs = (1.2 ± 0.2)×10−3 s−1); with 40 equivalents, the observed pseudo-first order rate constant is (6.2 ± 0.5)×10−3 s−1. No new 19F and 31P NMR signals can be observed under these conditions. In contrast, neither Bu4NCl nor radical traps such as 9,10-dihydroanthracene affect reductive elimination of 3; furthermore, the rate remains unchanged in the presence of excess oxidant (PhICl2 acts as a phosphine scavenger), precluding the scenario in which small amounts of PPh3 liberated from 3 induce reductive elimination. These observations are consistent with two operative pathways to biaryl: one independent of added PPh3 and another involving phosphine coordination. In the absence of added phosphine, reductive elimination occurs directly from square-planar complex 3. In the presence of excess PPh3, 3 likely undergoes associative ligand exchange to generate cation 4, which rapidly reductively eliminates biaryl. This process is first-order in PPh3 (Figure 3, right.) Since 4 is unobservable by NMR, reductive elimination from this complex must be substantially faster than from 3, given its bulkier steric environment and cationic charge. When 10 equivalents of PPh3 and 40 equivalents of Bu4NCl are added, the rate remains unchanged from that of the phosphine-accelerated reaction, suggesting that reductive elimination from 4 is substantially faster than the reverse ligand exchange. At −52 °C, association of PPh3 (k = 0.019 ± 0.001 s−1 M−1) and reductive elimination from 4 (k ≥ 0.22 s−1, see Supporting Information) are remarkably facile processes. We cannot definitively discount the possibility of reductive elimination from a 5-coordinate intermediate following coordination of PPh3, but this scenario is unlikely for two reasons. Presumably, alkyl Au(III) complexes should behave similarly in the presence of excess PPh3, but their reductive eliminations are instead drastically slowed. Secondly, while Yamamoto has demonstrated that a cis-arylmethylnickel(II) complex with a chelating bis(phosphine) undergoes accelerated reductive elimination in the presence of PPh3 via a 5-coordinate species[2], comparisons with complex 3 are imperfect due to the potentially labile Au-Cl bond.

Figure 3

Unexpected phosphine acceleration on the decay of 3 at −52 °C. a,,The rate of decay of 3 (slope of the natural log plot) increases with [PPh3] ( (0 equiv PPh3); (10 equiv PPh3); (20 equiv PPh3); (30 equiv PPh3); (40 equiv PPh3)). b, A linear relationship between [PPh3] and kobs for the decay of 3 in the presence of excess PPh3 (pseudo-first order conditions), indicating that ligand exchange from 3 to 4 is first-order in PPh3. Error bars represent the inherent error in interpreting 19F NMR signal intensity (a) or the standard deviation over three separate experiments (b).

To our knowledge, reductive eliminations from diarylgold(III) complexes 3 and 4 are among the fastest observed C-C bond-forming reductive couplings at any transition metal center at temperatures below −20 °C. For comparison, rate constants for reductive elimination of diarylplatinum(II)[19] and palladium(II)[12] complexes have been reported between 10−5 − 10−3 s−1 at 95 and 50 °C, respectively, although Pd-catalyzed Kumada-Corriu couplings can be achieved at temperatures as low as −65 °C[43]. In some cases, diarylplatinum(II) systems only reductively eliminate biaryl at temperatures above 100 °C[28]. Reductive elimination rate constants of ~0.1 s−1 have also been reported for Ni-catalyzed oxidative homocouplings at −35 °C[6], although the presumed diarylnickel(II) intermediate is never observed. Indeed, for a 3rd row metal, the rates of aryl-aryl reductive elimination from Au(III) are particularly impressive[44], outcompeting even C-C bond coupling in alkynylarylpalladium(IV) complexes (k ~ 10−3 s−1 at −35 °C). An Eyring analysis over a 29 °C range reveals that the reductive elimination from 3 not only has an unusually small enthalpic barrier, but a small entropic contribution to the activation energy as well (ΔH‡ = 17.2 ± 0.2 kcal/mol; ΔS‡ = 2.0 ± 0.8 e.u.); these kinetic parameters suggest a transition state resembling 3 in which the Au-C(sp2) bonds are mostly conserved.

It is not immediately obvious why aryl and chloride ligands drastically change the mechanism of C-C bond reductive elimination at Au(III). Generally, the barrier to aryl-aryl reductive elimination is lower than that for alkyl-alkyl reductive elimination[8], while the weaker σ-donating ability of aryl and chloride ligands relative to alkyls may increase the barrier to phosphine dissociation. The notion that “Au(III) and Pt(II) catalysts show similar reactivity”[42] is not necessarily true when redox processes are involved: at greater than 90 °C, Pt(II) biaryl complexes undergo reductive elimination at rates comparable to those observed for Au(III) at approximately −50 °C. The reason for this discrepancy in rates may lie in pronounced relativistic effects, which result in Au(I) having a strong preference for linearity over other transition metals[42]. Since the product of reductive elimination from a square planar complex is necessarily 2-coordinate, and the transition state is partway to linearity, the activation energy of reductive elimination of Au(III) to Au(I) may be less than that for other transition metals. This is reasonable considering the early transition state to reductive elimination of 3.

Biaryl reductive elimination from bimetallic gold complexes

The unexpected redox behavior of mononuclear Au(III) species led us to investigate analogous reactivity of bimetallic gold complexes. Others have reported that bimetallic complexes with three-atom linkers do not undergo two-electron oxidation at a single Au(I), but instead undergo formal one-electron oxidation at two Au(I) centers to generate a species stabilized by a Au(II)-Au(II) bond[28-38]. Electronic cooperation between metal centers in bimetallic complexes offers a lower barrier oxidation pathway relative to a two-electron oxidation at a single metal center[45], allowing, for example, oxidation of dicationic A-frame Au(I) complexes by oxidants typically unreactive toward monocationic Au(I), such as benzoyl peroxide[34]. Furthermore, electrochemical studies and DFT calculations suggest that aurophilic interactions reduce the oxidation potential of bimetallic Au(I) species relative to mononuclear complexes[27].

To assess the role of metal-metal bonding in the C-C bond reductive elimination of oxidized gold, we prepared a bimetallic Au(I) complex stabilized by a bis(diphenylphosphinoamine) ligand Ph2P-NR-PPh2 (PNP)[46,47]. X-ray diffraction analysis of PNP[Au(4-F-C6H4)]2 (5, R = CH3) indicates an aurophilic interaction (Au(I)-Au(I) distance = 3.0357(2) Å) that is likely conserved in solution (See Supporting Information). Laguna has shown that direct oxidation of PNP[Au(C6F5)]2 (R = H) with Cl2 generates symmetric bimetallic Au(II) complex PNP[Au(C6F5)Cl]2, which is stable at room temperature; crystallographic analysis establishes a Au(II)-Au(II) bond (2.576(2) Å) and a trans relationship between chloride ligands[38].

The analogous low-temperature oxidation of 5 with PhICl2 generates symmetric bimetallic Au(II) complex 6 (19F NMR and 31P NMR singlets at −120.4 ppm and 83.5 ppm, respectively), which undergoes slow reductive elimination at temperatures below −30 °C (Figure 4). At −23 °C, 6 undergoes first-order decay (k = (1.6 ± 0.3)×10−4 s−1) with concomitant formation of 4,4’-difluorobiphenyl. Without observing a mixed-valent intermediate, we cannot kinetically distinguish between a pathway involving rearrangement and reductive elimination, and one involving bimetallic reductive elimination via a 4-centered transition state[48]. However, Laguna has reported that similar PNP-supported binuclear perfluoroarylgold(II) complexes rearrange over several hours to mixed-valent Au(I)/Au(III) species[38]. Given fast aryl transmetalation and reductive elimination of mononuclear Au(III), as well as the scarcity of reported binuclear reductive elimination, we therefore favor a mechanism involving isomerization of 6 to mixed-valent species 7, which then undergoes fast, unobservable reductive elimination. The kinetic parameters for the rate-limiting isomerization of 6 to 7 (ΔH‡ = 16.9 ± 0.4 kcal/mol; ΔS‡ = −7.8 ± 1.6 e.u.) are reflective of a low-barrier process with a small entropic penalty (Figure 6). This step is first-order in 6, and unaffected by excess Bu4NCl. Complex 6 also oxidizes excess ligand, presumably via chloronium transfer; therefore, phosphine dissociation is likely not a prerequisite for either isomerization or reductive elimination, since PNP(AuCl)2 is formed quantitatively when no excess ligand is added, and no decomposition of liberated ligand (PNP + 6) is observed. These results are consistent with a mechanism involving intramolecular aryl transfer from 6 to generate 7, which then directly reductively eliminates to PNP(AuCl)2 and 4,4’-difluorobiphenyl.

Figure 4

Oxidation of bimetallic Au(I)/Au(I) complexes is under kinetic control, affording Au(II)/Au(II) intermediates which isomerize to mixed-valent Au(I)/Au(III) species and reductively eliminate biaryl from a single metal center. Solutions of PNP[Au(4-F-C6H4)]2 (5) and dppm[Au(4-F-C6H4)]2 (8) undergo fast oxidation by PhICl2 upon thawing to generate intermediates stabilized by intramolecular Au(II)-Au(II) bonds. Complex 6 undergoes isomerization slower than 9, perhaps due to decreased ligand flexibility. Since isomerization for PNP-supported complexes is rate-limiting, reductive elimination from 7 is unobservable. In contrast, the rates of isomerization and reductive elimination of dppm-supported complexes are comparable, allowing kinetic analysis of both steps to preclude the possibility of a bimetallic reductive elimination directly from 9.

Figure 6

Five-atom linker dppp discourages formation of Au(II)/Au(II) intermediates upon oxidation of bimetallic dppp[Au(4-F-C6H4)]2 (11). Solutions of 11 undergo fast oxidation by PhICl2 upon thawing to afford directly the unobservable Au(I)/Au(III) complex 12, which isomerizes to 13. Reductive elimination from 13 affords 4,4’-difluorobiphenyl.

We reasoned that N→P π-donation in PNP-type ligands results in a barrier to N-P bond rotation that becomes significant at low temperature[49]. A more flexible ligand, such as 1,2-bis(diphenylphosphino)methane (dppm), could presumably facilitate intramolecular aryl transfer from a bimetallic Au(II) intermediate, resulting in a lower barrier to isomerization. Oxidation of dppm[Au(4-F-C6H4)2]2 (8) at −78 °C by PhICl2 affords symmetric bimetallic Au(II) species 9 (19F NMR and 31P NMR singlets at −120.3 ppm and −18.2 ppm, respectively), which is consumed faster than analogous PNP complex 6 (Figure 4). Indeed, at −52 °C, 9 undergoes first-order decay, with fast appearance and slow consumption of mixed-valent Au(I)/Au(III) complex 10, and appearance of 4,4’-difluorobiphenyl at approximately the same rate as the disappearance of 9 (Figure 5). Kinetic modeling precludes any scenario in which a bimetallic reductive elimination occurs directly from 9. Instead, 9 likely undergoes irreversible isomerization to 10 (k = (4.6 ± 0.1)×10−4 s−1), which then reductively eliminates 4,4’-difluorobiphenyl at a comparable rate (k = 3.9×10−4 s−1). At −30 °C, both reactions are immeasurably fast. While PNP-supported Au(I)/Au(III) species 7 could not be observed, these findings support a similar reductive elimination mechanism of PNP-supported complexes involving the intermediacy of such mixed-valent species as proposed above.

Figure 5

Monitoring oxidation of dppm[Au(4-F-C6H4)]2 (8) to Au(II)/Au(II) intermediate 8, isomerization to Au(I)/Au(III) species 10 and reductive elimination of 4,4’-difluorobiphenyl. a, Reaction profile for the isomerization of 9 to 10 and reductive elimination of biaryl from 10 at −52 °C, monitored by 19F NMR. b, Selected region of 19F NMR spectrum (470 MHz, CD2Cl2) of oxidation of 8 at −52 °C. Complex 8 has been fully consumed. The equivalent fluorides of Au(II)/Au(II) species 9 are clearly visible at −120.3 ppm, as are the inequivalent fluorides in Au(I)/Au(III) species 10 at −118.2 and −118.6 ppm. Error bars represent the inherent error in interpreting 19F NMR signal intensity.

Using the dppm-supported system discussed above as a benchmark, we probed the behavior of a bimetallic system with a five-atom linker which could perturb intramolecular metal-metal interactions. We envisioned that a longer linker (such as 1,3-(diphenylphosphino)propane (dppp)) between Au atoms might preclude formation of a 7-membered metallacyclic Au(II)-Au(II) intermediate upon oxidation. Indeed, the only product of low-temperature oxidation of dppp[Au(4-F-C6H4)]2 (11) by PhICl2 is mixed-valent species 13 (two 1:1 19F NMR singlets at −118.3 and −118.6 ppm and two 1:1 31P NMR singlets at 27.3 and 21.7 ppm), presumably formed via fast isomerization of 12 (Figure 6). Even at −80 °C, no signals consistent with a symmetric bimetallic Au(II) species are observed, suggestive that a mixed-valent species is formed directly upon oxidation. Upon warming to −52 °C, 13 undergoes first-order reductive elimination to dppp(AuCl)2 and 4,4’-difluorobiphenyl (k. = (3.0 ± 0.1)×10−5 s−1). Hence, the kinetic profiles of oxidation and reductive elimination of bimetallic complexes bearing linkers that restrict metal-metal interactions resemble that of mononuclear species.

Importantly, these studies reveal that access to a kinetic bimetallic Au(II) product of oxidation does not necessarily compromise the rate of reductive elimination from Au(III); in fact, the rate of reductive elimination from dppm-supported 10 is about an order of magnitude faster than that of 13, perhaps due to increased steric crowding as a result of a shorter linker, aurophilic interactions between the Au centers in 10, or a combination thereof. Even Au(II)/Au(II) → Au(I)/Au(III) isomerization for PNP and dppm-supported complexes is a facile process at temperatures below 0 °C.

Conclusion

We have uncovered several unexpected properties of oxidized gold that are contrary to established ideas. First, C-C bond reductive elimination from Au(III) does not necessarily require phosphine dissociation. By avoiding this rate-limiting process, biaryl reductive eliminations from Au(III) are remarkably fast, and occur as low as −52 °C. Furthermore, by judicious choice of ligand, the barrier to oxidation of bimetallic Au(I) complexes may be reduced, allowing access to symmetric intermediates with Au(II)-Au(II) bonds. When stabilized by dppm, these complexes undergo fast isomerization and reductive elimination at very low temperature; when stabilized by more rigid PNP-type ligands, they may be arrested, although at slightly higher temperatures, isomerization and reductive elimination become facile.

The stoichiometric behavior of the complexes reported in this study inform the development of catalysis involving Au(I)/Au(III) redox cycling. For instance, that phosphine dissociation is not required for C-C bond formation may slow or entirely preclude deleterious ligand oxidation by excess oxidant under conditions relevant to catalysis. Efforts to develop gold-catalyzed transformations using relatively weak oxidants compatible with substrates should focus on bimetallic Au(I) precatalysts with three-atom linkers, such as PNP or dppm. Given that transmetalation to Au(I) can be achieved at temperatures as low as −78 °C50, and that Au(I) oxidation, isomerization and reductive elimination are facile processes below 0 °C, catalysis may be achieved at temperatures low enough to avoid unwanted side reactions between oxidant and substrate. Investigations probing ligand and electronic effects on the reductive elimination at Au(III), and stoichiometric reactivity of bimetallic Au(II) complexes are ongoing to further our nascent understanding of the fundamental chemistry of oxidized gold.