Mechanism of Photoredox-Initiated C−C and C−N Bond Formation by Arylation of IPrAu(I)−CF₃ and IPrAu(I)−Succinimide

Abstract

Herein, we report on the photoredox-initiated gold-mediated C(sp²)−CF₃ and C(sp²)−N coupling reactions. By adopting gold as a platform for probing metallaphotoredox catalysis, we demonstrate that cationic gold(III) complexes are the key intermediates of the C−C and C−N coupling reactions. The high-valent gold(III) intermediates are accessed by virtue of photoredox catalysis through a radical chain process. In addition, the bond-forming step of the coupling reactions is the reductive elimination from cationic gold(III) intermediates, which is supported by isolation and crystallographic characterization of key Au(III) intermediates.

Graphical Abstract

■ INTRODUCTION

The merger of photoredox catalysis and transition metal catalysis has enabled novel modes of reactivity via organic radical intermediates and odd-electron transition metal intermediates.¹ While this synergistic combination has been incredibly successful in upgrading the arsenal of synthetic chemists,² current mechanistic understanding of key bond-forming events remains limited. Transition metal catalysts and intermediates are subject to various visible light-induced reactions such as charge-transfer, electron-transfer, and energy-transfer that can complicate mechanistic analysis. For example, in the case of nickel/photoredox dual catalysis, the Doyle group proposed photolysis of aryl nickel(III) chloride intermediates and metal to ligand charge transfer (MLCT) of aryl nickel(II) halide intermediates as elementary steps.³ These light-induced processes, combined with the potential for odd electron intermediates and the instability of high-valent transition metal complexes, can make the details of the bond-forming step often difficult to ascertain.

A number of possibilities exist for the mechanism of bond formation in combined photoredox-transition metal catalysis, including a traditional reductive elimination,⁴ radical substitution,^5,6 or radical recombination (Scheme 1A).⁷ For example, copper/photoredox dual catalysis has been used by the MacMillan group to facilitate the formation of C−N bonds. These reactions are proposed to proceed via reductive elimination from a copper(III) intermediate (Scheme 1B).⁸ On the other hand, under the photoinduced Ullmann C−N coupling conditions reported by the Fu and Peters groups, the prevailing explanation for C−N bond-forming step is radical substitution reaction between copper(II) intermediates and carbon-centered radical intermediates (Scheme 1C).⁹ In this latter case, EPR and radical clock experiments are consistent with copper(II) and aryl radical species; however, the formation of a copper(III) intermediate has yet to be directly observed or excluded from a light-induced transformation. These mechanistic considerations are challenging to address due to the short lifetimes of radical and postulated high-valent copper(III) intermediates, especially under metallaphotoredox catalysis conditions.

Scheme 1.

Potential Bond-Forming Steps of Metallaphotoredox Catalysis (A) Copper(III) Intermediates Proposed for Photochemical C−N Coupling Reactions (B and C)

Light-promoted reactivity, particularly involving aryl diazonium salts,^10,11 has also recently been leveraged to enable gold-catalyzed transformations. Moreover, gold complexes have been employed as photoredox catalysts for transformations of alkyl halides that are similar to the aforementioned copper-catalyzed processes.¹² Theorectical calculations support a mechanism involving the generation of gold(III) intermediates;¹³ however, as is the case with copper catalysis, isolation of a photoredox-generated coordinatively unsaturated gold(III) complex capable of bond-formation through reductive elimination¹⁴ has proven challenging.¹⁵

Recent studies from our group demonstrated that reductive elimination of trifluoromethylgold(III) complexes is a viable strategy to forge C−CF₃ bonds;¹⁶ therefore, we hypothesized that diamagnetic gold(III) complexes generated from a photoredox-mediated oxidative addition might be isolated and characterized. Moreover, in direct analogy to Kochi’s seminal mechanistic studies on the use of gold(I/III) as a mechanistic probe for gaining insight into carbon−carbon bond formation from copper(III), isolation of the gold(III) complex may provide insight into the copper(III) complexes proposed as intermediates in metallaphotoredox cross-coupling catalysis.¹⁷ Trapping of these intermediates, presumably as less reactive and detectable d8-square planar complexes, would also allow for examination of the viability of the bond-forming event. Ultimately, on the basis of our understanding of the intermediates, we hoped to elucidate the bond-forming step of gold-mediated photoreodox catalysis and, eventually, to provide mechanistic insight to related heretofore uncharacterized metallaphotoredox processes.

Finally, we realized if mechanistic evidence supported the reductive elimination as the bond-forming step, we could also take advantage of these findings to examine other unprecedented reductive elimination reaction at gold(III). For example, we were especially intrigued by C(sp²)−N coupling reactions due the dearth of examples of C−N reductive elimination from gold(III) complexes and the ambiguity over the copper(III) intermediates in the copper-catalyzed C(sp²)− amide coupling reactions. Although copper(III)-mediated C(sp²)−N reductive elimination is supported by stoichiometric reactions of well-defined aryl−copper(III) complexes,¹⁸ and the formation of free aryl radical intermediates during the copper-catalyzed C(sp²)−amide coupling reaction is not plausible,¹⁹ the formation of copper(III) intermediates during the catalytic reactions remains inconclusive.²⁰ In this study, we provide mechanistic support for the formation of high-valent gold intermediates during photoredox-initiated gold-mediated C(sp²)−CF₃ and C(sp²)−N coupling reactions.

■ RESULTS AND DISCUSSION

We decided to investigate the reaction mechanism of photoredox-initiated gold-mediated C(sp²)−CF₃ coupling reaction as a consequence of the following consideration. All three proposed bond-forming scenarios (Scheme 1A) are plausible with a trifluoromethyl group, and, as a result, the accessible reaction mechanism is not biased.²¹ Also, if transition metal trifluoromethyl complexes are involved, undesired side reactions such as hydrogen atom transfer and β-hydride elimination from the resulting trifluoromethylgold(III) intermediates could be largely ignored.

With this in mind, we sought conditions for the photoredox-induced coupling of trifluoromethylgold(I) complex 1 with 4-nitrophenyl diazonium tetrafluoroborate. The desired C(sp²)−CF₃ coupling product was obtained in 60% yield using nitromethane as a solvent, 1 equiv of 1, and 1.5 equiv of 4-nitrophenyl diazonium tetrafluoroborate (Table 1, entry 1). Control experiments support that both the ruthenium complex and light are required for the reaction to proceed (Table 1, entries 2 and 3). The analogous iodonium electrophile was not as efficient as the diazonium salt (Table 1, entry 4), and significantly less of the desired product was formed when the Ru(bpy)₃(PF₆)₂ was replaced by other photoredox catalysts (Table 1, entries 5−9). The exclusion of air and water from the reaction media was critical for efficient reaction (Table 1, entries 10 and 11).

Table 1.

Control Experiment^a,b

entry	variation from the reaction condition	yieldc,d
1	none	60%
2	no catalyst	<1%
3	dark and 40 °C instead of blue LED and rt	<1%
4	ArI(Mes)OTf instead of ArN2BF4	22%
5	Ir(ppy)2(dtbbpy)PF6 instead of Ru(bpy)3(PF6)2	11%
6	Ir{dF(CF3)ppy}2(dtbbpy)PF6 instead of Ru(bpy)3(PF6)2	18%
7	Ru(bpz)3(PF6)2 instead of Ru(bpy)3(PF6)2	19%
8	Cu(dap)2Cl instead of Ru(bpy)3(PF6)2	<1%
9	Mes−AcrBF4 instead of Ru(bpy)3(PF6)2	7%
10	under air instead of three freeze−pump−thaw cycles	24%
11	CH3NO2:H2O = 96:4 (v/v) instead of CH3NO2	<1%e

^aReactions run at 8 μmol scale.

^bThe temperature increases up to 40°C.

^c19F NMR yield by using PhOCF₃ as an internal standard.

^dIn triplicate.

^e9% of IPrAuCF₃ and 55% of Au^IIICF₃ were observed.

When water was present in the reaction mixture, instead of the desired product, another CF₃-containing species was observed by 19F NMR (55% yield in Table 1, entry 11). We postulated that this trifluoromethyl-containing species might be the four-coordinate gold(III) intermediate with one of the coordination sites occupied by water, thereby inhibiting reductive elimination (Scheme 2). This hypothesis led us to evaluate other coordinating aprotic polar solvents. In the event, replacing nitromethane with acetonitrile provided the analogous CF₃-containing gold(III) acetonitrile complex. In this case, other CF₃-containing species such as the starting gold(I) complex, the coupling product, and any other side products were not detected using ¹⁹F NMR (Figure 1).

Scheme 2.

Intercepted Gold(III)−CF₃ Intermediates^a

^aAr = 4-nitrophenyl.

Figure 1.

¹⁹F NMR spectra of gold-mediated C(sp²)−CF₃ coupling reaction, solvent = CH₃CN.

This new species, 4, was analyzed by HRMS, ¹³C NMR, and ¹H NOESY (see the Supporting Information). Two new quartets in the ¹³C NMR spectrum with ³J_C−F values of 25 and Hz, respectively, support that the IPr, CF₃, and 4-nitrophenyl are all bound on the same gold atom but in a different stereochemical relationship. ¹H NOE between IPr and 4-nitrophenyl supports that they are cis to each other, and therefore we predicted the CF₃ group was trans to the IPr ligand. After the acetonitrile-bound gold(III) intermediate was characterized by NMR studies, the gold(III) aquo complex was crystallized as the tetrafluoroborate salt. The structure was analyzed by single-crystal X-ray diffraction and further supported the aforementioned spectroscopic characterizations (Figure 2).¹⁴

Figure 2.

Crystal structure of 3-BF₄. Thermal ellipsoid plot is drawn at 50% probability. Hydrogen atoms and solvents are omitted for clarity.

On the basis of this evidence for the formation of a gold(III) complex as an intermediate, we sought to unravel the mechanism of photoredox-promoted oxidative addition. To gain a deeper insight into the role of the visible light in this transformation, the emission spectrum of the light source, Kessil H150, was compared to the absorption spectra of all chemical compounds in the reaction mixture. Although there are examples in which gold complexes initiate radical reactions as photosensitizers,¹² we observed that the radiated visible light was exclusively absorbed by the ruthenium photoredox catalysts in our system (Figure 3). This conclusion is further supported by the previously performed control experiment (Table 1, entry 2).

Figure 3.

(Top) Three possible modes of sensitization with visible light. Paths A and B were ruled out. (Bottom) UV−vis absorption spectra of all chemicals (concentration = 0.05 mM in CH₃NO₂) and emission spectrum of the light source, Kessil H150.

To explore the elementary step following the excitation of the photoredox catalyst, a photoluminescence quenching experiment was performed. The Stern−Volmer relationship supports the hypothesis that only the aryldiazonium salt reacts with the excited photoredox catalyst (Figure 4, see the Supporting Information). An electron-transfer process is supported by the evolution of nitrogen gas upon radiation of visible light in the absence of gold complexes.²²

Figure 4.

(Top) Two possible modes of photoluminescence quenching. Path E was ruled out. (Bottom) Stern−Volmer relationship. [Ru(bpy)₃(PF₆)₂] = 0.05 mM in CH₃NO₂.

The generation of odd-electron intermediates from photoluminescence quenching suggests that oxidative addition could proceed through a gold(II) intermediate. Two distinct gold(II) intermediates were considered: a neutral or a cationic gold(II) species (Figure 5). The cationic pathway would presumably result from oxidation of the gold(I) complexes by the photoredox catalysts. To clarify whether the Ru(III) intermediate is capable of oxidizing IPrAuCF₃, the cyclic voltammogram of both Ru(bpy)₃(PF₆)₂ and IPrAuCF₃ was measured and showed that the redox potential of Ru(II)/ Ru(III) is not adequate to oxidize IPrAuCF₃ to the corresponding cationic gold(II) intermediate (Figure 5, see the Supporting Information). As a result, we envision that the free aryl radical reacts directly with 1 to generate a neutral (trifluoromethyl)arylgold(II) intermediate.

Figure 5.

(Top) Two possible reaction mechanisms of oxidative addition. (Bottom) Cyclic voltammogram of Ru(bpy)₃(PF₆)₂ and IPrAuCF₃, scan rate = 100 mV/s.

With the implication of the neutral gold(II) intermediate and the direct observation of a trifluoromethylgold(III) species, we sought to ascertain the oxidant responsible for the single electron oxidation from the gold(II) to gold(III). To this end, the quantum yield of oxidative addition was mearsured.²³ By radiating 3.3 mol % of photons over 20 min in a fluorometer at room temperature, the nitromethane-ligated trifluoromethylgold(III) complex was observed in 22% yield; but, the reductive elimination product was not detected (Figure 6, see the Supporting Information). The quantum yield of 6.6 implies that the oxidative addition involves a chain process, wherein the gold(II) intermediate is mainly oxidized by aryldiazonium. Furthermore, in sharp contrast to the CH₃CN and H₂O-bound gold(III) complexes, the CH₃NO₂-bound gold(III) complex readily underwent quantitative reductive elimination upon heating up to 40 °C (Figure 6). This observation suggests that the activation barrier of reductive elimination might be overcome by heat from the light source.²⁴

Figure 6.

(Top) Revised reaction mechanism of oxidative addition. The major operating pathway is path I. Path J was not observed. (Bottom) Quantum yield and reductive elimination of CH₃NO₂-bound gold(III) intermediate. Yields are reported in triplicate.

During quantum yield measurements, we also uncovered a chain terminating event. The presence of unreacted IPrAuCF₃ and 4-nitrophenyl diazonium salts indicates that the chain propagation is interrupted. According to a previous report on gold(II) intermediates,⁸ we proposed the release of trifluoromethyl radical from the gold(II) intermediate as the chain terminating event (Figure 6, path K). Additionally, given that this process competes with oxidation of the gold(II) intermediate, the electronic properties of aryldiazonium are expected to influence the yield of the reaction. As we hypothesized, a decrease in product yield was observed as the σ_para of substituents inverts from positive values (Table 2, entries 1−6) to negative values (Table 2, entries 8−10). Moreover, to support the release of trifluoromethyl radical, we attempted to intercept the trifluoromethyl radical with radical traps. Addition of 70 equiv of benzene to the optimized reaction condition produced trifluoromethylbenzene as a major product (Scheme 3, see the Supporting Information).²⁵

Table 2.

Electronic Effect on Aryldiazonium^a,b

entry	X	yieldc,d
1	NO2	60%
2	CO2Me	48%
3	Ac	43%
4	Br	46%
5	Cl	52%
6	F	44%
7	H	19%
8	Me	17%
9	tert-Bu	16%
10	OMe	15%

^aReactions run at 8 μmol scale.

^bThe temperature increases up to 40 °C.

^c19F NMR yield by using PhOCF₃ as an internal standard.

^dIn triplicate.

Scheme 3.

Radical Trapping Experiment

Having gained evidence for the stepwise formation of a gold(III) intermediate as relevant to the C−C bond formation in the photoredox-promoted gold-mediated C(sp²)−CF₃ coupling reaction, we sought to apply these mechanistic insights to examine the possibility that a similar pathway is involved in previously unknown photoredox driven C(sp²)−N bond formation.

In the event, the product (6) of the photoredox-initiated gold-mediated C(sp²)−N coupling reaction was obtained in 68% yield from gold(I)−succinimide (5) by only slight modification of the reaction conditions previously employed for trifluoromethylation (Scheme 4). It is notable that other gold−amides did not undergo the C−N bond-forming reaction with the same efficiency; rather, competitive protodeauration (formation of an N−H bond) was observed. Notably, the relative stability of the N-centered radical species that would be generated by Au−N bond scission appears to be irrelevant to the efficiency of the gold-mediated C(sp²)−N coupling reactions. On the basis of additional experiments (see the Supporting Information), we propose that the nature of the amide ligand is important for suppressing uncatalyzed reactions of the gold(I)−amide, such as direct oxidation by the photoredox catalyst or reaction with electrophilic diazonium salt. This hypothesis is consistent with our previous observations that gold(I)−amides show increased nucleophilicity.²⁶

Scheme 4.

Photoredox-Initiated Gold-Mediated C(sp²)−Succinimide Coupling Reaction^a

^aReactions run at 8 μmol scale. The temperature increases up to 35 °C. ¹H NMR yield by using residual CHD₂NO₂ as an internal standard. In triplicate.

To further explore the possibility that high-valent gold(III) intermediates are involved in the C(sp²)−N coupling reaction, experiments to intercept the potential gold(III) imide intermediate were untaken (Scheme 5). These experiments resulted in the successful isolation and characterization of a gold(III) aquo succinimide complex (Figure 7). The isolation of this gold(III) species from a photoredox-initiated process is consistent with the hypothesis that the photoredox-promoted gold-mediated C(sp²)−succinimide coupling reaction proceeds by a mechanism closely related to that established for the photoredox-initiated gold-mediated C(sp²)−CF₃ coupling reaction: oxidative addition followed by gold(III)-centered C(sp²)−N reductive elimination.

Scheme 5.

Intercepted Gold(III)−Succinimide Intermediate^a

^aAr = 4-nitrophenyl.

Figure 7.

Crystal structure of 7-BF₄. Thermal ellipsoid plot is drawn at 50% probability. Hydrogen atoms and solvents are omitted for clarity.

■ CONCLUSION

In summary, the unique characteristics of gold have simplified the mechanistic understanding of metallaphotoredox-mediated trifluoromethylation. This reaction was initiated by visible light in the presence of a photoredox catalyst. Evidence was gained in support of a mechanistic scenario involving addition of the aryl radical to trifluoromethylgold(I) complex to generate a gold(II) species that undergoes a single electron oxidation to access the gold(III) intermediate. This oxidation likely occurs by reaction of the gold(II) species with the aryldiazonium salt, thereby generating an aryl radical. This chain mechanism is consistent with the measured quantum yield (6.6) for the process. After radical chain oxidative addition, the gold(III) intermediate, which was isolable in coordinating solvents, is converted to the desired product via reductive elimination (Figure 8). We propose that the reductive elimination, which regenerates gold(I), is the bond-forming step of photoredox-initiated gold-mediated C(sp²)−CF₃ and C(sp²)−N coupling reactions. To the best of our knowledge, the analogous copper(III) intermediates in copper/photoredox dual catalytic trifluoromethylation and amination reactions have not been isolated or characterized. Thus, the use of gold, as another member of group 11, may provide useful insight into the mechanism of these metallaphotoredox-catalyzed transformations.

Figure 8.

Summary of proposed reaction mechanism, Solv = CH₃NO₂.