Reductive Elimination Reactions in Gold(III) Complexes Leading to C(sp3)–X (X = C, N, P, O, Halogen) Bond Formation: Inner-Sphere vs SN2 Pathways

The reactions leading to the formation of C–heteroatom bonds in the coordination sphere of Au(III) complexes are uncommon, and their mechanisms are not well known. This work reports on the synthesis and reductive elimination reactions of a series of Au(III) methyl complexes containing different Au–heteroatom bonds. Complexes [Au(CF3)(Me)(X)(PR3)] (R = Ph, X = OTf, OClO3, ONO2, OC(O)CF3, F, Cl, Br; R = Cy, X = Me, OTf, Br) were obtained by the reaction of trans-[Au(CF3)(Me)2(PR3)] (R = Ph, Cy) with HX. The cationic complex cis-[Au(CF3)(Me)(PPh3)2]OTf was obtained by the reaction of [Au(CF3)(Me)(OTf)(PPh3)] with PPh3. Heating these complexes led to the reductive elimination of MeX (X = Me, Ph3P+, OTf, OClO3, ONO2, OC(O)CF3, F, Cl, Br). Mechanistic studies indicate that these reductive elimination reactions occur either through (a) the formation of tricoordinate intermediates by phosphine dissociation, followed by reductive elimination of MeX, or (b) the attack of weakly coordinating anionic (TfO– or ClO4–) or neutral nucleophiles (PPh3 or NEt3) to the Au-bound methyl carbon. The obtained results show for the first time that the nucleophilic substitution should be considered as a likely reductive elimination pathway in Au(III) alkyl complexes.


■ INTRODUCTION
The advances in the field of Au(I)/Au(III) catalysis during the last decade 1−15 have boosted research on fundamental aspects of gold redox reactions. 16−23 Thus, the oxidative addition of substrates containing C−halogen, C−N 2 + , or strained C−C bonds to Au(I) complexes, regarded as the most challenging step of these catalytic cycles, has received considerable attention. 24−30 By contrast, reductive eliminations in Au(III) complexes have been less studied, despite the fact that the degree of selectivity in this step dramatically affects the outcome of the catalytic processes. 31 Pioneering studies carried out during the 1970s 32−34 revealed that reductive elimination in square planar Au(III) complexes can occur (a) through tricoordinate intermediates formed after ligand dissociation or (b) directly from the tetracoordinate complexes (Scheme 1A). 24,25 Alternatively, a concerted mechanism where C−C and B−F bonds form simultaneously has been proposed for the reaction of Au(III) alkyl fluorido complexes with arylboronic acids (Scheme 1B). 35 Whereas the experimental evidence of pathways (a) 32−34,36−39 and (b) 35,40−48 is abundant for C−C coupling reactions, mechanistic data on reductive elimination processes leading to the C−heteroatom bond formation are scarce. The available information has been obtained on the dissociative reductive eliminations of alkyl, 49,50 aryl, 38,51 or trifluoromethyl halides 52 and nondissociative reductive eliminations of ArX derivatives (X = NR 2 , PR 3 + , OR, SR, halogen). 53−59 Alkyl esters of strong oxyacids such as triflic, perchloric, or nitric acids are powerful electrophiles, capable of alkylating organic 60−62 or metal-based 63−66 nucleophiles. In particular, an oxidative addition of MeOTf has been invoked to explain the formation of ethane or toluene in the reaction of [AuR(IMes)] with MeOTf (Scheme 1C). 67 The reverse reaction, namely, the reductive elimination of organic triflates, perchlorates, or nitrates from metal complexes is thus very unlikely. Nevertheless, it has been postulated for the Bi-or Cu-catalyzed formation of aryl or vinyl triflates 68,69 and in the reaction of [Au(CF 3 ) 2 (Me)(IPr)] with Me 3 SiOTf (Scheme 1D). 70 Direct reductive elimination of an alkyl nitrate or tosylate has been observed in cyclometalated Pd(IV) complexes (Scheme 1E). 71 Herein, we report a study on reductive elimination reactions from Au(III) complexes of the type [Au(CF 3 )(Me)X(PR 3 )], which lead to C−C, C−N, C−P, C−O, or C−halogen bond formation. Uncommon reductive elimination products, such as methyl fluoride, triflate, perchlorate, and nitrate, have been observed. Mechanistic and computational studies show that, depending on the ligand X, the classical dissociative pathway or an S N 2-type pathway operates in these reactions.
Complex 3a rapidly reacted in chloroform or dichloromethane with strong protic acids, such as triflic, concentrated aqueous perchloric or nitric, trifluoroacetic, hydrobromic, or hydrochloric acid to give methane and complexes [Au(CF 3 )-(Me)(X)(PPh 3 )] (X = OTf (4a), OClO 3 (5a), ONO 2 (6a), OC(O)CF 3 (7a), Br (8a), Cl (9a)) as the main reaction products (Scheme 2). Similarly, the reaction of 3b with triflic or hydrobromic acid gave [Au(CF 3 )(Me)(X)(PCy 3 )] (X = OTf (4b), Br (8b)). Remarkably, only one of the methyl ligands was protonated even in the presence of an excess of acid. No significant reaction was observed between 3a and hydrofluoric acid, acetic acid, phenol, or 4-methoxythiophenol. The protonation of the robust Au−C(sp 3 ) bonds 72 of 3a or 3b with acids is in line with the previous results of Kochi and coworkers on the protonolysis of complexes [Au-(alkyl) 36 where only one of the Au−alkyl bonds cis to the phosphine is protonated, in agreement with the mutually exerted large trans effect of these alkyl ligands.
Complexes 6a, 7a, 8a, and 8b were isolated in good yields. The isolated samples of 8a contained small amounts (4−7%) of the corresponding isomer 8a′, where the PPh 3 and Me ligands are mutually trans (see below). In contrast, the attempts to isolate 4a, 4b, or 5a gave impure oils, which we attribute to their lower stabilities (see the Supporting Information).
The outcome of the reaction of 3a with triflic acid depended on the solvent. Thus, the Au−Me bond acidolysis was the dominant process in dichloromethane, chloroform, or toluene, whereas the isomerization to cis-[Au(CF 3 )(Me) 2 (PPh 3 )] (3a′), with the concurrent reductive elimination of ethane, was the main process in tetrahydrofuran (THF) or acetone (Scheme 2). This reaction occurs even in the presence of a substoichiometric amount of acid (10%) but not when acetic acid was used instead of triflic acid.
These observations can be rationalized by considering the influence of the solvent on the pK a of triflic acid. 73 Thus, the irreversible Au−Me acidolysis needs a strongly acidic medium and therefore it takes place only in those solvents where the pK a of the acid is lowest (dichloromethane, chloroform, and toluene). In contrast, in THF or acetone, the pK a of triflic acid would not be low enough to protonate the methyl carbon. Then, other reaction pathways would come into play, where the acidic medium facilitates the isomerization and reductive elimination reactions of the Au(III) complexes (see below).
The reaction of a mixture of isomers 8a and 8a′ with AgF gave AgBr and the corresponding fluorido complexes 10a and 10a′ (Scheme 3), which were unambiguously identified in solution by NMR spectroscopy, but could not be isolated in pure form. The reaction of in situ-generated 4a and KI gave a mixture containing mainly PPh 3 , [Au(CF 3 )(Me)(I)(PPh 3 )] (11a and 11a′), and another gold complex containing the CF 3 and CH 3 ligands but lacking the PPh 3 ligand, which was tentatively identified as cis-or trans-K[Au(CF 3 )(Me)I 2 ] (12) (Scheme 3). MeI and [AuI(PPh 3 )] were also detected by NMR spectroscopy in the reaction mixture. The reaction of 4a with PPh 3 cleanly gave cis-[Au(CF 3 )(Me)(PPh 3 ) 2 ] (13), which was isolated in good yield (Scheme 3).
Single crystals were obtained from an in situ-generated solution of 5a. However, the X-ray diffraction analysis showed that the crystal contained the salt [Au(CF 3 )(Me)(OH 2 )-(PPh 3 )]ClO 4 (5a·H 2 O), instead of the expected perchlorato complex (Figure 1), suggesting that under the reaction conditions, 5a and H 2 O are in equilibrium with 5a·H 2 O. In the crystal structure, each ClO 4 − anion is hydrogen-bonded to two coordinated water molecules to form ···O−Cl−O···H− O−H··· chains along the a axis. The Au−OH 2 distance (2.1562(17) Å) is almost identical to the value found in cis-[Au(Me) 2 (OTf)(OH 2 )] (2.157(6) Å). 74 The crystal structure of 6a ( Figure 1) shows that the nitrato ligand is coordinated to gold through one of the oxygen atoms. The Au−ONO 2 distance (2.133(2) Å) is similar to that found in [Au- (2.090(4) Å). 76 In both 5a·H 2 O and 6a, the trifluoromethyl and phosphine ligands are in a mutual trans disposition. The 19 F and 31 P{ 1 H} NMR spectra of 3a−9a, 11a, 3b, 4b, and 8b showed a doublet and a quartet, respectively, with a large 3 J PF value (62.4−72.8 Hz) characteristic of a mutually trans arrangement of the phosphine and trifluoromethyl ligands. The 31 P{ 1 H} NMR spectrum of 13 showed the presence of two inequivalent and mutually coupled 31 P nuclei with an additional splitting due to 31 P− 19 F coupling, indicating a cis configuration. The 19 F 3 C and 31 PPh 3 resonances of 10a and 10a′ show coupling with the gold-bound fluorine nucleus, and the 19 F NMR spectrum of the mixture displays two signals at −223.7 (10a) and −246.6 ppm (10a′), which fall in the same region as those of previously reported Au(III) fluorido complexes. 49 Table 1). Thermal isomerization was observed as a secondary process in most cases. Upon heating, the concentration of isomers 5a′−10a′ increased rapidly and then remained steady at the equilibrium proportions shown in Table  1, until both isomers completely transformed into the corresponding reductive elimination products. Exceptions to this behavior were the triflato complex 4a, which gave very small amounts of its isomer 4a′, and the PCy 3 complexes 3b, 4b, and 8b, for which no isomerization was observed. The fastest decompositions were observed for the triflato and perchlorato complexes (4a, 4b, and 5a), which were consumed in 0.5−4.5 h at 50°C. In contrast, the decomposition of 6a, 7a, 8a, 9a, or 10a required higher temperatures and longer times. Finally, the decomposition of the PCy 3 complexes 3b and 8b required more energetic conditions compared to their PPh 3 counterparts 3a and 8a. Au(I) complexes [Au(PR 3 ) 2 ] + (R = Ph or Cy) were formed as secondary products in most cases. These products arise from hydrolysis or decomposition of the resulting Au(I) complexes 1a and 1b (see the Supporting Information). Small amounts (5%) of phosphonium salts (PMePh 3 )X (X = OC(O)CF 3 , Cl, Br) were observed in the decompositions of 7a, 8a, and 9a. Similarly, complex 13 underwent a quantitative reductive elimination of (PMePh 3 )OTf after heating for 2 h at 80°C or 4 h at 60°C (Scheme 4).
The configurational assignment of isomers 3a′−11a′ is supported by their smaller 3 J PFcis (9.1−11.8 Hz) and larger 3 J PHtrans (8.9−9.7 Hz) values with respect to those of 3a−11a ( 3 J PFtrans 62.4−72.8 Hz; 3 J PHcis 5.8−6.6 Hz), as well as by the larger 31 P− 13 CH 3 coupling constant of 7a′ ( 2 J PCtrans = 91.6 Hz) compared to that of 7a ( 2 J PCcis = 3.8 Hz). 79 Mechanistic Studies. The concentrations of the reactants and products during the thermal decomposition of representative complexes (3a, 8a, and 13) were monitored by NMR spectroscopy. In addition, the effect of the presence of an additional amount of PPh 3 was studied.
The decomposition of the dimethyl complex 3a was monitored at 100°C (Figure 2). The formation of ethane from 3a was drastically inhibited by the addition of PPh 3 (0.2 or 1 equiv), which is in agreement with a mechanism where PPh 3 dissociation gives the tricoordinate intermediate [Au-(CF 3 )(Me) 2 ], which decomposes to give ethane (Scheme 5). Remarkably, the rate of consumption of 3a increased with time until approximately half-conversion, suggesting the acceleration of the reaction by one of the reaction products. Indeed, the reaction was faster in the presence of added 1a (Figure 2). Heating of a solution of the bromido complexes 8a (96%) and 8a′ (4%) at 80°C led to the rapid equilibration of both isomers at an 82:18 ratio, respectively, which remained constant during the reaction ( Figure 3). As observed for 3a, the rate of formation of 1a increased with time until halfconversion, suggesting the acceleration of the reaction by a reaction product. In the presence of one additional equivalent of PPh 3 , both the consumption of 8a and 8a′ and the formation of MeBr were slower, in agreement with a PPh 3dissociative pathway ( Figure S83). However, in these conditions, the major reaction product was (PMePh 3 )Br (Scheme 6). Similarly, the heating of [Au(CF 3 )(Me)(Cl)-(PPh 3 )] (9a) in the presence of PPh 3 (5 equiv) gave mainly (PMePh 3 )Cl.
The slower reductive eliminations of the tricyclohexylphosphine complexes 3b and 8b are also in line with phosphine dissociation being the rate-determining step and are attributed to the higher donor ability of PCy 3 compared with that of PPh 3 .
The decomposition of 13 into 1a and (PMePh 3 )OTf was monitored at 60°C in CDCl 3 (Figure 4). In marked contrast with the behavior of 3a and 8a, the depletion of 13 showed a first-order dependence until an 85% conversion and, importantly, was accelerated by the added PPh 3 (the Scheme 4. Thermal Decomposition and Isomerization of the Au(III) Complexes All experiments were carried out in CDCl 3 except the decompositions of 3a, 3b, and 7a, which were carried out in D 8 -toluene (all spectra are given in the Supporting Information). b Necessary time for the consumption of at least 95% of the starting Au(III) complex. c Equilibrium proportion of isomers 4a′−10a′. d In situ-generated from 3a or 3b and HX. e The conversion was 6.6% after 14 h. f Determined by integration of the 1 H NMR spectra measured before and after the reaction. g Indirectly estimated from the resulting concentration of 1a. h The yield was not accurately determined in these cases. The room-temperature 31 P NMR spectrum of a mixture of 13 and PPh 3 did not show evidence of pentacoordinated species. Instead, only three sharp signals corresponding to 13 and free PPh 3 were observed ( Figure S103). To test the possibility of a nucleophilic attack, 13 was reacted with 2 equivalents of NEt 3 at 50°C. In these conditions, 1a and a mixture of (PMePh 3 )OTf and (NEt 3 Me)OTf in a 0.78:1 molar ratio formed (Scheme 7a). Besides, no signs of NEt 3 coordination or substitution of PPh 3 by NEt 3 were found in the NMR spectra. Overall, these observations point to the external attack of PPh 3 or NEt 3 on the methyl carbon being the main reaction pathway.
In the absence of added phosphine, the formation of (PMePh 3 )OTf from 13 could possibly occur through the dissociation of PPh 3 followed by nucleophilic substitution on the methyl carbon (Scheme 7b). In agreement with this, the reductive elimination of the phosphonium salt was faster in solvents with a higher coordinating ability (reaction times at 60°C : 1.5 h in acetone, 0.75 h in acetonitrile). The fast dissociation of PPh 3 was evidenced by the 31 P{ 1 H} NMR spectrum of 13 at 60°C as a broadening of the signal of the 31 P trans to the methyl ligand, with the loss of P−P and P−F    Figure S100). In addition, the observed first-order dependence of the consumption rate of 13 agrees with this mechanism if a steady concentration of the intermediate is assumed (Figure S96). The formation of (PMePh 3 )X (X = Cl, Br) from 8a or 9a in the presence of PPh 3 could follow a similar pathway, although in this case a fraction of the observed phosphonium salts could be originated by the reaction of the formed MeBr or MeCl with PPh 3 . The formation of small amounts of phosphonium salts during the thermal decompositions of 7a, 8a, and 9a (Table 1) suggests that the S N 2 pathway could be a secondary reductive elimination route in these cases.
In marked contrast, the reductive eliminations of MeOTf or MeOClO 3 from 4a, 4b, or 5a were significantly faster, suggesting a different reaction pathway. In addition, complexes 4a and 4b show comparable decomposition times, meaning that phosphine dissociation is not rate-determining in these cases. Considering the low coordinating abilities of OTf − and ClO 4 − , the degree of anion dissociation is expected to be higher for 4a, 4b, or 5a compared with the rest of the studied complexes and thus could reasonably be the starting point of the observed reactivity.
Computational Study. For an understanding of the decomposition reactions of the Au(III) complexes bearing a labile anionic ligand, the reductive elimination of MeOTf from 4a was modeled using DFT calculations at the B3LYP/(6-31G**+LANL2DZ) level in chloroform solution (see details in the Supporting Information). Different pathways involving the dissociation of PPh 3 or OTf − as the first step were calculated and compared ( Figure 5). In both cases, the relaxed scan of the potential energy as the Au−P or Au−O distance was elongated to 4.5 or 3.2 Å, respectively, produced a Morsetype curve with a gradual increase in energy as the ligands are separated from the metal ( Figure S110). Therefore, no transition state can be located using this theoretical model. These scans appear to indicate a higher activation barrier for the dissociation of PPh 3 compared to that of OTf − . However, the calculated free-energy change for the dissociation of OTf − to give the cationic tricoordinate intermediate [Au(CF 3 )(Me)-(PPh 3 )] + (Int1 + ) is significantly higher relative to the dissociation of PPh 3 to give [Au(CF 3 )(Me)(OTf)] (Int2), the latter being slightly exergonic. This can be explained by a higher entropy increase upon the dissociation of PPh 3 and, more importantly, an additional stabilization of the Int2 fragment due to the chelating coordination of the OTf − ligand.
The C−O coupling step could occur from Int1 + through two different mechanisms. The first one involves a threecentered transition state with a pseudotetrahedral coordination around the gold atom (TS1A), having a very high free energy. The second one involves an S N 2-like transition state resulting from the nucleophilic attack of the OTf − anion on the methyl ligand (TS1B) and provides a much more favorable pathway because it has virtually the same free energy as the previous dissociation step. On the other hand, the C−O coupling from Int2 would occur through a highly energetic tricoordinate transition state (TS2), which makes the PPh 3 dissociation pathway clearly unfavorable. The attempts to model the concerted reaction pathways through pseudotetrahedral or planar transition states formed from 4a or 4a′, respectively, converged to a transition state very similar to TS1A.
Thus, on the basis of these calculations and the abovepresented experimental data, the C−O reductive couplings from 4a, 4b, or 5a occur most likely via dissociation of the labile anionic ligand and subsequent nucleophilic attack on the metal-bound methyl group (Scheme 8). Previous studies have revealed that this is a feasible pathway for the reductive elimination of Csp 3 −X (X = N, O, Cl, Br, I) coupling products in Rh(III), 80,81 Pd(II), 82 Pd(IV), 71,83,84 and Pt(IV) 85−88 alkyl complexes. Interestingly, the studies on Pt(IV) complexes showed that reductive elimination is faster for more electron-  86 despite their lower nucleophilicity. 89 The S N 2 reductive elimination mechanism has been previously considered for Au(III) complexes, but to the best of our knowledge, no experimental evidence has been provided. Thus, in a computational study of the Au-catalyzed methane oxidation, Periana and co-workers proposed that the formation of the O−CH 3 bond could take place through the nucleophilic attack of a free HSO 4 − anion on an intermediate Au(III) methyl complex. 90 Later, Bercaw and co-workers discarded an S N 2-type pathway for the reductive elimination of MeI from [Au(Me)(I) 2 (IPr)]. 50 The lowest unoccupied molecular orbital (LUMO) isosurface of Int1 + (Figure 6) provides additional support for the proposed mechanism. It is a σ* orbital mostly distributed along the Au−C bond, with large external lobes, implying that both the C and Au atoms can display Lewis-acidic behavior. The electrophilicity of the Me group in this species is also supported by the above-discussed nucleophilic attacks of PPh 3 or NEt 3 on this group. Mechanistic studies indicate that the reductive eliminations of ethane or methyl bromide from [Au(CF 3 )(Me)(X)(PR 3 )] (X = Me, Br) take place from a tricoordinate intermediate formed by phosphine dissociation. In contrast, in the cases of the analogous complexes with X = OTf or ClO 4 , and the cationic complex cis-[Au(CF 3 )(Me)(PPh 3 ) 2 ]OTf, experimental evidence and computational modeling are consistent with a reaction pathway involving X − or PPh 3 dissociation (respectively) followed by nucleophilic attack to the gold-bound carbon. This S N 2 reductive elimination mechanism has not been documented before in Au(III) complexes and should be considered as a potential pathway in Au(I)/Au(III)-catalyzed reactions, in particular, in those leading to C(sp 3 )−heteroatom coupling products.