Ruthenium(II)-Catalyzed Homocoupling of α-Carbonyl Sulfoxonium Ylides Under Mild Conditions: Methodology Development and Mechanistic DFT Study

A mild ruthenium(II)-catalyzed homocoupling of α-carbonyl sulfoxonium ylides was developed and the detailed mechanism was understood based on DFT calculations in the current report. The catalytic system utilizes the α-carbonyl sulfoxonium ylide as both the directing group for ortho-sp2 C-H activation and the acylmethylating reagent for C-C coupling. Various substituents are compatible in the transformation and a variety of isocoumarin derivatives were synthesized at room temperature without any protection. The theoretical results disclosed that the full catalytic cycle contains eight elementary steps, and in all the cationic Ru(II) monomer is involved as the catalytic active species. The acid additive is responsible for protonation of the ylide carbon prior to the intramolecular nucleophilic addition and C-C bond cleavage. Interestingly, the intermediacy of free acylmethylation intermediate or its enol isomer is not necessary for the transformation.


INTRODUCTION
Under transition metal catalysis, the sulfoxonium ylides have found wide applications in synthetic chemistry (Li et al., 1997). These species could be used as efficient carbene precursors by elimination of dimethyl sulfoxide (DMSO) by activation of the ylide C-S bond with metal (Bayer and Vaitla, 2018;Cheng et al., 2018). This strategy has recently found important applications in transition metal-catalyzed C-H activation reactions (Scheme 1) (Gulias and Mascarenas, 2016;Wang et al., 2016;Sambiagio et al., 2018), as sulfoxonium ylides possess the advantages of easy availability of starting materials and safe operation in reactions compared with the alternative approach with diazo precursors (Davies and Manning, 2008;Xia et al., 2017;Clare et al., 2019;Wen et al., 2019;Zhou et al., 2020). In this context, since the initial independent reports by the Li (Xu et al., 2017a) and Aïssa (Barday et al., 2017) groups, interesting acylmethylation methods with sulfoxonium ylides as the acylmethylating reagents have been developed under the catalysis of rhodium (You et al., 2018;Xu et al., 2019;Yu J. et al., 2019;Tian et al., 2020), ruthenium (Karishma et al., 2019;Li H. et al., 2019;Fu et al., 2020), and other transition metals (Ji et al., 2018;. Notably, tandem intramolecular annulations of the in-situ generated acylmethylation products GRAPHICAL ABSTRACT| Ru(II)-catalyzed homocoupling of alpha-carbonyl sulfoxonium ylides.
In most cases the sulfoxonium ylide functioned as a traceless bifunctional directing group, which were removed in terms of DMSO elimination during the course of annulation with alkynes (Xu et al., 2017b;Hanchate et al., 2019;, anthranils , allenoates (Lou et al., 2019), and alkenes (Kommagalla et al., 2020). However, when using oxa/azabicyclic olefins as coupling partners, chemo-divergent couplings were achieved by the Li group , and the sulfoxonium ylide moiety was retained in the C-H alkylation product that controlled by the introduction of PivOH. The retention of the sulfoxonium ylide was also found in a recent work by Fan and coworkers , in which the naphthalenone derivatives were synthesized from Rh(III)-catalyzed cascade reactions of sulfoxonium ylides with α-diazocarbonyl compounds.

RESULTS AND DISCUSSION
We initiated the investigation by optimizing the reaction conditions for the homocoupling of α-carbonyl sulfoxonium ylide 1a to form isocoumarin 2a ( Table 1) under Ru(II) catalysis. It was found that 36% of the NMR yield of 2a could be obtained when the reaction was catalyzed by 5 mol% of [RuCl 2 (p-cymene)] 2 with 0.2 equivalent AgOAc and 1 equivalent KOAc in trifluroethanol solution under air atmosphere at room temperature (entry 1). No reaction was observed if the catalyst was changed to RuCl 2 (PPh 3 ) 3 , [Cp * Rh(CH 3 CN) 3 SbF 6 ] 2 , or [Cp * RhCl 2 ] 2 (entry 2). Similar or worse yields resulted if the AgOAc is replaced by other silver salts (entries 3-6).
Among different solvents screened with AgSbF 6 as the silver additive (entries 7-11), dichloroethane was found to be the most effective to afford a 52% yield in 2a. Based on this result, we changed the KOAc additive to K 2 CO 3 and Cs 2 CO 3 but no positive result was obtained (entries 12-13). However, improvement of the yield to 65% could be achieved when using 1 equivalent NaOAc instead of KOAc (entry 14), and a similar yield was obtained if the amount of NaOAc was reduced to 0.5 equivalent (entry 15). A better reaction was found by adding 0.5 equivalent pivalic acid to the system (entry 16), and an 82% NMR yield of 2a was obtained by increasing the AgSbF 6 to 0.5 equivalent (entry 17). However, the yield would decrease if the AgSbF 6 was increased to 1 equivalent (entry 18). Control experiments showed that both the Ru(II) and silver salt are essential to the homocoupling (entries 19-20), and the efficiency of the reaction would be dramatically reduced in the absence of NaOAc (entry 21).
With the optimal conditions in hand, the scope of this ruthenium(II)-catalyzed homocoupling protocol with respect to different α-carbonyl sulfoxonium ylide derivatives was investigated ( Table 2). While the 2a was isolated in a 76% yield in reaction of 1a, the substitution of electron-donating methyl, ethyl, and t-butyl groups at the para position of the benzene ring was found to have positive effects on the efficiency, affording the corresponding isocoumarins 2b-d in good yields. However, other substrates with other electron donating groups, including chloromethyl, methoxyl, phenyl, trifluoromethoxyl, and trifluoromethylthio, resulted in slightly lower yields of products 2e-2i. The meta methyl group in 1j does not have notable influence on the formation of 2j, however, substrate 1k, having an ortho methoxyl group, delivered the 2k in moderate yield, probably due to the steric effect of the substituent in this case. When both meta positions of 1l are substituted by methyl groups, a 58% yield of 2l was isolated.
The effects of electron-withdrawing group on the reactions were also investigated. When the α-carbonyl sulfoxonium ylides were substituted by ester, trifluoromethyl, or nitro group at the para position, the desired products were obtained in 48-76% of yields (2m-2o). Various halides could be tolerated in the reactions, delivering the products in moderate to good yields (2p-2x). While the meta-fluoro-substituted precursor 1w underwent para C-H activation selectively, interestingly, poor selectivity was observed in the reaction with the chloro-containing analog 1x, forming an 82% yield of isolable products 2x-p and 2x-o in 5:4 ratio. The toleration of halogens could be useful for further functionalization of the products. In addition, ylide 1y containing the naphthalene ring was also compatible, affording a 70% yield of the 2y.
To show the synthetic application of the catalytic homocoupling, a gram-scale synthesis of 2a was performed, and a high yield was achieved with a reduced loading of the Ru(II) catalyst (Scheme 2A). The cross-coupling between aromatic and alkyl α-carbonyl sulfoxonium ylides was tested by the reaction of an equimolar mixture of 1n and 1n' under standard conditions (Scheme 2B), which resulted in 2n and 2n' in a 1.4:1 ratio, indicating that introducing an alkyl group at C4 of the isocoumarin is possible (more examples are given in the SI). To probe the reaction mechanism, a deuterium labeling experiment was carried out with 1a in presence of 2 equiv of CD 3 OD (Scheme 2C). After 4 h, a 49% yield of 2a-D was isolated, in which deuterium incorporation only occurred at C4, but no deuterium incorporation was observed in the recovered 1a. This indicated that the C-H activation step should be irreversible under the current conditions.
To better understand the experimental results, DFT calculations were carried out to highlight the details of the transformation (Figures 1-3) (Hou et al., 2017;Jiang J. et al., 2019;Shu et al., 2019). According to the theoretical results, the reactant complex IM1 formed exergonically from substrate 1a and cationic monomeric LRu(OAc) + (L = p-cymene), which was produced in the catalytic system of [RuCl 2 (p-cymene)] 2 , AgSbF 6 , and NaOAc (Figure 1) (Xie et al., 2018a;Xie H. et al., 2018). Calculations found that if the neutral complex LRu(OAc) 2 was used, one anionic ligand should be dissociated from the Ru(II) to form a stable reactant complex, indicating a generation of cationic species is more favorable. From IM1, the ortho-C-H cleavage directed by the carbonyl functionality occurs via the CMD process (TS1) with an activation barrier of 20.4 kcal/mol and leads to metallated intermediate IM2 endergonically. IM3 is formed by releasing HOAc prior to the incorporation of another 1a to form complex IM4 through interaction between the ylide carbon and the Ru(II). From the latter intermediate, the C-S cleavage via TS2 becomes facile with a small barrier of 10.6 kcal/mol. This step forms Ru-carbene intermediate IM5 slightly exergonically and eliminates DMSO concurrently. The migratory insertion of the carbene moiety into the Ru-C bond requires a barrier of 17.7 kcal/mol via TS3. The profile in Figure 1 disclosed that TS2 and TS3 are much lower in energy than TS1 and the formation of the six-membered ruthenocycle IM6 is highly exergonic, suggesting that the C-H activation step is irreversible and is consistent with the deuterium-labeling experiment.
of an enol intermediate was also studied. The η 3 oxallyl complex IM9 and O-bound enolate complex IM10 are 1.4 and 11.8 kcal/mol higher in energy than C-bound enolate complex IM6, respectively. The high energy of IM10 is probably due to the strong interaction between the Ru(II) and the phenyl group, which leads to a puckered structure and dearomatization of the phenyl ring. The complexation of HOAc with IM10 forms IM11 by H-bonding, the proton transfer from HOAc to the enolate oxygen is very facile with a barrier of 1.0 kcal/mol via TS5 and generates the complex IM12 slightly endergonically. The free enol intermediate IM13, 11.1 kcal/mol higher in energy than IM6, could be released by the incorporation of another 1a, from which the reactant complex IM1 is regenerated. Tautomerism between IM13 and IM8 could be possible via an intramolecular process involving HOAc as the proton shuttle as shown in TS6 with an activation barrier of 20.6 kcal/mol, while tautomerism by intramolecular 1,3-H shift requires a much higher barrier of 48.9 kcal/mol from IM13 (See SI for more details). However, the energy of TS6 is 31.7 kcal/mol above that of the global minimum IM6 1 . It was supposed that the protonation of the α carbon of IM11 could be another possible pathway to complex IM7. This could be realized via TS7, but a 1 It should be noted that the enol IM13 could be formed in the system as it is in equilibrium with IM6 according to the relative low energies for related intermediates and TS in Figure 2, and in experiments the enol-keto tautomerism is very comment. However, further transformations from free IM8 and IM13 are less favorable as compared with the pathway in Figure 3. Details are given in the Supplementary Material. relatively high activation barrier of 28.2 kcal/mol is still required from IM6.
While the above results indicated that the generation of acylmethylation intermediate IM8 should be difficult under current conditions 1 , we found the pathway initiated by protonation of the anionic ylide carbon in IM6 by HOAc is the most energetically favorable (Figure 3). Accordingly, the barrier for the protonation via TS8 is 23.0 kcal/mol, leading endergonically to IM14 in which the acetate is associated with both the carbonyl carbon and the Ru atom. IM14 undergoes a very facile C-O dissociation via TS9 to form IM15, from which the intramolecular nucleophilic addition via TS10 requires a small barrier of 6.9 kcal/mol and forms the C-O bond of the 6membered heterocycle in intermediate IM16. In the following step, C-C bond cleavage occurs via TS11 with a barrier of 13.5 kcal/mol, this generates product complex IM17 and eliminates dimethylsulfoxonium methylide (DSM) concurrently. In the last step the formation of product 2a and regeneration of reactant complex IM1 could be realized by a ligand exchange reaction of IM17 with 1a. Thus, the protonation of the ylide carbon by HOAc via TS8 is the most difficult step in the whole reaction. This explains why the acid additive is required for promoting the reaction.
Based on the above results, the full catalytic cycle for the transformation contains eight elementary steps as shown in Figure 4. Upon the formation of cationic reactant complex A, the first step is the acetate-assisted C-H activation to form a five-membered ruthenacycle B. The incorporation of another 1a by ylide coordination generates σ-complex C, which undergoes

DMSO elimination to form carbene intermediate D.
From this, C-C bond formation by migratory insertion generates a sixmembered ruthenocycle E. Then, protonation of the ylide carbon by HOAc leads to intermediate F. The following step is an intramolecular nucleophilic addition which creates G, from which the DSM elimination by C-C cleavage occurs to deliver product complex H. In the last step, releasing isocoumarin product 2a and regenerating complex A is completed by a ligand exchange.

CONCLUSION
In conclusion, we have established a mild ruthenium(II)catalyzed homocoupling of α-carbonyl sulfoxonium ylides and carried out a detailed mechanistic investigation using DFT calculations. The methodology enables the efficient synthesis of a variety of isocoumarin derivatives under air conditions at room temperature. Theoretical results uncovered that the Ru(II) catalyst is involved in all steps of C-H activation, C-C coupling, C-O formation, and C-C cleavage, and the intermediacy of free acylmethylation intermediate or its enol isomer was not necessary for the intramolecular nucleophilic cyclization process. The mechanistic information could have implications for better understanding related tandem reactions in other catalytic systems.

Computational Details
All DFT calculations were carried out with the Gaussian 09 suite of computational programs (Frisch et al., 2013). The geometries of all stationary points were optimized using the B3LYP hybrid functional (Lee et al., 1988;Becke, 1993a,b) at the basis set level of 6-31G(d) for all atoms except for Ru, which was described by the relativistic effective core potential basis set of Lanl2dz. Frequencies were analytically computed at the same level of theory to obtain the free energies and to confirm whether the structures were minima (no imaginary frequency) or transition states (only one imaginary frequency). The solvent effect of toluene was evaluated by using the SMD polarizable continuum model by carrying out single point calculations at the M06/6-311+G(d,p) (SDD fur Ru) level (Zhao and Truhlar, 2008a,b). All transition state structures were confirmed to connect the proposed reactants and products by intrinsic reaction coordinate (IRC) calculations. All the energies given in the text are relative free energies corrected with solvation effects.

MATERIALS AND METHODS
Commercially available materials were used as received, unless otherwise noted. 1 H NMR and 13 C NMR spectra were measured on a Bruker-400 or Bruker-500 instrument, using CDCl 3 as the solvent with tetramethylsilane (TMS) as an internal standard at room temperature. Chemical shifts are given in δ relative to TMS, the coupling constants J are given in Hz. Melting points were measured on an X4 melting point apparatus and uncorrected. HRMS analysis was measured on a Bruker micrOTOF-Q II instrument (ESI) or a Waters GCT Premier instrument (EI-TOF).

Typical Procedure for the Synthesis of α-Carbonyl Sulfoxonium Ylides 1
Under N 2 , trimethylsulfur iodide (3.3 g, 15 mmol, 3 equiv) was suspended in dry THF (25 mL) in a flame-dried 100 mL round bottom flask that was protected from light with aluminum foil. Potassium tert-butoxide (2.24 g, 20 mmol, 4 equiv) was added and the mixture was stirred at reflux for 2 h. After cooling to room temperature, benzoyl chloride (5 mmol, 1 equiv) in THF (10 mL) was added. The mixture was stirred at reflux for another hour and then filtered at room temperature through a plug of celite before all volatiles were removed under vacuum. Purification by flash chromatography afforded sulfur ylide 1a.