P,N-Chelated Gold(III) Complexes: Structure and Reactivity

Gold(III) complexes are versatile catalysts offering a growing number of new synthetic transformations. Our current understanding of the mechanism of homogeneous gold(III) catalysis is, however, limited, with that of phosphorus-containing complexes being hitherto underexplored. The ease of phosphorus oxidation by gold(III) has so far hindered the use of phosphorus ligands in the context of gold(III) catalysis. We present a method for the generation of P,N-chelated gold(III) complexes that circumvents ligand oxidation and offers full counterion control, avoiding the unwanted formation of AuCl4–. On the basis of NMR spectroscopic, X-ray crystallographic, and density functional theory analyses, we assess the mechanism of formation of the active catalyst and of gold(III)-mediated styrene cyclopropanation with propargyl ester and intramolecular alkoxycyclization of 1,6-enyne. P,N-chelated gold(III) complexes are demonstrated to be straightforward to generate and be catalytically active in synthetically useful transformations of complex molecules.


General Procedure for Testing of Catalytic Activity.
The catalytic activity of the different Au(I) and Au(III) complexes described in Section 1.2-1.3 were evaluated in the two model reaction, cyclopropanation, and alkoxycyclization of a 1,6-enyne, as described below.
Alkoxycyclisation of a 1,6-enyne: Dimethyl(E)-2-(prop-2-yn-1-yl)-2-styrylmalonate (10 mg, 0.037 mmol) was dissolved in methanol (0.6 mL) and added to the gold(III) catalyst (5 mol-%). The reaction progress was monitored by 1 H NMR or TLC. The crude product was dried and purified by silica column chromatography (n-pentane:ethylacetate 10:1, R f = 0.21). Reactivity data and yield for the different Au-catalyst are presented in the main text, Cyclopropantion of styrene with propargylester: The propargyl ester (10 mg, 1 equiv.) and styrene (4 equiv.) were dissolved in d-DCM (0.6 mL) and added the gold-catalyst (5 mol-%) dissolved in d-DCM. The reaction progress was monitored by 1 H NMR or TLC, while the cis:trans ratios were determined by 1 H NMR. The crude product was purified by silica column chromatography. Reactivity data, yield and diastereoselectivity for the different Au-catalyst are presented in the main text,

Crystallographic data
Single crystals of (   NOTE: The Au atoms are slightly pulled towards each other but the distance of ca. 3.5 Å is fairly long for aurophilic interactions, i.e. only weak interactions might be responsible for that.  NOTE: Notable trans effect on the Cl atoms: 2.2619(12) and 2.3408 (12) trans to N and P, respectively.  NOTE: Same trans effect as above. 2.2608(18) and 2.3321 (19), trans to N and P. S13

Computations
All calculations were carried out with the Gaussian 16 program, revision C01, 12 employing the B3LYP exchange and correlation functional [13][14][15][16] in conjunction with the D3 empirical dispersion correction with Becke-Johnson damping. 17 Solvent effects were included using the SMD implicit solvent model, with dichloromethane as solvent (ε=8.93), 18 while in case of charged complexes, the counterion was omitted from the calculation and the complex of interest was computed with either a single (monomer) or a double (dimer) positive charge. The integration grid consisted in a pruned (99,590) grid, which provides consistent results irrespective of the molecular orientation and is the default in Gaussian 16.
For all geometry optimizations, the cc-pvdz basis set 19 was used on all atoms but gold, for which the MDF60 Stuttgart/Cologne effective core potential in conjunction with the cc-pvdz-pp basis was used instead. [20][21] In all cases, the optimized geometries were checked with calculation of the molecular hessian to ensure the nature of the structure, be it equilibrium or transition state. The solvent effects were included in these optimizations unless otherwise stated. Thermochemical corrections to electronic energies in order to obtain Gibbs free energies were computed at this level of theory using the quasi harmonic approximation, 22 thereby setting all low vibrational modes (<100cm -1 ) to the threshold value of 100cm -1 . These values were obtained using the GoodVibes program. 23 The single point electronic energies were obtained at a higher level of theory, namely using the larger cc-pvtz basis set [19][20][21] instead of the cc-pvdz one. Therefore, all Gibbs free energies reported in the main manuscript are obtained by the sum of the electronic energy at triple-zeta level, computed on the structure optimized with the double-zeta level basis and with thermochemical corrections from the double-zeta calculation. An additional set of single point calculations for the free energy profile reported in Scheme 2 of the main manuscript was carried out using the M06 exchange and correlation functional 24 in conjunction with the D3 empirical dispersion correction without Becke-Johnson damping. 25 The M06 functional is known to outperform other functionals in gold chemistry, 26 and thus represents a good test to support the results obtained with B3LYP. The free energy profile obtained with the M06 functional is reported in Figure SX. As can be seen, the qualitative features of the profile are the same as in Scheme 2 of the main manuscript, where TS1 still remains the rate-determining step of the reaction. In general, the effect of using M06 over B3LYP is to decrease the energy of all intermediates and transition states by 2 to 3.5 kcal/mol. All NMR chemical shifts reported in the main manuscript were obtained using the Gauge-Independent Atomic Orbital (GIAO) NMR method 27 with the cc-pvtz basis set and calculated according to the formula where σ ref is the isotropic magnetic shielding constant of a reference compound (nitromethane for 15 N NMR and phosphoric acid for 31 P NMR) and σ that of the sample. Note that the reference compounds were calculated at the same level of theory as the samples. In order to ensure that the choice of basis set is adequate, we report a comparison in Table S6 of the NMR coordination shifts obtained for a 1-Au(I) with respect to ligand 1 using three difference basis sets: aug-cc-pvdz, 28 cc-pvtz, aug-cc-pvtz. 28 In the case of three systems, ([2-4-Au(I)]SbF 6 ) 2 , no equilibrium structure was found in the geometry optimization when using the solvent model. Thus, the NMR chemical shifts reported in table 1 of the main manuscript for those three complexes are computed on the structures optimized without the effect of the solvent. In order to ensure that the obtained structures are of good quality, we report in Table S7 a comparison of the coordination shifts obtained for ligand 1, with structures optimized with and without solvent effects. Note that the solvent model was taken into account when computing the NMR chemical shifts, it was only during the geometry optimization that it was omitted.

50.52
As we see, the differences are minimal, in particular for the dimer.
For the [2-Au(III)]SbF 6 , one imaginary frequency of -3.8 cm -1 was still present upon several attempts to obtain the true minimum, however this is ascribed to the solvent effects, since optimization without SMD yields a very similar geometry with all frequencies positive. The NMR value reported in table 1 of the main manuscript refers to the optimized structure with SMD.
As it is explained in the main manuscript, 1-Au(I) as well as [1-Au(I)]SbF 6 have three (quasi-) isoenergetic isomers each. In Table S8 we report the NMR coordination shifts for all the isomers, whereas in table 1 of the main manuscript we only report a single, indicative value (highlighted in boldface in Table S8). For 2-4-Au(I), only the isomer with the gold atom above the oxazoline plane was considered (cf. Figure 3a of main text), and consequently, for [2-3-Au(I)]SbF 6 only the dimer formed with this isomer was calculated (cf. Figure 3d of main text). In contrast, for [4-Au(I)]SbF 6 , given the availability of the X-Ray crystal structure, the isomer formed by one monomer having the gold atom above the oxazoline and one monomer having it below the oxazoline, was calculated. The value reported in table 1 of the main text is the average between the two different NMR signals, however this difference is small, 2.7 ppm. Finally, all structures obtained for the reaction mechanism from ([1-Au(III)]SbF 6 ) 2 to [1-Au(III)]SbF 6 (Scheme 2 in the main text), the solvent effect was omitted during the geometry optimization and it was included in the single point calculation at the triple-zeta level.
All computed structures obtained in this work are available as xyz files as part of the Supporting Information. In table S9 we report the energy contributions, the number and values of imaginary frequencies for all structures involved in the mechanism reported in Scheme 2 of the main manuscript and Figure SX. In Table S10, we report the energy contributions for the different isomers of 1-Au(I) as well as [1-Au(I)]SbF 6 (no imaginary frequencies are reported because all structures correspond to true minima with only real and positive frequencies). In Table S11, we report the energy contributions of all other structures used to rationalize the mechanism proposed in Scheme 2 as well as the mechanism involving acetonitrile. Table S11. Energy contributions (in Hartree), number and value of imaginary frequencies for the remaining structures used to rationalize the mechanism depicted in Scheme 2 of the main manuscript and that involving acetonitrile. The first two columns refer to the electronic energy obtained at different levels of theory, vdz stands for cc-pvdz, while vtz for ccpvtz. These energies include the contribution of dispersion and, for the triple zeta case, the contribution of solvation. ZPVE is the zero-point vibrational energy, H the enthalpy and G the Gibbs free energy. Note that the names used in tables S9-S11 reflect the names given to the xyz files available as Supporting Information.     Ph OAc O Figure S44. The 1 H NMR spectrum of 9-trans acquired at 25 o C in CD 2 Cl 2 at 600 MHz.