Dioxygen insertion into the gold(i)–hydride bond: spin orbit coupling effects in the spotlight for oxidative addition

O2 insertion into a Au(i)–H bond occurs through an oxidative addition/recombination mechanism, showing peculiar differences with respect to Pd(ii)–H, for which O2 insertion takes place through a hydrogen abstraction mechanism in the triplet potential energy surface with a pure spin transition state. We demonstrate that the spin-forbidden Au(i)–hydride O2 insertion reaction can only be described accurately by inclusion of spin orbit coupling (SOC) effects. We further find that a new mechanism involving two O2 molecules is also feasible, and this result, together with the unexpectedly high experimental entropic activation parameter, suggests the possibility that a third species could be involved in the rate determining step of the reaction. Finally, we show that the O2 oxidative addition into a Au(i)–alkyl (CH3) bond also occurs but the following recombination process using O2 is unfeasible and the metastable intermediate Au(iii) species will revert to reactants, thus accounting for the experimental inertness of Au–alkyl complexes toward oxygen, as frequently observed in catalytic applications. We believe that this study can pave the way for further theoretical and experimental investigations in the field of Au(i)/Au(iii) oxidation reactions, including ligand, additive and solvent effects.


1.Hydrogen abstraction mechanism.
For investigating the hydrogen abstraction mechanism, we performed two potential energy surface scans, one for the triplet and one for the closed shell singlet spin state, using the oxygenhydrogen (hydride) distance as the reaction coordinate, and optimizing all the other degrees of freedom.
The energy profiles is shown in Fig. S1 (top): the separated reactants, corresponding to a very large O-H distance, interact forming a weakly bound van der Waals complex with an O-H distance of 2.80 Å. Upon closer approach of oxygen to the metal hydride, on the triplet PES, the energy increases constantly, reaching a value of about 23 kcal mol -1 for an O-H distance of 1.00 Å. At this point the H is almost abstracted from gold, since the Au-H distance is 2.20 Å compared to 1.62 Å in LAuH. The energy of the isolated LAu and OOH radicals is about 16 kcal mol -1 higher, suggesting that at this point on the reaction path an interaction between the LAu and OOH fragments is still present. The singlet profile starts at an energy about 20 kcal mol -1 higher than that of the isolated reactants (the splitting between the 3 Σ g and 1 Δ states of O 2 is +27.6 kcal mol -1 at this level of theory, suggesting that at 2.8 Å an interaction between O 2 and LAuH occurs), and, as oxygen approaches hydrogen, the energy slightly increases until an O-H distance of about 1.50 Å. The energy then slightly decreases till 1.35 Å; from this point a sharp lowering of the energy takes place, leading directly to the product, after significant rearrangement of the relative position of the O 2 fragment and the gold complex. The one-dimensional relaxed scans displayed in Fig. S1 (top) show that at an O-H distance of roughly 1.35 Å the triplet and singlet PES have the same energy, but this is only a lower bound to the energy of the MECP, because the structure of the singlet and triplet complexes, except for the identical O-H distance, are significantly different (Fig. S1 bottom). In Fig. S1 (top) we put also the MECP and TS SOC for comparison (as green and red point respectively). Fig. S1: top) PESs scan (triplet and closed shell singlet spin) using the O-H distance (in Å) as reaction coordinate; MECP and TS SOC are also shown as a green and red point respectively; bottom) geometry of the complex at the O-H distance of 1.35 Å for the singlet (left) and triplet (right) PESs.

O 2 metal coordination mechanism.
In Fig. S2 the plot of the PESs scan (triplet and closed shell singlet) using the Au-O distance (in Å) as reaction coordinate for the O 2 metal coordination mechanism is shown. In the Figure the structure of the complex at Au-O = 1.8 Å within the singlet PES is reported. This geometry closely resembles the product of an oxidative addition, and therefore it does not represent the intermediate for the O 2 metal coordination pathway.

3.Oxidative addition mechanism.
In Fig. S3 the MECP for the oxidative addition using the experimental carbene is shown. The energy of this MECP is +17.9 kcal mol -1 with respect to the isolated reactants, very close to the energy of the MECP using the computational model (+18.2 kcal mol -1 ). Also the geometries are very similar. In Fig. S4 the MECP for the oxidative addition with LAuCH 3 and O 2 as reactants, the product of the oxidative addition (singlet) and the structure with maximum energy in the triplet PES scan for the rearrangement process with O 2 are shown. The oxidative addition is feasible, with a ΔE ≠ of +19.3 kcal mol -1 (18.2 kcal mol -1 for the MECP with LAuH), and is endothermic with a ΔE of +4.9 kcal mol -1 (2.7 kcal mol -1 for the MECP with LAuH), while the calculation of the maximum energy point of the rearrangement process gives a ΔE of +34.6 kcal mol -1 .

4.Reliability of the theoretical method.
In order to assess the reliability of the level of theory used in this work, we performed additional SOC calculations, aimed at first testing the basis set effects on the geometries and the energies (absolute values of LAuH, O 2 , transition states and activation barriers) of the three analyzed mechanisms. We then evaluated the D3-BJ dispersion effects on the energies. We further explored the effects of a few most popular XC functionals on the energies, by comparing 3 GGA (BP86, PBE and BLYP), one meta-GGA (TPSS), one hybrid (B3LYP), and one meta-hybrid (M06) functionals for the three mechanisms studied. Finally, we estimated the solvent effects (benzene) by performing implicit solvation calculations using COSMO model. The results of this study, while assessing the possible error bars one would get by changing any of basis set, exchangecorrelation functional, D3-BJ dispersion correction and solvation (benzene) items, both substantiate the reliability of our DFT SOC (BP86-D3-BJ/DZP) level of theory to describe the system for the purpose of this work and confirm the chemical insight we obtain, namely that the oxidative addition is preferred over the H-abstraction mechanism.

Basis set effects.
We performed geometry optimizations using the same computational details as specified in the text (SOC BP86-D3-BJ), but increasing the basis set from DZP to TZP for all atoms. The geometries of the reactants and the transition states (TS SOC) for the three mechanisms, optimized using a DZP and a TZP basis set, are compared in Fig. S5. and transition states (TS SOC) geometries optimized with a DZP and a TZP basis set for the three mechanisms (Hydr. Abstr. = hydrogen abstraction, mechanism a); Oxid. Addit. = oxidative addition, mechanism c); 2 O 2 molecules = mechanism c) with the additional O 2 molecule) . Figure S5 clearly demonstrates that the basis set quality does not affect the geometries of the species. For the purposes of our paper, the DZP basis set for all atoms (TZP for gold) guarantees a high degree of reliability for geometry optimization calculations.
The energies of all the species optimized with TZP and the single point energies using a TZP and a TZ2P basis set on the DZP optimized structures are compared with the corresponding energies of the species optimized with DZP in Table S1, where activation (electronic) energy barriers are also reported for the 3 considered mechanisms. From Table S1 we observe that basis set quality does have an effect on the activation energy barriers. Comparison of the third and fourth sets of data, reporting results of single point energy calculations performed at TZP and at T2ZP level of theory on the geometry optimized at DZP, respectively, with the first set of data shows that the activation energy barriers for the 3 mechanisms increase from DZP to TZP (from +1.4 to +3.3 kcal/mol) but are almost converged from TZP to TZ2P (from +1.4 to +1.6 kcal/mol), with the activation energy barrier for mechanism c) + O 2 being the most sensitive to basis set quality. Interestingly, single point energy calculations performed with TZP basis set on DZP optimized geometries gives activation energy barrier values almost identical to those obtained from TZP optimized geometries.
Importantly, the same activation energy barrier trend, namely ΔE ≠ a) > ΔE ≠ c) > ΔE ≠ c) + O 2 is calculated with all the considered basis sets.

D3-BJ dispersion effects.
The D3-BJ dispersion effects are evaluated by performing single point energy calculations on reactants and TS SOC DZP-optimized structures, using the TZ2P basis set and the BP86 functional with (BP86-D3-BJ) and without dispersion corrections (BP86). Results are compared in Table S2 where activation (electronic) energy barriers for the 3 mechanisms are also shown. We see that Inclusion of D3-BJ dispersion effects accounts for a non negligible lowering of the activation energy barriers which decrease by 1.5 kcal/mol for the hydrogen abstraction (mechanism a)), by 2.5 kcal/mol for the oxidative addition (mechanism c)) and by 4.0 kcal/mol for the oxidative addition involving two O 2 molecules. Importantly, the same activation energy barrier trend, namely ΔE ≠ a) > ΔE ≠ c) > ΔE ≠ c) + O 2 is calculated also without inclusion of D3-BJ dispersion effects.

XC commonly used functionals (PBE, BLYP, TPSS, B3LYP, M06) effects.
The XC functional effects have been investigated by performing single point energy calculations with a TZ2P basis set and SOC, without inclusion of D3-BJ dispersion effects, on the BP86-D3-BJ DZP optimized reactants and transition state structures for the 3 mechanisms. We selected 5 different popular XC as representative of 2 GGA (PBE and BLYP), 1 meta-GGA (TPSS), 1 hybrid (B3LYP) and 1 meta-hybrid (M06) functional. The results are shown in Table S3, including activation (electronic) energy barriers, and compared with the BP86 outcomes. We see that within the GGA functionals set the activation energy barriers for the 3 mechanisms are only slightly affected by the functional-type: with respect to BP86 calculated activation barriers, PBE lowers the 3 values in the range from -1.2 to -3.0 kcal/mol, and BLYP slightly lowers the value for mechanism a) (-0.8 kcal/mol) and increases the values for mechanisms c) and c) + O 2 (+1.1 and +0.4 kcal/mol, respectively). The meta-GGA TPSS functional increases the energy barriers for mechanisms a) and c) + O 2 (+3.6 and +2.1 kcal/mol, respectively) and slightly decreases the barrier for mechanism c) (-0.3 kcal/mol). Both the hybrid B3LYP and meta-hybrid M06 functionals significantly increase all the three energy barriers, with B3LYP in the range from +9.3 to +18.4 kcal/mol, and M06 in the range from +7.8 to +10.0 kcal/mol. Inclusion of D3-BJ dispersion on B3LYP calculations decreases all of the activation energy barriers of the same amount as that calculated for BP86 (from -1.4 to -4.0 kcal/mol).

Solvation (benzene) effects.
We studied solvent effects on this reaction by performing SOC single point energy calculations with BP86 (without inclusion of D3-BJ dispersion) and with a TZ2P basis set on the DZP optimized reactants and transition state structures for the 3 mechanisms, using COSMO model (solvent = benzene, as in experiment). The results, reported in Table S4, where are compared to the gas-phase ones, show that inclusion of solvent effects decreases significantly the activation barrier only for the hydrogen abstraction mechanism by 4.9 kcal/mol, whereas the activation energy barriers for the oxidative addition mechanisms (with one and two dioxygen molecules) are only marginally affected (a decrease of 1.5 kcal/mol and an increase of 0.1 kcal/mol are calculated for the two paths, respectively). Importantly, again the same activation energy barrier trend, namely ΔE ≠ a) > ΔE ≠ c) > ΔE ≠ c) + O 2 , is calculated also with inclusion of solvent (benzene) effects.