In this paper we study the gas-phase hydrogen abstraction reaction between fluorine atoms and silane in a three-step process: potential energy surface, kinetics and dynamics. Firstly, we developed for the first time an analytical full-dimensional surface, named PES-2021, using high-level explicitly-correlated ab initio data as the input. PES-2021 represents a continuous and smooth potential with analytical gradients and includes intuitive concepts (stretching and bending nuclear motions). Based on the PES-2021 quasi-classical trajectory (QCT) calculations were performed to analyse the kinetics and dynamics. Secondly, in the kinetics study at room temperature we observed a very fast reaction with a rate constant of 3.90 × 10-10 cm3 molecule-1 s-1, reproducing the scarce experimental evidence. Finally, the third step is the dynamics study, which was performed under two different conditions, a temperature of 77 K and a collision energy of 2.5 kcal mol-1, for direct comparison with experiments. In the first case, we found the largest fraction, 44%, deposited as HF(v) vibration, where the most populated states were HF(v = 2, 3), both results reproducing the experimental evidence. The largest discrepancy with the experiment was found in the HF(j) rotational distribution, where hotter distributions were found, this discrepancy being associated with limitations of the QCT method. The second case, E = 2.5 kcal mol-1, was a state-to-state correlated study and, therefore, more difficult. The theory overestimates (again) and consequently underestimates, respectively, the rotational and vibrational fractions of the HF(v,j) product as compared with experiments. While experimentally the SiH3 product appears excited only in the umbrella mode, ν2 = 0-5, correlated with the HF(v) co-product vibrational excited in v = 3 and 4, theoretically a wider vibrational distribution is found in both products, and these distributions have, obviously, an influence on the product correlated speed distributions. However, the product correlated angular distribution is well reproduced. In general, these results allowed us to test the capacity of PES-2021 + QCT tools to simulate the experimental evidence, revealing that agreement is better when average properties are compared, making the comparison worse when state-to-state properties are compared. Different causes of the theory/experiment discrepancies were analysed, and it was found that they are due, mainly, to limitations of the QCT method.
Read full abstract