On the automation of VRC-TST simulations: strategies to determine wave function guesses, exploration of black box methodologies, and application to test systems Luigi Crisci 1,2 , Andrea Della Libera 3 , Carlo Cavallotti 3 , Nadia Rega 2,4,5 and Vincenzo Barone 1,6 1 Scuola Normale Superiore, Italy; 2 Dipartimento di Scienze Chimiche, Universitá di Napoli Federico II, Italy; 3 Dipartimento di Chimica, Italy; 4 Scuola superiore Meridionale, Italy; 5 Centro di Ricerca Interdipartimentale sui Biomateriali (CRIB), Italy; 6 Istituto Nazionale di Fisica Nucleare (INFN) sezione di Pisa, Italy Theoretical research on the reactivity of singlet potential energy surfaces (PES) is critical for the entire gas- phase kinetics community. It is useful in combustion to forecast fragmentation patterns of molecules present in biomass components, as well as in atmospheric chemistry and astrochemistry to examine reaction pathways that can be accessible following the recombination of two radicals. Its correct description, however, is a difficult undertaking since the quickest exit (or entry) channels on singlet PESs are often barrierless reactions. In these cases, determining rate constants can be quite difficult because it either necessitates computationally expensive trajectory calculations involving high quality multidimensional analytic PESs, or it requires the use of a variational form of transition state theory (TST). The latter technique, whose most advanced form is Variable Reaction Coordinate Transition State Theory (VRC-TST), [1,2] necessitates the stochastic sampling of a 6-Dimensional PES using multireference ab initio methods if accuracy is demanded. Neither VRC-TST nor multireference ab initio computational techniques have black box implementations at the moment. The current work focuses on the creation of techniques for making VRC-TST theory more accessible by enhancing the stability of the stochastic multidimensional sampling step. We concentrate on two distinct characteristics of VRC-TST simulations. VRC- TST computations are done on the PES utilizing multireference simulations in its most common implementation. One feature of this technique that hinders implementation is the requirement that each single point energy (SPE) estimation conducted on the PES uses the same active space in order to be consistent (AS). However, developing an approach that ensures the selection of the same active area for the enormous number of SPE determinations required by VRC-TST theory, frequently in the tens of thousands, is problematic. In this paper, we describe a method we devised to ensure that electrical structure computations always converge to the same AS. To reach these findings, we link VRC-TST computations with EStokTP [3] to create a CASPT2 input in which the structure is translated to internal coordinates so that it may use a reference guess wave function to systematically converge to the required electronic structure. The second way we suggested to improve VRC-TST robustness is to use Density Functional Theory (DFT) for sampling the PES, again utilizing reference wavefunctions as guesses. When applied to barrierless processes, correction potentials based on multireference computations are used to improve DFT predictive capabilities. In this paper, we discuss the implications of using alternative functionals to estimate rate constants for a variety of systems. References 1. L. B. Harding, Y. Georgievskii, S. J. Klippenstein, J. Phys. Chem. A , 109, 4646-4656 (2005). 2. Y. Georgievskii, L. B. Harding, and S. J. Klippenstein, VaReCoF 2016.3.23. 3. C. Cavallotti, M. Pelucchi, Y. Georgievskii, S.J. Klippenstein, J.Chem.Theory Comput. 4. 15, 1122–1145 (2018).
P05
© The Author(s), 2022
Made with FlippingBook Learn more on our blog