Small molecules atomistic simulations: from QCT to machine learned models

The understanding of chemical processes and their underlying mechanism has been fundamental in chemical and physical sciences. From large-scale astronomical phenomena to the evolution of microscopic organisms. The characterization and study of such a system has given new insight to experiments, which presents its own difficulty for systems that are bordering the technical possibilities.
With the improvement of computer resources, simulations are being conducted at conditions and settings that are just not possible with current experimental techniques. This work focuses on hypersonic reentry conditions where temperature can reach $T > 10000$ K and local chemistry processes have non-linear characteristics. In such settings, high accuracy observable are of vital importance for the simulation and modeling community, in particular with the lack of accurate experimental reference for all species and processes involved. Multi-scaled simulation can provide a solution for the in-depth accuracy needed for such coarse-grain observables.
The first chapter is a historical background of the previously conducted studies and modeling efforts. The second chapter focuses on the theoretical background of the construction and representation of high-fidelity potential energy surfaces (PES), ab-into electronic structure methods and a brief background into Quasi-Classical simulations (QCT).
The second part of the thesis shows the results and implementation of high-fidelity PESs in QCT simulations of atom + diatom reactions. The third chapter shows our results for the thermal and vibrational relaxation rates for the ${\rm N}(^4{\rm S}) +{\rm O}_2({\rm X}^3\Sigma^-_g) \leftrightarrow {\rm O}(^3{\rm P}) + {\rm NO}({\rm X}^2\Pi)$ reaction over a wide temperature range. The fourth chapter shows our investigation for the thermal and vibrational relaxation rates from 15 K to 20000 K for the C($^{3}$P) + O$_{2}$($^3\Sigma_{g}^{-}$) $\leftrightarrow$ CO$_{2}$ $\leftrightarrow$ CO($^{1}\Sigma^{+}$)+ O($^{1}$D)/O($^{3}$P) reaction including five electronic states.
The third part of the thesis shows the implementation of our newly developed state-to-distribution (STD) model which predicts product state distributions from a given initial state. The fifth chapter shows the implementation of the STD model for the ${\rm N}(^4{\rm S}) +{\rm O}_2({\rm X}^3\Sigma^-_g) \leftrightarrow {\rm O}(^3{\rm P}) + {\rm NO}({\rm X}^2\Pi)$ reaction using the quartet electronic state and the sixth chapter is a further implementation to machine-learned (ML) based on spectroscopic assignment. In chapter seven we show the initial steps in preparing an iterative model that can cycle through the different processes involved in a complete air chemistry system.
The last chapter shows the overall conclusion and discussion of this work.