Machine learning for mixed quantum-classical dynamics and transition metal chlorides oxidation potential via density functional theory

Abstract

[EMBARGOED UNTIL 8/1/2024] Chemical phenomena in our daily life are understood to be the result of physical interactions between matter. These interactions are governed by physical laws that mostly has been understood and well-described through the principle of quantum mechanics (QM). In 1929, Dirac stated4: "The fundamental laws necessary for the mathematical treatment of a large part of physics and the whole of chemistry are thus completely known, and the difficulty lies only in the fact that application of these laws leads to equations that are too complex to be solved." QM in principle is capable of providing a description to any given system and thus predict a likely outcome or an accurate measurable observation for that system. Many would agree that while the physical laws itself is well understood and the equation governing the interaction is readily constructable, solving these equations for many chemical systems still require approximations due to computational power restriction. This in turn limit the accuracy of the predicted outcomes, which can affect our capability to understand the given system. An example of a system that represent an important chemical research area is the interaction between light and molecules. The interaction between these matters governs the study of excited state molecules. Many natural processes, such as photosynthesis,5 to important technologies, such as photovoltaic cells6,7 and spectral analysis8-10 arise from understanding these interactions. When photons interact with a molecular system, depending on the suitability of the wavelength of the given light, the molecular system can absorb the photon energy sending its electron to a different electronic state and transition the system to an excited state. After this photoexcitation, the system can undergo various decay which include radiative decay or nonradiative pathway to other electronic states. In a radiative decay (or photoluminescence), the system release photon with energy that corresponds to the difference in energy between the two electronic states. Whereas in a nonradiative decay, the system undergoes an internal conversion or an intersystem crossing. Internal conversion is a transition between electronic states of the same spin multiplicity. (e.g., singlet to singlet or triplet to triplet.) An intersystem crossing is a nonradiative transition between electronic states of different spin multiplicities. (e.g., singlet to triplet or triplet to singlet.) It is understood that intersystem crossing is facilitated by spin-orbit coupling which tends to be small for small systems. However, it is also has been shown that it can occur at short times and compete with internal conversion even in small molecules,11 particularly organometallics with heavy elements.12 Furthermore, triplets are understood to have longer lifetimes that allow organic molecules to undergo phosphorescence with good quantum efficiency.13,14 A hundred years since its development, QM have developed several methods to approximate the interaction between particles to a varying degree of accuracy. Some of those method includes Hartree-Fock (HF),15,16 Kohn-Sham density functional theory (DFT),17,18 configuration interaction (CI),19 and coupled-cluster (CC).20 As the accuracy of these approximation method increases, so is the computational cost to solve the QM equations. The exponential advancement of computing technology has led to increasingly powerful computer over the years. This progress has enabled us to study larger systems more effectively using an increasingly accurate method. However, for many large systems, studying their excited state accurately remained computationally challenging due to the manifold of degenerate states and their couplings. Recent advancements of new approximative method through machine learning (ML) and their increasingly accurate prediction at fraction of the computational costs,21,22 is a promising method to provide solutions for excited state dynamical study. In an effort to contribute to the development of this new chemical frontier, this dissertation aims to presents a methodical approach in developing a good initial dataset to train ML model that are capable of good accurate chemical prediction. In addition to the works in ML, this dissertation also includes collaborative efforts with Dr. Young's group. We performed DFT calculations that provide support to the mechanistic picture they established for the surface reactions responsible for the oxidative molecular layer deposition (OMLD), a mechanism that enables polymer growth on the surface of thin films for various electrochemical applications including energy storage,23-25 sensors,26,27 textiles,28-31 and desalination.32 Building upon this insight, we sought to benchmark oxidation potentials of various transition metal chlorides using DFT method. This investigation aims to enable other transition metal chloride oxidants to be selectively paired with various monomers thereby facilitating novel chemistry to be explored via OMLD mechanism. This work also hopes to motivate further computational studies to further explore and better characterizes various metal-halides compounds that can been used as oxidant for OMLD mechanism.