Modeling the Energetic Landscape of Transition Metal Complexes via Electronic Structure Theory and Chemically-Informed Artificial Intelligence Methodologies

Abstract

Methods used to computationally study the electronic structures of complex chemical systems are ever-evolving to address the desire for increased accuracy and reduced computational cost. Modern advancements have given rise to sophisticated methodologies, such as Density Functional Theory (DFT), and offer a means in which to evaluate systems ranging from simple organic materials to transition metal nanoparticles. Part I of this thesis will employ DFT in order to simulate spectroscopic signatures of bimetallic platinum(II) complexes and analyze the electronic structure as a function of the complex ligand. Through this study, a general trend can be extracted to define a set of design rules for building Pt(II) dimer complexes with desirable electron transfer behavior. Part II of this work will introduce a physics-informed reinforcement machine learning (RL) algorithm that has been designed to learn from physically-motivated actions and seek to address the cost challenges of studying large metal-hydride systems by adapting the RL algorithm to simulate the electronic landscape. Electronic structure theory will supplement the RL algorithm, in the metal-hydride material application, to improve the quality of the simulated physical properties and aid in training the model.