Date of Award
Open Access Dissertation
Chemistry and Biochemistry
This dissertation presents the progress of two independent projects. Chapter 2 and Chapter 3 focus on the first project, which involves material exploration utilizing machine learning techniques. We explore the potential use of cobaltocenium (CoCp+2) derivatives as metal cations in anion exchange membranes (AEMs) for alkaline fuel cells, highlighting their superior thermal and alkaline stability compared to ammonium derivatives. The stability of CoCp+2 can be fine-tuned by varying the substituent groups attached to the cyclopentadienyl ring (Cp) in CoCp+2 .These derivatives encompass a variety of electron-donating and electron-withdrawing groups as substituents on both Cp rings of the CoCp+2 cation. To overcome the experimental challenges associated with synthesizing and characterizing all possible derivatives, we employ computational tools integrated with machine learning techniques to predict their stability. We use bond dissociation energy (BDE) as a proxy to measure the stability of CoCp+2 derivatives. In Chapter 2, we discuss the selection of optimal molecular descriptors for predicting BDE and the subsequent development of a theory model based on a relatively small dataset. This model is further refined into deep learning neural networks. In Chapter 3, we expand the dataset to gain deeper insights into the stability of CoCp+2 derivatives and employ standard machine learning models to predict BDE, thereby achieving more accurate predictions while reducing computational time.
Chapter 4 focuses on the second project, which investigates the reactivity of aromatic systems towards radicals. Understanding substitution reactions between aromatic systems and radicals is challenging due to the highly reactive nature of radicals and their interaction with delocalized Π electrons in the ring system. Existing models based on electron density, frontier molecular orbital theory (FMO Theory), and unpaired spin population of related radicals explain the reactivity of closed-shell molecules with free radicals but fall short in explaining reactions involving more complex molecules. To address this limitation, we propose a hypothesis that spin polarizability can effectively predict the reactivity of aromatics. To validate this hypothesis, we develop a theory model based on the Coupled Perturbed Hartree-Fock Theory.
Wetthasinghe, S. T.(2023). Computational Studies of Bond Dissociation Energies and Organic Reaction Mechanisms. (Doctoral dissertation). Retrieved from https://scholarcommons.sc.edu/etd/7424