First-principles method development and design of complex 2D materials for renewable energy applications

Abstract

Modulation of a material's dimensionality enables novel physics at the atomic scale. Exploiting this effect creates opportunities to design and manufacture highly functional materials for specific engineering applications. As such, 2D materials are an exciting material group due to their unique properties compared to their 3D counterparts. Currently, research is focused on understanding how these low dimensional materials can perform as photovoltaics, catalysts, and high strength materials. The first goal of this thesis is to understand and design the properties of complex 2D materials for novel applications in renewable energy. The second goal is to develop new methods that will enable accurate and efficient investigation of the fundamental electronic structure properties of these and other complex materials. In this thesis, we study the underlying physics of an exciting class of materials broadly referred to as transition metal phosphates (TMPs). These materials are of interest for engineering applications because of their 2D properties, ease of solution processing, and ability to form 2D monolayers. Interestingly, they form crystalline materials composed of alternating layers of TMPs and organic molecules, enabling a wide range of material properties. Additionally, TMPs exist in a variety of compositions including zirconium, titanium, vanadium, zinc, tin, and a number of other metal cations. This range of cations presents an opportunity to study a rich set of properties and potential applications within the framework of TMPs. To study these materials, we employ density functional theory (DFT) computations to investigate the properties of TMPs and TMP-based heterostructures. Using DFT, we develop a framework for the understanding and control of the band gap, band alignment, and other properties within TMP-organic heterojunctions. This work enables new pathways for the realization of cheap and efficient photovoltaic materials as well as applications to broader engineering fields concerned with precise control of band energies. In performing this study, we also address several critical limitations of DFT. While DFT is highly accurate at studying many materials properties, it has significant limitations in studying time variant and excited-state properties. Further, computationally, DFT does not scale linearly with the system size, imposing significant roadblocks to study large systems. To enable the study of these complex material properties, method development represents a significant portion of this work. Artificial neural network (ANN) approaches represent an emergent method in the field of Material Science. Exploiting this trend, we develop ANN methods to reduce the computational complexity and cost of DFT simulations. By combining large datasets of relatively small DFT calculations, we develop high dimensional potentials for large-scale molecular dynamics (MD) calculations. This enables the prediction of DFT-accurate energies in large and time-variant systems for a fraction of the computational cost. Additionally, DFT relies on accurately understanding the relationship between functionals of the charge density even though the explicit form of some functionals are sometimes unknown. To address this shortcoming of DFT, we develop machine-learning methods as a novel way to learn complex functionals. Understanding this process may allow for linear speedup in DFT calculations, possibly opening enabling 'orbital-free' DFT. In concluding this thesis, we deploy our computational framework to learn both analytical potentials as well as functionals of the charge density. We use these developed methods to study a range of material properties of interest to the engineering sciences including the bandgap and mechanical properties of 2D and bulk materials. This method could enable significant advances in the computational material science field by enabling researchers to study systems not possible with classical approaches.