Rational Design of Interfacial Properties of Transition Metal Dichalcogenides for Device Applications using ab initio Calculations

Abstract

This dissertation presents a computational investigation of two-dimensional (2D) material heterostructures with the goal of understanding and engineering their electronic behavior for device applications, with a particular focus on memristive systems. Using density functional theory (DFT) in combination with data-driven modeling, this work explores how interfacial interactions, electrode selection, and structural phase transitions govern the properties of transition metal dichalcogenide (TMD)-based devices. First, proximity-induced charge transfer is examined in heterostructures composed of RuCl$_3$ and semiconducting TMDs or graphene. Due to the large work function mismatch, significant p-type doping is induced in the adjacent 2D layer, with carrier densities on the order of $10^{13}$ cm$^{-2}$. This behavior is shown to be robust under structural perturbations such as out-of-plane strain and is accurately described by an analytic electrostatic model based on layer-resolved properties. These results indicate the effectiveness of high work function materials in tuning carrier concentrations in 2D systems.

Next, the mechanism of non-volatile resistive switching in single-crystal TMD memristors is investigated. Using first-principles DFT and nonequilibrium Green’s function (NEGF) transport calculations, the switching behavior is linked to a structural phase transition from the semiconducting 1H phase to the metallic 1T' phase and its impact on electronic transport. Nudged elastic band (NEB) calculations are used to complement these results by evaluating transition pathways and confirming that defects and interfacial effects can reduce the activation barrier, supporting a physically consistent picture of low-energy switching. Together, these results establish a direct connection between atomic-scale structure and device-level functionality. Building on this understanding, a high-throughput computational framework is developed to systematically evaluate a wide range of electrode/TMD heterostructures. By combining DFT calculations with machine learning techniques, including feature screening via random forests and interpretable modeling using the Sure Independence Screening and Sparsifying Operator (SISSO), key descriptors governing binding energies and phase stability are identified. These descriptors enable the construction of predictive models that relate electrode properties, such as electronic structure and work function alignment, to device-relevant energetics. The results reveal clear design principles for optimizing TMD-based memristors. In particular, electrode materials from groups 10–12 of the transition metal series are identified as favorable due to their ability to stabilize the metallic phase while maintaining appropriate interfacial interactions. More broadly, this work demonstrates how combining first-principles calculations with interpretable machine learning can provide physically grounded insight into complex materials systems, enabling the rational design of next-generation electronic devices.