Computational design and analysis of nanostructured materials for neuromorphic engineering

Abstract

We study electronic structure of nanometer-scale materials using electronic structure theories like density functional theory. Nanoscale two-dimensional intelligent materials under strain, and their applications as electronic devices for neuromorphic computing is investigated. Combination of two-dimensional materials with other materials are in the scope of this work. Density functional theory calculations are done using Quantum ESPRESSO simulation suite. Designed and analysed neuromorphic electronic materials are simulated and compared with available technological data.Thesis has significance in terms of the choice of materials; secondly, in terms of the realization of homostructures, and heterostructures, and understanding the mechanical strain effects in these structures; thirdly, in terms of the application area; namely, neuromorphic electronic devices for memories. The overall purpose and the scope of the thesis can be listed chapter-wise. In Chapter 2, "Single atom precise, ultrafast, and universal emulation of biological synapses using atomically thin vertical heterostructures": We realize the current voltage-like characteristics of heterostructures using density functional theory and Boltzmann transport methods, which can reveal the hysteresis characteristics. Moreover, we used time dependent density functional theory to show characteristics of synapses. We show a healthy synapse with N vacancy and dysfunctional synapses with pristine heterostructure. We heal the dysfunctional synapses using N intercalation. Here, a single atom can manipulate the behavior of different synapses. Created heterostructures have different abilities of learning, memory, and so on. One can use these heterostructures to realize certain brain regions on chips because every region of the brain is responsible and superb in one ability mostly. In Chapter 3,"Mimicking bacterial learning and memory in tungsten based two-sided single layers of WSeO, WSeS, WSeSe, and WSeTe": The current literature about neuromorphic materials is based on showing how one can resemble to synapses and neurons of Human using two-dimensional materials. In this chapter, we are showing for the very first time to the best of our knowledge how to use 2D materials to mimic the characteristics of the bacterial learning and memory. We developed methods to show this in 2D materials using a quantum memristor and phase-change like mechanisms. Our modeling is directly comply with the gene regulatory network's governing equation of bacteria, also physical model resemble to the real world bacteria, which is supported by experimental biological letters. In Chapter 4,"Single-electron-precise tailoring of a resistive-switching device with transfer printing": We focus on how to replace the electrochemically active top electrode of a conductive filament forming resistive switching device with an inert electrode. For this, we study a molecular junction to design and model a resistive switching device based on a single-electron box effect, where the molecular junction composed of self-assembled monolayers. Resistive switching is establish through the penetrating Au atoms from the inert top electrode after transfer printing process. Here, Coulomb blockade effect through Au island and tunneling through the self-assembled monolayers are the two effective phenomenon to explain the resistive switching. To conclude, a molecular junction is studied using the methods of density functional theory and environmental effects are modeled using the Quantum ESPRESSO's module environ. In Chapter 5, "Computational Analysis of Device-to-Device Variability in Resistive Switching devices through Single-Layer Hexagonal Boron Nitride and Graphene Vertical Heterostructure Model": We show a simple model for defects and their effects at the interfaces of atomically thin heterostructures by varying the interlayer distance between two atomically thin materials forming the heterostructure. This model will allow us to gain insights into variations in current-voltage characteristics of resistive switching devices compose of metal-insulator-metal vertical structures using the methods of density functional theory. We believe this approach of modeling interfaces with varying distances and showing how it affects the current-voltage characteristics is first time interpreted in this way. Also, the last subsection reveals a very important and easy way to mimic the neurons. In Chapter 6,"Atomistic origins of compound semiconductor synthesis with computational neuromorphic engineering": This article can be considered a follow-up to our previously published article in the Journal of Applied Physics titled 'Mimicking Bacterial Learning and Memory in Tungsten-based Two-sided Single Layers of WSeO, WSeS, WSeSe, and WSeTe.' In this new chapter, we apply a similar methodology to another bacterium. However, the primary objective of our new article extends beyond this. Our aim is to demonstrate the potential for memristivity during compound semiconductor synthesis. It is widely recognized that the fabrication process using chemical vapor deposition based methods lacks a comprehensive understanding in terms of chemical kinetics. We endeavor to elucidate how growth processes exhibit learning behavior and possess memory. This is achieved through the analysis of the smallest potentially meaningful subunit of the growth: the resonant tunneling diode structure. Furthermore, we illustrate how the convergence of various computational methods—such as tight-binding, density functional theory, transfer matrix, and Boltzmann transport theory—can contribute to the design of future multinary memristors. Specifically, $sp^3s^*$ semiempirical tight-binding methods are predicting energy band gaps more accurately than density functional theory at a low computational cost and allow the investigation of compound semiconductors. Our work throughout this thesis is mainly based on density functional theory. Here, we will consider electron density to investigate the electronic structure of the materials. In this way, researchers found functionals, which are connecting the density with the energy. Density functional theory is based on interacting electrons in an external potential. The ground state energy is determined as the unique functional of the electronic density, which means that the ground state electron density is enough to construct the Hamiltonian of the system, and it verify that the construction of many electron wave function is not required to calculate the ground state properties. To conclude, we believe that the computational studies of nanoscale two-dimensional materials using the methods of applied computational materials science can enhance the performance of the neuromorphic devices.