Computational methods for predicting kinetics of 2D phase change materials and optimizing tight-binding parameters
Abstract/Contents
- Abstract
- Computational methods are playing an increasingly influential role in the design and analysis of materials. The need to describe materials properties in a reasonable timescale has made it important for scientists to develop and use computational methods to both create new methods to elucidate previously unknown information about a material, as well as develop faster methods to do these computations in a feasible amount of time. This thesis focuses on both aspects of developing new methods to learn new materials properties, as well as faster methods to compute materials properties in a reasonable timescale. The first half of this thesis focuses on two-dimensional (2D) materials, a family of materials that are only a few atoms thick yet possess a number of intriguing properties, including properties that do not appear in the 3D bulk materials from which they originate. These 2D materials exist in multiple different phases, and this phase change ability points towards applications in electronic and optical devices as well as information and memory storage. However, the most mature and predictive modeling capabilities are only aimed at the calculation of thermodynamic qualities. Kinetic effects have historically been more difficult to describe accurately, but are crucial to the ability to engineer materials. We develop a method to further understand the kinetics of the phase changes in 2D materials including MoS$_2$ and MoTe$_2$, where structural changes between the 2H and 1T' phases have been reported under a variety of stimuli. Curiously, these 2D transition metal dichalcogenides are found to have various geometric constraints that prevent solving for nucleation kinetics in a conventional manner using the commonly employed Wulff construction method. Here an approach is developed to account for these constraints, allowing for the prediction of the nucleating equilibrium crystal energy and shape of one phase within the other. Combined with density functional theory (DFT), we describe the subsequent predictions of kinetic timescales under a variety of conditions, including different temperatures and electrostatic gating. These predictions point to strategies to engineer kinetics in these phase change materials, an important capability for a practical phase change material. In the second half of this thesis, we discuss approaches to make computationally efficient, yet accurate predictions for materials properties using density functional tight-binding. We make more computationally efficient predictions, including for atomization energies and forces, for C, H, N, and O containing molecules with respect to first principles DFT calculations. We use density functional tight-binding theory (DFTB) as the underlying foundation, and combine this with policy-guided Monte Carlo reinforcement learning methods to obtain parameterizations with a root-mean squared test set error of 0.01 -- 0.06 eV/atom with respect to DFT values, and at a fraction of the computational cost. We utilize both stochastic and deterministic methods for exploration and exploitation of the parameter space. This results in sampling of the full parameter space, subsequently allowing estimation of parameter sensitivity and correlation between parameter variables. Unlike common approaches that optimize only the pair potential terms, we also perform a holistic optimization over numerous parameters including bond integrals, pair potentials, and on-site energies.
Description
| Type of resource | text |
|---|---|
| Form | electronic resource; remote; computer; online resource |
| Extent | 1 online resource. |
| Place | California |
| Place | [Stanford, California] |
| Publisher | [Stanford University] |
| Copyright date | 2019; ©2019 |
| Publication date | 2019; 2019 |
| Issuance | monographic |
| Language | English |
Creators/Contributors
| Author | Krishnapriyan, Aditi |
|---|---|
| Degree supervisor | Reed, Evan J |
| Thesis advisor | Reed, Evan J |
| Thesis advisor | Cai, Wei, 1977- |
| Thesis advisor | Pop, Eric |
| Degree committee member | Cai, Wei, 1977- |
| Degree committee member | Pop, Eric |
| Associated with | Stanford University, Department of Materials Science and Engineering |
Subjects
| Genre | Theses |
|---|---|
| Genre | Text |
Bibliographic information
| Statement of responsibility | Aditi S. Krishnapriyan. |
|---|---|
| Note | Submitted to the Department of Materials Science and Engineering. |
| Thesis | Thesis Ph.D. Stanford University 2019. |
| Location | https://purl.stanford.edu/vn877dy4757 |
Access conditions
- Copyright
- © 2019 by Aditi Krishnapriyan
- License
- This work is licensed under a Creative Commons Attribution Non Commercial 3.0 Unported license (CC BY-NC).
Also listed in
Loading usage metrics...