Keywords:
Density Functional Theory, DFT, Computational Materials Science, Optimization, High Performance Clusters (HPC), Thermoelectric properties, Optical Properties, Raman IR spectrum, VASP, Quantum Espresso, BoltzTrap, Machine Learning, Parameter Optimization, Data Analysis
Synopsis
Introduction
My freshman year Quantum Physics course had inspired my interest in the subject, leading to my work on Density Functional Theory (DFT) based simulation of materials. I knew that experience in computational modeling of materials would come in handy later. I worked on calculating thermoelectric properties of doped Mg2Si using VASP and BoltzTrap. In addition to literature survey, attending group meetings, learning software (including Linux, VESTA, Qt Plot, parallel programming for high performance computing cluster (HPC) and so on), I attended Professor’s courses on Quantum Mechanics (EP 252) and Computer Programming Lab (PH 413) based on Fortran for graduate students and took Condensed Matter Physics (PH 310). I learnt how to optimize code, document everyday’s progress and developed good coding practices. Specifically, we implemented PBE, PBEsol, MetaGGA based scf for optimizing parameters to finding band gap closest to that experimentally reported in literature. I compared the electronic band structure for magnetic doping with different configurations leading to ferromagnetic and antiferromagnetic materials and determined the stable phase. Transport calculations were done using BoltzTrap using a rigid shift. Data analysis to find the reason in high figure of merit was remaining to be done by the time I left in the summer. The concept behind DFT- Hohenberg-Kohn’s Variational Characterization of Eigenvalues and eigenfunctions for the density functional leading to Kohn-Sham equations was to be encountered 2 years later when I would study Spectral Theory.
Next, the material unit cell had a large number of atoms and I had to calculate the optical properties (IR and Raman Spectrum) using open source software. I used Quantum Espresso and Virtual NanoLab. This being an experimental group, I was the only theory person so I had to independently learn how to operate the new software, get it integrated with the HPC and optimize parameters to save time and computational effort. This was a steep learning curve for me as I ran into all kinds of errors and every small bug in the new code had to be solved by myself. In a month, I was ready with the optimal convergence to start optical calculations. This required using PHonon package which gave new problems in convergence. Getting convergence took a structured attack along with studying the principles behind algorithms.
Problems at the crossroads of modern civilization have increasingly emphasized the importance of computational methods for materials discovery and for acceleration of materials development.
Abstract
Future directions and Applications
Materials informatics and Cross disciplinary collaboration with Engineering departments like Physics, Mathematics, Chemistry, Computer Science are significant.