Author ORCID Identifier

https://orcid.org/0000-0002-4324-0106

Semester

Summer

Date of Graduation

2024

Document Type

Dissertation

Degree Type

PhD

College

Eberly College of Arts and Sciences

Department

Physics and Astronomy

Committee Chair

Aldo Humberto Romero

Committee Co-Chair

Paul Cassak

Committee Member

Paul Cassak

Committee Member

Matthew Johnson

Committee Member

Srinjoy Das

Abstract

This dissertation explores predicting the physical properties of solids using first-principles methods, with a focus on Density Functional Theory (DFT). DFT uses the electronic density within a material to predict its properties, simplifying the treatment of electron-electron interactions and allowing the study of realistic systems with a balanced treatment between accuracy and computational efficiency. Additionally, machine learning (ML) is employed to create correlations between some physical properties of solids and other properties or parameters that are more difficult to calculate.

The main problem addressed in this study is adjusting the parameters in the Strongly Constrained and Appropriately Normed (SCAN) semilocal density functional to accurately predict the electronic properties of various solid materials. SCAN is an exchange-correlation (XC) functional used in DFT to handle the quantum mechanical exchange and correlation effects in electron-electron interactions. The electronic nature of solids (metal, semiconductor, or insulator) is defined by the value of the electronic bandgap. Although SCAN improves electronic bandgap predictions compared to other XC functionals, it still falls short of experimental values. The goal is to develop a dynamic XC functional, called d-SCAN, focused on predicting electronic bandgaps by tuning the inner parameters to match experimental bandgap values of various semiconductors and insulators. This aims to provide insights into the relationship between SCAN parameters and the material's electronic behavior.

Key findings are that the current SCAN XC functional cannot always match the experimental bandgaps of some materials. The two main exchange parameters, metallicity and bonding, are found to be particularly influential in balancing accurate bandgap predictions with the lattice parameters of solids. Improving the accuracy of bandgap predictions also enhances the precision of bandgap-related properties like the optical dielectric constant and dielectric function, which are essential for understanding light interaction with matter. Additionally, it is observed that d-SCAN generally performs better than other XC functionals in opening electron bandgaps for doped materials.

Machine learning techniques are used to predict the values of metallicity and bonding parameters based on the physical properties of materials. Results suggest a significant correlation between these parameters and the physical properties of materials. The ML model identified covalency, space group, and bond strength as the top three key features. The ML model outperforms traditional methods in both accuracy and variability.

Overall, this research highlights the importance of fine-tuning XC functional parameters to accurately predict the physical properties of materials and represents a significant step towards developing a new dynamic ML-driven XC functional specifically for semiconductors and insulators.

Download

Share

COinS