Date Approved

9-26-2023

Embargo Period

9-26-2025

Document Type

Thesis

Degree Name

Master of Science

Department

Electrical and Computer Engineering

College

Henry M. Rowan College of Engineering

Advisor

Dwaipayan Chakraborty, Ph.D.

Committee Member 1

Wei Xue, Ph.D.

Committee Member 2

Huaxia Wang, Ph.D.

Keywords

applied artificial intelligence, applied machine learning, bandstructures, feature vectors, non-volatile memory

Subject(s)

Computers--Design

Disciplines

Computer Engineering | Electrical and Computer Engineering

Abstract

As industry shifts increasingly towards post-Moore architectures, the importance of identifying non-CMOS devices is of growing importance. However, a large bottleneck towards the discovery of these devices is the high cost-of-entry to their fabrication and testing, despite existing literature providing a large number of successful non-volatile devices and the materials used in their construction. This thesis leverages this literature in tandem with the availability of single-value chemical and physical properties, as well as many-value bandstructures of these materials through the Materials Project database. A method of classifying these materials for classical ML classification is presented, using simple feature vectors of these single value chemical and physical properties, as well as multiple methods of classification which incorporate the many-value calculated bandstructures for these materials. After performing this classification, various explainable AI methods are implemented to further identify material characteristics that contribute to the non-volatile classification of a candidate material.

Available for download on Friday, September 26, 2025

Share

COinS