39th Annual Student Research Conference
Predicting Band Gap in Semiconductors Using Machine Learning
Yashwinder Rajput
Dr. Scott Thatcher and Dr. Rasanjali Jayathissa, Faculty Mentors
Bandgap is the energy required to make a semiconductor electrically conductive. It acts as an energy barrier; electrons must jump this gap to allow electrical conductivity. While conductors have zero bandgap and insulators have a very high bandgap, semiconductors have a moderately small bandgap which can be reduced by methods such as straining, pressure, heating, alloying, doping, under electrical field, or by creating quantum wells. Bandgap can be measured experimentally, or modelled mathematically using Density Functional Theory, and other first-principles methods; it, however, under- or over-predicts and is computationally costly. Machine Learning models predict it faster and much cheaper using elemental properties of constituent elements of a semiconductor and their bulk and structural properties using models such as Regression, Random Forests, Support Vector Regression.
Keywords: machine learning, material science, data science, semiconductors, physics , chemistry, statistics
Topic(s):Physics
Statistics
Presentation Type: Oral Presentation
Session: -5
Location: MG 2001
Time: 11:00