Semester of Graduation

Fall 2025

Degree

Master of Chemical Engineering (MChE)

Department

Chemical Engineering

Document Type

Thesis

Abstract

The development of alloyed metal films for high-performance optoelectronic devices, like ultrafast Schottky barrier photodetectors, is key to advancing technologies in ultrafast detection and communication. However, exploring these materials experimentally becomes increasingly difficult as the number of alloy components grows. For example, moving from binary to ternary or quaternary systems creates a combinatorial explosion of possible compositions, making it nearly impossible to test them all in reasonable timeframe. Additionally, there’s a lack of frameworks that not only predict the properties of these alloys but also explain why certain composition perform better- particularly in terms of band hybridization and its impact on hot carrier generation and lifetime.

To address this, we propose a framework that combines transfer learning (TL) with physics-based modeling. TL allows us to leverage knowledge from simpler or well-studied systems to guide the exploration of more complex alloys, reducing the need for extensive experimentation. But more than just predicting properties, this approach helps us understand the underlying physics -such as how band hybridization influences electronic structure and hot carrier dynamics. This insight is critical for identifying which alloy compositions are most promising and why, rather than relying on trial and error.

In this work, we develop a physics-informed model validated by spectroscopic ellipsometry, X-ray diffraction, and ultrafast pump-probe spectroscopy. This model generates synthetic data to train machine learning surrogate models, which are then refined using TL to adapt to new alloy systems. The goal is to optimize key properties like hot carrier lifetime and optical properties while providing clear physical explanations for why certain compositions excel. By bridging prediction with understanding, this approach accelerates the discovery of high-performance materials for next-generation optoelectronic devices.

Date

11-3-2025

Committee Chair

Kevin M. McPeak

Download

Available for download on Thursday, November 02, 2028

Share

COinS