First Principles and Machine Learning Guided Search for Extreme Thermal Conductivity Materials - College of Engineering - Purdue University

Project Description

Extreme (high and low) thermal conductivity is a key material property desired for many applications. Existing high or low thermal conductivity materials suffer one or more drawbacks including limited performance, high cost, unfavorable interfacial resistance, etc. Moreover, discovery of such materials had been largely based on trial and error process, which is expensive and slow. Accelerated development based on state-of-the-art first principles calculations and machine learning methods are highly desirable. Recently, Ruan has developed the general formulation of four-phonon scattering and predicted its significance in both high and low thermal conductivity materials, such as in boron arsenide (BAs) with a thermal conductivity of 1400 W/m-K which was confirmed by experiments later. It has since been accepted by the community that four-phonon scattering theory should be considered when searching for high and low thermal conductivity dielectric and semiconductor materials. In this project, the Gilbreth fellow will work in the area of thermal conductivity or other relevant topics of mutual interest. The fellow will be co-mentored by Professors Xiulin Ruan and Guang Lin, with their combined expertise on thermal transport and machine learning. Potential tasks include but are not limited to high throughput search of high and/or low thermal conductivity materials with predictive first principles calculations that incorporate 3- and 4-phonon scattering. Further, considering that such predictions are computationally expensive, we will develop novel machine learning approaches such as neural networks and genetic algorithms to significantly accelerate the prediction and gain unique insights.

Start Date

Spring 2026

Postdoc Qualifications

Co-advisors

Bibliography