Identification and Prediction of Photoplasticity in Semiconductors Using Feature Engineering and Machine learning

TL;DR

This study combines high-throughput nanoindentation, physics-informed feature engineering, and machine learning to identify and predict photoplasticity in semiconductors, enabling design of light-responsive mechanical properties.

Contribution

It introduces an interpretable machine learning framework with physics-informed features to predict photoplasticity, advancing understanding and engineering of light-responsive semiconductor mechanics.

Findings

Identified key features like bandgap and refractive index for predicting photoplasticity.

Developed transferable design rules for light-tunable mechanical behavior.

Provided a practical pathway for engineering light-responsive semiconductor devices.

Abstract

Photoplasticity, the light-induced change in plastic deformation, plays a pivotal role in the mechanical durability and manufacturing of semiconductor materials. Yet, its governing mechanisms remain incompletely understood, owing to the interplay of coupled multiphysics factors. Here, we conduct high-throughput nanoindentation measurements to compile a dataset of paired hardness values in dark and light conditions. Then, we engineer physics-informed descriptors spanning electrical, mechanical, and optical properties, and identify the ten most informative features, including bandgap, breakdown field, and refractive index, to enable an interpretable machine learning framework that yields transferable design rules for light-tunable semiconductor mechanics. By identifying and predicting photoplasticity in semiconductors, this work provides a practical pathway for extracting mechanism-linked, transferable guidelines to engineer light-responsive mechanical behavior in semiconductor materials and devices.