Computational Design of Quantum Defects via Machine Learning Acceleration - College of Engineering - Purdue University

Project Description

The screening of quantum defects in semiconductors using density functional theory (DFT) is limited by the expense of large calculations at advanced levels of theory and efficiently calculating properties such as zero-phonon line wavelengths and Huang-Rhys factors. A targeted search over compounds and defects of interest with DFT accelerated by state-of-the-art machine learning interatomic potentials (MLIPs) provides a suitable solution. Building on recent MLIP+ DFT workflows, this project will drive the discovery of stable point defects in 3D and 2D binary oxides and sulfides which create charged states in the band gap and show high excitation energies and transition dipole moments. A wide net will be cast in terms of semiconductor compositions and possible defects including vacancies, self-interstitials, anti-site substitutions, extrinsic substitutional or interstitial defects, and defect complexes that are combinations of native and extrinsic defects. The postdoctoral researcher will perform first principles simulations and develop descriptor-based property prediction models with uncertainty quantification as well as fine-tuned MLIPs for driving rapid and efficient simulations. The outcome would be the generation of a dataset of defects with promise for single photon emission and quantum applications. Results will drive collaborative experimental testing and more in-depth simulations involving beyond-DFT methods and excited states.

Start Date

January 1, 2027

Postdoc Qualifications

Experience in running atomistic simulations, training machine learning models on materials datasets, using materials databases and research tools, density functional theory simulations for defects in semiconductors, and using foundation interatomic potentials.

Co-advisors

Arun Mannodi Kanakkithodi, Assistant Professor, MSE: https://engineering.purdue.edu/MSE/people/ptProfile?resource_id=239950
Guang Lin, Professor, ME & Math: https://www.math.purdue.edu/~lin491/

Bibliography

[1] Xiong et al., Sci. Adv. 9, eadh8617 (2023)
[2] Rahman et al., APL Machine Learning. 2, 016122 (2024)
[3] Rahman et al., Physical Chemistry Chemical Physics. 28 (17): 10718–10730 (2026)
[4] Radova et al., npj Comput Mater 11, 237 (2025)
[5] Zhang et al. Proc Math Phys Eng Sci. 474 (2217):20180305 (2018)