Quantum-enhanced Neural Exchange-Correlation Functionals
Published in arXiv (Top 3 Best Papers at MoML 2024 at MIT), 2024
We introduced a differentiable Kohn-Sham DFT framework using quantum neural networks (QNNs) to learn exchange-correlation functionals. Our QNN-based functionals achieve chemical accuracy on both seen and unseen molecular systems, outperforming classical ML baselines using fewer parameters. This work provides a foundation for integrating quantum models into density functional theory, enabling differentiable quantum-enhanced simulations in JAX. We open-sourced this work as a package, QEX.
A related patent was filed in 2024.
Recommended citation: Igor O. Sokolov, et al. (2024). "Quantum-enhanced Neural Exchange-Correlation Functionals." arXiv:2404.14258
Download Paper