Using differentiable programming to obtain an energy and density-optimized exchange-correlation functional
arXiv:2106.04481
Abstract
Using an end-to-end differentiable implementation of the Kohn-Sham self-consistent field equations, we obtain an accurate neural network-based exchange and correlation (XC) functional of the electronic density. The functional is optimized using information on both energy and density while exact constraints are enforced through an appropriate neural network architecture. We evaluate our model against different families of XC approximations and show that, for non-empirical functionals, there is a strong linear correlation between energy and density errors. We use this correlation to define a novel XC functional quality metric that includes both energy and density errors, leading to a new, improved way to rank different approximations. Judged by this metric, our machine-learned functional significantly outperforms those within the same rung of approximations.
References in corpus (8)
- PyTorch: An Imperative Style, High-Performance Deep Learning Library
- Climbing the Density Functional Ladder: Non-Empirical Meta-Generalized Gradient Approximation Designed for Molecules and Solids
- Gaussian Error Linear Units (GELUs)
- Recent developments in the PySCF program package
- Benchmark tests of a strongly constrained semilocal functional with a long-range dispersion correction
- Kohn-Sham equations as regularizer: building prior knowledge into machine-learned physics
- DeePKS: a comprehensive data-driven approach towards chemically accurate density functional theory
- Learning from the Density to Correct Total Energy and Forces in First Principle Simulations