All-electron machine-learned orbital-free density functional for periodic CHNO systems

Key points

• Electrostatics are evaluated analytically in an atom-centered Gaussian basis, while an equivariant network learns the kinetic and exchange-correlation energy from energies and gradients of non-ground-state densities.1
• Minimizing the functional from a superposition of atomic densities recovers the Kohn-Sham ground state: on held-out crystals, the density error falls from 6.7% to 0.4% of the electron count and the energy error reaches 0.25 mHa per atom, half that of a same-architecture interatomic potential.1
• A single functional can be trained for both isolated molecules and periodic systems, and the model generalizes to larger unit cells.1
• Together, these results demonstrate the viability of machine-learned all-electron orbital-free DFT for variational density optimization and energy evaluation in fixed-geometry solids.1

Sources (1)

• [1]All-electron machine-learned orbital-free density functional for periodic CHNO systems
  arXiv (AI, ML, NLP, CV, robotics, multi-agent) · Oct 13, 08:47 AM

  We introduce STRUCTURES-P, a machine-learned all-electron orbital-free density functional for periodic CHNO systems.

  Electrostatics are evaluated analytically in an atom-centered Gaussian basis, while an equivariant network learns the kinetic and exchange-correlation energy from energies and gradients of non-ground-state densities.

Extractive summary: sentences quoted from the sources.