Physics-Informed Long-Range Coulomb Correction for Machine-learning Hamiltonians

1. 1Derive closed-form long-range Hamiltonian matrix elements using variational decomposition of the electrostatic energy
2. 2Apply variationally consistent mapping from the electrostatic energy to the nonorthogonal atomic-orbital basis
3. 3Integrate the physics-informed correction into machine-learning models for electronic Hamiltonians
4. 4Validate the corrected models against experimental or reference data

Researchers and developers working on machine-learning models for electronic Hamiltonians, particularly those in materials science and physics, can benefit from this method to improve accuracy and efficiency

💡 Variational decomposition of electrostatic energy enables accurate and efficient modeling of long-range Coulomb interactions in polar crystals and heterostructures

💡 Correcting long-range Coulomb interactions in ML Hamiltonians with physics-informed methods!