Research questionHow can machine-learning interatomic potentials capture long-range electrostatics and polarization in ionic, polar, and interfacial systems?Short-range interatomic models often miss electrostatic interactions and polarization that extend beyond local atomic environments. This limits their accuracy for systems whose energies and responses depend on long-range charge and dipole behavior. Latest papersRecent research connected to this question, newest first.Polarizable atomic multipoles for learning long-range electrostaticsThe source studies a semi-local framework that learns environment-dependent atomic multipoles and response terms from energies and forces, without direct supervision of electrical responses. Evidence spans four benchmarks and four short-range MLIP architectures, including bulk water, MAPbI3, water–air interfaces, and ferroelectric HfO2; reported outputs include Born effective charges, infrared and Raman spectra, vibrational sum-frequency spectra, LO–TO splitting, and polarization switching.research paper · Sep 1, 2026