NEPHONON: An Efficient Phonon Calculator Based on Neuroevolution Potentials

Phonon calculations are pivotal for understanding the thermodynamic and dynamical properties of materials, yet density functional theory (DFT) approaches often incur prohibitive computational costs for large-scale systems such as moiré superlattices. We introduce NEPHONON, an open-source high-performance package designed to quickly calculate phonon properties of extended systems with near-DFT accuracy. By leveraging the computational efficiency of machine-learned Neuroevolution potentials, NEPHONON enables the rapid generation of second-order force constants even in systems with thousands of atoms. Beyond standard phonon band structures, density of states, and group velocity, NEPHONON implements the simulation of iso-frequency phonon surfaces and inelastic neutron scattering spectra S ( Q ,E). Furthermore, the code provides access to full phonon eigenvectors, facilitating advanced analyses of topological chiral phonons and visualization of vibrational modes. Through benchmarks on twisted bilayer phosphorene, graphene, and copper, we demonstrate the package's capabilities, confirming that NEPHONON is a powerful tool for rapidly exploring complex phononic phenomena in materials.