Theory-guided Data-driven Understanding of Temperature-dependent Phonons

Phonons, quantized vibrations of atoms in crystal lattices, are fundamental to understanding heat conduction, thermal expansion, and quantum phenomena such as phase transitions and superconductivity. While large computational materials databases have advanced our understanding of vibrational properties, they predominantly rely on the harmonic approximation. To account for temperature-dependent phonons arising from anharmonicity, we present a unified computational framework integrating state-of-the-art machine learning algorithms, high-throughput first-principles calculations, and advanced anharmonic lattice dynamics simulations. Our approach enables rapid and extensive exploration of temperature-dependent phonons across the periodic table, fostering a theory-guided and data-driven understanding of chemical and structural trends in anharmonic phonon renormalization at finite temperatures. Using this framework, we identify materials exhibiting strong temperature-dependent phonons for detailed analysis of their impact on thermal transport and lattice dynamics properties. This work represents a significant advance in our ability to predict and understand phonon behavior under realistic conditions, with implications for materials design and engineering.