Research on heat transport under high temperature and pressure is significant for planetary science. Compared with ambient conditions, the intricate phonon physics emerges at extreme conditions due to the substantial variation in lattice dynamics with high temperature (T) and pressure (P), proposing challenges for accurate modeling of thermal conductivity. Recently, we combined machine-learning potential with molecular dynamics to efficiently investigate the thermal transport of several solids in a broad P-T range within first-principles accuracy. A series of novel phonon transport behaviors were revealed, such as partial failure of phonon particle image in magnesium silicate, anomalous heat transport induced by superionic phase transition in ice, and competitive mechanism of multi-phonon scattering in magnesium oxide. Our work provides fundamental insights into phonon transport mechanisms under extreme conditions and guides modeling thermal conductivity in the interior of planetary physics.

thermal conductivity,phonon,extreme conditions,machine learning,molecular dynamics