The thermal conductivity of ice polymorphs constitutes a critical parameter in multidisciplinary research spanning cryobiology, atmospheric physics, and planetary science. However, the intricate structures and phonon dynamics pose significant challenges to understanding thermal transport in ice polymorphs across diverse phase diagrams. To bridge this knowledge gap, we combined deep learning potential with molecular dynamics simulations to investigate the thermal transport in ice polymorphs with ab initio accuracy across 150–400 K and ${10}^{-3}–10\mathrm{GPa}$. Pronounced anisotropic behaviors and nonmonotonic pressure dependence of thermal transport are demonstrated. Phonon dynamics analysis reveals the structure-phonon relationship and rich phonon behaviors, including nonuniformity-induced phonon hybridization, distortion-induced phonon softening, and disorder-induced phonon scatterings. Configurational local entropy quantification identifies the complexity of the hydrogen bond network as the governing factor in thermal transport variation. Our Letter provides fundamental insights into microscopic energy transport in complex hydrogen bond systems.

Close

PrevNext

View figure in article

• Lattice thermal conductivity
• Phase diagrams
• Ice
• Ab initio molecular dynamics
• Machine learning