Twisting‐Induced Phonon Localization and Ultralow Thermal Conductivity in Penta‐PdTe ₂ Bilayer Revealed by a Universal Machine‐Learning Potential

The recent synthesis of penta-PdTe₂ sheet highlights the growing interest in pentagonal topologies for developing new 2D materials. Distinct from previously synthesized 2D pentagonal materials, penta-PdTe₂ possesses the heaviest elemental composition to date, rendering its phonon transport properties particularly intriguing. In this study, the universal machine-learning potential, NEP89, is fine-tuned using high-accuracy first-principles data spanning monolayer, bilayer, and twisted configurations of penta-PdTe₂ and achieving an energy prediction error within 2.3 meV atom^-1. Leveraging the homogeneous non-equilibrium molecular dynamics simulations with the fine-tuned machine-learning potential and Wigner transport theory, the phonon transport behavior and lattice thermal conductivity is systematically investigated. These results reveal that interlayer stacking reduces the bilayer thermal conductivity to 28.00% of the monolayer value, while interlayer twisting induces a further reduction to 76.19%, leading to an ultralow thermal conductivity of (0.30 ± 0.059) W m^-1 K^-1. These findings demonstrate that the interplay of pentagonal topology, strategic elemental composition, and twist engineering provides an effective route for tuning phonon transport in 2D materials.