Physical Origin of High Lattice Thermal Conductivity in β-Si3N4: A Machine Learning Potential study

Silicon nitride (Si 3 N 4 ) has garnered significant interest owing to its excellent high-temperature stability and high fracture toughness. Nevertheless, the underlying physical mechanism behind the lattice thermal conductivity difference between its α - and β -phase remains unclear. To address this, a high-precision Neuroevolution Potential (NEP) was constructed for both phases using training dataset obtained from ab initio molecular dynamics calculations. Based on this potential, the lattice thermal conductivity for both phases within the 300-1100 K temperature range was calculated by the homogeneous non-equilibrium molecular dynamics (HNEMD) method which naturally includes higher-order phonon scattering and applied quantum corrections. The calculated thermal conductivity at room-temperature along the c-axis is 162.04 ± 9.63 W/mK for β -Si 3 N 4 , which is significantly higher than that of α -Si 3 N 4 (87.03 ± 1.69 W/mK). Analyses of phonon dispersion relation, phonon group velocities, and phonon lifetimes reveal that β -Si 3 N 4 exhibits both larger phonon group velocities and longer phonon lifetimes in the dominant heat transport frequency range (0-15 THz). More importantly, crystal symmetry analysis revealed that the β -Si 3 N 4 (hexagonal, P6_3/m) has higher structural symmetry than the α -Si 3 N 4 (trigonal, P31c). This not only improves the phonon group velocity, but also significantly extends the phonon lifetime, thus contributing to its higher intrinsic thermal conductivity. Besides, the lattice thermal conductivity of both phases decreases as temperature rises, primarily due to enhanced anharmonic scattering that shortens phonon lifetimes. This work reveals the physical origin of the higher intrinsic lattice thermal conductivity of β -Si 3 N 4 , providing a theoretical foundation for optimizing silicon nitride-based thermal management materials.