Unraveling Thermal Transport Mechanisms in Monolayer CrSi₂N₄ Using Machine-Learned Potentials

Abstract

Efficient heat dissipation is vital for the reliability and performance of next-generation nanoelectronics based on two-dimensional MA₂Z₄(M= Mo, W, V, Nb, Ta, Ti, Zr, Hf or Cr, A = Si , Ge, and Z = N, P , As) semiconductors. While their electronic and mechanical properties have been extensively characterized, the microscopic physics governing thermal transport in these complex septuple-atomic-layer structures remains elusive. In this work, we systematically investigate the intrinsic lattice thermal conductivity of monolayer CrSi₂N₄ by developing Machine-Learned Potentials trained on first-principles data. This framework captures many-body interatomic interactions with quantum-mechanical accuracy, enabling rigorous assessment of phonon dynamics via large-scale molecular and lattice dynamics calculations. We predict a high room-temperature thermal conductivity of approximately 372 W·m⁻¹·K⁻¹ for CrSi₂N₄, positioning it as a promising heat-spreading candidate. Mode-resolved analyses reveal that heat transport is dominated by in-plane acoustic phonons; the flexural modes undergo pronounced anharmonic renormalization, significantly suppressing their contribution to total heat flux. Compared with MoSi₂N₄, the disparity in thermal conductivity of CrSi₂N₄ originates from altered lattice anharmonicity and a constricted three-phonon scattering phase space rather than atomic mass effects. These results provide insight into phonon transport in Cr-based MA₂Z₄ nitrides and demonstrate the effectiveness of Machine-Learned Potentials for predicting thermophysical properties of complex two-dimensional materials.