Crystal-like thermal transport in amorphous carbon

Thermal transport in amorphous carbon has attracted immense attention due to its extreme thermal properties: It has been reported to have among the highest thermal conductivity for bulk amorphous solids up to ~37 W m−1 K−1, comparable to crystalline sapphire (α-Al2O3). However, mechanism behind the high thermal conductivity remains elusive due to many variables at play. In this work, we perform large-scale (~105 atoms) molecular dynamics simulations utilizing a machine learning potential based on neural networks with first-principles accuracy. Through spectral decomposition of thermal conductivity which enables a quantum correction to classical heat capacity, we find that propagating vibrational excitations govern thermal transport in amorphous carbon (~100 % of thermal conductivity) in sharp contrast to the convention that diffusive vibrational excitations dominate thermal transport in amorphous solids. This remarkable behavior resembles thermal transport in simple crystals. Our work, therefore, provides a perspective that deepens our understanding of intermediate thermal transport mechanisms between the two ends of spectrum of solids: crystalline and amorphous solids.