Machine learning assisted understanding of the layer-thickness dependent thermal conductivity in fluorinated graphene

Manipulating thermal conductivity (κ) plays vital role in high-performance thermoelectric conversion, thermal insulation and thermal management devices. In this work, we using the machine learning-based interatomic potential and the phonon Boltzmann transport equation to systematically investigate layer thickness dependentκof fluorinated graphene (FG). We show that the latticeκof FG can be significantly decreased with Bernal bilayer stacking. Surprisingly, the further increasing of stacking layer can no longer affect theκ, however, theκis increased in the bulk configuration. The variation ofκcan be attributed to the crystal symmetry change from P-3m1 (164) at single layer to P3m1 (156) at multilayer. The decreasing crystal symmetry from single layer to bilayer resulting stronger phonon scattering and thus leading a lowerκ. Moreover, we also show that the contribution of acoustic mode toκdecreases with the increase of layers, while the contribution of optical mode toκis increased with increasing layers. These results provide a further understanding for the phonon scattering mechanism of layer thickness dependentκ.