Thermal Conductivity of Silicate Liquid Determined by Machine Learning Potentials

Silicate liquids are important agents of thermal evolution, yet their thermal conductivity is largely unknown. Here, we determine the thermal conductivity of a silicate liquid by combining the Green-Kubo method with a machine learning potential of ab initio quality over the entire pressure regime of the mantle. We find that the thermal conductivity of MgSiO3 liquid is 1.1 W m(-1) K-1 at the 1 bar melting point, and 4.0 W m(-1) K-1 at core-mantle boundary conditions. The thermal conductivity increases with compression, while remaining nearly constant on isochoric heating. The pressure dependence arises from the increasing bulk modulus on compression, and the weak temperature dependence arises from the saturation of the phonon mean free path due to structural disorder. The thermal conductivity of silicate liquids is less than that of ambient mantle, a contrast that may be important for understanding melt generation, and heat flux from the core.

Automated summary

The study combined the Green-Kubo method with machine learning potential to determine the thermal conductivity of a silicate liquid and found that it increases with compression and remains nearly constant on isochoric heating. The pressure dependence is due to the increasing bulk modulus on compression, while the weak temperature dependence is due to the saturation of the phonon mean free path caused by structural disorder.