Lattice thermal conductivity of hcp Fe-Si alloys determined by machine learning potentials

The discrepancy in the thermal conductivity (κ) of iron and iron-silicon alloys between experiments and theoretical calculations has hampered our understanding of the thermal state of the Earth’s core. The lattice thermal conductivity of Fe-Si alloys is crucial for reconciling computational and experimental results. Here, we utilized the machine learning method to compute the electronic (κe) and lattice (κl) thermal conductivity of hcp Fe-Si alloys at room temperature and high pressure. At 300 K and above 100 GPa, the lattice contribution is approximately 57% of the total thermal conductivity for the Fe-8.7wt% Si alloy. Increasing silicon content reduces the κ of the Fe-Si alloy. Additionally, we estimate the κ of liquid Fe-8.7 wt% Si alloy to be ~90 W/m/K at the core-mantle boundary, implying a relatively young inner core if Si were the major light element in the core.