[M-GI33] Data-driven geosciences

[MGI33-P02] Bayesian modeling of the equation-of-state by integration of various data sets for liquid iron in Earth’s outer core

*Taroujirou Matumura¹, Yasuhiro Kuwayama², Kenta Ueki³, Tatsu Kuwatani³, Ando Yasunobu¹, Kenji Nagata⁴, Shin-ichi Ito⁵, Hiromichi Nagao⁵ (1.Research center for computational design of advanced functional materials, AIST, 2.Department of Earth and Planetary Science, The University of Tokyo, 3.Research Institute for Marine Geodynamics, Japan Agency for Marine-Earth Science and Technology, 4.Research Center for Advanced Measurement and Characterization, National Institute for Materials Science (NIMS), 5.Earthquake Research Institute, The University of Tokyo)

Bayesian modeling of the equation-of-state (EoS) was demonstrated to constrain P-wave velocity (V_P) and density (ρ) of liquid iron under Earth’s outer-core conditions. We collected six data sets obtained by high-pressure (P) and high-temperature (T) experiments and ab initio molecular dynamics simulations in previous works. These data sets were integrated into one to estimate the parameters of EoS. However, as the type of observed data depends on the approach, integrated data set includes some unobserved data. To analyze such data set, we performed an analysis based on Bayesian inference. Our analysis successfully estimated the posterior probability distribution of the parameters and unobserved data by using the Hamiltonian Monte Carlo method. These posterior probability distributions enable us to calculate P– V_P andP–ρ profiles of liquid iron along the adiabatic P–T profile together with the associated credible intervals. Bayesian modeling of the EoS enables estimation of EoS’s parameters, integration of data sets that include unobserved data and evaluation of uncertainty ranges of physical properties, such as V_P and ρ, which are important for comparison with seismological properties of the core.