Mechanical Stability Of Body-Centered Cubic Iron At Earth's Core Pressure And Temperature Conditions Using A Molecular Graph Kernel Regression And Gaussian Process.

We used machine learning to study nonmagnetic BCC iron in conditions similar to thosefound in the Earthâ??s core. Our model predicts the stability of BCC iron at high temperatures and pressures by analyzing a dataset from ab initio molecular dynamics simulations. The framework uses Gaussian process regression to make predictions by comparing the similarity between molecules. This comparison is made using the marginalized graph kernel. To use the marginalized graph kernel, the framework applies a spatial adjacency rule to convert molecules into graphs, where the vertices and edges are labeled by elements and interatomic distances, respectively. We can use this model to forecast energy landscapes, identify the forces acting on iron atoms and recognize any instability pathways. Our findings suggest that BCC iron is stable in such extreme environments. This machine learning approach helps us better understand material behavior in challenging experimental settings, which is valuable for geophysical and materials science applications.