Faculty member co-authors paper on using neural networks to understand planet formation

Our faculty member Gabriele Morra co-authored a recent publication in Scientific Reports titled "Insights into core-mantle differentiation from bulk Earth melt simulations”. It describes the simulation of a bulk Earth melt system, formed by magma ocean and metal silicate at high pressure and high temperature, during the formation phase of a solid planet like Earth, Venus and Mars. The approach is innovative in the use of a deep neural network potential trained on first-principles data, which enables to scale ab-initio atomic dynamics of fluids to a much larger number of atoms compared to previous research. The predicted distribution gives new insights on the formation of our and other planets, and on the composition of the Earth's mantle and core. This work is mainly performed by Abin Shakya and Dipta Bhanu Ghosh under the supervision of Bijaya Karki (LSU) and is part of a NASA funded collaboration between Tulane University, LSU and UL Lafayette.

The open-access paper can be downloaded from here:

https://www.nature.com/articles/s41598-024-69873-8