Using Machine Learning to Improve Equations of State for Core-Collapse Supernovae and Neutron Star Mergers.

The equation of state for matter under extreme densities and temperatures is important for understanding the behavior of core-collapse supernovae and neutron star mergers. The equation of state is typically computed in tabular form for simulations. However, due to the large number of nuclear species involved, the numerical evaluation can be unstable, leading to acausal speeds of sound. We fix these tables with a machine learning algorithm called Gaussian process regression (GPR). GPR is used to correct the acausal points by interpolating over the Helmholtz free energy of nearby points that have causal speeds of sound. The rest of the entries in the EOS table can be derived from this interpolated free energy. We will demonstrate fixing a region of acausal points in the EOS table using GPR.