Neural network-based prediction of particle-induced fission cross sections for r-process nucleosynthesis trained with dynamical reaction models

TL;DR

This paper develops neural network models trained on theoretical reaction data to efficiently predict particle-induced fission cross sections, aiding large-scale r-process nucleosynthesis simulations.

Contribution

It introduces a neural network approach trained on INCL+ABLA model data to predict fission cross sections, addressing computational challenges in r-process modeling.

Findings

Neural networks accurately predict proton-induced fission cross sections.

Models demonstrate good generalization across diverse nuclei.

The approach reduces computational costs in r-process simulations.

Abstract

Large-scale computations of fission properties play a crucial role in nuclear reaction network calculations simulating rapid neutron-capture process (r-process) nucleosynthesis. Due to the large number of fissioning nuclei contributing to the r-process, a description of particle-induced fission reactions is computationally challenging. In this work, we use theoretical calculations based on the INCL+ABLA models to train neural networks (NN). The results for the prediction of proton-induced spallation reactions, in particular fission, utilizing a large variety of NN models across the hyper-parameter space are presented, which are relevant for r-process calculations.