https://doi.org/10.1140/epja/s10050-025-01661-y
Regular Article –Theoretical Physics
Employing deep-learning techniques for the conservative-to-primitive recovery in binary neutron star simulations
1
Institut für Physik und Astronomie, Universität Potsdam, Haus 28, Karl-Liebknecht-Str. 24/25, 14476, Potsdam, Germany
2
Departament de Fısica & IAC3, Universitat de les Illes Balears, Palma de Mallorca, 07122, Baleares, Spain
3
Institute for Gravitational and Subatomic Physics (GRASP), Utrecht University, Princetonplein 1, 3584 CC, Utrecht, The Netherlands
4
Nikhef, Science Park 105, 1098 XG, Amsterdam, The Netherlands
5
Max Planck Institute for Gravitational Physics (Albert Einstein Institute), Am Mühlenberg 1, 14476, Potsdam, Germany
a
henrique.gieg@uni-potsdam.de
Received:
10
March
2025
Accepted:
4
August
2025
Published online:
20
August
2025
The detection of GW170817, together with its electromagnetic counterparts, has proven that binary neutron star mergers are of central importance to the field of nuclear astrophysics, e.g., through a better understanding of the formation of elements and novel constraints on the supranuclear dense equation of state governing the matter inside neutron stars. Essential for understanding the binary coalescence are numerical-relativity simulations, which typically come with high computational costs requiring high-performance computing facilities. In this work, we build on recent studies to investigate whether novel techniques, such as neural networks, can be employed in the conversion of conservative variables to primitive hydrodynamical variables, such as pressure and density. In this regard, we perform – to the best of our knowledge – the first binary neutron star merger simulations in which such methods are employed. We show that this method results in stable simulations, reaching accuracies similar to traditional methods with an overall comparable computational cost. These simulations serve as a proof of principle that, in the future, deep learning techniques could be used within numerical-relativity simulations. However, further improvements are necessary to offer a computational advantage compared to traditional methods.
© The Author(s) 2025
Open Access This article is licensed under a Creative Commons Attribution 4.0 International License, which permits use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons licence, and indicate if changes were made. The images or other third party material in this article are included in the article’s Creative Commons licence, unless indicated otherwise in a credit line to the material. If material is not included in the article’s Creative Commons licence and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this licence, visit http://creativecommons.org/licenses/by/4.0/.
