Researchers have developed a new artificial intelligence-based simulator that can significantly improve the efficiency of modeling the formation of heavy elements in the universe. An international team at GSI/FAIR has created a machine learning model that enables more accurate and faster reproduction of complex nuclear reactions occurring during neutron star mergers and other large-scale cosmic events. The results of this work have been published in the journal Physical Review D.

Advancing the Modeling of Heavy Element Formation

Most heavy chemical elements are formed during extreme cosmic phenomena such as supernova explosions and neutron star mergers. These processes release the energy necessary for the creation of heavy nuclei through rapid neutron capture (the r-process). During the r-process, atomic nuclei quickly absorb free neutrons, some of which then turn into protons, facilitating the formation of heavy elements found in nature. Simulating such reactions requires significant computational resources, making this one of the most challenging tasks in nuclear astrophysics.

Using Artificial Intelligence to Accelerate Calculations

The new system, named RHINE (Rapid Heating Implementation in r-process Hydrodynamic Simulations with Neural Networks), utilizes deep neural networks to estimate the amount of energy released during r-process nuclear reactions in hydrodynamic simulations. This process, known as heating, affects the speed of matter ejection and the subsequent glow observed, for example, as a kilonova during neutron star mergers.

Instead of performing every nuclear calculation during each simulation, the artificial intelligence is pre-trained on a large library of benchmark calculations with complete nuclear reaction networks. After training, the system can accurately estimate heating rates using only a fraction of the computational resources. This approach allows for approximate calculations with minimal time and power consumption.

Outlook and Open Access

RHINE is expected to enable more detailed simulations in the future with significantly reduced computational costs. Improved models may help establish links between experiments at the upcoming FAIR research facility and astronomical observations of stellar explosions and neutron star mergers. The RHINE source code has been released as open source for use by other researchers. The project was partially funded by the European Research Council (ERC) and other organizations.