Where does the gold in rings and the uranium in reactors come from? The answer lies in the most powerful explosions in the Universe. An international team of scientists has created a breakthrough AI model, RHINE, which for the first time reconstructs the processes of heavy chemical element formation with high precision.

As reported by RBK-Ukraine, citing a study in the journal Physical Review, the new development allows us to peer into the heart of stellar catastrophes without overloading supercomputers.

The Magic of the r-process

The creation of elements heavier than iron is impossible in ordinary stars. This requires extreme conditions that arise during supernova explosions or neutron star collisions. At this moment, the so-called r-process — rapid neutron capture — is triggered.

The mechanism looks like a chain reaction: atomic nuclei instantly absorb free neutrons, which then turn into protons. This is how heavy elements are born, including gold, platinum, and uranium.

However, modeling this chaos has always been a nightmare for physicists. To calculate the process, one must simultaneously account for the behavior of thousands of isotopes. Until now, scientists have been forced to greatly simplify theoretical models, as there simply was not enough computing power for a full calculation.

How the RHINE model works

The developed RHINE model offers an elegant solution to this problem, combining nuclear physics and machine learning. The algorithm works on the principle of "learning from experience":

  • Preliminary training. The neural network is trained on a huge database of reference calculations containing a complete set of possible nuclear reactions.
  • Hydrodynamic integration. Unlike old methods, the model predicts the rate of nuclear energy release for any state directly during the launch of simulations.
  • Autonomy. AI allows separating complex nucleosynthesis calculations from the modeling of matter movement, which previously overloaded systems.

Thanks to this approach, the algorithm reproduces heat release processes with high accuracy. This is critically important, as heat release affects the speed of the expansion of cosmic matter and the characteristics of electromagnetic signals recorded by telescopes — so-called kilonovae.

Reality Check

Scientists confirmed the effectiveness of the new AI scheme by comparing simulation results with real astronomical observations. The historical case of August 17, 2017, was chosen as the benchmark.

On this day, in the lens-shaped galaxy NGC 4993, the collision of two neutron stars was recorded for the first time using gravitational waves. The Hubble Space Telescope documented the gradual fading of the kilonova flash caused by this event. The data obtained from RHINE perfectly matched reality.

The use of machine learning allowed saving a colossal amount of computing time without loss of accuracy. The creators of the project have already released the source code of the program to the public.

In the future, the RHINE model will become a link between laboratory experiments at the new FAIR accelerator complex and real observations of stellar explosions, helping humanity finally unravel the origin of matter.