Where do gold and uranium come from? LUMI simulations by Polish team uncover gamma-ray burst secrets

Gamma-ray bursts (GRBs) belong to the most energetic phenomena in the Universe. Their sources are jets launched from massive collapsing stars or black holes formed after the merger of binary systems – black holes and neutron stars. Understanding the mechanisms behind these phenomena requires advanced numerical simulations that exceed the capabilities of national computational infrastructure.

A project led by Prof. Agnieszka Janiuk from the Center for Theoretical Physics of the Polish Academy of Sciences (Warsaw, Poland)  utilised the LUMI supercomputer resources to model the so-called “central engines” of gamma-ray bursts.

Research aim and scope

The main goal of the project was to conduct large-scale simulations incorporating the evolving Kerr metric and self-gravity effects, where jets propagate through the dense stellar environment. This process lasts up to several hundred seconds of real time, although the most powerful HPC machines allow modelling only the first couple of seconds of such an event.

Shorter gamma-ray bursts (lasting below a couple of seconds) originate from black hole systems formed after the merger of two neutron stars. The jets then propagate through dense winds ejected from the accretion disk and through dynamic ejecta generated by the disrupted stars before the merger.

– What is most interesting is that we want to determine the sites of chemical enrichment in the Universe and our galaxy, and answer the question of where the heavy elements come from – for example, platinum, gold, uranium, explains Prof. Janiuk.

The winds and ejecta are sites of heavy element nucleosynthesis, which occurs through rapid neutron capture on seed nuclei. The radioactive decay of freshly synthesised elements powers a distinct electromagnetic transient – a kilonova.

Project results

The newest simulations focus on accelerating GRB jets to relativistic velocities and modelling their energetics, structures, and collimation mechanisms. The team used the HARM-EOS code, developed at the Center for Theoretical Physics PAS. The microphysics of the wind incorporates nuclear physics – namely, the tabulated equation of state of dense matter based on Helmholtz energy distribution.

– Nuclear reactions that physically modify the wind’s conditions affect its pressure and internal energy, while the magnetic field and neutrino heating drive wind acceleration and allow it to escape the black hole’s gravitational potential, says the researcher.

The complex physics requires powerful computational resources – millions of CPU hours. The dynamical ejecta are probed in the code by tracer particles, which store data on chemical composition and velocities at the trajectories of the outflows.

The results of the project completed last year on LUMI (grant PLL/2024/07/017501) were published early this year in the article Janiuk, Saji, Urrutia (2026, Astronomy & Astrophysics, 707, 307). It presents a suite of 2D and 3D GRMHD numerical simulations.

– We find that the accretion disks operating in the standard and normal accretion (SANE) mode can power GRB jets via neutrino annihilation if the disk-to-black hole mass ratio exceeds approximately 0.01 and the black hole spins rapidly, summarises Prof. Janiuk.

Slowly spinning black holes surrounded by massive post-merger disks can also power these jets and can serve as sites of efficient Lanthanide nucleosynthesis, responsible for the red or purple kilonova spectral components.

Why LUMI?

Large-scale numerical simulations require powerful hardware and run efficiently only when a sufficient number of cores are available. National resources in Poland were insufficient to run 3-D GR MHD simulations with fine grid resolution and detailed microphysics.

– Before, we had to rely on simplified modelling, limited to 2D setups or on simple adiabatic equations of state. Once we obtained access to LUMI via the preparatory grant, we were able to test the code on a large number of cores. We improved its performance by implementing a hybrid, MPI+OpenMP, parallelisation,” recalls the researcher.

The difference was clear when running the code at more than 1000 cores.

Application process and cooperation

The application process was not difficult – the PLGrid user portal is user-friendly, and the team was guided to properly fill in all sections. However, it required effort to write a detailed proposal with proper scientific background, technical description, test results, and justification of ‘economic readiness’.

– The latter is particularly tricky in our field – the fundamental sciences – as we do not aim to make commercial use of our results, notes Prof. Janiuk.

The overall performance of LUMI was evaluated very positively. Compilers and modules were available, the queueing system was efficient, and jobs were run in a timely manner. The cooperation with the user support team was assessed positively – responses to tickets were always substantive, and the researchers appreciated reminders about grant allocation deadlines and system maintenance periods.

Challenges and future plans

The team has just started a follow-up project in which models will cover a different case – a MAD (magnetically arrested accretion disk) in GRB and kilonova engines. The goal is to answer a fundamental question about the source of power for GRB jets – neutrinos or magnetic fields.

– We also plan to study further the self-gravitating collapsar disks with detailed microphysics. This is a cutting-edge project, and very few groups worldwide are doing that type of research – in fact, mostly in the US, says Prof. Janiuk.

In Europe, some groups (e.g. the team at the University of Amsterdam) are performing numerical relativity simulations of core-collapse supernovae and magneto-rotational mechanisms that may create GRB engines. This is a very recent but fast-developing field of research.

– The codes must be GPU-accelerated, so we envisage this as a promising path to further develop our numerical tools, concludes the researcher.

Summary

Prof. Janiuk’s project demonstrates how crucial the availability of world-class HPC infrastructure is for research on extreme astrophysical phenomena. Without LUMI resources, 3D simulations with full microphysics would be impossible, and answers to questions about the origin of heavy elements would remain out of reach.

Author: Kamil Mucha, Cyfronet