Our research highlights serve as a collection of feature articles detailing recent scientific achievements on GCS HPC resources. 

Researchers Peer into Dark Matter with MillenniumTNG Simulation
Research Highlight –

A research collaboration including a team based at the Max Planck Institute for Astrophysics has long leveraged world-class supercomputing resources to understand how our universe came to exist in its current form. Building on the successes of the previous “Millennium,” “Illustris,” and “IllustrisTNG” projects, the researchers are simulating dark matter in unprecedented detail in the context of the “MillenniumTNG” project.

In recent decades, a mix of technological advancements in observational facilities as well as computational power have ushered in an exciting, productive period for astrophysicists and cosmologists. Among the countless new theories and observations collected during this period, perhaps none has more fundamentally reshaped thinking about how our universe functions than the hypothesis that a mysterious form of matter called dark matter dominates the universe, making up a much larger percentage of matter in the universe than the so-called baryonic matter that comprises planets, stars, and living beings here on Earth.

A team of researchers led by Prof. Dr. Volker Springel of the Max Planck Institute of Astrophysics seeks to better understand both the origins of our universe as well as how materials spread throughout it today, as cosmologists must spend more effort understanding dark matter’s behavior and characteristics.

The team is well-equipped to probe the cutting edge of cosmology research: as part of the Millennium, Illustris, and IllustrisTNG consortia, it has worked for years on developing cosmological simulations on world-class high-performance computing (HPC) resources. The team has long turned to the Leibniz Supercomputing Centre (LRZ) in Garching near Munich in support of their simulation efforts. Recently, the researchers have used LRZ’s flagship supercomputer, SuperMUC-NG, as part of the next iteration of this research collaboration, the MillenniumTNG project, named as a nod to its predecessor projects and the insight they have provided for the team’s ambitious efforts to better understand dark matter.

“Our new project combines both approaches from previous work,” Springel said. “We have a simulation suite, which includes a very large hydrodynamics simulation like we did in the IllustrisTNG project, but at bigger volume. Then the Millenium simulation, which was done all the way back in 2005, gave us the ability to map out the cosmic web very well and analytically predict galaxy formation on top of it for an area of 500 megaparsecs cubed. 20 years later, we are now at the stage where we can do this very large volume that we did for the Millenium simulations, but with full physics included in the simulation.”

Over 700 papers have been published connected to the simulations done as part of this chain of successor projects. With MillenniumTNG, the team created the largest simulation of dark matter of its kind, and recently published a series of 10 articles in the Monthly Notices of the Royal Astronomical Society.

Dark matter comes into view

For the MillenniumTNG project, Springel and his collaborators designed a computational approach that could facilitate a full-physics simulation of a large slice of the universe, but their goal was to focus on better understanding dark matter’s influence on celestial bodies and materials moving through galaxies. In order to represent galaxy formation as accurately as they had in previous work, the team used its in-house AREPO code for accurately representing the complex hydrological dynamics in play during galaxy formation. The researchers also developed a new code, GADGET-4, that focused on high-resolution simulations of dark matter. For the first time, the team also included massive neutrinos in its simulations. Neutrinos are also known as “ghost particles,” because they are the most numerous particles in the known universe, yet they are so small and neutral in their interactions with other particles that scientists have been unable to accurately measure their mass. To create a hyper-realistic view of the universe, researchers must model neutrinos as part of their simulations.

The team was able to accurately simulate dark matter in a computational cube that is roughly 10 billion lightyears across. By adding in neutrinos, the researchers created a model of galaxies so large that they can extrapolate how effects seen in their simulation can impact the universe as a whole.

When simulations such as those running as part of MillenniumTNG work in tandem with next-generation experimental resources, such as the European Space Agency’s Euclid satellite, they provide cosmologists a powerful tandem of tools to answer burning questions in the field and bring scientists closer to measuring the mass of elusive neutrinos.

With our new simulation suite, we are able to develop new ways to test the influence of neutrinos on cosmological structures,” Springel said. “Then we can develop new observational probes or advise which probes to use by demonstrating their accuracy with simulations.”

Increased computing power unlocks new details in cosmological models.

For the MillenniumTNG simulations, the team used 122,880 of SuperMUC-NG’s computational cores to run its record-breaking simulations. While it focused on dark matter for the simulation, the team also included an analytical model that allowed for accurate rendering of galaxies and matter moving throughout the universe. By simulating such a large volume, the team feels confident in using these simulations as a direct comparison with new observational data.

Springel and his collaborators have worked closely with LRZ staff over the years to refine and optimize code performance on SuperMUC-NG, and LRZ has ensured that the team has been able to access large amounts of the machine at once for its demanding calculations.

The team has refined its computational approach over many years and is a voracious user of computing time. Moving forward in the MillenniumTNG project, it plans to leverage increased computing power to include even more detail in its simulations. Specifically, it hopes to add cosmic rays—high-energy particles racing through the universe at light speed—into the simulation suite. In addition, next-generation computing power will allow the team to include a host of complex physics in play for celestial bodies and interstellar materials in close proximity to black holes.

Between the power and promise of next-generation HPC systems and the increased insight offered by new observational tools, the team feels confident that MillenniumTNG will follow its parent projects to fascinate and challenge cosmologists around the world in their search for further insights.

-Eric Gedenk

This article originally appeared in the
Autumn 2023 issue of InSiDE magazine

Tags: MPI Garching LRZ Astrophysics