Largest-ever Computer Simulation of the Universe

Using one of the most powerful supercomputers in the world, astronomers have carried out the largest ever cosmological simulations. Known as Flamingo, the simulations trace the growth of the large-scale structure of the universe over 13.75 billion years. By comparing the simulations to actual observations, scientists hope to learn about the fundamental properties of the universe that govern its long-term behavior.

Shortly after the Big Bang, the universe was filled with an almost homogeneous soup made of atoms (mostly hydrogen and helium), dark matter particles, and almost massless neutrinos. Although the average matter density was decreasing over time as space expanded, dark matter nevertheless started to clump together under its own gravity. Atoms followed suit, eventually resulting in what we observe today: a “cosmic web” of clusters and superclusters of galaxies, each galaxy containing billions of nebulae, stars, and planets.

Ever since the 1980s, astronomers have tried to reproduce this process in computer simulations. Thanks to rapid developments in computer technology, these simulations have become ever more detailed, both in size (how much space is being simulated) and in resolution (how many particles, or “elements,” are being followed). Flamingo (a convoluted acronym of Full-hydro Large-scale structure simulations with All-sky Mapping for the Interpretation of Next Generation Observations) is the largest and most complex project so far.

Led by Joop Schaye (Leiden Observatory, The Netherlands), the international Flamingo team simulated the evolution within an expanding volume of space — a cube with sides 10 billion light-years long. In this volume, they traced the gravitational clumping of no less than 300 billion particles, each with the mass of some 130 million solar masses, equivalent to an average dwarf galaxy.

The simulations required a total of more than 50 million hours of computer time, distributed over the 30,000 processors that make up the DiRAC-COSMA8 supercomputer at Durham University in the UK. The results are published today in three papers in the Monthly Notices of the Royal Astronomical Society (see links below).

Apart from its huge size and high resolution, Flamingo sets itself apart from earlier simulations by incorporating much more than gravity alone. “Up to now, most of the cosmological computational simulations of our universe focused on modelling dark matter only, as it is the main matter component,” says cosmologist Guadalupe Cañas Herrera (ESA), who was not involved in the new work.

However, she explains, even though “normal” baryonic matter makes up only 15% of all the mass in the universe, it can have a big effect on how cosmic matter is distributed at small distances. For instance, galactic winds powered by supermassive black holes and supernova explosions may stall the growth of galaxies.

“The Flamingo project delivers cosmological simulations, including the behavior of baryons, by using a hydrodynamical approach,” says Cañas Herrera, referring to a method used to describe the behavior of gas flows.To accurately model phenomena such as galactic winds, the team calibrated their calculations against existing observations using machine learing.

Moreover, Flamingo also takes into account the behavior of cosmic neutrinos — elementary particles with a very small but non-zero mass.

“On all fronts, these simulations are the state of the art,” comments Koen Kuijken (Leiden Observatory). Kuijken, who was not involved in the studies, is a member of the Euclid consortium, which is behind the recently launched European space telescope. Euclid will measure the 3D distribution of billions of galaxies out to 10 billion light-years.

“Euclid absolutely needs these kinds of simulations,” Kuijken says. “They form the bridge between the fundamental parameters of the universe and the observations obtained by Euclid.”

So far, the Flamingo team has carried out 28 slightly different simulations, by tweaking cosmic parameters like the dark matter fraction, the neutrino mass, the influence of active galactic nuclei (known as AGN feedback), and the stellar mass function, that is, what percentage of stars fall in different mass ranges. By comparing the results with actual observations, astronomers hope to deduce the true values of these parameters and properties.

“Simulations are crucial for cosmologists,” says Cañas Herrera. “They are a key player in understanding how well theoretical models predict the behavior of our universe.”

One reason why astronomers are so eager to learn more about cosmic evolution is that cosmology is presented with two nagging problems. First: The current expansion rate of the universe (the Hubble constant) is higher than you would expect on the basis of observations of the cosmic microwave background (the “afterglow” of the big bang). This mismatch is sometimes called the Hubble tension. Second: The universe appears to be a bit less clumpy than the standard cosmological model predicts, a problem known as the S₈ tension.

“To fully understand if the tensions in the Hubble constant or in the S₈ [clustering] parameter are real, we need to verify that there is no bias introduced during the statistical analysis,” says Cañas Herrera. “Simulations play a fundamental role in this regard.”

In their third paper, the Flamingo team examines whether or not hydrodynamical processes like AGN feedback might explain the apparent S₈ tension. However, although the authors conclude that galactic winds indeed affect the large-scale structure of the universe, they write, “We find that baryonic effects are not sufficiently large to remove the S₈ tension.” If unknown systematic errors in the analysis can also be excluded, “then the exciting implication would appear to be that new physics […] is required.”

In other words: the new Flamingo simulations might indicate that something is wrong with our cherished standard model of cosmology.

References:

“The FLAMINGO project: cosmological hydrodynamical simulations for large-scale structure and galaxy cluster surveys”, J. Schaye et al., Monthly Notices of the Royal Astronomical Society, 2023.

“FLAMINGO: Calibrating large cosmological hydrodynamical simulations with machine learning”, R. Kugel et al, Monthly Notices of the Royal Astronomical Society, 2023.

“The FLAMINGO project: revisiting the S₈ tension and the role of baryonic physics”, I. McCarthy et al., Monthly Notices of the Royal Astronomical Society, 2023.