Astronomy researchers are using Gadi’s scale and performance to test the next generation of their star formation simulations. They are simulating the random motions found in magnetised clouds of interstellar gas. These are complex flows of hydrogen gas spread thinly throughout our own, and other galaxies in the Universe. The clouds are undergoing supersonic, turbulent motions, interacting through gravity and magnetic fields, and collapsing in some of the ultra-dense regions to form powerful stars, like the Sun.

Three panels showing elements of turbulence from galaxy formation simulations, on the left magnetic energy is highlighted in purple swirls, in the middle gas density is highlighted in red and black, and on the right turbulent energy is highlighted in reds and yellows.
Two-dimensional slices of the magnetic energy, gas density, and turbulent energy from a three-dimensional, highly-spatially resolved, supersonic, magnetohydrodynamic interstellar medium turbulence simulation run on Gadi. This simulation has over 12 billion grid elements, allowing the researchers to explore how the large-scales in the interstellar medium, which are pervaded by supernova explosions and bulk motions from galaxies, are coupled to the small-scales, where star formation has the potential to occur.

Computational simulations that run on the world’s biggest supercomputers aim to replicate the real turbulent phenomena as closely as possible. To do this, researchers need to use and create software specially designed to run for tens of millions of hours of computing time. Only then can the simulation start to cover the range of length, time and density scales that are essential to understand interstellar turbulence. This type of turbulence is so complex because it couples large and small scales, is a mixture of slow and ultra-fast motions, and has a gigantic range in density that spans many orders of magnitude. Getting an accurate computational representation of this turbulence therefore requires extremely high spatial and temporal simulation resolutions, and extremely efficient computational and processing tools.

Associate Professor Christoph Federrath from The Australian National University uses supercomputers around the world to run turbulence and star formation simulations. On Gadi, Federrath and PhD student, James Beattie tested their high-performance FLASH simulation package on 99,840 processors at the same time. By efficiently distributing large calculations across thousands of processors, researchers speed-up and transform the quality of research they can produce.

A collage of two images. On top is a simulation of galaxy formation turbulence showing blue and black structures, and on the bottom is a similar-looking real-world picture of a galaxy. of
A comparison between the gas densities from a magnetohydrodynamic interstellar medium turbulence simulation (top) run on Gadi, and an observation of the W3/W4/W5 molecular cloud and star forming complex from the Herschel space telescope (bottom). Many of the features and structures shown in the observation can be reproduced with high-resolution, magnetised, supersonic turbulent gas motions, shown in our simulation.

Mr James Beattie, says, “We wanted to push Gadi to its limits and run this code on more processors than we ever have before in Australia. This proof of concept shows that our version of FLASH works extremely well at large scales and that Gadi is highly-capable of running jobs that utilise a significant fraction of the whole machine. We are moving to ever more realistic simulations that take into account all of the complexity and chaos of real interstellar gas. These help inform astrophysical gas dynamics and star formation researchers all over the world. Gadi, and the next-generation of Australian supercomputers are going to be vital for helping us understand the nature of the turbulent interstellar medium.”

Testing high-performance software at extreme scales helps researchers develop their codes to run as efficiently as possible for the maximum scientific benefit. The mathematics and simulations of interstellar turbulence in galaxies are remarkably similar to fluid flow simulations used for understanding the movement of air and water on Earth. Thus, scientific and technological advances from one field benefit many others. Computational disciplines at all points on the research spectrum are working with Gadi and other leading supercomputers to optimise, improve and develop their codes for their ever-increasing scientific ambitions.