RESEARCH HIGHLIGHTS
Our research highlights serve as a collection of feature articles detailing recent scientific achievements on GCS HPC resources.
Researchers Use the Power of SuperMUC-NG to Map Our Near Universe
Research Highlight –
While countless research domains have benefitted greatly from the rise of high-performance computing (HPC), perhaps no other field has seen such rapid growth connected to HPC as cosmology and astrophysics. For many years, even the most driven and innovative scientists were limited in their study of space by what could be seen with observation facilities.
Improvements in computing technology have finally provided researchers with a method for not only verifying phenomena seen through observation, but also allowing scientists to model the behavior of dark matter and other materials difficult to observe experimentally. Researchers can also reach back in time to model how our universe came to be.
As a result, the field is in the midst of a renaissance period, with cosmologists seeking answers to previously unanswerable questions, creating computational models that allow them to compare simulation results with observation data of far-away objects, and simulating the movements of the particles, gasses, and plasmas that birthed our planet within increasing accuracy at ever-wider ranges. Recently, researchers have grown interested in better understanding how galaxy clusters are connected to one another by focusing on the “filaments” that serve as the connections between them.
The nearby Coma cluster—a cluster of at least 1,000 galaxies that lies relatively close to Earth (a mere 320 million light years away, on average)—is a major research focus of a multi-institution team of researchers centered at the Ludwigs-Maximillians-Universität München. The team gained access to the Leibniz Supercomputing Centre’s (LRZ’s) SuperMUC-NG supercomputer to computationally model the Coma cluster in unprecedented detail. With access to LRZ’s flagship supercomputer, the team was able to create the largest model of the Coma cluster to-date and plans to use it as a basis for even more intensive investigation of how our cosmic neighborhood came into existence and how early galaxies clustered together.
“This project is just one of numerous studies that can be conducted using this simulation,” said Dr. Jenny Sorce, lead researcher on the project. “We are able to look at different local objects including clusters and then compare results directly with observations. It is extraordinary when we consider that the first simulations of this kind were not all that long ago and had very few particles, but now we are able to produce these high-resolution simulations of our cosmic home.” The team published its results in Astronomy and Astrophysics.
Three 20 Mpc/h (~1 billion light-years) thick Supergalactic slices of the simulated box representing our local neighborhood in the vicinity of the Coma cluster of galaxies counterpart. More precisely, the simulated massive dark matter halo standing for the Coma cluster is in the center of the slices. The orange lines represent the filaments connecting Coma replica to the cosmic web. NE and W stand for the North-East and West simulated filaments lookalike of the observed ones. The gradient of color gives dark matter halo velocity field in Coma outskirt. Black arrows emphasize that matter is mostly fueled into the clusters via the filaments (blue cells alongside orange lines) with increasing collinearity with decreasing distance to the cluster. Image credit: Jenny Sorce
Dark matter as a guiding light
To accurately recreate the Coma cluster, the research team not only has to simulate a massive area—its largest simulations modeled a computational cube of the Coma cluster that corresponded to roughly 1 billion light years per side. Further, it is simulating dark matter—a object that we cannot actually see, but can “observe” by seeing its influence on other nearby bodies or on light-carrying photons.
For an accurate simulation, the team needed to include 8.5 trillion dark matter particles and their influence on physics in order to try and accurately create dark matter “halos” In its simulation, with the Coma cluster being the largest. While these halos are only hypothetical—since they are made of dark matter, our inability to confirm them observationally means that researchers can only prove their existence by charting a halo’s influence on nearby bodies—they seemingly have major influence on capturing or retaining plasmas, gases, and other solid materials nearby.
In fact, the team is focused heavily on ensuring that halo behavior is accurately represented in its simulations because its goal is not only to simulate the Coma cluster in accurately but ultimately to better understand the mechanisms that bind the Coma cluster to the wider universe. The so-called cosmic web is essentially a network serving as the main routes for materials to move across the universe.
Once the dark matter is well-represented in its simulation through use of the RAMSES computational code, the team then ran a “cosmic web finder” that allows it to find the minimum, maximum and tertiary “nodes,” or points where these interstellar pathways meet. The team’s simulations showed accurate agreement with observational data, and in the process, it was able to create the largest computational model to-date of Coma cluster and have plans to use that model as the basis to dive deeper.
HPC is a vital tool in piecing together the origin of the Coma cluster
After successfully modelling the Coma cluster in unprecedented detail, the team feels confident it now has a basis model that can be used to further study specific aspects of the Coma cluster. Moving forward, the team indicated that it would focus on using the model to focus on even higher redshift snapshots—redshift being when light and other waves get longer. Adding more details into its simulations will prove computationally challenging, but will allow the team to better understand how the Coma cluster became connected to the greater cosmic web and how specific connections have influenced the formation, evolution, and growth of the cluster.
In the interest of efficiency, the team works closely with LRZ staff to improve the memory demands from its simulation, working on methods to reduce the amount of memory-per-processor needed to run its simulation. Sorce is also working closely with collaborators in France to evaluate how artificial intelligence methods can benefit and expand the scope of the team’s research in the years to come.
-Eric Gedenk
Funding for SuperMUC-NG was provided by the Bavarian State Ministry of Science and the Arts and the German Federal Ministry of Education and Research through the Gauss Centre for Supercomputing (GCS).