Top to bottom: Maps of the Milky Way galaxy based on low-energy radio waves, visible light, high-energy gamma rays, and the new map based on neutrino emissions.

Simulations on PSC’s Bridges-2 System Help Identify Signals in Huge Antarctic Ice Sheet Neutrino Detector

We’ve learned a lot about the Universe from telescopes that are sensitive to anything from high-energy gamma waves to visible light to low-energy radio waves. But detecting such electromagnetic waves has limitations. Using PSC’s Bridges-2 system to simulate signals in their Antarctic IceCube detector, an international collaboration of scientists has now made the first map of the Milky Way galaxy using particles called neutrinos — the first map of a cosmic structure that didn’t depend on electromagnetic waves.


As our technology has progressed, we have devised new and more ingenious ways to observe and measure the Universe. Telescopes let us see objects in visible light; radio telescope dishes let us see new objects, as well as different behaviors by previously seen objects. Space launches allowed telescopes to have unprecedented clarity, as well as to see through opaque clouds using infrared light. Each of these leaps in technology literally opened new worlds for us. But they all detected electromagnetic waves, which can be distorted, absorbed, and generally scrambled by stuff in between us and what we’re trying to look at.

The first detection of gravitational waves in 2016 completely changed that. It represented a completely new way of looking. A year later, the IceCube Observatory in Antarctica made an equally momentous detection: the first pinpointing of an object out in space using weird particles called neutrinos. We now had three “messengers” to probe the universe with, each telling us different things about the objects that produced them.

“The original point [for IceCube] was this phenomenon called cosmic rays. [Scientists] discovered them over 120 years ago. But we had no idea where they were coming from … They don’t travel in straight lines. They’re being deflected so we can’t really point back to the sources. And then other messengers like gamma rays get absorbed [by] dust … So at the longest distances and highest energies anything from radio out to the gamma rays is being absorbed. It’s basically dark to us.” — Benedikt Riedel, University of Wisconsin

The IceCube Collaboration scored several firsts. First localization of a source of cosmic neutrinos. With colleagues using traditional telescopes, first co-detection of neutrinos and electromagnetic signals from a neutron star, pinpointing a source of cosmic rays. Simulations on PSC’s supercomputers helped them prepare for these discoveries. For their next step, the team wanted to take their revolutionary detector to a new level. They wanted to map the entire Milky Way galaxy. If successful, it would be the first cosmic map that didn’t depend on electromagnetic waves.

To make this happen, they once again turned to PSC, and the center’s Bridges-2 supercomputer.


To understand how PSC’s NSF-funded, ACCESS-program-allocated Bridges-2 supported IceCube’s work, you first must understand a little about neutrinos.

Neutrinos have mass, but just barely. They also have no electrical charge. So unlike the particles that make up normal matter, they’re what physicists call “weakly interacting.” Neither gravity, electrical charge, nor magnetic fields have much of an effect on them. Because of that, they rarely interact with matter. Right now, 100 trillion neutrinos are passing through your body every second. But if you live to be 80 years old, on average only one of them will have interacted with the matter in your body.

The IceCube neutrino detector, then, had its work cut out for it. Because such an incredibly tiny fraction of neutrinos interacts with matter, the scientists who designed IceCube had to put an immense amount of matter in the detector. They hit on the idea of taking roughly a cubic kilometer of Antarctic ice and drilling it to insert hundreds of detectors, sensitive to the blue Cerenkov radiation light expected from these rare collisions.

First, though, they had to work through a bunch of challenges. In theory, a neutrino could create a line of light as it crashed through the ice, allowing the detectors’ positions and times of detection to trace that line back to the neutrino’s cosmic source. But sometimes, the detection is more of a sphere. The scientists would also have to screen out detections due to backgrounds coming from cosmic ray interactions in the atmosphere. They’d also need to tell the difference between cosmic neutrinos from the Milky Way and ones from other sources.

“We do a lot of simulations. We take an idealized image of our detector and we say, ‘This is the response of our detector to this particle in this interaction.’ We simulate a response and then we compare that with our data … Where Bridges-2 comes in is [that] it simulates the light moving through the south polar ice coming from the neutrino interactions on Bridges-2’s GPUs, and then the spare CPU cycles can be used for anything from data analysis to particle generation.” — Benedikt Riedel, University of Wisconsin

Benedikt Riedel at the University of Wisconsin, a leading scientist in the IceCube Collaboration, oversaw the use of several systems to simulate how imperfections in the ice would affect the patterns of detection. Bridges-2 proved particularly adept at these simulations. Its ability to offer both powerful central processing units, or CPUs, and late-model graphical processing units, or GPUs, helped untangle the crazy particle showers expected, to show how they related to neutrinos passing through the ice sheet. The collaborators also used the large Frontera supercomputer at the Texas Advanced Computing Center, PSC’s partner in the ACCESS network of NSF-funded supercomputers.


Map of the Milky Way galaxy with the new neutrino map superimposed on a visible-light map.

Map of the Milky Way galaxy with the new neutrino map superimposed on a visible-light map.

Thanks in part to Bridges-2, the team was able to identify what patterns of detector activations in IceCube came from real cosmic neutrinos. The result was a map of our galaxy — the first such map using a new messenger other than electromagnetic waves. While the map is admittedly crude compared with the exquisite maps produced by visible-light- and infrared-detecting space telescopes, it provides the first opportunity to compare what the galaxy looks like using independent messengers. The team reported their results in the prestigious journal Science in July 2023.