Neutrinos from the Milky Way
- Allgemein
Seen with the naked eye, the Milky Way is a milky white band that stretches once across the entire sky. Using instruments other than our own eyes, we find that our own galaxy looks different in every wavelength range. For scientists these differences are immensely important, because in this way information about the structure of the Milky Way can be found. However, all previous observations of the Milky Way have one thing in common: they use photons for observation. But this does not have to be the case. Scientists at TU Dortmund University, together with their international colleagues, have demonstrated that the Milky Way can also be observed with the help of other particles: neutrinos.
The origin of cosmic rays is a mystery that has been driving physics for more than 100 years and around which a relatively young scientific subfield, the so-called astroparticle physics, has developed. To solve this mystery, physicists rely on various messenger particles, including high-energy gamma rays and neutrinos.
"The Milky Way is a known source of high-energy gamma rays and also contains potential sources of cosmic rays," says Professor Wolfgang Rhode, head of the TU Dortmund University research group leading the discovery. "Since gamma rays and neutrinos can be produced in the same physical processes, the Milky Way and in particular the galactic plane have long been a potential emitter of high-energy neutrinos."
However, the detection of these astrophysical neutrinos is anything but trivial and the list of challenges is long. First, there is the low interaction probability of the neutrinos, which must be collected by very large detectors. One of these detectors is the IceCube neutrino observatory, which has a volume of one cubic kilometer instrumented with more than 5000 photosensors. Second, there is the overwhelming background of atmospheric muons and neutrinos. IceCube, a neutrino observatory at the South Pole with an instrumented volume of one cubic kilometer, records 100 million muons from Earth's atmosphere for every astrophysical neutrino. In addition, most of the neutrinos are expected to come from the Galactic Plane direction from the southern sky, further complicating the background reduction.
"In order to be able to successfully pursue this novel way of observing the Milky Way nevertheless, one needs methods from the field of artificial intelligence. Deep neural networks, for example," says Dr. Tim Ruhe, astroparticle physicist and data science expert at TU Dortmund University. "Neural networks, once trained, have a comparatively short and, above all, stable runtime and can therefore be used in the analysis at an early stage." A distinct advantage over other methods. The AI methods used were developed over a period of more than ten years, including in a project of the Collaborative Research Center 876 led by Katharina Morik, Wolfgang Rhode and Tim Ruhe.
In the context of neutrino analysis, they have been used very successfully for reconstructing the energy and direction of the particles, among other things. "Due to the improved methods, we have about 20 times more neutrinos with an angular resolution improved by a factor of two in our final data set. For the analysis, this means that we are about three times more sensitive," said Mirco Hünnefeld, a doctoral student at TU Dortmund University and one of the scientists responsible for the success of the study.
"We have tried many different algorithms over the last twelve years," confirms Prof. Katharina Morik. "Always with the goal of detecting as many neutrinos as possible. Finally, a neural network gave us an analysis with really impressive precision."
All these successes are not yet sufficient for an observation of neutrinos from the Milky Way. The majority of the events in the final data set (about 84%) consist of atmospheric neutrinos. In order to still be able to extract a signal, statistical methods are used at this point to estimate the background based on the data itself and avoid potential uncertainties in the modeling. Next, the probability that the observed signal is a random fluctuation is determined. This is less than 0.001% in this case.
Although the events currently available are not sufficient to identify individual Galactic sources as neutrino emitters, the AI-assisted search for neutrinos from the Galactic plane direction has been extremely successful. The present results confirm that our understanding of the Milky Way, particularly with respect to cosmic ray propagation, is correct. And they open a new observational window into our own galaxy.