Figure 1: Artistic visualization of a heavy cosmic nucleus, originating from a far distant cosmic source, reaching Earth and interacting with its atmosphere. Image generated using Microsoft Designer
Using artificial intelligence (AI), researchers have unlocked novel insights into the mass composition of ultra-high-energy cosmic rays, the most energetic particles found in nature, paving the way for a deeper understanding of their origins and the extreme environments that birth them. The article has been published in the most recent issue of Physical Review Letters.
Cosmic rays are atomic nuclei, ranging from the lightest, hydrogen, to heavier elements like iron. Understanding their composition is crucial for piecing together the puzzle of where they come from and how they are accelerated to such incredible energies. Scientists at the Pierre Auger Observatory have made a significant breakthrough in studying ultra-high-energy cosmic rays, the high-energetic end of these mysterious cosmic particles bombarding Earth from the depths of space that reach energy unreachable with human-designed particle colliders.
For decades, scientists have relied on fluorescence telescopes to measure the mass composition of the cosmic particles. The instruments capture the faint fluorescent light emitted as cosmic rays interact with Earth's atmosphere, and induce giant particle cascades, so-called air showers. The fluorescent light reveals the depth at which the shower reaches its maximum number of particles, known as Xmax, which is a key indicator of the cosmic ray's mass. Heavier nuclei produce showers that reach their maximum higher in the atmosphere, while lighter nuclei, like protons, penetrate deeper, like a bowling ball and a golf ball dropped into water. The bowling ball makes a big splash higher up compared to a golf ball. Regardless of their precision, fluorescence telescopes have a limitation: they can only operate on clear, moonless nights, resulting in a small duty cycle.
The AI revolution: shedding new light on the mass composition of cosmic rays
To overcome this obstacle, scientists turned to the power of artificial intelligence. They trained a deep neural network (DNN), the working horse of AI, to analyze data from the Surface Detector (SD) of the Pierre Auger Observatory. This vast array of detectors, spread over 3,000 square kilometers, continuously monitors cosmic ray showers, regardless of the time of day or weather conditions.
To leverage the algorithms effectively, a central pillar in the analysis was the calibration of the algorithm using data from the fluorescence observations to set the absolute scale of the shower maximum, which can be thought of as tuning an instrument – the AI – to the correct pitch, accurately determined by the Fluorescence Detector (FD). The result is an unprecedented tenfold increase in the amount of data available for studying cosmic rays and their mass composition, equivalent to 150 years of running the observatory's FD.
This new data has revealed a fascinating picture of the cosmic ray mass evolution. At lower energies, the composition is light and mixed, consisting of various light atomic nuclei like protons or helium nuclei, confirming previous measurements with fluorescence telescopes. However, as energy increases and the data measurable with the telescope becomes less and less, the composition shifts, becoming heavier and purer. This implies that the most energetic cosmic rays we detect are heavier nuclei, likely a combination of elements like nitrogen or iron, corroborating the Collaboration's earlier findings that challenge the long-held assumption that nature's most energetic particles are mostly protons with high statistics.
Figure 2: Energy evolution of the mass composition (top) and flux of cosmic rays (bottom). The grey regions indicate the found breaks in evolution, raising exciting questions about the nature of cosmic rays.
Top: Composition measurements using fluorescence observations are shown as grey markers and compared to the new measurement (black points, top). The red (blue) lines indicate the expected measurement for a pure proton (iron) composition.
Bottom: Cosmic ray flux as measured using the SD of the Pierre Auger Observatory.
The evolution from a mix of light nuclei to a purer composition of heavy nuclei appears to follow a characteristic structure with three breaks. Remarkably, these breaks are found to be in proximity to features in the energy spectrum, which quantifies the energy distribution of measured cosmic rays. This could tell us that different sources and acceleration mechanisms may dominate at different energy ranges, opening new prospects to investigate various theoretical models.
A collection of both measurements is shown in Figure 2. The upper plot shows the evolution of the composition using the shower maximum Xmax with increasing energy. The lower plot shows the energy spectrum, indicating the number of particles measured at a specific energy. The grey bands show the regions where features have been identified. Their proximity can be easily recognized.
These findings reveal a more complex picture of the cosmic ray composition than previously understood, raising intriguing questions about the mechanisms that accelerate these particles to such extreme energies and the nature of the sources that produce them.
Towards solving the ultra-high-energy cosmic ray puzzle
The novel data set provided by the improved reconstruction algorithm is a cause for excitement about the future of cosmic ray research. By studying their anisotropy, scientists search for variations in the arrival directions of the cosmic rays, pointing towards regions in the sky to locate their origin. However, the charge of UHECRs causes them to be deflected by magnetic fields in our galaxy, making it challenging to trace them back directly to their sources, which have been unknown for more than a century.
With the new extra information, scientists can now combine the precise energy, arrival direction, and composition data to probe astrophysical models and search for the sources of the most energetic particles in our universe, for example, by selecting particles with low charge and high energy, least affected by magnetic fields.
The AugerPrime era
The AugerPrime detector upgrade involves adding scintillator detectors and radio antennas to the SD array and will shape the next decade in ultra-high-energy cosmic ray research. The upgraded instrument will enhance the sensitivity to the mass composition of UHECRs by allowing for simultaneous measurements of electrons and muons. Combining the power of the upgraded observatory with innovative algorithms as developed in the recent study will unlock a new era for examining these extreme particles, promising a deeper understanding of their origins and the extreme astrophysical environments that produce them.
Learn more:
Listen to our podcast discussing the recent findings on the cosmic ray composition.
(The podcast has been created using notebookLM and serves as an entertaining way to learn more about the research in a popular scientific way only. Therefore, it has only limited scientific accuracy. For details, please refer to the publications.)
Figure 3: Artistic representation of an astrophysical source emitting ultra-high-energetic cosmic rays into the cosmos.
Image generated using Microsoft Designer
Related papers:
Inference of the Mass Composition of Cosmic Rays with energies between 3 and 100 EeV using the Pierre Auger Observatory and Deep Learning
The Pierre Auger Collaboration, Phys. Rev. Lett. 134 (2025) 021001
[arxiv.org/abs/2406.06315] [doi: 10.1103/PhysRevLett.134.021001]
Measurement of the Depth of Maximum of Air-Shower Profiles with energies between 3 and 100 EeV using the Surface Detector of the Pierre Auger Observatory and Deep Learning
The Pierre Auger Collaboration, Phys. Rev. D 111 (2025) 022003
[arxiv.org/abs/2406.06319] [doi: 10.1103/PhysRevD.111.022003]
Disclaimer
The images in Figures 1 and 3 have been AI-generated using Microsoft Designer, which is licensed for non-commercial use. The podcast has been AI-generated using https://notebooklm.google.com.
