Figure 1: Contributions to the AMS computing efforts in the context of the search for antihelium by country. 
Figure 2: Picture of the International Space Station. The Alpha Magnetic Spectrometer (AMS) is a detector designed for precision spectroscopy of cosmic rays that was installed on the International Space Station (ISS) in May 2011 (Fig. 2). With dimensions of 5 × 4 × 3m3 and a weight of 7.5 tons, AMS is the largest cosmic-ray spectrometer ever built. Its construction began in 1995, and a successful prototype flight (AMS-01) aboard the Space Shuttle Discovery proved the feasibility of the detector concept in 1998. Led by Nobel laureate Professor Samuel Ting from MIT, AMS has been constructed and is now operated by an international collaboration of more than 200 scientists and engineers, from Europe, America and Asia. The overall construction costs, including the flight of AMS to the Space Station aboard Space Shuttle Endeavour, have amounted to 1.5 billion US dollars. AMS is the only magnet spectrometer in space and the largest instrument for basic research on the ISS. In Germany, RWTH Aachen has been strongly involved in the AMS project since its inception. One of the main components of AMS, the transition radiation detector (TRD), has been designed and constructed by the I. Physikalisches Institut B under the direction of Professor Stefan Schael. Today, the Aachen group, comprising 20 scientists and students, plays a major role in the analysis of the data gathered by AMS and in the operation and calibration of the instrument. Since their discovery in 1912, cosmic rays have held many surprises in stock for us, from the discovery of new elementary particles to the most violent processes taking place in the Universe and accelerating cosmic rays to enormous energies. As a multi-purpose instrument for the precision spectroscopy of cosmic rays, AMS was conceived to answer fundamental questions about our Universe: What is the nature of Dark Matter? What happened to the antimatter that must have been produced in the Big Bang? Where are cosmic rays accelerated and how do they propagate through the Milky Way? Answers to these questions will have a profound impact on our understanding about the inner workings of our Universe and help advance fundamental science. In particular, the search for dark matter complements the search for new elementary particles at the Large Hadron Collider (LHC) at CERN, Geneva. AMS has recorded over 240 billion individual particle crossings (called “events”), more than all previous cosmic-ray experiments combined. The raw data volume collected is on the order of 40 TB per year. AMS employs redundant subdetectors for particle identification and for energy or momentum measurements: the TRD, an electromagnetic calorimeter (ECAL), a ring-imaging Cherenkov counter (RICH), a silicon tracker and a time-of-flight system (TOF). Before any physics analysis of the data can be performed, the information from all these subdetectors has to be pieced together and complicated reconstruction algorithms have to be run for each of them. The resulting high-level data serves as the input for physics analyses and occupies a volume of 160 TB per year of AMS flight on disk. HPC resources are vital for the processing, calibration, and analysis
of this enormous dataset.
Project Details
Project term
May 1, 2022–December 31, 2024
Affiliations
RWTH Aachen University
Institute
I. Physikalisches Institut B
Principal Investigator
Methods
Monte Carlo simulations are an essential ingredient of the physics analyses of AMS data. The interactions of cosmic-ray particles entering AMS with the detector components are modelled together with the detector responses that convert energy depositions in the sensitive material to digitized output that has the same format as the event records stored by AMS for actual cosmic rays. The simulations are used to study reconstruction algorithms, extract effective acceptances for measurements of cosmic-ray fluxes, and determine the background levels due to misidentification or misreconstruction of particles. The contributions of the main AMS computing centers to the production of helium simulations used to study the charge confusion background in the search for antihelium are shown in Fig. 1. Thanks to the support of NHR4CES/CLAIX and Jülich
Supercomputing Center, the German contribution was 50 %.
Results
Over twenty-five publications from the AMS collaboration have appeared in the renowned Physical Review Letters, most of which have been selected as an Editor’s suggestion. The findings have received considerable attention among astrophysicists and triggered an enormous amount of theoretical work. The physics highlight of AMS in 2024 was the detailed study of cosmic-ray deuterons. Instead of being accelerated in supernova remnants like primary cosmic rays p and 4He, deuterons are conventionally thought to overwhelmingly originate from interactions of helium with the interstellar medium. But AMS has discovered that the cosmic-ray deuteron flux has a distinctly different rigidity (momentum per charge) dependence than that of (secondary) 3He, and instead the ratio of the deuteron flux to that of the (primary) proton flux becomes constant with rigidity at high rigidities. These unexpected observations indicate that cosmic deuterons have a sizable primary-like component.
Discussion
For the antihelium search, a main focus of our computing projects for many years, we have been able to produce a substantial amount of simulations in the latest version of the AMS detector description and reconstruction, in particular using an improved track fit based on the general broken lines algorithm, along with a vastly improved tracker alignment. We have also explored the potential of using Deep Learning methods for AMS event reconstruction. We first looked at the particle identification and energy regression with the ECAL because major improvements in these areas would allow extending the energy range of the positron flux. Using simulations of electrons and protons produced within our computing project, we have successfully trained convolutional neural networks (CNNs) for both tasks. Finally, our understanding of cosmic-ray physics benefits from new insights from particle physics obtained in the context of our computing project: we have presented the world’s first experimental limits on the Λb → 3He+X branching ratio. They already exclude the prediction by models created to explain the observation of antihelium candidates with AMS by nearly two orders of magnitude and significantly restrict the 3He abundance in cosmic rays.
Additional Project Information
DFG classification: 311 Astrophysics and Astronomy
Software: ROOT framework, Geant4
Cluster: CLAIX
Publications
Miguel Aguilar, et al,
Properties of Cosmic Deuterons Measured by the Alpha Magnetic Spectrometer,
https://doi.org/10.1103/PhysRevLett.132.261001
Miguel Aguilar, et al,
Solar Modulation of Cosmic Nuclei over a Solar Cycle: Results from the Alpha Magnetic Spectrometer,
https://doi.org/10.1103/PhysRevLett.134.051001
Miguel Aguilar, et al,
Antiprotons and Elementary Particles over a Solar Cycle: Results from the Alpha Magnetic Spectrometer,
https://doi.org/10.1103/PhysRevLett.134.051002
LHCb Collaboration, Antihelium production in decays,
https://cds.cern.ch/record/2905862
Thesis:
Robin Sonnabend,
Search for Antihelium Nuclei in Cosmic Rays with the AMS-02 experiment on the International Space Station,
PhD thesis, RWTH Aachen University, November 2023,
https://publications.rwth-aachen.de/record/973566
Hanna Meuten,
Untersuchung der zeitlichen Stabilität des Übergangsstrahlungsdetektors von AMS-02,
BSc thesis, October 2024
Benjamin Gudisch,
Zeitliche Stabilität des Elektronen-Flusses in der kosmischen Strahlung,
BSc thesis, August 2024
In cooperation with the Computer Graphics Group at RWTH (Prof. Kobbelt), we have performed a MSc thesis project investigating the use of Deep Learning autoencoders for particle identification in AMS:
Leonard Schramm,
Particle Identification with Autoencoders for the AMS-02 Experiment on the International Space Station,
MSc thesis, July 2024.