Periodic Reporting for period 3 - PeVSPACE (Direct Detection of TeV--PeV Cosmic Rays in Space)
Período documentado: 2023-02-01 hasta 2024-07-31
Cosmic Rays are charged particles, mostly protons (85-90%), helium (5-10%), heavier nuclei (1%), and electrons (<1%), coming from space and produced in the most energetic events in our Galaxy and beyond, in particular in Supernovae explosions. The exact origin of Cosmic Rays remains a mystery up to now, more than a century since their discovery by Victor Hess in 1912. The absolute majority of cosmic rays have the energy of about a few hundred Megaelectronvolts and are deflected away by the earth's magnetic field. Yet a smaller portion of them can reach energies almost a billion times higher than those of the particles accelerated at the Large Hadron Collider. Comic Rays are the ultimate laboratory of the Universe study, in particular of the events driving the evolution of Galaxies, the composition of life of stars, the structure of the interstellar medium, chemical composition and generation of elements in stars. It is generally believed that Galactic sources are capable of accelerating Cosmic Rays up to about Petaelectronvolt, while everything with higher energy comes outside the Galaxy. For a long time, a conventional hypothesis dominated that Galactic Cosmic Rays are produced in the Supernovae Remnants (SNR) and their spectrum represents a simple power-law with a fixed spectral index. Recent direct measurements performed in the past decade by space missions challenge the conventional models, revealing peculiar structures in the spectrum of various cosmic ray components at the energy of about Teraelectronvolt (TeV). This is suggestive of a new source type of cosmic rays or an unknown acceleration or propagation effect. Further high-precision measurements of spectra of different cosmic ray components beyond the TeV scale are essential for solving the puzzle of Galactic Cosmic Ray origin and will also help to pinpoint the possible Dark Matter signatures, which may manifest themselves in the spectrum of Cosmic Rays, notably electrons. Such measurements, however, are hampered by the systematic uncertainties due to the limited accuracy of Cosmic Ray particle reconstruction and identification techniques, and the relatively low precision of hadronic Monte-Carlo simulation models.
The main objective of this project is to radically improve and optimize the techniques for Cosmic Ray detection at TeV—PeV energy region, including particle reconstruction, identification, and simulation, using a state-of-the-art Artificial Intelligence (AI) approach. As a result, Cosmic Ray spectra and composition will be measured first with the DArk Matte Particle Explorer (DAMPE) space mission and then subsequently with the next-generation space instrument — High Energy Radiation Detector (HERD), with unprecedented precision, which could not be achieved otherwise. There are two main innovations in the project. First, the application of AI techniques in astroparticle physics will be pioneered at the highest energies, in an unconventional use case. Second, the Monte-Carlo hadronic simulation models will be tested and tuned using the detector data, for the first time at such high energies.
The main objective of this project is to radically improve and optimize the techniques for Cosmic Ray detection at TeV—PeV energy region, including particle reconstruction, identification, and simulation, using a state-of-the-art Artificial Intelligence (AI) approach. As a result, Cosmic Ray spectra and composition will be measured first with the DArk Matte Particle Explorer (DAMPE) space mission and then subsequently with the next-generation space instrument — High Energy Radiation Detector (HERD), with unprecedented precision, which could not be achieved otherwise. There are two main innovations in the project. First, the application of AI techniques in astroparticle physics will be pioneered at the highest energies, in an unconventional use case. Second, the Monte-Carlo hadronic simulation models will be tested and tuned using the detector data, for the first time at such high energies.
Machine Learning Research and Development have been performed resulting in the creation of a novel particle track reconstruction algorithm based on the adaptation of Deep Learning techniques. Preliminary studies demonstrate significant performance enhancement of the algorithm with respect to the classical (standard) tracking approach in the analysis of the DAMPE data. This opens the possibility of performing particle identification of proton and helium Cosmic Rays at the highest energies accessible by space-borne detectors of current and next-generation, with uncertainty not exceeding a few percent. An article describing the developed track reconstruction algorithm is submitted to a journal, it is currently under the review process by external referees.
A method of Cosmic Ray Electron (CRE) identification based on the Deep Neural Net (DNN) technique has been developed in cooperation with DAMPE colleagues. The method provides up to factor 4 improvement of proton background rejection compared to the standard methods and makes possible a precise CRE spectrum measurement up to unprecedented energy of about 10 TeV. The work is published in the Journal of Instrumentaion (2021). Consequently, a preliminary measurement of CRE spectrum using the developed method has been performed and is under internal review by the DAMPE collaboration. Further Research and Development of a Convolutional Neural Net (CNN) approach using the images of cosmic rays in the detector have been performed by the team members. In the current implementation, the CNN approach shows partial enhancement over the DNN method, likely reaching the theoretical performance limit for a given type of detector to which it was optimized (DAMPE). At the same time, it shows very high potential and serves as a stepping stone for research and development of the electron identification technique for the future HERD mission, using the unique 3D particle calorimeter images, to be held in the next phases of the project.
Work is in progress on repeating Cosmic Ray proton spectrum measurement with the latest DAMPE data, using different hadronic simulation models and the developed track reconstruction algorithm. The goal is to significantly improve the accuracy of the hadronic models, substantially reducing the related systematic uncertainty, which is a dominant factor limiting the accuracy of the Cosmic Ray proton spectrum measurement. A unique interface is being developed for the integration of the state-of-the-art hadronic simulation models from the Cosmic Ray Monte Carlo (CRMC) package into the detector simulation toolkit, Geant4, used both in DAMPE data analysis and HERD development and optimization. In particular, this work is at the core of the Cosmic Ray helium spectrum measurement performed by the DAMPE collaboration, published in Physical Review Letters journal (2021), appearing as an Editor's highlight.
Results of the project were presented in oral contributions at multiple international conferences, including EPS-HEP 2021 (online conference), CORSIKA 8 Air-Shower Simulation and Development Workshop 2022, Heidelberg, Germany, Connecting The Dots 2022, Princeton, USA, and TeV Particle Astrophysics (TeVPA) 2022, Kingston, Canada.
A method of Cosmic Ray Electron (CRE) identification based on the Deep Neural Net (DNN) technique has been developed in cooperation with DAMPE colleagues. The method provides up to factor 4 improvement of proton background rejection compared to the standard methods and makes possible a precise CRE spectrum measurement up to unprecedented energy of about 10 TeV. The work is published in the Journal of Instrumentaion (2021). Consequently, a preliminary measurement of CRE spectrum using the developed method has been performed and is under internal review by the DAMPE collaboration. Further Research and Development of a Convolutional Neural Net (CNN) approach using the images of cosmic rays in the detector have been performed by the team members. In the current implementation, the CNN approach shows partial enhancement over the DNN method, likely reaching the theoretical performance limit for a given type of detector to which it was optimized (DAMPE). At the same time, it shows very high potential and serves as a stepping stone for research and development of the electron identification technique for the future HERD mission, using the unique 3D particle calorimeter images, to be held in the next phases of the project.
Work is in progress on repeating Cosmic Ray proton spectrum measurement with the latest DAMPE data, using different hadronic simulation models and the developed track reconstruction algorithm. The goal is to significantly improve the accuracy of the hadronic models, substantially reducing the related systematic uncertainty, which is a dominant factor limiting the accuracy of the Cosmic Ray proton spectrum measurement. A unique interface is being developed for the integration of the state-of-the-art hadronic simulation models from the Cosmic Ray Monte Carlo (CRMC) package into the detector simulation toolkit, Geant4, used both in DAMPE data analysis and HERD development and optimization. In particular, this work is at the core of the Cosmic Ray helium spectrum measurement performed by the DAMPE collaboration, published in Physical Review Letters journal (2021), appearing as an Editor's highlight.
Results of the project were presented in oral contributions at multiple international conferences, including EPS-HEP 2021 (online conference), CORSIKA 8 Air-Shower Simulation and Development Workshop 2022, Heidelberg, Germany, Connecting The Dots 2022, Princeton, USA, and TeV Particle Astrophysics (TeVPA) 2022, Kingston, Canada.
For the first time, a Deep Learning solution is being applied for particle track reconstruction in space astroparticle missions at the highest energies. The preliminary results demonstrate significant performance enhancement beyond the state-of-the-art of the existing tracking algorithms. This result promises a clear advancement in the accuracy and quality of cosmic ray measurement in the next years. After ongoing deep scrutiny and verification, the results are planned to be published in the course of the project, with the potential of becoming a new standard in the field.
A Deep Learning solution is elaborated for the Cosmic Ray Electron identification, demonstrating significant advantage with respect to the existing techniques at the highest energies of direct Cosmic Ray Electron detection, opening now prospectives for high-energy Cosmic Ray Electron measurements, which could not be achieved without it. Further development in the course of the project, in particular, the application of deep learning to 3D particle images in space missions, which was never attempted before, demonstrates the clear potential of advancing the field of particle detection techniques beyond the state-of-the-art.
The project team is leading the development of a tool that solves a long-standing problem of particle simulation at energies beyond those of the Large Hadron Collider, with a significant impact and contribution to high-energy physics and astroparticle physics communities. The tool is part of the project research aimed at optimization and tuning of hadronic simulation models at the highest energies.
A Deep Learning solution is elaborated for the Cosmic Ray Electron identification, demonstrating significant advantage with respect to the existing techniques at the highest energies of direct Cosmic Ray Electron detection, opening now prospectives for high-energy Cosmic Ray Electron measurements, which could not be achieved without it. Further development in the course of the project, in particular, the application of deep learning to 3D particle images in space missions, which was never attempted before, demonstrates the clear potential of advancing the field of particle detection techniques beyond the state-of-the-art.
The project team is leading the development of a tool that solves a long-standing problem of particle simulation at energies beyond those of the Large Hadron Collider, with a significant impact and contribution to high-energy physics and astroparticle physics communities. The tool is part of the project research aimed at optimization and tuning of hadronic simulation models at the highest energies.