Reference : UMR5814-CORFEU-003
Workplace : ANNECY LE VIEUX
Date of publication : Tuesday, May 10, 2022
Scientific Responsible name : Pierre Aubert
Type of Contract : PhD Student contract / Thesis offer
Contract Period : 36 months
Start date of the thesis : 1 October 2022
Proportion of work : Full time
Remuneration : 2 135,00 € gross monthly
Description of the thesis topic
The experiment Cherenkov Telescope Array (CTA) will become, in a near future, the biggest gamma ray observatory in the world. Nowadays, the first prototype of Large Size Telescope (LST-1) is operational at La Palma site.
LAPP's effort, in the Array Control And Data Acquisition system, ACADA, allowed to develop and provide a real time analysis, based on simple and robust algorithms. This analysys is in deployment phase at La Palam site and treats data from LST-1 prototype, with few computing cores. In term, this algorithm will treat data from 13 to 19 telescopes at La Palama which represents about 17 GB/s and about 27 GB/s on the south site at Chile.
One mission of the real time analysis of CTA will consist to send alerts to multi-messenger observatories around the world. This analysis will be able to deal with fast repointing of LST telescopes, and evaluate as quick as possible is a recieved alert from another observatory is relevant and will be used to adjust telescopes pointing to ensure high quality detection. Alerts will concern mainly detection of transient phenomena such as Active Galactic Nuclei (AGN) ou gamma ray burst. A good sensitivity at low energy ables faster detection.
A reconstruction method, based on comparison of images from telescopes to pre-computed ones, improves a lot sensitivity at low energy. However, actual implementations are too slow and cannot be used in the real time analysis.
Position and Main Responsibilities
We therefore propose a PhD in computing sciences to optimise this reconstruction method to allow its use as real time analysis. Firstly, student will optimised the calssical algorithm with High Performance Computing (HPC) methods. Result will be tested with simulations and thank to LST-1 and with the future LST-2, 3 and 4, and also future Medium Size Telescopes (MST) data. Then, more powerful methods, based on HPC (hash tables, page rank, ect) will be developped and evaluated to allow real time analysis on the full site. Use of GPU can also be considered.
Theses algorithms would improve CTA sensitivity to extra galacit transient phenomena, and would allow better observation choice.
Such a method could also be used in long term to improve data volume reduction by selecting only relevant events.
Pattern search is quite common in high energy experiments, so inovative methods could be shared and reused in other context or experiments, such as HL-LHC, Virgo, Ligo, Dune, etc.
• team work
• presentation of work in international conferences
Mathematic knowledge :
• Krylov method / QR method / page rank
• Levenberg-Marquardt algorithm
Architectures knowledge :
• CPU (vectorization, cache level)
• GPU (architecture and programming models)
Programming languages :
• C++ (C++17)
• knoledge of python 3 is a plus
Compilation and program versionning :
• cmake 3
• g++-9 or newer
• git / gitlab
• knowledge of latest nvc++ functionnalities is a plus
he LAPP is a laboratory of the Institute of Nuclear Physics and Particle Physics (IN2P3), institute of the National Center for Scientific Research (CNRS) which coordinates programs in these fields. The LAPP is a joint research unit (UMR 5814) of the CNRS and the Université Savoie Mont-Blanc (USMB). More than 150 researchers, research teachers, engineers, technicians, administrators, students and foreign visitors work there. The research conducted at the LAPP aims to study the physics of elementary particles and their fundamental interactions, as well as to explore their links with the major structures of the Universe. The LAPP is involved in the experiments of the LHC (ATLAS, LHCb), neutrinos (SuperNémo, STEREO, LBNO, etc.), astrophysics (HESS, CTA) or gravitational waves (Virgo). The location of the laboratory 50 kms from CERN, and the presence of a theoretical physics laboratory in the same premises, make it a very attractive laboratory for all scientists in this field of research. The LAPP also houses an intermediate-sized mesocenter for computation and storage, but which contributes significantly to the data processing of large physics experiments such as ATLAS, LHCb or CTA.
We talk about it on Twitter!