Thesis

A next-generation gamma-ray observatory powered by machine learning techniques

• Call:

  PT-CERN Call 2020/2

• Academic Year:

  2020/2021

• Domain:

  Astroparticle Physics

• Supervisor:

  Ruben Conceição

• Co-Supervisor:

  Alberto Guillén

• Institution:

  Instituto Superior Técnico (Universidade de Lisboa)

• Host Institution:

  Laboratório de Instrumentação e Física Experimental de Partículas

• Abstract:

  The observation of the high-energy gamma-rays provides a unique window to explore the extreme energy universe and fundamental phenomena related to it. The observation of these rare phenomena implies the construction of huge compact experiments, with areas of 1 km^2, placed at altitudes that surpass the 5000 meters above sea level. These experiments should be able to fight the enormous hadronic background to observe the gamma-rays. Classically, this has been done by identifying muons, which led to detector options which are unfit for such an ambitious project. New ideas to build a next-generation experiment are currently being assessed by the Southern Wide-field Gamma-ray (SWGO) collaboration. At LIP, we have shown that the application of Machine Learning algorithms together with innovative detector solutions, can lead to a smaller, more capable next-generation detector which could revolutionize the field of gamma-rays. In particular, the time evolution of the detector response can be exploited to identify the nature of the particle entering the tank. While very promising, the implementation of these neural networks has many challenges ahead: the complexity of the problem and necessity to perform features engineering; use of simulations with incomplete accurate descriptions; resilience to spurious effects (backgrounds)... This work will be developed in close collaboration with the Computer Technology and Architecture department of Granada University.