Identificação de pares de fótons utilizando informações aneladas de calorimetria para a busca por ALPs no ATLAS

Abstract

Despite all the success of the Standard Model (SM) in particle physics, several phenomena remain unresolved, among them, it is worth mentioning, the baryonic asymmetry, the nature of the dark matter and the strong CP problem. These questions raise the idea that any deviation between the SM predictions and the experimental measurements opens room for new physics contributions. In this sense, less established particles such as the Higgs boson can have an important role in the relationship the between ordinary and the matter which is uncharged for the SM interactions. For extended models, such as Two-Higgs doublet, 2HDM, new analyses for the exotic decays of the MP Higgs boson, in lighter states, can be studied at the LHC, especially with axion-like particles, and their generalizations, which can be considered as dark matter candidates. Although well motivated, the study of the decay of light scalars into photon pairs is affected by the angular resolution of the detectors. For light particles with a large boost factor (βa ≈ 1), photon production occurs with a small opening angle, so the granularity of the ATLAS detector cells does not allow separation into two objects. To deal with the limitation on the resolution, techniques that involve the description of the lateral and longitudinal developments of the showers of these photons are useful. The technique of deposition of energy in rings, or annealing technique, exploits the conical structure of the shower development, to provide information regarding the characteristics of the photons that produce these showers. This work is being developed in collaboration with the H/γ ATLAS group, led by Prof. Bertrand Laforge from the LPNHE (Laboratoire de Physique Nucléaire et de Hautes Énergies). The project aims to develop a tool based on Machine Learning techniques for the identification of photon pairs, using information from the rings of the two most energetic cells of a given cluster, such as input. For the preparation of the two sets, the rings building algorithm, stored in the reconstruction section of the Athena framework, was adapted for this purpose. For this study samples of isolated photons, and from the decay of a π⁰ meson were used, in order to “simulate” the decay of a light particle, such as ALP, in a photon pair. In this study, the performances of different models of AutoEncoders, neural networks developed for reconstruction tasks, were evaluated. The two seeds approach and the application of neural networks can be used to improve the photon and jet reconstruction algorithms in the ATLAS Experiment. It can be also used for analyses with decays of new particles in photon pairs, following the current studies involving the detection of dark matter in accelerators.