Testing the nature of light and ultralight dark matter with deep learning
PT-CERN Call 2022/1
Felipe Ferreira de Freitas
Universidade de Aveiro
CIDMA - Centro de Investigação em Matemática e Aplicações da Universidade de Aveiro
This thesis proposes a synergy between the areas of strong gravity, cosmology and particle physics in order to explore the Dark Matter (DM) paradigm. The deep learning techniques necessary for these studies will be based on evolutionary optimization algorithms and computer vision methods recently developed both in the context of particle and gravitational wave identification. Contrary to the typical approach taken by the strong gravity community, this thesis aims at the construction of concrete particle physics models where ultralight DM particles can emerge either as pseudo-Nambu-Goldstone bosons or as vector bosons such that they can reveal themselves in the form of stable halos around compact astrophysical objects. However, such halos are only viable when the Compton wavelength of the field approximately matches the gravitational scale of the compact object. For the known astrophysical objects, such as black holes, whose mass ranges from 1-10¹⁰ solar masses, this means that such bosonic particles must have masses between 10⁻¹⁰-10⁻²⁰eV. The presence of an ultralight bosonic sector implies further bosonic particles that are not necessarily ultralight, and perhaps at the reach of the LHC if their mass is larger that 1 GeV. Furthermore, for models featuring neutrino mass generation mechanisms, a sterile neutrino can also become fermionic DM candidate with masses in the keV-MeV range. In this thesis, a catalog of models that can consistently predict light and ultralight bosons as well as sterile neutrinos will be constructed. These are extraordinary candidates to offer multi-component DM which will be studied and confronted with data both from the gravitational, collider, and direct detection channels.