Collider and Gravitational probes for New Physics in Multi-Higgs Models
PT-CERN Call 2020/2
Felipe Ferreira de Freitas
Universidade de Aveiro
CIDMA - Centro de Investigação em Matemática e Aplicações da Universidade de Aveiro
With the recent observation of gravitational waves (GW) by the LIGO-VIRGO collaboration and the Large Hadron Collider (LHC) RUN-3 scheduled to start at the middle of 2021 with twice the nominal luminosity and a center of mass energy of 14 TeV, a great opportunity for New Physics searches following a multi-messenger philosophy has emerged. In particular, models with extended scalar sectors feature the possibility of first order phase transitions (PT) capable of generating a stochastic background of primordial GW which, if detected, can become a gravitational portal and a probe for Beyond the Standard Model (BSM) physics. The LISA mission will offer a range of frequencies well into the mHz region, relevant for the electroweak (EW) PT. The non-trivial structure of such extended scalar sectors allows for multi-step phase transitions and corresponding GW signatures. This information, in synergy with collider physics, can offer complementary information on the quest for hints of a more complete description of particles and interactions. In this thesis we will consider models with additional electroweak Higgs doublets and singlets, supplemented by additional flavour or gauge symmetries, whose breaking takes place not far from the EW scale. We will investigate which particular BSM signatures are expected at the LHC for different benchmark models and how can this information be complemented by the stochastic background of GW, with focus on the future LISA mission. We will also consider models where vector-like fermions, extended neutrino sectors as well as composite Higgs scenarios will be studied along the same research strategy. These studies involve the development of intelligent algorithms based on Evolutionary searches for Deep Learning and Model Agnostic Meta-Learning (MAML). Such techniques can be used as a “control” environment for Monte-Carlo event generator software tools, such as MadGraph, Pythia and CosmoTransitions. One consequence of using these techniques is the possibility to accelerate the task of simulating data points for a given specific model. Simulations provided by MadGraph, Pythia and CosmoTransitions are rather CPU intensive and demand a fair amount of time to provide a significant statistics. One possible way to speed up the computation is to use Deep learning techniques, such as MAML, in order to capture the underlying features inherent of a given BSM scenario while leveraging such features to sample points more efficiently.