Search for hidden new physics signals at the Large Hadron Collider

The Standard Model (SM) of Particle Physics is currently the best theory to describe Nature at subatomic scales. However, despite its remarkable agreement with the data collected so far at colliders, the SM leaves several phenomena unexplained.

Searches for new physics have been carried out at the Large Hadron Collider (LHC) in the attempt to identify a more complete version of the SM, but those efforts have not yet produced the desired result. We cannot rule out the possibility of new physics searches being missing signals whose experimental signatures are more elusive than expected.

Thus, in this thesis, we study new physics signals that might be hidden in the LHC data, as well as SM extensions where those signals are predicted. In particular, we address multiboson production from a heavy Z ′ resonance in the framework of an U(1)′ extended next-to-minimal two-Higgs doublet model (UN2HDM).

Small excesses around 95 GeV observed in many searches for a new scalar in three different decay channels, γγ, τ τ and bb, motivate one further look into UN2HDMs. In this context, we check which anomalies can be explained by those models, including scenarios where excesses in one or two decay channels turn out to be statistical fluctuations.

The second goal of this thesis is to develop tools that can make new physics searches sensitive to a broader range of signals, including those with non-conventional experimental signatures. To this end, we introduce the concept of Mass Unspecific Supervised Tagging (MUST) for the identification of multi-pronged jets.

Using Machine Learning algorithms, namely Neural networks and Gradient Boosting, we show that among other benefits, jet tagging tools built upon MUST can discriminate many different types of multi-pronged jets in wide ranges of jet mass and transverse momentum.