Beyond the Standard Model at the Large Hadron Collider : from Simplified Model Constraints to Artificial Proto-Modelling for Dispersed Signals

Overview & Synopsis

The Standard Model (SM) of particle physics is undoubtedly one of the most successful models of modern physics. Nevertheless, many experimental and theoretical questions remain open and hint at the existence of beyond the Standard Model (BSM) physics. These questions include, for instance, the nature of dark matter and the hierarchy problem. Several experiments of various scales are currently searching for a manifestation of new physics. The ones carried out at the Large Hadron Collider (LHC) benefit from today's highest energy delivered by a particle accelerator. However, despite the intensive search program carried out by the LHC experiments, no compelling signal for new physics has been found.In the absence of any discovery, the obtained results are used to set constraints on a wide variety of BSM theories. One of these theories, studied in this thesis, is the minimal supersymmetric extension of the SM (MSSM), where each SM field has a supersymmetric (SUSY) counterpart differing by half a spin unit. The constraints set by the LHC experiments on the masses of SUSY particles are quite stringent. However, these limits are derived through a channel-by-channel approach and most of the time hold only in the context of simplified models, where only a handful of SUSY particles, with trivial mixings and decay branching ratios, are assumed to be kinematically accessible. Therefore, it is necessary to globally reinterpret these results to correctly estimate how they constrain realistic BSM scenarios.The approach adopted in this thesis is based on the reinterpretation tool smodels, which uses simplified model results to constrain more realistic models. The work of this thesis led to the extension of the smodels framework in several ways, on both the code and the database side. The most notable new feature is the combination of approximately uncorrelated analyses, enabling the confrontation of models against different experimental results in a consistent way through global likelihoods.In an exemplary phenomenological application, this is used to set global LHC constraints on the electroweak-ino sector of the MSSM through the dynamical combination of 16 LHC searches for which efficiency and acceptance maps are available in the smodels database. Particular attention is given to characterising which combinations maximise the sensitivity in different regions of the parameter space, how fluctuations in the data in individual analyses influence the global likelihood, and what is the resulting exclusion power of the combination compared to the analysis-by-analysis approach.Another objective of this thesis is to go beyond the top-bottom approach usually adopted in BSM searches. Instead of testing each theory against the data, the data is used to guide the search for BSM physics. Signs of new physics are here searched for by constructing so-called proto-models, a collection of simplified models not bounded by any higher-level theoretical assumptions, and by confronting them against the results in the smodels database. This relies on the dynamical combination of LHC searches to find mutually consistent excesses from dispersed signals while being consistent with their current limits. This thesis presents an improved version of this method, with a more furnished database, a refined statistical treatment, and the possibility to probe off-shell regions of the parameter space. A first application to the LHC searches for SUSY signatures from the production of electroweak-inos with compressed mass spectra is also discussed.