Electron studies and search for vector boson scattering in events with four leptons and two jets with the CMS detector at the LHC

This thesis reports the first experimental investigation into vector boson scattering (VBS) in the ZZ channel, where both Z bosons are required to decay into electrons or muons and are accompanied by at least two hadronic jets ($ZZ\mathrm{jj}\rightarrow\ell\ell\ell′\ell′\mathrm{jj}$, where $\ell, \ell′ = e$ or $\mu$). VBS is a key process for elucidating the physics of electroweak symmetry breaking and the role of the recently discovered Higgs boson. This study analyses $35.9\,\mathrm{fb} ^{−1}$ of proton–proton collisions collected with the CMS experiment at the CERN Large Hadron Collider at a center-of-mass energy of 13 TeV. A multivariate analysis (MVA) technique is exploited to separate the electroweak signal from the QCD irreducible background and to measure the signal strength $\mu$, i.e., the ratio of the observed number of events to the standard model expectation. The observed signal strength is $\mu = 1.39^{+0.86}_{−0.65}$ which excludes the background-only hypothesis at 2.7 standard deviations (1.6 standard deviations expected). Limits on physics beyond the standard model are derived in terms of anomalous quartic gauge couplings in the effective field theory approach, providing the most stringent constraints to date on the couplings for the operators T8 and T9. The ZZjj VBS analysis requires an accurate modeling of the signal and irreducible background processes, going beyond the existing simulations. Extensive work on generating and comparing the theory predictions from several Monte-Carlo event generators is presented. The detailed understanding of the signal and background kinematics is used to develop and systematically optimize a boosted decision tree (BDT) classifier. A matrix element discriminant is also developed and its classification performance compared to the BDT, finding comparable performance and indicating that the BDT is adequate. The signal extraction technique via a template fit of all ZZjj events also permits to constrain the normalization of the QCD background using the data. Multilepton analyses like the search for VBS in the ZZ channel depend on the ability to efficiently reconstruct and identify the final state leptons. This work presents the optimizations of the multivariate electron identification algorithms used in the first data at 13 TeV in 2015. A study on extending the use of tracking information in the MVA resulted in the reduction of the non-prompt electron background by up to 50 %. Monitoring the changes to the reconstructed electron objects and continuous optimizations allowed to improve or maintain the performance of the electron MVA ID algorithms, despite the harsher pileup conditions in the 2016 data. The electron efficiency measurements performed for the 2016 multilepton analyses in CMS are also documented.