Search for Anomalous Production of the Higgs Boson at the CMS Experiment and Exascale Computing for the LHC
Skip to main content
eScholarship
Open Access Publications from the University of California

UC San Diego

UC San Diego Electronic Theses and Dissertations bannerUC San Diego

Search for Anomalous Production of the Higgs Boson at the CMS Experiment and Exascale Computing for the LHC

Abstract

This dissertation presents a search for anomalous production of the Higgs boson with the Compact Muon Solenoid (CMS) experiment, together with two developments in reconstruction and computing for the High-Luminosity LHC (HL-LHC). The search targets vector boson scattering production of a Higgs boson with two vector bosons. It aims to constrain the quartic HHVV coupling, quoted relative to its Standard Model value: departures from unity grow the cross section quadratically and boost the final-state bosons. The search uses the full Run 2 dataset, 138 inverse femtobarns at a center-of mass energy of 13 TeV and 283 inverse femtobarns of Run 3 data at 13.6 TeV, in three channels: an all-hadronic state with three large-radius jets and two single-lepton states. The dominant multijet and top quark backgrounds are estimated from data with an ABCD method built on two decorrelated classifiers, and the signal is extracted from a simultaneous fit to several orthogonal signal regions per channel. The combination excludes a coupling of 1.5, with an observed 95 percent confidence level upper limit of 0.48 on that rate against an expectation of 0.64; the Run 3 data alone constrain the coupling between 0.612 and 1.387, improving on the published Run 2 result. The second development addresses charged-particle tracking at the HL-LHC using a machine learning approach: the first demonstration in CMS of a transformer-based model trained with object condensation, built on Line Segment Tracking triplets to reconstruct full track candidates. The model reduces the fake rate by roughly a factor of three and all but eliminates duplicates, at an efficiency comparable to the algorithmic baseline. The third concerns moving the resulting data. The HL-LHC will produce an unprecedented volume of data, so the wide-area network is no longer a best-effort service but a critical resource that must be managed. This work integrates the Rucio data management system with the SENSE software-defined networking service, so that bandwidth is allocated by transfer priority and reservations can be reconfigured mid-transfer without interrupting traffic in flight. It is demonstrated on a multi-domain testbed and in benchmarks at 400 Gbps.