Search for a dark photon and an invisible dark Higgs at the Belle II experiment

This thesis presents a search for the dark Higgsstrahlung process with invisible Higgs e+e− →A′(→ µ+µ−)h′(→ inv.) at the Belle II experiment. The search is motivated by hidden-sector models, also known as dark-sectors, poten tially populated by particles beyond the Standard Model (SM), including dark matter candidates. Dark sector models represent an extension of the SM: a minimal extension foresees a U(1)′ gauge symmetry which undergoes spontaneous symmetry breaking. As a consequence, two new bosons arise: a dark photon A′, associated with the gauge interac tion, and a dark Higgs boson h′, responsible for the symmetry breaking mechanism. These two particles can mediate interactions between the Standard Model and the dark sector acting as portals. The present work focuses exclusively on the search for these mediator particles, without directly probing other possible dark-sector states, which are assumed to lie at higher mass scales. In the scenario considered, the dark photon decays promptly into a pair of muons, while the dark Higgs escapes detection due to its very long lifetime. The analysis is performed using the full Belle II dataset collected during the years 2019-2022, also called Run-1 dataset, corresponding to an integrated luminosity of 365 fb−1 collected at the SuperKEKB e+e− collider operating at the Υ(4S) resonance. The signal signature consists of a dimuon pair associated with missing energy: both the dimuon invariant mass Mµµ and the recoil mass Mrec against the dimuon system are expected to show up as resonances, corresponding to the dark photon and the dark Higgs, respectively. The search is then performed by looking for localized excesses over the background in the two-dimensional plane defined by Mµµ and Mrec. A cut-and-count strategy is adopted, defining a large set of mass-dependent ellipti cal signal windows, optimized according to the detector resolution, in which to compare data and expected background yield. Background contributions from Standard Model processes are estimated using Monte Carlo (MC) simulation and data-driven corrections, while dedicated control samples are used to test the MC reliability. The statistical inter pretation relies on the computation of local p-values in each search window, as estimates of the significance, and, in absence of a signal, on the derivation of upper limits through a likelihood-based Bayesian framework validated with pseudo-experiments. Searches for the same process were performed at KLOE [1] with 1.65 fb−1 of data collected at and near the Φ resonance, and at Belle II with approximately 9 fb−1 of data. At the time of writing, the signal regions remain blinded. Therefore, this work focuses on the development of the analysis strategy and its validation on data and simulated control samples, on the evaluation of the sensitivity in terms of expected exclusion limits both on the process cross section and on the dark-sector parameter combination ϵ2 ×αD.