Angular analysis of the $B \to K^* \ell \ell$ decay in long shutdown 1 data at Belle II

Sumitted to PubDB: 2025-02-26

Category: Phd Thesis

Tags:

Principal Authors Martin Sobotzik
Date 2024-10-30
Belle II Number BELLE2-PTHESIS-2025-003
Abstract This thesis will present comprehensive studies of a workflow to analyze the B →K∗ℓℓ decay at Belle II. This includes the development of snapshot ensemble deep neural networks for classification, which aim to maximize the efficiency and purity of the data samples for each possible charge configuration of the decay channel. These networks allow the two dimensional discriminating fit on the uncorrelated Mbc and ∆E variables to reliably determine the signal and non-signal contributions. In the next step, as a novel approach, another deep neural network is developed to model the detector response function following the function approximation theorem. In order to deal with remaining non-signal events on a statistical level, the sWeights method is studied and used. By applying this method it is possible to determine the true signal shape of the target variables by calculating weights from the supplied PDF models of the uncorrelated discriminating variables. Using this method, the background contribution within the target variables has no longer be determined by simulation. In order to test the developed workflow, the differential decay rate is extracted in the three angular dimensions in four bins of q2, the di-lepton invariant mass squared. Exploiting transformation symmetries in the differential decay rate, the number of free parameters of the differential decay rate is reduced from eight down to three (FL, S3 and P′5) which are then determined on both simulation and real data.

Files