Pranav Bhagawath Prasad

Topic

Semileptonic Decays of B Mesons: Characterization and Comparison of ML-Based Semileptonic B Tagging

Department of Physics and Astronomy

Date & location

• Friday, October 9, 2026
• 2:00 P.M.
• Clearihue Building, Room B017

Examining Committee

Supervisory Committee

• Dr. Robert Kowalewski, Department of Physics and Astronomy, University of Victoria (Supervisor)
• Dr. J. Michael Roney, Department of Physics and Astronomy, UVic (Member)

External Examiner

• Dr. Katherine Pachal, Research Scientist, TRIUMF

Chair of Oral Examination

• Dr. Boualem Khouider, Department of Mathematics and Statistics, UVic

Abstract

Particle physics aims to understand the fundamental constituents of matter and the interactions that govern them. Although the Standard Model successfully describes a wide range of experimental observations, it cannot explain phenomena such as dark matter, neutrino masses, or the matter–antimatter asymmetry of the Universe. Precision measurements in flavour physics provide a powerful means of searching for physics beyond the Standard Model, as new particles may influence rare or suppressed processes through virtual contributions.

This thesis presents two studies performed using data collected by the Belle II experiment. The first investigates bremsstrahlung recovery techniques for electrons, which are essential for mitigating energy loss due to photon radiation in detector material. Two algorithms implemented within the Belle II software framework, correctBremsBelle and correctBrems, are evaluated using reconstructed 𝐵⁻ → 𝐽/𝜓(ℓ⁺ℓ⁻)𝐾⁻ decays. By scanning photon-energy thresholds and angular acceptance parameters, their performance is assessed through momentum-resolution studies, and data–Monte Carlo comparisons. Both algorithms significantly improve electron momentum reconstruction relative to the uncorrected case; however, correctBrems demonstrates substantially lower rates of photon misassociation and greater robustness against parameter variations while providing comparable improvements in 𝐽/𝜓 mass resolution. The study identifies optimal operating parameters of 𝐸_𝛾^min ≈ 20 MeV and maxAcceptance ≤ 3, supporting correctBrems as the preferred bremsstrahlung recovery method for future Belle II analyses.

The second study is on the comparison Full Event Interpretation (FEI) with a custom Simple semileptonic Tag for Belle II. The FEI uses a hierarchical decay-chain reconstruction in which intermediate particle candidates are ranked with gradient-boosted decision trees, allowing reconstruction of tag-side semileptonic decays such as 𝐵 → 𝐷^∗ℓ𝑣, while the custom Simple Tag uses a cuts-based approach. The performance of this FEI semileptonic tag is evaluated using Monte Carlo simulation and 365 fb⁻¹ of Belle II data. A study of tagging efficiency, background levels and data—Monte Carlo agreement is used to define a working point. For this working point, the FEI increases the tagging efficiency relative to the cut-based method by 74%. The agreement between the Belle II simulation and data is slightly better for the cut-based reconstruction.