Description
We examine a new tool utilising Graph Neural Networks (GNNs) for particle reconstruction, to enable investigation of rare and physically interesting $B$-meson decays. Belle II is an experiment at the SuperKEKB asymmetric electron-positron collider at KEK in Japan, that studies large numbers of pairs of $B$ mesons produced at the Upsilon(4S) resonance. These $B$ mesons decay in a multitude of ways, and can be used to search for new physics. By fully reconstructing and tagging one of the $B$ mesons, we gain insight into the kinematics and properties of the other $B$, providing a powerful method of suppressing backgrounds. The graph-based Full Event interpretation (graFEI) is an innovative alternative to the currently used tagging tool in Belle II, the Full Event Interpretation (FEI). By using GNNs, the graFEI can in principle provide access to the full spectrum of potential tagging $B$ decays. We demonstrate the graFEI’s capabilities in comparison with the traditional FEI, when tested on benchmark signal $B$ meson decays, showcasing the improvement in efficiency and background rejection that it can provide, and its potential as a useful upgrade for the Belle II analysis program.
| I am the presenting author | No |
|---|---|
| If you are not the presenting author, please give the presenting author's name: | Anthony Little |