Speaker
Description
Precise reconstruction of displaced vertices and associated decay chains is critical for many flavour-physics analyses, and searches for long-lived particles with lifetimes above 100 ps. In LHCb, these signatures often occur far from the high-resolution trackers, inside the dipole magnet region. Reconstructing these signatures with LHCb’s forward and segmented tracking system is challenging. It is only possible using short track segments with limited momentum resolution that require numerical extrapolation over long distances within a complex magnetic field. To meet these challenges in LHCb’s fully software trigger, these signatures are reconstructed with a new high-quality, fast-extrapolation strategy that caches precomputed states and interpolates between reference points. This avoids costly numerical integration and improves displaced-vertex yields and purity without affecting throughput, leveraging GPU texture-memory techniques. Machine learning techniques are used for fast online event selection. Offline, these advances and others are consolidated in a modular, extensible C++ decay-tree reconstruction framework integrated around LHCb’s software framework. Designed for complex constrained topologies and decays inside strong and non-uniform magnetic fields, it is also applicable to other experiments. Performance studies show large improvements in efficiency, invariant-mass resolution, and vertex resolution, expanding the physics reach for precision measurements and long-lived particle searches at LHCb.
| I read the instructions above | Yes |
|---|