Electron track reconstruction studies and improvement for LHCb's real-time analysis trigger

This thesis presents an improved neural-network-based Track-Matching algorithm to recover and reconstruct electron tracks more efficiently in LHCb's real-time analysis trigger. Electrons emit bremsstrahlung, which complicates their track reconstruction. Currently, the track-reconstruction tuning explicitly excludes electrons, since including them significantly lowers the reconstruction efficiency of other particles. The presented algorithm is intended to be run in a dedicated electron reconstruction in HLT2, thereby circumventing such issues entirely, while allowing all aspects of the algorithm to be optimised to electrons. It is demonstrated that the improved electron Matching algorithm allows for electron tracking efficiencies of above 90%, while simultaneously reducing the fake track fraction to below 15%.