Pattern-Based Electron Counting Algorithm for LDMX

Sammanfattning

The Trigger Scintillator (TS) of the Light Dark Matter eXperiment (LDMX) is responsible for counting the electrons in the incoming beam in order to accurately determine the trigger threshold used in identifying missing momentum events. Electrons are counted by reconstructing their approximately horizontal tracks through the TS. The software algorithm operates on a maximum vertical tolerance value, making it both computationally expensive and inflexible to a misalignment in the TS geometry. This thesis proposes an alternate counting algorithm using a data-driven lookup table (LUT) containing real electron patterns derived from simulation or data. This allows the tracking definition to naturally adapt to the TS geometry presented in the data. To write the LUT, electrons’ characteristic propagation patterns are first examined using truth-level information and then compared to unfiltered simulation data. It is shown that real tracks can be isolated from random combinatorics patterns by applying a minimum pattern frequency threshold to simulated data. At an optimal threshold of 0.0008, the data-driven LUT is able to achieve a maximum tracking efficiency 99.37% for a fake track rate 0.001668%, compared to the tolerance definition algorithm’s efficiency 98.00% and fake rate 0.00073%. This study demonstrates the feasibility of a data driven tracking method, with potential for further improvement through refinement of the frequency-threshold definition.