Optimization of the CA Track Finder for the STS Geometry with Sensors Overlapping

Our previous work on adaptation of the Cellular Automaton (CA) track finder [1] to the realistic STS geometry with overlapping sensors allowed to recover the high tracks-finding efficiency [2] which it has had for the simplified STS description with sensors of each station arranged at the same Z position (the “same-Z” geometry). Below we report on a detailed study of the behaviour of the track finder for the realistic STS geometry that addressed the problem of an increased number of clone and ghost tracks relative to the same-Z geometry. In the realistic STS geometry the sensors overlap, the hits within the same station have different Z coordinates, and also the detector contains more material. A larger number of hits (both true and fake) is produced in the sensor overlap regions giving more than 1 hit per track per station, additional hits also coming from the increased number of delta electrons. To reduce the combinatorics we optimized the Xand Y -windows used in the track candidate search based on the pull distributions, Fig. 1. The main component of the clones turned out to consist of short pieces of long MC tracks, broken due to hits, which did not fit into the Y -window at the step of the doublet creation. These broken tracks (∼ 4%) become effectively lost for analysis, since one of the pieces has low spatial accuracy, and the other has bad momentum resolution. From our studies we found that the increase in the number of clone tracks is not connected to the inhomogeneity of the magnetic field, but to the increase in the amount of multiple