Introduction

In the general concept of the HERA-B experiment track reconstruction in the Pattern Tracker is the first step in the reconstruction chain. The next steps, magnet tracking and track prolongation to the trigger chambers, are essentially track following procedures which use tracks reconstructed in the pattern tracker as seeds.

Figure 4.1: A typical simulated event: $J/\psi \rightarrow \mu ^+\mu ^-$ decay superimposed with two inelastic interactions. For details see text.

A typical simulated event demonstrating the large track density with the pronounced radial dependence is shown in Fig. 4.1. Evidently, not all tracks passing through the pattern tracker are straight-line high-momentum tracks coming from the vertex region. There are also a considerable fraction of low-momentum tracks affected by multiple scattering which come from secondary interactions and in-flight decays. This is clearly visible in the isometric closeup (right figure) of the Inner Tracker section of the event shown in the left part of Figure 4.1. Due to the large track density the probability of track overlap is relatively high and any tracking algorithm must routinely cope with clusters of hits from nearby tracks. In the Outer Tracker, the problem of track recognition becomes even more complicated due to the small cell size in the drift chamber which makes resolution of the left-right ambiguity particularly difficult.

In the past years members of the HERA-B collaboration have performed an extensive analysis of different track recognition methods [#!ranger!#,#!Schober!#,#!Borgmeier!#,#!Paus!#]. In particular, the reconstruction package RANGER based on track following and concurrent track evolution has been developed. After extensive tests on simulated data, RANGER was chosen as the default track reconstruction package for HERA-B.

RANGER was tuned to the detector parameters expected for the original experiment design [#!TDR!#]. As a result, it has faced serious problems when the actual hit efficiencies of the detectors turned out to be lower than expected in [#!TDR!#] (the most pessimistic estimates are 90% for the OTR and 86% for the ITR). At the same time hit resolutions were also found to be worse than in the original design (lower bound estimates are 500 $\mu$m in the OTR and 200 $\mu$m in the ITR). In addition an unexpected large fraction of channels (up to 7%) in the Outer Tracker was insensitive due to high voltage problems.

These circumstances motivated new and alternative approaches to track reconstruction which should be more robust and tolerant against hardware problems than RANGER.