This thesis is a study on the various aspects of track reconstruction at the
CDF II experiment, ranging from fast tracking at trigger level to precision
tracking in offline analyses.

Part of the work is devoted to studying the performance of SVT, the silicon
track processor which is part of the second level of trigger at CDF II. The
resolution, efficiency and fake rate of the tracker are estimated both in the
baseline configuration of the experiment, and in two other configurations,
which make use of Layer 00 --- a recently-added radiation-hard silicon layer,
placed immediately outside the beam pipe. As expected, introduction of Layer 00
improves track resolution considerably with respect to the baseline;
one of the two configurations is also shown to decrease the fake rate.

At the third level of trigger, and in offline analyses, CDF II reconstructs
tracks by means of several strategies. This thesis presents a general-purpose
tracking strategy, called ``Histogram Tracking'', which is used both in the
drift chamber and in the silicon vertex detector. This strategy is now part
of the official tracking code of CDF II.

In order to evaluate the performance of Histogram Tracking and of other
strategies, their results are compared to the output of an ideal track finder.
This tool provides a lower bound to the resolution of track parameters at
CDF II, as a function of momentum and detector occupancy. Strategies are also
evaluated in terms of their CPU time cost.

For the purpose of track reconstruction in the transverse plane of the
drift chamber, Histogram Tracking turns out to be a very reliable algorithm,
with a high efficiency even in high-occupancy situations. The complementary
strategy, Segment Linking, while being faster, is more sensitive to occupancy.

Once a track has been found in the drift chamber, it is extrapolated towards
the silicon vertex detector, and it is associated to a set of silicon hits.
In this situation, Outside-In --- a strategy based on Kalman filtering ---
proves to be the most accurate strategy; however, it is also very slow.
Histogram Tracking requires less time; it is less reliable in borderline
decisions, but most of the Histogram Tracking errors have a small impact on
the fit parameters of reconstructed tracks. Therefore, silicon Histogram
Tracking is being considered for use at trigger level, where speed is
essential.



Posted to /cdf/pub/thesis/cdf5561_tracking_performance.ps
    http://www-cdf.fnal.gov/thesis/cdf5561_tracking_performance.ps

Posted by gatti@cdfsga