Projections for Single Top Sensitivity

The first CDF single top analysis in Run II has been published in February 2005, Phys. Rev. D 71, 012005 (2005). We are working on several improvements of this analysis with respect to increasing the acceptance and improving the discrimination between signal and background. Potential improvements have been laid out on the CDF Single Top Workshop that took place at Fermilab on December 2 and 3, 2004. Three multivariate techniques that combine kinematic variables into a powerful discriminant have been tested. In addition, we investigate a matrix element method. Currently, a neural network technique seems to be the most promising approach.

The neural network combines 15 kinematic variables to a single network output ranging from -1 to 1. The sensitivity study is based on the Monte Carlo data sets used for the result published in February 2005. The same preselection cuts, trigger and reconstruction efficiencies as well as background estimates are assumed. The cuts on the top quark mass and on the E_T of the leading jet are replaced by a cut on the neural network output. The cut is chosen such, that S/sqrt(B) is maximized. The figure below shows the sensitivity expressed by S/sqrt(B) as a function of the integrated luminosity.



Plot as [ eps ] or [ jpg ]. This plot was accepted for public display in the Lepton plus Jets meeting on July 1, 2005. See the presentation given at that meeting for more details.

A three sigma evidence can be expected with 1.5 fb^-1. The signal would contain an average of 27.3 events.
With a data sample corresponding to 1 fb^-1 we expect to see a 2.4 sigma excess.
These projections do not take systematic uncertainties into account.