Top Quark Mass in Dilepton Channel Using a Matrix Element Method and Neuroevolution Selection

We present a measurement of the top quark mass in the dilepton channel using a technique in which we form a posterior probability for the mass as a product of the per-event differential cross-section for leading order top quark pair production. The calculation of the differential cross section for processes which produce background events are used to reduce the impact of background events in the sample. The events used in this analysis are selected using an evolutionary neural network that is directly optimized for precision in measurement of the top quark mass.

In 2.0 fb^-1 of Run II data, we expect a statistical uncertainty of 2.7 GeV if M_top=175.0 GeV/c². We measure

Documentation

• A note describing this analysis can be found here.
• Internal page for CDF collaborators

Tables and Plots

Description	jpg	pdf/eps
Kinematic validation of the preselection.		pdf
Kinematic validation of the neural network selection.		pdf
Expected sample composition after neural network selection		pdf
Final posterior probability density as a function of top pole mass for the 344 candidate events in data.		pdf
Expected distribution of errors from MC (Mt=172 GeV/c2) and measured error in data. 53% of pseudo-experiments had an error lower than that measured in data.		pdf
Response for pseudo-experiments of signal and background events.		pdf
Residual, pull mean, and pull width for varying top mass MC samples after scaling of statistical error. The error scaling factor is S=1.16		pdf
Table of systematic uncertainties.		pdf