A Neural Networks Search for Single Top Quark Production: Collider Detector at Fermilab

Measuring single top is an important test of the Standard Model, as its production cross section is directly proportional to Vtb^2, where Vtb represents the coupling between the top and the bottom quarks. In addition, it impacts the future searches for the Higgs boson, and has the potential to validate a number of phenomena outside of the Standard Model. We have developed a Neural Networks (NN) technique to identify single top events. Considering the single top signal to background ratio of less than 10%, the NN approach is more suitable than a cut based analysis as it allows combining information from multiple event variables. We devised a three-output perceptron with one intermediate layer, whose output nodes estimate signal and backgrounds'' probabilities. We searched for Standard Model single top production in the data collected by the Collider Detector at Fermilab and found a single-top cross section approximately 2.5 standard deviations larger than the theoretically predicted value. We employed a fully Bayesian treatment to set an upper limit on single top production of 24 pb at 95% C. L., including systematic uncertainties.