Münster 2017 – scientific programme

Parts | Days | Selection | Search | Updates | Downloads | Help

T: Fachverband Teilchenphysik

T 75: Top Quark 4 (Eigenschaften)

T 75.7: Talk

Wednesday, March 29, 2017, 18:15–18:30, JUR 4

Probing the ttγ analysis at √s = 13 TeV with ATLAS using Neural Networks — Boris Lemmer, Maria Moreno Llacer, Arnulf Quadt, Elizaveta Shabalina, and •Joshua Wyatt Smith — II. Physikalisches Institut, Georg-August-Universität Göttingen

Through the ttγ process we can measure the electromagnetic couplings of the top quark. Evidence of this process was seen at CDF with √s = 1.96 TeV, while observation occurred at the LHC at √s = 7 and √s =8 TeV, with increasing precision. The largest source of uncertainty comes from the estimate of the background originating from hadron-fakes. These are photons from hadrons or hadron decays that are misidentified as prompt photons. Both analyses relied on “conventional” background estimation of these hadron-fakes using data-driven techniques. More advanced methods can be used to improve this estimate and potentially reduce this uncertainty (and others) even further. New approaches making use of Neural Networks will be discussed. One approach is to build a discriminating variable to distinguish real prompt photons from hadron-fakes. A working point would then be chosen to maximize purity of prompt photons. The major advantage is that this is not necessarily analysis specific and thus can serve as a general tool for the ATLAS community. Another approach is to build an analysis specific classifier. This is where the topology of the whole ttγ event is important. Kinematics and isolation of the photon and other particles would play a major role in this classifier.