Parts | Days | Selection | Search | Updates | Downloads | Help
T: Fachverband Teilchenphysik
T 82: Flavor-Tagging, Jet-Kalibration
T 82.9: Talk
Thursday, March 28, 2019, 18:00–18:15, S09
The Track Classification Tool and the implementation of a new b-tagging algorithm for ATLAS — Carlo A. Gottardo, Sebastian Heer, Vadim Kostyukhin, •Ö. Oğul Öncel, Keshava Prasad, Andrea Sciandra, and Markus Cristinziani — Physikalisches Institut, Universität Bonn
Correctly identifying b-quark initiated jets (b-tagging) at large transverse momentum (pT > 1 TeV) will become increasingly important as ATLAS accumulates more data. Currently used b-tagging algorithms distinguish b-jets against light-jets. This approach is effective for low- and medium-pT jets, but results in performance degradation at high pT, where fragmentation dominates.
Instead of distinguishing two sets of tracks, the b-tagging performance can be improved by introducing explicitly the most important track categories and classifying tracks before the b-tagging step itself. The Track Classification Tool (TCT) is a multi-class multivariate discriminator that classifies tracks into one of the following three categories: heavy flavour, fragmentation or hadronic interactions and pile-up.
A newly developed b-tagging algorithm in ATLAS uses the TCT to classify tracks in a jet and based on this information creates a b-tagging score. It is found to be enhancing the b-tagging performance, particularly in the high-pT regime, in comparison to currently used b-taggers. The implementations will be described and comparative performance studies will be presented.