Heavy Flavour Jet Tagging in ATLAS.

The strongly interacting colored partons, i.e. quarks and gluons, never exist in their bare form in nature. During a hard scattering process, instantly after production (time scale orders of 1e-22 seconds), the partons fragments and hadronizes to form a spray of color neutral particles which are collectively known as hadrons and mesons. These particles are clustered through an iterative clustering algorithm and the output cluster is termed as jet. The jets are macroscopic representative of the partons. 

The jets originating from heavier b and c quark (heavy-flavor jets) have a structural difference from those originating from light quarks and gluon (light jets).  The heavy b/c-hadron (originating from hadronization of b/c-quark)  undergoes a secondary decay inside the jet. Thus a heavy-flavor jet poses a secondary vertex inside jet which is missing in the light jets. The geometric features of the secondary vertex along with track reconstruction parameters are used to identify heavy-flavor jets and separate them from light jets. This technique is termed heavy-flavor jet tagging or simply flavor-tagging. 

Flavor tagging has great importance in LHC physics analyses, especially Higgs boson interaction studies. The Higgs boson has the highest probability to decay into a pair of b-quarks. Many other physics studies related to the standard model dynamics and beyond standard model search experiments use flavor-tagging techniques. 

Modern machine learning techniques are heavily used in flavor-tagging to identify b-jets and c-jets from light jets. In recent times, Weizmann group has played a significant role in optimizing the flavor-tagging performance, using machine learning, within ATLAS [1]. In past, the team played a central role in building c-tagging techniques. At present, the team is responsible for software and algorithm development tasks for flavor-tagging within ATLAS experiment at CERN. 

In recent times, the group has developed a novel vertex finding algorithm using deep neural network techniques [2]. This work is expected to improve the flavor-tagging performance further.

# --------------------------- #

1. JHEP 08 (2018) 089

2. https://arxiv.org/abs/2008.02831