Identification of b-quark jets from gluon splitting:

Contact: Ariel Schwartzman

b-jets from gluon splitting are a major background to many Super-symmetric and Higgs searches in final
states containing b-jets.

Gluon splited b-jets have several distinct experimental characteristics that can be used to efficiently 
separate them from signal b-jets. They contain -on average- larger track multiplicity, opening angle, 
and calorimeter cluster multiplicity than single b-jets.

The SLAC group has proposed and developed a technique to identify merged b-tagged jets from gluon splitting, 
and distinguish them from single b-tagged jets. The method uses a Neural Network (NN) trained with tracking 
and calorimeter information.

The proposed project consist of improving the existing technique by properly accounting for the different 
transverse momentum spectra of signal and background b-jets (this involves the use of a weighted Neural 
Network algorithm), and the implementation and validation of the algorithm in the official ATLAS software. 
In a future step, we would like to develop methods to improve b-tag background estimation using the gbb NN 
output.