A study of systematic uncertainties in mistag calibration of b-taggers in ATLAS

Identification of jets originating from the heavy quarks (b-tagging) is one of the most fundamental techniques of the LHC experiments that is widely used in physics analyses. Recently, ATLAS has switched to the new graph neural network based b-tagging algorithm, GN2. The new algorithm provides excellent performance in terms of b-tagging efficiency versus mistag rate (probability to misidentify jets from light quarks and gluons as b-jets), but it also makes the task of calibrating the algorithm (deriving the scale factors reflecting the difference between the tagger efficiency in data and Monte Carlo simulation) very challenging. I will present my study of systematic uncertainties of the mistag scale factors derived using Z+jets events.