Reweighting heavy-flavor production fractions to reduce flavor modelling uncertainties for ATLAS

The ability to identify jets originating from $b$- and $c$-hadrons (flavor tagging) is one of the key experimental techniques that enables a wide range of searches and measurements by the ATLAS experiment. Generally, the tagging efficiencies in simulated Monte Carlo samples differ from the data and need to be corrected with dedicated calibration procedures. However, the tagging efficiencies also vary between the Monte Carlo shower simulations such as Pythia, Sherpa, and Herwig. These differences have a direct impact on the modeling uncertainties associated with flavor tagging so it is important to reduce them if possible. In this note we introduce a method of reweighting the heavy-flavor production fractions to a common world average, which eliminates one of the largest causes of differences in tagging efficiencies between the Monte Carlo shower simulations. The heavy-flavor production fractions are the rates at which $b$- and $c$-quarks hadronize into different species and the flavor tagging efficiencies in ATLAS have been found to depend on the hadron species inside a jet. The note presents heavy-flavor production fractions in the ATLAS Monte Carlo samples, the reweighting procedure and the impact of the reweighting on flavor tagging efficiencies. Moreover, the experimental uncertainties in the heavy-flavor production fractions can also be estimated with the same reweighting procedure which gives rise to a common way of estimating these systematic uncertainties in ATLAS.