Performance of the b-jet identification in ATLAS

A performant identification of the jets originating from $b$-quarks is one of the key ingredients which allows for the diverse physics program of the ATLAS experiment. Algorithms for the $b$-jet identification in ATLAS are exploiting long lifetime and high mass of the $b$-hadrons, as well as the information on the tracks associated with jets. The performance of these algorithms is measured in data to enable reliable usage in the physics analyses. In this talk we present the performance of the multivariate technique MV1 for $b$-jet identification which as its inputs takes information from the other $b$-jet identification algorithms. The efficiency of the MV1 algorithm is measured using dileptonic top pair events and is based on a recently developed likelihood approach. This approach allows to exploit per-event flavour correlations between the two jets, taking into account the transverse momentum dependence of the $b$-jet identification algorithms. Correction factors which take into account differences in the $b$-jet identification efficiencies in simulation and data are derived as a function of jet transverse momentum and pseudo-rapidity. All the results are presented using the proton-proton collision dataset at centre of mass energy of $\sqrt{s}=8$~TeV corresponding to an integrated luminosity of $\mathcal{L}\approx$20.3~fb$^{-1}$.