Wing Sheung Chan

Object reconstruction and identification 47 When the parameter P = − 1 , the algorithm is called the anti- k t algorithm, which is the standard algorithm for jet reconstruction in ATLAS. The anti- k t algorithm prioritises the grouping of low-energy clusters into a close-by high-energy cluster. This avoids reconstructing multiple high-energy jets which are close to each other and results in circular-cone-shaped jets of maximum radius R . The anti- k t algorithm exhibits a number of desirable properties. By construction, the algorithm is collinear- and infrared-safe. To be collinear-safe means that if particles in a jet are split into multiple particles that travel in the same direction, the reconstructed jet will not change as a result. To be infrared-safe means that soft emissions cannot radically impact the jet reconstructions. Moreover, the anti- k t algorithm has been shown to be particularly insensitive to underlying events and pile-up. These are very important qualities for a jet reconstruction algorithm as they ensure that the observed jets can faithfully represent the particles from a hard collision event that initiated them. They also make comparisons with theoretical predictions practical as QCD processes at low-energy scales, which are difficult to calculate, cannot significantly alter the reconstruction of energetic jets. On top of all the above mentioned merits, the anti- k t algorithm, implemented using the FastJet library [74] , is also computationally fast. 3.1.2. Flavour tagging The LHC produces an abundance of top quarks. In the search for Z → `τ decays, events such as t ¯ t → b ¯ bW + W − → b ¯ b` ± τ ∓ ν ¯ ν are a significant contribution to the SM background if they could not be effectively identified. The top quark decays via weak interaction into a bottom quark and a W boson 99.8% of the time. Therefore, events with top quarks can be effectively identified by recognising b -quark-initiated jets ( b -jets), a process also known as b -tagging. While a b quark always hadronises immediately after its production, the b -hadrons created from the hadronisation have an appreciable lifetime of around 1 . 5 ps . This relatively long lifetime allows a 50-GeV b -hadron to travel approximately 3mm on average before it decays. Because of this, a significantly displaced secondary vertex is expected in b -jets, which is the most important feature that is exploited by b -tagging algorithms. In the ATLAS experiment, b -tagging is performed using several low-level algorithms that provide intermediate outputs which are then fed into a high-level tagging algorithm that creates a final discriminant using multivariate analysis (MVA) [75] . There are four low-level algorithms: IP2D, IP3D, SV1 and JetFitter . The IP2D and IP3D algorithms are likelihood-based algorithms that exploit the large impact parameters of tracks from the b -hadron decay. While the IP2D algorithm only makes use of the transverse impact parameter significance, the IP3D algorithm also takes into account the impact parameter significance in the z -direction. The SV1 and JetFitter algorithms are both vertex reconstruction algorithms. The SV1 algorithm explicitly reconstructs a single displaced secondary vertex in a jet, whereas the JetFitter attempts to reconstruct the full b -hadron decay chain by reconstructing the multiple vertices.