Wing Sheung Chan

46 Object reconstruction and identification 3.1.1. Jet finding and reconstruction If we were to manually and visually find a jet in an event, it would be rather simple and intuitive – a jet is just a collimated shower of particles. Yet, of course, in order to analyse the billions of recorded events, jet finding must be an automated and objectively defined process. The algorithm used to reconstruct the jets must be able to deliver reproducible results that can be confronted with theories and other experiments. For this, a topological clustering algorithm [71] and a sequential recombination algorithm known as the anti- k t algorithm [72] are used in the ATLAS experiment. The reconstruction of a jet begins with the grouping of energy deposits in the calorime- ters into so-called “clusters”. Calorimeter cells which have a high signal-to-expected-noise ratio ( ζ ) are used to seed the clusters. Calorimeter cells that are topologically connected to a seed and with ζ passing a certain threshold are joined with the seed to form an initial cluster. The cluster is then grown by repeatedly collecting neighbouring calorimeter cells with ζ above the threshold into the cluster. Each cluster can be interpreted as a massless pseudoparticle which has its own energy and momentum. The energy of a cluster is the signal-weighted sum of the energies in the cells and the direction of a cluster is that of the signal-weighted barycentre of the cells. Furthermore, the shape of a cluster can be quantitatively described by its moments, which are important variables for identifying the origin of a jet. The kinematic properties of clusters are initially calibrated at the electromagnetic scale. Since the calorimeters in the ATLAS detector are non-compensating, i.e. they respond differently to electromagnetic and hadronic particle showers, calibration is needed to correct the scale for hadronic clusters. For that, a local hadronic cluster calibration (LC) scheme is used [73] . In a sequential recombination algorithm, jets are reconstructed by repeatedly combining clusters based on their geometric proximity and energies. For every pair of clusters, the “distance” d ij = min( p 2 P T ,i , p 2 P T ,j ) (∆ R i,j ) 2 R 2 , (3.1) where i and j are indices for the two clusters, p T is the transverse momentum of the cluster, ∆ R is the distance between clusters in the rapidity–azimuth space, and P and R are constant parameters of the algorithm, between them is calculated. In addition, the distance d i B = p 2 P T ,i (3.2) between every cluster and the beam is also calculated. Then, repeatedly, the smallest of { d ij } and { d i B } is identified. If it is a d ij , the two clusters are combined into a “protojet”. Otherwise, the cluster/protojet with the smallest d i B is considered a completely reconstructed jet and is removed in subsequent iterations. The parameter R is usually referred to as the jet radius or cone size and is the maximum angular distance between a cluster in a jet and the central axis of the jet. In ATLAS, R is usually chosen to be 0.4.