Jets/MET Performance with the combination of Particle flow algorithm and SoftKiller

The main purpose of my work is to study the performance of the combination of Particle flow algorithm(PFlow) and SoftKiller(SK), “PF+SK”. ATLAS experiment currently employes Topological clusters(Topo) for jet reconstruction, but we want to replace it with more effective one, PFlow. PFlow provides us with another method to reconstruct jets[1]. With this algorithm, we combine the energy deposits in calorimeters with the measurement in ID tracker. This strategy enables us to claim these consistent measurements in a detector come from same particles and avoid double counting. SK is a simple and effective way of suppressing pile-up[2]. This way, we divide rapidity-azimuthal plane into square patches and eliminate particles lower than the threshold 𝑝"#$%, which is derived from each 𝑝",' so that the median of 𝑝" density becomes zero. Practically, this is equal to gradually increasing 𝑝"#$% till exactly half of patches becomes empty. Because there is no official calibration on PF+SK so far, we have tried to derive our own calibration that performs well with PF+SK and to compare it with the performance of the official one. Especially as for MET, this is the first study of PF+SK (reference [3] is the prior research in PF+SK for jet performance). Because our calibration target is always jets, we derive new calibration with dijet samples where all we have is jets by Monte-Carlo simulation(MC), no reconstructed data included in this process. We calculate jet response of the events (non-closure response) and derive a calibration which corrects this response (closure response) and ideally leads this response to unit value. Then by applying the calibration to reconstructed events (in this report, I focus on ttbar events), we can look into its ultimate performance, “performance study”. In this study, in addition to PFlow and SoftKiller, we calculate various combinations of jet reconstruction, Topo, and pile-up suppression, like Voronoi subtraction(Vor) and Constituent subtraction(CS) [4].