Transcript Slide 1
Highlights of the SiD-Iowa Particle Flow Algorithm • Previous (LOI) version of PFA was for up to 500 GeV collider energy. – Even at 500 GeV performance sufficient, but could be improved. • Established a ground-up approach: – Targeted diagnostics for each piece of the algorithm to evaluate each piece. • Photon reconstruction: – Once photons are reconstructed, the hits are taken out from use. – An anti veto is in place which checks “photon-hits” for fakes and treats them as hadrons. • Sub-cluster categories (clump purity) – Clump (sub-cluster) purity was not good enough: make smaller (don’t use NN). • Linking probabilities of sub-clusters (discriminating variables, likelihood) – Use identical clustering for linking probability and shower reconstruction. – Add likelihood method with several discriminating variables. • Shower reconstruction (two passes): IN PROGRESS – Form a high purity skeleton with all tracks treated similarly. – Add hits in second pass with adjudication between nearby showers. Highlights of the SiD-Iowa Particle Flow Algorithm Energy weighted Purity (clumps) Energy fraction (clumps) baseline new baseline new ECAL 90% 94% 15.3% 16.2% HCAL 80% 87% 32.2% 29.0% Overall 84% 90% 47.5% 45.2% Clump reconstruction performance Variables in likelihood used for linking Discrimination improves significantly when angles b and c are added in the likelihood