Transcript Erik
Update FTK Meeting 07/13/06 Erik Brubaker U of Chicago Outline • Tools Signal Samples – Sherpa samples – LVL1 Parametrizations • Physics Sherpa Samples LVL1 Params FTK Performance – Hhh4b • Plans LVL1 Rate LVL2 Rate Optimize FOM with rate constraints July 13, 2006 FTK Meeting Analysis FOM 2 Sherpa samples • Last meeting, thought final samples were being generated. • Several problems uncovered since then. July 13, 2006 FTK Meeting 3 Sherpa problem #1 • 22-only sample (red) is harder than 22+23 (green)! • Problem in sherpa 1.0.8—parton shower starting scale not clearly defined for 22 alone. July 13, 2006 FTK Meeting 4 Drop 22 ME in sherpa • All the physics processes we look at so far have 4 final state objects. • In a 22+23 sample, events produced by the 22 ME contribute negligibly to trigger rates. • Drop 22 ME to increase effective statistics. July 13, 2006 FTK Meeting 5 Sherpa problem #2 (related to #1?) Events per pb-1 • Sample without 22 ME has harder 4th jet. • 4th jet is produced by PS in either case—expect very similar distributions. • Plots normalized by cross-section. July 13, 2006 FTK Meeting 6 Sherpa problem #3 (from Kohei) • Lowering the ycut of the sample should only increase the derived trigger rate by covering more phase space. • But rate(ycut=10) < rate(ycut=25)! • Related to other problems? Revisit with recent fixes. July 13, 2006 FTK Meeting 7 Additional sherpa questions • Execution time in my setup is longer than the authors see. Strangely, t(MI-on) < t(MI-off). • Why not throw this all out the window and use Pythia or Alpgen? – Point was to avoid pythia dijet+PS to describe 3rd, 4th jets. But pythia can and should be used as a baseline to compare to. • Help from Kohei for jobOptions. – Alpgen is complicated in different ways. SUSY folk are trying to get a validated multijet sample out of it. Could be an option. July 13, 2006 FTK Meeting 8 tt LVL1 parametrization • For BG processes with real ts, would be useful to have parametrization of LVL1 t trigger response to ts. • Improvements in ATLAS DDM mean I can access datasets… – SingleTaupt20, 50, 100 copied over. – Can’t use samples with filters! • Next step is to run full latest simulation and make response plots for matched ts. July 13, 2006 FTK Meeting 9 Multi-threshold machinery • Rigorous determination of the probability for a given event to pass a given set of jet trigger thresholds (given the LVL1 jetjet parametrizations) is somewhat complicated. • Code available now for the general case at: http://hep.uchicago.edu/~brubaker/ftk/ July 13, 2006 FTK Meeting 10 LVL1 rates for Hhh4b 3rd & 4th jet threshold • Example threshold scan, two thresholds, using full parametrization machinery. • Valid for any 4-jet FS. • Uses Sherpa MC with known bugrates too low! 40 60 80 100 1st & 2nd jet threshold 40 60 80 3.0±0.6 2.2±0.4 1.6±0.4 0.035±0.013 0.031±0.009 (5±3)e-5 Rate @ 1033 cm-2 s-1 July 13, 2006 FTK Meeting Bad MC! 100 1.1±0.3 0.023±0.008 (4±1)e-5 (4±2)e-9 11 Plans • • • • • Make and check “final” sherpa samples. Use pythia sample as bg rate cross-check. Make tt LVL1 parametrization. Understand Hhh4b cross-section vs mH, mh. Perform threshold scans and optimize observation significance for Hhh4b with and without FTK. • Focus, finish, document by July 26. July 13, 2006 FTK Meeting 12