170 likes | 189 Views
This analysis compares the cross-sections and event distributions of Photon+Jet events generated with Alpgen and Pythia in CMSSW 1_6_8. The main differences and possible reasons for discrepancies are discussed.
E N D
+jet Events from Alpgen vs Pythia in CMSSW 1_6_8 Michael Anderson University of Wisconsin March 27, 2008
+jet Datasets Cross-section (pb)69327.91558.992.016.34.12.2 • Alpgen: • Cross sections from AlpgenSummer07 • Pythia: • Cross sections from CSA07EventWeightProducer /ph1j_20_60-alpgen/CMSSW_1_6_7-CSA07-1201165474/RECO /ph1j_60_120-alpgen/CMSSW_1_6_7-CSA07-1202106391/RECO/ph1j_120_180-alpgen/CMSSW_1_6_7-CSA07-1200559977/RECO /ph1j_180_240-alpgen/CMSSW_1_6_7-CSA07-1200560452/RECO/ph1j_240_300-alpgen/CMSSW_1_6_7-CSA07-1200560032/RECO /ph1j_300_7000-alpgen/CMSSW_1_6_7-CSA07-1203060287/RECO /CSA07AllEvents/CMSSW_1_6_7-CSA07-Gumbo-B1-PDAllEvents-Skims1-photonjets_AODSIM/USER
Pythia Skim • The main differences between these pythia and alpgen samples: • Pythia events were in a skim - so they had to pass trigger filters! • ALPGEN normalized to 100pb-1 by hand • Pythia normalized to 100 pb-1 by the CSA07EventWeightProducer • (Used processID to select only +jet events in pythia skim)
Generated Objects • From each event, selected: • Highest Et with • Et > 20 GeV • Highest Pt jet with • Pt > 20 GeV • 2.5 < < 3.8 • Used genParticleCandidatesand iterativeCone5GenJetsPt10
Photon Et • Highest Et Generated photon • Remember: Pythia went through trigger filters & used event weight producer • but still - why the bumps? Events/bin GeV
Photon and • Dips in for pythia due to reconstruction not passing trigger filters
Photon R to Jet • Between generated and jet • Jet with • Pt > 20 GeV • 2.5 < < 3.8 Events/bin R
Generated Jet • Number of Generated Jets with • Pt > 20 GeV Events/bin GeV
Generated Jet Pt Events/bin GeV
Next Steps • Implement the same selection to the Alpgen sample and see if the shape is similar to the PYTHIA sample • Begin looking at reconstruction level photons & jets
CSA07EventWeightProducer Handle<double> genCrossSect; evt.getByLabel( "genEventRunInfo", "PreCalculatedCrossSection", genCrossSect); cross_section = *genCrossSect; . . . // photon + jets if ( processID == 14 || processID == 18 || processID == 29 ) { if (pthat > 0 && pthat < 15) { (*weight) = cross_section / 0.3E+06;} if (pthat > 15 && pthat < 20) { (*weight) = cross_section / (0.52E+06); } if (pthat > 20 && pthat < 30) { (*weight) = cross_section / (0.6E+06); } if (pthat > 30 && pthat < 50) { (*weight) = cross_section / (0.51E+06); } if (pthat > 50 && pthat < 80) { (*weight) = cross_section / (0.52E+06); } if (pthat > 80 && pthat < 120) { (*weight) = cross_section / (0.53E+06); } if (pthat > 120 && pthat < 170) { (*weight) = cross_section / (0.56E+06); } if (pthat > 170 && pthat < 300) { (*weight) = cross_section / (2.0E+05); } if (pthat > 300 && pthat < 500) { (*weight) = cross_section / (3.0E+04); } if (pthat > 500 && pthat < 7000) { (*weight) = cross_section / (3.0E+04); } }
Weighing Events • Used CSA07EventWeightProducer by Filip Moortgat to get event weights in skims https://twiki.cern.ch/twiki/bin/view/CMS/CSA07GeneratorInformation
CSA07EventWeightProducer • When different processes get mixed together into a big soup, they cannot all go in with their proper cross sections (we would only have QCD), so some processes are reduced with respect to others. To compensate for this reduction, a "weight" can be given to each event. In an analysis, this weight should be included when calculating numbers of events of a given process passing a certain selection, or when plotting a certain event variable integrated over various processes. • In principle, the "event weight" is given by the cross section times luminosity divided by the number of events of a given process that are in the soup. When a generator level filter has been applied, this artificially enhances the (effective) cross section, so the event weight should be multiplied by the filter efficiency. • In practice, for the CSA07 soup, the calculation of the event weights is non-trivial. The number of events of every process that actually make it into the soup has to be exactly known. Some processes are overlapping (e.g. Min Bias contains also soft QCD) and care needs to be taken to avoid double counting.
Photon + Jet Processes Pythia Alpgen Only selecting these processes from skim using processID