1 / 24

Update on tt-bar signal and background simulation

This update provides a summary of the last meeting, including information on MC@NLO simulation, running in ATLAS, comparison to LO generators, and issues with AlpGen and Pythia. It also discusses the top mass reconstruction, changes in the Pt(tt-bar) system in Pythia, and the W+jets background.

sshumate
Download Presentation

Update on tt-bar signal and background simulation

An Image/Link below is provided (as is) to download presentation Download Policy: Content on the Website is provided to you AS IS for your information and personal use and may not be sold / licensed / shared on other websites without getting consent from its author. Content is provided to you AS IS for your information and personal use only. Download presentation by click this link. While downloading, if for some reason you are not able to download a presentation, the publisher may have deleted the file from their server. During download, if you can't get a presentation, the file might be deleted by the publisher.

E N D

Presentation Transcript


  1. Update on tt-bar signal and background simulation Stan Bentvelsen

  2. Resume of last meeting • MC@NLO: • Matching NLO calculations of QCD process with parton shower MC simulation • Fully exclusive events generated • Hard emissions treated as in NLO • Soft emissions handled by MC shower (Herwig) • No ‘double counting’ between these two • Running in ATLAS: • Create event file using ‘runNLO’ program (extern) • Contains kinematic of hard NLO process • Interface to Herwig via McAtNLO_i

  3. 13.5% 86.5% Resume: Weights ‘standard’ tt production process -1706 • Weights: ±w • ‘unweighted’ events, up to a sign!(practically weight ±1) • Efficient event generation possible • NLO distributions (without MC showering) are non-physical tt production cross section MCatNLO: 842 pb HERWIG: 458 pb PYTHIA: 490 pb (nb: no consistent pdf’s!)

  4. Resume: Comparison to LO generators • Pt(tt system) • Herwig & MCatNLO agree at low Pt, • At large Pt MCatNLO ‘harder’ • PYTHIA completely off Same distribution on linear scale All distributions normalised to 1

  5. AlpGen ‘hard multiparton’ generator • Many hard processes – with extra n-jets (‘light jets’) • E.g.: tt+n-jets, W+n-jets • Exact (LO) matrix element • Alpgen generates file with hard scattering • To be fed into Herwig/Pythia shower MC’s • Problems: (AlpGen v1.3 & Herwig_i-00-01-18) • Compiler optimization problems on Linux gcc 3.2 • Works fine under gcc 2.96; subtle problem! • Solution: -Compiler optimization flag change to –O (was –O2) -Use f90 version of the generator • Interface in Herwig not comply to Alpgen V1.3 • On list t.b.d. for next release (do not know actual status) • Private version working

  6. Alpgen: tt+1jet • Inputs • Mtop=175 • 1 extra light jet • Jet: Pt>10, ||<2.5, R>0.4 • Initial grid 3 * 200000 • Events: 40.106 • Produced 60 samples • Production • Un-weighting to single lepton (e,,) decay • Effective : 293 pb • 1.9 106 events generated (8 10-4 efficiency) • 18.1% (351000) events pass first selection • ETmiss>20 GeV, lepton (e, ) Pt>20 and >=4 jets Pt>40

  7. tt-system alpgen affected by extra gluon Previously problems, now solved! Histograms normalized to unity Extra jet: Pt-min = 10 GeV |η| < 2.5 R>0.4 AlpGen tt+1-jet production Alpgen looks ok!

  8. Top mass reconstruction • Simple kinematic reconstruction • First selection: • Event ETmiss>20 GeV • 1 lepton (e,) with pT>20 GeV • At least 4 jets (cone size R=0.4) with ||<2.5 and pT>40 GeV • Reconstructed W: • |W(reco)-W(true)|<20 GeV • 1 b-tagged jet: Opening angle (b,W) < (b,l) • 2 b-tagged jets: Combination with maximum resulting Pt for top • ‘Commissioning’ (i.e. no b-tag):Exactly 4 jets

  9. Reconstructed top mass Changes wrt Herwig (selection wrt previous)

  10. Pt tt-bar system in Pythia • Suggestion by Sjöstrand: • Increase ISR phase space for Pythia generator • Set process scale Q2=s • MSTP(32) = 10 • Raise maximum scale of initial shower to s • MSTP(68) = 2 (default: maximum scale upto Q2) • Do not use cone restrictions from ISR to top quarks • MSTP(67) = 0 • Events generated; results not yet shown to author. Equiv. upto s?

  11. Various pythia options Pythia0: All 3 options set Pythia1: MSTP(32) = 10 Pythia2: MSTP(68) = 2 Pythia3: MSTP(67) = 0 Pythia ‘overshoot’ hard Pt spectrum by opening phase space Cone restriction little effect by itself Pt tt-bar system in Pythia • None of the Pythia options describe the hard Pt spectrum as in Herwig or MC@NLO (n.b: NLO ME calculations coincide at high Pt with MC@NLO)

  12. Pt tt-bar system in Pythia Changes wrt Standard pythia (selection wrt previous) ~20% variation

  13. Top mass with pythia • Large differences in Pt spectrum for various Pythia settings • Upto 20% difference in final selection efficiencies • Effect on resulting top mass less dramatic • Does these settings have consequences for other processes in Pythia? • Need to get opinion of Sjöstrand • No final conclusion on this yet

  14. W+jets background • Most important background: W+n jets • Leptonic decay of W, and n=4 extra jets • In Pythia only relevant process: qq’W (+q(g) ) • No ‘hard’ matrix element for 4 extra jets • I.e.: 3 or 4 extra jets need to be generated by • Fragmentation • Decays • Detector response • Reconstruction • MC@NLO has NLO qq’ W+X • No ‘hard’ matrix element for 4 extra jets • Generated 350k events, only 1 event passed first selection • lepton (e, ) Pt>20 and >=4 jets Pt>40 • Alpgen does have ‘hard’ matrix element for 4 extra jets Very unlikely and no reliable rate nor distributions

  15. Due to small generation efficiencies in Alpgen: Use local NIKHEF LCG grid Currently 30% of total LCG grid This will change soon Total 240 CPU’s Mix of PIII: 0.8, 1.2, 2.0 and 2.6 GHz machines NIKHEF data processing facility AlpGen jobs running!

  16. NIKHEF data processing facility • For alpgen event generation (+Atlfast): • Many tries to debug ‘job submission’ • Taking advantage of ‘empty farm’ • Total submitted jobs: 2303 • Total GHzHrs (equivalent hours on 1 GHz machine): 15469 (!) Large fraction of ‘playing around’ as well…

  17. Alpgen: W+4jets • Main use of background production • Inputs • W+4 extra light jets • Jet: Pt>10, ||<3.0, R>0.3 • No lepton cuts • Initial grid: 200000*3 • Events: 150·106 • Jobs: 198 • Production: • Un-weighting to W lepton (e,,) decay • Effective : 4390 pb • 108401 events generated (3.6 10-6 efficiency) • 2.57% (2784) events pass first selection • ETmiss>20 GeV, lepton (e, ) Pt>20 and >=4 jets Pt>40

  18. Alpgen: W+4jets (2) • Main use of background production • Inputs • W+4 extra light jets • Jet: Pt>10, ||<2.5, R>0.4 • No lepton cuts • Initial grid: 200000*3 • Events: 150·106 • Jobs: 98 • Lower maximum weight by factor 10 (?? Can I do this??) • Un-weighting to W lepton (e,,) decay • Effective : 2430 pb • 380740 events generated (2.6 10-5 efficiency) • 3.41% (13002) events pass first selection • ETmiss>20 GeV, lepton (e, ) Pt>20 and >=4 jets Pt>40

  19. Alpgen: W+4jets (3) • Main use of background production • Inputs • W+4 extra light jets • Jets: Pt>10, ||<2.5, R>0.4 • Lepton: Pt>30, ||<3.0, Etmiss>30. • Initial grid: 200000*3 • Events: 200·106 • Jobs: 100 • Lower maximum weight by factor 10 (?? Can I do this??) • Un-weighting to W lepton (e,,) decay • Effective : 106 pb • 39810 events generated (2 10-6 efficiency) • 25.8% (10264) events pass first selection • ETmiss>20 GeV, lepton (e, ) Pt>20 and >=4 jets Pt>40 Chosen too large for fair comparison with other data sets Data set still included in next plots just for comparison

  20. AlpGen W+4jet comparisons All histograms normalized to unity

  21. Reconstructed top mass • Normalized according to same luminosity • Large difference of Alpgen3 due to hard lepton, Pt>30 • Difference Alpgen1 and Alpgen2 only from  and R Some more work needed to check these statements. E.g. make harder cuts on parton level data sets 1 and 2 to see if coincides exactly with data set 3

  22. Luminosity: 10 pb-1 MC@NLO signal Alpgen1 sample Luminosity: 150 pb-1 MC@NLO signal Alpgen2 sample Top signal + background

  23. Top signal + background • Requiring 1 b-tag, 150 pb-1: • No mis-tag rate included…

  24. Request for DC2 (proposal) • Use for commissioning studies • Initial detector layout • Signal tt-bar events: 1 fb-1 (10%) • 2 event generators • MC@NLO 830k events (eff slightly less due to w<0) • Pythia 830k events (i.e. NLO normalisation!) • Full decay modes W’s • Background events: 200 pb-1 • Alpgen, according to Alpgen1 sample • Alpgen W+4jet 600k events • Only leptonic decays of W (e,)

More Related