light jet energy scale from Wjj See D. pallin: Tatra workshop ATLAS phys workshop (Rome) Top meetings http://agenda.cern.ch/askArchive.php?base=agenda&categ=a041267&id=a041267s4t4/transparencies http://agenda.cern.ch/askArchive.php?base=agenda&categ=a044738&id=a044738s11t4/transparencies http://agenda.cern.ch/askArchive.php?base=agenda&categ=a057774&id=a057774s1t3/moreinfo J schwindling Top meetings http://agenda.cern.ch/askArchive.php?base=agenda&categ=a057774&id=a057774s1t6/transparencies http://agenda.cern.ch/askArchive.php?base=agenda&categ=a061436&id=a061436s0t1/transparencies http://agenda.cern.ch/askArchive.php?base=agenda&categ=a062180&id=a062180s0t1/transparencies
Wjj sample in lepton+jets events Select an enriched W sample from ttbar (lepton+jets) events Selection with 1 or 2 btag typically 3000 W/fb-1 (2btag, _btag=0.6) x2 if 1 btag ( 80%-90%) purity (1 or 2 btag) Different Selection if 1btag W mass shift linked to JES Jets energies between 40 and 400 GeV j1 j2 l(e,m) n t1 t2 W1 W2 b1 b2 Rome AOD 500pb-1
Method 1 (D.Pallin) Principle to obtain Constraint 1 Mjj=MWPDG Constraint 2 the calibration function is universal No hypothesis on function , no MC inputs build W mass distributions in bin of jet energies extract the peak values for each bins (n bins) Deduce (E) from MW(E)
Method 1 W mass distributions MW Ejet (GeV) Rome AOD ‘cone 04’
Method 1: Deducing (E) from MW(E) Extraction of n values (Ebin) Several possible methods chi2 fit Iterative method (equalisation of MC/DATA distribution) Results (R) reproduce well the initial calibration (theo) as function of E 1% precision with 1fb-1 EPart / E E
Method 1: BUT the calibration function as function of E is not the true calibration E_jet / E_part % E with MC calib=1 Bias is within 1% above 40 GeV But need to be corrected huge effect below 40 GeV EPart / E Eparton E recons How to correct? origine of bias : purely statistical from : Ejet shape+ Ejet resolution negligeable if resolution(E)<<E important when resolution(E)~E easy to compute if shape and resolution known Shape given since W sample is 80% pure Resolution could be computed from W width (or obtain from Z,+jets samples)
Method :1 ongoing work extraction of jet energy resolution from W width evaluations of systematics transportation of calibration to other samples JES from 20 GeV Increase size of initial sample Note in preparation
Method 2: templates (J Schwindling) h: 297 / √E 11 mrad Idea more or less « à la CDF » Will be presented in jet/ETmiss meeting in Trigger/Physics week end of May Idea: generate mjj template histograms by smearing W qq quarks with: Expected angular resolution b x expected energy resolution energy correlation between jets energy scale a Resolution (mrad) j: 224 / √E 10 mrad Expected angular resolution Fitted on Rome ttbar sample 4100 Equark (GeV) Expected Energy resolution s(Ej – Eq) (GeV) s(E) = 3.8 GeV + 0.063xE Eq (GeV) Jet - quark s(E1) 10.7 s(E2) 9.1 s(E1) s(E2) 14.0 s(E1+E2) 16.5 (Error bars ~ 0.2 GeV) + 18 %
Meth2 :Expected performance Fit both energy scale a and resolution b With enough statisitics, can do this as a function of energy (and/or h ?) Example: apply templates on independent set of smeared quarks with: Different energy scale(s) Different energy resolution Results (templates smeared quarks): Can fit average energy scale to ~0.4% with 1fb-1 With more statistics (10 fb-1), could fit energy scale as a function of E In principle allows also to check energy resolution as a function of E Injected a 1% Reconstructed a Injected resolution 3.0 + 0.080xE Fitted resolution Default resolution 3.8 + 0.063xE
Meth 2: Towards real data PT cut = 40 GeV All jj combinations Only 2 light jets + 150 < mjjb < 200 Only 2 light jets Try to fit the templates (smeared PYTHIA quarks) on jj pairs from MC@NLO + Geant 4 + reconstruction (Rome sample 4100) Event selection (tt jjb lnb): 1 isolated lepton with PT > 20 GeV ETmiss > 20 GeV 2 b-tagged jets with PT > 40 GeV 2 and only 2 light jets with PT > 40 GeV 150 GeV < mjjb < 200 GeV purity ~ 83 %, ~ 1200 jj pairs for 500 pb-1 Expect energy scale (Ejet / Eq) ~ 0.96 – 0.97 b = 1
Meth 2: Conclusion / questions Results (templates sample 4100) a = 0.958 ± 0.005 b = 1.07 ± 0.05 Fit of a seems very stable: ~ independent of qq smearing, ~ independent of mtop (± 0.4 % uncertainty on Escale for Dmtop = ± 5 GeV), independent of combinatorial background b sensitive to angular resolution, correlation between jet energies … cannot claim to measure energy resolution We need to understand better: Origin of energy correlation between jets Angular correlation ? Influence of underlying event E scale versus pileup