Deteration of the Top- Quark Mass frohe Distribution in Dileptonic Top-Pair Events at s= 8 TeV.1.14 - Motivation - - Event Selection & mlb Definition- - Extraction: Method - - Results - - Summary - 1
Motivation Inclusive tt cross sections used to extract mt,pole at NNLO Taking advantage of different slope of σpred(mt) and σmeas(mt) Next step: Extract from differential cross sections (PLB 78 (14) 496) Possibly more sensitive to mt, more clear theoretical interpretation Step to simultaneous mt, pdf, αs extraction Observable sensitive to mt, measurement + prediction Reference: PAS-TOP4-14
Event Selection and mlb Definition 3
Event Selection Full 1, 8 TeV dataset (19.7 fb ) Signal: MadGraph+Pythia (7 masspoints) Selection opposite charged isolated leptons eµ Remove low m ll < GeV high p T jets 1 b-tagged jet N evt obs pred 18 16 14 1 8 6 4 1.5 1.5 3 data ttv/γ non-w/z Wjets tw/tw ll DY/VV tt signal MC syst+stat 1 3 4 5 N b N evt /GeV obs pred 6 5 4 3 data ttv/γ non-w/z Wjets tw/tw ll DY/VV tt signal MC syst+stat 1.5 1.5 4 6 8 14 16 18 p T,b N evt /GeV obs pred 7 data 6 ttv/γ non-w/z 5 Wjets tw/tw ll 4 DY/VV 3 tt signal MC syst+stat 1.5 p 1.5 4 6 8 14 16 18 p T,l1 4
Observable mlb, Definition on generator and reconstruction level leading b: Combines production rate and decay shape Leading b-jet: smaller JES uncertainties Lepton kinematics known to high precision Visible phase space (vis. PS) b-quark froop / b-jet η <.4, pt > 3 GeV, : use smaller of combinations x l: Status 3 ME leptons froop (e/µ/tau( e,µ)) / selected oppositely charged lepton pair η <.4, p T > GeV N evt /GeV /N mt N mt 45 4 35 3 5 15 5 1.5 1.5 19.7 fb =17.5 GeV =166.5 GeV =169.5 GeV =175.5 GeV =178.5 GeV 5 15 5 3 35 (8 TeV) Simulation MadGraph+Pythia normalized to NNLO+NNLL 5
Observable mlb, () Good data/mc agreement Low background Only small contribution of migrations from outside visible PS (reco) 35 3 5 15 5 Simulation.1.8.6.4. N evt /GeV obs pred 5 4 3 19.7 fb data ttv/γ non-w/z Wjets tw/tw ll DY/VV tt signal MC syst+stat 1.5 1.5 5 15 5 3 35 (8 TeV) τ decays, efficiencies included in response matrix 4 6 8 14 16 18 (gen) Strong correlation between reconstructed and generated mlb, 6
Folding Use reconstruction level distributions Compare to MadGraph on reconstruction level Fold theory: Separate matrix for each systematic variation / mt Reflects: Detector modeling uncertainties Signal modeling uncertainties (reconstruction efficiencies) (reco) 35 3 5 15 5 Simulation 4 6 8 14 16 18 (gen).1.8.6.4. No additional unfolding/regularization uncertainties No stat. correlations between bins of reconstructed distribution Here: Fold theory within visible phase space 7
Top-Mass Extraction: Method 8
Modeling Likelihood (mt) Fit mt-dependence of prediction, Npred, with nd order polynomial Includes backgrounds Use statistical uncertainty (mostly background) from central point σpred Parametrization uncertainties not included here ( see systematics) N evt 38 36 34 3 3 3 data prediction Data yield Ndata: constant with σdata 8 166 168 17 17 174 176 178 18 χ 5 4 3 Define estimator 1 - -3-4 -5 166 168 17 17 174 176 178 18 Extract top mass from imum of χ (mt) For total rate or in bins of any distribution (mlb) 9
Method Overview N evt 45 4 35 3 5 3 15 38 5 36 1.5 34 1.5 3 N evt /GeV N mt /N mt m t =169.5 GeV =175.5 GeV data =178.5 GeV prediction Simulation =17.5 GeV =166.5 GeV 5 15 5 3 35 prediction: either or + data N evt 3 38 36 34 3 3 8 data prediction 3 8 166 168 17 17 174 176 178 18 For each bin: 166 168 17 17 174 176 178 18 Derive dependence of pred (fit) Construct χ (mt) Combine bins No stat correlations sum χ over bins Detere Repeat for each syst. variation rate shape & rate shape χ 5 4 3 1 - -3-4 -5 166 168 17 17 174 176 178 18
Top-Mass Extraction: Results 11
Extraction of mt (MadGraph+Pythia) N evt /GeV 1/N tot.16.14.1.1.8.6.4. shape χ 165 17 175 18 data 5 15 5 3 35 8 6 4 =178.5 GeV =17.5 GeV =166.5 GeV N evt /GeV 6 5 4 3 shape & rate χ 5 15 165 17 175 18 data 5 15 5 3 35 5 =178.5 GeV =17.5 GeV =166.5 GeV only stat uncertainties MadGraph+Pythia + background processes Extraction performed for shape only, shape & rate and rate-only Repeated for each systematic uncertainty 1
Extraction using MCFM Use MCFM to simulate mlb, distribution in. vis PS Fold to detector level and extract top mass normalized (rate underestimated) Pure MCFM NLO / LO NLO production + decay 1 Real and virt. corrections in decay & production NLO production + decay Real and virt. corrections in production only (Closest to MadGraph ME level) Plain LO (no FSR) decay decay 1 Uncertainties similar to MadGraph+Pythia 13
Results Mlb shows low mt-sensitivity wrt. higher orders in production Overall consistent results from shape, shape & rate and rate based extraction Doant uncertainties (experimental/theory): Rate: normalization uncertainties, JES signal normalization Shape & rate: partial cancellation of JES uncertainties signal normalization Shape smallest total uncertainty JES Q scale N evt /GeV 1/N tot.16.14.1.1.8.6.4. χ 165 17 175 18 data 5 15 5 3 35 8 6 4 =178.5 GeV =17.5 GeV =166.5 GeV 14
Summary 15
Summary Extracted top mass from reco-level distributions & total rate with MadGraph+Pythia From shape: = 17.3 ±1.3 GeV All MadGraph+Pythia mass -values consistent Partial cancellation of uncertainties in shape+rate extraction N evt /GeV 1/N tot.16.14.1.1.8.6.4. χ 5 15 5 3 35 8 6 4 165 17 175 18 data =178.5 GeV =17.5 GeV =166.5 GeV Extracted using folded MCFM LO/NLO No unfolding/regularization effects Low sensitivity of to production mechanism Shift due to radiation in decay 16
BACKUP 17
Closure Tests: Pseudo-Experiments Based on fitted Npred(mt) (per bin) Data toys: Random yield (Poisson) around Npred Same statistics as data Prediction toys: Signal - Random yield (Poisson) with tt MC statistics - Uncorrelated for each mass point Background - Random yield (Poisson) with background MC statistics - Correlated for each mass point Scaled back to luosity and added Extract mt and evaluate mt,in vs. mt,out (x 5 ) No bias Stat uncertainties estimated correctly private work rate =169.5 GeV peak:.1, width:.98 =17.5 GeV peak:, width:.98 =175.5 GeV peak: -., width:.98-6 -4-4 6,out -,in / Δ stat rate & shape =169.5 GeV peak:.5, width:.99 =17.5 GeV peak:.5, width:.98 =175.5 GeV peak:.4, width:.98-6 -4-4 6,out -,in / Δ stat shape =169.5 GeV peak:.3, width:.99 =17.5 GeV peak:.4, width: 1.1 =175.5 GeV peak:.3, width: 1-6 -4-4 6,out -,in / Δ stat 18