Transcript Slide 1
V+jets m/e studies with 21X: Status and plans General EWK Meeting 20-01-09 Christopher Rogan California Institute of Technology for Marcella Bona, Emanuele Di Marco, Ran Feldesh, Joseph D. Lykken, Paolo Meridiani, C.R., Chiara Rovelli, Maurizio Pierini, Ilaria Segoni, Maria Spiropulu, Thiago Tomei, Lukas Vanelderen, Marco Zanetti ◉ goal of the analyses: ◎ candle with the Z mass: validate e and m reconstruction ◎ application examples: ◍ MC tuning with the dN/dnjets and dN/dpT(Z) ◍ check and correction of MET ◍ ratios good for start-up: systematics cancel ◉ references ◎ CMS notes for 16X analyses: Z+jets & W+jets Alpgen Validation: CMS AN-2008/091 ◉ Z(ee)+jets Candle Analysis: CMS AN-2008/092 ◉ Z(mm)+jets Candle Analysis: CMS AN-2008/095 ◉ W(en)/Z(ee)+jets Ratio Analysis: CMS AN-2008/096 ◉ W(mn)/Z(mm)+jets Ratio Analysis: CMS AN-2008/105 ◉ ◎ previous EWK talks on 16X analysis: summary by M.Pierini Nov 11th 2008 ◉ note release: E. Di Marco, C. Rogan, and I. Segoni: Oct 17th 2008 ◉ ◎ V+jets talks: e and m analyses: I. Segoni and M.Zanetti: Nov 14th 2008 ◉ m selection: M. Bona: Nov 28th 2008 ◉ V+j so far Christopher Rogan - EWK Meeting 20-01-09 3 ◉ samples: 2.1.X and Fall09 ◎ signals, other Z/W decays andtt events: MADGRAPH ◎ QCD: “InclusiveMu15” sample (muon enriched QCD) ◉ general strategy for the analyses single non isolated muon trigger muon selection: optimization in the next slides Z mass window [60, 120] GeV/c2 closest-to-the-leading-muon PV selection jet clustering (calo jets and track jets: SisCone with 0.5) with E/pT and h cuts (30/15 GeV and 3/2.4, respectively) ◎ 2D maximum likelihood fit: Z mass and sin(angleMET, MHT) for Z+jets; transverse mass and sin(angleTS-jet) ◎ floating the shape parameters for low jet multiplicities and fixing for the high multiplicities ◎ sPlots for verify shapes on data ◎ ◎ ◎ ◎ ◎ ◉ muon selection in 16X ◎ ◎ ◎ only global muons pT > 15 GeV/c vertex variables and isolation variables simultaneous optimization: ◉ vertex variables: Dxy/s(Dxy) and Dz(m-PV) ◉ isolation variables: tracking isolation i piT/pmT ECAL isolation R<0.35 ETECAL HCAL isolation R<0.5 ETHCAL ◍ extensive study to select the variables and to establish the optimization procedure: done in 16X and now repeated in 21X production, the PV is displaced in x so we can calculate the distance wrt the PV but the error is calculated wrt (0,0,0) so we started looking at: Dxy better for the electrons that have typically poorer tracking resolution about the same discrimination for muons (wrt the significance) Hcal isolation: H0 excluded and zero suppression introduced Cut variables R<0.35 ETECAL i piT/pmT Dz(m-PV) R<0.5 ETHCAL Dxy Christopher Rogan - EWK Meeting 20-01-09 6 Optimization on the W selection Optimize the Punzi Significance: Consider all backgrounds and scan over all cut values simultaneously, with rates normalized to 100 pb-1 Consider only events in ML fit ‘signal region’ Christopher Rogan - VecBos + jets 15-01-09 7 R<0.35 ETECAL <2.25 GeV R<0.5 ETHCAL <2.8 GeV Optimization scans requiring: i piT/pmT <0.15 Christopher Rogan - VecBos + jets 15-01-09 8 Optimization scans requiring: Dz(m-PV) <0.11 cm Christopher Rogan - VecBos + jets 15-01-09 Dxy(PV) <0.005 cm 9 Optimized cut values Calo jets Track jets Values used 0.15 0.2 0.2 Values used 0.11 0.11 0.11 2.0 2.0 2.0 3.1 4.5 4.5 0.1 0.11 0.11 0.0045 0.0045 0.0045 (0.11) 2.25 2.75 2.5 (1.6) 2.8 2.8 4.1 4.0 (6.0) 0.11 0.1 0.11 (0.08) 0.005 0.005 0.005 (n.a.) Christopher Rogan - VecBos + jets 15-01-09 10 Optimized cut values Calo jets 2.8 Track jets Values used 0.2 0.15 0.2 2.25 2.75 2.5 Values used 0.11 0.11 0.11 2.5 2.0 2.0 2.0 4.5 4.0 2.5 4.5 4.5 0.11 0.1 0.11 0.1 0.11 0.11 0.0055 0.005 0.005 0.0045 0.0045 0.0045 Christopher Rogan - EWK Meeting 20-01-09 11 Changing muon pt cut W selection Christopher Rogan - EWK Meeting 20-01-09 12 Changing muon pt cut Z selection Christopher Rogan - EWK Meeting 20-01-09 13 Analysis strategy Common requirements: Single non isolated HLT trigger (HLT_Ele15_LW_L1R) pixel-match GSF electron reconstruction electron identification* electron isolation* electron - PV compatibility calo-jet clustering (SisCone, ΔR=0.5) electron(s) from W(Z) cleaning from calo-jets jet counting * for Z, asymmetric id+iso W-specific requirements: Z-specific requirements: exactly 1 electron (against Z+j) Z mass veto (decouple W+j dataset from ≥2 electrons (against W+j) Z+j) Z mass range (decouple Z+j dataset from MET > 20 GeV (against QCD) W+j) mT(W) > 30 GeV (against QCD) yields and ratio determination: Maximum Likelihood fit efficiency correction of yields, if needed Christopher Rogan - EWK Meeting 20-01-09 14 why re-iteration with 2.1.X ? electron reconstruction details changed w.r.t. 1.6.X. (i.e. CSA07) tracker material changed some electron identification related to tracker - ECAL match can be changed among others, H/E definition changed pre-shower is excluded HO is excluded and zero-suppression in HCAL has been introduced H/E is affected HCAL isolation is affected PV is displaced in x w.r.t. (0,0,0) Disclaimer! In this iteration, possibly no new variables w.r.t. the ones used in CSA07 analysis are used. Various new quantities have been developed and released, will be tested in W+j/Z+j @ next round Christopher Rogan - EWK Meeting 20-01-09 15 Electron identification We use the set of variables from egamma POG, with selection depending on electron classification (based on fbrem) tracker-ECAL match ECAL-HCAL match ECAL cluster shapes we compare 2.1.X - 1.6.X and re-optimize selection for golden / showering both barrel and endcap this defines the W electron and the first Z electron (i.e. tight) loose egamma identification on the other (asymmetric ID) Christopher Rogan - EWK Meeting 20-01-09 16 Efficiency of ele - id Cuts optimized simultaneously to maximize W signal significance against QCD QCD e.m. enriched + b,c→electrons used The optimized cuts Single electron efficiency on W+≥0j selection ε(CSA07) ε(Summer08)-*- total eID: 75.5 +/- 0.2 79.0 +/- 0.2 egamma loose the optimization is driven by W+j. For Z+j, the highest pT electron is identified with the same criteria to allow efficiency cancelation in Wj/Zj the ID of the second is based on the same variables, but looser Christopher Rogan - EWK Meeting 20-01-09 17 S and B yields (W+j) yields, normalized in 100 pb-1, up to electron definition. large stat error Christopher Rogan - EWK Meeting 20-01-09 18 S and B yields (Z+j) yields, normalized in 100 pb-1, up to electron definition Christopher Rogan - EWK Meeting 20-01-09 19 Z+jets (MADGRAPH) m(e+e-) lineshapes ≥1 jet ≥2 jet ≥3 jet Cruijff = Gaussian + exp tails for signal, parabola for inclusive bkg tt+jets (MADGRAPH) consistent parameters for signal lineshape as a function of jet multiplicity Christopher Rogan - EWK Meeting 20-01-09 20 mT(W) lineshapes W+jets tt+jets Z+jets QCD ≥2 jet ≥1 jet Cafu = Crystal ball + Gaussian with same mean decreasing importance Christopher Rogan - EWK Meeting 20-01-09 21 slope correction e-Slope=6.8±0.5 W/Z = 1.09±0.20 Z+j e-Slope=7.5±1.2 Yields efficiency-corrected W+j Yields uncorrected e-Slope=6.9±1.1 e-Slope=6.8±0.5 W/Z = 1.01±0.19 ⇒ eff changes the slope, but effect on double ratio is within uncertainty Christopher Rogan - EWK Meeting 20-01-09 22 slope correction e-Slope=8.0±0.4 W/Z = 0.91±0.05 Z+j e-Slope=7.32±0.09 Yields efficiency-corrected W+j Yields uncorrected e-Slope=6.82±0.08 e-Slope=7.4±0.3 W/Z = 0.93±0.04 ⇒ eff changes the slope, but effect on double ratio is within uncertainty Christopher Rogan - EWK Meeting 20-01-09 23 ◉ for the candle analysis: ◎ we start from the optimization values obtained for the ratio in the W selection we can afford to loosen the cut to increase efficiency ◉ muon pT > 10 GeV/c ◉ vertex variables: Dxy <0.02 cm & Dz(m-PV) <0.15 cm ◉ isolation: i piT/pmT <0.30 (on the second leg) Christopher Rogan - VecBos + jets 15-01-09 24 Background ≥1jet ≥1jet Christopher Rogan - EWK Meeting 20-01-09 25 Signal ≥1jet ≥2jet ≥3jet ≥4jet Christopher Rogan - VecBos + jets 15-01-09 26 Background ≥1jet ≥2jet ≥3jet ≥4jet Christopher Rogan - VecBos + jets 15-01-09 27 Signal ≥1jet ≥2jet ≥3jet ≥4jet Christopher Rogan - VecBos + jets 15-01-09 28 Background ≥1jet ≥3jet ≥2jet ≥4jet Christopher Rogan - VecBos + jets 15-01-09 29 ◉ for the electron candle analysis: we can afford to loosen the cut to increase efficiency ◉ electron pT > 10 GeV/c ◉ loose electron ID on both legs ◉ Best Z candidate choice (highest pt electrons) ◉ vertex variables: Dxy <0.04 cm & Dz(m-PV) <0.12 cm ◉ isolation: i piT/pmT <0.15 (on both legs) Christopher Rogan - VecBos + jets 15-01-09 30 Signal ≥1jet ≥2jet ≥1jet ≥2jet Christopher Rogan - EWK Meeting 20-01-09 31 Background ≥1jet ≥2jet ≥1jet ≥2jet Christopher Rogan - EWK Meeting 20-01-09 32 EXTRA SLIDES Christopher Rogan - EWK Meeting 20-01-09 33 Madgraph/Alpgen Production changes For tt+j parton pt 20, matched at 30 MADGRAPH – KT MLM matching x-section 317pb @ 10 TeV, x-section 750 pb @ 14 TeV Parton pt 70, matched at 20 GeV, ALPGEN –MLM (old, IC) Matching x-sec 450 pb @14 TeV – if 20 GeV parton pt x-sec 754 pb@14 TeV, x-section 270 pb@ 10 TeV [15% compared to MAD, MAD>ALPGEN but different scales, PDFs see below] Scale (m_top)^2 at MADGRAPH Scale (M_T)^2=sum(m_top^2+P_T^2) ALPGEN Lower Q^2 pushes the cross section up (MAD>ALPG) (10%) PDF CTEQ6L NLO and corresponding a_s change in MADGRPAH PDF CTEQ5L LO ALPGEN NLO PDF pushes x-section up (MAD>ALPG) (10%) Christopher Rogan - EWK Meeting 20-01-09 34 34 Comparison of isolation variables: CSA07 - FALL08 Christopher Rogan - VecBos + jets 15-01-09 35 Comparison of vertex variables: CSA07 - FALL08 significance Christopher Rogan - VecBos + jets 15-01-09 36 isolation variables: tracking isolation normalized: i piT/pmT ECAL isolation: R<0.35 ETECAL HCAL isolation: R<0.5 ETHCAL ◉ tracking isolation: tracks within DR < 0.5 ◈ at least 5 associated hits ◈ |Dz(track-PV)| < 0.1 cm |Dxy/s(Dxy)| < 5 ◈ 500 MeV/c < pT < 500 GeV/c ◈ ECAL isolation: EM calotowers in the 0.35 cone around the m m energy is not subtracted HCAL isolation: hadronic calotowers in the 0.5 cone Christopher Rogan - VecBos + jets 15-01-09 37 W selection efficiencies Christopher Rogan - VecBos + jets 15-01-09 38 Z selection efficiencies Christopher Rogan - VecBos + jets 15-01-09 39 Eseed/Pout Major point: track pout estimation depends by the tracker material knowledge barrel endcap Christopher Rogan - EWK Meeting 20-01-09 40 Δηvtx GSF-track - ECAL supercluster match larger tails for non-showering non-showering endcap bumps for η>0, η<0: tilt changed in the geometry? barrel endcap Christopher Rogan - EWK Meeting 20-01-09 41 ΔΦvtx GSF-track - ECAL supercluster match barrel endcap Christopher Rogan - EWK Meeting 20-01-09 42 H/E CSA07: H/E = sum(H rechits) / supercluster energy in cone ΔR=0.1 no HCAL zero-suppression Summer08: H/E = max (H rechit / basic cluster ), the H rechit is the one above the ECAL basic cluster HCAL zero - suppression barrel endcap Christopher Rogan - EWK Meeting 20-01-09 43 s9/s25 Cluster shape variable barrel endcap Christopher Rogan - EWK Meeting 20-01-09 44 σηη Cluster shape variable (cluster width) changed thresholds? barrel endcap Christopher Rogan - EWK Meeting 20-01-09 45 R<0.35 ETECAL Dz(m-PV) R<0.5 ETHCAL i piT/pmT Dxy(PV) Optimization scan requiring: Christopher Rogan - VecBos + jets 15-01-09 46 R<0.35 ETECAL Dz(m-PV) R<0.5 ETHCAL i piT/pmT Dxy(PV) Optimization scan requiring: Christopher Rogan - VecBos + jets 15-01-09 47 R<0.35 ETECAL Dz(m-PV) R<0.5 ETHCAL i piT/pmT Dxy(PV) Optimization scan requiring: Christopher Rogan - VecBos + jets 15-01-09 48 Optimization Some comments: Currently we are not including a sphericity cut or MET cut in the optimization (we will include this once I finish the W MET calibration using the Z candle) As a result, we are cutting more tightly on these five variables relative to what we will do when including MET or sphericity - I anticipate that the signal efficiency will go up with Ultimately, we will also do the optimization in the signal region of the ‘angular variable’ which we use in the fit - we can’t do this until we have the W MET calibration. Hence these are probably not the ‘final’ values, but they demonstrate the procedure/philosophy of the optimization Christopher Rogan - EWK Meeting 20-01-09 49 what looks different Most of the electron ID variables looks very similar to CSA07. (see backup) larger differences in: H/E, σηη H/E: definition changed, was H/E in a cone ΔR=0.1, now is HoverECALseed/Eseed σηη: noise threshold for covariance matrix calculation changed (w0=4.2 → 4.7): expected larger width σηη barrel H/E barrel Christopher Rogan - EWK Meeting 20-01-09 50 what’s looks different: Δηvtx double peak for golden (i.e. when the super-cluster = 1 basic cluster) in the endcap peak>(<)0 for η<(>)0 true also evaluating Δη=ηclus-ηgenele it is a cluster problem difference 1.6.X - 2.1.X: 2.1.X has not the preshower: the cluster position (depth in crystal) accounts for it with a parameter hypothesis check. In 3.0.X preshower is back Δηvtx(3.0.X) ηclus-ηgenele the double peak disappears η Christopher Rogan - EWK Meeting 20-01-09 51 electron isolation Comparison of the variables used in the CSA07 notes Traker: sum(pTtracks)/pTele , with internal cone veto internal cone veto found to be good and robust external cone ΔR=0.4: all similar, chosen the intermediate one ECAL: sum(ET5x5 crystals-ETsuper-cluster)/ETsuper-cluster in a cone ΔR=0.4 electron footprint ETsuper-cluster not optimal, move to rechits isolation soon tracker veto cone ECAL rechits removal HCAL: sum(ETrechits)/ETsuper-cluster in a cone ΔR=0.4 See summary from M. Le Bourgeois: http://indico.cern.ch/contributionDisplay.py?contribId=4&confId=4 3146 Christopher Rogan - EWK Meeting 20-01-09 52 isolation: distributions tracker HCAL ECAL ε(CSA07) isolation *- tracker: *- ECAL: *- HCAL: 91.8 +/- 0.4 90.3 +/- 0.4 93.6 +/- 0.4 Christopher Rogan - EWK Meeting 20-01-09 ε(Summer08)- 89.025 +/- 0.294.748 +/- 0.291.989 +/- 0.2 53 vertexing variables Used to suppress events not coming from Primary Vertex (PV): dz = zele-zPV dxy = distance in transverse plane w.r.t. PV RMS=210μm RMS=66μm Christopher Rogan - EWK Meeting 20-01-09 54 jet multiplicity dependency (1) -estimate on Fall08 samples with HLT efficiency variations in different multiplicities change the slope V+(n-1)j / V+(n)j but mostly cancels out in double ratio W+nj / Z+nj (W ele and 1st ele of Z have the same selection) Z+jets: no variations at all Christopher Rogan - EWK Meeting 20-01-09 55 jet multiplicity dependency (2) re-estimate on Fall08 samples with HLT W+jets: 2% drop at each added jet Christopher Rogan - EWK Meeting 20-01-09 56 maximum likelihood fit signal, backgrounds yields extracted on data with extended, maximum likelihood fit for Z+≥1jet also Z lineshape parameters extracted on data Z+jets: 1dim fit: P=PDF(mee) total number of events entering the fit (i.e. extended likelihood) W+jets: 1dim fit: P=PDF(mTW) Ni=signal and backgrounds yields Z+jets: i=signal, tt W+jets: i=signal, tt+QCD, Z+jets Christopher Rogan - EWK Meeting 20-01-09 57 Tracking Isolation i piT/pmT comparison wrt the jet bin Christopher Rogan - VecBos + jets 15-01-09 58 ECAL Isolation R<0.35 ETECAL comparison wrt the jet bin Christopher Rogan - VecBos + jets 15-01-09 59 HCAL Isolation R<0.5 ETHCAL comparison wrt the jet bin Christopher Rogan - VecBos + jets 15-01-09 60 vertex variable: Dz(m-PV) comparison wrt the jet bin Christopher Rogan - VecBos + jets 15-01-09 61 vertex variable: Dxy comparison wrt the jet bin Christopher Rogan - VecBos + jets 15-01-09 62 W selection results Calo jets 16X numbers: CAVEAT different cuts and different energy! Christopher Rogan - VecBos + jets 15-01-09 63 W selection results Track jets Christopher Rogan - VecBos + jets 15-01-09 64 Z selection results Calo jets 16X numbers: CAVEAT different cuts and different energy! Christopher Rogan - VecBos + jets 15-01-09 65 Z selection results Track jets Christopher Rogan - VecBos + jets 15-01-09 66 W slopes Calo jets Christopher Rogan - EWK Meeting 20-01-09 67 W slopes Track jets Christopher Rogan - EWK Meeting 20-01-09 68 Z slopes Calo jets Christopher Rogan - EWK Meeting 20-01-09 69 Z slopes Track jets Christopher Rogan - EWK Meeting 20-01-09 70 Calo jets Track jets Christopher Rogan - EWK Meeting 20-01-09 71 W/Z Ratio Christopher Rogan - EWK Meeting 20-01-09 72