Measurement of the Top quark mass

Transcription

1 Measurement of the Top quark mass Martijn Mulders CERN Tevatron Turns Twenty

2 The top quark Superficially like the other quarks 40x heavier than b: mt > mw : t --> bw Lifetime shorter than hadronization time Yukawa coupling ~ 1 Only quark with important coupling to the Higgs Is top special? or is it the only natural quark? 2

3 Probing the EW vacuum µ GF EW measurements (α, GF, mz, mw,... mt) Precise predictions Standard Model: πα 1 GF = (1 mw / mz ) mw 1 r W e νe νµ µ W Sensitive to radiative corrections t νµ νe H W W W b δmw mt2 δmw ln(mh) Used for predictions on top Constraints on Higgs, new physics!! W b t W W H W 3 e

4 The road to the Higgs... PhD at NIKHEF: W mass, DELPHI Top mass, D0, Fermilab LHC CERN 4

5 'History' of the top mass... 5

6 D0 CDF Tevatron Linac Booster p-bar Main Injector/ Recycler --> NEW: Electron Cooling! 6

7 total recorded ~1.1 fb-1 Results 7

8 Integrated Luminosity Projection Design We are here Base 8

9 Top production Top quark pair production via strong interaction 85% Cacciari etal, JHEP 0404:068, pb (1.96TeV, mt=175gev/c2) RunII 30% higher than RunI 15% Single top quark production via weak interaction s-channel: 0.88pb Sullivan, Phys.Rev.D70:114012,2004 t-channel: 1.98pb 9

10 Top quark decay & identification All jet 46% t Wb 100% Need to reconstruct and identify Electrons, muons, jets, b-jets and missing transverse energy Exploit : b-tagging, and/or Kinematic distributions dilepton (e+µ) 4.5% Lepton + jet 29% 10

11 DØ & CDF Tracking in magnetic field Precision tracking with silicon Calorimeters Muon chambers CDF Excellent Hermeticity Excellent tracking resolution 11

12 CDF double b-tagged event 12

13 Let s see what we have DØ RunII Preliminary, 363pb-1 for σtt= 7 pb 1 tag 2 tags 3j, 1tag 4j, 1tag 3j, 2tag 4j, 2tag Expected bkg 71 ± 9 22 ± 3 7±1 1.5±0.3 Observed events σ tt = ( stat + syst ) ± 0.5(lumi ) pb 13

14 Without b-tagging: Use of kinematical distributions HT = scalar sum of Jet pt's (watch out: correlated with Top mass and Jet Energy Scale!!) two examples use seven of these event shape variables Neural Net 14

15 topological X-section Fit to neural network output for top and W+jets background: Sensitivity similar to btagged analysis, but larger sample used CDF 347 pb-1 σ tt = 6.3 ± 0.8(stat ) ± 1.0(syst + lumi)pb 15

16 Top firmly established CDF 3 publications 16

17 CDF 17

18 W Helicity from t Wb Decays Examines the nature of the twb vertex, probing the structure of weak interactions at energy scales near EWSB Stringent test of SM and its V-A type of interaction. Uses boosted W from top decays mu-mu 18

19 W helicity result dilepton f+ = (stat) (syst) CDF F0= (stat+syst) F+ < 95% C.L. (162 pb-1) Mlb l+jets Mlb l+jets F+ = / 0.13 (stat) +/ 0.07 (syst) F+ < 95% CL (230 pb-1) PRD 72, (2005) 19

20 R You Standard Model Top? Probing the assumption Br(t Wb)=1 Br (t Wb) Vtb 2 2 R= = = Vtb Br (t Wq ) Vtb + Vts + Vtd q = b, s or d-quark Vtb = to C.L. R= to True in SM assuming three generations of quarks 20

21 Measurement Use the ability to identify jets with a distinguished secondary vertex: b-tagging Njet=3 The number of b-tagged jets depends strongly on R and btagging efficiency We classify the ttbar sample based on the number of btagged jets The relative rates of events with 0/1/2 b-tags are very sensitive to R Br(t Wb)=1 and σtt=7 pb Result is obtained from a binned maximum likelihood fit to data for Njet =3 and Njet =4 Simultaneous fit to R and cross section 21

22 Result 230 pb-1 Br ( t Wb ) 19 = ( stat + syst ) Br (t Wq ) The most precise measurement to date σ tt = ( stat + syst ) ± 0.5(lumi ) pb Model independent measurement (stat) (syst) CDF 22

23 !! w e N Top or exotic quark? First measurement of top quark charge tt-->w+w-bb allows for t charge 2e/3 or 4e/3 Result: 4e/3 excluded at 94% CL compatible with 2e/3 lepton+jets channel, 17 double tagged events, 366pb-1 Kinematic fit, mt=175 GeV, correct jet assignment in 79% of events Measure jet charge of b-jet and bbar-jet, t-->bw, W charge known from leptonic decay--> top charge

24 What about single top? Much more challenging than ttbar! B-tag and extensive use of kinematical distributions tt likelihood discriminant / NN / decision tree W+jets 24

25 newest (best) result Getting close... s-channel: 0.88pb? t-channel: 1.98pb? CDF (162pb-1): s: < % C.L. t: < % C.L. s+t: < % C.L. PRD D0 (230pb-1) Neural Network s: < % C.L. t: < % C.L. PLB 622, 265 (2005) 25

26 Top physics is sexy! Many analyses!!! 26

27 Back to the Top Mass 27

28 Lepton+jets channel Signature Selection One isolated lepton pt>20 GeV four jets pt>20 GeV ETmiss>20 GeV Features: Relatively high Br Manageable background Perfect for studies of top properties Backgrounds W + jets Multijet 28

29 Extracting the Mass: Use kinematic constraints Improve measurement beyond detector resolution, solve ν Σpx = Σpy = 0 mt = mt, mw+ = mw- = 80.4 GeV But multiple jet/neutrino solutions: 2-12 possible jet assignments Double neutrino solution x 2...?? choose one? 0 tag: 12 jet assignments 1 tag: 6 jet assignments 2 tag: 2 jet assignments Signal or background? Different approaches: Template, Matrix Element, or Ideogram method... 29

30 Traditional Template method Data Plot lowest χ2 mass Wbb MC tt MC Datasets Mass fitter Signal/background templates Data Likelihood fit Result 30

31 Traditional Template Method Match data histogram to MC 'Templates' for different top masses... choose signal + BG Template that fit best (likelihood fit) Advantage: Auto-calibration, nice plot Dis-advantage: Limited information per event (typically 1 number: mass) May require large MC statistics 31

32 Run I: Matrix Element Method probability to observe a set of kinematic variables x for a given top mass dnσ is the differential cross section Contains matrix element squared W(x,y) is the probability that a parton level set of variables y will be measured as a set of variables x 1 n Psgn ( x; mt ) = d σ ( y ; mt ) dq1 dq2 f ( q1 ) f ( q2 ) W ( x, y ) σ ( mt ) f(q) is the probability distribution than a parton will have a momentum q Normalization depends on mt Includes acceptance effects b Integrate over unknown q1,q2, y q q t t 32

33 Run I: Matrix Element Method probability to observe a set of kinematic variables x for a given top mass dnσ is the differential cross section Contains matrix element squared W(x,y) is the probability that a parton level set of variables y will be measured as a set of variables x 1 n Psgn ( x; mt ) = d σ ( y ; mt ) dq1 dq2 f ( q1 ) f ( q2 ) W ( x, y ) σ ( mt ) P(x,mtop) ~ ftop Psgn(x,mtop) + (1-ftop) Pbg(x) 33

34 Combine events D0 Run 1 Measurement: Mt= 180.1±3.6±3.9 GeV Nature Vol 429, Page

35 Advantages Possible to use (much) more statistical information per event: Possibility that event is signal or background Integration includes sum over all possible jet assigments Integral over detector resolutions Properly treats all the above ambiguities... and a bit lucky too... D0 Run 1 Measurement: Mt= 180.1±3.6±3.9 GeV Nature Vol 429, Page

36 End of Run I : (22 tt events) mtop = 178.0±4.3 GeV/c2 36

37 'History' of the top mass... 37

38 Run II Template Lepton + jets channel without b-tag e or μ with pt>20 GeV/c 4 jets with pt>20 GeV No b-tag requirement Kinematic fit, require lowest χ2<10 Low bias discriminant (DLB) using topological variables require DLB>0.4 - tt W+jets Plot lowest χ2 solution from fit, compare data to binned MC templates 94 tt- candidates selected, S/B ~ 1/ MM.Mulders, =169.9±5.8 stat sys GeV /c top CERN

39 An idea from DELPHI times: 39

40 Ideogram method Developed at NIKHEF ('97/'98 N.Kjaer, M.Mulders, I.v.Vulpen) for W mass measurement / SM Higgs search in DELPHI, all-jets channel W LEP: similar challenges, ambiguous events: Multiple jet assignments (but not as many) Backgrounds (but not as much) Measure invariant masses with kinematic fit Use statistically powerful event-by-event likelihood 'A lot of work' for 10-15% improvement in statistical error ~ 1 extra year of LEP running More to be gained with complex ttbar events??? 40

41 Ideogram method Same event selection as Template (w/o cut on D) Same Kinematic fit Keep fitted mass mi, estimated error σi and χ2i for all 24 jet/neutrino solutions Construct event likelihood taking into account all jet combinations and the probability that the event is background Event likelihood like Matrix Element method, but ~ times faster in CPU time!!! 41

42 Ideogram Likelihood 24 L m t = i =1 w i [ P evt G m ', i, m i BW m ', m t dm ' 1 P evt BG m i ] 42

43 Ideogram Result 30% improvement ~ factor 2 in statistics! Summer

44 Ideogram developments Include b-tagging in Ideogram likelihood... inclusive fit 0, 1, 2 tag P evt S S S S = 1 P evt B samp B D B ptfrac B ntag 370 pb-1 2 i njets tag? w i =exp j=1 p i, j 2 b-tagging weights improve separation correct vs wrong jet permutations 24 use of b-tags in discriminant improves separation signal vs background inclusion 'wrong jet permutation' shape improves purity fit, slope and pull 0 b-tag 1 b-tag 2 b-tags Nov,2005 L m t, P samp = i =1 wnikhef P colloquium, S m i, m t 1 P evt BG m i[ evt i i ]

45 Ideogram JES fit JES sensitivity from W mass = 80.4 GeV constraint in fit Re-do kinematic fit for different values of overall JES scaling factor One sample of selected events file with results file with results file with results file with results file with results file with results for each value of JES, run fit program on all events, + calculate event likelihoods glue likelihoods from different files together in one 2D plot --> Fit JES factor and top mass 45

46 Ideogram Summary Ideogram approach works beautifully for Top mass From 'complex' W mass analysis to 'simplified' and super-fast Matrix Element method Soon update with b-tagging, 370 pb-1, and JES fit! Need to develop method further to stay competitive with Matrix Element method... e.g. Proper 5-jet treatment, include separate b-jet decay hypotheses Future of Ideograms in D0 in hands of Pieter Houben (NIKHEF) and Michele Weber (FNAL)... Stay tuned! 46

47 Template + b-tagging Lepton + jets channel with b-tag using 'SVT' secondary vertex tagger Systematic Uncertainties Jet Energy Scale Gluon Radiation Signal Model Jet Energy Resolution Calibration Background Model b-tagging Trigger Bias Limited MC Statistics Total one or more b-tagged jets 4 jets with pt>15 GeV No cut on low bias discriminant DLB 60 tt candidates selected, S/B ~ 3/1 Δmtop (GeV/c2) -5.3/ M top =170.6±4.2 stat ±6.0 sys GeV /c 2 47

48 CDF Jet Energy Scale Correct jet energy as well as possible for calorimeter effects (offset / non-linearity) underlying hadronic activity and out-of-cone showering Events samples used for checks/calibration: γ + jet, di-jet (data and MC) Relative calibration uncertainty data/mc is what counts for Top mass Dominant systematic uncertainty on Top mass A lot of work done to better understand the jet energy scale in Run II CDF: Run II systematic JES uncertainty comparable to or better than Run I Run II '2004' New Run II Similar work done (and ongoing) in DØ Jet Energy Scale group Run I central calorimeter 0.2 < η <0.6 48

49 Jet Energy Scale JES l W+ b-jet Is di-jet scale applicable to multi-jet environment? Also look at invariant mass of W jets in ttbar events Advantage: Will get more precise with more ttbar stats Works in multi-jet environment Is part of Top you are trying to measure --> good correlation Remaining uncertainty: b-jet scale! n t t W- jet Mjj(W) bjet jet 49

50 Template fit with 'in situ' JES Use 1D Templates of fitted top mass And 1D Templates of W->jj mass before fit Combined fit to extract mtop, JES, purity(?) Auto-calibration of 'wrong jet permutation' + background shape 50

51 2-tag sample: only one jet assignment for W-->jj! Fit to data (JES) 51

52 Fit to data (mtop) And allow background to float freely: 52

53 Result with mtop & JES fit Split samples according to b-tags combine with the traditional JES based on di-jet, photon+jet etc. A bit lucky... CDF = (stat) (JES+syst) GeV/c2 = (total) GeV/c2 mt = (stat + JES) ± 1.3(syst)GeV / c 2 single best result Submitted to PRL hep-ex/ Submitted to PRD hep-ex/ Submitted to NIM A hep-ex/

54 Run II Matrix Element Still without use of b-tagging Use overall JES scale only from 'in situ' W-mass constraint! Mtop, JES, purity fitted simultaneously 2 c / V e )G (syt.6 1 ) S JE + tat (s.4 4 ±.5.7 calibrated = t m + calibrated 54

55 55

56 Tevatron Combination Only use best analysis in each channel, each experiment Mt=172.7±2.9 GeV/c2 Stat uncertainty: 1.7GeV/c2 Syst uncertainty: 2.4GeV/c2 hep-ex/ Top quark Yukawa coupling to Higgs boson gt=mt 2/vev =0.993±

57 What does it do? From M.Gruenewald's HEP2005 talk 57

58 Top vs Higgs mass From M.Gruenewald's HEP2005 talk 58

59 Combined systematics GeV/c2 Result Stat. 1.7 JES 2.0 Sig. Model 0.9 Bkgd. Model 0.9 Multi-Interaction 0.3 Fit Method 0.3 MC Generator 0.2 Total Syst. 2.4 Total Error 2.9 Basic improvement by 1/ L - L 1fb-1 soon! - Further improvement on JES by direct b-jet JES calibration by Z bb or γ+jet events. Current b-jet JES taken same as generic jet + additional uncertainty according to LEP/SLD measurements. Sig./Bkgd. Modeling (ISR/FSR/Q2 dependence etc.) can be improved by using Tevatron data. Di-lepton and all-hadronic channel will improve, with independent systematics 59

60 Projection for uncertainty on top quark mass Expect uncertainty < 2 GeV per experiment: CDF Aimed for luminosity of Tevatron Run II 60

61 Top beyond Tevatron LHC will have unlimited top statistics! Should allow improvements of all Tevatron top physics measurements But it will require good detector understanding! Interesting object of study in itself Calibration tool Tagger of Interesting events Background to New Physics 61

62 CONCLUSIONS Top Turns Ten Top physics is top Top mass: how many (Ideo)grams? Precision era for the mass measurement ( GeV) Expect uncertainty < 2 GeV / exp! Expect SM single top In Ten years top turned from one-loop correction to particle, to a tagger/tool for other interesting (new?) physics! What is next? 62

63 What is next? Number of Physicists Destined to become a 'background' / calibration tool?? Tevatron now ~ years ago! LHC Year Discovered 63