Measurement of the t t Production Cross Section at CDF Using Lepton + Jets Events With Secondary Vertex b-tagging. Henri Bachacou

Transcription

1 Measurement of the t t Production Cross Section at CDF Using Lepton + Jets Events With Secondary Vertex b-tagging Henri Bachacou U.C. Berkeley / LBNL Henri Bachacou UCSC Seminar, Oct. 19 th 24 1

2 Outline Top Quark Physics at Hadron Collider and Motivation for this Analysis The Tevatron and The Collider Detector at Fermilab Description of the Analysis: Secondary Vertex B-Tagging Background Estimates New Results with Run II Data Outlook Henri Bachacou UCSC Seminar, Oct. 19 th 24 2

3 The Top Quark: A Brief History Long search for the top quark in the 8 s and early 9 s. The top quark was discovered at the Tevatron in 1995, completing the third generation of quarks. u c t d s b e µ τ ν e ν µ ν τ Surprise (hinted by LEP precision measurements): top mass 175 GeV/c 2. Top mass is of the order of the EW scale! This large mass is responsible for the top quark properties. It is also a hint that the top quark may play a special role in the SM and EW symmetry breaking. Henri Bachacou UCSC Seminar, Oct. 19 th 24 3

4 Some Properties of the Top Quark Large mass large radiative corrections on W mass: m t is a very important EW parameter LEP1, SLD Data LEP2, pp Data 68% CL m W [GeV] α M W ( M Z cosθ W ) 2 = ( 175GeV )2 + αln( m2 H ) m t m 2 W m H [GeV] Preliminary m t [GeV] Large mass Very short life-time ( 1 24 s.) decays before hadronization. i.e. Γ t 1.4 GeV Λ QCD 4 MeV Makes it the only quark that can be observed in a free state. Henri Bachacou UCSC Seminar, Oct. 19 th 24 4

5 What do we know about the Top Quark? It exists. It s heavy. (precision measurement: 3% level) That s about it... Other properties are not well-measured. Production rate (this talk), branching ratios, W polarization from top decay are measured with large uncertainties. Charge, life-time (width) are predicted by the SM but have not been measured! Most of these measurements are sensitive to physics beyond the Standard Model At LHC, Top will be the main background to new physics We need precision measurements Henri Bachacou UCSC Seminar, Oct. 19 th 24 5

6 Motivation for this Analysis Why measuring the t t production cross-section? Test of production mechanism of top: Is strong interaction for top the same as for other quarks? New physics may affect the t t production cross-section. Potential 1% effect from Supersymmetry. Compare cross-section measurement for different decay channels: does top decay to some exotic particle? E.g. charged Higgs could enhance the di-lepton cross-section measurement (t H + b and H + τν). Define the top sample and understand its content: is there only top in there? Higgs, Supersymmetry could hide in the top sample. Prerequisite for top mass measurement. In the process, understand the W+jets sample: crucial for Higgs search. Henri Bachacou UCSC Seminar, Oct. 19 th 24 6

7 Top Quark Production at Hadronic Machines, and Decay Channels At the Tevatron, t t pairs are produced mostly through q q annihilation (85%) and 15% through gluon fusion At LHC, gluon fusion dominates. t W b almost exclusively in the SM. W lν l or q q Depending on W decays, three different channels are observable: Di-lepton : Both W lν l. BR = 4%. Clean signal but low statistics. Lepton+Jets : 1 W lν l and 1 W q q. BR = 3%. Larger background, but larger branching ratio. Hadronic : Both W q q. BR = 46%. Difficult, due to large QCD background. (in this talk: lepton = electron or muon) Tau decay channels are also studied, but very difficult. Henri Bachacou UCSC Seminar, Oct. 19 th 24 7

8 p p collider at Fermilab (Illinois). Upgraded for Run II: Main Injector and Antiproton recycler Runs at s = 1.96 TeV. (compared to 1.8 TeV in Run I) Expect 3% increase in σ t t The luminosity has been significantly increasing since the beginning of Run II. Record luminosity: cm 2 s 1 (compared to cm 2 s 1 in Run I) Run II goals: base goal : 4.4 fb 1 by end of 9 design : 8.5 fb 1 by end of 9 The Tevatron Henri Bachacou UCSC Seminar, Oct. 19 th 24 8

9 The Collider Detector at Fermilab (CDF) General purpose detector with 4π coverage Tracking system in 1.4 T magnetic field: Silicon Vertex Detector and Tracking Drift Chamber (COT) Lead/Scintillator and Iron/Scintillator sampling calorimetry Muon system (drift chambers and scintillators) Henri Bachacou UCSC Seminar, Oct. 19 th 24 9

10 The Silicon Vertex Detector Main Run II upgrade Necessary for precise measurement of tracks impact parameter and secondary vertex reconstruction. Silicon micro-strips. Made of 3 sub-systems: L : Innermost layer, glued on the beam pipe (radius 1.3 cm). SVX II : 5 double-sided layers (9 and 1.2 o stereo angle strips provide z measurement). ISL : Intermediate Silicon Layer, 2 double-sided (small angle) layers. Extends coverage to η = 2. Henri Bachacou UCSC Seminar, Oct. 19 th 24 1

11 Silicon Detector: A long commissionning... Several problems had to be solved: stuck cooling lines in ISL, bit errors in data, analog pickup noise in L However, some modules are permanently damaged. Two main sources of casualties: Jumper failure : wire connecting φ and z side hybrids breaks due to Lorentz force creating by 1.4T field. Effect enhanced by some special DAQ conditions (readout rate at wire mechanical resonance; high current) AVDD2 failure : sudden loss of communication with chip (analog) front end. Occured mostly during beam incidents Improved protection against beam dump into the detector. Thermal cycle also seem to play a role. Silver epoxy connection to the chip suspected. The module chip chain is severed at affected chip. Now running smoothly: 92% of modules operational (85% with <1% error rate) Henri Bachacou UCSC Seminar, Oct. 19 th 24 11

12 Silicon Detector Online Monitoring ( SVXMon ) SVXMon runs on a fraction of events and identifies readout problems in real time. Provides plots, error logs, alarms to shift crew and silicon experts. Ex: send a signal to DAQ to re-initialize the detector when necessary. Wedge # ISL Cell Id L & ISL Chip Status Map Strip Mean Charge for Detector W 2 4 -A PHI (ADC Counts) ISL Strip Number L Bulkhead, Phi Side Bulkhead, Z Side Accumulated from 16:4:4 till 18:12:5, 176 events Run No data! Failure! OK Checks Disabled Updated 19:16:18 Run Accumulated from 17:51:5 till 18:12:5, 176 events Henri Bachacou UCSC Seminar, Oct. 19 th 24 12

13 Outside-In tracking algorithm: seeded by a drift-chamber track. Silicon Detector: Performance Look for axial hits along the path of the COT track, going inward. Add stereo hits in the second iteration. Tracking efficiency 94% ( 3 radial hits) Uncertainty on impact parameter 25µm Impact Parameter d Interaction Point L (not used for this analysis) significantly improves d resolution, especially at low momentum. Stand-alone tracking (silicon only) still under development. Thanks to the ISL, it should extend silicon tracking up to η = 2 SVX/ISL only W/ L (error includes beam size 3µm) Henri Bachacou UCSC Seminar, Oct. 19 th 24 13

14 Analysis Overview Method: Simple counting analysis: σ t t = N t t L = N obs N bkd A t t ɛtag t t L N obs : Number of observed events N bkd : Number of expected background events A t t ɛ tag t t : Acceptance x b-tag efficiency = fraction of produced t t events that are actually detected L : Integrated Luminosity I will describe (in this order): Event Selection. b-tagging algorithm, estimate of b-tagging efficiency and acceptance. Estimate of backgrounds. Henri Bachacou UCSC Seminar, Oct. 19 th 24 14

15 Signature of the t t Lepton+jets channel Top quarks decay to W+b two b-jets W q q two light jets W lν l (l=e or µ) one lepton and large missing energy Several processes (may) have the same signature: W+jets, W lν l can be greatly reduced by identifying b-jets W b b,w c c,w c + jets: irreducible QCD (multi-jets): W is faked (either by fake lepton, or semileptonic B decay). lepton less isolated, less missing energy than in t t t t events tend to be more energetic and more central than backgrounds. Henri Bachacou UCSC Seminar, Oct. 19 th 24 15

16 Event Selection Select W lν l : Exactly 1 High-p T lepton: p T > 2 GeV Lepton must be isolated: I.4 < 1% I.4 = Isolation = Energy found in a R.4 radius about the lepton, divided by lepton energy. Missing E T > 2 GeV require 3 jets of E T > 15 GeV and η < 2 Tag b-jets with Secondary Vertex Tagger: require 1 tagged jet Reject background further: Total Transverse Energy > 2 GeV (next slide) Henri Bachacou UCSC Seminar, Oct. 19 th 24 16

17 Event Selection Optimization: Rejecting the Background Try to reach the highest sensitivity on the measurement. Total (transverse) Energy in the event: H T = Scalar Sum of Jets E T, Lepton p T, Missing E T is very discriminative. Requiring H T > 2 GeV rejects > 1/3 of background, keeping 96% of t t signal. Events / 2 GeV Top Mistags QCD Wc Wcc Wbb S/B S / S / S+B 2 S+B+σ(B) H T (GeV) H T Cut (GeV) H T Cut (GeV) Selection S/B (expected) No b-tag 1/5 With b-tagging 2 b-tag and H T 3 Henri Bachacou UCSC Seminar, Oct. 19 th 24 17

18 Secondary Vertex B-Tagging Algorithm Take advantage of the long life-time of B hadrons: cτ 45 µm 1) Select good quality tracks with large impact parameter. 2) Try to reconstruct a vertex. 3) Tag vertices with large (transverse) decay length significance: L xy σ L xy > 3 L xy is signed w.r.t jet axis: Positive tag: sec. vertex in the direction of the jet Negative tag: sec. vertex behind interaction point Use negative tag rate to estimate the mistag rate Y (cm) Secondary VTX Lxy Reference VTX from beamline Jet GeV Jet5 2 GeV Jet GeV L xy Jet4 25 GeV L xy = 3 mm = 2.3 mm I.P. Jet Dir CDF II Preliminary Secondary Vertex HT = 358 GeV MET 66.7 GeV Electron 72.6 GeV Jet2 5.4 GeV X (cm) Henri Bachacou UCSC Seminar, Oct. 19 th 24 18

19 B-Tagging Efficiency Measurement (I) Measure efficiency independently in MC and data to calibrate the MC. Control sample: back-to-back jet events with electron within a jet. Sample has large Heavy Flavor content: electron from semi-leptonic decay of b and c hadrons. Measure tagging rate of Electron-Jet in events with Tagged Away-Jets. Difficulty: Measuring precisely the Heavy Flavor content of the sample. Novel technique: use conversions to estimate the fraction of light flavor jets in the double-tag sub-sample. Henri Bachacou UCSC Seminar, Oct. 19 th 24 19

20 B-Tagging Efficiency Measurement (II) Entire Sample Tagged Away Jet Composition of electron jets: H.F. Conversions Fake Electrons Assume light jet composition is the same whether away jet is tagged or not Normalize light jet fraction with conversions: Flight a+ = N conv a+ N F conv light Still need to measure the heavy flavor content of the entire sample, with two methods: Identify D K ± π Identify muons from double-semi-leptonic decays in the electron jets. Events /.5 cm L SVX L SVX L1 SVX L2 SVX L3 SVX L4 SVX Readout ISL L Forward ISL L Central ISL L1 ISL Outer screen COT inner cyl. Radius (cm) χ 2 / ndf / 56 Prob.1274 ND ± 68.3 c 1.216e+4 ± c1-1.75e+4 ± 57.5 c ± Reconstructed K-Pi mass (GeV/c^2) Henri Bachacou UCSC Seminar, Oct. 19 th 24 2

21 B-Tagging Efficiency Measurement (III) Efficiency measured in MC and data: Note: this is the efficiency for semileptonic decays. MC reproduces the data well, but slightly overestimates the b-tagging efficiency: Scale Factor : ɛdata B ɛ MC B = 82 ± 6 % The t t MC is tuned with the Scale Factor... Efficiency for tagging at least one jet in a t t event (l+ 3 jets): ɛ t t 1 tag = 53.4±.3±3.2% B-Tagging Efficiency Data MC Electron Jet E T (GeV) Scale Factor 2 χ / ndf / 7 Prob Electron Jet E T (GeV) Henri Bachacou UCSC Seminar, Oct. 19 th 24 21

22 Acceptance Pre-tag acceptance A t t evaluated with Pythia MC. Include trigger efficiency, z cuts, and other scale factors: A t t = A MC t t ɛ data z ɛ data z ɛdata trigger Event tagging efficiency: ɛ t t 1 tag = 53.4±.3±3.2% Overall Acceptance: A t t ɛb tag t t ɛ data leptonid ɛ MC leptonid = (3.84 ±.4)% Quantity Relative error (%) Energy Scale 4.9 PDF 2. ISR/FSR 2.6 MC modeling 1.4 Lepton ID 5. B-tagging 6. (electrons and muons combined; includes branching ratios) Henri Bachacou UCSC Seminar, Oct. 19 th 24 22

23 Backgrounds Key issue of this analysis: Understanding the b-tagged W+jets sample composition We use both data and MC to evaluate the backgrounds. W b b,w c c,w c: Monte Carlo provides Heavy Flavor fraction of W+jets, normalization from data W +light jets: a light jet is wrongly tagged ( mistag ) Mistag rates measured in multi-jet control sample QCD (multi-jets): W faked either by fake lepton, or semileptonic B decay Use non-isolated lepton control sample Single Top, W W, W Z, ZZ: from MC Henri Bachacou UCSC Seminar, Oct. 19 th 24 23

24 W+Heavy Flavor Background, ALPGEN Tuning Wb b, Wc c, Wc + jets processes have same signature than t t. Cross-sections only known at Leading-Order not accurate enough. Normalize the overall W+jets cross-section from the pre-tag W+jets data sample. Use Monte Carlo to estimate HF fractions = W+HF / W+jets. ALPGEN: Matrix Element leading order MC. HERWIG for hadronization. Two issues... Gluon radiation from HERWIG creates phase space overlaps between W+n partons and W+(n+1) partons ALPGEN samples reject hard gluon radiation, by matching Matrix Element partons with jets. ALPGEN validation and tuning: It would be nice to measure the HF content of W+1 jet and W+2 jets samples from data, but not enough statistics look at multi-jet sample instead. Henri Bachacou UCSC Seminar, Oct. 19 th 24 24

25 ALPGEN Tuning: Heavy Flavor in Multi-Jet Sample Compare ALPGEN to data in control sample of multi-jet events. In data, fit pseudo cτ = L 2D M vtx /p vtx T to extract contributions to tags: b, c, symmetric mistags (resolution), material+long life-time Template shapes from MC. Conclusions: Material and life-time mistags create an asymmetry: mistag rate = (1.2 ±.1) neg. tag rate ALPGEN Heavy Flavor fraction is much lower than data: Data = (1.5 ±.4) ALPGEN/HERWIG 1 5 positive tag excess Could it be that Leading Order calculation underestimates Heavy Fla- 1 vor fraction? (cm) material interactions + c quark jets + b quark jets pseudo-cτ Henri Bachacou UCSC Seminar, Oct. 19 th 24 25

26 Mistags Mistag = positive tag of a light jet. W+light jets are an important background. Parametrize the negative tagging rate in multi-jet samples as a function of Jet E T, track multiplicity, η, φ, and ΣE T. Correct for the mistag asymmetry: mistag rate = (1.2±.1) negative tag rate Mistag rate at 1% level. Apply the parametrized rate to jets in the pre-tag W+jets sample. Tag Rate Tag Rate Jet Et (GeV) Number of Good Tracks Henri Bachacou UCSC Seminar, Oct. 19 th 24 26

27 QCD Background: Estimate before b-tagging ( pre-tag ) Background of multi-jet events faking a W: Fake lepton, or real b-jet and lepton from semi-leptonic decay of b or c hadrons. Missing E T can be faked by cracks, or due to neutrino from semi-leptonic decay. Method: Assume Missing E T and Lepton Isolation are uncorrelated. Isolation = Energy found in a R.4 radius about the lepton, divided by lepton energy. Divide sample in 4 regions of the Missing E T - Iso. space. Estimate # QCD events in signal region (D): N pre tag QCD = B.C A e-channel: 2% of W+ 3 jets sample is QCD bgd. µ-channel: only 7.5%. MET 2 15 D B.1.2 C A Isolation Henri Bachacou UCSC Seminar, Oct. 19 th 24 27

28 QCD Background: Tagged Sample Estimate tagged QCD background by assuming the tagging rate is the same in the low and high Missing ET regions, and calculate tag rate from region B: N tag QCD = B.C.T agrate(b) A Alternatively: use tagged events only in the four regions. Low statistics, but promising method: N tag QCD = Btag.C tag A tag Combine two methods. MET 2 15 D B.1.2 C A Isolation Henri Bachacou UCSC Seminar, Oct. 19 th 24 28

29 Data Sample Analysis based on 162 pb 1 accumulated between Feb. 22 and Sep. 23. W + 1 jet W + 2 jets W + 3 jets W + 4 jets W + 3 jets W + 4 jets H T > H T > 2 GeV Electrons Muons Events/ 4 GeV W+jets events W Transverse Mass in W+ 1-jet: Transverse Mass (GeV/c 2 ) Henri Bachacou UCSC Seminar, Oct. 19 th 24 29

30 Background Summary H T > 2 GeV Jet multiplicity W + 1 jet W + 2 jets W + 3 jets W + 4 jets Pretag Mistags 4.9 ± ± ± ±.3 Wb b 37. ± ± ± ±.4 Wc c 13.7 ± ± ±.3.5 ±.2 Wc 34.5 ± ± 2..7 ±.2.3 ±.1 WW/WZ/ZZ,Z ττ 2.2 ± ±.4.3 ±.1.1 ±.3 QCD 24.3 ± ± ± ±.3 single top 2.6 ± ±.5.8 ±.1.2 ±.2 Z+HF 1.1 ±.3.6 ±.2.1 ±.5 Total ± ± ± ±.8 Corrected Total ± ± ± 1.8 Data Check background estimate in 1 and 2 jet events. Signal Region: 3 jets and H T 2 GeV. Expect 13.5±1.8 background events, observe 48 events. Number of events single top diboson Number of jets in W+jets Wbb Wcc Wc non-w mistags tot bkgd ± 1σ -1 data (162 pb ) Henri Bachacou UCSC Seminar, Oct. 19 th 24 3

31 Event Display of a W+5 Jets Event Candidate 1 Jet GeV CDF II Preliminary Secondary Vertex HT = 358 GeV.8 MET 66.7 GeV Jet5 2 GeV L xy = 2.3 mm Y (cm).6.4 Jet GeV L xy = 3 mm I.P. Electron 72.6 GeV.2 Jet4 25 GeV Jet2 5.4 GeV X (cm) Henri Bachacou UCSC Seminar, Oct. 19 th 24 31

32 Measurement of the t t Cross-Section: Based on the 48 candidate events with 3 or more jets and H T > 2 GeV, we measure a crosssection of: Number of events σ t t = (stat.)+.9.6 (syst.) pb Background Background errors Background + tt (5.6pb) Bkgnd + tt errors -1 Data (162 pb ) Number of jets in W+jets Syst. Err on σ t t Acceptance 1% (incl. B-tagging 6%) Luminosity 6% Bgd 5% m t (GeV/c 2 ) σ (pb) ± ± ± 1.1 Henri Bachacou UCSC Seminar, Oct. 19 th 24 32

33 Does it look like top? Data (162 pb ) Top (6.7 pb) Mistags QCD Wc Wcc Wbb H T (GeV) 22 2 Data 18 Top (6.7pb) 16 Mistags 14 QCD 12 Wc 1 Wcc 8 6 Wbb Tagged Jet E T (GeV) Data Top (6.7pb) Mistags QCD Wc Wcc Wbb Pseudo-cτ (cm) Henri Bachacou UCSC Seminar, Oct. 19 th 24 33

34 Measurement of σ t t with double-tag events Look at the sub-sample of events with 2 tagged jets. Very pure signal: S/B=9 Interesting check of the B content of W+jets sample. Backgrounds are evaluated with similar technique: Wb b dominates. Mistags are dominated by events with 1 real tag & 1 mistag apply mistag rate to single-tag sample, and normalize to expected Wq q content (single-tag sample is rich in t t; also avoid double-counting with Wb b, Wc c ). QCD: no double-tag events in low Missing E T region set an upper limit. Jet multiplicity 2 jets 3 jets 4 jets Single top.4±.8.15±.3.4±.1 WZ.15±.4.2±.1.1±.1 Wb b 2.76±.86.64±.18.21±.6 Wc c.2±.8.5±.2.3±.1 Mistag/QCD.14 ±.4.16 ±.4.11 ±.3 Total 3.65 ± ±.23.4 ±.9 Corrected Total 3.6± ±.3 Data Henri Bachacou UCSC Seminar, Oct. 19 th 24 34

35 Measurement of σ t t with double-tag events Based on the same 162 pb 1. 8 candidates dominated by statistics. Number of Double tagged events mistags Wbb Wcc WZ Single top ttbar (5.pb) Tot bkgd ± 1σ Data (162 pb -1 ) σ t t = (stat.) (syst.) pb 2-jet events: < 2σ excess Not seen in the single-tag sample Number of jets in W+jets Henri Bachacou UCSC Seminar, Oct. 19 th 24 35

36 Outlook At the Tevatron: The Tevatron will still have the monopoly on top physics for a few more years. Much more data is expected. For this analysis: many systematics will scale down with statistics (lepton id, b- tagging efficiency). Improved b-tagging: use L and forward tracking to extend b-tagging up to η = 2 Crucial for double-tagging efficiency We should keep our eyes open for new physics! At LHC: The LHC will be a top factory (8 1 6 /year at low luminosity!) statistics will not be an issue. Precision measurements of mass, width, branching ratios... Top signal will also be the main source of background for many analyses crucial to understand it! Henri Bachacou UCSC Seminar, Oct. 19 th 24 36

37 Conclusion σ t t has been measured with significantly larger statistics than in Run I, at a new center-of-mass energy ( s = 1.96 TeV). So far, results are consistent with a Standard Model t t signal with m t 175 GeV (σt t SM = pb): Single Tag measurement (m t = 175 GeV): σ t t = (stat.)+1..7 (syst.) pb Measurement with double-tag events (m t = 175 GeV): σ t t = (stat.) (syst.) pb Tevatron Run II is now at full speed. LHC is coming up. Some very exciting years ahead of us! Henri Bachacou UCSC Seminar, Oct. 19 th 24 37

38 Backup Slides Henri Bachacou UCSC Seminar, Oct. 19 th 24 38

39 Low E T electron control sample: MC vs Data comparison(i) Electron E T (GeV) MC Data Electron Jet E T (GeV) Electron p T (GeV/c) Away Jet E T (GeV) Henri Bachacou UCSC Seminar, Oct. 19 th 24 39

40 Low E T electron control sample: MC vs Data comparison (II) MC Data Number of tracks in jet Number of tracks in fit Pseudo-c τ (cm) Secondary Vertex Mass (GeV/c ) Henri Bachacou UCSC Seminar, Oct. 19 th 24 4

41 B-Tag Efficiency Measurement: c/b fraction and systematics Fraction of Tags Light Charm Bottom Data Tag Mass (GeV/c 2 ) Source uncertainty (%) F HF 3.5 FHF a method 3. mistag subtraction 3. E T dependence 2.5 B-decay 1.2 total systematic error 6.2 data statistics 3.2 MC statistics 3.6 Total 7.8 Henri Bachacou UCSC Seminar, Oct. 19 th 24 41

42 Observed cross section (pb) ALPGEN W+j prediction -1 Bkgd-corrected data (162 pb ) Number of jets in W+jets Henri Bachacou UCSC Seminar, Oct. 19 th 24 42

43 Number of events tt Zcc Zbb mistags Tot pred ± 1σ Data (162 pb Number of jets in Z+jets -1 ) Henri Bachacou UCSC Seminar, Oct. 19 th 24 43

44 ALPGEN Tuning: gluon splitting Could the discrepency be due to an anomaly in the contribution from different diagrams (gluon splitting vs flavor creation... )? Gluon splitting produces pairs of HF jets close to each other. Check φ distribution between two closest jets in 3-jet sample. ALPGEN gives reasonnable agreement with data. The 1.5 factor does not seem to depend on φ relative contribution between diagrams does not seem to be the issue. We tune ALPGEN by multiplying Wb b and Wc c fractions from ALPGEN by (1.5 ±.4). Events /.17 radians Events /.17 radians Excess tag rate /.17 radians 25 tagged events mistag prediction double-tagged MC prediction Min. φ between jets excess tag rate ALPGEN MC Min. φ between jets Henri Bachacou UCSC Seminar, Oct. 19 th 24 44