-
Measurement of the Top Quark PairProduction Cross Section using
a Topological Likelihood and using Lifetime b-Tagging
page page 11Tobias Golling
Tobias Golling - SCIPP - August 24th 2004
with the DØ Detector in Run II at the Fermilab Tevatron at a center of
mass energy of √s = 1.96 TeV
-
Overview
Physics Motivation Tevatron & DØ Top Quark Production and Decay Overview over the Analyses Methods
1) Common W+Jets Preselection
2a) Topological Analysis 2b) Lifetime b-Tagging Analysis
Summary & Outlook
page page 22Tobias Golling
tt
tt--
WW++
WW - -
bb
bb--
qq
q'q'
µ
µ
pppp --
-
The Top Quark in the Standard Model
page page 33Tobias Golling
Why is the Top Quark so interesting ? completes the quark sector large mass m
top ~ 180 GeV / c2
short lifetime ~ 5 · 10-25 s sensitive to physics beyond the Standard Model
Discovery of the Top Quark in 1995by the CDF and DØ Collaborations.
Top-antitop production cross section test of perturbative QCD sensitive to
top decays to New Physics particles (e.g. charged Higgs) Physics beyond the SM decaying to top-antitop
Higgs-Boson coupling to fermions: f ~ m
f
t ~ 1
-
The Top Quark in the Standard Model
page page 44Tobias Golling
1
Corrections to W and Z boson mass from top quark and Higgs boson loops allow prediction of the top quark and the Higgs boson mass
Comparison of precision EW measurement ↔ corrections⇒ prediction of top quark mass
zero order
discoveryobservation
-
The Top Quark in the Standard Model
page page 55Tobias Golling
corrections to W and Z boson massfrom top quark and Higgs boson loopsallow prediction of the top quark and the Higgs boson mass
corrections: ~ Mt
2
corrections: ~ ln (MH)
Goal for Tevatron in Run II:M
t = 3 GeV
MW
= 20 MeV
-
The Tevatron at Fermilab
page page 66Tobias Golling
Main Injector & Recycler
Tevatron
Booster
p p
p source
Chicago
_
_
CDF DØ
p p
√s =1.96 TeV ∆t = 396 ns
Run I 1992-95 Lint
~ 125 pb-1
Run II 2001-09(?) 40 × larger dataset at increased energy
Only place to study the top Only place to study the top
quark before the LHC eraquark before the LHC era
-
DØ Detector at Fermilab
page page 77Tobias Golling
new silicon and fibre tracker new ~2 T solenoid
upgraded muon system upgraded trigger/DAQ
-
The Silicon Microstrip Tracker
page page 88Tobias Golling
axial, double-sided small-angle stereo and double-sided 90° detectors 800k channels, SVX2 readout
Silicon Microstrip Tracker (SMT): 6 barrels, 16 disks tracking out to ~ = 3
-
Data Set
page page 99Tobias Golling
analyseddata
installation finishedand detector
commissioning Analysed data setin Run I (1992-1995):
~ 125 pb -1
Analysed data setin Run II (08/2002-10/2003)for winter and summer '04
conferences:~ 140 - 165 pb -1
Data set in Run II until 2009:
~ 4-9 fb -1
N = σ ∙ ÿ∫ L dt
Run I record
-
Top Quark ProductionTop quarks mainly produced in pairs at Tevatron and LHC
page page 1010Tobias Golling
Tevatron LHCtypical S/B: ~1/1 10-100
qq gg
ttbarNLO (pb) qq->tt gg->tt
Run I (1.8 TeV) 4.87±10% 90% 10%
Run II (2.0 TeV) 6.70±10% 85% 15%
LHC (14 TeV) 803±15% 10% 90%
σtt
Run I:
ttbar=5.7±1.6 pb
(statistics limited)
-
Top Quark Decay
lepton
neutrino
b -jet
b -jet
jetjet
page page 1111Tobias Golling
τ+Xµ+jetse+jetse+ee+µµ+µhadronic
1.4 % 1.4 %2.8 %
14.8 %
14.8 %
22 %44 %
Top quarks decay predominantlyTop quarks decay predominantly
(~100%) to a W-Boson and a b-quark(~100%) to a W-Boson and a b-quark
Top-Antitop SignaturesTop-Antitop Signatures
determined by the W decay modes:determined by the W decay modes:
`dilepton channel'`dilepton channel'
~5% :~5% : 2 jets, 2 charged leptons, 2 neutrinos2 jets, 2 charged leptons, 2 neutrinos
`lepton+jets channel'`lepton+jets channel'
~30%:~30%: 4 jets, 1 charged lepton, 1 neutrino4 jets, 1 charged lepton, 1 neutrino
- Large statistics compared to dilepton channel- Large statistics compared to dilepton channel
- Clear signature compared to all-jets channel- Clear signature compared to all-jets channel
`all-jets channel'`all-jets channel'
~40%:~40%: 6 jets 6 jets
always 2 jets are b-jetsalways 2 jets are b-jets
I will concentrate on theI will concentrate on the
'muon+jets channel''muon+jets channel'
-
page page 1212Tobias Golling
Analyses Overview – 2 Methods
QCD
W+jetstt-
QCD
W+jetstt-
QCD
W+jets
tt-
Topological Likelihood Method: Isolate and estimate signal content, fit ttbar fraction by shape comparison
Apply Lifetime b-Tagging(Counting Experiment): Further enhance signal content, subtract estimated backgrounds
Common W+jets selection(W decays leptonically to muon or electron) Enhance signal content
1. 2.
-
page page 1313Tobias Golling
Hard scatter primary vertex
Isolated high p
T lepton
Neutrino reconstructedas missing transverse energy
≥ 3 jets with p
T > 15 GeV
and || < 2.5
Signature - W+Jets Selection
tt
WW++bb
tt--
WW - - bb--
qq
q'q'
ℓ
ℓ
pppp --
-
page page 1414Tobias Golling
Jet Reconstruction and IdentificationThe jet reconstruction and identification efficiency in data is notreproduced by the MC: DØ identifies jets in the calorimeter only (cone-algorithm) Determine efficiency in data and in MC and determine MC correction factor Use pT-balanced physics processes
q
q-
, Z, J/,
tracks calorimeter cluster
}}jet and (Z, J/, ): back-to-back in balanced in p
T
not balanced in due to z-boost
Efficiency determination: Select the away-object (,Z, J/, ) Select jet reconstructed only from tracks back-to-back to (Z, J/, ) check how often an associated calorimeter jet is found
gjet
-
Muons are reconstructed in the muon system and in the tracking systemMuons coming from a leptonic decay of a W boson tend to be isolated from jets have a relatively high transverse momentum
page page 1515Tobias Golling
Muon Isolation
R=22Loose muon isolation:
R , je t 0.5
Tight muon isolation:
E Tcalorimeter cells R0.4−E T
calorimeter cells R0.1
P T 0.08
E Ttracks R0.5
P T 0.06
Muon = MIP(Minimum Ionizing Particle)
-
1. Topological AnalysisStatus Moriond 04
page page 1616Tobias Golling
-
Instrumental Background (fake lepton, fake MET):QCD multijet production
Electron Fakes:Electrons faked by (electromagnetic) jets
Muon Fakes:Muon-fakes are real muons which are fakely isolated(muons from semileptonic b-decays, where the b-jet is not reconstructed)
page page 1717Tobias Golling
Backgrounds
Physics Background:Electroweak W production W ℓ+
ℓ
additional ≥ 4 jets from ISR
W+1jet W+2jets ...
MET Fakes:Misreconstructed calorimeter energy
-
Solve linear system of equations for NQCD
and NW+ttbar
tight lepton:
Nloose
= NQCD
+ NW+ttbar
Ntight
= QCD
* NQCD
+ W+ttbar
* NW+ttbar
QCD
= 8% W+ttbar
= 82%
QCD
W+jetstt-
QCD
W+jetstt-
Determine instrumental QCD multijet background purely from data
page page 1818Tobias Golling
loose lepton:
Determination of QCD Background
-
Topological Analysis OverviewQCDW+jets
QCD
QCD W+jetsW+jets
loose Wselection+ ≥ 4 jets
tight Wselection+ ≥ 4 jets
DetermineQCD
Background
Combine topological event information in a likelihood discriminant, and perform a fit to the data
tt-
tt-
tt-
tt-
page page 1919Tobias Golling
-
Define variables which describe the event topologyCriteria: - Good separation power - Low sensitivity to the jet energy scale (dominant systematic uncertainty)
WW++qq
q'q'
qq q'q'µ µ --qqqqtt
tt--
WW++
WW - -
bb
bb--
qq
q'q'
µ
µ
pppp --
Jets: high-energetic isotropic
Jets dominated by QCD-Bremsstrahlung: low-energetic in the forward region
page page 2020Tobias Golling
ElectroweakW production
ttbar signal
Event Topology
-
page page 2121Tobias Golling
Topological Variables
Sphericity
AplanarityKTmin
'
HT2'
Variables describing theangular distributions ofthe physics objectsin the final state:
Sphericity Aplanarity
Ratios of energy dependent variables:
HT2' (describes centrality)
KTmin
' (=Rmin
(jet,jet)·ETmin/E
TW)
-
Si = ttbar signal
Bi = W+jets background
i runs over the 4 topological variables
L D=∏ i S i
∏ i S i∏ i B i
page page 2222Tobias Golling
Likelihood Discriminant (LD)
Describe data by a linear combination of ttbar, W+4jet and multijet (QCD)
Constrain QCD by using the loose and tight system of equations
Fit the relative fractions
-
page page 2323Tobias Golling
Fit linear combination of QCD (inverted tight selection in data), W+4jet and ttbar to data
Likelihood Fits for µ+Jets & e+Jets
µ+jetsL=144pb-1
e+jetsL=141pb-1
Likelihood Discriminant Likelihood Discriminant
-
page page 2424Tobias Golling
Kinematic Distributionse+jets
µ+jets
-
page page 2525Tobias Golling
Dominant uncertainties: Jet energy scale Jet reconstruction efficiency
ResultComparison with Run I & Summer 03
p p t t Xl jets =7.2−2.4
2.6 stat −1.71.6 syst ±0.5 lum i pb
p p t t X jets =6.0−3.0
3.4 stat −1.61.6 syst ±0.4 lum i pb
p p t t Xe jets =8.8−3.7
4.1 stat −2.11.6 syst ±0.6 lum i pb
=N t t
BR⋅L⋅selBR=0.44
sel=10%L=143 pb−1
Run I summer 03
-
sum m er 03
sum m er 03
page page 2626Tobias Golling
Summer 03 NOW:Large improvement instatistical and systematicuncertainties
Outlook: Publication this fall L = 230 pb-1 Improved systematics
Summary of Topological Analysis& Outlook
-
2. Lifetime b-Tagging AnalysisStatus ICHEP 04
page page 2727Tobias Golling
-
page page 2828Tobias Golling
Count the number of tracks with Count the number of tracks with large positive DCA significancelarge positive DCA significance Jet is tagged if Jet is tagged if NN
trackstracks((>2) > 3 or>2) > 3 or
NNtrackstracks
((>3) > 2>3) > 2
Explicitly reconstruct vertices whichExplicitly reconstruct vertices which are significantly displaced from theare significantly displaced from the primary vertexprimary vertex VV 00 Filter: Filter: Remove track pairs in the mass Remove track pairs in the mass windows corresponding to Kwindows corresponding to K
SS, ,
and photon conversions (and photon conversions ( -> e -> e++ee--))
Counting Signed Impact Parameter (CSIP)Counting Signed Impact Parameter (CSIP) Secondary Vertex Tagger (SVT)Secondary Vertex Tagger (SVT)
B-Tagging at DØ
I will concentrate on the SVT resultsI will concentrate on the SVT results
DCA
Secondary VertexPrimaryVertex
(DCA = Distance to Closest Approach)
-
page page 2929Tobias Golling
Estimate QCD background purely from data:using the loose and tight system of equations(Can be done before or after tagging)
Backgrounds
Subtract small backgrounds using known cross sections: Diboson production: WW l+jets, WZ l+jets, WZ jjll, ZZ lljj Single top production in s- and t-channel Z l+jets
Main background: W+jets Use W+jets sample generated with ALPGEN and interfaced to PYTHIA Rely on ratios of the cross sections of the generated sub-processes:
W+light W+c W+cc W+b W+bb
Apply a matching procedure of partons from the matrix element generator and reconstructed jets
eliminate double counting of the exclusive processes reduce sensitivity to parton generation cuts
tightloose
-
page page 3030Tobias Golling
Philosophy of this B-Tagging Analysis
N tag=N untagged×P eventtag
P eventtag n≥1tag =1−∏ jets 1− jet flavor E T ,
Present versions of DØ MC do not reproduce the b-tagging efficiency and mistag rate observed in data
Tagging algorithm not applied in MC
Instead: Measure jet tagging efficiencies in data (for jet flavor = b,c,light) Parameterize as a function of jet E
T and :
jet(flavor)(E
T,)
Apply tagging parameterisations jet(flavor)
(ET,) to the Monte Carlo
Derive event tagging probabilities: Ptagevent
Determine number of expected tagged events in the MC: Ntag
In particular for the W+njets background:
P W njetstag =∑flavor F flavor P W njets flavor
tagF flavor P W b b j j
tag =42%
P W j j j jtag =1%
-
page page 3131Tobias Golling
Estimate ttbar production cross section from observed excess in the actual number of tagged events with 3 and ≥4 jets with respect to the background expectation Optimum use of the statistical information:
single tags and double tags
Visualize B-Tag Philosophy
P t ttag=45% P t t
double tag=14%
-
page page 3232Tobias Golling
Challenge: Select pure sample of b-jets in data
Enrich dijet data sample in bbbar QCD heavy flavor production:Select jets which contain muons (muon-in-jet)
2 Methods: Use Shape of P
Trel to discriminate
between heavy and light flavor P
Trel = transverse momentum of the
muon relative to the (µ+jet) axis
Apply Lifetime Tagger and Soft Lepton Tagger to two samples (purely from data) muon-in-jet muon-in-jet with away jet lifetime tagged
(further enriched in heavy flavor) Clearly solvable system of 8 equations with 8 unknowns
B-Tagging Efficiency
jet PTrel
µ
µ+jet
B-Tagging Efficiency in Data
-
page page 3333Tobias Golling
Mistag Efficiency +light flavor
Mistag Efficiency = Tagging efficiency for light flavor jets “Mis-Tagging” not due to lifetime but due to decay length resolution
ligh t flavor+ =SF heavy flavor×SF long lived×
-
Idea: Measure negative tagging efficiency -
in a multijet data sample Correct for heavy flavor jets SF
heavy flavor
(which have a higher negative tagging rate than light flavor jets) Correct for the remaining long lived particles which are not present in negative tagged jets SF
long lived
Primary Vertex
JetTracks
SecondaryVertex
SecondaryVertex
positive tag negative tag
-
µ-
IP
SV
Jet 1
IP
SV
Jet 2MTC
mip signalmip signalin calorimeterin calorimeterr
page page 3434Tobias Golling
IP
SV
Jet 1
Muon+Jets Candidate Event
Jet 3
Jet 4
Jet 5
-
page page 3535Tobias Golling
Result for SVT Tagger
=8.2−1.21.3 stat −1.6
1.9 syst ±0.5 lum i pb
Dominant systematic uncertainties: Jet energy scale Jet reconstruction efficiency b-tagging efficiency in data
Single Tags: Double Tags:
Luminosity =164 pb-1
Control bins
-
page page 3636Tobias Golling
Summary & Outlook
Topological Analysis: World best systematics for a topological analysis Forthcoming publication with 230 pb-1
B-Tagging Analysis: Just got ready for ICHEP Forthcoming publication with improved systematics
Understanding of data set:Understanding of data set:jets, jets, leptonsleptons, met, b-tagging, triggering,..., met, b-tagging, triggering,...Development of powerful high massDevelopment of powerful high massanalysis techniques analysis techniques Prepared for Prepared for LHC LHC
Direct input toDirect input tovarious analyses...various analyses...
-
page page 3737Tobias Golling
R=B(tWb)/B(tWq) Measurement
In the SM, unitary of the CKM matrix:In the SM, unitary of the CKM matrix: R=∣V tb∣
2
∣V tb∣2∣V ts∣
2∣V td∣2=∣V tb∣
2
R=0.7−0.240.27 stat −0.10
0.11 syst
R=0.65−0.300.34 stat −0.12
0.17 syst
SVT:SVT:
CSIP:CSIP:
CSIPCSIP
CSIPCSIP
SVTSVT
SVTSVT
tttt versus R (R fixed in the fit) versus R (R fixed in the fit)
68% and 90% CL contours68% and 90% CL contours
Good agreement with SM: R=1Good agreement with SM: R=1
R deduced from the numberR deduced from the numberof single tagged and the of single tagged and the number of double tagged eventsnumber of double tagged events
-
page page 3838Tobias Golling
Top Mass MeasurementMeasurements in l+jets channel (~150 pbMeasurements in l+jets channel (~150 pb-1-1))- - template methodtemplate method uses templates for signal and background mass spectra uses templates for signal and background mass spectra- - ideogram methodideogram method uses analytical likelihood for event to be signal or uses analytical likelihood for event to be signal or background – for each selected eventbackground – for each selected event
template methodtemplate method
ideogram methodideogram method
systematics limited:systematics limited:- jet energy scale- jet energy scale- ttbar modeling- ttbar modeling- W+jets modeling- W+jets modeling- ...- ...
-
W-
W0
sum
page page 3939Tobias Golling
Helicity of the W in ttbar Events
FF++
-
Backup Slides
page page 4040Tobias Golling
-
Probability for a jet to be tagged conveniently broken downin two components: Probability for a jet to be “taggable” (=“taggability”) Probability for a taggable jet to be tagged (=“tagging efficiency”)
Motivation:Decouple tagging efficiency from issues related to
tracking efficiency calorimeter noise
which are absorbed in the taggability
Taggability: Probability to match (R
-
Challenge: Select pure sample of b-jets in data
Enrich dijet data sample in bbbar QCD heavy flavor production:Select jets which contain muons (muon-in-jet): b (->c) -> µ (b-jets) c -> µ (c-jets) /K -> µ (light flavor jets)
PTrel
= transverse momentum of the
muon relative to the (µ+jet) axis
PTrel
discriminates between b-jets and non-b-jets
Templates: Shape of P
Trel distribution for c- and b-jets from MC
Shape of PTrel
distribution for light flavor jets from data
(random tracks in a jet)
jet PTrel
µ
µ+jet
page page 4242Tobias Golling
B-Tagging Efficiency in Data
-
1.) ST vs. NT (Single Tag vs. No Tag): PTrel
Fit
Derive b-fraction in data from PTrel
fit to the templates
before and after lifetime tagging the muon jet:
2.) DT vs. ST (Double Tag vs. Single Tag): PTrel
Fit
Same as 1.) but require that away-jet is lifetime tagged to further enrich the sample in b-jets:
3.) System8: Solely from data 2 samples with different b-fractions
muon-in-jet muon-in-jet with away jet lifetime tagged
2 tagging algorithms soft muon tagger SVT tagger
Solvable system of 8 equations: before and after tagging with any or both of the taggers
b−data =
N − jettag F b−
tag
N − jet F b−
b−data =
N Dtag F b−Dtag
N away− jettag F b−
tag
use Method 3. fo r b−data in c l
data=inc lM C b−
data
b−M C
page page 4343Tobias Golling
B-Tagging Efficiency in Data – 3 Methods
-
ttbar MC:
page page 4444Tobias Golling
inc ldata=in c l
M C b−data
b−M C
Inclusive b-Tagging Efficiency
-
Method 1: Determine N
QCD in untagged
sample (topological approach)
Event tagging Probability PQCD
from data
NQCD
tag = PQCD
NQCD
Caveat:P
QCD has to be determined on
sample with same heavy flavor fraction same kinematicsas the signal sample
Method 2: Determine N
QCDtag directly in
tagged sample (topological approach)
Caveat: Large statistical uncertainty One has to verify that
QCDtag =
QCD
large statistical uncertainty
page page 4545Tobias Golling
QCD Background Determination
-
Use W+jets samples generated by ALPGEN interfaced to PYTHIA Don't use absolute values of cross sections, but rely on their ratios Apply matching procedure between Matrix Element partons and reconstructed jets (à la Mangano)
reduce double counting and sensitivity to parton generation cuts
Remarks: Use matching of Matrix Element partons and parton-jets in the future LO estimate of Wbb / Wjj reproduced by NLO ? Qualified “yes” for scale choices >= 50 GeV and pT cuts of about 15 GeV or greater
page page 4646Tobias Golling
Flavor Composition of W+Jets
-
page page 4747Tobias Golling
Kinematic Distributions of Tagged Events
-
page page 4848Tobias Golling
Result for CSIP Tagger
=7.2−1.21.3 stat −1.4
1.9 syst ±0.5 lum i pb
Dominant systematic uncertainties: Jet energy scale Jet reconstruction efficiency b-tagging efficiency in data
Single Tags: Double Tags:
Luminosity =164 pb-1
Control bins
-
page page 4949Tobias Golling
DØ Tracking System
-
page page 5050Tobias Golling
Particle Identification at DØ
Had.Layers
Tracking Systemwith magnetic field
Calorimeterinduces showersin dense material
InnerSilicon
Detectors
Muon Detectors
InteractionPoint
Absorber Material
Curvature > Momentum
Electrons
Quarks and Gluons
Muons
Jets
EM Layers