Top production and branching ratios

1
Top production and branching ratios

• For DØ collaboration
• Elizaveta Shabalina
• University of Illinois at Chicago
• Wine and Cheese seminar at FNAL
• 09/16/05

2
Outline

• Introduction
• Top pair production
• Dilepton channel
• Leptonjets
• Event kinematics method
• B-tagging method
• Top branching ratio
• Top pair production in all hadronic channel

3
Top quark physics today

• Tevatron was built more than a decade ago to
  discover top quark successfully achieved in
  1995
• Run I cross sections
• CDF and D0
• Precision (25) was severely statistically
  limited

• At present
• ?s 1.96 TeV - 30 higher production rate
• much higher luminosity
• Current goal deliver precision measurements
• Theoretical prediction of cross section 6.5
  accuracy
• Tev2000 study precision of ttbar cross section
  measurement
• Five WC seminars since June 1st are dedicated to
  top physics

4
Top cross section - motivation

• Important test of perturbative QCD
• Higher production rate ttbar resonances
  (topcolor)
• Measure in different channels
• Exotic top decays (to charged Higgs or light
  stop) different cross sections in different
  channels
• Dilepton to ljets cross sections ratio
  tests top decays without W boson in final state
  

• Measure with different methods
• b-jet tagging method assumes Br (t ? Wb) 1
• an implicit use of the SM prediction
  Vtb0.9990 ? 0.9992 (at 90C.L.)
• Topological method is free from this assumption
• Using both test of top decays without b
  quark in the final state

Test of Standard Model
5
Top production

• Standard model pair production through strong
  interactions
• Standard model electroweak production (single top)

Discovered in Run I

• 6.77 0.42 pb for mtop 175 GeV

To be observed in Run II
6
and decay

• Very short lifetime ? decays as a free quark
• Br (t ? Wb) ? 100

• W decay modes determine top quark final state
• Dilepton (ee, µµ, eµ)
• Both Ws decay leptonically
• BR 5
• Lepton (e or µ) jets
• One W decays leptonically, another one
  hadronically
• BR 30
• All-hadronic
• Both Ws decay hadronically
• BR 44
• thad X
• BR 23

7
DØ detector
All detector subsystems are important for high
quality top quark measurements

• Electrons - energy clusters in EM section of the
  calorimeter and track in the central tracking
  system
• Muons - track segments in muon chambers and
  track in the central tracking system
• Jets - clusters of energy in EM and hadronic
  parts of calorimeter

8
Tuning simulation to data

• Monte Carlo simulation is used to calculate
  selection efficiencies and to simulate event
  kinematics
• improve the agreement between data and MC
• additional smearing of the reconstructed objects
• correction factors derived from comparison of
  Z?ll data and MC events and applied to MC
• Systematic uncertainties from uncertainties on
  the smearing parameters and/or from the
  dependence on detector regions, various jet
  environment

RMS
SF
9
Electron and muon identification

• Electron
• Deposit gt90 of energy in the EM calorimeter
  within a cone of ?Rlt0.2 relative to the shower
  axis
• Isolated the ratio of the energy in the hollow
  cone 0.2 lt ?R lt 0.4 to the reconstructed electron
  energy 15
• Transverse and longitudinal shower shapes
  consistent with those expected for an electron
• Reconstructed track found within ?Rlt 0.5 from the
  shower position in the calorimeter
• Discriminant combining information from central
  tracking system and calorimeter is consistent
  with the expectations for a high-pT isolated
  electron

• Muon
• a muon track segments are matched inside and
  outside of the toroid
• timing (from associated scintillator hits) is
  within 10 ns of the interaction ? muon originates
  from primary vertex
• a track reconstructed in the tracking system
  belonging to event vertex is matched to the muon
  candidate found in the muon system
• Isolated in calorimeter and in the tracking
  system isolation criteria are different for
  dilepton and ljets analyses

l o o s e
l o o s e
t i g h t
t i g h t
10
Dilepton channels

• Selection
• At least two jets (pTgt20 GeV, ylt2.5)
• Two charged opposite sign leptons (pTgt15 GeV e
  ?lt1.1 or 1.5lt?lt2.5 µ ?lt2)
• Lepton quality tight µ, tight e in ee,
  loose e in eµ (electron discriminant
  distribution in data is used to extract ttbar
  signal)
• Large missing ET in ee and µµ channels no cut in
  eµ
• and further selections are optimized for
  each channel to account for difference in
  backgrounds

• Signature

p
?
p
?
11
Backgrounds

• Drell-Yan background rejection
• ee
• veto events with 80ltMeelt100
• gt35 GeV(gt40 GeV) for Meegt100 GeV (Meelt80 GeV)
  
• µµ
• gt35 GeV
• is tightened at low and high values of
  azimuthal distance ?f( µ, )
• Remove events with ?f(µ, )gt175

• Physics
• Leptons from W/Z decay and missing ET from
  neutrinos WW/WZ, Z/?????ll
• Estimated from MC
• Instrumental
• jet or lepton in jet fakes isolated lepton (QCD,
  Wjets)
• missing ET originates from resolution effects,
  misreconstructed jet or lepton or noise in
  calorimeter (Drell-Yan processes Z/??ee(µµ) (eµ
  channel is not affected)

12
Backgrounds

• Fake in Z/?? ee(µµ) primary background
  in ee(µµ) channels
• spectrum in MC Z events agrees well with
  data
• µµ directly from simulation
• ee fake rate is measured in ?jet events
  multiplied by the number of data events that fail
  the selection but pass all others in MC
• ?2 cut on fit to Z hypothesis (µµ)

• Fake electron (Wjets, QCD events)
• Fake rate ? from data sample dominated by fake
  electrons (2 loose EMs, low , outside Z
  mass window)
• Measure fraction of loose electrons that pass
  tight criteria
• Fake isolated muon (muons from heavy flavor
  decays)
• Use loose dimuon events
• One non-isolated muon
• Measure probability that the other is isolated
• Multiply by number of loose-tight events in data

13
eµ channel

• The cleanest channel
• Optimized to minimize total error
• Optimal cut removes Z/???? background
• Extract fake electron background from the fit to
  the observed distribution of electron LH in data
• Shape for real electrons from Z ? ee data
• Shape for fake electrons from background
  dominated sample (anti-isolated muon, low missing
  ET)

real electrons
fake electrons
14
Results
Background control bin
ttbar signal
15
Dilepton events properties
eµ
Electron likelihood distribution for data events
after full selection
combined
for ?tt 7 pb
16
Cross section calculation

• ee and µµ channels counting experiments
• Define likelihood for each channel based on
  Poisson probability that expected number of
  signal background events mj is compatible with
  observed Njobs
• where
• eµ channel extended unbinned likelihood method

- electron likelihood distributions for signal
and background events
- number of physics background events
xi value of electron likelihood for an electron
in each event
Fit simultaneously cross section and Nfake
17
Cross sections
ee
µµ
eµ
For dilepton channel combination minimize the sum
of negative log-likelihood functions for
individual channels
370 pb-1
combined dilepton _at_ m_top 175 GeV
18
Systematic uncertainties
Comparable contributions from all sources
19
Leptonjets channel

• Signature

• Selection
• One isolated lepton (pTgt20 GeV e ?lt1.1 or
  1.5lt?lt2.5 µ ?lt2)
• At least four jets (pTgt20 GeV, ylt2.5)
• gt20 GeV and not collinear with lepton
  direction in transverse plane

jet
_
?
_
?
p
b
jet
jet
jet
20
Sample composition
Multijet background
Estimate amount of QCD from Matrix Method
Nloose
Ntight
21
Discriminant function definition

• Only ttbar and Wjets simulated events are
  used to build discriminant
• Kinematic properties of multijet background are
  similar to Wjets

Transform topological variables to be less
sensitive to statistical fluctuations in regions
of rapid variations
Build logarithm of the signal to background
ratios and fit with polinomial
22
Discriminating variables

• HT scalar sum of the pT of four leading jets
• Centrality ratio of scalar sum of jet pT to
  scalar sum of jet energies
• Aplanarity
• Sphericity

• Set of variables is chosen
• to provide the best separation between ttbar and
  Wjets background
• to have the least sensitivity to the dominant
  systematic uncertainties
• Only 4 highest pT jets are used to build
  variables

• kTmin?RjjminpTmin/ETW, ?Rjjmin ? maximum
  separation between pairs of jets, ETW scalar
  sum of lepton pT and , pTmin pT of the lower
  pT jet

Linear combination of the eigenvalues of a
normalized momentum tensor
23
Discriminant function

• Fit modeled discriminant function distribution
  to that of data
• Extract Nttbar, Wjets and multijet events in
  the sample

By construction background peaks at 0, signal
at 1
24
Cross section

• Define
• where Poisson probability
  density for n observed events given µi predicted,
  i runs over all bins of the discriminant, niobs
  content of bin i as obtained in selected sample
• Expected number of events in bin i is a function
  of number of ttbar, W and QCD
  events in the selected sample
• f - fractions in bin i of the ttbar, W and QCD
  discriminant templates
• Second term implements Matrix Method constraint
  on number of QCD events via the Poisson
  probability of the observed number of events in
  loose but not tight sample

25
Results
240 pb-1
ejets
µjets
ejets
µjets
26
Results combined
For leptonjets channel combination minimize the
sum of negative log-likelihood functions for
individual channels
Sample composition 38 ttbar 44 Wjets 18
multijet background
240 pb-1
combined _at_ m_top 175 GeV
27
Event kinematics
signal dominated
Background dominated
28
Systematic uncertainties
By far the largest systematic uncertainty comes
from the Jet energy calibration, 90 of total
error
29
Leptonjets channel with b-tagging

• event has two b-jets
• b-jets in background processes are seldom
• Use this feature to discriminate signal from
  background
• Dramatically improves signal-to-background ratio
• Signature of a b decay is a displaced vertex
• Forms long lifetime of B-hadrons (c? 450µ)
• B-hadrons travel Lxy 3mm before decay with
  large charged track multiplicity

• Use same selection as in topological analysis but
  
• Relax cut on jet transverse momentum pT gt 15 GeV
• Use events with njet?3
• Use events with one and two jets as control
  samples for background estimation

QCD
Wjets
30
b-tagging algorithm - SVT

• Reconstructs secondary vertex
• ?2 tracks with pT?1GeV, ?1 SMT hit, impact
  parameter significance gt3.5
• Removes tracks associated with K0S, ?0 and photon
  conversions (? ? ee-)
• Positive tag
• Secondary vertex within a jet with a decay length
  significance Lxy/?Lxygt7
• Negative tag
• Secondary vertex within a jet with a decay length
  significance Lxy/?Lxylt?7 (due to resolution
  effects)

Impact parameter
31
Tagging rates

• b-tagging efficiency
• Measured in dijet data events for jets with muon
  inside
• Compare two samples with different heavy flavor
  content (increased by tagging the away jet)
• Tag jets with two tagging algorithms SVT and SLT
  (SLT soft muon with pTrelgt 0.7 GeV inside a
  jet)
• Solve system of 8 eqs to extract semileptonic
  b-tagging efficiency
• Use MC to correct measured efficiency to that for
  inclusive b-decays

• Light tagging rate
• Measure negative tagging rate in dijet events
  (low missing ET)
• Correct for long-lived particles in light jets
• Heavy flavor contribution in dijet events
• c-tagging rate
• From MC corrected with the SF derived for
  b-tagging

32
Backgrounds

• Calculate QCD (non-W) contribution from Matrix
  Method
• Subtract small backgrounds (single top, VV, Z???)
  using known cross sections
• Separate W from ttbar using difference in
  their tagging probability
• Interpret excess in observed tagged events with
  ?3 jets over predicted background as ttbar
  signal

small bkgr
33
Event tagging probability
DØ RunII Preliminary, 363pb-1

• Use MC to calculate event tagging probability
• Depends on the flavor composition of the jets in
  the final and on the overall event kinematics
• Apply the tagging rates measured in data to each
  jet in MC based on its flavor, pT and y
• For Wjets, use the ALPGEN MC to estimate the
  fraction of the different Wheavy flavor
  subprocesses.

34
Results
Background dominated
35
Kinematics of llets tagged sample
DØ RunII Preliminary, 363pb-1
36
Cross section

• Define likelihood based on Poisson probability
  that expected number of signal background
  events mj is compatible with observed Njobs
• The product is taken over 8 independent channels
  e/µ jets, one-/two-tags, 3rd and 4th jet
  multiplicity bins
• Multijet background in each tagged sample, and
  the corresponding samples before tagging, is
  constrained within errors to the amount obtained
  from Matrix Method

37
Result and systematic uncertainties

• Combined statistical and systematic error is
  obtained
• Individual contributions are obtained by
  refitting after fixing all but the Gaussian term
  under study

• Gaussian term for each source of errors is
  included (nuisance parameter method)
• Each source is allowed to affect the central
  value of the cross section

• Systematic and statistical uncertainties are the
  same 11
• Main sources
• JES and jet ID
• B-tagging efficiency in data
• W fractions
• Luminosity

363 pb-1
38
Branching ratio

• Probe the assumption Br(t?Wb)1
• CKM matrix element Vtb0.9990?0.9992 _at_90 C.L.
  
• R0.9980?0.9984. True in SM assuming
• Three quark generations
• CKM matrix is unitary
• For expanded CKM matrix Vtb0.07?0.9993 _at_90
  C.L.
• CDF measurement
  162pb-1

39
Method

• Split selected sample into 3 categories 0,1 and
  ?2 tags
• Predicted number of ttbar events depends on R
• Fit R and ?tt from the number of observed tagged
  events and the event kinematics in 0 tag sample
  

• Compute probabilities to observe 0, 1 and ?2 tags
  for each final ttbar state
• Combine to obtain
• Pn-tag(R), n-tag0, 1, ?2
• Use topological discriminant in 0 tag sample with
  ?4 jets to determine ttbar content

2 b-jets
1 b, 1 light jet
2 light jets
40
Fitting procedure

• Perform binned maximum likelihood fit to data in
• 10 bins of discriminant of ljets 0 tag, Njet?4
• 2 bins of ljets 0 tag, Njet3
• 4 bins of ljets 1 tag, Njet3, ?4
• 4 bins of ljets 2 tag, Njet3, ?4
• Statistical fluctuations of the multijet
  background are taken into account by additional
  12 Poisson terms (0,1, ?2 tags, nj3, ?4, ejets,
  µjets)
• Nuisance parameter method to include systematic
  uncertainties

Njet3
Njet?4
41
Result
Statistical uncertainty dominates
Potential for improvement include dilepton
events
42
All hadronic channel

• Signature 6 jets, 2 b-quark jets
• All decay products should be visible in the
  detector, no energetic neutrinos produced
• Six jet multijet production rate is many orders
  of magnitude larger than ttbar
• Impossible to extract signal without tagging
  b-jets SVT algorithm is used

• Njets ? 6, pTgt15 GeV
• Suppress multiple interactions (second
  interaction is also hard QCD process)
• Reject events with several hard primary vertices
  gt3 cm apart
• At least 3 jets assigned
• Jet is assigned to PV if at least 2 tracks from
  it come from PV
• Removes 32
• Reject bb background
• ?R(tagged jets)gt1.5

43
TRF and neural network

• Derive TRF (tag rate function) in the 6-jet data
  sample (ttbar contribution is 0.3) in 4 bins of
  HT 0?200, 200 ? 300, 300 ? 400, ?400 GeV
• Parameterize as a function of jet pT, ?, f,
  position of primary vertex along the beam
• Compare predicted and observed tagging rates and
  obtain correction factor

• Select a set of variables discriminating signal
  from background
• Avoid as much as possible JES dependent variable
• Use smallest possible number of input variables
• Combine into Neural network

44
Discriminating variables
Variables are designed to address different
aspects of the background

• Energy Scale HT
• Event Shape aplanarity
• Soft non-leading Jets ET56 geometric mean of
  the transverse energies of the 5th and 6th
  leading jet

• Rapidity lth2gt - weighted RMS of h of 6 leading
  jets
• Top Properties
• Mmin3,4 the second smallest dijet mass
• Mass likelihood, ?2-like variable calculated from
  MW, ?W, ?top

45
Cross section calculation

• Background was estimated on the sample containing
  signal correct cross section
• ?TRF probability to tag ttbar MC event using
  TRF
• ?btag probability to tag ttbar event using b,c
  and light tagging rates

46
Cross section and uncertainties
At mtop 175 GeV, 350 pb-1
55 relative error
Potential for improvement make better use of
double tagged events
JES error dominates 70 of total systematic
error
CDF 311 pb-1 40 relative error
47
Summary
Accepted for publication in PLB
Best precision 16 ljets/btag at 363 pb-1
Work in progress on combination of the latest
results up to 370 pb-1
CDF combined up to 350 pb-1 13 relative error
48
From TeV2000 to reality

• Do we meet expectations?
• For 363 pb-1
• predicted 180 b-tagged events (scaled from 500
  per fb-1)
• Observed 140 (241 tagged event, 101 expected
  background)

• Can we do better?
• Data quality
• Improved calorimeter calibration
• Improved performance of SMT is crucial
• Improved simulation
• Optimization
• Better tools
• Neural network lifetime b-tagger
• Fighting major sources of systematic
  uncertainties

49
Glance into the future
Total error on ljets/btag channel

• Assumptions
• Errors from limited MC statistics are set to 0
• Luminosity dependent and constant terms
• JES
• B-tagging efficiency
• Lepton identification
• Limiting factors
• Luminosity (6.5)
• Heavy flavor fractions (5.9)
• Solutions
• Measure ratio of ttbar to Wjets cross section
• Combine channels

This will be replaced by a real plot
50
Conclusion

• The precision of the latest top pair production
  cross section measurements rapidly approaches
  accuracy of theoretical prediction and will allow
  to probe Standard Model
• With combination of measurements in different
  channels and using different methods we have an
  excellent opportunity to exceed the precision
  limit set by TeV2000 11 for 1 fb-1
• and the one for 10 fb-1 5.9 but with less
  luminosity!

This is a challenge. Lets go for it!