1. W boson helicity measurement
in top pair events in Dø
Bertrand Martin
Workshop on Top Physics : from the Tevatron to the LHC
10/20/07

2. Outline Motivations Analysis guideline Event selection
lepton + jets , dilepton
likelihood discriminant
Analysis technique
cos(θ*) template
Results with 1 fb-1 or _

3. Motivations Standard Model : ν θ* t b W+ l+
Due to the observed parity violation, charged current (W boson) only couples left handed particles
V-A structure of the E.W. current in the S.M. Lagrangian
W boson helicity : sensitive to a possible V+A component :
→ Measured through the cos(θ*) distribution
θ* : angle between the top quark flight direction and
the charged lepton momenta in the W rest frame : t b W+ l+ ν θ*
S.M. Fraction :
Left handed f- ~ 0.30
Longitudinal f0 ~ 0.70
Right handed f+ ~
A non zero f+ could sign new physics…
C.-R.Chen, F. Larios, and C.-P. Yuan, Phys. Lett. B 631, 126 (2005)

4. Analysis guideline 2 final states : Measured dilepton lepton+jets
Fixed to the SM value
Select a data sample enriched in ttbar candidate events
(estimate physics and instrumental background contamination)
Reconstruct the event kinematics :
top quark & W boson leptonic decay
compute cos(θ*)
Generate MC ttbar samples with fixed f0 (0.70)
- different f+ values
Estimate the f+ value that best reproduces the observed cos(θ*) distribution in data

5. Multivariate selection (likelihood discriminant)
Event selection
Dilepton ( ee , eμ , μμ )
2 high pT leptons > 15 GeV
≥ 2 high pT jets > 20 GeV
Main backgrounds :
- Drell-Yan : Z/γ* + jets
- Diboson (WW,WZ,ZZ)
- Fake lepton
Lepton (e, μ) + jets
Only 1 high pT lepton > 20 GeV
≥ 4 high pT jets > 20 GeV
Missing ET > 20 GeV
& MET not aligned with the lepton
Main backgrounds :
W + jets
QCD multijet production
Preselection
Multivariate selection (likelihood discriminant)
See next slide

6. Likelihood discriminant (l+jets)
The variables used to discriminate signal and background must :
be well modeled in the MC (K.S. proba > 5%)
have different shapes between S and B
Good MC modeling
11 “potentially good” variables
( topology, kinematics, b-tagging… )
discriminants can be built out of 11 variables…
Discriminating ?
Not aware that TMVA exists …?

7. Likelihood discriminant (cont’d)
Best Lt
e + jets
Fit ln(s/b)
Likelihood optimization : ensemble tests
Np observed data events after the preselection
→ Create a “mock data set” of the same size ,
picking up MC (ttbar , W+jets) and QCD events
according to preselection efficiencies
→ Fit the ttbar fraction. Repeat 1000 times.
Best Lt : gives the smallest error on the ttbar fraction
Optimal cut : highest discrimination S/√(S+B)
Lt is used to estimate the
data sample composition,
and enrich in ttbar
( Lt cut )

8. cos(θ*) templates Reconstruction of the W leptonic decay(s)
l+jets : 1 neutrino – kinematically constrained : 1 solution , 1 unknown = pz(v)
Kinematic fit used for the (lepton – b jet) pairing, assuming Mtop=172.5 GeV
dilepton : 2 neutrinos unknown momenta – underconstrained
Mtop assumption, algebraic resolution & average over the possible (lepton,jet) pairings
Success in solving the equations : % (nominal rate)
95 % if lepton and jet energies are smeared
After the likelihood cut, the cos(θ*) distribution is built for :
Data
ttbar, for different f+ (V-A / V+A)
Background
Goal : find which f+ value best reproduces the data distribution

9. What it looked like some time ago…
Analyses results
l+jets (1 fb-1)
Dilepton (1 fb-1)
Now…
What it looked like some time ago…
l+jets
230 pb-1
Dilepton
370 pb-1
l+jets
Later…

10. f+ measurement Likelihood maximization ( w.r.t. ns and nb,i ) :
Gaussian term for the background normalization
How well does this f+ hypothesis
match with the cosθ* data distribution
Observed data
Predicted average
nb,i and σb,i obtained after Lt cut
Poissonian probability to observe in the bin j
dj data events with a predicted average of

11. Extracting f+
l+jets
Dilepton
l+jets and dilepton results are marginally consistent (2.4σ stat)
→ Assumed it originates from statistical fluctuation
→ Proceed to combination
Combination

12. Results with 1 fb-1
Combined 1 fb-1
The likelihood maximization does not guarantee f+ > 0 !
Bayesian approach : use a prior probability density
flat for f+  [0 , 0.30] and null in the non-physical region
Confidence interval [f+ min , f+ max] for a given Confidence Level

13. Summary
With 1fb-1 of analyzed data, the combined lepton+jets and dilepton measurements of the right handed W fraction f+ is :
Published result for 370 pb-1 : PHYSICAL REVIEW D 75, (R) (2007)
Compatible with the predicted Standard Model value : f+ = 1.36 x 10-3
The analyses will be updated with an integrated luminosity of ~ 2 fb-1
( with a more optimized multivariate selection )
The best CDF preliminary results for lepton + jets with ~ 1.7fb-1 , using cos(θ*) :

14. Backup slides

15. The D0 experiment
Results on the W helicity are shown for ~1fb -1 of analyzed data, while D0 has ~ 2.8 fb -1 on tape

16. QCD estimation The “Matrix Method”
2 samples used to estimate the QCD background
preselection sample
“loose” sample :
all preselection cuts except lepton quality
likelihood for electrons
isolation for muons

17. Systematic uncertainties
Varying parameters can affect both :
the data sample composition
(different selection efficiency of the likelihood discriminant)
the shape of cos(θ*) distributions
Effect on the fitted f+ : studied with pseudo-experiments (varying the parameters in the peudo-dataset)