Measurement of the ttbar Cross Section in the Lepton + Jets Channel Using Simultaneous Kinematic Fits in 2.7 fb-1 of CDF Data


Andrew Ivanov *, Tom Schwarz
UC Davis
* now at Kansas State

Charles Plager
UCLA / FNAL

Michelangelo Mangano
CERN

Elizabeth Sexton-Kennedy
FNAL

Gavril Giurgiu, Petar Maksimovic, Mark Mathis*, Sal Rappoccio
Johns Hopkins University
*Now at College of William and Mary

Email the authors

CDF public conference note

Abstract
Event selection
Signal and background mdoeling
Systematics
The fit and results
Comparison to previous method
Kinematic validation plots

ttbar Cross Section Results for 2.7 fb-1

Note: All results assume Mttbar = 175 GeV/c2.

Using the KIT Flavor Separator and nJet distribution:
σ ttbar 7.64 ± 0.57 (stat+syst) ± 0.45 (lumi) pb


Method III Fit

Method III Fit Njet Distribution


Abstract

We present a measurement of the top pair production cross section in ppbar collisions at 1.96 TeV, with an integrated luminosity of 2.7 fb-1 at the CDF experiment on the Fermilab Tevatron. We use the a neural-net based flavor separator (the KIT Flavor Separator) and the nJet spectrum to simultaneously measure the the top cross section and background normalizations in the lepton plus jets sample. We measure the top pair production cross section to be
σ ttbar = 7.64 ± 0.57 (stat+syst) ± 0.45 (lumi) pb
The total uncertainty is 9.5%, surpassing the Tevatron Run II goal of 10%. An added benefit of this procedure is the inverse scaling of the systematic uncertainties with integrated luminosity.

Event Selection

We use the standard CDF lepton + jets selection
We apply an additional requirement that the event passes a QCD Veto cut, to remove most of the QCD background. It makes cuts on the transverse mass of the W as well as a quantity called the Missing ET Significance, taking advantage of the fact that in QCD events, these quantities are different typically due to the mismeasurement of the missing ET.

Signal and Background Modeling

Each of these samples is run through a neural-net based flavor separator - the Karlsruhe Institute of Technology (KIT) Flavor Separator. We then assemble the signal and background samples into templates - flavor separator distributions in a given jet- and tag-bin - as seen below. The full set of templates is available here.

1 Jet 1 Tag Template

Including Systematics

In order to estimate systematics, we first make additional templates with the systematic effect in question shifted (e.g., Mistag rate + 20% and Mistag rate - 20%). We then look at the event yields of these shifted templates relative to the yields of our default, nominal templates. Those relative yields are interpolated to define a function - this function parameterizes the event yield as a function of the shift, Rx, in the systematic effect. An example of these functions is shown below, and many more can be found here. These functions are included in the fit as multiplicative factors to the template normalizations.

JES Wbb Polynoid

Final Fits of the KIT Flavor Separator Distributions

The fitter is a binned Poisson likelihood fitter, based on the MINUIT package of ROOT. The fitter was tested by running several thousand pseudoexperiments. The results of those PEs are available here.

Fits for the 2.7 fb-1 data sample follow below.

The input QCD normalization is obtained from a fit to the missing ET distribution before the missing ET cut is applied. The ttbar and W+jets normalizations float freely in the fit. The EW and QCD normalizations are constrained to 30% and 10%, respectivly.

Method III Fit Output of the KIT Flavor Separator after the fit, split by the number of jets and tags.
Method III Fit Log Scale Output of the KIT Flavor Separator after the fit, split by the number of jets and tags, on a log scale.
Method III Fit NJet Distribution
 The flavor separator distribution has been integrated in each jet- and tag-bin, to yield the total number of events in each.

Comparison to Previous Method

We compare the uncertainties we obtained in this analysis to a previous analysis, which used the same amount of data. This allows us to directly see the improvement of using this new technique. The results of this comparison are shown in the table below. Note that the previous analysis did not estimate an uncertainty due to Q2 or color reconnection, nor did they have an uncertainty due to the KIT Flavor Separator. The uncertainty for our heavy flavor correction is included in the statistical uncertainty.

Kinematic Validation Plots

These plots are shown using the signal and background normalizations obtained from the fit. Note that these variables were not used in the fit, and so we do not expect these distributions to look as good as the ones above. These provide evidence that the procedure is correct.

Missing ET in the 1-jet, 1-tag bin
Missing ET in the 2-jet, 1-tag bin

Missing ET in the 4-jet, 1-tag bin
Missing ET in the 4-jet, 2-tag bin



Invariant transverse mass of the W in the 1-jet, 1-tag bin
Invariant transverse mass of the W in the 2-jet, 1-tag bin

Invariant transverse mass of the W in the 3-jet, 1-tag bin
Invariant transverse mass of the W in the 4-jet, 2-tag bin



Transverse momentum of the lepton in the 1-jet, 1-tag bin
Transverse momentum of the lepton in the 2-jet, 1-tag bin

Transverse momentum of the lepton in the 4-jet, 1-tag bin
Transverse momentum of the lepton in the 4-jet, 2-tag bin



ET of the leading jet in the 4-jet, 1-tag bin
ET of the leading jet in the 2-jet, 2-tag bin

ET of the second jet in the 2-jet, 1-tag bin
ET of the second jet in the 4-jet, 2-tag bin



Two-dimension displacement of the first tagged jet in the 1-jet, 1-tag bin
Two-dimension displacement of the first tagged jet in the 5-jet, 1-tag bin

Secondary vertex mass in the first tagged jet in the 1-jet, 1-tag bin
Secondary vertex mass in the first tagged jet in the 5-jet, 1-tag bin


Last modified on 04/23/10 by Mark Mathis