Measurement of the WW Production Cross Section with Jets at CDF Using 9.7 fb-1 of Data

Matthew Herndon, Will Parker

Please see CDF Public Note 11098 for more details about this analysis.

You can email the authors with any questions you have regarding this analysis.

Cross Section Measurement

At CDF using 9.7 fb-1 of data we measure the WW production cross section in the dilepton channel. We find the WW production cross section to be:

.tex


.eps


.eps


.tex

Abstract

WW production is an interesting process in the gauge sector being produced both by radiation from quarks and multiple gauge boson coupling. It is also a critical background for measurements of Higgs production with decay to WW bosons. The WW cross section is measured using similar selection to the HWW search. The WW cross section measurement is extended to include events with one and two or more jets. A neural net is trained to distinguish WW signal from background. The WW cross section is then extracted using a maximum likelihood method to fit the WW and background neural net shapes. The WW cross section is presented both inclusive and differentially in jet multiplicity. We additionally subdivide the one jet region by leading jet ET. The analysis uses the full 9.7 fb-1 CDF dataset. The measured cross section is 14.0 ± 0.6(stat) +1.6/-1.3(syst) ± 0.8(lumi) pb, for a total uncertainty of +1.9/-1.6 pb.

Signal Region

We select events with two leptons (e, μ) of opposite sign and significant missing transverse energy. This is characteristic of two W bosons decaying leptonically to two leptons and two neutrinos (which escape the detector resulting in "missing" transverse energy). We separate events into categories of zero, one, or two or more jets, before fitting we further separate one jet events into low ET, mid ET, and high ET categories. The tables below show numbers of events from signal and background processes after fitting with systematics errors from tables below, as well as the number of events observed in data.

.tex

.eps

0 Jets Signal Region:

.eps .eps

1 Jet Signal Region:

.eps .eps

2 or More Jets Signal Region:

.eps .eps

Control Regions

In order to insure that our backgrounds are well modeled we construct a number of control regions that are kinematically similar to or the same as our signal region, but orthogonal.

  • Drell-Yan control region:

    Requires low missing tranverse energy and a dilepton mass between 76 and 106 GeV and vetoes e-μ events.

.eps .eps .eps

  • Same Sign control region:

    We reqire two leptons of the same sign. This region checks the modeling of the Wgamma and fake lepton (W+jets) backgrounds.

.eps .eps .eps

  • ttbar control region:

    We reqire two or more jets, and that at least one jet be btagged. This region checks the modeling of the ttbar background.

.eps .eps .eps

Systematic Uncertainties

We assess many systematics on our signal and background processes. Systematics in a single row are correlated unless otherwise noted.



.tex



.tex



.tex



.tex



.tex

Neural Nets

We train a NeuroBayes® neural net to discriminate WW events from background in the 0, 1 and 2+ jets regions.

0 Jet Neural Net Input Variables:


.eps .eps .eps .eps
.eps .eps .eps .eps
.eps

1 Jet Neural Net Input Variables:


.eps .eps .eps .eps
.eps .eps .eps .eps

2 or More Jet Neural Net Input Variables:


.eps .eps .eps .eps
.eps .eps .eps .eps
.eps .eps .eps .eps
.eps .eps .eps .eps
.eps

Likelihood Fit and Results

We divide the 1 jet region by ET and fit the neural net output distributions via a binned maximum likelihood method. We construct the likelihood function from a product of Poisson probabilities for each neural net bin and Gaussian constraints for each systematic.


           .tex

Where ni is the number of data events in the i-th bin. μi is the total expectation in the i-th bin, given by


           .tex

For a process k and a systematic SC, f ck is the fractional uncertainty assigned to the systematic, and (NExpk)i) is the expected number of events from process k in bin i. We evaluate shape uncertainties for WW Scale and JES and DY JES by vertical interpolation:


           .tex

αk is fixed to one, except for the αWW in each bin which float freely, and from which we extract the cross section measurement. We fit simultaneously across analysis bins. The table below shows event yields after fitting. The figures below show the neural net output distributions after fitting, and the distribution of pseudoexperiments.

.eps .eps .eps
.eps .eps .eps