Single-Top Search using a Likelihood Function Technique

1.5/fb Single-Top Search Using a Multivariate Likelihood Function Technique

Sarah Budd, Catalin Ciobanu, Thomas Junk, Tony Liss, Michael Wren
University of Illinois at Urbana-Champaign

Abstract

This webpage documents the multivariate likelihood function search for single-top production with CDF II data corresponding to 1.5fb-1 of integrated luminosity. A top mass of 175 GeV is assumed. A p-value for single top production of 3.1*10^{-3} is measured, which corresponds to a 2.74 Sigma excess over the Standard Model background. The best-fit value for the combined s+t-channel production of single top quarks assuming the Standard Model ratio of their production cross sections is Sigma(s)+Sigma(t) = 2.72+1.29-1.12 Separately, the s-channel and t-channel cross sections are fit to be 1.14+1.36-1.14 pb and 1.29+1.21-0.99 pb. From the one-dimensional cross-section fit, we extract |Vtb|=0.97+0.21-0.19(exp) +- 0.07(theor). Assuming a flat prior in |Vtb|^2, we obtain |Vtb|>0.52 at the 95% CL.

Documentation

The public note for this analysis is availible at http://www-cdf.fnal.gov/physics/new/top/confNotes/cdf8964_STLF_pub.pdf

Significance using 1-D t-channel Likelihood Function

Fit to Data Results

Extraction of |Vtb|

From the 1D fit to sigma_s+sigma_t, assuming that both sigma_s and sigma_t are proportional to |Vtb|, that sigma_s/sigma_t is as predicted by the Standard Model, and that the top quark decays to Wb 100% of the time, we obtain |Vtb|=0.97+0.21-0.19(exp) +- 0.07(theor). Further assuming a prior flat in |Vtb|^2 in the range 0<|Vtb|^2<1 we perform a Bayesian analysis and find that |Vtb|>0.52 at the 95% CL.

Exclusion of Anomalous Single-Top Production

Observed upper limit: 5.14 pb at the 95% CL
Median expected upper limit: 2.06 pb (in the absence of single-top production)

Changes to the Analysis since 1fb-1 Result

Top Group wide changes

Used new MC with minbias overlay and a new Alpgen version (v2.1.0) for several major backgrounds (W+Jets and Z+Jets).
Improved jet reconstruction (jet energy now corrected to hadron level)

Single Top group changes

New Neural Net b-tagger optimized for updated MC.
Background estimate redone using new MC. In addition, we now calibrate the Heavy Flavor fraction using a sideband data region. The new background estimate and a plot of the number of jets is shown below.
Updated veto for non-W (QCD) events.

Changes specific to the Likelihood Function Analysis

Replaced Kinematic Fitter with Kinematic Solver. By constraining the lepton momentum, the b-jet direction, the b-jet invariant mass, and the p_t of the top quark, we find that the number of constraints on the b-jet and neutrino system equals the number of unknowns. We then analytically solve for b-jet energy and neutrino p_z instead of using Minuit to fit for these solutions. The old kinematic fitter failed to converge for a fraction of events, and the new kinematic solver fixes this feature.
A coding error was discovered and fixed in ordering the input variables passed to the MADGRAPH matrix element calculation.
Updated the variable used as inputs to the Likelihood Function. We replaced the t-channel and s-channel versions of the top mass (used in the t-channel and s-channel LF respectively) with the chi-squared output from the kinematic solver. We also added several variables to the s-channel search, including Mlvbb, the s-channel MADGRAPH matrix element, and the chi-squared output from the kinematic solver.