Measurement of the relative fraction of ttbar events produced via gluon-gluon fusion in 955 pb-1 of data


Authors

Ricardo Eusebi
(FNAL)

Eva Halkiadakis
Sunil Somalwar
Jared Yamaoka
(Rutgers)

Abstract
We present a measurement of the relative fraction of ttbar events produced via gluon-fusion to the total number of ttbar events, which we call Cftrue, using a template method based on Neural Networks probabilities. Using a total integrated luminosity of 955 pb-1 we find Cftrue< 0.33 at 68% confidence level, and Cftrue< 0.61 at 95% confidence level.

Analysis

The total production cross section of a given physical process can, in general, be measured simply by counting events in specific final-state channels. The contribution from different production processes to the total production cross section is, however, very hard to measure. Typically, the spin information of the decay products is washed out by the hadronization process of quarks, hence blurring out any difference in the kinematic properties of the different production processes.


The Standard Model predicts gg produced events and qq produced events will occur with relative fractions of 15% and 85% respectively. The theory has large uncertainties. For example, the contribution from gg fusion at the Tevatron can vary up to a factor of 2 (from 10-20%).


The top quark, with a mass of about 175 GeV, has an expected lifetime of approx 0.5 x 10-24 seconds; which is an order of magnitude faster than the typical hadronization time. As a consequence, the spin information carried by the top quark is preserved in the decay products, allowing the different production processes to retain their kinematic characteristics in the final state. We use a neural network to maximize the discrimination between gg and qq produced events. You can find a description of the variables that go into the neural network here.
This sample is composed mainly of three processes: ttbar produced via gluon-fusion, ttbar produced via quark annihilation, and background from the W boson produced in association with other quarks. Top quark events have b quarks in the decay. CDF has a method to identify the b quarks called tagging, and we split our sample into subsamples of 1 tag and 2 tags.

Neural network for the three components of our sample. These are used to fit our data for the correct fraction of gg produced top events. 1 Tag.


Neural network for the three components of our sample. These are used to fit our data for the correct fraction of gg produced top events. 2 Tag.


To obtain our result, we created a large number of simulated data samples with the varying fractions of gg produce top events. These simulated samples, or "pseudo-experiments" as we call them, are then fit to the neural network templates. We then use a statistical method called Feldman-Cousins (FC) to create a map from the fitted value of our pseudo-experiments to the real gg production fraction. More info on Feldman-Cousins Method can be found here.

Like any measurement there are sources of uncertainty. In our case our main source of uncertainty comes from statistics, we don't have a large number of ttbar events. Other sources of uncertainty come from systematics, which are uncertainties based on our experimental method. Systematics that we consider are: how well do we measure the energy of the decay products of the top, how well do we model the initial state and final state radiation in our simulation, how well do we know the background shape and composition, and how well our simulation models more complicated ways of producing top events (so call next to leading order (NLO) processes).


These are FC bands. The true gg fractions (Y axis) Vs a quantity obtained from the NN fit (X axis). The red line indicated the NN fit obtained from data, and its intersection with the green band indicates the 68% confidence level limit. The intersection with the blue line indicates the 68% confidence level limit if we were not affected by systematics. (EPS)



Here you can see the 68% confidence level limit and the 95% confidence level limit. (EPS)




Here is the fit of to the neural network for the 1 tag events. (EPS)


Here is the fit of to the neural network for the 2 tag events. (EPS)

A full set of plots for all the variables that go into the neural network can be found here.


Results

Fraction of gg produce events. Integrated luminosity 955 pb-1

More Information


Jared Yamaoka
Last modified: Thurs May 17 11:56:50 CST 2007