# Abstract:

We measure the forward-backward asymmetry in \(b\bar{b}\) pairs at large \(b\bar{b}\) mass using jet-triggered data and jet charge to identify \(b\) from \(\bar{b}\). As a function of \(m(b\bar{b})\), the asymmetry is consistent with both zero and with the standard model predictions.

More details about the analysis can be found in CDF Note 11092.

# Monte Carlo model:

We use dijet Monte Carlo generated with pythia to describe both our signal and the light-jet background. This model reproduces the overall kinematics of the data well.

A full set of model validation plots can be found in the Model validation section.

# Sample purity:

We estimate the purity of the sample by studying the rate at which light jets are \(b\) tagged. This rate is found from a fit to the mass of the vector sum of tracks associated with the identified secondary vertex. The templates in the fit are produced from our pythia Monte Carlo. The per-jet mistag rates are then used to calculate the per-event \(b\bar{b}\) fraction, or the sample purity.

# Jet charge

In order to define the forward-backward asymmetry, we need to be able to identify \(b\) from \(\bar{b}\). We do this using the momentum-weighted sum of the charges of tracks associated with the jets. This provides some separation between \(b\) and \(\bar{b}\), but there is still some charge confusion.

We quantize the individual jet charges, and take the difference between the two quantized charges. From the rates of single-jet charge confusion, we can derive the per-event rates of charge confusion.

# Background asymmetry

We use a data sideband to estimate the asymmetry of the light-jet background. To do this, we look for jets with \(b\) tags derived from a looser version of our tagging algorithm. These jets are kinematically similar to the jets in our signal region, but they are depleted in signal.

# Mass smearing

Because we measure the energies of the jets with a finite resolution, our estimate of the \(b\bar{b}\) mass is also limited in resolution. We estimate this resolution using \(b\bar{b}\) events in our pythia dijet Monte Carlo. We produce a matrix which describes the probability that an event with a \(b\bar{b}\) mass in one bin will be measured with a dijet mass in a different bin.

# Results:

We use a Bayesian technique to extract the hadron-jet level asymmetry result. We use a formula to relate the background asymmetry, the charge confusion rate, the sample purity, the mass smearing, and the signal asymmetry to the observed number of forward and backward events in the data. This likelihood, along with prior probability distributions representing our estimate of each parameter with its uncertainty, is sufficient according to Bayes' theorem to define the posterior probability distribution. We estimate this posterior using Markov chain Monte Carlo.

We marginalize the posterior over all parameters except for the signal asymmetry. The marginal posterior probability density is compared to the SM theory prediction and to the prediction of a model containing a low-mass axigluon. The result is consistent with zero, the SM, and the axigluon model with a mass of 345 GeV\(/c^2\). The 200 GeV\(/c^2\) axigluon model is inconsistent with our measurement at more than 95%.

# Model validation:

Because we employ a Monte Carlo simulation for calibrations and in the correction to the particle-jet level, we must verify that the model adequately describes our data. The model is quite good, and we verified that it was adequate by reweighting the model to match the data and verifying that this produced only a negligible shift in the results.