Z Decays to b Quark Pairs

Z decays to b quark pairs are not exactly a unknown piece of physics. Since 1992 the LEP experiments have detected several millions of them, and more have come from the polarized beams of the SLC. The process is thus very well understood; the Z is one of the best known particles, and there can be no surprise in the thereabouts. At a proton-antiproton collider this particular process has never been seen before, though. The UA2 collaboration published in 1987 an analysis of jet data where they could spot the combined signal of W and Z decays to dijet pairs, but the decay of the Z to b quarks was not separated from the other hadronic decays.

A Z -> bb signal is interesting at the Tevatron because it mimics the topology of direct Higgs boson production, and its most probable decay (if the Higgs mass is lower than 130 GeV). Moreover, a large sample of these decays would allow a nice calibration tool for the top mass measurement.

To search for Z decays to b quark pairs, we start from a dataset collected by triggers requiring the presence of a central muon candidate with transverse momentum exceeding 7.5 GeV. This dataset provides lots of events containing b quark decays, and is advantageously free from any bias on jet energies. A sample with electrons would also suggest itself, but electrons are more difficult to identify than muons when they are embedded in a high energy jet, making the expected number of Z events in a clean electron sample half of that expected in our muon sample.

After requiring the presence of two secondary vertices from $b$ decay in the two leading jets (hereafter "SVX tags"), we apply tight kinematic requirements on two variables that discriminate the electroweak production from QCD backgrounds on the basis of the amount and topology of the radiation surrounding the two leading jets, as can be seen here.

We have devised some improved jet momentum corrections, to be used on top of the standard jet corrections in use at CDF, to be used on the very particular events featuring a bb final state with a semileptonic decay to a energetic muon. These jet momentum corrections, hereafter named MT corrections, use the muon momentum, the jet charged fractions and the missing transverse energy of the event to improve the jet momentum resolution. They improve the relative mass resolution for a Z signal in PYTHIA and HERWIG Monte Carlo datasets by about 50%.

We apply our MT corrections and perform a counting experiment using a very precise and reliable data-driven background prediction. We find an excess of 70+-20 events around the Z mass (check it on this table); the excess has a nice gaussian shape in the mass distribution, as can be seen in this nice plot.

As a check that the MT corrections perform as expected, we can see what happens to the bump-like excess if we do not make use of them: the bump has indeed a lower average mass this time, and a larged width, as the MC predicts: check it in this plot.

To select the data the cuts on the kinematic variables have been optimized to give the best expected signal significance of an excess due to Z decay, as shown here and here . The binning of the mass distribution is also optimized to get a significant excess in a single bin, as shown in this plot.

The excess of events behaves as expected also when studied as a function of the two kinematic variables used to select them, as shown here and here. The significance of the excess, evaluated as the number of standard deviations corresponding to the probability that the observed excess was due to a upward background fluctuation, amounts to 3.23 sigma. This result is obtained taking into account a 4% systematic uncertainty on the background prediction: it has been estimated in many ways, one of which is to predict events with only one SVX tag (that do not contain a significant fraction of Z signal) using events with no SVX tags. The agreement of observed and predicted mass spectra, shown here, is impressive.

A unbinned likelihood technique is then used to fit the dijet mass spectrum to the sum of a background shape (obtained from data failing the 2 SVX tags requirement) and a free gaussian shape, as shown here. The results for both peak mass and width are in very good agreement with expectations for the Z signal, and the fitted signal turns out to be 91 +- 30 (stat.) +- 19 (syst.) events.


Back to the Recent Work frame