Measurement of the top pair cross section in dileptonic channels with a hadronic tau and t→bτν branching ratio, with 9.0 fb-1


Primary Authors: M. Corbo, S. Lammel, A. Savoy-Navarro

Public Note of the current analysis


Abstract


We present an analysis of tt dilepton events with one hadronic tau decay. The analysis is based on 9.0 fb-1 of proton-antiproton collisions recorded with a low pT electron or muon plus isolated track trigger at CDF. Background from misidentified tau leptons is estimated using jet enriched data samples. We measure the t t production cross section and top into tau branching ratio. We find σtt = 8.2 ± 2.3 (stat.) +1.2-1.1 (sys.) ± 0.5 (lum.) pb, and BR = 0.120 ± 0.030 (stat.)+0.022-0.019 (sys.) ± 0.007 (lum.); these are up to date the most accurate results in this top decay channel and are are shown to be in good agreement with the results performed at the Tevatron considering all the other decay channels of the top. The branching ratio is measured in both a combined approach and separating single tau and di-tau events with a log likelihood method. This is the first time an attempt is made to separate single tau and di-tau events. The results on cross section and BR are in good agreement with the expectations of the SM within the experimental uncertainties. The measured branching ratio can thus be used to limit new physics, like the charged Higgs boson.


Signal and Background Events


We estimate the physics background of our selection using simulated data samples. We rely on the simulation to describe the overall selection acceptance but we apply scale factors to account for small mismodeling. For what concerns the background with fakes of the tau candidate we have two major sources: events where QCD induced jets are misidentified as tau decay products and a smaller contribution of events where an electron or a muon is misidentified as a tau. We implemented a technique to evaluate the background with QCD hadron jets faking taus which is based on the calculation of the probability of jets to pass the tau identification cuts, generally named tau fake rate. The component to the background caused by electrons or muons misidentified as taus is evaluated through simulated data.

The measurement of the fake rate is obtained as average between the fake rate of the leading jet and the subleading jet (second jet with highest transverse energy). We consider as systematic uncertainty the disagreament between results of the leading and the subleading jets



The misidentification rate of jets in tau candidates as function of the calorimeter transverse energy. On the left 1 track jets, on the right 3 track jets.




Event Selection


The requirements of our selection of top pair events are:

  • exactly 1 tau candidate;
  • exactly 1 electron or 1 muon candidate;
  • opposite electric charge between the tau and the other lepton;
  • 2 or more jets;
  • missing transverse energy > 10 GeV.

To remove background from Drell-Yan processes we apply a veto. We use a cluster based rejection method for the electron misidentified as tau and a track based method to reduce muons in the tau sample. We consider all jets with high electromagnetic energy fraction, EHAD/EEM > 0.9, in pseudorapidity region |η| < 2 as potential non identified electrons. We remove only those events where the electron/cluster invariant mass is 86 < M < 96 GeV.

We consider each minimum ionizing particles as potential non-identified muon. We reject events with muon plus MIP invariant mass 76 < M < 106 GeV and M < 15 GeV.

We then select events using a variable defined as scalar sum of the ET of the objects in the event:



where ETτ is the cluster ET of the tau decay products, ETjeti is the ET of jets i.

The next table summarizes the uncertainty in the selection efficiency due the propagation of the systematic uncertainties, expressed in percentage.



The result of the kinematic selection is summarized below.



Below we show the basic characteristic kinematic distributions. The figures correspond to the selection on the electron plus isolated track sample and CMUP and CMX muon plus isolated track samples.







To remove the contamination we require at least one secondary vertex tag from the tight SECVTX algorithm. We select jets in simulated events, which are matched in ΔR < 0.4 with b flavoured hadrons. This class of jets is used to compute the number of genuine SECVTX tags.

The jets having a SECVTX tag could contain a contamination of light flavoured jets erroneously tagged as b induced jets, commonly called ``mistags''. For what concern the MC simulated samples we do not rely on the SECVTX selection of light flavoured jets. The jets in simulated events not matched with b flavoured hadrons are considered as possible SECVTX mistags and are weighted by the mistag probabilities.

We repeated the study of the systematic uncertainties at this stage of the selection: kinematic selection and at least secondary vertex tag. The next table summarizes the uncertainty in the selection efficiency due the propagation of the systematic uncertainties. They are expressed in percentage.



The result of the kinematic selection is summarized below.



Below we show the basic characteristic kinematic distributions. The figures correspond to the selection on the electron plus isolated track sample and CMUP and CMX muon plus isolated track samples.







Events with misidentified taus show high pT electron or muon, missing transverse energy and a b induced jet. To study in more detail the dominant contributions in the misidentified tau background we use tt and W+ bb Monte Carlo samples. We identified the top pair decay in sigle lepton to be the major source of background with tau fakes.

To separate signal production from the main background of misidentified taus, we look for variables that distinguish the two sources. We choose two variables from the tau identification requirements like the ratio of the tau cluster energy and the momenta of the tracks, and the sum of the track momenta in the tau isolation cone.

We use the kinematic of top pair decay in single lepton mode to identify variables which allow to distinguish them from our signal events. We use
  • the missing transverse energy,
  • the transverse mass of the system composed by electron plus missing transverse energy,
  • the transverse energy of the third highest ET jet.

The method we implement is known as log-likelihood ratio discriminant: the tool is easily obtained combining one-dimensional distribution templates of background and signal events. We report the distribution on data.



After the requirement ln(LR) > 0, a significant reduction can be noticed in the number of background events with fake taus.







We repeated the study of the systematic uncertainties at this stage of the selection: kinematic selection, at least secondary vertex tag and the requirement on the logarithm of the likelihood ratio ln(LR) > 0. The next table summarizes the uncertainty in the selection efficiency due the propagation of the systematic uncertainties. They are expressed in percentage.



The result of the kinematic selection, the secondary vertex requirement and ln(LR) > 0 is summarized below.





Results


We measured the top pair production cross section using the equation:


,

where Nsel indicate the number of observed events, Nbg is the number of background events, εττ and ετl are the selection efficiencies for the ditau and tau plus lepton channel, L is the instantaneous luminosity;

,


,

are the branching ratio in the lepton plus tau and ditau channel respectively.

We measure the top pair cross section and we propagate the systematic uncertainties individually. We correlate systematics among channels, but we treat each source as uncorrelated. We include also the statistic uncertainty from the selection of the Monte Carlo events. The result is:



The measurement is in agreement with the previous CDF measurements.



We can derive from this cross section measurement:



We measure the branching ratio of top into tau using as signal events the top pair decay into lepton plus tau. The branching ratio is given by the equation

,

where we assume top branching ratio into lepton neutrino plus b quark to be equal to the branching ratio of W boson into tau neutrino. We use for this measurement the top cross section obtained with the combination of the results of the CDF collaboration: 7.5 +- 0.5 pb. We obtain:

,




Ditau channel discrimination


Our aim is to perform a measurement of the branching ratio of top into tau, neutrino and b-quark which uses the information of the top pair decay in ditau mode. For this reason we first separate the lepton plus tau component using a second Log-Likelihood ratio discriminant. We use the simulated events of top pair into ditau as signal sample and the simulated events of top pair into lepton plus tau as background sample.

The probability distribution function we use are obtained normalizing the distributions of

  • the transverse mass of the lepton plus missing ET,
  • the azimimuthal angle between the lepton and the missing ET,
  • the lepton transverse energy.

The ditribution of expected and observed events is given by the following Figure.

,

We proceed fitting the event expectations to the data. We use as nuissance parameters of the fit the rate systematics from each source of systematic separately (secondary vertex tag and mistag together). The uncertainty on the top pair cross section is considered as well: 7% from the last CDF top pair cross section combination, 7.5 +- 0.5 pb. We want to obtain the measurement of the BR of top into tau, neutrino and b-quark, which is obtained as unconstrained paramenter in the fit. The expectation of top pair decay channel in lepton plus tau is set to be linearly dependent on the BR, the expectation of the ditau channel is set to be quadratically dependent instead.

We obtain from the fit

,

We can variate the branching ratio of top into tau, neutrino and b-quark and refit the distribution. We obtain from the chi squared profile the limit of the branching ratio at 95% C.L., and we report:

,