We have analyzed electroweak single top production using the full CDF Run II data set collected up to the end of the Tevatron run in September 2011, corresponding to a total integrated luminosity of 9.1 fb ^-1. The cross section being so small and the processes that mimic the signal so large, CDF and D0 have not been able to measure the cross section of this interesting process with full-prove significance. In order to increase the scanty statistics, we look here at events rejected by previous analyses, i.e. events where there are no identified leptons, or where &tau s are reconstructed as a jet. We thus rely solely on the signature of high PT jets and large missing transverse energy. Being statistically independent of the lepton+jets sample, this sample provides, albeit with low precision, an independent measurement of the single top production cross section. Eventually, the result of this measurement can be combined with the lepton+jets one and help reaching a solid single top cross section measurement at the Tevatron.

Many Standard Model processes can produce a final state with large missing tranverse energy and jets, such as single top (our signal), top pair production, W/Z + jets, diboson production. In addition, QCD multijet production can mimic this signature due to severely mismeasured jets which appear to have large MET. Since the QCD heavy flavour production cross-section is orders of magnitude higher than that of the signal, it constitutes the biggest background in this search. Additionally light flavour jets can be falsely identified as b-jets (commonly referred to as "mistags").

Having no identified leptons in the final state, backgrounds are many orders of magnitude larger than the signal even after requiring large missing transverse energy and b-tagged jets in the final state. It is thus necessary to develop an event selection which reduces backgrounds to a more manageable size before trying to build a discriminant to measure the single top cross section. The QCD multijet production is the only background where the MET is mainly instrumental. In a first step we study the kinematics of these events and implement a multivariate technique to cut them out as much as possible.

We develop in a second step a discriminant to act against tt background. Top pair production gives a large contribution to total background, especially when both the leading jets in the final states are required to be b-tagged. In addition, the produc- tion of real top quarks makes this background more signal like if we take into account variables connected to the top mass, that are instead particularly useful in separating single top from all other non-top backgrounds.

In the final step, we use again a machine learning technique to discriminate the signal from the surviving backgrounds, and scan its output distribution to measure the single top production cross section in the MET+jets final state.

lepton veto = use loose identification cuts to reject events with isolated leptons

MET > 50 GeV

Number of jets = 2 or 3 and one of the leading jets (j1 or j2) central (|&eta| < 0.9).

Events with a larger number of jets are rejected

&Delta R(j1 , j2 ) > 1

ET (j1 ) > 35 GeV, ET (j2 ) > 25 GeV

Events are b-tagged either 1S, SS, or SJ

To model the single top dynamics and predict both shape and rate, we use a sample generated with Powheg. The official sample is actually composed of three subsamples, one s-channel sample, a LO and a NLO t-channel sample. This process is characterized by an accurately predicted cross section, used to derive its normalization.

We use the LO Pythia cross-sections scaled by a k-factor corresponding to the ratio between the NLO and LO cross-section prediction in MCFM3 to predict both shape and rate of diboson production. The LO MCFM predictions are also consistent with those from Pythia and are reported to be in good agreement with the data. To estimate the contribution of mistagged light flavor diboson events in the tag sample we apply the corresponding mistag matrix to diboson light flavor MC samples, vetoing events with a real b- or c-quark from HEPG. This process is characterized by an accurately predicted cross section, used to derive its normalization.

Top pair-production yields a significant contribution to the background in the pre- selection region, especially in the double-tag sample. Semi-leptonic top decays are energetic, bear large MET and high jet multiplicity. We use Monte Carlo samples generated for the Top Group to model the shape of this background. The top-antitop events were generated with Pythia with Mtop = 172.5 GeV and normalized with the cross section measured by CDF: 7.71+-0.51 pb.

V+jets background was generated with Alpgen+Pythia. As for diboson production, the contribution of mistagged light flavor V+jets events was determined by applying the mistag rate matrix to W/Z+lf samples vetoing events with a real b- or c-quark from HEPG. Since the imprecise theoretical prediction for the cross section of this process, we derive its normalization from data.

Due to the largecross-sections, it is practically impossible to generate enough statistics to simulate all QCD processes. To deal with this problem a method for estimating QCD background from data was developed. This technique allows us to estimate not only heavy flavor QCD production, but also processes with a light flavor jet falsely tagged as a b-quark. Additionaly it allows us to model Single SecVTX tagged data sample, which adds additional sensitivity to the analysis. We derive QCD multijet normalization from data.

Estimation of QCD and V+jets content

We fitted QCDNN distribution in order to estimate the normalization of both QCD and V+jets backgrounds. We choose this variable to perform the fit because the QCD background is well separated from the other electroweak processes and its normalization can be derived with a good precision. The fit is performed by using a binned likelihood fit (TFractionFitter function in ROOT).

We train an artificial neural network (NN), a multilayer perceptron (MLP) fed with 13 kinematic variables, to separate the signal from the main background: QCD multijet production. We cut on this output at QCDNN > 0.35, to form the Pre-selection 2 region.

Input Variables

MPT

DPhiMET.MPT

MinDPhiMET.Ji

J12HRF

Sphericity

QCDNN output

QCDNN.singletop

We train an artificial neural network (NN), a multilayer perceptron (MLP) fed with 12 kinematic variables, to separate the signal from top pair production background. We cut on this output at TTNN > 0.3, to form the Signal region.

Input Variables

SUMJETPT

MVJ12

MJJJ

J1Et

J2Et

TTNN output

TTNN.singletop

11 variables are fed to a MLP trained with events with at least one tight tag (SecVTX)to obtain the final NN discriminant. We then use a likelihood profile of this discriminant to measure the production cross section of single top events.

MVJ2

MVJ123

HT

SIGNN output

SIGNN3.singletop

SIGNN3.singletop

Systematic uncertainties are split in normalization uncertainty and shape uncertainty. The normalization uncertainty relfects changes to the event yield due to the systematic effect while the shape uncertainty reflect changes to the template histograms. Both of these effects can be included, depending on the source the systematic uncertainty.

The table below summarizes the systematic uncertainties and their effects.

We apply our analysis to 9.1 fb ^-1 of CDF Run II data and measure the single top production cross-section.

Cross-section measurement

The result of the binned maximum likelihood fit is shown below. All sources of systematic uncertainties (normalization and shape) are included.

Posterior probability density