Measuring the Top Pair Background to the Higgs Search in the MET+b-jets Channel

Daniela Bortoletto, Qiuguang Liu, Fabrizio Margaroli and Karolos Potamianos (Purdue University)  [Contact]
public note

Abstract

We present the first measurement of the top pair production cross section in events with large missing transverse energy, two or three high-p_T jets, where at least one is identified as a b-jet. A veto on loosely identified electrons and muons is applied. We reduce the dominant QCD multijet background using neural network techniques, and then use another neural network to isolate the top pair signal from the remaining backgrounds. Analyzing 5.7 fb^-1 of data, we measure top pair production cross section with σ_{top pair} = 7.12^+1.20_-1.12,(stat.+syst.) pb where we assume mass_top of 172.5 GeV.

Introduction

Top quark has been discovered in 1995 at Tevatron by CDF and D0 experiments. The top pair production cross section has been measured in lepton+jets, dilepton and all-hardronic decay modes. We measure the top pair production cross section in events with large missing tranverse energy (MET) and 2 or 3 high-p_T jets, where at least one is identified as a b-jet. A veto on loosely identified electrons and muons is applied. This is the first measurement for top pair production with this signature, in which ~66% events decay in lepton+jets mode, and the other 34% decays in dilepton mode. As a complementary to existing measurements, it can be combined with those to achieve greater precision and test more stringently QCD NLO predictions.

MET+b-jets channel is also a very intereting channel to searches for new physics. For example, it is one of the most sensitive channel for low mass Higgs search at Tevatron [1], as well as other searches like SUSY/leptoquark analyses. And top pair is a significant background in these physics searches. In this analysis, we employed many state-of-the-art techniques, and followed the strategies which have been used in low mass Higgs search analysis.

To improve the signal-to-background ratio, we select jets identified as originating from b quarks using b-tagging algorithms. Even after these requirements, the ratio is still too low to achive sensitivity to top pair production. We further exploit the kinematic and topological characteristics of top pair events using neural networks to isolate the signal from dominant QCD background and subsequently from the remaining backgrounds.

Event Selection

Without identified lepton information, the largest background to MET+jets analyses is the QCD multijet background. QCD multijet has very high production rate at a hadron collider. Although these processes generally do not produce neutrinos, mismeasured jet energy do result in a significant imbalance of transverse energy. Furthermore, QCD b quark production yields neutrinos whenever one b-hadron decays semi-leptonically, thus giving additional missing transverse energy. Due to the mismeasurement, most such events will have MET align with their 2^nd or 3^rd jet. Thus we are using these topological characteristics (DPhi(MET,j_2,3)>0.4 and DPhi(MET,j₁)>1.5) to reduce a large part of QCD multijet at the first stage.

The b-tagging algorithms we are using in this analysis are SecVTX [2] and JetProb [3]. SecVTX is a b-tagging algorithm based on secondary vertex reconstruction, and JetProb however tags a jet depending on the probability that all tracks associated with a jet come from the primary interaction vertex.

We accept events with exactly one SecVTX-tagged jet (1S), two SecVTX-tagged jets (SS), and one SecVTX-tagged and one JetProb-tagged jet (SJ).

After preselection, the multijet production is still the dominant background. We thus use a neural network approach to further reduce the QCD multijet. Neural network technique exploits the correlations among the many observables which provide discrimination between signal and backgrounds. In this analysis, we use 15 variables as the QCD Neural Network inputs, which inherit from the previous Single Top analysis [4] . The distribution for these 15 variables after preselection is shown as follow,

The NN output is shown as follow,

We choose QCDNN > -0.5 to define the signal region.

Final Discriminant

Though a good signal/background ratio has been achieved by using the QCDNN, we are still interested to develop a final discriminant to separate signal bins and background bins as much as possible, since we know backgrounds like W/Z+jets have very uncertainties. These uncertainties can worse the measurement, especially when we are using a binned likelihood technique to measure its cross section.

We use Neural Network again to develop the final discriminant. Here, we have 5 NN input variables,

The final discriminant is shown as follow,

Systematic Uncertainties

Systematic uncertainties are split in normalization uncertainties and shape uncertainties. The normalization uncertainty reflects 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 normalization uncertainties are summarized in the following tables. We also assign a systematic shape uncertainty to the multijet model due to possible signal contamination.

Cross Section Measurement and Results

We use a binned likelihood technique to measure the cross section. The likelihood function L is given by the product of the likelihood for each of the different sub-tagging categories L_c, where L_c is definded as,

where n_i is the data count in that particular bin and n_bins is the number of bins in the distribution which is scanned to look for an excess of signal-like events. The prediction in each bin is a sum over signal and background contributions [4]:

In this analysis, using 5.7fb^-1 data collected by CDF at Tevatron, we measure σ_{top pair} = 7.12^+1.20_-1.12,(stat.+syst.) pb where we assume mass_top of 172.5 GeV.

References:
[1], Search for the Higgs Boson Using Neural Networks in Events with Missing Energy and b-Quark Jets in pp collisions at √s=1.96 TeV, Phys. Rev. Lett. 104, 141801 (2010)
[2], Measurement of the tt production cross section in pp collisions at √s=1.96 TeV using lepton + jets events with secondary vertex b-tagging, Phys. Rev. D 71, 052003 (2005)
[3], Measurement of the tt production cross section in pp collisions at √s=1.96 TeV using lepton + jets events with jet probability b-tagging, Phys. Rev. D 74, 072006 (2006))
[4], Search for single top quark production in pp collisions at √s=1.96 TeV in the missing transverse energy plus jets topology, Phys. Rev. D 81, 072003 (2010)