Fake E T miss

Fake missing transverse energy is defined as the difference between reconstructed and trueE_T^miss. It can induce significant backgrounds from many different source:

• beam-gas scattering and other machine backgrounds;

• a displaced interaction vertex;

• hot, noisy, or dead calorimeter cells/regions;

• mis-measurements in the detector itself, mainly due to: high-pT particles that escape detec-tion outside the fiducial acceptance of the detector; undetected energy deposits in detector

(GeV)

calibration at EM scale

missT

calibration at EM scale

missT

E^miss_T global calibration E^missT global calibration+cryostat E

ATLAS

Figure 9.7: Linearity of the reconstructed E_T^miss, as a function of the average true E_T^miss for different physics processes, taken from Ref. [26]. In the left plot the results were obtained from: Z →τ τ for the data point with average true E_T^missof 20GeV ;W → eν andW → µν for the data point at35GeV ; semi-leptonictt¯decays for the point at68GeV ;A → τ τwithmA = 800GeV for the point at124GeV

; and finally SUSY decays with a typical mass scale of1TeV for the average trueE_T^missof280GeV. All results shown in the right plot were obtained from the processA→τ τwithmA= 800GeV.

cracks or inactive material; or the limited detector resolution (in particular energy fluctua-tions of high-pT jets).

A detailed discussion of the strategies to remove sources of fakeE_T^miss in early data and to mea-sure theE_T^missresolution and scale can be found in Ref. [26]. The main concepts considered for removing fakeE_T^miss are: requiring a minimum azimuthal angular separation between theE_T^miss vector and all jets in the event; employ track-jets (reconstructed from inner detector tracks) as a complementary measure to the standard calorimeter-based jets. In the latter approach, one can for example impose the track-jet momentum to match (in a certain window) the corresponding stan-dard jet energy. Another approach to clean fakeE_T^missis to mask certain detector regions which are known to malfunction, and or require theE_T^missvector not to point toward these regions.

9.5.3 Performance

The performance results shown here are only for the cell-based E_T^miss algorithm, which is the current default in ATLAS. The cell-basedE_T^miss algorithm together with the refined calibration has also been used for the SUSY studies of this thesis.

Fig.9.7shows theE_T^missresponse linearity as a function of the trueE_T^miss, as obtained by differ-ent physics processes. The linearity is defined as the difference of true to reconstructedE_T^miss, normalised to the trueE_T^miss.

The use of uncalibrated calorimeter cells, i.e. at the EM energy scale, leads to a bias of about 30%. As expected, this bias is significantly smaller forW →eνandW →µνevents (data point at35GeV) due to the reduced hadronic activity. TheE_T^missreconstructed from globally calibrated calorimeter cells, and including the muon correction term, shows a bias of about 5%. Further including the cryostat term reduces the bias to1%, except forE_T^miss from theW → eν process.

Finally, the use of the refined calibration together with the muon and cryostat corrections terms

9.5.3 Performance 133

0 200 400 600 800 1000 120014001600 18002000Σ

Resolution (GeV)

Figure 9.8: Resolution ofE_T^miss in the range low to medium (left) and low to to high (right) values of total transverse energy (P

ET), taken from Ref. [26]. Both variables E_T^miss and P

ET were obtained from the cell-based E_T^miss algorithm with the refined calibration. The solid lines correspond to the fits σ= 0.53pP

ET through the points fromZ →τ τ events (left) andσ= 0.57pP

ET through the points fromA →τ τ events (right). The points of theA→τ τ process were obtained withmAranging between 150GeV and800GeV. The QCD jets correspond to a sample with a560< pT <1120GeV range for the hard scattering.

leads to aE_T^misslinearity below1%, including theW →eνprocess.

The right plot of Fig.9.7shows the E_T^misslinearity as obtained alone from theA → τ τ process withm_A= 800GeV. One observes a negative bias for small trueE_T^missvalues. SinceE_T^missis by design a positive quantity, smallE_T^missvalues cause a negative bias.

TheE_T^missresolution is shown as a function of the total transverse energy (P

E_T) in Fig.9.8. The resolution is obtained from a Gaussian fit to the difference of reconstructed to trueE_T^miss, in each PE_T bin. Solid lines represent fits toσ=a·pP

E_T, which describes the observed stochastic behaviour of theE_T^missresolution. The fittedaparameter was found to be between0.53and0.57, for the different physics processes andP

E_T ranges. It is noteworthy that theE_T^missreconstructed from SUSY events shows a similar resolution.

The performance ofE_T^misswill be determined (and the simulations validated) with the first colli-sion data. Several studies have been proposed:

• Minimum bias events can be used to monitor and diagnoseE_T^miss reconstruction problems in the very beginning;

• W → eν andW → µν decays accompanied by jets can be used to test theE_T^miss recon-struction and determine theE_T^missscale in-situ in the20−150GeV range;

• Z →τ τ can be used in conjunction with theZmass constrain to determine theE^miss_T scale in-situ;

• Z →eeandZ →µµprocesses can be used to test forE_T^missbiases, and the resolution;

• Finally, semi-leptonict¯tdecays can be used to test theE_T^missreconstruction in a busy envi-ronment, which is relevant for SUSY studies.

Chapter 10

Search for Supersymmetry in the Inclusive One-Lepton Channel

The one-lepton inclusive channel is among the most prominent modes for SUSY searches with early LHC data. The one-lepton requirement reduces background from QCD jets, leavingt¯t(91%) andW+jets (8%) processes as the dominant and subdominant SM backgrounds, respectively.

This chapter is devoted to the detailed description of the one-lepton SUSY search channel. This analysis along with the simulated datasets (described in Chapter8) was part of the so-called com-puting system commissioning (CSC) programme [26].¹The author significantly contributed to the SUSY note of the CSC-book, namely to the one-lepton search mode, the scan and optimisation section, and to a lesser extent also to the zero-lepton channel.

Section10.1explains the objects and variable definitions. The one-lepton event selection cuts are discussed in Section10.2, where also the corresponding event flow is given for the SM background processes, and the SUSY benchmark points. The trigger efficiencies for the signal are described in Section10.3.

In Section 10.4 the statistical procedure to derive a signal significance, and systematic back-ground uncertainties are discussed. The estimated discovery potential is then presented for several mSUGRA models, including a parameter scan. The expected number of background events is obtained directly from MC simulation, and systematic uncertainties are assigned per process type (t¯t, W + jets, etc.). Finally, Section10.6 considers a multivariate analysis technique which is compared to the baseline one-lepton analysis.

1As mentioned in Chapter8: note that the results differ slightly from those in the SUSY CSC chapter “Prospects for Supersymmetry Discovery Based on Inclusive Searches”, because high statistics Alpgent¯tsamples are used here, whereas the CSC plots have been obtained with the MC@NLO top samples.

135

10.1 Object and Variable Definition

The signature of the inclusive one-lepton SUSY channel is based on one isolated electron or muon, multiple hard jets, and large missing transverse energy. The corresponding reconstruction algo-rithms are described in Chapter 9. This section details the employed reconstruction parameters, the quality and identification cuts, as well as the isolation requirements imposed on the various objects.

Electrons are “only” required to satisfy the medium purity cuts since the background from the production of QCD multijets is expected to be already significantly reduced by the event selection of four hard jets and largeE_T^miss, as described later. A strong jet rejection is therefore not needed in the electron identification cuts, leading to a considerable gain in signal efficiency.

Muons are obtained from combined inner detector tracks and muon spectrometer tracks. Tagged muons are discarded since they improve the muon performance only in the low-pT region.

The expected high jet multiplicity in typical SUSY decays favours the use of a jet algorithm with a narrow cone size. The tower-cone jet algorithm was chosen, despite its known shortcomings (infrared and collinear unsafe), to conform to the current ATLAS default.

The objects used are fully defined as follows:

• Electrons are reconstructed with the default e/γ algorithm. They must satisfy the set of medium cuts (cf. Section9.2), and the quality cuts: |η|< 2.5andp_T >10GeV. Further-more, the calorimeter-based isolation requirement is imposed: less than10GeV of energy in a cone of size∆R= 0.2around the electron (excluding the electron energy).

• Muonsare reconstructed using the default Staco algorithm. They are required to be com-bined muons with a matchingχ²below100. If more than one inner detector track matched the stand-alone muon track, then only the one with the smallest distance in ∆R is kept (best-match flag). The same quality cuts and isolation criteria as for electrons are applied:

|η|<2.5,p_T >10GeV, andE_T^cone<10GeV with cone size∆R= 0.2.

• Jetsare reconstructed as cone-tower jets with a cone size ofR = 0.4, and are required to satisfy:|η|<2.5, andp_T >20GeV. They are calibrated using the H1-style method.

• Missing transverse energy(E_T^miss) is reconstructed using the default calorimeter-based al-gorithm with the refined calibration and including the cryostat and muon terms.

• Tausandphotonsare not considered separately, but treated as jets.

In the remainder of this analysis the word “lepton” is used to denote isolated electrons and muons.