22 pages

Measurement of the top quark pair cross section with ATLAS in pp collisions at sqrt(s) = 7 TeV using final states with an electron or a muon and a hadronically decaying tau lepton

Please download to get full document.

View again

of 22
All materials on our website are shared by users. If you have any questions about copyright issues, please report us to resolve them. We are always happy to assist you.
Measurement of the top quark pair cross section with ATLAS in pp collisions at sqrt(s) = 7 TeV using final states with an electron or a muon and a hadronically decaying tau lepton
    a  r   X   i  v  :   1   2   0   5 .   2   0   6   7  v   1   [   h  e  p  -  e  x   ]   9   M  a  y   2   0   1   2 EUROPEANORGANISATIONFORNUCLEARRESEARCH(CERN) CERN-PH-EP-2012-102 Submittedto:PhysicsLettersB Measurement of the top quark pair production cross sectionwith ATLAS in  pp collisions at √   s  =  7  TeV using final states withan electron or a muon and a hadronically decaying  τ  lepton TheATLASCollaboration Abstract Ameasurementofthecrosssectionoftopquarkpairproductioninproton − protoncollisionsrecordedwiththeATLASdetectorattheLHCatacentre-of-massenergyof7TeVisreported.Thedatasampleusedcorrespondstoanintegratedluminosityof2.05  fb − 1 .Eventswithanisolatedelec-tronormuonanda  τ leptondecayinghadronicallyareused.Inaddition,alargemissingtransversemomentumandtwoormoreenergeticjetsarerequired.Atleastoneofthejetsmustbeidentifiedasoriginatingfroma  b quark.Themeasuredcrosssection,  σ tt   =  186 ± 13 (stat.) ± 20 (syst.) ± 7( lumi. ) pb ,isingoodagreementwiththeStandardModelprediction.  Measurement of the top quark pair cross section with ATLAS in  pp  collisionsat  √  s  =  7 TeV using final states with an electron or a muonand a hadronically decaying  τ  lepton The ATLAS Collaboration Abstract A measurement of the cross section of top quark pair production in proton − proton collisions recorded with the ATLAS detectorat the LHC at a centre-of-mass energy of 7 TeV is reported. The data sample used corresponds to an integrated luminosity of 2.05 fb − 1 . Events with an isolated electron or muon and a  τ  lepton decaying hadronically are used. In addition, a large missingtransverse momentum and two or more energetic jets are required. At least one of the jets must be identified as originating from a  b quark. The measured cross section,  σ tt   =  186 ± 13 (stat.) ± 20 (syst.) ± 7(lumi.) pb, is in good agreement with the Standard Modelprediction. Keywords:  top-quark physics, cross section, lepton + τ 1. Introduction Measuring the top quark pair ( tt  ) production cross section( σ tt  ) in di ff  erent decay channels is of interest because it canopen a window to physics beyond the Standard Model (SM).In the SM, the top quark decays with a branching ratio close to100% into a  W   boson and a  b  quark, and  tt   pairs are identifiedby either the hadronic or leptonic decays of the  W   bosons andthe presence of additional jets. ATLAS has previously used thesingle-lepton channel [1], and the dilepton channels includingonlyelectronsandmuons[2]to performcross-sectionmeasure-ments.The large cross section for  tt   productionat the LHC providesan opportunity to measure  σ tt   using final states with an elec-tron or a muon and a  τ  lepton with high precision. The  σ tt   inthis channel has been measured at the Tevatron with 25% pre-cision [6] and recently by the CMS Collaboration at the LHCwith 18% precision [7]. A deviation from  σ tt   measured in otherchannels would be an indicationof non-StandardModel decaysof the top quark, such as a decay to a charged Higgs (  H  + ) anda  b  quark with  H  + decaying to a  τ  lepton and a  τ  neutrino,or contributions from non-Standard Model processes [3, 4, 5]. ATLAS has set upper limits on the branching ratio of top quarkdecays to an  H  + bosons decaying to a  τ  lepton and a neutrino[8].This analysis uses 2.05 fb − 1 of data collected by ATLAS attheLHC from  pp collisionsat acentre-of-massenergyof7TeVbetween March and August 2011. After application of kine-matic selection criteria that require one top quark to decay via W   →  ℓν (where ℓ   iseitheramuonoranelectron)andidentifica-tion of a jet as srcinating from a  b  quark ( b -tag), the dominantbackground to the  tt   →  ℓ   +  τ  +  X   channels with the  τ  leptondecaying hadronically is the  tt   →  ℓ   +  jets channel in which a jet is misidentified as a hadronic  τ  lepton decay. Therefore,  τ lepton identification ( τ  ID) is critical for separating signal andbackground. The  τ  ID methodology employed in this analysisexploits a multivariate technique to build a discriminant [9].A boosted decision tree (BDT) algorithm is used [10]. Thenumber of   τ  leptons in a sample is extracted by fitting the dis-tributions of BDT outputs to background and signal templates.The results are checked using an alternative method, referred toas the “matrix method”, based on a cut on the BDT output. 2. ATLAS Detector The ATLAS detector [11] at the LHC covers nearly the en-tire solid angle around the collision point. 1 It consists of aninner tracking detector surrounded by a thin superconductingsolenoid, electromagnetic (EM) and hadronic calorimeters, andan external muon spectrometer incorporating three large super-conducting toroid magnet assemblies. The inner tracking de-tector provides tracking information in a pseudorapidity range | η |  <  2 . 5. The liquid-argon (LAr) EM sampling calorimeterscover a range of   | η |  <  3 . 2 with fine granularity. An iron–scintillator tile calorimeter provides hadronic energy measure-ments in the central rapidity range ( | η |  <  1 . 7). The endcapandforwardregionsare instrumentedwith LAr calorimetersforbothEM andhadronicenergymeasurementscovering | η |  <  4 . 9.The muon spectrometer provides precise tracking informationin a range of  | η |  <  2 . 7. 1 Atlas uses a right-handed coordinate system with its srcin at the nominalinteraction point in the centre of the detector and the z-axis along the beampipe. The x-axis points to the centre of the LHC ring, and the y-axis pointsupwards. The azimuthal angle  φ  is measured around the beam axis and thepolar angle  θ   is the angle from the beam axis. The pseudorapidity is definedas  η  =  − ln[tan( θ/ 2)]. The distance  ∆  R  in  η  −  φ  space is defined as  ∆  R  =   ( ∆ φ ) 2 + ( ∆ η ) 2 . Preprint submitted to Elsevier May 10, 2012  ATLASusesa three-leveltriggersystemtoselect events. Thelevel-1 trigger is implementedin hardwareusing a subset of de-tector information to reduce the event rate below 75 kHz. Thisis followed by two software based-trigger levels, level-2 andthe event filter, which together reduce the event rate to about300 Hz recorded for analysis. 3. Simulated Event Samples Monte Carlo (MC) simulation samples are used to optimiseselection procedures, to calculate the signal acceptance and toevaluate contributions from some background processes.For the  tt   and single top-quark final states, the next-to-leading-order (NLO) generator MC@NLO [12] is used with atop-quarkmassof172.5GeVandwiththeNLOpartondistribu-tion function (PDF) set CTEQ6.6 [13]. The “diagram removalscheme” is used to remove overlaps between the single top-quark and the  tt   final states. The  tt   cross section is normalisedto the prediction of HATHOR (164 + 11 − 16  pb) [14], which employsan approximate next-to-next-to-leading-order (NNLO) pertur-bative QCD calculation.For the background channels, MC samples of   W  /  Z  , singletop-quark events and diboson  WW  ,  WZ  , and  ZZ   events (all inassociation with jets) are used.  W  +  jets events and  Z  /γ  ∗ +  jetsevents (with dilepton invariant mass  m ℓ  + ℓ  −  >  40 GeV) are gen-erated by the ALPGEN generator [15] with up to five outgoingpartons from the hard scattering process, in addition to the vec-tor bosons. 2 The MLM matching scheme of the ALPGENgenerator is used to remove overlaps between matrix-elementand parton-shower products. Parton evolution and hadronisa-tion is handled by HERWIG [16], as is the generation of dibo-son events. The leading-order PDF set CTEQ6L is used for allbackgrounds described above.All samples that use HERWIG for parton shower evolutionand hadronisationrely on JIMMY [17] for the underlyingeventmodel. The  τ -lepton decays are handled by TAUOLA [18].Thee ff  ectofmultiple  pp interactionsperbunchcrossing(“pile-up”) is modelled by overlayingsimulated minimum bias eventsover the srcinalhard-scatteringevent [19]. MC events are thenreweighted so that the distribution of interactions per crossingin the MC simulation matches that observed in data. The aver-age number of pile-up events in the sample is 6.3. After eventgeneration, all samples are processed with the GEANT4 [20]simulation of the ATLAS detector, the trigger simulation andare then subject to the same reconstruction algorithms as thedata [21]. 4. Data and Event Selection The eventselectionuses the same objectdefinitionas in the tt  cross-section measurement in the dilepton channel [2] with theexception of a  τ  candidate instead of a second electron or muoncandidate and some minor adjustments. The electrons must be 2 The fraction of events with  m ℓ  + ℓ  −  <  40 GeV is estimated to be less than0.2% of the total after all selections. isolated and have  E  T  >  25 GeV and  | η cluster |  <  2 . 47, exclud-ing the barrel-endcap transition region (1 . 37  <  | η cluster |  <  1 . 52),where  E  T  is thetransverseenergyand η cluster  is the pseudorapid-ity of the calorimeter energy cluster associated with the candi-date.Theelectronis definedas isolatedifthe  E  T  depositedinthecalorimeterandnot associated with the electronin a cone in  η - φ space of radius  ∆  R  =  0 . 2 is less than 4 GeV. The muons mustalso be isolated and have  p T  >  20 GeV and  | η |  <  2 . 5. For iso-latedmuons,boththecorresponding  E  T  andtheanalogoustrackisolation transverse momentum (  p T ) must be less than 4 GeVin a cone of   ∆  R  =  0 . 3. The track isolation  p T  is calculatedfrom the sum of the track transverse momenta for tracks with  p T  >  1 GeV around the muon. Jets are reconstructed with theanti- k  t   algorithm [22] with a radius parameter  R  =  0 . 4, startingfrom energydeposits (clusters) in the calorimeter reconstructedusing the scale established for electromagnetic objects. These jets are then calibrated to the hadronic energy scale using  p T -and  η -dependent correction factors obtained from simulation[25]. The jet candidates are required to have  p T  >  25 GeVand  | η |  <  2 . 5. Jets identified as srcinating from a  b  quark ( b -tag) by a vertex tagging algorithm are those that pass a decaylength significance cut corresponding to an e ffi ciency of 70%for  b -quark jets from  tt   events and a 1% e ffi ciency for light-quark and gluon jets [2, 26]. The missing transverse momentum is constructed from thevector sum of all calorimeter cells with  | η |  <  4 . 5, projectedonto the transverse plane. Its magnitude is denoted  E  missT  . Thehadronicenergyscale is usedfortheenergiesofcells associatedwith jets;  τ  candidates are treated as jets. Contributions fromcells associated with electrons employ the electromagnetic en-ergy calibration. Contributions from the  p T  of muon tracks areincluded, removing the contributions of any calorimeter cellsassociated with the muon. 4.1.  τ  Reconstruction and Identification The reconstruction and identification of hadronically decay-ing  τ  leptons proceeds as follows:1. the  τ  candidate reconstruction starts by considering each jet as a  τ  candidate;2. energy clusters in the calorimeter associated with the  τ candidate are used to calculate kinematic quantities (suchas  E  T ) and the associated tracks are found;3. identification variables are calculated from the trackingand calorimeter information;4. these variables are combined into multivariate discrimi-nants and the outputs of the discriminants are used to sep-arate jets and electrons misidentified as  τ  leptons decayinghadronically from  τ  leptons.Details are given in Ref. 9. In this analysis the outputs of BDTdiscriminants are used.Reconstructed  τ  candidates are required to have 20 GeV  <  E  T  <  100 GeV. They must also have  | η |  <  2 . 3, and one, twoor three associated tracks. A track is associated with the  τ  can-didate if it has  p T  >  1 GeV and is inside a cone of   ∆  R  <  0 . 4around the jet axis. The associated track with highest  p T  must2  have  p T  >  4 GeV. Thechargeis givenbythesumofthechargesof the associated tracks, and is required to be non-zero. Theprobability of misidentifying the  τ  lepton charge sign is about1%. The charge misidentification rate for muons and electronsis negligible.If the  τ  candidate overlaps with a muon (  p T  >  4 GeV, no iso-lation required) or an electron candidate within  ∆  R ( ℓ,τ )  <  0 . 4,the  τ  candidate is removed. To remove electrons misidentifiedas  τ  leptons, an additional criterion is used that relies on a BDTtrained to separate  τ  leptons and electrons (BDT e ) using sevenvariables shown to be well modelled by comparing  Z   →  e + e − and  Z   →  τ + τ −  events in data and in MC simulation. The vari-ables were chosen after ranking a large set by their e ff  ective-ness. 3 The most e ff  ective variables for BDT e  are  E  /  p , theEM fraction (the ratio of the  τ  candidate energy measured inthe EM calorimeter to the total  τ  candidate energy measured inthe calorimeter), and the cluster-based shower width. The BDToutput tends to be near 1 (0) if the  τ  candidate is a  τ  lepton(electron). The  τ  candidate is required to satisfy BDT e  >  0 . 51;85% of reconstructed  τ  leptons decaying hadronically satisfythat requirement as measured in  Z   →  τ + τ −  events. The addi-tional rejection for electrons is a factor of 60.The majority of objects reconstructed as  τ  candidates in amulti-jet environment are jets misidentified as  τ  leptons (fake τ ). Another BDT (BDT  j ) based on eight variables is usedto separate  τ  leptons in  τ  candidates with one track (denoted τ 1 ) from such jets. For candidates with more than one track(denoted  τ 3 ) BDT  j  includes ten variables. The most e ff  ectivevariables for BDT  j  are calorimeter and track isolation, cluster-based jet mass, and the fraction of energy within  ∆  R  =  0 . 1 of the jet axis. The BDT  j  distributions are fit with templates forbackground and signal to extract the number of   τ  leptons inthe sample. Details are given in Section 6. The fake  τ  back-ground in the  τ 3  sample is significantly higher than in the  τ 1 sample, leading to very di ff  erent BDT  j  distributions. Hence in-dependentmeasurementsare carriedoutfor  τ 1  and τ 3  candidateevents and the results are combined at the end. If there is a  τ 1 and a  τ 3  candidate in the event, the  τ 1  candidate is kept as theprobability that the  τ 1  is a  τ  lepton is much higher. If there aretwo  τ 1  or  τ 3  candidates, both are kept. 4.2. Event Selection For this analysis, events are selected using a single-muontrigger with a  p T  threshold of 18 GeV or a single-electron trig-ger with a  p T  threshold of 20 GeV, rising to 22 GeV duringperiods of high instantaneous luminosity. The o ffl ine require-ments are based on data quality criteria and optimised usingMonte Carlo simulation: •  a primary vertex with at least five tracks, each with  p T  > 400 MeV, associated with it; 3 The e ff  ectiveness is quantified by quadratically summing over the changein the purity between the mother and daughter leaves for every node in whichthe given variable is used in a decision tree. •  one and only one isolated high-  p T  muon and no identifiedelectronsforthe  µ + τ  channel,oroneandonlyoneisolatedelectron and no isolated muons for the  e + τ  channel; •  at least one  τ  candidate (as defined in Section 4.1); •  at least two jets not overlapping with a  τ  candidate, i.e. ∆  R ( τ,  jet)  >  0 . 4; •  E  missT  >  30 GeV to reduce the multi-jet background, andthe scalar sum of the  p T  of the leptons (including  τ ), jets,and  E  missT  must be greater then 200 GeV, to reduce the W  +  jets background.The  ℓ   +  τ  samples are divided into events with no jets iden-tified as a  b -quark jet (0  b -tag control sample) and those withat least one such jet ( ≥  1  b -tag  tt   sample). The 0  b -tag sampleis used to estimate the background in the  ≥  1  b -tag  tt   sample.Each sample is split into two, one with the  τ  candidate and  ℓ  having the opposite sign charge (OS), and the other one with  τ and  ℓ   having the same sign charge (SS). While the  τ  candidatesin the SS samples are almost all fake  τ  leptons, the OS sampleshave a mixture of   τ  leptons and fake  τ  leptons. The numbers of observed and expected events in the above samples are shownin Table 1. All processes contribute more events to OS than SSbecause of the correlation between a leading-quark charge andthe lepton charge, except for the multi-jet channel contributionwhich has equal number of OS and SS events within the uncer-tainties. The  ℓ  +  jets entry includes the contribution from eventswith  τ  leptons when the  τ  candidate is actually a fake  τ . The τ  entries require the reconstructed  τ  candidate be matched to agenerated τ  lepton. Thematchingcriterionis ∆  R  <  0 . 1betweenthe  τ  candidate and the observable component of the generated τ  lepton.To estimate the multi-jet background from data, an event se-lection identical to the  µ + τ  ( e + τ ) event selection except for aninverted muon (electron) isolation cut is used to obtain a multi- jet template for the shape of the transverse mass,  m T . 4 Thenormalizationofeachselecteddatasampleisobtainedbyfittingthe  m T  distribution of the selected data samples with the multi- jet template and the sum of non-multi-jet processes predictedby MC, allowing the amount of both to float. The uncertaintyon the multi-jet background is estimated to be 30%. However,because of the subtraction method discussed in Section 5, themulti-jet background plays no role in the cross-section mea-surement. Thereare small di ff  erencesbetween the total numberof events predicted and observed which motivate using data asmuch as possible to estimate the background.As one can see from Table 1, the  τ  leptons are almost allin the OS sample and come mainly from two sources:  Z   → τ + τ − , which is the dominant source in the sample with 0  b -tag, and  tt   →  ℓ   +  τ  +  X   which is the dominant source in thesample  ≥  1  b -tag. The sources of fake  τ  leptons are also quitedistinct between the 0  b -tag and the  ≥  1  b -tag samples: thefirst is mainly  W  /  Z  +  jets with small contributions from otherchannels, the second is mainly  tt  . 4 m T  =   (  E  ℓ  T   +  E  missT  ) 2 − (  p ℓ   x  +  E  miss  x  ) 2 − (  p ℓ   y  +  E  miss  y  ) 2 . 3
Related Documents
View more...
We Need Your Support
Thank you for visiting our website and your interest in our free products and services. We are nonprofit website to share and download documents. To the running of this website, we need your help to support us.

Thanks to everyone for your continued support.

No, Thanks

We need your sign to support Project to invent "SMART AND CONTROLLABLE REFLECTIVE BALLOONS" to cover the Sun and Save Our Earth.

More details...

Sign Now!

We are very appreciated for your Prompt Action!