1 Study of Top-Antitop Production with the ATLAS Detector at the LHC DPG Frühjahrstagung, A. Bangert, T. Barillari, S. Bethke, N. Ghodbane, T. Göttfert, R. Härtel, S. Kluth, S. Menke, R. Nisius, S. Pataraia, E. Rauter, P. Schacht, J. Schieck

2 The Large Hadron Collider E CM = 14 TeV L = cm -2 s -1

3 Experiments at the LHC

4 The ATLAS Detector Lead / liquid argon electromagnetic sampling calorimeter. Electron, photon identification and measurements. Hadronic calorimeter. Scintillator-tile barrel calorimeter. Copper / liquid argon hadronic end-cap calorimeter. Tungsten / liquid argon forward calorimeter. Measurements of jet properties. Air-core toroid magnet Instrumented with muon chambers. Muon spectrometer. Measurement of muon momentum.. Inner Detector surrounded by superconducting solenoid magnet. Pixel detector, semiconductor tracker, transition radiation tracker. Momentum and vertex measurements; electron, tau and heavy-flavor identification.

5 Why Study the Top Quark? Measurement of top-antitop production cross section provides a test of QCD. Accurate measurement of m t is valuable as input to precision electroweak analyses. A precision measurement of m t and m W can be used to constrain the Higgs mass. Top pair production will provide a significant background to future measurements. It is important to measure the production rate precisely. LEP, Tevatron, SLD: m W = ± Tevatron: m t = ± 2.1 GeV

6 Top Quark Physics Γ(t → Wb) ~ 100% Hadronic channel: tt → Wb Wb → jjb jjb Γ = 44.4% Semileptonic channel: tt → Wb Wb → lvb jjb Γ = 50.7% Dileptonic channel: tt → Wb Wb → lvb lvb Γ = 4.9% LHC: 10% quark-antiquark annihilation 90% gluon fusion

7 Samples Semileptonic ttbar events are signal. Dileptonic ttbar events are background. csc , σ=461 pb, L=1072 pb -1 NLO event generator / Herwig Detector simulation performed using Athena Hadronic ttbar events are background. csc , σ=369 pb, L=105 pb -1 NLO event generator / Herwig Detector simulation performed using Athena Inclusive W+N jets events are main physics background. rome , σ=1200 pb, L=136 pb -1 LO event generator Alpgen Generated using Athena Reconstructed using Athena

8 Selection Cuts Each event was required to have: Exactly one isolated, high-p T electron or muon. For e “isolated” meant E ∆R=0.45 < 6 GeV For μ “isolated” meant E ∆R=0.20 < 1 GeV p T (l) > 20 GeV, |η| < 2.5 At least four jets. Jets are reconstructed using k T algorithm with D = 0.4. |η| < 2.5 p T (j 1 ) > 40 GeV, p T (j 2 ) > 40 GeV p T (j 3 ) > 20 GeV, p T (j 4 ) > 20GeV Missing E T > 20 GeV No b-tagging was required. Considered events in the semileptonic ttbar channel where the lepton is an electron or a muon. Γ = 29.9% j ν μ W W b b t t g g p p j

9 Event Reconstruction For each event: Consider all jets with p T > 20 GeV. Create all possible 3-jet combinations. Select that 3-jet combination with maximum p T to represent the reconstructed hadronic top quark. Consider the 3 jets used to reconstruct top quark. Form the 3 possible 2-jet combinations. Take the 2-jet combination with the maximal p T to represent the reconstructed hadronic W boson.

10 Hadronic Top and W Boson Mass Mass of Hadronic Top m t fit = ± 1.4 GeV m t Tevatron = ± 2.1 GeV m t MC = GeV Mass of W Boson m W fit = 79.6 ± 0.5 GeV m W LEP = ± 0.03 GeV

11 Optimization of k T Jet Algorithm A successive combination algorithm such as k T recursively combines objects (particles, calorimeter cells, towers, clusters) with “nearby” momenta into jets. “Nearby” is defined in η- φ space: ∆R = √((∆η) 2 + (∆ φ) 2 ) Two objects will be merged into one jet if ∆R < D 2. The parameter D must be chosen properly for the process in question. If D is too small, objects which should be combined into a jet may be excluded. If D is too large, objects which should not be merged into a single jet may be merged. D = 0.4D = 0.7D = 0.9D = 1.0D = 0.1 Invariant Mass Distribution of the W Boson Jet reconstruction algorithm was run on Monte Carlo truth after hadronization. No detector simulation was performed.

12 Optimization of kT Jet Algorithm Jet reconstruction algorithm was run on Monte Carlo truth after hadronization. No detector simulation was performed. D = 0.4 is “reasonable”. W Mass as function of D ParameterTop Mass as function of D Parameter

13 Used 10% of ttbar signal sample as “data”. Used remaining 90% of sample as Monte Carlo. L MC = 971 pb -1 L data = 106 pb -1 Only ttbar background was considered. Consistency check: ε data = ± ε MC = ± For Monte Carlo: ε e = %, ε μ = % ε τ = 4.86 %, ε ll = % Monte Carlo Cross Section Studies

14 Monte Carlo Cross Section Studies σ data = N data / L data ε e MC For semileptonic ttbar events with electron or muon: σ data = 250 ± 3 (stat) pb Consistency check: Known from Monte Carlo: σ MC = 248 pb L data = 100 pb -1 gives statistical precision of 1%. Statistics will not be an issue in this channel at the LHC.

15 Summary Studies were performed in semileptonic channel with electron or muon. Jet reconstruction was performed using k T algorithm with D = 0.4. Efficiencies: For semileptonic signal: ε e = 27.4 %, ε μ = 25.3 % For semileptonic background with tau lepton: ε τ = 4.4 %, For dileptonic and hadronic ttbar background: ε ll = 13.6 %, ε jj = 0.1 % For W+N jets background: ε W+Njets = 14.4 % Purity is π = 0.26 No correction has yet been made for jet energy scale. Estimate of systematic errors on top mass and cross section is next on our agenda.

16 Backup Slides

17 Goals of the ATLAS Experiment Investigate mechanism of mass generation. Detect and measure properties of Higgs boson. Search for evidence of Supersymmetry. Probe the nature of matter- antimatter asymmetry and CP violation. Investigate nature of dark matter and dark energy. Perform precision measurements of top quark properties. Operate at L = cm -2 s -1 Provide as many different types of event signatures as possible.

18 Why Study the Top Quark? Top is the heaviest known fundamental fermion. Top will couple more strongly to the Higgs boson. Top provides logical point of departure for studies of the nature of fermion mass generation. Dhiman Chakraborty Mass hierarchy of the elementary particles.

19 Jet Energy Scale Factors affecting jet reconstruction include: Physics factors Initial- and final-state radiation Fragmentation Underlying event, minimum bias events, pile-up Detector performance factors Electronic noise Magnetic field Non-linear calorimeter response Lateral shower size Dead material and cracks Longitudinal energy leakage (high-pT jets) Reconstruction Jet reconstruction algorithm Out-of-cone energy loss

20 Successive Combination vs. Cone Algorithm Successive combination algorithm recursively groups objects (particles, calorimeter cells, towers, clusters) with “nearby” momenta into larger sets of objects. “Nearby” is defined in (E T, η, φ ) space. Initial sets contain one object each; final sets are the jets. Never assigns a single object to more than one jet. Merging jets is not an issue: the algorithm performs this automatically. Geometry of jet boundaries can be complicated. Does not yield regular shapes in η- φ plane. Jet cross sections are likely to exhibit smaller higher-order and hadronization corrections. S. Ellis, D. Soper, CERN-TH.6860/93 Cone algorithm defines jet as set of objects whose angular momentum vectors lie within a cone centered on the jet axis. Jet cones can overlap such that one object is contained in more than one jet. Merging jets is an issue. Cone jets always have smooth, well defined boundaries. Jet cross sections may have larger higher-order perturbative corrections. Parameters can be adjusted such that the inclusive jet cross section resulting from application of successive combination algorithm is essentially identical to inclusive jet cross section obtained by applying cone algorithm.

21 Fit to Top Mass, W Mass m t fit = ± 1.5 GeV m t PDG = ± 3.3 GeV m W fit = ± 1.14 GeV m W PDG = ± 0.03 GeV Jets were reconstructed using Cone4 jet algorithm.

22 Cross Section Estimate Number of initial events in each channel was known from Monte Carlo. Number of final events in each channel after selection cuts was obtained from Monte Carlo. Efficiency was computed for each channel: ε MC = N f MC / N i MC Fraction of events in each channel was calculated using Monte Carlo: F e MC = N f MC (e) / N f MC Fraction of final events per decay channel and efficiency obtained from Monte Carlo were assumed to be valid for “data”. N f data (e) = F e MC N f data, N f data (μ) = F μ MC N f data δN e = √N e, δσ e = δN e / L data ε e, δσ = √(δσ e 2 + δσ μ 2 ) δ ε = √(ε (1 - ε) / N i )