Early top physics with ATLAS

1. Early top physics with ATLAS • Introduction: top physics environment at LHC • What can we expect in the first years..? • Experimental ingredients for top physics at LHC • Electron and muon identification, jets, ET miss and b-tagging • Top-pair cross-section measurements • Initial ‘rediscovery’ and beyond • Single top cross-section measurements • Reconstruction of top pair events and the top mass • Conclusions • Concentrating on ‘early’ measurements and not on couplings and rate decays - covered later… • All results shown can be found in the ATLAS ‘CSC’ book CERN-OPEN-2008-020 Richard Hawkings (CERN) LHC top physics workshop, Lisbon 13/3/09 Richard Hawkings

2. Why top quark physics? • Top quark within Standard Model: • It exists! Measure its fundamental parameters (production cross-section, mass, charge, couplings, etc.) • Electroweak corrections typically  mt2 – interesting for model builders • Heaviest known quark, least studied, some peculiar properties • Decays involve real rather than virtual W • Decays before hadronises – spin/polarisation information is preserved • Top quark beyond the Standard Model: • Top may be produced in new particle decays (t-tbar resonances, heavy H …) • Top quarks may decay in peculiar ways, e.g. tH+b • Top is a ‘template’ for many new physics topologies • Complex decay signatures involving leptons, missing energy, multi-jets, b-jets • Understand the detectors, develop the tools needed for hunting for exotic things • Top production will be a background to many new physics processes • Hleptons, SUSY, more exotic things •  Understanding top physics is essential in many searches Richard Hawkings

3. Top quark production at LHC • At 1033 cm-2s-1 (‘nominal’ low luminosity), get 1 top pair/second, or 8M/year • 30% of these decay to {e/}b jjb or {e/}b{e/}b final states, good trigger efi • At 1034 cm-2s-1 (full LHC design luminosity before SLHC upgrade), get 10 Hz top pairs … O(100M)/year • But background from pileup at 1034 will make precision studies difficult - analyses like top mass cocentrate on 1033 scenario strong t-tbar pair production electroweak single top production (13%) (dominant) (87%) stt(th)=830±100 pb @ 14 TeV st(th)320 pb@ 14 TeV Richard Hawkings

4. Data samples in the first year of running • LHC will start with lower performance • Initial luminosity 1031-1032 cm-2s-1 • CM energy ≤10 TeV in 2009-10 • What can we expect…? • Current plan is to start in October 2009, run through to autumn 2010 • Then shutdown to complete consolidation and prepare 14 TeV(?) • Tt cross-section ~400 pb at 10 TeV • Hope for O(100) pb-1 analysable data from 2009-10 run • Very competitive with final Tevatron ttbar sample Events in 50 pb-1 (ATLAS commissioning selection) +good W mass, 100 pb-1 • Results here assume 14 TeV! • ~factor 2 more L needed at 10 TeV Richard Hawkings

5. Accessing top physics at LHC • Typical analysis - semileptonic top decay • Hard (ET> ~20 GeV) isolated lepton • Electron and/or muon identification • Missing energy ET>~20 GeV • Both lepton and ET miss similar to W production • Hadronic W-decay: • 2 hard (ET>20-40 GeV) light quark jets • Known mjj=mw – calibration of jet energy scale • 2 hard (ET>20-40 GeV) b quark jets • b-tagging essential tool in top identification • Some opportunities for b-tag calibration • Explore/commission full range of detector capabilities (leptons, ET miss, jets, b-tagging) • Main background is W+multijet production • Multijet with fake lepton will also be important • Both difficult to simulate, need data normalisation • In dilepton channel, Z+jets and diboson prodn t t • Single top: similar ingredients, typically study semileptonic decay ({e/} b ) Richard Hawkings

6. Ingredients - electron identification • Leptons are key handle on top decays • Detector optimised for electron/muon ID • Huge background (~105) from jet production • Electrons identified by: • Shower shape in EM calorimeter • Matching high pT track in tracking detector • Transition radiation in TRT (e/ separation) • Sources of electrons • Fakes from jets (especially at low ET) • Electrons from heavy flavour (b,ce) in jets • ‘Prompt’ electrons from W/Z (and top) • Top (and W/Z) physics requires isolation • Low activity in a cone around electron • E.g. ET<6 GeV in cone R<0.2 • Effectively removes heavy flavour, leaving W,Z Electron candidates/100pb-1 Rejection of non-isolated e in top events Barrel, 30 GeV Endcap, 17 GeV Richard Hawkings

7. Ingredients - electron trigger and efficiency pT(e) in top decay • Electron also used to trigger event • L1 calorimeter trigger with fast sums • L2 and EF based on full detector info • Good efi for 25 GeV threshold - around 85% of offline selected electrons • Majority of tWe have pT>20 GeV • Need to measure trigger and selection efficiency for electrons from data • Use copious Zee decays to measure ‘generic’ efficiency with ‘tag and probe’ • Select one electron with standard cuts • Look for a second electron which makes Z mass, see how many pass ID cuts • Corrections needed to subtract fake b/g • Method can be used for trigger/offline • Corrections needed for Ztop • Different isolation performance due to presence of extra jets • Eventually take from data, need statistics Electron trigger efficiency (wrt offline) Probe e Z 1-2% error for 100pb-1 Tag e Richard Hawkings

8. Ingredients - muon identification and efficiency • Muons reconstructed in external muon spectrometer with toroid field • Matching to inner detector tracks reduces fake rate and improves pT resolution • Fake rate for pT>20 GeV ~10-3 in top-pair events - K/ decays in flight and punch-through • Further improvements by cutting on match 2 • Efficiency ~90% but losses due to coverage • As for electrons, need isolation to remove muons from heavy flavour decays Efficiency pT()>10 GeV • Muon trigger: • Dedicated chambers (RPC, TGC) at L1 • Full info at L2/EF • Measure efi using tag-and-probe Z • As for electrons Muon trigger efficiency #fake muons/tt event Richard Hawkings

9. Ingredients - jet reconstruction and calibration Particle types in jets • Quarks from t/W decay give jets in detector • Want to reconstruct energy and direction… • Fragmentation and hadronisation • Complex detector response to different particles; different technologies, dead material, cracks • Effects of clustering, jet algorithms, out of cone • Effect of ISR/FSR, underlying event and pileup • Jet efficiency (ET threshold), resolution, scale • Just calculating acceptance (N jets > threshold) will take time and data  photons K,p,n • /Z-jet balance to determine ET scale • Few % statistical precision with 100 pb-1 • Ultimate goal ~1% PT balance vs energy Zee jet Z+1 jet event Richard Hawkings

10. Ingredients - B-jet energy scale and missing ET • Complication for b-jets (important in top) • 19% of b-jets contain a muon + neutrino • b, bc decays accompanied by  • Effect does not occur in light jets - systematic underestimate of b-jet ET scale • If muon found in jet, can correct (O(10%)) • Similar effect also for electrons… • Missing ET measurement - sensitive to energy carried off by neutrinos in transverse plane • Genuine ET-miss in top events ( from W) • Remove b/g from QCD multijets with fake e/ • Stages of ET-miss calculation/refinement: • Calorimeter cells sum (noise suppression, calib) • Addition of identified muons (fake , ID/calo info) • Refine calo info for jets; estimate energy in dead regions • Painstaking work to commission… • Can also use phsyics processes to help with calibration (including semileptonic top decays) b-jet energy scale, Jets with  ET-miss resolution evolution 500 GeV dijets tt{e,}bjjb W{e,} Z Richard Hawkings

11. Ingredients - b-tagging using top-pair events • Top-pair events offer source of b-jets for use in b-tagging efficiency calibration • Method I: Count number of events with 0,1,2,3 b-tagged jets in cross-section seln • Can determine b and c along with tt • Get b to ~5% (incl. syst) in 100 pb-1 • Method II: Exploit the topology / kinematics of the top-pair event to select the leptonic top b-jet, without using its b-tag info • Selection methods exploiting mass info, kinematic fit or jet/lepton kinematics • Jet samples ~70-90% pure in b flavour, have to estimate and subtract background • Can then study distributions of b-tagging input and output variables, calculate b • Get b to 5-10% in 200 pb-1, can also look at distributions (e.g. b vs jet ET and ) • b determined in-situ, complementary to dijet-based methods also used at Tevatron b-vetos b-tag unbiased b-jet leptonic top mass Abs stat error, 200 pb-1 Richard Hawkings

12. Rediscovering the top quark • Detector performance will take data and time to understand • For initial top quark ‘rediscovery’… emphasis on simple selections • Select events with high ET lepton, missing ET and 4 jets (ET>20-40 GeV) • Backgrounds from W+jets events, QCD multi-jet with fake lepton, single top • W+jet and multijet events can be reduced by requiring 1 or 2 b-tagged jets (b~ 50-70%) • But also interesting to look at simplified analyses without b-tagging: • Find combination of 3 jets which represents the hadronic top decay • Criteria such as largest sum pT, good W mass between 2 of the jets +good W mass, 100 pb-1 Largest sum pT, 100 pb-1 Richard Hawkings

13. Top cross section in semileptonic channel • Extract signal using two methods: • Event counting (need background estimate) • Fit to 3-jet hadronic top mass distribution • Reduces sensitivity to W+jets background normalisation and jet energy scale • ISR/FSR (acceptance) modelling also important • Once b-tagging is available, use it: +jets Statistical signficance vs L • With 100 pb-1, expect to commission b-tag and understand efficiency to ~5% • Helps with W+jets background reduction and combintorial background e+jets +jets • 5 sensitivity from ~ 20pb-1 • Very clear signal at 100 pb-1, even with x2 background Richard Hawkings

14. Top cross-section in dilepton channel • Di-lepton signal (lb lb) potentially very clean: • Background from Zee,  can be removed with mass cut, leaving Z, lepton+jets ttbar and diboson (WW/ZZ/WZ) • e channel particularly clean (only Z is from ) • Require ET-miss + 2 jets… no need for b-tag • Cross-section extracted using counting, likelihood or template (ET-miss vs njet) Dilepton mass 100 pb-1 Jet multiplicity • Systematics dominated by jet ET scale, ISR/FSR • Clear signal even with 10-20 pb-1 • Probably the ‘discovery’ channel Stat+syst significance Richard Hawkings

15. Properties of initial top-pair events • Once top-pair signal established, look at properties of events for non-SM behaviour • Use semileptonic selection - study pT and distribution of top… can also start to look for extra jets etc • Study top-antitop invariant mass distribution • Constrained fit to t-tbar system using known W and top masses (without b-tagging - could be used to reduce combinatorics) • t-tbar mass resolution 5-9% from 200-850 GeV Hadronic top pT Hadronic top  t-tbar invariant mass Richard Hawkings

16. Top cross-section using tau decays b-tag weight of highest pT jet • Expect top-pair states involving  same as e, • Branching ratio could be enhanced: tH+b • Important background for e.g. SUSY searches • Decay tt {e,}b had b (‘leptonic’) • Similar to semileptonic selection, but with identified hadronic tau replacing jets from Wqq • Get additional rejection using 1 or 2 b-tagged jets • S/B 1:10 without b-tagging, get 1:1 with 1 b-tag • ~50 events, in 100 pb-1, b/g W+jets and single top • Can also use this channel to study -identification • Decay channel ttqqb hadb (‘hadronic’) • No electron or muon - need to trigger on some combination of , N-jets and ET-miss • Require ≥4 jets, of which 2 must be b-tagged, plus identified hadronic  • Expect about 300 candidates in 100 pb-1, with S/B~20 • Use to study  identification and trigger (for events triggered on jets/ET-miss) • Both analyses are challenging - requiring well understood ID, jets, b-tagging,… • Will come later than the ‘discovery’ analyses with electrons and muons Richard Hawkings

17. Single top production at LHC • Electroweak top quark production - contrast to pair production • Sensitive to new particles (e.g. H+, W’) and flavour changing neutral currents • Important background to many new physics searches (lepton, missing energy) • Overall cross section is large (c.f Tevatron), can distinguish contributions: • Large backgrounds from top pair production, also W+multijet and QCD jet events • At LHC - attempt to measure all production modes (s-chan & Wt challenging) • Can then extract |Vtb| and study polarisation, charge asymmetries, searches … • Basic event signatures: high ET lepton, missing ET, restricted number of (b) jets • Plenty of events, but large backgrounds which will have to be understood from data s-channel: t=111 pb t-channel: t=24712 pb Wt-channel: t=662 pb ( c.f. top-pair tt=83050 pb ) Richard Hawkings

18. Single top production: t-channel cross-section • Require lepton, missing ET and one b-jet from the top quark decay • Infer top quark mass using missing ET • Jet from light quark is forward, can require this jet and/or veto additional central jets • Second b-jet is usually soft - below ET cut b-jet pT BDT value • O(1k) events per fb-1, similar size tt background  large systematics (jet E scale,b,b/g) • Can be reduced by multivariate techniques - e.g. Boosted Decision Tree with event shape variables • Measurement to ~10% precision possible with 10 fb-1 • Then get |Vtb| to ~5% ATLAS Mtop BDT>0.6 Richard Hawkings

19. Single top production: W-t and s-channel • Much smaller signal cross-sections, very large background • Especially from top-pair events where some particles are missed • W-t: channel: Single b-jet; look for two light jets consistent with W decay (l-j channel), or second lepton from leptonic W decay • Can use control region with similar kinematics but rich in top-pairs (e.g. require extra b-jet) to estimate background, cancel systematics • s-channel: Two b-jets, lepton + missing ET, no other high ET jets W-t chan s chan • In both cases, multivariate techniques can be used to enhance signal significance • Some representative analysis results - note small S/B and large systematics • Mainly from background - b-tagging/vetos, jet energy scales, PDFs, .. • Need O(10) fb-1 of data and careful background studies to establish 5 signals Richard Hawkings

20. Top reconstruction - for mass and more • Top mass, angular analysis need reconstruction of ttbar final state • 4-vectors of all or some particles • Kinematic fits, exploiting constraints • E.g. W-mass in light jets from Wqq should give known W mass • Extra ‘in-situ’ calibration of light jet energy scale for the top-pair topology • Essential for top mass systematics Energy scale correction Reconstructed W-mass • Analyses typically require ≥1 fb-1 • Make tight selections (4 jets ET>40 GeV) + b-tags • Various other cuts on reconstructed W and top • E.g. extract hadronic top mass from mjjb distribution • Statistical error 0.3 GeV from 1 fb-1 data • Other techniques, e.g. global 2fit also involving leptonic side of event, under development • Important for spin correlations, polorisation etc … Hadronic top mass Richard Hawkings

21. Top mass measurement • Top mass quickly systematics dominated • C.f. Tevatron current average 1.2 GeV • Need to control light and especially b-jet JES to O(1%) to be competitive and exploit statistics (0.3 GeV stat error in 1 fb-1) • This will take time and data… • Control ISR/FSR from data distributions, e.g. jet multiplicity • For early data … can extract top mass with relaxed b-tag requirements • Start to learn about techniques and data Light jet multiplicity ISR/FSR changes Hadronic top mass,1-btag events Richard Hawkings

22. Conclusions • LHC now has a firm ‘recovery’ schedule and expects to start in late 2009 • Hope for 100-200 pb-1 data, mainly at 10 TeV, by autumn 2010 • This gives a top pair data sample very comparable to that of the Tevatron • But we have to achieve a mature understanding of the detector ... will take time • On the other hand, can make first observations with first ~20 pb-1 • Top analysis requires and exercises much of ATLAS object reconstruction • Electron and muon trigger and identification • Jet reconstruction and energy scale • Missing ET reconstruction • Tagging of b-jets • … all this will also help with understanding top as a background for searches • Initial measurements • Focus on top-pair cross section in dilepton, semileptonic and then other channels • Interesting measurements can already be done with 100-200 pb-1 • Start to develop tools to go further - top reconstruction, single top backgrounds • Then move on to ‘new’ territory: single top, top properties, rare decays … Richard Hawkings