Search for Standard Mod el Higgs in Two Photon Final State at ATLAS

1. Search forStandard Model Higgsin TwoPhotonFinal State at ATLAS HYEON JIN KIM JUNE 25, 2010

2. Outline • Standard Model and Higgs Boson • ATLAS at the LHC • Photon and Electron Identification (ID) in ATLAS with a Covariant Matrix based Method (H-matrix) • A Data Driven Method for Photon Identification Efficiency Measurement • Photons Showers in ATLAS Cosmic-ray Muon Data • Prospect for Higgs Search using H➝  • Conclusions Search for H → γ γ

3. STANDARD MODEL AND HIGGS BOSON

4. Standard Model • The Standard Modelof particle physics describes elementary particles and how they interact. • Matter is made of fermions. ✒ Leptons and quarks • The interactions between elementary particles ✒ The electromagnetic, the strong and the weak interactions ✒ mediated by particles called gauge bosons. • Gauge invariance requires all these particles to be massless! ✒ Experimental values of the masses of the W and Z gauge bosons suggest otherwise. ✒∼85-90 times the mass of a proton Search for H → γ γ

5. Higgs Mechanism • Higgsmechanismcangivemass to W, Z bosonswithoutbreaking gauge invariance. ✒ It keeps the photon massless. • The Higgs boson has not yet been observed experimentally. ✒ The search for the Higgs boson is one of the most important goals of the ATLAS experiment at the Large Hadron Collider (LHC). • Despite the prediction of Higgs, the theory does not provide a direct estimate of its mass. ✒ The LEP experiments have set a 95% confidence level (CL) lower limit on the Higgs mass: mH > 114.4 GeV @ 95% CL. Search for H → γ γ

6. Higgs production & decay mode at LHC Gluon Fusion • Large center of mass energy ✒Gluon fusion is the dominant production channel over the whole range • H ➝ γγ has relatively small branching ratio but a clean signature with two back-to-back photons ✒ The EM calorimeter identify electrons and photons against overwhelming background from hadronic jets Vector Boson Fusion Search for H → γ γ

7. ATLAS at the LHC

8. Lake of Geneva CMS Airport LHCb ALICE ATLAS LHC (Large Hadron Collider) • s = 14 TeV(7 times higher than Tevatron) • search for new massive particles up to m ~ 5 TeV • Ldesign = 1034 cm-2 s-1 (>102 higher than Tevatron) • search for rare processes with smallσ (N = Lσ ) Search for H → γ γ

9. ATLAS (A Toroidal LHC Aparatus) Installed just across the CERN main site, 92 m below ground. Tile barrel Tile extended barrel LAr EM end-cap Proton Beam Proton Beam LArhadronic end-cap LAr EM barrel LAr forward calorimeter Search for H → γ γ

10. Electromagnetic (EM) Calorimeter accordion-shaped Measure energy and direction of  & e • Presampler(|| < 1.8) ✒ correct for energy lost in dead material in front of calorimeter. • The barrel EM calorimeter (||<1.475) ✒ 1st layer :/0 separation,  position measurement ✒ 2nd layer :main energy measurement ✒ 3rd layer : high energy shower tails Hadron/EM separation • The end-cap calorimeter (1.375<||< 3.2) Search for H → γ γ

11. PHOTON AND ELECTRON IDENTIFICATION IN ATLAS WITH A COVARIANT MATRIX BASED METHOD (H-matrix)

12. Motivation • e/ identification with high efficiency and high rejection against jets is crucial for H ➝ search • Measure and understand e/ ID efficiencies and background rejection ✒ Improve Higgs search efficiency • Selection of e/ from jetsis based on & their characteristic features in EM calorimeter. • Covariant matrix technique (H-matrix ) w/ used at DØ for e-ID ✒ Used EM shower shape correlations ✒ Proved to be a great tool for e/ and hadrons separation • Benefit from fine cell sizes of ATLAS Calorimeter • Improve e/ ID by generating a 2 quantity utilizing H-matrix technique Search for H → γ γ

13. Principle of H-matrix • H-Matrix exploits the correlation in transverse and longitudinal electromagnetic shower shapes. • The correlations between the variables are reflected through the covariant matrix, M, whose elements are • e/-likeness is presented by χ2 a given candidate object m. where, Hij (covariant matrix)-1and yi is discriminating variables. • A shower that closely resembles an e or  shower will have a m2 ~ n dim. χγ2 Search for H → γ γ

14. Discriminating Variables (i) • Longitudinal Variables ✒ Fractional energies in each layer of EM cal (fi, i =0~3). ✒ Fractional energy in 1st layer of hadronic cal (f4) • Transverse Variables ✒ Fractional energy in shower core (R37). ✒ The second layer of EM calorimeter a. The Ratio of energy in 3x7 over 7x7 (R) & in 3x3 over 3x7 (Rφ) b. Lateral width in (ω2) f4 ω2 R f2 Search for H → γ γ

15. Discriminating Variables (ii) ΔE = Emin – Emax2 • Transverse Variables • ✒ The first layer of EM calorimeter • a. Shower width over 3 strips around maximal energy deposit (ω3stips) • b. Shower width over 20 strips (ωtot1) • c. Energy fraction outside of shower core (Fside) • d. Energydifference (ΔE) & fractional energy of 2nd maximal energy (Rmax2) in 1st layer EM calorimeter Fside Rmax2 ω3strips Search for H → γ γ

16. Mean values & Covariant Matrix M • Build H-matrix separately for e and  ✒ The shower shapes are different b/w electron and photon. ✒ The single electron and photon MC samples • The discriminating variables have energy and position (in η) dependences • The covariant matrix is parameterized by using energy & η. 200 GeV electrons Photons at 1.52 < |η| < 1.8 Search for H → γ γ

17. Available instrumentation in  region • 12 subdivisions of  based on granularities & the thickness of absorber in EM calorimeter. Search for H → γ γ

18. Parameterization • Mean values are parameterized as function of energy. ✒ f3 : ✒ others : ✒ f4, ΔE and Rmax2: Linear interpolation between the two adjacent energy points • Determine the energy dependence of M ✒ Linear interpolation between the two adjacent energy points Single electrons samples at 0.6 < ||< 0.8 1.0 < ||< 1.2 Search for H → γ γ

19. H-matrix for Background • To enhance H-matrix performance, H-matrix is need to be optimized using information of both signal and background (i.e. jets) ✒ Log-likelihood ratio : the exact shapes of background for p.d.f. construction ✒ Cut-based algorithm: sets the cuts values by looking at both the signal and backgrounds shapes • To incorporate background shape, H-matrix for background is built. • γ-jet samples in different ET ranges are used for jet H-matrix. • Photon H-matrix incorporate jet H-matrix. ✒ χ2 = ½ (χγ2 -χjet2) for combined photon and jet H-matrix Jet H-matrix Obtained from H → γγ (mH = 120 GeV) & dijet samples χjet2 Search for H → γ γ

20. H-matrix Performance • Efficiency and rejection ✒ Efficiency (ε) = ✒ Rejection = • Samples : generated by pythia and full detector simulation ✒ Z→ee ✒ H →γγ with mH = 120 GeV ✒ Highly EM dijet sample • Using means values of shower variables and covariant matrix M, χ2 is • calculated by Photon-jet H-matrix Photon H-matrix Electron H-matrix χγ2 χe2 Search for H → γ γ

21. Electron H-matrix Performance • H-matrix performance compared to cut-based method ✒ The tune ofχe2cut values in different ET bins ✒ Provides the same efficiency at the cut-based in each ETbin. • Rejection power calculated ✒ H-matrix shows better rejection. Search for H → γ γ

22. Photon–jet H-matrix Performance • Efficiency change with ET ✒ Adjust the χγ2 cut value in different ET bins to make the efficiency constant with respect to ET of the photon. • Tuned χ2 cut values to give same efficiency as cut-based algorithm ✒ Its performance equivalent to that of cut-based algorithm. Photon H-matrix Photon–jet H-matrix Photon-jet H-matrix Combined photon and jet H-matrix enhance photon H-matrix performance. Search for H → γ γ

23. Systematic Uncertainty • Effect of data to simulation • discrepancies in shower shapes • ✒ Shower property differ in data from their • distribution from simulation • ✒ Measure the potential effect of data to • simulation discrepancies on the performance • of the H-matrix • Effect ofdifferent EM scales in data and • simulation • ✒ EM scale uncertainty is known to about 2% • In early period of ATLAS data taking. • ✒ Emeasured /Etrue is 1.00 ±0.02 Discrepancies : 0.5% ~ 5% Emeasured = Etrue (1+0.02) Emeasured = Etrue (1- 0.02) Search for H → γ γ

24. A DATA DRIVEN METHOD FOR PHOTON IDENTIFICATION EFFICIENCY MEASUREMENT

25. Motivation • Pure photon sample is important ✒ To measure photon efficiency ✒ To understand shower shapes • No known physics process to give pure photon samples. • Z→μμγis source of pure photons at high center of mass and luminosity. • This method tested by using Z→μμMC sample • Z→μμγ events in Z→μμare selected by : ✒ Photon non collinear to muons ✒ Invariant mass cut of 2 muons and photon 3-body decay ISR (Initial state radiation) FSR (Final state radiation) γradiation Search for H → γ γ

26. Simulation Samples and Event Selection • Sample (√s = 10 TeV) ✒ Signal : Z→μμ generated by pythia , alpgen and full detector simulation ✒ Background • tt (mc@nlo + jimmy) • gg→WW →μμ(gg2WW + jimmy) • bb→μμ(pythiaB) • Z➝μμγevent selection ✒ Cut A: select 2 muons per event (ETμ1> 20 GeV, ETμ2> 6 GeV) ✒ Cut B: cut A+ iso cut to muon (Et cone < 5 GeV) ✒ Cut C: cut B + 1 loose photon per event (ETγ> 10 GeV) ✒ Cut D: cut C+ ΔR cut (Min(ΔR(μ1,γ), ΔR(μ2,γ)) >0.2) where ΔR = √(Δφ2 + Δη2) ✒ Cut E: cut D+ Mass cut (81 < Mμμγ < 101GeV) Search for H → γ γ

27. Kinematic Variables Photons frZ →μμγ, Photons frH → γγ (mH = 120 GeV/c2) Search for H → γ γ

28. Shower Shape Variables Photons frZ →μμγ Photons frH → γγ (mH = 120 GeV/c2) Required 30 < ET < 40 GeV Fside • Photons from Z→ mmγhave identical • shower shape to photons from • H→ γγ. • These photons can be used to • measure photon efficiency in data. f0 Search for H → γ γ

29. Efficiency ofγin Z➝μμγevents • Photon efficiency from Z→μμγshows consistent behavior with H→γγ • The photons in Z→μμγcan be used to measure photon efficiency. Cross-section = 1143 pb forZ→ μμ Efficiency of cut-based ID vs. ET Efficiency of H-Matrix vs. ET H→ γγ Z→μμγ H→ γγ Z→μμγ Search for H → γ γ

30. Statistical Uncertainty • The expected statistical error versus luminosity for different ET bins. • This graph takes only into account the 1/√Nbehavior of the • statistical error in each bin √s = 10 TeV For 200 pb−1 2% for 10 < ET < 20 GeV, 3% for 20 < ET < 40 GeV, 4% for 40 < ET< 60 GeV. Search for H → γ γ

31. Sources of Background Photons in pp➝Z➝ μμEvents 81< M < 101 GeV 85< M < 95 GeV Search for H → γ γ

32. Contamination by Other Standard Model Processes • Integrated luminosity : 200pb-1 • Center of mass energy : √s = 10 TeV Mμμγ (GeV) Search for H → γ γ

33. Extrapolation to other Samples, Different η distribution • The photon identification efficiency is dependent on η of the • photon. • ✒ The total efficiency will not be correct if the photon efficiency derived from thephotons from Z→μμγis applied to samples with different ηdistribution. • To estimate the size of this effect, the photon efficiency is calculated • in a γ-jet sample in different η bins Search for H → γ γ

34. PHOTON SHOWER IN ATLAS COSMIC-RAY MUON DATA

35. Motivation • To understand actual detector and e/γ ID performance, we need to compare in data and simulation ✒ Low level calorimeter quantities (noise, resolutions, …) ✒ Intermediate level quantities (energy fractions, shower variables) ✒ High level (efficiency and rejection of electron/photon ID tools) • The ATLAS cosmic data from 2008 contain O(108) events • Large sample of EM showers in ATLAS cosmic data ✒ The high-energy bremsstrahlungphotons produced from cosmic muons ✒ Actual shower distribution in data • Unlike photon from collision, photon from cosmic-ray muon has different energy profile in each layer of EM calorimeter ✒ Shower shape distribution may differ ✒ Study shower shape with place where photon emitted from cosmic muon. Search for H → γ γ

36. Photon Emission Point (i) • Find photon emission point from muon in cosmic ✒ Photon shower variables are sensitive depending on where photon radiated from muon track ✒ Track parameters of muon & positions of EM shower in the1st and 2nd layers ✒ The interception point of muon track and photon trajectory • Xγ and Yγ : the transverse coordinates of the emission points of the photons from μ→μγ • Rγ = √Xγ2 + Yγ2 ✒ equivalent to the true value of μ → μγvertex radius Rvertex Rγ =√Xγ2 + Yγ2 Search for H → γ γ

37. Photon Emission Point (ii) • The most probable value for Rγ corresponds to middle of EM • Photon have less probability to be recognized as photon if the shower occurs in different region of ATLAS Rγ -Rvertex Search for H → γ γ

38. Cosmic Data and Simulation • Data: Triggered by L1Calo and IDCosmic data streams ✒ merged two data stream with removing double counted events • Cosmic simulation : produced with inner detector volume Cut Flow Search for H → γ γ

39. Comparison of Kinematic Variables • Kinematic variables of photon in MC & data have good agreement. • Impact parameter d0 of muon shows difference in MC & data ✒ It could be the differences of trigger & tracking efficiencies in MC & data Search for H → γ γ

40. Shower Shape Comparison • Longitudinal shower variables • Most distributions of shower • variables are in agreement • between data & MC • Data is well modeled by • simulation. • Z→ μμγallows to do similar studies in collision data with much higher precision Search for H → γ γ

41. PROSPECT FOR HIGGS SEARCH USING H ➝ 

42. Signal and Background Process • Signal process (√s = 10 TeV) ✒ Gluon fusion process : masses of Higgs are 115, 120, 130, and 140 GeV/c2 • Background processes (√s = 10 TeV ) ✒ Irreducible background • Two prompt photons from qq→γγor gg→γγq • Bremsstrahlung from qg→qγ→qγγ . ✒ Reducible background a. γ-jet events where a leading π0 in a jet has been misidentified as one photon b. jet-jet events where both jets have been misidentified as photons. • Signal selection ✒ η cuts : |η| < 1.37 or 1.52 < |η| < 2. 37 (remove crack regions) ✒ pT cut on photons : pTγ1 > 40 GeV, pTγ2> 25 GeV Search for H → γ γ

43. Kinematic Variables Subleading photon Leading photon √s = 10 TeV √s = 10 TeV H ➝γγ(mH = 140 GeV/c2) H ➝γγ(mH = 120 GeV/c2) Search for H → γ γ

44. Signal Significance • Signal significance : expected number of signal and background events. • Requiring event preselection, mass cut mH ± 1.4σ and H-matrix cut. • Using the fake rates and efficiency for different χ2 cut values the expected number of photon events at 100 fb−1 and √s = 10 TeV are obtained. Search for H → γ γ

45. Conclusions (i) • A covariant matrix based algorithm has been developed to identify electrons and photons. ✒ Exploits the high granularity of the ATLAS EM calorimeter and use correlations between various EM shower shape variables ✒ To enhance further the jet rejection a jet H-matrix has been built and combined with the photon H-matrix. ✒ The electron H-matrix shows significantly better performance than the current standard ATLAS cut-based electron. ✒ The combined photon-jet H-matrix shows also excellent jet rejection but is comparable to the ATLAS cut-based method • Prior to the LHC beam, considerable amount of cosmic data were collected with the ATLAS detector ✒ Provided a great opportunity to study the detector response and compare it to the predictions of the simulation ✒ The shower shapes of photon in the cosmic-ray data and simulation are in good agreement. ✒ The shower shape variables produced by the geantsimulation can be relied on. Search for H → γ γ

46. Conclusions (ii) • A new process has been studied that will allow to measure the photon identification efficiency directly with the atlas data. ✒ Using bremsstrahlung photons in Z → μμγevents a pure sample of prompt photons can be isolated. ✒ It is possible to precisely extract efficiencies for photons with ET below 40 GeV • H → γγis very challenging and will only be possible to discover this signal once the atlas and the lhchave reached their design performance. ✒ The photon H-matrix provides a powerful tool to reject the γ +jet and jet-jet backgrounds to the H → γγsignal. • The validation of the photon H-matrix should become possible in data with the Z → μμγevents well before the search for H → γγbecomes statistically competitive. • With the upcoming atlas data it will soon be possible to measure the photon identification performance in real data (Now ready for data in Higgs search) Search for H → γ γ

47. Backup

48. +35 MH = 85 GeV/c2 -27 The Latest Result From Tevatron • The LEP EW working group has combined the experimental data from several precision EW measurements of SM parameters. ✒ the LEP experiment has set a 95% confidence level (CL) limit on the Higgs mass: mH> 114.4 GeV @ 95% CL. Search for H → γ γ

49. g p p H g H➝  channel • Very rare decay (BR ~10-3) • Background ✒ Irreducible:  continuum ✒ Reducible: -jet and jet-jet • Keys ✒ Excellent energy and angular resolutions, ✒ Excellent  efficiency, jet rejection ✒ High granularity and response uniformity ✒ H width negligible, resolution dominated by detector ✒ σ(M)/M Search for H → γ γ

50. ATLAS Coordinate System • The ATLAS Coordinate System is a right-handed system • The x-axis pointing to the centre of the LHC ring, the z-axis following the beam direction and the y-axis going upwards. ✒ At ATLAS positive z points towards LHCb. ✒ The azimuthal angle φ = 0 corresponds to the positive x-axis and φincreases clock-wise looking into the positive z direction. ✒ The polar angle θ is measured from the positive z axis. Search for H → γ γ