Diphoton searches in ATLAS Marco Delmastro CNRS/INP LAPP on behalf of the ATLAS Collaboration Les Rencontres de Moriond EW 6 Marco Delmastro Diphoton searches in ATLAS
Why diphoton searches? Clean signal over smooth and well known background (e.g. H(5)à γγ) Several extensions of the Standard Model predict high-mass states decaying to two photons Benchmark models HDM Spin- analysis e.g. extended Higgs sector ü 5 physical states h, H, A, H ± ü Under certain conditions, scalar and/or pseudo-scalar states can have sizable branching ratio to diphoton Spin- analysis e.g. Randall-Sundrum graviton Model predicts tower of Kaluza-Klein graviton states with TeV mass scale Phenomenology ü m G* = mass of lightest KK excitation ü κ/m Pl = dimensionless coupling to SM fields Marco Delmastro Diphoton searches in ATLAS
Recap of latest Run results Search for scalar diphoton resonances in the mass range 65-6 GeV with the ATLAS detector in pp collision data at s = 8 TeV Phys. Rev. Lett., 78 Search for high-mass diphoton resonances in pp collisions at s = 8 TeV with the ATLAS detector Phys. Rev. D 9, 4 (5) ] [GeV γ dn / dm γ ATLAS s = 8 TeV, Ldt =. fb Data Continuum+H fit (m X = 5 GeV) Continuum+H fit (m X = 5 GeV) Continuum+H fit (m X = 5 GeV) 8 H(5) 6 4 Data 8 Continuum+H fit (m X = 5 GeV) Continuum component of the fit 6 5 5 5 4 5 6 7 m Events/bin Significance 5 4 - - - Control region ATLAS Data Total b ackground Reducible background syst stat (reducible) syst stat (total) RS, k/ M Pl =., m G* =.5 TeV RS, k/ M Pl =., m G* =. TeV L dt =. fb s = 8 TeV 4 5 m γ Marco Delmastro Diphoton searches in ATLAS
Overview of ATLAS TeV analyses Common pre-selections & photon identification ü E T γ > 4 GeV, E T γ > GeV ü Precision region of EM calorimeter: η <.7,.7.5 excluded ü Tight photon identification based on shower moments in EM calorimeter ü Photon isolation (calorimeter cone + track isolation) SPIN- ANALYSIS Optimized for Higgs-like signal ü E T γ >.4 m γγ, E T γ >. m γγ +% significance for m X > 6 GeV Effectively deplete forward regions As model-independent as possible ü Limit on fiducial cross section Search range ü m X = [ GeV TeV] ü Γ X /m X = [% - %] SPIN- ANALYSIS Loose selection ü E T γ > 55 GeV, E T γ > 55 GeV Preserve acceptance at high mass Use RS graviton as (kinematic) benchmark Search range ü m G = [5 GeV TeV] ü κ/m Pl = [.-.] ü Γ G /m G ~ [.% - %] Γ G ~.44 (κ/m Pl ) Marco Delmastro Diphoton searches in ATLAS 4
Photon energy calibration MV regression to calibrate photon cluster energy, optimized on MC ü i.e. Eur.Phys.J. C74 (4) 7 ü EMC longitudinal layers inter-calibration from from data + additional uncertainty Mostly affecting constant term ü Energy scale and resolution corrections validated with TeV Zà ee events At E T γ > - GeV, resolution dominated by constant term ü c =.6% -.5% E E = p a E b E c Entries / GeV 5 4 Data MC Stat. Syst. Unc. + - Z e e Diboson - Z τ + τ Top + - Z e e TeV, 85 pb Uncertainties ü Energy scale: ±(.4%-%) ü Energy resolution (E Tγ = GeV): ±(8%-%)...9.8 m ee 7 8 9 m ee Marco Delmastro Diphoton searches in ATLAS 5 Data/Pred.
Photon identification and isolation Identification ü 85% (E T ~5 GeV) -95% (E T ~ GeV) ü Uncertainty: full data/mc difference ±% - ±5% for E T > 5 GeV η dependent (tight) ε ID.95.9.85.8.75.7.65 Simulation Pythia prompt photon MC s = TeV η <.6 iso E T < 4 GeV Converted Unconverted.6 4 E T 8 Isolation ü Calorimeter (ΔR=.4) E T iso <. E T γ +.45 GeV ü Track p T iso (ΔR=.) p T iso <.5 E T γ ü Uncertainty: full data/mc difference 5 5 5 5 Marco Delmastro Diphoton searches in ATLAS E T -. x E T 6 Events 5 5 5 5 < p < 45 T. < η <.6 iso Data Bkg template Sherpa Pythia
Sample composition Estimated for both selections using xd-sideband and Matrix Method ü As in SM diphoton cross section measurements (e.g. New J. Phys. 5 () 47) SPIN- ANALYSIS SPIN- ANALYSIS P γγ = 9 + -8 % P γγ = 94 + -7 % [/GeV] dn/dm γγ s= TeV,. fb Spin- Selection Data yield Estimated yield Estimated γ j+jγ yield Estimated jj yield [/GeV] dn/dm γγ s= TeV,. fb Spin- Selection Data yield Estimated yield Estimated γ j+jγ yield Estimated jj yield m γγ m γγ.9.9.8.8.7 Matrix method.7 Matrix method.6 xd sidebands.6 xd sidebands.5 4 6 8 4 6 8.5 4 6 8 4 6 8 m γγ m γγ Marco Delmastro Diphoton searches in ATLAS 7 fraction γγ fraction γγ
Events / GeV Signal modeling SPIN- ANALYSIS Heavy Higgs-like model ü Narrow-width (Γ X = 4 MeV) 6 ü Large-width (Γ X % m X ) Powheg line-shape assuming SM couplings convoluted to detector response (ggf) m X ± Γ X Double-Sided Crystal Ball ATLAS Simulation Preliminary s = TeV, X m X = 6 GeV 8 Γ X /m X = 6% Events / GeV 5 5 SPIN- ANALYSIS RS-graviton-like model ü κ/m Pl =. (Γ G =.% m G ) to measure and parameterize detector response (DSCB) Analytical convolution of theoretical line-shape with detector response ATLAS Simulation Preliminary s = TeV, G* γγ m G* = GeV k/m Pl =. (Γ G* /m G* = 5.8%) 4 5 5 55 6 65 7 m γγ 8 9 4 m Marco Delmastro Diphoton searches in ATLAS 8
Background modeling SPIN- ANALYSIS Functional form à sidebands ü Family of nested functions with increasing d.o.f. ü Validated on MC+data template ü All function parameters free Spurious signal uncertainty ü S+B fits on MC+data template SPIN- ANALYSIS MC+data template à high mass ü Irreducible (γγ) à MC DIPHOX NLO parton level SHERPA γγ including detector simulation, reweighted to DIPHOX m γγ ü Reducible (γj, jγ, jj) à data Inverting tight shower shape criteria Varying loose criteria Mixed according to data-driven purities F-test on binned data ü Validate need of additional d.o.f. (k= chosen) Marco Delmastro Diphoton searches in ATLAS 9
Results SPIN- ANALYSIS SPIN- ANALYSIS Events / GeV 4 ATLAS background-only fit Preliminary Data Background-only fit Spin- Selection Events / GeV 4 ATLAS background-only fit Preliminary Data Background-only fit Spin- Selection Data - fitted background 5 4 6 8 4 6 5 5 4 6 8 4 6 878 events (m γγ > GeV) m γγ Data - fitted background 5 4 6 8 4 6 8 5 5 4 6 8 4 6 8 566 events (m γγ > GeV) Marco Delmastro Diphoton searches in ATLAS m γγ
Results SPIN- ANALYSIS SPIN- ANALYSIS Γ X /m X [%] 8 6 4 Spin- Selection 4.5.5.5.5 Local significance [σ] ATLAS k/m Pl..5..5..5 Preliminary s = TeV,. fb Spin- Selection.5.5.5.5 Local significance [σ] 4 6 8 4 6 m X Largest deviation from B-only hypothesis ü m X ~ 75 GeV, Γ X ~ 45 GeV (6%) ü Local Z =.9 σ ü Global Z =. σ m X = [ GeV - TeV] Γ X /m X = [% - %] 6 8 4 6 8 m G* Largest deviation from B-only hypothesis ü m G ~ 75 GeV, κ/m Pl ~. (Γ G ~ 6% m G ) ü Local Z =.6 σ ü Global Z =.8 σ m X = [5 GeV.5 TeV] κ/m Pl = [..] Marco Delmastro Diphoton searches in ATLAS
/N dn/d Njets Properties of sideband and excess regions SPIN- ANALYSIS.6.5.4... m γγ = [6-7] GeV m γγ = [7-84] GeV m γγ =[84- ] GeV Spin- Selection Data (6 GeV< m < 7 GeV) Sherpa (6 GeV< m < 7 GeV) Data MC γγ 4 5 6 7 8 9 /N dn/d Njets.8.7.6.5.4... Spin- Selection Data (7 GeV< m < 84 GeV) Sherpa (7 GeV< m < 84 GeV) 4 5 6 7 8 9 /N dn/d Njets.8.7.6.5.4... Spin- Selection N jets Data (m > 84 GeV) Sherpa (m > 84 GeV) 4 5 6 7 8 9 SPIN- ANALYSIS N jets N jets N jets /N dn/d Njets.5.4. Spin- Selection Data (6 GeV< m < 7 GeV) Sherpa (6 GeV< m < 7 GeV) /N dn/d Njets.4. Spin- Selection Data (7 GeV< m < 84 GeV) Sherpa (7 GeV< m < 84 GeV) /N dn/d Njets.5.4. Spin- Selection Data (m > 84 GeV) Sherpa (m > 84 GeV)...... 4 5 6 7 8 9 4 5 6 7 8 9 4 5 6 7 8 9 anti-k t R-.4 N jets y j <.4 à p T j > 5 GeV + JVT <.64 Marco Delmastro Diphoton searches in ATLAS N jets.4 < y j < 4.4 à p T j > 5 GeV N jets
] [GeV γγ /N dn/dpt ] [GeV γγ /N dn/dpt Properties of sideband and excess regions SPIN- ANALYSIS.5..5..5..5..5 m γγ = [6-7] GeV m γγ = [7-84] GeV m γγ =[84- ] GeV Spin- Selection 4 5 6 SPIN- ANALYSIS Spin- Selection Data (6 GeV< m < 7 GeV) Sherpa (6 GeV< m < 7 GeV) Data MC γγ γγ pt Data (6 GeV< m < 7 GeV) Sherpa (6 GeV< m < 7 GeV) 4 5 6 γγ pt ] [GeV γγ /N dn/dpt ] [GeV γγ /N dn/dpt..5..5.4 Data (7 GeV< m < 84 GeV)...8.6.4. Spin- Selection Spin- Selection Data (7 GeV< m < 84 GeV) Sherpa (7 GeV< m < 84 GeV) 4 5 6 γγ pt Sherpa (7 GeV< m < 84 GeV) 4 5 6 γγ pt ] [GeV γγ /N dn/dpt ] [GeV γγ /N dn/dpt...8.6.4..4...8.6.4..4 Data (m > 84 GeV) Spin- Selection 4 5 6 Spin- Selection p T γγ Sherpa (m > 84 GeV) Data (m > 84 GeV) γγ pt Sherpa (m > 84 GeV) 4 5 6 γγ pt Marco Delmastro Diphoton searches in ATLAS
Limit on fiducial cross-section 4 4 BR [fb] 95% CL Upper Limit on σ fid BR [fb] 95% CL Upper Limit on σ fid 4 Observed CL s limit Expected CL s limit Expected ± σ Expected ± σ Γ X /m X = % Spin- selection 4 6 8 4 6 8 Observed CL s limit Expected CL s limit Expected ± σ Expected ± σ m X Γ X /m X = 6 % Spin- Selection 4 6 8 4 6 8 m X BR [fb] 95% CL Upper Limit on σ fid BR [fb] 95% CL Upper Limit on σ fid Observed CL s limit Expected CL s limit Expected ± σ Expected ± σ Γ X /m X = % Spin- selection 4 6 8 4 6 8 E γ T >.4 m γγ, E γ T >. m γγ 4 E iso T (ΔR=.4) <.5 E γ T + 6 GeV m X ± Γ X Observed CL s limit Expected CL s limit Expected ± σ Expected ± σ SPIN- ANALYSIS m X m X Γ X /m X = % Spin- selection 4 6 8 4 6 8 Marco Delmastro Diphoton searches in ATLAS 4
Limit on production cross section 95% CL limits on σ BR(G* γγ) [fb] Observed CL s limit Expected CL s Expected ± σ Expected ± σ limit Spin- Selection G*, k/m Pl =.5 95% CL limits on σ BR(G* γγ) [fb] Observed CL s limit Expected CL s Expected ± σ Expected ± σ limit SPIN- ANALYSIS Spin- Selection G*, k/m Pl =. 5 5 5 5 m G* 5 5 5 5 m G* 95% CL limits on σ BR(G* γγ) [fb] Observed CL s limit Expected CL s Expected ± σ Expected ± σ limit Spin- Selection G*, k/m Pl =. 95% CL limits on σ BR(G* γγ) [fb] Observed CL s limit Expected CL s Expected ± σ Expected ± σ limit Spin- Selection G*, k/m Pl =. 5 5 5 5 m G* 5 5 5 5 Marco Delmastro Diphoton searches in ATLAS 5 m G*
Compatibility with 8 TeV data 8 TeV data re-analyzed: latest Run γ calibration + same Run selections + TeV analysis methods Events / GeV 4 SPIN- ANALYSIS Data Background-only fit Spin- Selection s = 8 TeV,. fb Events / GeV 4 ATLAS SPIN- ANALYSIS Preliminary Data Background-only fit Spin- Selection s = 8 TeV,. fb Data - fitted background 4 6 8 4 6 8 5 5 5 4 6 8 4 6 8 m γγ.9 σ at m X = 75 GeV, Γ X /m X = 6% Compatibility with TeV scalar ü gg (scaling: 4.7) à compatibility:. σ ü qq (scaling:.7) à compatibility:. σ 4 6 8 4 6 8 5 5 5 5 4 6 8 4 6 8 No significant excess Compatibility with TeV graviton ü gg à compatibility:.7 σ Marco Delmastro Diphoton searches in ATLAS 6 Data - fitted background ü qq à compatibility:. σ m γγ
Summary Search for new resonances decaying to diphotons performed with. fb TeV data, with two analyses targeting spin- and spin- scenarios Most of the γγ spectrum consistent with B-only hypothesis Largest deviation from background-only hypothesis observed in broad region around 75 GeV, with global significance. (.8) σ for the spin- (spin-) analysis Numerous cross-checks of events with masses ~ 75 GeV performed 8 TeV data re-analyzed using latest Run calibration, compatibility with TeV results assessed ü Scalar. σ (gg). σ (qq) ü Graviton.7 σ (gg). σ (qq) More data needed to verify excess origin: looking forward to 6 LHC run! Marco Delmastro Diphoton searches in ATLAS 7
Additional information Marco Delmastro Diphoton searches in ATLAS 8
Systematic uncertainties Marco Delmastro Diphoton searches in ATLAS 9
Background modeling Relative uncertainty.4.5..5. total sys reducible irreducible purity isolation SPIN- ANALYSIS.5..5 5 5 5 5 m γγ Marco Delmastro Diphoton searches in ATLAS
Acceptance SPIN- ANALYSIS Fiducial region ü E γ (truth) T >.4 m γγ, E γ(truth) T >. m γγ ü E iso (truth) T (R=.4) <.5 E γ (truth T ) + 6 GeV ü m X ± Γ X ggf as baseline ü Difference to alternative production modes as systematics (ggf, VBF, WH, ZH, tth) As model-independent as possible C X = 55%-7% for m X = -7 GeV SPIN- ANALYSIS Total selection efficiency RS graviton as benchmark A G C G = 45%-6% for m G = 5- GeV Source of uncertanty Scalar Common Graviton Photon energy scale and resolution negligible Marco Delmastro Diphoton searches in ATLAS
Compatibility between spin- and spin- analyses Events in spin- analysis are subset of events in spin- analysis for m γγ > 85 GeV, so two analyses are not independent ü Resampling techniques (bootstrap) Union dataset is sliced into N blocks with events in each block. N blocks are randomly picked without considering possible duplication (i.e. the same block could be selected for more than one time). Resampled dataset will then pass spin- and spin- selections, outcome spectra are fed into max LL S+B fit assuming same signal hypothesis Procedure is repeated many times until decent statistics are accumulated for compatibility check Assuming spin- signal:. σ Assuming spin- signal:.9 σ Marco Delmastro Diphoton searches in ATLAS
8 TeV analysis selections Marco Delmastro Diphoton searches in ATLAS
* γγ /N dn/dcosθ Properties of sideband and excess regions SPIN- ANALYSIS.5.5.5.5 Data (6 GeV< m < 7 GeV) 4 Spin- Selection....4.5.6.7.8.9 SPIN- ANALYSIS Sherpa (6 GeV< m < 7 GeV) * cosθγγ * γγ /N dn/dcosθ 4.5.5.5.5 Spin- Selection Data (7 GeV< m < 84 GeV) Sherpa (7 GeV< m < 84 GeV)....4.5.6.7.8.9 * cosθγγ * γγ /N dn/dcosθ 6 5 4 Spin- Selection cosθ * γγ m γγ = [6-7] GeV m γγ = [7-84] GeV m γγ =[84- ] GeV Data (m > 84 GeV) Sherpa (m > 84 GeV)....4.5.6.7.8.9 * cosθγγ * γγ /N dn/dcosθ 5 4 Spin- Selection Data MC γγ Data (6 GeV< m < 7 GeV) Sherpa (6 GeV< m < 7 GeV) * γγ /N dn/dcosθ 7 Data (7 GeV< m < 84 GeV) 6 5 4 Spin- Selection Sherpa (7 GeV< m < 84 GeV) * γγ /N dn/dcosθ 8 6 4 Spin- Selection Data (m > 84 GeV) Sherpa (m > 84 GeV)....4.5.6.7.8.9 * cosθγγ....4.5.6.7.8.9 * cosθγγ....4.5.6.7.8.9 * cosθγγ Marco Delmastro Diphoton searches in ATLAS 4
] [GeV miss /N dn/de T Properties of sideband and excess regions SPIN- ANALYSIS.6.5.4... m γγ = [6-7] GeV m γγ = [7-84] GeV m γγ =[84- ] GeV Spin- Selection 4 6 8 4 SPIN- ANALYSIS Data (6 GeV< m < 7 GeV) Sherpa (6 GeV< m < 7 GeV) Data MC γγ miss E T ] [GeV miss /N dn/de T.6.5.4... Spin- Selection Data (7 GeV< m < 84 GeV) Sherpa (7 GeV< m < 84 GeV) 4 6 8 4 miss E T ] [GeV miss /N dn/de T.8.7.6.5.4... Spin- Selection E T miss Data (m > 84 GeV) Sherpa (m > 84 GeV) 4 6 8 4 miss E T ] [GeV miss /N dn/de T.6.5.4... Spin- Selection Data (6 GeV< m < 7 GeV) Sherpa (6 GeV< m < 7 GeV) 4 6 8 4 miss E T ] [GeV miss /N dn/de T.5 Data (7 GeV< m < 84 GeV).4... Spin- Selection Sherpa (7 GeV< m < 84 GeV) 4 6 8 4 miss E T ] [GeV miss /N dn/de T.45.4.5..5..5. Spin- Selection Data (m > 84 GeV) Sherpa (m > 84 GeV).5 4 6 8 4 miss E T Marco Delmastro Diphoton searches in ATLAS 5
EOYE NW results Local p-value 4 5 σ σ σ σ 4σ Observed 4 6 8 4 6 8 m X BR [fb] 95% CL Upper Limit on σ fid Observed Expected ± σ ± σ 4 6 8 4 6 8 m X Largest deviation from B-only hypothesis ü m X ~ 75 GeV ü Local Z =.6 σ ü Global Z =. σ m X = [ GeV - TeV] Marco Delmastro Diphoton searches in ATLAS 6
Recap of latest Run results Search for scalar diphoton resonances in the mass range 65-6 GeV with the ATLAS detector in pp collision data at s = 8 TeV Phys. Rev. Lett., 78 Search for high-mass diphoton resonances in pp collisions at s = 8 TeV with the ATLAS detector Phys. Rev. D 9, 4 (5) ] [GeV γ dn / dm γ 95% CL limit on σ fid BR [fb] ATLAS s = 8 TeV, Ldt =. fb Data Continuum+H fit (m X = 5 GeV) Continuum+H fit (m X = 5 GeV) Continuum+H fit (m X = 5 GeV) 8 6 4 8 6 Data Continuum+H fit (m X = 5 GeV) Continuum component of the fit 5 5 5 4 5 6 7 ATLAS internal s = 8 TeV, Ldt =. fb Observed Expected ± σ ± σ m Events/bin Significance 5 4 - - - Control region ATLAS Data Total background Reducible background syst stat (reducible) syst stat (total) RS, k/ M Pl =., m G* =.5 TeV RS, k/ M Pl =., m G* =. TeV L dt =. fb s = 8 TeV 4 5 [pb] )γ BR(G* σ - - m γ ATLAS Observed limit Expected limit s= 8 TeV Expected ± σ G* γ Expected ± σ k/ M Pl =. k/ M Pl =. k/ M Pl =.5 k/ M Pl =. L dt =. fb 4 5 6 m X.5.5.5 m G* [TeV] Marco Delmastro Diphoton searches in ATLAS 7-4
Recap of latest Run results Search for Extra Dimensions in diphoton events using proton-proton collisions recorded at s = 7 TeV with the ATLAS detector at the LHC New J. Phys. 5 () 47 Events/bin ATLAS L dt = 4.9 fb s = 7 TeV - - Control region data Total Background Reducible Background syst stat (total) syst stat (reducible) RS, k/ M Pl =., m G =.5 TeV ADD, GRW, M S =.5 TeV Significance - 4 m γ Marco Delmastro Diphoton searches in ATLAS 8
Data sample & Selections Luminosity & Trigger ü Trigger: E T γ 5 GeV, E T γ > 5 GeV + loose EM shower identification ~99% efficient w.r.t. final selections ü. fb ± 5% SCALAR E T γ >.4 m γγ, E T γ >. m γγ 79 events (m γγ > 5 GeV) 878 events (m γγ > GeV) GRAVITON E T γ > 55 GeV, E T γ > 55 GeV 566 events (m γγ > GeV) Marco Delmastro Diphoton searches in ATLAS 9
Background modeling IRREDUCIBLE born box parton fragmentation REDUCIBLE jets in γj and jj events with a neutral meson decaying in collimated photon pairs Marco Delmastro Diphoton searches in ATLAS
A photon showers in the EMC. Most of its energy is lost in Pb Electrons in EM shower ionize LAr Ionization electrons produce current Current is collected, amplified, shaped, sampled and digitized for each EMC cell E in-situ intercalibration MVA calibration cluster corrections clustering electronic calibration Photon energy scale is adjusted to EM scale from Zà ee events Cluster energy is corrected for loss to get photon energy Cluster energies are corrected for detectors effects Cells are grouped in clusters Energy in a cell is reconstructed from signal samples Marco Delmastro Diphoton searches in ATLAS
ATLAS liquid argon electromagnetic calorimeters sampling calorimeter Pb-LAr S X Cells in Layer ϕ η =.45.5 Trigger Tower η =. η = S 47 mm 6X E E = p a E % 5 MeV.7% b E c S 5 mm 4.X.7X ϕ=.45x4 Trigger Tower ϕ =. S (Presampler) Energy loss correction S (Strips) γ/π separation 4. X S 6.8mmx 4 =47.mm 7.5mm/8 = 4.69 mmm η =. ϕ =.45 Square cells in Layer S (Middle) Main energy deposit 6 X S (Back) High energy showers X ϕ Cells in PS η ϕ =.5. η =.5 Strip cells in L ayer Marco Delmastro Simulation of Electron and Photons in ATLAS η
ATLAS Inner Detector + IBL! Properties of track from charged particles ü momentum ü charge Transition Radiation ü e/pi discrimination e/γ discrimination γ conversion reconstruction Marco Delmastro Simulation of Electron and Photons in ATLAS
Electrons and photons in ATLAS Electron and photon candidates in ATLAS are built from EM sliding window fixed-size rectangular clusters, that can be associated to tracks ϕ ü η φ =.75.75 (Barrel),.5.5 (Endcap) ü Unifed between electrons and photons in 5 ü Size is trade-off between energy leakage and noise pickup ϕ η η Cells in EM clusters are calibrated to EM scale ü Including expected average sampling fraction, from simulated electron at GeV Several corrections needed for e and γ, mostly based on simulation ü e.g. energy losses outside cluster, losses in upstream material, variable sampling fraction Marco Delmastro Simulation of Electron and Photons in ATLAS 4
Photon conversion reconstruction Candidate photon conversion vertices reconstructed from tracks pre-selected as loosely matching EMC clusters ü or tracks ü -track conversions from tracks that missing the hit in innermost ID layer Photon reconstruction expected efficiency ü ~ 98% for photons E T > 5 GeV ü > 99% for unconverted photons ü ~ 95% for converted (R < 8 cm) Expected fraction of converted photons ü ~ % at η - ~ 45% η.6 Relative fraction of reconstructed photon conversion depends on: ü Material upstream EMC ü In MC, on conversion model Marco Delmastro Simulation of Electron and Photons in ATLAS 5
Photon conversion reconstruction Fraction of photon candidates.8.6.4. Unconverted Converted Single track conversion Double track conversion s = TeV, L dt = 9 pb γ γ η <.7 (.7 < η.5 excluded), E γ T > 5 GeV Fraction of photon candidates. Unconverted Converted.8.6.4. Single track conversion Double track conversion s = TeV, L dt = 9 pb γ γ η <.7 (.7 < η.5 excluded), E γ T > 5 GeV 5 5 5 5 4 45 5.5.5.5.5 γ E T γ η Marco Delmastro Diphoton searches in ATLAS 6
Photon pointing z position of diphoton primary vertex obtained by combining average beam-spot position with photon pointing, enhanced by using tracks from photon conversions with conversion radii in Si volume ü Resolution ~5 mm in z direction NN discriminant with Σp T,Σp T, diphoton balancing with vertex tracks, trajectory from calorimeter segmentation (z pointing) to choose best vertex candidate ü After this procedure contribution of the opening angle resolution to the mass resolution is negligible. ü Efficiency to reconstruct the correct primary vertex within ±. mm is about 88%. ε PV.8.6.4. Data Z ee, ignoring lepton tracks MC Z ee, ignoring lepton tracks MC H γγ (ggf), m =5 GeV H 5 5 5 Number of Primary Vertices Marco Delmastro Diphoton searches in ATLAS 7
Photon identification S ( Back ) ϒ candidate π candidate S ( Middle ) S ( Strips ) Presampler η Marco Delmastro Diphoton searches in ATLAS 8
Photon identification Marco Delmastro Diphoton searches in ATLAS 9
Photon identification 5 (tight) ε ID.95 Simulation (tight) ε ID.95 Simulation.9.9.85.8.75.7.65 Pythia prompt photon MC s = TeV η <.6 iso E T < 4 GeV Converted Unconverted.85.8.75.7.65 Pythia prompt photon MC s = TeV.6 < η <.7 iso E T < 4 GeV Converted Unconverted.6 4 E T 8.6 4 E T 8 (tight) ε ID.95.9.85.8.75.7.65 Simulation Pythia prompt photon MC s = TeV.5 < η <.8 iso E T < 4 GeV Converted Unconverted.6 4 8.6 4 E T E T Marco Delmastro Diphoton searches in ATLAS 4 (tight) ε ID.95.9.85.8.75.7.65 Simulation Pythia prompt photon MC s = TeV.8 < η <.7 iso E T < 4 GeV Converted Unconverted 8
Photon isolation Calorimetric isolation energy corrected event-by-event ü Leakage of photon cluster ü Underlying event and pileup contributions Average correction for PV ~54MeV Cacciari, Salam and Soyez, JHEP 4, 5 (8) Cacciari, Salam and Sapeta, JHEP 4, 65 () Per-event isolation efficiency.95.9.85.8 H (ggf), m s = 8 TeV s = TeV ATLAS Simulation Preliminary H = 5 GeV.75 5 5 5 Number of primary vertices Marco Delmastro Diphoton searches in ATLAS 4
Photon isolation 5 Events 5 5 5 5 < p < 45 T. < η <.6 Data Bkg template Sherpa Pythia Events 5 5 5 5 5 < p < 45 T.6 < η <.7 Data Bkg template Sherpa Pythia 5 5 5 5 iso E T -. x E T 5 5 5 5 iso E T -. x E T Events 4 8 6 4 5 < p < 45 T.5 < η <.8 Data Bkg template Sherpa Pythia Events 5 5 5 < p < 45 T.8 < η <.7 Data Bkg template Sherpa Pythia 5 5 5 5 5 5 5 5 iso E T -. x E T E T -. x E T Marco Delmastro Diphoton searches in ATLAS 4 iso
Background estimates xd-sidebands L L sample, leading candidate D isolation template fit Identification cut TIGHT non-tight Control region C Signal region A D B Control region Control region events / GeV 9 8 7 6 5 4 ATLAS Data, s γ > 6 GeV E T γγ γj jγ+jj γγ+γj+jγ+jj = 7 TeV, Ldt = 7 pb (leading photon) -5 5 5 5 5 isol E iso T, -5 5 5 5 iso E T, Identification cut TIGHT non-tight Control region C Signal region A A sample, sub-leading candidate D B Control region Control region events / GeV 7 6 5 4 ATLAS Data, s γ > 6 GeV E T γγ jγ γj+jj γγ+γj+jγ+jj = 7 TeV, Ldt = 7 pb (sub leading photon) -5 5 5 5 5 E iso T, -5 5 5 5 iso E T, Marco Delmastro photon and diphoton production at ATLAS 4
Background estimates Matrix Method Passes or Fails isolation cut ε i = probability for a γ to pass isolation cut (data-driven) f i = probability for a jet to pass isolation cut (data-driven) accounting for the correlation of the isolation energy of the γ candidates Event weights Marco Delmastro photon and diphoton production at ATLAS 44
Do we need an additional free parameter? One might be tempted to use a more complex function to fit the data ü If a more complex function f b (x; {p b }) embeds the simpler one f a (x; {p a }) (e.g. there exist a choice of {p b } such that f b (x; {p b }) = one f a (x; {p a }), then the function with more degrees of freedom will have a smaller Χ The problem is to decide whether f b is more motivated than f a ü Eyes are not good judges ü F-test à the function F has Fisher distribution f(f; p, p ) if the added parameter is not improving the model One rejects the hypothesis that the additional parameter if useless if P <.5, where P is the probability of observing a F value at least as extreme as the one in data, if drawn from a Fisher distribution with the same degrees of freedom M. Delmastro on behalf of many High mass diphoton resonance search 45
Statistical procedure Marco Delmastro Diphoton searches in ATLAS 46
Double-Sided Crystal Ball function Arbitrary Units σ CB Gaussian Distribution σ CB.α Low ATLAS σ CB.α High X γγ Power Law -n Low ~(-m ) Power Law -n High ~(m ) m X 5 54 56 58 6 6 64 66 68 m Marco Delmastro Diphoton searches in ATLAS 47
Signal modeling Events /( GeV) ATLAS Simulation Internal s = TeV, X m X = 6 GeV = 4 MeV 8 Γ X Events / 5 GeV 5 ATLAS simulation internal m G = GeV k/m Pl =. 6 5 4 5 Events /( GeV) ATLAS Simulation Internal s = TeV, X m X = 6 GeV Γ X /m X = 6% 8 6 54 56 58 6 6 64 66 m γγ Events / GeV 5 5 96 98 4 6 8 m γγ ATLAS m G = GeV k/m Pl =. simulation internal 4 5 5 55 6 65 7 m γγ 8 9 4 m Marco Delmastro Diphoton searches in ATLAS 48
JVF Jet Vertex Fraction Marco Delmastro Diphoton searches in ATLAS 49