Search for a Standard Model Higgs boson in the H ZZ ( ) * 4l decay channel with the ATLAS Experiment at CERN Giacomo Artoni Seminario Finale Dottorato XXV Ciclo 5// Thesis Advisor: Prof. Carlo Dionisi
The Large Hadron Collider LHC design parameters: s = 4 TeV L = 34 cm s September 8: incident in sector 34, running in reduced mode : s = 7 TeV ] Total Integrated Luminosity [fb 7 ATLAS Online Luminosity s = 7 TeV 6 5 4 3 LHC Delivered ATLAS Recorded Total Delivered: 5.6 fb Total Recorded: 5.5 fb 8/ 3/4 3/6 3/8 3/ Day in Peak luminosity L 4 33 cm s Integrated luminosity L = 5. fb : s = 8 TeV Peak luminosity L 8 33 cm s Integrated luminosity L = 6.3 fb (midjune)... Giacomo Artoni Seminario Finale Dottorato XXV Ciclo 5// ] Total Integrated Luminosity [fb 8 6 4 8 6 4 ATLAS Online Luminosity LHC Delivered ATLAS Recorded Total Delivered: 5. fb Total Recorded: 4. fb s = 8 TeV 8/3 7/4 7/5 7/6 7/7 6/8 6/9 Day in
The Large Hadron Collider The increased instantaneous luminosity in mainly obtained by increasing the number of colliding bunches in the LHC This translates directly into more interactions happening for each bunch crossing: almost doubled mean number of interactions per event This effect is usually referred to as pile up Decreases detector performance and makes simulation more complicated We already beyond detector design! Peak interactions per crossing /.] Recorded Luminosity [pb 5 45 4 35 3 5 5 5 8 6 4 ATLAS Online Luminosity Jan Apr Jul s = 7 TeV s = 7 TeV s = 8 TeV Oct detector design Jan Apr Jul Month in Month in Month in ATLAS Online Luminosity s = 8 TeV, s = 7 TeV,, <µ> =. 5 5 5 3 35 4 Oct Jan Apr Ldt = 4. fb Ldt = 5. fb Mean Number of Interactions per Crossing Jul Oct, <µ> = 9. Giacomo Artoni Seminario Finale Dottorato XXV Ciclo 5// 3
The ATLAS detector Tracking: Pixel detector, Semiconductor tracker, Transition Radiation Tracker in a solenoidal magnetic field of ~ T EM Calorimeter: lead / liquid argon Hadronic Calorimeter: lead / liquid argon iron / scintillating tile (depending on region) Muon Spectrometer: Resistive Plate Chambers and Thin Gap Chambers (trigger) Drift Chambers (precision) in a toroidal magnetic field Multi Purpose detector at the LHC: designed to discover the Higgs boson and explore the TeV energy scale Giacomo Artoni Seminario Finale Dottorato XXV Ciclo 5// 4
Working at a hadron collider Protons are not elementary particles, interactions happen at the parton level p E = E P x E = E P x p No knowledge on the system s initial energy or momentum along the beam y The event is closed kinematically only in the transverse plane! Projection of momentum on the xy plane (pt) and azimuthal angle (φ) Instead of the polar angle (θ), we use the pseudorapidity: η = ln(tan(θ/)) φ x Differences in pseudorapidity are invariant under boosts along the z axis! Giacomo Artoni Seminario Finale Dottorato XXV Ciclo 5// 5
The Higgs boson 95% CL Limit/SM χ 9 8 7 6 5 4 3 Tevatron LEP Exclusion 5 5 5 3 LEP Exclusion LEPATLAS Exclusion LEP 95% CL Tevatron 95% CL Theory uncertainty Tevatron Run II Preliminary, L. fb ATLAS Exclusion ATLAS Exclusion SM= Observed Expected w/o Higgs ± s.d. Expected ± s.d. Expected Tevatron G fitter SM Fit including theory errors Fit excluding theory errors ATLASCMS Exclusion M H Jul ATLASCMS Exclusion 3σ σ σ Missing piece of the Standard Model The Higgs mass is a free parameter of the theory Limits on the production cross section from: Indirect searches Global fit of the electroweak precision data collected at LEP Direct searches Using data from LEP, Tevatron and LHC CMS Exclusion ATLASCMS Exclusion June 3 4 5 6 7 8 9 (GeV/c ) Giacomo Artoni Seminario Finale Dottorato XXV Ciclo 5// 6
The low mass scenario Channel S/B Expected events σ( ) (GeV) * H ZZ( ) 4l 5 BR [pb] VBF H ± WH l bb s = 8TeV ± WW l qq WW l l LHC HIGGS XS WG * H WW( ) lνlν. 59 3 H γγ.3 9.6 ZZ l l qq ZZ l l ZZ l l l l H ττ. 6 poor H bb. 5 poor 3 4 ZH l l bb l = e, µ = e, µ, q = udscb tth ttbb 5 5 M H Giacomo Artoni Seminario Finale Dottorato XXV Ciclo 5// 7
H ZZ ( * ) 4l decay channel Very clean signature provided by the four leptons (electrons and muons) in the final state Needs high lepton acceptance, as well as high reconstruction and identification efficiencies Needs optimal energy and momentum resolution on e/µ Two different sources of background: a.u. /.5 GeV..8.6.4. Simulation = 3 GeV Gaussian fit H ZZ* 4µ ( s = 8 TeV) m = (9.5 ±.4) GeV = (.3 ±.4) GeV fraction outside ± : 6% without Z mass constraint * Irreducible (mainly for >m Z ): pp ZZ( ) 4l Reducible (especially for <m Z ): Zjets and tt Main improvements in and : Use of GSF electrons Extension of the analysis at low New background estimation methods Major issues for the analysis of data: Redo electron identification in order to ensure better performance (also with high pile up) Giacomo Artoni Seminario Finale Dottorato XXV Ciclo 5// a.u. /.5 GeV.7.6.5.4.3.. 8 9 3 4 5 Simulation = 3 GeV Gaussian fit H ZZ* 4e ( s = 8 TeV) m = (7.86 ±.7) GeV = (.76 ±.6) GeV fraction outside ± : 7% without Z mass constraint m 4l m 4l 8 9 3 4 5 8
Gaussian Sum Filter This new reconstruction algorithm for electron tracks can account for energy losses due to bremsstrahlung Energy loss is modeled by a sum of gaussians At each surface the track state is convolved with material effects This algorithm provides: Improved resolution for all track parameters belonging to the transverse plane Smaller dependence on the material in the Inner Detector Improved mass resolution ] Radiation length [X Measurement Surface.5.5.5 Simulation Services TRT SCT Pixel Beampipe Material Surface 3 3 Giacomo Artoni Seminario Finale Dottorato XXV Ciclo 5// 9
Gaussian Sum Filter This new reconstruction algorithm for electron tracks can account for energy losses due to bremsstrahlung Energy loss is modeled by a sum of gaussians At each surface the track state is convolved with material effects This algorithm provides: Improved resolution for all track parameters belonging to the transverse plane Smaller dependence on the material in the Inner Detector Improved mass resolution Resolution on d /σ(d )* from simulation ) for d / (d.4 Simulation s= 7 TeV. Z ee (Standard) Z ee (GSF).8.6.4..5.5.5.5.5.5 truth *d represents the closest distance in the xy plane between the track and the beam line Giacomo Artoni Seminario Finale Dottorato XXV Ciclo 5//
Gaussian Sum Filter This new reconstruction algorithm for electron tracks can account for energy losses due to bremsstrahlung Energy loss is modeled by a sum of gaussians At each surface the track state is convolved with material effects This algorithm provides: Improved resolution for all track parameters belonging to the transverse plane Smaller dependence on the material in the Inner Detector Improved mass resolution Arbitrary Units Arbitrary Units.4.< y <.5.5< y <.. Simulation.< y <.5 s = 7 TeV.5< y <...< y <.5.8.6.4. Standard electrons 8 6 4 4.7.6.5.4.3.. Simulation s = 7 TeV. < y <.5.5 < y <.. < y <.5.5 < y <.. < y <.5 (m m e e m e J/ PDG )/ e GSFrefitted electrons 6 4 4 (m m e e m e J/ PDG )/ e Giacomo Artoni Seminario Finale Dottorato XXV Ciclo 5//
Electron identification in Developed a new identification menu specifically for our analysis The so called Loose is the one used for the analysis of data This new menu has been called MultiLepton, since it can be used by all analysis with a high number of leptons in the final state An identification menu is a set of cuts on sensitive variables to discriminate between signal and background hadrons: π ± or K ± which fake the electron behavior conversions: real electrons but coming from the pair production of a photon The MultiLepton makes full use of the new features of GSF algorithm: this yields better and more stable performance! Giacomo Artoni Seminario Finale Dottorato XXV Ciclo 5// Efficiency Rejection on Hadrons.5.95.85 3 5 5 ATLAS Work in Progress Simulation s= 8 TeV version.9 of Loose version of Loose MultiLepton 5 5 5 3 ATLAS Work in Progress Simulation s= 8 TeV Loose () average (.98) Loose () average (.946) MultiLepton average (.946) Loose () average (4.5) Loose () average (3.6) MultiLepton average (4.8) n vtx 3 3 φ (rad)
Analysis strategy 6 Analysis Requirements (optimized for low mass Higgs): quadruplets: two oppositecharge, same flavor lepton pairs p T,,3,4 >, 5,, 6 (7) GeV for µ (e) mthr 47.5 35.5 4 6 8 m4l Leading dilepton (the one closest to m Z ) mass: 5 < m < 6 GeV Subleading dilepton mass: m thr (m 4l ) < m 34 < 5 GeV, m thr = 7.5 5 GeV Quadruplet is rejected if alternative sameflavor, opposite charged pair gives m ll <5 GeV ΔR ll >. if l and l have the same flavor, otherwise ΔR ll >. Track and calorimeter isolation, as well as d /σ(d ) cuts applied to all leptons Four different types of quadruplets: 4µ, 4e, µe and eµ (leading dilepton first) background contaminations (and thus estimations) are completely different if the subleading dilepton is a µ µ pair or an e e pair! We will focus on the estimation of the background in the µe and 4e channels Giacomo Artoni Seminario Finale Dottorato XXV Ciclo 5// 3
llee background estimation Categories from MC: the additional electrons can be: real electrons (e), hadrons or fakes in general (j), conversions (γ) or semileptonic decays of heavyflavor quarks (Q) Categorize electrons in data (using Transition Radiation Tracker, number of hits in Blayer* and energy in first sampling of the EM calorimeter) and MC (using truth information) Check (with e ± e and e ± e ± pairs) data/mc agreement when relaxing identification criteria on subleading pair and then extrapolate to signal region using efficiencies extracted from MC E = Electron C = Conversion F = Fake *innermost layer of the Inner Detector Giacomo Artoni Seminario Finale Dottorato XXV Ciclo 5// 4
llee background estimation Maximum likelihood fit to sensitive variables: use samecharge subleading dileptons * (in order to remove ZZ( ) contribution) relax analysis requirements on least energetic electron constrain Q contribution from MC use hadron/conversion discriminating variables to understand the composition in this control region extrapolate to the signal region using efficiencies derived for each category on MC Events Events/. 4 Data s = 8 TeV: Ldt = 5.8 fb Total f Q 8 6 4 3 5 5 3 ATLAS s = 8 TeV: Preliminary E = Electron C = Conversion F 5= Fake Ldt = 5.8 fb µe blayer n hits Data Total f Q µe.5..5..5.3.35.4 TRT Ratio Giacomo Artoni Seminario Finale Dottorato XXV Ciclo 5// 5
Background estimation control Enlarge signal region by removing isolation and d /σ(d ) cuts on subleading dilepton Normalize MC expectations to the background estimations made with data driven methods Control Data/MC agreement on m (leading dilepton mass) and m34 (subleading dilepton mass) distribution Events/4 GeV 8 6 4 µ µ /e e µ µ s = 7 TeV: Ldt = 4.8 fb s = 8 TeV: Ldt = 5.8 fb Data ZZ Zjets tt WZ Syst.Unc. Events/5 GeV 7 6 5 4 3 µ µ /e e µ µ s = 7 TeV: Ldt = 4.8 fb s = 8 TeV: Ldt = 5.8 fb Data ZZ Zjets tt WZ Syst.Unc. Events/4 GeV 8 6 4 µ µ /e e e e s = 7 TeV: Ldt = 4.8 fb s = 8 TeV: Ldt = 5.8 fb Data ZZ Zjets,tt Syst.Unc. Events/5 GeV 7 6 5 4 3 µ µ /e e e e s = 7 TeV: Ldt = 4.8 fb s = 8 TeV: Ldt = 5.8 fb Data ZZ Zjets,tt Syst.Unc. 6 8 m 4 6 8 m 34 6 8 m 4 6 8 m 34 Giacomo Artoni Seminario Finale Dottorato XXV Ciclo 5// 6
Results of the event selection Expected Background Expected Signal 5.±.8 5.3±.8 Data 3 Events/5 GeV 35 3 5 5 s s Data (*) Background ZZ Background Zjets, tt Signal (m =5 GeV) H Signal (m =5 GeV) H Signal (m =9 GeV) H Syst.Unc. ATLAS = 7 TeV: Ldt = 4.8 fb = 8 TeV: Ldt = 5.8 fb Preliminary (*) H ZZ 4l In the window < m 4l < 3 GeV 5 5 5 m 4l Giacomo Artoni Seminario Finale Dottorato XXV Ciclo 5// 7
95% CL limit on / SM (*) H ZZ 4l s=7 TeV, Ldt =4.8 fb s=8 TeV, Ldt =5.8 fb Observed CL s Expected CL s ± ± Exclusion plots 95% CL upper limits on the Standard Model Higgs boson production cross section as a function of, divided by the expected SM Higgs boson cross section 95% CL limit on / SM (*) H ZZ 4l s=7 TeV, Ldt =4.8 fb s=8 TeV, Ldt =5.8 fb Observed CL s Expected CL s ± ± 3 4 5 6 3 4 5 6 7 8 Expected exclusion : 464 GeV and 765 GeV Observed exclusion : 36 GeV and 746 GeV Giacomo Artoni Seminario Finale Dottorato XXV Ciclo 5// 8
Significance of the excess What is the probability that a backgroundonly experiment is more signallike than the observed data? Local p Obs. combined Exp. combined Obs. Exp. Obs. Exp. ATLAS (*) H ZZ 4l s=7 TeV: Ldt =4.8 fb s=8 TeV: Ldt =5.8 fb Local p Obs. combined Exp. combined Obs. Exp. Obs. Exp. ATLAS (*) H ZZ 4l s=7 TeV: Ldt =4.8 fb s=8 TeV: Ldt =5.8 fb σ σ σ σ 3 3σ 3 3σ 4 5 3 4 5 6 4σ 4 5 3 4 5 6 7 8 4σ Expected deviations for ~5 GeV.5σ.σ.6σ Observed deviations for ~5 GeV.3σ.7σ 3.4σ Giacomo Artoni Seminario Finale Dottorato XXV Ciclo 5// 9
Signal strength Signal strength (µ) 4 3.5 3.5.5.5.5 Best Fit ln λ(µ) < s = 7 TeV: Ldt = 4.8 fb s = 8 TeV: Ldt = 5.8 fb 3 4 5 6 7 8 Signal Strength (µ) is defined as the ratio between the observed signal rate fitting data and the rate expected from the Standard Model at a given µ 4 3.5 3.5.5.5 Ldt = 4.8 5.8 fb s = 7 and 8 TeV (*) H ZZ llll best fit ln (m ln (m,µ) <.3 H,µ) < 6. H 4 6 8 3 In the D profile likelihood fit to signal strength and the best fit is given by: = 5 GeV µ =.3 ±.6 Giacomo Artoni Seminario Finale Dottorato XXV Ciclo 5//
Combined ATLAS results Local p 3 4 (*) ATLAS (a) H ZZ 4l σ 3 σ 95% CL Limit on µ ATLAS s = 7 TeV: Ldt = 4.64.8 fb s = 8 TeV: Ldt = 5.85.9 fb ± σ ± σ Observed Bkg. Expected Local p Local p 5 6 7 3 4 5 6 7 3 4 5 6 4 σ 5 σ s = 7 TeV: Ldt = 4.8 fb s = 8 TeV: Ldt = 5.8 fb σ 4 σ Exp. Obs. Exp. Obs. (b) H γ γ 4 σ 5 σ σ 3 σ s = 7 TeV: Ldt = 4.8 fb s = 8 TeV: Ldt = 5.9 fb (*) (c) H WW lνlν 3 σ 7 5 σ Exp. s = 7 TeV: Ldt = 4.7 fb 8 Obs. s = 8 TeV: Ldt = 5.8 fb 9 5 5 3 35 4 45 Local p Signal strength (µ) 3 4 5 6 7 8 9.5.5.5 (a) (b) (c) 5 CL s Limits 3 4 5 Sig. Expected Observed Observed ln λ(µ)< σ 3 σ 4 σ 5 σ 6 σ 3 4 5 Giacomo Artoni Seminario Finale Dottorato XXV Ciclo 5//
What have we found? Consistent excess of events in the best three channels lead to the discovery of this new particle It is a boson, since it decays into pairs of bosons It is neutral, since it decays in neutral final states It has spin, since it decays into pairs of photons (LandauYang theorem) Its production rate is compatible with the one of the Higgs Can we say it is the Higgs boson? Does it decay into fermions? What are its spin and parity? ATLAS W,Z H bb s = 7 TeV: Ldt = 4.7 fb H H WW H H ZZ (*) (*) s = 7 TeV: Ldt = 4.8 fb s = 8 TeV: Ldt = 5.9 fb Combined s = 7 TeV: s = 8 TeV: s = 7 TeV: Ldt = 4.64.7 fb l l s = 7 TeV: Ldt = 4.7 fb s = 8 TeV: Ldt = 5.8 fb 4l s = 7 TeV: Ldt = 4.8 fb s = 8 TeV: Ldt = 5.8 fb Ldt = 4.6 4.8 fb Ldt = 5.8 5.9 fb µ =.4 ± More data are needed to answer these questions, stay tuned....3 = 6. GeV Signal strength (µ) Giacomo Artoni Seminario Finale Dottorato XXV Ciclo 5//
Conclusions * The H ZZ( ) 4l channel is considered the golden channel for the Higgs search Low rate but clean signature, S/B and very good invariant mass resolution We improved the ATLAS results by working on: The new electron reconstruction algorithm, Gaussian Sum Filter The new identification menu for electrons in, MultiLepton The new background estimation techniques for the µe and 4e channels * The H ZZ( ) 4l analysis is able to observe a 3.4σ significant excess at 5 GeV The combination of all ATLAS searches achieves a significance of 5.9σ We have discovered a new particle! We need more data from the LHC to confirm if this is the Standard Model Higgs boson... Giacomo Artoni Seminario Finale Dottorato XXV Ciclo 5// 3