Search for the SM Higgs Boson in H γγ with ATLAS

Transcription

1 Search for the SM Higgs Boson in H γγ with Heberth Torres, LPNHE - Paris on behalf of the Collaboration Second MCTP Spring Symposium on Higgs Boson Physics University of Michigan 17/4/1 1

2 Introduction H γγ is one of the most sensitive channels at low mass at the LHC. The only region not excluded after LEP, Tevatron, and early LHC Higgs searches. This is also the mass range favored by the Electroweak SM fit LEP 95% CL Tevatron 95% CL neglects correlations G fitter SM AUG OR CMS 95% CL AND CMS 99% CL 6 4 Theory uncertainty Fit including theory errors Fit excluding theory errors M H 1

3 SM Higgs Boson production & decay g g (pp H+X) [pb] (a) gg H 1 H pp H (NNLO+NNLL QCD + NLO EW) pp qqh (NNLO QCD + NLO EW) pp WH (NNLO QCD + NLO EW) pp ZH (NNLO QCD +NLO EW) Higgs boson production at the LHC q q (b) VBF q q H q q (c) V H s= 7 TeV W/Z H LHC HIGGS XS WG g g (d) t th gg H is the dominant process, known at NNLO, still with large uncertainty ~15% VBF has a distinctive signature, two forward jets and higher Higgs pt, known at ~NNLO with ~5% uncertainty gg H VBF t t H Higgs boson decay H γγ: low BR ~.%, but clear signature σxbr ~4 15 GeV BR [pb] s = 7TeV - H + SM ± WW l qq + - WW l l ZZ l l qq - VBF H + l = e, µ ± WH l bb = e, µ, + - ZH l l bb q = udscb H γγ decay through loops ZZ l l + - ZZ l l l + - l M H LHC HIGGS XS WG 11 pp tth (NLO QCD) - W/ZH tth M H 3

4 Simple signature channel Event selection: - γ trigger ET > GeV - Data quality requirements - Two photons: E 1 > 4 GeV HCal ID ECal E > 5 GeV η < 1.37 & 1.5 < η <.37 Calorimeter-based identification Isolated from hadronic activity

5 3 = 36cm= 6 σγ L 5 pb4s 1 cm s 1 6 nbc Nevtf = n 35 pb bc = 6 nbc f nbc 4s 1 35 (8)pb (8) First sampling very finely segmented in η: precision η measurem pb cm s cm pb cm s cm= 6 Photon reconstruction and π rejection. = 6 s 35 pb s 35 pb (9) (9) End of the EM shower Photons are seeded by clusters in the EM E T,EM () calorimeter. ET = ET = ET,EM () A lead/liquid-argon TEM calorimeter TE T,electron E E (11) ut =ut = E (11) T,electron with accordion geometry (crack-less) % σe σe%.7%, η < 3. E.7%, η < 3. E E E %(17%) Currently %(17%) 1.%(1.8%) 1.%(1.8%) barrel (endcap) E E (1) (1) Layer 3:.5x.5 Layer :.5x.5 Main energy deposit (13) (13) φ Photon reconstruction efficiency ~98% γ/π separation Layer 1:.3x.1 η EM Calorimeters Early EM showers correction Scott Snyder (BNL) Presampler:.5x.1 Clusters are classified as electrons, converted or photons to trackingjuly, Electron/photon unconverted performance at accordingichep information. Reconstruction of a photon producing an electron-positron pair (photon conversion) 5

6 Photon Identification Photon ID: based on longitudinal and lateral shower shape Cuts on 9 discriminant variables The fine granularity in S1 7 allow high rejection of photon pairs from πφ Calorimetric isolation: Sum of E of calo cells in a cone ΔR<.4, excluding the cluster cells. Out-of-core energy leakage corrections. Ambient energy correction event per event (to reduce the effects of underlying event and pile-up). Cone Ring: 6 S3 S s = 7 TeV, " Ldt = 878 nb (Uncalibrated) sum of cells outside of 5 7 central core: In this case: R = R= η + φ <.4 φ + η <.4 Need to correct for out-of-core leakage Also need to account for non-perturbative effects... Calorimeter Isolation Data Simulation S1 (fake!) Simulation (isolated prompt!) Simulation (non-iso prompt! ) PS.1 Entries/1 GeV Need to correct for out-of-core 4 leakage.1 η Entries/1 GeV π Data Simulation (fake!) Simulation (isolated prompt!) Simulation (non-iso prompt!) -5 5 s = 7 TeV, 15 Ldt = 878 nb (Uncalibrated) sum of cells outside of 5 7 central core: In this case: R = φ + η <.4 experimental cut (3 GeV) Also need to account for non-perturbative effects... s = 7 TeV, " Ldt = 878 nb Data Isolation Simulation (fake ) Simulation (isolated prompt ) Simulation (non-iso prompt ) Cut: ET isol < 5 GeV M. Hance 14 / 44 Les Houches Winter Workshop- 16 February 11 nce 14 / 44 Les Houches Winter Workshop- 16 February Isolation

7 Signal efficiency and yields Using several MC generators and Geant4 for full detector simulation. gg H & VBF Powheg interfaced to Pythia for showering and hadronization. W/ZH & tth Pythia. Photon ID efficiency: MC Shower shape variables shifted to match the data. Efficiencies cross-checked in data with Z ee and Z llγ (l = e,μ) Isolation efficiency also cross-checked/corrected to match data (Z ee studies) Gluon fusion re-weighted for destructive interference between gg γγ and gg H γγ. Gluon fusion events re-weighted to match Higgs boson p T froqt σ BR [fb] Signal events Efficiency [%]

8 Photon energy calibration The energy is calibrated separately for converted and unconverted photons. Calibration based on MC simulation of the detector (calorimeter simulation tuned in test-beam). Good material description verified with in situ measurements Then, an η dependent correction is applied. It is obtained from a global fit to Z ee data. Corrections ~ ±1% The MC is corrected to match the resolution in data. 8 Events / 1 GeV Entries / mm Preliminary Data -.66 < < -. MC conversion candidates MC true conversions Preliminary Data 11, data =1.76 ±.1 GeV s=7 TeV, Ldt = 4.6 fb MC =1.59 ±.1 GeV <.47 R [mm] Data Fit result Zee MC m ee

9 Diphoton angle reconstruction /.5 GeV 1/N dn/d The resolution of the angle measurement is dominated by the reconstruction of the primary vertex z position (the IP spread of ~5.6 cm would add ~1.4 GeV in mass resolution) Combine information from both photons. Use conversion track IP or the calorimetric pointing in the case of unconverted photons. Preliminary *=1 m gg H, m =1 GeV H Calorimetric pointing use the shower barycenters in the 1st and nd layer. z resolution with pointing ~15mm in the barrel More robust against pileup than using the recoil tracks Fit truth vertex p T Calo/Conv pointing Entries 15 Data 11 MC () Preliminary Ldt = 1.8 fb s = 7 TeV z z CaloPointing [mm]

10 Signal invariant mass resolution The signal mass peak is modeled with a Crystall Ball core + a small amplitude Gaussian to describe the tails. Expected FWHM = 4.1 GeV and σcb = 1.7 GeV The resolution changes depending on the photon η position and conversion Both photons unconverted /.5 GeV 1/N dn/d status, so we separate in categories ~15% improvement of the sensitivity Both central photons η <.75 H = 1 GeV Unconverted central, low p CB = 1.4 GeV FWHM = 3.4 GeV 3.4 GeV /.5 GeV 1/N dn/d H = 1 GeV Unconverted rest, low p Rest CB = 1.7 GeV FWHM = 4. GeV 4. GeV Full Width at Half Maximum (FWHM) quoted in red At least one photon in the barrel-endcap transition 1.3 < η < 1.75 At least one converted /.5 GeV 1/N dn/d H = 1 GeV Converted central, low p CB = 1.6 GeV FWHM = 3.9 GeV /.5 GeV 1/N dn/d H = 1 GeV CB =. GeV FWHM = 4.7 GeV 3.9 GeV.4 Converted rest,.3.3 Converted transition 4.7 GeV. 5.9 GeV..1.1 /.5 GeV 1/N dn/d.6.5 H.4 = 1 GeV CB =.3 GeV FWHM = 5.9 GeV

11 p γγ discriminant variable p γγ : component of pt γγ orthogonal to the thrust axis of the diphoton pair. p T p p Tl p T p T 1 thrust axis t = pγ 1 T pγ T p γ 1 T pγ T VBF and W/ZH production have a harder p γγ spectrum than gg H and the background. Gluon fusion have also larger p γγ tail than the background. Categorizing events in low and high p γγ (separating at 4 GeV). It give 5-% improvement of sensitivity depending on mass hypothesis 11

12 Categorization Splitting in 9 categories according to: photon conversion status η position of photons exploits also signal-background kinematic differences p γγ It gives categories with different mass resolution and signal/background ratio. Category σ CB FWHM N S N D S/B η - conversion categories [ [ [ [ [ Unconverted central, low p Unconverted central, high p Unconverted rest, low p Unconverted rest, high p Converted central, low p Converted central, high p Converted rest, low p Converted rest, high p Converted transition All categories Cats. with largest S/B ratio 1

13 Diphoton sample The results presented correspond to the 11 data, 4.9 fb after DQ requirements. 489 diphoton candidates ( < mγγ < 16 GeV) Expected 7 signal events at 1 GeV Large background smoothly decreasing Events / GeV Selected diphoton sample Data 11 Background model SM Higgs boson m H = 1 GeV (MC) 5 4 s = 7 TeV, 3 Data - Bkg m

14 Background processes q q q Irreducible: diphoton QCD production:! q! g g!!! Reducible: One or two jets fragmenting into a leading neutral hadron (mainly π ) faking photons. Jet rejection O(5). γ-jet events Dijet events " q! q g! Drell-Yan events, when both electrons are misidentified as converted photons. 14 σ (pb) R R jj γj γγ H γγ 1

15 Background decomposition Extraction of 4 background components, γγ, γ-jet, jet-γ and jet-jet, through different data-driven methods. D fit of the two photons calo-isolation distributions (PDFs obtained from data) Events / ( 1 GeV ) Events / ( 1 GeV ) Preliminary Data 11, s = 7 TeV, Ldt = 1.8 fb,1, > 4 GeV, E > 5 GeV E T Preliminary Data 11, s = 7 TeV, Ldt = 1.8 fb,1, > 4 GeV, E > 5 GeV E T T j j+jj j +j+j+jj T j+jj +j+j+jj leading γ j j+jj +j+j+jj iso E T,1 sub-leading γ xd sideband, one for the leading and another one for the sub-leading. "j!" #" C Photon ID "" jj jx!" #" "j $" %" D $" A %" B γx "j "j non-isolated candidates jx jj jx jj Isolation B$F!! G$'+-3& 41&34.$ N A sig = candidates failing photon ID Very low correlation between isolation and photon ID N A " N B # N C N D iso E T, 15

16 Background decomposition results γγ γj jj Drell-Yan Events 16 ± 1 53 ± ± ± 8 Fraction (71 ± 5) % (3 ± 4) % (5 ± 3) % (.7 ±.1) % The results from the different decomposition methods are in good agreement. The Drell-Yan background is also estimated through a data-driven method. (71 ± 5) % pure γγ The background composition is extracted as a function of the mass bin by bin. ] [GeV dn/d Preliminary Data 11 s = 7 TeV, +DY data j data jj data Statistical error Total error

17 Data mass distribution Events / GeV Events / GeV Events / GeV s = 7 TeV, Unconverted central, low p Data 11 Exponential fit s = 7 TeV, Unconverted rest, high p Data 11 Exponential fit s = 7 TeV, Converted rest, low p Data 11 Exponential fit Exponential background model Events / 4 GeV Events / GeV Events / GeV s = 7 TeV, Converted central, low p Data 11 Exponential fit s = 7 TeV, Unconverted central, high p Data 11 Exponential fit s = 7 TeV, Converted rest, high p Data 11 Exponential fit Events / GeV Events / 4 GeV Events / GeV s = 7 TeV, s = 7 TeV, Converted central, high p Data 11 Exponential fit s = 7 TeV, Unconverted rest, low p Data 11 Exponential fit Converted transition Data 11 Exponential fit

18 Background modeling Differences between the true background shape and the exponential model would bias the fitted signal strength and the test statistic. Background underestimation induces fake excesses and prevents limit setting. Background overestimation masks out-going excesses and induces fake limits. Possible bias estimation: It has been estimated by using MC Resbos samples (including photon efficiency effects). The possible bias is also cross-checked fitting the data sample with functions with more degrees of freedom than the single exponential. They are estimated to be ±(.1-7.9) events depending on the category. It is taken into account in a conservative way: Add a systematic uncertainty to the background model. Include a term in the likelihood function that allows a signal-like component in the background. Its normalization is left floating and is gaussian constrained to this uncertainty in each category. 18

19 uncertainties Systematic Signal event yield Photon reconstruction and identification ±11% Effect of pileup on photon identification ±4% Isolation cut efficiency ±5% Trigger efficiency ±1% +1 8 % Higgs boson cross section (scales) Higgs boson cross section (PDF+α s ) ±8% Higgs boson p T modeling ±1% Luminosity ±3.9% Signal mass resolution Calorimeter energy resolution ±1% Photon energy calibration ±6% Effect of pileup on energy resolution ±3% Photon angular resolution ±1% Signal mass position Photon energy scale ±.7 GeV Signal category migration Higgs boson p T modeling ±8% Conversion rate ±4.5% Background model ± (.1 7.9) events Dominant experimental uncertainty on the signal yields. Cross-checked on data with Z llγ (l = e,μ) Dominant uncertainty on the mass resolution. Determined on data with Z ee and J/Psi ee events. Uncertainty on the Higgs mass position due to the imperfect knowledge on the photon energy scale. From PDFs and scale variations in HqT From uncertainty on the amount of material in front of the calorimeter The systematic uncertainties are taken into account by introducing nuisance parameters with constraints in the likelihood function. All the systematic uncertainties are treated as fully correlated between the nine categories, except for the background uncertainty. 19

20 Results The largest excess is observed at 16.5 GeV: Local significance of.8σ, Global significance of 1.5σ (probability to observe a background fluctuation equivalent or larger anywhere in the explored mass range 15 GeV) Signal strength μ ~ σsm, considering the uncertainty, it is still compatible with the SM prediction. Events / GeV Data - Bkg Selected diphoton sample Data 11 Background model SM Higgs boson m H = 1 GeV (MC) s = 7 TeV, m µ 3 Best fit H ± 1 1 Local p Observed p 1-1 SM H expected p Data 11, s = 7 TeV - -3 Data 11, s = 7 TeV Observed p (with energy scale uncertainty)

21 Results Expected limit with 95% CL: σsm. Observed limit with 95% CL: σsm. Excluded with 95% CL: GeV and GeV. 95% CL limit on / SM Observed CL s limit Expected CL s limit ± 1 H ± Data 11, s = 7 TeV

22 Conclusions and Outlook Using a nine category analysis, taking into account the conversion status and η position of photons, and the p of the diphoton pair. H γγ alone starts setting exclusion limits with 95% CL: GeV and GeV. An excess is observed at 16.5 GeV, with.8σ of local significance and 1.5σ of global significance, compatible with a SM Higgs boson. With the data expected to be collected in 1 at 8 TeV (15- fb ), this channel will have the sensitivity to exclude or observe a SM Higgs boson at low mass.

23 References Collaboration, Search for the Standard Model Higgs boson in the diphoton decay channel with 4.9 fb of pp collisions at sqrt(s)=7 TeV with, Phys.Rev.Lett. 8 (1). Collaboration, Search for the Standard Model Higgs boson in the two photon decay channel with the detector at the LHC, Phys.Lett.B 75 (11). LHC Higgs Cross Section Working Group, S. Dittmaier, C. Mariotti, G. Passarino, R. Tanaka (Eds.), Handbook of LHC Higgs Cross Sections, CERN- 11- arxiv: (11) and arxiv: (1). and CMS Collaborations, LHC Higgs Combination Group, Procedure for the LHC Higgs boson search combination in summer 11, ATL-PHYS- PUB-111 (11). 3

24 Backup 4

25 Signal invariant mass resolution /.5 GeV 1/N dn/d H = 1 GeV Unconverted central, low p CB = 1.4 GeV FWHM = 3.4 GeV /.5 GeV 1/N dn/d.1.1 H m.8 H = 1 GeV.6.4. Unconverted central, high p CB = 1.4 GeV FWHM = 3.3 GeV /.5 GeV 1/N dn/d H = 1 GeV Unconverted rest, low p CB = 1.7 GeV FWHM = 4. GeV /.5 GeV 1/N dn/d H = 1 GeV Unconverted rest, high p CB = 1.6 GeV FWHM = 3.9 GeV /.5 GeV 1/N dn/d H = 1 GeV Converted central, low p CB = 1.6 GeV FWHM = 3.9 GeV /.5 GeV 1/N dn/d H = 1 GeV Converted central, high p CB = 1.5 GeV FWHM = 3.6 GeV /.5 GeV 1/N dn/d H = 1 GeV Converted rest, low p CB =. GeV FWHM = 4.7 GeV /.5 GeV 1/N dn/d.8.7 H.6 = 1 GeV Converted rest, high p CB = 1.9 GeV FWHM = 4.5 GeV /.5 GeV 1/N dn/d.6.5 H.4 = 1 GeV.3..1 Converted transition CB =.3 GeV FWHM = 5.9 GeV

26 Pileup robustness Mass resolution robust against pileup Background composition extracted with data-driven technique. Equivalent purities in low and high pileup data taking periods /.5 GeV 1/N dn/d Preliminary gg H, m =1 GeV H Fit µ 6 6 < µ 9 < µ µ > ] Recorded Luminosity [pb 4 Online 11, s=7 TeV Ldt=5. fb * = 1. m, <µ> = 11.6 * = 1.5 m, <µ> = 6.3 Events/pb Preliminary Data 11 s = 7 TeV, +DY data j data jj data Stat.+syst. error *=1.5 m *=1. m All Mean Number of Interactions per Crossing 6 Period

27 Signal strength µ 3 Best fit H ± Data 11, s = 7 TeV µ 5 Best fit ± 1 Data 11, s = 7 TeV -5 - Unconv. Central low p Unconv. Central high p Unconv. Rest low p Unconv. Rest high p Conv. Central low p Conv. Central high p Conv. Rest low p Conv. Rest high p Conv. Transition H

28 Local p Excess at 16.5 GeV Observed p 1 SM H expected p Data 11, s = 7 TeV Observed p Observed p Expected SM p Expected SM p (-conversion) (Inclusive) (-conversion) (Inclusive)

29 at 16.5 GeV Excess The γγ excess is in coincidence with a smaller excess observed in the four leptons channel in. Local p 1 11 Preliminary - L dt ~ fb -3 3 s = 7 TeV Obs. Comb. (ESS) -4 Exp. Comb. Exp. H Exp. H bb -5 Obs. Comb. Exp. H 4l Obs. H 4l Obs. H Exp. H ll Obs. H ll Obs. H bb Exp. H Obs. H