Yoshikazu NAGAI (CPPM, Aix Marseille Université)

Transcription

1 Search for the Standard Model Higgs boson decaying into a b quark pair at ATLAS Yoshikazu NAGAI (CPPM, Aix Marseille Université) CPPM Seminar October 20, 2014

2 Contents Introduction Analysis overview Result of a search for the H bb Summary & Prospects for the upcoming high luminosity LHC 2

3 July 4th, 2012 H Observation of a new boson Observation of a new particle at a mass of around 125 GeV by ATLAS (5.1 ) & CMS (5.0 ) September, 2012 October 8th, 2013 Nobelprize.org PLB 716(2012)

4 Decay modes of this new boson ATLAS CMS H YES! YES! H ZZ (*) YES! YES! Bosonic decay modes H WW (*) YES! YES! H Evidence (4.5 ) Evidence (3.2 ) H bb?? Lepton Quark Fermionic decay modes This seminar reviews a search for the H bb decay mode 4

5 Is this new boson the SM Higgs boson? property measurements Mass (H, ZZ 4l) Phys. Rev. D. 90, (2014) G fitter: Combined (H and H ZZ ( * ) 4l): Direct measurements of top quark and W mass are compatible with the EWK global fit including Higgs mass in the fit 5

6 Is this new boson the SM Higgs boson? property measurements Spin/Parity (H, ZZ ( * ) 4l, WW ( * ) l l ) Phys. Lett. B. 726 (2013) Spin 0 + vs H Spin 2 + JP = 0, 1 +, 1, 2 + hypotheses are excluded > 2 (For spin 2 hypothesis, we tested one benchmark scenario) 6

7 Is this new boson the SM Higgs boson? property measurements Couplings (all production modes, all decay modes) ATLAS CONF (2014) No significant deviations from SM prediction Strong constraint from bosonic decay modes Caveat: H bb does not contain latest results = ( x Br) measured / ( x Br) SM 7

8 Does the new boson decay into a b quark pair? Branching ratio of the SM Higgs boson H H It predominantly decays to a b quark pair, if the new boson is the SM Higgs boson 8

9 Previous studies on VH Vbb ATLAS CMS Tevatron 2011 (7TeV) 2011 (7TeV) (8TeV) Upper limit: < 4.6 x SM Expected limit: < 4.0 x m H = 125 GeV Some evidence? but has yet to be confirmed Signal m H = 125 GeV Obs. 2.1 Exp. 2.1 Signal m H = 125 GeV Obs. 2.8 Exp

10 Higgs boson productions at LHC 8 TeV H = 125 GeV ggh = pb ggh VBF = pb VBF WH ZH bbh tth WH = pb ZH = pb bbh = pb tth = pb 10

11 How do we look for H bb at LHC? QCD (di) jets production ggh, H bb VBF, H bb Huge QCD jet production, > 4 5 orders of magnitude Very tough at the LHC 11

12 How do we look for H bb at LHC? W or Z + di jets production tt production Better S/B Can suppress QCD jets with (W/Z) leptonic decays Will present in this seminar WH/ZH, H bb 12

13 How do we look for H bb at LHC? tt with >= 4 jets Not enough sensitive in Run 1 Important to address Higgs Yukawa coupling for Run2 & beyond Not covered in this seminar bbh/tth, H bb 13

14 LHC and integrated luminosity proton proton collisions at 7 TeV (2011) and 8 TeV (2012) The peak instantaneous luminosity at 8 TeV is 7.7 x cm 2 s 1 LHC ATLAS recorded: > 5.0 fb 1 (2011) and > 23 fb 1 (2012) In this seminar, full data analysis is presented 14

15 A challenging environment 2011: 11 vertices with Z 2012: 25 vertices with Z Number of pile up = (Cross section) [cm 2 ] x (Luminosity) [cm 2 s 1 ] x (bunch spacing) [s] (50 ns!!) Need very good resolution to resolve vertices & very fast trigger system to keep interesting events 15

16 A Toroidal LHC Apparatus (The ATLAS detector) p p The ATLAS coordinate z r x y Polar angle is expressed by pseudorapidity = ln tan( /2) Person 16

17 The ATLAS detector (cont d) Inner tracking detectors & solenoid magnet <2.5 B Silicon (Pixels & strip) Transition radiation straw chamber Superconducting solenoid magnet For H bb analysis, transverse impact parameter resolution is very crucial for b jets identification (d 0 ) ~ 20 m 17

18 Calorimeters The ATLAS detector (cont d) particles EM calorimeter (Liq. Ar+Pb) Had calorimeter (Fe+scintillator, Liq.Ar+Cu) Gap less structure allows good object measurements for jets/electrons and missing transverse energy (E T miss ) 18

19 The ATLAS detector (cont d) Muon spectrometers and Toroid magnet B Consists of fast trigger chambers and position sensitive chambers Superconducting coils of toroid magnets Both independent/combined with ID measurement is possible for muon 19

20 The trigger system L1: hardware based use calorimeter and muon detector information define Regions of Interest (coarse granularity) 50ns (20MHz) ~1.5 MB/event HLT: consists of L2 & EF L2: Software based (Special fast algorithms) use full detector granularity inside RoI including tracker information EF: Software based (Offline algorithms) use full detector granularity 20

21 Contents Introduction Analysis overview Result of a search for the H bb Summary & Prospects for the upcoming high luminosity LHC 21

22 Signal categorization 0 lepton 1 lepton 2 lepton Decay Z W e / W Z ee / Z Branching fraction 20% 11% / 11% 3.3% / 3.3% Number of leptons Signal yield (20fb 1, 8 TeV) ~960 ~900 / ~900 ~160 / ~160 m H = 125 GeV Before any selection applied 22

23 How does the event look like? 2 lepton candidate 0 lepton candidate leptons (e, ) jets from b quark missing energy ( ) 1 lepton candidate Yoshikazu NAGAI (CPPM) CPPM Seminar 23

24 Binning of the signal regions Signal binning Separate signal regions based on S/B : maximize analysis sensitivity The idea is to split the analysis in bins of jet multiplicity and vector boson p T S/B Number of jets 2 jets 3 jets 0 lepton S/B lepton S/B lepton S/B p T V 0 lepton 1 lepton 2 lepton 24

25 Two types of analysis Event Selections Dijet mass analysis: the mass of the dijet system of b jets (m bb ) as the final discriminating variable (Cross check of MVA result) Multivariate analysis (MVA): Boosted decision tree (BDT) combines various kinematic variables in addition to the m bb (Nominal result) Basically focus on MVA analysis in this seminar, except new features compared to the previous ATLAS analysis Common selection R cut values are optimized for the angular separation between two jets, as a function of W/Zp T MVA p TV (GeV) ( * ) > 120 R(jet1, jet2) >0.7 ( * ) 0 lepton selection is only GeV with m bb analysis W/Z W/Z b jet b jet b jet b jet 25

26 1 lepton selection Event Selections (cont d) H T and E T miss cuts to suppress tt and QCD multjet background MVA p TW (GeV) > 120 trigger primary: single e/ secondary: E T miss m TW (GeV) ( * ) ( * ) MVA analysis uses m T W as an input variable H T (GeV) > 180 E T miss (GeV) > 20 2 lepton selection m ll cut to select on shell Z boson candidates MVA p TW (GeV) > 120 trigger single e/ or di lepton (ee/ ) m ll (= m Z ) (GeV) Single trigger Offline p T (GeV) E T miss trigger Offline E miss T (GeV) New strategy to gain signal E miss T trigger 26

27 0 lepton selection Event Selections (cont d) Several angular cuts applied to suppress QCD multijet background m bb MVA p TV (= E miss T ) (GeV) > 120 trigger E T miss trigger E T miss trigger R(jet1, jet2) > 0.7 (for p TV < 200) track E miss T (GeV) > 30 > 30 (E miss T, p miss T ) < /2 < /2 min[ (E miss T, jet)] > 2.2 > 2.8 (jets p T ) (GeV) > 120 (2 jets) N/A (3 jets) > 120 (2 jets) > 150 (3 jets) Further QCD MJ cuts applied New signal category for the latest analysis Former analysis only for E T miss > 120 GeV ZH bb E T miss QCD multijet b jets b jets Track E T miss Fake E T miss OR Fake track E T miss 27

28 b jet identification Separate b quark jet (b jet) from other flavor jets (light jet, charm jet) b quark jet identification (b tagging) performance is crucial for H bb analysis Secondary vertices Primary vertex O(mm) B hadrons (B 0, B +, etc ) travels a few mm before it decays with unique topology Displaced vertex Large impact parameter (d 0 ) Cascade structure (b c x primary secondary tertiary) soft e/ from semi leptonic decay of B hadrons (~40%) MVA to combine information all together 28

29 b jet identification an artificial neural network algorithm: MV1 (MV1c) light jet vs b jet W+b/c/l response b jet eff. 50% c jet eff. 4% l jet eff. < 0.1% b jet eff. 80% c jet eff. 33% l jet eff. 3% Separate signal regions based on MV1c 29

30 b jet identification an artificial neural network algorithm: MV1 (MV1c) light jet vs b jet W+b/c/l response b jet eff. 50% c jet eff. 4% l jet eff. < 0.1% b jet eff. 80% c jet eff. 33% l jet eff. 3% Separate signal regions based on MV1c 30

31 b jet identification an artificial neural network algorithm: MV1 (MV1c) light jet vs b jet W+b/c/l response b jet eff. 50% c jet eff. 4% l jet eff. < 0.1% b jet eff. 80% c jet eff. 33% l jet eff. 3% Separate signal regions based on MV1c 31

32 Shape modeling with MC simulation Modeling overview Z QCD multijets (fake lepton, fake MET) from data 32

33 Shape modeling with MC simulation Modeling overview QCD multijets (fake lepton, fake MET) from data 33

34 Shape modeling with MC simulation Modeling overview QCD multijets (fake lepton, fake MET) from data 34

35 Shape modeling with MC simulation Modeling overview QCD multijets (fake lepton, fake MET) from data 35

36 Shape modeling with MC simulation Modeling overview QCD multijets (fake lepton, fake MET) from data 36

37 Background modeling Modeling of p TV is crucial: sub divide signal region based on p T V correction applied correction applied 37

38 Background modeling Modeling of p TV is crucial: sub divide signal region based on p T V correction applied Shape uncertainty p TV dependent modeling collection correction applied Treat similarly for other background 38

39 Improve m bb resolution The most important input for MVA analysis Muon in jet collection Kinematic likelihood fit in 2 lepton channel (no intrinsic E T miss except for b semileptonic decay) total resolution improvement ~30% 39

40 BDT optimization Input variables are optimized for each signal region separately 0 lepton: 10 (12) variables for 2 jet (3 jet) bin Including 1 lepton: 11 (13) variables for 2 jet (3 jet) bin m bb 2 lepton: 12 (14) variables for 2 jet (3 jet)bin p T V R(b 1,b 2 ) BDT 40

41 Fit model Binned maximum likelihood fit performed on the BDT across regions determine signal yield & background normalization Impact of syst. uncertainties described by nuisance parameters across regions constrain systematic uncertainties through global fit 0 lepton 1 lepton 2 lepton 0 lepton 1 lepton 2 lepton Constrain physics background through lepton categories (e.g. 0 lepton Z+jets constrained from 2 lepton) 41

42 Fit model Binned maximum likelihood fit performed on the BDT across regions determine signal yield & background normalization Impact of syst. uncertainties described by nuisance parameters across regions constrain systematics through global fit 2 jet 3 jet Constrain background flavor composition (e.g. b tagging syst. for b/c/light flavor and Z+bb/cc/light normalization ) Constrain 2 jets vs 3 jets difference 42

43 Fit model Binned maximum likelihood fit performed on the BDT across regions determine signal yield & background normalization Impact of syst. uncertainties described by nuisance parameters across regions constrain systematics through global fit p T V correction applied Shape uncertainty p TV dependent modeling collection correction applied Constrain background modeling of p T dependence (e.g. modeling of W+jets ttbar as a function of p T (V) ) 43

44 Fit model Binned maximum likelihood fit performed on the m jj /BDT across regions determine signal yield & background normalization Impact of syst. uncertainties described by nuisance parameters across regions constrain systematics through global fit 0 lepton 1 lepton 2 lepton 2 jet x x x p V T jet Fit performed through the BDT bins 44

45 Contents Introduction Analysis overview Result of a search for the H bb Summary & Prospects for the upcoming high luminosity LHC 45

46 Our analysis techniques are valid for H bb signals? 46

47 Analysis validation with di boson signal Validation of analysis techniques is very important Can use di boson signals (WZ, ZZ), which produces exactly same final states Well established in the SM and can be used as the standard candle Z 0 lepton 1 lepton 2 lepton BDT (VZ) BDT (VZ) BDT (VZ) Data and background+signal yield are compatible 47

48 Di boson signal strength = ( x Br) measured / ( x Br) SM Results are compatible with the SM. Successfully re observed di boson process. Signal excess Observed: 4.9 Expected:

49 Let s go Good luck! 49

50 Fit results (8 TeV, MVA analysis) 0 lepton 1 lepton 2 lepton BDT (VH) BDT (VH) BDT (VH) Data and background + signal yields are compatible Background normalizations are expressed as scale factors to the pre fit normalization 50

51 Fit results (8 TeV) sorted as a function of S/B S/B = 8e Pull of data compared to BG only or BG+signal Both hypotheses look compatible with data 51

52 H bb signal strength (7 TeV + 8 TeV) = ( x Br) measured / ( x Br) m H = GeV 52

53 Upper limit & significance (7+8 TeV) Observed limit: < 1.2 x SM Expected limit: < 0.8 x m H = 125 GeV Signal excess Observed: 1.4 Expected: m H = 125 GeV Close to 3 sensitivity achieved in Run1 53

54 Systematics uncertainties What is the main source of systematic? 54

55 Systematics uncertainties What is the main source of systematic? Blue box expresses the size of impact on the post fit with decreasing order Leading systematics are: W+jets modeling Z+jet modeling Signal modeling b jet & jet For upcoming LHC Run2 & beyond, systematic uncertainty reduction will be important 55

56 Results submitted to JHEP Preprint is available arxiv: Yoshikazu NAGAI (CPPM) CPPM Seminar 56

57 Contents Introduction Analysis overview Result of a search for the H bb Summary & Prospects for the upcoming high luminosity LHC 57

58 Prospects toward HL LHC 2 years after 4th July 2012 seminar, we are still on a journey toward the nature of this Higgs boson 58

59 A more challenging environment lhc project today 59

60 VH Vbb at 300 fb 1 / 3000 fb 1 arxiv: fb fb 1 Scenario 1: Run1 condition ATL PHYS PUB lepton 3000 fb 1 2 lepton 3000 fb 1 60

61 VH Vbb at 300 fb 1 / 3000 fb 1 ATL PHYS PUB fb 1, < > = 60 1 lepton 2 lepton lepton Significance ( ) fb 1, < > = lepton 2 lepton lepton Significance ( ) These results assume jet energy scale uncertainty reduction H bb observation can be expected in upcoming LHC run 0 lepton is not included in this prospect 61

62 Summary Analysis validated using di boson process For di boson (WZ, ZZ) Observed: 4.9 Expected: 6.3 (W/Z)H (W/Z)bb analysis using full ATLAS Run1 data presented For m H = 125 GeV Observed: 1.4 Expected: 2.6 Successfully re observed di boson process = ( x Br) measured / ( x Br) SM Close to 3 evidence sensitivity and we had a bit of bad luck (or real deficit from new physics? ) LHC will restart soon with higher energy and luminosity H bb observation can be expected soon if it exists! Improving background understanding is crucial for analysis Run2 and beyond Stay tuned!! 62

63 Backup 63

64 Latest mass measurements Mass (H, ZZ 4l) Phys. Rev. D. 90, (2014) H : H ZZ 4l: 64

65 Latest spin/parity measurements Spin/Parity (H, ZZ 4l, WW l l ) Phys. Lett. B. 726 (2013) Spin 0 + vs Spin 0 (H ZZ) Spin 0 + vs Spin 2 + (H WW) 65

66 H spin/parity observable Collins Soper frame 66

67 Spin 2 models arxiv: 1001:3396 We choose 2 m+ model among possible models for spin 2 hypothesis 67

68 Constraint on the new physics MCHM4 model: MCHM5 model: 68

69 Triggers for analysis Trigger selection 0 lepton: trigger by missing transverse energy (E T miss ) threshold: 80 GeV 1 lepton: Primary trigger by single lepton trigger (e/ ) threshold: 18~24 GeV Secondary trigger by E T miss to compensate inefficiency 2 lepton: trigger by single lepton trigger (e/ ) or di lepton trigger (ee/ ) threshold: 18~24 (12~13) GeV for single (di lepton) trigger E T miss trigger Single trigger Offline MET (GeV) Single e trigger 69

70 0 lepton new signal region E T miss GeV (low E T miss bin) Difficulty: analysis using trigger turn on region and huge QCD multi jet BG Developed special QCD multi jet rejection technique: Likelihood ratio method QCD multijet like Signal like QCDmultijet Likelihood ratio constructed from kinematics variable, such as (jet1, jet2) QCD multijet background is negligible after cut m bb of low E T miss bin is used for nominal analysis 70

71 Fit results (7 TeV, m jj analysis) 0 lepton 1 lepton 2 lepton Data and background + signal yields are compatible 71

72 Fit results (8 TeV, m jj analysis) 0 lepton 1 lepton 2 lepton Data and background + signal yields are compatible 72

73 Correlation b/w BDT & m jj 73

74 BDT input kinematics 74

75 Kinematics fitter 75

76 Systematics What is the main source of systematic? The systematic uncertainties are listed in decreasing order of their impact on the postfit impact on The deviations of the fitted nuisance parameters from its nominal value. error bars express postfit uncertainties relative to their nominal uncertainties 76

77 Systematics What is the main source of systematic? The systematic uncertainties are listed in decreasing order of their impact on the postfit impact on The deviations of the fitted nuisance parameters from its nominal value. error bars express postfit uncertainties relative to their nominal uncertainties The normalization parameters which are freely floating in the fit 77

78 Systematics What is the main source of systematic? The systematic uncertainties are listed in decreasing order of their impact on the postfit impact on The deviations of the fitted nuisance parameters from its nominal value. error bars express postfit uncertainties relative to their nominal uncertainties The normalization parameters which are freely floating in the fit Leading systematics are: W+jets modeling Z+jet modeling Signal modeling b jet & jet 78