H -> 4 H -> 4 in the low mass region in the low mass regionE.Meoni, L.Larotonda, M.Antonelli, F.CeruttiE.Meoni, L.Larotonda, M.Antonelli, F.Cerutti

H -> 4 H -> 4 in the low mass region in the low mass regionE.Meoni, L.Larotonda, M.Antonelli, F.CeruttiE.Meoni, L.Larotonda, M.Antonelli, F.Cerutti

• Introduction

• ATLFAST++ and MOORE/MuID performance

• Irreducible background rejection

• Reducible bkg. Rejection

• Status and prospects

• Introduction

• ATLFAST++ and MOORE/MuID performance

• Irreducible background rejection

• Reducible bkg. Rejection

• Status and prospects

IntroductionIntroductionIntroductionIntroduction

• Study started more then 1 year ago with twofold goals:– Validate ATLFAST++ and ATHENA Moore/MuID– Improve analysis w.r.t. TDR by using

multivariate techniques against irreducible (ZZ->4) and reducible (tt and Zbb) backgrounds

• Started with ATHENA release 6.0.3• Mass region studied: MH[130-180] GeV

(low mass is the most challenging because of the off-shell Z and higher bkg)

• Here results showed for MH=130 GeV

• Study started more then 1 year ago with twofold goals:– Validate ATLFAST++ and ATHENA Moore/MuID– Improve analysis w.r.t. TDR by using

multivariate techniques against irreducible (ZZ->4) and reducible (tt and Zbb) backgrounds

• Started with ATHENA release 6.0.3• Mass region studied: MH[130-180] GeV

(low mass is the most challenging because of the off-shell Z and higher bkg)

• Here results showed for MH=130 GeV

IntroductionIntroductionIntroductionIntroduction

• Samples produced with PYTHIA 6.2 with the exception of Zbb (ACERMC ME)

• Filter: 4 with Pt>4 GeV and ||<2.7

• Samples produced with PYTHIA 6.2 with the exception of Zbb (ACERMC ME)

• Filter: 4 with Pt>4 GeV and ||<2.7

Produced Equiv. Int. Luminosity

H 4 (m=130 GeV) 50 K 1.5 10^5 fb-1

H 4 (m=150 GeV) 50 K 9.8 10^4 fb-1

H 4(m=180 GeV) 50 K 1.8 10^5 fb-1

ZZ4 115 K 1.4 10^4 fb-1

(Z/*)(Z*/ *)-> 2µ2 27 K 8.5 10^4 fb-1

(Z/*)bb, with (Z/*)->2 µ + 2 filter 93 K 9.7 10^2 fb-1

tt->WbWb, with W-> 698 K 6.3 10^3 fb-1

Introduction: Pt and Introduction: Pt and spectra spectraIntroduction: Pt and Introduction: Pt and spectra spectra

ATLFAST++, MH=130 GeV

Pt(GeV)

Introduction: analyses chainIntroduction: analyses chainIntroduction: analyses chainIntroduction: analyses chain

Analysis with Multivariate methods

COMMON PRESELECTION 4 muons, null total charge : 2 with pT > 20 GeV and | | < 2.5 2 with pT > 7 GeV and | | < 2.5

Couples µ+ µ- with invariant mass : M12= Mz ± 15 GeV M34> 20 GeV ( mH= 130GeV )

M12= Mz ± 10 GeV M34>30 GeV ( mH= 150GeV )

M12=Mz ± 6 GeV M34>60 GeV ( mH= 180GeV )

Lepton isolation cuts (single variable cuts)

Lepton isolation cut likelihood/NN (with isolation variables) cut

Angular cut likelihood/NN (with angular variables and M12 & M34) cuts

TDR analysis

Mass window cut (mH 2 ) to compute significance

Mass window cut (mH 2 ) to compute significance

Introduction: software codesIntroduction: software codesIntroduction: software codesIntroduction: software codes

• ATLFAST++ (object oriented version of ALTAS fast simulation implemented in ATHENA framework)

• ATHENA: MOORE/MuID with muon spectrometer in standalone and combined – started with version 6.0.3 many bugs found– latest results with 7.0.2

• First step check of general performance– Efficiency– Pt resolution– MH resolution

• ATLFAST++ (object oriented version of ALTAS fast simulation implemented in ATHENA framework)

• ATHENA: MOORE/MuID with muon spectrometer in standalone and combined – started with version 6.0.3 many bugs found– latest results with 7.0.2

• First step check of general performance– Efficiency– Pt resolution– MH resolution

Selection efficiencySelection efficiencySelection efficiencySelection efficiency

• Acceptance after kinematic cuts (4m and M12 and M34 cuts):– ATLFAST++: 33.0%– TDR: 33.5%– MOORE/MuID combined 6.0.3: 9%– Inefficiency concentrated in low Pt region

• Acceptance after kinematic cuts (4m and M12 and M34 cuts):– ATLFAST++: 33.0%– TDR: 33.5%– MOORE/MuID combined 6.0.3: 9%– Inefficiency concentrated in low Pt region

Muid CombinedMuid CombinedAthena6.0.3Athena6.0.3

Selection efficiencySelection efficiencySelection efficiencySelection efficiency

• Improved with version 7.0.2– MOORE/MuID combined 7.0.2: 23%– Inefficiency concentrated eta~2 region

• Improved with version 7.0.2– MOORE/MuID combined 7.0.2: 23%– Inefficiency concentrated eta~2 region

Muid CombinedMuid CombinedAthena7.0.2Athena7.0.2

Mass resolutionMass resolutionMass resolutionMass resolution

• Performance muon spectrometer:– TDR: 2.7 GeV– MOORE 7.0.2: 3.0 GeV

• Combined (including Z mass constraint):– TDR: 1.4 GeV– ATLFAST++: 1.5 GeV– MOORE/MUID comb 7.0.2: 1.7 GeV

• Performance muon spectrometer:– TDR: 2.7 GeV– MOORE 7.0.2: 3.0 GeV

• Combined (including Z mass constraint):– TDR: 1.4 GeV– ATLFAST++: 1.5 GeV– MOORE/MUID comb 7.0.2: 1.7 GeV

Irreducible bkg.: ZZ->4Irreducible bkg.: ZZ->4Irreducible bkg.: ZZ->4Irreducible bkg.: ZZ->4

• Multivariate analyses: in addition to MH, M12 and M34 there are other 9 independent kinematic variables (12 in total)

• Try to select variable sensitive to the spin and parity of the signal • Combine all variables with multivariate techniques: likelihood and NN

• Multivariate analyses: in addition to MH, M12 and M34 there are other 9 independent kinematic variables (12 in total)

• Try to select variable sensitive to the spin and parity of the signal • Combine all variables with multivariate techniques: likelihood and NN

Likelihood function (and neural network) with 11 variables:- Angle of the decay planes of the two Z in Higgs rest frame (see ATL-COM-PHYS-2003-001,Buszello et al.)- Angle between - in Z rest frame and Z boost in Higgs rest frame (both for on-shell Z and off-shell Z) (see ATL-COM-PHYS-2003-001,Buszello et al.)- Angle between Z (both on-shell and off-shell) direction in Higgs rest frame and the Higgs boost- Angle between the two + in Higgs rest frame - Angle between the two - in Higgs rest frame - Angle between the two of Z (both on-shell and off-shell)- Invariant masses of the two + couples (M12 and M34)

Likelihood function (and neural network) with 11 variables:- Angle of the decay planes of the two Z in Higgs rest frame (see ATL-COM-PHYS-2003-001,Buszello et al.)- Angle between - in Z rest frame and Z boost in Higgs rest frame (both for on-shell Z and off-shell Z) (see ATL-COM-PHYS-2003-001,Buszello et al.)- Angle between Z (both on-shell and off-shell) direction in Higgs rest frame and the Higgs boost- Angle between the two + in Higgs rest frame - Angle between the two - in Higgs rest frame - Angle between the two of Z (both on-shell and off-shell)- Invariant masses of the two + couples (M12 and M34)

Angle between the decay planes of the two Z in Higgs rest frame

Angle between - in the Zrest frame and Z boost in Higgs rest frame

HH44

HH44 HH44

HH44 HH44

HH44

ZZZZ44ZZZZ44 ZZZZ44

ZZZZ44 ZZZZ44

ATLFASTATLFAST ATLFASTATLFASTATLFASTATLFAST

ATLFASTATLFASTATLFASTATLFAST ATLFASTATLFAST

FULL REC.FULL REC.

FULL REC.FULL REC.FULL REC.FULL REC.

FULL REC.FULL REC.FULL REC.FULL REC.

FULL REC.FULL REC.ZZZZ44

Angle between on-shell Z direction in Higgs rest frame and Higgs boost

Results with Fast simulationResults with Fast simulationResults with Fast simulationResults with Fast simulation

Results for Results for

L=30 fbL=30 fb-1 -1 inside inside

mass windowmass window

ATLFAST++TDR CUTS

ATLFAST++LIKELIHOOD

ATLFAST++NEURAL NETWORK

H ZZ* 4µ 4.40 5.71 5.62

(Z/*)(Z*/ *) 4µ 0.97 1.46 1.34

Poisson SignificancePoisson Significance 3.03.0 3.33.3 3.43.4

Improvement: mainly coming from Improvement: mainly coming from M12M12 and and M34M34 optimization optimization

angles relevant only at higher MHangles relevant only at higher MH

Improvement: mainly coming from Improvement: mainly coming from M12M12 and and M34M34 optimization optimization

angles relevant only at higher MHangles relevant only at higher MH

Reducible backgroundReducible backgroundReducible backgroundReducible background

• 2 out of 4 muons not isolated in tt and Zbb background• 2 out of 4 muons not isolated in tt and Zbb background

Likelihood (and neural network) with 6 variables:Likelihood (and neural network) with 6 variables:-the 2 largest normalized impact parameters(IP) in trasverse plane of the 4 IP-the 2 largest pT reconstructed inside a cone of R=0.2 around the 4 µ tracks-the 2 largest total transverse energy depositions in calorimeters (EM+HC) in a cone of R=0.2 around the 4µ tracks

We have added in CBNT ntuple block of Moore/Muid the energy deposition in cones We have added in CBNT ntuple block of Moore/Muid the energy deposition in cones of different radii around the “muon track”of different radii around the “muon track”

““muon track” defined in 4 ways: moore trk, muid statandalone trk, muid combined trk, muon track” defined in 4 ways: moore trk, muid statandalone trk, muid combined trk, iPat trkiPat trk

Best results with: energy of radius R=0.2 around “iPat” trkBest results with: energy of radius R=0.2 around “iPat” trk

SignalSignalttbarttbarZbbZbb

After pT & After pT & cuts cutsand m12 & m34 cutsand m12 & m34 cuts

SignalSignalttbarttbarZbbZbb

SignalSignalttbarttbarZbbZbb

Largest IPLargest IP

Largest pT Largest pT

Largest energy loss Largest energy loss Around iPat trackAround iPat track

Signal eff. Zbbar Rej. ttbar Rej.

Likelihood Likelihood 81%81% 101 101 ± 17± 17 1733 ±5331733 ±533

Neural Network 81% 75 ± 12 1014 ± 203

TDR TDR 81%81% 110 ± 50110 ± 50 1200 ± 3501200 ± 350

ATLFAST 76% 170 ± 40 590 ± 120

After pT & After pT & cuts cutsand m12 & m34 cutsand m12 & m34 cuts

LikelihoodLikelihoodNeuralNeural NetworkNetwork

Final Results: L=30 fbFinal Results: L=30 fb-1-1Final Results: L=30 fbFinal Results: L=30 fb-1-1

MUID

H H 4 4(Ang. lik & isol lik cut )

MUID

H H 4 4 lep lep(results of H 4

Multiplied by 4)

ATLFAST++

H H 44(Ang lik &ATLF’s isol. cuts)

TDR

HH4lep4lep

Signal (mH=130GeV)

4.45 17.8 5.71 11.4

ZZ 4 1.34 5.36 1.46 2.27

ZZ 2 0.12 0.48 0.21 0.20

zbbar 0.45 1.80 0.34 0.12

ttbar 0.01 0.04 0.02 0.02

BKG 1.92 7.68 2.03 2.61

PoissonPoisson

SignificanceSignificance2.42.4 5.05.0 3.03.0 4.8 4.8

Mass plotsMass plotsMass plotsMass plots

SignalZZ 4 ZZ 2 ttbarZbb

SignalZZ 4 ZZ 2 ttbarZbb

All channels

All channels

After preselection (4After preselection (4,Qtot=0,pT& ,Qtot=0,pT& cuts)cuts)

After overall analysis (preselection+ ang. lik cut+isol lik cut)After overall analysis (preselection+ ang. lik cut+isol lik cut)

-Preselection (as in TDR) : 4 , total charge =0, pT & cuts- Angular cut: likelihood- 11 variables -Isolation cut : likelihood– 6 variables

Conclusions and ProspectsConclusions and ProspectsConclusions and ProspectsConclusions and Prospects

• ATLFAST++ and MOORE/MuID (7.0.2) comb. performance studied on H->4 (low mass): worse performance then TDR, still low efficiency at ||~2 to be understood

Prospects Prospects • Add Noise and pileup, relevant for lepton isolation• Control samples to study lepton isolation variables

on data: tt-> WWbb: W->l W->jj select b jet with Mbjj=Mtop (b forced to leptonic decay)

• Wait for bug fixes ?• Produce documentation: ATLAS note and SN• Participation to DC2 validation very important:

– New digitization– New simulation GEANT4– New output data format– New reconstruction release

• ATLFAST++ and MOORE/MuID (7.0.2) comb. performance studied on H->4 (low mass): worse performance then TDR, still low efficiency at ||~2 to be understood

Prospects Prospects • Add Noise and pileup, relevant for lepton isolation• Control samples to study lepton isolation variables

on data: tt-> WWbb: W->l W->jj select b jet with Mbjj=Mtop (b forced to leptonic decay)

• Wait for bug fixes ?• Produce documentation: ATLAS note and SN• Participation to DC2 validation very important:

– New digitization– New simulation GEANT4– New output data format– New reconstruction release