John Marshall 350GeV HZ Analysis 1

350GeV HZ Recoil Analysis

LCD WG Meeting, 19/06/2012

J.S. Marshall, University of Cambridge

John Marshall 350GeV HZ Analysis 2

HZ Analyses

s = 500GeVLint = 500fb-1

mH =120GeVNo polarization

• Relevant processes for this study are the recoil reaction e+e-HZHff, commonly called Higgsstrahlung.

• Model independent recoil analysis:

• Z decays to charged leptons. Identify leptons and compute recoil mass:

• Model dependent analysis, with direct Higgs mass reconstruction:

• HZqq channel, hadronic decay of Z

• HZ channel, Z decays to neutrinos

sEMsM ZZrecoil 222

John Marshall 350GeV HZ Analysis 3

Previously...

DescriptionmH

(MeV)nsig (%)

1. 500GeV HZ Recoil, X

1390 12.51

2. 500GeV HZ Recoil, eeX

2233 20.46

3. 500GeV HZqq 149.1 1.940

4. 500GeV HZ 245.0 2.626

1. 2.

3. 4.

John Marshall 350GeV HZ Analysis 4

HZ Recoil: 350GeV vs. 500GeV

Process descriptio

n

Whizard process

ID

500GeV cross-sec

350GeV cross-sec

X (signal) hzmumu 2.451 fb 4.855 fb

ff  e2e2ff 4389 fb 4753 fb

eeX (signal) hzee 2.450 fb 4.850 fb

eeff  e1e1ff 6047.78 fb 4847 fb

Generator-level cuts for background

samplesff and eeff

pT l+l- > 10GeV, cut on transverse momentum

(vector sum of two leptons)

|cos l+/l-| < 0.95, cut on angle of either of leptons

500GeV cross-

sec (gen cuts)

350GeV cross-

sec (gen cuts)

2.451 fb 4.855 fb

832.9 fb 913.4 fb

2.450 fb 4.850 fb

1854 fb 1608 fb

• Move from 500GeV to 350GeV associated with increase in signal cross-section.

• e1e1ff background falls, whilst only modest rise for e2e2ff background.

• Recoil mass peak noticeably sharper, with radiative effects tail reduced.

Reco. lepton id.

John Marshall 350GeV HZ Analysis 5

350GeV HZ Recoil: Selection

40 GeV < Mdl < 120 GeV95 GeV < Mrecoil < 290 GeV60 GeV < PTdl

• Simple selection cuts before TMVA training:

• Lepton selection as previously described:

hzmumu e2e2ff hzee e1e1ff

Efficiency

98.1% 79.7% 97.0% 73.3%

John Marshall 350GeV HZ Analysis 6

350GeV HZ Recoil: Selection

TTdlTbal PPP

)/.( cosacop -12121 TTTT PPPP )/.( cosacol -1

2121 PPPP

John Marshall 350GeV HZ Analysis 7

350GeV HZ Recoil: BDT

)(/ BSS NNN

John Marshall 350GeV HZ Analysis 8

350GeV HZ Recoil: Fitting

• Use Simplified Kernel Estimation to approx. signal shape by sum of many Gaussians.

• Transformation x’ = x – mH allows sensitivity to mH. Scaling allows sensitivity to nSig.

• Fit shape of selected background: 4th order polynomial seems OK – see later checks.

• Scaling distribution allows sensitivity to nBkg. Background shape does not change.

• Fit uses MINUIT to vary mH, nSig and nBkg. A predicted distribution is created for these parameters and compared to the data, producing a 2 value.

John Marshall 350GeV HZ Analysis 9

350GeV HZ Recoil: X

X

• Fluctuate high statistics signal sample, and the smooth background fit, to produce test samples.

• Fit different test samples and examine fitted values and reported precisions:

Mean RMS

mH 119,945.4 MeV 131.4 MeV

mH 131.2 MeV

6.3 MeV

nsig 4.76 % 0.18 %

John Marshall 350GeV HZ Analysis 10

350GeV HZ Recoil: eeX

Mean RMS

mH 119,893.4 MeV 294.5 MeV

mH 303.1 MeV

30.9 MeV

nsig 8.11 % 0.62 %

• 1000 test samples for eeX channel:

John Marshall 350GeV HZ Analysis 11

350GeV HZ Recoil: Brem. Recovery

Improves peak entries, but broader due to use

of cluster energy

Removes tail

John Marshall 350GeV HZ Analysis 12

350GeV HZ Recoil: Brem. Recovery

Mean RMS

mH 120,018.3 MeV

404.6 MeV

mH 395.1 MeV

36.1 MeV

nsig 7.72 % 0.56 %

• Brem. recovery increases no. of entries in peak, but does increase peak width:

John Marshall 350GeV HZ Analysis 13

Comparison with ILC Results

• Event generation using PYTHIA• Beam Pol. (e-: -80%, e+: +30%),• s=350 GeV, L=175 fb-1

• Details taken from LCWS10 talk in Beijing, by Hengne Li, available via this link

• Event generation using WHIZARD• No beam polarization• s=350GeV• Scale to L=175 fb-1 for comparison

Reaction Cross-section

hzmumu 4.855 fb

e2e2ff 4753 fb

e2e2ff (pre-cuts) 913 fb

Reactions

hzmumu e2e2ff e2e2ff (pre-cuts)

Ninitial 850 831,775 159,785

Nselected 399 n/a 483

John Marshall 350GeV HZ Analysis 14

Comparison with ILC ResultsStarting point:No selection cuts

Selection must work harder than at ILC, but achieves similar S/B

Reco quantities,full selection

Pick true or reco leptons, relatively small difference

mH nsig

ILC 110 MeV 5.2 %

CLIC 131 MeV 4.8 %

• Comparison of results for X channel:

John Marshall 350GeV HZ Analysis 15

Comparison with ILC Results

• Try to understand remaining shape differences between recoil mass distributions at ILC/CLIC, so obtained ILC/CLIC E’ distributions from Frank. Use distributions to obtain E’ weight: CLICILC.

mH nsig

ILC 110 MeV 5.2 %

CLIC 131 MeV 4.8 %

CLIC, with E’ weight 87 MeV 4.3 %

John Marshall 350GeV HZ Analysis 16

Fit Robustness• Quick test of remaining free parameters in analysis.

Look at reported Higgs mass precision as function of:

Number of bins used for simplified Kernel estimation.

Number of bins used in fit to recoil mass distribution.

Mass range considered in fit; experiment with different windows around true Higgs mass.

John Marshall 350GeV HZ Analysis 17

Fit Robustness

1. Pol4a 2. Pol4b 3. Exponential

4. Gaussian 5. Pol5 6. Landau

mH

(MeV)nsig

(%)

1 2 3 4 5 6120

124

128

132

136

140

1 2 3 4 5 62

3

4

5

6

100 testsfor each