1. Modified HEEP ID for Z’ SUSY Search
Norbert Neumeister, Hwidong Yoo
Purdue University
Identification e/g meeting
January 29, 2014

2. Outline Introduction Signal modeling Event selection: HEEP id
Merged electron
Modification of HEEP id variables
Electron isolation

3. Introduction Non-typical Z’ search A baryonic Z´ resonance model
4 leptons in the final state
μμμμ, μμμe, μμee, μeee, eeee
V. Barger, H.S. Lee, Phys. Rev. D 85, (2012)
Several presentations in EXO non-hadronic meetings
In the talk at Dec. we fully discussed the electron id
Asked to discuss details of the modified HEEP id in E/g meetings that we have used in this search

4. Signal Modeling Z’ → 4 leptons Generator: CalcHep 3.4.1 + Pythia
Tree level matrix element (ME) calculation by CalcHEP
Model files provided by authors of theory paper
Generate 10 mass bins (12000 events each) for expected limit calculations
M = 750, 1000, 1250, 1500, 1750, 2000, 2250, 2500, 2750, 3000 GeV
Five channels: μμμμ, μμμe, μμee, μeee, eeee
Simulation and reconstruction
Use Full-Sim:
PV smearing, pile-up mixing, L1/HLT emulation, physics object reconstructions
CMSSW_5_3_4 with Summer12 S10 Pile-up scenario
Apply pT > 45 GeV (leading), pT > 30 GeV (sub-leading) with |eta| < 2.4
Performance is based on the signal mc samples

5. Event Selection: HEEP ID
Apply HEEP id for all electrons in each channel
Use modified HEEP isolation
Modify the HEEP id to improve the efficiency for boosted electrons
Same feature as for muon, but in electron case, the boosted two electrons are merged as a single electron in the reconstruction of the ECAL supercluster
Modified HEEP id by boosted Z→ee analysis doesn’t work in our analysis
Additionally there are several id variables to introduce inefficiency for such boosted electrons
Z’, M = 3 TeV

6. Boosted Z→ee AN2012-168 for boosted Z→ee analysis
Inefficiency starts in current standard electron reconstruction in barrel and endcap
Apply the cut for dEta < and dPhi < 0.3
In high mass sample most of electrons are rejected by the cuts in our case AN

7. Merged Electron Problem happens in current supercluster algorithm
Hybrid algorithm in barrel
Fixed bar of 3 or 5 crystals in eta direction
Dynamic search in phi direction
Fixed 5*5 crystals in endcap
In supercluster reconstruction, the ECAL rechits from two electrons are merged as one in many cases if they are close to each other and are reconstructed as a single supercluster
Only ECAL rechits used in the supercluster are stored in RECO/AOD format sample. Other rechits are not.
There is no way to separate the merged supercluster in order to match to the tracker tracks
Therefore, in current standard supercluster algorithm, we can not avoid this merged electron

8. Merged Electron: Example Event
Example event for the merged electron
Truth tracks
e1: pt = GeV
e2: pt = 27.8 GeV
e3: pt = GeV
e4: pt = GeV
e3+e4 is close to E3
Two reco electrons
E1: et = GeV
E2: et = 32.0 GeV E1 E2
From e4
One reco electron
E3: et = GeV
From e3
Two ECAL rechit towers shown but they are merged in supercluster and the supercluster is used to reconstruct electron
Reconstructed E3

9. Merged Electron ID (1)
Although the electrons are merged in the supercluster, a gsfTrack is usually reconstructed for the 2nd electron (that is merged to the 1st electron)
Check the dR(gsfTrack, electron)
Use gsfTrack with pt > 30 GeV
E/p of the merged electron should be larger than 1.0 because the 2nd electron is merged in the supercluster
Probability for merged electrons significantly increases for very high pT electrons
Apply this algorithm to only very high pT electrons

10. Merged Electron ID (2)
Selection: dR < 0.25 and E/p > 1.5 for electron et > 500 GeV
Et < 100
Black: merged electron
Red: non-merged electron
100 < Et < 500
500 < Et < 750
750 < Et < 1000
No merged electron, et < 500 GeV
Significant contribution from non-merged electron up to E/p 1.5
Et > 1000
Use Z’ signal MC sample with various masses

11. Modification of HEEP ID: dEtaIn
Difference in eta between the tracker position (measured in inner layer) and the eta of the supercluster
Sensitive variable for the merged electron
HEEP id: (EB), (EE)
Modification: 0.02 for EB and EE
Significant contributions above in high pT electrons EB EE

12. Modification of HEEP ID: H/E
Ratio of the HCAL energy of the CaloTowers in a cone of radius 0.15 to the electron’s supercluster ECAL energy
HEEP id: 0.05
Modification: 0.1
Significant contributions above in pT < 1000 GeV EB EE

13. E(2x5)/E(5x5), E(1x5)/E(5x5)
Ratio between supercluster energy (2x5 or 1x5) / 5x5
HEEP id: E(2x5)/E(5x5) > 0.94 or E(1x5)/E(5x5) > 0.83
Drop this variable
Electrons with E(1x5)/E(5x5) < 0.83
Significant contribution < 0.94

14. Electron Isolation Try both original HEEP iso and modified HEEP iso
Cut values are same but isolation definitions are different
Track iso < 5 GeV
ECAL iso + HCAL depth1 iso – 0.28*rho
< 2.0 GeV * Et in barrel
< 2.5 GeV in endcap, for Et < 50 GeV
< 2.5 GeV *(Et-50) in endcap, otherwise
Both isolations have significant inefficiency in high electron et
In origianl HEEP iso, there is strange feature (wiggles) in the high et distribution
Modified iso has stable efficiency up to 500 GeV but it is significantly decreasing in the higher et region.

15. Difference of Isolation Efficiency
Tracker isolation is strongly dependent on the Z’ mass
This feature is caused that two muons from same mother particle are close to each other by the boost
M = 1750
Replace by 4elec
M = 3000
M = 750

16. Iso Efficiency with dR
All
M = 1750
Replace by 4elec
M = 3000
M = 750

17. Modified HEEP Isolation
In the algorithm, we need to apply the HEEP id
Need to change cut parameters of HEEP id used in the Thomas’s package, same as what I am using in this analysis
Certainly this HEEP id cuts introduce the inefficiency.
The feature of strange wiggle is caused by the boost, same as muon
Trk Iso still introduces additional inefficiency in high et region
Use modified ECAL iso + HCAL depth1 iso
Change the factor by instead of 0.03 to increase efficiency in high et region

18. Summary

19. Back Up

20. Model 1 SM quark SM leptons 4G quarks 4G leptons U(1)B 1/3 -4
4 lepton Z´ resonance model without having a corresponding lepton pair signal
Use generic SUSY framework
But B (baryon number) is not preserved in the SUSY framework and proton is unstable
Consider gauged B by introducing U(1)B
Motivated to protect proton stability in SUSY
But in gauging B, additional fermions for anomaly cancellation
One possibility: an entire 4G family with B for all quarks
Under this condition
Proton decay (ΔB = 1) never occurs
Z´  ee, μμ, ττ are absent. (no dilepton resonance)
Z´  ν4ν4*, where ν42l is possible (4 lepton resonance), where ν4 is the lightest of the 4G states
Multilepton Z´ resonance without dilepton resonance is possible with large cross section
SM quark
SM leptons
4G quarks
4G leptons
U(1)B
1/3 -4 ~ ~ ~ ~

21. Datasets
2012 Dataset
SingleMu, DoublePhotonHighPt and MuEG PDs to collect muons, electrons and emu events
Use Jan22 ReReco and corresponding golden JSON file
Full 2012 dataset: 19.7/fb MC
Signal: private production (see next slide)
Backgrounds
Use Summer12 MC samples
DY (Powheg, high mass samples), ttbar, tW/tbarW, dibosons (WW, WZ, ZZ), W+Jets, QCD
Full list in next slide

22. MC Samples
Drell-Yan
/DYTo**_M_##_TuneZ2star_8TeV_pythia6/Summer12-PU_S7_START52_V9-v1/AODSIM ** = EE or MuMu
## = 20, 200, 500, 800, 1000, 1300, 1600
Diboson
/WW_TuneZ2star_8TeV_pythia6_tauola/Summer12_DR53X-PU_S10_START53_V7A-v1/AODSIM /WZ_TuneZ2star_8TeV_pythia6_tauola/Summer12_DR53X-PU_S10_START53_V7A-v1/AODSIM /ZZ_TuneZ2star_8TeV_pythia6_tauola/Summer12_DR53X-PU_S10_START53_V7A-v1/AODSIM
ttbar
/TT_CT10_TuneZ2star_8TeV-powheg-tauola/Summer12_DR53X-PU_S10_START53_V7A-v2/AODSIM
Drell-Yan tautau:
/DYToTauTau_M-20_CT10_TuneZ2star_8TeV-powheg-pythia6/Summer12_DR53X-PU_S10_START53_V7A-v1/AODSIM W+Jets: /WJetsToLNu_TuneZ2Star_8TeV-madgraph-tarball/Summer12-PU_S7_START52_V9-v1/AODSIM
QCD MC
EMEncriched
/QCD_Pt_**_EMEnriched_TuneZ2star_8TeV_pythia6/Summer12_DR53X-PU_S10_START53_V7A-v1/AODSIM
**: 20-30, 30-80, , , , 350
MuEnriched
/QCD_Pt-**_MuEnrichedPt5_TuneZ2star_8TeV_pythia6/Summer12_DR53X-PU_S10_START53_V7A-v1/AODSIM
**: 20-30, 30-50, 50-80, , , , , , , , 1000

23. HEEP ID Common ecalDriven = 1 H/E < 0.05 |dPhiIn| < 0.06
Inner layer lost hits <= 1 EB
E(2x5)/E(5x5) > 0.94 or E(1x5)/E(5x5) > 0.83
|dEtaIn| < 0.005
|dxy| < 0.02 EE
|dEtaIn| < 0.007
Sigma_ieta_ieta < 0.03
|dxy| < 0.05
No isolation cuts are applied in our analysis