1. Electron reco and identification improvements for 17.2 rel & H  ZZ analysis Fany Dudziak ISU group meeting Focus Talk - June 4th 2012 1

2. Introduction New electron reconstruction in 17.2 New electron PID dedicated to the HSG2 analysis Cross checks about misalignments effects on the new track match Δη and Δφ res What about adding a new electron category to handle the conversions ? Big summary 2

3. The new electron reconstruction : rel 17.2 Big effort since few years to improve the electron reco Miss an efficient brem fit : GSF has been chosen (efficient, « CPU reasonable ») Studies about the track to cluster matching (my thesis)  way of improvement Both are used in the new reco Before 17.2 : 2 kinds of electrons: el_pt … el_GSF_pt Now only el_pt but THEY ARE GSF We need a new electron Identification corresponding to this new reco. 3

4. Motivations for the new reco 4 Track reco efficiency Fraction of radiated energy in the ID vs eta Lot of material in the ID  high brem effect mainly at large eta(projectif) The track reco suffers from the brem + affects the pt measurement + bigger losses at low pt + big ambiguity between electron/photon + charge misidentification Brem fit refitted the track surching for kicks in the path, and estimating the energy loss  Very CPU consuming  Not used for the trigger !! Add the Δφ res to improve again

5. Importance of low pt electron in the H  ZZ  4l analysis 5 ET1 ET5ET4 ET2

6. Reminder about the Δφ rescaled Δφ (track - calo_s2) extrapolated from perigee, P rescaled to the E_cl (much less sensitive to brem) 6 minbias

7. Impact of the new reco 7 Old reco (17.0) GSF without Δφ res GSF + Δφ res (17.2)

8. Impact of the new reco 8

9. The MultiLepton Menu (I) Reconstruction/egamma/egammaAnalysis/egammaAnalysisUtils/trunk/ Root/MultiLeptonDefs.cxx ReconstructionegammaegammaAnalysisegammaAnalysisUtilstrunk RootMultiLeptonDefs.cxx Dedicated to multilepton analysis involving low pt electrons Separation in 2 categories of electrons : low brem and high brem (as we have the information from GSF) Goal: Flat ID efficiency vs eta, pt, and pile-up, wrt the reconstructio Better rejection of fakes (hadrons, photons) than the loose++ ID but with similar efficiency. Robust wrt to pile-up. 9 bool passMultiLepton (double eta, double eT,double rHad, double rHad1, double Reta, double w2, double f1, double f3, dou ble wstot, double DEmaxs1, double deltaEta, int nSi, int nSiDeadSensors, int nPix,int nPixDeadSenso rs, double deltaPhiRes, double dpOverp, double TRratio, int nTRTTotal,int nBlayerHits, bool expectBl ayer, bool debug )

10. The MultiLepton Menu (II) 10 Giacomo Artoni and Kate Wahlen are the student who wrote the menu I was asked by Christos to do some checks, as I already did on the first data Check on the track info : With GSF no more outlier hits, but now we have the deadSensorHits info  should we use that? Answer is yes : Nothing important on efficiency or rejection, just independant from detector condition Efficiency wrt container (no cut)Iso electronBkg electronhadrons nPixHits++nSCTHits>=799.15%60.80%89.55% Idem + Dead Sensors (Pix & SCT)99.16%60.85%89.58%

11. 11 The MultiLepton Menu (III) Check of the misalignment effect on the track-match variables

12. Misalignment : overview of the study 12 First look at the 2012 data with 17.2 reco (m1xxx tag) runs : 200987, 201191, 200842, 200982, 200913, 201052, 200965, 201006, 200804, 200863, 200967, 201138, 201190, 200926, 201113, 201120 Comparison with MC « à la » 17.2 : mc11_7TeV106046.PythiaZee_no_filter.merge.NTUP_EGAMMA.e815_s1272_s1274_r3395_r3417_p948/ « Tag & Probe » very basic (no grl, no pileup reweighting, I just want to look at the shapes)  Tag is tight, with pt>20 GeV, eta in the acceptance,  Probe, with pt>20 GeV has only the calo and track cuts  We look at the trackmatch Δη, Δφ res and Silicon hits.  Cut around the Z peak (+/- 10 GeV)

13. Misalignment : cross-checks pile-up and Z mass 13 More pile-up in the MC ! Good Zmass reconstruction with this quick tag & probe Looks like we have mor background in the data  tails.

14. Misalignment : eta and pt comparison 14

15. Misalignment : what are we sensitive to? 15 All the effect of misalignment between the calo and the ID : Sagging of the electrodes  Δφ (need to separate charges and quadrants to see it) Pear shape  all Tilt  all Shift wrt (0,0,0)  Δη

16. Misalignment : Δφ res check 16 Data --- MC Q>0 Q<0 -π/4< ϕ <π/4 5π/4< ϕ <7π/ 4 3π/4< ϕ <5π/4 π/4< ϕ <3π/4 Everything is normal wrt the misalignement and charge asymetry, tails do to background electrons 16

17. BARREL eta>0 17 -π/4< ϕ <π/4 3π/4< ϕ <5π/4 π/4< ϕ <3π/4 -π/4< ϕ <π/4 5π/4< ϕ <7π/4 3π/4< ϕ <5π/4 π/4< ϕ <3π/4

18. BARREL eta<0 18 -π/4< ϕ <π/4π/4< ϕ <3π/4 3π/4< ϕ <5π/4 -π/4< ϕ <π/4 5π/4< ϕ <7π/4 3π/4< ϕ <5π/4 π/4< ϕ <3π/4

19. End cap eta>0 19 -π/4< ϕ <π/4π/4< ϕ <3π/4 3π/4< ϕ <5π/4 -π/4< ϕ <π/4 5π/4< ϕ <7π/4 3π/4< ϕ <5π/4 π/4< ϕ <3π/4

20. End cap eta<0 20 -π/4< ϕ <π/4π/4< ϕ <3π/4 3π/4< ϕ <5π/4 -π/4< ϕ <π/4 5π/4< ϕ <7π/4 3π/4< ϕ <5π/4 π/4< ϕ <3π/4

21. Misalignment : Δη check 21 We cut bigger than 0.005 so it’s ok. Not yet understood for 2012 data!

22. Misalignements : conclusion 22 Δφ res is well discribed by the MC, we have a very good agreement, mainly in the Barrel. The effects that we can see are too small to impact the Identification and create asymetries. Δη has some discrepancies in the End Cap that need to be understood, but that are very small for what we are interested in, and we are not sensitive to them 2012 data electrons benefit from the new reconstruction that allows more silicon hits in the tracks

23. Going on optimizing … 23 Adding a new electron category conversion like??

24. Motivations : 24 New menu for the electron ID dedicated to the H4l analysis. A cut on the Blayer hits is rejecting a lot of conversions : Rej : 6  14 But costs a lot on efficiency (~2% /electron  >5% loss in 4e final state) We know that the electron candidates from conversions have a Δφ res shifted in the positive side. Try to select a new catagory of electron failing the Blayer cut and cutting tighter (than the std menu) on the positive side of Δφ res

25. Δφ res for conversions 25

26. Using B-layer information 26 Creation of a new electron category : Electrons with no B-Layer hit Cut at 92% signal on Δφ res positive side (don’t touch the negative side ) MultileptonMultilepton + Blayer cut Multilepton + new category Efficiency for Zee electrons 89.97%88.13%89.71% Conversion rejection 6.0115.906.82 -0.3% instead of -2% for the e efficiency  0.7% loss / 5.2% in a 4e final state +13.5% of rejection  gain of 30% for 2 conversions

27. Results : electron efficiency 27

28. Results : conversion rejection 28

29. Results : conversion rejection 29 We can use Δφ res in combinaison to the Blayer cut to disantangle conversions and electrons without any Blayer hit. We want to have the best efficiency for the Higgs research so we cutted not to hard In the future, as we know it is feasible we can play on this cut to rejection the conversions more or less depending on the efficiency we want Multilepton macro in egammaAnalysisUtils-00-03-09

30. Summary plots 30

31. Efficiency : Et 31 Multilepton Multilepton + new BL category

32. Efficiency : eta 32 Multilepton Multilepton + new BL category

33. Efficiency : pile-up 33 Multilepton Multilepton + new BL category

34. Conversion rejection : Et 34 Multilepton Multilepton + new BL category

35. Conversion rejection : eta 35 Multilepton Multilepton + new BL category

36. Conversion rejection : pile-up 36 Multilepton Multilepton + new BL category

37. Hadrons rejection : Et 37 Multilepton Multilepton + new BL category

38. Hadrons rejection : eta 38 Multilepton Multilepton + new BL category

39. Hadrons rejection : pile-up 39 Multilepton Multilepton + new BL category

40. CONCLUSION First cross checks for 2012 HSG2 analysis : Improvement of 44% for the H4e analysis efficiency Great thing done at the end 40