1. W→e and W→e +jets: a data-driven selection method By Alessandro Tricoli ATLAS UK SM Meeting 4 th June 2008 In Collaboration with M. Wielers (RAL) D. Prieur (RAL) T. Guillemin (LAPP)

2. Alessandro Tricoli, RAL 2 ATLAS UK SM, 4 th June 2008 Overview  W/Z incl. cross section measurement:  Study for early (10 31 lumi) and later (10 33 lumi) running.  Standard Cut-based selection  Data-driven selection method  Application to W+jets cross section measurement  Conclusions This work is part of the CSC notes W and Z cross section measurements and W/Z + jets cross section measurements

3. Alessandro Tricoli, RAL 3 ATLAS UK SM, 4 th June 2008 PART I Inclusive W→e cross section measurement

4. Alessandro Tricoli, RAL 4 ATLAS UK SM, 4 th June 2008 Data Samples Signal: Pythia PythiaWenu DS no. 5104 No Evt. 187650 Backgrounds: Pythia PythiaZee DS no. 5144 No Evt. 109900 PythiaWtaunu DS no. 5106 No Evt. 158350 JF17 DS no. 5802 No Evt. 4890386  Reconstruction ATHENA v12.0.6  Offline Analysis: AOD & custom AAN-tuples  use panda to submit analysis jobs

5. Alessandro Tricoli, RAL 5 ATLAS UK SM, 4 th June 2008 Cut-based Selection  On-line/off-line electron ID :  early data (50 pb -1 ):  trigger e20 (10 31 menu) [Monika W.’s code to extract 10 31 decision from AOD]  ele-ID: medium isem  higher lumi (1 fb -1 )  trigger e25i (10 33 menu)  ele-ID: tight isem  W selection:  acceptance and cracks removal cuts:  =1.37-1.52 and  2.4  Electron E T >25 GeV  Missing-E T >25 GeV  Jet Veto: Jets E T <30 GeV (Tower CONE  R=0.7, ele-jet overlap removed)  no recoil P T cut applied: no improvement on B/S only exception QCD, due to poor stat.: no trigger sel., but corrected normalization

6. Alessandro Tricoli, RAL 6 ATLAS UK SM, 4 th June 2008 Cut-based Event Selection Early data: e20, medium isem, lumi 50 pb -1 Bkg contamination can be rather small after cuts, but large uncertainties on QCD background (both theoretical and experimental) 50 pb -1 Higher Lumi: e25i, tight isem, lumi 1 fb -1 This Cut-Based selection completely relies on Monte Carlo

7. Alessandro Tricoli, RAL 7 ATLAS UK SM, 4 th June 2008 Data-driven W Event Selection  Problems with Cut-Based:  difficult estimation of QCD background after all cuts:  not enough MC stat. to trust MC (plenty of data)  amount and shape under W peak difficult to estimate:  ETmiss and jet Veto remove most of backgrounds, but leave ‘some’ under W peak => can only rely on MC estimate We want to Minimise our dependence on MCs Measure backgrounds from data Use simple and reliable method

8. Alessandro Tricoli, RAL 8 ATLAS UK SM, 4 th June 2008  Proposed W Selection :  Electron Trigger selection (e20 or e25i)  Electron Identification (medium or tight)  acceptance and cracks removal cuts:  =1.37-1.52 and  2.4  Electron E T >25 GeV  Missing-E T >25 GeV  Jet Veto: jet E T <30 GeV  Z->ee Background Removal:  cut on Mee invariant mass  QCD Background Removal:  fit ETmiss spectrum in high purity QCD sample  subtract QCD spectrum away Replaced by: Etmis and jet veto remove most of backgrounds, loose knowledge of shapes Data-driven W Event Selection

9. Alessandro Tricoli, RAL 9 ATLAS UK SM, 4 th June 2008 Z->ee Removal R(had/EM) in HEC/EMEC R(had/EM) in FCal M T Before Z->ee RemovalM T After Z->ee Removal Z→ee contamination reduced from B/S=24.5% to 3.0% Negligible effect on W→e, W→ , QCD and their shapes QCD Jet TruthEle Matching Jet TruthEle Matching Jet QCD Jet W ->  Z -> e - e + QCD W -> e W ->  Z -> e - e + QCD W -> e

10. Alessandro Tricoli, RAL 10 ATLAS UK SM, 4 th June 2008 Data-driven QCD subtraction Method  QCD Subtraction Method: Fit QCD background from a pure QCD sample to access distr. tails under W peak  Find a pure QCD Control Sample:  fake-photons: photon trigger (g20) and look at photon container  W→e signal swamped by QCD events (98% purely QCD)  Fit ETmiss from QCD Control Sample in range ETmiss>10 GeV  ETmiss shape for photon and electron samples must be similar  Normalise Fit to electron-sample in side band (10 GeV< ETmiss <22.5 GeV)  Subtract normalised fit under W peak: ETmiss>22.5 GeV

11. Alessandro Tricoli, RAL 11 ATLAS UK SM, 4 th June 2008 Data-Driven QCD Subtraction Pure QCD Control Sample W ->  Z -> e - e + QCD W -> e High purity QCD sample Photon/Electron Shape Study  QCD Control Sample:  g20 trigger, medium photon isem, E T >25 GeV, acceptance and crack cuts  98% purely QCD  for ETmiss>30 GeV small jacobian peak from W→e  ~19% of QCD  for ETmiss>40 GeV small jacobian peak from W→e  : ~42% of QCD  next slide explains hot to get rid of this contamination  ETmiss shape of QCD in Control Sample same as in Signal Sample Similar shape in electron and photon samples for ETmiss>10 GeV QCD only

12. Alessandro Tricoli, RAL 12 ATLAS UK SM, 4 th June 2008 Data-Driven QCD Subtraction W→e contamination in QCD Control Sample (ETmiss>30 GeV) Anti-Isolation cut can remove most of this W→e  contamination: Tried etcone (Calo Isolation in  R<0.45 Cone) W→e contamination can be also tested with data by an iterative signal+bkg fit (fit signal in electron sample and bkd in photon sample) etcone/E T phot QCD W -> e High purity QCD sample W ->  Z -> e - e + QCD W -> e etcone/E T phot >0.15 No visible distortions On ETmiss shape of QCD by Anti-Isolation cut W→e contamination reduction:  ETmiss>30 GeV: from ~19% to ~5%  ETmiss>40 GeV: from ~42% to ~6.4%  Improvement with data:  make etcone cut tighter  add track anti-isol cut

13. Alessandro Tricoli, RAL 13 ATLAS UK SM, 4 th June 2008 Data-driven QCD subtraction Fit and Normalization  Fit ETmiss in QCD Control Sample  exp(ax)*(1+bx 2 ) best fit  fit stable with different parametrisations and fit ranges  Normalisation of Fit to electron sample  normalisation in side band ETmiss=10-22.5 GeV  dominated by MC stat. uncertainty Side-band For norm. Signal region Normalised Fit with error band compared to QCD fake-electron data points Fit compared to QCD Control Sample data points (photon sample)

14. Alessandro Tricoli, RAL 14 ATLAS UK SM, 4 th June 2008 Data-driven QCD subtraction QCD subtraction results Early Data: e20, medium isem, lumi 50 pb -1 Higher Lumi: e25i, tight isem, lumi 1 fb -1 ETmiss spectrum after QCD subtraction compared to W→e  W→  Z→ee (no QCD)  QCD subtraction accuracy 0.1%-1.2%  uncertainty dominated by MC statistics (~4%) W ->  Z -> e - e + W -> e

15. Alessandro Tricoli, RAL 15 ATLAS UK SM, 4 th June 2008 Cross section measurement  Background contaminations and Acceptance after data-driven selection  Global Trigger and Electron Eff. from Tag&Probe (Ellie D., Mike F., Maria F., Guillaume K.)  Lumi uncertainty assumed: 10% at low lumi, 5% at higher lumi  Theory (NNLO, CTEQ6.1M) 20.5 nb: good agreement with our results!!

16. Alessandro Tricoli, RAL 16 ATLAS UK SM, 4 th June 2008 PART II W→e  + Jets cross section measurement

17. Alessandro Tricoli, RAL 17 ATLAS UK SM, 4 th June 2008 Data Samples W+jets Analysis Signal: Alpgen+Jimmy  Generator Filter: N jets ≥ 1 with P T >20 GeV AlpgenJimmyWenuNp0 DS no. 6101 No Evt. 30000 AlpgenJimmyWenuNp1 DS no. 6102 No Evt. 39100 AlpgenJimmyWenuNp2 DS no. 6103 No Evt. 48670 AlpgenJimmyWenuNp3 DS no. 6104 No Evt. 48400 AlpgenJimmyWenuNp4 DS no. 6105 No Evt. 31700 AlpgenJimmyWenuNp5 DS no. 6106 No Evt. 9700 Backgrounds: Pythia PythiaZee DS no. 5144 No Evt. 109900 PythiaWtaunu DS no. 5106 No Evt. 158350 ttbar_Pythia DS no. 5568 No Evt. 324800 JF17 DS no. 5802 No Evt. 4890386  Reconstruction ATHENA v12.0.6  Offline Analysis: AOD & custom AAN-tuples  use panda to submit analysis jobs

18. Alessandro Tricoli, RAL 18 ATLAS UK SM, 4 th June 2008 W +Jets Cut-based Selection  W Selection :  Trigger menu e25i applied:  one isolated e ±, tuned for efficiently select e ± with E T > 25 GeV  only exception QCD, due to poor stat.: no trigger sel.  Electron Identification: Medium  cracks removal  =1.37-1.52 and  2.4  Electron E T >25 GeV  Trask Isolation (to reduce QCD):  tracks in  R<0.2 cone around electron: n.track≤4 and  p T ≤4GeV  Missing-E T >25 GeV  Jets:  Jet Algorithm: Tower CONE  R=0.4  E T >20 GeV  ele-jet overlap removal:  R=0.4

19. Alessandro Tricoli, RAL 19 ATLAS UK SM, 4 th June 2008 W -> e + Jets cut-based selection W ->  Z -> e - e + QCD W -> e ttbar Cumulative Jet Multiplicity W Selection: e25i Trig+ isEM medium+ E T ele >25GeV + Track Isol. + Miss E T >25GeV Jet Selection: Tower CONE 04 Jet Et > 20 GeV Ele-jet overlap removal:  R=0.4 Lumi = 1 fb -1  QCD dominates at lower Jet Mult.  ttbar dominates at larger Jet Multi. Jet Multiplicity

20. Alessandro Tricoli, RAL 20 ATLAS UK SM, 4 th June 2008 W -> e and backgrounds W ->  Z -> e - e + QCD W -> e ttbar after e25i + isEM + E T ele >25GeV + Track Isol+ Miss E T >25GeV  ETmiss cut removes most of Zee bkg and large part of QCD  QCD tails under the W MT peak  QCD tails difficult to estimate (limited MC stat, large uncertainty on x-sect etc.) Lumi = 1 fb -1 Data-driven method necessary

21. Alessandro Tricoli, RAL 21 ATLAS UK SM, 4 th June 2008 Data-driven Before Zee removal After Zee Removal W ->  Z -> e - e + QCD W -> e ttbar Lumi = 1 fb-1  Follow same procedure as in inclusive measurement:  Z→ee Removal event-by-event  QCD subtraction for each jet multiplicity x-sect as function of jet mult. with QCD subtracted away  In addition data-driven selection for ttbar background:  See Maria’s talk

22. Alessandro Tricoli, RAL 22 ATLAS UK SM, 4 th June 2008 Conclusions  Standard Cut-based W Selection (a la TDR):  large uncertainties on amount and shape of QCD background  Alternative Data-Driven W selection:  replace ETmiss and jet veto cuts with  explicit Z->ee removal by cutting on e-e pair invariant mass  good rejection of Z->ee background is possible from 25% to ~3%  fit of QCD ETmiss distribution  QCD background subtraction: accuracy ≤1% (±4% MC stat uncertainty)  accurate inclusive W cross section meas. both with early data and higher lumi.  Apply similar Data-Driven selection on W+jets  also ttbar data-driven background removal (See Maria’s talk)

23. Alessandro Tricoli, RAL 23 ATLAS UK SM, 4 th June 2008 EXTRAS