1. 1 Heavy Flavor Backgrounds in view of Top Quark Measurements at D0 Elizaveta Shabalina University of Illinois at Chicago Berkeley Workshop on Boson+Jets production Berkeley, 03/26/08

2. 03/26/08 E. Shabalina Berkeley Boson+Jets Workshop 2 Outline Introduction Introduction History History When we were using unmatched AlpgenWhen we were using unmatched Alpgen “How to” for matched Alpgen “How to” for matched Alpgen W+jets in single top W+jets in single top W+jets in top pairs W+jets in top pairs W+jets in WH W+jets in WH MCFM calculations: MCFM calculations: Z+jets scale factorsZ+jets scale factors W+jets scale factorsW+jets scale factors Other issues Other issues

3. 03/26/08 E. Shabalina Berkeley Boson+Jets Workshop 3 V+jets and top pair production Dilepton channel Lepton+jets channel Z(ee/μμ)+jets - main background in ee and μμ channels, Z()+jets in eμ Very important at the early stages of event selection After final selection Z+jets contribution is small Measurements do not usually use b-tagging Flavor composition is not important Channel providing the most precise measurements W+jets - main background Very important at all stages of selection Most of measurements use b-tagging Knowledge of flavor composition is a limiting factor for precision measurements

4. 03/26/08 E. Shabalina Berkeley Boson+Jets Workshop 4 V+jets in single top s-channel t-channel Overwhelming W+jets background Understanding of all components is crucial for the analysis

5. 03/26/08 E. Shabalina Berkeley Boson+Jets Workshop 5 History W background model: Alpgen 1.3+Pythia 6.2 W background model: Alpgen 1.3+Pythia 6.2 No built-in matching No built-in matching Generate samples with 1,2,3 and 4 partons: Generate samples with 1,2,3 and 4 partons: Wj, Wjj, Wjjj, WjjjjWj, Wjj, Wjjj, Wjjjj Wc, Wcj, Wcjj, WcjjjWc, Wcj, Wcjj, Wcjjj Wcc, Wccj, WccjjWcc, Wccj, Wccjj Wbb, Wbbj, WbbjjWbb, Wbbj, Wbbjj Massive c-quark Massive c-quark No parton level cuts on heavy quarks (b,c) except for c-quark in single c production No parton level cuts on heavy quarks (b,c) except for c-quark in single c production Apply ad-hoc matching procedure: match partons to reconstructed jets Apply ad-hoc matching procedure: match partons to reconstructed jets Keep events in the samples if the number of reconstructed jets equals the number of matrix element partons Keep events in the samples if the number of reconstructed jets equals the number of matrix element partons W(cc) and W(bb): events with one c-(b-) jet coming from double c(b) production W(cc) and W(bb): events with one c-(b-) jet coming from double c(b) production two c(b) quarks merged in one jet one of c(b) jets outside acceptance Normalize W+jets to data before b-tagging Normalize W+jets to data before b-tagging PRD 74, 112004 (2006) PRD 74, 112004 (2006) Top pair cross section measurement with 420 pb -1 This approach was used only for ttbar cross section measurement Single top and Higgs used Wjj and Wbb samples only

6. 03/26/08 E. Shabalina Berkeley Boson+Jets Workshop 6Fractions Uncertainties: Matching cone size R (0.5 vs 0.7): 2% on Wc, 5% on other HF Cross check with limited sample of matched Alpgen – 20% PDF (20 eigenvector pairs from CTEQ6M) factorization scale (2 and 1/2) heavy quark mass: 0.3 GeV from nominal Correction factor K HF =1.050.07: K HF =(Wbb/Wjj) NLO /(Wbb/Wjj) LO J.M.Campbell and R.K.Ellis PRD 74, 112004 (2006) PRD 74, 112004 (2006) J.M.Campbell and J.Huston hep-ph/0405276 Applied to all HF contributions except Wc

7. 03/26/08 E. Shabalina Berkeley Boson+Jets Workshop 7 Result Uncertainty from W fraction calculation: 0.28 pb (4%), one of the two largest uncertainties, comparable with b- tagging efficiency measurement in data Background dominated

8. 03/26/08 E. Shabalina Berkeley Boson+Jets Workshop 8 What has changed? 1 fb -1 analysis: Uniform approach throughout the collaboration Uniform approach throughout the collaboration Alpgen 2.05+Pythia 6.323 with MLM matching for W+jets and Z+jets Alpgen 2.05+Pythia 6.323 with MLM matching for W+jets and Z+jets light partons jets are MLM matched (pT>8 GeV, Rjj>0.4) Due to the bug in Wc process we generate Wc within W+light Due to the bug in Wc process we generate Wc within W+light c-quarks are masslessc-quarks are massless Switched from adding generated minimum bias events to overlaying zero bias data events Switched from adding generated minimum bias events to overlaying zero bias data events We changed the tagger: SVT to Neural Network (NN) tagger We changed the tagger: SVT to Neural Network (NN) tagger

9. 03/26/08 E. Shabalina Berkeley Boson+Jets Workshop 9 Know-how I Generate 14 samples: W+bb+Nlp, with N=0,1,2 or ≥3 W+cc+Nlp, with N=0,1,2 or ≥3 W+Nlp, with N=0,1,2,3,4 or ≥5 (includes W+c+jets) Individual samples can not be used any more! Add all parton-jet bins together with weights F i

10. 03/26/08 E. Shabalina Berkeley Boson+Jets Workshop 10 Know-how II Sample has to be frozen Large relative weights Complicates statistics Post processing: Data qualityData quality selection due to zero bias overlay Discard events with additional heavy quarks created by Pythia Generate 14 samples: W+bb+Nlp, with N=0,1,2 or ≥3 W+cc+Nlp, with N=0,1,2 or ≥3 W+Nlp, with N=0,1,2,3,4 or ≥5 No skimming Book keeping is a nightmare!

11. 03/26/08 E. Shabalina Berkeley Boson+Jets Workshop 11 From generated MC to data Generate multi-parton MEs with Alpgen Generate multi-parton MEs with Alpgen Add showering and hadronization from Pythia Add showering and hadronization from Pythia b-fragmentation modelb-fragmentation model underlying event modelunderlying event model Run full D0 detector simulation and reconstruction Run full D0 detector simulation and reconstruction Add zero bias events to match luminosity profile in data Add zero bias events to match luminosity profile in data Apply to simulated events: Apply to simulated events: JES Jet removal Smear jets, electrons and muon Propagate to missing ET Correction factors: Correction factors: Trigger efficiency Electron and muon ID efficiency For b-tagging: For b-tagging: Taggability RF Tag rate function Reweightings: lumi, z vertex, etc Reweightings: lumi, z vertex, etc

12. 03/26/08 E. Shabalina Berkeley Boson+Jets Workshop 12 Data-driven procedure (first applied for single top evidence) Simplified background model: Three components: multijet, W+jets, top pairs Z+jets and diboson contributions are implicitely included in W+jets Do not rely on cross sections provided by LO Alpgen simulation Normalize W+jets (Wbb+Wcc+Wjj) to data before tagging in each jet multiplicity bin e/μ+jets channels are consistent within uncertainties PRL 98, 181802 (2007) Top pair contribution for nj  3

13. 03/26/08 E. Shabalina Berkeley Boson+Jets Workshop 13 Data-driven - II PRL 98, 181802 (2007) The ratio of heavy to light fraction is expected to be different at LO and NLO With matched Alpgen data supported HF fraction correction Extract from subsamples with 0 b-tags with negligible signal Cross check in W+1 jet sample with b-tag Assumptions: Scale factor is the same for Wbb and Wcc No correction to Wc contribution is applied Uncertainty: Shape difference between Wcc and Wbb Intrinsic uncertainty on LO xs Difference between 0 tag and signal sample

14. 03/26/08 E. Shabalina Berkeley Boson+Jets Workshop 14 Wc production Measured: Predicted by Alpgen for massless c-quark: Larger fraction of Wc can significantly affect HF scale factor ~80% difference

15. 03/26/08 E. Shabalina Berkeley Boson+Jets Workshop 15 W+jets in tt  lepton+jets The most precise measurements – can’t afford large uncertainties The most precise measurements – can’t afford large uncertainties Tighter selection cuts than in single top Tighter selection cuts than in single top Complete background model: W+jets Z+jets, dibosons, single top Complete background model: W+jets Z+jets, dibosons, single top First step is the same: normalize W+jets background to data before b-tagging in each jet multiplicity bin First step is the same: normalize W+jets background to data before b-tagging in each jet multiplicity bin Similar procedure than single top, more refined Similar procedure than single top, more refined Use background dominated samples to derive HF scale factor: Use background dominated samples to derive HF scale factor: 1 jet bin, 0 and 1 tags1 jet bin, 0 and 1 tags 2 jet bin, 0, 1 and 2 tags2 jet bin, 0, 1 and 2 tags

16. 03/26/08 E. Shabalina Berkeley Boson+Jets Workshop 16 HF scale factor calculation

17. 03/26/08 E. Shabalina Berkeley Boson+Jets Workshop 17 HF scale factor studies Sensitive to extra backgrounds included Sensitive to extra backgrounds included Sensitive to selection cuts on jets (single top ||<3.4, ttbar ||<2.5) Sensitive to selection cuts on jets (single top ||<3.4, ttbar ||<2.5) Consistent between e/μ channels Consistent between e/μ channels Studied several working points of NN algorithm Studied several working points of NN algorithm Separated Wc for W+lp and varied by 20% (not enough?) Separated Wc for W+lp and varied by 20% (not enough?) Dependence on the multijet background contribution Dependence on the multijet background contribution Tried to extract corrections to Wcc and Wbb separately Tried to extract corrections to Wcc and Wbb separately Measured HF scale factor (uncertainty from all deviations observed in the studies): Measured HF scale factor (uncertainty from all deviations observed in the studies): S HF = 1.170.18 S HF = 1.170.18 Used for summer results Used for summer results Switched to Alpgen 2.12 Switched to Alpgen 2.12 Several bug fixes including the bug in Wcc/Wbb generation Compared shapes: the only noticeable difference is R(j,j) Cross sections differ by a factor of 2! New S HF factor ~1.9 New S HF factor ~1.9

18. 03/26/08 E. Shabalina Berkeley Boson+Jets Workshop 18 Top pair results Precision measurements Simultaneous cross section and R=B(tbW)/B(tqW): Simultaneous cross section and R=B(tbW)/B(tqW): 0.59 pb - total systematics (w/o lumi) 0.59 pb - total systematics (w/o lumi) 0.22 pb (3%) from W HF fraction 0.22 pb (3%) from W HF fraction third largest (after JES and lepton identification) third largest (after JES and lepton identification) Top mass Not sensitive Not sensitive 25% variation of S HF changes mass by 0.05 GeV 25% variation of S HF changes mass by 0.05 GeV

19. 03/26/08 E. Shabalina Berkeley Boson+Jets Workshop 19 WH Data-driven procedure similar to top Data-driven procedure similar to top Apply same normalization factors to W+jets and Z+jets Apply same normalization factors to W+jets and Z+jets Overall normalization factor 1.7 applied to (Wl+Wcc+Wbb+Zl+Zcc+Zbb) Overall normalization factor 1.7 applied to (Wl+Wcc+Wbb+Zl+Zcc+Zbb) HF scale factor on top of it is 1.03 for Alpgen 2.05 HF scale factor on top of it is 1.03 for Alpgen 2.05 Good consistency between e/mu channels and various operating points of NN b-tagger Good consistency between e/mu channels and various operating points of NN b-tagger Consistent with ttbar analysis for the same version of Alpgen (i.e. 1.17) Consistent with ttbar analysis for the same version of Alpgen (i.e. 1.17) Small difference can be due to different approaches to tagging Small difference can be due to different approaches to tagging Direct tagging + data-MC scale factors (Higgs)Direct tagging + data-MC scale factors (Higgs) Tag rate functions (Top)Tag rate functions (Top) Work in progress

20. 03/26/08 E. Shabalina Berkeley Boson+Jets Workshop 20 Understanding scale factors Let’s try to define precisely and factorize the effects K-factor: theoretical (N)NLO/LO ratio of cross sections S-factor: empirical, comes on top of K to bring MC in agreement with data MC should be initially normalized to luminosity, and all correction (a.k.a. scale) factors should be applied (trigger, ID…) K HF -factor: theoretical by how much heavy flavor production should be increased on top of K and possibly S S HF -factor: empirical, comes on top of K or S, and K HF, to bring MC in agreement with data, after b-tagging

21. 03/26/08 E. Shabalina Berkeley Boson+Jets Workshop 21 Z+jets (MCFM) Z/+jets in Z mass window [60,130] NNLO cross section with NNLO PDFs with MRST2004 NNLO PDF : 256.6+5.112.0 pb [Hamberg et al] The total ALPGEN cross section is 199.5 pb From two numbers above K=1.3 with 5% uncertainty from PDF ProcessLONLOK (pb)(pb) Z+0lp(incl)1822421.33 Z+2lp(incl) 22 291.29 Zcc+0lp(incl)1.453.132.16(massless quarks) Zbb+0lp(incl)0.470.931.96(massless quarks) J.-F.Grivaz, private communication No additional S factor is needed for Z+jets inclusive to agree with data in ttbar analysis in dilepton channel. Use the same normalization in l+jets channel

22. 03/26/08 E. Shabalina Berkeley Boson+Jets Workshop 22 Z+HF (MCFM) K HF = K(Zbb+0lp(incl) / K(Z+2lp(incl)) where the K-factors are NLO/LO with massless quarks K HF (bb) = 1.96 / 1.29 = 1.52 K HF (cc) = 2.16 / 1.29 = 1.67 In Top we do not have sensitivity to these factors ZH analysis: Alpgen 2.05 S HF =1.5 Alpgen 2.12 S HF =1.52.0 There are possible (probable) dependences on generation cuts which may be responsible for the additional factor of 2 found in recent analyses J.-F.Grivaz, private communication

23. 03/26/08 E. Shabalina Berkeley Boson+Jets Workshop 23 ZH control plots

24. 03/26/08 E. Shabalina Berkeley Boson+Jets Workshop 24 W+jets (MCFM) K=1.3 (same as for Z+jets) K HF (cc) = K HF (bb) = 1.47 Data supports additional empirical factor ProcessLONLOK (pb)(pb) W+0lp(incl)186424561.32 W+2lp(incl) 222 2841.28 Wcc+0lp(incl) 1.89(massless quarks) Wbb+0lp(incl)561041.88(massless quarks) J.-F.Grivaz, private communication

25. 03/26/08 E. Shabalina Berkeley Boson+Jets Workshop 25 Other issues We’ve been using Alpgen extensively for the last couple of years to simulate W/Z+jets We’ve made several mistakes when using Alpgen There were some bugs found on the way and we had to invent work around Slow turn around times Alpgen release  DØ release  Production (large 0lp samples, slow 5lp samples)  Postproduction  Analyses 6-12 months Limits our ability to generate sufficient samples to study systematics Limits our ability to generate sufficient samples to study systematics

26. 03/26/08 E. Shabalina Berkeley Boson+Jets Workshop 26 Systematics Shape uncertainties are very important, especially for statistically limited analyses using discriminants Shape uncertainties are very important, especially for statistically limited analyses using discriminants Higgs, single topHiggs, single top Unmatched Alpgen was providing reasonable systematics from factorization scale. Matched Alpgen shows very little sensitivity to it. How should we estimate this uncertainty? Unmatched Alpgen was providing reasonable systematics from factorization scale. Matched Alpgen shows very little sensitivity to it. How should we estimate this uncertainty? PDF uncertainties: PDF uncertainties: CTEQ6.1M PDF error set: 1 central PDF and 40 error PDFs ( variations of 20 eigenvectorsCTEQ6.1M PDF error set: 1 central PDF and 40 error PDFs ( variations of 20 eigenvectors Use reweighting to estimate uncertaintiesUse reweighting to estimate uncertainties Does reweighting affect/destroy matching? Does reweighting affect/destroy matching?

27. 03/26/08 E. Shabalina Berkeley Boson+Jets Workshop 27 Conclusions W/Z background normalization is an issue in the majority of Top and Higgs analyses W/Z background normalization is an issue in the majority of Top and Higgs analyses Data-driven methods were proven to work but we are clearly missing an understanding of what is going on Data-driven methods were proven to work but we are clearly missing an understanding of what is going on Help from theorists is necessary and very welcome Help from theorists is necessary and very welcome

28. 03/26/08 E. Shabalina Berkeley Boson+Jets Workshop 28 Backup

29. 03/26/08 E. Shabalina Berkeley Boson+Jets Workshop 29 B-jet identification Top, Higgs signal contain b- jets Top, Higgs signal contain b- jets Most of backgrounds do not Most of backgrounds do not B-hadron lifetine ~ 1 ps B-hadron lifetine ~ 1 ps B-hadron travels L xy ~1 mm before decay B-hadron travels L xy ~1 mm before decay Combine properties of reconstructed secondary vertexes and displaced tracks in 7-variable network Combine properties of reconstructed secondary vertexes and displaced tracks in 7-variable network Working point: efficiency ~54%, fake rate ~1%Working point: efficiency ~54%, fake rate ~1% tt event tagging probability ~70% tt event tagging probability ~70% 11/26/07 E. Shabalina RAS conference 29

30. 03/26/08 E. Shabalina Berkeley Boson+Jets Workshop 30