1. C. Ciobanu, page 1 Single-top MC generator studies at CDF Workshop on Top Physics: from the Tevatron to the LHC LPSC Grenoble, October 19, 2007 Catalin Ciobanu LPNHE-Universités de Paris 6&7 / IN2P3-CNRS l Short Introduction l MadEvent vs TopReX l MadEvent vs ZTOP l MadEvent vs MC@NLO

2. C. Ciobanu, page 2 Single Top Production l B.W. Harris et al.: Phys. Rev. D 66, 054024, Z. Sullivan hep-ph/0408049 l Compatible results: Campbell et al, Phys. Rev. D 70, 094012 (2004). l N. Kidonakis, Phys.Rev. D 74, 114012 (2006) s-channel production (W*) t-channel production (Wg fusion) s 1/2 =1.96TeVNLO Cross-sections t-channel1.98±0.25 pb s-channel0.88±0.11 pb s-channel production (W*) l At the Tevatron, top quarks are: äMostly produced in pairs (7pb): å qq annihilation (85%) å gg fusion (15%) äAlso electroweak (single-top): å s-channel å t-channel  Wt associated production M top = 175 GeV/c 2

3. C. Ciobanu, page 3 LP 07 Status ANN 1 fb -1 Likelihood 1.5 fb -1 Matrix El. 1.5 fb -1 CDF Run II Preliminary Single Top Evidence at D0 Phys. Rev. Lett 98, 181802 (2007)

4. C. Ciobanu, page 4 Monte Carlo Issues l Obviously, background modeling and estimation comes first! l What about signal modeling? äAt CDF, we studied MadEvent, TopRex, and MC@NLO (Run I single- top analyses used Pythia signal samples) äRemarkable progress in MC generators since the beginning of Run II… äGenerous help from the MC/pheno/theory community l Test different generators vs each other: äLook at final state particle distributions: å At generation level å After parton showering (PS) å After detector simulation and reconstruction äLook at event yields

5. C. Ciobanu, page 5 t-channel samples l MadEvent, TopRex: äMix LO (initial state b-quark) and NLO (initial state gluon) samples äManually matching the two samples (a la CMS 2000/065, PRD 70, 114012) å Why need matching (what is different between LO and NLO)? t-channel production (Wg fusion) l Matching based on Pt(b2): äPt(b2)<K use LO events äPt(b2)>K use NLO events äBy definition imperfect

6. C. Ciobanu, page 6 t-channel samples l MadEvent, TopRex: äMix LO (initial state b-quark) and NLO (initial state gluon) samples äManually matching the two samples (a la CMS 2000/065, PRD 70, 114012) å Why need matching? t-channel production (Wg fusion) l Matching based on Pt(b2): äPt(b2)<K use LO events äPt(b2)>K use NLO events äBy definition imperfect

7. C. Ciobanu, page 7 t-channel samples l MadEvent, TopRex: äMix LO (initial state b-quark) and NLO (initial state gluon) samples äManually matching the two samples (a la CMS 2000/065, PRD 70, 114012) å Why need matching? t-channel production (Wg fusion) l Matching based on Pt(b2): äPt(b2)<K use LO events äPt(b2)>K use NLO events äBy definition imperfect

8. C. Ciobanu, page 8 I. MadEvent vs TopRex l Commissioned TopRex for the Tevatron l Looked at distributions of Pt and Eta of: äLepton, neutrino, b-quark from top decay, 2nd b-quark (b2), light q äOther variables providing good S/B discrimination å eg Q x  LO t-channel NLO t-channel

9. C. Ciobanu, page 9 MadEvent (red) vs TopReX (black) l Light quark (generator level) in the LO t-channel samples:

10. C. Ciobanu, page 10 MadEvent (red) vs TopReX (black) l Light quark (generator level) in NLO t-channel samples:

11. C. Ciobanu, page 11 l Top quark slightly harder in P T (and more central) in TopReX l Conclusion: t-channel samples (esp NLO) somewhat different: äNLO fraction of events is small compared to LO äLight quark more central in TopReX; top quark also, to a smaller extent MadEvent (red) vs TopReX (black)

12. C. Ciobanu, page 12 II. Comparisons to ZTOP l Also compared the MadEvent distributions against the NLO kinematic distributions (no events) given by the ZTOP program. l ZTOP gives pt and  of top and the leading jets (top was not decayed) l For the most part, shapes agree well, eg 2 nd b-jet in t-chan: l Worst agreement: Second leading light jet distributions (next page)

13. C. Ciobanu, page 13 l 2 nd leading light jet (few ME processes not included our MC) Comparisons to ZTOP l Acceptance agreement looks good:

14. C. Ciobanu, page 14 III. Comparisons to MC@NLO l MC@NLO single-top code available relatively recently – we used v3.3: äIncludes spin correlation between FS top and bottom quarks äMatching is not done by hand! äHERWIG used for PS (our default sample was MadEvent+Pythia) l Same procedure as before: äLook at final state particles and record the agreement äLook at acceptances in different jet bins

15. C. Ciobanu, page 15 MadEvent vs MC@NLO l Nice agreement. It can be seen that (after PS+full simulation) äIn general MC@NLO means lower by ~1 GeV

16. C. Ciobanu, page 16 l Very good agreement in all kinematic distributions, before and after detector simulation l For example, light quark jet in t-chan events: MadEvent vs MC@NLO

17. C. Ciobanu, page 17 l Reconstructed top and Q x Eta in t-channel events: MadEvent vs MC@NLO l Acceptances? l In the 2 jet bin the agreement is good: äs-channel: 3.43% (ME) vs 3.67% (MC@NLO) ät-channel: 2.41% (ME) vs 2.43% (MC@NLO) äAcross all jet bins – satisfactory agreement (next page)

18. C. Ciobanu, page 18 MadEvent vs MC@NLO MC@NLO MadEvent N(tight jets) Accept (%) Tight jets: Et > 20 GeV, |  | < 2.8 Require at least one b-tagged jet l Acceptances by jet bin are slightly different – esp. 3 jet bin l All jet bins together: s-chan: 9%, t-chan 4% l Understand the size of this effect on the cross section measurements for the next round – when the 3 jet bin will be included. s-ch t-ch

19. C. Ciobanu, page 19 Summary l Several generators studied for single-top samples at CDF: l MadEvent, TopRex, with manual matching for t-channel: äPhys. Rev. D, 71 012005(R) (2005) l Agreement with ZTOP NLO distributions reasonably good l MC@NLO versus MadEvent comparisons: äVery similar distributions äSlightly different distribution of acceptance by Njet bin (especially 3jet bin) l Many thanks to the theo/pheno/MC people – Tim Stelzer, Fabio Maltoni, Scott Willenbrock, Steve Mrenna, Zack Sullivan, Sergey Slabospitsky, Stefano Frixione, Bryan Webber, and others for providing expertise and support

20. C. Ciobanu, page 20