First Evidence for Electroweak Single Top Quark Production

1. XLIInd Rencontres de Moriond QCD and High Energy Hadronic Interactions La Thuile, 17-24th March 2007 First Evidence for Electroweak Single Top Quark Production Leonard Christofek on behalf of the D0 Collaboration Leonard Christofek, Brown University

2. Single Top Theory and Motivation s-channel (tb) t-channel (tqb) Associated production (tW) NLO = 0.88 ± 0.11 pb NLO = 1.98 ± 0.25 pb NLO ~ 0.21 pb Too small at Tevatron • Study the Wtb coupling in top quark production • measure |Vtb| directly (no assumption of 3 families) • test unitarity of CKM matrix • Cross section sensitive to new physics • Background to SM Higgs production • Test of techniques to extract small signal from dominant background Leonard Christofek, Brown University

3. Window to new physics t-channel (tqb) Associated production s-channel (tb) c, W’, b, New heavy boson, New strong dynamics Modified Wtb coupling Flavor changing neutral currents (FCNC) Each new process would enhance single top quark production! Leonard Christofek, Brown University

4. Event Signature High pT jet High pT lepton Missing transverse energy High pT jet (“b-tagged”) High pT jet (“b-tagged”) • One isolated high transverse momentum lepton • Missing transverse energy • Two to four high transverse momentum jets • One or two jets associated with b-quark: • Labeled as a “b-tagged” jet Leonard Christofek, Brown University

5. Backgrounds • Top pair production normalized to NNLO cross section • Multijet background modeled using data with a non-isolated lepton and jets • W+jets (Wbb, Wcc, Wjj) and multijet background normalized to data before b-tagging Leonard Christofek, Brown University

6. Analysis Strategy Binned Likelihood Cross Section or Limits Leonard Christofek, Brown University

7. Background Model • Signal acceptances: tb=(3.2 ± 0.4)% tqb=(2.1 ± 0.3)% • Single top signal smaller than total background uncertainty • Cross section uncertainty dominated by statistical uncertainty Leonard Christofek, Brown University

8. Bayesian Neural Networks Hidden Nodes Input Nodes Output Node • Use 24 variables for training networks (input nodes). • Train network on signal and background simulated events: • Signal tends to one and background tends towards zero (output node). • Average many different networks for stability (hidden nodes). • Less prone to over-training. tqb Wbb Leonard Christofek, Brown University

9. Matrix Element • Use full kinematical information contained in event (i.e. the • four vectors from the lepton and jets). • Compute the probability for that event configuration to occur. • Backgrounds: Wbb, Wcg, Wgg, Wbbg Leonard Christofek, Brown University

10. Decision Trees • Goal: recover events that fail a simple cut-based analysis • Use 49 variables for training: most discriminating variables M(alljets), M(W,b-tag1), cos(b-tag1,lepton), Q(lepton)*(untagged1) • Decision tree output for each event = leaf purity: NS/(NS+NB) • Train network on signal and background simulated events: • Signal tends to one and background tends towards zero • Boosting: retrain 20 times to improve “weak classifier” Leonard Christofek, Brown University

11. Expected and Observed Results First evidence for single top quark production! SM compatibility is 11%. Leonard Christofek, Brown University

12. Discriminant Output Leonard Christofek, Brown University

13. Extracting |Vtb| • Assuming standard model production: • Pure V-A and CP conserving interaction: f1R = f2L = f2R = 0. • |Vtd|2 + |Vts|2 << |Vtb|2 or B(t Wb) ~ 100%. 0.68 < |Vtb| < 1 at 95%CL (f1L = 1) |Vtb f1L| = 1.3 ± 0.2 Leonard Christofek, Brown University

14. Combination Correlation matrix Highly correlated. Preliminary  = 4.8 ± 1.3 pb Significance = 3.5  Leonard Christofek, Brown University

15. Summary • First evidence for single top quark production. • With 3.5  significance. 4.8 ± 1.3 pb • First direct measurement of | Vtb | • Now starting to analyzing full 2 fb-1 data set. • Onto observation! Leonard Christofek, Brown University

16. Backup Transparencies Leonard Christofek, Brown University

17. Outline • Theory and motivation for single top • Event signature • Backgrounds • Multivariate analyses: • Bayesian Neural Network, Matrix Elements, Decision Trees • Cross section and signal significance • Combination of measurements • Measurement of the CKM element |Vtb| • Summary Leonard Christofek, Brown University

18. Discriminating Variables Leonard Christofek, Brown University

19. Single Top Monte Carlo • Based on COMPHEP • Reproduces NLO kinematic distributions • PYTHIA for parton hadronization Leonard Christofek, Brown University

20. Systematic Uncertainties Full treatment of systematics. Each systematic is turned off and the cross section reevaluated. The contribution of each systematic is estimated by taking subtracting these two widths in quadrature. Leonard Christofek, Brown University