Single Top Physics at the Tevatron Enrique Palencia Fermilab for the CDF and DØ Collaborations Rencontres de Moriond EW 2009, La Thuille (Italy), March 7-14, 2009
Fasten your seatbelt!!! • In this talk, you will see brand new single top results using the latest amount of data available per collaboration • ..... reporting the first observation of the single top quark production!!!!! ♦ You will see a lot of 5-sigma analyses! • By CDF and DØ, independently, at the Tevatron • Major achievement of both collaborations Enrique Palencia, Fermilab Rencontres de Moriond EW 2009, La Thuille (Italy), March 10, 2009 1 of 20
Why Single Top? • Single top quark is produced via electroweak interaction but has not been observed SO FAR ♦ σ SM (t-channel/tqb) = 1 . 98 ± 0 . 25 pb (m top = 175 GeV) ♦ σ SM (s-channel/tb) = 0 . 88 ± 0 . 11 pb (m top = 175 GeV) ♦ σ SM ( t ¯ t ) = 6 . 7 ± 0 . 8 pb (via strong interaction) ♦ B.W. Harris et al. , Phys. Rev. D 66, 054024 (2002) Z. Sullivan, Phys. Rev. D70, 114012 (2004) • Test of the Standard Model ♦ Direct measurement of | V tb | ♦ Top quark properties: polarization, spin, W helicity,... ♦ Same final state as WH • Sensitive to new physics ♦ Search for W ′ , H + (s-channel signature) ♦ Search for FCNC,... Enrique Palencia, Fermilab Rencontres de Moriond EW 2009, La Thuille (Italy), March 10, 2009 2 of 20
��� Event Selection and Backgrounds • Top decays most of the times to Wb • W + 2 or 3 (4 in DØ) energetic jets • One high p T isolated lepton (electron or muon) from the leptonic decay of the W / T , • Large missing transverse energy, E from the neutrino Z/Dib tt non- W Wbb • At least one jet identified as b-tagged Mistags W+HF jets (Wbb/Wcc/Wc) Wcc • Main backgrounds: W +Heavy Flavor, Wc W +mistags, t ¯ •W +j et s nor m al i zat i on f r om dat a and t , QCD, diboson heavy f l avor ( HF) f r act i on f r om MC Enrique Palencia, Fermilab Rencontres de Moriond EW 2009, La Thuille (Italy), March 10, 2009 3 of 20
Why it took so long? Experimental Challenge! • Single top hidden under huge background ⇒ counting experiment is NOT possible • Multivariate analyses needed to discriminate single top from backgrounds ♦ No single observable to see single top ♦ Will show several multivariate analyses in next slides Enrique Palencia, Fermilab Rencontres de Moriond EW 2009, La Thuille (Italy), March 10, 2009 4 of 20
Search Strategy Enrique Palencia, Fermilab Rencontres de Moriond EW 2009, La Thuille (Italy), March 10, 2009 5 of 20
Likelihood Function (LF) • Combines several sensitive variables into a single one • 7 (10) variables used in the 2 (3) jet bin: H T , Q × η , M jj , cos(l, j), log (ME t − chan )... CDF Run II Preliminary, L=3.2 fb -1 Events/0.155 Data Wbb W+LF 160 160 s-channel ttbar NonW Monte Carlo Scaled to Prediction t-channel Wc+Wcc Z+jets,Diboson 140 140 120 120 N sig ( x i ) p sig 100 100 ( x i ) = i ( x i ) , i N sig ( x i )+ N bkg i i 80 80 60 60 i =1 p sig Π nvar ( x i ) L = i 40 40 Π nvar i =1 p sig ( x i )+Π nvar i =1 p bkg ( x i ) i i 20 20 -3 -2 -1 0 1 2 3 Q* η Enrique Palencia, Fermilab Rencontres de Moriond EW 2009, La Thuille (Italy), March 10, 2009 6 of 20
Likelihood Function: Results -1 CDF Run II Preliminary, L=3.2fb CDF Run II Preliminary, L=3.2 fb -1 Events Events s-channel 10 5 t-channel Pseudoexperiments B 3 3 Wbb 10 10 ttbar S+B WC 10 4 mistag ZJETS Diboson nonW 10 3 Median S+B 2 2 10 10 10 2 10 Observed 10 10 1 -100 -80 -60 -40 -20 0 20 40 60 80 -2lnQ 0 0 0.1 0.1 0.2 0.2 0.3 0.3 0.4 0.4 0.5 0.5 0.6 0.6 0.7 0.7 0.8 0.8 0.9 0.9 1 1 L L tchan tchan Lum. (fb − 1 ) LF Exp. sign. Obs. sign. Cross Section (pb) 1.6 +0 . 8 3.2 4.0 σ 2.4 σ − 0 . 7 Enrique Palencia, Fermilab Rencontres de Moriond EW 2009, La Thuille (Italy), March 10, 2009 7 of 20
Neural Networks: Results -1 All Channels CDF II Preliminary 3.2 fb Candidate Events Candidate Events 80 80 single top MC normalized to SM prediction 70 70 t t 60 60 Wb b +Wc c 300 300 Wc 50 50 Wq q 40 40 Diboson 30 30 Z+jets 20 20 200 200 QCD 10 10 data 0 0 0.6 0.6 0.7 0.7 0.8 0.8 0.9 0.9 1 1 100 100 0 0 -1 -1 -0.5 -0.5 0 0 0.5 0.5 1 1 NN Output NN Output Lum. (fb − 1 ) NN Exp. sign. Obs. sign. Cross Section (pb) 3.2 5.2 σ 3.5 σ 1.8 ± 0.6 4.7 +1 . 2 2.3 4.1 σ 5.2 σ − 0 . 9 Enrique Palencia, Fermilab Rencontres de Moriond EW 2009, La Thuille (Italy), March 10, 2009 8 of 20
Matrix Elements (ME) • Compute, for each event, the probability for signal and background hypotheses • Use full event kinematic information • Calculate probabilities for signal and backgrounds • Build a discriminant b · P sig ( � x ) EPD = b · P sig ( � x )+ b · P b − bkg ( � x )+(1 − b ) · P nonb − bkg ( � x ) P sig ( � x ) D ( � x ) = x ) , (separate for s and t channels) P sig ( � x )+ P bkg ( � Enrique Palencia, Fermilab Rencontres de Moriond EW 2009, La Thuille (Italy), March 10, 2009 9 of 20
Matrix Elements: Results schan -1 CDF Run II Preliminary, L=3.2 fb tchan wbb wcc mistag ww 500 500 wz zz zjets Candidate Events Candidate Events nonW ttbarlj 400 400 ttbardil 80 80 Candidate Events Candidate Events 60 60 300 300 Normalized to Prediction 40 40 20 20 200 200 0 0 0.7 0.7 0.75 0.75 0.8 0.8 0.85 0.85 0.9 0.9 0.95 0.95 1 1 Event Probability Discriminant Event Probability Discriminant 100 100 0 0 0 0 0.2 0.2 0.4 0.4 0.6 0.6 0.8 0.8 1 1 Event Probability Discriminant Event Probability Discriminant Lum. (fb − 1 ) ME Exp. sign. Obs. sign. Cross Section (pb) 2.5 +0 . 7 3.2 4.9 σ 4.3 σ − 0 . 6 4.3 +1 . 0 2.3 4.1 σ 4.9 σ − 1 . 2 Enrique Palencia, Fermilab Rencontres de Moriond EW 2009, La Thuille (Italy), March 10, 2009 10 of 20
Boosted Decision Trees (BDT) • Sequence of binary splits using the discriminating variable which gives best sig-bkg separation • Leaf nodes are classified as sig-like or bkg-like depending on majority of events ending up in the respective leaf • Use large number of input variables ♦ Non-discriminating variables are automatically ignored, but do not degrade the performance • Boosting algorithm improves the discrimination power and statistical stability ♦ Events misclassified during a DT training are given a higher weight in the next DT training Enrique Palencia, Fermilab Rencontres de Moriond EW 2009, La Thuille (Italy), March 10, 2009 11 of 20
Boosted Decision Trees: Results s-channel -1 CDF Run II Preliminary, L=3.2 fb t-channel W+light 120 120 120 120 120 120 120 120 120 120 120 120 25 25 W+charm W+bottom 20 20 Events Events Non-W 100 100 100 100 100 100 100 100 100 100 100 100 15 15 Candidate Events Candidate Events Candidate Events Z+jets 10 10 Diboson 5 5 80 80 80 80 80 80 80 80 80 80 80 80 tt 0 0 Data 0.5 0.5 0.6 0.6 0.7 0.7 0.8 0.8 0.9 0.9 1 1 BDT BDT Normalized to Prediction 60 60 60 60 60 60 60 60 60 60 60 60 40 40 40 40 40 40 40 40 40 40 40 40 20 20 20 20 20 20 20 20 20 20 20 20 0 0 0 0 0 0 0 0 0 0 0 0 -1 -1 -1 -1 -1 -1 -1 -1 -1 -1 -1 -1 -0.5 -0.5 -0.5 -0.5 -0.5 -0.5 -0.5 -0.5 -0.5 -0.5 -0.5 -0.5 0 0 0 0 0 0 0 0 0 0 0 0 0.5 0.5 0.5 0.5 0.5 0.5 0.5 0.5 0.5 0.5 0.5 0.5 1 1 1 1 1 1 1 1 1 1 1 1 BDT Output (2 jets, 1 tag) BDT Output (2 jets, 1 tag) BDT Output (2 jets, 1 tag) Lum. (fb − 1 ) BDT Exp. sign. Obs. sign. Cross Section (pb) 2.1 +0 . 7 3.2 5.2 σ 3.5 σ − 0 . 6 3.7 +1 . 0 2.3 4.3 σ 4.6 σ − 0 . 8 Enrique Palencia, Fermilab Rencontres de Moriond EW 2009, La Thuille (Italy), March 10, 2009 12 of 20
Recommend
More recommend
