SingleTop at CDF Catalin Ciobanu Wolfgang Wagner for
Single-Top at CDF Catalin Ciobanu, Wolfgang Wagner for the CDF Single-top Group Te. V 4 LHC meeting October 20, 2005 C. Ciobanu, page 1
First Run II Analysis l Phys. Rev. D 71: 012005, 2005 l Look in the W+2 jets channel: Ø Ø Ø l 1 lepton with ET>20 Ge. V, | |<1. 0 missing transverse energy: MET>20 Ge. V 2 jets : ET> 15 Ge. V, | | < 2. 8 at least one b-tag (displaced sec. vertex) Veto Z, dilepton, conversion events Topological cuts: Ø 140 < Ml b <210 Ge. V/c 2 (combined and separate searches) Ø leading jet ET > 30 Ge. V (separate search for t-channel only) l Backgrounds: non-top and tt C. Ciobanu, page 2
Single-Top Analyses l Two distinct analyses: combined and separate searches l Combined Search: Ø Signal: s-channel and t-channel single-top events Ø Both cross-sections proportional to |Vtb|2 Ø Exploits distributions similar for s- and t-channels: Ø HT = the total transverse energy in the event (ET lep + MET + ET jet ) l Separate Search: Ø 1. Signal = t-channel (s-channel is a background) Ø FCNC couplings, anomalous V+A contributions to the W-t-b vertex, etc. Ø Q • variable (Q = lepton charge, = pseudorapidity of non b-tagged jet) Ø Q • asymmetric in t-channel events: N(Q • >0) = 2* N(Q • <0) Ø 2. Signal = s-channel (t-channel is a background) Ø Heavy charged vector bosons W’, CP-violation effects within MSSM, Kaluza-Klein excited W-boson within MSSM Ø Double b-tags – simple counting C. Ciobanu, page 3
Combined Search l Two-variable analysis: cut on reconstructed top mass Ml b then fit the total transverse energy HT Normalized to unit area Process HT [Ge. V] S/B = 12 % Number of events/162 pb-1 tt 3. 8 ± 0. 9 Non-top 30. 0 ± 5. 8 Sum Background 33. 8 ± 5. 9 t-channel 2. 8 ± 0. 5 s-channel 1. 5 ± 0. 2 Sum Single-Top 4. 3 ± 0. 5 Sum Expected 38. 1 ± 5. 9 Observed 42 C. Ciobanu, page 4
Combined Search Results A-priori, no syst: 12. 4 pb A-priori, w/ syst: 13. 6 pb A-posteriori w/ syst: 17. 8 pb MPV( units) 2. 7 +1. 8 -1. 7 MPV(pb) 7. 7 +5. 1 -4. 9 C. Ciobanu, page 5
Separate Search Results t-channel: A-priori: 11. 2 pb A-posteriori: 10. 1 pb s-channel: A-priori: 12. 1 pb A-posteriori: 13. 6 pb Channel MPV( units) t-channel 0. 0 +2. 4 MPV(pb) 0. 0 +4. 7 -0. 0 4. 6 +3. 8 -0. 0 s-channel 5. 2 +4. 3 -4. 3 C. Ciobanu, page 6
Next Steps? l Dec 04 CDF Workshop Ø Make sure we model signal correctly Ø Z. Sullivan, T. Stelzer, E. Boos, S. Slabospitsky Ø Plan for the next iteration Ø Increase acceptance Ø Increase S/B Ø Mulivariate techniques l Next publication - aim for observation: Ø Previously: set limits on anomalously high signal rates Ø Null hypothesis: background+SM single-top Ø Test hypothesis: background+very large signal rates Ø Currently: Ø Null hypothesis: backgrounds only, no signal Ø Test hypothesis: background + SM signal C. Ciobanu, page 7
NN b-tagging t-channel 5% s-channel 3% 11% 32% 12% Mistags (u, d, s) 25% C. Ciobanu, page 8
Kinematic Fitter l Kinematic fitter: allow pb, b and ET , to vary within uncertainties l 4 fits: 2 b-jet assignment + 2 pz solutions Can use this 2 for – choosing the b from top (~80% correct) Calculate matrix element-like quantities Then, form a combined probability l l l Ø Different variables for t-channel and s-channel C. Ciobanu, page 9
Likelihood Method l l 6 Variables: Ml b , ME(t-chan), cos( l, jet), Mjj, HT, Q* Using the PRD samples, need 1. 7 fb-1 for a 3 evidence on the t-channel. For s-channel, need good variables… t-channel Likelihood function C. Ciobanu, page 10
Neural-network Method l l 3 -layer Neural Network Neuro. Bayes@ program Ø (Run I single-top PRD: JETNET) l l 15 variables input to 3 -layer net Best 10 variables: C. Ciobanu, page 11
Neural-Network Method C. Ciobanu, page 12
Matrix Element Method C. Ciobanu, page 13
Matrix-Element Method l l l Main reference: Bernd Stelzer’s thesis: Matrix element from Mad. Event Transfer functions – double Gaussian parametrization EPD = Ps/Pb Making several assumptions – 1. 2 fb-1 for 3 C. Ciobanu, page 14
Conclusions l LHC 4 Te. V – MC help from Sergey Slabospitsky (CMS) l In progress: Ø Increase acceptance (forward electrons) Ø Increase signal purity (NN b-tagger) Ø Use multivariate techniques: Ø Matrix element, Neural-Nets, Likelihood l l We should be ready for >1 fb-1 – switch to discovery mode. No discovery without reducing the background uncertainties C. Ciobanu, page 15
- Slides: 15