1. 1 A Feasibility Study for a Strange Sea Asymmetry Analysis at ATLAS: update II Laura Gilbert and Jeff Tseng 13/12/07

2. 2 OUTLINE 1)Reminder: Analysis technique: W+D* Selection and Results 2)Background Round-up 3)Irreducible Backgrounds: Results 4)Electroweak Backgrounds: Results so far 5)QCD Backgrounds: Results so far 6)Summary and Continuing Work

3. 3 D* D* + W Search: Technique Select W candidate (K) π Reconstruct D 0 →K - π + : pT(K) > 1 GeV, pT(π) > 1.5 GeV D 0 vertex displaced. π B Add prompt pion, pT(π B ) > 0.5 GeV Consider 3 sign correlations: (K - with π +, K - with π B +, π B + with e - ) Consider 3 sign correlations: (K - with π +, K - with π B +, π B + with e - ) Plot reconstructed D*-D0 mass difference = 145.4MeV Plot reconstructed D*-D0 mass difference = 145.4MeV Optimise cuts: Optimise cuts: m(D0reco)- m(D0true)< 40MeV m(D0reco)- m(D0true)< 40MeV Signed Lxy > 0.35cm Signed Lxy > 0.35cm π B impact parameter significance d0/σ(d0)<3 π B impact parameter significance d0/σ(d0)<3 d0(K)*d0(π)<0cm 2 d0(K)*d0(π)<0cm 2 D0 impact parameter <0.2cm D0 impact parameter <0.2cm D* pT>6GeV, |η| 6GeV, |η|<2.5 s g W c cg W s Branching ratios: D* + →D0π + 67.7% D0 → K - π+ 3.8% c→D* 25.5% c→e 9.6% s c W g s g W c cg W s s c W NLO W production

4. 4 Signal sample: Results NB. Comphep cross section calculation suggests there should be 7000 events / fb -1, i.e. the reconstruction efficiency is ~1.5%. Low pT of batchelor pion is especially significant. Reconstructed Unsmeared Real D*s

5. 5 Quick comparison of electron reconstruction efficiencies in full and fast simulation Comparison of CSC sample with ATLFAST: Comparison of CSC sample with ATLFAST: Generator level filter: pT e > 10 GeV, |η| 10 GeV, |η|<2.7 Samples 5250, 5254 fully simulated and reconstructed in (11.0.42): 50000 of each W +, W - Samples 5250, 5254 fully simulated and reconstructed in (11.0.42): 50000 of each W +, W - Same samples reconstructed in ATLFAST (12.0.3), 3 million of each. ATLFAST has 100% electron reconstruction efficiency in inner detector acceptance region. Same samples reconstructed in ATLFAST (12.0.3), 3 million of each. ATLFAST has 100% electron reconstruction efficiency in inner detector acceptance region. Assume ~70% of ATLFAST reconstructed electrons will be reconstructed in full simulation/data Assume ~70% of ATLFAST reconstructed electrons will be reconstructed in full simulation/data

6. 6 Round-up of Signals υ, υ generated. 2 events pass selection / fb -1, i.e. < 8 events /fb -1 at 95% CL. 3 million of each W - →τ - υ, W + →τ + υ generated. 2 events pass selection / fb -1, i.e. < 8 events /fb -1 at 95% CL. Approximately 30% of electrons reconstructed in ATLFAST will be lost in full sim υ υ This brings our totals to 84 ± 25 events down from 120 ± 35 for W → eυ and < 6 events /fb -1 for W →τυ.

7. 7 Background Round-up I 12 5

8. 8 1) Irreducible backgrounds From d, b contributions to W+c production. No b diagrams contribute to signal (truth), <<1 events/fb -1 from comphep cross section calculations. d diagrams account for ~12 events/fb -1 (including electron selection efficiency of 70%). d diagrams produce 5% number asymmetry (and associated momentum asymmetry).

9. 9 Electroweak backgrounds Of sample of 2 million events, 45 pass all selection cuts *except* MET cut. Including 30% of electrons that are not reconstructed in full sim this gives <5.5 events /fb -1. υ 3) NLO Z → ττ production with one electron lost: likely to contribute <1 event /fb -1 by analogy with above and W→τυ 2) NLO Z →ee production with one electron lost

10. 10 Electroweak backgrounds 4) Z →cc: one c decays semileptonically, the other produces a D*. Cross-section is 2.2nb (by comparison with Z →ee), fractional cross section is 3.5pb for the full decay channel, corresponding to 3500 events / fb -1. Signal reconstruction efficiency was ~1.5% corresponding to ~50 events here. However electron is not isolated and batchelor pion d 0 will be larger, so it should be much lower. Z →bb: total relevant fractional cross section 8.6pb (130 events?) 1) one b decays semileptonically, the other produces a D*: this will produce the incorrect sign correlation for the electron and batchelor pion. Electron sign mis-reconstruction (hard brem, asymmetric photon) 2.35 ± 0.10% in full sim. Fractional cross-section 3.5pb. υ 2)One b decays to D*e υ X: fractional cross section is 5.1pb for the full decay channel. Correct sign but can cut on angle between D* and e. π B ) will be larger, Here electrons and D*s likely to be at much lower pT than signal, d 0 (π B ) will be larger, electrons will not be isolated. Generate to LO in Pythia to check (back-to-back events are worst case)

11. 11 Electroweak backgrounds 5) W →cs: c produces D*, electron from one of the jets. Cross section 44nb (from total W production 140nb, TDR). Electrons should be quite rare and unlikely to pass isolation cuts. υ 6) W →cb: one quark produces a D*, the other an electron (again probably not isolated). Cross section is low, ~2nb (similar to W → eυ) Generate these to check, probably small.

12. 12 Electroweak backgrounds 7), 8) and 9): WW, ZZ and WZ. Presented previously. Working from HERWIG cross sections (Alan Barr) and branching ratios. No selection cuts applied. Contribution from these backgrounds is negligible. Electroweak backgrounds look manageable!

13. 13 Background Round-up II

14. 14 QCD backgrounds 10) q+q→w+g: g →cc (bb). If a D* of the opposite sign to the W is produced in the gluon jet this could be a background. In the signal sample “truth” <2 of the passing events /fb -1 resulted from gluon splitting. Taking into account electron reconstruction efficiencies this corresponds to <5 events /fb -1 at 95% CL.

15. 15 QCD backgrounds υ 11) q+q→W+b: This constitutes a background when the b quark jet includes a D*. These events do not pass signal selection in the W → eυ samples, giving a limit of < 2 events /fb -1 at 95% CL. 12), 13) q+q→Z+g, q+q→Z+b: In each case no events are found to pass in the NLO Z →ee sample, corresponding to <1.2 events /fb -1. By analogy with W sample likely actually to be <<1.

16. 16 QCD backgrounds 14), 15) and 16) cc, bb, tt. Cross sections: 1.45mb (Pythia), 3.3mb (MC@NLO), 0.8nb (Pythia) cc is a background when one c decays semi- leptonically, the other to a D*. Electron not isolated bb when D* and electron from same b, or different bs with electron charge misidentified. Electron not isolated, in first case D* and e in same hemisphere. higher batchelor pion impact parameter. tt decays into bWbW. Much lower cross section, higher batchelor pion impact parameter. Should be simple to reject if bb background is rejected.

17. 17 QCD backgrounds 14), 15) and 16) cc, bb, tt. Cross sections: 1.45μb (Pythia), 3.3mb (MC@NLO), 0.8nb (Pythia) Samples of size ~10 9 /10 12 needed to reduce cc, bb backgrounds to <10 events/fb -1. Taking a different tactic: Count number of electrons passing and number of D*s separately. Assume independence (true for cc, not necessarily for bb but good approximation cc: preliminary check on 431 k events: 13 electrons pass, 6 D* candidates in selection region (3 real). Multiplying probabilities suggests background rate of <0.5 events / fb -1. bb cross section is factor of three larger however and will require larger sample ~10 7 events for similar reduction, and careful consideration of correlations.

18. 18 QCD backgrounds 17) D* with fake Ws. From sample 5802 (dijet + fake electron, full simulation) ~191mb. Same cross section problem as above, will use same approach. Back of envelope suggests if N e N D* <600 in sample of 3.4x10 6 events then this will amount to <10 events / fb -1.

19. 19 Summary 12 5

20. 20 Summary

21. 21 Summary W/Z →qq backgrounds remain to be studied. Will use Pythia since back-to-back case most important, good approximation. Expected small. Fake electron with jets sample ongoing. Expected to be manageable. qq backgrounds may still be large, in particular bb. Might require study in real data.