1. Fully Inclusive b->s  Status Report UCSC: Eisner, Schmitz, Schumm Edinburgh: Tinslay Pisa: Bucci, Walsh Royal Holloway: Brown, McMahon SLAC: Jessop, Ozcan John Walsh INFN-Pisa Outline I.Introduction II.New Features III.BB Background Determination IV.Plans/Conclusions

2. BaBar Italia – Aprile 2003 Capri 2 Physics of b->s  Sensitive to New Physics due to internal loop. SM theoretical prediction on fairly solid ground: –Gambino and Misiak: 3.57 +- 0.30 x 10 -4 Very low expected CP violation Distribution of m Xs (and hence E  *) depends m b and  p, Fermi motion of b quark inside B meson

3. BaBar Italia – Aprile 2003 Capri 3 Key Points of Inclusive Analysis Select high energy photons: the minimum photon energy is an important aspect of the analysis –Lower energy threshold  less model dependence higher BB background  higher systematic uncertainty. Backgrounds – few handles to suppress them compared to exclusive analyses –Continuum events Photon from initial state radiation, ISR   0 (  ) decay photon –BB events   (  ) decay photon (90-95%) hadronic fakes (5-10%)

4. BaBar Italia – Aprile 2003 Capri 4 Analysis Strategy Photon selection – similar to K*  analysis Background suppression –shape cuts, lepton tags,  0 /  vetoes Off-resonance data to subtract remaining continuum background  safe, although sacrifice statistical precision Use control samples to study BB background: –dedicated  0 /  skim –hadronic control sample –use to correct BB MC estimates

5. BaBar Italia – Aprile 2003 Capri 5 Branching ratio measurement (56 ifb on-peak): B(b  X s  ) = 3.88 +- 0.36 (stat) +- 0.37 (syst) + 0.43 – 0.23 (model) Systemics from ICHEP result –Photon Selection 3.7 % –Event Shape Cuts negligible –Lepton Tags 2.0 % –Normalization 1.1 % –BB background 7.8 % Rely on MC simulation to estimate the BB background Use a control sample to establish how well MC simulation models data, derive correction factors ICHEP 2002 Result

6. BaBar Italia – Aprile 2003 Capri 6 New Features for 2003 Use full (up to 2002) data set: –80 ifb On-resonant data –9.6 ifb Off-resonant data Improve event-shape selection by combining several variables optimally into Fisher Discriminant Improved treatment of BB background: dedicated measurement of  0 and  yields in BB events (Pisa group) Reduce minimum photon energy cut – 2 GeV is the goal Determination of photon spectrum as well as branching ratio

7. BaBar Italia – Aprile 2003 Capri 7 Fisher Discriminant – Event Shape Have found it useful to combine several event shape variables into a Fisher Discriminant –R2’/R2 (R2’ is 2 nd FW moment with photon removed from calculation  for ISR removal) –18 energy flow cones around photon direction Gives significant improvement in S 2 /(S+B’) compared to ICHEP analysis Introduces small energy- dependence in efficiency  negligible increase in model dependence Fisher > 0.575

8. BaBar Italia – Aprile 2003 Capri 8 Determining Background from BB ICHEP treatment – –use control sample of events that fail the  0 /  vetoes –compare rates in data and MC as function of photon energy –Drawbacks: limited statistics  need to extrapolate from low energy bins mismatch in bsg and control sample p0/h energy due to inverting vetoes

9. BaBar Italia – Aprile 2003 Capri 9 Improved BB Background Determination New   /  skim: –increase statistics with dedicated  0 /  skim with lower photon energy threshold (1 GeV instead of 1.5 Gev) Method: 1.Select candidates using all the standard bs  selection cuts, except for  0 /  vetoes 2.Perform fits of the  0 /  mass distributions in bins of  0 /  energy, thus determining the background subtracted  0 /  energy spectra. 3.Compare continuum subtracted data spectrum with MC BB spectrum and take ratio as “correction factors”, in bins of  0 /  energy.

10. BaBar Italia – Aprile 2003 Capri 10 Results –  0 mass fits energy spectrum data/MC ratio  0 Energy (GeV)

11. BaBar Italia – Aprile 2003 Capri 11 Results –  mass fits energy spectrum data/MC ratio  Energy (GeV)

12. BaBar Italia – Aprile 2003 Capri 12 Validation: Mass fits Compare fit yield with number of truth-matched candidates On MC we find that fit yields are systematically less than the number of “true” candidates. Typical yield about 85-90%. Find that using Novosibirsk does not help significantly, culprit is long tails in mass distribution Similar problem with  fits Is problem with fit or with truth-matching?

13. BaBar Italia – Aprile 2003 Capri 13  2 vs Ghit Truth-Matching Our original skim produced with  2 truth matching We compared with a small sample run with GHIT matching Qualitatively, it appears that the Ghit performance is significantly better Decided to reproduce  0 /  skim with GHIT matching All candidates Truth-matched  0 s  2 matchingGhit matching

14. BaBar Italia – Aprile 2003 Capri 14  0 mass fits – Log scale The better truth- matching required a better fit function Fit function: –signal: double gaussian with power-law low energy tail –background: polynomial (of degree 1, 2 or 3) Use MC and high-statistics untagged sample to help determine some parameters Fit yields now within 3% of true value

15. BaBar Italia – Aprile 2003 Capri 15  mass fits – Log scale Fit function: –signal: gaussian with low and high energy power law tails –background: polynomial (of degree 1, 2 or 3) Fewer energy bins due to low statistics Again fit yields within 2-3% of true value

16. BaBar Italia – Aprile 2003 Capri 16 Yields from Low Statistics Sometimes it’s necessary to obtain a  0 yield from very low statistics. In this case, we estimate yield by simple counting in the signal region, with sideband subtraction. We are currently validating this technique. There seems to be a slight bias, which we are investigating. Estimate - Truth (Estimate – Truth)/error

17. BaBar Italia – Aprile 2003 Capri 17 Next Steps Produce new BB background correction factors for  0 and  as soon as the last skim jobs are complete Correct BB MC and perform checks on unblinded control regions: –1.5 < E  * < 1.9 GeV of standard sample (lower side-band) –1.0 < E  * < 2.5 GeV of  0 /  sample Extract photon energy spectrum Unblind and move towards publication

18. BaBar Italia – Aprile 2003 Capri 18 Conclusions The inclusive bs  analysis is moving forward, although at a slower pace than was hoped An inclusive measurement of  0 and  yields in BB events to a precision of a few percent has proved more difficult than originally envisioned, but we feel things are under control We have improved continuum background suppresion with a Fisher Discriminant and studyied the associated systematic effects We are converging towards a new result and hopefully it will be available soon.