1. Preliminary results for the BR(K S  M. Martini and S. Miscetti

2. Talk Layout Short summary of strategy for the measurement DATA-MC QCAL calibration Signal/background fit repeated in different conditions: - with cos(  ) - without cos(  ) Determination of efficiency for the signal Study of normalization sample Preliminary estimate of BR First look at background shapes Plans/prospects

3. Data sample and preselection We have analized 1.6 fb -1 of DATA (2001-2002-2004 and part of 2005 sample).  400 pb -1 still missing on 2005. whole production of neukaon MC 2001-2002 used for the bkg (450 pb -1 ) ksr04 used for the signal Started using the preliminary sample 100 pb -1 of the 2004 MC. Not yet for shapes.. Checking rates only. From NA48 results: BR(K S  ) = 2.78 x 10 -6, we expect to have tagged 1550 signal events with Kcrash. Kcrash events Preselection: consists of requiring 2 “and only 2” prompt photons with E  >7 MeV, cos(  ) < 0.95 and T<min(5 , 2 ns) Qcal veto

4. DATA BKG Example fit 2d chi2... - K S tagged from Kcrash (122 x 10 6 events) - 2 prompt photons required (  496000 events) The major background is constituted by K S  2  0 with 2 lost photons. To disentangle signal from background we use: - Kinematic fit (  2 <20) - We then look at the scatter plot M  vs  , where: -    Opening angle between the two photons in the K S cms -    Reconstructed  mass

5. DATA SIG Example fit 2d chi2... - K S tagged from Kcrash (122 x 10 6 events) - 2 prompt photons required (  496000 events) The major background is constituted by K S  2  0 with 2 lost photons. To disentangle signal from background we use: - Kinematic fit (  2 <20) - We then look at the scatter plot M  vs  , where: -    Opening angle between the two photons in the K S cms -    Reconstructed  mass

6. After splash filter events with 3, 4, 5  prescaled of 400 QCAL  (-5 <  Tqcal < 5) ns Data vs MC: QCAL rates (2001-2002) DT=Tqcal-Rqcal/c (ns)

7. After splash filter events with 3, 4, 5  prescaled of 400 QCAL  (5 <  Tqcal < 35) ns DT=Tqcal-Rqcal/c (ns) Data vs MC: QCAL rates (2004-2005)

8. Extraction of losses... -We defined two windows, early and late in  T: - Early: (-50 ; -40) ns - Late: (70 ; 80) ns TOT events After splash filter Event in window Ploss in window N  =2,3,4 weighted mean

9. Extraction of losses... - For the moment we have only used the early window (we can use late fraction as systematic) - since we have difference between 2001-2002 and 2004-2005 sample, we calculate different values of Ploss: 2001-2002 Ploss = (4.85  0.07)% 2004-2005 Ploss = (15.7  0.07)% -  QCAL is evaluated as:  QCAL (DATA) = 1 - Ploss

10. QCAL data/MC efficiency - For each period all numbers with N  =2,3,4 fit with the following stuff... - we calculate the ratio: We found compatible value of R for the different DATA sample Qcal vetoed DATA Qcal vetoed MC

11. QCAL data/MC efficiency results Using the results on Ploss for the different DATA sample, we can extract the QCAL efficiency: Sample  QCAL (DATA) 2001-2002 (95.16  0.07)% 2004-2005 (84.23  0.07)%

12. MC checks for QCAL efficiency QCAL rejected events: 1)ALL events 2)Cos  accepted events 3)Cos  rejected events 4)Energy of accepted events

13. MC checks for QCAL efficiency QCAL survived events: 1)ALL events 2)Cos  accepted events 3)Cos  rejected events 4)Energy of accepted events

14. MC checks for QCAL efficiency We still have events impinging the QCAL that survived QCAL cut. We can improve the simulation studying these events. cos  >0 ; cos  <0

15. Signal-background fit 2001-2002 sample weights from fit DATA -- MC all Signal Background

16. Signal-background fit 2004-2005 sample weights from fit DATA -- MC all Signal Background

17. Fit using costheta (preliminary) cos  2001-2002 sample: comparison without and with cos  in the fit cos 

18. 2004-2005 sample: comparison without and with cos  in the fit cos  Fit using costheta (preliminary)

19. 2001-2002 sample: comparison without and with cos  in the fit cos  Fit using costheta (preliminary)

20. 2004-2005 sample: comparison without and with cos  in the fit cos  Fit using costheta (preliminary)

21. Fit results and analysis efficiency Sample2001-20022004-2005 Nsig 143.9  20.1462.8  34.7  (  2 )=(63.3  0.7)% The analysis efficiency (  2 cut after kcrash and acceptance selection) is the same for the two samples since up to now we have used MC 2001-2002 only. :

22. Signal Acceptance, Total Efficiency Using KSR04 MC production, we evaluate the signal efficiency requiring KL-far events and counting events with N  =2 Using the standard efficiency curves, we obtain:  ACC)(N  =2) = 83.2  0.2 stat (1) The systematic error has been evaluated varying the cone (0.6, 0.7, 0.8) and using the maximum variation from (1):  ACC  (N  =2) = 83.2  0.2 stat  0.1 sys  tot (kcrash given) =  (acc) *  (qcal) *  (ana) For the moment statistics and systematics together.

23. Normalization sample Kcrash with N  =4, N  =3,4,5 prescaled of 400. Splash filter applied. Stability plot shown with N  = 4 SampleInt. Luminosity 2001158 pb -1 2002189 pb -1 2004748 pb -1 2005527 pb -1

24. Kcrash counter stability (2001-2002) 20012002

25. 20042005 Kcrash counter stability (2004-2005)

26. K S  2  0 efficiency Using a sample of 160 Kevents, extracted from 2001 and 2002 statistics, we calculate K S  2  0 efficiency using events with a KLfar definition:  (N  =2) = ( 65.0  0.02 stat )% Using the same method applied for the signal, we can evaluate a first systematics on this parameter:  (N  =2) = ( 65.0  0.2 stat  0.1 sys )%

27. Kcrash Final normalization Using K S  2  0 efficiency, we can extract the number of Kcrash of the normalization sample Total number of Kcrash: 159.8 x 10 6 We can compare this results with Ncrash obtained integrating N  = 3, 4, 5: Ncrash(3, 4, 5) = 159.5 x 10 6 SampleNcrash 2001 15.83 x 10 6 2002 18.91 x 10 6 2004 74.58 x 10 6 2005 50.46 x 10 6

28. First BR estimate  TOT (2001-2002) = (50.1 ± 0.6)% Ncrash = 34.7 x 10 6 BR(K S  2  0 ) = (31.05 ± 0.14)% BR(K S   ) = 2.57 x 10 6  TOT (2004-2005) = (44.4 ± 0.5)% Ncrash = 125.1 x 10 6 BR(K S   ) = 2.59 x 10 6 Combined result: BR(K S   ) = (2.58 ± 0.17) x 10 6

29. Fast simulation of Background To study the fit uncertainty as a function of MC statistics we have developed a method based on “hit or miss”. The procedure is only based on MC signal and background. Recipe: - use the original 2d-distribution from sig and bkg, to create 2 smoothed distribution - Use hit or miss to create N different distribution for signal and background for different MC statistics - create a fake data distribution using sig and bkg from hit or miss with entries from fit - repeat the fit procedure N times for each statical point.

30. Fast simulation of Background Metti qualche plot di preparazione per hit or miss

31. Hit or miss Signal and bkg statistical error as a function of the used MC statistics Actual stat: Mcfact=1 Using twice MC statistics we can lower signal uncertainty of a factor 10%

32. Hit or miss Stability of signal and bkg event as a function of MC used statistics.

33. Look at background shapes DATA-MC comparison for bkg enriched samples with  2 > 100, ?

34. Plans-prospects We need to study the systematics on the spectra shapes: 1) apply MC energy scale 2) calibration check with background dominated samples 3) calibration with K S  2  0 4) effect of DATA-MC differences on QCAL efficiency We will process the few missing pb -1 of data and the 2004-2005 MC generated so far. We are working on “fixing” the QCAL simulation to answer to point 4) Study on the tag bias Meeting with referees Start writing documentation and planning for a pre-xmass blessing