1. A more realistic photon simulation in Geanfi Capri 24-5-03, Kloe Phys Workshop III Motivations: A better description of shower simulation for photons and a “standard” recipe to recover the data-MC discrepancy needed to : - attain a 0.1% error on K s  0  0 - measure rare K s decays in neutral channels - parametrize response /resolution for Kinematic Fitting Four main tasks : - selection of Machine background events - insertion of Machine background events in MC data - survey of existing data-MC discrepancy - correction-adjustment of MC simulation By: C.Bloise, M.Martini, S.Miscetti, M.Moulson, T.Spadaro... S.Miscetti @ KPW3

2. Differences with last year bkg simulation (ACCELE + MBCKADD)  Use a unique selection sample ( e  e    from FILFO) for evaluation of DCH, EMC, QCAL machine background ( bias found in mbckadd where the usage of L3BHA events was providing events already cleaned of out of time clusters ! )  Study side-bands (early/late) of T-R/c to measure rate of accidental events but extend the selection to the whole time spectrum & inserting hits for the 3 detectors at the same time. All correlations preserved.  Closely follow the rate dependence along the run by selecting events of “same machine activity” in groups of few pb -1. The background is inserted in MC on a run by run basis. Finally... BGG stream created! Insertion in MC weighted with the effective cross section of BGG selector. SELBKG & INSERT: basic ideas S.Miscetti @ KPW3

3. Two clusters in the barrel with:  TW ( 5  T, 1 ns)  Ecl > 480 MeV  Etot > 960 MeV   * cuts   z cuts  Vclu cut Etot (1+2) (MeV)  T (ns) Still a lot of mess in the “IN TIME” window SELBKG : selection of golden events S.Miscetti @ KPW3

4.  T (ns)  R(cm) Once “clean” clusters have been selected we look for all “residual” ones looking at all combinations of DRij vs DTij between the “clean” and “residual” clusters.   (  )  Flying clusters (shower fragments)  splitted clusters 3 categories SELBKG : looking for the machine bkg events S.Miscetti @ KPW3 Fiducial region

5. After golden clusters selection & restricting the search to the fiducial region:  pieces of collision remain in the c0,c1,c2 bands for single hits.  comparing the E vs cos  plot for early band and cj bands a set of 2D weights is found.  Nice agreement found in all distributions after applying these 2D weights SELBKG : survey distributions after weighting tecnique 1 hit  1 hit Early Timesc0 c1 c2 bands S.Miscetti @ KPW3

6. SELBKG : survey distributions after weighting tecnique II S.Miscetti @ KPW3 solid line (early-band) black dots (Cj bands) C0 C1 C2C0,C1,C2 C1 C2 C0,C1,C2

7. INSERT: two bugs found in TSKT & SmearT0 fixed last week:  now times can be negative  correction for att in EMC and space2time relation for DCH close to reality.  overlap between accidental and MC hits tested at first order both for tracking &clustering INSERT: inserting BGG events inside MC events SELBKG technique applied to groups of few pb -1 to follow closely the machine activity along the run. BGG stream DONE! S.Miscetti @ KPW3

8. Tests of insert bgg done with MC PRE-Production T-R/c behaviour with /without t0_finder: enhancement  0 for bgg satisfying t0_finder ! otherwise correctly bunched Tests to be done for post-production are: energy slope hits multeplicity effects on Filfo effect on rec.efficiency INSERT: inserting BGG events inside MC events S.Miscetti @ KPW3

9.  EMC geometry closer to the real one (Barrel moved down inY of  1 cm)  Survey of existing data-MC discrepancy done! Most of the checks based on       sample. Other tests in progress with  & Bhabha’s samples. A lot of differences data-MC found.  Scale and Resolutions: - Energy resolution - Timing resolution - Non linearity of response - behaviour along cracks (  barrel, X EndCaps) - behaviour along Y ( attenuation lengths EndCaps)  Shower shape - X 0 simulation  Light Yield - Ea/Eb - Efficiency ? Toward a more realistic MC description of EMC S.Miscetti @ KPW3

10. Energy response and resolution vs E  DataMC 5.7%/  (E/GeV) 5.0%/  (E/GeV)  Energy scale is set better in data than in MC. Still contained below 1%  Non linearity better in data than MC (clufixene effect)  Stoc.term in resolution much better in MC S.Miscetti @ KPW3

11. DATA MC Energy response vs  Data show clear  -cracks + calib error around  = 0 S.Miscetti @ KPW3

12. Energy response vs MOD( ,15) crack effect much more visible if shown vs mod( ,15) S.Miscetti @ KPW3

13. Energy response & resolution: no crack vs crack zone As shown by profile there is a lowering in response in the crack. Integrating 1 degree around the crack the effect is of  –2% in response (some energy dependence ?) slightly worse resolution! S.Miscetti @ KPW3

14. Reconstruction efficiency: no crack vs crack zone No effect on efficiency data MC differ up to 70 MeV S.Miscetti @ KPW3

15. Also on Endcaps cracks are visible along X. Moreover, there is a miscalibration of response along Y at a level of 3-4 % Energy response in EndCaps: dependence vs Y S.Miscetti @ KPW3

16. Shower Shape: measurement of the decay length (data) N(x) = N 0 e (-x/ ) (1) with x the EMC depth. The distribution of the first plane fired fit with (1) in bins of   as a function of E . E  dependence of cross section as expected. “plateau” region E   250 MeV We derive = 1.5 cm X 0 = 7/9  1.2 cm S.Miscetti @ KPW3   (degrees)

17. Shower Shape: measurement of the decay length (MC) Following same description X 0  1.5 cm (MC) MC composition: 23 cm with Pb/Sc slabs of 0.11 cm (400  m/700  m) X 0 = 1.53 cm Fs(Mip) = 24% Fs(e.m.) = 14.5% Same exercize on data: 200 planes of 500  m Pb S.Miscetti @ KPW3   (degrees) 1/X 0 = F pb /X 0 pb +F sc /X 0 sc Great agreement! X 0 = 1.2 cm EMC in MC too light!

18. Measurement of light yield (MC) S.Miscetti @ KPW3 Two techniques used to get N pe Gaussian fit to the distributions: R = (Ea/Eb)  /S = (Ea-Eb)/(Ea+Eb) for Barrel cells fired @ Z=0. Assuming fluctuaction to be dominated by Poisson on Npe   v =  (2/Npe)  /S method more stable. we get 1.2 pe/MeV in scint @ calorimeter center. In agreement with light yield used in MC 25pe*0.145*0.33  (2./(1.2xE/MeV))  0.11  (2./(1.5xE/MeV))  0.12

19. Measurement of light yield (data) S.Miscetti @ KPW3 Using same technique on data sample we find Npe = 0.6-0.7/MeV i.e. 600-700 pe/GeV/side  50% of exp. light yield - 20% drop due to B-field - estimated @ CRS using time resolution and  (scint) - effect of light yield on energy resolution small. ( in operation we found less than extrapolating from CRS and TB )  (2./(0.60xE/MeV))  0.12  (2./(0.65xE/MeV))  0.12

20. S.Miscetti @ KPW3 Measurement of time resolution (data) Usual numbers: - stoc. term 57 ps/sqrt(E/GeV) - const.term 140 ps Stable since 1999 on MC - const.term  0 - stoc.term 50 ps/sqrt(E/GeV) much more light in MC still not a factor 2 E  (MeV)  T-R/c  (ns)  T (ns)

21.  Radiation length on data 1.26 cm ( vs 1.53 cm MC )  Light yield 0.6-0.7 pe/MeV ( 1.2 pe/MeV )  Sampling fraction used in MC for showers 14.5 %  Stoc.term of energy resolution lower in data than MC (5.7% vs 5.0 %)  Not recoverable only with difference in Npe  Stoc.term of time resolution 57ps (data) vs 50 ps (MC) S.Miscetti @ KPW3 Summary review of information in our hands missing description of holes and behaviour along Y reconstruction efficiency data lower than MC 

22. Geanfi simulation was adjusted in many steps varying : 1) X 0 lead thickness from 400 to 500  m 2) light yield lowered from 25 to 19 pe/MeV 3) adjusting sampling fraction in rec for each selected X 0 4) fastening the Time emission distribution and recalibrating MC T0s for each choice. 5) constant term of 140 ps add + Two new routines in MC path: EMCSIMULA: Adding holes CLUADJUST called in clufixene - fixing  calib-hole and attenuation lengths - flagging clusters to be killed to simulate obs.efficiency (only  ) S.Miscetti @ KPW3 Adjustements done in MC and Reconstruction

23. S.Miscetti @ KPW3 Comparison old vs new MC ( fo  ) events Barrel Barrel holes EndCaps Resolution OK, holes ok bug found on EndCap.. now fixed

24. S.Miscetti @ KPW3 Old vs New MC: dependence along 

25. S.Miscetti @ KPW3 Old vs New MC: dependence along X and Y

26. S.Miscetti @ KPW3 New MC: simulation of cluster efficiency for 

27. MC meeting LNF 23-1-03 S.Miscetti @ KPW3 Conclusions The new MC follows much more closely the details of cluster reconstruction and shower development in the EMC: energy and time resolution OK Non linearity in response similar. holes and other small details simulated shower shape OK inefficiency still a trouble. Forced imposing measured efficiency. simulation of mach background OK for all detectors (much more tests expected for post-mc production Slopes, rates of accidentals ) Much more testing also necessary on masses of neutral objects