Gael Rospabe Lapp 15/04/08 CaloSoft Meeting 1 Ecal calibration using  0 Sabine Elles/ Marie-Noëlle Minard/ Gaël Rospabé

Gael Rospabe Lapp 15/04/08 CaloSoft Meeting 2 Outline  Sample   0 method  De-calibration  Re-calibration  Results  Perspectives

Gael Rospabe Lapp 15/04/08 CaloSoft Meeting 3 Sample  MinBias 10 6 evts  Cf talk in december and next one : fisher for Ecal in stand alone  Cuts :  Pt  > 0.3GeV  Fisher (stand alone) > 0.15 => eliminate non photon clusters  Dist  1  2 eliminate bad combinatories  Gives ~1.8  0 /100events

Gael Rospabe Lapp 15/04/08 CaloSoft Meeting 4 Cut effects Eff = 80% Rej = 50% Eff = 65% Rej = 80% Dist < 500mm Eff = 90% Rej = 95%

Gael Rospabe Lapp 15/04/08 CaloSoft Meeting 5 Sample No corrections : -Shifted mass (124.6MeV) - large sigma (11MeV) - because no corrections applied  0 /100evts - S/B = 3 Applying corrections : Mass : 133.2MeV  = 9MeV With the L0 : - get 0.3/100evts - S/B = 3 => No need to apply it

Gael Rospabe Lapp 15/04/08 CaloSoft Meeting 6 Calibration method Based on the disintegration  0 ->  Clusters : 2E 1 E 2 (1-cos  12 ) = m 2 00 E1and E2 deposited energy for each photon  12 angle between the 2  Most energetic cell into the cluster ( = seed) Because can bring bias (correlation between two corrections) In practice : i = index for cells into the cluster c i = calibration coefficient e i = deposited energy E 1(2) = C i,1(2) e i,1(2)  2E 1 E 2 (1-cos  ) = m 2   (correction) = m / m 00 C n+1 = C n *correction i,1(2) Where n is the iteration index

Gael Rospabe Lapp 15/04/08 CaloSoft Meeting 7  Biases come from :  cells shared by several clusters  clusters between two regions of the Ecal  extended clusters  Solutions : adding others selections  only clusters with more than 70% of total energy into the seed  applying energy corrections only to the seed and taking into account other corrections on other cells  0.1<  0 mass < 0.17GeV : less than 20% of error  => 1  0 /100events Calibration selections

Gael Rospabe Lapp 15/04/08 CaloSoft Meeting 8 Effects of the cuts on the clusters Cuts on Pt  > (0.3GeV), Fisher (> 0.15), Dist  1  2 < 500mm and % of total energy into the seed make us keeping compact clusters Without cuts Pt, dist, fisher cuts + seed cut

Gael Rospabe Lapp 15/04/08 CaloSoft Meeting 9 Calibration Algorithm Cells grouped in 3x3 patterns for each region 9 random numbers for each region (vs Gaussian) Patterns reproduced along x & y axis cells de-calibrated in function of position in 3x3 patterns 2 consecutive cells with  de- calibration coefficients seed coefs only changed : no bias and correlations decalibration level can be chosen Detector de-calibration : Method : getting closer from “reality” (realistic data tacking and first mis-calibration) : de-calibration / re-calibration Outer Middle Inner

Gael Rospabe Lapp 15/04/08 CaloSoft Meeting 10 Outer region Inner regionMiddle region De-calibration In each region : 9 coefficients => 9 distributions In total : mean=120MeV  = 26MeV (red =initial distributions) m=120MeV  = 26MeV

Gael Rospabe Lapp 15/04/08 CaloSoft Meeting 11 Calibration algorithm Detector calibration : Different parameter evolutions are checked : mean (mass), sigma of mass distributions, and dispersions for each region  when for a given region stop criteria (see next transparencies) are satisfied : stop to re-calibrate this region and only going on for non-re-calibrated ones Distribution of calibration coefs : 9 distributions / region Nowadays: - a  0 is randomly selected into the sample - fill the dedicated histogram ( position into the 3x3 pattern) with it’s reconstructed mass - when one contains 1000  0 : Gaussian fit => gives mean value : m which is m 2 = 2E 1 E 2 (1-cos  12 ) - correction is calculated : - the correction is applied giving 2E’ 1 E’ 2 (1-cos  ) closer to m 2 00 m m  = 00

Gael Rospabe Lapp 15/04/08 CaloSoft Meeting 12 Coefficient re-adjustment 9 examples of distributions of the mean : - not very Gaussian - not centered on m - distributions fitted by a Gaussian - give us the correction to apply on the cell(s)

Gael Rospabe Lapp 15/04/08 CaloSoft Meeting 13Gael Rospabe Lapp 15/04/08 CaloSoft Meeting 13 How to stop/evolution of the calibration 3 variables to check : - evolution of the mean of the distributions - the  of the distribution - the dispersion Dispersion (Disp) is defined by : Disp = (mean – m bin ) 2 x Bin content Stop iterative process for one region when system stables over 4 quantities (0.1%) : -Correction factors - mean, -  and dispersion for global distribution and for each region  bin

Gael Rospabe Lapp 15/04/08 CaloSoft Meeting 14 How to stop/evolution of the calibration Oscillations of the mean : - over-corrected - corrected Diminution of the mean and dispersion Inner oscillates more than other regions Middle stabilizes faster than others

Gael Rospabe Lapp 15/04/08 CaloSoft Meeting 15 Results for the total Ecal m=135 MeV  = 9 MeV m i = 120MeV Even if it looks well calibrate d before the end it is not !! Because of the addition of the 3 regions

Gael Rospabe Lapp 15/04/08 CaloSoft Meeting 16 Results for the total Ecal m=135 MeV  = 10 MeV m=135 MeV  = 8 MeV m=135 MeV  = 11 MeV

Gael Rospabe Lapp 15/04/08 CaloSoft Meeting 17 Conclusions  Iterative method : correction of the mass  obtain resolution of 1% on each cell  Time 1753 seconds (~1/2h)  CPU seconds (~5mins)*2000/9=18h  Number of iterations : ~200  # of needed  0 = 1000/cell => ~10 7  0  with Fisher and selections 1.  0 /100evts => evts  Calibration time and iterations function of de- calibration