1. Status of calorimeter simulations Mikhail Prokudin, ITEP

2. Outline ► π ˚ reconstruction quality ► Cluster finding  requirements  2 approaches ► Shower library ► Simplified MC supporting design studies   by A. Chernogorov ► Next steps

3. π ˚ reconstruction quality ► Standard 25GeV AuAu UrQMD event ► γ from same π ˚ ► Smooth with  7%/sqrt(E) energy resolution  3cm position resolution  both RMS 2.627e-3 RMS 6.669e-3 RMS 6.692e-3 3cm position resolution doesn’t affect the reconstruction quality!

4. Requirements for cluster finding procedure ► Fitting routine  first approximation ► energy  calibration ► position  S-curves ► number of photons  number of local maximums in cluster  iterative procedure ► shower shape  shower library ► shower width ► Cluster finding  maximize cluster size ► use all possible information for fitting ► minimum 3 cells for cluster with 1 maximum  3 degrees of freedom per photon  occupancy ► hadron contamination  minimize number of maximums ► less parameters in fit  stability

5. Cluster unfolding. Naive approach ► Color: energy deposition ► Edging:  Green: clusters  Red: excluded by charged tracks ► Cluster definition:  connected area of cells with energy deposition above the threshold ► Threshold ~150MeV BTW, Real tracking!

6. Cluster finding. Naive approach ► Cluster definition:  connected area of cells with energy deposition above the threshold  threshold can be found from MC ► Occupancy differs ~10 2 times ► partially compensated by segmentation  lost clusters in periphery of calorimeter  create too large clusters in central region ► No chance to fit cluster with >7 peaks  central region is lost! ► Clusters with size <3 cells require additional information ALICE approach 25 UrQMD events Peaks per cluster

7. Cluster finding ► Find peak above threshold  peaks associated with charged track are removed from the consideration ► Construct an area with certain number of cells around the peak  it should contain 2x2 submatrix with maximum energy of 3x3 matrix around peak  size determined by ► reconstruction quality of single photon  at least 3 cells for reconstruction  at least 4 cells for unfolding ► occupancy ► Find areas intersecting with given area ► Construct a cluster PHENIX-like approach

8. Cluster unfolding ► Color: energy deposition ► Edging:  Green: clusters  Red: excluded by charged tracks ► Less clusters in central region with new algorithm ► Cluster size under control

9. Comparison Naive New25 UrQMD events Number of peaks per cluster lower using new approach Peaks per cluster

10. Conclusions for clustering algorithm ► Naive approach does not work  occupancy variations ► New (PHENIX-like) approach  number of peaks in cluster under control  less sensitive to occupancy ► Real tracking

11. Shower shape ► Analytical formula for shower shape approximation  ALICE and PHENIX experience  No memory consumption ► Best for multicore CPU  Poor quality for large incident angles ► Shower library  Fits exactly to the data  Requires a lot of memory ► (and CPU!)  Any incident angle  checking analytical approximation quality ► Shower width  Also needed for fitting procedure  Approximation or storing in library?

12. Shower library ► ECAL with very high segmentation (1x1mm 2 volumes)  use one shower multiple times  volumes merging procedure ► Transport photons for every:  Energy  Theta  Phi ► Eightfold decrease due to symmetry ► For low energies we have to generate more showers ► Larger fluctuations for low energy showers  10k for 0.49 GeV  2k for 16GeV

13. Volumes merging ► During shower library creation  Faster during reconstruction  Also can store RMS (see next slide)  Bigger library size  Cell size fixed ► Different data set for each cell size ► During reconstruction  Homogenous dataset ► Cell size not fixed  Information about RMS is lost ► Need analytical approximation for errors  More CPU required during reconstruction

14. Shower width ► Energy deposition in cluster cells are not independent  RMS value storing useless ► Need an analytical formula  correlations should be in! ► … h4 h5 h4 h5

15. Shower library ► 170 Mb disk space ► 300 Mb in memory ► Energy: 0.49, 1, 2, 4, 6, 9, 12, 16 GeV ► Theta: 2˚, 3.5˚, 5˚, 6.5˚, 8˚, 9.5˚, 11˚, 13˚, 16˚, 20˚, 24˚, 28˚, 32˚ ► Phi: 0˚, 10˚, 20˚, 30˚, 40˚ See CbmShowerLib class 6x6cm 2 cells

16. Shower library 12x12cm 2 cells ► Shower rotating on fly?  classical trade CPU vs. memory

17. Simplified MC supporting design studies ► Constructive details  steel band  dead material in front  fibers insertion ► Necessarily for design works  can be used in CBMROOT simulation Ex. 1. SS bands between modules on the ECAL response. Eγ=16 GeV, θ=3° Pb=Sc=1mm. 1- ▲, 2-*, 3-● Steel band effects by A. Chernogorov

18. Dead material in front of the calorimeter ► Calorimeter  200 layers  1 mm Pb + 1 mm Sc ► Absorber  C, Fe, and Cu studied  0.0 X 0, 0.2 X 0, 0.4 X 0, 0.8 X 0  distance to ECAL 2-10m by A. Chernogorov ECAL 5m γ Example: Fe, 5m

19. Conclusions and next steps ► Real tracking in calorimeter ► Studied different approaches to cluster finding  naive one does not work ► Shower library is complete  realization in SVN and AUG07 ► Unfolding procedure under (extensive) development  analytical formula for shower width ► with correlations  can be submitted to SVN any moment ► Design supporting MC studies under way  by A.Chernogorov

20. Procedure of γ reconstruction ► First approximation  energy ► calibration  position ► S-curves ► Cluster unfolding  cluster finding procedure  fitting routine ► shower shape  shower library ► LHCb like methods ► Pure γ, no background ► Simple and easy to check ► Test site for shower library routines ► Can be done in few month ► ALICE PHENIX-like methods ► Require much more effort  CALO parameters should be fixed? From September 2006 CBM collaboration meeting Done From September 2006 CBM collaboration meeting