1. Photon reconstruction and calorimeter software Mikhail Prokudin

2. Outline ► Calorimeter software development  photon reconstruction ► cluster finder ► simple reconstruction  UrQMD events  matching  Calorimeter drawing tools ► Cluster fitting  requirements ► Conclusions ► Next steps

3. Photon reco. Requirements ► Robust reconstruction of single photons ► Two close photons case:  robust reconstruction of parameters in case two separate maximums  separation one/two photons in case of one maximum ► Fast!

4. Cluster finder Cluster formation ► Remove maximums near charged tracks  real tracking ► Precluster:  formed near local maximum ► cut on maximum energy  find maximum 2x2 matrix near maximum  add a neighbor to local maximum cell with minimal energy deposition ► to add inromation  check precluster energy ► >0.5GeV ► Cluster: group of preclusters with common cells 4870 Central UrQMD Au+Au 25GeV CbmEcalClusterFinderV1. At SVN Requirements ► Clusters should be large  information for unfolding ► Clusters should be small  hadrons background

5. Cluster finder performance Benchmark ► 2x1 GeV photons ► 3x3 cm cells  inner calorimeter region ► 2-10° angle  geometry of inner calorimeter region

6. Simple reconstruction ► Energy:  calibration ► only energy in scintillator is visible ► Position:  S-curves ► χ 2 calculation for reconstructed photon CbmEcalRecoSlow. At SVN

7. UrQMD events with simple reco Subtraction of mixed events is necessarily! Invariant mass spectra True Mixed Gamma spectra

8. Simple reconstruction ► Robust reconstruction of single photons ► Two close photons case: ► occupancy  robust reconstruction of parameters in case two separate maximums  separation one/two photons in case of one maximum ► Fast! ► Need more complex reconstruction!

9. Matching ► Why?  check origin of cluster ► neutron clusters  physics processes ► π 0 and η decays ► prompt photons ► … ► Most simple method at moment ► Energy deposition in the cluster > 70% of cluster energy  γ /e is secondary also look for mother ► showers started before the calorimeter treated correctly  loss clusters with more than one maximums CbmEcalMatching. At SVN

10. Matching. Usage example π0π0 η π 0 born after IP (conversion)

11. Calorimeter drawing tool ► Draw  calorimeter structure  energy deposition in calorimeter  reconstructed tracks ► and energies  reconstructed photons ► energies ► and matched MC particles  clusters ► found approximation quality ► and χ 2 of cluster  MC tracks ► type (photon, neutron …) ► energy ► …and all at one picture!

13. Calorimeter drawing tools ► Photons  MC  Reconstructed ► * ► (Anti) neutrons ► Charged tracks  Reconstructed ► *  MC ► Secondary  Photons  Electrons CbmEcalQualityCheck. At SVN

14. Fitter requirements ► Robust reconstruction of single photons ► Two close photons case:  robust reconstruction of parameters in case two separate maximums  separation one/two photons in case of one maximum ► χ 2 criteria ► all analyzed approaches have failed to reconstruct photons energy/position correctly ► χ 2 shape should not depend on photon’s energy  same value for efficiency cut ► separation power of one/two photons in case of one maximum as a criteria  example: with 95% efficiency for clusters formed by single photon

15. Realization of fitter ► CbmEcalRecoSlow ► current version at SVN  no so slow actually! ► 40 sec per UrQMD event  first approximation ► CbmEcalRecoSimple  χ 2 minimization ► minimizer  TFitterMinuit ► shower shape (E pred )  shower lib ► σ 2 formula

16. Shower width ► Energy deposition in cluster cells are not independent  storing of RMS in shower library useless ► Analytical formula  with correlation ► ALICE  σ 2 =c 0  (E meas +c 1 ) ► no correlations! ► PHENIX  σ 2 =c 0  (E meas  (1- E meas /E cluster )  (1+k  sin 4 θE cluster )+c 1 ) ► correlations are in  Angle dependence ► shower library h4 h5 h4 h5

17. σ 2 formula ► σ 2 formula declared in configuration file ► … and parameters too  for easy change ► without recompilation  different formulas for different cell types ► … to maintain commonness ► Sum of photon’s energies fixed to energy of cluster  a switch in configuration file ► Parameters space is huge  σ 2 formula ► best c n could be computed if σ 2 formula is fixed  also parameters of cluster finder # Number of cells types for reconstruction types=4 # Number of constants for each type consts=2 # Use Ecluster, Emeas and Epred for measured cluster energy, measured cell energy and predicted cell energy respectively c0_1=0.008 c1_1=0.0016666 c0_2=0.008 c1_2=0.00345 c0_3=-1111 c1_3=-1111 c0_4=0.008 c1_4=0.0043333 sigma_1=c1*(Emeas*(1-Emeas/Ecluster)+c0) sigma_2=c1*(Emeas*(1-Emeas/Ecluster)+c0) sigma_3=-1111 sigma_4=c1*(Emeas*(1-Emeas/Ecluster)+c0) # if chi2 for cluster is less than no fitting chi2th=-1111 # Max iterations in fitting process maxiterations=1000 # Steps for calculation of gradients estep=0.0001 cstep=0.0005 # Fix sum of energies of cluster particles to energy of cluster fixclusterenergy=1 # # Cluster finder stuff # # Maximums belong to charged tracks should excluded? removecharged=1 # Minimum precluster energy minclustere=0.3 # Minimum energy of precluster maximum minmaxe=0.2 # An algorithm for preclustering: 0 --- default, 1 --- PHENIX like, # 2 --- ALICE like, 3 --- default, but remove low energy cells preclusteralgo=0 # Minimum cell energy mincelle=0.020 # Minimum size of precluster minsize=4 # Attach to cluster nearby cells with Edep>fMinCellE attachcells=0.1 Example of configuration file

18. Photon reconstruction. Merged photons Simple reconstruction Cluster fitting Fitting of clusters with two maximums allows us disentangle photons! σ 2 formula and parameters does not require much tuning!

19. Fitter requirements ► Robust reconstruction of single photons ► Two close photons case:  robust reconstruction of parameters in case two separate maximums  separation one/two photons in case of one maximum ► χ 2 criteria ► all analyzed approaches have failed to reconstruct photons energy/position correctly ► χ 2 shape should not depend on photon’s energy  same value for efficiency cut ► separation power of one/two photons in case of one maximum as a criteria  example: with 95% efficiency for clusters formed by single photon

20. χ 2 distributions. Single photons 95% 1 GeV95% 4 GeV σ 2 =c 0  (E meas  (1-E meas /E cluster )+c 1 ) c1=0.0005 95% 1 GeV95% 4 GeV σ 2 =c 0  (E meas +c 1 ) Shape of χ 2 for each energy looks Ok, but cut with 95% efficiency has different value! Need a different σ 2 formula!

21. Rejection power. Inner region σ 2 =c 0  (E meas  (1-E meas /E cluster )+c 1 )σ 2 =c 0  (E meas +c 1 )

22. Rejection power. Outer region σ 2 =c 0  (E meas  (1-E meas /E cluster )+c 1 )σ 2 =c 0  (E meas +c 1 ) Reconstruction in outer region is most sensible to σ 2 formula!

23. Conclusions ► Calorimeter software development in progress  Cluster finder: CbmEcalClusterFinderV1  Reconstruction: CbmEcalRecoSimple ► and CbmEcalRecoSlow  Matching: CbmEcalMatching  Quality check: CbmEcalQualityCheck

24. Conclusions ► Photons reconstruction  simple procedures are ready to use  more complicated procedures ► not too slow ► fit clusters with more than one maximum ► still have limited usability  σ 2 formula ► bad cluster rejection ► not trivial ► All presented calculations done using 2 computers  UrQMD transport, reconstruction, etc. ► 3.0 GHz Core 2 Duo ► 2.0 GHz Core 2 Duo (My laptop)

25. Next steps ► Reconstruction tuning ► Detector optimization  geometry  segmentation ► Detailed sensitivity studies for process with photons  π 0, η, χ c … ► e/ π separation with real tracking ► Detailed detector geometry  construction details  light collection efficiency