1. December, 16 20081 Calibration of electromagnetic calorimeter of Hall A DVCS experiment Eric FUCHEY Ph.D Student @ Laboratoire de Physique Corpusculaire UMR 6533 CNRS/IN2P3 Université Blaise Pascal Clermont-Ferrand

2. December, 16 20082 Outline 1- Motivations of calibration 2- Real data calibration –Method –Results 3- Simulated data calibration –Method –Results 4- Conclusion

3. December, 16 20083 Outline 1- Motivations of calibration 2- Real data calibration –Method –Results 3- Simulated data calibration –Method –Results 4- Conclusion

4. December, 16 20084 Experimental layout Calibration for ep->ep  0 Incident beam : E = 5.75 GeV during experiment –Scattered electron measured by HRS –2 photons issued from  0 measured by 132 PbF 2 blocks calorimeter –Recoil proton detected by proton array -> not used

5. December, 16 20085 Looking at real and simulated  0 data sets We measure We can reconstruct for each event a missing mass: with and We can either reconstruct for one block a missing mass distribution for all events with one photon hitting this block.

6. December, 16 20086 Mp2Mp2 Reconstructed missing mass distribution in each block depends on block position in calorimeter. Effects assumed to come only from calorimeter. => necessity to calibrate calorimeter to correct these effects.

7. December, 16 20087 For analysis, we need to fit  0 data set with  0 simulated set to extract cross section unconstistency of resolutions for real and simulated. => necessity to calibrate and to smear simulated.

8. December, 16 20088 Outline 1- Motivations of calibration 2- Real data calibration –Method –Results 3- Simulated data calibration –Method –Results 4- Conclusion

9. December, 16 20089 Method of calibration (real) Goal of calibration: to find for each block of the calorimeter a coefficient in order to reconstruct in each block a missing mass squared in accordance with M p 2. Coefficient of the form: for block#  (  = 0,…,89)

10. December, 16 200810 Method of calibration (real) For an event i, a photon is in block #a, the other is in block #b. Correcting each photon energy: Recomputing M X 2 with these corrected energies, and filling #a and #b blocks distributions.

11. December, 16 200811 Method of calibration (real) Fit distributions by a gaussian -> Mean of the gaussian => New missing mass squared value.  New calibration coefficient Correct 2 photons to reconstruct missing mass -> calibration coefficients for each block are correlated.  necessity to calibrate by steps.  necessity to estimate calibration quality for each step.

12. December, 16 200812 Corrections Before filling distributions, we apply some corrections: –By effects of resolution, missing mass squared and invariant mass ( ) are correlated –-> possibility to improve resolution. with R = 13 GeV –Additional cuts: 0.4 GeV 2 < M X 2 < 1.4 GeV 2 ; 0.105 GeV < m  < 0.165 GeV. k’ = pHRS +/- 4.5% ; r-function > 0.005 Each photon is in a calibrated block, and E  > 1.0 GeV -6.0 cm < z vtx < 7.5 cm

13. December, 16 200813 Quality calibration control Once iteration completed : –Plot reconstructed missing mass squared value for each block over all blocks –Fit by a constant –Check for deviation from reference value, and dispersion (RMS).

14. December, 16 200814 Results (real) Quick convergence from iteration 0 to 5-6 After, averaged M X 2 oscillates, RMS is stable. => Converged

15. December, 16 200815 Results (real) Ex: Iteration #11 Dispersion is low, Calibration has converged Pion mass deviation less than 1 MeV

16. December, 16 200816 Outline 1- Motivations of calibration –Experimental layout 2- Real data calibration –Method –Results 3- Simulated data calibration –Method –Results 4- Conclusion

17. December, 16 200817 Method of calibration (simulated) Same principle, except we try to match simulated with data: –Searching for a calibration coefficient: –Searching to bring each block simulated resolution (  of gaussian fit) to real block resolution. => also need for each block a « smearing » coefficient

18. December, 16 200818 Method of calibration (simulated) For an event i, a photon is in block #a, the other is in block #b. Correcting each photon energy: Smearing each photon energy: picking a random variable in a gaussian distribution.  : corrected energy,  : Recomputing M X 2 with these corrected energies, and filling #a and #b blocks distributions.

19. December, 16 200819 Quality calibration control Once iteration completed : –Plot M X 2 | simu - M X 2 | real for each block over all blocks and  (M X 2 | simu ) -  (M X 2 | real ) for each block over all blocks –Fit each plot by a constant

20. December, 16 200820 Results (simulated) Averaged  M X 2 converges until iteration 15, then stable RMS does not stop decreasing: <0.001 since iteration #30

21. December, 16 200821 Results (simulated) Since ~iteration #20, seems to be converged. Check out for resolution to conclude

22. December, 16 200822 Results (simulated) Quick convergence averaged  (M X 2 ) of from iteration 0 to 6-7, stable after. averaged  (M X 2 ) RMS stable since Iteration #15

23. December, 16 200823 Results (simulated) Invariant mass deviation due to calibration is still less than 1 MeV Since ~iteration #20, dispersion is low, calibration has converged. Only advantage to go further is to improve  M X 2 dispersion.

24. December, 16 200824 Outline 1- Motivations of calibration –Experimental layout –Consequences on measurements 2- Real data calibration –Method –Results 3- Simulated data calibration –Method –Results 4- Conclusion

25. December, 16 200825 Conclusion Our goal was to calibrate the calorimeter with our set of Kin3  0 data and our set of Kin3  0 simulated to be able to make consistent fit for analysis. We have reached regime of stability for both real calibration and simulated calibration iteration processes => Calibration has been succesfully completed.