1. HARPO Analysis

2. New reconstruction algorithm: no tracking Simulation
Quick reminder:
HARPO data
“Old” reconstruction
Preliminary results
New reconstruction algorithm: no tracking
Simulation
Resolution and polarimetry

3. HARPO reminders

4. NewSUBARU Beam

5. HARPO events

6. Main issue: Electronics saturation
Understood phenomenon
saturation on cumulated charge on single channel
response function not calibrated yet
Additional crosstalk
poorly understood
Affects a very large fraction of the events
energy dependent
flagging and removing is not an option
Important due to beam geometry
should not affect (much) cosmic photons

7. Example of saturation 74

8. Reconstruction v2.0 Clustering “classic” T and C clusters (Shaobo)
Tracking “Hough”
good for straight separated tracks (cosmics)
problems with shared vertex and scattering
Tracking “Kalman”
closest neighbour search in T and C direction
straight line fit
Track matching XZ ↔ YZ
independent of tracking method
problem with electronics saturation

9. Polarisation Pairs of matched tracks get POCA & DOCA
cut on distance between POCA and first signal in TPC
Combine data from runs with different TPC orientation
Ratio (P=100%)/(P=0)

10. Polarisation

11. Polarisation

12. Polarisation

13. Limitations Low efficiency ~10%=Nreconstructed/Non beam
Low angular resolution
straight line fit with large multiple scattering
benefit of Kalman lost around vertex
Difficult to improve
complex algorithm
very sensitive to parameter changes

14. Reconstruction v3.0

15. Reconstruction v3.0 Focus on finding vertexes No tracking
Simple bloc clustering
Find ROI (vertex candidates): “corner shape”
Refine vertex position
Find peaks in polar distribution around vertex
Vertex matching (same or switch)

16. Vertex Finding Reduce data: simple clustering Find “corner” shape
Draw circle
Find longest arc with no data
Select arcs longer than π
Refine position
CoG of RoI (circle)

17. Vertex Finder
φ>π
φ<π

18. Vertex Finding Simple Robust
local characteristics: works on complex events
Weaknesses
many fake vertexes (e.g. track ends)
needs continuous signal, low noise
Potential for improvement
better position

19. Vertex Fitting Polar charge distribution around vertex delta electron

20. Vertex Fitting
Clean up: keep only straight lines

21. Vertex Fitting
Peak finding: track directions

22. Vertex Fitting Simple Robust:
ignores obvious scattering and background
potential for small opening angle
Potential for improvement
better peak finding
use of distance info (focus on short distance for large opening angle, long distance for small)

23. Vertex Matching As before: compare charge profile
1: match vertexes if there are several with same Z position
2: match the tracks in the vertex (simple: only 2 possibilities)

24. Vertex Matching
Assign signal to tracks

25. Vertex Matching
Compare profiles: X(1,2) ↔ Y(1,2) “same”

26. Vertex Matching
Compare profiles: X(1,2) ↔ Y(2,1) “switch”

27. Vertex Matching
Same
Switch

28. Vertex Matching Simple (again) Robust (again)
small number of combinations
works if 1 of the 2 tracks has a problem (e.g. electronics saturation)
Potential for improvement ?

29. Example with saturation

30. Simulation

31. Simulation very basic Geant4 cube of gas, list of charged particles
complete TPC simulation
drift, electron capture, diffusion, gain fluctuations
HARPO geometry
pad/strips geometry, shaping, noise, saturation
Good description of the data
still needs to be properly calibrated

32. Results

33. Polarisation: v2.0

34. Polarisation: v3.0

35. Polarisation: Simulation

36. Polarisation: summary
Greatly improved efficiency
→ Higher statistics
Good agreement with simulation
Asymmetry smaller than expected
resolution?
matching?

37. Angular resolution

38. Angular resolution Agreement with theoretical prediction
relatively small contribution of tracking
Excellent agreement with simulation
effect of saturation dominates at high energy
Potential for improvement
estimation of track momentum
event 100% resolution should significantly improve

39. Conclusions
An new, simpler, reconstruction has boosted the performance of HARPO
+ more robust, higher efficiency
– relies on low noise, many fakes
The data is well described by simulation
better calibration needed for more direct comparisons
Performance is good without optimisation
It can probably still be improved

40. Future...

41. Future...
ST3G

42. Future...

43. Backup

44. Efficiency

45. Polarisation asymmetry