Advanced Reconstruction in Large Volume Liquid Scintillator Detectors Applied to LENA Björn Wonsak Universität Hamburg 09/06/15 1
Overview ● Tracking at high energies (GeV) ● Basic algorithm ● Performance ● Application to low energies (MeV) ● New techniques to improve robustness ● Positron discrimination 09/06/15 2
Motivation: Tracking at High Energies ν e appearance experiments: NC-background → Is it possible to identify the π 0 ? Reactor experiments Short-lived cosmogenics ( 9 Li/ 8 He) dangerous background Full veto produces too much deadtime → Identify places of high energy deposition (showers induced by muon) 09/06/15 3
Why no 3D Tracking (so far)? Point-like event: Light emitted in 4 π → no directional information Time between emission and detection = distance → Circles Point of light emission 09/06/15 4
Why no 3D Tracking (so far)? Point-like event: Light emitted in 4 π → no directional information Time between emission and detection = distance → Circles Point of light emission 09/06/15 5
Why no 3D Tracking (so far)? Point-like event: Light emitted in 4 π → no directional information Time between emission and detection = distance → Circles Point of light emission 09/06/15 6
Why no 3D Tracking (so far)? Track: Lots of emission points with different emissions times → No association between signal and emission time 09/06/15 7
My Basic Idea Assumption: ● One known reference-point (in space & time) ● Almost straight tracks ● Particle has speed of light Concept: ● Take this point as reference for all signal times 09/06/15 8
The Drop-like Shape Signal time = particle tof + photon tof → ct = |VX X| + n*|X XP| light light emission emission track X X V ertex path of (reference point light on track) P MT 09/06/15 9
The Drop-like Shape ct = |VX| + n*|XP| → drop-like form P V X X 09/06/15 10
The Drop-like Shape ct = |VX| + n*|XP| → drop-like form P Possible Possible origin of origin of light light V X X 09/06/15 11
Time Distribution Used as a smearing on the drop-like shape time in ns Convolution of Gaus and Exponential-Function 09/06/15 12
Time Distribution LS typically has more than one decay component, the smallest and dominate one in our case is 4.6 ns. time in ns Convolution of Gaus and Exponential-Function 09/06/15 13
Result 1 PMT The co colour- r-co code represe sents the The colour-code represents the pro robability y of light emi missi ssion fro rom m a probability of light emission from a given point given point y [cm] red = = high pro robability y red = high probability blue = = low pro robability blue = low probability x [cm] 09/06/15 14
Result a Few PMTs y [cm] x [cm] 09/06/15 15
Result 266 PMTs y [cm] x [cm] 09/06/15 16
Light Distribution (LD) Effects Some parts of each drop-like shape are more likely the origin of light, because: – they are closer – directly in front of the PMT → Need to consider: – solid angle of PMT area – attenuation – angular acceptance 09/06/15 17
Light Distribution (LD) Effects Some parts of each drop-like shape are more likely the origin of light, because: – they are closer – directly in front of the PMT → Need to consider: – solid angle of PMT area – attenuation – angular acceptance Finally I have to normalise the resulting pdf ! 09/06/15 18
Result all PMTs y [cm] 3 GeV muon Initial direction (1,-1,0) x [cm] 09/06/15 19
Probability Mask So far probabilities have been added! → correct for independent information However: Light signals are not completely independent from each other, because they belong to the same track. → Use “Result I” to weight all the single light contribution and re-normalise each of them! 09/06/15 20
Result I y [cm] 3 GeV muon Initial direction (1,-1,0) x [cm] 09/06/15 21
Result 2nd Iteration y [cm] x [cm] 09/06/15 22
Result 3rd Iteration y [cm] x [cm] 09/06/15 23
Result 9th Iteration y [cm] x [cm] 09/06/15 24
3D Topology Probability distribution projected into the xy plane Color: Total photon emission probability in arbitrary units → dE/dx seems accessible 09/06/15 25
Image Processing 3D-Presentation Binarisation Blob finding 3D Medial line Medial line Medial line XY-Projection XZ-Projection Ph. D. student Sebastian Lorenz Resolution < 20 cm 09/06/15 26
Computing One 3 GeV event, 20 cm bins, full light, 22 iterations in LENA → several hours (despite usage of adaptive mesh refinement) However: ● I'd like to go to 2 cm bins - because there should be enough light for this resolution ● In principle many more iterations are allowed But algorithm highly parallisable → GPUs, etc. 09/06/15 27
Current Status Large reconstruction campaign ongoing! Muons with 1-5 GeV: (first results) y r a ● Robustness → okay n i m ● Angular resolution: ~1.5° i l e r P Electron events under production Other event classes still to be studied Paper under preparation! Ph. D. student Sebastian Lorenz 09/06/15 28
Can also do it with Cherenkov Light 3 GeV muon, initial direction (1,-1,0) Preliminary Preliminary Bachelor student David Meyhöfer A few % of light in liquid Scintillator is Cherenkov light → using both could help pattern and partical identification Also suitable for water Cherenkov detectors! Perfect for WbLS! 09/06/15 29
Tracking at Low Energies (a few MeV) 09/06/15 30
Robust Iterations!? y x y x y x 09/06/15 31
New Procedure ● Divide detector in different parts ● Do reconstruction for each part ● Multiply results ● Use this as Probability Mask ● Go back to first step 09/06/15 32
Result 2nd Iteration z-projection y-projection y z x x 1MeV positron at center 09/06/15 33
Result 2nd Iteration (Zoom) Z-projection Y-projection (top view) (side view) y z x x 1MeV positron at center 09/06/15 34
Result 2nd Iteration Slice 241 XY-slice of 3d probability density distribution Y in cm X in cm 09/06/15 35
Result 2nd Iteration Slice 240 XY-slice of 3d probability density distribution Y in cm X in cm 09/06/15 36
Result 2nd Iteration Slice 239 XY-slice of 3d probability density distribution Y in cm X in cm 09/06/15 37
Result 2nd Iteration Slice 238 XY-slice of 3d probability density distribution Y in cm X in cm 09/06/15 38
Result 2nd Iteration Slice 237 XY-slice of 3d probability density distribution Y in cm X in cm 09/06/15 39
Result 2nd Iteration Slice 236 XY-slice of 3d probability density distribution Y in cm X in cm 09/06/15 40
Crystalisation of the Result ● Use well defined probability mask ● Do reconstruction for each photon ● Identify bin with highest probability ● Associate photon with this bin → number of photons from that bin 09/06/15 41
Crystalisation: 1 MeV Positron Color: Number of photons detected from that bin Y in cm X in cm 09/06/15 42
Crystalisation: 2 MeV Electron Color: Number of photons detected from that bin Y in cm X in cm 09/06/15 43
Crystalisation: 2 MeV Electron Color: Number of photons detected from that bin Y in cm X in cm 09/06/15 44
Electron vs. Positron Discrimination: First Try Results I Ratio R of light reconstructed near vertex vs. total light Electrons Positrons (At energies relevant for the IBD of reactor neutrinos.) R in % ● 3343 events of electron and positron events each ● Visible energy 1 – 5.5 MeV Notice: Used perfect vertex ● At the center of the detector → worst place position for this analysis ● LENA-MC → 250 photons per MeV 09/06/15 45
Electron vs. Positron Discrimination: At C-11 Energy Region Ratio R of light reconstructed near vertex vs. total light Electrons Positrons R in % ● 111 events of electron and positron events each ● Visible energy 1 – 2 MeV Notice: Used perfect vertex ● At the center of the detector → worst place position for this analysis ● LENA-MC → 250 photons per MeV 09/06/15 46
Remarks on Potential ● Possible improvements: ● So far only 250 p.e/ MeV → Borexino: 500 p.e/ MeV, JUNO: 1200 p.e/ MeV ● Faster scintillator ● Remove scattered light statistically ● Multivariate analysis ● Other ideas: ● Use time as 4th dimension ● Gradient information (Sobel-Filter) 09/06/15 47
Remarks on Potential ● Possible improvements: ● So far only 250 p.e/ MeV → Borexino: 500 p.e/ MeV, JUNO: 1200 p.e/ MeV ● Faster scintillator ● Remove scattered light statistically ● Multivariate analysis ● Other ideas: ● Use time as 4th dimension ● Gradient information (Sobel-Filter) 09/06/15 48
Eliminating Influence of Scattered Light ● Idea: Use probability mask and lookup tables to calculate for each signal the probability to be scattered → reweigh signals after each iteration 09/06/15 49
Eliminating Influence of Scattered Light ● Idea: Use probability mask and lookup tables to calculate for each signal the probability to be scattered → reweigh signals after each iteration Multiplication trick not used here y in cm x in cm Result before removal of scattered light! 09/06/15 50
Recommend
More recommend
