kno detector simulation
TRANSCRIPT
KNO Detector Simulation
SEO Ji-Woong
Sungkyunkwan Univ.
(On behalf of KNO software working group)
KNO 6th Workshop
2021.08.20
β’ Status
β’ KNO detector
β’ Neutrino event generator
β’ Detector simulation : WCSimβ’ Event display
β’ High Energy studyβ’ Vertex resolutionβ’ Angular resolutionβ’ Energy resolutionβ’ PID
β’ Low Energy studyβ’ Observable energy thresholdβ’ Energy resolution
β’ Plan
β’ Summary
Contents
2021-08-20 KNO 6th workshop 2
Status
β’ Software working groupβ’ Meeting on every Thursday 17:00
β’ Studied neutrino event generator
β’ Understanding water Cherenkov detector using MC
β’ Developing event reconstruction tools
β’ Manpowerβ’ 4 regular contributors
β’ Computing resourceβ’ SKKU : 8-cores CPU with Quadro P7000
β’ Sejong : 120-cores CPU with 2 RTX3090
2021-08-20 KNO 6th workshop 3
KNO detector
β’ Detector designβ’ Cylindrical shape
β’ Height : 54.8 m
β’ Diameter : 100.0 m
β’ Total water mass : 503 ktβ’ Active water mass : 384 kt
β’ 63,000 20-inch PMTs (40% coverage)
β’ ~1.5 deg. off axis angle (mt. Bisul)
β’ ~1,000 m overburden
2021-08-20 KNO 6th workshop 4
β’ GEANT4 doesnβt simulate neutrino interaction
β’ Two publicly available neutrino event generators were examinedβ’ GENIE
β’ NuWro
β’ Both neutrino event generators provide various neutrino interactionsβ’ Can input neutrino beam energy profile
β’ Provide various output file formats (text, nuance, etc.)
Neutrino event generator
2021-08-20 KNO 6th workshop 5
β’ Publicly available GEANT4 based water Cherenkov detector simulation program
β’ Relies on GEANT4 and ROOT
β’ SK & HK geometry and basic PMT information are available
β’ Provides simulated PMT hit informationβ’ Hit time, charge, hit positionβ¦
β’ Trigger, dark rateβ¦
β’ Use output file of an event generator (nuance format) as input
Detector simulation : WCSim
2021-08-20 KNO 6th workshop 6
β’ GUI event display is prepared
β’ Can draw event display easily
β’ Use WCSim output file
β’ Implemented simple functions
Event display
2021-08-20 KNO 6th workshop 7
β’ KNO fitβ’ Step by step fitter using maximum goodness method
β’ Mainly use hit time informationβ’ Time residual = PMT hit time β Time of flight from vertex
High Energy Software
2021-08-20 KNO 6th workshop 8
Pre-fitterLepton
directionPrecise
fitterLepton energy
Obtain vertex (x,y,z,t)
fastly
Use pre-fitterβs
information, find
lepton direction &
ring edge
Use lepton direction
and ring edge,
re-fit vertex (x,y,z,t)
Use vertex, lepton
direction and conversion
function, calculate lepton
energy
β’ Vertex resolution
High Energy Software
2021-08-20 KNO 6th workshop 9
β’ KNOfit tries to fit vertex using ring information
β’ Shows improvements in Ξπ between pre-fitter and precise-fitter
β’ Current best Ξπ β’ Peak ~45 cm
β’ 65% ~90 cm
Current best
65%
β’ Angular resolutionβ’ Ξπ of electron is slightly larger than Ξπ of muon
High Energy Software
2021-08-20 KNO 6th workshop 10
π πβ
65% 65%
β’ Energy resolutionβ’ Obtain lepton energy using conversion function
β’ Current best resolution : 1 GeV Muon ~5.7 %, 1 GeV Electron ~8.5 %
High Energy Software
2021-08-20 KNO 6th workshop 11
1 GeV π 1 GeV πβ
β’ Particle Identification (PID)β’ Muon and electron can be identified using ring pattern
β’ A muon typically produces a sharp-edged ring
β’ And an electron will produce a much fuzzier ring due to EM shower
β’ ~99% accuracy
High Energy Software
2021-08-20 KNO 6th workshop 12
π πβ
β’ Machine Learning PIDβ’ Two leptons produce
different ring patterns
β’ Identify lepton type using machine learning
β’ python 3.7 with pytorch
β’ Train and test using image of event display only
High Energy Software
2021-08-20 KNO 6th workshop 13
β’ Improved event reconstruction tool
β’ KNOfit is a βstep by stepβ goodness fitterβ’ vertex β lepton direction β momentum
β’ Developing new likelihood fitter based on the experience accumulated while developing KNOfit
β’ Simultaneous fitter - vertex (x,y,z,t), lepton direction (ΞΈ,Π€) and momentum (p)
β’ Expect better performance than KNOfit
High Energy Software
2021-08-20 KNO 6th workshop 14
Low Energy Software
2021-08-20 KNO 6th workshop 15
40% 20% 10%
Photo coverage 40 % 20 % 10 %
trigger pass 99 % 11 MeV 17 MeV 21 MeV
trigger pass 65 % 7 MeV 9 MeV 11 MeV
β’ KNO low energy thresholdβ’ KNOfit (also likelihood fitter) can not be used for low E (~O(10) MeV) events due to
small number of PMT hits
β’ Need to check observable energy thresholds for KNO geometry
β’ Expected energy resolution using low energy fitter
Low Energy Software
2021-08-20 KNO 6th workshop 16
β’ Develop likelihood fitterβ’ Developing as a one ring fitter : ~50% progressβ’ PID will not be included
β’ Develop low energy fitterβ’ Developing event vertex and lepton direction fittingβ’ Packaging for distribution
β’ Study of ML PIDβ’ Developing ML PID tool using various information of WCSimβ’ To be integrated into likelihood fitter
Plan
2021-08-20 KNO 6th workshop 17
β’ High energy event reconstruction
Super-K reconstruction performance
2021-08-20 KNO 6th workshop 18
Reference : An Analysis of the Oscillation of Atmospheric Neutrinos, doctoral thesis, Shimpei Tobayama, 2010
β’ KNO detector simulationβ’ Use WCSim simulation program based on GEANT4
β’ Event reconstructionβ’ KNOfit (step by step maximum goodness fitter) is available
β’ without PID process
β’ energy resolution : ~5.7% (muon) and ~8.5% (electron)
β’ Low energy threshold of KNO is about 7~11 MeV (current design)
β’ Advanced high energy fitter is being developed
β’ Work on machine learning PID is in progress
Summary
2021-08-20 KNO 6th workshop 19
backup
2021-08-20 KNO 6th workshop 20
nuWro Genie-MC
exported βnuanceβ file
from genie-mc : exist β-2β tag particles
β-2 = intermediate stateβ
from nuWro : exist βinfoβ line2021-08-20 KNO 6th workshop 21
β’ final muon (tag β0β) energy distribution
(using βnuanceβ file)
β’ both generator shows 9.x GeV peak
2021-08-20 KNO 6th workshop 22
nuWro Genie-MC
π
π
π
π
β’ Final particles (tag β0β) (using βnuanceβ file)
β’ Similar outputs
2021-08-20 KNO 6th workshop 23
nuWro Genie-MC
π
π0 π+
π
πΎπ0 π+
β’ genie-mc shows βgammaβ
2021-08-20 KNO 6th workshop 24
Cross-section
25
Carbon scattered cross-section
Looks very similar
2021-08-20 KNO 6th workshop
26
QE + RES + DIS QE + RES + DIS + MEC
Charged Current cross section
2021-08-20 KNO 6th workshop
27
QE + RES + DIS QE + RES + DIS + MEC
Neutral Current cross section
2021-08-20 KNO 6th workshop
Vertex Reconstruction using simple method
β’ Conceptβ’ Assume all photons from one point
β’ Set a virtual vertex in the Detectorβ’ can calculate distance from hit PMT
β’ obtain βTime of Flight (in the water)β with all hit PMTs
β’ calculate best vertex (x,y,z) combination using all of TOF information in vertex grid in detector
β’ In Super-K, already use similar concept vertex simple fitter
28
Real Vertex
(unknown)
Virtual Vertex
for calculation
Hit PMT
distance
2021-08-20 KNO 6th workshop
Vertex Reconstruction using simple method
β’ ππππ π ππ πππ’ππ π‘π = π‘π0 β
ππβπ
πβ’ π‘π
0 : digitized hit time of ith hit PMT
β’ ππ : position of ith hit PMT
β’ π : position of virtual vertex
β’ π : light speed in water (refractive index = 1.33)
β’ πΊππππππ π πΊ = Οπ exp(βπ‘πβπ‘0
2
2 1.5Γπ 2)
β’ π‘0 : event time (free parameter)
β’ π : typical time resolution (using 2.5 ns from SK)
29
Function of Vertex (x,y,z) and π‘0Search minimum ββGoodnessβ combination
2021-08-20 KNO 6th workshop
Concept
30
Search for Proton Decay Using an Improved Event Reconstruction Algorithm in Super-Kamiokande, Yusuke
Suda , PhD Thesis, University of Tokyo , Sep. 2017
1. Calculate particle direction using vertex
2. Calculate angle between particle direction and hit
PMT
3. Draw angle vs. p.e. distribution ( left (i) )
4. Find an angle contained maximum p.e. from 2nd
derivative of the distribution
5. It is the edge of Cherenkov Ring
2021-08-20 KNO 6th workshop
31
β’ Particle direction
ππ =
π
ππππ β π
ππ β π
β’ ππ βΆ πβππππ ππ π β π‘β πππ
β’ ππ βΆ π β π‘β πππ πππ ππ‘πππ
β’ π βΆ πππππ’πππ‘ππ π£πππ‘ππ₯
β’ Edge finder
π πππππ =ππππππ ππΈ π ππ
sin πππππα€
πππΈ(π)
πππππππ
2
ππ₯π β(πππππ β πππ₯π)
2ππ2
β’ πππ₯π βΆ ππ₯ππππ‘π πΆβπππππππ£ πππππ ( ~ 42Β°)β’ ππ βΆ ππππ’πππ πππ πππ’π‘πππ (~3Β° ? )
2021-08-20 KNO 6th workshop
Re-fit vertex using Ring information
β’ Conceptβ’ Assume all photons from one point (pre-fitter)
β add track, In & Out ring information
β’ From pre-fitted vertexβ’ Get the brightest ring and particle direction using
pre-fitted vertex
β’ Add track information, more precise vertex fitting
β’ Outside of ring, PMT hits are considered by scattered light
32
Pre-fitted
Vertex
Cherenkov angle
Hit PMT
track
ππ ππ
2021-08-20 KNO 6th workshop
β’ Super-K TDC fit formula
β’ Goodness β βout ringβ
33
π π = fraction of total charge which inside the Cherenkov ring
ππ πππ‘π‘ = 60 ns
2021-08-20 KNO 6th workshop
Neutrino energy reconstruction
β’ μ¬μ§ found more precise formula in T2K doctoral thesis
β’ Calculate neutrino energy using same sample
Reference : Otani, Masashi. "Measurement of Neutrino Oscillation in the T2K Experiment." (2012).
34
μ±νβs formula
2021-08-20 KNO 6th workshop
35
True Energy
Reco Energy
True Energy
Reco Energy
μ±νβs formula result T2K formula result
T2K formula shows better result
2021-08-20 KNO 6th workshop
36
μ±νβs formula result T2K formula result
Decrease difference
T2K formula shows better
reconstruction
performance
energy vs difference has
no correlation
2021-08-20 KNO 6th workshop
37
True vs Reco distribution shows similar distribution between
μ±νβs and T2K
Linear correlation
Not bad result
True E : True neutrino E
Reco E : calculated neutrino E using true
lepton E & direction
Linear correlation
Add μ±νβs energy resolution 4.5%
(gaussian smearing)
Not bad result, too
2021-08-20 KNO 6th workshop
Likelihood fitter study
β’ SK&HK event reconstruction algorithmβ’ fiTQun : maximum likelihood with a particle event hypothesis x
β’ Using observed information, construct the likelihood function
38
Main challenge is calculating these βmuβ
2021-08-20 KNO 6th workshop
39
Likelihood fitter study
2021-08-20 KNO 6th workshop
2021-08-20 KNO 6th workshop 40
2021-08-20 KNO 6th workshop 41