Special Issues on Neutrino Telescopy
Apostolos G. Tsirigotis
Hellenic Open UniversitySchool of Science & TechnologyParticle and Astroparticle Physics Research Group
Monte Carlo Development : Event Generation
•Neutrino Interaction Events
•Atmospheric Muon Generation
μ
ν
•Atmospheric Neutrinos
ν
•Cosmic Neutrinos
Earth
Monte Carlo Development : Event Generation
Neutrino Interaction (use of Pythia)
Neutrino Interaction Probability Survival probability
Effective Neutrino Flux = Neutrino Flux × Survival Probability × Neutrino Interaction Probability
Nadir Angle P
rob
abil
ity
of
a ν μ
to
cro
ss E
arth
Monte Carlo Development : Event Generation : Example
Muon neutrinos from AGN Jets
1 Diffuse Neutrino Flux (Mannheim 1995)
1
2
3
45
2 Effective Neutrino Flux (horizontal)
3 Effective Neutrino Flux (10 degrees below horizon)
4 Effective Neutrino Flux (20 degrees below horizon)
5 Effective Neutrino Flux (30 degrees below horizon)
GeV
-1cm
-2s-1
sr-1
Neutrino energy (GeV)
Monte Carlo Development : Detector Description (Geant4)
• Any detector geometry can be described in a very effective way
Simulation strategy: During the Detector description in the simulation, the whole group of PMTs are divided in nested subgroups:
•All the relevant physics processes are included in the simulation
All PMTs group
subgroup 1 subgroup 2
subgroup 3 subgroup 6subgroup 5subgroup 4
. . . . . . . . . . . . . . . . . . . .
•For each subgroup is defined a sphere that contains all the PMTs of this subgroup
•These spheres are used to speed up the simulation as it will be described later
•Use of Clustering Algorithm that ensures the minimum dispersion of the PMTs of each group
Monte Carlo Development : Detector Description : PMT Clustering Algorithm
find the center of mass m1 of group1
find the center of mass m2 of group2
Define 2 points inside the detector, p1 & p2
For all PMTs
Which point is closer ?
p1 p2
Add PMT to group1 Add PMT to group2
is m1=p1 and m2=p2
yes
no
converge
Monte Carlo Development : Detector Description : Working Example
Detector Geometry (1km3 Grid)
21x21 Strings21 Storey per String2 PMTs per Storey (Looking up and Down)
OM Geometry (15inch)
Charge Current Atmospheric νe (20GeV) interaction
18522 PMTs
50m between PMTs
Monte Carlo Development : Simulation Technique Cherenkov photon emission
Cherenkov photons are emitted only if the are going to hit a PMT
Use the nested groups in order to minimize computer time :
The photons cross the sphere containing a detector subgroup?
NODo not produce photons
YES
For each of the 2 subgroups of the previous group
YES
The photons cross the sphere containing a detector subgroup?
NODo not produce photons
…………….
The photons cross the sphere containing the whole detector?
YES
For each of the 2 subgroups of the previous group
NODo not produce photons
Monte Carlo Development : Fast Simulation
Angular Distribution of Cherenkov Photons
EM Shower Parameterization
Parameterization of EM Shower
•Longitudal profile of shower
•Number of Cherenkov Photons Emitted (~shower energy)
•Angular profile of emitted photons
Signal Simulation
PMT response to optical photons
Collective Efficiency Collective Efficiency
Single Photoelectron Spectrum
mV
Πρότυπος παλμός
Quantum Efficiency
Standard pulse
Monte Carlo Event Production
•Computer Farm with 15 computers (15 double xeons )
•We are currently installing 64 more computers (64 double opterons)
350 Gflops
Reconstruction Algorithms
•Clustering of candidate tracks
•Kalman Filter (novel application in this area)
Angular deviation (degrees)
Angular deviation (degrees)
1TeV muons
1TeV muons
Simulation Example
1 TeV Vertically incident muon
K40 Noise Hits
Signal Hits
(Hit amplitudes > 2p.e.s)
Fast Triggering Algorithms
Estimation of Information Rate
1km3 Grid (18522 15inch PMTs)
Information Rate = PMT Number * K40 Noise Rate * (Bytes/Hit)
= 18522 * 50kHz * 32
≈ 30GB/sec
Cannot be saved directly to any media
Charge & Multiplicity Characteristics
Charge/hit distribution
Number of pes
noise
signal
Multiplicity (signal)
Multiplicity (noise)
Number of active PMTs in 6 μs window
Number of active PMTs in 6 μs window
No Cut
1TeV Vertical Muons
Charge & Multiplicity Characteristics
Selection based on hits with at least 2 photoelectrons
Multiplicity (signal) Multiplicity (noise)
Information Rate = PMT Number * K40 Noise Rate * (Bytes/Hit)
= 18522 * 3kHz * 32
≈ 1.8GB/sec
By Using clustering like DUMAND the background rate is reduced by 75% (450 MByte/sec) and the signal hit has a higher than 80% probability to survive
Fast Triggering Algorithms
Estimation of Information Rate
1km3 Grid (18522 triplets of PMTs)
3 PMTs per hemisphere in coincidence
10nsec time window, 2 out of 3 coincidence
Each triplet’s total photocathode = 15inch PMT photocathode
Information Rate = PMT Number * K40 Noise Rate * (Bytes/Hit)
= 3* 18522 * 17 Hz * 32
≈ 30MB/sec
Triplet coincidence rate=17Hz (17kHz background per PMT)
Number of active triples
Background
Signal
1TeV Vertical Muons
Fast Triggering Algorithms
Use of the number of active triplets as fast selection trigger
Distributions normalized to 1
Fast Triggering Algorithms
Estimation of Event Rate and Efficiency
Eve
nt R
ate
(kH
z)
Cut to the number of active triplets
Eff
icie
ncy
Cut to the number of active triplets
180 kByte/event
10TeV
1TeV
Fast Triggering Algorithms
1TeV Vertical Muons
Use also the Dumand clustering:
Background
Signal
Number of active triples
Fast Triggering Algorithms
Estimation of Event Rate and Efficiency
Eve
nt R
ate
(kH
z)180 kByte/event
Cut to the number of active triplets
Eff
icie
ncy
Cut to the number of active triplets
1TeV
Fast Triggering Algorithms
Raw Hits
Absolute time
TimeStretching
Trigger Level
trigger
Accepted Interval
Triggering Method
36 PMs in 3 subcylinder
35 3” photomultipliers in a cylinder
Determination of photon direction, e.g. via multi-anodic PMs plus a matrix of Winston cones.
Large photocathode area with arrays of small PMTs packed into pressure housings
Alternative Options for photodetection