KLG4sim: A Full Monte Carlo Simulation for KamLAND

KLG4sim: A Full Monte Carlo
Simulation for KamLAND
MAND Workshop
June 2005
Lauren Hsu
Lawrence-Berkeley National Laboratory
Outline
I.
KamLAND Overview
II.
KLG4sim Overview
III.
KLG4sim Performance
i. Basic Checks
ii. Reconstructed Quantity Checks
KamLAND
KAMioka Liquid scintillator
Anti-Neutrino Detector
Detecting reactor anti-neutrinos 1 km
beneath Mt. Ikeyama
Inside the
Kamioka Mine
Surrounded by 53 Japanese Nuclear Reactors
A Reactor Anti-Neutrino Experiment
Disappearance measurement of - e
Sensitive to mixing between the first and second neutrino mass eigenstates
PRL 90 (2003) 021802
The KamLAND Detector
(1879)
KamLAND: Simultaneously tank of target protons (>1031) and a
sensitive calorimeter.
Anti-Neutrino Signal Detection
Coincident energy deposits are a distinct signature of inverse beta-decay:
e
+ p → e+ + n
Prompt Energy:
positron energy deposit
(K.E. + annihilation γ’s)
Delayed Energy:
n-capture releases 2.2
MeV γ, ~200 µs later
Selecting Candidate Events
Apply Time and Spacial Cuts to Obtain Candidate Coincidence Events
Candidate Coincidence
Events: ∆t = [0.5, 1000]µs
-e
energy obtained from
E = Eprompt + 0.8 MeV
Basic KamLAND Data
Reconstruction
Energy and Vertex Information are Crucial
Energy Reconstruction:
• Energy ∝ Number of Hit PMT’s
• Correction for Vertex Position
• Corrections for Quenching and
Cerenkov
Vertex Reconstruction
• Determined by Very Precise Timing of Hits (~ few ns):
• Inherent Detector Resolution ~15cm.
• Based on push-pull minimization algorithm.
Energy and Vertex fitter Calibrated w/ Co60, Ge68, Zn65, and AmBe deployed
along the z-axis.
Backgrounds
Correlated background from
13C(α,n)16O is troublesome
- α’s produced indirectly through
decay of 210Pb, a long-lived daughter
of 222Rn
13C(α,n)16O
Energy
Uncorrelated
backgrounds
from trace
radioactive
isotopes.
Fiducial Volume: R < 5.5m
Correlated backgrounds from
muon spallation:
whole detector
veto (2 s)
cylinder cut veto
(2 s)
Obtaining the Unoscillated Flux
Data obtained by special agreement from Japanese nuclear power plants
plus well understood theoretical calculations predicts number of anti-neutrinos
in the case of no-oscillations.
Unoscillated Flux can be determined to a few percent!
Observation of Spectral Distortion
Measurement of Energy Spectrum Distortion Due to -e Oscillation
(Latest KamLAND Result)
PRL 94 081802 (2005)
Unparalled Sensitivity to ∆m12
2
Extract Oscillation Parameters and Combine with Solar Data
PRL 94 081802 (2005)
PRL 94 081802 (2005)
Solar + KamLAND:
2
∆m12
+0.6
=7.9 -0.5 ×10-5
eV2,
tan2θ12
=0.4
+0.10
-0.07
Future Improvements: Reactor
Analysis
Further Improvements Require Reducing Systematic Uncertainty!
Systematic Unc. on Rate
%
Fiducial Volume
4.7
un-oscillated
2.5
e
spectrum (theor.)
Energy Threshold
2.3
Reactor Power
2.1
Cut Efficiency
1.6
Fuel Composition
1.0
Cross Section
0.2
Livetime
0.06
Total
7.1
Compare to statistical uncertainty: 6.7%
Better understanding of 13C(α,n)16O will also improve shape analysis
Future Improvements:
Reactor Analysis
A Full-Volume Calibration Device
• Improve Uncertainty on
Fiducial Volume
• Better Understanding of
Energy and Vertex Reconstruction
A Muon Tracker
Gold-Plated Muon Events
Cross-Check Muon Track Reconstruction
~200 events per day, x-y resolution of ~3cm
Future KamLAND: 7Be Phase
Detect e from 7Be via elastic
scattering (not a coincidence signal)
After Purification
Will require good understanding
of backgrounds!
Where Does KamLAND MC Fit Into
This?
Visualization of KamLAND
w/ KLG4sim
• Current and Past Uses: Diagnose
problems in energy estimation and
vertex reconstruction, simulate
backgrounds for 7Be solar phase
• Together w/ 4π calibration data,
use to extrapolate 4π calibration
points and cross-check fiducial
volume estimates.
• Improving our understanding of
how the detector works
and how backgrounds behave
The uses of a well-tuned, detailed, full-detector simulation are many!
II.
KLG4sim
KamLAND GEANT4 Simulation
What is KLG4sim?
Highly detailed, comprehensive and flexible detector simulation.
Still undergoing rigorous verification and some code development
• Built off GEANT4 toolkit
• Additional customized physics processes.
• Standalone Generators (co60, zn65, ge68, K, Th, U, Rn)
• Precurser to (Generic-Land GEANT4 Simulation) GLG4sim:
moving towards a merge
Other Good Features:
Flexible file output, user interface, multiple visualization tools
Originally written by various members of the KamLAND collaboration,
significant contributions from Glenn Horton-Smith
KLG4sim Overview
Detector
Descriptions
• Material Properties
• Geometry
Geant4
User Supplied
Algorithms
• Generators
• Customized Physics
Processes
KLG4sim
• Object Oriented
Framework and Kernel
Simulated Response of
KamLAND Detector
• Particle Interactions
with Material
• Electronics Readout
• Particle Transport
• Output Formats
• User Interface
User Supplied
Commands Via Macros
or Interactive Sessions
KLG4sim is built from more than just Geant4 routines…
It has a number of customized features
Customized Feature: Torus Stack
Geometry Class
A flexible tool for constructing non-standard geometrical shapes
Spherical top
Outside Torus
Edge
KamLAND
PMT in
KLG4sim
Inside Torus
Edge
Cylindrical base
KamLAND Balloon in KLG4sim
Example: Constructing a PMT Volume
by Stacking Toruses
Customized Feature:
Optical Photon Propagation
KL(GLG4)Scint:
• Emission of scintillation photons (as secondary particles)
• records the energy that is converted to scintillation light
Does photon interact with Scintillator?
KL(GLG4)OpAttenuation:
• Determines if photon is absorbed or scattered.
• If scattered, calculates new direction of propagation.
If Absorbed: KL(GLG4)Scint also takes care of reemission (at
equal or longer wavelength).
Emission and reemission decay times are currently the same.
Probabilities to attenuate, reemit and scatter are specified in materials file.
KLG4sim Modeling Light
Propagation in KamLAND
KLG4sim
Illumination Plot
reflected beam
Reflection Hitting
PMT #1268
scattered
and
reemitted light
380 nm
photon beam
Hitting
center of
PMT #258
Customized Feature:
PMT Sensitive Detector Model
A “G4Parametization” class that handles absorption, reflection
And transmission at the photocathode boundary
Features:
• Realistic photocathode geometry thanks to KLTorus Stack
• “Lux Level”: options detail allowing one to switch between:
- black (no reflection)
- infinitely thin photocathode, reflect, refract or absorb.
- finite photocathode thickness (implemented by Dario from Double Chooz).
• Handles weighted tracks (some breakdown for very high multiplicity events
as in the case of cosmic muons).
Customized Feature: Neutron
Diffusion and Decay
By default, Geant4 treats a thermal neutron from creation to capture as 1 event
KamLAND event window is ~500 ns, but neutron capture time is ~400
longer than this!
Solution:
• KLNeutronCaptureAndDecay
(GLG4NeutronDiffusionAndCapture)
- assigns a diffusion distance and
capture time based on data set in
materials file.
• KL(GLG4)DeferTrackProc – Sets
a finite time window for an “event”
and “delays” any persisting tracks
to the next event.
This doesn’t work well near a boundary…needs more work for θ13 experiments!
Customized Feature: Signal
Readout Simulation
Currently: KLG4sim decides whether a PMT registers a hit or not at the
photocathode and simulation basically ends there.
Last major area still in need of significant coding!
Electronic Signal Readout Simulation
•
•
•
•
A generic package that can be used for GLG4sim is being written.
PMT effects: Transit Time Spreads, dark noise, after-pulsing
Waveforms
Trigger Simulation
Shell is in the process of being written by Chris Jillings of KamLAND,
III. KLG4sim Performance
i. Basic Checks
ii. Reconstructed Quantity Checks
Simulation Verification
Always an ongoing process!
Optical Properties: (reflectivity, index of refraction, absorption,
reemission, scattering, scintillator decay times) are critical for accurate
simulation of energy and vertexing in KamLAND.
Main Areas of Concentration So Far:
• Geometry: Shadowing studies, and material reflectivity
verification studies.
The
Basics
• Optical Photon Propagation: light yields, and speed of light
• PMT model: Getting Help from GLG4sim users!
The
ultimate
test!
• Energy and Vertex Reconstruction
There’s still work to be done (especially on the energy scale)
Checking Geometry
View from lower chimney
KLG4sim (raytrace) – Glenn Horton-Smith
Geometry implemented in great detail, some small discrepancies persist.
Light Yield vs Distance to
Source
• Cross checked attenuation, scattering and reemission against
bench measurements and Borexino published data
• Still needed a little fiddling of attenuation length to make simulation agree.
Good agreement for accurate
energy reconstruction
Effective Speed of Light
Important for accurate vertexing
plot by Chris Jillings of KamLAND
Measuring the speed of
light with laser calibration
data
A 4% deviation causes a 25 cm bias near the edge of the fiducial volume!
….checking Reconstructed quantities
Vertex Resolution
KLG4sim vertex resolution is
better than data,
Not surprising since we don’t
(yet) model effects that cause
extra smearing.
Waiting for the readout sim!
Energy Deposit Source: Co60 (2γ’s = 2.5 MeV)
Vertex Reconstruction Along
the z-axis
<1% different!
Why an offset?
Is it a geometry
problem?
Energy Resolution
Determined to leading order by PMT occupancy
both in fairly good agreement
compton
scattering off capsule
material
Energy Deposit Source: Co60 (2 γ’s = 2.5 MeV)
Note energy of Co60 at z = 0 tuned to agreement in MC, but nothing else is constrained
Energy Reconstruction Along
the z-axis
Energy estimator has a mysterious bias that is under investigation…
Luckily KLG4sim reproduces it exactly!
Energy Scale
Not as much time spent on this so far…not bad, but could be better
co60 γ
~5% deviation
We need to do some more work!
Isotropic gamma’s
Birk’s Constant: cross check w/ KamLAND data (use α backgrounds)
Cerenkov Contributions: perfect optical propagation and efficiency for detection.
Summary
KLG4sim: Geant4 based, full detector simulation for KamLAND.
Customized features tailor simulation for that of a large-volume anti-neutrino
detector.
Striving for < few % disagreement w/ data to help reduce systematic
uncertainty in future KamLAND measurements.
KamLAND Calibration Data and KLG4sim Comparisons:
- <1% disagreement in vertex and energy reconstruction vs z
at fixed energy.
- ~5% deviations in energy scale, needs more work.
- GeV energy deposits (cosmogenic backgrounds), KLG4sim currently
not optimized
KLG4sim is modular and portable. Many useful features applicable for
research into future reactor experiments → GLG4sim.