Post on 09-Aug-2020
transcript
Detector Simulation and Performance
R. Bayes1,A. Bross3, A. Cervera-Villanueva2 , M. Ellis4,5, Tapasi Ghosh2
A. Laing1 , F.J.P. Soler1, Chris Tunnell6 and R. Wands3
1University of Glasgow, 2IFIC and Universidad de Valencia, 3Fermilab, 4Brunel University,5Westpac Institutional Bank, Australia, 6University of Oxford
September 21, 2012
R. Bayes (University of Glasgow) Detector Simulation and Performance September 21, 2012 1 / 17
1 Simulation Overview
2 Analysis
3 Outlook and Future Development
R. Bayes (University of Glasgow) Detector Simulation and Performance September 21, 2012 2 / 17
Introduction
Both νe and ν̄µ are present in the νSTORM beamFour oscillation detection modes are possible
1 νµ appearance: νe → νµ, µ− signal2 ν̄µ disappearance: ν̄µ → ν̄e, µ+ signal3 νe disappearance: νe → νµ, e− signal4 ν̄e appearance: ν̄µ → ν̄e, e+ signal
Central requirement is charge discriminationRequires a magnetic field and good detector efficiencyA magnetized iron neutrino detector fulfills these requirements fora µ± signal.
R. Bayes (University of Glasgow) Detector Simulation and Performance September 21, 2012 3 / 17
Introduction
Both νe and ν̄µ are present in the νSTORM beamFour oscillation detection modes are possible
1 νµ appearance: νe → νµ, µ− signal2 ν̄µ disappearance: ν̄µ → ν̄e, µ+ signal3 νe disappearance: νe → νµ, e− signal4 ν̄e appearance: ν̄µ → ν̄e, e+ signal
Central requirement is charge discriminationRequires a magnetic field and good detector efficiencyA magnetized iron neutrino detector fulfills these requirements fora µ± signal.
R. Bayes (University of Glasgow) Detector Simulation and Performance September 21, 2012 3 / 17
Simulation Overview
Detector Design
Detector consists of layerediron and scintillator planesIron plates 1 cm(2 cm) thick.Scintillator planes 2 cm thick.
Composed of scintillatorbars 1 cm thick and 1 cmwidth.Measure x and y position ateach plane.
Circular cross-section, 5 mdiameter.20 m long for 1 kton mass.Magnetization achieved withSCTL
IV. FAR DETECTOR - SUPERBIND
The Super B Iron Neutrino Detector (SuperBIND) is an iron and scintillator sampling
calorimeter which is similar in concept to the MINOS detectors [76]. We have chosen a
cross section of approximately 5 m in order to maximize the ratio of the fiducial mass to
total mass. The magnetic field will be toroidal as in MINOS and SuperBIND will also use
extruded scintillator for the readout planes. Details on the iron plates, magnetization, scin-
tillator, photodetector and electronics are given below. Fig. 33 gives an overall schematic
of the detector. We note that within the Advanced European Infrastructures for Detectors
Figure 33. Far Detector concept
at Accelerators (AIDA) project, whose time line runs from 2011 to 2015, detectors similar
to those planned for !STORM will be built and characterized at CERN. The motivation is
to test the capabilities for charge identification of ! 5GeV/c electrons in a Totally Active
Scintillator Detector and !5 GeV/c muons in a Magnetized Iron Neutrino Detector (MIND).
These detector prototypes will provide further experience in the use of STL technology, and
SiPMs and associated electronics, to complement the already large body of knowledge gained
through past and current operation of this type of detector.
28
R. Bayes (University of Glasgow) Detector Simulation and Performance September 21, 2012 4 / 17
Simulation Overview
SuperBIND Simulation
Based on MIND simulation forthe Neutrino FactoryNeutrino events simulated inGENIE.Detector geometry andmaterials simulated withGEANT4.Scintillator plane simulated asa polystyrene slab.
hits grouped into discretebars and attenuate indigitization.
Use toroidal field: model fromfit to simulation of field.
current represents approximately 80% of the critical current achieved at 6.5K in the STL
test stand assembled for the VLHC proof-of-principle tests.
Figure 33. Toroidal Field Map
C. Detector planes
1. Scintillator
Particle detection using extruded scintillator and optical fibres is a mature technology. MI-
NOS has shown that co-extruded solid scintillator with embedded wavelength shifting (WLS)
fibres and PMT readout produces adequate light for MIP tracking and that it can be manu-
factured with excellent quality control and uniformity in an industrial setting. Many exper-
iments use this same technology for the active elements of their detectors, such as the K2K
Scibar [74], the T2K INGRID, the T2K P0D, the T2K ECAL [75] and the Double-Chooz
detectors [76].
Our initial concept for the readout planes for SuperBIND is to have both an x and a y
view between each plate. The simulations done to date have assumed a scintillator extrusion
profile that is 1.0 ! 1.0 cm2. This gives both the required point resolution and light yield.
2. Scintillator extrusions
The existing SuperBIND simulations have assumed that the readout planes will use a rect-
angular extrusion that is 1.0 ! 1.0 cm2. A 1 mm hole down the centre of the extrusion is
provided for insertion of the wavelength shifting fibre. This is a relatively simple part to
manufacture and has already been fabricated in a similar form for a number of small-scale
applications. The scintillator strips will consist of an extruded polystyrene core doped with
blue-emitting fluorescent compounds, a co-extruded TiO2 outer layer for reflectivity, and
26
R. Bayes (University of Glasgow) Detector Simulation and Performance September 21, 2012 5 / 17
Simulation Overview
Reconstruction
Uses a Kalman filter for pattern recognition and track fitting.
Longest set of single hitsidentified as muon.Further hits filtered into track.Fits assume
range of track as estimate ofmomentum .sum of deviation of track fromstraight-line in magnetic fieldestimates the charge.
Only the muon track is fit.No hadron reconstruction fromdigitized events.
Position Along Detector Axis (in m)-10 -5 0 5 10
Rad
ial D
ista
nce
from
Det
ecto
r Axi
s (in
cm
)
0
50
100
150
200
250
Position Along Detector Axis (in m)-10 -5 0 5 10
Rad
ial D
ista
nce
from
Det
ecto
r Axi
s (in
cm
)
0
50
100
150
200
250
R. Bayes (University of Glasgow) Detector Simulation and Performance September 21, 2012 6 / 17
Simulation Overview
Reconstruction
Uses a Kalman filter for pattern recognition and track fitting.
Longest set of single hitsidentified as muon.Further hits filtered into track.Fits assume
range of track as estimate ofmomentum .sum of deviation of track fromstraight-line in magnetic fieldestimates the charge.
Only the muon track is fit.No hadron reconstruction fromdigitized events.
Position Along Detector Axis (in m)-10 -5 0 5 10
Rad
ial D
ista
nce
from
Det
ecto
r Axi
s (in
cm
)
0
50
100
150
200
250
Position Along Detector Axis (in m)-10 -5 0 5 10
Rad
ial D
ista
nce
from
Det
ecto
r Axi
s (in
cm
)
0
50
100
150
200
250
R. Bayes (University of Glasgow) Detector Simulation and Performance September 21, 2012 6 / 17
Simulation Overview
Software Summary
Software is modular.Parts are interchangeable.Information between simulation and reconstruction uses a "bhep"format.
Beam Flux
GENIE
GEANT4
mindG4 Digitization
BHEP
RecPack
mind_rec ROOT Tree
Event Selection
R. Bayes (University of Glasgow) Detector Simulation and Performance September 21, 2012 7 / 17
Analysis
Charge Current Selection
Identify muon signals.Reject tracks from NC eventsand shower processes.Analysis simplified to six cuts.
Successful reconstructionpµ < 1.6× Eµ.Track vertex before last 1 mof detector volumeFitted track includes >60%of candidate hits.Scaled curvature uncertaintyis CC event like.Number of candidate hits isCC event like.
Greatest analytic power inlikelihood cuts.
cuts_passed_sigEntries 3461334Mean 0RMS 0
Fraction of Events Passed by Cuts-610 -510 -410 -310 -210 -110 1
Reconstruction Success
Fiducial
Max Momentum
Fitted proportion
Track quality
CC Selection
CCµν signal from -µ
CCµνBackground from
NCµνBackground from
CCeνBackground from
NCeνBackground from
R. Bayes (University of Glasgow) Detector Simulation and Performance September 21, 2012 8 / 17
Analysis
Muon Selection from Uncertainty in Curvature
/(q/p)q/p
σ0 0.5 1 1.5 2 2.5 3 3.5 4
(1/N
) dN
/dx
0
0.01
0.02
0.03
0.04
0.05
. Correct Charge ID+µ
. Incorrect Charge ID-µ
q/pL-5 -4 -3 -2 -1 0 1 2
Occ
upan
cy
0
0.05
0.1
0.15
0.2
0.25
0.3
0.35
0.4 Tracks+µ CC, µν
Tracks-µ CC, µν
Tracks-µ CC, µν
Tracks+µ CC, µν
Distributions of |σq/pq/p | compiled for
CC events with correct chargeCC events with incorrect chargeNC events
Distributions used to define aquantity
Lq/p = logP(σq/p/(q/p)|CC))
P(σq/p/(q/p)|NC))
Allow events with Lq/p > 0.5.A "weak" cut to remove signalfrom background.
R. Bayes (University of Glasgow) Detector Simulation and Performance September 21, 2012 9 / 17
Analysis
Muon Selection from Number of Hits
Number of Hits in Trajectory0 20 40 60 80 100 120 140
Pro
babi
lity
0
0.1
0.2
0.3
0.4
0.5
0.6 CCµν NCµν CCµν
L_1-4 -2 0 2 4 6 8 10 12
Occ
upan
cy
0.05
0.1
0.15
0.2
0.25
0.3
0.35 Tracks+µ CC, µν
Tracks-µ CC, µν
Tracks-µ CC, µν
Tracks+µ CC, µν
Number of candidate hits in muontrajectory compiled for
CC eventsNC events
Distributions used to define aquantity
LCC = logP(Ncand |CC))
P(Ncand |NC))
Allow events with LCC > 6.5A very strong cut to removebackground.Also good at eliminating lowenergy signal.
R. Bayes (University of Glasgow) Detector Simulation and Performance September 21, 2012 10 / 17
Analysis
Efficiency and Background Rejection
1 cm Plate
True Neutrino Energy0 0.5 1 1.5 2 2.5 3 3.5
Fra
ctio
nal E
ffici
ency
0
0.05
0.1
0.15
0.2
0.25
0.3
0.35
0.4
0.45
2 cm Plate
True Neutrino Energy0 0.5 1 1.5 2 2.5 3 3.5
Fra
ctio
nal E
ffici
ency
0
0.05
0.1
0.15
0.2
0.25
0.3
0.35
0.4
1 cm Fe plates and 2 cm Fe plates considered.1 cm plate initially favoured to improve energy threshold.Rejection of charge mis-ID events better in 2 cm plate.Improvement due to the larger magnetic field.
R. Bayes (University of Glasgow) Detector Simulation and Performance September 21, 2012 11 / 17
Analysis
Efficiency and Background Rejection
1 cm Plate
True Neutrino Energy (GeV)0 0.5 1 1.5 2 2.5 3 3.5 S
urvi
ving
Fra
ctio
n of
Bac
kgro
und
-610
-510
-410
-310CC eventµν ID from -µ
NC eventµν ID from -µ
CC eventeν ID from -µ
NC eventeν ID from -µ
2 cm Plate
True Neutrino Energy (GeV)0 0.5 1 1.5 2 2.5 3 3.5
Sur
vivi
ng F
ract
ion
of B
ackg
roun
d
-610
-510
-410
-310CC eventµν ID from -µNC eventµν ID from -µCC eventeν ID from -µNC eventeν ID from -µ
1 cm Fe plates and 2 cm Fe plates considered.1 cm plate initially favoured to improve energy threshold.Rejection of charge mis-ID events better in 2 cm plate.Improvement due to the larger magnetic field.
R. Bayes (University of Glasgow) Detector Simulation and Performance September 21, 2012 11 / 17
Analysis
Detector Response
Signal Response, 2 cm Plate
Reconstructed Neutrino Energy (GeV)0 0.5 1 1.5 2 2.5 3 3.5 4 4.5 5
Tru
e N
eutr
ino
Ene
rgy
(GeV
)
0
0.5
1
1.5
2
2.5
3
3.5
4
0
0.02
0.04
0.06
0.08
0.1
0.12
Background Response
Reconstructed Neutrino Energy (GeV)0 0.5 1 1.5 2 2.5 3 3.5 4 4.5 5
Tru
e N
eutr
ino
Ene
rgy
(GeV
)
0
0.5
1
1.5
2
2.5
3
3.5
4
0
5
10
15
20
25
30
-610×
Full energy reconstruction still lacking.
Eν =mNEµ + 1
2(m2N′ −m2
N −m2µ)
mN − Eµ + pµ cos θfor QES events, or
Eν = Eµ + Ehad , Ehad is smeared byδEhad
Ehad=
0.55√Ehad
+ 0.03
R. Bayes (University of Glasgow) Detector Simulation and Performance September 21, 2012 12 / 17
Analysis
Sensitivity to Sterile Neutrinos
0.0 0.5 1.0 1.5 2.0 2.5 3.0 3.5 4.0Neutrino Energy [GeV]
0
1
2
3
4
5
6
7
8
NumberofEvents
/0.30GeV
νµapp. bkg.: 6.0 ev
νµapp. sig.: 61.2 ev
Above results synthesized byChris Tunnel into sensitivities10σ goal is reasonableachieved.
Only statistical uncertaintiesincluded.Consideration of systematicerrors required.
Figure 46. Contour in sterile parameter space associated with !e ! !µ appearance. Assumed is1.8"1018 stored µ+ at p = (3.8±0.38) GeV/c and a detector at 2 kilometers with a fiducial mass of1.3 kilotonne. A smearing matrix is used corresponding to 2 cm steel plates. The 150 m integrationstraight and detector volume are integrated over. The CPT-conjugate of the LSND best-fit regionis shown.
the energy of the ring if the decay ring cost become excessive. As the cuts-based detector
performance improves, the various background rejections (Fig. 48 and 49) allow for further
overall optimizations with respect to physics reach. The tools have been developed that will
allow us to optimize over all components of !STORM.
2. Disappearance channels
Since disappearance measurements are very sensitive to the signal normalization, additional
near detectors have been proposed in !̄e disappearance reactor experiments to measure
"13 [105, 106]. These near detectors are supposed to be as similar as possible to the far
detectors, where the main purpose is to control the uncertainty on the reactor neutrino
fluxes. This concept has been well established, and can be found in all of the state-of-the-
art reactor experiments, such as Double Chooz, Daya Bay, and RENO. For !STORM, the
situation is very similar: while the flux is well under control, cross sections " e!ciencies
must be measured by a near detector. However, since oscillations may already take place in
the near detector, the oscillation parameters need to be extracted in a self-consistent way in
a combined near-far fit [107]. In fact, the near and far detectors may even swap the roles:
while for "m2 # 1 eV2, the near detector e#ectively measures the cross sections and the far
detector the oscillation, for "m2 $ 10 eV2, the near detector measures the oscillations and
46
R. Bayes (University of Glasgow) Detector Simulation and Performance September 21, 2012 13 / 17
Outlook and Future Development
Short Term Progress
Improvements made parasitically to NuFact MIND developmentFit multiple trajectories.
Allow for muon reconstruction at lower momenta.Identify set of hadron hits.
Introduce multi variate analysis for CC selectionUse more variables than NhitsPossible variables include mean energy deposition and variation inenergy deposition.
Quantify systematic uncertaintiesBackground due to cosmic rays.Cross-section uncertainties.Fiducial uncertainty.
R. Bayes (University of Glasgow) Detector Simulation and Performance September 21, 2012 14 / 17
Outlook and Future Development
Multiple Trajectory FitsSome events11
DISDIS
DIS DIS
R=
p (X2
+Y
2)
Figures from Tapasi Ghosh,4th Annual EUROnu Meeting.
Secondary tracks observed inDIS and QES events.Reduces event into series oftrajectories
Longest set of hits identified.Hits filtered into trajectory.Repeat with remaining hits.Stop when less than 5 hitsare left.
Working on the best way touse this information.
R. Bayes (University of Glasgow) Detector Simulation and Performance September 21, 2012 15 / 17
Outlook and Future Development
Multiple Trajectory Fits
Some events11
DISDIS
DIS DIS
R=
p (X2
+Y
2)
Figures from Tapasi Ghosh,4th Annual EUROnu Meeting.
Secondary tracks observed inDIS and QES events.Reduces event into series oftrajectories
Longest set of hits identified.Hits filtered into trajectory.Repeat with remaining hits.Stop when less than 5 hitsare left.
Working on the best way touse this information.
R. Bayes (University of Glasgow) Detector Simulation and Performance September 21, 2012 15 / 17
Outlook and Future Development
Multiple Trajectory FitsSome Events12
QES QES
R Y
Figures from Tapasi Ghosh,4th Annual EUROnu Meeting.
Secondary tracks observed inDIS and QES events.Reduces event into series oftrajectories
Longest set of hits identified.Hits filtered into trajectory.Repeat with remaining hits.Stop when less than 5 hitsare left.
Working on the best way touse this information.
R. Bayes (University of Glasgow) Detector Simulation and Performance September 21, 2012 15 / 17
Outlook and Future Development
Multiple Trajectory FitsSome Events12
QES QES
R Y
Figures from Tapasi Ghosh,4th Annual EUROnu Meeting.
Secondary tracks observed inDIS and QES events.Reduces event into series oftrajectories
Longest set of hits identified.Hits filtered into trajectory.Repeat with remaining hits.Stop when less than 5 hitsare left.
Working on the best way touse this information.
R. Bayes (University of Glasgow) Detector Simulation and Performance September 21, 2012 15 / 17
Outlook and Future Development
Multi-variate Analysis
Track Quality [F]-4 -2 0 2 4
0.25
7 F
/ (1
/N)
dN
0
0.5
1
1.5
2
2.5
3
3.5
4 SignalBackground
U/O
-flo
w (
S,B
): (
0.1,
0.1
)% /
(0.1
, 0.1
)%
Input variable: Track Quality
Hits in Trajectory [F]20 40 60 80 100 120
2.97
F /
(1/N
) dN
0
0.005
0.01
0.015
0.02
0.025
U/O
-flo
w (
S,B
): (
0.0,
0.0
)% /
(0.0
, 0.0
)%
Input variable: Hits in Trajectory
Fractional Energy Depostion [F]0.2 0.4 0.6 0.8 1
0.02
56 F
/ (1
/N)
dN
0
2
4
6
8
10
U/O
-flo
w (
S,B
): (
0.0,
0.0
)% /
(0.0
, 0.0
)%
Input variable: Fractional Energy Depostion
Mean Energy Deposition [F]2 4 6 8 10 12
0.30
5 F
/ (1
/N)
dN
0
0.2
0.4
0.6
0.8
1
1.2
U/O
-flo
w (
S,B
): (
0.0,
0.0
)% /
(0.4
, 0.1
)%
Input variable: Mean Energy Deposition
Variation of Energy Deposition [F]0.2 0.4 0.6 0.8 1
0.02
54 F
/ (1
/N)
dN
0
0.5
1
1.5
2
2.5
3
3.5
4
4.5
U/O
-flo
w (
S,B
): (
0.0,
0.0
)% /
(0.0
, 0.0
)%
Input variable: Variation of Energy Deposition
Momentum (GeV/c) [F]1 2 3 4 5 6
0.15
6 F
/ (1
/N)
dN
0
0.1
0.2
0.3
0.4
0.5
U/O
-flo
w (
S,B
): (
0.0,
0.0
)% /
(0.0
, 0.0
)%
Input variable: Momentum (GeV/c)
Approximate NC background analysisVariables in CC background analysis are notsuitable.
Considered 6variablesTrained for CC andNC bkgnd rejectionStill needs work
betterunderstanding ofvariabledistributions
R. Bayes (University of Glasgow) Detector Simulation and Performance September 21, 2012 16 / 17
Outlook and Future Development
Multi-variate Analysis
Track Quality [F]0 200 400 600 8001000120014001600180020002200
55.1
F /
(1/N
) dN
00.0020.0040.0060.008
0.010.0120.0140.0160.0180.02
0.022 SignalBackground
U/O
-flo
w (
S,B
): (
0.0,
0.0
)% /
(0.0
, 0.0
)%
Input variable: Track Quality
Hits in Trajectory [F]20 40 60 80 100 120
2.97
F /
(1/N
) dN
0
0.020.040.06
0.080.1
0.120.14
0.160.180.2
0.22
U/O
-flo
w (
S,B
): (
0.0,
0.0
)% /
(0.0
, 0.0
)%
Input variable: Hits in Trajectory
Fractional Energy Depostion [F]0.2 0.4 0.6 0.8 1
0.02
56 F
/ (1
/N)
dN
0
0.5
1
1.5
2
2.5
3
3.5
4
U/O
-flo
w (
S,B
): (
0.0,
0.0
)% /
(0.0
, 0.0
)%
Input variable: Fractional Energy Depostion
Mean Energy Deposition [F]2 4 6 8 10 12 14 16 18
0.47
F /
(1/N
) dN
0
0.1
0.2
0.3
0.4
0.5
0.6
0.7
0.8
0.9
U/O
-flo
w (
S,B
): (
0.0,
0.0
)% /
(0.0
, 0.7
)%
Input variable: Mean Energy Deposition
Variation of Energy Deposition [F]0.2 0.4 0.6 0.8 1
0.02
53 F
/ (1
/N)
dN
0
0.5
1
1.5
2
2.5
3
3.5
4
4.5
U/O
-flo
w (
S,B
): (
0.0,
0.0
)% /
(0.0
, 0.0
)%
Input variable: Variation of Energy Deposition
Momentum (MeV/c) [F]0 500 10001500200025003000350040004500
310×
1.19
e+05
F /
(1/N
) dN
0
1
2
3
4
5
6
7
8
9
-610×
U/O
-flo
w (
S,B
): (
0.0,
0.0
)% /
(0.0
, 0.0
)%
Input variable: Momentum (MeV/c)
Approximate NC background analysisVariables in CC background analysis are notsuitable.
Considered 6variablesTrained for CC andNC bkgnd rejectionStill needs work
betterunderstanding ofvariabledistributions
R. Bayes (University of Glasgow) Detector Simulation and Performance September 21, 2012 16 / 17
Outlook and Future Development
Multi-variate Analysis
Cut value applied on KNN output0 0.2 0.4 0.6 0.8 1
Eff
icie
ncy
(P
uri
ty)
0
0.2
0.4
0.6
0.8
1
Signal efficiency
Background efficiency
Signal puritySignal efficiency*purity
S+BS/
For 400 signal and 40000 background isS+Bevents the maximum S/
13.1806 when cutting at 0.9876
Cut efficiencies and optimal cut value
Sig
nif
ican
ce
024
6
81012
14
Approximate NC background analysisVariables in CC background analysis are notsuitable.
Considered 6variablesTrained for CC andNC bkgnd rejectionStill needs work
betterunderstanding ofvariabledistributions
R. Bayes (University of Glasgow) Detector Simulation and Performance September 21, 2012 16 / 17
Outlook and Future Development
Summary
We have a simulation of a MIND developed for the neutrino factoryMIND simulation has been used to develop SuperBIND.Detector can achieve sterile neutrino physics goals.
In absence of knowledge of systematics 10σ.Next steps:
Develop hadron reconstruction.Develop multi-variate analysis.Quantify systematics.Make improved user interface.
R. Bayes (University of Glasgow) Detector Simulation and Performance September 21, 2012 17 / 17