Measurement of the mixing angle θ 13 using the reactor neutrino for the Double Chooz experiment Fumitaka Sato A dissertation submitted in partial fulfillment of the requirements for the degree of Doctor of Science Department of Physics, Tokyo Metropolitan University 1-1 MinamiOsawa, Hachioji, 192-0397 Tokyo, Japan August, 2012
110
Embed
Measurement of the mixing angle using the reactor …dchooz.titech.jp.hep.net/~fsato/thesis-history/thesisNov...Measurement of the mixing angle θ13 using the reactor neutrino for
This document is posted to help you gain knowledge. Please leave a comment to let me know what you think about it! Share it to your friends and learn new things together.
Transcript
Measurement of the mixing angle θ13
using the reactor neutrino
for the Double Chooz experiment
Fumitaka Sato
A dissertation submitted in partial fulfillment
of the requirements for the degree of
Doctor of Science
Department of Physics, Tokyo Metropolitan University
1-1 MinamiOsawa, Hachioji, 192-0397 Tokyo, Japan
August, 2012
2
Abstract
The Double Chooz is a reactor anti-neutrino experiment which aims to measurethe neutrino mixing angle θ13. In order to measure or constrain θ13, the overallsystematic errors have to be controlled at the one or sub-percent level. DoubleChooz has two neutrino detectors of identical structure placed undergrounds ofnear (L∼400 m) and far (L∼1050m) location from the Chooz reactors to cancelout many uncertainties associated with neutrino flux and spectrum, detector re-sponse and efficiencies. Construction of the far detector had been finished in 2010.Physics data taking was started in 2011 after the detector commissioning.
In this thesis, XXX days of data recorded in the 20XX-20XX running periodwere analyzed for the measurement of neutrino mixing angle θ13. The oscillationanalysis is performed by evaluating both the deficit of reactor anti-neutrino andthe distortion of neutrino energy spectrum.
We observed XXX...YYY...ZZZ.
ii
Contents
1 Introduction 1
2 Physics Overview 3
2.1 Neutrino in the Standard Model . . . . . . . . . . . . . . . . . . . . . . . . 3
2.1.1 Over view of the Standard Model . . . . . . . . . . . . . . . . . . . 3
layers of 32 plastic scintillator strips (5 × 2 × 320 cm) with wavelength-shifting fibers.
Scintillation lights generated in a scintillator strip is collected through fibers and detected
by multi-anode PMTs. Figure 3.7 and 3.8 show schematic views of OV module and its
arrangement. OV detector can reconstruct vertices where muons interact by coincidences
of different layers and different X-Y dimensional modules with very high resolution (∼few cm). Furthermore, muon track can be reconstructed by coincidence of upper and
lower plane. Total dimension is 6.4 × 12.8 m2 for the lower plane and 3.2 × 6.4 m2 for
upper plane. In addition to IV detected muons, this extended detector provide efficiency
for near-miss muons which could not be detected by IV whereas cause fast neutrons from
interaction with surrounding rock. For the near detector, larger area of OV detector with
11.0 × 12.8 m2 will be implemented because of higher rate muons due to the shallow
depth at the near detector laboratory. In this thesis, only lower plane of OV detector was
in operation. The trigger rate of OV lower plane is ∼2.7 kHz.
3.2.4 Calibration system
The calibration system plays an important role in precise experiments. We must accu-
rately know neutrino signal efficiency and its energy since the θ13 is measured by observing
a few percent of deficit in neutrino rate and it energy distortion with respect to the pre-
diction. Several calibration systems are implemented in the Double Chooz detector to
achieve the precise measurement of θ13, as follows:
Light injection system
Inner Detector Light Injection system (IDLI) is embedded on the Inner-detector PMT
and used for PMT gain and timing calibration. The light from LEDs are transported into
26
6 March 2009 Double Chooz Meeting 2
3225 mm
3625 mm
Mirrored fiber ends
Scintillator strips
Fiber routing
Al skin for module
Fiber holder
M64
FE card
Figure 3.7: A schematic view of the layout of scintillator strip in a OV module.
X layer
Y layer
Figure 3.8: Arrangement of OV modules. Top : X layer modules which provides vertex
position of Y. Bottom : Y layer modules laying on X layer modules provides vertex
position of X.
the detector through an optical fiber arranged along edge of µ-metal (Fig. 3.9). We drive
three types of LEDs emitting different wavelength; λ =385, 425 and 475 nm. Light with
λ =385 nm will be absorbed by the scintillator and then re-emitted. Light with λ =425
27
and 475 nm can pass through the scintillator and reach directory to the PMTs. Intensity
of light is monitored by PIN photodiode and can be controlled from one photoelectron to
several hundred photoelectron level. At the end of fiber, there is a diffuser which spread
the light over the angle of about 22 degrees, or a quartz fibers providing a more narrow
light (about 7 degrees) called pencilh beam. Diffused light is used for PMT gain and
timing calibration and the stability check.
Figure 3.9: Picture of IDLI fiber.
Figure 3.10: Illustration of diffused light. Figure 3.11: Illustration of pencil light.
Radioactive source and deployment system
Response of liquid scintillator and detector are depending on various factors such as
energy, kind of particles (α, β, γ), or vertex of the interaction. Hence, several kind of
28
radioactive sources with deployment system are embedded in the Double Chooz detector.
• 68Ge
68Ge decays by the electron capture to 68Ga, then which decays to stable 68Zn
by e+-decay. Finally, two annihilation gammas which has 1.022 MeV in total are
produced. This energy corresponds to the minimum prompt signal for IBD reaction,
thus allowing to calibrate the efficiency of the trigger threshold at different positions
to make sure all IBD positrons are accepted.
• 252Cf
252Cf emits several neutrons with average multiplicity of 3.76. It can be used to
study neutron efficiency and position dependence of that, in particular close to the
boundary between target and gamma catcher which is called spill-in-out effect. The
neutron energy spectrum of 252Cf is softer than the one of the AmBe source and has
an average of approximately 2.1 MeV.
• 137Cs
137Cs emits 0.662 MeV mono-energetic γ-ray that can be used to calibrate scintil-
lator energy scale with half-life of 30.07 years.
• 60Co
60Co emits 1.17 and 1.33 MeV γ-rays with half-life of 5.27 years.
Level diagram of each isotopes are shown in Fig 3.2.4.
Z-axis deployment system
The Z-axis deployment system allows the radioactive sources to be deployed in the target
along the central axis of the detector from the glove box. This system is used to calibrate
energy response in the target.
Guide tube system
Guide tube is a double teflon tube to deploy a radioactive sources into the Gamma-catcher
region. The tube is installed along with ν-target and γ-catcher acrylic vessels (Fig. 3.13).
At the boundary of the ν-target region, spill-in effect, which the IBD interaction occurs
in γ-catcher region but neutron got into the target region and produce 8MeV gammas,
should be take into account. This system is used for the study of energy response in the
γ-catcher and spill-in effect.
29
60 28Ni
00+
1332.5162+
2158.642+ ~0.0
007 2158.5
7 E
2
~0.0
08 826.0
6 M
1+E
2
0.2
4 1332.5
01 E
2
stable
0.713 ps
0.59 ps
60 27Co
0!
0.230% 7.2
0.0084% 7.4
2+ 58.59
10.47 m
Q"#=2823.90.24%
68 31Ga
01+ 67.629 m
68 32Ge !
100% 5.0
0+ 0
270.82 d
QEC
=106
137 56Ba
03/2+
661.66011/2Ð 85.1
661.6
60 M
4
stable
2.552 m
137 55Cs!
5.6% 12.1
94.4% 9.61
7/2+ 0
30.07 y
Q"#=1175.63
248 96Cm
00+43.382+
143.84+
298.86+
506.08+ 207.2
E
2
~0.0
019 155.0
E
2
~0.0
13 100.4
E
2
0.0
148 43.3
8 E
2
3.40!105 y 121 ps
78 ps
33 ps
13.2 ps
252 98Cf "
81.6% 1.0
15.2% 3.24
0.23% 65
~0.0019% ~1200
~6!10-5% ~2600
0+ 0
2.645 y
Q#=6216.87
96.908%
Figure 3.12: Level diagram of radioactive isotope 60Co, 68Ge, 137Cs, and 252Cf
[?] used in Double Chooz calibration source deployment.
3.3 Electronics and DAQ systems
Electronics of the Double Chooz is shown in Figure 3.14 [?]. Details of each components
are describes following paragraph.
3.3.1 Photo multiplier tube and HV splitter
Scintillation light generated from neutrino interactions is observed by 390 of 10-inch PMTs
on the buffer wall. Double Chooz adopts special low background PMT (Hamamatsu
R7081MODASSY). This PMT is developed based on R7081 used in IceCube experiment[?]
[?]. Design figure and quantum efficiency as a function of wave length are shown Figure
3.15. Basic properties are summarized in Table 3.3. The glass of PMT formed with
30
Target
-catcher
Detector equator (Z=0)
Target wall
-catcher wall
Buffer
Guide Tube
Figure 3.13: An image of Guide tube.
platinum coating glass furnace achieves very low radioactive contamination of 238U, 232Th
and 40K and provide low background condition in the experiment. Each PMT is protected
by a µ-metal against magnetic field and angled in order to ensure a uniform detector
response for the signals from target volume.
PMTs in ID and IV have a single cable for reducing dead volume in the detector and to
avoid ground-roop effects. It reduces cost as well. Hence, the single cable has to carry both
signals and high voltage supply. Splitter circuit, which is combination of high-pass and
low-pass filter, is developed and manufactured by CIEMAT (Centro de Investigaciones
Energeticas Medioambientales y Tecnologicas, Spain) to separate the signal from high
voltage and for noise reduction. The circuit diagram of high voltage splitter is shown in
Fig. 3.16.
Item Specification
Wave length region 300 ∼ 650 nm
Photo cathode Bialkali
Peak wavelength 420 nm
Diameter φ253 mm
Number of dynodes 10
Glass weight ∼ 1,150 g
Table 3.3: Basic specification of R7081
31
Energy deposit
ID-PMTHamamatsu
R7081MODASSY390 PMTs (10Ó)
IV-PMTHamamatsu R1408
78PMTs (8Ó)(from IMB)
HV-SplitterCIEMAT(custom)
HV-SupplyCAEN
SY1527LCA1535P
FEEgain~7
(custom)
VME Crate~16 FADC cards
DAQ software in Ada
Trigger & Clock SystemID: Energy
IV: Energy & pattern62.5MHz clock
MVME3100
22m ID26m IV
18m
ID 16:1IV !6:1
PMT Splitter FEE
HV Trigger
!-FADC500MHz
CAEN-V1721
"-FADC125MHz(custom)
Computers
Figure 3.14: Electronics of the Double Chooz.
Figure 3.15: Design of Hamamatsu R7081 MOD-ASSY and its quantum efficiency as a
function of wave length.
3.3.2 High voltage system
Double Chooz adopted an universal multichannel power supply system manufactured
by CAEN [?]. Figure 3.17 shows the picture of HV main frame SY1527LC and A1535P
module. This HV is used for 390 PMTs for the Inner detector and 78 PMTs for Inner veto.
Thus, total 468 channel of HV are needed. The SY1527 frame has 16 slots for module input
and A1535 has 24 channels for high voltage output. Main properties are summarized in
Table 3.4. In order to uniform the gain of PMTs, the HV system have to provide different
32
Figure 3.16: The circuit diagram of the splitter. Combination of high-pass and low-pass
filter separate signal and HV. Additionally, noise from HV system can be reduced.
voltages to PMTs individually. In Double Chooz, precise measurement of neutrino energy
is a essential to improve the sensitivity and realize the precise measurement of θ13. The
energy is reconstructed from the signal charge of PMTs, hence the high voltage which
directly affects PMT gain is taking an important role in the experiment.
The precision of the output voltage, long term stability of that and HV produced noise
level have been tested. CAEN high voltage system shows good performance to use for
Double Chooz experiment[?].
Figure 3.17: Picture of High Voltage crate and module.
33
Polarity Positive
Output Voltage 0∼3.5 kV
Max. Output Current 3 mA
Voltage Set/Monitor Resolution 0.5 V
Current Set/Monitor Resolution 500 nA
Hardware Voltage Max 0∼3.5 kV
Hardware Voltage Max accuracy ±2 % of Full Scale Range
Software Voltage Max 3.5 kV
Software Voltage Max accuracy 1 V
Ramp Up/Down 1 ∼ 500 V/sec, 1 V/sec step
Voltage Ripple <20 mV typical; 30 mV max
Voltage Monitor vs. Output Voltage Accuracy typical: ±0.3 % ±0.5 V
max:±0.3 % ±2V
Voltage Set vs. Voltage Monitor Accuracy typical: ±0.3 % ±0.5 V
max:±0.3 % ±2V
Current Monitor vs. Output Current Accuracy typical: ±2 % ±1 µA
max:±2 % ±5 µA
Current Set vs. Current Monitor Accuracy typical: ±3 % ±1 µA
max:±2 % ±5 µA
Maximum output power 8W(per channel, soft ware limit)
Power consumption 310 W @ full power
Table 3.4: Properties of CAEN A1535P module
34
3.3.3 Front End Electronics and Flash ADC
Signals from PMTs are separated from high voltage by splitter then sent to front end
electronics (FEE). The FEEs amplify the signals from the Inner detector by a factor of
7.8 to match the dynamic range of following Waveform digitizers(Flash ADC). On the
other hand, signals from the Inner veto events are amplified by a factor of 0.55. The
gain factor is smaller than that of ID since muon events emit large amount of scintillation
lights in IV. In addition, FEE reduces noise in the incoming signal and keep the baseline
voltage stable.
The FEE also sums up analog signals for 8 channels and send stretcher signal to the
trigger system. The stretcher signal has a pulse height proportional to the charge sum of
analog signals.
Signals from FEE are send to CAEN VX1721 flash ADCs shown in Figure 3.18, those
were developed by CAEN SpA with APC (Astro Particule et Cosmologie, Paris)[?].
Each module houses eight channels for input with dynamic range of 1000 mV (8 bit
resolution). The 500 MHz sampling rate provides 2 ns timing resolution. Each channel has
2 MB memory split into pages. The number of pages is adjustable. In case of 1024 pages,
each one can store 2048 samples for a total of 4 µs of digitized data. In the experiment,
time window is set to 256 ns for extra data reduction.
Figure 3.18: Picture of CAEN VX1721 flash ADC board.
3.3.4 Trigger system
ID trigger system consists of three trigger boards (TB) and one trigger master board
(TMB)[?]. Two identical trigger board named TB-A and TB-B are implemented for
inner detector (ID) and one trigger board (TB-IV) is implemented for the inner veto (IV).
Figure 3.19 shows a schematic overview of the system. TB-A and TB-B has 13 inputs,
with each input being an analog sum of 16 PMTs formed by the Front End Electronics
(only one input has 3 PMTs signal sum). TB-IV has 18 inputs with 3 ∼ 6 PMTs signal
sum.
In each input, trigger condition is checked at the end of each 32 ns clock cycle and
to release a trigger signal in case of a fulfilled condition. Discrimination is performed by
35
evaluating pulse height of summed analog signals.
Trigger master board receives digitized trigger signals from each trigger boards and
send trigger signal to FADC boards acting on logical OR operation. External trigger and
its input to the Trigger master board is also implemented for the detector calibration.
ID Group APMTs
ID Group B
Inner Veto
PMTs
PMTs
ExternalTrigger
FEE
FEE
FEE
FEE
FEE
VME
VME
VME
Trigger Board A
Trigger Board B
Trigger Board Veto
TriggerMasterBoard
Fan Outs
EventNumber
TriggerWord
Trigger
Inhibit
SystemClock
LDF
LDF
NIM
NIM
LVDS
VME
(62.5 MHz)
(32 bit)
(32 bit)
(TMB)
16x
18x
TB out
TB out
8x
8x
TB out
8x
7x(calibration, ...)
16x
3x
12x
1x
12x
3x 1x
3x − 6x
Trigger System
Interfaces to the Outer Veto and Muon Electronics are not shown
(TB A)
(TB B)
(TB V)
Inn
er Veto
78 P
MT
s390
PM
Ts
Inn
er Detecto
r
DAQ
Figure 3.19: schematic overview of the trigger system.
36
Chapter 4
Event Reconstruction and Detector
Calibration
In this chapter, event reconstruction for the Double Chooz is presented. In Double Chooz,
390 PMTs detects scintilaltion light generated from energy deposit inside the detector.
Signals from PMT are amplified and digitized by FADCs. Firstry, we sums up digitysed
pulses from each PMTs by impremented pulse reconstruction algorithm. Secondary, in-
tegrated charge is converted to number of photo-electrons by deviding total charge by
calibrated PMT gains. Reconstructed P. E. is finaly transrated to reconstructed energy
considering event vertex, non-linearity of FADC, and stability. Overview of event re-
construction flow is shown in Fig. 4.1. The detail of each reconstructions and detector
calibration method is presented in this chapter. Additionally, muon track reconstruction
method is also presanted.
Energy deposit
PMTs
MeV P. E. Charge
ChargeP. E.
FADCElectronics
Pulse reconstruction
PMT gain
MeV
Energy reconstruction - non linearity - vertex - stability
TrueReconstructed
Figure 4.1: Schematic view of event reconstruction flow.
37
4.1 Pulse reconstruction
Pulse resonstruction and charge calculation tool for DC is impremented and called DCRe-
coPulse[?]. The main purpous of this tool is to provide us total charge and timing infor-
mation of the pulses that we observed. The DCRecoPulse performs following functions.
Baseline calculation
This is the first step to get correct charge and timing information. Two method for baseline
calculation are impremented. One is performed by making use of external triggere event
produced every second(1Hz clock cicle). The mean of all ADC values is computed and
then the sample with the largest deviation from the mean is removed, thus pushing the
mean of the remaining ADC values to the attest region of the readout. This process is
iterated until largest and lowest ADC values have same deviation, within a tolerance of 1
ADC count (or DUI). This method is called External baseline method. Another method
called Floating baseline method is also impremented. In this method, baseline calculation
is performed every self-triggered physics events by taking baseline samples in the biggining
of readout window. Only 10 bins of FADC (20ns) are used for calculation. Both of those
methods have cirtain advantage and also disadvantage.
The External baseline method allows more precise estimation in a typical, however
suffers from baseline shift occures after huge muon-like signals. On the other hand, the
Floating baseline method is more stable against baseline fluctuation but has some draw-
back in accuracy of calculation due to its smallness of integrated window. For example, if
some pre-pulse or dark noise arise in this region, baseline could not be calculated correctly.
Moreover, large pulse get across the FADC time window also hide baseline in integrated
part.
We adopt hibrid method extracting good point of both methods. Namely, the numbers
of both method, mean and RMS value of the baseline, are calculated, then the values of
Floating method is adopted by default. However, if RMS of Floating method is more
than 0.5 DUQ2 larger than that of Extrnal method, former one is considered unreliable,
and the number of External method is adopted.
Pulse charge calculation
After the baseline subtraction, total charge is calculated by integrating ADC counts inside
the fixed-size time window. The time window slides to analyse another part of waveform.
The window position that has the maximum integral is assumed to contain the pulse.
In principle, the algorithm then reiterates and searches for possible other pulses in the
38
Mean 244.5RMS 0.1093
Amp (DUI)243 243.5 244 244.5 245 245.5 246
Ent
ries
0
100
200
300
400
500
600
700
Mean 244.5RMS 0.1093
Mean 125RMS 0.9086
Num. samples120 122 124 126 128 130
Ent
ries
0
100
200
300
400
Mean 125RMS 0.9086
Figure 4.2: Pedestal mean estimation of a sample of 1000 simulated 1PE pulses
with a pedestal level of 244.5 DUI. Left: pedestal mean estimation according to
the External baseline method (solid line), and to the Floating baseline method
assuming a 40 ns window (dashed line). Right: number of time samples used
for the pedestal estimation when using the External baseline method.
waveform, until the the maximum integral in a window falls below the threshold.
Qmin = nσ · σped ·√WS; (4.1)
where nσ is the constant number defined by user, σped is RMS of baseline and WS is the
size of time window set to 112ns by default.
Pulse timing analysis
The DCRecoPulse compute the following timing caracteristics for each found pulse.
• Start time
Time corresponds to 30% of the maximum amplitude before it is reached.
• End time
Time corresponds to 20% of the maximum amplitude, after it was reached.
• Maximum amplitude time
Time corresponding to maximum of the pulse
• Rise time
Time defined as the difference between the maximum amplitude and start times.
39
• Fall time
The fall time is defined as the difference between the end and maximum amplitude
times
4.2 Vertex reconstruction
Vertex reconstruction for the Double Chooz detector is performed by maximum likelihood
method using charge and timing information[?, ?]. Events will be reconstructed is assumed
to be a point-like source produces isotropic light of strength per solid angle Φ (photon/sr).
The expected light at any given PMT can be calculated with simple imaging detector
model, where light propagation is only affected by pure attenuation.
µi = εi × Φ× Ωi × exp(−ri/λ), (4.2)
where εi is quantum efficiency, Ωi is the solid angle subtended by the PMT, ri is the
distance from the source, and λ is the characteristic attenuation length. Assuming the
angular responce function of the ith PMT to be f(cos ηi), where ηi is the angle of incidence
of the light with respect to the ith PMT normal. The solid angle subtended by the ith
PMT with radious R can be written with a approximation (R ¿ ri) as
Ωi = πR2 × f(cos ηi)
r2i
. (4.3)
The optical model is used to predict the amount of light the PMTs see. It is fully
characterized λ and f cos(η). These allow then to calculate the total amount of light
created in an event, which is basically proportional to its total energy. They are essentially
the probability of µi measuring a certain charge where is expected. The timing likelihood
is also obtained simplified detector model as
tpredi = t0 +
ri
cn(4.4)
where cn is the effective speed of light in the scintillator. The event likelihood is defined
as
L(X) =∏qi=0
fq(0;µi)∏qi>0
fq(qi;µi)ft(ti; tpredi , µi) (4.5)
where the first product goes over the PMTs that have not been hit, while the second
product goes over the remaining PMTs that have been hit (i.e., have a non-zero recorded
charge qi at the registered time ti ). fq(qi;µi)is the probability to measure a charge qi
given an expected charge µi, and ft(ti; tpredi , µi) is the probability to measure a time ti
given a prompt arrival time tpredi and predicted charge µi. These are obtained from MC
40
simulations. The task of the event reconstruction is to find the best possible set of event
parameters Xmin which maximizes the event likelihood L(X).
The ideally simplified concept and method of vertex reconstruction is presanted so far,
however, the responce of real detector is complicated in fact. For the cahrge likelihood
function, quantum efficiency of PMTs are not uniform even on their own photocathode
and must be depending on incident angle of photons. Collection efficiency (efficiency of
first diode for generated photoelectron) is also have PMT dependence. Moreover, PMT,
electronics, readout systems have finite energy resolutions of cource. For the timing
function, the situation is much more complicated. The pulses from individual PMTs may
take different time to propagate through the different lengths cables and internal delays
in the electronic circuits. So that, it is required so called “T0 calibration”, which would
equalize the time offsets of individual PMTs and correct the raw pulse time reported by
the RecoPulse. In addition, the time profile of the light emission by a scintillator is not a
delta-function, and has a sharp rise and then decays exponentially with one or more time
constants. The light propagation itself suffers from exponential attenuation, reflections or
scattering at the boundaries between different media inside the detector. To make things
worse, the speed of light is wave-length dependent and may generally differ in different
media. Those parameter in Monte Carlo should be tuned from real calibration data. The
UV laser calibration system with multiple intensities is under preparation. The system
can be used for detector calibration and MC tuning in future.
Figure 4.3, 4.2 shows performance of vertex reconstruction on calibration source data.
As it mentioned at bigging of this section, this algorithm construct with point like source
assumption hence can reconstruct only point like energy deposit events. Events which
widely depositting energy, like muon crossing the detector could not be reconstructed
correctry. Other reconstruction algorithms for muon track reconstruction are also impre-
mented in DC and described in next section.
4.3 Energy reconstruction
4.3.1 PMT gain calibration
PMT gain calibration is a first step for the energy reconstruction and important for all
analysis. This gives a number of photo-electron(P.E.) reconstructed from charge observed
by each PMTs. Due to the uncertainty on the baseline of FADC, observed gain is differs
according to amount of signals. This behavior is called “gain non-linearity”. In order to
resolve this problem, two method for extract the gain for both low-charge and high-charge
signals are performed. After that, two method are combined to obtain so-called linearized
P.E.. The PMT gain calibration is performed using diffused light from lightinjection
41
[mm] true - XrecX-1000 -500 0 500 1000
Eve
nts
0
2000
4000
6000
8000 MC
DATA
Figure 4.3: The vertex destribution for Co-60 events. Source was positioned at the
detector center. Data histogram is background-subtracted.
Z position[mm] -1500 -1000 -500 0 500 1000 1500
Bia
s [m
m]
-50
0
50
100
DATA
MC
Z position [mm] -1500 -1000 -500 0 500 1000 1500
Res
olut
ion
[mm
]
60
80
100
120
140
160DATA
MC
Figure 4.4: Reconstruction bias and resolution are defined as mean and sigma
oftained by gaussian fitting of Figure 4.3. Left: Reconstruction bias as a
function of Z position. Right: Resolution of vertex reconstruction as a function
of Z positoin.
system (IDLI) with low and high intensity.
Single P.E. calibration
The diffused light of IDLI system is useful for illuminate all PMTs. Wavelength of emitted
light is set to 425nm optimesed for allowing PMTs to give a maximum quantum efficiency.
The Single P.E. calibration is performed by fitting the obtained single P.E. peaks(Fig. 4.5).
Hence, in order to obtain pure single P.E. peak, very low intencity light which produce
much lower signal than single P.E. level in avalage is used. The charge distribution
is considered to obey a Poisson and Gaussian distribution. Poisson component models
42
the PMT behavior and Gaussian component, which includes PMT gain, cames from the
resolution of PMT.
F (x) =2∑
n=1
Ne−µµn
√2πnσ1n!
exp
−1
2
(x− na
σ1
√n
)2, (4.6)
where, N is number of single P.E. signals, a is single P.E. peak (= gain), σ1 is single peak
resolution, µ is the expected number of occurrences based on Poisson statistics and n is a
number of P.E.. In this case, only one and two P.E. signals are taken into account. Fig.
4.5 is a example of single P.E. fitting.
Figure 4.5: Example of PMT Gain extraction from single P.E. peak fitting. Four free
parameters (N, µm Gain, σ), one and two P.E. signals are taken into account.(Abe-kun
ni kireina plot wo morau)
Multi P.E. calibration
The Multi P.E. calibration method can provide the gain from higher light signals[?].
Several advantages can be expected by making use of this method; it is not necessary to
know the precise form of the single P.E. spectrum, and it can be used in all light level.
Furthermore, higher signals is less affected by noise contaminating signals and reduce the
FADC non-linearlity.
This is a classic method to calculate the gain of PMT uses a constant avarage number
of photpelectrons N per injection. The gain is obtained from only the variance of charge
distribution. Square of standerd deviation σ of charge distribution is composed of several
kinds of factors.
σ2 = σ2poisson + σ2
spe + σ2noise + ... (4.7)
43
where, σpoisson is fluction of the number of P.E. emitted per light injection which
follows the Poisson distribution, σspe is variation of chrge obtained from each photp-
electron namely it denotes resolution of a PMT, and σnoise is a facter came from noises.
If the signal level is high, that is to say, photon and photo-electron statistics is high,
Poisson distribution can be approximated by Gaussian distribution and noise level can be
negrected.
σ2poisson ' k2N, σ2
spe ' α2k2N,
The gain of PMT(k) is obtained as,
σ2 = σ2poisson + σ2
spe + σ2noise + ...
= k2N(1 + α2)
k =σ2
µ
1
1 + α2(µ = kN)
where µ is a mean number of observed charge and α is a constant relating property of
PMT. The constant parameter α cauld not be derived by this method itself, but can be
determined from Single P.E. calibration.
Linearized P.E. calibration
Combining Single P.E and Multi P.E. calibration, PMT gain as a function of observed
charge is obtained as shown in Fig. 4.6.
The distribution is functionalize with three parameters slope, intersection and con-
stant. The gain is obtained as a function of observed charge by this function.
4.3.2 Time offset calibration
Each channel has different time offset coused by the acquisition system such as; transit
time of each PMT, slightly different length of sinal cables. Relative time offset of each
channel is important information for event reconstruction and especially for the vertex
reconstruction. By applying time offset calibration, the hit time of the PMTs can be
estimated more precisely and more precise reconstruction can be obtained.
Time offset is measured using IDLI calibration system with 425 and 475 nm light,
which dose not exite the scintillator, to avoide the uncertainties of light emmition time
of the scintillator. High intensity light from 8 LEDs is injected periodicaly to cover all
PMTs in the detector.
The pulse time with maximum amplitude is defined as observed time and is corrected
for event-to-event trigger differences using external trigger signals (T obsi −T ext
i ). By fitting
the observed time distribution, mean observed time for given PMTs are extracted. After
44
Charge (arbitrary units)
0 50 100 150 200 250 300 350 400 450
Gai
n (c
harg
e a.
u./P
E)
45
50
55
60
65
70
75
80
85
90
readout gain
slope (non-linearity)slope (non-linearity)
Figure 4.6: Example of extracted PMT Gain as a function of observed charge for one PMT.
Each point cprrespond to different data with different intensity. Low intensity region is
fitted with a linear function while a region above 200 DUQ is fitted with constant [?].
that, mean observed time are plotted as a function of dostance between LEDs and PMTs
for 8 LED sample, the time of flight from the LED to the PMT are estimated and by
fitting the plots by linear function. The slopes are obtained from fitting for each LED
samples and mean value of each slope, which indicates the expected speed of light in the
detector, is calculated. The individual plots are re-fitted again with fixed slope. Finaly,
the relative time offsets are obtained by subtracting expected time(T exti + T ToF
i ) from
observed time for given PMT.
The IDLI runs for time offset calibration are taken every 24 hours (First X month from
data taking starts, it is taken every 12 hours) then, calibration constants are calculated
and applied for every 24 hours period.
4.3.3 Energy reconstruction
The visible energy(Evis) provides the absolute calorimetric estimation of the energy de-
posit per trigger. Evis is calcurated from observed total calibrated PE.
Evis = PEm(ρ, z, t)× fmu (ρ, z)× fm
s (t)× fmMeV , (4.8)
where PE =∑
i pei =∑
i qi/gaini(qi). Coordinates in the detector are ρ and z, t is
time, m refers to data or Monte Carlo and i refers to each good channel. The correction
factor fu, fs and fMeV correspond, respectively, to the spacial uniformity, time stability
and PE/MeV calibrations. PE is a sum of all good channel flaged by waveform analysis.
45
Figure 4.7: Left : Distribution of pulse observed time from external trigger. Right :
Observed time distribution as a function of distance between LED ans PMT.(Abe-kun ni
kireina plot wo morau)
Few channels are flagged not good sporadically and are excluded from the visible energy
calculation. Four stages of calibration are carried out to render Evis. Absolute energy
scale factor fmMeV is determined usong 252Cf source deployed at detector center. Neutrons
emitted from 252Cf source are captured by hydrogen and several number of γ-rays are
emitted. Then, PE/MeV factor is obtained by matching this peak to 2.223MeV energy
deposit. The absolute energy scale are found to be 229.9 PE/MeV and 227.7 PE/MeV
for the data and MC respectively. Each step for correction factor is described in next
paragraph.
Non-uniformity correction
The PE responce is position dependent for both MC and data due to several factors such
as; acceptance of PMTs, detector structure non-uniformity (chimney, acrylic vessel and
its support structure) and difference of attenuation length of liquid in different region.
Detector responce map is applied to cancel the PE bias depending on the interaction
vetex position in the detector. The capture peak on H of neutrons from spallation and
antibutrino interactions provides, for data and for MC respectively, a precise and copious
calibration souorce to charactarize the response non-uniformity over full volume. Correc-
tion factor is defined as fractional responce for each position with respect to the detector
center.
fmu (ρ, z) =
PEmHcapture(0, 0)
PEmHcapture(ρ, z)
. (4.9)
46
Figure 4.8 is the detector responce map applied to data. A 2D-interpolation method was
developed to provide a smooth application of the calibration map at any point (ρ, z).
(m)ρ0 0.2 0.4 0.6 0.8 1 1.2 1.4 1.6 1.8 2
z (m
)
-2
-1.5
-1
-0.5
0
0.5
1
1.5
2
0.84
0.86
0.88
0.9
0.92
0.94
0.96
0.98
1
Target
GC
Figure 4.8: Detector responce map, in cylindrical coordinates(ρ, z) as sampled with spal-
lation neutrons capturing in H across the ID. Responce variations are quantified as the
fractional responce with respect to the detector center. Largest deviation in ν-target are
up to 5%. (Bernd ni MC plot wo moratte naraberu)
Syatematic uncertainty relative to MC is estimated using Gd-captured events. After
the Non-uniformity correction, detector responce map is recomputed using Gd-captured
events and the relative difference between each detector map is defined as
2× EMCi − Edata
i
EMCi + EData
i
, (4.10)
where i is a bin number of the detector map. The RMS deviation of the relative difference
distribution is used as the estimator of the non-uniformity systematic uncertainty, and is
0.43%.
Stability correction
The detector responce was found to vary in time due to two effects, which are accounted
for and corrected by the term fms (t). First, the detector responce can change due to
variation in readout gain or scintillator responce. This effect has been observed as a +2.2%
monotonic increase over 1 year using the responce of the spalaltion neutrons capturing
on Gd within the ν-target, shown in Fig 4.9. Stability correction factor is defined as,
fs(t) =PEm(t0)
PEm(t). (4.11)
47
where t0 is defined as the day of the first Cf source deployment, during August 2011. This
factor is applied to only data since MC is stable.
Elapsed Days0 50 100 150 200 250 300
Ene
rgy
Dev
iatio
n (%
)
-5-4-3-2-1012345
Pea
k en
ergy
(M
eV)
7.6
7.8
8
8.2
Figure 4.9: Time evolution of Gd captured peak position.
After the stability correction using Gd-capture peaks, H-capture peak shows remaining
instability. Systematc uncertainty in stability correction is estimated from flactuation of
H-capture peak from soalation neutrons shown Fig 4.10. RMS in distribution of relative
energy variation is defined as instability syatematics and the value is 0.61%.
Elapsed day0 50 100 150 200 250 300
Ene
rgy
devi
atio
n [%
]
-5-4-3-2-1012345
Pea
k en
ergy
[MeV
]
2.15
2.2
2.25
2.3
Energy variation (%)-4 -3 -2 -1 0 1 2 3 4
Ent
ries
/ 5 d
ays
0
2
4
6
8
10
12
14
16 Mean : -0.349
RMS : 0.606
Figure 4.10: Stability of the reconstructed energy as sampled by the evolution in responce
of the spallation neutron H-capture after Gd stability correction. Left:Time evolution
plot. Right: 1-D projection histogram. Systematics in tability correction is estimated
from RMS deviation of right histogram.
48
Non linearity systematics
After the all energy correction, data/MC discrepancies in the absolute energy scale can
still arise from the relative non-linearity across the prompt energy spectrum. This possi-
bility was explored by using all calibration sources in the energy range 0.7 - 8 MeV with
deployments along the z-axis and guide tube. Some relative non-linearity was observed (¡
0.2% / MeV) but the pattern diminished when integrated over the fill volume. A 0.85%
variation consistent with this non-linearity was measured with the z-axis calibration sys-
tem, and this is used as the systematic uncertainty for relative non-linearity. Systematic
uncertainties in energy scale are summarised in Table 4.1.
Error (%)
Relative Non-Uniformity 0.43
Relative Instability 0.61
Relative Non-Linearity 0.85
Total 1.13
Table 4.1: Systematic uncertainties on energy scale.
4.4 Muon track reconstruction
Muon tracks are reconstructed by different algorithm from point like events. Several
methods using information from the different detector are deveropped.
4.4.1 ID muon reconstruction
Muon reconstruction algorithm based on ID information is deveropped by Hamburg uni-
versity. The muon track is reconstructed by PMT hit timing assuming straight trajectory
inside the detector and spherical light fronts emitted along the muon track. Maximum
likelihood method determins most probabli tracks and outputs muon entry point(θin, φin)
and exit point(θout, φout). This algorithm requires the muon passes at least γ-catcher
volume and deposit large energy. Therefore, buffer or IV clipping muon caould not be
reconstructed by this algorithm and need to rely on other reconstruction method. Figure
4.11 shows reconstruction performance using OV hit information as a reference.
4.4.2 IV muon reconstruction
Another reconstruction algorithm using only IV information reconstruct muons more
higher efficiency. Even OV or IV clipping muons can be reconstructed. The method
49
FDEGF*+,>,HDE, IDEGI*+,>,HDE,
Figure 4.11: Resolution of muon entry point projected on OV serfice. (Atode jibunnde
kireina plot wo tsukutte haru)
is a Maximum likelihood using charge and timing information from IV PMTs. Likelihood
function is generated from MC simulation. Reconstruction performance is shown Fig.
4.12.
K*5/+(
K:$5
P>QRNH(70(
P>QRNS(70
Figure 4.12: Atode sasikaemasu.
Performance of each reconstruction method is summarized in Table 4.2.
Algorithm resolution on OV(cm) efficiency (%)
ID based XXX XXX
IV based XXX XXX
Table 4.2: Reconstruction performance for each method.suuji dasimasu
50
Chapter 5
Monte Carlo simulation
5.1 Electron anti-neutrino generation
5.2 Detector simulation
5.3 Readout system simulation
51
52
Chapter 6
Selection of neutrino candidates
In order to select neutrino candidates, we require the delayed coincidence, which satisfies
two signals (prompt and delayed) and the time correlation. Derivation of θ13 is performed
by comparing data and MC spectrum. Hence, the discrepancy between data and MC
becomes mandatory parameter for oscillation analysis as described in Chap. 8. The selec-
tion efficiency and those systematic uncertainties were estimated using calibration source
data.
6.1 Strategy for neutrino selection
The event characteristics of neutrino interactions by the inverse beta decay are two in-
dependent signals by a positron and a neutron, respectively, and the time correlation
between those. As presented in Sec. 3.1, selection criteria are based on the following
requirements:
• Energy of prompt signal is to be produced by neutron energy and positron annihi-
lation,
• Energy of delayed signal is to be 8 MeV from neutron capture on Gd.
• Time correlation is to be 30 µs in average.
For event selection, we veto all signals within 1 ms after a muon, since many un-
expected effects occur. Moreover, additional cuts to reduce backgrounds, such as light
emitted by PMT itself, are required. The selection criteria based on these characteristics
are described in the Sec. 6.4.
53
6.2 Data sample
Double Chooz started the official data taking in April 2011. Figure 6.1 shows the data
taking history since data taking start. The data sample collected with the Double Chooz
far detector during the period from April 2011 to March 2012 are used in this thesis. The
total amount of run time is 251.27 days, as shown in Tab. 6.1. For a few weeks in August
to September 2011, the radioactive source data were taken for the detector calibration,
so that the corresponding runs in this period were excluded from the neutrino analysis.
Moreover, data with different configrations for several detector studies, such as PMT noise
test, were not used. In addition, the data sets are limited to those in which the electronics
have been operating properly. In the beginning of September 2011, one PMT produced
the PMT noise with high rate, and we turned off the HV for the PMT. Data with the
high rate noise before turning off HV were also removed from the analysis.
Double Chooz catches neutrino from two reactors only. Both reactors sometimes are
down for the mentainance. This oppotunity gives us to estimate backgrounds purely, by
assuming no neutrino signal. During the period used in this thesis, there were twice of
such a chance, which corresponds to Oct 2011 and May-June 2012. The backgground
analysis using such a data set is described in Chap. 7.