The first result of the MiniBooNE neutrino oscillation experiment

1
The first result of the MiniBooNE neutrino
oscillation experiment
Teppei Katori for the MiniBooNE collaboration
Indiana University McGill University, Montreal,
May., 09, 07
2
The first result of the MiniBooNE neutrino
oscillation experiment
outline1. Neutrino
oscillation2. LSND experiment3. Neutrino
beam4. Events in the detector5. Cross section
model6. Oscillation analysis7. Systematic error
analysis8. The MiniBooNE initial results9. Low
energy excess events10. Future plans
3
1. Neutrino Oscillation
4
1. Neutrino oscillation
The neutrino weak eigenstate is described by
neutrino Hamiltonian eigenstates, n1, n2, and n3
and their mixing matrix elements.
The time evolution of neutrino weak eigenstate is
written by Hamiltonian mixing matrix elements and
eigenvalues of n1, n2, and n3.
Then the transition probability from weak
eigenstate nm to ne is
So far, model independent
5
1. Neutrino oscillation
From here, model dependent formalism. In the
vacuum, 2 neutrino state effective Hamiltonian
has a form,
Therefore, 2 massive neutrino oscillation model is
Or, conventional form
6
2. LSND experiment
7
2. LSND experiment
LSND experiment at Los Alamos observed excess of
anti-electron neutrino events in the anti-muon
neutrino beam. 87.9 22.4 6.0 (3.8.s)
LSND Collaboration, PRD 64, 112007
8
2. LSND experiment
3 types of neutrino oscillations are found LSND
neutrino oscillation
Dm21eV2 Atmospheric neutrino oscillation
Dm210-3eV2 Solar neutrino oscillation
Dm210-5eV2 But we cannot have so many Dm2!
We need to test LSND signal MiniBooNE experiment
is designed to have same L/E500m/500MeV1 to
test LSND Dm21eV2
9
2. MiniBooNE experiment
Keep L/E same with LSND, while changing
systematics, energy event signature P(nm-ne)
sin22q sin2(1.27Dm2L/E) MiniBooNE is looking for
the single isolated electron like events, which
is the signature of ne events
MiniBooNE has - higher energy (500 MeV) than
LSND (30 MeV) - longer baseline (500 m) than
LSND (30 m)
10
3. Neutrino beam
11
3. Neutrino beam
MiniBooNE extracts beam from the 8 GeV Booster
12
3. Neutrino beam
8GeV protons are delivered to a 1.7 l Be target
within a magnetic horn (2.5 kV, 174 kA)
that (increases the flux by ? 6)
p-
p
p
p-
13
3. Neutrino beam
Modeling Production of Secondary Pions - 5 l
Beryllium target - 8.9 GeV proton beam momentum
Data are fit to a Sanford-Wang parameterization.
HARP collaboration, hep-ex/0702024
14
3. Neutrino beam

• Neutrino Flux from GEANT4 Simulation
• MiniBooNE is the ne appearance oscillation
  experiment
• Intrinsic ne ?ne sources
• m ? e ?nm ne (52)
• K ? p0 e ne (29)
• K0 ? p e ne (14)
• Other ( 5)

ne/nm 0.5 Antineutrino content 6
15
4. Events in the Detector
16
4. Events in the Detector
The MiniBooNE Detector - 541 meters downstream
of target - 3 meter overburden - 12 meter
diameter sphere (10 meter fiducial
volume) - Filled with 800 t of pure mineral oil
(CH2) (Fiducial volume 450 t) - 1280 inner
phototubes, - 240 veto phototubes Simulated
with a GEANT3 Monte Carlo
17
4. Events in the Detector
The MiniBooNE Detector - 541 meters downstream
of target - 3 meter overburden - 12 meter
diameter sphere (10 meter fiducial
volume) - Filled with 800 t of pure mineral oil
(CH2) (Fiducial volume 450 t) - 1280 inner
phototubes, - 240 veto phototubes Simulated
with a GEANT3 Monte Carlo
541 meters
Booster
18
4. Events in the Detector
The MiniBooNE Detector - 541 meters downstream
of target - 3 meter overburden - 12 meter
diameter sphere (10 meter fiducial
volume) - Filled with 800 t of pure mineral oil
(CH2) (Fiducial volume 450 t) - 1280 inner
phototubes, - 240 veto phototubes Simulated
with a GEANT3 Monte Carlo
19
4. Events in the Detector
The MiniBooNE Detector - 541 meters downstream
of target - 3 meter overburden - 12 meter
diameter sphere (10 meter fiducial
volume) - Filled with 800 t of pure mineral oil
(CH2) (Fiducial volume 450 t) - 1280 inner
phototubes, - 240 veto phototubes Simulated
with a GEANT3 Monte Carlo
20
4. Events in the Detector
The MiniBooNE Detector - 541 meters downstream
of target - 3 meter overburden - 12 meter
diameter sphere (10 meter fiducial
volume) - Filled with 800 t of pure mineral oil
(CH2) (Fiducial volume 450 t) - 1280 inner
phototubes, - 240 veto phototubes Simulated
with a GEANT3 Monte Carlo
Extinction rate of MiniBooNE oil
21
4. Events in the Detector
The MiniBooNE Detector - 541 meters downstream
of target - 3 meter overburden - 12 meter
diameter sphere (10 meter fiducial
volume) - Filled with 800 t of pure mineral oil
(CH2) (Fiducial volume 450 t) - 1280 inner
phototubes, - 240 veto phototubes Simulated
with a GEANT3 Monte Carlo
22
4. Events in the Detector
nm charged current quasi-elastic (nm CCQE)
interaction is the most abundant (40) and the
fundamental interaction in MiniBooNE detector
MiniBooNE detector (spherical Cherenkov detector)
muon like Cherenkov light and subsequent decayed
electron (Michel electron) like Cherenkov light
are the signal of CCQE event
Cherenkov 1
e
m
n-beam
12C
Cherenkov 2
n
p
(Scintillation)
23
4. Events in the Detector
19.2 ms beam trigger window with the 1.6 ms
spill Multiple hits within a 100 ns window form
subevents nm CCQE interactions (nn ? mp)
with characteristic two subevent structure
from stopped m?nmnee
24
4. Events in the Detector
Times of hit-clusters (subevents) Beam spill
(1.6ms) is clearly evident simple cuts eliminate
cosmic backgrounds Neutrino Candidate Cuts lt6
veto PMT hits Gets rid of muons gt200 tank PMT
hits Gets rid of Michels Only neutrinos are left!
Beam and Cosmic BG
25
4. Events in the Detector
Times of hit-clusters (subevents) Beam spill
(1.6ms) is clearly evident simple cuts eliminate
cosmic backgrounds Neutrino Candidate Cuts lt6
veto PMT hits Gets rid of muons gt200 tank PMT
hits Gets rid of Michels Only neutrinos are left!
Beam and Michels
26
4. Events in the Detector
Times of hit-clusters (subevents) Beam spill
(1.6ms) is clearly evident simple cuts eliminate
cosmic backgrounds Neutrino Candidate Cuts lt6
veto PMT hits Gets rid of muons gt200 tank PMT
hits Gets rid of Michels Only neutrinos are left!
Beam Only
27

• Muons
• Sharp, clear rings
• Long, straight tracks
• Electrons
• Scattered rings
• Multiple scattering
• Radiative processes
• Neutral Pions
• Double rings
• Decays to two photons

4. Events in the Detector
28
4. Events in the Detector

• Muons
• Sharp, clear rings
• Long, straight tracks
• Electrons
• Scattered rings
• Multiple scattering
• Radiative processes
• Neutral Pions
• Double rings
• Decays to two photons

29

• Muons
• Sharp, clear rings
• Long, straight tracks
• Electrons
• Scattered rings
• Multiple scattering
• Radiative processes
• Neutral Pions
• Double rings
• Decays to two photons

4. Events in the Detector
30
4. Events in the Detector

• Muons
• Sharp, clear rings
• Long, straight tracks
• Electrons
• Scattered rings
• Multiple scattering
• Radiative processes
• Neutral Pions
• Double rings???
• Decays to two photons???
• Looks like the electron (the biggest misID)

31
5. Cross section model
32
5. Cross section model
Predicted event rates before cuts (NUANCE Monte
Carlo)
Casper, Nucl.Phys.Proc.Suppl. 112 (2002) 161
33
5. Cross section model
CCQE (Charged Current Quasi-Elastic) - 39 of
total - Events are clean (few particles) -
Energy of the neutrino (EnQE) can be
reconstructed from - Scattering angle
cosq - Visible energy Evisible
34
5. Cross section model
The data-MC agreement in Q2 (4-momentum transfer)
distribution is not great We tuned nuclear
parameters in Relativistic Fermi Gas model
Smith and Moniz, Nucl.,Phys.,B43(1972)605
Q2 fits to MB nm CCQE data using the nuclear
parameters MAeff - effective axial mass
EloSF - Pauli Blocking parameter Relativistic
Fermi Gas Model with tuned parameters
describes nm CCQE data well (paper in preparation)
35
5. Cross section model
data-MC ratio is flat through entire kinematic
space made by CCQE interaction
36
6. Oscillation analysis
37
6. Blind analysis
The MiniBooNE signal is small but relatively easy
to isolate The data is described in
n-dimensional space
hit time
veto hits
energy
38
6. Blind analysis
The MiniBooNE signal is small but relatively easy
to isolate The data is described in
n-dimensional space
hit time
NC
veto hits
high energy
energy
The data is classified into "box". For boxes to
be "opened" to analysis they must be shown to
have a signal lt 1s. In the end, 99 of the data
were available (boxes need not to be exclusive
set)
39
6. Blind analysis

• Intrinsic ne ?ne sources
• m ? e ?nm ne (52)
• K ? p0 e ne (29)
• K0 ? p e ne (14)
• Other ( 5)

p ? m nm
K? m nm
Since MiniBooNE is blind analysis experiment, we
need to constraint intrinsic ne background
without measuring directly (1) m decay ne
background (2) K decay ne background
m ? e nm ne K? p e ne
ne/nm 0.5 Antineutrino content 6
40
6. Blind analysis
hit time
(1) measure nm flux from nmCCQE event to
constraint ne background from m decay nmCCQE is
one of the open boxes. Kinematics allows
connection to p flux, hence intrinsic ne
background from m decay is constraint.
NC
veto hits
high energy
energy
p ? m nm
En-Ep space
m ? e nm ne
41
6. Blind analysis
hit time
(2) measure high energy nm events to constraint
ne background from K decay At high energies,
above signal range nm and ne -like events are
largely due to kaon decay
NC
veto hits
high energy
energy
p ? m nm
example of open boxes - nmCCQE - high energy
event - CCp - NC elastics - NC po - NC
electron scattering - Michel electron etc....
K? m nm
signal range
n events Dominated by Kaon decay
K? p e ne
42
6. MiniBooNE oscillation analysis structure
Start with a GEANT4 flux prediction for the n
spectrum from p and K produced at the target
Predict n interactions using NUANCE neutrino
interaction generator Pass final state
particles to GEANT3 to model particle and light
propagation in the tank Starting with event
reconstruction, independent analyses form (1)
Track Based Likelihood (TBL) and (2) Boosted
Decision Tree (BDT) Develop particle ID/cuts to
separate signal from background Fit
reconstructed EnQE spectrum for oscillations
detector model
Simultaneous Fit to nm ne
Pre-Normalize to nm Fit ne
43
6. Track-Based Likelihood (TBL) analysis
This algorithm was found to have the better
sensitivity to nm?ne appearance. Therefore,
before unblinding, this was the algorithm chosen
for the primary result Fit event with
detailed, direct reconstruction of particle
tracks, and ratio of fit likelihoods to identify
particle Separating e from m positive (negative)
likelihood ratio favors electron (muon) hypothesis
44
6. Track-Based Likelihood (TBL) analysis
Separating e from po electron hypothesis is
tested in 2 dimensional likelihood ratio space
45
6. Track-Based Likelihood (TBL) analysis
TBL analysis summary - Oscillation analysis uses
475MeVltElt1250MeV
46
6. Boosted Decision Tree (BDT) analysis
Boosted Decision Trees - data learning method
(e.g., neural network,...) - 100 input
variables from point like model event
reconstruction - combined many weak trees (
1000 weak trees) to make strong "committee" -
Designed to classify signal and background Signal
oscillation ne CCQE events , Background
everything else (misID)
Output PID variables
47
6. Boosted Decision Tree (BDT) analysis
BDT analysis summary - Oscillation analysis uses
300MeVltElt1600MeV - PID cut is defined each EnQE
bin
PID cut
48
7. Systematic error analysis
49
7. Error analysis
We have two categories of backgrounds
nm mis-id
intrinsic ne
(TB analysis)
50
7. Error analysis
Handling uncertainties in the analyses
What we begin with...
... what we need
For a given source of uncertainty, Errors on a
wide range of parameters in the underlying model
For a given source of uncertainty, Errors in
bins of EnQE and information on the
correlations between bins
51
7. Error analysis
Handling uncertainties in the analyses
What we begin with...
... what we need
For a given source of uncertainty, Errors on a
wide range of parameters in the underlying model
For a given source of uncertainty, Errors in
bins of EnQE and information on the
correlations between bins
"multisim" nonlinear error propagation
Input error matrix keep the all correlation of
systematics
Output error matrix keep the all correlation of
EnQE bins
52
7. Multisim
Multi-simulation (Multisim) method many fake
experiments with different parameter set give the
variation of correlated systematic errors for
each independent error matrix total error matrix
is the sum of all independent error matrix
Roe et al., Nucl.,Phys.,B43(1972)605
53
7. Multisim
cross section parameter space
QE s norm
ex) cross section uncertainties
MAQE 6 Elosf
2 QE ? norm 10
correlated
MA
uncorrelated
Elo
Input cross section error matrix
cross section error for EnQE
repeat this exercise many times to create smooth
error matrix for EnQE
54
7. Multisim
cross section parameter space
QE s norm
ex) cross section uncertainties
MAQE 6 Elosf
2 QE ? norm 10
correlated
MA
uncorrelated
Elo
Input cross section error matrix
cross section error for EnQE
repeat this exercise many times to create smooth
error matrix for EnQE
55
7. Multisim
Output cross section error matrix for EnQE
cross section error for EnQE
Oscillation analysis use output error matrix for
c2 fit c2 (data - MC)T (Moutput)-1 (data - MC)
56
7. Multisim
ex) cross section uncertainties
MAQE 6 Elosf
2 QE ? norm 10 QE ? shape function
of E? ??e/?? QE ? function of E? NC ?0
rate function of ?0 mom MAcoh, coh
?????25 ? ? N??rate function of ? mom 7
BF EB, pF 9 MeV, 30 MeV ??s
10 MA1? 25 MAN?
40 DIS ? 25 etc...
determined from MiniBooNE ?? QE data
determined from MiniBooNE ?? NC ?0 data
determined from other experiments
57
7. Multisim
Total output error matrix Mtotal M(p
production) M(p- production)
M(K production) M(K0
production) M(beamline model)
M(cross section model) M(p0
yield) M(dirt model)
M(detector model) M(data stat)
Oscillation analysis c2 fit c2 (data - MC)T
(Mtotal)-1 (data - MC)
58
8. The MiniBooNE initial results
59
8. Box opening procedure

• After applying all analysis cuts
• 1. Blind c2 test for a set of diagnostic
  variables.
• 2. Open up the plots from step 1.
• 3. Blind c2 test for a fit to EnQE , without
  returning fit parameters.
• 4. Compare EnQE in data and Monte Carlo,
  returning the fit parameters.
• box opening (March 26, 2007)

60
8. Step 1. Blind c2 test
12 variables are tested for TBL 46 variables are
tested for BDT All analysis variables were
returned with good probability except... Track
Based Likelihood analysis c2 probability of
Evisible fit is 1 We change the energy cut
gt475MeV for oscillation fit
61
8. Step 2. Open the blind c2 test plots
Opening 12 plots for TBL and 46 plots for BDT
BDT
TBL (EnQEgt475 MeV)
MC contains fitted signal at unknown level
62
8. Step 3. Blind c2 test for a fit to EnQE
This is the c2 oscillation fit TBL (EnQEgt475
MeV) c2 Probability of fit 99 BDT analysis
c2 Probability of fit 52
Step 4. Open the box...
63
(No Transcript)
64
8. The MiniBooNE initial results
TBL analysis TBL show no sign of an excess in the
analysis region (where the LSND signal is
expected from 1 sterile neutrino
interpretation) Visible excess at low E
BDT analysis BDT has a good fit and no sign of an
excess, in fact the data is low relative to the
prediction Also sees an excess at low E, but
larger normalization error covers it
65
8. The MiniBooNE initial results
BDT analysis Counting Experiment 300ltEnQElt1600
MeV data 971 events expectation 1070 ?33
(stat) ? 225 (sys) events significance -0.38 s
TBL analysis Counting Experiment 475ltEnQElt1250
MeV data 380 events expectation 358 ?19
(stat) ? 35 (sys) events significance 0.55 s
66
8. The MiniBooNE initial results
The observed reconstructed energy distribution
is inconsistent with a nm?ne appearance-only
model
Energy-fit analysis solid TB dashed
BDT Independent analyses are in good agreement.
Excluded region
67
9. Low energy excess events
68
9. Excess at low energy region?
Our goals for this first analysis were - A
generic search for a ne excess in our nm beam, -
An analysis of the data within a nm?ne
appearance-only context Within the energy
range defined by this oscillation analysis, the
event rate is consistent with background.
The observed low energy deviation is under
investigation, including the possibility of more
exotic model - Lorentz violation - 32
sterile neutrino - extra dimension etc...
69
9. Lorentz violation model
Science American (Sept. 2004)
Lorentz and CPT violation is a predicted
phenomenon from Planck scale physics
Characteristic signals of Lorentz violation in
neutrino oscillation physics is... 1. sidereal
variation of oscillation data 2. anomalous energy
dependence 3. CPT violation etc...
vacuum Hamiltonian for fermion
am
70
9. Lorentz violation model
Science American (Sept. 2004)
Lorentz and CPT violation is a predicted
phenomenon from Planck scale physics
Characteristic signals of Lorentz violation in
neutrino oscillation physics is... 1. sidereal
variation of oscillation data 2. anomalous energy
dependence 3. CPT violation etc...
vacuum Hamiltonian for fermion
SSB
am
71
9. Lorentz violation model
Science American (Sept. 2004)
Lorentz and CPT violation is a predicted
phenomenon from Planck scale physics
Characteristic signals of Lorentz violation in
neutrino oscillation physics is... 1. sidereal
variation of oscillation data 2. anomalous energy
dependence 3. CPT violation etc...
vacuum Hamiltonian for fermion
SSB
am
72
9. Lorentz violation model
Under the active (particle) Lorentz
Transformation
73
9. Lorentz violation model
Under the active (particle) Lorentz
Transformation
by definition, "a" is insensitive to active
transformation
Lorentz violation can be seen in fixed coordinate
system Physics (ex. neutrino oscillation)
depends on the sidereal motion
74
9. Lorentz violation model
Under the passive (observer) Lorentz
Transformation
75
9. Lorentz violation model
Under the passive (observer) Lorentz
Transformation
76
9. Lorentz violation model
Under the passive (observer) Lorentz
Transformation
Lorentz violation cannot be seen by observers
motion (coordinate transformation is
unbroken) Physics doesn't depend on observers
motion
77
9. Lorentz violation model
Neutrino oscillation sidereal time dependence
Kostelecky and Mewes PRD70(2004)076002
Lorentz violation fit for LSND
LSND neutrino oscillation data is consistent with
no Lorentz violation (but cannot rule out)
TK and LSND collaboration, PRD72(2005)076004
78
9. Lorentz violation model
It is possible to construct the global model for
neutrino oscillation including Lorentz and CPT
violation
TK et al, PRD74(2006)105009
effective Hamiltonian
This simple model is consistent with all existing
neutrino oscillation data, including solar,
atmospheric, reactor, and LSND.
m2 1.04 ? 10-3 eV2 mass term a -2.4 ? 10-19
GeV Lorentz and CPT violation term c 3.4 ?
10-17 Lorentz violation term
79
9. Lorentz violation model
It is possible to construct the global model for
neutrino oscillation including Lorentz and CPT
violation
TK et al, PRD74(2006)105009
effective Hamiltonian
This simple model is consistent with all existing
neutrino oscillation data, including solar,
atmospheric, reactor, and LSND. It predicts
signal at low E region for MiniBooNE Fit with
Lorentz violation model is under study.
80
10. Future plans
81
10. Future plans
A paper is submitted to PRL. Many more papers
supporting this analysis will follow, in the
very near future nm CCQE production p0
production MiniBooNE-LSND-Karmen joint analysis
We are pursuing further analyses of the
neutrino data, including... an analysis which
combines TBL and BDT more exotic model for the
LSND effect MiniBooNE is presently taking data
in antineutrino mode. Further questions of
MiniBooNE are answered by SciBooNE including...
cross check of intrinsic ne
measurement reduce cross section
uncertainty reduce p flux
uncertainty etc...
What is SciBooNE?
82
10. SciBooNE
SciBar detector (used by K2K experiment) is
shipped from Japan. Goal of SciBooNE is to
measure the cross section around 0.8GeV, where
the most important energy scale for T2K
experiment.
MINOS, NuMI
K2K, NOvA
MiniBooNE, T2K, SciBooNE
83
10. SciBooNE
SciBar detector Extruded scintillators with WLS
fiber readout by multi-anode PMT
Extruded scintillator (15t)
3m
Multi-anode PMT (64 ch.)
3m
Wave-length shifting fiber
1.7m
84
10. SciBooNE
EC (electron catcher) lead with scintillation
fiber "Spaghetti" calorimeter to see electrons
EM calorimeter
3m
3m
MRD (Muon Range Detector) iron plates with X-Y
scintillator panels Measure the muon momentum up
to 1.2GeV
1.7m
85
10. SciBooNE
SciBar installation
86
10. SciBooNE
MRD installation
Stay tuned...
87
BooNE collaboration
University of Alabama Los
Alamos National Laboratory Bucknell University
Louisiana State
University University of Cincinnati
University of Michigan University of Colorado
Princeton University Columbia
University Saint Marys
University of Minnesota Embry Riddle University
Virginia Polytechnic Institute Fermi
National Accelerator Laboratory Western
Illinois University Indiana University
Yale University
Thank you for your attention!
88
10. Back up
89
1. Neutrino oscillation
Neutrino oscillation is a quantum interference
experiment (cf. double slit experiment).
n1
n2
If 2 Hamiltonian eigenstates, n1 and n2 have
different phase rotation, they cause quantum
interference. For double slit experiment, if
path n1 and path n2 have different length, they
have different phase rotations and it causes
interference.
90
1. Neutrino oscillation
Neutrino oscillation is a quantum interference
experiment (cf. double slit experiment).
Um1
n1
Ue1
nm
ne
n2
n1
n2
If 2 neutrino Hamiltonian eigenstates, n1and n2,
have different phase rotation, they cause quantum
interference. For massive neutrinos, if n1 is
heavier than n2, they have different group
velocities hence different phase rotation, thus
the superposition of those 2 wave packet no
longer makes same state as before
91
2. LSND experiment
In terms of the oscillation probability, P(nm-ne
)0.264 0.067 0.045
Under the 2 flavor massive neutrino oscillation
model, one can map into Dm2-sin22q space
(MS-diagram) This model allows comparison to
other experiments Karmen2 Bugey
92
3. Neutrino beam
4 ?1012 protons per 1.6 ms pulse delivered at up
to 5 Hz. 5.58?1020 POT (proton on target)
93
3. Neutrino beam
Modeling Production of Secondary Kaons K Data
from 10 - 24 GeV. Uses a Feynman
Scaling Parameterization. K0 data are also
parameterized.
In situ measurement of K from LMC agrees within
errors with parameterization
94
3. Stability of running
Full n Run
Observed and expected events per minute
95
4. Calibration source
Laser flask system
Muon tracker and scintillation cube system
96
(No Transcript)
97
5. Cross section model
Events producing pions CCp Easy to tag due to 3
subevents. Not a substantial background to the
oscillation analysis. NCp0 The p0 decays to 2
photons, which can look electron-like
mimicking the signal... lt1 of p0 contribute
to background.
(also decays to a single photon with 0.56
probability)
98
6. Precuts
data MC
Both Algorithms and all analyses presented
here share hit-level pre-cuts
Only 1 subevent
Veto hits lt 6
Tank hits gt 200
And a radius precut Rlt500 cm (where
reconstructed R is algorithm-dependent)
99
6. Track-Based Likelihood (TBL) analysis
This algorithm was found to have the better
sensitivity to nm?ne appearance. Therefore,
before unblinding, this was the algorithm chosen
for the primary result Fit event with
detailed, direct reconstruction of particle
tracks, and ratio of fit likelihoods to identify
particle Fit event under the different
hypotheses - muon like - electron like Fit is
characterized by 7 parameters Fit knows -
scintillation, Cherenkov light fraction - wave
length dependent of light propagation -
scattering, reemission, reflection, etc - PMT
efficiencies
q,f
E
t,x,y,z
light
100
6. Boosted Decision Tree (BDT) analysis

• Events are reconstructed with point-like model
• Construct a set of analysis variables
• (vertex, track length, time cluster, particle
  direction, event topology, energy, etc)

101
6. Boosted Decision Tree (BDT) analysis
A Decision Tree
Variable 1
(sequential series of cuts based on MC study)
(Nsignal/Nbkgd)
bkgd-like
signal-like
Variable 2
9755/23695
bkgd-like
sig-like
Variable 3
30,245/16,305
1906/11828
7849/11867
sig-like
bkgd-like
20455/3417
This tree is one of many possibilities...
9790/12888
etc.
102
6. Boosted Decision Tree (BDT) analysis
Boosted Decision Tree
- a kind of data learning method (e.g., neural
network,...) - training sample (MC simulation)
is used to train the code - combined many weak
classifiers ( 1000 weak trees) to make strong
"committee"
(k-2)th
(k-1)th
(k1)th
kth decision tree
Boosted Decision Trees
103
6. Boosted Decision Tree (BDT) analysis
Fake data sample
Example of classification problem
Two ways to use decision trees. 1) Multiple cuts
on X and Y in a big tree, 2) Many weak trees
(single-cut trees) combined
The goal of the classifier is to separate blue
(signal) and red (background) populations.
2) Many weak trees (single cut trees) only 4
trees shown
1) Development of a single decision tree
104
6. Boosted Decision Tree (BDT) analysis
Single decision tree
500 weak trees committee
Classified as signal
Classified as background
Boosting Algorithm has all the advantages of
single decision trees, and less suceptibility to
overtraining.
105
7. Error analysis
Use of low-signal/high background energy bins
In both analyses, high energy bins constrain ne
background
TB
up to 3000 MeV
signal range
BDT
In Boosted Decision Tree analysis Low energy bin
(200ltEnQElt300 MeV) constrains nm mis-ids p0,
D?Ng, dirt ...
signal
106
7. Error analysis
We constrain p0 production using data from our
detector
This reduces the error on predicted
mis-identified p0s
Because this constrains the D resonance rate, it
also constrains the rate of D?Ng
Reweighting improves agreement in other variables
107
7. Error analysis
Other Single Photon Sources
Neutral Current n N ? n N g
negligible
From Efrosinin, hep-ph/0609169, calculation
checked by Goldman, LANL
Charged Current
lt 6 events _at_ 95 CL
n N ? m N g
where the presence of the g leads to
mis-identification
Use events where the m is tagged by the michel
e-, study misidentification using BDT algorithm.
108
7. Error analysis
External Sources of Background
Dirt Events
n interactions outside of the detector Ndata/NMC
0.99 0.15
Event Type of Dirt after PID cuts
EnhancedBackgroundCuts
Cosmic Rays
Measured from out-of-beam data 2.1 0.5 events
109
7. Error analysis
Summary of predicted backgrounds for the final
MiniBooNE result (Track Based Analysis)
110
7. Multisim
How the constraints enter...
Two Approaches
TB Reweight MC prediction to match measured nm
result (accounting for systematic error
correlations)

• BDT include the correlations of nm to ne in the
  error matrix

Systematic (and statistical) uncertainties are
included in (Mij)-1
(i,j are bins of EnQE)
111
7. Multisim
Error Matrix Elements

• N is number of events passing cuts
• MC is standard monte carlo
• a represents a given multisim
• M is the total number of multisims
• i,j are EnQE bins

BDT
Total error matrix is sum from each source.
TB ne-only total error matrix BDT nm-ne total
error matrix
112
7. Multisim
For each error source, Multisims are generated
by 2 different methods (1) Event weight
multisim standard Monte Carlo is
reweighted. (2) Generated multisim hit-level
simulations are used.
1000 event weighted multisims for K production
70 generated multisims for detector Model
Number of events passing cuts in bin
500ltEnQElt600 MeV
113
8. the MiniBooNE initial results
Two points on interpreting our limit

• 1) There are various ways
• to present limits
• Single sided raster scan
• (historically used, presented here)
• Global scan
• Unified approach
• (most recent method)
• 2) This result must be
• folded into an
• LSND-Karmen
• joint analysis.
• We will present a full joint analysis soon.

Church, et al., PRD 66, 013001
114
8. the MiniBooNE initial results
MiniBooNE-LSND combined fit
MiniBooNE is incompatible with nm?ne appearance
only interpretation of LSND at 98 CL
115
9. Lorentz violation model
Science American (Sept. 2004)
In a field theory, Spontaneous Symmetry Breaking
(SSB) create nonzero expectation value for fields
(e.g., Higgs) in a true vacuum In string theory,
there is a possibility that some Lorentz tensor
fields get the nonzero vacuum expectation value
(Lorentz symmetry is spontaneously broken)
Kostelecky and Samuel PRD15(1989)683
vacuum Hamiltonian for fermion
SSB
am
116
9. Lorentz violation model
PRD74(2006)105009
Lorentz and CPT violating effective Hamiltonian
This model is consistent with all existing
neutrino oscillation data, including solar,
atmospheric, reactor, and LSND.
117
9. Lorentz violation model
PRD74(2006)105009
Lorentz and CPT violating effective Hamiltonian
This model is consistent with all existing
neutrino oscillation data, including solar,
atmospheric, reactor, and LSND.
118
9. Lorentz violation model
PRD74(2006)105009
Lorentz and CPT violating effective Hamiltonian
This model is consistent with all existing
neutrino oscillation data, including solar,
atmospheric, reactor, and LSND.
119
9. Lorentz violation model
Standard Model Extension (SME) self-consistent
low-energy effective theory with Lorentz and CPT
violation within conventional QM (minimum
extension of QFT for Particle Lorentz violation)
Colladay and Kostelecky ('97)
Modified Dirac Equation (MDE)
SME parameters
Direction dependence Hamiltonian with SME
parameters has direction dependent physics, so,
it is important to fix the coordinate system to
describe the effect (Lorentz violating physics is
consistent with any frames!)
120
9. Lorentz violation model
Kostelecky and Mewes ('03)
Effective Hamiltonian for neutrino oscillation
usual Hamiltonian (3X3)
Lorentz violating terms (3X3)
Short Baseline Approximation good approximation
for small oscillation probability experiment
(LSND)
sidereal frequency sidereal time
Sidereal variation is described with 5 coupled
SME parameters (combination of aL and cL )
121
9. Lorentz violation model
Goal
Fit the sidereal distribution of 186
candidate events (oscillation and background
signals) with the equation of sidereal depending
oscillation probability
coordinate dependent direction vector