Zelimir Djurcic Physics Department Columbia University Status of MiniBooNE Status of...
-
Upload
asher-douglas -
Category
Documents
-
view
217 -
download
3
Transcript of Zelimir Djurcic Physics Department Columbia University Status of MiniBooNE Status of...
Zelimir DjurcicZelimir Djurcic
Physics DepartmentPhysics Department
Columbia UniversityColumbia University
Status of MiniBooNEStatus of MiniBooNEExperimentExperiment
WIN07, CalcuttaWIN07, Calcutta
January 15-20,2007January 15-20,2007
MiniBooNE consists of about 70 scientists from 16 institutions.
Y. Liu, D.Perevalov, I. Stancu Alabama S. Koutsoliotas Bucknell R.A. Johnson, J.L. Raaf Cincinnati T. Hart, R.H. Nelson, M.Tzanov, E.D. Zimmerman, M.Wilking Colorado A. Aguilar-Arevalo, L.Bugel, L. Coney, J.M. Conrad, Z. Djurcic, J. Monroe, K. Mahn, D. Schmitz, M.H. Shaevitz, M. Sorel, G.P. Zeller Columbia D. Smith Embry Riddle L.Bartoszek, C. Bhat, S J. Brice, B.C. Brown, D.A. Finley, R. Ford, F.G.Garcia, P. Kasper, T. Kobilarcik, I. Kourbanis, A. Malensek, W. Marsh, P. Martin, F. Mills, C. Moore, E. Prebys, A.D. Russell, P. Spentzouris, R. Stefanski, T. Williams Fermilab D. C. Cox, A. Green, T.Katori, H.-O. Meyer, C.
Polly, R. Tayloe Indiana G.T. Garvey, C. Green, W.C. Louis, G.McGregor, S.McKenney, G.B. Mills, H. Ray, V. Sandberg, B. Sapp, R. Schirato, R. Van de Water, D.H. White Los Alamos R. Imlay, W. Metcalf, S. Ouedraogo, M. Sung, M.O. Wascko Louisiana State J. Cao, Y. Liu, B.P. Roe, H. Yang Michigan A.O. Bazarko, P.D. Meyers, R.B. Patterson, F.C. Shoemaker, H.A.Tanaka Princeton A. Currioni, B.T. Fleming Yale P. Nienaber St. Mary’s U. of Minnesota E. Hawker Western Illinois U. J.Link Virginia State U.
MiniBooNE MiniBooNE CollaboratioCollaboratio
nn
Zelimir Djurcic-WIN2007
Before MiniBooNEBefore MiniBooNE
Zelimir Djurcic-WIN2007
LSND took data from 1993-98 - 49,000 Coulombs of protons - L = 30m and 20 < E< 53 MeV
Saw an excess ofe :87.9 ± 22.4 ± 6.0 events.
With an oscillation probability of (0.264 ± 0.067 ± 0.045)%.
3.8 significance for excess.
ee
eOscillations?
Before MiniBooNE: The LSND Before MiniBooNE: The LSND ExperimentExperiment
Signal: p e+ n
n p d (2.2MeV)
e
Zelimir Djurcic-WIN2007
Kamioka, IMB, Super K, Soudan II, Macro, K2Km2 = 2.510-3 eV2
Homestake, Sage, Gallex, Super-KSNO, KamLAND m2 = 8.210-5 eV2
This signal looks very differentfrom the others...• Much higher m2 = 0.1 – 10 eV2 • Much smaller mixing angle• Only one experiment!
Current Oscillation Current Oscillation StatusStatus
In SM there are only 3 neutrinos
m13
m12
m23
2 2 2
2 2 221 32 31
Three distinct neutrino oscillation signals,
with
For three neutrinos,
expect
solar atm LSNDm m m
m m m
Zelimir Djurcic-WIN2007
• Want the same L/E• Want higher statistics• Want different systematics• Want different signal signature and
backgrounds
Fit to oscillation hypothesis
Backgrounds
Confirming or Refuting Confirming or Refuting LSNDLSND
Need definitive study of e at high m2 … MiniBooNE
Zelimir Djurcic-WIN2007
MiniBooNEMiniBooNE
((BooBoosterster N Neutrinoeutrino E Experiment)xperiment)
Zelimir Djurcic-WIN2007
magnetic horn: meson focusing
decay region: , K
absorber: stops undecayed mesons
“little muon counters:” measure K flux
in-situ
→
e?
50 m decay pipe
magnetic focusing horn
FNAL 8 GeV Beamline
Search for Search for ee appearance in appearance in beambeam
e e ??????
Use protons from the 8 GeV booster Neutrino Beam <E>~ 1 GeV
MiniBooNE MiniBooNE Detector:Detector:12m diameter 12m diameter spheresphere950000 liters of 950000 liters of oil oil (CH2)1280 inner PMTs1280 inner PMTs240 veto PMTs240 veto PMTs
Zelimir Djurcic-WIN2007
Few words on:Few words on:-Neutrino Flux-Neutrino Flux-Cross-section-Cross-section
-Detector Modeling-Detector Modeling
• mainly from • <E> ~ 700 MeV
predicted flux
intrinsic e
• ~10-3
• + e+ e
• K+ 0 e+ e (also KL)
Flux at MiniBooNE Flux at MiniBooNE DetectorDetectorFlux simulation uses Geant4 Monte Carlo
Meson production is based on Sanford-Wang parameterization of p-Be interaction cross-section.
p
• E910: , K production @ 6, 12, 18 GeV
w/thin Be target• HARP: , K production @ 8 GeV w/ 5, 50, 100% thick Be target
MINOS, NuMIK2K, NOvAMiniBooNE, T2K
Super-K atmospheric
Predictions from NUANCE
- MC which MiniBooNE uses - open source code - supported & maintained by D. Casper (UC Irvine)
- standard inputs
- Smith-Moniz Fermi Gas - Rein-Sehgal 1 - Bodek-Yang DIS
Low Energy Low Energy Cross Cross SectionsSections
Imperative is to preciselypredict signal & bkgd ratesfor future oscillation experiments
We need data on nuclear targets!(most past data on H2, D2)Current cross-section studies devoted to understanding Current cross-section studies devoted to understanding
of the backgrounds in the MiniBooNE appearance signal.of the backgrounds in the MiniBooNE appearance signal.
Zelimir Djurcic-WIN2007
Neutrino Interactions in the Neutrino Interactions in the DetectorDetector
e n e- pWe are looking for We are looking for e e
::Current Collected data:700k neutrino candidates (before analysis cuts) for 7 x 1020 protons on target (p.o.t.)
If LSND is correct, we If LSND is correct, we expect several hundred expect several hundred ee (after analysis cuts) from (after analysis cuts) from for for e e oscillations.oscillations.
- 48% QE- 31% CC +
- 1% NC elastic- 8% NC 0
- 5% CC 0
- 4% NC +/-
- 4% multi-
NUANCE MC generator NUANCE MC generator converts the flux into converts the flux into event rates in event rates in MiniBooNE detectorMiniBooNE detector
Zelimir Djurcic-WIN2007
Detector ModelingDetector Modeling
Detector (optical) model defines how light of generated event is propagated and detected in MiniBooNE detector
Sources of light: Cerenkov (prompt, directional cone),and scintillation+fluorescence of oil (delayed, isotropic)
Propagation of light: absorption, scattering (Rayleigh and Raman) and reflection at walls, PMT faces, etc.
Strategy to verify model:External Measurements: emission, absorption of oil, PMT properties.Calibration samples: Laser flasks, Michel electrons, NC elastic events.Validation samples: Cosmic muons (tracker and cubes).
Michel electrons from decay: provide E calibration at low energy (52.8 MeV), good monitor of light transmission, electron PID
0 mass peak: energy scale & resolutionat medium energy (135 MeV), reconstruction
We have calibration sources spanning wide range of energies and all event types !
12% E res at
52.8 MeV
Energy CalibrationEnergy Calibration
cosmic ray + tracker + cubes: energy scale & resolution at high energy (100-800 MeV), cross-checks track reconstruction
provides tracks of known length → E
e
PRELIMINARYPRELIMINARY
Zelimir Djurcic-WIN2007
How to Detect and ReconstructHow to Detect and ReconstructNeutrino EventsNeutrino Events
Zelimir Djurcic-WIN2007
Main trigger is an accelerator signal indicating a beam spill.Information is read out in 19.2 s interval covering arrival of beam and requests of various triggers (laser, random strobe, cosmic…).
Detector OperationDetector Operation
The rate of neutrinocandidates was constant:1.089 7 x 10-15 /P.O.T.
Zelimir Djurcic-WIN2007
Detector Operation and Event Detector Operation and Event reconstructionreconstruction
To reconstruct an event:-Separate hits in beam window by time into sub-events of related hits
-Reconstruction package maximizes likelihood of observed charge and time distribution of PMT hits to find track position, direction and energy (from the charge in the cone) for each sub-event
No high level analysis needed to see neutrino events
Backgrounds: cosmic muons and decay electrons
->Simple cuts reduce non-beam backgrounds to ~10-3
Electronics continuously record charges and times of PMT hits.
Zelimir Djurcic-WIN2007
0 →
Michel e-
candidate
beam candidate
beam 0
candidate
Čerenkov rings provide primary means of identifying products of interactions in the detector
n - p
e n e- p
p p 0
n n
Particle IdentificationParticle Identification
Particle Identification IIParticle Identification II
Search for oscillation
e n e- p
events is by detection of single electron like-rings, based on Čerenkov ring profile.
muon
Angular distributions of PMT hits relative to track direction:
electron
PRELIMINARY
PRELIMINARY
Signal Separation from Signal Separation from BackgroundBackground
Reducible
NC 0 (1 or 2 e-like rings)
N decay (1 e-like ring)
Single ring events
Irreducible
Intrinsic e events in
beam from K/ decay
0→Signal N
Search for O(102) e oscillation events in O(105) unoscillated events
Backgrounds
Zelimir Djurcic-WIN2007
Two complementary approaches
for reducible background
“Simple” cuts+Likelihood: easy to understand
Boosted decision trees: maximize sensitivity
Background Rejection and Blind Background Rejection and Blind AnalysisAnalysis
MiniBooNE is performing a blind analysis:
We do not look into the data region where the oscillation candidates
are expected (“closed box”).We are allowed to use:
– Some of the info in all of the data– All of the info in some of the data(But NOT all of the info in all of the data)
Zelimir Djurcic-WIN2007
Boosted decision trees: • Go through all PID variables and find best
variable and value to split events.• For each of the two subsets repeat the process• Proceeding in this way a tree is built.
• Ending nodes are called leaves.• After the tree is built, additional trees are built with the leaves re-weighted.• The process is repeated until best S/B separation is achieved.• PID output is a sum of event scores
from all trees (score=1 for S leaf, -1 for B
leaf).
Boosting PID AlgorithmBoosting PID Algorithm
Boosted Decision Trees at MiniBooNE:Use about 200 input variables to train the trees-target specific backgrounds-target all backgrounds generically
Boosting Decision Tree
Muons Electrons
PRELIMINARYPRELIMINARY
Reference NIM A 543 (2005) 577.Reference NIM A 543 (2005) 577.
Zelimir Djurcic-WIN2007
Likelihood ApproachLikelihood Approach
Apply likelihood fits to three Apply likelihood fits to three hypotheses:hypotheses:-single electron track-single electron track-single muon track-single muon track-two electron-like rings (-two electron-like rings (0 0 event event hypothesis )hypothesis )Form likelihood differences using minimized –logL
quantities: log(Le/L) and log(Le/L)
Compare observed light distribution to fit prediction:
Does the track actually look like an electron?
log(Le/L)
log(Le/L)<0 -like events
log(Le/L)>0 e-like events
PRELIMINARYPRELIMINARY
Zelimir Djurcic-WIN2007
CCQE and CCQE and 00 Analysis Analysis
MiniBooNE Quasi-Elastic MiniBooNE Quasi-Elastic DataData
epne pn
12C-
beam
ell
llQE
PEM
mMEE
cos
2
2
1 2
l
CCQE events are used because one can use CCQE kinematics to reconstruct the neutrino energy – one can look at neutrino energy spectraWe are looking for an oscillation signal in an EQE distribution of electron eventsOne can use an EQE distribution of muon events to understand our models
measure visible E and from mostly Čerenkov () + some scintillation light (p)
90% purity sampleMain bkgd: CC+
(+ absorbed)
p
nScintillation
Cerenkov 1
12C Cerenkov 2
e
Compare data to the Smith Moniz model implemented
in NUANCE for
CCQE events
n → - p
Zelimir Djurcic-WIN2007
Deficit is seen in the data for low values of the momentum transfer, Q2
Similar effects have been seen in otherchannels and by other experiments
Given the Fermi gas model approximation used one can imagine deficiencies – particularly in the low Q2 (very forward) kinematic region
Use data sample to adjust available parameters in present model to reproduce data: only – e differences are due to lepton mass effects, vs. e
With the high statistics and resolutions attainable at MiniBooNE, the MiniBooNE data will be used in the future to carefully study this and other models of CCQE interactions
MiniBooNE Quasi-Elastic MiniBooNE Quasi-Elastic DataData
0 0 ’s Background ’s Background DeterminationDetermination
e appearance: 0 production important because background to →e
if ’s highly asymmetric in energy or small opening angle (overlapping rings) can appear much like primary electronemerging from a e QE interaction!
0→Signal N
Reconstruction of Reconstruction of ππ00 results in results in excellent Data/MC agreement.excellent Data/MC agreement.We use Data to reweight (i.e. We use Data to reweight (i.e. tune) NUANCE rate prediction as tune) NUANCE rate prediction as a function of a function of ππ00 momentum. momentum.
PRELIMINARYPRELIMINARY
We measure rate of We measure rate of ππ00 in the data sample out of the oscillation in the data sample out of the oscillation regionregionand extrapolate it into the oscillation region.and extrapolate it into the oscillation region.
The reconstructed The reconstructed γγγγ mass distribution is mass distribution is divided into 9 momentum divided into 9 momentum bins.bins.
MC is used to unsmear MC is used to unsmear the data:the data:1.1. In bins of true In bins of true
momentum vs. momentum vs. reconstructed reconstructed momentum, count MC momentum, count MC events, over BG, in the events, over BG, in the signal window.signal window.
2.2. Divide by the total Divide by the total number of number of ππ00 events events generated in that true generated in that true momentum bin.momentum bin.
3.3. Invert the matrix.Invert the matrix.
4.4. Perform a BG subtraction Perform a BG subtraction on the data in each on the data in each reconstructed reconstructed momentum bins.momentum bins.
5.5. Multiply the data vector Multiply the data vector by the MC unsmearing by the MC unsmearing MatrixMatrix
Monte Carlo Events Passing Analysis Cuts
All eventsEvents with no
π0
Data Un-smearing and efficiency Data Un-smearing and efficiency correctioncorrection
Zelimir Djurcic-WIN2007
The Corrected Data DistributionThe Corrected Data Distribution
The corrected The corrected ππ00 momentum distribution is softer momentum distribution is softer than the default Monte Carlo. The normalization than the default Monte Carlo. The normalization discrepancy is across all interaction channels in discrepancy is across all interaction channels in MiniBooNE.MiniBooNE.
From this distribution From this distribution we derive a reweighting we derive a reweighting function for Monte Carlo function for Monte Carlo events. events.
Ratio of Ratio of datadata and and MCMC points points scaled to equal numbers of scaled to equal numbers of events.events.
MC: Generated MC: Generated distribution.distribution.
Data: Corrected to true Data: Corrected to true momentum andmomentum and 100% efficiency.100% efficiency.
Zelimir Djurcic-WIN2007
Reweighting improves Reweighting improves data/MC agreement.data/MC agreement.
The plots are:The plots are:
• Decay opening angleDecay opening angle
• Energy of high energy Energy of high energy γγ
• Energy of low energy Energy of low energy γγ
• ππ angle wrt the beam angle wrt the beam
The disagreement cos The disagreement cos θθππ may be due to coherent may be due to coherent ππ00 production which we fit production which we fit for.for.
Reweighting MC to DataReweighting MC to Data
Zelimir Djurcic-WIN2007
The Resulting The Resulting ππ00 MisID MisID DistributionDistribution
The resulting misID The resulting misID distribution is softer in distribution is softer in EEνν QE.QE.
Also there are less Also there are less misID events per misID events per produced produced ππ00 than in the than in the default Monte Carlo.default Monte Carlo.
The error on misID yield The error on misID yield is well below the 10% is well below the 10% target.target.
This is not the final PID cut set!
PRELIMINARYPRELIMINARY
Zelimir Djurcic-WIN2007
Cross-ChecksCross-Checks
Important Cross-Important Cross-check… check…
… comes from NuMI events detected in MiniBooNE detector!
MiniBooNE
Decay Pipe
Beam Absorber
We get e , , 0 , +/- , ,etc. eventsfrom NuMI in MiniBooNE detector, allmixed together Use them to Use them to
checkcheck
our our ee reconstruction and PID reconstruction and PID separation!separation!
Remember that MiniBooNE
conducts a blind data
analysis! We do not look in
MiniBooNE data region
where the osc. e are
expected…
The beam at MiniBooNE from NuMI is significantly enhanced in e from K decay because of the off-axis position.
NuMI events cover whole energy region relevant to e osc. analysis at MiniBooNE.
Example of use of the events from NuMI Example of use of the events from NuMI beam beam
Boosted Decision Tree
Likelihood Ratios
e/
e/
PRELIMIN
ARY
PRELIMIN
ARY
PRELIMIN
ARY
PRELIMIN
ARY
Data/MC agree through background and signal regions
PRELIMIN
ARY
PRELIMIN
ARY
Zelimir Djurcic-WIN2007
Appearance Signal Appearance Signal and Backgroundsand Backgrounds
Zelimir Djurcic-WIN2007
Arb
itra
ry U
nit
s
Oscillation e
Example (fake) oscillation signal
– m2 = 1 eV2
– sin22 = 0.004
Fit for excess as function of reconstructed e
energy
Appearance Signal and Appearance Signal and BackgroundsBackgrounds
PRELIMIN
ARY
PRELIMIN
ARY
Zelimir Djurcic-WIN2007
Arb
itra
ry U
nit
sAppearance Signal and Appearance Signal and
BackgroundsBackgroundsMisID • of these…… • ~83% 0
– Only ~1% of 0s are misIDed
– Determined by clean 0 measurement
• ~7% decay – Use clean 0
measurement to estimate production
• ~10% other– Use CCQE
rate to normalize and MC for shape
PRELIMIN
ARY
PRELIMIN
ARY
Zelimir Djurcic-WIN2007
Arb
itra
ry U
nit
sAppearance Signal and Appearance Signal and
BackgroundsBackgroundse from +
• Measured with CCQE sample– Same parent + kinematics
• Most important low E background
• Very highly constrained (a few percent)
PRELIMIN
ARY
PRELIMIN
ARY
p+Be+
e
+
e+
Zelimir Djurcic-WIN2007
Arb
itra
ry U
nit
sAppearance Signal and Appearance Signal and
BackgroundsBackgrounds
e from K+
• Use High energy e and to normalize
• Use kaon production data for shape
PRELIMIN
ARY
PRELIMIN
ARY
Zelimir Djurcic-WIN2007
Arb
itra
ry U
nit
sAppearance Signal and Appearance Signal and
BackgroundsBackgrounds
High energy e
data• Events below
1.5 GeV still in closed box (blind analysis)
PRELIMIN
ARY
PRELIMIN
ARY
Zelimir Djurcic-WIN2007
Combined Fit Combined Fit (Example)(Example)
Combined fit Combined fit constrains constrains uncertainties uncertainties common to common to and e
2==I,J(OI-PI)(CIJ)-1(OJ-PJ)
Systematic error matrixSystematic error matrixCIJ includes estimated includes estimated
systematic uncertaintiessystematic uncertainties
CIJ = e
e e
PRELIMIN
ARY
PRELIMIN
ARY
PI =PI (sin2(2),m2)
Scan Scan sin2(2)e,m2,with
sin2(2)x=0; calculate2 value over ande
bins:bins:
Zelimir Djurcic-WIN2007
Reconstructed visible Reconstructed visible muon energy (left) muon energy (left) muon neutrino energy muon neutrino energy (right) using CCQE (right) using CCQE data.data.
Error bands show both statistical Error bands show both statistical and systematic errorsand systematic errors
Evaluating Evaluating SystematicsSystematics
PRELIMIN
ARY
PRELIMIN
ARY
Zelimir Djurcic-WIN2007
LSND best fit sin22 = 0.003 m2 = 1.2 eV2
MiniBooNE Oscillation MiniBooNE Oscillation SensitivitySensitivity
MiniBooNE aims to cover LSND region.
We are currently finalizing work on systematicerror (i.e. error matrix)that combines the error sources(flux, or measured rate, detector
modeling) of signal and the background componentsto predict sensitivity to oscillation signal
Zelimir Djurcic-WIN2007
Total accumulated dataset 7.5 x 1020 POT, world’s largest dataset in this energy range.
Jan 2006: Started running with antineutrinos.
Detected NuMI neutrinos – using in analysis.
Oscillation Analysis progress: we are preparing
to open the closed “oscillation box”.
SummarySummary
Zelimir Djurcic-WIN2007
Backup SlidesBackup Slides
Zelimir Djurcic-WIN2007
– Sterile Neutrinos • RH neutrinos that don’t interact (Weak == LH only)
– CPT Violation• 3 neutrino model, manti-
2 > m2
• Run in neutrino, anti-neutrino mode, compare measured oscillation probability
– Mass Varying Neutrinos• Mass of neutrinos depends on medium through which it
travels
– Lorentz Violation• Oscillations depend on direction of propagation• Oscillations explained by small Lorentz violation• Don’t need to introduce neutrino mass for oscillations!• Look for sidereal variations in oscillation probability
Explaining the LSND Explaining the LSND resultresult
Zelimir Djurcic-WIN2007
World p+Be MeasurementsWorld p+Be Measurements
• E910: , K production @ 6, 12, 18 GeV w/thin Be target
• HARP: , K production @ 8 GeV w/ 5, 50, 100% thick Be target
kinematic boundary of HARP measurement at exactly 8.9 GeV/c
● black boxes are the distribution of + which decay to a that passes through the MiniBooNE detector
HARP ResultsHARP Results
HARP (CERN)Data taken with MiniBooNE target slugs using 8 GeV beamResults on thin target just added (Apr06).
Further improvement in flux prediction expected soon with HARP thick target and K data
Zelimir Djurcic-WIN2007
Exclusive channels are handled separately and use differing, appropriate models
Total cross-sections are then the sum of all relevant exclusive channels
Nuclear effects of hadrons propagating through the nucleus are considered to give you an expected final state condition
The most critical exclusive channel for the MiniBooNE oscillation search is the charged-current quasi-elastic interaction
NUANCE models CCQE events using the relativistic Fermi gas model of Smith and Moniz as a framework
The next most critical exclusive channels are the NC production of NC 0's
NUANCE uses the resonant and coherent p0 production models of Rein and Sehgal
About NUANCEAbout NUANCE
Zelimir Djurcic-WIN2007
• NuMI ’s sprayed in all directions.• K and decays at off-axis
angle:
p beam , K
22
,
2
2,
2
1
1
K
K
EM
m
E
• Opportunity to check the /K ratio of yields off the target.
~110mrad to MiniBooNE
NuMI events at NuMI events at MiniBooNE MiniBooNE
Zelimir Djurcic-WIN2007
Production of the Production of the 0 0 ’s’s
Resonant 0 production
N N=(p,n)
0 N’
Coherent 0 production
A A 0 0→
In addition to its primary decay In addition to its primary decay NN, , the the resonance has a branching resonance has a branching fraction fraction of 0.56% to of 0.56% to NN final state. final state.
e appearance: 0 production important because background to →e
if ’s highly asymmetric in energy or small opening angle (overlapping rings) can appear much like primary electronemerging from a e QE interaction!
0→Signal N
Zelimir Djurcic-WIN2007
ReweightedReweighted
UnweighteUnweightedd
The fit coherent fraction The fit coherent fraction is higher after is higher after reweighting.reweighting.
This was expected based This was expected based on the additional peaking on the additional peaking in the reweighted cos in the reweighted cos θθ distribution.distribution.
The reweighted fit does The reweighted fit does much better in the much better in the important forward region.important forward region.
Fit coherent, resonant, and background components to the Fit coherent, resonant, and background components to the datadata
Coherent Fit EffectCoherent Fit Effect