Double Chooz Sensitivity Analysis: Impact of the Reactor ...8/1/14 Outline • Neutrino Mixing: A...

37
Double Chooz Sensitivity Analysis: Impact of the Reactor Model Uncertainty on Measurement of Sin 2 (2θ 13 ) Elizabeth Grace 1 8/1/14

Transcript of Double Chooz Sensitivity Analysis: Impact of the Reactor ...8/1/14 Outline • Neutrino Mixing: A...

  • Double Chooz Sensitivity Analysis: Impact of the Reactor Model

    Uncertainty on Measurement of Sin2(2θ13)Elizabeth Grace

    1 8/1/14

  • 8/1/14

    Outline• Neutrino Mixing: A Brief Overview

    • Goal of Double Chooz

    • Detector Schematic

    • Fit Framework and Statistical Analysis

    • My Work: Sensitivity Study of Reactor Uncertainty Impact

    2

  • 8/1/14

    Neutrino Flavor Mixing

    • P(νe→νe) = 1 - sin2(2θ13) sin2(1.27 ∆m2 L/E)

    • Function of ∆m2 and L/E, which varies and causes oscillation

    • θ13 is“mixing angle” and determines amplitude

    • Standard Model prediction: ∆m2 = 0, no mixing

    3

  • 8/1/14

    “Mixing Angle”

    • Why do we use sin2 (2θ13) instead of some constant? In other words, what angle does θ13 refer to?

    • Coordinate transformation from axes of states of definite mass to definite flavor described by rotation matrix

    • Mass eigenstate ≠ flavor eigenstate

    θν1

    ν2νe-νµ

    Two-neutrino approximation

    P(νe→νe) = 1 - sin2(2θ13) sin2(1.27 ∆m2 L/E)

    4

  • 8/1/14

    Goal of Double Chooz• Measure sin2 (2θ13) using anti-νe- source: nuclear reactors

    • Measure anti-νe- flux as a function of L/E

    • Neutrino flux goes like L-2 , neutrino energy peaks at around 3 MeV

    • Deficit in anti-νe- indicates oscillation

    5

  • 8/1/14

    Detector Schematic

    6

  • 8/1/14

    Signal: IBD

    prompt

    delayed

    νe- + p n + e+

    7

    _

  • 8/1/14

    Target • Innermost cylinder:

    target containing 0.1% Gd + liquid scintillator

    • Gd has high neutron cross section for absorption: 
Gd + n —> γγ characteristic gammas detected

    8

  • 8/1/14

    • Second cylinder is “gamma catcher” containing liquid scintillator: allows gammas escaping target to deposit some of their energy in scintillator

    • >95% gamma containment

    γ Catcher

    9

  • 8/1/14

    • Photomultiplier Tubes: Turn a light signal into an electron signal via photoelectric effect and amplification

    • PMTs are surrounded by B field shielders to cancel earth’s magnetic field so the e- do not bend

    • 390 PMTs

    • Surrounded by buffer oil for radiation (PMTs, external sources)

    Buffer

    10

  • 8/1/14

    Inner µ Veto

    • Contains 70 tons of liquid scintillator

    • 78 PMTs

    • Purpose is to detect muons in order to veto events caused by them

    11

  • 8/1/14

    Outer µ Veto• Similar purpose to inner

    veto: signal rejection

    • Plastic scintillator strips

    • 2 modules: X and Y

    • 2 layers per module: offset like bricks to pinpoint muon location

    12

  • 8/1/14

    Background (Noise)• Uncorrelated: Accidentals • Correlated: Fast n and stopping µ,

    9Li and 8He • 0.5 < Ee+ < 10 MeV • 6 < En < 10 MeV • 1 μs < Δt < 100 μs

    13

    W+

    νe-_

    p n

    e+prompt

    delayed

    IBD signal

  • 8/1/14

    Motivation: My Analysis

    • Have been using a single detector for 2-3 years in Chooz experiment

    • Obtained results with uncertainties

    14

  • 8/1/14

    Motivation: My Analysis• Problem: Inexact prediction of

    reactor flux (initial neutrino flux)

    • Solution: Replace prediction with measurement

    • New, identical detector closer to reactor source (Near Detector)

    15

    • Want to know: What will the resultant uncertainty in the measurement of sin2(2θ13) be?

    • Answer: Conduct sensitivity study in order to predict it

  • 8/1/14

    Overview: χ2• Main component of sensitivity study

    • Measurement of how expectations compare to results, i.e. goodness of a fit

    • For random, independent, Gaussian variables, the sum of their squares (i.e, chi-squared) follows a well-known distribution

    • Energy bin contents are such variables and thus follow the distribution

    16

  • 8/1/14

    Overview: χ2• Binning scheme: discretize the data

    into 250 keV-wide energy bins

    • Pull parameter (nuisance parameter): Term that is not of interest but must be accounted for, so it floats in the fit

    • Pull term: penalty term that changes as you attempt to optimize chi-squared

    • Covariance matrix: a matrix containing information on how the error in the i

    th

    and jth

    elements change together

    17

  • 8/1/14

    Building χ2

    18

    + …

    Sums the difference between predicted and observed over B, the number of bins

    Covariance matrix: contains information about the way

    that the error in bin i is related to the error in bin j

    Nipred is the number of predicted events in energy bin i. Niobs is the number of observed events in energy bin i.

  • 8/1/14

    Building χ2: ∆m2

    19

    + …

    This term factors in any error in our measurement of ∆m2 by adding a penalty when ∆m2≠∆mee2 and scaling it by the uncertainty

    ∆m2 is the parameter we allow to float in the DC fit

    uncertaintyMeasured value of ∆mee2

    (derived from MINOS result) used to constrain our data

  • 8/1/14

    Building χ2

    20

    Each α is a pull term for the relevant background error: 9Li and 8He; fast neutrons and stopping muons; and systematic accidentals.

    We subtract 1 because the value of α

    would be 1 if there were no deviation from the prediction.

    + …Nipred = ∑Niν +αiLi+He NiLi+He+ αiFN+SM NiFN+SM+…

    where Niν = Niν (sin2(2θ13), ∆m2, …)

    Far Detector

    Near Detector

  • 8/1/14

    Covariance Matrix• In our model, there are a lot of uncertain parameters. Within the set of possibilities contained

    within our uncertainties, what happens to the bin contents with respect to one another, i.e., how do they correlate?

    • ∆i∆j = matrix element of Mij where ∆i is the way that the predicted bin contents would be different if a parameter in our prediction were changed

    • Ex. How would an incorrect neutrino flux prediction affect the contents of energy bin i?

    • Off-diagonal elements describe the covariance, i.e., the amount that the ith bin’s error correlates with the jth bin’s error

    • Ex. After changing neutrino flux prediction, if bin i changed in the positive direction and bin j would change in the negative direction, then the bins are anti correlated

    • Pull terms can often be written in the form of a covariance matrix: α is Nobs

    , 1 is Npred

    , and Mij-1

    contains σ

    2

    !

    21

  • 8/1/14

    Covariance Matrices• Mij is a sum of other covariance matrices:

    Mij = Mijstat

    + Mijreactor

    + Mijefficiency

    + MijLi+He shape

    + Mijacc(stat)

    • Mijstat

    : contains info about the statistical spread around the

    prediction, so there is no covariance (i.e., it is a diagonal matrix)

    • Mijreactor

    : contains uncertainty in reactor prediction

    • Mijefficiency

    : normalization uncertainties (translates spectrum up or down)

    • MijLi+He shape

    : shape error in Li+He background spectrum

    • Mijacc(stat)

    : diagonal matrix containing statistical info about accidentals

    22

  • 8/1/14

    Sensitivity Studies• Create data with a known

    sin2(2θ13) and plot ∆χ2(sin2(2θ13),α, ∆m

    2,…) where ∆χ2 = χ2 - χ2min

    • ∆χ2 will be zero at the known sin2(2θ13) value

    • Go up by 1 on the ∆χ2 axis to find the locations of ±1σ

    • Use this to estimate the future ±1σ for the real data

    23

    Minuit

  • 8/1/14

    Sensitivity Studies• Want to know: What will the uncertainty in the

    measurement of sin2(2θ13) be when we use near detector instead of reactor prediction?

    • To find out: scale reactor error matrix by a multiplicative coefficient and plot ±1σ error in sin2(2θ13)

    • Considered FD only and ND+FD cases

    • Plot the ±1σ error in sin2(2θ13) as a function of time for several different scaling factors

    24

  • 8/1/14

    Results

    25

    One detectorTwo detectors

    50% error10% error

    no added error100% error

    Reactor Flux Assumptions

  • 8/1/14

    Results (Far Detector Only)

    26

    Reactor Flux Assumptions50% error

    10% errorno added error

    100% error

  • 8/1/14

    Conclusions

    • Second detector adds a high level of certainty in the measurement of sin2(2θ13) even with little to no knowledge of the reactor flux

    • Expected result: was motivation behind adding a second detector

    27

  • 8/1/14

    AcknowledgementsMike Shaevitz

    Rachel Carr

    Leslie Camilleri

    Georgia Karagiorgi

    John Parsons and the NSF

    28

  • 8/1/14

    Backup

    29

  • 8/1/14

    Assumptions• Inputs for far detector are based on the most recent

    DoubleChooz analysis (DCIII), and near detector are based on DCIII knowledge (conservative prediction)

    • Signal from detectors scaled for live time and baseline using DCIII

    • Background Rates and Uncertainty

    • Energy Scale Uncertainty

    30

  • 8/1/14

    InputsBackground

    TypeFD Rate (per

    day)ND Rate (per

    day)Spectrum

    ShapeShape

    uncertainty

    9 0.97 ± 0.29 4.37 ± 1.29 DC-IIIDC-III (100%

    correlated between rxtrs)

    FN+SM 0.60 ± 0.052.00

    (different from DC-III value: 3.18 ± 0.27)

    flat none

    Accidentals 0.0701 ± 0.0054 0.21 ± 0.02 DC-IIInone (factored into rate unc)

    31

  • 8/1/14

    InputsParameter Uncertainty (same for ND and FD)

    ND-FD Correlation Coefficient

    Detection Efficiency 0.62% on signal normalization 0.94

    Energy Scale 1.00% on linear energy scale 0

    Reactor Flux 1.73% normalization + shape1 (different from DC-III

    value of 0.99)

    ∆m -4.1%, 3.69% on ∆m N/A (single parameter)

    32

  • 8/1/14

    Accidentals• Accidental coincidences

    • Example:

    • Gamma enters the detector

    • Cosmic muon hits a nearby nucleus and releases a neutron, which is detected

    • Not linked temporally

    33

  • 8/1/14

    Fast Neutrons• Fast neutron: high energy

    neutron

    • Cosmic µ hits a nearby nucleus and ejects spallation neutron

    • Fast neutron enters target:

    • causes a proton recoil, which mimics e+ signal

    • is captured, producing a gamma signal

    34

  • 8/1/14

    Stopping Muons• Cosmic muon enters the detector from

    above

    • Prompt signal from muon track (ionization from scintillator): muons produce scintillation light that is in the positron energy range

    • Delayed signal from decay of muon into a Michel electron

    • ∆t is the muon lifetime (distribution with mean within the allowed signal range)

    • Don’t apply cuts because we want to keep the signal

    35

  • 8/1/14

    9Li + 8He• Cosmic muon hits the target

    and produces Li-9 or He-8

    • Result is electron (mimics e+) and neutron

    • Should be able to eliminate with µ veto, but their half life is so long that we would miss data

    36

  • 8/1/14

    9Li + 8He, cont.• 9Li is 85% of this type of

    background

    • Half life is 113 ms (8He) and 178 ms (9Li)

    • Vetoing all the data after seeing a muon for the amount of time it takes these to decay would result in continuous veto

    • Must deal with this in analysisDecay chain for 9Li

    37