Neutron-Antineutron Oscillation in
Liquid Argon Time Projection Chambers
Jeremy Hewes, for the DUNE collaboration
IOP HEPP & APP conference
10th April 2017
1
Neutron-antineutron oscillation
Antiproton annihilation in bubble chamber
(Phys. Rev. 101.909)
•
Baryon number violating process, ΔB = 2.
•
n-n̄ transition bound inside nucleus.
•
Immediate annihilation with another nucleon in nucleus.
•
Motivations: Baryogenesis, neutrino mass.
•
Current lifetime limit by Super-Kamiokande (see next slide)
⇡
⇡
+
⇡
0
⇡
⇡
+
Jeremy Hewes - n-n̄ in LArTPCs
2
Super-Kamiokande measurement
•
•
•
Water Cherenkov detector with 22.5kt fiducial mass.
•
Expected 24.1 background events, observed 24.
•
Signal selection efficiency 12.1%
•
Can DUNE, with larger FV & tracking info, do better?
32
Lifetime limit 1.9x10 yrs at 90% CL
for bound neutron.
8
Equivalent to 2.7x10 s at 90% CL for
free neutron.
Jeremy Hewes - n-n̄ in LArTPCs
3
Deep Underground Neutrino Experiment
•
Very large LArTPC detector with four 10kt
modules — under construction.
•
1.5km underground at SURF in Lead, SD.
•
Primary physics goals: δCP, mass ordering.
•
Additional goals: supernova neutrinos,
nucleon decay.
60m
15m
Jeremy Hewes - n-n̄ in LArTPCs
4
Benefits of underground LArTPC
•
LArTPC technology
provides mm-level spatial
resolution.
Sense Wires
U V Y
•
•
•
Liquid Argon TPC
Excellent dE/dx resolution.
Deep underground
detector to protect from
cosmogenic
backgrounds.
Primary background
expected to be
atmospheric neutrinos.
V wire plane waveforms
Charged Particles
Cathode
Plane
g
in
I
om
nc
ut
Ne
o
rin
Edrift
Y wire plane waveforms
t
5
Simulation of n-n̄ oscillation in argon nucleus
GENIE simulation
n n̄ → π+ π- 3π0
•
Event generator released in GENIE v2.12.0.
•
Branching ratios from p̄ annihilation data.
•
Simulation of nuclear effects:
•
Fermi momentum, binding energy, FSI.
Simulated event display
in LArTPC
Jeremy Hewes - n-n̄ in LArTPCs
6
Signature in LArTPCs
•
Expect ~2GeV energy, ~300MeV momentum
over entire event from initial annihilation.
•
During FSI, ~50% of visible energy absorbed
by nucleus, net momentum smeared out.
Jeremy Hewes - n-n̄ in LArTPCs
7
Convolutional neural networks
•
Investigating convolutional neural
nets for event selection.
•
Building on existing CNN studies in
LArTPCs — see arXiv:1611.05531
n-n̄ MC image
Atmospheric ν MC image
Jeremy Hewes - n-n̄ in LArTPCs
8
Convolutional neural networks
WORK IN PROGRESS
•
Train network with signal &
background MC images.
•
Test network with separate samples.
•
CNN assigns each image a score.
WORK IN PROGRESS
•
Count how many signal & background
events pass a selection cut on this score.
•
Propagate efficiency & background rate to
sensitivity calculation (360 kt yr exposure).
Jeremy Hewes - n-n̄ in LArTPCs
9
DUNE sensitivity
•
•
n-n̄ oscillation sensitivity
depends on exposure,
selection efficiency
and background rate.
Calculate sensitivity to
free n-n̄ oscillation as a
function of efficiency and
background rate, for
exposure of 360 kt yrs.
P ( |nobs ) = A
Z
Z Z Z
5 x SK limit
WORK IN PROGRESS
e
90%
0
P ( |nobs )d = 0.9
(
3 x SK limit
✏+b)
( ✏ + b)nobs
P ( )P (✏)P (b)d d✏db
nobs !
Γ = Oscillation width
nobs = No. events observed
A = Normalisation constant
Jeremy Hewes - n-n̄ in LArTPCs
λ = Exposure
ε = Selection efficiency
b = Background rate
10
Summary
•
Studying prospects for n-n̄ oscillation in DUNE.
•
GENIE event generator developed & released.
•
Studying topology in LArTPCs using MC.
•
Investigating use of convolutional neural networks to
separate signal from background.
•
Early studies hint at potential for DUNE to improve on
current limits by sizeable margin.
Jeremy Hewes - n-n̄ in LArTPCs
11