Spallation Tagging Techniques in Super-Kamiokande Scott Locke - - PowerPoint PPT Presentation

Spallation Tagging Techniques in Super-Kamiokande

Scott Locke Super-Kamiokande Collaboration TAUP 2019 2019/09/13

Super Kamiokande Detector

• 50kton water Cherenkov
• Ultrapure Water
• 22.5 kton fiducial volume (32 kton inner

detector volume)

• 11, 129 20-inch PMTs (inner), 1885 of 8-inch

PMTs (outer)

• Phase Period inner PMTs coverage
• SK-I 1996-2001 11,146 40%
• SK-II 2002-2005 5,182 19%
• SK-III 2006-2008 11,129 40%
• SK-IV 2008-2018 (same as SK-III with new

electronics)

• SK-V 2019-2020
• SK-Gd 2020-
• ~400 kt-years exposure
• Detects neutrinos from many sources
• T2K far detector

41.4 m 39.3 m 1000 m

2019/09/13 Scott Locke - University of California, Irvine 2

Introduction to Muons in SK

• ~2 Hz muon rate
• Muons typically deposit a few GeV

in the detector

• Even though in many instances

muons behave as a MIP, it isn’t always the case

• Large amounts of energy can be

deposited, especially in the form

• f a shower
• These showers may cause

spallation

5.6 Mpe ≈ 1 TeV deposited within detector Picture of Control Room Event Display

2019/09/13 Scott Locke - University of California, Irvine 3

Spallation

• Spallation – Muons initiate showers of secondary

particles (e, n, γ, π , …) which capture on/break apart nuclei, creating unstable isotopes, and those isotopes decay feigning a desired signal

• Lifetime can be short (τ ~ O(.001s)) to relatively

long-lived (τ ~ O(10s))

• Troublesome when trying to perform specific analyses

2019/09/13 Scott Locke - University of California, Irvine 4

Daughter Particles Spallation Decay False Signature

• Y. Zhang et et al. (Super-Kamiokande Collaboration),Phys. Rev. D 93, 012004 (2016)

Why do we care?

• Inhibits us from lowering the threshold

for Diffuse Supernova Neutrino Background (DSNB) analysis and is a lingering background

• At ~1σ excess from last time analysis was

done

• Largest remaining background in the

solar neutrino spectrum

• Although roughly 90% is cut, it accrues

~20% deadtime as a result

• Background in Day/Night analysis
• Reducing the background mitigates this

background shaping

• In other words, pretty much in all the

continuous Low Energy analyses

• Not much is known in terms of the

actual physics of spallation production

• What is happening in those showers?

DNSB spallation efficiencies

Personal Work

Phys.Rev. D85 (2012) 052007 2019/09/13 Scott Locke - University of California, Irvine 5

Showering Muons

• The muon energy loss rate is:

dE/dx = α(E) + β(E)E

• α term corresponds to continuous ionization

energy loss

• β term corresponds to the radiative processes
• Usual two types of showers:
• EM and Hadronic showers
• Hadronic showers still have large EM component due

to π0 decay into γγ

• Currently use this light to try and identify

showering muons

• Calculate likelihoods to determine if spallation

2019/09/13 Scott Locke - University of California, Irvine 6

Downward going muons through center

• f the detector
• S. W. Li and J. F. Beacom, Phys. Rev. C 89, 045801(2014)

Improvements and Modifications

• Results presented are with respect to the solar spallation tagging
• Would need separate treatment for DSNB search, but basic principles still

apply

• Using Wideband Intelligent Trigger (WIT) system to tag hadronic

showers after muons, specifically look for neutrons

• Using multiple events to tag spallation
• Revisiting likelihoods set in SK-I, and updating
• New likelihoods?

2019/09/13 Scott Locke - University of California, Irvine 7

Defining Parameters

• Transverse distance (lt):
• Distance of closest approach of event to track
• In all plots, lt is plotted as lt2 for flat phase space
• Longitudinal distance (ln):
• Distance along track in reference to another point
• Taken as (xi – xavg) for neutrons
• Time difference (dt)
• Time from muon to candidate
• Time ordering dictates signal vs BG
• Muon before candidate → Signal
• Candidate before Muon → BG
• Multiplicity:
• Number of candidate events for a muon (neutrons or

spallation candidates)

• Residual Charge (resq):
• Excess light from muon, above minimum ionization
• Etot – (EMIP per cm)*(track length)
• cos(Θsun)
• Cos of the angle between reconstructed vertex

direction and path from the sun

5/9/2019 Scott Locke - University of California, Irvine 8

Muon Track lt (transverse distance) Neutron/Spallation Candidate x (distance along track)

Neutron Captures

• Neutrons from hadronic showers capture on H and emit a 2.2 MeV γ,

but detection efficiency is very low (~7 detected photons)

• Use WIT to look for these events
• Events are below 3.49 MeV Kinetic, which is the trigger threshold for SK
• WIT has lower trigger threshold, allowing for chance to tag neutron captures
• AFT trigger (DSNB search) disabled for muons to save CPU time
• Not the case for WIT
• Look for events within 500 μs of muon, and 5m of muon track
• Efficiency is still not great, WIT triggers on 11 hits above dark-noise and event

reconstruction is less reliable at this low of energy

9/11/2019 Scott Locke - Spallation Studies in Super Kamiokande 9

Neutrons after Muons

• When making an initial cut on event reconstruction goodness, definite cloud can be seen in lt vs ln
• Reminder: ln is (xi – xavg) for neutrons
• For a clear dt distribution, an additional cut on lt (<1.5m) was made
• Just for trying to have good fit to the neutron signal, reduce BG
• τ = 208.2 +/- 1.84
• A little over 1σ from AmBe measurement

2019/09/13 Scott Locke - University of California, Irvine 10

1 2 3-4 5-19 20-39 40+

Neutron Multiplicity

PRELIMINARY

Making a Cut

• Use MC to make a more intelligent quality cut
• n neutrons
• Simulate 2.2 MeV γ’s in SK, compare well to poorly

reconstructed events

• 2 + 1 cut on clouds
• 2+ events within 500μs/5m, 1+ events making

quality cuts

• Only parameterize cloud with “good” events
• Taking shape of neutron cloud into

consideration

• Shift coordinate for cloud, use muon direction as

z-axis, and project cloud back to track

• Cylindrical coordinates
• Cut for 60s (30s for 2 neutron showers)
• Cut on cloud size as function of multiplicity
• Have big cuts in short time
• 7.5m sphere for 200ms, 5m sphere for 2s
• Downside: No WIT data available for this

analysis before late Oct 2016

2019/09/13 Scott Locke - University of California, Irvine 11

Muon Centered

x y z (muon track)

PRELIMINARY

Multiple Spallation

• A little more novel approach
• Do not expect multiple solar neutrinos within the same area in a

short time frame

• Cut events within a small timeframe and small area of each other
• 4m and 60s
• Only use events that would make final sample, without spallation cut and

patlik cut, and above 5.49 MeV kinetic

• Patlik is more sensitive to spallation
• Can cut ~45% of spallation this way, with minimal deadtime
• Great, because it can be applied retroactively

2019/09/13 Scott Locke - University of California, Irvine 12

Revisit Likelihoods

• Normal solar spallation cut is done by

log likelihood in lt, residual charge, and dt

• Updating muon-fitter
• Former muon fitter struggled to fit high

energy muons, and had worse track correlation for spallation

• Refit for SK-IV data and the change in

fitter

• Different fitter will have different

distribution, cover any changes since early days of SK-I

• Big difference in resq likelihood
• Change lt → lt2
• More intuitive with flat phase space
• Find there to be a dt dependence in lt2
• lt2 likelihood binned in time and charge
• dt has minimal difference
• Fit for 7 major spallation isotopes

2019/09/13 Scott Locke - University of California, Irvine 13

New Old Sig BG 0-100 ms 100ms-3s 3s-30s

Fraction

Frequency

New Old PRELIMINARY

Bringing it All Together

• For time with WIT available, applying neutron cloud,

multiple spallation, updated likelihood:

• Maintain efficiency, minimize deadtime
• Less than half the deadtime!
• For time without WIT, apply multiple spallation and

updated likelihood (change likelihood cut value)

• ~250 days added exposure for SK-IV alone
• Plots shown for >5.99 MeV kinetic
• Solid line is signal, dashed is background (backwards dt)

2019/09/13 Scott Locke - University of California, Irvine 14

Method Efficiency (%) Deadtime (%) OLD 89.97 19.61 Cloud 54.34 1.18 Multiple 46.48 1.27 Spal-like 81.08 7.70 Cl + Mu + Like 90.05 9.65

Remaining Events Old New Spallation Effectiveness During WIT Period All

OLD

NEW

Spalike only Cloud only Mult only OLD Accidentals Cloud Accidentals Spalike Accidentals

All SK-IV OLD NEW

Updated spalike and multi cut

PRELIMINARY

SK-IV Updated Peak (New spallation Cut)

2019/09/13 Scott Locke - University of California, Irvine 15

9.6% More events 5.3% Relative Statistical Error reduction PERSONAL WORK

Moving Forward

• There are a couple more small

investigations to complete:

• Trying to use dE/dx (charge

deposition along track) and ln in likelihood

• Using muon classifications in

likelihoods

• lt2 and multiple tracks
• In the future, effect of SK-Gd
• n neutron tagging efficiency
• Efficiency should be greatly

increased

• However, PMT after pulsing may

cause issues

2019/09/13 Scott Locke - University of California, Irvine 16

~50% capture efficiency

.01% 10t Gd3(SO4)2 in SK .1% 100t Gd3(SO4)2 in SK .001%

~90% capture efficiency 100% 80% 60% 40% 20%

Capture Efficiency on Gd 1% Dissolve Gadolinium into SK J. Beacom and M. Vagins, PRL93, 171101 (2004)

PERSONAL WORK

Summary

• Successfully tagged hadronic showers after a muon
• These showers can successfully spallation, with minimal deadtime
• However, neutron tagging efficiency too low to use by itself
• Multiple spallation, simple yet effective
• Can be retroactively applied to all SK-IV and earlier phases
• Revisited current spallation likelihoods, with small, intuitive improvements
• For when neutron data is available, more than factor 2 reduction in

deadtime, while maintaining spallation tagging efficiency

• Looking to the future:
• Introduction of SK-Gd will improve neutron tagging efficiency, while also introducing

its own issues

• Sets groundwork for future experiments needing a more stringent spallation cut (e.g.

Hyper-K)

2019/09/13 Scott Locke - University of California, Irvine 17

Thank You!

2019/09/13 Scott Locke - University of California, Irvine 18

Backup

2019/09/13 Scott Locke - University of California, Irvine 19

ln by multiplicity

2019/09/13 Scott Locke - University of California, Irvine 20

# of Neutrons All > 1 2 3 4 5 6

• Different high purity cuts
• Not the ones used for final

cloud tagging

• For the use of getting much

better signal to noise for dt, ln, and lt investigation

2019/09/13 Scott Locke - University of California, Irvine 21

No Additional Cuts goodness > 0.5 lt < 1.5m lt and goodness cuts

Spallation in Neutron Clouds

• All are background subtracted
• dt is fit for N16, and had τ ~ 10.5s +/- 0.14 (N16 τ = 10.3)
• Vertex correlation taking center of neutron cloud for events passing

goodness cuts

2019/09/13 Scott Locke - University of California, Irvine 22

Case II

• Maintain deadtime, maximize

efficiency

• Same size peak, lower plateau of BG

2019/09/13 Scott Locke - University of California, Irvine 23

Old Case II Method Efficiency (%) Deadtime (%)

Old

89.97 19.70

CL+MU+Like (Case II)

93.84 19.71