Status and Recent Results of the Acoustic Neutrino ... · AMADEUS too small, “2D-geometry”...

Robert Lahmann ARENA 2014�Annapolis, 09-June-2014

Status and Recent Results of the Acoustic Neutrino Detection Test System AMADEUS of ANTARES

2

Outline

●  Introduction: Acoustic Neutrino Detection and AMADEUS ●  Ambient Noise and Transient Background Investigations ●  Lessons Learned ●  Conclusions and Outlook

ARENA 2014, Annapolis - Robert Lahmann - 9-June-2014 2

3

Acoustic Detection of Neutrinos


-0.08-0.06-0.04-0.02

0 0.02 0.04 0.06 0.08

0.1 0.12

-0.04 -0.02 0 0.02 0.04

Pres

sure

[Pa]

Time [ms]

1011GeV @ 1000m

Hadronic cascade: ~10m length, few cm radius Pressure field: Characteristic “pancake” pattern Long attenuation length (~5 km @ 10 kHz)

Thermo-acoustic effect: (Askariyan 1979) energy deposition ð local heating (~µK) ð expansion ð pressure signal

ν ~1o

Time [ms]

Pres

sure

[Pa]

Bipolar Pressure Signal (BIP)

The AMADEUS System of the ANTARES Detector


AMADEUS : •  Total of 6 “acoustic storeys” •  Total of 36 hydrophones •  Continuous sampling •  Online filter selects ~1% of data volume for storage

ANTARES site: •  2500m depth, 30km offshore

ANTARES site

5

Operation of AMADEUS

●  Main objective: feasibility study for a potential future large-scale acoustic neutrino detector ●  Investigate background conditions ●  Determine energy threshold for neutrino detection ●  Devise high efficiency, high purity neutrino detection algorithms

●  Data from first line with acoustic sensors: Dec 2007 Data from two lines: Nov. 2009 – Oct. 2010 Since April 2013 (new position of IL)


6

ANTARES: New Geometry since April 2013


“Instrumentation Line” was redeployed at new position:

Distance between lines with acoustic storeys

220m 150m

7

Background for Acoustic Detection in the Sea


Transient background

Bipolar Pressure Signals (BIPs)

ðDetermines fake neutrino rate Suppress by � clustering

� signal classification � fiducial volume cuts

ðDetermines intrinsic energy threshold Use “effective volume” for estimate Depends on “sea state” (surface agitation) (see talk by Dominik Kiessling)

Ambient noise

Frequency [ kHz ]5 10 15 20 25 30 35 40 45 50

/Hz

]2

Paµ

Pow

er le

vel [

dB

re 1

0

10

20

30

40

50

60

70Sea state 0

Sea state 2

Sea state 4

Model ss0

Model ss2

Model ss4

Measured noise spectra and model

Signal (a.u.), schematic

Transient Background: Properties


•  Very diverse Shipping traffic, marine mammals, … ð perform signal classification

•  Mostly originating from near surface ð “straight forward” approach: Impose cut based on source location

~500

m

“qu

iet z

one”

em

issi

ons,

re

flect

ions

Transient Background: Position Reconstruction


•  For events selected by online filter, reconstruct direction for individual storeys

•  When directions reconstructed by more than one storey get source location

Data: 156 days of measuring time from Nov. 2009 to Oct. 2010

Source Localization


Problem: Small size of AMADEUS device ð large errors in z, despite good angular

resolution for direction reconstruction: in zenith

in azimuth Solution: Project positions to sea surface and remove event clusters from moving sound emitters

Δθ = 0.6± 0.2

Δϕ =1.6± 0.2

Cluster Analysis of Moving Sound Emitting Objects


x [m]

y [m

]

Nov. 2009

Oct. 2010

Signal Classification with Machine Learning Algorithms

●  Classification: neutrino candidate (BIP) background

●  Different algorithms have been investigated: ●  Random Forest ●  Boosted Trees ●  Naïve Bayes ●  Decision Tree ●  Support Vector Machine

●  Recognition Error: ●  For individual sensors < 10% ●  For clusters of sensors < 2%

best performance

12 ARENA 2014, Annapolis - Robert Lahmann - 9-June-2014

Spatial Distribution of Transient Background


Dep

th z

[m]

Distance r [m]

-2500-2000-1500-1000

-500 0

500 1000

0 2000 4000 6000 8000 10000 1e-08

1e-07

1e-06

1e-05

0.0001

0.001

Even

t den

sity

[m-3

]

All clustered events

Dep

th z

[m]

Distance r [m]

-2500-2000-1500-1000

-500 0

500 1000

0 2000 4000 6000 8000 10000 1e-08

1e-07

1e-06

1e-05

0.0001

0.001

Even

t den

sity

[m-3

]

All reconstructed events 15·103 km−3 year−1

After signal classification and cluster analysis 100 km−3 year−1

14

Search for a Fiducial Volume - Motivation

●  Using signal classification and cluster analysis for the identification of neutrino-like bipolar signals

●  Remaining events density: ~100 events/km3/yr ●  Need for further reduction ð cut on the volume


Dep

th z

[m]

Distance r [m]

-2500-2000-1500-1000

-500 0

500 1000

0 2000 4000 6000 8000 10000 1e-08

1e-07

1e-06

1e-05

0.0001

0.001

Even

t den

sity

[m-3

]

15

z [m

]

Distance r [m]

-2500-2000-1500-1000

-500 0

500 1000

0 2000 4000 6000 8000 10000 12000 1e-50 1e-45 1e-40 1e-35 1e-30 1e-25 1e-20 1e-15 1e-10 1e-05 1

Even

t den

sity

[m-3

]

Search for a Fiducial Volume - PSF

Optimize fiducial volume for minimal background content: ●  Point spread function calculated from MC Simulations ●  Deconvolution of the PSF using Richardson-Lucy-Algorithm


Preliminary

16

Search for a Fiducial Volume - Cut Strategy

●  Optimization problem: ●  Minimal number of events and ●  Maximal remaining volume after applying the cut

●  Using a Genetic Algorithm to solve the optimization problem ●  Remaining event density after cut: ~0.05 events/km3/yr

(but volume closest to sensors is removed)


z [m

]

Distance r [m]

-2500-2000-1500-1000

-500 0

500 1000

0 2000 4000 6000 8000 10000 12000 1e-50 1e-45 1e-40 1e-35 1e-30 1e-25 1e-20 1e-15 1e-10 1e-05

Even

t den

sity

[m-3

]

Preliminary

Further Reduction of Transient Background

Search for characteristic geometry of pressure field from neutrino interaction (“pancake”) ●  AMADEUS too small, “2D-geometry” ●  investigations with Monte Carlo simulations (input from AMADEUS) ●  KM3NeT: Combined system for acoustic positioning and neutrino

detection planned ð test bed for algorithm development

17

See talk by Dominik Kiessling

ARENA 2014, Annapolis - Robert Lahmann - 9-June-2014

18

Effective Volume


Ve↵ =

Pp(E,x, ep)�sel

NgenVgen

Effective Volume

Number of Neutrinos: 107

Volume in which the Neutrinos are generated: 1200km3

Probability of the neutrino reaching the vertex

only counted if signal is detected;

Earth

ν

Water

10°

Earth density model (PREM)

19

AMADEUS Effective Volume


Level 1:

•  low ambient noise •  minimal filter

Level 2: •  noise model (annual distr.) •  std. filter

10-5

10-4

10-3

10-2

10-1

100

101

9 9.5 10 10.5 11 11.5 12

V eff(

E) [k

m3 ]

E [log(E/Gev)]

Preliminary

Veff(Ei) Level 1Random Coincidences Level 1Veff(Ei) Level 2Random Coincidences Level 2

Preliminary

AMADEUS: Lessons Learned

●  Ambient background: GZK neutrinos (for pure proton flux) detectable, reduction of threshold crucial ð bigger detector, use signals from more sensors

●  Transient noise: High level of background (mainly dolphins); High level of reduction already achieved with AMADEUS, for competitive flux limits recognition of “acoustic pancake” crucial

●  Road ahead: Apply knowledge about ambient noise and transient background data to simulations: ●  KM3NeT acoustic system for positioning/neutrino detection ●  large scale fiber-based acoustic neutrino telescope?

(see talk on behalf of E.J. Buis)


Conclusions and Outlook

•  Ambient noise: Smaller effect on neutrino detection than assumed

•  Transient background: Strong suppression achieved, further reduction requires larger detectors

•  Monte Carlo simulations developed and energy threshold derived from effective volume of AMADEUS

•  Next step KM3NeT: Combined system for acoustic positioning and neutrino detection planned

21 ARENA 2014, Annapolis - Robert Lahmann - 9-June-2014

Thank you for your attention


Status and Recent Results of the Acoustic Neutrino ... · AMADEUS too small, “2D-geometry”...

Documents

Transcript of Status and Recent Results of the Acoustic Neutrino ... · AMADEUS too small, “2D-geometry”...