Detecting neutrinos from Gamma-Ray Bursts with IceCube

Garmt de Vries-Uiterweerd

Universiteit Utrecht

Outline

Neutrino astronomy
Amanda/IceCube
From waveforms to hits
From hits to tracks
From tracks to GRB detection
Conclusions

Neutrino astronomy: Motivation

Perfect messengers

Protons and photons: absorption, deflection, CMBR
Neutrinos: Can cover huge distances, point back to source

Useful probe

Detection (or non-detection!) of neutrinos gives information on source
Provides information on particle physics (e.g. oscillations)
Expect the unexpected

Neutrino production in Gamma-Ray Bursts

Particle acceleration in relativistic jets

Gamma-Ray Bursts

Extremely powerful explosions (~ 10⁵¹ erg)
Extragalactic, isotropically distributed
~ 100 yr^-1
Timescale: ms to min

Generic model

Collapsing star or merger of compact objects
Accretion disc, powerful jets
Relativistic shocks, particle acceleration
Electromagnetic: Synchrotron radiation, GeV photons
Hadronic: PeV photons, neutrinos

Detection principle

Weak interaction → need large detector
Indirect detection of neutrino via interaction products
Interaction types:
- Neutral current
- Charged current ν_e
- Charged current ν_μ
- Charged current ν_τ

Detection principle

Neutral current

Hadronic shower
Neutrino escapes unnoticed

Detection principle

Charged current ν_e

Hadronic shower
Electromagnetic shower from electron

Detection principle

Charged current ν_μ

Hadronic shower
Čerenkov cone from energetic muon

Detection principle

Charged current ν_τ

Hadronic shower
Second hadronic shower from τ decay

Current and future neutrino telescopes

Antarctica

Amanda
IceCube

Mediterranean

Antares
Nemo
Nestor
KM3Net

Other

Baikal
Super-Kamiokande

Amanda and IceCube

Amanda

19 strings, 1500–2000 m
677 Optical Modules
10–20 m between OMs, 55–75 m between strings
Analog signal transmission

IceCube

80 strings, 1450–2450 m
4800 Digital Optical Modules
17 m between DOMs, 125 m between strings
Signal digitised in DOM and sent to surface

From waveforms to hits

Transient Waveform Recorder (TWR)

In Amanda, signal arrives at surface as wide pulse: waveform
Multiple pulses possible
Goal: analyse each pulse individually
- Leading edge (LE)
- Time over threshold (TOT)
- Charge

From waveforms to hits

Simplest case: single peak

Determine peak location
Determine point of steepest rise
Extrapolate tangent to baseline → LE
Determine where signal drops below baseline again → TOT
Integrate signal → charge

From waveforms to hits

Slightly less simple: two peaks

Two separate peaks with extracted features

Determine individual peak locations
Determine range for each peak (range boundary: minimum between the peaks)
Determine LE, TOT and charge for each peak individually

From waveforms to hits

Challenge: overlapping peaks

Two overlapping peaks, well distinguished

Determine LE peak 2 with new baseline: 0.5×(minimum between peaks)
Adjust charge integration ranges
Extra charge from other peak compensates for missing charge outside integration region

End of charge integration for peak 1: upper end of peak range, or baseline crossing, whichever comes first
Start of charge integration for peak 2: lower end of peak range, or LE, whichever comes last

From waveforms to hits

Too hard: peaks overlap too much

Two overlapping peaks, ill distinguished

Peaks overlap too much → analysis for peak 2 problematic
If LE for peak 2 too far before lower end of peak range, merge peaks

From waveforms to hits

Very simple: optical channels

Optical channels: narrow pulses
Very simple peak finding: signal above given threshold → peak

From waveforms to hits

Analog Transient Waveform Digitizer (ATWD)

Digitised in DOM → very narrow pulses
Hit extraction very easy

From hits to tracks

First guess method: direct walk

Create Track Elements (TE) from pairs of hits: minimum distance, maximum time difference
Find associated hits: hits that are close enough to TE
Select Track Candidates (TC): TEs that fulfill certain quality criteria
Combine TCs into jets: maximum opening angle between TCs in a jet
Merge jets
Order merged jets by quality factor → tracks

From hits to tracks

Full reconstruction

Log-likelihood maximisation
First guess solutions used as seed
Resolution: ≲ 1° (IceCube) to few degrees (Amanda)

From tracks to GRB detection

Background

Atmospheric μ (100000s per hour)
Atmospheric ν (10s per hour)

GRB signal

At Earth, low ν fluence
only 1–10% of GRBs will induce one signal ν

→ Need method to combine data from many GRBs

Signal rate: Halzen & Hooper 1999, Astrophys. J. 527 (1999) L93

Stacking

Take all data in 2 hour time window around GRB trigger
For each event, determine time with respect to trigger
Put all events from all GRBs into single histogram
Background events distributed uniformly
Signal events cluster

Reducing background

Cut on opening angle with respect to GRB position
Only look at GRBs in Northern hemisphere
For observed stacked time profile: plausibility of background-only hypothesis
Use Bayesian statistics to assess significance

Background only

Background + signal

Assessing significance

Bayes’ theorem

p(B|A) = p(B) \frac{p(A|B)}{p(A)}

Evidence

Data D, hypothesis H, unspecified alternative hypothesis H'

\frac{p(H|D)}{p(H'|D)} = \frac{p(H)}{p(H')} \frac{p(D|H)}{p(D|H')}

e(H|D) ≡ 10 log

₁₀ p(H|D) p(H'|D)

e(H|D) = e(H) + 10 log

₁₀ p(D|H) p(D|H')

Assessing significance

Introduce ψ

ψ ≡ −10 log

₁₀ p(D|H)

ψ' ≡ −10 log

₁₀ p(D|H') Since 0 ≤ p(D|H) ≤ 1:

e(H'|D) = e(H) + ψ − ψ' ≤ e(H) + ψ

→ No alternative H' can be supported by data more than ψ

→ ψ quantifies degree of belief in H, given the data D

Assessing significance

Applying ψ analysis to stacked time profile

D = {n_i}, number of events in each bin i

H: uniform distribution

n: number of events

m: number of time bins

p(D) = \frac{n!}{n}

₁! ⋯ n_m! m⁻¹

ψ = −10 [ log

₁₀n! − ∑ log₁₀ n_k! − log₁₀m ∑ n_k]

ψ distributions

Even in case of uniform background, observed time profile will probably not be exactly uniform
Which ψ values are compatible with H?
Distribution of possible ψ values depends only on n and m

ψ distributions

20 background events; 2, 5, 10, 20, 50, 100 signal events

ψ distributions

50 background events; 2, 5, 10, 20, 50, 100 signal events

ψ distributions

100 background events; 2, 5, 10, 20, 50, 100 signal events

ψ distributions

200 background events; 2, 5, 10, 20, 50, 100 signal events

ψ distributions

500 background events; 2, 5, 10, 20, 50, 100 signal events

ψ distributions

1000 background events; 2, 5, 10, 20, 50, 100 signal events

Significance

Observe number of events n
Determine ψ distribution for n background events
Determine ψ_obs value of observed time profile
Determine fraction f of cases with ψ > ψ_obs
f quantifies significance of observation

Optimisation

Apply cuts (position, quality factors, etc.)
Goal: minimise number of background events, keep maximum number of signal events
For all possible cut settings:
- Determine number of observed events from data
- Determine corresponding ψ distribution
- Determine number of signal events in one bin needed for 5σ significance
- Determine flux for which 5σ is reached
Choose cut settings for which needed flux is minimum

(This is currently work in progress)

Conclusions

Neutrino astronomy is an important and interesting field
IceCube is ready for serious science
Time stacking is a promising method for GRB detection
Last details to be worked out, analysis to be performed soon

Detecting neutrinos from Gamma-Ray Bursts with IceCube

Garmt de Vries-Uiterweerd

Universiteit Utrecht

Outline

Neutrino astronomy: Motivation

Perfect messengers

Useful probe

Neutrino production in Gamma-Ray Bursts

Gamma-Ray Bursts

Generic model

Detection principle

Detection principle

Neutral current

Detection principle

Charged current νe

Detection principle

Charged current νμ

Detection principle

Charged current ντ

Current and future neutrino telescopes

Antarctica

Mediterranean

Other

Amanda and IceCube

Amanda

IceCube

From waveforms to hits

Transient Waveform Recorder (TWR)

From waveforms to hits

Simplest case: single peak

From waveforms to hits

Slightly less simple: two peaks

From waveforms to hits

Challenge: overlapping peaks

From waveforms to hits

Too hard: peaks overlap too much

From waveforms to hits

Very simple: optical channels

From waveforms to hits

Analog Transient Waveform Digitizer (ATWD)

From hits to tracks

First guess method: direct walk

From hits to tracks

Full reconstruction

From tracks to GRB detection

Background

GRB signal

Stacking

Reducing background

Background only

Background + signal

Assessing significance

Bayes’ theorem

Evidence

Assessing significance

Introduce ψ

Assessing significance

Applying ψ analysis to stacked time profile

ψ distributions

ψ distributions

20 background events; 2, 5, 10, 20, 50, 100 signal events

ψ distributions

50 background events; 2, 5, 10, 20, 50, 100 signal events

ψ distributions

100 background events; 2, 5, 10, 20, 50, 100 signal events

ψ distributions

200 background events; 2, 5, 10, 20, 50, 100 signal events

ψ distributions

500 background events; 2, 5, 10, 20, 50, 100 signal events

ψ distributions

1000 background events; 2, 5, 10, 20, 50, 100 signal events

Significance

Optimisation

Conclusions

Charged current ν_e

Charged current ν_μ

Charged current ν_τ