1. Rolling Search For a GRB Cascade Signal
Brennan Hughey
March 19, 2005

2. Our Story so far……
The Rolling cascade search scans the entire data sample (year 2001) for a
significant clustering of events above background
The signal MC spectrum is νe-produced cascades from a Waxman-Bahcall
(Band function) spectrum with break energies at 7.7 X 104 and 1.0 X 107 GeV
Background rejection is accomplished in 3 steps:
The high energy filter is applied – cuts on nhits and OMs with > 2 hits
Ndird(muon fit)/nhits – a loose cut is applied on this parameter
6 variable svm cut – a support vector machine (which functions in a manner
similar to a neural network) is used for further reduction
Total background rejection ~106

3. Two Searches
BATSE experiment finds two classes of bursts, so we have two searches:
1 second and 100 seconds

4. Support Vector Machine Selection
SVMs were chosen such that the total chance of getting a false detection was less than 1% for both searches in the whole analysis – p value of .01 in the event of a detection – 1 second search svm retains 80% of signal and requires 3 events for significance. 100 second search svm retains 53% of signal and requires 4 events for significance.
___ - signal Monte Carlo _____ - real data _ _ _ _ - background MC (dCorsika)
keep
keep

5. Background rates
Number of surviving events are shown as a function of run number

6. Poissonian Distribution of Events
Chi2 test yields a result of 76% for both plots.

7. Poissonian Distribution of Events
Chi2 test yields a result of 75% for left plot and 86% for right plot.

8. Check - Bins With 2 Events
Occurrence of bins with 2 events is not significant, but provides a useful check.
Real data is compared to distribution produced by a Toy Monte Carlo
18 or more bins with 2 events will occur > 5% of the time, so this is a little on the
high side, but not a statistically significant fluctuation

9. Cascade Effective Volume
Use of uncontained events allows effective volume greater than detector size (but means no directional reconstruction)

10. Neutrino Effective Area

11. Expected Distribution of Detected Signal Energies

12. Comparison Between my Search and Satellite-Coincident Searches
Disadvantage
Requires multiple events for a detection.
This means far lower likelihood of getting lucky from a faint source: 1X10-4 events fluctuates to 1 way more often than 3, so requires a particularly bright event even more than other analyses
Limit on Waxman-Bahcall style flux from 667 equivalent sources only ~7X10-6 Gev/cm2 s sr
(This will look better if a more realistic distribution of fluences is used)

13. Comparison Between my Search and Satellite-Coincident Searches
Advantages:
Can pick up GRBs missed by satellites -particularly useful given that CGRO went down in early 2000
Can detect γ-ray dark choked bursts as described in 2001 Meszaros and Waxman paper: astro-ph/
Choked bursts would produce ν as normal GRBs but photons would fail to escape. Frequency unknown but could be as high as ~100 times conventional GRB rate.
Utilizes cascade channel and therefore searches all sky rather than 2π (okay, Ignacio’s analysis can do this too)

14. Sphere of Sensitivity Starting with a GRB at redshift of z=1
Bring closer: 2 effects happen
Geometric effect: Fluence at Earth proportional to 1/r2
This makes a huge difference in the expected event rate.
Relativistic effect: Spectrum changes shape as it moves closer: Ebν increases as 1/(1+z)2.

15. Sphere of Sensitivity: long bursts
Closest measured GRB redshift z=.009 (anomalous burst)
Center of Virgo Cluster
Fluence for “typical” burst used by Waxman-Bahcall

16. Sphere of Sensitivity: short bursts

17. Coincidences With Other Experiments: IPN3
With framework already in place, a check for temporal coincidence with satellite detections is fairly straightforward
81 bursts during livetime from IPN3 catalogs time estimates obtained mostly from Konus-Wind lightcurves
For 100 second search cuts:
Probability of 2 background events during burst times 0.01.
Probability of 3 background events during burst times 5X10-4.

18. Coincidences With Other Experiments: MILAGRO
I will also be collaborating with the MILAGRO air shower array on AMANDA-MILAGRO coincidence studies

19. Backup Slide: cuts

20. Backup Slide 50 TeV 43.43% 100 TeV 48.99% 300 TeV 56.24% 700 TeV
61.01%
1100 TeV
64.08%

21. Backup Slide: cuts

22. Backup Slide: cuts

23. Backup Slide: ndird cut