MG12

1
RECENT RESULTS OF THE IGEC2 COLLABORATION SEARCH
FOR GRAVITATIONAL WAVE BURST
Massimo Visco on behalf of the IGEC2
Collaboration
2
OUTLINE OF THE TALK

• IGEC2 collaboration detectors
• IGEC2 activity during past years
• Data analysis methods
• Results of second data exchange of IGEC2
  2005-2007
• Data quality
• Analysis parameters optimization
• Results
• Conclusion and perspectives of the IGEC2
  observatory

3
IGEC2
International Gravitational Events
Collaboration ALLEGRO - AURIGA - ROG
(EXPLORER-NAUTILUS)
4
A 4-ANTENNAE OBSERVATORY

• The four antennas see an identical signal,
  independently of the source and time

5
SENSITIVITY OF IGEC DETECTORS

• The best sensitivity is reached around 900 Hz

6
IGEC 1 search for burst signals 19972000

• First experience using the data of 5 bar
  detectors ALLEGRO, AURIGA, EXPLORER, NAUTILUS
  and NIOBE. In four years 29 days of four-fold
  coincidences- 178 days of three-fold coincidences
  - 713 days of two-fold coincidences
• After the last upgrades the resonant detectors
  have resumed the operations at different times
  EXPLORER in 2000, AURIGA in 2003, NAUTILUS in
  2003, ALLEGRO in 2004. NIOBE no longer in
  operation.

7
IGEC2 search for burst signals 2005-

• First analysis- from May to November 2005 when no
  other observatory was operating. Based on
  three-fold coincidences. No detection
• Second analysis from November 16th, 2005 to
  April 14th, 2007 Based on three and four-fold
  coincidences. No detection
• Future analysis - on April 14th 2007 ALLEGRO
  ceased data taking. Since then the three European
  detectors gathered new data yet to be analyzed.

8
DATA ANALYSIS METHODS
The analysis is based on time coincidence among
candidate events selected in each detector.

• The events selected by each group using filter
  matched to ? signals are characterized by Fourier
  amplitude H and arrival time ti
• h(t) H ? (t- ti)
• The data are exchanged after adding a secret
  time shift to arrival time ti.
• A statistical distribution of the accidental time
  coincidences number is calculated using lists of
  candidate events obtained from the original ones
  adding many different time shifts.
• The analysis parameters (search threshold,
  coincidence window) are fixed a priori using
  the accidental coincidences analysis.
• Finally the groups exchange the secret times and
  the search for real coincidences is performed.

9
IGEC 2 2nd period Nov 16th, 2005 Apr 14th,
2007

• The analysis is based on a composite search, an
  OR of five different configurations four and
  three-fold coincidences.
• To our knowledge this is the longest reported
  period of fourfold coincidence observation.
• The background was fixed at 1 event/century
  equally divided in the four configurations (0.2
  event/century each).
• The data of 2007 became available later, they
  were analyzed using slightly different SNR
  thresholds.

10
OPERATION TIME NOV 16th 2005APR 14th, 2007
515 days
Number of detectors in coincidence Exclusive observation time Analyzed time
0 0 d ---
1 1.6 d ---
2 31.0 d ---
3 188.8 d 482.4 d
4 293.5 d 482.4 d
Full coverage
94 of time useful for analysis
57 of time with 4 detectors
11
AMPLITUDE OF THE EXCHANGED DATA in terms of
Fourier amplitude H

• SNR gt 4.5 for AURIGA
• SNR gt 4.0 for EXPLORER and NAUTILUS
• H gt 1.1 10-21 Hz-1 for ALLEGRO

12
EVENTS AMPLITUDE DISTRIBUTIONS

• SNR gt 4.5 for AURIGA
• SNR gt 4.0 for EXPLORER and NAUTILUS
• H gt 1.1 10-21 Hz-1 for ALLEGRO

13
DATA QUALITY DISTRIBUTIONS OF EVENTS

• Few events/day with SNRgt7
• Few very large events (SNRgt30) on the whole period

14
TUNING OF ANALYSIS PARAMETER
The R factor must be maximized. It depends on
shape and energy of the different signals

• Analysis target are
• a false alarm low enough to select significant
  candidate events (1 event /century)
• a reasonable detection efficiency for the
  searched signals (to be evaluated by software
  injections)
• The parameters to be tuned are
• events SNR selection threshold
• time coincidence windows

15
TIME UNCERTAINTY
The time windows were chosen large enough to
include not only ?-like signals. By software
injection we tested the response also to damped
sinusoids h(t)h0 sin(2 ? f0 t) e-t/? ?(t)
Statistical uncertainty 95 of coincidences
retrieved with a 25 ms windows Systematic
biases the time bias is within 15 ms for ? lt30
ms
16
TIME COINCIDENCE WINDOW
The maximum light travel time between detectors
is 2 ms for European detectors 20 ms European
- United States detectors The chosen time
windows were 40 ms for European detectors
coincidences 60 ms for European - United States
detectors coincidences
17
SNR SELECTION

• Once the time windows were fixed, we tuned the
  SNR thresholds to the required false alarm.
• We used different thresholds for each
  configuration and for each detector equal for
  ALLEGRO, EXPLORER, NAUTILUS and higher by a
  factor 1.5-1.8 for AURIGA

18
SAMPLE DETECTION EFFICIENCY AU-EX-NA
w60ms

• An efficiency of about 50 is reached with
  different signals at amplitudes hrss of (710-20
  - 510-19)

19
BACKGROUND EVALUATION

• In order to highlight possible data correlation
  the background analysis was implemented using
  more than one time shift.
• We used 13 different time shifts from 0.12s to
  3s.
• For each shift value we performed about 12
  million of time lags.

Averaged false alarms with their standard
deviations
The experimental false alarm error is larger than
the statistical one But this does not effect our
analysis
20
BACKGROUND EVALUATION

• A precise evaluation of errors, including
  systematic effects, was made possible by
  calculating false alarms with different time
  shifts.

Configuration P(N0) P(N1) P(N2) P(N3)
AL AU EX 0.998049 0.000046 (1.9490.046) 10-3 (2.020.49) 10-6 lt810-8
AL AU EX 0.998049 0.000047 (1.949 0.047) 10-3 (1.904 0.091) 10-6 (1.2410.088) 10-9
AL AU NA 0.997840 0.000102 (2.150.10) 10-3 (2.230.35) 10-6 lt8 10-8
AL AU NA 0.997840 0.000102 (2.19 0.10) 10-3 (2.340.22) 10-6 (1.690.24) 10-9
AL EX NA 0.998299 0.000057 (1.7000.057) 10-3 (1.49 0.29) 10-6 lt8 10-8
AL EX NA 0.998299 0.000057 (1.7000.057) 10-3 (1.4480.097) 10-6 (8.240.83) 10-10
AU EX NA 0.998325 0.000034 (1.6740.034) 10-3 (1.510.26) 10-6 lt8 10-8
AU EX NA 0.998325 0.000034 (1.6740.034) 10-3 (1.400.057) 10-6 (7.850.48) 10-10
AL AU EX NA 0.998402 0.000024 (1.5980.024) 10-3 (2.13.9) 10-7 lt9 10-8
AL AU EX NA 0.998403 0.000024 (1.5950.023) 10-3 (1.2750.038) 10-6 (6.790.30) 10-10

• Each first row contains experimental occurrence
  probability from time shifts
• Each second row contains Poisson probability
  using the experimental mean
• The two values are fully compatible

21
FINAL RESULTS
NO COINCIDENCE given a false alarm of 1/century

• The collaboration established a priori to make
  available the coincidences found with no
  selection, at high false alarm, for further
  analysis with other experiments.

Configuration Operation time (days) Accidental number Coincidences number
AL AU EX 361.8 4.290.01 3
AL AU NA 390.6 5.150.01 5
AL EX NA 308.7 10.230.01 8
AU EX NA 308.7 2.340.01 4
AL AU EX NA 293.5 (7.660.01)10-3 0
22
CONCLUSION

• Nowadays interferometric detectors have reached
  a sensitivity at least one order of magnitude
  better than bar detectors and no further upgrade
  is scheduled.
• The IGEC observatory is presently capable of
  unattended, low cost operations with high duty
  cycle and low false alarm.
• Interferometric detectors have scheduled
  up-grades in the near future and an important
  increase in sensitivity is expected.
• At present the role of bar detectors is to
  guarantee the coverage for rare but powerful
  events with specific attention to the periods not
  covered by interferometers.