Joint LIGO-Virgo data analysis Inspiral and Burst

1
Joint LIGO-Virgo data analysisInspiral and Burst

• Summary of the first project results
• Overview of the future activities
• M.-A. Bizouard (LAL-Orsay)
• on behalf of the LSC-Virgo working group

Hanford
Livingston
Virgo
2
Goals - introduction

• Benefits of using multiple detectors
• Decrease false alarm
• Increase sky coverage
• Source location reconstruction with at least 3
  detectors
• Confirm or eventually kill a golden candidate
• Before real data joint analysis many issues
• Different search pipelines comparison
• Different sampling rate and sensitivities
• What are the expected benefits and performance of
  the network?
• How well can we estimate signal parameters/source
  location
• Goal of the working group
• Address all the potential issues
• Coincidence and coherent analysis
• Define a strategy for a burst/inspiral discovery?

Project Ia

Project Ib
SIMULATED DATA!
3
Coincidence search simulated data set

• Noise 24 hours of Virgo, Hanford and Livingston
  noise at nominal sensitivities
• Signals
• Inspiral 1-3 M from 2 Galaxies
• NGC 6744 at 10 Mpc
• M87 at 16 Mpc (Virgo cluster)
• Random polarization orbital plane inclination
• 2PN chirp inspiral generation
• Burst Galactic center direction
• 2 core collapse simulated waveforms
• 2 Gaussian peaks
• 2 CosineGaussians
• Random polarization
• Normalization only one event seen at SNR gt 10
• In the 3 ITFs over 24 hours

4
Inspiral search

• Pipelines
• Virgo MultiBandTemplateAnalysis (MBTA) flat
  search (Merlino)
• LIGO flat search
• Cross-check that all pipelines are working well
  and have similar performance running on Virgo and
  LIGO data.
• Solve few signal generation discrepancies
• G value must be the same
• Chirps length must be computed in the same way at
  2PN
• inspiral end frequency definition LSO or ISCO

5
Inspiral coincidence analysis resultsHanford-Livi
ngston-Virgo network performance
MBTA efficiencies

• Single interferometer results
• SNR threshold at 6
• False alarm rate 0.1 Hz
• Coincidence
• Require time and mass coincidence
• Triple coincidence
• False alarm in 24 hours 0
• Double coincidence
• False alarm in 24 hours 1
• Adding Virgo gives 25 increase in efficiency
  for M87

H L V H?L?V
61 62 56 75
M87 (16 Mpc) NGC 6744 (10 Mpc)
HLV 24 48
quite high ? source location often possible!
M87
HL HV LV HL?HV?LV
42 32 30 56
6
Inspiral source reconstruction

• Timing accuracy and binary parameters dependence
  issue
• Improvement obtained by requesting a mass
  correlation between the 3 ITFs template triggers
  (pseudo coherent follow up)

Fitted direction using the same template for the
3 ITFs
Without asking any mass correlation in The 3
ITFs triggers
M87
NGC
7
Inspiral search open issues

• Timing estimation bias observed in LIGO / Virgo
  pipelines
• Signal time domain generation both for Virgo and
  LIGO data sets
• Template LIGO stationary phase approximation
  (frequency
  domain)
• Virgo Fourier transform of time domain
  templates

• MBTA / LIGO SNR ratio

8
Burst search - pipelines

• Time frequency
• Power Filter (PF), Q Transform (QT), Kleine Welle
  (KW)
• Time domain
• Mean Filter (MF), Alternative Linear Fit Filter
  (ALF)
• Correlators
• Gaussian templates (PC), Complex Exponential
  Gaussian templates (EGC)
• A battery of filters used to cover a large
  variety of possible waveforms.
• Performance comparison different efficiencies
  according to the waveform
• robustness tests
• Can we gain by combining the different filters?
• AND/OR analysis?

9
Burst LIGO-Virgo network sky coverage
Source in the direction of the Galactic center
Burst SNR seen in each ITF as function of time
Virgo and LIGO ITFs do not see the Galaxy
center at the same time ? is there an
interest of coincidence analysis?
24 hours
10
Burst coincidence search resultsperformance of
the HLV network

• Example A2B4G1 waveform
• Single interferometer results
• Best efficiency among 5 filters
• False alarm rate 0.1 Hz
• (10 000 FA per day)
• Coincidence
• Require time (and frequency) coincidence
• Double coincidence
• False alarm 10-6 Hz
• Triple coincidence
• False alarm 10-6 Hz

efficiency
H L V
63 60 55
efficiency
HL HV LV HL?HV?LV
41 22 22 60
efficiency
Adding Virgo to LIGO increases the network
efficiency by 50
HLV
19
11
Burst AND/OR analysis (on-going work)
OR can we recover events by combining several
algorithm triggers?
At high FAR (0.1Hz) small efficiency
increase At low FAR no gain
PRELIMINARY!
12
Filter robustness tests

• Goal assess filters performance over a class of
  signals spanning the duration, central frequency
  and frequency band parameters

Example PF signal SNR10
Band pass White noise SineGaussian
DFM

• Some signal location depends on the parameters
  definition
• A filter can have different efficiency response
  depending on the waveform
• shows that the 3 parameters do not fully
  describe a signal and/or the
• response of a filter

13
Burst source location

• 2 classical methods using arrival time and SNR
  information of triple coincidence triggers. (?2
  minimization and likelihood maximization).
• Comparison to be done!
• Talks of F. Cavalier S. Klimenko

Example burst from GC using the ?2 method using
triple coincidence events over 24 hours (HLV
eff19) GC a GC 266.4 dGC
-28.98 Rec a GC 266.4 /- 0.70
dGC -28.98 /- 0.97
angular error 1
14
Present activities

• 2 papers in preparation containing all results
  obtained so far
• Coherent analysis under test and/or development
• Burst
• 2 LIGO pipelines (Likelihood and NULL streams
  methods)
• 1 Virgo pipeline (J. Sylvestre PRD 68 (2003))
• Inspiral
• 1 LIGO pipeline
• 1 Virgo pipeline
• LIGO coherent inspiral parameter estimation
  (Markov chains)

S. Klimenko talk A. Searle talk

S. Bose, S. Dhurandhar A. Pai Int. J. Mod Phys
D9, 25 (2000)
15
The next project real data exchange

• Project 1 demonstrates the benefits and the
  feasibility of a joint data analysis
• Pipelines are ready both in LIGO and Virgo
• Project 2 real data analysis!
• 2a exchange of 3 hours of data for technical
  validation (S4 and C7 f.i.)
• 2b exchange of 24 hours of coincident real data
  
• coincidence and coherent burst and
  inspiral searches
• face real noise issues!
• However, many issues still open
• which data? (similar sensitivity? or at least
  above few hundreds of Hz)
• ? when? wait for Virgo is back this
  spring!
• joint data analysis coordination the LSC-Virgo
  data analysis group
• publication policy?
• detector knowledge spreading enhancement?

16
Conclusions

• Inspiral and burst pipelines succesfully tested
  (up to source location reconstruction)!
• Coincidence analysis shows clear benefits to add
  Virgo to LIGO network.
• Coherent analysis in progress
• Work scientifically sound!
• Consensus for real data exchange in the next
  years !
• Political agreements to be signed
• Good prospect for joint data set and analysis
  in 2006