Status of TAMA data analysis

1
Status of TAMA data analysis
Hideyuki Tagoshi (Osaka Univ.) on behalf of the
TAMA collaboration
2
Outline

• TAMA300 data taking history
• TAMA data analysis
• Inspiral analysis
• Burst analysis
• Ringdown analysis
• Summary

3
Data taking run (1)- Observation runs -

• TAMA observation runs

Data Taking Data Taking Objective Observation time Typical strain noise level Total data (Longest lock)
DT1 August, 1999 Calibration test 1 night 3x10-19 /Hz 1/2 10 hours (7.7 hours)
DT2 September, 1999 First Observation run 3 nights 3x10-20 /Hz 1/2 31 hours
DT3 April, 2000 Observation with improved sensitivity 3 nights 1x10-20 /Hz 1/2 13 hours
DT4 Aug.-Sept., 2000 100 hours' observation data 2 weeks (night-time operation) 1x10-20 /Hz 1/2 (typical) 167 hours (12.8 hours)
DT5 March, 2001 100 hours' observation with high duty cycle 1 week (whole-day operation) 1.7x10-20 /Hz 1/2 (LF improvement) 111 hours
DT6 Aug.-Sept., 2001 1000 hours' observation data 50 days 5x10-21 /Hz 1/2 1038 hours (22.0 hours)
DT7 Aug.-Sept., 2002 Full operation with Power recycling 2 days 25 hours
DT8 Feb.-April., 2003 1000 hours Coincidence 2 months 3x10-21 /Hz 1/2 1157 hours (20.5 hours)
DT9 Nov. 2003 - Jan., 2004 Automatic operation 6 weeks 1.5x10-21 /Hz 1/2 558 hours (27 hours)
Todays talk
4
Data taking run (2)- Obervable range -
Observable range
Detectable distance for binary inspirals
(SNR10, optimal direction and polarization)
1.4 Mo binary inspirals
DT6 33kpc DT8 42kpc DT9 72kpc (30kpc
on average)
Now, TAMA300 covers most part of our Galaxy
DT6
DT9
DT8
5
TAMA data analysis- overview -
Inspiral of compact binaries waveforms
are well-known (chirp) Bursts from stellar-core
collapses, etc wavefoms are not known
precisely Black holes quasi-normal mode
damped sinusoidal waves (ringdown) Pulsars
continuous periodic waves Others veto
analysis, etc.
6
Inspiral analysis (1)
Inspiral phase of coalescing compact binaries are
promising target because expected event rate of
NS-NS merger for LCGT and advLIGO is a few
within 200Mpc / year, and because waveforms are
well-known, etc.
chirp signal
amplitude
time
7
Inspiral analysis (2)- Matched filtering -

• Detector outputs
• h(t) known gravitational waveform
  (template)
• n(t) noise
• Matched filter
• 
  Sn(f) noise power spectrum
• We need to introduce fake event reduction
  method because of non-Gaussian noise
• Fake event reduction by ?2 selection

a measure of the deviation of events from real
signal.
Parameters (mass, coalescence time, ) are not
known a priori. We must search the parameter
space.
8
Inspiral analysis (3)-?- ?2 relation -
TAMA triggers vs Galactic signals
TAMA triggers
?

• We found that the ?- ?2 relation is
  different between the
• non-Gaussian triggers and the simulated
  Galactic signals.
• Thus, we can distinguish them and reduce the
  fake event rate
• produced by non-Gaussian noise.

9
Inspiral analysis (4)- DT8 result -
mass region1-3Msolar
Mass region 1-3Msol
Log10Number of events
Threshold
Set False alarm rate to 0.8 event/yr
10
Inspiral analysis (5) -Upper limit to the
Galactic event rate -

• Threshold12.5 (S/N 9)
• (fake event rate 0.8 /
  year)
• Detection efficiency from Galactic event
  simulation
• We obtain upper limit to the average number of
  events which exceed the threshold by standard
  Poisson statistics analysis
• NUL 2.3 (C.L. 90)
• Observation time T 1163 hours

Upper limit to the event rate
events/hour
29 event/yr (C.L. 90 )
(1-3Msolar)
11
Inspiral analysis (6)- summary -

• DT6(2001) (1038 hours)
• Range 33kpc
• 83 events/yr (1-2Msolar)
• DT8(2003) (1163 hours)
• Range 42kpc
• 29 events/yr (1-3Msolar)
• DT9(2003-4) (558 hours)
• Range 72kpc
• analysis is not finished

Initial results of DT8 analysis Takahashi et al.
Class.Quant.Grav. 21 (2004) S697
12
Inspirla analysis (7)- Coincident analysis -
(20m IFO, Kamioka)
Coincident analysis of DT6 data of TAMA and LISM
was done.
orientation latitude
longitude TAMA 225 35.68N
139.54E LISM 165 36.25N
137.18E
Kamioka (LCGT, CLIO site) 220km west from Tokyo

• Distance between TAMA and LISM 220km

• Maximum delay of signal arrival time0.73msec

• Relation between TAMA and LISM arms direction

Kamioka
220km
Tokyo (NAOJ)
IIAS
13
Inspiral analysis (8)- Coincident analysis -
Data length 275 hours
H.Takahashi et al. PRD70, 042003 (2004)
TAMA triggers
LISM triggers
compare require consistency
DT6 results
99.96 triggers are removed by coincident
analysis
The number of remained triggers are consistent
with the accidental coincidence (no detection)
14
Inspiral analysis (9)- Coincident analysis -

• We demonstrated the power of coincident analysis
  by real data.
• Many technical issue were tackled.

Now, LIGO-TAMA coincident analysis which targets
the inspiral signals is now underway by
LIGO-TAMA joint working group (S. Fairhurst,
H.Takahashi, et al.) Please see the poster by
Takahashi.
15
Burst analysis (1)- overview -

• Excess-power filter analysis (Ando et al.
  gr-qc/0411027)

• Target
• Unmodelled Stellar-core collapse, etc
• Ref. waveforms by numerical sim.
• Schemes
• Excess-power filter
• Fake reduction
• Veto with auxiliary channel
• Time-scale selection
• Galactic simulation
• ? detection efficiency
• ? Upper limit for event rate

Detectable range 300 pc (optimal direction,
polarization)
16
Burst analysis (2) - Excess power filter -

• Burst filter Excess power filter

Evaluate signal power in given time-freqency
regions
Spectrogram
Time- Frequency plane (spectrogram)
Freq. sum
Signal !!
Assumptions for signal time scale,
frequency band Robust for waveform uncertainties
17
Burst analysis (3)- Target waveforms -

• Burst waves by stellar core collapse.

Reference waveforms general relativistic,
conformal flat, axisymmetric simulation
by Dimmelmeier et al. (DFM) 26
waveforms
H.Dimmelmeier et al, Astron. Astrophys. 393
(2002) 523.
Amplitude h rss 4 x 10-22 /Hz1/2
(at Galactic center 8.5kpc) Energy
E tot 9 x 10-8 Moc2
Common characteristics Short burst waves
Spike wave 1msec Duration time lt30msec

• Determine time-frequency band
• Fake reduction

18
Burst analysis (4)- Fake reduction -

• Fake reduction, Injection test

Two veto methods
Burst signal lt 100 msec Most detector noises gt
a few seconds
1. Time-scale selection
Confirm that false dismissal rate is small by
injection tests
remove long-duration triggers
2. Veto with monitor channels
Correlated bursts in intensity monitor channel
Effective to short spikes
19
Burst analysis (5)- Analysis results -

• Analysis results

Trigger rate with vetoes
Improvement in rates with veto
analyses Better in DT9 than DT8 Fake rate
30 100 times Sensitivity 3-6 times
DT8
DT9 (before veto)
Much larger than results with
Gaussian noise
DT9
DT6
Still many fake events
20
Burst analysis (6)- Results -

• Results of Galactic injection test

DT9, 2nd half 200 hours (Christmas,
new-year holidays) Better noise level Stable
environment
Event-selection threshold SNRgt2.9
Detection efficiency 1.5x10-5 Observation
result 7.5x10-2 events/sec
21
Burst analysis (7)- Summary -

• Burst-wave analysis with TAMA300 data
• TAMA300 DT9, 200hours of data
• Excess Power filter, Fake reduction
• Galactic event simulation

Too large for real events Originate in residual
fake triggers
Details can be found in Ando et al.,
gr-qc/0411027
Other activities of burst analysis in TAMA
22
Ringdown analysis (1)
inspiral-merger
Ringdown
Binary, SN expl.
BH formation
Kerr BH
core collapse
QNMs
perturbed BH
Waveform Damped sinusoid (Quasi-normal modes)
h(t)exp(-pfct/Q)sin(2pfct)
Fitting formula for the least damped QNM by
Echeverria (1989)
central frequency
M mass
a angular momentum (non-dimension)
Quality factor
Probe for BH direct observation BH physics in
inspiral-merger, core collapses, ... Good SNR
expected, 100_at_10kpc (TAMA sensitivity)
23
Ringdown analysis (2)- template space -
s(f) signal noise h(f) template Sn(f) noise
power spectrum
Template space (parameter space) construction in
(fc, Q) plane which is slightly more efficient
than previously proposed methods. (Nakano et
al., PRD68, 102003 (2003), PTP 111, 781 (2004) )
Q
682 templates
(SNR loss lt 2)
This template space is effectively independent to
Sn(f) because of its narrow band nature.
fc
24
Ringdown analysis (3)- Event selection -
True signal exponential tail
symmetric around the local maximum
Fake triggers exp rising, no tail
Time t
Time Domain Cuts
-cut
filter output
expected tail
Time t
-cut
True signal ---gt smaller
Time t
25
Ringdown analysis (4)- Detection probability -
for Galactic events
Lower Mass
Larger Mass
100
50
Assumption E0.03MBHis radiated by QNM
gravitational wave
10
26
Ringdown analysis (5)- Galactic event rate -
Tobs Hours DT6 959 DT8 1086 DT9 430
Preliminary
fc gt 1500Hz
(M lt 20Msolar)
Integrated over Q axis
27
Ringdown analysis (6)- Summary -
BH ringdown is promising GW source
Matched filtering code developed
TAMA has good sensitivity to detect Galactic
events, detection probability gt 10
DT6, DT8, DT9 analysis is almost done
See Tsunesada et al. gr-qc/0410037 Initial
results which include detection probability,
pamameter estimation errors, etc. and poster by
Tsunesada for more discussion of the results
28
Summary

• Inspiral analysis, burst analysis, ringdown
  analysis have been developed and the results for
  DT6-9 are now being obtained.
• The tasks needed to be done
• inspiral lower and higher mass search,
  spin, etc.
• burst other filters, more
  effective fake reduction method,
• ringdown more effective fake reduction
  method.
• All of these are now under
  investigation.
• Coincidence analysis
• LIGO-TAMA joint analysis for inspiral and
  burst is in progress.
• ROG-TAMA (bar-interferometer)

• Other activity
• Continuous wave search (target 1987A
  remnant) K.Soida et al., Class. Quantum Grav. 20
  (2003) S645
• ALF filter analysis, Wavelet method,
• Veto analysis,

29
End