1. K.Somiya Detection of blackhole ringdown signals
using quasi normal mode harmonics
(Status talk of A04:15H00778)
5th annual symposium of the innovative area
Dec. 2016
Tokyo Institute of Technology
K.Somiya and S.Atsuta
K.Somiya

2. BH inspiral and ringdown
In GW the inspiral signal is clear while the ringdown
is not so clear (ringdown freq ~ 250Hz).
-> Is it possible to increase the SNR for the ringdown?
-> Is it possible to detect BH only from its ringdown signal?

3. aLIGO current sensitivity and KAGRA's goal sensitivity
We recently changed a design of
the suspension and the thermal
noise peak moved away from the
ringdown frequency of 60Ms BH.
A 60Ms ringdown is quite suitable
for the current aLIGO sensitivity.

4. Can we detect a QNM alone?
gravitational wave
time
It is not easy to distinguish a signal from nonstationary
noise as decaying sine waves are too common in the nature.
Using not only the 22 mode but also its harmonics, we may
be able to increase the SNR for the ringdown.

5. Ringdown signal decay rate and amplitude
(Average spin before collision)
Leaver 1985
Kamarestos 2012
QNM frequencies was derived long time ago.
QNM amplitudes can be estimated now with NR.

6. What is the best way to use HOMs?
(a) coincidence search
(b) MF with HOMs
Data
Data
Matched filter
with a wave with modes
Matched filter
with a 22 mode
Matched filter
with a 33 mode
Compare the arrival time
(a) would be no good as SNR with 33-mode alone is low
We have selected (b)

7. What can we expect from the MF with HOMs?
Detection probability may increase
False-alarm rate may decrease
Signal-to-noise ratio may increase

8. Permutation test with aLIGO data
Download the data around GW (no signal)
Shuffle the data to exclude non-Gaussian noise
Include ringdown signal at 22 and 33 modes
Matched filter with 22 mode only or with modes

9. Analysis method Code : Matlab Data length : 1 sec
Signal amplitude : given so that SN~3 with 22 mode MF
Signal waveform : decay sine
Mass ratio before collision : 1.4Ms/60Ms
Harmonics mode : 33 only
Initial phase difference : fixed to zero

10. If parameters for 33 mode are unknown, the ROC curve
Results 1: ROC curve
*ROC=Receiver Operating Characteristics
Blue: 22-mode only
Red: modes
detection probability
detection probability
Higher-order mode parameters are unknown
Higher-order mode parameters are given
false alarm rate
false alarm rate
If parameters for 33 mode are unknown, the ROC curve
is not so different from the one with 22 mode only.
If parameters for 33 mode were known, the ROC curve
would be improved with the 33 mode.

11. Results 2: Signal-to-noise ratio
Blue: SNR / 22 SNR
Red: mean
SNR improves with the 33 mode included.

12. Summary and future work
We investigated the possible advantage of using higher-order harmonics of QNM in the ringdown signal analysis with aLIGO Gaussian noise
ROC curve does not change much if HOM parameters are not given; does improve it the parameters are given
SNR improves even if the HOM parameters are not given
Analysis with non-Gaussian noise is to be done

13. Acknowledgement
This study is supported by MEXT Grant-in-Aid for Scientific Research on Innovative Areas , A04 “多様な観測に連携する重力波探索データ解析の研究”

14. Supplementary

15. KAGRA sensitivity << Old curve (calculated in 2013)
The peak at 120Hz is for vertical suspension thermal noise
Recoil mass is suspension from the intermediate mass
New curve (calculated in 2016) >>
Sapphire suspension is longer
Temperature is 23K
Vertical isolation by Sapphire blade spring
Recoil mass is suspended from the intermediate recoil mass