New method for enhanced efficiency in detection of gravitational waves from supernovae using coherent network of detectors

Supernovae in our universe are potential sources of gravitational waves (GW) that could be detected in a network of GW detectors like LIGO and Virgo. Core-collapse supernovae are rare, but the associated gravitational radiation is likely to carry profuse information about the underlying processes driving the supernovae. Calculations based on analytic models predict GW energies within the detection range of the Advanced LIGO detectors, out to tens of Mpc for certain types of signals e.g. coalescing binary neutron stars. For supernovae however, the corresponding distances are much less. Thus, methods that can improve the sensitivity of searches for GW signals from supernovae are desirable, especially in the advanced detector era. Several methods have been proposed based on various likelihood-based regulators that work on data from a network of detectors to detect burst-like signals (as is the case for signals from supernovae) from potential GW sources. To address this problem, we have developed an analysis pipeline based on a method of noise reduction known as the harmonic regeneration noise reduction (HRNR) algorithm. To demonstrate the method, sixteen supernova waveforms from the Murphy et al. 2009 catalog have been used in presence of LIGO science data. A comparative analysis is presented to show detection statistics for a standard network analysis as commonly used in GW pipelines and the same by implementing the new method in conjunction with the network. The result shows significant improvement in detection statistics.

Figure 1

The figure shows 8 supernovae waveforms used in the study from Murphy et al. 2009 catalog [9] with progenitor masses 12 and 15.

Figure 2

The figure shows 8 supernovae waveforms used in the study from Murphy et al. 2009 catalog [9] with progenitor masses 20 and 40.

Figure 3

The analysis pipeline starts with obtaining the raw data from three LIGO detectors (H1: Hanford 4k and L1: Livingstion 4k.) Supervovae signals are added to the data stream. The prepared data streams are now subjected to the data conditioning step where data are bandpassed between 50 Hz and 2048 Hz and all narrowband noise in this range is suppressed. The conditioned data now passes through the TSNR and HRNR module. The output from this module is supplied as the input to the network analysis pipeline. The network analysis pipeline yields the detection statistics in accordance with Eqs. (58) and (59).

Figure 4

The waveform (grw_12_2, left panel) is injected into the data stream (right panel) with a scale factor of 30. The x-axis represents time in seconds and the y-axis is the amplitude.

Figure 5

The figure shows the spectrograms of the signal $(\text{grw}\_12\_2)+\text{noise}$ after being conditioned without the $\mathrm{TSNR}+\mathrm{HRNR}$ denoising effect (top left panel) and the same with the inclusion of the $\mathrm{TSNR}+\mathrm{HRNR}$ denoising (top right panel.) The x-axis represents time in seconds and the y-axis represents frequency in Hz. The bottom row of figures shows the radial statistics (as given in equation [17]) for the analysis performed without the proposed denoising (left panel) and with the $\mathrm{TSNR}+\mathrm{HRNR}$ denoising (right panel.) The x-axis represents time in seconds and the y-axis represents the radial distance.

Figure 6

The figure shows the radial statistics as a function of distance for four waveforms from the catalog. The x-axis represents the distance in kpc. The y-axis represents the detection snr defined from Eq. (59). The detection snr is equal to the average value of the radial distance $R_{\text{rad}}$ subtracted from the maximum value of the same that corresponds to the event detected. The lower curve (in blue) is the operating characteristic for the original pipeline without the implementation of the proposed $\mathrm{TSNR}+\mathrm{HRNR}$ denoising module. The upper curve (in red) represents the same with the incorporation of the $\mathrm{TSNR}+\mathrm{HRNR}$ denoising.

Figure 7

The figure shows the radial statistics as a function of distance for four waveforms from the catalog. The x-axis represents the distance in kpc. The y-axis represents the detection snr defined from Eq. (59). The detection snr is equal to the average value of the radial distance $R_{\text{rad}}$ subtracted from the maximum value of the same that corresponds to the event detected. The lower curve (in blue) is the operating characteristic for the original pipeline without the implementation of the proposed $\mathrm{TSNR}+\mathrm{HRNR}$ denoising module. The upper curve (in red) represents the same with the incorporation of the $\mathrm{TSNR}+\mathrm{HRNR}$ denoising.

Figure 8

The figure shows the radial statistics as a function of distance for four waveforms from the catalog. The x-axis represents the distance in kpc. The y-axis represents the detection snr defined from Eq. (59). The detection snr is equal to the average value of the radial distance $R_{\text{rad}}$ subtracted from the maximum value of the same that corresponds to the event detected. The lower curve (in blue) is the operating characteristic for the original pipeline without the implementation of the proposed $\mathrm{TSNR}+\mathrm{HRNR}$ denoising module. The upper curve (in red) represents the same with the incorporation of the $\mathrm{TSNR}+\mathrm{HRNR}$ denoising.

Figure 9

The figure shows the radial statistics as a function of distance for four waveforms from the catalog. The x-axis represents the distance in kpc. The y-axis represents the detection snr defined from Eq. (59). The detection snr is equal to the average value of the radial distance $R_{\text{rad}}$ subtracted from the maximum value of the same that corresponds to the event detected. The lower curve (in blue) is the operating characteristic for the original pipeline without the implementation of the proposed $\mathrm{TSNR}+\mathrm{HRNR}$ denoising module. The upper curve (in red) represents the same with the incorporation of the $\mathrm{TSNR}+\mathrm{HRNR}$ denoising.