Astrophysical signal consistency test adapted for gravitational-wave transient searches

Gravitational-wave astronomy is established with direct observation of gravitational wave from merging binary black holes and binary neutron stars during the first and second observing run of LIGO and Virgo detectors. The gravitational-wave transient searches mainly separate into two families: modeled and modeled-independent searches. The modeled searches are based on matched filtering techniques, and model-independent searches are based on the extraction of excess power from time-frequency representations. We have proposed a hybrid method, called wavegraph that mixes the two approaches. It uses astrophysical information at the extraction stage of model-independent search using a mathematical graph. In this work, we assess the performance of wavegraph clustering in real LIGO and Virgo noise (the sixth science run and the first observing run) and using the coherent WaveBurst transient search as a backbone. Further, we propose a new signal consistency test for this algorithm. This test uses the amplitude profile information to distinguish between the gravitational-wave transients from the noisy glitches. This test is able to remove a large fraction of loud glitches, which thus results in additional overall sensitivity in the context of searches for binary black-hole mergers in the low-mass range.

Figure 1

Bias ${\mu }_{\mathrm{grid}}$ and standard deviation ${\sigma }_{\mathrm{grid}}$ of the observed pixel amplitude vs the scale parameter $M$, estimated using LIGO O1 data.

Figure 2

The $\xi /{\eta }_c$ distribution for selected O1 noise, S6 noise, and CBC injections. The panel (a) shows the distribution of consistency statistics $\xi$, and panel (b) shows the scatter plot of $\xi /{\eta }_c$ vs ${\eta }_c$. The dashed black line in both plots shows the proposed $\xi /{\eta }_c$ threshold.

Figure 3

Wavelet graph for three BBH parameter space generated using wavegraph algorithm. Left panel for $R_1$ space, middle panel for $R_2$ space, and right panel for $R_3$ space. The color shows the ${\overline{w}}_s$ for selected wavelets.

Figure 4

Background false alarm rate (FAR) vs the statistic ${\eta }_c$. No noise vetoes has been applied here, besides the signal consistency test described in Sec. 3. The solid, dotted, and dashed curves correspond to $R_3$, $R_2$, and $R_1$ regions, respectively. The lines in cyan correspond to cWB with wavegraph (WG), while the lines in black corresponds to cWB with WG and consistency test (CT). The horizontal dashed-red line shows the reference FAR level of ${10}^{-8}\mathrm{Hz}$ (0.3 event per yr).

Figure 5

Background false alarm rate (FAR) vs the statistic ${\eta }_c$. Noise vetoes discussed in Appendix have been applied. The solid, dotted, and dashed curves correspond to $R_3$, $R_2$, and $R_1$ regions. The orange (resp. black) color is for cWB (resp. cWB with wavegraph and consistency). The horizontal dashed-red line indicates the reference $\mathrm{FAR}={10}^{-8}\mathrm{Hz}$ (0.3 event per yr).

Figure 6

Cumulative distribution of the network correlation for the recovered events. The orange (black) line is for cWB ($\mathrm{cWB}+\mathrm{WG}+\mathrm{CT}$). The solid, dotted, and dashed lines are for the $R_3$, $R_2$, and $R_1$ regions, respectively.

Figure 7

Detection efficiency (in percent) vs the injected network SNR given for cWB (orange stars) and for cWB with wavegraph (black circles) with 1-sigma error bars. (a)–(c) panels correspond to the $R_1$, $R_2$, and $R_3$ simulations, respectively.

Figure 8

Fraction of extra detected events by the combined pipeline $\mathrm{cWB}+\mathrm{WG}+\mathrm{CT}$ compared to cWB alone for the $R_1$ (green-dashed), $R_2$ (orange-dotted), and $R_3$ (blue-solid) regions as a function of the FAR threshold (Hz).