Multi-detector null-stream-based $\chi^2$ statistic for compact binary coalescence searches

We develop a new multi-detector signal-based discriminator to improve the sensitivity of searches for gravitational waves from compact binary coalescences. The new statistic is the traditional $\chi^2$ computed on a null-stream synthesized from the gravitational-wave detector strain time-series of three detectors. This null-stream-$\chi^2$ statistic can be extended to networks involving more than three detectors as well. The null-stream itself was proposed as a discriminator between correlated unmodeled signals in multiple detectors, such as arising from a common astrophysical source, and uncorrelated noise transients. It can be useful even when the signal model is known, such as for compact binary coalescences. The traditional $\chi^2$, on the other hand, is an effective discriminator when the signal model is known and lends itself to the matched-filtering technique. The latter weakens in its effectiveness when a signal lacks enough cycles in band; this can happen for high-mass black hole binaries. The former weakens when there are concurrent noise transients in different detectors in the network or the detector sensitivities are substantially different. Using simulated binary black hole signals, noise transients and strain for Advanced LIGO (in Livingston and Hanford) and Advanced Virgo detectors, we compare the performance of the null-stream-$\chi^2$ statistic with that of the traditional $\chi^2$ statistic using receiver-operating characteristics. The new statistic may form the basis for better signal-noise discriminators in multi-detector searches in the future.