During 2015-2019 the American Laser Interferometer Gravitational-Wave Observatory (LIGO) and European Virgo detected about 100 binary black hole mergers.  These observations predict that every minute a gravitational-wave signal from a coalescing pair of black holes and/or neutron stars should pass by the Earth every minute. Most of those signals will be far below the noise level but upgraded LIGO and Virgo detectors could observe one such merger every day. But the situation is entirely different with Cosmic Explorer and Einstein Telescope (planned next generation (XG) observatories) which could be sensitive to majority of these signals. The problem is that these signals could also last far longer in XG observatories and the random superposition of hundreds of signals sweeping across a detector’s sensitive band could cause a confusion foreground. I recently wrote about how this foreground could masquerade stochastic backgrounds produced in the early universe but they may also degrade the visibility of individual signals.

In a recent arXiv article, Shichao Wu and Alexander Nitz at the Albert Einstein Institute looked at the problem and found that the confusion foreground (red trace in the left plot) can overwhelm the instrument noise (gray trace in the left plot) and significantly reduce the distance, equivalently the cosmological redshift,  up to which Cosmic Explorer and Einstein Telescope could detect coalescing binaries. Indeed, they find that the loss in the case of Cosmic Explorer could be between 15-35% (in the case of Einstein Telescope 8-21%) depending on the true rate of mergers, which is currently uncertain plagued by small number of statistics of LIGO-Virgo detections. They also find that binary neutron stars are largely responsible for this degradation in sensitivity (pink trace in the right plot).

Luckily, not all is lost. It turns out that a simple “detect and subtract” method, which we had found is not sufficient to mitigate foreground contamination of stochastic backgrounds, actually works pretty well in reducing the confusion noise. This would bring the red trace in the left panel pretty close to the gray curve and hence largely alleviate the loss in range.

This is very encouraging but we need to show that this would also work for data that is not as clean as simulated Gaussian noise and replete with noise stationarities and other gravitational-wave signals whose rate we don’t know.