Prototype Stochastic Gravitational Wave Background Recovery in the LISA Global Fit Residual

Abstract

The Laser Interferometer Space Antenna (LISA) mission poses a difficult parameter estimation challenge: the sources will be so dense in both time and frequency that they all must be fit simultaneously in a `global fit'. Successful tests of global fit efforts on synthetic datasets have been recently reported, recovering extra-galactic black hole mergers and galactic binaries, including the $\mathtt{GLASS}$ pipeline in arXiv:2301.03673. Injected stochastic sources, however, have so far been absent in these datasets. In this work we report our development of a stochastic search pipeline ready for inclusion in future tests of the global fit, capable of detecting or placing limits on a wide variety of possible cosmologically- and astrophysically-inspired SGWBs. The code uses short-time Fourier transforms (STFTs) to allow for inference despite the non-stationarity of the noise. We quote results using both purely synthetic confusion noise and two $\mathtt{GLASS}$ residuals, and quantify the impact of the residuals' non-gaussianity on injected signal recovery and on setting upper limits. We find that, if not properly mitigated, non-gaussianities can preclude setting accurate SGWB upper limits and lead to false detections. We also stress that the narrow-band non-gaussianities we find do not affect all sources equally, and many narrower-band, cosmologically-inspired SGWBs are more sensitive to non-gaussianity than others.