Presentation #211.05 in the session Gravitational Waves and Lensing.
Around 2037, the Laser Interferometer Space Antenna (LISA) will launch into heliocentric orbit. LISA is expected to detect merging supermassive black-hole binaries, superimposed upon millions of white-dwarf (WD) binary systems in our Milky Way. The WD galactic background will present a set of independent, slowly-evolving monochromatic signals: a forest of lines in the LISA frequency spectrum. This galactic background may enable large-scale surveys of Milky Way in addition to distinct tests of General Relativity, facilitated by the long duration and high spatial resolution of the signals. The NASA-developed Global Bayesian Fit (GBF) to the WD background can model tens of thousands of these WD sources. This presentation describes statistical approaches that could complement the GBF by providing alternative, rapid source detection with high computational efficiency. Such techniques are adapted from the Peta-scale computing costs of continuous-wave searches on Laser Interferometer Gravitational-wave Observatory (LIGO) data, encompassing a family of matched-filtering/cross-correlation statistics and related methods that marginalize over parameters in the Bayesian fit. Realistic trade-offs are considered between low-dimensional, multi-modal likelihood surfaces in contrast to high-dimensional surfaces, as in the GBF, that would be unimodal in the absence of noise. Iterative approaches to parameter estimation in the multi-modal case, based on LIGO continuous-wave techniques, may provide for rapid identification of prominent spectral lines from the WD galactic background. Identifying and estimating parameters from the LISA background is of critical importance to multi-messenger astronomy collaborations, both in conjunction with electromagnetic observations and, for some source types, ground-based gravitational-wave observatories. Estimates will often be time-sensitive, and a finite number of events are likely in the nominal four-year mission duration. LISA Data Challenge datasets will be used to validate these computational methods. This project is funded by Laboratory Directed Research & Development (LDRD) 20230448ECR and this abstract is approved for release under LA-UR-23-23589.