Speaker
Description
The LISA space mission, set to launch in the mid 30s, is expected to open a new window on the “gravitational wave universe”. Thanks to its exceptional sensitivity in the low frequency spectrum ~10^-4-10^-1 Hz, it will observe a variety of different sources all at the same time: from massive black hole binaries to extreme mass ratio inspirals or Galactic compact binaries. Among these, Galactic compact binaries of double white dwarves, emitting nearly monochromatic waves, are expected to be the dominant source in the mHz band. Of ~10^7 binaries in the Milky Way, a small fraction of tens to thousands will be actually resolvable by LISA. The unresolved ones are indeed expected to produce a foreground (“confusion”) stochastic noise which must be accounted for in the recovery of the other sources.
Recent works have shown that different astrophysical mechanism driving the binary evolution can affect the overall astrophysical population. In particular, different interacting binary populations can produce sensibly different foreground noises. However, the extraction of the astrophysics from confusion noise still remains a non-trivial challenge in the context of the Global Fit.
In this talk, I will describe an alternative approach to overcome this challenge based on Simulation Based Inference. By simulating several catalogue realizations from different astrophysical populations we train a Neural Posterior Estimator to map the simulated confusion noises to the parameters of the populations. Our model represents a first step towards the linking of astrophysical populations to the output of a Global Fit analysis.