Steward Observatory/NSF’s NOIRLab Joint Colloquium Series: Cosmology and astrophysics from field level inference of multi-probe observations

Abstract: In cosmological studies, our goal is to constrain the initial conditions and cosmic evolution by forward modeling cosmic probes and comparing them with observations. To exhaust the available information, it's essential to forward model the entire field of observations. This approach, however, faces two primary challenges. First, high energy feedback from supernovae and supermassive black holes, which significantly alters the matter field, is poorly understood. I will show how the Sunyaev-Zel’dovich (SZ) effect, which probes ionized gas thermodynamics, aids in calibrating this feedback. I'll introduce a new GPU-accelerated model for joint modeling of gas thermodynamics and matter distribution that significantly enhances cosmological constraint precision. Second, the high-dimensional nature of field-level inference requires a differentiable pipeline for efficient parameter space exploration. I will present how machine learning-based models can be used for this purpose, which, when combined with joint baryon-matter modeling, forms a foundation for a field-level inference pipeline with multi-probe observations.

In-person and Virtual