Lokken: Boundless Baryons: the anisotropic superclustering of cosmic gas / Storey-Fisher: Learning Precise Relationships between Dark Matter and Galaxies by Imposing Exact Physical Symmetries

Nov 28, 2022 - 11:00 am to 12:00 pm

Campus, Varian 355

Martine Lokken (University of Toronto) / Kate Storey-Fisher (NYU) In Person Only


The large-scale distribution of baryons in the cosmic web contains a wealth of cosmological and astrophysical information. At low redshifts, the anisotropy of filaments and superclusters depend not only on how dark matter and dark energy shape structure formation, but also on the small-scale galactic feedback processes that eject and heat gas far beyond galaxies. However, most of the gas at z<2 is challenging to observe due to its highly ionized state and intermediate densities and temperatures. This diffuse gas leaves faint imprints in the cosmic microwave background through the thermal Sunyaev-Zel’dovich (tSZ) effect that are well below the noise in current data. In this talk, I will discuss why regions of strong superclustering, as defined by galaxy data, are ideal laboratories to measure the tSZ effect from the diffuse baryons. I will explain the oriented stacking technique and show how it boosts the signal-to-noise of large-scale tSZ measurements while also incorporating the characteristic anisotropy of superclusters. Finally, I will discuss how recent predictions of the diffuse tSZ signal made using The Three Hundred hydrodynamical simulations relate to previous and ongoing measurements using the Atacama Cosmology Telescope + Planck tSZ data.


Mapping dark matter halos to the properties of the galaxies they host in cosmological simulations is a key challenge, required for both performing cosmological inference and understanding galaxy formation. Machine learning (ML) methods have shown promise for this problem, but these typically do not respect the exact physical symmetries that cosmological simulations (and our universe) obey. In this talk, I will present a new approach to enforcing symmetries in ML tasks: the core idea is that input features can be contracted into “invariant scalars” that remain the same under transformations. This can be used to construct a symmetry-preserving description of dark matter halos that contains detailed phase-space information. I will show that this results in precise predictions of galaxy properties as well as the halo assembly history. This will be critical for the analysis of upcoming galaxy surveys with complex target selection.