Bubbles and wizards in the intergalactic dark forest: constraining fundamental physics through cosmological simulations and observations of the intergalactic medium

Abstract

The intergalactic medium (IGM) holds most of the baryons in the Universe and occupies nearly all of its volume, making it a long path length over which feebly interacting new physics can act. This thesis uses large cosmological hydrodynamical simulations to characterise the IGM and to assess what it can and cannot tell us about physics beyond the Standard Model.

Using the IllustrisTNG simulations, we show that feedback from supernovae and accreting supermassive black holes inflates over-magnetised bubbles extending tens of megaparsecs beyond collapsed structures. These bubbles form at z ≲ 2, occupy more than ten per cent of the volume at z = 0, and carry magnetic fields orders of magnitude stronger than unaffected gas of the same electron density. Because their field strength is set by processes internal to galaxies, it is insensitive to the primordial seed field across thirteen orders of magnitude, which both makes it predictable and makes the IGM an unreliable record of primordial magnetogenesis. We quantify the consequences for the conversion of cosmic microwave background photons into axions, and show that roughly half of 10²⁰ eV protons are deflected by more than one degree over 100 Mpc, with implications for the identification of ultra-high-energy cosmic ray sources.

We then construct a continuous lightcone through the FLAMINGO simulations and compare the resulting anisotropies with unprocessed Planck data to constrain resonant photon–dark photon conversion in the IGM. Finally, we address the Lyman-α forest as a probe of dark matter. We present a phenomenological recipe connecting the three-dimensional matter power spectrum to the observed one-dimensional flux power spectrum, validated against a dedicated suite of warm and cold dark matter simulations, and demonstrate that warm dark matter free streaming and pressure smoothing are degenerate: a reionization history can be constructed for which the two scenarios are observationally indistinguishable. We also present specwizard, a public Python package for computing mock quasar absorption spectra, and use it to infer the redshift evolution of the photo-ionization rate.