Variance-aware model discrimination with the Sunyaev-Zel'dovich effect of the 21 cm background (SZE-21cm)

TL;DR

This paper explores the potential of the Sunyaev-Zel'dovich effect of the 21 cm background (SZE-21cm) to distinguish different Cosmic Dawn scenarios using semi-numerical simulations and a new separability index, providing a model-independent upper bound for experiments.

Contribution

It introduces a novel, intrinsic, and instrument-free separability index for SZE-21cm spectra to differentiate Cosmic Dawn models, accounting for astrophysical variance and observational limitations.

Findings

Heating timing variations produce the strongest spectral separations.

X-ray hardness variations are nearly degenerate with the fiducial case after variance inclusion.

The framework can map anomalous global-signal morphologies into SZE-21cm signatures.

Abstract

The sky-averaged redshifted 21 cm signal from Cosmic Dawn is a uniquely sensitive tracer of early heating and ionisation, but it remains challenging to measure directly. The Sunyaev-Zel'dovich effect of the 21 cm background (SZE-21cm) provides a complementary route: Comptonisation of the incident low-frequency background by hot electrons in galaxy clusters produces a spectral distortion that can be recovered as a difference between the line of sight through a galaxy cluster and a nearby blank-sky reference, and is therefore naturally compatible with interferometric observations. We use semi-numerical simulations of the global 21 cm background, together with a relativistic scattering kernel built from the Maxwell-Juttner electron distribution, to assess how well the SZE-21cm separates physically distinct Cosmic Dawn scenarios. The model suite varies star-formation efficiency, X-ray spectral hardness, heating timing, and the suppression of low-mass sources; SZE-21cm spectra are computed over 45-200 MHz. Separability is quantified through feature-based summaries, standardised residual spectra, and a band-integrated pairwise separability index that combines seed-to-seed scatter with an astrophysical variance layer derived from coeval ON-aperture / OFF-annulus statistics. The resulting indices are intrinsic and instrument-free, and therefore upper bounds on what a real experiment with finite thermal noise can achieve. Heating-timing variations produce the strongest separations, while the X-ray hardness variations explored here are nearly degenerate with the fiducial case once the ON-OFF variance is included. We also propagate an EDGES-like benchmark curve as a morphological stress test of the framework, not as a physically validated model, illustrating how anomalous global-signal morphologies would map into the SZE-21cm.