Testing the cosmic distance-duality relation with localized fast radio bursts: a cosmological model-independent study

TL;DR

This study tests the cosmic distance-duality relation using localized fast radio bursts and supernova data, employing neural networks for model-independent reconstruction, and finds no significant deviations from the relation.

Contribution

It introduces a neural network-based, model-independent method to test the CDDR with FRBs and supernovae, providing new constraints on cosmological distances.

Findings

No significant deviation from CDDR detected.

Neural network effectively reconstructs distance measures.

Constraints on host galaxy dispersion measure obtained.

Abstract

We test the Etherington cosmic distance-duality relation (CDDR), by comparing Type Ia supernova (SNIa) luminosity-distance information from the Pantheon+ compilation with an angular-diameter-distance reconstructed from localized Fast Radio Bursts (FRBs). The core of our methodology is a data-driven reconstruction from FRBs using artificial neural networks (ANNs): we infer a smooth mean extragalactic dispersion-measure relation and use its redshift derivative to recover $H(z)$ and hence $D_\mathrm{A}^{\rm FRB}(z)$ without assuming a parametric form for the expansion history. Possible deviations from CDDR are parameterized through three one-parameter models of $\eta(z)\equiv D_\mathrm{L}/[(1+z)^2D_\mathrm{A}]$. We implement two complementary likelihoods: (i) a direct approach using individual SNIa with the full Pantheon+ covariance, and (ii) a machine-learning approach in which we reconstruct the SN Hubble diagram on the FRB redshift grid, propagating SN and FRB uncertainties into non-diagonal covariance matrices via Monte Carlo and bootstrap realizations. Within the FRB reconstruction, we anchor the mean extragalactic dispersion measure at $z=0$, which yields a data-driven constraint on the average host/near-source contribution $\mathrm{DM}_{\rm host}=128.8\pm 34.1\,\mathrm{pc\,cm^{-3}}$ ($3\sigma$ of statistical confidence). We find that both likelihood implementations give consistent posteriors and no statistically significant evidence for departures from CDDR at the current precision.