Learning to See: Applying Inverse Recurrent Inference Machines to See through Refractive Scattering

TL;DR

This paper introduces a neural network approach to mitigate interstellar scattering effects in EHT images of Sagittarius A*, enabling visualization of structures below the nominal resolution limit.

Contribution

The study applies inverse recurrent inference machines to improve the resolution of EHT images by reducing scattering distortions.

Findings

Mitigated scattering effects at 1.3 mm wavelength

Revealed structures at 5 microarcseconds scale

Achieved resolution below the nominal 24 microarcseconds limit

Abstract

The Event Horizon Telescope (EHT) has produced horizon-resolving images of Sagittarius A* (Sgr A$^*$). Scattering in the turbulent plasma of the interstellar medium distorts the appearance of Sgr A$^*$ on scales only marginally smaller than the fiducial resolution of EHT. Therefore, this process both diffractive blurs and adds stochastic refractive substructures that limits the practical angular resolution of EHT images of Sgr A$^*$. We utilized a novel recurrent neural network machine learning framework to demonstrate that it is possible to mitigate interstellar scattering at wavelengths of $1.3\,{\rm mm}$ near the galactic center up to structures at the scale of $5\mu{as}$ well below the nominal instrumental resolution of EHT, $24\,\mu{\rm as}$.