Retrieving Internal Kinematics of Galaxies with Deep Learning using Single-Band Optical Images

TL;DR

This paper demonstrates that deep learning models can accurately predict the internal kinematic properties of galaxies, such as velocity dispersions and rotational velocities, solely from single-band optical images, revealing hidden structural information.

Contribution

The study introduces a novel deep learning approach to infer galaxy internal velocities from optical images, achieving near-resolution accuracy without spectroscopic data.

Findings

Velocity dispersions measured within 29 km/s accuracy

Rotational velocities measured within 69 km/s accuracy

Optical galaxy structures encode detailed internal kinematic information

Abstract

Using deep machine learning we show that the internal velocities of galaxies can be retrieved from optical images trained using 4596 systems observed with the SDSS-MaNGA survey. Using only $i$-band images we show that the velocity dispersions and the rotational velocities of galaxies can be measured to an accuracy of 29 km~$\rm{s}^{-1}$ and 69 km~$\rm{s}^{-1}$ respectively, close to the resolution limit of the spectroscopic data. This shows that galaxy structures in the optical holds important information concerning the internal properties of galaxies and that the internal kinematics of galaxies are quantitatively reflected in their stellar light distributions beyond a simple rotational vs. dispersion distinction.