Connecting optical morphology, environment, and HI mass fraction for low-redshift galaxies using deep learning

TL;DR

This paper demonstrates that deep convolutional neural networks can accurately predict a galaxy's neutral hydrogen content from optical images, revealing morphology-gas content connections and environmental effects.

Contribution

It introduces a CNN-based method to estimate HI mass fraction from optical images, outperforming previous estimators and providing interpretability through Grad-CAM visualizations.

Findings

CNN accurately predicts HI mass fraction from optical images.

The morphology-HI connection is stable in low/intermediate-density environments.

The CNN outperforms previous estimators across multiple catalogs.

Abstract

A galaxy's morphological features encode details about its gas content, star formation history, and feedback processes, which play important roles in regulating its growth and evolution. We use deep convolutional neural networks (CNNs) to learn a galaxy's optical morphological information in order to estimate its neutral atomic hydrogen (HI) content directly from SDSS $gri$ image cutouts. We are able to accurately predict a galaxy's logarithmic HI mass fraction, $\mathcal{M} \equiv \log(M_{\rm HI}/M_\star)$, by training a CNN on galaxies in the ALFALFA 40% sample. Using pattern recognition (PR), we remove galaxies with unreliable $\mathcal{M}$ estimates. We test CNN predictions on the ALFALFA 100%, xGASS, and NIBLES catalogs, and find that the CNN consistently outperforms previous estimators. The HI-morphology connection learned by the CNN appears to be constant in low- to intermediate-density galaxy environments, but it breaks down in the highest-density environments. We also use a visualization algorithm, Gradient-weighted Class Activation Maps (Grad-CAM), to determine which morphological features are associated with low or high gas content. These results demonstrate that CNNs are powerful tools for understanding the connections between optical morphology and other properties, as well as for probing other variables, in a quantitative and interpretable manner.