Beyond single tracers: Convolutional neural network-based inference of galaxy mass profiles from combined gas and stellar kinematics

¹ Instituto de Astrofísica de Canarias (IAC), Calle Via Láctea s/n, 38205 La Laguna, Tenerife, Spain
² Universidad de La Laguna, Avda. Astrofísico Fco. Sánchez s/n, 38206 La Laguna, Tenerife, Spain
³ New York University Abu Dhabi, PO Box 129188, Abu Dhabi, United Arab Emirates
⁴ Center for Astro, Particle and Planetary Physics, New York University, Abu Dhabi, United Arab Emirates
⁵ Max Planck Institute für Astronomie, Königstuhl 17, 69117 Heidelberg, Germany

^★ Corresponding author: This email address is being protected from spambots. You need JavaScript enabled to view it.

Received: 29 October 2025
Accepted: 17 February 2026

Abstract

Aims. We investigated whether combining gas and stellar kinematic maps provides measurable advantages in recovering galaxy mass profiles compared to using single-component maps alone. We used deep learning models to leverage this joint information.

Methods. We developed a probabilistic convolutional neural network (CNN) framework trained and tested on mock galaxy kinematic maps from multiple cosmological simulation suites. Our model was trained on gas-only, stars-only, and combined gas and stellar velocity maps, thus allowing the direct comparison of performance across tracers. To assess robustness, we included simulations with differing feedback models and galaxy properties.

Results. Combining gas and stellar maps reduces the dispersion in the inferred mass profiles by up to a factor of ~1.5 compared to models using either tracer independently. The CNN architecture effectively captures complementary information from the two components. However, we find limitations in generalising between simulation suites, with reduced performance when applying models trained on one suite to galaxies from another.