The spatially-resolved effect of mergers on the stellar mass assembly of MaNGA galaxies

Abstract

Understanding the origin of stars within a galaxy - whether formed in-situ or accreted from other galaxies (ex-situ) - is key to constraining its evolution. Spatially resolving these components provides crucial insights into a galaxy's mass assembly history. We aim to predict the spatial distribution of ex-situ stellar mass fraction in MaNGA galaxies, and to identify distinct assembly histories based on the radial gradients of these predictions in the central regions. We employ a diffusion model trained on mock MaNGA analogs (MaNGIA), derived from the TNG50 cosmological simulation. The model learns to predict the posterior distribution of resolved ex-situ stellar mass fraction maps, conditioned on stellar mass density, velocity, and velocity dispersion gradient maps. After validating the model on an unseen test set from MaNGIA, we apply it to MaNGA galaxies to infer the spatially-resolved distribution of their ex-situ stellar mass fractions - i.e. the fraction of stellar mass in each spaxel originating from mergers. We identify four broad categories of ex-situ mass distributions: flat gradient, in-situ dominated; flat gradient, ex-situ dominated; positive gradient; and negative gradient. The vast majority of MaNGA galaxies fall in the first category - flat gradients with low ex-situ fractions - confirming that in-situ star formation is the main assembly driver for low- to intermediate-mass galaxies. At high stellar masses, the ex-situ maps are more diverse, highlighting the key role of mergers in building the most massive systems. Ex-situ mass distributions correlate with morphology, star-formation activity, stellar kinematics, and environment, indicating that accretion history is a primary factor shaping massive galaxies. Finally, by tracing their assembly histories in TNG50, we link each class to distinct merger scenarios, ranging from secular evolution to merger-dominated growth.

Figures (19)

• Figure 1: Schematic of the neural network architecture used in this work to predict 2D spatially resolved maps of the ex-situ stellar mass fraction from observable 2D maps of stellar mass density, velocity, and velocity dispersion. A diffusion model is trained to learn the underlying distribution of ex-situ maps using the MaNGIA training set, conditioned on the corresponding observable maps (after a series of preprocessing steps, for details see Section \ref{['preprocessing']}).
• Figure 2: Reconstruction of the ground-truth from the diffusion model for 3 mock galaxies from the MaNGIA test set. The first three columns display the input observable maps provided to the diffusion model: stellar mass, stellar velocity, and velocity dispersion. These maps have been forward-modeled through the MaNGIA pipeline and subsequently normalized to retain only gradient information. The fourth column shows the ground-truth 2D ex-situ stellar mass fraction map from the TNG50 cosmological simulation. The fifth and sixth columns present two individual samples from the diffusion model. The seventh column displays the median prediction from 100 samples. Finally, the last two columns show the model uncertainty (computed as the 84th–16th percentile range) and the residual between the median prediction and the ground-truth map, respectively.
• Figure 3: Metrics of accuracy of the diffusion model trained on the MaNGIA dataset. The recovery of the global ex-situ stellar mass fraction from 2D predicted map of the model vs. the true global value of the ex-situ stellar mass fraction for the train and test set (left). The aggregated ex-situ stellar mass fraction from the predicted 2D map vs. the aggregated value from the ground-truth 2D map for the train and test set (middle). The recovery of the regional trends in the predicted map vs. the ground-truth map for the train and test set integrated at different annuli (right). The green solid line denotes the median recovery, and the shaded region covers 68 percent of the test set, representing the prediction scatter. For reference, we also include the same comparison for 1,000 training set galaxies in blue, underlining that the model is not overfitting.
• Figure 4: Comparison of the integrated 2D predictions with the 1D predictions from 2024NatAs...8.1310A. (a) The recovery of the global ex-situ stellar mass fraction from 2D predicted map of the model vs. the predicted global value of the ex-situ stellar mass fraction for the MaNGA dataset from previous work. Each point represents a galaxy, and the solid blue line traces the median of the distribution. (b) The median difference of the two predictions as a function of stellar mass. The orange error bars display the median uncertainty per stellar mass bin from the 1D predictions. While the residual between predictions shows a slight increase towards higher stellar masses, it is still covered by the produced uncertainties. The shaded regions enclose 68% of all data.
• Figure 5: The stacked radial profiles of the ex-situ stellar mass fraction for 10,000 MaNGA galaxies as predicted from the diffusion model. (a) The median of the stacked radial profiles separated in 3 stellar mass bins. (b) The median of the stacked radial profiles in each group, following the classification based on the central ex-situ gradient. The shaded regions enclose 68% of all data.