From Cluster Core to Splashback: Linking Dynamical Structure to Multidimensional Galaxy Evolution in the Coma Cluster

We investigate galaxy evolution across the full dynamical structure of the Coma cluster using the GalWCat19 spectroscopic cluster catalog combined with SDSS-based value-added galaxy properties. We measure a splashback radius of Rsp = 2.94 ± 0.16 h−1 Mpc. Using specific star formation rate (sSFR), color offset from the red sequence (∆(g − r)RS), and bulge-to-total ratio (B/T) as independent diagnostics, we find a coherent environmental transition from quiescent, red, bulge-dominated galaxies in the cluster core to increasingly star-forming, blue, disk-dominated populations at larger radii. We further introduce a three-dimensional framework in the joint (log sSFR, ∆(g − r)RS, B/T) space and develop a peak-based classification scheme that extends beyond traditional one-dimensional galaxy classifications. This framework identifies two dominant populations: red, quiescent, bulge-dominated galaxies, which account for 51% of the joint-analysis sample, and blue, star-forming, disk-dominated galaxies, which account for 29%. The remaining ∼ 20% of galaxies occupy transitional or mixed states that connect these two principal populations. The relative fractions of these populations change strongly near the splashback radius, where the red, quiescent, bulge-dominated population declines rapidly and the blue, star-forming, disk-dominated population becomes increasingly dominant. These results show that the splashback boundary is not only a dynamical boundary, but also a critical evolutionary transition zone. Overall, our findings suggest that galaxy evolution in Coma is not a purely binary transformation, but instead proceeds through continuous multidimensional pathways in which star formation quenching, color evolution, and morphological transformation occur on different timescales while remaining closely linked to the cluster dynamical structure.

Improvements to AI Systems:

1. Multidimensional Classification Beyond Binary Labels

• Improvement: Replace traditional single-variable galaxy classifiers (e.g., quiescent vs. star-forming) with a peak-based clustering algorithm that operates in a joint 3D space (sSFR, color offset, B/T).

• Capability: The AI can automatically identify not just two dominant populations but also transitional/mixed states, providing a continuous, probabilistic classification of galaxies rather than forcing discrete categories.

2. Splashback-Aware Environmental Transition Modeling

• Improvement: Integrate the measured splashback radius (Rsp = 2.94 ± 0.16 h−1 Mpc) as a learnable boundary parameter in a hierarchical model that predicts galaxy properties as a function of cluster-centric radius.

• Capability: The AI can detect sharp, non-linear transitions in galaxy evolution (e.g., quenching rate, morphological change) at dynamically defined boundaries, enabling more accurate predictions of galaxy fate in clusters.

3. Asynchronous Process Decoupling

• Improvement: Train a multi-task neural network with separate output heads for star formation quenching, color evolution, and morphological transformation, each with its own timescale parameter.

• Capability: The AI can infer that these processes are not simultaneous—e.g., a galaxy may become red before it becomes bulge-dominated—and can predict the order and lag times between them for individual galaxies.

4. Dynamic-Structure-Coupled Evolution Predictor

• Improvement: Use the cluster’s full dynamical structure (not just radius) as input features, including local density, velocity dispersion, and infall history derived from the GalWCat19 catalog.

• Capability: The AI can forecast a galaxy’s evolutionary state based on its current dynamical context, improving predictions for galaxies in merging or accreting subclusters.

5. Continuous Pathway Discovery via Manifold Learning

• Improvement: Apply a variational autoencoder (VAE) to the joint 3D feature space to learn a low-dimensional latent manifold that represents continuous evolutionary pathways.

• Capability: The AI can generate intermediate galaxy states (e.g., red but still disk-dominated) and interpolate between populations, revealing the most likely transition sequences and their timescales.

6. Boundary-Conditioned Population Fraction Estimator

• Improvement: Train a regression model that takes the splashback radius as an input parameter and outputs the fraction of each population (red/bulge, blue/disk, transitional) as a function of radius.

• Capability: The AI can generalize to other clusters by predicting their splashback radii from observable properties, then immediately estimating the expected population fractions—enabling cross-cluster comparisons without new spectroscopy.

What the Improved AI System Can Do:

• Automatically classify galaxies into continuous evolutionary states, not just binary categories.

• Predict the exact radius at which quenching and morphological transformation accelerate, using the splashback boundary as a learned feature.

• Decouple the timing of star formation cessation, color change, and bulge growth for individual galaxies.

• Generate synthetic galaxy populations that follow realistic multidimensional pathways, useful for simulations.

• Transfer learned evolutionary rules from Coma to other clusters, given only their dynamical structure.

We investigate galaxy evolution across the full dynamical structure of the Coma cluster using the GalWCat19 spectroscopic cluster catalog combined with SDSS-based value-added galaxy properties. We measure a splashback radius of R sp = 2.94 plus or minus 0.16 h-1, Mpc. Using specific star formation rate (sSFR), color offset from the red sequence ((g-r) RS), and bulge-to-total ratio (B/T) as independent diagnostics, we find a coherent environmental transition from quiescent, red, bulge-dominated galaxies in the cluster core to increasingly star-forming, blue, disk-dominated populations at larger radii. We further introduce a three-dimensional framework in the joint (sSFR, (g-r) RS, B/T) space and develop a peak-based classification scheme that extends beyond traditional one-dimensional galaxy classifications. This framework identifies two dominant populations: red, quiescent, bulge-dominated galaxies, which account for 51% of the joint-analysis sample, and blue, star-forming, disk-dominated galaxies, which account for 29%. The remaining about20% of galaxies occupy transitional or mixed states that connect these two principal populations. The relative fractions of these populations change strongly near the splashback radius, where the red, quiescent, bulge-dominated population declines rapidly and the blue, star-forming, disk-dominated population becomes increasingly dominant. These results show that the splashback boundary is not only a dynamical boundary, but also a critical evolutionary transition zone. Overall, our findings suggest that galaxy evolution in Coma is not a purely binary transformation, but instead proceeds through continuous multidimensional pathways in which star formation quenching, color evolution, and morphological transformation occur on different timescales while remaining closely linked to the cluster dynamical structure.