The Shape of the Vertical Action Distribution Locates the Scatterers that Heat the Galactic Disc

The paper investigates what mechanism heats the Milky Way's stellar disc vertically, by analyzing the shape of the distribution of vertical action (Jz) in a coeval stellar population. The central theoretical result is that the shape of this distribution records where along a star's orbit the scattering occurs. The authors derive that for any height distribution of scatterers, the distribution of vertical action is p(Jz) ∝ exp[−(Jz/J0)(2−b)], where b is the logarithmic slope of the diffusivity in action, D ∝ Jz b. Scatterers filling the volume give b = 1 and an exponential distribution, while scatterers confined to a thin midplane layer give b = 1/2 and a sharper, three-halves stretched exponential cutoff.

The paper fits this model to 7589 low-α red clump stars from APOGEE with Gaia astrometry, between 5 and 10 kpc and 2 and 8 Gyr old, leaving the heating history free. The measured exponent is b = 0.51+0.06−0.07, consistent with the thin-layer prediction but not the volume-filling one. The authors state: "Measured on 7589 low-α red clump stars from APOGEE with Gaia astrometry, the scale left free at every age, b = 0.51+0.06−0.07. A molecular layer of the measured thickness predicts 0.44–0.54 once its flaring, the encounter-speed dependence and a realistic vertical potential are all included. Scatterers filling the volume predict 0.85–1.00 under the same treatment."

Comparing the 2–4 Gyr heating amplitude with the present molecular surface density gives an effective scatterer mass of 2.7 × 10 6 M⊙. The paper notes: "Dividing by the observed molecular surface density gives an effective scatterer mass of 2.7 × 10 6 M⊙ from the combined 2–4 Gyr sample. Cloud catalogues and mass functions give 1.4–3.4×10 6 M⊙ for the same quantity, a factor of 2.5 among themselves, and the stellar value falls inside that range." Older stars have experienced more of the Galaxy's gas-richer past; correcting for that history brings all four age bins to 1.9–2.8 × 10 6 M⊙, inside the range cloud catalogues and mass functions give.

The paper concludes: The Milky Way's disc is heated near the plane, by an evolving population of objects of giant-molecular-cloud mass. It also rules out volume-filling scatterers such as bending waves, dark substructure, or 10 6 M⊙ black-hole haloes as the main vertical heating agent, since these would give b = 1. The paper explains that a sample without ages would return b ≃ 1 even when the truth is 1/2, due to mixing of populations of different ages, which explains the classical b = 1 result of Binney & Lacey (1988). The free scales grow with stellar age as J0 ∝ τ 1.12, indicating the disc heated faster in the past than a constant rate allows.

Improvements to AI Systems:

1. Physics-Informed Generative Models for Stellar Dynamics:

• Train an AI to generate synthetic stellar phase-space distributions (e.g., actions, positions, velocities) that explicitly incorporate the derived functional form p(J z) proportional to [-(J z/J 0) 2-b] with b as a free parameter.

• The AI can then infer b and J 0(tau) from noisy, incomplete observational data (e.g., APOGEE + Gaia) using Bayesian neural networks or normalizing flows, directly embedding the theoretical prediction into the likelihood.

2. Hierarchical Bayesian Age–Action Decomposition:

• Build an AI that jointly models stellar age, metallicity, and vertical action distributions without assuming a fixed heating history.

• The system can deconvolve mixed-age populations to recover the true b even when age information is sparse or uncertain, avoiding the classical bias toward b=1 from population mixing.

3. Automated Discriminant Analysis for Heating Mechanisms:

• Develop a classifier that, given a measured b and its uncertainty, automatically distinguishes between thin-layer scatterers (e.g., giant molecular clouds, b about 0.5) and volume-filling scatterers (e.g., dark substructure, b about 1).

• This AI can flag inconsistencies in future surveys (e.g., Gaia DR4, 4MOST) and prioritize follow-up observations for anomalous regions.

4. Time-Dependent Diffusivity Inference from Action Distributions:

• Use a recurrent or transformer-based AI to infer the time evolution of the diffusion coefficient D(J z, tau) from the age-dependent scale J 0(tau) proportional to tau 1.12.

• The AI can reconstruct the Milky Way’s past gas surface density and star formation history, providing a self-consistent model of disc heating over cosmic time.

5. Simulation-to-Observation Transfer Learning with Physical Priors:

• Train an AI on high-resolution N-body/hydrodynamic simulations of disc galaxies (e.g., FIRE, EAGLE) to predict b as a function of scatterer properties (mass spectrum, scale height, flaring).

• Then fine-tune on real data to infer the effective scatterer mass (e.g., 2.7 times 10 6 M) and its evolution, with uncertainty propagation from cloud catalogues and mass functions.

6. Uncertainty-Aware Extrapolation for Exoplanet and Galactic Habitability Models:

• Use the derived heating law to predict the vertical velocity dispersion of stars in other galaxies or in the Milky Way’s future, enabling AI-driven models of planetary system stability and habitability as a function of galactic environment.

7. Real-Time Anomaly Detection in Large Spectroscopic Surveys:

• Implement an AI that monitors incoming stellar catalogs for deviations from the predicted p(J z) shape, flagging potential systematics, unresolved binaries, or new heating mechanisms (e.g., satellite perturbations) in real time.

What the Improved AI System Can Do:

• Accurately measure the vertical heating exponent b from any stellar sample, even with age uncertainties or incomplete phase-space coverage.

• Distinguish between competing heating mechanisms (molecular clouds vs. dark matter substructure) with quantifiable confidence, using only kinematic data.

• Reconstruct the Milky Way’s past gas content and star formation history from present-day stellar actions.

• Provide robust predictions for stellar velocity dispersions in external galaxies, aiding in the interpretation of IFU surveys (e.g., MaNGA).

• Automatically correct for population-mixing biases that have historically plagued action-space analyses.

The Milky Way's stellar disc is thicker than the cold gas layer from which its stars form. What scatters stars onto orbits with greater vertical motion remains unresolved. Scatterers could fill the disc volume, as bending waves or dark substructure would, or could be confined to the midplane, as giant molecular clouds are. Scattering from a midplane layer occurs only during the fast plane-crossing phase and becomes less effective as that speed increases, whereas a volume-filling perturbation remains effective near the slow turning points. We derive the distribution of vertical action this leaves behind, for any height distribution of scatterers. The geometry turns out to enter only through the logarithmic slope of the diffusivity in action, D proportional to J z b, and solving the Fokker-Planck equation gives p(J z) proportional to [-(J z/J 0) 2-b]. Scatterers that fill the volume give b=1 and an exponential, a thin layer at the midplane gives b=1/2 and a sharper cutoff: the shape of p(J z) records where the scatterers sit, the growth of its scale how strongly they scatter. We fit this model to 7589 low- alpha red clump stars of Ting & Rix (2019) between 5 and 10 kpc and 2 and 8 Gyr old, leaving the heating history free. This yields b=0.51+0.06-0.07, consistent with the thin-layer prediction but not the volume-filling one. Comparing the 2-4 Gyr heating amplitude with the present molecular surface density gives an effective scatterer mass of 2.7 times10 6,M. Older stars have experienced more of the Galaxy's gas-richer past; correcting for that history brings all four age bins to (1.9-2.8) times10 6,M, inside the range cloud catalogues and mass functions give. The Milky Way's disc is heated near the plane, by an evolving population of objects of giant-molecular-cloud mass.