exoALMA XXIII. Estimating Disk and Planet Properties from Dust Morphologies with DBNets2.0

Abstract

The exoALMA large program provided an unprecedented view of the morphology and kinematics of 15 circumstellar disks, offering a biased but homogenous and well-characterized sample for population-level analysis. Continuum observations revealed numerous dust substructures, known to be potential signatures of embedded planets. We analyze the observed dust morphologies with the simulation-based inference tool DBNets2.0, assuming these are due to embedded planets at fixed locations, to infer the system properties. We estimate the putative planet mass, the disk $α$-viscosity, scale-height, and dust Stokes number that would reproduce 19 substructures in 13 of the 15 exoALMA disks. We compare our results with literature estimates derived with different methods, and find good agreement in most cases. We further explore the implications of the inferred disk properties for accretion, showing that for the Herbig stars in our sample, the implied viscous accretion timescales are too long to account for their observed stellar accretion rates. Regarding planet migration, our results favor inward migration, with only three putative planets expected to migrate outward. Finally, we check for correlations of the inferred disk and planet properties with the disks' gas-to-dust mass ratio, non-axisymmetry index, and masses of the gas, dust, and host stars, finding no remarkable trend.

Figures (15)

• Figure 1: Distributions of the four inferred properties for the full sample, excluding results considered uncertain or unconstrained. The shaded histograms show the medians of the individual posterior distributions, while the overlaid curves represent the stacked posteriors, normalized for comparison. Y-axis ticks refer to the histograms. Contributions from putative planets located in gaps (blue) and in cavities (pink) are shown separately.
• Figure 2: Overview of all inferred posterior distributions for the mass of the proposed planets (gray and red violins) and comparison with constraints provided in previous studies. The continuous error bars mark the 16th, 50th, and 84th percentiles of the respective distribution, while the dashed lines go from the 2.5th to the 97.5th percentiles. Red violins mark uncertain estimates. Right and left facing arrows mark, respectively, lower and upper limits, while crosses indicate estimates. Different colors are used to distinguish between the methods used to provide these constraints. HCI: high contrast imaging; SAM: sparse aperture masking; formula: empirical relationships with gap width and depth; simulations: numerical modelling of dust observations; kinematics (wake): estimates from localized kinematic perturbations in channel maps, tracing the planet wake; kinematics (rotation curves): estimates from gas pressure dips observed through rotation curves. Full references can be found throughout Sect. \ref{['sec:resump']} and Appendix \ref{['app:allsources']}.
• Figure 3: Inferred planet masses as a function of the putative planet's radial location. Blue points and histogram's bins mark DBNets2.0 planet mass estimates on the exoALMA sample presented in this paper. Respective disk names and assumed planet locations are listed in descending order according to the inferred planet mass. Asterisks mark uncertain or unconstrained estimates. Orange points and histogram's bins mark DBNets2.0 estimates on the larger sample of substructured disks considered in Ruzza2025DBNets2.0:Discs. Black points mark all confirmed exoplanets (data from https://exoplanet.eu).
• Figure 4: Comparison between the aspect-ratio estimates obtained with DBNets2.0 and the temperature-profile measurements reported by Galloway-Sprietsma2025ExoALMA.Structures. For this comparison, the midplane temperatures are converted into disk aspect ratios under the assumptions of vertical hydrostatic equilibrium and a vertically isothermal structure, consistent with the physical model used to train DBNets2.0. Faded points indicate sources with unreliable or unconstrained estimates. The red dashed line denotes perfect agreement between the two methods. The gray background marks parameter values that fall outside of DBNets2.0's scope.
• Figure 5: Degeneracy of DBNets2.0 planet mass estimate with the assumed planet radial location in the cavity of the J1604 disk. The overlayed orange lines mark the 16th, 50th and 84th percentiles of the inferred distribution. The vertical blue line marks the location of the putative planet assumed in this study. The vertical green dashed line indicates $R_\text{edge}/2$, where $R_\text{edge}$ is the radial location of the peak of the ring outside the cavity.