DiskMINT-GARDEN: Self-consistent Models to Estimate Disk Masses

We present DiskMINT-GARDEN, a grid of self-consistent models together with a fast, open source inference tool for disk masses. The grid is built on DiskMINT, a tool which couples hydrostatic disk structure, continuum/line radiative transfer, and a reduced CO chemical network including freeze-out, grain-surface conversion, and isotope-selective photodissociation. DiskMINT-GARDEN model grid spans a large range of stellar mass ($0.1-2.0M_{\odot }$), gas disk mass (${10}^{-5}-{10}^{-1}{M}_{\star }$), dust-to-gas ratio ($0.003-0.1$), and characteristic radius ($10-300\mathrm{a}\mathrm{u}$), and provides synthetic ALMA observables. We train a machine-learning regression model to infer the disk mass, dust-to-gas mass ratio, and disk size from the dust continuum and ${\mathrm{C}}^{18}\mathrm{O}$ line observations. Applying DiskMINT-GARDEN to archival ALMA data of 34 disks, we find gas masses in good agreement with dynamical and HD-based estimates. Comparing our results with estimates from chemical modeling using DALI, we find that their need for large-scale elemental or CO depletion can be accounted for by grain-surface chemistry implemented in DiskMINT, with CO conversion to CO${}_2$ being one of the main reactions. Therefore, extant data suggest little chemical processing due to disk evolutionary processes.