Modeling protoplanetary disk SEDs with artificial neural networks. Revisiting the viscous disk model and updated disk masses

Vol. 642
6. Interstellar and circumstellar matter

Modeling protoplanetary disk SEDs with artificial neural networks. Revisiting the viscous disk model and updated disk masses

by A. Ribas, C. C. Espaillat, E. Macias, and L. M. Sarro, 2020, A&A, 642, A171

There are still several open questions in our understanding of planet formation, including the main mechanism responsible for angular momentum transport in protoplanetary disks or their apparent low masses compared to those of exoplanetary systems. Investigating these types of issues requires disk models that account for many physical processes, which also makes them computationally demanding. As a result, modeling large samples of disks can usually only be achieved with simpler models. In this work, Ribas et al. combine physically motivated viscous disk models with artificial neural networks to drastically speed up their computation times, which allows for modeling of spectral energy distributions of protoplanetary disks with detailed models using a Bayesian framework. Applying this approach to 23 protoplanetary disks in the Taurus-Auriga star-forming regions yields high viscosity values for many of them, which is in contradiction with recent ALMA results. This suggests that mechanisms other than viscosity could play a significant role in the transport of angular momentum in disks. The disk masses derived using full radiative transfer models are also systematically higher than those derived from (sub)mm fluxes alone, which could help alleviate the observed discrepancy between the masses of disks and exoplanetary systems.