Presentation #102.351 in the session Poster Session.
Substructures such as gaps and rings are ubiquitous in protoplanetary disks. They can be induced by young planets in protoplanetary disks and used to infer the potential young planets’ properties. We developed convolutional neural networks (CNNs) to rapidly and directly infer the planet mass from radio dust continuum images. Hydrodynamical simulations have been used to study the relationships between the planet’s properties and these disk features. However, these attempts either fine-tuned numerical simulations to fit one protoplanetary disk at a time, which was time consuming, or azimuthally averaged simulation results to derive some linear relationships between the gap width/depth and the planet mass, which lost information on asymmetric features in disks.
Fitting 2D images directly using CNNs can solve these problems. We obtain planet mass and disk viscosity together with uncertainties as 0.16 and 0.23 dex, respectively. We can reproduce the degeneracy scaling α ∝ Mp3 found in the linear fitting method, which means that the CNN method can even be used to find degeneracy relationship. The gradient-weighted class activation mapping effectively confirms that PGNets use proper disc features to constrain the planet mass. We also provide programs for PGNets and the traditional fitting method from Zhang et al. online.