RSAA Colloquium: Ian Czekala (CSIRO)

Imaging the Planet Forming Environment with Differentiable Forward Models

Credit: https://github.com/iancze

In recent high angular-resolution observations of protoplanetary disks, astronomers have discovered radial and azimuthal substructures like gaps, rings, and spirals visible in both dust continuum and spectral line emission. 

In the first half of this talk, I will highlight the observational results from the MAPS and exoALMA recent large programmes on the Atacama Large sub/Millimeter Array (ALMA), which have given us the best look yet at the chemistry and kinematics of the planet-forming environment across a sample of protoplanetary disks. 

In the second half of the talk, motivated by the complexities inherent to ALMA spectral line datasets, I will discuss our work to build differentiable forward models for radio interferometric datasets. Neural network components provide a rich set of modular and composable building blocks that can be used to express the physical relationships between latent model parameters and observed data, including parameters which might otherwise be restricted to a one-time calibration procedure prior to image synthesis.