Regularized Maximum Likelihood Techniques for ALMA Protoplanetary Disk Imaging

The application of Regularized Maximum Likelihood (RML) imaging techniques to sub-mm interferometric observations can achieve higher angular resolution and superior image fidelity compared to traditional imaging methods like CLEAN, as shown in the analysis of the recent Event Horizon Telescope images of M87. We use the GPU-accelerated open source Python package MPoL to explore the behavior of various RML prior distributions (maximum entropy, sparsity, total variation, and total squared variation) on ALMA observations in order to understand how each prior impacts the final image and in which imaging cases certain priors work best, both individually and in combination with each other. We focus our exploration on protoplanetary disk morphologies and identify which prior choices are likely to be suitable for certain types of features, such as sharply defined rings or diffuse emission. Using RML methods to improve the resolution of protoplanetary disk observations (both future and archival) will open the door for the detection of deviations in disk structure caused by embedded planets. In order to obtain a final image product with the best possible angular resolution and image fidelity, we also develop and distill recommendations for incorporating the image validation procedure cross-validation (CV) into the RML workflow, and provide a comparison of RML and CLEAN images for the protoplanetary disk HD 143006.