Analyzing HST Fomalhaut b data using Non-negative Matrix Factorization

Originally believed to be an exoplanetary system, Fomalhaut A and its compation, Fomalhaut b, have undergone extensive investigation over the nature of the system. Prevailing theories support Fomalhaut b being a resultant dust cloud due to a collision between two planetesimals. To improve on exoplanetary imaging methods and further investigate Fomalhaut b, we decide to analyze the Fomalhaut system using nonnegative matrix factorization (NMF). Unlike previous examinations where point source-based post processing methods were used, NMF works best for detecting broad and spread sources such as disks and dust clouds. By using NMF, we work to examine Fomalhaut b and compare results from these previous works. Due to its iterative nature, NMF uses significant computing resources and compile times. To work around this, we implement Google's JAX NumPy and Just-In-Time (JIT) compilation to increase efficiency and speeds. By running code on accelerators (graphics processing unit) and JIT's ability to compile given functions all at once, this allows us to greatly improve on computing times. The data we use consists of different years of direct imaging of Fomalhaut through the Hubble Space Telescope coronagraph, that we then use NMF to process. After locating Fomalhaut b in these images, we then examine its make-up pixels for differences in relative size and brightness across different years.