Giant Exoplanets, Sirius, and Starlight Subtraction at Scale

Abstract

In a few short decades, we have learned that planets are not only possible but common around other stars. However, the mass and semi-major axis parameter space for which we can directly detect planets through high-contrast imaging remains largely non-overlapping with those of the transit and radial velocity methods that discovered the majority of these planets. In this dissertation, I present thermal-infrared (λ = 3.95 μm) coronagraphic imaging of Sirius through a vector-apodizing phase plate--180º coronagraph (gvAPP-180) with the MagAO/Clio instrument in search of exoplanets. The age and temperature of Sirius mean the contribution of instellation would significantly elevate any exoplanetary companion's effective temperature, and thus our ability to detect it in the infrared. Using over 10,000 frames of Sirius data as a benchmark, I show that co-adding (temporal downsampling) of high-contrast imaging data results in worse contrast limits at small separations. To enable analysis of fully-sampled data, I develop two new algorithms to achieve the optimal contrast: first, the KLIP downdate, which reduces the algorithmic complexity of KLIP (Karhunen-Lo\`eve Image Projection) from O(N^4) to O(N), resulting in a practical speed-up of 100x in benchmarks. Second, the PCA-Transpose or PCAT algorithm, which transposes starlight subtraction into the space of eigen-time-series to reduce self-subtraction and reduce computation time on large datasets. I discuss the optimal choice of hyperparameters for starlight subtraction, and the Bayesian optimization scheme I used to optimize the recoverable signal strength at every point. After optimization, I achieve post-processed contrast limits of 1.5 x 10^-6 to 9.8 x 10^-6 outside of 0.75 arcsec which correspond to planet masses of 2.6 to 8.0 M_J. These are combined with values from the recent literature of high-contrast imaging observations of Sirius to synthesize an overall completeness fraction as a function of mass and separation. After synthesizing these recent studies and our results, the final completeness analysis rules out 99% of ≥9 M_J planets from 2.5–7 AU. The future of exoplanet direct imaging lies in pushing towards bluer wavelengths and deeper contrasts to enable reflected-light imaging and—some day—directly image an exo-Earth. Towards this goal, I also present work I have done to develop data infrastructure and calibrations of the MagAO-X focal plane. MagAO-X is an extreme adaptive optics (ExAO) system for the Magellan Clay 6.5-meter telescope, but it is also a testbed for novel adaptive optics and coronagraphy techniques, and a pathfinder for ExAO on the Giant Magellan Telescope and the other extremely large next-generation telescopes.