Reading between the rainbows

: comparative exoplanetology through molecule agnostic clustering

Abstract

Rocky exoplanets are faint and difficult to observe due to their small size and low brightness compared to their host star. Accurately characterising exoplanet atmospheres could offer insights not only into the planetary demographics of rocky worlds in the universe, but also how our own Earth compares. Previous work simulating the observational spectra of planets with various models of atmospheric composition has constrained conditions under which specific molecules would be observable. I seek a different approach that does not depend on specific molecular features, but rather on correlating "agnostic" spectral characteristics to each other. In this work, I detail the creation and usage of the "Molecule Agnostic Spectral Clustering" (MASC) method. I created this tool as a novel way to examine exoplanet spectra using comparative exoplanetology. This technique functions by comparing the profiles of planetary spectra to each other, based on the idea that a like profile implies a like composition. Using a combination of dynamic bandpasses to quantify the profile of spectra and the HDBSCAN clustering algorithm to identify groups within the data, planetary samples are found to be grouped according to their atmospheric traits. Individual planets can be characterised as belonging to one planetary subtype or another depending on where they are found on a MASC plot. MASC can additionally be used to differentiate between CO2- and O2- dominated atmospheres by combining multiple outputs. I then further used MASC clusters to attempt to create a taxonomy for exoplanet based on the patterns found during analysis. MASC can be used in most contexts for spectra with a signal-to-noise ratio as low as 3, provided an appropriate database of known samples is available to perform comparative analysis with. This makes the MASC tool effective for swift, "first-glance" analysis of incoming spectra from space telescopes such as JWST or HWO.

Date of Award	1 Dec 2026
Original language	English
Awarding Institution	University of St Andrews
Supervisor	Mark Claire (Supervisor) & Claire Cousins (Supervisor)