Description
Molecular clouds are the cradle of stars: comprehending their composition and evolution is key to understanding the processes and materials that go on to form stars and planets. Ices make up 90% of the condensible molecular reservoir in such clouds and can only be observed as absorption features against a background stellar continuum. With thanks to its spectral resolution and sensitivity, JWST has provided excellent observations so far. To best analyse this data, it is critical to accurately model the background source continuum to then extract the superimposed absorption spectra of ices.
Sophisticated stellar atmosphere models exist, but they are often computationally expensive and time-consuming to fit to the observations. Consequently, for efficiency, it remains common practice to approximate the continuum spectra of background stars using simple polynomial functions or Gaussian Process (GP) regression. Over-simplified or poorly fit stellar models can introduce biases, adversely affecting the extraction of interstellar ice characteristics.
To solve the issue of underfitting the observations while keeping a reasonable run-time, I present a novel method for fitting model background spectra to tens of thousands of existing and future observations across multiple molecular clouds. I will focus on streamlining the model-fitting process to efficiently manage large volumes of data, and improving reliability of background source characterisation and determination of relevant parameters. I will describe using GPs to build a continuous grid of model spectra, and Monte Carlo algorithms to sample and fit observational data.
