< Back to previous page

Project

Probabilistically modelling the birth and death of stars: changing the game with fast and precise emulation-enabled radiative cooling

Electromagnetic radiation provides an efficient heating and cooling mechanism that plays a crucial role throughout astrophysics. Radiative cooling critically determines, for instance, both the birth and the death of stars: radiative cooling dictates (1) whether a giant molecular cloud can cool enough during its gravitational collapse to reach a sufficient density for stars to form, and (2) whether the outer layers of evolved stars can cool enough to provide the conditions for dust condensation to occur and hence for a wind to be initiated. Despite the paramount importance of radiative cooling, it is often only included in an approximate way since a full treatment is computationally too expensive. Additionally, existing formulations all critically lack any kind of uncertainty quantification, making it hard to draw stringent conclusions. In this project, we will develop a probabilistic method to compute radiative cooling that naturally allows for uncertainty quantification. Furthermore, we will alleviate the computational burden with a hybrid solver/emulator that uses expensive cooling only on a small part of the model and leverages a trained emulator to infer the result on the remaining part. This allows us to model radiative cooling at unprecedented speeds and with uncertainty quantification. We will demonstrate our new cooling approach in two applications: by tracing the thermochemical structure of stellar winds, and by quantifying the stability of star-forming clouds.

Date:21 Oct 2021 →  Today
Keywords:Radiative Transfer, Uncertainty Quantification, Trained Surrogate Models (emulation)
Disciplines:Astronomy and space sciences not elsewhere classified, Astronomy and astrophysics
Project type:PhD project