Starlight spectra

The evolution of massive stars to- ward their supernovae and remnant compact objects is poorly understood, especially for very massive stars with initial masses >100 solar masses (e.g., Brands et al. 2023). How do these stars evolve? What are their supernova progenitor properties and resulting supernova characteristics? How massive are the black holes these stars leave behind? We explore these questions by analysing the emitted light of stars, which allows us to characterize their properties. In turn, confronting stellar evaluation models with these measured properties of many stars helps us constrain the physics that controls the lives of stars, hence furthers our understanding of the way stars evolve and end. Progress is severely hampered by the complexity and CPU-intensive nature of spectral analysis, a problem that is particularly urgent given upcoming large-scale spectroscopic surveys, which will produce thousands of spectra of massive stars in galaxies in the Local Group and beyond. In this MSc project, we will develop a neural network (NN) approach to predict the spectra of massive stars. Exploratory NN-predictions of the spectra of these stars considering five stellar properties have yielded very promising results. Higher dimensional analysis (considering many more stellar properties) now needs to be developed, tested, explored and optimized. Developing a versatile and robust NN methodology for massive stars in the range 8-300 solar masses is the goal of this project. This is a great project for a student who’s excited about stars and interested in delving deeper into how we can use machine learning in astrophysical modelling, and would like to tinker with some neural networks.