Modelling globular clusters in the era of large space telescopes

by Arn Marklund (Université de Strasbourg)

Wednesday 4 Nov 2026, 13:30 → 14:30 Europe/Stockholm

FC 61 (AlbaNova Main Building)

Current and upcoming wide-field photometric surveys will observe hundreds of thousands of globular clusters (GCs), posing a key challenge: deriving reliable cluster properties using methods that scale to these data volumes. Meanwhile, GPU-accelerated direct N-body codes (e.g., NBODY6++GPU) now allow realistic simulations of >1M-star GCs. In principle, these models allow for direct one-to-one comparisons between simulated and observed clusters, but doing so in practice further requires methods that can generalise from limited data coverage. In this talk I will present a forward modelling algorithm, CREMANT, and a deep learning algorithm, pi-DOC. Combined, they allow us to transform simulation snapshots into multi-band photometric mock images, and to map a cluster's underlying properties (e.g., mass, age, ellipticity) directly from the image. Using archival HST data of the Andromeda galaxy, I will demonstrate the method’s reliability and scalability, and discuss its application to next generation wide-field surveys such as Euclid.