M.Sc. defenses Vito Tuhtan, Mikkel Christensen, Rami Al-Belmpeisi

Title: Simulated Analogues using Deep Learning

Defenses:
13:00 - 14:15 Vito Tuhtan presents “The physical structure of young stellar environments"
14:15 - 15:30 Mikkel Christensen presents “Creating synthetic observations of young stars"
15:30 - 16:15 Rami Al-Belmpeisi presents “Deep learning to identify simulated matches to observations"

Abstract:
Over the past few years, it’s become increasingly evident that star formation is a multi-scale problem, and therefore only global simulations that properly account for the connection from the large-scale gas flow to the accreting protostar can be used to understand protostellar systems. At the same time long wavelength interferometers (ALMA, NOEMA) are able to make observations with tens of AU resolution for the nearest young stellar objects (YSO).

Using high resolution simulations and post processing methods, we aim to bridge the gap between simulations and observations of binary YSOs. Our goal is to create synthetic observations and perform a down-selection from large datasets of synthetic images to a handful of matching candidates in a semi-automatic way. From synthetic observations we infer the underlying physics that drives the creation and evolution of YSO binaries.

We simulate the creation and evolution of YSO binaries with RAMSES, a 3D MHD adaptive mesh refinement code. By post-processing the simulation data with the radiative transfer code RADMC-3D, we produce synthetic observations. We deploy Deep Convolutional Neural Networks (DCNN) which analyzes an observation and the synthetic images database to perform the down-selection and predict several system parameters for the observed system.
We apply our method on the observations of systems IRAS-2A to find their simulated analogues. We describe the chosen simulated analogues and analyse their respective pre-collapse environment, evolutionary stage, accreting disk properties.

Advisors: Troels Haugbølle, Rajika Kuruwita
Censor: Thomas Greve, Cosmic Dawn Center, Danmarks Tekniske Universitet (DTU)