Seminars and Colloquia at ESO Garching and on the campus

The era of large surveys demands innovative approaches for the rapid and accurate analysis of vast spectral datasets. A promising direction to address this challenge lies in the synergy between physics-based simulations and machine learning. Astronomy, like many other scientific disciplines, relies heavily on simulations to encode our physical understanding of complex processes, generate hypotheses, design experiments, and explore causal relationships. Increasingly, this knowledge is embedded in forward models that produce realistic synthetic spectra, images, and data-cubes. A central and long-standing challenge, however, is the solution of the "inverse problem": inferring the underlying physical parameters from observational data. In this talk, I will present a powerful framework to tackle this problem based on simulation-based inference (SBI) accelerated by neural networks. I will show  preliminary results based on simulated spectra for upcoming instruments, along with ongoing SBI applications aimed at probing the interstellar medium in galaxies. Finally, I will discuss key open challenges, including model interpretability and domain adaptation, with particular emphasis on bridging the gap between synthetic training data and real observations.

Disk observations show ubiquitous substructure both at their surface and close to their midplane with gaps, rings, and spirals. While planets often explain such features, substructures are regions of local pressure maxima that can form dust traps, an essential initial condition for dust growth in line with planet formation theories. One proposed mechanism for generating these substructures is stellar-irradiation-driven shadowing, where vertical temperature perturbations cause the disk surface to puff up, casting shadows behind. We investigate the physical conditions for the existence of starlight-driven shadowing and test the viability of this "irradiation instability" argument with radiation hydrodynamical models using the PLUTO code. We also produce synthetic scattered light images with RADMC-3D to demonstrate observable signatures to such shadows and the disk’s vertical temperature structure. Our models underscore the importance of consistent modelling of dust dynamics for accurately resolving the irradiation absorption surface. We also present the multi species fluid dust method implemented in the PLUTO code, along with applications. Lastly, we present some preliminary results comparing different radiation methods used in the field.

Metal-poor galaxies provide a unique window into the physical conditions and chemical enrichment processes that govern star formation in nearly pristine environments. A subset of these systems exhibit spectra with extremely strong high-ionization emission lines that cannot be reproduced by standard stellar population models and, therefore, offer an ideal laboratory for testing the physical mechanisms that produce unusually hard ionizing radiation fields and extreme emission. These extreme emission line galaxies (EELGs) are often modeled under simplified assumptions, such as the low-density limit, and are widely used as benchmarks for interpreting elemental abundances and ionizing spectra across cosmic time. However, growing empirical evidence suggests that more extreme conditions at the heart of these sources are biasing our interpretations.

I will present new empirical methods to constrain the ionizing continua of EELGs from the JWST CLASSYIR Treasury Survey, which combines ultraviolet (UV) through mid-infrared emission lines to map the high-energy ionizing spectrum. These observations reveal radiation fields that are significantly harder and more structured than predicted by standard stellar population models, pointing to additional contributions from very massive stars, ultra-luminous X-ray sources (ULXs), and obscured AGN. At the same time, I will show that nebular conditions in these galaxies are far from uniform. Density stratification, particularly in highly ionized gas, can lead to systematic biases in temperature measurements and subsequent abundance determinations when using traditional low critical-density optical emission lines. As a result, even the long-standing “gold-standard” of metallicity measurements, the direct method, will be significantly biased in extreme environments.

Fortunately, UV diagnostics provide access to the densities and physical conditions of the high-ionization gas, enabling more robust determinations of temperatures and abundances. By combining UV and optical measurements, we can establish a physically consistent framework for interpreting local EELGs and connect them to high-redshift galaxies observed with JWST, which exhibit even more extreme ionization conditions, elevated densities, and enhanced N/O ratios. I will discuss the physical pathways that can drive rapid enrichment in relative abundances, and the implications for interpreting both local and distant galaxy populations.

Together, these results demonstrate that metal-poor EELGs expose the interconnected physics linking ionizing spectra, nebular conditions, and chemical enrichment across cosmic time, but only when interpreted with a self-consistent UV+optical framework.