Special Colloquium
Felix Glaser	SCHEDULED

Universität Würzburg

M 87’s nuclear region has been well observed by the GMVA and the EHT, hence its jet is theoretically the best constrained one, so far. I exploited this and compared a suite of GRMHD jet-launching simulations to the 86 GHz GMVA observation of M 87 by means of the jet-width measurements to that of the GRMHD simulations, down to the jet foot-point. In comparison, Cen A’s jet is far less constrained than that in M 87, due to varying inclination estimates, and an unresolved and optically thick nucleus in the mm-regime. However, like M 87 (and 3C 84), Cen A exhibits pronounced limb-brightening, with emission concentrated along the jet edges rather than the spine. We explored the physical origins of this phenomenon as a pilot study for the case of Cen A. The most natural explanation as we found with GRMHD simulations and advanced PIC-based GRRT calculations is that the electron momentum-distribution is not isotropic, as mostly assumed, but anisotropic. Furthermore, the implications of our model for future space-VLBI observations were investigated and will be discussed.

	Special Colloquium
Prof. Dr. Julie Hlavacek-Larrondo	SCHEDULED

Université de Montréal, Canada

The best place to study black hole feedback processes is in the hot atmospheres of galaxy clusters, which host the most massive black holes in the Universe and where we can directly image their impact on the surrounding medium. Yet major questions remain about how this black hole feedback operates, from gas cooling out of the hot atmosphere to ultimately fuelling the central black hole. I will first review our current understanding of this field and highlight recent observational advances that are revealing the complete self-regulated black hole feedback cycle. I will then focus on how machine learning and new data-analysis techniques can provide a complementary view of this feedback cycle. Using mock X-ray observations from the IllustrisTNG and TNG-Cluster simulations, we are exploring how information about the thermodynamic state of the intracluster medium can be extracted directly from these simulations. I will present recent work using deep learning and simulation-based inference, as well as new results showing that AstroCLIP, a foundation model trained on optical astronomical data, can be transferred to X-ray observations to predict cluster cooling times. Together, these studies illustrate the potential of combining observations, simulations and modern machine-learning techniques to extract the physics of black hole feedback from the rapidly growing datasets of current and next-generation observatories.