Inferring Black Hole Physics through Image and Video Reconstructions with the Event Horizon Telescope

This thesis focuses on inferring black hole physics through the development and application of advanced image and video reconstruction techniques using data from the Event Horizon Telescope (EHT). The EHT, a global Very Long Baseline Interferometry (VLBI) array, achieves unprecedented angular resolution by operating at millimeter wavelengths (primarily 1.3 mm and 0.87 mm), enabling the horizonscale study of supermassive black holes (SMBHs) like M87* in the galaxy M87 and Sagittarius A* (Sgr A*) at the center of our Milky Way. The research presented herein aims to extract quantitative astrophysical information from EHT observations, test predictions of General Relativity (GR) in the strong-gravity regime, and provide insights into the complex plasma dynamics governing accretion near these extreme objects. A core component of this research involved imaging the April 2018 EHT observations of M87* utilizing the Bayesian interferometric modeling framework Comrade. These analyses confirmed the persistence of the iconic asymmetric ring-like emission structure, consistent with theoretical expectations of the shadow cast by a Kerr black hole of approximately 6.5 × 109 solar masses (M⊙ ). The measured ring diameter of ∼ 43.3+1.5 −3.1 μas was found to be remarkably consistent with the 2017 EHT results, showing the stability of this gravitational feature. Moreover, a significant counter-clockwise shift in the position angle of the ring’s brightness asymmetry by approximately 30◦ was observed between the 2017 and 2018 epochs. This evolution is attributed to the turbulent dynamics within the accretion flow, a phenomenon seen in General Relativistic Magnetohydrodynamics (GRMHD) simulations. The robust uncertainty quantification provided by the Bayesian approach was crucial in assessing these features and investigating the challenges in constraining parameters like the compact flux density of M87*. Further investigation into the M87* images focused on quantitatively characterizing and understanding the origin of the observed ring ellipticity. Leveraging the enhanced (u, v)-coverage of the EHT 2018 array, which included the Greenland Telescope (GLT) for improved North-South baseline sampling, a consistent ellipticity (τ = 0.08+0.03 −0.02) and major axis position angle (ξ = 50.1+6.2 −7.6 degrees East of North) were measured across five independent imaging and modeling methods. To interpret this measured ellipticity, an extensive comparison was performed against a large library of GRMHD simulations, spanning a wide range of physical parameters such as black hole spin, electron distribution functions (thermal and non-thermal), magnetic field configurations (Magnetically Arrested Disk – MAD, and Standard and Normal Evolution – SANE), and ion-to-electron temperature ratios. This analysis revealed no statistically significant correlation between the black hole spin and the observed image ellipticity. Instead, the ellipticity was found to correlate with the fraction of non-ring emission, particularly in non-thermal models and those exhibiting higher jet emission. These results compellingly suggest that the ellipticity measured in the current EHT images of M87* is predominantly governed by astrophysical effects within the turbulent accretion flow and jet base, rather than being a direct manifestation of the gravitational lensing of an intrinsically elliptical photon ring due to black hole spin. Isolating purely gravitational contributions to ellipticity will necessitate future observations with enhanced capabilities or different observational strategies. The next frontier in black hole physics is to obtain the first event-horizon-scale videos, particularly of rapidly variable sources like Sgr A*. Addressing the distinct challenges posed by imaging such dynamic SMBHs, this thesis introduces ehteval, a rigorous and systematic software framework developed for the validation and evaluation of horizon-scale video reconstructions. The development of ehteval is particularly relevant now as the EHT collaboration works towards producing these first videos, making robust validation tools essential. This framework utilizes realistic synthetic EHT datasets generated from a comprehensive library of ground truth models, including static geometric models, dynamic geometric models with diverse kinematic and polarimetric properties, and GRMHD simulations. A comprehensive suite of evaluation metrics was developed and validated to quantitatively assess reconstruction fidelity across multiple domains. These metrics probe the goodnessof- fit to interferometric data products (e.g., closure phases, log-closure amplitudes, complex fractional polarization), the morphological accuracy of total intensity and linear polarization structures in both static (time-averaged) and dynamic (timevariable) components, using normalized cross-correlations; and the ability to recover specific dynamical features such as hotspot parameters (size, position, flux) and rotational pattern speeds. The ehteval framework is foundational for interpreting the first EHT video reconstructions of Sgr A*, and ensuring the robustness of scientific conclusions drawn from these complex datasets. Collectively, this thesis bridges the gap between raw interferometric observables and interpretable physical quantities by combining rigorous image modeling with feature extraction and simulation comparison. These tools offer a principled foundation for interpreting horizon-scale observations of black holes and provide the groundwork for exploiting future expansions of the EHT, including Space VLBI missions capable of resolving fine structures such as photon subrings. By enabling reliable characterization of dynamic and polarized features in black hole environments, this work not only strengthens our ability to constrain accretion physics and spacetime geometry but also contributes to the development of next-generation techniques in computational astrophysics.