While the Event Horizon Telescope (EHT) has given us revolutionary images of supermassive black holes in M87* and Sagittarius A*, the swirling plasma flows just inside those glowing rings remain hidden from direct view. A new preprint takes us deeper into that hidden realm. By slicing through a standard 3D general-relativistic magnetohydrodynamic (GRMHD) simulation and focusing on the equatorial plane, Argyrios Loules and collaborators reveal that magnetically arrested disks (MADs) undergo dramatic episodes where non-axisymmetric structures swell dramatically during magnetic flux eruption events. These bursts of activity are not minor turbulence; they are dominated by large-scale azimuthal modes, primarily m=1 and m=2, creating continent-sized plasma concentrations that span significant fractions of the orbit near the horizon.

This work, which remains a preprint and has not yet completed peer review, builds on more than a decade of MAD theory pioneered by researchers such as McKinney, Tchekhovskoy, and Narayan. Previous simulation studies often emphasized poloidal magnetic fields, jet launching, or time-averaged radial profiles. What this analysis makes clear—and what much early coverage of EHT results missed—is the explicit azimuthal morphology and its evolution. Popular descriptions of EHT data frequently discuss a smooth, crescent-shaped ring. Yet the inner accretion flow is anything but smooth during these eruptions. The simulation shows that horizontal field lines in low-density equatorial regions reconnect into vertical bundles. These bundles, filled with tenuous plasma, detach from the horizon and are propelled outward by magnetic buoyancy, carving out the large angular-scale features the Fourier analysis detects.

The methodology is purely numerical: the team post-processes data from one high-resolution 3D GRMHD run, extracting equatorial slices and decomposing the density and magnetic-field distributions into azimuthal Fourier modes. There is no observational 'sample size' in the astronomical sense; instead, the simulation provides millions of computational cells tracked over thousands of dynamical timescales near the innermost stable circular orbit. Limitations must be kept in mind: the run assumes ideal MHD without explicit radiative transfer, electron heating, or non-thermal particles that would directly influence what telescopes actually see. It also represents one specific set of initial conditions; real astrophysical disks may explore a broader parameter space of spins, magnetic fluxes, and inclinations.

Synthesizing these results with two key observational and theoretical pillars sharpens the picture. The landmark 2019 EHT papers (Astrophys. J. Lett. 875 L1 and related works, arXiv:1906.11238) demonstrated that the emission ring is brighter on one side, consistent with Doppler boosting from orbital motion. Yet those papers noted that time variability, especially the rapid flaring of Sgr A*, required additional explanation. Later simulation libraries, such as those compared in the 2020 Event Horizon General Relativistic Magnetohydrodynamic Code Comparison Project (arXiv:1909.14904) and subsequent MAD-focused papers (e.g., arXiv:2110.01509 by Wong et al.), showed that flux eruptions can drive powerful outflows. The new preprint supplies the missing quantitative link: the low-m modes quantify exactly how 'lumpy' the disk becomes, with the m=2 mode often appearing as two opposing dense regions that could produce the quasi-periodic brightness oscillations reported by EHT monitoring campaigns.

The physical analogy is striking. The reconnection and buoyant rise of vertical flux tubes mirrors solar coronal mass ejections, but scaled to gravitational radii around a spinning black hole. This process allows the black hole to reject excess magnetic flux that would otherwise choke accretion, a regulatory mechanism whose observational imprint may already be visible as rapid changes in millimeter flux and polarization. What earlier coverage often got wrong was treating the accretion flow as essentially axisymmetric or dominated by small-scale turbulence; the simulation demonstrates that during eruptions the largest-scale modes prevail close to the horizon, exactly where EHT is most sensitive.

These dynamic azimuthal structures therefore do more than decorate the disk—they likely dictate the observable light curves and polarization patterns that next-generation EHT arrays with improved time resolution will map. The findings tighten the connection between theoretical MAD states and real extreme astrophysical phenomena, suggesting that what looks like stochastic flickering in our data may actually be the predictable choreography of reconnecting, buoyant magnetic dragons being sloughed off by the black hole.