How does a spinning black hole launch a mildly relativistic outflow? Our 3D general-relativistic MHD simulations of accretion flows threaded by magnetic loops of alternating polarity, at black-hole spins a = 0, 0.5 and 0.9, expel discrete, sub-relativistic ejections, a mechanism that could be responsible for the delayed radio counterparts of tidal disruption events.

Spin is necessary: the non-rotating run drives no unbound outflow, while both rotating runs expel ejections through 50 r_g roughly every 500 M. The alternating polarity keeps the horizon flux from saturating and fills the disk with current sheets. The outflow is driven thermally and kinetically rather than magnetically, which bounds its terminal velocity at β ≤ 0.21.

Push poloidal flux inward and it eventually pushes back: flux piles up on the horizon until magnetic pressure balances the infalling gas, and accretion leaks through only in clumps. In 3D the flow never arrests completely, interchange instabilities keep feeding a persistent jet at η ≈ 140%.

Feed the disk alternating magnetic loops and it migrates between arrested states as loops reconnect near the horizon. Even an arrested funnel is not static: flux builds until it reorganizes in a burst, and the jet stutters.

Three-dimensional GRMHD simulations follow a magnetized jet as it propagates through the ejecta of a binary neutron-star merger. The outflow that breaks out has a hollow core of about 4° and an opening angle of at least 10°, confirming earlier 2D results, while the ejected matter contributes to the kilonova.

Modelling the non-thermal afterglow of the jet and fitting it to the observations and the superluminal motion of GRB 170817A gives an observer angle of about 35.7°. The same work computes the afterglow of the ejecta itself, and whether the matter behind the thermal kilonova can also produce detectable non-thermal emission.

In accretion flows threaded by several magnetic loops, current sheets form between loops of opposite polarity and tear into chains of plasmoids, compact magnetized islands where particles are accelerated.

A SANE multi-loop GRMHD model of Sgr A*, with no stable jet, lets us locate these current sheets and regions of strong magnetic turbulence and compute 230 GHz and 2.2 μm light curves from non-thermal particles. The resulting near-infrared flares match the observed flare durations and spectral index behaviour, linking Sgr A* flares to particle acceleration in plasmoid chains and at the disk boundaries close to the black hole.

Drive more flux through the funnel than a standard MAD can hold and magnetic pressure governs the entire inner flow. Ṁ onto the horizon collapses by orders of magnitude while the jet power barely moves.

Sustained for about 10,000 dynamical times, this magnetically arrested state globally suppresses accretion, and the jet power exceeds the energy supplied by accretion by more than a factor of 100, an efficiency not reached in earlier simulations.

ASKAP J1832−0911 pulses in radio and X-rays for two minutes every 44 minutes, too slow for an ordinary pulsar, too regular to be random flaring. Tilt a jet away from the black hole spin axis and frame dragging makes it precess like a slow gyroscope; every sweep across our line of sight is a pulse.

With the Lense–Thirring relation m r³ = 1.703 × 10⁸ a an intermediate-mass black hole (M ~ 10³–10⁵ M_☉) reproduces the 44-minute period at r ~ 5–30 r_g. The period is geometry, not a clock.