The plasma falling into M87 and Sgr A is too thin to collide with itself. In a normal accretion disk — the kind powering bright quasars — particles smash into each other constantly, exchanging energy and momentum, maintaining something resembling thermal equilibrium. Magnetohydrodynamics (MHD) describes this well: treat the plasma as a conducting fluid, solve the Navier-Stokes equations coupled to Maxwell's equations, and the results match observations.
In low-luminosity systems, this approximation fails. The accretion rate is so low that the mean free path between collisions exceeds the size of the disk. Individual particle orbits matter. Pressure becomes anisotropic — different in different directions — because nothing forces it toward isotropy. The plasma is collisionless, and MHD should not work.
Mehlhaff et al. present the first fully kinetic simulations of collisionless accretion with finite angular momentum onto a spinning black hole. Kinetic means they track the distribution function of the plasma — not bulk fluid quantities, but the full six-dimensional phase space of positions and velocities. This is enormously more expensive than MHD, but it captures physics that MHD cannot: pressure anisotropy, kinetic instabilities, non-thermal particle distributions.
The surprise is that the collisionless accretion flow looks remarkably like its fluid counterpart. The plasma self-organizes into a magnetically arrested disk (MAD) state — the configuration where accumulated magnetic flux near the black hole regulates the accretion rate. Kinetic instabilities grow fast enough to regulate pressure anisotropy, effectively restoring the fluid-like angular momentum transport that MHD assumes but that has no microscopic justification in a collisionless system.
The distinction matters for jets. In the Blandford-Znajek mechanism, spinning black holes extract rotational energy through magnetic fields threaded through the ergosphere, launching relativistic jets. This requires plasma in the jet funnel to carry current. The kinetic simulations show that accreting material penetrates the funnel inefficiently — pair creation (electron-positron production from photon-photon collisions near the horizon) may be necessary to supply the charge carriers that sustain the jet.
The practical consequence: MHD simulations of M87 and Sgr A have been producing results that look right for reasons that are now better understood. The kinetic instabilities that restore fluid-like behavior are fast enough to justify the fluid approximation, at least for bulk dynamics. The differences emerge in the jet — where the collisionless nature of the plasma may limit the jet power in ways that MHD cannot capture.