The battleground is subauroral geospace, the poorly charted zone lying equatorward of the bright auroral oval, bounded in the magnetosphere by the inner edge of the hot electron plasma sheet and the earthward tip of the ring current, the torus of energetic ions that swells around Earth during storms. In quiet times this region hosts the main ionospheric trough, a thinning of plasma just below the aurora, and the dim SAR arcs: subvisual 630-nanometer emissions of atomic oxygen, at intensities of 100 to 500 Rayleigh from altitudes above 300 kilometers, which require electron temperatures exceeding roughly 3,000 kelvin. When a substorm fires, however, latitudinally narrow channels of westward flow — subauroral ion drifts, or SAIDs, near midnight, and the broader subauroral polarization streams, or SAPS, at dusk — tear through the ionosphere. Because magnetized plasma drifts at the velocity E×B, poleward electric fields of 100 to 350 millivolts per meter drive ions at up to four kilometers per second, bundling cross-channel voltages of 30 to 40 kilovolts into channels barely one degree of latitude wide. Nearly every fast channel is later found on the poleward wall of the trough, at the ionospheric footprint of the plasmapause.

The classical account rests on the current-generator model advanced by Anderson and colleagues in 1993. Downward Region 2 field-aligned currents, generated by the pressure of the asymmetric ring current, flow into the low-conductance subauroral ionosphere, close through Pedersen currents, and thereby build the poleward electric field that maps outward into the magnetosphere. Because the ring current grows slowly as ions drift westward around the planet, the model predicts flows appearing roughly an hour after substorm onset, with particle boundaries smeared by energy-dependent drift. An earlier voltage-generator model of Southwood and Wolf instead predicted channels that start narrow at onset and widen with time. Satellites and radars contradict both: flows erupt within minutes of breakup, channels contract while preserving nearly constant voltage, and the hot electron plasma sheet terminates in a razor-sharp, energy-independent boundary unlike the broad, dispersive edge that test-particle physics demands. Early attempts to reconcile theory with CRRES satellite data had to invoke a rotational discontinuity tens of ion gyroradii wide sweeping inward at ten kilometers per second — assumptions untestable with a single spacecraft and, as Cluster later showed, wrong by an order of magnitude, for the boundary creeps along at merely half to eight-tenths of a kilometer per second.

The new framework begins in the magnetotail, where mesoscale plasma flows — bundles of kilovolt plasma pushed earthward at speeds of order 150 kilometers per second — race toward the inner magnetosphere, heralded by the auroral streamers that skywatchers chase. When such a flow strikes the plasmapause, the outer boundary of the cold, dense plasmasphere, its ions outrun its electrons and pile up positive polarization charge at the front. There the cold plasma short-circuits that charge: the flow’s electrons halt abruptly, carving the dispersionless plasma-sheet boundary that satellites observe, while the ions — the leading edge of the ring-current injection — keep penetrating. Plasma turbulence, dominated by lower hybrid waves, scrambles the ions’ orbits until they become effectively unmagnetized and slip across magnetic field lines deep into the plasmasphere, stopped only when a self-consistent electric potential builds ahead of them. The result is what the authors call a fast-time voltage generator: the SAID channel stretching between the electron stopping point and the nose-shaped tip of the freshly injected ring current. In this picture the plasmapause acts as a power plant, converting the kinetic energy of incoming flows into mesoscale electromagnetic energy.

The evidence is forensic. On 18 April 2005, minutes after a weak substorm began, the four Cluster probes crossing the near-equatorial magnetosphere found a channel whose outer boundary sat exactly where the cold plasma density climbed past a critical five to seven particles per cubic centimeter — and where one-kilovolt electron counts collapsed to zero. Between successive crossings the boundary marched inward at 1.8 kilometers per second, yet the cross-channel voltage held near 8 kilovolts, the signature of a constant-voltage source, while a ground-based radar watched the flow contract within two minutes. On 22 April 2004 a DMSP satellite crossed a SAID just six minutes after substorm onset, moments after dispersionless particle injections appeared at geosynchronous orbit. During a substorm on 6 May 2003, Cluster and DMSP together timed duskside ring-current injections that raced three hours in magnetic local time in under twenty minutes — far too fast for drifting ions — and THEMIS later caught a 150-kilometer-per-second earthward flow triggering the breakup that preceded similar injections. Magnetically conjugate passes show electric field profiles matching within a few percent when mapped along field lines, and the small-scale field-aligned currents that should span the whole channel in the current-generator picture instead cluster tightly at its poleward edge.

Duskside flows obey a different engine: the two-loop substorm current wedge, or SCW2L. As the aurora’s westward-travelling surge — a front roughly 100 kilometers wide sweeping duskward at three to ten kilometers per second — crosses a magnetic meridian, the upward Region 1 current at its head demands a compensating downward Region 2 current. That response drives the freshly injected ring-current ions sunward, capturing particles that gradient-curvature drift would leave creeping westward, while the Pedersen current closing the two loops amplifies the subauroral electric field. The front moves too fast for a steady state — several Alfvén bounce periods, minutes in total, would be needed for one — so broadband ultra-low-frequency pulsations are inherent to the circuit, and similar current loops even paint a dawnside counterpart across the post-midnight sky. The auroral and subauroral halves of a substorm, in short, form one electromagnetic machine whose intermittent bursts, recurring every five minutes or so, imprint themselves as irregular, wave-like structure on the flows below.

The same framework resolves a long-standing mystery of the red arcs. In the classical picture, SAR arcs shine because Coulomb collisions between ring-current ions and plasmaspheric electrons heat the electron gas that excites the 630-nanometer oxygen line. That process is orders of magnitude too slow to explain arcs that flare, bead, or jump position during substorm expansion — including the spectacular ten-kiloray arc photographed over Millstone Hill in October 1991 above 6,000-kelvin electrons. In the turbulent plasmasphere boundary layer, lower hybrid waves with measured amplitudes near one millivolt per meter exceed several-fold the threshold for chaotic ion orbits, and their Landau resonance with electrons yields suprathermal energy fluxes of a few times ten billion electron-volts per square centimeter per second — enough to push ionospheric electrons past 6,000 kelvin within minutes. Even the beaded green-red arcs, whose 557.7-nanometer emission demands excitation above 4.2 electron-volts, cannot be produced by collisions at all. Wave heating, not collisions, is the arc’s engine, operating precisely where the new framework places the ring current’s inner edge.

STEVE and the picket fence emerge at the extreme end of this evolution. As a SAID channel deepens — plasma upflows and charge-exchange reactions that convert oxygen ions into rapidly recombining molecular ions strip the ionosphere of density — the ionospheric feedback instability ignites. Over-reflection of Alfvén waves in the shear flow above roughly 120 kilometers altitude spawns small-scale dispersive Alfvén waves that bounce between conjugate hemispheres, threading kilometer-scale filaments of field-aligned current of 10 to 20 microamperes per square meter. Each filament carries an embedded parallel electric field that grows nonlinearly as the background field strengthens and the density thins — a dependence sharpened further by a shallow density valley between 120 and 200 kilometers altitude. Once depletion crosses a critical threshold, these fields heat electrons locally enough to excite both the 630-nanometer red line and, at energies above 4.2 electron-volts, the 557.7-nanometer green line of the picket fence. Swarm passes directly over STEVE arcs confirm the predicted current filaments, with Doppler-shifted oscillations at 0.4 to 0.6 hertz corresponding to wavelengths of 12 to 18 kilometers — explaining why the purple ribbon and its rayed companion look, and behave, nothing like particle-precipitation auroras.

Simulations back the story quantitatively. Solving a reduced two-fluid model of magnetosphere–ionosphere coupling in dipole coordinates, seeded with the fields and densities measured by spacecraft, the authors flew virtual satellites through their domain along real trajectories. For the 17 March 2015 storm, a virtual Van Allen Probe reproduced the observed ultra-low-frequency oscillations with dominant frequencies near 6 and 12 millihertz; for a June 1991 event, virtual CRRES and DMSP trajectories both peaked near 8 millihertz, matching the data down to the spiky electric fields sharpened by the feedback loop between currents and conductivity. The instability reaches a self-sustained state within about five minutes and persists as long as the driving field, with the ionosphere — not the magnetosphere — setting the frequencies. Saturation near 10 microamperes per square meter arises either from anomalous electron heating by the Farley–Buneman instability or from the decay of inertial Alfvén waves into three-dimensional convective cells, and the parallel electric fields scale with the driving field to a power approaching two.

The stakes reach beyond celestial aesthetics. Subauroral flows drive the scintillation that corrupts GPS and other navigation signals at mid-latitudes, and conventional global models — which treat particles as test bodies and neglect polarization fields — missed the first one to two hours of storm-time injections and flows entirely, simulating channels ten times too wide. The review’s message is that the plasmapause behaves like a power plant, converting the kinetic energy of incoming plasma flows into the mesoscale electric fields and currents that light SAR arcs and, at their most extreme, paint STEVE across the sky. Precursor SAR arcs, the authors note, consistently appear before STEVE, tracking the contraction and structuring of the underlying flow channel — evidence that the eerie transition from faint red glow to blazing purple ribbon is choreographed by Alfvénic turbulence in a depleted ionosphere. What skywatchers photograph as a beautiful anomaly, the authors argue, is the visible output of a multiscale electromagnetic circuit that physicists are only now learning to read.