11. Birkeland currents
Magnetic variations observed on the ground in the auroral zone are much larger than those at middle and low latitudes: swings of 500-1000 nT at auroral latitudes (out of about 60,000 nT) are much more common than 100 nT disturbances at the equator, which would be classified as fair-sized magnetic storms (Rufenach et al. , Figures 12 and 13, respectively). The strong polar disturbances are localized, suggesting that the currents producing them flow nearby, probably in the ionosphere.
Birkeland [1908, 1913; Bostrom, 1968; Stern, 1977; see also BH-1] noted that the direction of the disturbance field in the auroral zone tended to be perpendicular to auroral arcs. He concluded that large electric currents flowed lengthwise along the arcs, and speculated that those currents arrived along magnetic field lines at one end of the arc and returned to space by a similar route at the other end. An overall pattern inferred in this way was later mapped, especially by Silsbee and Vestine , and its currents were named auroral electrojets; they seemed to originate on the day side and to flow towards midnight along both sides of the auroral oval. Sugiura and Davis  combined the readings of about a dozen magnetic observatories around the auroral zone and extracted an "AE (auroral electrojet) index" which gauged the strength of the electrojets. Values of this index are now regularly compiled and often serve as indicators of substorms and of the level of magnetospheric agitation [Rostoker, 1972b; Mayaud,1980].
Because the ionosphere conducts electricity, the existence of a dawn-to-dusk polar electric field (Figure 5b, contours viewed as electric equipotentials) suggests that an electric current flows across the polar cap; the current might enter on the morning side of the polar cap and exit in symmetric fashion on the evening side, like the current in Figure 8. The pattern of E, however, also has fringe fields that extend equatorward of the oval, to field lines that are shorter and therefore thread parts of the magnetosphere closer to Earth. In a static electric field E = -grad V, if E// is negligible, it follows from (1) that B.grad V= 0 and hence that the electric potential V is constant along field lines. The fringe pattern then maps F and E to the near-earth magnetosphere.
Schield et al.  deduced from this an important new effect. The earthward flow in the tail predicted by both convection theories (Axford-Hines and Dungey) is associated (by (1)) with a dawn-to-dusk electric field E across the tail, which then maps along field lines to the polar cap, and the polar fringe pattern extends this E to nightside equatorial regions closer to the Earth. When convecting ions and electrons arrive near Earth, appreciable guiding-center drifts caused by the dipole-like internal field are added to their convective flow. These deflect the flow around the inner part of the magnetosphere, as was assumed by Axford and Hines  and as was claimed even earlier by Alfven [1939; see BH-1] .
However, the magnetic drifts move positive ions and electrons in opposite directions. Schield et al.  showed that as a result, if such drifts are added to the convective flow, the plasma no longer stays electrically neutral. This cannot be allowed to happen, because even a relatively tiny deviation from strict neutrality produces huge electric fields. The process may be halted in one of two ways: either E is modified in a way that keeps the plasma flow out of the region of strong magnetic drifts, or else electric currents arise along magnetic field lines (the easy flow direction in a plasma) and drain away the excess of electric charge. Both processes seem to occur.
The modification of E takes the form of "shielding," of an exclusion of E from the vicinity of the Earth, making the fringe-pattern in Figure (5b) narrower than what a calculation based solely on ionospheric conductivity would give. This was first calculated by Vasyliunas  and also by Jaggi and Wolf , who devised a way of simulating the process on a computer. The method was later expanded by Wolf and his group at Rice University in Houston with the "Rice Convection Model" (RCM) which simulates both the shielding and the neutralizing currents [Spiro and Wolf, 1984].