A New Magnetic Storm Model

Robert B. Sheldon, Harlan E. Spence

Boston University Center for Space Physics
725 Commonwealth Av, Boston, MA 02215
18 June 1997


Recent observations from the new perspective of the POLAR orbit have elucidated the crucial role that parallel electric fields and the ionosphere play in the development of a magnetic storm. During the main phase, we observed field-aligned beams of ionospheric ions that appear to be caused by deeply convecting plasmasheet ions. These ionospheric beams are a substantial fraction of the ring current density, suggesting a novel mechanism for ring current and Dst growth during the main phase of a magnetic storm. From these observations, we construct a new magnetic storm model that has implications for all aspects of a storm sequence, including ground, ionospheric and magnetospheric observations.


The elements of this new magnetic storm model have all been presented before, but they have lacked a coherent, causal chain, and above all, a quantifiable model that had predictive ability. For example, (Smith74) presented the time-dependent model of ring current injection, but were unable to predict how deep the injection would occur or subsequent Dst development of a storm. Similarly, Hamilton (1988) measured the enhanced ionospheric oxygen content of great storms, but were unable to predict which storms or how much oxygen was to be expected. Our model takes as its input the cross-tail potential at the inner edge of the plasmasheet and the plasmasheet density at this location, to predict a complete storm sequence. With a suitable inner tail model, we should be able to use only the solar wind density and electric field, Ey=Bz * Vx, to make these predictions.

Data Analysis

FIGURE 1: POLAR/CEPPAD/IPS data on April 15, 1996, displaying roll modulation of the counts in the 90o head in the energy bands from 15--1500 keV. Dotted white and blue lines are 90o and 60o pitchangles.

The POLAR spacecraft is in a polar, 9 x 2 Re orbit that on April 15 was in the noon-midnight meridian with perigee over the south pole. Thus it made two diagonal passes through the midnight and noon radiation belts. A typical pass shows an energetic particle population whose average energy is proportional to the |B|. This is the normal distribution for ions diffusing in L-shell from a source region in the plasmasheet. On this day the Comprehensive Energetic Particle and Pitch Angle Distribution (CEPPAD) (Blake95) experiment detected two nearly monoenergetic bands superposed on the night side ring current (figure 1): a population of trapped ions at ~90 keV, and a beam of field-aligned ions at ~40 keV.

A nearly monoenergetic trapped population is possible when a strong cross-tail electric field drives ions against the grad-B drift deep into the magnetosphere (Smith74). Such a ``nose'' event must be nearly 90deg trapped particles because of the large increase in |B| while convecting from the plasmasheet, which is the characteristic of the upper energy band in our data. Since nose events are highly correlated with storms, and storms are defined by Dst, but Dst is generally not immediately available, we turn to the preliminary Dst provided by the Kyoto University web site. After subtracting the ionospheric Sq contribution using the ``quietest day of the month'' (April 7/8) method, we find a moderate storm of at least -63 nT on the first hour of April 15. Additional support for the nose event identification came from the extensive GGS database.

Examination of WIND data (Ogilvie95) for April 14 showed that there were several Bz<0 periods lasting for 1 hour or less. Corresponding CU and CL derived from the CANOPUS array (Rostoker95) show that these periods led to substorms with riometer absorption signatures at auroral latitudes. However the storm trigger appears to be the strong southward turning of Bz< -10 occurring at 2000 UT, accompanied by a jump in the solar wind speed from 450 km/s to 600 km/s, which produced an even larger Bz in the compressed magnetic field of the magnetosheath. This period of strong southward Bz lasted more than 3 hours, effectively saturating the ability of the tail to shield out the polar cap potential. The IZMIRAN model (Papitashvili94) predicted in excess of 150 kV across the polar cap for these solar wind conditions.

The CANOPUS array detected a magnetic bay, a nearly equal response of AU and AL, suggesting that the current systems had moved equatorward, overhead of the magnetometers. Indeed, the Halley Bay magnetometer at L~4 (Dudeney95) showed a large H deflection with almost no Z deflection, indicative of strong overhead currents. The ionosphere responded strongly at this time (private communication, J. Aarons, 1996) as ionospheric scintillations were enhanced.

While CANOPUS riometers at L>6 recorded very little activity, the riometer at L=4.4 as well as the Halley Bay riometer at L~4, (private communication A. Rodger, 1996) showed an extremely intense and narrow absorption feature at this time, indicating that precipitation had penetrated to low latitudes, deep in the magnetosphere and down to E-layer ionospheric depths. From these observations, as well as complementary ground observations, we surmise that after several intense substorms had pumped up the plasmasheet, a strong convection field injected the plasma to at least L=4.4, which POLAR/ CEPPAD observed as a 90-100 keV band.

The band of 40 keV field-aligned ions, however, are harder to explain. They have the wrong pitchangles to have convected from the plasmasheet, because the adiabatic decompression involved in backtracing them to their origin would place them in the plasmasheet loss cone. Nor would ions of this energy have had access to the plasmasheet simultaneously with higher energy ions, since the magnetosphere is a ``notch filter'' for only one energy. This implies that these ions are trapped on closed drift orbits. But if they undergo the same processes as the adiabatically energized ring current, which can be seen simultaneously with the banded distribution, they would not be monoenergetic, nor would they track the energy of the nose ions so precisely. That is, if they had resided for any length of time in the magnetosphere, the same convection that brought in the plasmasheet ions would disperse these ions as well. Nor do these ions show the energy dispersion associated with substorm injections. Thus we conclude that these field-aligned ions are in situ accelerated during the time of the measurement.

``Zipper" events, (Fennell81,Kaye81), have the same bimodal pitch-angle distributions, though at somewhat lower energy. They found that the ``zippers'' were rich in O+, and concluded that they were observing beams coming from the ionosphere. Since beams are seen at auroral latitudes, they concluded that they were on flux tubes connected to auroral arcs. The composition experiment, POLAR/ CAMMICE, was turned off during this period, so we cannot determine the E>30 keV O+ content for April 15, however when it was switched back on, around April 19, it found an anomalously large amount of O+ in the ring current (RC) in two energy bands centered at 40 and 100 keV. Because of the short lifetime of O+ against charge exchange, such a measurement is consistent with a storm injection occurring only a few days previously, since between storms, CAMMICE detects no O+ enhancements. A similar storm on March 21, 1996 showed that the IPS zippers are simultaneous with the double peaked O+ spectrum observed by CAMMICE. POLAR/ TIMAS (Shelley95) (private communication W.K. Peterson, 1996) did detect E<20 keV O+ beams for this day. So we conclude that on April 15 we are observing O+ ions accelerated to ~30--40 keV by strong parallel electric fields in the ionosphere. Since these ions track ~50 keV below the nose ions, it appears there must be a causal connection.

Whipple77 argues that one can produce a field-aligned electric field if the electron and ion pitch angle distributions (PADs) are not identical. In our case, a hot monoenergetic nose ion superposed on a cold plasmasphere electron population will produce a parallel electric field of several kTe pointing away from the equator. That is, since a mirroring ion spends most of its time away from the equator, the electrons at the equator will experience a force pulling them toward higher latitudes. If the energetic ion density (or current density in a dynamic system) is greater than the plasmaspheric cold electron density, a second ion-dominated solution to the Whipple equations is possible that produces an ion space-charge potential at approximately the ion beam parallel energy. (Since the Whipple equations are derived for a neutral plasma, we must rederive these results using Poisson's equation for a non-neutral plasma. Intriguing evidence that such space charge is both possible and probable come from laboratory experiments on magnetized, trapped electron clouds (Hansen95). This space charge potential would completely expel electrons to the equator and could only be neutralized somewhere earthward on the flux tube where the cold electron density would again exceed the hot ions. At this point, the potential would drop from kV to a few kT_e, forming a double layer. Naturally this space charge should be shielded by neighboring flux tubes, generating a locally perpendicular electric field that may be manifest in the ionosphere by a ``polarization jet'' or a subauroral ion drift (SAID).

FIGURE 2: Peak fits to 96s averaged spectra of the 90deg head. Ring current ([]), nose ion (*), beam ion (triangle) energies, difference (O) and 10X the ratio (X) of nose:beam are plotted. ``Error'' bars are FWHM from fits, since fitting errors are negligible. Overplotted with different scales are: one half the model equatorial B-field intensity (dashed); and on linear scale the magnetic latitude from -40 to 40 degrees (dotted), with the equator marked with a vertical dotted line.

By measuring the energy of these beams as the spacecraft crosses through this region, we can map out some features of this potential structure. We have fit the spectra with a sum of three peaks: a Chapman layer function appropriate for the ionospheric beam, a Gaussian for the nose ions, and a log-space Gaussian for the ring current ions, achieving remarkably good 9 parameter fits to 16 energy points. We note that an asymmetric Chapman layer function, y=exp(1+x-exp(x)), is what one expects if the extraction potential is extended over some distance.

Figure 2 demonstrates how the three populations, beam, nose and RC, evolve during this orbit. The RC is adiabatically energized so that as the model equatorial magnetic field increases the energy increases, as noted by the squares and dashed line. A latitude effect in RC is apparent due to the more rapid loss of low energy ions at higher latitudes. This energization is not seen by the nose ions because they are not an equilibrium trapped population, rather we are seeing the nose ions that have access to this location, so that stronger B-field ions also began with smaller magnetic moment. A more detailed analysis of the pitch angles is needed to clarify the coupled energy-L dependence of these ions.

If the first solution to Whipple77 holds, then both the nose and the beam energies should move in concert as we move along the field line, and we should observe a constant difference between these two energies. We find instead that the ratio is more constant, that the beam ion energy tracks the nose ion energy with a nearly constant ratio of 1/2. This suggests that the beam energy is determined not by the electron thermal temperatures or anisotropies, but by the energy of the nose ions themselves, consistent with the second, ion-dominated solution to the Whipple77 equilibrium.

Such a parallel electric field should also accelerate magnetospheric electrons to 30 keV into the ionosphere, producing a riometer absorption event. Since the electric field is colocated with the nose plasma, the absorption must be a narrow strip in latitude but distributed in longitude, as seen in the CANOPUS data. This extended signature is reminiscent of SAIDs which have been previously identified with substorm injections. Our mechanism would identify a large SAID with a storm injection, and which may also account for the E-layer keV electron signature seen in subauroral riometer data.

We carefully distinguish between substorm and storm injections because the characteristic signatures of each are different. A substorm involves a reconfiguration of the magnetic field which produces a dB/dt electric field in the region between 6-10 Re near midnight. This produces two types of signatures in the POLAR data. The inductive electric field may in situ accelerate the entire plasma population to ~30 keV, which is observed in our instrument as an isotropic, dispersionless <50 keV enhancement over a restricted MLT and L-shell range. Or it may bring in plasmasheet material with a distinctive, highly energy dispersed signature. A storm injection, on the other hand, is observed as a nose event, a monoenergetic band of ions penetrating to subauroral L-shells and existing over a broad range of MLT determined by the duration of the cross-tail field. Energization of low energy plasma is not seen, but adiabatic energization of the nose ions occurs as the ions convect toward stronger B. From an ionospheric viewpoint, substorms are in the auroral zone, whereas storms penetrate down to sub-auroral latitudes albeit in a restricted latitude band. Thus we argue that these beams cannot be substorm generated.

Some other figures

POLAR/CEPPAD/IPS data on (a) April 14, and (b) April 15, and (c) April 16, 1996, displaying energy spectra summed over pitch angles from 20--1500 keV.

Successive 96s spectra (symbols) and three peak fit (line) for April 15 data set.


If our mechanism generates field aligned electric fields and oxygen rich beams for every storm injection, then we have elucidated a new model for magnetic storms. Since it is generally thought that magnetic storms are characterized by intense convection fields that bring plasmasheet plasma deep into the RC, then all storms should create parallel fields and extract oxygen from the ionosphere as well. Thus large storms should extract more oxygen than small storms, with a larger Dst effect. Similarly, Dst should change on minute bounce timescales, rather than the 10's of minutes convection timescales.

Several recent observations lend support for this theory. Analysis of the CRRES data set shows that there is a positive correlation between the magnitude of Dst and the O+ content of the RC ( M. Grande, COSPAR 96 proceedings to be published in Adv. Sp. Res.). A Dst prediction filter (Gleisner96) found that a neural network with one hidden layer, representing an unknown quadratic dependence on solar wind Vx,Bz, and density, could explain up to 84\% of the variance in Dst. If significant O+ is extracted during the main phase as we predict, one would expect such a non-linear dependence of Dst with Ey. And Kyoto 1-minute Dst show excursions much too rapid for convective timescales.

A study of storm-time PC1 pulsations by Mursula (COSPAR 1996 paper to be published in Adv. Sp. Res.) showed that the PC1 waves associated with ion-cyclotron waves during the most intense part of the main phase of a storm occurred primarily at dusk and the PC1 frequency dropped down to below the oxygen gyrofrequency. These observations are suggestive of an oxygen rich plasma occuring at dusk, the location of the deepest penetration of the nose ions and also in agreement with our theory. This extraction of ions and precipitation of electrons near midnight will generate an outward flowing current which then drifts westward with the bulk of the RC. We expect that the disappearance of the parallel field, occurring near the dusk terminator, will result in a downward current thus completing the loop of the partial ring current as measured by Suzuki85 using Magsat data. They concluded that 1/3--1/4 of RC amperage was observed in the partial RC. If we assume that half of the partial RC is carried by upward flowing ions, then we estimate that 1/6 - 1/8 of the total RC is composed of ions of ionospheric origin.

Thus we find that the mechanism described in the abstract not only provides a causal chain for the entire storm sequence, but has great predictive power in explaining many other observations not previously linked to storms.


We have attempted to construct a new model of magnetic storms that incorporates field-aligned electric fields as an intrinsic part of the storm development. A storm proceeds then with the following steps:
  1. A critical density plasmasheet may be required for a significant magnetic storm, in which case precursor substorm activity or a solar wind density enhancement would be a prerequisite.
  2. Large solar wind $Ey$\ produces a large polar cap electric field, which if persistant for an extended period ($>$3hr) exceeds the ability of the tail to shield out the electric field.
  3. This produces a strong cross-tail convection electric field that transports plasmasheet ions deep into the magnetosphere.
  4. The opposing $\nabla B$\ and $E\times B$\ drifts act as a selective filter permitting a single energy the deepest penetration, a monoenergetic ``nose'' event.
  5. This spatially narrow band of hot ions, compressed as it convects towards stronger B, can outnumber the local cold electron density at the equator and thus generate a field-aligned space-charge potential that attempts to confine the hot ions to the equator, a parallel electric field.
  6. At some point down the field line earthward of the equator, the growing cold electron density exceeds the hot ions and the potential is neutralized to approximately the electron thermal temperature, forming a double layer that very quickly extracts ionospheric ions, including H$^+$\ and O$^+$.
  7. Ion cyclotron waves, generated by this unstable particle population, then perpendicularly heat these ion beams and scatter them out of the loss cone so that they become a permanent part of the ring current.
  8. This entire structure is a ``convective instability'', a dynamic balance between the current of corotating cold electrons and convecting hot ions, which would rapidly decay if it were not for the continuous power input of the convection electric field. When the field switches off, the convection power source is removed, the parallel electric fields vanish, and the hot ions are trapped in the ring current and subsequently decay through charge exchange.


This study was supported by NASA contract NAS5-30368 and NSF grant ATM-9458424. The authors gratefully acknowledge K. Ogilvie, E. Shelley, J. Dudeney, G. Rostoker, B. Blake and T. Fritz for the use of their data.


TOC Comments? r*bs@rbsp.info
(Due to spamming, delete asterisks)