Journal of Geophysical Research, 89, 8919-8939, 1984
(Received March 27, 1984; revised June 21, 1984; accepted June 25, 1984)
Copyright 1984 by the American Geophysical Union.
Paper number 4A0918.
During times of high magnetic activity, significant changes are known to occur within the earth's magnetosphere. Specifically, the ring current particle population is known to be modified at these times as shown by the Dst index and by direct measurements of the energetic particle populations. It has been argued by Lyons and Williams  that the increase in the ring current population at storm times can be explained by inward convection of the preexisting plasma population. Mass composition data, however [Johnson et al., 1977; Balsiger, 1981; Johnson, 1981; Lennartsson et al., 1981; Hultqvist, 1982; Lundin et al., 1982a; Young et al., 1982], indicate that there is a change in the composition of the lower-energy ( < 17 keV/q) ring current population during times of increased geomagnetic activity, and it has been deduced that there is an enhanced injection of ionospheric plasma into the magnetosphere during storm times. Presumably, the increase in the ring current signature as observed at the ground is due to both these processes occurring at the same time; that is, enhanced inward convection of the preexisting trapped population together with an overall increase in the particle population due to injection. Although the relative contribution of the two processes to the storm time ring curent remains to be determined quantitatively, Johnson  and Johnson et al.  have argued that in some storms the ionospheric component will be an important contributor to the ring current energy density because of the buildup of the ionospheric component in the hot plasma population prior to the main ring current injection process.
Mass composition data are a useful indicator in assessing the relative significance of both processes. Unfortunately, most storm time mass composition data is restricted either to the lower-energy component of the ring current, typically less than 17 keV/q, or to energies of the order of 1 MeV [e.g., Spjeldvik and Fritz, 1981a, b]. Nevertheless, much can be learned concerning the effects of injection and convection on the magnetospheric ion population using such data. Strangeway and Johnson , for example, have investigated the mass composition of ion dispersion events as observed at the near-geosynchronous SCATHA spacecraft orbit. It was found that the dispersing ions showed features within the composition data indicative of at least some injection of ionospheric plasma into the equatorial magnetosphere. It was suggested by the authors that field-aligned currents associated with an inward propagating substorm front [Moore et al., 1981] could result in enhanced injection of ionospheric plasma at or near the substorm front.
As is mentioned above, Balsiger [l981] and Johnson  have reported on mass composition data acquired during storm times. Data from the first five major storms occurring in 1979 were described and discussed by Johnson  and Johnson et al. . One of the storm times investigated occurred during February 21-22, 1979, and for this period Balsiger  and Johnson  used data from severe spacecraft. Data from ISEE 1 and Prognoz 7 were used to monitor the ion composition in the near-equatorial magnetosphere over a large range of radial distances. The Prognoz 7 data have also been discussed by Lundin et al. [1982a] and Hultqvist . At geosynchronous altitudes, data from GEOS 2 and SCATHA were used. Low-altitude data from the polar-orbiting S3-3 spacecraff were also available. From this combined data set it was determined that there was a marked increase in singly charged oxygen (O), in both number and energy density, at or near the peaks in Dst associated with the storms occurring during this time period. It was also shown that this injection of ionospheric plasma occurred over a large region of the magnetosphere, in both local time and radial distance.
The subsequent transport and convection of the injected plasma is of considerable interest; it may be possible to address such questions as the importance of loss processes and trapping. With this end in mind, in this paper we investigate in considerable detail data from the near-geosynchronous SCATHA spacecraff and the polar-orbiting S3-3 spacecraft for the period of February 21-22, 1979. In the following section we shall describe the geomagnetic conditions during this time period together with the data coverage and instrumentation used to acquire the data.
Near-geosynchronous data from the SCATHA spacecraft will be described in section 3. In addition to the bulk parameters used to summarize the data, we shall also present the data in spectrogram format. The spectrograms allow us to understand in more detail some of the features apparent in the summary plots; for example, dispersionlike structures can be seen in the particle spectra.
In section 4 we shall present low-altitude data from the S3-3 spacecraft. We shall defer from presenting the data in spectrogram format until section 6 and initially consider number density plots only. We do this because the S3-3 data have not been previously published and it is easier to use a survey format to introduce the data. We shall show that there are some features in the S3-3 data that raise questions concerning the evolution of the low-altitude, 0.5- to 16-keV/q plasma population after the main phase of the magnetic storms.
Having described the data in more detail in sections 3 and 4, we shall go on to consider the importance of convection as an explanation of the signatures found in the low-altitude S3-3 data. In section 5 we shall briefly summarize some of the work concerning steady state convection models and point out those aspects of the theory which we consider relevant to the S3-3 data. Results from the theory shall be applied to the S3-3 data in section 6, and for this purpose we shall also display the data in a spectrogram format. Using the geosynchronous composition data as a monitor of the source of the injected plasma, we shall then discuss the results in section 7. A summary and conclusions are given in the last section of the paper.
In this paper we present mass spectrometer data from both the S3-3 and SCATHA spacecraft. A more detailed description of the SCATHA instrument and spacecraft orbit has been given by Kaye et al. . The S3-3 mass spectrometer characteristics have been described by Sharp et al. . We shall, however, summarize some of the features of the sapcecraft orbits and instrument capabilities here.
The SCATHA spacecraft is in a near-geosynchronous orbit with apogee at 7.8 R and perigee at 5.3 R. The orbital period is a little less than a day, the apogee drifting about 5° eastward per day. The spacecraft is spin stabilized with a spin period of approximately 60 s. The spin axis is in the orbital plane and is pointed perpendicular to the sun-earth line. The mass spectrometer is mounted with the look direction some 11° from the spin plane and consequently covers a large range of pitch angles in one spin period during relatively undisturbed magnetic conditions. The energy range of the mass spectrometer is from 0.1 to 32 keV/q, and the mass range includes protons (H) and singly charged oxygen (O). Mass composition data for alpha particles (He) are also routinely obtained, but we shall not present those data here, as the count rates for these two species are usually considerably lower than for H and O.
The S3-3 spacecraft is in a polar orbit with apogee at 8000 km altitude and perigee at 240 km altitude. The spacecraft is also spin stabilized with a spin period of approximately 20 s. The mass spectrometer is mounted perpendicular to the spin axis, which is itself perpendicular to the orbital plane, and so an almost complete pitch angle scan is available. The mass spectrometer covers the same mass range as the SCATHA mass spectrometer, but the energy range is 0.5 to 16 keV/q.
Figure 1 summarizes the data coverage and geomagnetic conditions as determined by Dst for February 21-23, 1979. The bottom panel shows Dst variation as a function of time with the bar segments showing those time periods for which data were used from the SCATHA and S3-3 spacecraft. Two major storms occurred on February 21, near 0900 UT and 2200 UT. The recovery phase for the first storm was fairly short, whereas the recovery phase for the second storm was much longer. There is some evidence for suggesting that there are additional ring current enhancements during this recovery phase, minima in Dst also occurring-near 0100 and 0900 UT on February 22.
Fig. 1. Data coverage of SCATHA and S3-3 spacecraft during February 21 and 22, 1979. The bottom panel shows Dst and the time periods for which data were used in the storm time study. Those periods marked MS in the SCATHA coverage indicate magnetosheath encounter. The top part of the figure shows orbital segments for both spacecrah projected in L shell and magnetic local time. Two representative magnetopause locations have been included for reference.
As indicated by the bars, the SCATHA data coverage is only broken at those times in which the spacecraft encounters the magnetosheath (MS). Data are available at this time, and in fact the magnetosheath encounters were determined from mass composition information, as shall be described in the next section. Data from S3-3 were only used when the spacecraft was near apogee. In total, data from 10 orbits were used, the first three orbits covering the time period following the first storm at 0900 UT on February 21, with the rest occurring during the second storm period.
The top part of Figure 1 shows the orbital segments for both spacecraft projected into magnetic local time (MLT) and L shell. Noon is at the top of the figure, and two typical magnetopause locations have been included for reference. The SCATHA orbital segment is shown for February 21, the numbered tick marks giving the universal time (UT). For this day, apogee in these coordinates occurs near 0900 MLT, and it will be noted that the orbit lies very close to the high solar wind pressure magnetopause. This is somewhat fortuitous, but does indicate that this is the local time sector in which magnetosheath encounters are most likely. The line elements show the S3-3 orbital segments corresponding to those times shown in the bottom part of the figure. The solid curves correspond to orbits occurring on February 21, and the dashed curves show orbits for February 22. The numbers labeling the orbital segments give the UT to the nearest hour at which the data were obtained.
Figure 2 [after Johnson, 1981] summarizes the mass composition as observed at the SCATHA orbit. At the bottom of the figure, Dst and Kp for February 21 and 22, 1979, are shown. The vertical dashed lines near 0900 and 2200 UT on February 21 and 0100 on February 22 indicate minima in Dst. Above these curves, four panels of mass composition data are given. From top to bottom, we plot the O/H energy density ratio at 32 keV (the maximum energy measured by the mass spectrometer), the energy density ratio integrated over the full energy range (0.1-32 keV), the proton number density, and the oxygen number density. The number densities are integrated over the full energy range. The shaded regions labeled MS indicate magnetosheath encounter.
Fig. 2. Summary of SCATHA data for the storm period of February 21 and 22, 1979. Mass composition data are shown in the top four panels, and Dst and K are shown in the bottom panel [after Johnson, 1981].
Johnson [l981] pointed out that near the time of maximum excursion in Dst the O number density increases markedly, whereas the proton number density is relatively constant. The increase in oxygen abundance is also observable in the energy density ratio. Even at 32 keV there is some evidence for suggesting that around the peak in Dst, oxygen is the major contributor to the energy density.
Although moment integrals of the ion species distribution functions are useful as a summary, the integrating process averages out energy dependencies in the particle flux. Plates 1a and 1b show energy-time spectrograms for February 21 and 22, respectively. These plates show differential energy flux measured in keV cm s sr keV for protons in the top plot and oxygen in the bottom plot as indicated by the labels on the right-hand side of each plate (DEF H and DEF O, respectively). The abscissa is universal time (UT), and it should be noted that the small tick marks occur at 15-min intervals, this being the time period over which the composition data were averaged to produce the spectra. Energy, measured in keV and plotted on a logarithmic scale, is ordinate. For consistency with other data presentations, such as those used by DeForest and McIlwain  and Mauk and McIlwain [l974], in which both ion and electron fluxes are shown, the energy scale is reversed with energy increasing downward. The flux intensity corresponding to each color is given by the color scale on the right of each spectrogram. It should be noted that both protons and oxygen have different intensity scales, the values being chosen such that the same color value at a particular energy W and time implies approximately equal number densities for some W. The difference in scale is a factor of 5, whereas the square root of the mass ratio is 4. The fluxes shown in Plates 1a and 1b have been averaged over all pitch angles, since no pitch angle information is available for this time period.
Plate 1a. Energy-time spectrogram of SCATHA data for February 21, 1979. Differential energy flux for both protons (top panel) and oxygen (bottom panel) are shown as a function of energy and universal time (UT). For consistency with other presentations, the energy scale is plotted downward. Ephemeris information has also been included at the bottom of the plate.
Plate1b. Energy-time spectrogram of SCATHA data for February 22, 1979. Similar to Plate 1a.
At the bottom of each plate we plot ephemeris information. The light blue curve gives L as a function of UT. On February 21, for example, apogee occurs near 1400 UT. The yellow curve gives MLT-UT. Universal time has been subtracted out, as the net drift in MLT with respect to UT is of the order of half an hour per day. Since the orbit is slightly eccentric, there is a diurnal variation in MLT-UT. [ The following sentence has been corrected from the originally published version, which gave an incorrect MLT - UT correspondence.] For ease of reference, symbols are plotted on this curve at 1800 MLT (near 0100 UT), 2400 MLT (near 0500 UT), 0600 MLT (near 1100 UT), and 1200 MLT (near 1930 UT). The symbols represent the earth phase as viewed from the spacecraft at these local times.
The most striking features in Plate 1a is the large proton fluxes at energies of < 10 keV around 1600 to 2000 UT, with a corresponding drop in oxygen flux. These fluxes are due to spacecraft entry into the magnetosheath. Large fluxes of alpha particles (not shown) are also observed at this time. As is noted above, we have not included data from these times in the calculation of number densities as shown in Figure 2.
The first storm on February 21 occurred around 0900 UT. The spectrogram shows that there is considerable flux of O in this time interval. There is an increase in flux for both species just prior to 0600 UT, when Dst first begins to decrease, but the proton fluxes stay at this initial value. After 2000 UT, both proton and oxygen fluxes again show an increase in flux, with oxygen showing the more spectacular changes.
The time sequence is continued in Plate lb. It is apparent that oxygen is again the dominant ion up to 0600 UT. This was shown in Figure 2, but we note the dispersionlike signatures in both particle species, which cannot be deduced from number density signatures alone. After 0600 UT, proton fluxes increase, with maximum flux occurring around the 10-keV level. There is no similar increase in the oxygen fluxes. This feature is also present in the proton number densities as shown in Figure 2. While the oxygen number densities are falling off, there is an increase in proton density around 0600 UT. The dip in Dst near 0900 UT may be associated with this.
One last point concerning Plate lb is the magnetosheath encounter near 1500 UT. Data from this time interval were not included in Figure 2. Apart from this particular signature, both particle species show the development of such features as the minimum at intermediate energies in both proton and oxygen fluxes. This minimum is probably the "proton deep minimum" as discussed by McIlwain , which is usually attributed to loss processes preferentially removing those particles with small azimuthal drift velocities or whose drift orbits penetrate to low altitude.
In summary then, the SCATHA mass composition data show that significant amounts of oxygen are injected into the magnetosphere during the main phase of the first two storms on February 21 and 22, 1979. Energy-time spectrograms show that the oxygen is injected over a large energy range, so that the oxygen ions are a major contributor to both the number density and the energy density of the near-geosynchronous plasma in the 0.1- to 32-keV energy range. We have not attempted to determine the mechanism which results in the enhanced oxygen content, but it should be noted that Johnson  and Balsiger  have reported the observation of marked oxygen flux increases throughout the near-equatorial magnetosphere during the storm period described here.
Having described the near-geosynchronous mass composition data in the previous section, we shall now consider the low-altitude S3-3 data. In general, the near-geosynchronous SCATHA data show local time effects, whereas data acquired from the polar-orbiting S3-3 spacecraft tend to show L shell dependent effects, except at very high latitudes, where the orbit may cross local time more rapidly than L shell. The S3-3 orbit segments shown in Figure 1 indicate that for the data presented in this section, the orbits were roughly constant local time.
In a similar manner to the previous section we first present number density plots as a summary of the mass composition data. As shown in Figure 1, there were three S3-3 pass during the main phase and recovery phase of the first storm, the peak in Dst being near 0900 UT on February 21, 1979. Number densities for these three passes are shown in Figure 3 for both protons and oxygen as a function of L shell. The number densities were calculated using the full energy range of the S3-3 mass spectrometer (0.5 - 16 keV/q). The open symbols in each panel are upper limits for the number densities. We have used the 3 value as the upper limit, when the error on the measurement due to counting statistics is greater than 33%.
Fig. 3. Composition data from the S3-3 spacecraft for the first storm on February 21, 1979. Number densities of oxygen and protons are plotted as a function of L shell for three orbits acquired during the first storm on February 21. The top panel shows data obtained near the peak in Dst (at 0900 UT), and the other panels give data acquired during the recovery phase. Open symbols indicate upper limits to the number density due to poor statistics.
First, we note that oxygen is the major component at this time, which is similar to the geosynchronous data. Second, both particle species show a falloff in number density as L decreases. There is an additional feature which we shall discuss further below. This is the movement to lower L of the drop-off in number density as time progresses. The bottom panel in Figure 3, from 1404 to 1427 UT, shows data acquired some 6 hours into the recovery phase of the first storm. The SCATHA data show that the oxygen number density is beginning to decrease at this time, although it should be remembered that the SCATHA spacecraft is near apogee, and also for this particular day, close to the magnetopause. Nevertheless, that both species still show quite high number densities at low L may be significant.
If it is assumed that the signature in Dst is due to inward convection of preexisting plasma populations, as suggested by Lyons and Williams , with the subsequent recovery in Dst resulting from decay of the ring current plasma, the lowest excursion in L might be expected to occur near the peak in Dst. This does not appear to be the case, although care should of course be taken when using ring current changes to explain the signatures at S3-3 altitudes. The S3-3 mass spectrometer measures particle flux only up to 16 keV/q and so does not cover the bulk of the ring current. Also, the plasma measured at S3-3 altitudes will be confined to within a few degrees pitch angle at the equator. We shall, however, investigate this feature further in the next section, where we consider energy-L spectrograms of the S3-3 data.
To show that the movement to lower L during the recovery phase of the storm is not just associated with the first storm, we have plotted number densities for the first three passes for which data were available during the second storm, which peaked around 2200 UT on February 21, 1979. These are shown in Figure 4. The data plotted in the top panel were acquired prior to the peak m Dst, whereas the two botton panels show data some 6 and 9 hours into the recovery phase respectively. It is interesting to note that besides showing the movement to lower L as a function of time, there is also some evidence for different source and/or loss rates for the different particle species. The proton and oxygen number densities are roughly equal in the top panel, especially at the higher L shells, but during the recovery phase the proton number density is at least an order of magnitude less than the oxygen number density. This signature is presumably a function of both enhanced injection of ionospheric plasma and different loss rates for both species. The loss rate due to charge exchange, for example, is larger for protons than for oxygen in this energy range [Tinsley, 1976].
Fig. 4. Composition data from the S3-3 spacecraft for the second storm on February 21, 1979. Similar in format to Figure 3. Data shown in the top panel were acquired during the main phase of the storm. whereas the other data were obtained during the recovery phase.
In Figure 5 we show data for four passes acquired well into the recovery phase of the second storm. As can be seen from Figure 1, there is still some activity at this time, and there appear to be two additional enhancements of the ring current near 0100 and 0900 UT on February 22. In the previous section we pointed out that there was an increase in proton flux observed at near-geosynchronous altitude around this time. Again, we note that at S3-3 altitudes the dominant species is oxygen, although the proton number density is much higher than that shown in Figure 4.
Fig. 5. Composition data from the S3-3 spacecraft acquired during the recovery phase of the storm on February 21 and 22. Similar to Figure 3.
The variation of number density as a function of L shell displayed in Figure 5 shows no consistent trends as time progresses. In subsequent sections we shall argue that the signatures shown in Figures 3 and 4 are due to the recent injection and subsequent convection of ionospheric plasma. It is possible that recently injected plasma is not observed at lower L shells in the predusk local time sector because of time of flight effects. Presumably, the magnetosphere is not so disturbed during a long recovery phase, and the convection pattern within the magnetosphere is fairly static. Drift time effects may consequently be less significant, and the number density distributions shown in Figure 5 are less likely to show significant trends.
In summary, the data presented in this section indicate that for the magnetic storms of February 21 and 22, 1979, a significant amount of ionospheric plasma (for which oxygen is an indicator) is injected into the low-altitude magnetosphere. Johnson  and Balsiger  reported from five satellite measurements that ionospheric plasma appears to be injected throughout the magnetosphere at this time, and we have shown that this is indeed the case for both the SCATHA and the S3-3 orbit. Additionally, there is some structure associated with the number density enhancements, which we assume are a result of both injection and convection processes, as will be discussed in the following sections.
As an initial step in discussing the storm-related signatures presented in the previous section, we shall review some convection theory. The work of Cowley and Ashour-Abdalla [l976] is a convenient basis for this discussion, although we note that other workers, such as Chen , have also described some of the ion convection signatures associated with static drift models. It should be noted that as a first approximation we are assuming that a steady state convection model is adequate for describing the effects seen in the S3-3 composition data. This is a somewhat questionable assumption, especially since it is likely that the cross-tail convection field and the ambient magnetic field undergo significant changes during storm times. We argue, however, that the major changes occur during the main phase of the storm and only slow changes (which we shall ignore) occur during the recovery phase.
Following Cowley and Ashour-Abdalla  (hereinafter referred to as paper 1), we shall assume a uniform cross-tail electric field and a dipole magnetic field to model drifts. Particle drift paths conserve total energy , and so for ions,
(equation (1) in paper 1), where L is the L parameter normalized to L*, is the azimuth measured anticlockwise from local noon, and is the ion energy W normalized to the characteristic energy W*. L* is the L shell value at dusk for which zero-energy particles have stagnant flow, that is, d/dt = 0 and dL/dt = 0. Both L* and W* are related to the potential scale for the corotation electric field and the cross-tail electric field:
where is the potential scale for the corotation field ( 92 kV) and E is the cross-tail electric field in kilovolts per earth radius.
In order that (1) may be used to determine those values of L and that have the same total energy as L and , the functional dependence of on L must be determined. This is given by equations (2a) and (2b) in paper 1, which have the limits
for 90° and 0° equatorial pitch angle particles, as can be seen from conservation of the first and second adiabatic invariants.
Lastly, we must determine the L and corresponding to stagnant flow. From equations (3) and (4) in paper 1,
where F() is a function of equatorial pitch angle as described in paper 1 and shown in the work of Cowley and Ashour-Abdalla . The limits of F() are 3 and 2 for 90° and 0° pitch angles, respectively.
When both d/dt = 0 and dL/dt = 0, the flow stagnates. From (3b), stagnation occurs only at dusk and dawn ( = /2).We can solve (3a) to give the first adiabatic invariant and second adiabatic invariant J as a function of L for which the flow stagnates at dusk or dawn:
from 90° and 0° particles, respectively. The dawn stagnation point is given by choosing the positive sign in (4a) and (4b).
From (4a) and (4b) it can be seen that for all or J there is a single stagnation point on the dawnside of the magnetosphere, whereas there are two stagnation points on the duskside for L < 1. As was pointed out in paper 1, the dawnside stagnation point is an x-type stagnation point, whereas the duskside stagnation points consist of an x-type and an o-type stagnation point, with the o-type being at lower L. On the duskside for L > 1, and J are negative, corresponding to solutions for electrons, and only one x-type point exists.
Equations (1), (2a), (2b), (4a), and (4b) are sufficient for us to determine the energies of ions whose drift paths may map to the dawn and dusk stagnation points for any particular value. of L and . It should be noted that not all solutions of these equations correspond to valid drift paths. However, for a given particle invariant, the energy is a minimum on the duskside and a maximum on the dawnside at the same L. Drift paths lie on lines of constant energy and must remain between the two extrema given by the dawnside and the duskside energy. Consequently, we can eliminate those solutions of the above equations which do not correspond to drift paths mapping to the stagnation points.
Following paper 1, it is useful to classify the different types of drift path associated with the convection model. In Figure 6 we have plotted all values of normalized energy and L shell whose drift paths map to the x-type stagnation points for 0° equatorial pitch angle ions. The stagnation flow lines have been plotted assuming a constant local time at 1600 MLT which is representative of those local times for which S3-3 data were acquired. The dashed curves give two contours of the constant second invariant. We shall discuss the values of J for which these are plotted below.
Fig. 6. Different types of drift orbit for 0° equatorial pitch angle particles. The diflerent drift regimes have been calculated using a dipole magnetic field with a uniform cross-tail convection electric field plus corotation. The boundaries have been plotted as a function of normalized energy (W/W* = ) and normalized L shell (L/L* = L) and have been calculated assuming 1600 MLT. See the text for more details and a description of the different drift regimes.
In general, the orbits are either closed or open, and the boundary between these two regimes is often referred to as the Alfvén layer. As a starting point for defining the types of orbit, it is useful to consider the zero-energy case. For zero energy the closed orbits have azimuthal drift velocities in the same direction as corotation, and we have consequently labeled these types of orbit as "closed, corotation," although it should be noted that the ions do not drift at the corotation velocity. At higher L the drift paths become open, and some are "open through dawn," whereas others are "open through dusk." It will be noted that we have extended the analyis in paper 1 to draw this distinction, since the boundary between these two drift regimes also maps to the duskside x-type stagnation point and so may produce a signature in particle spectrograms.
As J increases, but is less than some critical value J (which we describe below), some additional type of closed orbit exists. At sufficiently low L the drift motion of finite energy particles is dominated by magnetic field drifts, which for 0° pitch angle is curvature drift. The orbits dominated by curvature drift are labeled "closed, curvature" in Figure 6. Between closed, corotation and closed, curvature, there is a regime labeled "banana." Particles in this regime have closed orbits also, but these orbits are closed around the o-type stagnation point on the duskside and never pass through the dawn meridian.
So far, all the closed orbits that we have considered lie within the Alfvén layer, and this Alfvén layer is given by drift paths mapping to the duskside x-type stagnation point. Consequently, one may consider there to be a single closed orbit regime, although the types of drift path within that regime can be complicated. With increasing J, the dawnside stagnation point and the o-type stagnation point on the duskside move to increasing L, whereas the duskside x-type stagnation point moves to decreasing L. When J = J, the drift orbit boundaries coalesce, and so at higher J there are two closed orbit regimes. One of these orbit regimes is labeled "banana, isolated" in Figure 6. Particles in this region still have closed orbits around the o-type stagnation point, but the region is separated from what we might still call the Alfvén layer, if we insist that the Alfvén layer marks the boundary between open orbits and closed orbits around the earth. Now the Alfvén layer is given by drift paths mapping to the dawnside stagnation point.
The separation of the banana orbits from the other closed orbits allows the penetration of ions to low L from the magnetotail, as discussed by Chen  and Cowley and Ashour-Abdalla . The latter compared the model calculations with the data of Smith and Hoffman  as an explanation of the "proton nose," although it was necessary to invoke loss processes to explain the apparent lack of penetration to very low L.
Returning to Figure 6, there is an additional important value of J: J, at which the o-type and x-type stagnation points on the duskside are colocated and so vanish. Above this value of J there are only two types of orbit, closed, curvature and open through dusk. It should be noted that the picture is slightly different when we consider orbits on the dawnside of the magnetosphere for high J since the open orbits can be open through dawn also.
So far, we have only considered 0° pitch angle particles. The basic morphology of the types of drift motion is the same for all pitch angles, although the actual energies mapping to the stagnation points will depend on pitch angle. The two limiting cases are 90° and 0° equatorial pitch angle, and we show stagnant flow lines for these two cases in Figure 7. The different types of drift motion are again labeled, but we now make no distinction between the different types of closed orbit. The major effect is apparent in the boundary between the open trajectories through dusk and dawn.
Fig. 7. Different drift regimes for 90° and 0° equatorial pitch angle particles. No distinction has been made between the different types of closed orbit (cf. Figure 6). The boxes drawn in the figure indicate how the S3-3 mass spectrometer energy range (0.5-16 keV/q) and the L shell range from 3 to 8 map to the normalized energy and L shell coordinate system for 0.5-, 1.-, 2.-, and 4.-kV / R cross-tail convection fields. The scale for the 0.5 convection field is the lower left; the scales move to the upper right-hand corner with increasing convection field. The boundaries have again been calculated for 1600 MLT.
As Figure 7 is intended to relate the drift boundaries to the S3-3 data shown in the previous section, we have plotted the limiting curves for 1600 MLT, as in Figure 6. Additionally, we have shown how the energy range of the S3-3 mass spectrometer and a reasonable range in L values map to the dimensionless units W/W* and L/L* for four values of cross-tail convection field strength. The lowest value of convection field used (0.5 kV / R) corresponds to the leftmost set of axes, with axes for increasing convection field strength moving to the upper right-hand side of the figure. The other three values of convection field are 1., 2, and 4. kV / R
From Figure 7 it can be seen that the greatest difference between stagnant flow lines occurs at higher L and is only significant at low L shell values for the larger convection field strengths, of the order of 4 kV/ R. For the data shown in section 4, the equatorial pitch angle of particles which are locally mirroring at the S3-3 location is typically less than 30° at L = 3 and is less than 15° by L = 5. Also, the simple model assumptions are most likely to break down for stagnant flow lines which map to stagnant points some distance from the spacecraft location, as is the case for the boundaries in the upper right of the figure. Any improvement in the estimate of stagnant flow lines using locally mirroring particles is probably offset by inaccuracies introduced by this eflfect, and in most cases it is reasonable to use 0° pitch angle particles to determine the energies which map to stagnant flow points from the spacecraft location.
In general, we can refer to the boundary between open and closed orbits as shown in Figure 7 as the Alfvén layer. On comparison with Figure 6, part of the closed/open boundary actually marks the transition between open orbits and closed banana orbits. However, the boundary curves are sufficiently close that identification with the Alfvén layer is probably adequate. It should be noted that this part of the "Alfvén layer boundary" marks a region of oppositely directed flow. Lower-energy ions drift eastward while higher-energy ions drift westward. The flow transition actually occurs over the banana orbit regime shown in Figure 6, but this region is small in the parameter space used, and we can consider the region of "shear" to be given by the convection boundary shown in Figure 7. In comparison the lower-energy part of the Alfvén layer is not marked by any similar drift shear.
While the boundaries between open orbits through dusk and dawn do not have any shear associated with them, these boundaries indicate regions where flow has been diverted around the Alfven layer. The boundary between open orbits through dusk is due to diverted flow about the isolated banana orbits. We shall consequently refer to boundaries between open drift trajectories as "diverted flow boundaries."
When presenting the low-altitude S3-3 data, we used the number densities as a means of summarizing the composition data. In the previous section we discussed convection models and described some of the convection boundaries to be expected in the magnetosphere. To do so, it was necessary to consider how the convection boundaries vary as a function of particle energy and position. Moment integrals of the particle distribution function mask such structure, and if we are to explain the signatures seen in the number density plots in terms of convection boundaries and drift motions, it is necessary to consider the particle spectra.
Plate 2 shows particle spectrograms corresponding to the three orbits shown in Figure 3. The plate contains six panels, each horizontal pair showing the proton and oxygen spectra for an individual orbit. The format of each panel is the same in this plate and in subsequent plates, and we shall describe only one panel in detail for reference.
Plate 2. Energy-L shell spectrograms of the S3-3 composition data for the first storm on February 21. Differential number flux is plotted in both a three-dimensional representation and a color spectrogram. The data correspond to the number density plots given in Figure 3. Theoretical drift boundaries have been included in the color spectrograms. The figure is described in more detail in the text.
As indicated by the label on the panel, the panel in the top leff of the plate shows proton spectra, acquired from the S3-3 spacecraft between 0826 and 0841 UT on February 21, 1979 Directly below this label, the proton spectra are displayed in a three-dimensional format. The vertical axis is differential number flux, plotted logarithmically from 10 to 10 particles cm s sr keV. It should be noted that although the upper limit on the axis is 10, higher fluxes are plotted to higher fluxes are plotted to scale, whereas fluxes below 10 are truncated to this lower limit. The particle flux is shown as a function of energy scaled logarithmically from 0.5 to 16 keV/q and L shell scaled linearly. The energy range is so chosen because this covers the energy range of the S3-3 mass spectrometer and furthermore the energy channels are spaced uniformly when plotted on a log scale. For the data shown in Plate 2, L varies from 3 to 6. This range can be different from plate to plate, whereas both the flux scale and the energy scale are constant.
Below the three-dimensional plot the same data are plotted using a color scale representation. We have not labeled this plot because of lack of space, but the energy scale and L range are the same as for the three-dimensional plot. Furthermore, we have not included a color reference scale, since the color levels can be directly compared with the corresponding values in the three-dimensional plot. The actual colors used are the same as those shown in the color scales in Plates 1a and 1b For completeness we note that the color associated with a particular flux value is scaled to the logarithm of the flux value with a flux value of 10 being black and a flux value of 10 being red. We have chosen to display the data using both types of spectrogram since both have different attributes. The three-dimensional plot is a more quantitative method of displaying data, whereas the color plot allows us to qualitatively compare data with theoretical models. We can consider the three-dimensional plot as a constraint when interpreting features in the color spectrograms.
The last element in the panel is the projection of the convection boundaries, as described in the previous section, onto the color spectrogram. The theoretical boundaries are given by the white line traces, with the label to the left showing the convection fields and particle pitch angle used in the model. As has been discussed in the previous section, we have assumed that the particles can be well represented by 0° pitch angle particles at the spacecraft location. For this particular case, we have chosen two convection field strengths, 3.0 and 2.5 kV / R It should be noted that in general the part of the convection boundary trace which goes to low energy does so at lower L values for higher convection field strengths (cf. Figure 7). That is, the convection boundary trace corresponding to the 3.0-kV / R convection field is the leftmost trace in each panel.
The convection boundaries shown in Figure 7 were plotted at a constant magnetic local time (MLT) as a function of L and . In Plate 2, rather than use some average local time for a particular orbit, we have used the MLT and L shell as given by the S3-3 ephemeris. This produces some distortion in the convection boundaries when compared with those shown in Figure 7, as MLT is not constant along an orbit. Nevertheless, the general form of the boundaries will be similar.
When displaying the number densities corresponding to the three orbits shown in Plate 2, we noted that there appeared to be a movement to lower L shell as time progressed. Continued injection of new ionospheric plasma at the point of observation is probably not responsible for this phenomenon, since the data were obtained during the recovery phase of the first storm on February 21, 1979. It is apparent from the spectrograms that the increase in number density at lower L is associated with a flux enhancement, but this flux enhancement does not occur at all energies. In fact, for the protons this enhancement is more marked at energies below 2 or 3 keV, with some enhancement at energies around 10 keV. There is a distinctive minimum between these energies, and the location of the minimum in the spectra is also a function of L shell. For the oxygen fluxes this minimum is also observed, but the flux enhancement is greater for the higher energies than for the lower energies.
The convection field strengths used in Plate 2 were chosen because they gave a reasonable qualitative fit to the minima seen in the spectra at lower L values. Bearing in mind that at present we have not considered the implications of attributing a particular feature in the particle spectra to a convection boundary, the 2.5-kV / R convection field model appears to give a better fit to the proton signatures, whereas the 3.0-kV / R model is a slightly better fit for the oxygen spectra. Given that the fit is at best qualitative, the difference between the two models is probably insignificant and gives only an estimate of the convection field strength. It is worth noting that the convection field strength is reasonable; Kivelson  has shown that a convection field near 2 kV/RE is to be expected at high Kp values.
The drift boundary shown in Plate 2 which is roughly constant in energy corresponds to a diverted flow boundary. There appears to be some evidence in the color spectrograms of a signature in the proton fluxes associated with this boundary. The minimum is not as marked when considering the three-dimensional representation of the data, and so a certain amount of caution should be exercised when interpreting this feature. Nevertheless, that such a feature is present adds support to the supposition that the structure in the particle spectra is due to convection effects.
It should be noted that the close correspondence of a minimum in the particle spectra with a convection boundary as predicted by theory is not in itself sufficient to explain the presence of the minimum. If we assume that loss processes are not significant during the early recovery phase of the storm, then additional spatial and/or temporal variations in the ion distributions must be invoked to explain the signatures. For example, the flux enhancement of the proton fluxes at low energies and low L shells during the recovery phase implies that the proton fluxes were low in the early afternoon local time sector prior to the main phase of the storm since the drift boundaries indicate that these lower-eriergy protons have drifted eastward.
Another point concerning the low-energy proton fluxes is the fact that the minimum around L = 4 appears to be associated with a drift boundary that maps to the duskside stagnation although these protons are drifting eastward and the S3-3 spacecraft was around 1600 MLT when the data were acquired. Consequently, some radial dependence to the low- energy proton fluxes may also have to be invoked.
Turning to the oxygen data shown in Plate 2, the major enhancement at low L occurs at the higher energies. From the model we infer that these ions are on open trajectories, convecting through dusk. This plasma must have been injected tailward of the observation point.
Summarizing Plate 2, the minimum apparent in the spectra at low L shells is consistent with time of flight effects, with the lower-energy protons convecting eastward and the higher- energy oxygen ions convecting westward. These two populations are drifting in opposite directions; the minimum in the spectra is a signature of this oppositely directed flow. Additionally, there is some evidence for the existence of diverted flow for the protons at higher L shells. The presence of this signature implies that these protons have convected from the nightside. The minimum at higher L is not as marked in the oxygen spectra, which may be explained if oxygen rich plasma has been injected over a large range in local times.
Plate 3 shows data for the second storm on February 21. The data shown in the top two panels were acquired during the main phase of the storm, near 2000 UT, whereas the data shown in the bottom four panels were acquired during the recovery phase. There is some evidence for a minimum in the particle spectra in the top two panels, ond we have shown the drift boundaries for a 2.0-kV / R convection field. Assuming the model is correct, then the lower-energy fluxes at high L are produced by particles on open orbits drifting through dawn. The boundary between open orbits through dusk and dawn does not appear to have any marked signature in the particle fluxes.
Plate 3. Energy-L shell spectrograms for the second storm on February 21. Similar to Plate 2. The corresponding number density information is given in Figure 4.
The data acquired during the recovery phase, shown in the bottom four panels, indicate that the density enhancement shown in Figure 4 is due to an increase in the lower-energy fluxes at low L. Minima in both the proton and oxygen spectra are apparent. The proton flux levels are considerably lower than the corresponding oxygen fluxes, and the proton minima are much wider in energy. Considering the oxygen data shown in the middle panel, the minimum is very narrow and runs from L = 2, W 15 keV to L = 7, W 1 keV. To explain this minimum in terms of convection, we have included drift boundaries for 4.0- and 2.0-kV/R convection fields. The 4.0-kV/R trace is the leftmost in each display. The 4.0-kV/R convection field model appears to give a reasonable fit to the minimum in the oxygen spectra. The minimum corresponds to diverted flow of particles on open drift paths, with the lower-energy ions drifting through dawn and the higher-energy ions drifting through dusk. It is also tempting to attribute the decrease in flux at L < 3 shown in the bottom right-hand panel to the Alfvén layer boundary. However, a 4.0-kV/R field is rather large, and this boundary does not have any corresponding signature in the oxygen spectra acquired near 0400 UT.
A 2.0-kV/R field can also model the minima in both the proton and oxygen spectra fairly well. If this field strength is correct, then at lower L values the minimum in the spectra corresponds to the shear associated with the banana orbit regime, whereas at higher L the boundary is again given by the diversion of flow around the Alfvén layer. There is no corresponding particle signature associated with the Alfvén layer at lower energies. Since this part of the Alfvén layer boundary is given by eastward convecting ions whose drift path maps to the duskside stagnation point, we should not expect any corresponding particle signature if time of flight effects alone are responsible for these signatures.
The data in Plate 3 show that if the minima in the spectra are associated with a diversion of flow, then this signature is relatively insensitive to the choice of convection field. It appears then that we could assume a convection field strength in the range of 2 to 4 kV/R. Following Kivelson , it is more likely that the convection field is nearer to 2 kV/R, rather than 4 kV/R.
When data acquired well into the recovery phase of the storm on February 22 were considered, no significant trends were found in the number density plots, as shown in Figure 5. This is consistent with time of flight effects being less noticeable long after the initial injection of the drifting plasma, which we assume occurred sometime around the main phase of the storm. Nevertheless, convection boundary signatures should still be present, although the primary cause of these signatures will be loss processes such as pitch angle diffusion and charge exchange. Plates 4a and 4b show spectrograms corresponding to the number density plots shown in Figure 5. We have again plotted convection boundaries on the color spectrograms.
Plate 4a. Energy-L shell spectrograms for the first two orbits shown in Figure 5. Similar to Plate 2.
Plate 4b. Energy-L shell spectrograms for the second two orbits shown in Figure 5. Similar to Plate 2.
In Plate 4a it can be seen that the leftmost trace, which marks the Alfvén layer for a 3.0-kV/R convection field, has an associated particle signature in both the proton and oxygen spectra. In addition to this feature, decreases in particle flux may be associated with the flow diversion around the Alfvén layer. These signatures are not as marked, and so some caution should be exercised, but that these features are present supports the supposition that the structure in the spectra can be interpreted in terms of convection boundaries. We have included drift boundaries assuming a 2.0-kV/R convection field for comparison with data presented in In Plate 4b.
The data shown in Plate 4a were acquired close to dusk, whereas the data in Plate 4b were obtained on orbits near 1500 MLT. For this reason, in Plate 4b the diverted flow drift boundary covers a much larger range in L shells for the two convection field strengths chosen. The location of the flow boundary for open orbits is somewhat insensitive to the choice of convection field strength. However, the Alfvén layer boundary, which lies in the lower left part of the spectrograms, suggests that the 2.0-kV/R convection field strength models the proton signatures somewhat better. The oxygen spectra do not show as clear a loss signature at the low L shells and low energies. Presumably, shielding effects are becoming more important, and so a decrease in convection field strength may be expected. The model does not take into account shielding, but the model will give a rough approximation, and the fact that the direction of change in convection field strength is consistent with increased shielding is encouraging.
The signatures seen in the particle spectra shown in Plates 4a and 4b are presumably due to loss processes affecting those particles whose drift velocities are slow. As the data were obtained well into the recovery phase of the storm, we do not expect to see density enhancements associated with time of flight effects for recently injected plasma. We have included the spectrograms for this time period, however, to give additional evidence for our interpretation of the particle signatures as due to convection drift boundaries. It is interesting to note that to explain the features seen in the particle spectra, we have had to include both the boundaries between open and closed orbits and the boundaries signifying diverted flow for open drift paths.
As has been mentioned in the previous section, interpreting the particle signatures as observed at S3-3 altitudes in terms of convection boundaries has implications for the distribution of the observed ions at the time of the main phase of the storm. We shall discuss this point in more detail here, taking into account the other ion composition data available at this time.
When describing the multiple spacecraft observations associated with the storm period we have considered, Johnson  stated that the magnetosphere appeared to fill with O around the main phase of the storms, in that marked increases in oxygen density were observed at several different spacecraft locations. From our analysis of the S3-3 data we must deduce that the injection of ionospheric plasma into the magnetosphere at these times has some local time and radial dependence to it. Another point to be considered is whether the ions observed at both SCATHA and S3-3 altitudes were indeed recently injected from the ionosphere or were present within the magnetosphere for some time prior to the observation, but at larger radial distances than those covered by the spacecraft. Lennartsson and Sharp  have shown that there is a steep gradient in the oxygen to proton density ratio at quiet times, the ratio decreasing by an order of magnitude as L increases from 5 to 10. This does indicate that for the first storm, at least, most of the oxygen ions observed at SCATHA and S3-3 have been recently injected into the magnetosphere.
The density enhancement observed at lower L shells at S3-3 altitudes during the recovery phase of the first storm on February 21, as shown in Figure 3, is mainly due to the enhanced lower-energy fluxes, as shown in Plate 2. From the discussion in the previous section we have deduced that the lower-energy ions are on closed drift paths and moreover are drifting in the same direction as corotation. The time delay associated with the flux enhancement at low L implies that the lower-energy ions were located westward of local noon at the time of the main phase of the storm. Although oxygen is the major component of the plasma in terms of number density, the major flux enhancement of eastward convecting ions is observed in protons.
Although Ghielmetti et al.  have shown that upward flowing ions (both beams and conies) are preferentially observed in the dusk local time sector, Gorney et al.  found that low-energy conies are preferentially observed on the dayside at S3-3 altitudes during times of low magnetic activity. It may be that these conies are in fact the source for the low-energy ions observed to be drifting eastward during the recovery phase of the storm. We note that Sharp et al  have suggested that a dayside source is required to explain the observations of precipitating oxygen at 800 km altitude. At times of high activity, however, conies are observed at all local times. Additionally, the dayside conies are most often seen at high latitudes around the polar cusp region.
Additional support for an ionospheric source for the low- energy ions observed at low L shells is the work of Lundin et al. [1982b] and Hultqvist [1983a, b]. These authors have deduced that significant amounts of ionospheric plasma (specifically O ) are injected into the dayside magnetosphere during magnetic storms, using data from Prognoz 7. In his review article, Hultquist [1983a] specifically addressed data for the storm period of February 21, 1979. He deduced that O was injected over a wide latitude range. The data shown in Figure 3 appear to agree with this conclusion, but the data in Plate 2 indicate that there are differences in the spectra for each particle species. Not all the oxygen ions responsible for the density enhancement have convected from the dayside.
One explanation for the dayside source of ions is the injection from the ionosphere, as discussed above. However, one must invoke compositional changes as a function of latitude, since the Prognoz 7 data show O to be the dominant ion at higher L shells. Alternatively, the eastward convecting ions could be part of a preexisting magnetospheric plasma population that has been convected to low L shells because of an enhanced convection field. These ions are then trapped on closed orbits as the convection field relaxes during the recovery phase.
An additional complication to be considered when discussing the source of the low-energy plasma is that in Plate 2 both species show a flux signature which appears to be related to a duskside stagnation point (i.e, the Alfvén layer flow boundary). However, these particles are on drift paths which pass through dawn. The presence of a flux signature near the flow boundary consequently implies that the processes responsible for the injection of the plasma are affected by the presence of the flow boundary. That is, the prestorm Alfvén layers control the storm time phenomena. Specifically, the processes responsible for the extraction of ionospheric plasma during the main phase of the storm may be different inside and outside side of the Alfvén layer. For example, we could assume some height dependence on the extraction process which would be reflected in compositional and/or flux changes in the injected plasma as a function of location with respect to the Alfvén layer.
The presence of the Alfvén layer-related signature appears to support the assumption that the observed ions have all been injected from the ionosphere. However, the discussion presented here is only indicative, and modified convection of a preexisting plasma population cannot be totally discounted as a possibility for explaining the low-energy, low-L shell observations shown in Plate 2.
The picture is simpler for those particle fluxes observed at S3-3 altitudes which we have deduced as being on open drift paths. The proton spectra presented in Plate 2 indicate that the protons were initially in the night sector during the main phase of the storm, as the spectra show signatures associated with both the dawnside and duskside stagnation points. The lack of a dawn-related signature in the oxygen spectra implies that the injection of these ions was not confined to just the night sector. From the SCATHA data shown in Plate la, enhanced oyxgen fluxes are observed for some time after 0600 UT. The first peak in Dst occurred around 0900 UT. Because of the magnetosheath encounter following 1600 UT, it is not certain that this flux enhancement continues through to the dayside magnetosphere from SCATHA data alone. However, Prognoz 7 data [Johnson, 1981; Hultqvist, 1982] show that large fluxes of oxygen ions were also observed in the dayside magnetosphere after 1100 UT, when the spacecraft first enters the magnetosphere. It should further be noted that the flux enhancements are indeed temporal and SCATHA data from the previous day (not shown) show low oxygen fluxes throughout the orbit.
During the second storm period on February 21, both SCATHA and S3-3 composition data again show a large increase in oxygen content. Dispersion signatures are observed in the energy-time spectrograms of SCATHA data following 2000 UT as shown in Plates 1a and 1b, for both protons and oxygen. In Plate 1b, two dispersion signatures can be seen in the oxygen data, the first passing through 100 eV near 0200 UT on February 22 and the second passing through 100 eV at 0400 UT. This suggests that additional oxygen ions are injected into the magnetosphere at these times, augmenting the preexisting ionospheric plasma injected during the first storm time period. At SCATHA altitudes the enhanced oxygen fluxes are observed primarily in the afternoon and evening local time sectors. Because of the magnetosheath encounter prior to 2000 UT on February 21 the initial dispersionlike signature may be due to spatial structure, rather than temporal effects. The protons also show dispersion signatures at this time.
The data acquired from the S3-3 spacecraft also impose some constraints on the local times at which these enhancements occurred. Signatures in the data shown in Plate 3 indicate that at this time the convection field is of the order of 2.0-kV/R. The density enhancements shown in Figure 4 are due to the eastward convection of low-energy ( 1 keV) ions. By way of contrast to the previous storm, no signature corresponding to the Alfvén layer is observed in the ion spectra. This may be because the second storm occurred in an active period, rather than following a substantial period of low activity. S3-3 data imply that at higher energies and higher L shells, most of the flux enhancement initially occurred in the postdusk local time sector, as evidenced by the signature given by the boundary between open and banana orbits and by the boundary signifying diverted flow of open orbits around the Alfvén layer. Data acquired in the early morning local time sector (not shown) imply that the injection of higher-energy ionospheric plasma was further confined to the premidnight local time sector.
Some consideration must be given to the fact that at the higher L shells the low-energy particles which are convecting eastward have a signature which is related to the dawnside stagnation point. For the first storm we deduced that the lower-energy ions were probably injected in the prenoon local time sector. For the data shown in Plate 3 some of these ions must have been injected at even earlier local times, close to or earlier than dawn. To verify this point, some form of drift-time analysis should be performed to show that particles which are at fairly high L shells can convect from dawn to the S3-3 spacecraft location in a reasonable amount of time. We can give a rough estimate of the time scales involved. From Figure 1 we note that for the second storm, Dst begins to decrease around 1800 UT. The data shown in the middle two panels of Plate 3 were acquired near 0400 UT at roughly 1500 MLT. The low-energy ions will convect at roughly the corotation velocity and so were near dawn when Dst first showed a decrease.
We can attempt to summarize the requirements for the source locations and characteristic energies of the plasmas observed at the S3-3 orbit that are necessary if we are to interpret the signatures associated with the two storm periods on February 21 and 22, 1979. First, the characteristic energy of the injected plasma should increase with increasing L shell. That is, at low L shells, most of the plasma has energies less than a few (say 5) keV, whereas at higher L shells, the plasma can consist of ions with energies from 100 eV up to some tens of keV. Second, as the characteristic energy increases, the local time of injection moves from near noon through dawn to the premidnight local time sector. The highest-energy particles are injected in the postdusk premidnight sector; the lowest-energy particles are primarily injected in the morning sector, although some low-energy particles are injected at most local times. One point should be made concerning the "injection" of plasma. We have not distinguished between enhanced convection to lower L shells and the direct injection of new ionospheric plasma into the magnetosphere. We can use the composition data to guide us in our deductions concerning the sources of the plasma for the first storm period, but this is not as simple for the second storm period. A much more quantitative analysis of the drift and loss mechanisms is required to clarify this point, as we do not know how much of the recently injected ionospheric plasma is still present within the magnetosphere.
The major constraint on the source characteristics of the plasmas detected is that very little flux is injected in the afternoon local time sector, where the S3-3 observations were made. This is apparently in contradiction to the results of Ghielmetti et al.  and Gorney et al. , who showed that ions were injected into the magnetosphere throughout the dusk local time sector. If the location of the more energetic ion beams and conies is confined to the auroral oval, then this contradiction may not be as serious as first thought. Most of the observations considered in this study were made on L shells whose invariant latitude was less than 70°. Since ions drift to higher L shells on the dayside, only those ions injected on the nightside auroral oval are likely to be observed in the L shell range we have used.
The SCATHA data shown in Plates 1a and 1b indicate that for the first storm this may be the case, since enhanced higher-energy fluxes are initially observed after dusk. It should be noted that this could be a universal time effect; Dst only begins to decrease around 0600 UT, when SCATHA is near local midnight. For the second storm the data are not clear. Because of the magnetosheath encounter it is difficult to determine when freshly injected plasma is first observed. The presence of dispersionlike signatures does imply recent injection of plasma, although from a single spacecraft alone, we cannot determine uniquely the temporal and spatial history of the injection process. However, the SCATHA data are not inconsistent with the injection in the postdusk local time sector of most of the new ~ 10-keV ionospheric plasma observed below L 7, with subsequent drift motions bringing this plasma to the afternoon local time sector.
After the disturbed period, the number density data in Figures 2 and 5 show no general trends, the number densities are roughly constant at SCATHA altitudes, and although there is some variation in the number densities at S3-3 altitudes, features such as those observed during the storm time periods are not present. The SCATHA data shown in Plate 1b and the S3-3 data shown in Plates 4a and 4b show the formation of signatures in the particle spectra which are probably due to loss processes such as charge exchange and pitch angle diffusion. The spectrograms of SCATHA data indicate the formation of the "deep minimum." This can be more readily seen in the oxygen data after 1600 UT. The S3-3 spectrograms show that convection boundary signatures are still present, but time of flight effects do not appear to be the primary cause for these signatures.
For the first storm on February 21, 1979, the S3-3 composition data imply that at higher L shells (L 4), oxygen was injected in both the nightside and the morning local time sector. This conclusion was drawn since the main convection feature in the oxygen spectra was due to the stagnation of flow through the duskside of the magnetosphere and there was little evidence for a similar feature due to the dawnside stagnation point. The proton spectra did show such a feature, however, and so presumably were confined to the nightside sector during the main phase of the storm, at which time we assume the plasma enhancement initially took place.
In addition a low-energy enhancement was observed at L 4 some 3 hours into the recovery phase of the first storm. Time of flight considerations lead us to suggest that the presence of the minimum in the particle spectra at low energies near L = 4 was due to the process responsible for the injection of the plasma being affected by preexisting boundaries within the magnetosphere. That no similar minimum was observed in the particle spectra following the second storm on February 21, 1979, supports this inference. Presumably, the ion Alfvén layers have not reestablished themselves following the preceding storm.
The low L shell density enhancement as observed at the S3-3 orbit during the second storm was again due to convection of low-energy plasma from the dawnside of the magnetosphere. For this storm the amount of oxygen at low L shells was much higher than in the previous storm. While injection of ionospheric plasma on the dayside is a possible cause for this signature, we could also attribute this enhancement solely to enhanced inward convection of magnetospheric plasma. The previous storm injected large amounts of ionospheric plasma into the nightside magnetosphere.
At the same time as low-energy ions were observed by S3-3 at low L shells, higher-energy ions were observed convecting from the nightside of the magnetosphere following the second storm. The presence of both dusk- and dawn-related signatures in the spectra implies that the ions observed at higher L shells were initially injected in the nightside magnetosphere. The SCATHA mass composition data are consistent with significant injections of ionospheric plasma in the nightside magnetosphere at the higher L shells because of the presence of multiple dispersionlike features in the particle spectra.
As a justification for assuming that convection drift boundaries are responsible for the signatures observed in the S3-3 ion data, we included data acquired well into the recovery phase of the second storm in this period. The number density data did not show any trends, such as those observed earlier in the recovery phase of the storms, which we have attributed to time of flight effects. Nevertheless, correspondences between drift boundaries and features within the particle spectra were apparent. Presumably, loss processes are more important at this time, and the loss of ions with low drift velocities is responsible for the illumination of the boundaries by minima in the particle spectra. Additionally, the data acquired well into the recovery phase show some evidence for the onset of shielding. We noted that later in the recovery phase a 2.0-kV/R cross-tail electric field seemed to give a better fit to the particle spectra, whereas somewhat earlier a 3.0-kV/R convection field gave a better fit.
In conclusion we have shown that significant amounts of ionospheric plasma are injected into the magnetosphere during storm time periods. Drift analysis indicates that ionospheric plasma is being injected into the nightside magnetosphere over the energy range of under 1 keV to some tens of keV, around L 5. While injection may also occur at L > 8, these ions do not convect to the S3-3 orbits used in the present study. At lower L shells, L 4, lower-energy ( 5 keV) plasma is injected into the dayside magnetosphere. Time of flight calculations indicate that these ions are primarily injected in the prenoon local time sector. Subsequent convection results in an enhancement of plasma density at low L during the recovery phase, rather than the main phase of the storms occurring on February 21 and 22, 1979. At present, we cannot determine whether the ionospheric plasma injection at low L shells is due to enhanced inward convection of plasma previously injected in the nightside magnetosphere or is due to extraction of ionospheric plasma on the dayside around the time of the main phase of the storm. Detailed time-dependent convection models are required if a more quantitative explanation of the particle signatures is to be found. We note that significant effort is required to produce time-dependent convection models such as that reported by Wolf et al.  when modeling the magnetic storm of July 27, 1977. Particle populations representing the ring current and plasma sheet were included in such a model. It would be interesting to consider the evolution of additional sources of plasma, such as those inferred from the present study, specifically the large amounts of ionospheric plasma which we deduce to have been injected into the magnetosphere.
Acknowledgments. One of us (R.J.S.) would like to thank S. W. H.
Cowley of Imperial College, London, for supplying drift analysis
programs. Extensive use was made of this software in developing the
analysis techniques presented here. This work was funded by NASA
under contract NASW 3395, the Office of Naval Research under
contracts N000014-76-C-0444 and N000014-78-C-0479, the National
Science Foundation under grant ATM 8317710, and the Lockheed
independent research program.
The Editor thanks R. Lundin and T. E. Moore for their assistance in evaluating this paper.
Balsiger, H., Composition of hot ions (0.1-16 keV/e) as observed by the GEOS and ISEE mass spectrometers and inferences for the origin and circulation of magnetospheric plasmas, Adv. Space Res., 1, 289, 1981.
Chen, A. J., Penetration of low-energy protons deep into the magnetosphere, J. Geophys. Res., 75, 2458, 1970.
Cowley, S. W. H., and M. Ashour-Abdalla, Adiabatic plasma convection in a dipole field: Variation of plasma bulk parameters with L, Planet. Space Sci., 23, 1527, 1975.
Cowley, S. W. H., and M. Ashour-Abdalla, Adiabatic plasma convection in a dipole field: Proton forbidden-zone effects for a simple electric field model, Planet. Space Sci., 24, 821, 1976.
DeForest, S. E., and C. E. Mcllwain, Plasma clouds in the magnetosphere, J. Geophys. Res., 76, 3587, 1971.
Ghielmetti, A. G., R. G. Johnson, R. D. Sharp and E. G. Shelley, The latitudinal, diurnal, and altitudinal distributions of upward flowing energetic ions of ionospheric origin, Geophys. Res. Lett., 5, 59, 1978.
Gorney, D. J., A. Clarke, D. Croley, J. Fennell, J. Luhmann, and P. Mizera, The distribution of ion beams and conics below 8000 km J. Geophys. Res., 86, 83, 1981.
Hultqvist, B., Recent progress in the understanding of the ion composition in the magnetosphere and some major question marks, Rev. Geophys. Space Phys., 20, 589, 1982.
Hultqvist, B., On the dynamics of the ring current, J. Geophys., 52, 203, 1983a.
Hultqvist, B., On the origin of the hot ions in the disturbed dayside magnetosphere, Planet. Space Sci., 31, 173, 1983b.
Johnson, R. G., Review of the hot plasma composition near geosynchronous altitude, Proceedings of Spacecraft Charging Technology 1980 Conference, edited by N. J. Stevens and C. P. Pike, NASA Conf: Publ., NASA CP-2182, 412, 1981.
Johnson, R. G., R. D. Sharp, and E. G. Shelley, Observations of ions of ionospheric origin in the storm-time ring current, Geophys. Res. Lett., 4, 403, 1977.
Johnson, R. G., R. J. Strangeway, E. G. Shelley, J. M. Quinn, and S. M. Kaye, Hot plasma composition results from the SCATHA spacecraft, in Energetic lon Composition in the Earth's Magnetosphere, edited by R. G. Johnson, p. 287, Terra Scientific, Tokyo, 1983.
Kaye, S. M., R. G. Johnson, R. D. Sharp, and E. G. Shelley, Observations of transient H and O bursts in the equatorial magnetosphere, J. Geophys. Res., 86, 1335, 1981.
Kivelson, M. G., Magnetospheric electric fields and their variation with geomagnetic activity, Rev. Geophys. Space Phys., 14, 189, 1976.
Lennartsson, W., and R. D. Sharp, A comparison of the 0.1-17 keV/e ion composition in the near equatorial magnetosphere between quiet and disturbed conditions, J. Geophys. Res., 87, 6109, 1982.
Lennartsson, W., R. D. Sharp, E. G. Shelley, R. G. Johnson, and H. Balsiger, lon composition and energy distribution during 10 magnetic storms, J. Geophys. Res., 86, 4628, 1981.
Lundin, R., B. Hultqvist, N. Pissarenko, and A. Zackarov, The plasma mantle: Composition and other characteristics observed by means of the Prognoz-7 satellite, Space Sci. Rev., 31, 247, 1982a.
Lundin, R., B. Hultqvist, E. Dubinin, A. Zackarov, and N. Pissarenko, Observations of outflowing ion beams on auroral field lines at altitudes of many earth radii, Planet Space Sci., 30, 715, 1982b.
Lyons, L. R., and D. J. Williams, A source for the geomagnetic storm main phase ring current, J. Geophys. Res., 85, 523, 1980.
Mauk, B. H., and C. E. Mcllwain, Correlation of Kp with the substorm-injected plasma boundary, J. Geophys. Res., 79, 3193, 1974.
Mcllwain, C. E., Plasma convection in the vicinity of the geosynchronous orbit, in Earth's Magnetospheric Processes, edited by B. M. McCormac, p. 268, D. Reidel, Hingham, Mass., 1972.
Moore, T. E., R. L. Arnoldy, J. Feynman, and D. A. Hardy, Propagating substorm injection fronts, J. Geophys. Res., 86, 6713, 1981.
Sharp, R. D., R. G. Johnson, and E. G. Shelley, The morphology of energetic O ions during two magnetic storms: Temporal variations, J. Geophys. Res., 81, 3283, 1976.
Sharp, R. D., R. G. Johnson, and E. G. Shelley, Observations of an ionospheric acceleration mechanism producing energetic (keV) ions primanly normal to the geomagnetic field direction, J. Geophys. Res., 82, 3324, 1977.
Smith, P. H., and R. A. Hoffman, Direct observations in the dusk hours of the characteristics of the storm time ring current particles during the beginning of magnetic storms, J. Geophys. Res., 79, 966, 1974.
Spjeldvik, W. N., and T. E. Fritz, Observations of energetic helium ions in the earth's radiation belts during a sequence of geomagnetic storms, J. Geophys. Res., 86, 2317, 1981a.
Spjeldvik, W. N., and T. E. Fritz, Energetic heavy ions with nuclear charge Z 4 in the equatorial radiation belts of the earth: Magnetic storms, J. Geophys. Res., 86, 2349, 1981b.
Strangeway, R. J., and R. G. Johnson, Mass composition of substorm-related energetic ion dispersion events, J. Geophys. Res., 88, 2057, 1983.
Tinsley, B. A., Evidence that the recovery phase ring current consists of helium ions, J. Geophys. Res., 81, 6193, 1976.
Wolf, R. A., M. Harel, R. W. Spiro, G.-H. Voigt, P. H. Reiff, and C.-K. Chen, Computer simulation of inner magnetospheric dynamics for the magnetic storm of July 29, 1977, J. Geophys. Res., 87, 5949, 1982.
Young, D. T., H. Balsiger, and J. Geiss, Correlations of magnetospheric ion composition with geomagnetic and solar activity, J. Geophys. Res., 87, 9077, 1982.
Return to the top of the document
Go to R.J. Strangeway's homepage
Converted to html by P.R. Schwarz and Shaharoh Bolling
Last Updated Aug. 10, 2000