The ion experiment onboard the Interball-Aurora satellite ; initial results on velocity-dispersed structures in the cleft and inside the auroral oval

The Toulouse ION experiment  ̄own on the Russian Interball-Aurora mission performs simultaneous ion and electron measurements. Two mass spectrometers looking in opposing directions perpendicular to the satellite spin axis, which points toward the sun, measure ions in the mass and energy ranges 1±32 amu and 0±14 000 eV. Two electron spectrometers also looking in opposing directions perform measurements in the energy range 10 eV±20 000 eV. The InterballAurora spacecraft was launched on 29 August 1996 into a 62.8° inclination orbit with an apogee of 3 RE. The satellite orbital period is 6 h, so that every four orbits the satellite sweeps about the same region of the auroral zone; the orbit plane drifts around the pole in 9 months. We present a description of the ION experiment and discuss initial measurements performed in the cusp near noon, in the polar cleft at dusk, and inside the proton aurora at dawn. Ion-dispersed energy structures resulting from time-of̄ight eects are observed both in the polar cleft at 16 hours MLT and in the dawnside proton aurora close to 06 hours MLT. Magnetosheath plasma injections in the polar cleft, which appear as overlapping energy bands in particle energy-time spectrograms, are traced backwards in time using a particle trajectory model using 3D electric and magnetic ®eld models. We found that the cleft ion source is located at distances of the order of 18 RE from the earth at about 19 MLT, i.e., on the  ̄ank of the magnetopause. These observations are in agreement with  ̄ux transfer events (FTE) occurring not only on the front part of the magnetopause but also in a region extending at least to dusk. We also show that, during quiet magnetic conditions, time-of̄ight ion dispersions can also be measured inside the dawn proton aurora. A method similar to that used for the cleft is applied to these auroral energy dispersion signatures. Unexpectedly, the ion source is found to be at distances of the order of 60±80 RE, at the dawn  ̄ank of the magnetosphere. These results are discussed in terms of possible entry, acceleration, and precipitation mechanisms.


Introduction
Direct measurements of suprathermal plasma in the earth's magnetosphere are essential for the understanding of the ®lling of the magnetosphere with ionospheric and solar wind plasmas, for the processes at the root of electromagnetic energy dissipation in the auroral zones, and for solar wind energy accumulation in the magnetospheric tail. Simultaneous measurements performed in key regions such as the ionosphere and the magnetospheric outer boundaries are critical for understanding the dynamics of the complete system. Such simultaneous large-scale measurements are one of the main aims of the Interball project, which consists of two widely spaced main satellites, the tail probe (TP) (apogee $200 000 km) and the auroral probe (AP) (apogee $20 000 km), the tail probe being associated with a subsatellite for mesoscale measurements (Galeev et al., 1996).
The ION instrument¯own onboard the AP satellite was developed to measure auroral ions and electrons. This instrument allows detailed studies of : · the dynamics of the polar cap and of the auroral zone plasma and ®eld under the in¯uence of interplanetary medium changes and internal processes such as substorms; · the acceleration and precipitation mechanisms of electrons and ions (H + , He + , He ++ , O + ) in discrete and diuse structures and especially at the footprint of the magnetospheric boundaries; · the local heating of ions from the ionosphere and their ejection into the magnetosphere; · the ionospheric signatures of the acceleration and transport of suprathermal particles at the magnetopause and in the magnetospheric tail; · the processes of turbulent heating and acceleration of particles by powerful ground-based VLF-transmitters.
In this paper, after a description of the experiment, we focus on the ionospheric signatures of the solar plasma entry and acceleration processes in the magnetosphere using mainly hydrogen and electron data. Other initial ION data pertaining in particular to ionospheric ion out¯ow can be found in the companion paper by Dubouloz et al. (1998).

Basic measurement capabilities
The ION instrument consists of two energetic ion composition spectrometers (ION-1 and ION-2), and of two energetic electron spectrometers . Figure 1 shows the block diagrams of the sensor heads. The two ion and two electron spectrometers view opposite directions, perpendicular to the spin axis of the satellite, which is directed toward the sun. However the viewing directions of the ion and electron spectrometers are dierent. The spin period of the satellite is $120 s.
ION-1 and ION-2 provide routine measurements of H + and O + , scanned successively either in the energy range $0±14 000 eV/Q or in the range $0±120 eV/Q, where Q is the ion charge. Four ions, H + , He ++ , He + , and O + , or all masses between 1 and 64 amu, can also be measured. The geometrical factors of the two spectrometers are identical and depend on the energy, ranging from 2 to 6´10 A3 á E(keV) cm 2 á ster á keV where E is the ion energy. The mass resolution at the focus, waDw, is greater than 10 ( Fig. 2).
The experiment electronics includes a low-voltage power converter, and a data processing unit, based on an 80C86 CPU in charge of the data acquisition and compression and the interface with the telemetry systems. The DPU also commands the electronics for the sensor operational modes and the devices for communication with the onboard magnetometer and with the satellite service systems.

Ion composition analyzers
The design of the detector heads of the ION-1 and ION-2 mass spectrometers is based on the principle of the Wien ®lter, where the mass ®ltering is performed by crossed electric and magnetic ®elds (e.g., Ghielmetti, 1974). The energy over charge (ia) selection is made by a cylindrical electrostatic analyzer placed in front of the ®lter (Fig. 1). The design of the ION-1 and ION-2 sensors is quite similar to that of the mass spectrometers previously¯own onboard GEOS-1/2, ISEE-1, DE-1, AUREOL-3, and AMPTE/CCE (Balsiger et al., 1976;Shelley et al., 1978Shelley et al., , 1981Shelley et al., , 1985Bosqued et al., 1982). Particles enter the ion spectrometers through a collimator. All positive ions are then accelerated through a potential of A2000 V. The remainder of the analysis system is maintained at this negative centerline potential. Therefore, ions enter the ®rst electrostatic analyzer with an energy per charge i int (eV) i ext 2000. This analyzer selects particles with a ratio i int a 4X94. The voltage applied to the analyzer plates is programmable in 32 steps allowing the subdivision of the energy range from 2000 to 16 000 eV/Q internal to the analyzer. In terms of ion energy outside the analyzer, the mean particle energy per charge is given by: The FWHM energy pass-band is $4% of the internal energy ( Fig. 3) so that: Di I ai I increases greatly at low energies and for i 0, all ions with energies between the satellite potential and $70 eV are selected by the electrostatic analyzer. Between the end of the collimator and the entrance of the electrostatic analyzer, a retarding potential analyzer (RPA) can be switched on to subdivide this energy range. The RPA can be used to select ions into 31 energy bands between V RPA,i and 70 eV (V RPA,i being the voltage applied to the RPA central grid). However this mode is rarely used, the Hyperboloid experiment being dedicated to the analysis of ions with energies <80 eV (Dubouloz et al., 1998). After ia selection ions undergo an wa selection in the magnetic analyzer, which consists of a second cylindrical electrostatic de¯ection system placed between the poles of a Samarium-Cobalt magnet, with a Fe-Co yoke, giving a $1000-G ®eld. The i and f ®elds are nearly perpendicular and arranged so that their de¯ections on the ions are additive. The equation of motion for a normally incident ion, traveling along the central line of the magnetic analyzer, is: where n is the radial electric ®eld at , is the ion charge, w the ion mass, and f the magnetic ®eld. The internal energy per charge i int is ®xed by the ®rst electrostatic analyzer. At each energy step, the electric ®eld in the magnetic analyzer is programmed in order either to scan 2 or 4 ions (H + , He ++ , He + , O + ) or to step through 64 values in such a way that the same mass always appears at the same step number, independently of the ion energy per charge. At the end of the selection system the ion strikes one of the two channeltrons used to multiply the particle charge. Ampli®ers convert the charge, collected at the outputs of the channeltrons, into a detectable digital pulse. The count rate delivered by one or two channeltrons during each mass step is sent to the telemetry system. The mass spectrometers have been calibrated in the CESR vacuum facilities using a large ion beam to obtain the energy, angle, and mass transmissions at various energies, and to construct the voltage tables necessary for the¯ight software. Figures 2 and 3 illustrate the main characteristics of the spectrometers, i.e., the mass separation ability at the focus and the energy resolution at high energy from which, using Eq. (3), the energy resolution at low energies can be deduced.

Electron analyzers
Two spectrometers, ION-3 and -4, measure electron uxes in the energy range $10±20 000 eV. They consist of two 127°cylindrical electrostatic analyzer heads mounted on the spinning spacecraft body with (1) ®elds of view instantaneously opposed (anti-collinear) to each other, (2) narrow, symmetrical energy bandwidth of Diai 6%, (3) narrow ®eld of view, and (4) two directional geometrical factors. These dierent values result from the use of two dierent channeltrons at the output of each analyzer. Each channeltron has its associated ampli®er and discriminator (Fig. 1).
The outer analyzer plate of the analyzers is grounded, while the inner one sweeps positive potentials, from $1 to $2700 V. The analyzer constant is k iaV 7X5(eV/V). The electron energy range is Fig. 3. Energy resolution of the ION-2 mass spectrometer at high energy (10 keV) scanned in 128 elementary steps. The count rate, for 32 energy steps summed over four consecutive elementary steps, is sent to the telemetry systems. Charge ampli®ers are used to convert the charge, collected at the outputs of each channeltron, into a detectable voltage pulse. The spectrometers have been calibrated to obtain the transmission as a function of energy and angle. Figure 4 gives the main results of the calibration, i.e., the energy and angular resolutions.

Operating modes
The mass spectrometers ION-1 and -2 can be operated in 31 dierent modes: 25 energy modes to measure the energy range from $30 to 14 000 eV. During each energy step 1, 2, or 4 dierent ions are analyzed. The mass analysis voltage can be either a single value or made up of three closely spaced steps. 1 energy mode to step 64 values of the mass range between 1 and 64 amu. 4 modes during which the voltage of the central grid of the RPA varies to allow the detection of cold ions. 2 test modes to allow the veri®cation of the experiment.
Each mode can be spin-synchronized or free running. The count rates delivered by each of the two channeltrons of each spectrometer can be transmitted to the telemetry. It is also possible to transmit the information coming from a single channeltron or the sum of the two channeltrons. Energy spectra have 8, 16, or 32 channels, and 16, 32, or 64 spectra per spin are sampled. Furthermore, during each spin, 16 magnetic ®eld vectors are transmitted to the ION sensor from the onboard magnetometer. Moreover, during direct telemetry sessions, the ION experiment sends 8, 4, or 2 magnetic ®eld vectors per second to the telemetry.
The electron detectors can be operated in three main modes, with 128, 64, or 32 energy sweeps per spin. The energy table can be changed by command to avoid the detection of high¯uxes of photoelectrons. Furthermore at the entrance of the analyzer, a set of grids can be used to repel secondary electrons.

Scienti®c results
As the ®rst scienti®c results, we will present (1) examples of cusp crossings near noon, (2) cleft measurements showing time-dependent magnetosheath plasma injections near 16 hours MLT, and (3) keV quiet-time auroral proton precipitation characterized by time-ofight signatures. Superposed ion dispersion curves in the cleft are shown to be indicative of pulsed magnetic reconnection acting on the side of the magnetosphere (MLT = 19 hours, R = 18 R E ). The ion energy dispersions inside the proton aurora are similarly used to estimate the source distance in the tail which, unexpectedly, is located very far from the earth, at about 80 R E on the dawn¯ank of the tail. These results are discussed in terms of possible entry, acceleration, and precipitation mechanisms.

Dayside polar cusp
A typical polar cusp crossing during a southward IMF episode f z $ À5nT is presented in Fig. 5. Data are taken along orbit 791 on 7 March 1997 from 18:00 to 18:30 UT, when Interball-Aurora¯ew from the dayside auroral regions to the nightside. From the top to the bottom, Fig. 5 presents the energy-time spectrogram of electrons in the energy range from 30 to 20 000 eV, the energy-time spectrogram of hydrogen ions in the energy range from 30 to 14 000 eV, and their pitch angle. Fluxes expressed in particles (cm 2 á s á ster á keV) A1 are color coded. Several characteristic features are apparent: Fig. 4. Energy and angular resolutions of the electron detectors ION-3/ 4. The angle a is measured in a plane perpendicular to the analyzer cylindrical plates; b is measured in a plane perpendicular to this 1. Equatorward of the cusp, from 21:10:00 to 21:26:00 UT, the satellite passed through a proton aurora showing an energy decrease from about 10 000 to 4000 eV, when the latitude increased. This trend is very common in the dayside auroral oval, resulting from the adiabatic heating of the plasma of the inner plasma sheet followed by its strong grad B drift in quasi-dipolar ®eld topology (see Galperin et al., 1976Galperin et al., , 1978Sauvaud et al., 1981;Onsager and Mukai, 1995). 2. At energies lower than about 1 keV, high¯uxes of protons were recorded. Particles at these energies drift eastward under the eect of the electric ®eld (convection and corotation). The origin of these low-energy ions can be either the ionosphere or the solar wind (see Sauvaud et al., 1981). 3. A gap in ion¯uxes which displayed an energy dependence with latitude was detected equatorward of the cusp. No counts were measured, meaning thē uxes were below the measurement threshold, i.e., below % 2 Á 10 5 keV/(cm 2 á s á ster á keV). This kind of gap is commonly observed in the morning and dayside auroral region onboard Interball-Aurora. To our knowledge, this feature has not been previously reported in the literature. A dedicated study is devoted to this structure, which results from an extremely long drift time (>30 h) for the ions which could be able to reach the gap; the ions are lost by wave-particle interactions and/or charge exchange (Kovrazhkin et al., 1998). 4. The electron energy-time spectrogram indicates the transition from auroral to magnetosheath-like electrons at 21:26:00 UT. Intense electron¯uxes were detected after that time while ion¯uxes of magnetosheath origin were detected 1 min later, ®rst at energies of the order of 10 keV. The most intense electron precipitation lasts until about 21:32 UT.
5. At the equatorward edge of the cusp a clear energydependent dispersion of the protons was recorded. The low-energy cut-o can be assigned to the reconnection occurring at low latitude on the magnetopause (e.g., Hill and Rei, 1977;Onsager et al., 1993;Alem, 1997). Coinciding with the energydispersed structure and with intense electron precipitation, up¯owing low-energy protons were detected at 21:28:30 UT. We will discuss this speci®c structure later. 6. Between 21:26 and 21:45 UT the polar cusp hydrogen precipitation showed a classical energy-latitude dependence in the energy range 14 keV to $100 eV. This energy-dispersed structure is already well documented for southward-directed interplanetary magnetic ®elds (e.g. Burch et al., 1982;Onsager et al., 1993). It results from the general northward convection of the plasma and from the large extension of the entry region on the magnetopause. A¯ux modulation with pitch angle was also detected. This modulation can be understood simply; starting from a ®xed position on the magnetopause, only the ions with the same drift length in a direction perpendicular to the ®eld line can reach the satellite position. For lowenergy ions the main drift is the E´B one, and the drift length is proportional to the time of¯ight between the magnetopause and the satellite. For a nearly ®xed satellite position, the ION experiment is thus expected to measure low-energy particles at low pitch angles and higher-energy particles at higher pitch angles coming from the same magnetopause location (see Onsager et al., 1993;Alem, 1997). Note however that in the foregoing simpli®ed description time-dependent eects are not taken into account. Figure 6 presents another cusp crossing on 25 February 1997 (IMF f z $\simÀ2nT), also in the noon sector (12±14.4 hours MLT) from low to high latitudes (73°±79.5°). The main features described for Fig. 5 are easily recognizable. Oxygen ion¯uxes are displayed here in the third panel from the top. Upwelling hydrogen ions were detected starting at 18:08 UT when the satellite encountered magnetosheath electrons and before magnetosheath protons were detected. Upwardowing oxygen ions are detected simultaneously with the ®rst hydrogen out¯ow. Inside the cusp proper, oxygen ion¯uxes up to 2 keV appeared each time the detector was downward-looking, starting at pitch-angles of about 140°. This indicates that the main heating mechanism was acting below the satellite, possibly accelerating the ions perpendicular to the magnetic ®eld. Note that the occurrence of most energetic up¯owing oxygen ions corresponds to the more intense electron precipitation. The detection of O + out¯ows in the cusp is in agreement with the statistical analysis of DE-1 satellite data performed by Kondo et al. (1990). These authors reported an enhancement of ion conics corresponding to the cusp and the cleft regions on the dayside. It must also be stressed that both cusp crossings showed high¯uxes of protons, indicating an intense heating of ionospheric hydrogen. Perpendicular ion energization is caused by the essentially perpendicular components of electric ®elds oscillating within some frequency range (see the review by AndreÂ and Yau, 1997). Several wave modes can cause perpendicular ion energization ranging from static structures and waves much below the gyrofrequencies to waves at and just above the ion gyrofrequencies to waves above the lower hybrid frequency (Chang and Coppi, 1981;Borovski, 1984;Klumpar, 1986;Lysak, 1986;AndreÂ , 1997). A correlative study of ion heating detected by the ION Fig. 8. Detail of Fig. 7. Energy-time spectrogram of protons between 02:45 and 02:54 UT on 24 February 1997. At the bottom, the pitch-angle of the ions. Note the serpentine structure starting around 02:47 UT and the two maxima seen between 02:50:30 and 02:53 UT experiment and low-frequency wave measurements performed onboard Interball-Aurora is underway (see, Perraut et al., 1998). It must be stressed that several mechanisms can concur to accelerate the ions, one of them being the electric ®eld caused by the nonequipotentiality of reconnected ®eld lines (Wing et al., 1997).

Dusk polar cleft
Ion and electron measurements performed on 24 February 1997 between 01:30 and 04:45 UT are given in Fig. 7 (K p = 0+, 2+). The satellite, coming from the dayside, ®rst encountered an auroral proton population, then reached invariant latitudes higher than 74°between approximately 02:25 and 03:45 UT in the dusk polar sector and passed again through the auroral precipitation region in the evening side of the oval. Between 01:00 and 02:30 UT and between 04:10 and 04:40, a very clear latitudinal energy variation for ions was detected, namely the keV ion average energy decreased when the latitude increased. This dispersion, seen at local times of about 11±13 MLT and then around 19 MLT, can be understood in terms of plasma heating inside the plasma sheet (see Galperin et al., 1978;Sauvaud et al., 1981). Electron measurements con®rm that during these intervals the satellite was inside a weak plasma sheet electron population. The average ion energy can be expected to vary as 1aL 2 due to conservation of the second adiabatic invariant: where L is the ®eld line length A comparison of the data with the predicted 1aL dependence on the dayside expected from the T-89 magnetic ®eld model (Tsyganenko, 1989) is indicated by a thick dashed line on the ion energy-time spectrogram in Fig. 7. The agreement is good, the model corresponds to up 2. At 19:00 UT the agreement is much worse (not shown). We tentatively interpret this discrepancy as due to the dynamical behavior of the magnetosphere, since a substorm occurred at 04:00 UT after a very quiet period. Note that the electron average energy also strongly increased after 04:10 UT, although the latitudinal/local time extent of this high¯ux region was more limited than that of the ions. This dierence can be linked to the local time of the satellite, 19 hours MLT, i.e., the local time sector where ions are expected to drift from the midnight sector while high-energy electrons are drifting eastward after substorm injection.
A very dierent region was detected at high latitudes. Here the ion and electron temperatures are much lower, and¯uxes are high, comparable to those of the magnetosheath. We identify this region, encountered from $02:23 to $03:47 UT, as the polar cleft. According to the Interball-Tail satellite located at dusk in the solar wind, the IMF f z component changed ®rst from positive values (5 nT) to values near zero and slightly negative at about 02:17 UT. Then, starting at $02:45 UT the IMF f z became strongly negative ($A5 nT) with a northward episode for 3 min just before 03:00 UT. In agreement with these measurements, a substorm growth phase was detected from provisional AE data starting several minutes after 03:00 UT. When Interball-Aurora entered the cleft, the IMF f z was already close to zero. The IMF f x component was positive ($6 nT) and the f y component negative (A2 to A3 nT).
Note that inside the cleft a succession of bursts of high energy (10 keV) protons is detected (Fig. 7). As an example, Fig. 8 gives a detailed view of the proton energy-time spectrogram taken between 01:45 and 01:53 UT. Energy/pitch angle structures are now clearly apparent. In this time-interval, two main injections of ions were detected, both starting at high energies, Fig. 9. Result of trajectories computed backward in time for the main ion energy/pitch angle structure displayed in Fig. 8. A 3D model of magnetic and electric ®elds is used (see text). At the top, the right black curve is the energy of the maximum ion¯uxes between 02:47 (T = 0) and 02:50:30 UT. Energy variations when backward computations are performed are indicated. Next panel: ion pitchangle and its backward variations for selected energies. The three bottom panels give the variation of the radial distance, magnetic local time, and magnetic latitude of the particles forming the energy/pitch angle structure measured after T = 0. Note the focal point at about T = A130 s around 10 keV, showing a general energy decrease with time. During this 9-min time-interval, the satellite was nearly stationary near the apogee (magnetic local time 16 hours, invariant latitude 76.6°, altitude 2.95 R E ). The ion structures showed a pitch-angle modulation: the energy, while generally decreasing, increased again when the spectrometer was viewing particles coming from the earth along the magnetic ®eld, thus producing a kind of serpentine-shaped structure. Note that the apparent injection duration was about 5 min, while the recurrence injection time was less, varying from 1.5 to 3.5 min. This produces a characteristic double maximum at a given time in the energy-time spectrogram. The dispersion shape of the ion structure starting around 02:47 UT is very clear and reminiscent of a time of¯ight dispersion eect. We ®rst tested this possibility by ®tting a simple dispersion law t t 0 vav through the experimental points disregarding the pitch-angle variations. The ®t is good and indicates the ions were traveling a distance of the order of 25 R E before reaching the satellite. We implicitly make the assumption that the particle pitch angle went close to zero quickly at higher altitudes. This is a reasonable assumption taking into account the variation of the earth's magnetic ®eld (see Fig. 9). We have tested this with preliminary calculations using a full 3D model of particle trajectories in the T-89 (Tsyganenko, 1989) magnetic ®eld model with a convection electric ®eld model of Volland (1978) with a cross polar cap potential drop of 60 keV [for a complete description, see Delcourt et al. (1989)]. The results are displayed in Fig. 9. The trajectories have been computed backward in time for about 250 s. Initial energies and pitch-angles are taken into account. The main result is that all particles started simultaneously at T 0 ± 125 s from a region located 18 R E from the earth, at a magnetic local time of 18.5 hours, i.e., very close to the dusk magnetopause. According to the model, the latitude of the injection is close to 20°. It is tempting to attribute this injection to sporadic and recurrent reconnection controlled by the IMF (e.g., Sauvaud et al., 1980;Newell and Meng, 1991). This result complements the work of Carlson and Torbert (1980), who found a source in the subsolar region from sounding rocket measurements using time-of-¯ight dispersion structures, and of Lundin et al. (1991) who similarly inferred sources in or near the dayside magnetopause boundary layer from VIKING data. Further studies on this subject are underway with Interball data.

Dawn proton auroral precipitation
The Interball-Aurora satellite's high apogee and relatively low inclination (62.8°) allow it to stay at nearly constant auroral latitudes for tens of minutes. We took advantage of this feature to study time-dependent phenomena over polar regions. This study was performed for an auroral pass of the satellite occurring on 27±28 September 1996, during a quiet period (up 2ÀY 2). The satellite¯ew from the early morning side (2.4 hours The bottom curve is the proton pitch angle MLT), near midnight, to the late-morning side of the oval (8.3 hours MLT), reaching the highest latitudes at 06 MLT (72.50°). A general view of the auroral particle data taken along the satellite path is presented in Fig. 10 which displays the electron and proton energy-time spectrograms, and the proton pitch angle variation.
The electron spectrogram shows several general features. (1) Fly between about 23:45 and 01:10 UT, a population of plasma sheet electrons with a temperature of several keV and intense energy¯uxes; (2) during the same time-interval, the appearance of an electron population with energies lower than 1 keV; (3) on both sides of this traversal of the auroral region, equatorward of the main precipitation, the presence of a population of electrons extending to very low energies, 60 eV here, around 23:45 and 01:03 UT, and (4) the continuation toward low latitudes of the plasma sheet population with reduced energy¯uxes (between 23:30 and 23:45 UT, and between 01:10 and 01:30 UT).
The ion spectrogram exhibits several features which can be related to the electron one: (1) in the early morningside at low latitudes corresponding to the electron plasma sheet extension to low latitudes, an 8-keV proton nose as reported by Smith and Homan (1974) from observations onboard the Explorer 45 satellite near the equatorial plane, (2) bi-directional protons with energies lower than 3 keV are simulta-neously detected, (3) a proton aurora with keV particlē uxes after 23:50 UT, ®rst smoothly distributed around 7±10 keV, then starting to show clear energy dispersed structures at $00:22 UT, (4) out¯owing low-energy protons, mainly between 00:22 and 00:42 UT, (5) an intense bi-directional low-energy ion population with a maximum energy reaching several keV; this population was still detected with an energy not exceeding 500 eV up to 01:30 UT, i.e., in the plasma sheet extension at small L values. Starting around 00:58 UT there was a gap in proton¯uxes between 1.5 and 7 keV; this kind of ion feature, very common in the morning auroral region, is linked to the excessive drift time of the particles which should reach the gap on the basis of their drift path geometry (Kovrazhkin et al., 1998).
Detailed electron and proton energy-time spectrograms between 1 and 14 keV, for the 10-min period between 00:34 and 00:44 UT are given in Fig. 11. Overlapping ion energy dispersion curves are clearly visible. They appear in the detector upper energy level about every 1.5 min, at $00:34:33, $00:36:00, $00:37:42, and ®nally $00:39:45 UT. Each structure is detected for about 4 min. Note that the last one stopped when a broader proton population began to be detected at 00:41:30. The distribution function f v k Y v c of the ions in the energy range 30 to 14 000 eV is given in Fig. 12 for the 1-min time-interval starting at 00:38:31 UT. Note that the energy-dispersed structure appears as a bump in the distribution function. Because of the slow satellite rotation, the maxima are slightly shifted from one distribution function cut to another. Note the low-energy ion out¯ow, clearly visible at 160°pitchangles.
As in the case of the cleft energy-dispersed structures, we applied a simple time of¯ight model to the data. The ®t gave good results and surprisingly long travel lengths, ranging from 77 R E at 00:34 UT to 100 R E at 00:37 UT. For the satellite location at 00:35 UT, the T-89 magnetic ®eld model gives a ®eld line length to the equatorial plane not exceeding 20 R E . However, the magnetic ®eld model is an average, and furthermore the satellite is located in a very speci®c region at dawn, where a small change in latitude will cause the satellite to pass from quasi-dipolar ®eld lines to ®eld lines going to the magnetospheric tail through the¯ank. In order to con®rm our preliminary time-of-¯ight calculations (which do not take into account the particle pitch angle), we used the following procedure: (1) Compute the trajectory of particles forming the energy dispersed structures backward in time in a 3D magnetosphere with electric and magnetic ®eld models. (2) Iterate for increasing invariant latitudes searching for a possible initial time when the particles were all grouped in the outer magnetosphere. The ®nal result is displayed in Fig. 13 for the ion dispersion ®rst appearing at 00:37 UT. Here, for an invariant latitude of 75.65°, the ions have a common origin in the tail, at radial distances of the order of 70±80 R E , a magnetic local time of 02 hours and a latitude of about 0°. This estimate is qualitatively well supported by the general aspect of the data. In Fig. 10 it must be noted indeed that energy-dispersed ion structures are ®rst detected when the electron maximum energy reached a minimum, i.e., when the satellite was connected to the most distant plasma sheet it reached on that orbit. Other trials for lower latitudes of the measurement point gave dierent results with nonconverging curves, in disagreement with the time-ofight nature of the observed dispersion. Note that although the initial MLT is only 02 hours, it corresponds to the tail¯ank at 80 R E ( GSM À35 R E ).
The ions are injected, then the precipitation stops and the cycle continues. For distances as large as 80 R E we expect the ion orbits in the tail to be nonadiabatic. To obtain high¯uxes inside the loss cone is dicult; only a very small part of the distribution function is pitch angle diused (Delcourt et al., 1996). Another possibility is that a parallel electric ®eld acts to align the particles. Such ®elds are expected at the interface of the plasma sheet and low-latitude boundary layer where the antisunward and sunward plasma¯ows mix. The¯ows shear and associated divergent electric ®elds can set up parallel electric ®elds and the boundary motion/instability could be linked to AlfveÂ n or kinetic waves acting as a source modulation. More theoretical work is clearly needed fully to understand the observations.

Summary
The ION experiment has been returning auroral ion and electron data since the end of September 1996. The satellite drifts around the pole in about 9 months, so that a full local time coverage has in principle been available since July 1997. However, due to operational constraints, mainly satellite reorientations, another 9month period is needed to accomplish a good coverage. Periods of correlated measurements onboard Interball-Tail and Interball-Aurora both in the nightside or in the cusp/cleft have been identi®ed and are under study. Already new results have been obtained with the ION experiment from measurements made in the altitude range 10 000±20 000 km. One is the detection of an ion gap in the morning and noon sectors due to an excessive drift time of the ions which could reach these regions. Also a clear determination of the dusk¯ank injection location of ions into the cleft has been deduced from data and from simple modeling. Finally, measurements made during quiet times inside a dawn proton aurora reveal that the source of the poleward boundary of the aurora is pulsating and located on the¯ank of the magnetosphere at distances of the order of 80 R E , presumably at the boundary between the antisunwardowing LLBL and the plasma sheet. For the last two examples of results we have taken advantage of the near stationarity of the satellite near apogee. Other results can be found in the companion paper by Dubouloz et al. (1998), particularly concerning the ionospheric counterpart of velocity dispersed ion structures (VDIS) observed at the polar edge of the nightside auroral oval.