Properties and origin of energetic particles at the duskside of the Earth’s magnetosheath throughout a great storm

. We study an interval of 56 h on January 16 to 18, 1995, during which the GEOTAIL spacecraft traversed the duskside magnetosheath from X @ ) 15 to ) 40 R E and the EPIC/ICS and EPIC/STICS sensors sporadically detected tens of energetic particle bursts. This interval coincides with the expansion and growth of a great geomagnetic storm. The ﬂux bursts are strongly dependent on the magnetic ﬁeld orientation. They switch on whenever the B z component approaches zero ( B z @ 0 nT). We strongly suggest a magnetospheric origin for the energetic ions and electrons streaming along these ‘‘exodus channels’’. The time proﬁles for energetic protons and ‘‘tracer’’ O + ions are nearly identical, which suggests a common source. We suggest that the particles leak out of the magnetosphere all the time and that when the magnetosheath magnetic ﬁeld connects the spacecraft to the magnetotail, they stream away to be observed by the GEOTAIL sensors. The energetic electron ﬂuxes are not observed as commonly as the ions, indicating that their source is more limited in extent. In one case study the magnetosheath magnetic ﬁeld lines are draped around the magnetopause within the YZ plane and a dispersed structure for peak ﬂuxes of di(cid:128)erent species is detected and interpreted as evidence for energetic electrons leaking out from the dawn LLBL and then being channelled along the draped magnetic ﬁeld lines over the magnetopause. Protons leak from the equatorial dusk LLBL and this spatial di(cid:128)erentiation between electron and proton sources results in the observed dispersion. A gradient of energetic proton intensities toward the Z GSM = 0 plane is inferred. There is a permanent layer of energetic particles adjacent to the magnetosheath during this interval in which the dominant component of the magnetic ﬁeld was B z .


Introduction
This work concerns the key-role of the B z component of the magnetic ®eld in the detection of energetic burst-like ¯uxes of magnetospheric origin within the duskside magnetosheath.Exodus of energetic particles out of the magnetospheric cavity takes place only via distinct channels along the magnetic ®eld lines, particularly from 15 to 40 R E downstream.The magnetosheath region is divided into a permanent layer of particles adjacent to the magnetopause and ephemeral channels of particles streaming away from the magnetopause.We identify features favouring leakage through the magnetopause and show a very good association between the energetic O + and H + time pro®les for the ®rst time in the nightside magnetosheath.In general, a deeper understanding of the energetic particles properties within the magnetosheath is gained.
The magnetosheath as an ``autonomous region'' has received less attention than the other parts of geospace and this is especially true for the nightside magnetosheath.Hones et al. (1972) reported the existence of 10 to 30 keV protons in the magnetosheath plasma and suggested the magnetosphere as the source of this population.The ®rst comprehensive survey of ³100 keV ions in the magnetosheath is that reported by West and Buck (1976).They found that the energetic ion ¯uxes correlated with both geomagnetic activity and enhanced turbulence in the magnetosheath magnetic ®eld.Detailed correlations of proton ¯uxes with particular substorm phases were unsuccessful.They observed power law ion spectra in the magnetosheath near the magnetopause quite similar to those in the nearby magnetosphere.The source of ³100 keV magnetosheath ions could be the magnetosphere and/or the energization of low-energy protons in the magnetosheath, the bow shock, or the upstream wave region.Sarris et al. (1976) found that energetic (E p ³ 290 keV; E e ³ 220 keV) proton and electron bursts are a semi-permanent feature of the near-Earth environment both within and outside the magnetotail and concluded that the energetic magnetosheath ions are of a nonthermal origin.Sarris et al. (1978) presented simultaneous observations from three spacecraft and argued that proton bursts occur nearly simultaneously inside the magnetosphere, in the magnetosheath, and upstream from the bow shock.They suggested that energetic protons and electrons most likely are accelerated inside the plasma sheet and then propagate to various regions.Asbridge et al. (1978) reported that 3 to 40 keV ions are present sporadically in the magnetosheath and proposed they are produced in the vicinity of the bow shock.Williams (1979) presented strong evidence for signi®cant but sporadic magnetosheath energy ¯ow in the antisolar direction using 50± 220 keV ions.Williams et al. (1988) studied two magnetosheath traversals, one on each of the dawn and dusk ¯anks of the magnetosphere.They concluded that the magnetosheath plasma often has at least two major components, the shocked solar wind seen at energies £5 keV (which behaves independently of the magnetic ®eld direction) and the magnetospheric (plasma sheet) particles seen at energies ³5 keV (which are highly modulated by the magnetic ®eld).Also they found that, at times, the magnetosheath plasma possibly consisted only of the shocked solar wind, with the observed highenergy tail perhaps due to ambient energetic solar ions or acceleration processes at the bow shock.In one case study, at the dawn magnetosheath $50 R E downstream, Scholer et al. (1984) observed, with the ISEE-3 spacecraft, electron and proton bursts occurring almost simultaneously.In general, it is worth noticing that although streaming ¯uxes of energetic particles are extensively studied in the dayside magnetosheath (see Table 1 in Sibeck et al., 1987a), corresponding work for the nightside magnetosheath is missing.
In this article with ®ne time-resolution data-sets, we establish the phenomenon of the exodus of magnetospheric energetic population via distinct channels.We determine their morphology and properties and proceed further in order to monitor processes over the magnetopause using their microstructure.

Spacecraft locations and instrumentation
The period to be examined closely is the interval from the beginning of day 16, 1995, to 08:00 UT on 18 January, 1995, (i.e. 56 h in all).During this interval the GEOTAIL spacecraft traversed the dusk magnetosheath from X = A14.5 to A39.9 R E and terminated its trajectory by repeatedly crossing the magnetopause boundary.Projections of the spacecraft positions on the YZ GSE and XY GSE planes are given in Fig. 1a, b.Also shown for reference are the nominal magnetopause and bow shock positions for average solar wind conditions.It is seen on the YZ GSE plane that the spacecraft gradually shifted from Z GSE = A3.4 to A6.4 R E during the interval.The latter leads to the inference that the GEOTAIL spacecraft essentially remained close to the ecliptic plane.
The identi®cation of dierent plasma regimes was made with plasma parameters obtained via the Solar Wind and Hot Plasma Analyzers (SWA and HPA respectively) of the Comprehensive Plasma Instrument (CPI, see Frank et al., 1994).These data are shown in the GSE system of coordinates with 48 s resolution.The CPI consists of three plasma analyzers: (1) an electrostatic analyzer for hot electrons and ions (HPA), (2) a complementary electrostatic analyzer for cool plasmas (SWA) with high bulk speeds such as those found within the solar wind and magnetosheath, and (3) an ion composition analyzer for identi®cation of ion species such as H + , He + , He ++ , and O + .The energy-per-unit charge ranges for these analyzers are 1.3 V to 48.2 kV, 145 V to 6830 V, and 1.3 V to 48.2 kV, respectively.For the present analysis we utilize magnetic ®eld data from the MGF experiment (see Kokubum et al., 1994) with 3 s resolution in the GSM system.The B z component of the IMF is given via the MFI experiment on board the WIND spacecraft (see Fig. 1a, b.GEOTAIL trajectories projected on the YZ GSE and XY GSE planes and corresponding to the days 16 (solid circles), 17 (open circles) and 18 (until 08:00 UT, solid triangles), 1995.Average bow shock and magnetopause positions are also shown Lepping et al., 1995).The energetic particle dierential ¯uxes are provided via the Energetic Particles and Ion Composition (EPIC) instrument (see Williams et al., 1994).EPIC consists of sensors ICS and STICS, while the time resolution is declared in each plot varying from 3 s to 96 s.The STICS sensor provides $ 4p angular coverage, composition and spectral observations, with charge state determination for all ions from 30 keV to 230 keV/e, and mass per charge measurements ³7.5 keV/e.The ICS sensor provides ¯ux, composition, spectra, and angular distributions over two polar angles of the elemental species protons through iron from ³50 keV to 3 MeV along with angular distributions in one plane of electron ¯uxes ³38 keV and ³110 keV.The two polar planes correspond to two nearly identical ion heads (north and south) with conical collimators, where the north (south) is centred 23°degrees above (below) the ecliptic plane.The ``head 0'' corresponds to the north collimator.

Region identi®cation and geomagnetic activity
The identi®cation of the dierent plasma regimes traversed by the GEOTAIL spacecraft has been made on the basis of the CPI experiment.In Fig. 2 are displayed from the SWA the density (in cm A3 , top panel), the three components of ion velocity V x , V y and V z (in kms A1 , third, fourth and ®fth panels respectively), as well as the ion average energy (in eV, second panel) from the HPA.The high levels of density and velocity of the tailward ion ¯uxes ensure that the spacecraft remains inside the magnetosheath.In contrast, after about 05 UT of day 18, 1995, at times, the very low ion SWA densities, as well as the dramatically enhanced kinetic temperatures of the HPA, indicate recurrent magnetopause crossings.
The whole time interval we are interested in is associated with the main phase of a magnetic storm.The sixth panel of Fig. 2 displays the D st index with a minimum value of A95 nT giving evidence about a suciently intensi®ed ring current almost reaching the bottom threshold of A100 nT designated for a great storm (Gonzalez et al., 1994).The B z component of the IMF, as recorded on board the WIND spacecraft at (X, Y, Z) GSE @ (150, A75, A10) R E , is shown in the bottom panel of Fig. 2. It is noticed that the AB z component exceeds the threshold of A10 nT for an interval of $9 h, leading us to classify the storm considered as clearly falling into the great category (see Gonzalez and Tsurutani, 1987).

An interval with recurrent ``exodus channels''
There were numerous energetic particle bursts in the magnetosheath throughout the whole interval.We are focusing only on a few of them.Figure 3 is a ®rst example, where the three components, as well as the magnitude of the magnetic ®eld (3 s resolution), are shown along with the dierential ¯uxes of the 58± 77 keV energetic protons (6 s resolution).The ¯uxes are spin ($3 s) averaged.The vertical dashed lines and the marked arrows emphasize the B z and peak ¯ux association.In general, with dominant B z component, the magnetic ®eld structure is devoid of energetic particles, whereas with near zero B z massive exodus of energetic particle population takes place via the opened corridor of the magnetic tube.Whenever the B z approaches zero, it seems that a magnetic connection is established between the spacecraft and the magnetotail.For instance, at $01:32 UT of day 16, 1995, a burst peaked at the time when the B z changes sign.
At this point, before the presentation of additional cases of exodus channels, it is important to look at the burst-like ¯uxes in connection with the CPI data.Throughout the interval studied, the GEOTAIL space- craft is actually moving parallel to the model magnetopause.Thus, it is conceivable to explain the GEOTAIL observations in terms of a series of short-term encounters with a ¯uctuating magnetopause, i.e. the periods with energetic particle bursts and magnetic ®eld discontinuities may correspond to such sporadic encounters.In order to analyze this possibility we examine distinct burst examples with the accompanying high-resolution plasma data from the SWA instrument.One particular case is shown in Fig. 4, where the intense ¯ux of energetic protons is shown in the bottom panel.The magnetic ®eld measurements (upper panels) are shown using the same format as in Fig. 3.The plasma density (®fth panel) and velocity (sixth panel), as well as the ¯ow direction remain essentially undisturbed revealing a core magnetosheath plasma.Phi (/) and theta (5) are the azimuthal and polar angles (in degrees) of the plasma ¯ow.There is no indication in the plasma data for intermittently approaching the magnetopause.Consequently, we conclude that these particle bursts are related only to prominent magnetic ®eld changes and not to magnetopause motion.Fig. 3. Vector magnetic ®eld measurements with 3 s resolution from the MGF experiment in GSM system of coordinates, along with the EPIC/ICS 58±77 keV energetic proton ¯uxes for the period 00:30± 03:00 UT of day 16, 1995.About ten successive bursts of energetic protons occurred whenever B z @ 0 nT.The arrows and dashed lines show the times when the ¯uxes peaked along the B z trace Fig. 4. Vector magnetic ®eld measurements and the highest resolution available data from the CPI/SWA experiment for 16:00±16:45 UT of day 16, 1995.The energetic protons burst is shown in the bottom panel.The plasma density and velocity (that is, the magnitude, azimuth / and latitude 5 angles) do not show any signi®cant burstassociated variations

A distinct burst with velocity dispersion structure
During the interval from 01:25 to 01:40 UT of day 16, 1995, the GEOTAIL spacecraft was located at (X, Y, Z) GSE @ (A15.9, 29.1, A3.6) R E .The top panel of Fig. 5 displays the magnitude of the magnetic ®eld (thick line), as well as the B z component, while the second and third panels show the azimuth u and latitude 5 angles.The direction of zero / and 5 points sunward, while the directions of 90°for / and 5 point duskward and northward, respectively.It is evident that the vector magnetic ®eld scans gradually all the directions from northward to southward.Most importantly, the interval of near zero B z is accompanied with energetic 58±77 keV proton (fourth panel), 187±222 keV CNO ion (®fth panel) and ³38 keV electron (bottom panel) ¯uxes.When the energetic electron ¯uxes are peaked the vector of the magnetic ®eld essentially lies on the YZ plane.The maximum of energetic electrons essentially precedes the peak ¯uxes of protons.The electron ¯uxes reach a peak $150 s prior to the CNO ions.The constructed average angular distributions of the peak ¯uxes for the electrons (³38 keV), protons (58±77 keV) and the CNO group ions (187±222 keV) are shown in Fig. 6a, b and c, respectively.The electron ¯uxes are mainly directed duskward parallel to the average magnetic ®eld direction.The same is valid for the observed proton ¯uxes which show distinct ®eldaligned streaming.The heavy CNO ions show a dominant antisunward convective component in the direction of the plasma ¯ow in the duskward magnetosheath.An interpretation for the observed dispersion is given at the discussion section.

The electron channels are rare compared to those of proton
This case study, shown in Fig. 7 oers an ideal example of recurrent bursts of energetic protons, which at the same time demonstrates that the extent of the electron source may be signi®cantly smaller than that of the proton source.The extent of the source is of critical importance because a spatial restriction may lessen the chances for the electrons to be transported and detected downstream in magnetosheath.GEOTAIL was located at (X, Y, Z) GSE = (A37.5,27.4,A6.2) R E .We note in particular the following: 1.The electrons display two distinct exodus channels of an estimated width of $6 R E (derived by the local plasma velocity and burst duration), whereas the protons display ®ve clearly distinct channels and the two major ¯uxes occurring simultaneously with the electron bursts.The ions of the CNO group essentially do not demonstrate any clear channel structure.2. The ®ve marked bursts of protons occurred with an equal number of excursions along the 5 or AB z traces.The 5 excursions towards zero are not accompanied by similar ones in the / or the magnitude B traces.Proton ¯uxes are observed when 5 is less than 30°a nd peaked with 5 minima.3. The average angular distribution for the energetic electron bursts ``K3'' and ``K4'', on the ecliptic plane, is shown in Fig. 8.The projected vector of the magnetic ®eld on the ecliptic plane is also included.
The almost undisturbed ®eld deviates about A27°f rom the X-axis.The constructed distribution concerns the >110 keV energetic electrons because only this electron channel provides measurements in 16 sectors.This distribution is typical for all the energetic particles, as well.The second angular distribution of Fig. 8 depicts the 7.5±230 keV energetic protons from the EPIC/STICS sensor on the XZ plane.The EPIC/STICS sensor provides measurements in six equal 26.7°polar sectors (from +80°to A80°).The overall inferred result is that the energetic An interpretation of the observed dierentiation between energetic proton and electron ¯uxes may be as follows: The electron source is limited to lower latitudes of the magnetopause, and therefore, it is only when B z = 0 nT that the draping lines over the postnoon magnetopause thread through the GEOTAIL site.An alternative interpretation may be supported assuming a wavy magnetopause surface.That is, the ®ve channels may correspond to equal approaches of the magnetopause to the spacecraft position.The diering gyro-radii of electrons and protons would explain very well why both species are not always observed simultaneously.A drawback in this consideration is that the streaming protons must be observed prior to and after the streaming electrons, which is not supported by the observations.Moreover, the latter interpretation cannot account for the occurrence of variations along the B z or 5 trace in close association with the bursts.

The magnetosheath region adjacent to the magnetopause
At the magnetopause neighbourhood the magnetosheath energetic particles demonstrate a dramatically dierent behaviour compared to that exhibited in the evanescent exodus channels.During the time interval 00±08 UT of day 18, 1995, the spacecraft terminates its duskside magnetosheath traverse and repeatedly crosses the magnetopause.Figure 9 shows the three components of the GEOTAIL magnetic ®eld (top three panels) concurrently with the energetic 58±77 keV protons (fourth panel), 187±222 keV CNO ions (®fth panel) and >38 keV electrons (sixth panel).Before the ®nal entry (at $07:40 UT) a few brief intervals of plasma sheet are encountered and shown by the negative (about A20 nT) excursions of the B x component of the magnetic ®eld.Figure 9 clearly demonstrates that the detection of magnetosheath energetic particles in the neighbourhood of the LLBL is not dependent on the condition B z @ 0 nT.Intense particle ¯uxes on the On the basis of these measurements, it could be inferred that a permanent layer of energetic particles always exists just outside the magnetopause (at least under the present magnetic ®eld geometry).Throughout this layer the heavy ions ¯ux level displays almost a ``plateau''.The angular distributions inside the layer are distinctly dierentiated from the typical ones observed in the exodus channels (e.g.those in Fig. 8).The distributions here display the following features: (a) a profound symmetry of the tailward streaming protons around the X-axis on the ecliptic plane (see Fig. 10a as a typical example); (b) the energetic protons stream tailward almost perpendicular to the dominant B z component of the local magnetic ®eld and (c) the major ¯uxes of the 7.5±230 keV energetic protons (Fig. 10b) are concentrated on the ecliptic plane with a small southward ¯ux along the magnetic ®eld.
It seems that the draped magnetosheath magnetic ®eld lines around the magnetosphere con®ne the escap-ing particles to a narrow layer near the magnetopause.In this context, some energetic particles with a prevailing component of velocity along the magnetic ®eld lines will be quickly lost northward or southward.The rest of particles are almost trapped on ®eld lines convecting downstream.The permanent layer particles may be energized via a merging mode because the local magnetosheath magnetic ®eld is strongly southward.In the past, it has been suggested that a permanent layer of energetic electrons is present just outside the dayside (Meng andAnderson, 1970, 1975) and nightside magnetopause (Baker and Stone, 1978).Meng et al. (1981) found that, at the duskside magnetosheath, the energetic electrons are essentially restricted to a narrow band adjacent to the LLBL region, whereas the energetic protons are very extended around the high latitudes of magnetotail.

Energetic O + and H + ions display similar time pro®les
In the past, in an exceptional case with very active geomagnetic conditions, a brief burst ($20 min) of energetic O + ions was observed streaming predominantly into the Sunward direction upstream of the Earth's bow shock (MoÈ bius et al., 1986;Krimigis et al., 1986).
Also, singly ionized suprathermal (E < 17 keV/q) oxygen is found, at least occasionally, to be present in the sub-solar magnetosheath (Peterson et al., 1982).In this work, intense ¯uxes of energetic O + ions at the nightside magnetosheath are commonly observed.Moreover, a time coincidence between the observed major bursts of O + and H + ions can be clearly seen.Figure 11 shows the 58±77 keV energetic proton ¯uxes in parallel with the 7.5±230 keV singly ionized oxygens for an interval of 14 h of day 17, 1995.Singly ionized oxygen O + ions are of ionospheric origin.The observation of similar time-intensity pro®les is of critical Fig. 9.During this interval from 00 to 08 UT on day 18, 1995, the spacecraft approaches and eventually repeatedly crosses the magnetopause boundary.The magnetotail domain is better diagnosed via the large negative excursions along the B x trace.Three indicative crossings occurred at $05:06, 05:37 and 06:36 UT.Fluxes of energetic particles adjacent to the magnetopause magnetosheath are detected with prevailing negative B z component suggesting the existence of a permanent layer just outside the LLBL region.The bottom panel shows the ratios of O + /H + and He ++ /H + Fig. 10a, b.Two average angular distributions in format similar to that of Fig. 8 demonstrating a dierent behaviour as the magnetopause boundary is approached.On the XY GSE plane the 58±77 keV protons convect tailward and almost perpendicular to the local magnetic ®eld.On the XZ GSE plane the vector magnetic ®eld deviates $16°from the southward direction and although the dominant ¯ux is the tailward one on the ecliptic plane, a minor southward ¯ux along the magnetic ®eld is also observable importance leading us to consider that both O + and H + species have the same origin.
The O + /H + ratio in the bottom panel of Fig. 9 (solid circles) of the 7.5±230 keV tailward streaming protons and oxygen ions shows that, at times, the O + ions are as abundant as the protons at the nightside magnetosheath.The ratio of O + /H + is at least one order of magnitude higher than the ratio of He ++ /H + presented in the same panel (open circles).The He ++ ions are of solar origin.At the plasma sheet the ratio He ++ /H + decreases when the geomagnetic activity level increases, whereas the ratio of O + /H + responds to the opposite direction (Sharp et al., 1982).In conclusion, the similarity of the magnetosheath oxygen and proton time pro®les along with the increased ratio of O + /H + , as it is the case with increased activity inside the magnetosphere, support the magnetospheric origin of both ions in the exodus channels.

Magnetosheath spectra
In the following we present two energy spectra corresponding ®rst to the region adjacent to the magnetopause permanent layer and the other to a wide exodus channel.The particle spectra provide some clear indications for certain aspects of the magnetospheric dynamics and particle properties determining their origin, energization mechanisms, transport paths etc.In Fig. 12a the open and solid circle symbols give the spectra displaying the relative energetic proton abundances both in the plasma sheet and the magnetosheath.The magnetosheath spectrum is the overall response within the intervals 02:00±05:00 UT and 05:45±06:30 UT, while that of the plasma sheet is the average spectrum for 06:50±07:10 UT of the same day 18, 1995.The almost similar spectral shape, as well as the fact that the gradient of energetic particles is directed toward the magnetopause probably are manifestations of their magnetospheric origin.The plasma sheet may be an adequate source of the magnetosheath population.Certainly, between the two spectra there is a time progression.In Fig. 12a the spectra of magnetosheath O + and He ++ ions are added.Additional notable observations are as follows: 1. ``Tracer'' ions of ionospheric-magnetospheric origin are detected inside the magnetosheath with energies as low as $9.5 keV. 2. The ¯uxes of solar origin He ++ ions are much lower than those of O + or H + .At $30 keV the percentage of H + /He ++ is about 1%. 3.At energies ranging from 20 to 60 keV the H + and O + ions have nearly the same abundances.In general, the proton spectrum is softer than the O + one.
The spectra in Fig. 12a correspond to the region adjacent to the magnetopause.A dierent point of view is given via the spectra of Fig. 12b, corresponding to a broad exodus channel from 10:45 to 11:45 UT of day 17, 1995.A prominent inference is that the ionosphere O + ions are channelled (similarly to the protons) away from the magnetopause at all energies above $15 keV.This seems to agree with Paschalidis et al. (1994), who have found signi®cant magnetospheric contribution in the dayside magnetosheath reaching energies as low as $10 keV.At this lower threshold, the ratio of O + /H + is $0.17.It is evident that the high energy oxygen ions have a higher chance to escape outside the magnetosphere and be transported away.

Discussion
We have examined a duskside magnetosheath traversal by the GEOTAIL spacecraft from about X = A15 to A40 R E .We are interested in the issues of morphology and properties of the ``exodus channels''.The origin of the outward channelled magnetosheath energetic particles is an outstanding problem, too.The detection of O + ions in magnetosheath is not something new by itself.Similarly, spectra constructed in the magnetosheath could be found in past works.In this all these are used as tools and serve for a better understanding of the phenomenon of the exodus of magnetospheric particles.Again, we point out that the large majority of magnetosheath works are restricted to the dayside rather than the nightside region.A fundamental feature of the observed particle exodus is that it occurred simultaneously with the evolution of a great storm leading to high abundance of ionospheric ions inside the magnetosphere (see Lennartsson and Sharp, 1982;Daglis, 1997), and the following release of O + ions outward in the magnetosheath.Therefore, these tracer ions provide good evidence for the path through which the dominant ions have been transported to the spacecraft.
Away from the magnetopause, the energetic proton ¯uxes are switched on or o depending essentially upon the condition that the magnetic ®eld remains on or out of the B z @ 0 nT plane.The in situ evidence justifying the key-role of this condition could be as follows: ®rst, the systematic and recurrent detection of ¯uxes each time the B z component of the magnetic ®eld approaches zero (irrespectively of the recorded B x and B y components) as in Fig. 4. Second, the magnitude and azimuthal angle of the magnetic ®eld do not change in many of the observed ¯uxes (see Figs. 4 and 7).And third, the fact that the energetic proton ¯uxes are highly sensitive even to micro-excursions of the B z component (B z or 5 trace).
The role of the B z component in controlling the escaping magnetospheric particles is understood as follows: the magnetic ®eld lines drape the dayside magnetosphere at low latitudes and are loaded with the escaping magnetospheric population.These particles stream along the magnetic ®eld lines and are subsequently measured by the EPIC sensors.Equatorial magnetosheath magnetic ®eld lines typically make their closest approach to the magnetosphere at local times near 1500 LT (see Sibeck et al., 1987a).If the B z component is signi®cant then the streaming particles will be quickly displaced well above or below the ecliptic plane, failing to encounter the GEOTAIL detector almost on the ecliptic plane.Hence, the B z component is the most critical parameter which favours or precludes whether the energetic particles are channelled far away downstream.On this basis it is conceivable that: (a) the IMF largely modulates the space and dictates the suitable condition under which the transport and detection of particles away from the magnetopause is possible; (b) the condition B z @ 0 nT establishes a magnetic connection between the spacecraft and the magnetosphere; and (c) the magnetosphere as the supplying source needs to be ®lled up with energetic particles in order to provide them.
It was reported (West and Buck, 1976) that magnetosheath ¯uxes occurred with depressed and turbulent magnetic ®eld.Instead, in this work intense bursts are observed even with almost unperturbed magnetic ®eld magnitudes (see the bursts of Figs. 4 and 7).The turbulent character may contribute to the particle dispersion and propagation, but the most important factor remains the magnetic ®eld orientation.
An interpretation of the observed dispersion structure in peak ¯uxes of Fig. 5 may be as follows: as the vector magnetic ®eld scans progressively all the directions from northward to southward lying at the same time on the YZ plane, the magnetic ®eld lines may be connected with the sources of the energetic particles.The energetic electrons peak precedes that of the protons and therefore the electrons source must be displaced slightly more northward than the protons one.Most probably, the electrons follow the trajectories along the draping magnetic ®eld lines outside the magnetopause over the YZ cross-sectional plane of the magnetotail.The magnetospheric energetic electrons (E > 25 keV) drift eastward, hit the pre-noon magnetopause and leak out (Anagnostopoulos et al., 1986;Sibeck et al., 1987b;Sibeck and McEntire, 1988).The latter statement is also supported by the fact that the majority of energetic proton (electron) intense ¯ux events are found on the dusk (dawn) magnetotailmagnetosheath (see Meng et al., 1981;Krimigis and Sarris, 1979).The escaping electrons in the dawn magnetosheath via the draped magnetic ®led lines over the magnetopause are transported to the GEOTAIL site.Conversely, the magnetospheric energetic protons are considered to leak out from the lower latitudes of the dusk magnetopause.Therefore the dispersion between protons and electrons is considered to be a spatial eect.The energetic heavy CNO ions and protons have the same source.The best contact with the ions source must be anticipated with zero latitude angle of the magnetic ®eld.The observed delay ($70 s) between the peak ¯uxes of the CNO ions and protons is considered to be a temporal eect corresponding to the dierent travel times.The faster protons arrived earlier than the CNO ions.In addition, the magnetic ®eld direction was changing gradually and this causes a further lengthening of travelling path for the CNO ions (and therefore a further increase of the needed travel time).
This interpretation is further supported by an additional distinctive point: the proton ¯uxes prior to the peak display a Sunward ¯ow character.The angular distribution of Fig. 6d demonstrates such a situation.And this feature may be a genuine result of a BXÑU anisotropy due to a large north-south gradient in the proton intensities (see Sarafopoulos and Sarris, 1991), since the plasma sheet energetic particles are expected to escape outside the magnetopause primarily in the vicinity of the Z GSM = 0 plane and the GEOTAIL is located to the north of that plane, at (X, Y, Z) GSM = (A15.5,28.5, 6.6) R E .This time the magnetic ®eld points eastward.
The claim about the magnetospheric origin of energetic particles in exodus channels could be supported by the following arguments: 1.The detected major magnetosheath ¯uxes of streaming protons along the magnetic ®eld lines display the same time pro®le as that of O + ions.In addition, the plasma sheet population of energetic particles alone suces to warrant the lower observable ¯uxes in magnetosheath.The density gradient of energetic particles is directed toward the magnetopause (Williams et al., 1988;Traver et al., 1991;Paschalidis et al., 1994;Eastman and Christon, 1995).
2. The ``exodus channel'' emerges each time the extrapolated local magnetic ®eld lines are draped around the lower magnetotail latitudes.In one case study, at the dayside magnetosheath, a similar observation is referred to by Sibeck et al. (1987a): the maximum ¯ux was related to the equatorial region of greatest particle escape.
3. It could be argued that the energetic population streaming along an ``exodus channel'' could be energized via the shock drift acceleration (SDA) mechanism.In this case the shock drift accelerated protons at E p ³ 290 keV and then transmitted downstream must present a double-peaked anisotropy nearly perpendicu-lar to the magnetic ®eld (see Anagnostopoulos and Kaliabetsos, 1994).In the present work, we failed to observe such a characteristic signature as the ®ngerprint of the SDA mechanism.For example the average angular distribution for the EPIC/ICS proton channel P6 (i.e.228 £ E p £ 342 keV) corresponding to the two exodus channels ``K3'' and ``K4'' of Fig. 7 is shown in Fig. 8. Instead of a double-peaked angular distribution, a profound ®eld aligned anisotropy is seen.
4. The Fermi acceleration is widely considered as another candidate mechanism working over quasi-parallel bow shocks and producing energetic particles, which in turn, could be swept downstream in the magnetosheath (Scholer et al., 1980;Fuselier, 1994).
Our prominent examples of exodus channels, like those shown in Fig. 7, took place at the dusk-side magnetosheath, although the typical ``Archemedian spiral'' angle of the IMF would favour the Fermi acceleration over the dawnside part of the bow shock.In addition, the Fermi acceleration mechanism is unable to produce energetic electrons, like those observed in exodus channels.
In Fig. 12 it is shown that the magnetosheath and magnetospheric spectra resemble each other greatly, and therefore one must conclude that escaping magnetospheric particles dominate any population of magnetosheath particles energized by merging.But the in situ measurements show something more against the merging process: ®rst, in the presented case study of Fig. 5 the burst occurred with B z @ 0 nT and dominating the B y component.Thus the lines drape around the magnetopause over the YZ plane and our interpretation of the observed dispersion structure for dierent species is based on the notion of leakage.We think that the nonmerging mode model is able to produce such a structure.Second, in most cases (see Figs. 4 and 7), however, the bursts occurred with B z @ 0 and dominating the B x component.In this magnetic ®eld geometry the magnetosheath extrapolated lines are draped at the post-noon magnetopause perpendicularly to the local magnetosphere magnetic ®eld.Therefore, the merging mechanism is also unlikely to work.So a much better explanation is that the particles are leaking out all the time and that when the magnetic ®eld connects the spacecraft to the tail, they stream away and are observed at the spacecraft.Scholer et al. (1984), in one case study at the dawn side of the distant magnetosheath ($50 R E downstream) observed energetic electrons and protons occurring almost simultaneously in bursts of $2 min duration.These particles are considered streaming along Earthrooted magnetic ®eld lines (that is, as remnants of ¯ux transfer events).In this work the magnetosheath particle bursts are scrutinized further.An objection in respect to this interpretation, and in the light of our observations, is that the reconnection process must work continuously over the magnetopause and the resulting streaming particles are seen whenever B z @ 0. Anagnostopoulos et al. (1986) in their Fig. 5 show magnetospheric ions on outer dusk magnetosheath magnetic ®eld lines which do not encounter the magne-topause.This may understood as follows: ®rst, this ®gure is a sketch drawn mainly on the base of measurements of escaping particles upstream of the bow shock.Second, a major implication of the ®gure seems to be a steep dawnward gradient of energetic ions in the dusk magnetosheath, which is largely unproved at present.Third, they have used 5.5 min averages of energetic particles, whereas the duration of our detailed studied channels usually are less than 5 min.Fourth, we have assumed that the typical duskside magnetosheath magnetic ®eld lines drape the post-noon magnetopause.In the present work streaming particles are commonly observed, except from the case of electrons detected when the B y component of the magnetic ®eld was dominant (see Fig. 5).
It is worth noticing that a paradoxical result that might be reached by hypothetical statistical surveys in the magnetosheath.Because the ¯uxes of energetic particles are strongly dependent on the magnetic ®eld direction, it follows that during intervals of great storms when the B z -negative component of the magnetic ®eld clearly dominates, no ¯uxes could be detected wellinside the magnetosheath.Conversely, during long periods of lower geomagnetic activity and suitable geometry of the magnetosheath magnetic ®eld persistent ¯uxes could occur.
Whenever the magnetospheric particles escape, the magnetosheath magnetic ®eld guides them toward the bow shock.Those particles that succeed in traversing the magnetosheath to the bow shock provide a source of particles for upstream events.(Sarris et al., 1976;Scholer et al., 1981;Anagnostopoulos et al., 1986;Sarris et al., 1987;Sibeck et al., 1987b;Baker et al., 1988).

Synopsis
Our observations concerning the origin and properties of energetic particles in the duskside magnetosheath lead us to the following conclusions: (a) the bursty character of the ¯uxes depends mainly from the B z component orientation (which, in turn, is probably dictated by the solar wind magnetic ®eld topology; (b) the magnetospheric origin of energetic particles is established on the basis of spectra and O + ¯uxes; (c) the observation of a dispersion structure occurring under a distinct magnetic ®eld geometry supports the leakage process against the merging one for the origin of energetic particles in the magnetosheath; (d) Sunward magnetosheath detected ¯uxes provide evidence for the presence of a gradient of energetic protons toward the Z GSM =0 plane; and (e) in general, the energetic protons stream along the magnetic ®eld lines, except in the layer adjacent to the magnetopause, where they merely convect perpendicular to the local ®eld.

Fig. 2 .
Fig. 2. Comprehensive Plasma Instrument (CPI) data are used to identify the spacecraft traversed regions.From top to bottom the ion density and average energy density, the three components of plasma velocity, the ground-based Dst index and the IMF B z component (as recorded onboard WIND via the MFI experiment) are shown.The density and velocity have measured by the Solar Wind Analyser, while the average ion energy by the Hot Plasma Analyser

Fig. 5 .
Fig. 5.A distinct burst of energetic particles (bottom three panels) along with magnetic ®eld data (upper three panels) are shown for the short period 01:25±01:40 UT of day 16, 1995.The upper panel shows the magnetic ®eld magnitude as well as the B z component (thinner line).The azimuthal u and latitude h angles are displayed in second and third panels.The three arrows mark the times of peaks for the energetic 58±77 keV protons (fourth panel), 187±222 keV CNO ions (®fth panel) and >38 keV electron ¯uxes (bottom panel) demonstrating a species dependent dispersion structure.Around the peak ¯uxes the B z component gradually changes sign

Fig. 6
Fig. 6. a, b and c average angular distributions over the ecliptic plane XY correspond to the three peak ¯uxes of dierent energetic particle species presented in Fig. 5.The energetic electrons and protons are directed duskward along the local magnetic ®eld direction.The angular distribution d displays an Earthward ¯ux (in magnetosheath) resulting from an intensity gradient of energetic protons toward the equatorial plane

Fig. 7 .
Fig. 7. From top to bottom the energetic >38 keV electrons, 58± 77 keV protons (second panel), 187±222 keV CNO ions (third panel), the magnetic ®eld magnitude and the AB z component (fourth panel), the azimuth (®fth panel) and the latitude (sixth panel) angles of the magnetic ®eld are shown.Five distinct bursts of energetic protons that occurred in association with B z @ 0 nT are marked with the letter K

Fig. 11 .
Fig. 11.Dierential ¯uxes of 52±77 keV energetic protons and 7.5± 230 keV energetic O + ions during an interval of 14 h of day 17, 1995, when the D st index continuously decreases.The magnetospheric origin ions as well as the energetic protons display almost similar time pro®les suggesting their common source