POLAR spacecraft observations of helium ion angular anisotropy in the Earth's

New observations of energetic helium ion fluxes in the Earth’s radiation belts have been obtained with the CAMMICE/HIT instrument on the ISTP/GGS POLAR spacecraft during the extended geomagnetically low activity period April through October 1996. POLAR executes a high inclination trajectory that crosses over both polar cap regions and passes over the geomagnetic equator in the heart of the radiation belts. The latter attribute makes possible direct observations of nearly the full equatorial helium ion pitch angle distributions in the heart of the Earth’s radiation belt region. Additionally, the spacecraft often re-encounters the same geomagnetic flux tube at a substantially off-equatorial location within a few tens of minutes prior to or after the equatorial crossing. This makes both the equatorial pitch angle distribution and an expanded view of the local off-equatorial pitch angle distribution observable. The orbit of POLAR also permitted observations to be made in conjugate magnetic local time sectors over the course of the same day, and this afforded direct comparison of observations on diametrically opposite locations in the Earth’s radiation belt region at closely spaced times. Results from four helium ion data channels covering ion kinetic energies from 520 to 8200 KeV show that the distributions display trapped particle characteristics with angular flux peaks for equatorially mirroring particles as one might reasonably expect. However, the helium ion pitch angle distributions generally flattened out for equatorial pitch angles below about 45°. Significant and systematic helium ion anisotropy difference at conjugate magnetic local time were also observed, and we report quiet time azimuthal variations of the anisotropy index.

Except for times of sudden impulsive injections or shock-like events, radiation belt energetic ions at MeV kinetic energies convect so slowly across geomagnetic ®eld lines that the long time scale diusive transport dominates dynamic behavior. To evaluate steady state radial and pitch angle distributions of helium ions in the interior of the radiation belts, one must consider a balance between transport and losses (e.g. Nakada and Mead, 1965;Cornwall, 1972). In an early Russian work, Tverskoy (1971) modeled the L-shell location of the equatorially mirroring peak¯uxes to occur where the characteristic time scales of cross-L transport and of local losses coincide. For helium ions at E~1 MeV, the eective local loss time scale is similar to, or smaller than, the cross-L diusive transport time scale at L<3.5 (e.g. Fritz, 1978, 1981) at the geomagnetic equator, and the L-shell range where this is true expands towards higher L-shells with lower equatorial pitch angles. This is so because charged particles with low equatorial pitch angles encounter a denser bounceaveraged exosphere, and thus suer comparatively greater collisional interactions and losses (e.g., quantitative evaluation by Smith and Bewtra, 1978;Smith et al., 1981).
Moreover, ions with dierent pitch angles observed at L = 3 (for example) come from dierent parts of the ion spectrum in the source population at higher L-shells (since the parallel ion kinetic energy adiabatically varies as L A2 while the perpendicular energy varies as L A3 in the dipole B-®eld approximation), and this couples the observable radiation belt ion anisotropy with both the outer zone source spectrum and the energy and pitch angle dependent cross-®eld diusive transport rate. It is therefore to be expected that the ion pitch angle distributions within the radiation belts should be determined by the combination of pitch angle dependent source mechanisms, by pitch angle dependent loss processes, by the cross-L transport characteristics, and to some extent by electric and magnetic L-shell splitting in the interior of the radiation belts.
It is known that plasma waves in the whistler mode help control the structure of the electron radiation belts (i.e., Lyons and Thorne, 1973), and plasma waves in the ion-cyclotron mode are certainly important for protons and other ions in the lower KeV energy range (e.g., Taylor and Lyons, 1976;Joselyn and Lyons, 1976;and others), yet there is little reliable information to unambiguously establish an importance of wave scattering of MeV ions. Part of the reason for this has been the limited availability of detailed pitch angle distributions of dierent ion species in the central parts of the radiation belts where these ions are abundant. The observational ®ndings presented herein may help remedy this empirical dearth.
Except for large injection events and magnetic shock transitions through the magnetosphere, helium ions can diusively populate the radiation belts from an external source region such as the solar wind and solar energetic particles (e.g., Cornwall, 1972;Schulz and Lanzerotti, 1974;Lyons and Evans, 1976;Fritz and Spjeldvik, 1978, 1979Fritz, 1978, 1981;Sheldon and Hamilton, 1993). Additionally, such ions (both 3 He and 4 He) can also be generated in situ in the inner radiation zone by local nuclear interactions (e.g., Chen et al., 1994Chen et al., , 1996aPugacheva et al., 1996;Selesnick and Mewaldt, 1996;Gusev et al., 1996; and references therein). There have been several modeling studies of geomagnetically trapped energetic helium ions (e.g., Tverskoy, 1971;Cornwall, 1972;Krimigis and Van Allen, 1976;Fritz and Spjeldvik, 1978, 1979, 1982Panasyuk and Vlasova, 1981;Spjeldvik and Fritz, 1978, 1983Sheldon and Hamilton, 1993;Chen et al., 1994Chen et al., , 1996aSelesnick and Mewaldt, 1996;Spjeldvik, 1996b;Pugacheva et al., 1996;and others). For radiation belt protons, there exist empirical models compiled by NASA as well as a survey of proton pitch angle anisotropies with local time, energy and L-shell (e.g., Garcia and Spjeldvik, 1985). But for radiation belt ions heavier than hydrogen, detailed data to carry out a local time variation assessment have until now not been available. Using data from the well-instrumented PO-LAR spacecraft in the NASA/ISTP program, we here study details of the MeV helium ion anisotropies and azimuthal asymmetries in the radiation belt region.
In an earlier conference paper we have reported on azimuthal helium ion asymmetries in the radiation belts , showing shifts in the radial location of the helium ion¯ux maxima and observed variations in helium ion¯ux intensities with magnetic local time. In this work we investigate the character of the observed angular anisotropies of hundreds of keV and several MeV helium ions. The CAMMICE/HIT instrument does not signi®cantly distinguish between 3 He and 4 He, so our results pertain to the overall helium ion population.

Helium ion observations
The POLAR spacecraft was launched on February 24, 1997, and it achieved a near polar orbit with an initial inclination~86°to the equator, an initial perigee of~1.8 Earth radii over the Earth's south pole, an initial apogee of~9 Earth radii over the north pole, and a spacecraft spin time scale of six seconds. Among the extensive instrumentation on the POLAR spacecraft is the CAMMICE instrument package which contains the heavy ion telescope (HIT) consisting of a stack of solid state detectors and electronic discriminators. HIT is mounted perpendicular to the spacecraft spin axis where it operates with an angular and temporal resolution of 16 sector samples per spacecraft spin period. The spacecraft spin axis orientation is nominally perpendicular to the orbital plane such that instruments with pointing direction perpendicular to the spin axis sample virtually all directions angular to the geomagnetic ®eld in each spin as POLAR crosses the geomagnetic equatorial plane. We here report observations of geomagnetically trapped helium ions in four energy ranges (data channels HID5±8) from 520 to 8200 keV ion kinetic energy. These are speci®ed in Table 1. The geometric factor of the CAMMICE/HIT instrument is g = 9.13´10 ±3 cm 2 sr, and the HIT aperture opening angle is 8°. Except for the highest energy channel which extends up to 8200 keV, these POLAR helium ion channels are similar in energy coverage and functionality to the helium ion detector channels, Da1, Da2, Da3 and Da4 utilized on the previous Explorer-45 spacecraft in magnetospheric near-equatorial orbit (e.g. Spjeldvik, 1978, 1982). The POLAR spacecraft made particle and ®eld observations during the low solar activity in the spring, summer and autumn of 1996, and POLAR continues to be operational as of this writing. In the interior of the radiation belts, this time interval constitutes a rather long period of geomagnetic quiescence with the D st ring current activity index rarely depressed by more than a few tens of nanoTesla between a minor magnetic`s torm'' on January 13 (with provisional minimum D st = A88 nT) and another``storm'' on October 23, 1996 (with provisional minimum Dst = A110 nT). Thus the period from the launch of POLAR to the October 1996 storm represents an opportunity to study the detailed structure of the relatively undisturbed ion radiation belts during sunspot cycle minimum conditions.

Angular and orbital data sampling
The POLAR CAMMICE/HIT instrument accumulated counts in the solid state detector channels with a resolution of 16 accumulation samples per spacecraft spin. Directionally, this means that each count sample sweeps out 22.5°of the total spin angular sweep, in addition to the instrument aperture angle of about 8°. For near-simultaneous on-and o-equatorial observations on essentially the same geomagnetic¯ux tube, it is possible to gain a complementary view of the local pitch angle distribution that corresponds to an expanded resolution of the distribution at low equatorial pitch angles. The schematic illustration in Fig. 1 shows the parametrized trajectory of the POLAR spacecraft on July 28, 1996 in terms of the ephemeris L-parameter, the geomagnetic latitude (MLAT), and the geomagnetic local time (MLT). It can be seen that on its inbound leg POLAR ®rst crosses the geomagnetic equator (MLAT~0) at about UT = 16:00 h just below L = 4, and since it orbits essentially in the plane of the local geomagnetic ®eld lines, POLAR then exits from the trapping region a few tens of minutes later by traversing virtually the same¯ux tubes at o-equatorial latitudes at about UT = 16:40. The subsequent path of POLAR (during UT = 17:40 to 18:30) cuts through the radiation belts again on the other side of the Earth, and this traversal occurs in reverse order at approximately conjugate geomagnetic local times after POLAR having swung over the south pole of the Earth. There are twò`L -shell minima'' (but not orbital perigees) on either side of the Earth (®rst at MLT~17 h and then at MLT~05 h).
The L-shell parameter used here is the spacecraft ephemeris L-value computed from the IGRF 95 model geomagnetic ®eld. A more physical L-parameter for speci®c epochs, varying geophysical current systems and event situations is being developed elsewhere (Richard Selesnick, personal communication, 1997), and that improved L-parameter may conceivably organize spacecraft data better, particularly at high L-shells in the auroral zone and beyond. However, within the quiet time trapping region there appears to be little dierence between the nominal and the speci®c epoch L-parameters as calculated by R. Selesnick. Consequently, this work uses the nominal L-parameter for data organization deep within the radiation belts.
To establish the helium ion pitch angle distributions, the available magnetic ®eld direction information (magnetic ®eld elevation angle and azimuth angle in despun spacecraft coordinates) provided in the routine spacecraft ephemeris once per spin period was used. For quiet time interior radiation belt observations, at L = 2.5 to 7, this routine six second resolution magnetic ®eld direction information is deemed adequate since temporal variations in the magnetic ®eld on time scales of the spacecraft spin period were either virtually absent or insigni®cant during this geomagnetically very calm period.
A set of examples of the directly observable angular distributions from the July 28, 1998 passage of POLAR through the radiation belt region at UT = 18:35 close to the geomagnetic equator at L~3 is presented in Fig. 2. These distributions are single-spin``snapshots'' of the observed CAMMICE/HIT dual half-spin counts per spin sector in data channels HID5±8. With orbital plane nominally parallel to the plane of ideal dipole Note. Channel HID5 is also marginally sensitive to protons (from dual proton pile-up), and is not to be used to establish helium ion spectral shapes when such pile-up is present Fig. 1. ISTP/GGS POLAR spacecraft ephemeris parameters for the period UT = 12 to 24 h on July 28, 1996. The solid curve depicts the nominal ephemeris L-shell, the dotted curve the geomagnetic latitude, MLAT, and the dashed curve the geomagnetic local time, MLT, of the instantaneous spacecraft position. During this period, the POLAR spacecraft passed through the geomagnetic equatorial plane twice, ®rst just after UT = 16:00 h near L = 3.2, and then again after UT = 18:00 h near L = 2.4. Notice that``minimum in L-shell'' is not the same at orbital perigee geomagnetic ®eld lines and the POLAR spin axis nominally perpendicular to the orbital plane, it is possible to sample ion distribution gradients in L-shell.
Due to the large gyro radii of energetic heavy ions, it is possible that radial gradients present in the spatial ion distributions can give rise to dierent count rates when the instrument detects ions that come from the upgradient direction compared to those that come from the down-gradient direction (inward and outward looking detector aperture), and this eect will be superimposed on the normal statistical¯uctuations in the helium ion count sampling. The dashed pitch angle histograms show observed counts for magnetospheric radial outward pointing of the HIT-instrument and dotted histograms are for magnetospheric radially inward recordings (considering the locations of the ion gyro centers), and the average of these count rates would approximate the helium ion¯ux intensity at the spacecraft location. Also shown in Fig. 2, the corresponding smooth curve ®ts to the data versus local pitch angle, a, represent least mean-square best ®ts sin N (a) distributions for each half-spin, and the smooth solid line ®t is the average of these two ®ts. The empirically determined exponent, N, represents an overall measure of the pitch angle anisotropy, commonly known as the pitch angle anisotropy index, although as we shall see, this single parameter is an incomplete descriptor for the entire equatorial pitch angle distribution. As one would reasonably expect in the heart of the radiation belt con®nement region, these are indeed geomagnetically stably trapped distributions exhibiting angular maxima for locally mirroring ions. Since the POLAR spacecraft crossed the geomagnetic equator near the L-shell of these data samples, these local helium ion pitch angles, a, are also essentially the equatorial pitch angles, a 0 . Thus, the observed peaks in the count rates are not local peaks o the geomagnetic equator, but in fact the actual peaks of the entire equatorial helium ion pitch angle distribution.
The observed anisotropy index N-values that one may extract from the ®ts in this ®gure and in similar plots are generally found to vary with instrumental energy pass band from about N~4 at E = 0.52± 1.15 MeV to N~9 at E = 2.4±8.2 MeV. The exempli-®ed trend of ®nding higher pitch angle anisotropies with higher helium ion energies is qualitatively similar to the observed statistical 1972 helium ion data result from Explorer-45 . But quantitatively, the deduced helium ion N-values in this 1996 data sample are smaller than the corresponding 1972 result from two solar cycles earlier when the empirical relation: Observed ISTP/GGS POLAR observations of the helium ion pitch angle distributions in helium ion data channels HID5±8 obtained at L~3 and at UT = 18:25 h on July 28, 1996 for B/B 0 1.1. The dashed line histograms are for the magnetospheric radially outward looking half-spin and the dotted histograms are for the inward looking half-spin. The corresponding smooth curves are local sin N (a) ®ts to these angular ion distributions, and the solid curve is the mean of the two b N = 7 + 9.1 ln(E/E 0 ) with E 0 = 1 MeV was found (i.e., Fritz and Spjeldvik, 1981). The reason for the dierence in these results may be related to the dierent observational epoch and possibly also to the dierent magnetic local times of the sampling (the Explorer-45 results were predominantly from the noon-to-early-dusk MLT sector, while the POLAR results on July 28, 1996 are from the dawn-dusk plane).

Mapping to the geomagnetic equator
From a spacecraft orbit point of view, near-equatorial observations are only possible over a limited L-shell range around the POLAR crossing of the geomagnetic equator. Yet it is desirable to obtain reasonably reliable estimates of the equatorial distributions and¯ux magnitudes over all L-shells for which sucient helium ion count statistics are available to do so. Any energetic particle distribution function, f, can be expanded in a series of base functions. From some theoretical considerations, power series, Bessel functions, trigonometric functions, Gaussian distributions, or other special functions may often seem desirable to describe the angular dependence. It is also possible to expand the pitch angle distribution in a power series in sine and cosine functions of the local particle pitch angle, a: f a R I i0 g i sin i a h i cos i a where g i and h i are constants to be empirically determined.
In a trapped energetic ion distribution exhibiting symmetry around a dominant peak at a = 90°, one might expect the terms proportional to cos i (a) to be relatively much smaller than the sinusoidal terms. For mathematical simplicity, one might set h i = 0 for all iindex values, and the remainder of the sum is most often approximated by a single dominant term with i = N so that one assumes jg i j ( jg x j for i-index values other than i = N. In earlier literature, this use of a single term had either not been reasoned at all or was sometimes posteriorily``justi®ed'' by goodness of the empirical ®ts, especially when detailed information to conclude otherwise was not available. From Liouville's theorem it follows that if the equatorial pitch angle distribution truly follows a sin N (a 0 ) form, then the local pitch angle distribution likewise follows a sin N (a) form. Of course, there is no compelling reason why a single term sinusoidal power relation should be a complete anisotropy descriptor, and the one-term trigonometric representation is a practical idealization that is at best an approximation suggested empirically by the data. Nevertheless, it is one representation that is most frequently used to describe geomagnetically trapped energetic ion distributions in terms of a single anisotropy index.
When this functional form is assumed a priori, then the relation between local and equatorial quantities appears particularly simple: f 0 (a 0 ) = f 0 (a 0 = p/2) sin N (a 0 ) and f (a) = f (a = p/2) sin N (a) (e.g., Roederer, 1970). Here f 0 (a 0 ) and f (a) are the equatorial and the local pitch angle distribution functions on the same geomagnetic¯ux tube. By Liouville's theorem, f 0 (a 0 ) = f(a), and the local and equatorial pitch angles are connected by the geomagnetic mirror equation. Based on this, the mapping from locally observed distributions to equatorial distributions in a dipolar magnetic ®eld becomes a simple approximation for the distribution function: f 0 (a 0 = p/2) = f(a = p/2) [B/B 0 ] N/2 , where B and B 0 are the local and equatorial magnetic ®eld induction values on the same geomagnetic¯ux tube. The mapping is similar for particle count accumulations, count rates (CPS) and particle¯uxes (J).
Since we cannot, with a single spacecraft, simultaneously measure both B and B 0 , it is necessary to take the B/B 0 -ratio from a magnetic ®eld model. Well within the interior of the Earth's radiation belts one may use the IGRF 95 model, although not too much inaccuracy in the ratio-values is suered by using the simple dipole ratio: B/B 0 = (4±3 cos 2 k) 1/2 /cos 6 k where k is the instantaneous geomagnetic latitude of the spacecraft. But the simplicity conveyed here is quite alluring since it assumes and depends on the speci®c functional form of the pitch angle distribution. In fact, mapping to the geomagnetic equator from high B/B 0 -values can go astray if the actual angular distributions deviate signi-®cantly from the simple one-term trigonometric representation form. In the following exposition we explore these aspects from an empirical point of view.
As noted, the characteristics of the POLAR orbital trajectory can be used to examine the helium ion pitch angle distribution from two dierent observational points of view. The top panel of Fig. 3 exempli®es the pitch angle distribution results in one of the CAM-MICE/HIT channels, helium channel HID5 (E = 0.52± 1.15 MeV) at L = 4±4.25 for the near-equatorial parts of the spacecraft trajectory at UT = 15:50±15:55 on July 28, 1996 where B/B 0~1 .06. Fitting a sinusoidalpower function to the distribution yields N~3.8, and at ®rst glance the functional ®t based on least mean squares deviation from the data appears reasonably good (solid curve is an unweighted ®t and the dotted curve is a ®t weighted by a factor proportional to the data values themselves). But the apparent good ®t can be deceptive, however, for it is in both ®tting procedures dominated by the near-equatorial count accumulations. Indeed, if one looks more closely, it becomes apparent that the ®ts are not quite as good for low equatorial pitch angles where essentially all count accumulations within the indicated pitch angle selection bins systematically fall above the sinusoidal ®t-curve. This is a clear indication that the simple one-term functional form sinusoidal distribution does not fully describe the helium ion pitch angle distribution down to low equatorial pitch angles.
As the POLAR spacecraft proceeded to travel on a more o-equatorial trajectory (at UT = 16:30 to 18:00), it encountered virtually the same geomagnetic¯ux tubes and was aorded o-equatorial views of the same¯ux tube helium ion distributions at L = 4±4.25 with a much higher B/B 0 -value of~8. The corresponding local helium ion angular distribution is depicted in the lower panel of Fig. 3, and it shows a much lower eective angular anisotropy where the local single-term sinusoidal functional ®t to the observed pitch angle distribution yields N = 1.9, about half of the anisotropy index-value exhibited by that¯ux tube at the geomagnetic equator.
From the speci®c quantitative results exempli®ed already, the entire equatorial helium ion pitch angle distribution in the interior of the radiation belts clearly cannot be described by a single anisotropy index, N. This situation is repeatedly encountered in the four energy channels our data set, and it brings about a sobering cautionary view regarding the use on any singlè`a nisotropy index'' to fully describe radiation belt ion distributions. The distributions are more complicated than the simplest one-term``theoretical'' description would suggest. Indeed, we ®nd this result in all of the POLAR CAMMICE/HIT helium ion data channels and over a wide range of L-shells. Consequently, this observed¯attening of the angular distribution below an equatorial pitch angle of about a 0~4 5°must be a characteristic feature of the helium ion pitch angle anisotropy at the lower equatorial pitch angles for radiation belt helium ions, at least within the upper keV and lower MeV energy range.
Another way to extract ion pitch angle distribution information is to convert local pitch angles, a, to equatorial pitch angles, a 0 = arc sin [(B/B 0 ) 1/2 sin a] using model B/B 0 -values, and to plot f 0 (a 0 ) = f(a) from this. The draw-back of this technique is that the full equatorial pitch angle distribution is not observable from o-equatorial locations, and one is often left with an incomplete distribution exhibiting a gap around a 0 = 90°. Figure 4 shows an example of the results from this technique. The upper panel in this ®gure depicts the angular distribution in the HID5 channel (E = 0.52± 1.15 MeV) observed at L = 3.81 on the same July 28, 1996 POLAR pass with B/B 0~1 .004, the middle panel shows the corresponding result at L = 3.34 with B/B 0 = 1.14, and the lower panel illustrates the situation at L = 3.27 with B/B 0 = 1.76. Inward looking (reconed by gyro center location) half-spins correspond to dashed curves and data values with up-ticks, and outward looking half-spins are shown as dotted curves and data values with down-ticks. The solid smooth curve is the mean of the two half-spin ®ts to the distributions, and would be representative of the distribution at the spacecraft location.
As one can glean from Fig. 4, the range of unobservable equatorial pitch angles, a 0 , widens considerably with increasing B/B 0 -values so that a``®ll-in-®t'' of a functional form sin N (a 0 ) becomes impractical and quite unreliable when B/B 0 > 2. On the other hand, B-®eld model-based conversion to equatorial pitch angles for the use of equatorial pitch angles for binning pitch angle distribution data in energy, L-shell and magnetic local time is, of course, valuable in an eort to produce statistical average distributions based on long-term sampling of steady state distributions.

Observations of anisotropy variations with location
Even though the``anisotropy N-index'' has been found to be at best an incomplete measure for the entire helium ion pitch angle distribution, it may still be utilized as a coarse overall descriptor of approximate validity for high equatorial pitch angles, at roughly a 0~9 0°45°, and to approximately map the near-equatorial¯uxes to the geomagnetic equator when B/B 0 <2. To obtain an overview of the variation of the helium ion pitch angle distribution with L-shell and to explore any dawn-dusk dierences, we studied the pitch angle anisotropy variation over the central section of the radiation belts where energetic helium ions are suciently plentiful to allow meaningful angular sampling and for those orbital segments of the POLAR spacecraft where B/B 0 <2.
A juxtaposition of pitch angle anisotropy variation with the L-parameter is presented in Fig. 5. This ®gure depicts the dusk (MLT~17 h; UT = 14:45±16:30) and dawn (MLT~05 h; UT = 18:00 to 19:45) sector radial pro®les of the N-index deduced (with the restrictions noted above) for the four energy pass bands of the In the dusk sector and at the lowest energies we calculate pitch angle anisotropy N-values (by least mean squares deviation ®ts) between 3 and 6 when the spacecraft is fairly near the equator. There is a slightly higher pitch angle anisotropy around L = 5. A similar situation is also seen in panels 2±4 for data channels HID6±8 (with energies in the range 1.15±8.2 MeV) where a small local anisotropy minimum is seen to occur just beyond L = 4. However, the radial pro®les of Nvalues are dierent in the dawn sector. Here the orbit of the POLAR spacecraft allows sampling down to about L = 2.5, and these panels show that the helium ion pitch angle anisotropy increases fairly systematically with lower L-shells in all of our four data channels. The HID5±8 channels show lower pitch angle anisotropy (N~1 to 2 at L~5) at dawn than at dusk (where N~4 to 5 at L~5), and all the helium ion channels suggest a brief``plateau'' around L~3.5 in the otherwise negative N-versus-L gradient.
In the earlier data report , we computed equatorially mirroring¯uxes from the locally mirroring¯uxes, and we found that there are systematic geomagnetic local time dierences in the¯ux intensities. Speci®cally, we tested whether or not there might be any temporal evolutionary trends in the helium ion¯ux data, in overall¯uxes or in the shape of the angular distributions. If present, such variability might tend to obscure a study of anisotropy variations with geomagnetic local time based on data from the slow precession of the POLAR spacecraft over several months. We found essentially no systematic temporal variation of the helium ion¯uxes over the study interval.
The dawn-dusk dierences found in the POLAR energetic helium ion data, together with the lack of a discernible temporal evolutionary trend in the data set, point to the desirability of carrying out an azimuthal survey covering the anisotropy at all magnetic local times. Since the orbital plane of the POLAR spacecraft precessed in magnetospheric coordinates, the spacecraft progressively sampled all local times over the geomagnetically quiet data period April through October 1996, making a complete survey around the Earth feasible. The azimuthal asymmetry analysis was carried out with 40 spacecraft orbits distributed over this time interval. We determined the helium ion anisotropy N-values from least-mean-square ®ts to the near-equatorial local pitch angle distributions, and for control we also carried out ®ts to the distributions in local-to-equatorial mapped pitch angle whenever feasible.
The result of carrying out the anisotropy analysis for these POLAR spacecraft orbits in dierent orbital planes as the spacecraft precessed in magnetic local time sectors are shown in Fig. 6. Here the MLT distribution of the near-equatorial N-values obtained at L = 4±4.25 during April through October 1996 is depicted with long vertical error bars indicating less than perfect sinusoindal-power ®ts to the local distributions. Count rate limitations also account for a larger spread in the scatter plots of deduced N-values for the higher helium ion energies.
Generally we ®nd that rather small N-values (N~2) are systematically observed in the early dawn sector and in all the four helium ion data channels (HID5±8). In contrast, signi®cantly higher anisotropy N-index values Fig. 4. POLAR observations of helium ion¯uxes in the HID5 channel plotted versus equatorial pitch angle, a 0 . The three panels depict distributions over a complete spacecraft spin at L = 3.81, 3.34 and 3.27 with B/B 0 = 1.004, 1.14 and 1.76 respectively. There is an increasing range of unobservable a 0 -values with increasing B/B 0 . The ®tted curves assume a sin N a 0 dependence in each spacecraft half-spin, and the solid smooth curve is the mean ®t of these two ®ts are observed in the late afternoon sector. These are found to vary signi®cantly with helium ion energy from N~3.5 (in channel HID5 at 0.52±1.15 MeV), N~4.25 (in channel HID5 at 1.15±1.8 MeV), N~5.5 (in channel HID7 at 1.8±2.4 MeV), to N~6 (in channel HID8 at 2.4±8.2 MeV). Although there is a signi®cant spread in this scatter-plot distribution of helium ion anisotropy Nindex values, there is a clear systematic variation with magnetic local time that appears unrelated to any temporal evolution of the trapped distribution. Table 2 gives a more extensive summary of the deduced N-values for dierent energies, L-shells and magnetic local times. Beyond the apparent MLT-trends in the data, limitations on count statistics restricts the accuracy of these results when¯uxes are low, and this is particularly true in the highest energy data channel which has the lowest count statistics.

Summary and discussion
Results from the CAMMICE/HIT-detector on the POLAR spacecraft have revealed new detailed information on the angular distributions and magnetic local time structure of radiation belt helium ion anisotropies at energies in the range from 0.52 to 8.2 MeV ion kinetic energy. From these observations we ®nd that: 1. Helium ion¯uxes in the upper keV and lower MeV kinetic energy range show classic geomagnetic trapping con®guration with the highest¯ux intensities at the geomagnetic equator; 2. A single pitch angle``anisotropy N-index'' based on an assumed distribution f 0 (a 0 ) = f 0 (a 0 = p/2) sin N (a 0 ) does not describe the entire equatorial pitch angle distribution outside of the atmospheric bounce loss cone, but appears empirically to be at least a fair descriptive measure for equatorial pitch angles, a 0~9 0° 45°. The distributions are found to partially¯atten out beyond this angular range; 3. Within the approximate range of descriptive validity during the April through October 1996 epoch, energetic helium ion¯uxes are generally anisotropic with simple-measure anisotropy N-index values in the range of N = 3 to 9, typically increasing with increasing ion energy in the heart of the radiation belt region. The MeV helium ion anisotropy was systematically found to be higher in the dusk sector than in the dawn sector.
4. Additionally, as earlier reported, although not explicitly exhibited here, equatorial peak helium¯uxes are generally found to be located in the vicinity of L = 3 in this energy range with the higher energies exhibiting radial peaks more Earthward, and the location of the¯ux peak varied from MLT dawn to dusk sector so that the dawn-side peaks were found at lower L-shells than the dusk-side peaks (for details, see ; Both the geomagnetic and the geoelectric ®elds are known to cause L-shell splitting eects (e.g., Roederer, 1970;Roederer and Schulz, 1971;Stern, 1971;Schulz, 1972;Kivelson and Southwood, 1975a,b), and there will be some eects of these drift shell splitting processes both in the interior of the trapping region and in the outer parts that encounter its boundaries. It is possible that the observed azimuthal asymmetry of MeV helium ions may be associated with asymmetries in the magnetic ®eld itself and/or with the geoelectric ®eld dawn-dusk structure. Additionally, the Earth has internal magnetic asymmetries that are manifested as higher magnetic multipole contributions with increasing contributions closer to the Earth. But eects of the latter nature It is, in principle, also possible that systematic errors in the magnetic ®eld models (compared with the real ®eld experienced by magnetospheric particles), even during geomagnetically quiet times, could give rise to apparent azimuthal asymmetries in the observed trapped particle distributions. There is at this time no easy way to quantify such a possible error source.
Another possibility is that the observed helium ion azimuthal MLT asymmetry may, in part, be the result of the large spectral and spatial gradients found in the trapped ion¯uxes themselves, combined with the rather modest (at best, tens of kV) dawn-dusk geoelectric potential ®eld. And if so, then it is even possible that a small shift in the energy spectrum due to this apparently modest eect might appear as a signi®cant¯ux variation seen with instruments of ®xed energy thresholds for ion detection, and so masquerade as anisotropy and peak ux variations. Such an interpretation is mildly supported by the earlier ®nding that the radial peak¯uxes show an azimuthal maximum in the dusk sector for 0.52±1.8 MeV helium, no dusk extremum for 1.8± 2.4 MeV helium, but an azimuthal dusk minimum for 2.4±8.2 MeV helium ions (e.g., Fig. 8 in . A more remote possibility for the interior of the radiation belts is that the observed¯attening of the energetic helium ion pitch angle distributions for low equatorial pitch angles might possibly be related to the structure of the radially driven source mechanism for these ions. Anderson et al. (1997) have studied nonadiabatic behavior of protons and other ion species in the near-Earth magnetic tail region. It is suggested from this work that the eective source function for protons as well as for helium and other heavy ions would be related to chaotic pitch angle scattering in the neutral sheet as these ions are injected onto geomagnetically stably con®ned trajectories.
Detailed helium ion combined diusion-and-drift modeling of these aspects appears desirable to see if a geoelectric ®eld asymmetry eect interpretation for energetic helium ions has any merit in the heart of the Earth's radiation belt region, or if a chaotically driven source process has any signi®cant eect. Conversely, if the observed azimuthal anisotropy variation features should be predominantly the result of errors in the magnetic ®eld model, then the ion observations themselves may be used to probe the magnetic ®eld asymmetry properties in the inner magnetosphere.
Acknowledgements. This work was supported in part by a NASA grant NAG5-2578 to Boston University. The authors are grateful to the entire ISTP/GGS POLAR spacecraft technical management team for data acquisition and routine processing of the telemetry data stream. One of us (WNS) enjoyed the Visiting Scholar guest privileges and hospitality at Boston University during his sabbatical leave from Weber State University. We also want to acknowledge the signi®cant contributions of Bryan Laubscher, Robert Hedges, Rose Vigil, and Gina Lujan on the CAMMICE/HIT sensor system at the Los Alamos National Laboratory; of Rocky Koga, Patricia Lew, Norman Katz, and Bill Crain on the CAMMICE/HIT data processing unit at the Aerospace Corporation; and of the administrative support and interest provided by Donald D. Cobb at the Los Alamos National Laboratory. We also acknowledge valuable Email conversations with Dr. Richard Selesnick at the Aerospace Corporation regarding validity and correction procedures for the L-shell coordinate.
Topical Editor K.-H. Glassmeier thanks a referee for his help in evaluating this paper.