High-resolution ion observations made in recent years, by the TIMAS instrument on the Polar satellite and other instruments, reveal a dynamic and finely structured plasma sheet, at least at high latitude. This study invokes multipoint Cluster observations with the CIS CODIF instruments (ion composition and distribution function) to determine whether transverse density gradients can be of the order of keV proton gyroradii scale size, as suggested by the TIMAS observations. It is shown that the plasma sheet is indeed prominently filamentary and that the proton density with 40 eV ≤ E ≤ 40 keV can vary by Δn = 0.4 cm−3 across less than five average proton gyroradii at R ≈ 5 RE (average E ≈ 7.5 keV at the time). This compares favorably with typical 10-km-size (or less) auroral structures when projected earthward.
 Space plasma probes passing from Earth's low-density tail lobes into the plasma sheet, whether by their own orbital motion or by plasma dynamics, usually encounter some fairly sharply bounded population of protons with energies of several keV to tens of keV, the mean energy often decreasing with time over a few minutes to tens of minutes, before the probes enter the plasma sheet proper. In addition to energy dispersion, these protons often show some degree of initial earthward-directed pitch-angle distribution. This particle population, known as the plasma sheet “boundary layer,” or PSBL, has been reported on since the late 1970's [e.g., Parks et al., 1979; Forbes et al., 1981; Spjeldvik and Fritz, 1981; Parks et al., 1984; Eastman et al., 1985; Bosqued et al., 1993]. Conversely, when the probes enter a tail lobe, or polar cap, from the plasma sheet, a similar population is often the last keV-type protons observed, although their energy and pitch-angle dispersions are not necessarily reversed in time [Lennartsson et al., 2001; Sauvaud and Kovrazhkin, 2004; Keiling et al., 2004]. Early estimates of the PSBL thickness at 10 to 20 Earth radii (RE) distance ranged from ∼1 RE [Eastman et al., 1985, and references therein] to less than a 35-keV proton gyroradius [Spjeldvik and Fritz, 1981].
 The notion of an essentially planar layer bounding the plasma sheet on each of the northern and southern sides was called into question by Huang et al. . Using electrostatic analyzers (LEPEDEAs) on the ISEE 1 and IMP 7 and 8 satellites, measuring up to 45 keV/e, they identified several high-latitude proton structures as being approximately aligned in the GSE (geocentric solar ecliptic) XZ plane (north-south) rather than the XY plane (dawn-dusk) and referred to these as “filaments.” They confirmed the dawn-dusk sense of density gradients with proton data from the two lowest energy channels of the ISEE-1 MEPI solid state detector, between 24 and 65 keV, by comparing the fluxes of protons having gyro centers on opposite sides of the ISEE 1 [Williams, 1979], thus yielding further evidence of gyroradii scale size. For the most part (∼80%), the Huang et al. results were inferred from data with ca 8.5-min time resolution, otherwise with 2.1-min resolution.
 The “filamentary” nature of the high-latitude plasma sheet has become increasingly conspicuous with improving time resolution of the particle analyzers, as has its dynamic nature. Lennartsson et al.  and Sauvaud and Kovrazhkin , using instruments on the polar orbiting Polar and Interball-Auroral satellites, respectively, established that the boundary regions are made up of multiple proton bursts, each with its energies dispersed over time rather than latitude, although the highest-latitude structures may sometimes be the most energetic overall. Typical thermal energy dispersion rates of 8 keV → 1 keV in two to ten minutes indicate a source at some 15 to 75 RE from the satellite (along the magnetic field). Burst densities are ∼0.1 to 1 cm−3, and the source appears to be near the equatorial plane [Lennartsson et al., 2001].
 These proton flows have numerous sharp density gradients transverse to the magnetic field. With 12 s averaging (paired 6-s s/c spin cycles), the Polar data of Lennartsson et al.  and Lennartsson , obtained with the TIMAS ion mass spectrometer (toroidal imaging mass-angle spectrograph) at energy-per-charge between 15 eV/e and 33 keV/e and geocentric location between 4 and 7 RE, show many large flux amplitude variations from one 12-s sampling to the next, often by orders of magnitude. Assuming that the plasma convects at no more than a few tens of km s−1 past the satellite, these variations translate to density gradient scale lengths of the order of 10-keV proton gyroradii at these high altitudes.
 The CODIF ion mass spectrometers (composition and distribution function) on the polar orbiting Cluster satellites [Rème et al., 2001], with perigees and apogees at about 4 and 19 RE, observe the same kind of sharp sample-by-sample proton density variations at 8 s time resolution (paired 4-s spin cycles) as the Polar TIMAS instrument, while also making simultaneous multipoint measurements.
 Each of the CODIF instruments is similar to the TIMAS in that it allows for about 360° instantaneous field of view and 4π angular coverage each spin cycle. The energy range is also similar at 15 eV/e to 38 keV/e. Ion flux-versus-energy spectra are typically derived and displayed as 8-s two-spin averages, although the nominal (all-energy-angle) resolution is a single 4-s spin.
 Of the four identical CODIF instruments initially mounted on the Cluster satellites, one per satellite, only three were successfully activated on orbit. These are on SC 1, 3 and 4. When their separations are sufficiently small, the three-point CODIF data confirm that keV protons may have large density gradients across a few gyroradii. This short paper is limited to data taken near the Cluster perigee at R ≈ 4 to 7 RE, allowing for the most direct comparison with the Lennartsson  Polar TIMAS results.
Figure 1 and Table 1 show typical nighttime data from the near-perigee portion of the Cluster orbits. Geomagnetic activity is briefly elevated here, with the three-hour Kp being 4−, 3+ and 3 between 06 and 15 UT, having been about 1 and 2 for the preceding 54 hours. The SC separations are among the largest used to date.
Table 1. Separations in GSE at 1000 UT in Figure 1
1 − 4
1 − 3
4 − 3
 Dense flux in the center of each panel, at L < 6, is due to ions on longitudinal drift orbits (ring current), except for O+ below 1 keV, which includes some noise from penetrating radiation belt MeV electrons. The weaker and much more structured flux at larger latitudes is due to high-energy ion bursts from above, mostly protons, and lower-energy (<1 keV here) up-flowing ions from the ionosphere (bottom energy channel on SC1 and SC4 has some spurious counts).
 Although not obvious on this time scale, the structured high-energy proton flux does contain few-minute duration segments of falling energy dispersions, identical to what is seen in Polar TIMAS data [e.g., Lennartsson et al., 2001]. What is obvious, however, especially on the left (south side), are the many differences between the bottom three panels. At these rather large separations, the three SC encounter entirely different proton bursts, indicating spatial scale sizes that are less than (or much less than) one RE. Large differences are typical of the Cluster CODIF observations, when separations are a thousand km or more.
 The greater latitudinal extent and stronger intensities of bursts observed in the southern hemisphere on this day, especially in the second panel, are probably related to the elevated activity (Kp = 4− and 3+). Such correlations are fairly common, but proton bursts do occur at any level of activity [cf. Lennartsson et al., 2001].
Figure 2 shows a 40-min near-perigee interval with multiple overlapping (mixing) proton bursts with almost uninterrupted energy dispersion (∼10 min each). This is on a rather quiet day, with a current Kp of 2 and preceding Kp of 2+, 2, 1−, and 1−. As the bottom panel shows, there is some initial earthward field-alignment of the flux (180° pitch angle in the southern hemisphere) associated with each burst, but this kind of dispersion is never as pronounced as the energy dispersion this close to Earth, because of the rapidly converging magnetic field lines (rapid mirroring). As in Figure 1, the upward flow of ionospheric-origin protons is energized and intensified in association with the bursts.
Figure 3 and Table 2 show a 10-min close-up of the second burst with data from all three satellites. Table 2 gives one important reason for choosing this particular day, namely the very small SC separations. In fact, out of some 100 nighttime perigee passes examined from 2002 through 2004 (15 from 2002), this one has the smallest separations, while also having simultaneous data of equal time resolution from all three satellites.
Table 2. Separations in GSE at 1250 UT in Figure 3
1 − 4
1 − 3
4 − 3
 At these small separations, the three CODIF instruments do observe what can be reasonably called the “same” proton bursts, although the finer details differ. It should be mentioned that the dark orange to red colors at several keV energy correspond to a sum of hundreds of counts per pixel, so counting statistics do not define those varying details. Red color at energies well below 1 keV means only a few tens of counts per pixel, however.
 One detail of note is the beginning of red color in the main dispersion, at about 124555 UT in the top panel, 124612 UT in the middle, and 124640 UT in the bottom panel, as indicated by arrow heads. The delays of SC4 and SC3 relative to SC1, 17 s and 45 s, respectively, are within less than a factor of two of what could be expected, in a superficial sense, from the satellites' time of flight alone. Specifically, the magnetic field GSE vector (at 1246 UT at SC1) is B ≈ (−410, 80, −150) nT, the satellites' nearly identical velocity vectors are v ≈ (−1.2, 2.1, 3.8) km s−1, and B and v are almost perpendicular at 87°. If measured from a planar surface perpendicular to v that contains SC1, then SC4 lags SC1 by 30 s and SC3 lags by 54 s. While the shorter times inferred from the Figure 3 spectra do not account for either the 8-s color pixel width or the slight difference in s/c spin phase (∼±1 s), they may reflect a non-planar burst edge, as well as plasma E × B drift.
Figure 4 shows two minutes of proton densities and mean thermal energies, derived by taking velocity moments from each of the three sets of data, including energies between 40 eV and 40 keV, around the times that the main burst is first encountered. Of special note is the large difference between the SC1 and SC4 densities at about 1246 UT, when SC1 enters the burst, ahead of SC4. This difference, Δn = 0.4 cm−3, is measured while the two satellites are situated on magnetic field lines that are a mere 135 km apart, that is, ∣Δr × B∣ B−1 ≈ 135 km. These densities are each a positively weighted sum of thousands of counts and have negligible statistical errors (∼1% from Poisson statistics).
 Being that the mean thermal energies are about 5 keV at the time and the pitch-angle distribution roughly isotropic (Figure 2), the average proton has about 7.5 keV energy (three degrees of freedom), and the local average gyroradius at 90° pitch angle is 28 km. Hence, the Δn = 0.4 cm−3 difference occurs across about 4.8 average proton gyroradii, and the gradient could well be steeper yet. Actually, the peak differential number flux is at about 6 keV (at E = kT for Maxwell-Boltzmann distribution) in Figure 3, suggesting a more appropriate mean energy of 9 keV and thus a mere 4.4 average gyroradii gradient width (or less).
 More than a hundred orbits in 2002–04 with nighttime perigee crossings were examined. A total of 87 had extensive and simultaneous data from at least two CODIF instruments. These two- or three-point observations at R ≈ 4 to 7 RE are each very often consistent with multiple narrow spatial proton structures in the indirect sense that proton flux varies strongly from one 8-sec sampling to the next. This is true at both high and low proton energy (Figure 3). These observations likewise confirm the dynamic nature of proton bursts (e.g., Figure 1). Except for early 2002, the satellite separations are generally from ten to more than a hundred times larger than the near-perigee average proton gyroradii, however.
Table 3 compares location and plasma parameters of the Figures 3 and 4 proton burst, in early 2002, with those used with a “generic” kind of burst by Lennartsson . The latter “event” was configured to be representative of multiple proton bursts observed on two consecutive Polar orbits on 23 and 24 October 1997. The second-day Polar observations showed several bursts with peak number flux at energies above 10 keV, but the 8 keV thermal energy was selected for modeling purpose, as being a more normal case. The electron thermal energy of 1 keV was selected somewhat arbitrarily for lack of actual Polar electron measurements at sub-keV energy.
 The most “provocative” aspect of the generic Polar event is perhaps its large local plasma density gradient of 0.2 cm−3 across 250 km, or 4.5 average proton gyroradii. This was not really measured with any of the many proton bursts, given the single-point nature of the observations, but it was judged to be grossly consistent with observed temporal flux variations and reasonable E × B drift speeds. As Table 3 shows, the Cluster event has an even steeper density gradient, one that is actually measured, of 0.4 cm−3 across (at most) about the same number, 4.8 (or 4.4), of average proton gyroradii. To attain a “sufficient” density gradient at the Polar satellite, it was assumed that the proton source in a model equatorial tail (at B = 8 nT) had a gradient of 0.2 cm−3 across about 7000 km, or 3.5 local gyroradii. The Cluster event in Table 3 justifies this approach in a more tangible fashion.
 The Lennartsson  modeling was meant to demonstrate the effects of differential mirroring of protons and electrons along density gradients, due to their very different gyroradii, effects that include the generation of strong transverse electric fields and local demagnetizing of the protons, as well as ambient “cold” ions, like the O+. This, it was argued, could explain the observed very large amplitude (50–100 mV m−1) electric field fluctuations near the O+ gyro frequency at Polar [see Lennartsson, 2003, Figures 3 and 4]. These effects are elaborated on by Lennartsson .
 Data from the electric-field and wave experiments (EFW) on Cluster (T. Johansson and F. Mozer, private communication, 2007) do not show comparably large fluctuations during the two-min Figure 4 time interval, however, although the field (<20 mV m−1) is noisy and significantly different at the three satellites (not shown). Another and very possibly related difference from the Polar events is the apparently ten-fold larger density of ambient cold plasma at this time, of the order 10 cm−3, according to preliminary data on total electron density available on CDAWeb from the Whisper sounders [Décréau et al., 2001]. This plasma may have damped the fluctuations.
 O. W. Lennartsson thanks the Cluster CIS Teams at UC Berkeley and UNH, Durham, for making their data analysis software available, and K. J. Trattner at the Lockheed Martin ATC for many helpful discussions. This work was supported by NASA under grant NAG5-13211.