Structure of the near-Earth plasma sheet during tailward flows

A detailed analysis of successive tailward flow bursts in the near-Earth magnetotail ( X∼−19RE) plasma sheet is performed on the basis of in-situ multi-point observations by the Cluster spacecraft on 15 September 2001. The tailward flows were detected during a northward IMF interval, 2.5 h after a substorm expansion. Each flow burst (Vx<300 km/s) was associated with local auroral activation. Enhancements of the parallel and anti-parallel ∼1 keV electron flux were detected during the flows. The spacecraft configuration enables to monitor the neutral sheet ( Bx≈0) and the level ofBx≈10–15 nT simultaneously, giving a possibility to distinguish between closed plasmoid-like structures and open NFTE-like surges. The data analysis shows NFTElike structures and localized current filaments embedded into the tailward plasma flow. 3-D shapes of the structures were reconstructed using the four-point magnetic filed measurements and the particle data.


Introduction
Several to ten minutes-long fast flows in the magnetotail, known as bursty bulk flows (BBFs, Angelopoulos et al., 1992), are observed to be predominantly Earthward in the near-Earth magnetotail (radial downtail distances R<20 R E ) Correspondence to: A. Runov (arunov@igpp.ucla.edu) and mainly tailward beyond this distance (Nagai et al., 1998b;Baumjohann et al., 1999). Recent statistics of BBFs at the Cluster orbit (R∼19 R E ) shows that only 22% of the observed fast flows were tailward (Nakamura et al., 2004). These observations indicate that the fast flows are, most likely, generated by electromagnetic acceleration of plasma in the reconnection region usually located tailward of 20 R E (e.g. Baumjohann et al., 1990;Nagai et al., 1998b). This process is the essential part of the Near-Earth Neutral Line (NENL) model of a substorm (e.g. Baker et al., 1996). In this context, tailward flows observed in the near-Earth magnetotail seem intriguing. Their appearance implies either enhanced magnetic activity (e.g. Miyashita et al., 2005;Nagai et al., 2005) or a mechanism that differs from acceleration at a NENL (e.g. Schödel et al., 2001).
Magnetic reconnection (MR) is, generally, a change of the magnetic field topology with magnetic field X-line formation and plasma acceleration via the magnetic tension force in the vicinity of X-line. The bulk outflows resulting from MR carry the oppositely directed magnetic field loops: The magnetic field is directed northward within the Earthward outflow and southward within the tailward outflow. One of the predictions of the NENL model, is the formation of a magnetic field structure, known as a plasmoid (e.g. Hones, 1979). Although both tailward and Earthward traveling plasmoid-like structure were observed (e.g. Zong et al., 2004), the term "plasmoid" is usually used for closed magnetic structures (magnetic islands), manifested as the northsouth bipolar variation of the magnetic field, embedded into tailward bulk flow (e.g. Mukai et al., 1996;Ieda et al., 1998).
Published by Copernicus Publications on behalf of the European Geosciences Union.
The presence of even a small cross-tail magnetic field component (B y ) leads to an appearance of a more complex 3-D structure, called "flux rope" (e.g. Slavin et al., 2003), instead of a 2-D magnetic loop, with the B y sign correlating with that in the IMF. It was also found in simulations that a strong B y component, or core-field, may be generated in the course of plasmoid evolution (Hesse et al., 1996). Alternatively, a large B y may be generated due to localized reconnection of sheared magnetic fields (Shirataka et al., 2006).
The energetic electrons, generated in the course of MR (e.g. Hoshino et al., 2001), may be used to probe the magnetic field structure. They provide information about changes in the open and closed magnetic field line configuration. Particularly, the observation of the enhanced electron flux in association with the magnetic field bipolar variation was interpreted as the indication of the closed O-type magnetic structure (Zong et al., 2004).
The bipolar variation of the magnetic field may be explained in the frame of so called Nightside Flux Transfer Events (NFTE, Sergeev et al., 1992, 2005) -a magnetic surge in the magnetotail plasma sheet containing accelerated plasma, generated by a reconnection pulse. Contrary to a plasmoid, NFTE implies a locally open magnetic field structure. The bipolar variation in the NFTE case is asymmetric: the flow burst predominantly carries the unipolar magnetic field (southward in the case of a tailward flow). A turn to another polarity, observed above and below the neutral sheet, is due to the magnetic field compression by the flow. A spacecraft crossing the NFTE related magnetic loop near the neutral sheet will detect a unipolar variation. Numerical simulations show that, at early stage, a plasmoid may have a structure, locally open at the tailward side, resembling NFTE (Abe and Hoshino, 2001).
Theoretical models of fast pulsed reconnection, resulting in NFTEs, predict formation of a pair of slow shocks, separating inflow and outflow regions (e.g., Petschek, 1964;Semenov et al., 2004). The slow shocks were observed in the distant tail (X<−100 R E ) on the lobe-plasma sheet boundary (e.g. Saito et al., 1995) and on the plasmoid boundaries at X∼−100 R E (Mukai et al., 1996). Recently, Eriksson et al. (2004) reported observations of reconnection related shocks in the near-Earth tail. The shear stress balance (Walén test, e.g. Khrabrov and Sonnerup, 1998) at fast flows, implying plasma acceleration by magnetic tension and predicting near-Alfvénic velocity of the accelerated plasma in the de Hoffmann-Teller reference frame, was considered as evidence of reconnection on the magnetopause (Paschmann et al., 1979;Sonnerup et al., 1981;Paschmann et al., 1986) and in the magnetotail (Øieroset et al., 2000;Eriksson et al., 2004).
The observations of tailward flows in the near-Earth magnetotail, discussed in literature, attribute the flows to near-Earth reconnection (e.g. Sergeev et al., 1995;Nagai et al., 1998a;Miyashita et al., 2005). Alternative hypotheses of a near-Earth flow onset are based on the current disruption (CD) concept (e.g. Lui, 1996). The CD framework is based on an idea of local decrease of the cross-tail current in the near-Earth magnetotail at −8 to −12 R E due to development of instability after near-Earth current sheet thinning down to a critical thickness. (see Lui, 2004, for a review). The resulting inductive electric field leads to the drift motion of plasma. This model predicts strong "turbulent" fluctuations of the magnetic field (δB∼B), ion energization perpendicular to the magnetic field and field-aligned electron beams (Lui, 1996). Southward magnetic field tailward of CD, produced by local generation of dawnward current, may lead to tailward bulk flow (Lui et al., 2006). Another possible mechanism of the near-Earth CD involves a development of the ballooning instability at the inner edge of the plasma sheet (Roux et al., 1991;Voronkov, 2005;Roux et al., 2006). This model predicts quasi-periodic variations of the magnetic field with the azimuthal wave number k y (k x , k z ). The tailward plasma flow is due to electric drift. The CD region expands tailward at a speed of 200-300 km/s (Ohtani et al., 1992) as a front of the magnetic field dipolarization. The turbulent CD-related tailward flow does not contain a plasmoidlike magnetic structure or any orderly magnetic field pattern (Lui et al., 2006). The j×B-force components are expected to be rather chaotic within a plasma flow due to the turbulent CD. The ballooning-based CD apparently can produce a more regular magnetic structure.
In this paper we discuss in situ Cluster observations of a set of successive tailward flow bursts in the near-Earth magnetotail plasma sheet during 03:30-04:30 UT on 15 September 2001. The measurements at four probes, forming the tetrahedron with the scale <2000 km, give the possibility to identify spatial structures with scales comparable or larger than the probes separation and discriminate between plasmoid-like structures, NFTEs, and waves. Furthermore, four-point magnetic field measurements enable the estimation of the magnetic field gradient and magnetic tension inside the flows, which may help to understand the process of their formation.

Instrumentation
For the analysis of the tailward flows, observed by Cluster at X∼−19 R E , we use magnetic field data from Cluster Flux Gate Magnetometer (FGM, Balogh et al., 2001), ion data from the Cluster Ion Spectrometry (CIS, Rème et al., 2001) experiment and electron data from the Plasma Electron And Current Experiment (PEACE, Johnstone et al., 1997). The ion data are provided by Cluster 1 and 3 (called C1 and C3 further on) Hot Ion Analyzer (HIA), with energy range ∼5 eV/e-32 keV/e, and the Composition Distribution Function (CODIF) instrument with energy range 20-40 keV/e and by the Cluster 4 (C4) CODIF instrument. No ion data are Fig. 1. Solar wind dynamic pressure, IMF B y and B z , AL-index, magnetic field at geostationary orbit, magnetic field (B x and B z ) and ion bulk velocity (V x ) observed by C4 (blue) and C3 (red) in the magnetotail during 00 -06 UT on September 15, 2001. GSM coordinate system is used for the vector components.
-14.9, 4.4] R E , GSM, at 0400 UT) and WIND satellites, AL-index from the Kyoto monitor, H e and H p magnetic field components at geostationary orbit, X and Z components of the magnetic field and the X-component of the ion bulk velocity in the magnetotail from the Cluster spacecraft. The GSM coordinate system is used throughout this paper. The substorm during 0000 -0100 UT was associated with a southward IMF and Earthward bursty bulk flow, detected by Cluster. At ∼0050 UT, B z at Geotail turned northward and stayed predominantly northward, fluctuating between 0 and 10 nT during several hours, with several short turns south-  provided by Cluster 2 (C2). The PEACE instrument enables the measurement of the electron flux with energy 10 eV-26.5 keV; the PEACE data are provided by all four spacecraft. The telemetry transmission worked in the burst mode from 00:30 till 04:30 UT during the event discussed below, providing one ion and one electron distribution functions per spacecraft spin (∼4 s), and the magnetic field with maximum time resolution is 67 Hz. Figure 1 presents an overview plot of the interplanetary media, and the magnetosphere state history during 00:00-06:00 UT on 15 September 2001: The solar wind dynamic pressure P dyn at the WIND satellite orbit ([53.5, −38.9, 19.4] R E , GSM, at 04:00 UT), IMF B y and B z at Geotail ([9.1, −14.9, 4.4] R E , GSM, at 04:00 UT) and WIND satellites, AL-index from the Kyoto monitor, H e and H p magnetic field components at geostationary orbit, X-and Zcomponents of the magnetic field and the X-component of the ion bulk velocity in the magnetotail from the Cluster spacecraft. The GSM coordinate system is used throughout this paper. The substorm during 00:00-01:00 UT was associated with a southward IMF and Earthward bursty bulk flow, detected by Cluster. At ∼00:50 UT, B z at Geotail turned northward and stayed predominantly northward, fluctuating between 0 and 10 nT during several hours, with several short turns southward after 03:40 UT. IMF B y at Geotail and WIND showed similar variations with lag time of 5 min. The solar wind dynamic pressure (P dyn ) decreased gradually from 8 down to 3 nPa during 03:00-04:00 UT. A drop of H e from ∼100 down to 70 nT with a simultaneous increase of H p was observed by GOES 8 at the onset of the first flow burst (03:41 UT). IMF B z at Geotail and WIND was positive (∼10 nT). A set of jump-like variations in B y and B z IMF were detected by WIND and Geotail between 03:20-03:40 UT. Some minor activations (with |AL|<100 nT) correspond to the flow bursts. Distinct auroral activations (pseudo-breakups and small substorms), corresponding to the tailward flow bursts observed in the plasma sheet, were detected by the CANOPUS stations (not shown, see Voronkov et al., 2006, for detailed description and timing) and the IMAGE satellite.

Event description
Further in this paper we will discuss the Cluster measurements during the three tailward flows observed between 03:40-04:07 UT (highlited interval in Fig. 1). The state of the magnetosphere is changed rapidly at about 04:07 UT (Voronkov et al., 2006). This change is, most likely, triggered by short southward IMF turn, detected by WIND and Geotail at about 04:00 UT (see Fig. 1). The two intensive tailward flows at 04:10 and at 04:25 UT, as well as the flow reversal at about 05:10 UT, studied by Xiao et al. (2006), are associated with substorm signatures, while the less intensive flow bursts during 03:40-04:07 UT correspond to rather pseudo-breakup activations (Voronkov et al., 2006).
During 03:30-04:10 UT Cluster stayed in the magnetotail plasma sheet at [−18.9, 3.4, −3.1] R E (barycenter), forming a nearly regular tetrahedron with the largest inter-spacecraft distance of 1750 km. Figure 2 presents the spacecraft configuration with respect to the tetrahedron barycenter and the summary plot of the Cluster observations during this interval. The intervals corresponding to the three distinct successive tailward flow bursts, detected by Cluster, are marked A, B, and C, respectively. Figure 3 presents IMAGE-WIC aurora observations, corresponding to the three intervals.
Before the first tailward flow was detected at 03:41 UT, plasma sheet was quiet (no ion velocity exceeding 50 km/s was observed, Fig.2) and relatively hot, with ion energy ∼10 keV. The magnetic gradient in the current sheet was directed along Z (B x at C3 was smaller than that at the other three spacecraft), j y >(j x , j z ). The current sheet halfthickness, estimated using the Harris function (Harris, 1962), varied between 1-2 R E (6000-10 000 km).
The ion energy increased between 03:41-03:51 UT and between 03:59-04:06 UT, somewhat exceeding the energy range of the CIS instrument (40 keV). large-amplitude variations (up to 10 nA/m 2 , often with sign change), manifesting a complex, 3D structure of the magnetic field within the flow bursts. The current sheet was thinning down to ∼3000 km during intervals B and C. Ion moments computed from HIA (C3) and CODIF (C1 and C4) ion distributions are very similar. The increases of the convective electric field E c = −V × B indicate a large rate of the magnetic flux transport during the three flow bursts. Thus, the criteria of the flow bursts (|E c | > 2 mV/m, e.g., Nakamura et al., 2001) are satisfied.

Detailed analysis of Cluster observations
In this section we analyze the magnetic field, ion moments and electron ET spectra during the three tailward flow intervals separately. We visually inspect the low-pass filtered (cut-off frequency is 1/8 Hz) 1-s averaged magnetic field time series and corresponding vector derivatives (magnetic field curl, magnetic tensions and magnetic pressure gradient) plot- flow bursts correspond to negative variation of B z . Variations of B y were up to ∼20 nT, and negative in the northern half of the current sheet (B x >0) and positive in the southern one (B x <0). The current density, derived from fourpoint magnetic field measurements (e.g. Chanteur, 1998), increased during intervals B and C up to j ∼10 nA/m 2 from j ∼3-5 nA/m 2 before the flow onsets, but fluctuated around ∼4 nA/m 2 during interval A. j y remained positive, indicating a constant presence of the cross-tail current during the flow bursts; j x and j z showed large-amplitude variations (up to 10 nA/m 2 , often with sign change), manifesting a complex, 3-D structure of the magnetic field within the flow bursts. The current sheet was thinning down to ∼3000 km during intervals B and C. Ion moments computed from HIA (C3) and CODIF (C1 and C4) ion distributions are very similar. The increases of the convective electric field E c =−V×B indicate a large rate of the magnetic flux transport during the three flow bursts. Thus, the criteria of the flow bursts (|E c |> 2 mV/m, e.g. Nakamura et al., 2001) are satisfied. From top to bottom: Magnetic field magnitude; X, Y , and Z components of the magnetic field at the four spacecraft; standard deviation of B; X, Y , and Z components of the electric current density j=µ −1 0 ∇×B; The angle between j and B; X, Y , and Z components of the j×B force, magnetic pressure gradient and magnetic field tensions; ion and total pressures; X and Y components of the ion bulk velocity; and Y and Z components of the E c =−V×B electric field. The GSM coordinate system is used for vectors.

Detailed analysis of Cluster observations
In this section we analyze the magnetic field, ion moments and electron ET spectra during the three tailward flow intervals separately. We visually inspect the low-pass filtered (cut-off frequency is 1/8 Hz) 1-s averaged magnetic field time series and corresponding vector derivatives (magnetic field curl, magnetic tensions and magnetic pressure gradient) plotted in the GSM coordinate system to identify spatial UT B x at all four probes dropped to very low values, indicating an entrance to the thick sheet with uniformly weak magnetic field. The electric current density (j=µ −1 0 ∇×B) varied between −4 and 4 nA/m 2 , with j x >0 and a set of bipolar variation of j x and j z till ∼03:45 UT. Intervals of the two distinct magnetic field variations (between 03:41:40-03:43:00 UT and between 03:43:00-03:46:00 UT) are marked by A1 and A2, respectively. During interval A1, the angle between j and b ( b, j) deviates from 90 • first to 30 • at 03:42 UT, then to 140 • , then gradually turns to the direction perpendicular to b. During interval A2, b, j varies between 100 • and 180 • . Cluster started to detect a tailward bulk flow with V x <−50 km/s at ∼03:41 UT. The Y-component of the bulk velocity was negligible comparing with V x . The flow decays at ∼03:47 UT. Until 03:44:40 UT, the j×B force was mainly contributed by the magnetic pressure gradient (∇P b ), the magnetic tension force (µ −1 0 (B·∇)B) was a factor of 2 smaller than ∇P b ; the ∇ z P b -component was the largest one and positive at B x >0. Both the ion pressure P i and the sum of ion and magnetic pressure P t =nkT i +B 2 /2µ 0 gradually increased until ∼03:42 UT. At ∼03:42 UT the ion pressure dropped down, showing a broad minimum during 03:43:30-03:45:30 UT. The total pressure decreased after 03:42 UT too, then increased with a local maximum at ∼03:43:30 UT. During 03:43:30-03:45:30 UT the total pressure was 15% larger than the ion pressure. The cross-tail (Y ) and vertical (Z) components of the −V×B electric field showed a broad maximum with E y ∼4 mV/m and a minimum with E z ∼−5 mV/m between 03:43:30-03:45:30 UT. The magnetic field fluctuations, calculated from non-filtered FGM data as standard deviations of B within 12 s (3 spins) sliding windows increased simultaneously with the bulk flow, but do not exceed 0.3 B during intervals of significant flux transfer (E c >2 mV/m).
Performing the deHoffmann-Teller analysis (HT analysis, Khrabrov and Sonnerup, 1998) Eriksson et al., 2004). Thus, no significant plasma acceleration in the H T frame, required by the fast reconnection model, was detected.
The electron pitch-angle ET spectrograms from the PEACE instrument at all four spacecraft are shown in Fig. 5 During interval A1 all four spacecraft detected significant and similar shape variations of the Y and Z magnetic field components, while amplitudes of B x variations did not exceed ∼5 nT. B y at the barycenter, previously slightly negative, first increasee to 4 nT then decreased down to −9 nT, then increases again to 4 nT. B z first increased up to 5 nT then decreased down to −2 nT.
Minimum Variance Analysis (MVA, e.g. Sonnerup and Scheible, 1998)  and exhibits the bipolar north-to-south variation. C1, C2, and C3 showed a short increase of |B|, while C4 showed a local minimum of |B|.
The similarity of the B max trace shapes at the four probes enables timing analysis (Harvey, 1998) of the B max . Fourpoint timing gives a unique solution of linear system for three components of the velocity normal to the planar boundary crossed by spacecraft. In our case, the dominant component of the plasma velocity is tailward (Fig. 4). We assume that the magnetic structure is transported by the plasma flow, i.e. moves tailward. Thus, a deviation of the timing normal from pure tailward direction gives a tilt of the boundary in XY and XZ planes, indicating a curved shape of the tailward moving magnetic structure encountered by the spacecraft constellation.  Fig. 7c. The first boundary is moving tailward and northward, indicating Cluster encounters the tailward moving surge with the magnetic field elevation. The second one moves tailward and southward, indicating Cluster exit the magnetic surge. The Y-component of the normal to the first boundary is negative, however small, while for the second boundary, n y is positive and large. These results may be interpreted as signatures of crossing the tailward moving magnetic structure with an ellipsoid-like boundary. The spacecraft crosses the structure close to its dawn-side edge, entering through dawnward moving boundary and exiting through duskward moving one. This boundary is shown by dashed line on the interpretation sketch (Fig. 7c) be interpreted as signatures of the left-handed flux rope like structure (Slavin et al., 2003). Yet, the MVA results disagree with classic constant-α model (see, e.g. Slavin et al., 2003;Henderson et al., 2006). Since the spacecraft crossed an upper (northern) part of this structure only, we can not state that it is closed, forming a magnetic island. It also may be an NFTE-type (Sergeev et al., 1992) or CD-related (Ohtani et al., 2004) surge of the current sheet. The electric current deviated from the perpendicular direction to ∼150 • , and |j×B| has a local minimum during A1. The X-aligned scale of this structure, estimated from the average V x during 03:42:00-03:42:50 (∼170 km/s) and the duration of the structure observation, is ∼8000 km ∼1.25 R E . During interval A2 the total magnetic field first increased up to 15 nT, varies at this level, and decreased down to 1 nT. Since B x is rather stable during the major part of this interval, except for the the drop at ∼03:45:30 UT, the increased of |B| is due to large |B y | and |B z | (both are negative). Timing analysis of the |B|-traces (Fig. 4)  .55]·101 km/s. Thus, show that the first boundary moved tailward, dawnward and southward while the second one moved tailward, duskward and northward. In a volume, bounded by these fronts, the magnetic field rotated from the Earthward-duskward-northward to Earthwarddawnward-southward enhancing its strength due to increase of cross-tail component magnitude, anti-parallel to the electric current. During a short time between ∼03:43:15-03:43:30 the magnetic tension and the magnetic pressure gradient were comparable and the Z-component of the magnetic tension force is negative (Cluster was in northern half of the plasma sheet). Then, the j×B force was dominated by the magnetic pressure gradient, and the Z-component of B- tension changes its sign to positive. Since ∇ z P b dominates and j y >0 and relatively stable, the spacecraft probe a sheetlike structure with the normal close to Z GSM . Figure 8 presents the magnetic field vectors at the four spacecraft in XY-and XZ-planes at 03:43:00 UT (ahead of the structure), 03:44:00, 04:45:00 and 03:45:30 UT. The bottom panel presents an interpretation scheme: the solid lines present the magnetic filed field force lines. Cluster crossed an X-elongated magnetic field surge with the field-line curvature first directed along −Z then along +Z. The structure was bounded in the cross-tail direction, and Cluster enters through its dawn-side boundary and crossed this object near its dawn-side edge (see Fig. 7c). This explains the fact, that most duskward probe (C2) started to detects the electron flux enhancement earlier and detected it longer than the others (see Fig. 5). The duration of this structure is about 130 s and average ion velocity ∼315 km/s, which gives the structure length ∼6 R E . These observations may be interpreted as the signatures of the post plasmoid plasma sheet (e.g. Ieda et al., 1998) with an X-elongated southward magnetic field loop. increased from ∼10 to ∼20 nT. B x is positive at C1, C2 and C4 (the northern group), while the most southern probe (C3) measured B x fluctuating near zero level. B x at the northern group increased around 03:50:00 UT, while B x at C3 slightly decreased, down to −4 nT at this time. B y shows a remarkable north-south difference: B y at the northern group displayed a pronounced minimum with the value of −17 nT around 03:50:00 UT (∼15 s early at C2), while B y at C3 slowly varied around zero with a local maximum (∼2 nT) at this instance. B z at the northern group exhibited a clear bipolar variation, changing from ∼4 to ∼−10 nT. B z at C3 showed no positive variation, gradually decreasing from approximately zero to ∼−10 nT. The Y-component of the current density was positive, while j x and j z (j y >|j x |, |j z |) showed bipolar variations (j z changes sign from negative to positive). j was close to be perpendicular to the magnetic field ahead of the structure; b, j varied between 90 • and 120 • inside the structure, because of the duskward rotation of b with ∇B directed mainly northward (j deviation from the nominal direction did not exceed 20 • around 03:50:00 UT). The Z-component of the j×B force was negative (B x >0 at the Cluster barycenter; |(j×B) z |>|(j×B) x |, |(j×B) y |). The magnetic field gradient was a factor of 7 larger than the Btension force. The major contribution to the force was from ∇ z P b . The X-component of the B-tension force changed sign from positive to negative, corresponding to the B z change. The Y-and Z-components of the B-tension force were approximately of the same value as the X-component, and changed their signs from negative to positive (Y ) and from positive to negative (Z), indicating that Cluster crossed a 3-D structure. The total pressure at C1 and C4 had a local maximum near 03:49:45 UT, while P t at C3 shows no significant change. The ion thermal pressure slightly decreased. The magnetic field variations (increase of |B|, strong B y <0, north-to-south variation of B z ), observed by the northern group of the spacecraft (C1, C2, and C4) may be interpreted as the signatures of a left-handed flux rope (Slavin et al., 2003). Staying near B x =0, C3, however, showed different signatures: B y ≥0, B z <0 (no bipolar signatures), no increase of P b . According to the flux rope interpretation, one expects the same direction of the core filed in the center and at the northern half of the rope. The proper interpretation should explain the B y difference, and the absence of bipolar signatures in C3 B-trace.

Interval B
Since the average direction of the current density vector was close to Y (e j =[0.37, 0.80, −0.19]) and the average direction of the magnetic pressure gradient, pointing along the normal, was close to Z (e pb =[0.22, 0.10, 0.94]), the GSM system is a good proxy for the natural coordinates. Figure 11a-c presents a multi-point view of the magnetic field structure in XY , XZ, and Y Z planes. The observations are interpreted in the following way. During 03:49:30-03:50:30 UT Cluster crossed a magnetic loop, closed from the tailward side with the north-south size comparable or larger than the Cluster inter-spacecraft separation, embedded into a more thicker current sheet C3, crossing the the structure near it's axis did not detect a northward B z , showing B z <0, while the northern group (C1, C2, C4) detected north to south B z variation. The northward B z turn is due to local magnetic field compression by the plasma flow. This can explain the increase of the magnetic pressure, detected by the northern group. Thus, the situation resembling rather NFTE with the magnetic structure similar to quasi-plasmoid configuration obtained in simulations of nonlinear plasmoid evolution (Abe and Hoshino, 2001), than a flux rope. However, the observed structure is more complex, including an out-ofplane magnetic field component (B y ) directed dawnward in the northern half and, likely, duskward in the southern half of the loop (C3 at B x ≤0 detects B y ≥0). The length of this structure along X-direction, estimated using V x ∼150 km/s and the structure duration, ∼60 s, is ∼1.5 R E . The magnetic field variation as well as the electron flux enhancement was first detected by C2, then by C3, C1 and finally by C4. This indicates that the magnetic structure was bounded in cross-tail direction, and crossed by Cluster near its dawn-side boundary, like it shown in Fig. 7c (valid for northern half of the structure only). In this case, C2 stayed within the structure longer than than the other probes, which can explain a longer interval of enhanced electron flux, observed by C2. The enhancement of bi-directional electron flux indicates, that the loop is closed further downtail. with maximum velocity of 400 km/s. Magnetic field variations, associated with significant flux transfer (E c ∼3 mV/m, |E cz |>E cy ) were detected during 04:00:30-04:02:00 UT (dashed-line box). The magnetic field absolute value increased from ∼5 to ∼15 nT while B x decreases. B y at C1, C2, and C4 (all are situated in the northern part of the plasma sheet) experienced a negative variation with amplitude of 10 nT, while B y at C3, located near the neutral sheet in the southern part of the sheet (−5<B x <−1 nT), exhibited a positive B y variation with amplitude of 10 nT. All four probes detected bipolar variation of B z with B z ∼0 at the maximum of |B y |. These magnetic field variations corresponded to a local enhancement of the X-component of the current density. The j×B force had a local maximum in Z-and a bipolar variation in the Y-components.

Interval C
[j×B] x stayed close to zero until ∼04:01:20 UT then turned negative simultaneously with the decrease of the total and ion pressures. The center of the bipolar B z variation coincided with the local maximum of P t . The current density vector was directed mainly perpendicular to the instantaneous magnetic field: (( bj) varied between 80-120 • ).
HT analysis of the C1 data during 03:59:40-04:02:00 UT results with V H T =[−208.7, 56.14, −57.86] km/s with well defined H T -frame: cc=0.98, s=1.01. Walèn test with cc=0.511 and s=0.226 again shows vanishing acceleration of plasma in the H T -frame. Figure 13 presents electron ET spectrograms for interval C. During 04:01:30-04:02:30 UT, when the above described magnetic field structure was observed, all the four spacecraft detected local enhancement of parallel and antiparallel electron fluxes with energies of 0.5-3 keV. The flux enhance-ment was first detected by C1 and C3, then by C2 and C4, contrary to the observations during interval B. Some, but smaller enhancement of the perpendicular flux was also detected mainly after 04:02:00 UT. Again a sufficient decrease of low energy <200 eV electron flux after 04:01:30 UT was detected. The average magnetic field direction during the electron flux increase was [0.21, −0.19, −0.64] at C1, [0.34, −0.27, −0.59] at C2, [−0.52, 0.51, −0.66] at C3, and [0.53, −0.49, −0.53] at C4. The electron flux features were the same at C3, situated in the southern half of plasma sheet detecting duskward magnetic field, and at C1, C2, C4, located in the northern half and detecting dawnward magnetic field. The increase of electron flux started about 10-15 s later than B z =0 was detected (∼04:01:15 UT).
Although the B y and B z signatures during 04:01:00-04:02:00 UT were similar to that during interval B, the B x Ann. Geophys., 26,[709][710][711][712][713][714][715][716][717][718][719][720][721][722][723][724]2008 www.ann-geophys.net/26/709/2008/ behavior, and timing of the bi-polar B z variations at the four probes are different. The difference between B x at the northern group and that at C3 ( B x ) decreased while B y increased, indicating a presence of X-directed current. Figure 14 shows projections of the magnetic field vectors at the four spacecraft onto XY , XZ, and Y Z planes at 04:01:00 and 04:01:30 UT. At 04:01:00 UT Cluster crossed the loop with northward directed magnetic field (B z >0), and at 04:01:30 UT Cluster crosses the loop of southward directed magnetic field (B z <0). Timing of B z time series shows that the maximum B z was first detected by C1 then by C3, C2, and, finally, by C4. The lag time vector (with respect to C1) is dt=[6., 5., 8.] s, which gives the normal velocity V n =[−0.91, 0.42, 0.04]·210 km/s. The minimum B z was detected with the lag dt=[6, 4, 10] s with respect to C1. This gives V n =[−0.93, 0.28, 0.23]·170 km/s: contrary to observations during interval B, the both boundaries move duskward with respect to the spacecraft. The magnetic field observations showed signatures of a current filament, with j x >j y , transported by the ion bulk flow. Cluster crossed the magnetic structure entering through its dusk-side boundary with (B·∇)B<0, and exiting through its dawn-side boundary, detecting (B·∇)B>0. The interpretation scheme is shown in Fig. 14d and e. Although the magnetic tension magnitude is comparable with that of the magnetic pressure gradient, the structure was not force-free, because the B-tension force was directed mainly along ±Y , while the ∇P b was directed mainly along Z. Current density vector did not significantly deviate from the B-perpendicular direction. The cross-tail size of this structure may be roughly estimated using mean value of V y at 04:01:00-04:01:45 UT, equals to ∼60 km/s and the lag between negative and positive (B·∇)B peaks, ∼45 s. This yields L y =2700 km∼0.5 R E . The scale along X estimated form V x ∼200 km/s and the structure duration is about 1.5 R E .

Discussion and conclusions
We have shown in-situ observations of three successive 2-7 min long tailward flow bursts with velocities of about 400 km/s detected in the near-Earth magnetotail at X GSM =−18.9 R E near the midnight meridian. In contrast with previously discussed observations of tailward fast flows in the near-Earth magnetotail (Nagai et al., 1998a;Eriksson et al., 2004;Miyashita et al., 2005), these three flow bursts were observed during northward IMF. IMF B z was northward (of 5-10 nT) during ∼2.5 h before the first tailward flow was observed. IMF B y was mainly dawnward and fluctuating. The first flow was detected 2.5 h after the large substorm followed by B z decreasing after a strong dipolarization (see Fig. 1), accompanied by the gradual increase of the magnetic gradient in the current sheet. Thus, the activity in the near-Earth magnetotail plasma sheet, leading to tailward flows, was, likely, internally triggered.  Although the B y and B z signatures during 0401:00 -0402:00 UT were similar to that during interval B, the B x behavior, and timing of the bi-polar B z variations at the four probes are different. The difference between B x at the northern group and that at C3 (∆B x ) decreased while ∆B y increased, indicating a presence of X-directed current. Fig. 14   The magnetic field fluctuations σ (B) were about 0.4 B during the flow bursts, therefore turbulent current disruption, implies σ (B)∼ B (e.g. Lui, 1996;Lui et al., 2006), hardly can be a mechanism of the flow generation. j y is positive and increases during the flow intervals B and C, thus, the ballooning mechanism, implying the generation of the duskdawn component of the electric current (Roux et al., 1991(Roux et al., , 2006, can not be considered as the source of these two tailward flow bursts either. During interval A, however, j y was oscillating between −1 to 3 nA/m 2 with a peak-to-peak period ∼30 s. This may be interpreted as the quasi-periodic generation of the dusk-dawn current. We also found the systematic lag between the magnetic field fluctuations at C1 and C2 (separated mainly along Y ), indicating azimuthal propagating waves, predicted by the models of CD (Roux et al., 1991;Lui, 1996). No signatures of fast reconnection (see, e.g. Semenov et al., 2004) were found during the three flow bursts: Plasma velocity in the de Hoffmann-Teller frame did not exceed 25% of the Alfvèn speed, which is much smaller than the acceleration, predicted by fast reconnection model www.ann-geophys.net/26/709/2008/ Ann. Geophys., 26, 709-724, 2008 (see, e.g. Eriksson et al., 2004). Thus, the tailward flows were generated rather due to a weak reconnection on closed field lines (or in the course of X-line formation by the ballooning type perturbation). The three flow bursts were associated with a local auroral activation in the MLT sector of the Cluster foot point, ∼5 • equatorward. The time relationship between plasmoids and auroral pseudo-breakups was studied by Ieda et al. (2001). They found that the earliest plasmoid was observed at X=−28 R E about 2 min before the auroral brightening. Typically, plasmoids are observed 0-2 min after the brightening. This was interpreted that reconnection occurs before the auroral activation, and a "young" plasmoid is observed before or simultaneously with the brightening if the spacecraft is located near the reconnection site. It follows from this model, that during intervals A and may be C Cluster was situated near the reconnection site, while during interval B reconnection occurs closer to the Earth. It should be noted, however, that this interpretation is based on a 2-D model with a plasmoid infinity long in cross-tail direction. Our observations show that the magnetic structures, observed during tailward flows in the near-Earth tail, are localized in the Y GSM direction. Nakamura et al. (2001) studied the relative timing between the bulk flow bursts (mainly Earthward), observed in the near tail, and localized auroral activations. They found that on average, the aurora precedes the flow activations by 0-3 min. If the auroral activations starts within 1 h MLT difference and 2 • latitude distance from the spacecraft foot point, the time delay is reduced down to less than 30 s. Our results for the tailward flows are similar: The auroral activation during intervals A, with the minimum delay with respect to the maximum |E c | observation, was observed around the MLT of the Cluster foot point; delays for B and C activations, observed within ±0.5 h MLT, are of −2.5 and +3 min, respectively.
Each observed flow burst corresponds to a similar magnetic field variations. The magnetic field turns dawnward (duskward) in the northern (southern) half of the plasma sheet. The spacecraft, situated above the neutral sheet detected a more or less pronounced north-to-south bipolar variations of the magnetic field. All probes show the southward magnetic field around minima of the flow velocity. The magnetic field magnitude increases by ∼10 nT, and this increase is due to an enhancement of B y . A total pressure enhancement, considered as the signature of a plasmoid , was observed during the interval B only, when the ion pressure was nearly constant. The increase of the magnetic pressure was compensated by a decrease of the ion pressure during intervals A and C.
The important and novel aspect of this case study is that, contrary to the case studies with small spacecraft tetrahedron (e.g. Eastwood et al., 2005;Henderson et al., 2006), the spacecraft configuration gives the possibility to monitor the neutral sheet vicinity B x ≈ 0 and the plasma sheet at B x ≈10-15 nT simultaneously, allowing to distinguish be-tween a closed plasmoid-like magnetic field configuration and an NFTE-like one, with a tailward-open magnetic surge. The analysis of the four-point magnetic field measurements have shown that the aforementioned bipolar north-to-south magnetic field variations, commonly interpreted as signatures of plasmoid or flux ropes, are not necessarily indications of a closed O-type magnetic field structure. They may be signatures of the NFTE-like surge (interval B) or indicate a crossing of an Earthward-duskward directed current filament (interval C).
In contrast to results, reported by Zong et al. (2004), the bi-directional electron fluxes were detected after the bipolar variation detections. They associated, therefore, with the NFTE-like southward magnetic field loops, indicating that they are closed further in tail. The spacecraft enters into the flux tubes, containing the bi-directional electron flux, after crossing the flow boundaries, manifesting as the bi-polar magnetic field variations.
The B y signs in these cases are different in the northern and southern halves of the plasma sheet. Thus this is neither a core field, assumed to be the same direction in both halves in the flux rope models, nor the guide field enhanced due to pile-up effect in localized reconnection of sheared magnetic field with a finite cross-tail length of X-line (Shirataka et al., 2006), but corresponds rather to the quadrupolar magnetic structure due to the Hall effect near X-line (Nagai et al., 1998b), It was shown, that the Hall currents, existing in the ion diffusion region, are closed by the system of the fieldaligned currents (FAC), which may exist far away from the X-line, keeping the cross-tail magnetic field component, negative in the northern and positive in the southern halves of the sheet (e.g. Fujimoto et al., 2001;Treumann et al., 2006). This cross-tail field, originated by the Hall currents, may be enhanced by the magnetic field compression within fast flow bursts. The tailward moving filament of the Earthwardduskward directed current crossed by Cluster during interval C (see Fig. 14) may also be a part of the FAC system, related to the Hall currents in the reconnection region, localized in the cross-tail direction.
The important outcome of this work is the reconstruction of 3-D shapes of the magnetic structures, embedded into the tailward plasma flows. Using the spacecraft separation in the cross-tail direction, we found that the NFTE-like surges (intervals A and B) have ellipsoidal shapes. The spacecraft cross their dawn-side boundaries. cially supported by the German Bundesministerium für Bildung und Forschung and the Zentrum für Luft-und Raumfahrt under contracts 50 OC 0104.
Topical Editor I. A. Daglis thanks two anonymous referees for their help in evaluating this paper.