The reconnection site of temporal cusp structures

[1] The strong precipitating particle flux in the cusp regions is the consequence of magnetic reconnection between the interplanetary magnetic field and the geomagnetic field. Magnetic reconnection is thought to be the dominant process for mass, energy, and momentum transfer from the magnetosheath into the magnetosphere. Observations of downward precipitating cusp ions by polar orbiting satellites are instrumental in unlocking many questions about magnetic reconnection, e.g., their spatial and temporal nature and the location of the reconnection site at the magnetopause. In this study we combine cusp observations of structures in the precipitating ion-energy dispersion by the Cluster satellites with Super Dual Auroral Radar Network radar observations to distinguish between the temporal and spatial magnetic reconnection processes at the magnetopause. The location of the cusp structures relative to the convection cells is interpreted as a temporal phenomenon caused by a change in the reconnection rate at the magnetopause. The 3-D plasma observations of the Cluster Ion Spectrometry instruments onboard the Cluster spacecraft also provide the means to estimate the location of the reconnection site. While an earlier study of a spatial cusp structure event revealed bifurcated reconnection locations in different hemispheres as origins for the precipitating ions creating the cusp structures, the same method applied to the temporal cusp structures in this study shows only a single tilted reconnection line close to the subsolar point. Tracing the distance to the reconnection site provides not only the location of the reconnection line but can also be used to distinguish between spatial and temporal cusp structures.


Introduction
[2] The magnetospheric cusp regions always play an important role in the study of magnetic reconnection at the Earth's magnetopause.Whatever process occurs at the magnetopause, a signature of the process can be found in the precipitating magnetosheath ions observed in the cusp [e.g., Lockwood and Smith, 1994;Onsager et al., 1993] as well as down to the ionosphere where the foot points of all magnetopause field lines converge to a relatively confined space [e.g., Frey et al., 2003].Plasma observations in the cusp in addition to in situ observations of field-aligned plasma flows close to the magnetopause (consistent with the occurrence of magnetic reconnection) have demonstrated that the fundamental process of magnetic reconnection occurs somewhere at the magnetopause for any condition of the interplanetary magnetic field (IMF) [e.g., Sonnerup et al., 1981;Gosling et al., 1991;Kessel et al., 1996;Phan et al., 1996;Fuselier et al., 2000a].
[3] Precipitating ions in the cusp exhibit a distinctive dispersion profile caused by the joint action of newly opened convecting field lines and different flight times of the ions from the injection point at the magnetopause to the observing satellite.For a southward interplanetary magnetic field and poleward convecting flux tubes, lower-energy particles arrive at successively higher latitudes.This velocity filter effect for reconnection during southward IMF conditions was predicted by Rosenbauer et al. [1975] and first observed by Shelley et al. [1976].
[4] For a steady rate of reconnection at the magnetopause, the ion energy of downward precipitating ions should show a smooth and continuous latitudinal dispersion on these open field lines.This smooth dispersion is not often observed.Instead these dispersions show complicated structures known as ''stepped'' or ''staircase'' cusp ion signatures with variations in flux levels and sudden changes in the energy of the precipitating ions [e.g., Newell and Meng, 1991;Escoubet et al., 1992].These cusp steps have been interpreted as both spatial and temporal structures.
[5] While spatial cusp structures are the result of crossing the boundary between spatially separated flux tubes [e.g., Onsager et al., 1995;Wing et al., 2001;Trattner et al., 2002a], temporal structures are caused by variations of the reconnection rate with periods of little or no reconnection that are interspersed with periods of continuous reconnection.The existence of temporal cusp steps has been predicted and is very well described in the pulsating cusp model [e.g., Cowley et al., 1991;Lockwood and Smith, 1990;Smith et al., 1992] and supported by in situ observations at the magnetopause [e.g., Russell and Elphic, 1979;Farrugia et al., 1998;McWilliams et al., 2004;Escoubet et al., 2006] and transient reconnection signatures in the ionosphere [e.g., Lockwood andSmith, 1989, 1994].This pulsating cusp model reduces to a steady state cusp with constant reconnection rate at one limit, and a cusp observed as a series of discrete events (bursts) at the other limit [Smith and Lockwood, 1990].
[6] Multisatellite observations in combination with remote sensing methods like the FUV imager on the Imager for Magnetopause-to-Aurora Global Exploration (IMAGE) mission or the Super Dual Auroral Radar Network (Super-DARN) radar arrays have been used to distinguish between spatial and temporal cusp structures [e.g., Onsager et al., 1995;Wing et al., 2001;Trattner et al., 2002bTrattner et al., , 2005]].The recent progress in determining the location of the reconnection site from 3-D plasma observations in the cusp, applied onto multiple cusp steps, provides an additional method which also distinguishes between spatial and temporal cusp structures.This so-called low-velocity cutoff method [e.g., Onsager et al., 1990] uses the velocity difference between precipitating and mirroring ion distributions in the cusp to estimate the distance to the reconnection site, which is subsequently traced back from the satellite position to the magnetopause along the T96 model magnetic field lines [Tsyganenko, 1995].Using this method on a Cluster double cusp event previously identified to be a spatial cusp structure resulted in bifurcated reconnection locations in different hemispheres for the two cusp structures [Trattner et al., 2005].
[7] It is the purpose of this study to investigate the spatial or temporal nature of ''step-up'' cusp structures observed by the Cluster satellites SC1 and SC4 on 23 September 2001.
Step-up cusp structures are sudden increases in the energy of precipitating ions and are further used to determine the location of the reconnection site for this phenomenon in order to demonstrate that the low-velocity cutoff method can also be used to distinguish between spatial and temporal cusp structures without requiring an additional source of measurements.Validating this aspect of the low-velocity cutoff method will provide a method for additional studies to analyze the periodicity of cusp structures as a function of solar wind and IMF conditions and the location of the reconnection line.

Instrumentation
[8] In this paper we present proton observations from two Cluster Ion Spectrometers (CIS) [Reme et al., 2001].These spectrometers provide high-precision 3-D distributions for the major ion species at energies from a few eV/e to about 40 keV/e every 4 s (1 spin period) in the high-time resolution mode.The CIS instruments consist of two different analyzers; the Hot Ion Analyzer (HIA) and the time-of-flight Composition and Distribution Function Analyzer (CODIF).This study will focus on proton observations by the CODIF instrument on board SC1 and SC4 which crossed the cusp at about the same time, making them ideal for documenting the periodicity of a series of step-up cusp structures and their related changes in the reconnection rate as well as their convection to higher latitudes.The CODIF instrument on satellite SC3 was operational but the observations are omitted in this study since SC3 was about 40 min behind the leading pair of satellites and encountered a different cusp precipitation profile (the sensor on SC2 failed early in the mission).All four identical Cluster satellites were launched in pairs on 15 July and 9 August 2000 into orbits with a perigee of $4 R E , an apogee of $19.7 R E , and an inclination of 90° [Escoubet et al., 2001].The Cluster observations are supplemented with simultaneous radar observations by the eight operating northern hemisphere SuperDARN radars.
[9] In addition to the Cluster and SuperDARN data, IMF and the solar wind conditions were observed by the ACE Magnetic Field Instrument (MFI) [Smith et al., 1998] and the ACE Solar Wind Experiment (SWE) [McComas et al., 1998], respectively.The IMF and solar wind data are provided by the ISTP key parameter web page.
[13] Cluster SC4 enters the cusp about one minute prior to SC1 at 11:07:14 UT but is not subjected to the same boundary motions that causes satellite SC1 to cross the open-closed field line boundary several times.However, SC4 encounters a larger series of step-up cusp structures with the first at 11:14:19 UT and 76°INVLAT (not marked), the second at 11:19:24 UT at 77.22°INVLAT (4a) and the third at 11:26:33 UT at 78.94°INVLAT (4b).Only the last two cusp step-up structures, which are in phase with the previously discussed structures 1a and 1b, are marked with vertical white lines.
[14] The first step-up cusp structure observed by SC4 was not encountered by SC1, which is most probably caused by the periodicity of this particular event.[15] It should be noted that the $7 min periodicity of the cusp steps for the 23 September 2001 event is in agreement with the 8 min average periodicity of FTE's observed at the magnetopause [Rijnbeek et al., 1984;Russell et al., 1996].A similar result was reported by Lockwood and Wild [1993], where they found that the time between successive FTE's varies from 1.5 to 18.5 min with an average of 8 min and a most common value of 3 min.Smith and Lockwood [1990] pointed out that the reconnection rate variations thought to be responsible for magnetopause FTE's should also cause the cusp to vary on timescales of about 2-20 min.
[16] Figure 2 shows solar wind conditions for the Cluster cusp crossing on 23 September 2001 observed by the ACE/ SWE and MFI experiments.The solar wind data in Figure 2 have been propagated by about 45 min to account for the travel time along the Sun-Earth line from the ACE spacecraft to the magnetopause.For the solar wind travel time calculation we have used bow shock and magnetopause locations derived from using the measured solar wind conditions together with Farris and Russell's [1994] bow shock model and Petrinec and Russell's [1996] magnetopause model.In addition, a reduced solar wind speed of 1/4 the upstream speed was assigned between the bow shock and the magnetopause.The location of the ACE spacecraft at about X (GSE) = 224 R E , Y (GSE) = À26 R E , and Z (GSE) = À10 R E for the time interval used in this study.
[17] Figure 2 (top) shows a highly variable solar wind density N between about 7 and 23 cm À3 .For the time of the Plotted are H + omnidirectional flux measurements (1/(cm 2 s sr keV/e)).The satellites encounter two distinctive ion-energy dispersion signatures typical for a cusp crossing during southward IMF conditions.These ion-energy dispersion signatures have been identified as temporal cusp features by Trattner et al. [2007a] and are marked as 1a and 1b and 4a and 4b for SC1 and SC4, respectively.cusp structures the solar wind density is about 20 cm À3 .The solar wind velocity V x observed at ACE was about 535 km/s (Figure 2 (middle)).The IMF components B x (black curve), B y (green curve), and B z (colored area) are shown in Figure 2 (bottom).For the time of interest the components average to about (3, À10, À3) nT, which represents an IMF clock angle of 254°.
[18] The combination of the Cluster observations together with simultaneous ground observations by the SuperDARN radar array has previously demonstrated the importance of multispacecraft/multiinstrument studies to understand and successfully separate the temporal from the spatial scale [e.g., Trattner et al., 2003].Figure 3 shows two composite plots that combine the temporal and spatial separations of the Cluster spacecraft in the cusp with ionospheric convection cells derived from SuperDARN radar observations.Figure 3 (top and bottom) are centered on the time when SC1 encounters the step-up cusp structures 1a and 1b, respectively.
[19] For Figure 3, the blue and red curves show the magnetic foot points of Cluster satellites SC1 and SC4 in the ionosphere, respectively.The foot points are plotted on top of geographical maps of the Greenland area.Overlaid along the magnetic foot points are the omnidirectional H + differential flux measurements (1/(cm 2 s sr keV/e)) shown in Figure 1.Measurements from the lowest-energy channel of the Cluster/CIS instruments are plotted at the magnetic foot points with higher energies extending away from the foot points.The color coding for the spectrograms is the same as in Figure 1.
[20] To clearly separate the temporal from the spatial scale, only 14 min wide sections of the color spectrograms shown in Figure 1 are attached to the magnetic foot points.White curves in the color spectrograms mark the position of the respective satellites for the time index.The extension of the color spectrogram equatorward of the white curves shows past measurements while the extension of the color spectrograms poleward of the white curves depicts future measurements of the CIS instruments when the Cluster satellites have progressed to those positions along the magnetic foot points.
[21] White solid curves in Figure 3 represent the average location of the auroral oval.Also shown is the ionospheric convection streamlines (black curves) derived from Super-DARN radar observations, where solid and dashed curves represent the dusk and dawn convection cells, respectively.These convection cells are derived from the radar line-ofsight data.The ionospheric equipotential flow streamlines Plotted are the solar wind density N observed to be about 7 cm À3 with several sudden enhancements up to 23 cm À3 , the solar wind velocity V x to be about 535 km/s, the variable magnetic field components B x (black curve) to be about 3 nT, B y (green curve) to be about À10 nT, and B z (color-coded area with blue representing southward IMF and red representing northward IMF) to be about À4 nT for the period of interest.Vertical black lines indicate the times when Cluster satellite SC1 encountered the two distinctive ion cusp steps 1a and 1b shown in Figure 1.
have been calculated using the technique of Ruohoniemi and Baker [1998], and are presented as equipotential contours.Here the fit to the line-of-sight data is made to a sixthorder spherical harmonic expansion, with the fit stabilized by a statistical pattern keyed to the upstream IMF data from the ACE satellite, delayed by 45 min to allow for the propagation time from the spacecraft to the magnetopause [Ruohoniemi and Greenwald, 1996].Such an equipotential map represents a best estimate of the global ionospheric flow response to the magnetopause processes sensed by the in situ spacecraft.
[22] The Cluster satellites over Greenland moved poleward and oblique to the ionospheric convection direction as derived from the radar data.At 11:18:36 UT (Figure 3 (top)), SC1 encountered the first step-up cusp structure (1a).SC4, positioned just downstream and about 1°poleward of SC1 along the convection path, encountered a similar step marked (4a) about 1 min later.
[23] The same scenario repeats for the second set of stepup cusp structures shown in Figure 3 (bottom).SC1 encountered the second step-up cusp structure at 11:25:24 UT (1b).The first pair of cusp structures 1a and 4a are still visible in the color spectrogram at lower latitudes.SC4, positioned again just downstream and poleward of SC1, encountered a similar step marked (4b) about one minute later and 1°higher in latitude.Observing a similar step feature about 1°higher in latitude and about 1 min later than the low-latitude satellite is consistent with a convecting temporal cusp structure as predicted by the pulsed reconnection model [see also Lockwood andSmith, 1989, 1990].
[24] Both step-up cusp structures observed by one spacecraft seem to have convected 1°poleward in the direction of the convection path within about 1 min (1.19°/min and 0.98°/min for the first and second pair of step-up cusp structures, respectively).That represents a convection velocity of about 1.8 km/s, in agreement with the observed convection speed in the ionosphere of about 1.2 km/s, as measured by the SuperDARN radars [see also Lockwood et al., 1990;Pinnock et al., 1993].
[25] Both Cluster cusp crossings are further investigated to determine the location of the reconnection line at the magnetopause for the cusp structures in an effort to establish if these temporal reconnection pulses originated at the same reconnection line.As described above, precipitating ions in the cusp exhibit a so-called time-of-flight effect which, under southward IMF conditions, causes particles with decreasing energy to arrive at successively higher latitudes [e.g., Rosenbauer et al., 1975;Shelley et al., 1976].This time-of-flight effect of cusp precipitating ions together with a method first used by Onsager et al. [1990Onsager et al. [ , 1991] ] in the Earth's plasma sheet boundary layer (to estimate the distance to the tailward reconnection site) can be used to estimate the distance to the dayside reconnection site.The procedure is generally known as the low-velocity cutoff method and was successfully used in several studies to determine the location of the dayside reconnection site White curves represent the average location of the auroral oval, while square symbols along the magnetic foot points represent 15 min tick marks.Overlaid on the magnetic foot points are 14 min wide sections of the Cluster/CIS H + flux measurements presented in Figure 1, with increasing energy away from the foot points and centered on the actual position of the Cluster satellites at 11:18.36 UT and 11:25.34UT (marked by white curves within the color spectrograms).SC1 and SC4 are deep inside the cusp and are moving oblique to the convection direction of the dusk convection cell.Cluster satellite SC4 is always located at higher latitudes than SC1 and encounters the same sequence of cusp steps observed by SC1 about 1 min later and 1°h igher in latitude; the typical signature of a temporal convecting cusp structure [Trattner et al., 2007a].
[26] For an observing satellite in the cusp at any time, protons near the low-velocity cutoffs in the parallel and antiparallel propagating populations originate near the reconnection site.The equal arrival times of the parallel and antiparallel propagating ions at these cutoffs and the known distance from the satellite to the mirror point in the ionosphere are used to estimate the distance to the reconnection line.The distance to the reconnection site X r is defined by: where X m is the distance to the ionospheric mirror point determined by using the position of the Cluster satellites in the cusp and tracing the geomagnetic field line at this position down to the ionosphere along model magnetic field lines of the T96 model.V e is the cutoff velocity of the precipitating (earthward propagating) ion distribution, and V m is the cutoff velocity of the mirrored ion distribution [e.g., Onsager et al., 1990;Fuselier et al., 2000b], determined from the three-dimensional proton observations by the CIS instrument in the cusp.
[27] Apart from using the T96 model to define X m in equation ( 1), the calculation of the distance to the reconnection site is model independent.However, to locate the reconnection site at the magnetopause and to determine the local conditions at the site requires additional modeling.
[28] The calculated distance to the reconnection site X r is traced back to the magnetopause along T96 model magnetic field lines, starting at the position of the Cluster satellites in the cusp.This study assumes that no electric fields directed along the reconnected field lines are present which would affect the distance calculation.A detailed discussion about the errors related to the field line tracing of the calculated distance to the magnetopause and the errors in the determination of the local conditions at the reconnection site found by Trattner et al. [2007b].
[29] Figure 4 (top) shows a two-dimensional cut through the three-dimensional distribution measured by the CIS instrument on Cluster SC1 for the time interval from 11:13:25 UT to 11:13:37 UT on 23 September 2001.Plotted is flux in the frame where the bulk flow velocity perpendicular to the magnetic field is zero.The plane of the two-dimensional cut contains the magnetic field direction (y axis) and the axis parallel to the Sun -Earth line.Three-dimensional flux measurements from the CIS instrument within ±45°of this plane are rotated into the plane and averaged by preserving total energy and pitch angle to produce the distribution in Figure 4 (top) [see also Fuselier et al., 2000b;Trattner et al., 2007b].
[30] Below the two-dimensional distribution is a cut through the distribution along the magnetic field direction (along the y axis of Figure 4 (top)).The solid line shows the measured flux level for this cut.For Figure 4, distributions with positive velocities are moving parallel to the geomagnetic field toward the ionosphere, while distributions with negative velocities are moving away from the ionosphere, antiparallel to the magnetospheric field.
[31] The peak of the precipitating magnetosheath distribution in Figure 4 is easily identified as the only peak at positive velocities at about 420 km/s.At negative velocities, three peaks are identified, representing the mirrored magnetosheath distribution at À670 km/s; an additional peak at À380 km/s which is lower than the velocity of the precipitating ion peak, and the ionospheric ion outflow distribu-tion around 0 km/s, which is often observed at these latitudes [e.g., Yau et al., 1985;Peterson et al., 2001].
[32] The precipitating and mirrored ion peaks are marked with vertical solid lines (Figure 4 (bottom)) and horizontal dashed lines (Figure 4 (top)).The low-velocity cutoff of the distributions is defined at the low-speed side of each beam where the flux is 1/e lower than the peak flux [see also Fuselier et al., 2000b;Trattner et al., 2005Trattner et al., , 2007b].The ions with velocities below the cutoff velocities of the precipitating and mirrored beams are either injected onto the open convecting field lines later and closer to the observing satellite or represent ionospheric ion outflow.
[33] To ensure a clear reproducible identification of the 1/e cutoff velocity in this region, especially for the mirrored distribution close to other distributions and ionospheric outflow peaks, the two-dimensional cuts of the Cluster/CIS observations are fitted with Gaussian distributions (blue curves) that are subsequently used to mark the 1/e reduced flux levels at the low-speed sides of the peak fluxes.
[34] Figure 5 shows the distance of the reconnection line versus time for all Cluster CIS1 (Figure 5 (top)) and CIS4 (Figure 5 (bottom)) 3-D distributions of the 23 September 2001 cusp crossing for which clearly separated low-velocity cutoffs could be observed.The uncertainties in the distance calculations arise primarily from the uncertainties in measuring the low-velocity cutoffs and are defined as half the difference between the velocity at the peak flux and the cutoff velocity [e.g., Fuselier et al., 2000b].The distances to the reconnection sites over the entire cusp crossings are relatively stable with a weighted average of 11.8 ± 3.1 R E for SC1 and 9.6 ± 1.8 R E for SC4.
[35] These distance calculations show that the distance from the satellite to the injection point at the magnetopause remained stable for the entire cusp crossing, including the temporal step-up cusp structures.This is in contrast to an earlier investigation about the reconnection site of spatial cusp structures [Trattner et al., 2005] where the reconnection locations for the two step-up cusp structures mapped to different hemispheres.This seems not to be the case for these temporal cusp structures.
[36] It should be noted that it is in general easier and more reliable to perform the distance calculation close to the open-closed field line boundary.At higher latitudes away from the open-closed field line boundary, the precipitating, mirrored and ionospheric outflow distributions start to merge into one almost isotropic distribution.Because of this effect, the distance calculations for the 23 September 2001 Cluster cusp event also contains distance calculations from ion measurements prior to the observations of the stepup cusp structures but misses values from cusp structure 4b.However, all measurements that led to a distance calculation show similar values, pointing to the same reconnection location.
[37] Figure 6 shows the magnetopause shear angle as seen from the sun and the location of the reconnection site (black square symbols) for the SC1 and SC4 cusp crossings on 23 September 2001.For the determination of the magnetic shear angles across the dayside magnetopause, we use the Cooling et al. [2001] analytical model at the magnetopause as the external (magnetosheath) magnetic field.The Cooling et al. [2001] magnetic field model is a restricted version of the Kobel and Flu ¨ckiger [1994] more general model (which is an analytic representation of the magnetic field throughout the magnetosheath).For the internal (magnetosphere) magnetic field, the T96 model at the Sibeck et al. [1991] ellipsoidal magnetopause is employed.Although these two magnetopause shapes are similar, they are not identical.Thus, a mapping of the draped magnetosheath field conditions along the boundary normal onto the Sibeck et al. [1991] magnetopause is performed to bring these two shapes into agreement.
[38] Red areas in Figure 6 represent regions where the geomagnetic fields and the draped IMF are antiparallel while blue and black areas represent regions where the merging fields become parallel.The antiparallel reconnection regions for the Cluster cusp crossing are bifurcated and located in the southern dusk and northern dawn region (IMF clock angle of 254°).The black circle represents the location of the terminator plane as it intersects the magnetopause.The dashed white line represents the line of maximum magnetic shear across the magnetopause.
[39] The reconnection site for the Cluster cusp event is marked in Figure 6 with black square symbols.The location of the symbols was determined by tracing the calculated distance to the reconnection site (Figure 5) along T96 model magnetic field lines back to the magnetopause.The symbols cluster around the subsolar point and far away from the antiparallel reconnection sites at high latitudes.During the equinoxes, the subsolar region is also crossed by the line of maximum magnetic shear.This component reconnection tilted X-line is consistent with a statistical study of 130 Polar cusp crossings during southward IMF conditions by Trattner et al. [2007b] who showed that the reconnection line for events with a significant IMF B Y component forms a tilted X-line along the line of maximum magnetic shear across the dayside magnetopause.
[40] This study shows that temporal cusp structures originate at the same reconnection line as the result of significant changes in the reconnection rate, in agreement with the pulsed reconnection model [e.g., Cowley and Owen, 1989].In contrast, spatial cusp structures originate from different reconnection lines [Trattner et al., 2005].Determining the reconnection location on the magnetopause for cusp structures can therefore also be used to distinguish between spatial and temporal phenomena.

Summary and Conclusion
[41] The merging of space-based particle measurements with global imaging technology and ground-based measurements has proven to be tremendously helpful in providing new insights into the nature of magnetic reconnection at the magnetopause.Particularly useful are the techniques for distinguishing spatial from temporal processes, avoiding the ambiguity of single satellite observations.Despite the success of the pulsed reconnection model [e.g., Lockwood et al., 1998] which explains the cusp and the reconnection process as a series of discrete reconnection pulses, multisatellite observations during stable solar wind and IMF conditions also showed spatial features not consistent with the pulsed reconnection model (e.g., DE-1 and DE-2 [Onsager et al., 1995], Interball-Polar [Trattner et al., 1999], DMSP [Wing et al., 2001], FAST-Polar [Trattner et al., 2002a[Trattner et al., , 2002b]], Polar-SuperDARN-IMAGE [Trattner et al., 2005]).
[42] In this study we have analyzed the 23 September 2001 Cluster cusp crossing in an effort to determine its spatial or temporal nature and document the location of the reconnection site from which the observed step-up cusp structures originated.An earlier Cluster study of spatial cusp structure by Trattner et al. [2005] revealed that spatial cusp structures originated in different hemispheres at bifurcated reconnection lines in the antiparallel reconnection region.This result was also confirmed with images from the FUV instrument on the IMAGE spacecraft which showed the typical ionospheric response to precipitating cusp ions originating at the antiparallel reconnection region in different hemispheres.
[43] Combining Cluster SC1 and SC4 ion observations in the cusp with the SuperDARN radar observations reveals that the step-up ion-energy dispersions are convecting within the dusk convection cell and do not cross the boundary between the convection cells.The convection speed of the cusp structures was determined to be about 1°/min which translates into 1.8 km/s, in agreement with the convection speed measured by the SuperDARN radar in the ionosphere of about 1.2 km/s.
[44] In addition the reconnection site for the temporal cusp structures has been determined.The reconnection site allows for distinction between the reconnection models.These two models are (1) antiparallel reconnection where shear angles between the magnetospheric field and the IMF are near 180°and (2) component reconnection where shear angles between the merging fields are as low as 50° [ Gosling et al., 1982[ Gosling et al., , 1990].
[45] One popular component reconnection model is the tilted neutral line model [e.g., Sonnerup, 1974;Cowley, 1976;Sonnerup et al., 1981;Cowley and Owen, 1989;Fuselier et al., 2002] which predicts that a neutral line runs across the dayside magnetosphere through the subsolar region regardless of the magnitude of the B y component.The magnitude of the B y component would only determine the tilt of the neutral line relative to the equatorial plane.
[46] Both reconnection scenarios have a profound impact on the location of the X-line and the plasma transfer into the magnetosphere.Earlier studies provided clear evidence that either the antiparallel reconnection scenario [e.g., Gosling et al., 1991] or the component reconnection tilted X-line scenario [e.g., Onsager and Fuselier, 1994;Sandholt et al., 1998;Fuselier et al., 1997;Onsager et al., 2001] occur at the magnetopause.
[47] In a recent statistical study by Trattner et al. [2007b], the authors have tested the antiparallel and tilted X-line reconnection models by investigating 130 Polar cusp crossings during southward IMF conditions during variable IMF and solar wind conditions.The study revealed that, depending on IMF conditions, either type of reconnection can occur.For southward IMF conditions (or up to 20À30°f rom the southward direction), reconnection occurs where the merging fields are exactly antiparallel, often bifurcating to the northern and southern cusp depending on the dipole tilt and the IMF B X component.For all other IMF orientations a tilted X-line crossing the dayside magnetopause is formed which basically follows the location of the line of maximum magnetic shear across the magnetopause.The study also revealed a very pronounced seasonal effect for the tilted X-line model that ''pushes'' the reconnection location to the south or to the north in the northern hemisphere summer and winter months, respectively.Exceptions are large IMF B x cases with jB x j/B > 0.7, for which reconnection switches back to antiparallel reconnection.
[48] To calculate the distance to the reconnection site for the cusp ion-energy dispersion events observed by Cluster satellites, we use 3-D ion observations by the CIS instruments to extract the low-velocity cutoffs of the downward precipitating and the mirrored magnetosheath distributions in the cusp.This information is used in a time-of-flight model [see Onsager et al., 1990;Fuselier et al., 2000b] to calculate the distance to the reconnection site.
[49] The weighted average distances to the reconnection sites are 11.8 ± 3.1 R E and 9.6 ± 1.8 R E for SC1 and SC4, respectively.These calculated distances to the reconnection site are relatively stable over the entire cusp pass, including the temporal step-up structures.Temporal convecting cusp structures originate at the same reconnection line in contrast to spatial cusp structures, which originated from vastly different locations in different hemispheres.The unchanged location of the reconnection line confirms that the step-up structures seen in Figure 1 are entirely temporal in nature.To create a step-up cusp structure in ion-energy dispersion profiles originating at the same magnetopause location, a drastic change in the reconnection rate must have occurred.
[50] The calculated distance to the reconnection site was subsequently traced back to the magnetopause along the geomagnetic field starting at the position of the Cluster satellites by using a semiempirical magnetic field model (T96).The study revealed that the temporal ion-energy dispersion events for the 23 September 2001 Cluster cusp event originated in the vicinity of the subsolar point, in agreement with the tilted X-line model.The result is also in agreement with the statistical study by Trattner et al. [2007b] who determined that for a cusp event during equinox with an IMF clock angle of 254°, the reconnection line would be along the line of maximum magnetic shear crossing the subsolar point (tilted X-line).
[51] This result demonstrates that the distance to the reconnection line can also be used to distinguish spatial from temporal cusp structures.In addition, with the ability to distinguish between the spatial and temporal nature of cusp structures, we now can systematically document the periodicity of temporal cusp structures as a function of solar wind and IMF conditions and the location of the reconnection line, and therefore also ascertain the variability of the reconnection rate.
[52] The periodicity of the cusp structures for the 23 September 2001 event was about 7 min which is in agreement with the average periodicity of about 8 min reported for FTE's observed at the magnetopause [Rijnbeek et al., 1984;Russell et al., 1996;Lockwood and Wild, 1993].Le et al. [1993] reported no evidence that FTE's are directly correlated with the southward turning of the IMF but seem to occur more frequently when IMF Bz is steady.This result was supported by Russell et al. [1996], who found no evidence for a connection between outside drivers and the occurrence rate of FTE's.In contrast, Lockwood and Wild [1993] suggested that FTE's are driven by IMF variations in which the Bz component becomes more southward.Such a behavior was evident during the 23 September 2001 Cluster cusp crossing as shown in Figure 2.For the time of interest the southward IMF B z component increased sharply twice with a periodicity of about 7 min in sequence with the appearance of cusp steps.
[53] Furthermore, recent observations suggested that temporal cusp structures caused by variations of the reconnection rate at the magnetopause (FTE's) are related to component reconnection sites [Fuselier et al., 2007].This relationship can be tested by applying the low-velocity cutoff method to temporal structures.
[54] Acknowledgments.We acknowledge the use of ISTP KP data- base.Solar wind observations were provided by D. J. McComas (ACE/ SWE).Observations of the interplanetary magnetic field were provided by N. Ness (ACE/MFI).The work at Lockheed Martin was supported by NASA contracts NAS5 -30302, NAG5 -12218, NNG05GE93G, and NNG05GE15G and a grant by the National Science Foundation under grant 0503201.We would like to thank the SuperDARN PIs (W. A. Bristow, P. Dyson, R. A. Greenwald, T. Kikuchi, M. Lester, M. Pinnock, N. Sato, G. Sofko, J.-P.Villain, and A. D. M. Walker) for providing the coordinated Cluster support radar modes which were running during this spacecraft conjunction.
[55] Amitava Bhattacharjee thanks Charles Farrugia and another reviewer for their assistance in evaluating this manuscript.
(top)) and SC4 (Figure 1 (bottom)) for the cusp crossings on 23 September 2001.Both Cluster satellites crossed into the cusp over Greenland and observed precipitating ions for the time interval from about 11:00 UT to 11:50 UT.During the cusp crossing, the satellite covered an MLT range from 11:40 to 12:08, an invariant latitude (INVLAT) range from 74.6 to 78.9°, and a geocentric distance from 4.7 to 5.1 R E .White regions in the color-coded flux plot indicate regions with flux levels above the maximum indicated flux level in the color bars.[11] The transition from closed field lines to open field lines is characterized and clearly visible by the sudden flux increase at about 1 keV/e.SC1 encounters precipitating magnetosheath ions for the first time at 11:08:15 UT and is subjected to several motions of the open-closed field line boundary which causes the satellite to exit and reenter the cusp three times before remaining on open cusp flux tubes at 11:11:25 UT.
The step-up cusp structures for the 23 September 2001 Cluster cusp crossing are separated by about 7 min (time difference between 1a and 1b and between 4a and 4b).By subtracting 7 min from the time SC1 encountered cusp structure 1a we estimate a time of about 11:12 UT; close to the time SC1 crossed the open-closed field line boundary for the final time.At this point SC1 encountered a newly opened flux tube but missed the change in the reconnection rate that launched the temporal cusp step encountered by the poleward located SC4 satellite at about 11:14 UT.The timings and positions when and where cusp structures are encountered by the Cluster satellites are also influenced by the unknown motion of the open-closed field line boundary.

Figure 1 .
Figure 1.Cluster/CIS observation from spacecraft SC1 and SC4 for the cusp crossing on 23 September 2001.Plotted are H + omnidirectional flux measurements (1/(cm 2 s sr keV/e)).The satellites encounter two distinctive ion-energy dispersion signatures typical for a cusp crossing during southward IMF conditions.These ion-energy dispersion signatures have been identified as temporal cusp features byTrattner et al. [2007a] and are marked as 1a and 1b and 4a and 4b for SC1 and SC4, respectively.

Figure 2 .
Figure 2. Solar wind parameter measurements by ACE/SWE and MFI for the Cluster cusp crossing on 23 September 2001.The solar wind data have been propagated by about 45 min to account for the travel time from the ACE spacecraft to the magnetopause.Plotted are the solar wind density N observed to be about 7 cm À3 with several sudden enhancements up to 23 cm À3 , the solar wind velocity V x to be about 535 km/s, the variable magnetic field components B x (black curve) to be about 3 nT, B y (green curve) to be about À10 nT, and B z (color-coded area with blue representing southward IMF and red representing northward IMF) to be about À4 nT for the period of interest.Vertical black lines indicate the times when Cluster satellite SC1 encountered the two distinctive ion cusp steps 1a and 1b shown in Figure 1.

Figure 3 .
Figure 3. Composite plot of Cluster satellite magnetic foot points overlaid on a geographical map of Greenland and ionospheric convection streamlines for 23 September 2001 at the times when SC1 encountered cusp structure (top) 1a at about 11:18 UT and (bottom) 1b at about 11:25 UT.White curves represent the average location of the auroral oval, while square symbols along the magnetic foot points represent 15 min tick marks.Overlaid on the magnetic foot points are 14 min wide sections of the Cluster/CIS H + flux measurements presented in Figure1, with increasing energy away from the foot points and centered on the actual position of the Cluster satellites at 11:18.36UT and  11:25.34UT (marked by white curves within the color spectrograms).SC1 and SC4 are deep inside the cusp and are moving oblique to the convection direction of the dusk convection cell.Cluster satellite SC4 is always located at higher latitudes than SC1 and encounters the same sequence of cusp steps observed by SC1 about 1 min later and 1°h igher in latitude; the typical signature of a temporal convecting cusp structure[Trattner et al., 2007a].

Figure 4 .
Figure 4. Two-dimensional representation of the three-dimensional H + ion flux distribution observed by the CIS instrument on-board Cluster/SC1showing the (top) velocity space distribution in a plane containing the magnetic field direction (y axis) and the plane parallel to the Sun-Earth line and (bottom) one-dimensional cut of the distribution above along the magnetic field direction.Precipitating magnetosheath ions move along the magnetic field toward the ionosphere with a velocity of about 420 km/s (marked with a dashed line (Figure4(top)) and a solid line (Figure4(bottom))), while the mirrored distribution from the ionosphere is observed at about À670 km/s.Also marked are the 1/e cutoff velocities at the low-speed side of the precipitating and mirrored distributions.Both distributions are fitted with Gaussian distributions (blue curves) to ensure consistent 1/e velocity cutoff definitions.

Figure 5 .
Figure 5.The distance to the reconnection line for ion-energy dispersions observed by the Cluster (top) SC1 and (bottom) SC4 on 23 September 2001.Each diamond represents a 3-D measurement by the CIS instrument, analyzed with the method described in this paper.These distances are subsequently used in the field line trace to follow the first closed T96 field line at the same local time position of the respective satellites back to the magnetopause.

Figure 6 .
Figure 6.The magnetopause shear angle as seen from the Sun, based on the magnetic field direction of the T96 model and the IMF conditions at the magnetopause for the 23 September 2001 Cluster cusp crossing.The IMF was draped around the magnetopause using the model by Cooling et al. [2001].The circle represents the magnetopause shape at the terminator plane.The dashed white curve represents the location of the line of maximum magnetic shear across the magnetopause.Square symbols represent the locations of the X-line at the magnetopause based on the measurements from (left) CIS1 and (right) CIS4.The locations were determined by tracing the calculated distances to the X-line (Figure 5) back along the geomagnetic field line in the T96 model, starting at the position of the Cluster satellites in the magnetosphere.