Substorm onset dynamics in the magnetotail as derived from joint TC-1 and Cluster data analysis

This study investigates two substorm onset events with favorable constellations of spacecraft, TC-1 and Cluster, separated by several Earth radii. The substorms have been identiﬁed in both auroral regions. One is believed to be triggered by a northward turning of the interplanetary magnetic ﬁeld (IMF), while the other occurs under generally southward IMF. For both events, Cluster was located tailward of TC-1, but observed the dipolarization at earlier time for one event while at later time for the other. The timing difference of dipolarization at different positions could be explained by the earthward or tailward propagation of the ﬁeld disturbances in the radial direction. And the earthward dipolarization front was found in one case to bounce back and forth at TC-1. The earthward propagation was accompanied by a fast earthward plasma ﬂow for the 21 September 2005 event. The event analysis suggests that substorms can be quite different from case to case due to differences in the solar wind condition and magnetotail conﬁguration.


Introduction
A magnetospheric substorm is a complicated phenomenon that is not fully understood.One of the most controversial topics concerns the exact physical conditions that initiate a substorm.A number of models have reported different onset source locations in the magnetotail.For example, the cross-tail Copy right c The Society of Geomagnetism and Earth, Planetary and Space Sciences (SGEPSS); The Seismological Society of Japan; The Volcanological Society of Japan; The Geodetic Society of Japan; The Japanese Society for Planetary Sciences.current instability (CCI) model regards the near Earth region (6-10R E ) as the initial region (Lui, 1996).
It postulates that plasma instabilities generate a cross-tail current disruption and a divergence into the ionosphere via field-aligned currents (FACs) to form the substorm current wedge (SCW).After that, a rarefaction wave is launched tailward, which makes the midtail magnetic configuration more stretched, thus more favorable for reconnection to occur.Therefore, reconnection is started after the substorm onset.Alternatively, the near-Earth neutral line model (NENL) predicts that the substorm is initiated at a further distance down-tail (>20R E ), where reconnection takes place (Baker et al., 1996).The formation of the SCW and the associated auroral intensifications are considered as secondary effects of the reconnection.The cause of the SCW can be explained in terms of the braking of the fast earthward bursty bulk flows (Shiokawa et al., 1998) and the pressure gradient driven current induced by the flow braking (Birn et al., 1999).The CCI and NENL models can be distinguished by comparing the time at which reconnection is observed in the tail, utilizing in situ satellite data, with the time when the substorm onset is detected at near-Earth locations.If reconnection is observed first far down-tail, the NENL model is more appropriate (Liou et al., 2002).
During a substorm, dramatic changes take place in the Earth's magnetic field, the magnetospheric and ionospheric currents, and auroral displays, which can be revealed by appropriate measurement techniques.During the past decades, great advances have been achieved in observational techniques for magnetospheric physics.The Double Star mission, TC-1 and TC-2, China's first satellite mission to study the magnetosphere (Liu et al., 2005), combined with the four European Cluster satellites, set up an excellent constellation for exploring Geospace.The magnetic local time of the Double Star at apogee is identical to that of the Cluster spacecraft.In a devoted study, Nakamura et al. (2005) used simultaneous observations of Cluster and Double Star to investigate the relationship between the dipolarization and fast flows.By using multipoint analysis techniques, they determined that the dipolarization front was mainly propagating dawnward.The role of the midtail fast flows in the dissipation process turned out to be quite different, depending on the conditions of the interplanetary magnetic field (IMF).However, detailed ionospheric signatures related to this propagation of the disturbance were not presented.Wang et al. (2006) reported a typical substorm onset event with a fortunate constellation of Cluster and Double Star near the current sheet, which could be one "textbook" example of a substorm due to its distinct character.Both ionosphere and magnetotail observations have shown a clear dawnward propagation of the substorm current wedge.These previous studies have exhibited a current wedge with a large radial extent, which may be another class of substorm that is different from the CCI model and the NENL model.More events need to be found to confirm the propagation of the substorm disturbances.
However, due to the spatially localized nature of substorms in the magnetotail (Nakamura et al., 2005), the number of events with Cluster/TC-1 conjunctions of observations of substorm onset features are quite limited.
In this paper, we present two substorm onset events when both Cluster and TC-1 observed the field dipolarization but at different radial positions.Coordinated observations from geosynchronous satellites, in addition to ground-based magnetometer networks, complement each other quite suitably in the auroral and magnetotail regions where substorm processes are effective.

The 28 September 2004 event
The first substorm onset took place at 22:07:46 UT on 28 September 2004 as identified from WIC observations on the IMAGE satellite (Frey and Mende, 2007).About 35 minutes before the onset, a northward turning of the IMF arrived at the magnetopause (the solar wind data are measured by the ACE satellite, which has been time shifted to the magnetopause with the minimum variance method outlined in Weimer et al. (2003), which might be the interplanetary trigger for the substorm (Lyons, 1996).During the substorm, the magnetic activity was quiet, with a minimum D ST index of -6 nT, a maximum Kp of 1+, and a minimum AL of -246 nT.
Figure 2 shows three auroral images of the south polar region from IMAGE FUV-WIC in the frame of MLat and MLT.The time resolution of the aurora observations is 2 minutes.It can be seen that a prominent auroral brightening occurred at ∼22:07:46 UT (22:10 MLT, -68.48 • MLat).In the subsequent tens of minutes, the aurora expanded both azimuthally and poleward.These are typical auroral substorm features.
Ground magnetometer recordings can also exhibit typical features of a substorm onset (Lühr et al., 1998).The 1 minute averaged geomagnetic X, Y, and Z components measured by five magnetometers during the substorm are shown in Figure 3.The background magnetic field during quiet time has been subtracted by using monthly median values.The positive variations of the X component at LER seemed to begin at 22:06 UT, which was the onset time in the north.This onset time is within the uncertain range of the IMAGE onset (22:05:46 -22:07:46 UT).Therefore, we can assume that the substorm occurred almost simultaneously in both hemispheres.The figure also shows a sharp drop in the X component (positive magnetic northward) at the station SCO (71.44 • MLat, 22:19 MLT) after the aurora breakup.
This indicates an intense westward electrojet.At LRV (64.8 • MLat, 21:57 MLT), after a shortlived positive perturbation in the X component, a major negative magnetic deflection was observed.This is indicative of the passage of a westward travelling surge followed by a westward electrojet.In the surge head an upward FAC is expected (e.g.Kirkwood et al., 1988;Lühr et al., 1998).All this is consistent with the previous CHAMP magnetic field studies (Wang et al., 2005).It can be seen that there was a second sharp drop of the X component after 22:30 UT, which was first registered by the AMK station, followed by the SCO station.The Z component deflection (positive downward) was positive at stations north of LRV and negative south of it, implying an average location of the westward electrojet slightly north of LRV.The IMF B y can cause a hemispheric asymmetry of the MLT of the substorm onset (e.g.Wang et al., 2007).During the period of the onset we had IMF B y ∼-4 nT.This negative B y will cause the onset in the north to shift to later local time than in the south.Unfortunately, with the limited number of ground stations it is difficult to determine the exact MLT and MLat locations of the north onset.However, the large negative X drop indicates that the center of the westward electrojet is very close to SCO (22:20 MLT), which is a little later than in the south as observed by the IMAGE FUV (22:10 MLT).
Corresponding to this magnetic field activity on the ground caused by a westward electrojet, dispersionless injections at geostationary orbit were observed by the LANL satellites.Figure 4 shows the energetic electron flux data with 10 seconds time resolution from LANL 01A, located in the premidnight sector (∼22:30 MLT) at the time of the auroral breakup, and LANL 02A, located postmidnight (∼02:50 MLT) at the time of the auroral breakup.LANL 01A detected a significant electron injection at ∼22:11 UT, while LANL 02A detected it about 1 minute later.Another significant electron injection occurred after 22:30 UT as observed subsequently by these two satellites.Thus, both magnetic ground stations and geosynchronous satellites recorded two onsets of substorm intensification.
Figure 5 shows the magnetotail data in the GSM coordinate system from TC-1 and Cluster between 21:30 and 23:00 UT.The time resolution of the TC-1 is 6 seconds and that of the Cluster is 10 seconds.
TC-1 was below the current sheet, as can be determined from the negative B x , while the Cluster satellites were above the current sheet.For a better localization of the dipolarization of the magnetic field, the elevation angle is shown in Figure 6.Around 22:06 UT (2 minutes before the auroral breakup) Cluster observed a clear magnetic dipolarization.There were some wave like oscillations in the TC-1 magnetic field around 22:04 UT, which can be identified as the onset time of the dipolarization at TC-1.
Around onset, no significant plasma flow was detected (see Figure 5 bottom).However, for the general features of the field dipolarization, there was a rather similar overall variation of the elevation angle at TC-1 and Cluster.A cross-correlation analysis was performed over the period from 22:00 to 22:30 UT in order to determine the lag time between TC-1 and Cluster.Figure 7 shows the correlation coefficients as a function of lag time between TC-1 and the four Cluster satellites.The lag time indicates that on average, for the overall dipolarization feature, TC-1 responded about 2 minutes delayed with respect to Cluster.It should be noticed that both Cluster and TC-1 observed another field dipolarization shortly after 22:30 UT (see Figure 6).It is clear in this case that TC-1 observed the onset of the disturbance after Cluster.The signature at TC-1 was intermittent.This means that the dipolarization front was first moving earthward past Cluster and TC-1, then retreated tailward for a while past TC-1 and finally moved earthward again where it retained.This feature can be confirmed by the B y variations, which will be discussed further in Section 3. From the above analysis the time of the first initialization can be determined.The start of this initial substorm activity was observed first at TC-1, then Cluster, then ground magnetometer, auroral imagers, then finally at LANL.Dipolarization at TC-1 occurred earlier than at Cluster indicating that the substorm is initiated in the near-earth region.The fact that TC-1 observed the overall dipolarization features later than Cluster indicates that after the near-earth initiation the far tail reconnection occurred, which propagated the large scale disturbances earthward, from Cluster to TC-1.The time line of activity for this substorm around 22:07:46 UT is listed in Table 1.

The 21 September 2005 event
The second typical substorm onset presented here took place at 14:05:55 UT (01:42 MLT, -67.6 • MLat) on 21 September 2005 as identified by IMAGE WIC auroral observations (Frey and Mende, 2007).One hour before the substorm onset, IMF B z had an almost constant magnitude of -2 nT.
Positive IMF B x and B y underwent minor variations (less than 2 nT).The solar wind dynamic pressure stayed roughly constant.During the substorm the magnetic activity was normal, with a minimum D ST index of -27 nT, a maximum Kp of 2, and a minimum AL of -268 nT. Figure 9 shows three auroral images just 2 minutes before, at, and after the substorm onset.The auroral breakup was reported to occur at 14:05:55 UT in the postmidnight sector.It should be noted that there was another auroral brightening in the premidnight sector, which originated at 13:30 UT (auroral images are not shown).This premidnight auroral breakup diminished around 14:08 UT.At the time of the auroral breakup, the three magnetometers shown in Figure 9 were about 2 hours away from the conjugate footprint region of the auroral breakup.During the period of the onset, the IMF B y is ∼4 nT.The positive B y will cause the onset in the north to shift to earlier local time than in the south, thus bringing it closer to the magnetometers.The magnetic variations of the X, Y, and Z components, with 1 minute time resolution throughout the substorm, are shown in Figure 10.Both CHD and TIX recorded a drop in the X component, which started around 13:30 UT.This was associated with the premidnight auroral breakup.Following that there was a two step drop.The first slow drop was still associated with the premidnight substorm auroral breakup.The second rapid drop with much larger magnetic deflections (14:22 UT at CHD and 14:25 UT at TIX), was associated with the substorm electrojet under study.But the onset times of the ground magnetic disturbance associated with the substorm around 14:06 UT were not clear due to the overlapping substorm activity.A remarkable feature in the Y component at all three stations was a sinusoidal variation (see Figure 10).It drifted westward and equatorward over the stations.We can determine a westward propagation velocity for the current structures of ∼0.08 • /s (1.28 km/s) from the signal delay between TIX and CHD using a cross correlation analysis.We can also determine the equatorward shift velocity to be ∼0.04 • /s (0.64 km/s) from CHD and ZYK.The Z component deflection (positive downwards) was negative at all the stations implying a location of the electrojet polewards of the magnetometer stations.The observations clearly show that this substorm also occurred in both hemispheres.Corresponding to the recorded auroral and ground based signatures, a small but significant electron injection was observed by LANL 1994-084 at 14:06 UT in the postmidnight sector around 03:00 MLT (see Figure 11).But no electron injections were observed associated with the premidnight auroral breakup around 13:30 UT.
Figure 12 shows magnetotail data in the GSM coordinate system from TC-1 and Cluster between 13:30 and 15:30 UT.TC-1 and the Cluster satellites are above the current sheet at the start of the time period, as could be deduced from the positive B x component.Details of the dipolarization are presented in Figure 13, where elevation angles of the magnetic field are shown.At 14:04 UT, TC-1 observed a clear magnetic dipolarization.The Cluster spacecraft measured a B x decrease ∼15 minutes before the auroral breakup, while they observed only small fluctuations in the B z component (see Figure 12).This is an indication that the quartet approached and partly crossed the cross-tail current sheet.B z began to increase about 7 minutes before the auroral breakup, although somewhat intermittent.When looking at the elevation angle in Figure 13, it is possible to estimate the onset of field dipolarization at Cluster to be 13:59 UT, which was 7 minutes before the auroral breakup.The field dipolarization was found to be associated with fast earthward busty bulk flows (see Figure 12 bottom).Neither TC-1 nor Cluster observed field disturbances associated with the premidnight auroral breakup around 13:30 UT, indicating a psuedobreakup event.From the above analysis, the time sequence of the substorm activity around 14:06 UT was first Cluster, then TC-1 and LANL.As mentioned earlier, the onset time of ground based measurements is not clear due to the repeated auroral breakups.The time line of activity for this substorm is also listed in Table 1.

Discussion and Conclusion
The observed characteristics in both the auroral zone and magnetotail region during two well developed substorms have been presented.The magnetic fields, the geosynchronous plasma distribution, ionospheric currents, and auroral displays have shown rapid changes.TC-1 and Cluster were not only separated in the radial direction but also in the longitudinal direction, i.e., Cluster was eastward (∼2 R E or 25 • in long.) for the 28 September 2004 event, while westward (∼7 R E or 33 • in long.) for the 21 September 2005 event.For the 2004 substorm event, TC-1 observed the onset of the field dipolarization earlier than Cluster, but observed the general dipolarization feature later than Cluster.For the 2005 substorm events, Cluster observed the field dipolarization earlier than TC-1.
For the 28 September 2004 event, TC-1 observed some field disturbances ∼2 minutes earlier than Cluster, indicating that current instabilities were initiated in the inner region then propagating tailward.
For the general field dipolarization feature, TC-1 is clearly ∼2 minutes later than Cluster, indicating an earthward expansion of the large scale field disturbances.This large scale field disturbance might be caused by reconnection in the far tail, which occurred later than the current disruption in the inner region.To understand the front propagation of field disturbances in a relatively simple way, satellites have been projected to the ground for comparison, as shown in Figure 1.If the time sequence was dominated by the radial distance, starting from the inner earth region, the expected time sequence would be TC-1 to Cluster.The subsequent far-tail reconnection propagated the disturbances further earthward, that is, from Cluster to TC-1 to LANL-01A.Therefore, we may conclude that a temporary tailward propagation of the field disturbance front can be observed, but this is embedded in an overall earthward expanding dipolarization front.
For the 21 September 2005 event, a timing difference of 7 minutes in the magnetic dipolarization signatures at TC-1 and Cluster could also be associated with the earthward propagation of the field disturbances.Figure 8 shows the distribution of satellites and magnetometers.The succession of activation was Cluster -TC1 -LANL.On the first glimpse, it seemed that either dawnward or earthward relation between the ground-based recordings at Abisko (65.1 • MLat, 00:09 MLT) and the TC-1 B y component on an expanded scale At 22:37 UT, the dipolarization front retreated tailward past TC-1.
The positive deflection of B y was consistent with an upward FAC sheet tailward of the spacecraft.
Only 2 minutes later, the dipolarization front swept back earthward, accompanied by a steep negative gradient in B y .This indicated a final earthward passage of the upward FAC.In summary, TC-1 had the fortunate opportunity to observe an earthward propagating dipolarization front that reversed its direction for a short while before it finally passed earthward.We may conclude that temporary tailward propagations of the current disruption front can be observed, but these are still embedded in an overall earthward expansion.As seen in Figure 14, on 21 September 2005, the Cluster B y component went negative from 14:06 UT onward, suggesting an earthward passage of a downward FAC.The TC-1 B y followed the trend at Cluster with a time delay, again implying the earthward passage of the FAC sheet.
After 14:25 UT the B y component at TC-1 was steeply rising.This may be interpreted as a duskward propagation of the current wedge center line passed TC-1.
Our initial aim was to make use of the multi-spacecraft configuration to find support for any of the competing substorm onset models.This could not be achieved conclusively by these two case studies.However, some preferences for one of the models may be deduced.One event showed that TC-1 observed field turbulence earlier than the Cluster quartet, thus, supporting the CCI model.However, the other event showed that TC-1 observed field dipolarization later than the Cluster quartet, thus, supporting the NENL model.Both events suggest the importance of the reconnection in the far tail with the large-scale activity propagating from the midtail to the inner magnetosphere.Even though this study focused mainly on dipolarization in the magnetotail, associated parameters such as plasma flow in the magnetotail were also considered in differentiating between the two substorm models (see Figure 5 and 12).For the 21 September 2005 event, the earthward expansion of the field dipolarization was found to be associated with fast earthward busty bulk flows (see Figure 12).This fast earthward flow together with a positive B z was believed to be related to an earthward moving flux rope type signature (Nakamura et al., 2005).However, no significant plasma flow was detected during the 28 September 2004 event (see Figure 5), which can be explained by the fact that the spacecrafts were not located in the central plasma sheet.To further test the casual relationship of substorm onsets more spacecrafts in radial constellation near the current sheet are needed.

Figure 8
Figure8shows the northern polar ionospheric footpaths of the Cluster and TC-1 satellites, the

Fig. 10 .
Fig. 10.Magnetograms from the 210 • Magnetic Meridian for the substorm under study.The vertical dashed blue line denotes the time of auroral breakup.

Fig. 12 .
Fig. 12. Magnetotail observations of TC-1 and Cluster.In the top three panels are shown Bx, Bz in GSM coordinates observed by TC-1, and the Cluster quartet.The bottom panel shows the velocity of hot ion observed by Cluster HIA.The vertical dashed blue line denotes the time of auroral breakup and green, ∼7 minutes ealier, the time of field dipolarization at Cluster.

Fig. 13 .
Fig. 13.Top panel shows the elevation angle magnetic field in GSM coordinates observed by TC-1, below by Cluster quartet.The vertical dashed blue line denotes the time of auroral breakup the red line, start of dipolarization at TC-1, and green, ∼7 minutes ealier, the time of field dipolarization at Cluster. 22

Fig. 15 .
Fig. 15.Ground and space-based observations of a Pi2 pulsation.The upper curve shows the magnetic field variations of the northward component at Abisko and the lower the By component variations at TC-1.

Table 1 .
Time of substorm disturbances Fig. 14.TC-1 and Cluster observation of the By components in GSM coordinates on 28 September 2004 (top) and on 21 September 2005 (bottom).The vertical dashed blue line denotes the auroral breakup time.