In the previous section we have presented measurements from the ISTP/GGS Polar and Wind satellites and the SuperDARN radar network. Here we compare them with the predictive features of models for the development of theta auroras described here and in the literature. We first consider the November 5 event, which is the only one having timely data from both VIS and MFI, and then the September 6 event which also has an IMF history similar to that suggested by Newell and Meng [1995]. The November 1 and May 7 events as well as two events reported by Huang et al. [1989] and by Craven et al. [1991] and most importantly the five events found by Cumnock et al. [1997] provide counter examples to the model of Newell and Meng but support our model.
The IMF for November 1 event shown in Figure 4 turned southward at 14:34 CUT after a four-hour northward period, then returned to northward at 15:19 CUT. The polar cap arc first appeared at 14:46 UT, ~ 12 mins after the southward IMF reached the dayside magnetopause. IMF By was positive during the interval of theta auroral configuration. This IMF pattern is similar to requirements for the model by Newell and Meng [1995]. It also agrees with predictions of our model, described in section 2, for the formation of theta aurora with regard to the polarity of IMF By and the temporal variations of Bz (see Figure 2). Moreover, the polar cap arc started on the dawn side auroral oval and moved across the polar cap toward the dusk for positive By, as predicted by the model. In the segments of the transpolar arcs traversed by Polar from 15:06 to 15:17 UT and 15:41 to 15:51 UT, the plasmas measured by Hydra, CEPPAD, and TIMAS show spectral characteristics similar to those detected in the poleward part of the auroral oval (08:25-09:15 UT). Both energies and energy fluxes of particles in the transpolar arc are much higher than those in the adjacent polar rain. These suggest that the transpolar arc regions and the poleward part of nightside auroral oval very likely map to the same region in the tail, i.e., BPS, and are on closed magnetic field lines as suggested by Frank et al. [1982, 1986] and Peterson and Shelley [1984]. In both regions, electron distributions measured by Hydra show counter streaming beams at energies of several tens of eV (Figures 5a, 5c, and 5d), indicating a closed magnetic topology, with the smallest gyroradius particle measurements to date. This is consistent with our model's prediction regarding the plasma source for the transpolar arcs. The energies of the counter-streaming electron beams in the first theta auroral arc is higher than those in the second, and may be why the first arc is brighter than the second.
The cross polar cap potential is anticorrelated with the IMF measurements when the latter are shifted by ~ 35 mins as shown in Figure 6. This time lag is estimated by equating the beginning of the polar cap potential increase with the detection of southward IMF at the spacecraft. The delay is somewhat longer than the 26 min delay for propagation of the IMF from Wind to the magnetopause. This is due to the Alfvén travel time between the magnetopause and the ionosphere and the response time of the ionosphere on a global scale. In this figure, the first period of small and irregular electric fields (before 14:43 UT) is associated with the first northward IMF period; the interval of stronger and more regular polar cap potential corresponds to the southward IMF conditions; the last interval (after 15:21 UT) corresponds to the second northward IMF period. The calculated ionospheric plasma convection evolved as the IMF changed, and correlates with the evolution of the theta auroral arc. These correlations are also predicted by the model as consequences of the dayside merging variations. The plasma convection inferred from EFI and MFE measurements was sunward during the two transpolar arc encounters and antisunward in the adjacent polar cap regions. This too is consistent with earlier reports [e.g., Frank et al., 1986] and our model that the source for the transpolar arcs is in the BPS. We note that the transpolar arc traversed by Polar was very narrow, about 0.5° wide in latitude and the associated regions in the ionosphere are even more narrower. The SuperDARN radar network can not resolve such small spatial features. Thus, it could not detect sunward convection in the theta auroral arc (Figure 7).
For the September 6 event, the IMF Bz pattern shown in Figure 9 is similar to that of the November 5 event. According to our model, the transpolar arc should have started ~ 06:52 UT after IMF turned southward at the subsolar magnetopause (06:42 CUT) plus the ionosphere response time. However, VIS images are not available from 06:00 to 07:00 UT to verify this prediction. After 07:12 UT, the theta aurora observed by VIS had already developed. There was about one hour interval of southward Bz and large negative By (08:25-09:19 CUT) during which the transpolar arc moved dawnward across the polar cap as predicted by the model. Near the end of the interval, the arc almost vanished into the dawn side of the auroral oval. It reappeared after Bz turned positive at 09:19 CUT. This behavior is consistent with the newly open flux being less effectively driven by the merging line under northward than the southward IMF conditions. Similar to the November 5 event, plasmas measurements from three particle instruments show similarities in both energies and energy flux in the transpolar arcs and the poleward part of the nightside auroral zone. This suggests that these regions had the same BPS source region in the magnetotail. Moreover, ion composition measurements from TIMAS show that low energy ions of four species H+, O+, He+ and He++) were flowing away from the Earth in the transpolar arc regions. The dominant species of precipitating ions are H+ and O+. These are also consistent with theta auroras being on closed field lines, as suggested by early observations [Peterson and Shelley, 1984] and our model.
Unlike the previous two events, the IMF pattern during the November 1 event, shown in Figure 10, was persistently northward for more than 10 hours, except three brief southward excursions of less than 5 mins duration. We first examine the possibility of the observed theta aurora being triggered by the Bz changes. During the negative Bz interval beginning at 23:55 CUT, By was almost zero. Therefore, the motion of newly merged open flux was mainly determined by Bz. The newly open flux would move tailward and simply join the existing open flux. Thus, this southward turning of the IMF could not have triggered the transpolar arc. For the 23:16 CUT case, Bz was negative for only 2 mins, and thus could not have initiated the theta aurora. The interval starting at 22:17 CUT is even less likely. Since these southward turnings are too brief, this event can not be explained by the Newell and Meng model. Nevertheless, the theta arc could have been triggered by a change in IMF By. One possible candidate is the negative turning of By at 22:25 CUT. This is like the case illustrated in Figure 3 and discussed in the model section. Considering the solar wind propagation and the time required for open fluxes to penetrate to the region previously occupied by the auroral oval, the theta aurora should have begun to develop at ~ 22:35 UT, about two hours before the first available VIS image (00:46 UT).
For the May 7 event, IMF Bz turned southward at 23:15 CUT and stayed negative for more than three hours. Based on our model, the theta aurora should appear at about 23:25 UT. However, no auroral image is available to verify this prediction. Nevertheless, this event shows a theta aurora did occur during the period of persistently southward IMF conditions, similar to the event reported in Figure 1 of Craven et al. [1991]. These events cannot be explained by the Newell and Meng model, either. Although Bz was continuously southward, the theta aurora persisted to about 01:10 UT. According to our model, it would continue because By changed polarity at 00:18 CUT before the arc was washed away by newly opened flux driven by the dayside merging associated with the southward IMF. This effect is similar to the situation described in Figure 3, but not considered by Newell and Meng [1995].
In the past, studies of theta auroras have been hampered by a lack of simultaneous observations of the polar cap and solar wind IMF. However, in the literature there are two reports of theta auroral event along with simultaneous IMF measurements. In the event on March 25, 1982, Huang et al. [1989] observed a theta aurora beginning at 05:02 UT. The onset of the transpolar arc must have been about 10 mins earlier since the observed transpolar arc had already developed. This puts the initiation time at ~ 04:52 UT. As shown in Figure 1 of Huang et al. [1989], IMF Bz from ISEE 3 turned negative for first time at ~ 05:10 CUT. This southward turning could not have triggered the event. However, there was a change of polarity and magnitude in By at ~ 04:30 CUT, after which it oscillated rapidly. We speculate this By jump activated the transpolar arc.
The other theta auroral event with simultaneous IMF measurements was reported by Craven et al. [1991]. According to their Figures 1 and 5, a polar arc was observed by DE 1 at 17:42 UT. The first southward turning of the IMF, following two hours of northward Bz, was at 17:48 UT (~ 17:54 CUT). This is at least ~ 12 mins after the polar arc had developed. The southward turning of the IMF could not have triggered the observed theta auroral arc. We note that By reversed from negative to positive polarity at about 17:06 UT while Bz was still positive. This jump very likely triggered the polar arc that later evolved into a transpolar arc. See the second example (Figure 3) of our model described in section 2.
More direct, convincing evidence for the predictions of our model has been provided by Cumnock et al. [1997]. Using the imaging data from DE 1 and IMF measurements from ISEE 3, they found five events of the theta aurora that occurred in the northern hemisphere after IMF By changed sign while Bz remained positive. In two events (days 82/021 and 82/028), the theta auroral arc started on the duskside auroral oval for By changing from positive to negative and then moved dawnward across the polar cap that is consistent with our model (Figure 3). The evolution of arc for the other three events (days 82/021, 81/329, and 81/363) was in the opposite direction as predicted by our model for By changing reversely. However, the DE 1 auroral images have a rather low time resolution (12 mins) that can not be used to assess our model.
The November 1 event is a possible example of a steady state theta aurora during northward IMF conditions on closed, sunward convecting field lines [e.g., Reiff and Burch, 1985; Kan and Burke, 1985; Toffoletto and Hill, 1990; Burch et al., 1992]. However, the observed dawn-dusk motions of the polar arcs as shown in the November 5 and September 6 events and other previously reported events [e.g., Frank et al., 1985, 1986; Craven et al., 1991] can not be explained by steady-state models. Nor can the two events showing theta auroras during extented periods of southward IMF be explained by steady-state models. Furthermore, for the four events presented here, the VIS images did not show any theta auroral arcs during the prolonged northward IMF interval before either Bz or By changed. Our time-dependent model of theta auroras predicts their observed dawn-dusk motions. Because the dayside merging driving the convection of newly opened flux erodes the magnetopause, regions in the ionosphere that were previously closed and mapped to the plasma sheet became open and a strip of closed field lines peeled off from the auroral oval, as illustrated in Figures 2 and 3. The detached closed regions then move from dawn to dusk for positive By in the northern hemisphere, and vice versa for negative By. The direction is reversed in the southern hemisphere. In addition, the model predicts the polar arcs that later evolve into the transpolar arcs, are generated at the dawn (dusk) sector in the northern (southern) hemisphere for positive By, and vice versa for negative By. Therefore, when IMF Bx ~ 0 and dayside merging occurs in both hemispheres, simultaneous polar cap arcs exist in two hemispheres having different source regions on different magnetic field lines and show opposite dawn-dusk motions. The simultaneous theta auroras in both hemispheres reported by Craven et al. [1991] showed these effects. Theta auroras occurring during persistently southward IMF conditions are also predicted by our model as long as two regions of open field lines remain separated. This can be achieved when IMF By occasionally changes polarity, as in the May 7 event and the event reported by Craven et al. [1991]. We assume that merging takes place continuously or may stop when IMF conditions are not favorable. The merging site jumps when the IMF orientation changes. Excluding the time intervals of IMF changes, the configuration described in the model is rather quasi-stationary.
It is important that |By| 
|Bz| for ~ 10 mins after the IMF turning.
Otherwise, the newly open region would expand the existing open region and the splitting of the previously closed flux region would not occur.
The IMF measurements from Wind/MFI for the four theta auroral events presented here and the two reported events with simultaneous IMF observations, all satisfied this condition.
Furthermore, it is essential to have a contracted polar cap before initiating the theta aurora to facilitate a closed region peeling off from the flank of the auroral oval.
Such configurations are usually achieved after prolonged periods of northward IMF Bz.
After the IMF changes orientation, the merging line jump on the dayside magnetopause has to be significant enough so that newly opened region does not simply join the previously open region.
The merging line's location depends on IMF conditions, the solar wind speed and plasma pressure.
Therefore, the changes of IMF Bz or By components have to be large enough to change the merging line's location significantly.
Moreover, as suggested by our model, it would be more effective for bifurcating polar cap if the IMF By and Bz change polarity together after a prolonged period of northward IMF.
There are mainly three factors that determine whether a visible transpolar arc is present. They are plasma convection, electron distributions in the source region and the parallel acceleration. The convection, determined by the dayside merging, decides the magnetic topology and how regions in the high latitude ionosphere map to the magnetotail. Even if the magnetic topology is favorable for visible transpolar arcs, the energy fluxes of the electrons are the ultimate determinant of the visibility of the theta bars. So far no one knows what are the causes of the large-scale potential structure associated with the acceleration of precipitating electrons. Although the convection is rather steady, as the potential structure changes and electrons are not accelerated effectively to the ionosphere, the transpolar arcs would disappear. Also if electron distributions in the source region change so that precipitating fluxes are low, the ionospheric excitations would be subvisual. Therefore, it happens sometimes that auroral measurements are not inconsistent with plasma measurements when particle fluxes are present but diffuse, no perceptible theta bar can be found. Because the polar cap arcs are on closed field lines, with enough precipitating keV electrons, polar cap arcs on the same field lines may be present simultaneously in both hemispheres. Multiple-satellite observations of simultaneous or nearly simultaneous presence of the arcs in the same local time sector of two hemispheres by Gorney et al. [1986], Mizera et al. [1987], and Obara et al. [1988] are such examples.
Please send questions, comments, or suggestions about the paper to:
Shen-Wu Chang
Department of Physics and Astronomy, The University of Iowa, Iowa City, IA 52242
Phone:(319)335-3828; Fax:(319)335-1753;
swc@space-theory.physics.uiowa.edu