Locations of night side precipitation boundaries relative to R2 and R1 currents

Transcription

1 JOURNAL OF GEOPHYSICAL RESEARCH, VOL. 115,, doi: /2010ja015444, 2010 Locations of night side precipitation boundaries relative to R2 and R1 currents S. Ohtani, 1 S. Wing, 1 P. T. Newell, 1 and T. Higuchi 2 Received 9 March 2010; revised 30 April 2010; accepted 7 June 2010; published 15 October [1] The present study statistically examines the location of various precipitation boundaries introduced by Newell et al. (1996) relative to R2 and R1 currents on the night side. Results are summarized as follows: (1) The electron and ion zero energy boundaries, b1e and b1i, are located mostly inside the R2 current at dusk to midnight and near the equatorward boundary of the R2 current or farther equatorward at midnight to dawn. (2) The maximum energy flux of ion precipitation (b2i) occurs inside the R2 current irrespective of magnetic local time. (3) The occurrence distributions of the most equatorward (b3a) and poleward (b3b) electron acceleration events indicate that monoenergetic electron precipitation is mostly confined in the upward R1 current at dusk tomidnight, whereas at midnight to dawn, it is more widely distributed including the downward R1 current. (4) The transition between structured and unstructured electron precipitation (b4s) tends to occur around the R2/R1 demarcation, but its occurrence distribution has extending tails. (5) The distributions of the poleward boundaries of the electron and ion auroral ovals, b5e and b5i, are centered around the poleward boundary of the R1 current but have extending tails especially at midnight to dawn. Result 1 suggests that the overlap between the ring current and the plasmasphere is more significant at dusk than at dawn. Result 2 indicates that the b2i boundary can be used as an identifier of the R2 current. Results 3 5 suggest that the R1 current is more structured than the R2 current and that the field aligned current structure is more complex at midnight to dawn than at dusk to midnight. Citation: Ohtani, S., S. Wing, P. T. Newell, and T. Higuchi (2010), Locations of night side precipitation boundaries relative to R2 and R1 currents, J. Geophys. Res., 115,, doi: /2010ja Introduction [2] Field aligned currents (FACs) and particle precipitation are two fundamental elements of the magnetosphereionosphere (M I) coupling, and their global distributions have been studied for many years. The polar distribution of FACs consists of a pair of zonal current systems, which are known as region 1 (R1) and region 2 (R2) systems [Iijima and Potemra, 1976]. The R1 current flows downward and upward on the morning and evening sides, respectively, whereas the R2 current is located on the equatorward side of the R1 current and has the opposite polarity at a given local time. Particle precipitation has also been addressed in terms of two zonal regions, that is, CPS and BPS [Winningham et al., 1975]. CPS is characterized by unstructured ion and electron precipitation that extends to a high energy (>1 kev) range. BPS is located adjacently poleward of CPS and is characterized by structured and often mono energetic 1 Johns Hopkins University Applied Physics Laboratory, Laurel, Maryland, USA. 2 Institute of Statistical Mathematics, Tokyo, Japan. Copyright 2010 by the American Geophysical Union /10/2010JA electron precipitation. It is now generally accepted that CPS and BPS do not correspond one to one with the highaltitude central plasma sheet and plasma sheet boundary layer. [3] The size of the auroral oval significantly varies depending on various conditions such as geomagnetic activity [e.g., Feldstein and Starkov, 1967] and interplanetary magnetic field (IMF) orientations [e.g., Holzworth and Meng, 1984]. The polar distributions of the FAC [e.g., Weimer, 2001; Papitashvili et al., 2002; Anderson et al., 2008] and particle precipitation [Newell et al., 2004] constructed from large data sets also reveal clear dependence on the IMF. Therefore, it should be reasonable to expect that the spatial distributions of FAC and particle precipitation change following each other. In other words, we expect that the locations of various precipitation boundaries are organized by the locations of large scale FACs (and vice versa). [4] Here it is important to compare the locations of precipitation boundaries with those of FACs for each auroral oval crossing and then to statistically examine their distributions rather than to compare the statistical distributions of particle precipitation and FACs. The reason is twofold. First, we believe that the variability of precipitation boundaries relative to large scale FACs carries important 1of13

2 Figure 1. The number of DMSP crossings of two (red), three (blue), and four (green) FAC sheet structures, for which particle precipitation boundaries are identified, at least for electrons. The positive and negative numbers indicate those of orbits with the most equatorward current flowing into and out of the ionosphere, respectively. information. Second, the locations of precipitation boundaries relative to large scale FACs may be different for different FAC structures (this is indeed the case for the b2i boundary as will be shown later). Such relationships can be addressed only if we compare FACs and particle precipitation for individual orbits. [5] In the present study we statistically examine particle precipitation boundaries relative to the R2 and R1 currents. The locations of eight precipitation boundaries introduced by Newell et al. [1996] are compared for each auroral oval crossing with FAC boundaries identified by an automated procedure developed by Higuchi and Ohtani [2000a, 2000b]. A focus is placed on the nightside (18 magnetic local time, MLT < 06). Our accompanying paper, Wing et al. [2010], reports a similar study but focusing on the dayside with a different set of precipitation boundaries [e.g., Newell and Meng, 1988]. [6] In section 2, we describe magnetic field and particle measurements made by DMSP satellites, which we use in this study. We also explain the event selection procedure in section 2. In section 3 we examine two example events, one at premidnigt (MLT 22.4) and another one at early morning and address how electron and ion precipitation signatures differ in different (R2 or R1) FAC systems or on different sides of midnight. In section 4 we examine how the latitudes of the FAC and precipitation boundaries change as a function of MLT. Section 5 is the primary part of this study, in which we statistically examine the locations of the precipitation boundaries relative to the boundaries of R2 and R1 currents. We discuss the results in terms of magnetospheric convection, structure, and the structuredness of large scale FACs. We summarize the study in section Data Set and Event Selection [7] In the present study we use magnetic field and particle precipitation data acquired from DMSP F7 ( ), F12 ( ), F13 ( ), F14 ( ), F15 ( ), and F16 ( ) satellites. All DMSP satellites are on Sun synchronous orbits (F13 approximately in a dawn dusk orientation and others in prenoon premidnight orientations) at about 840 km in altitude with orbital inclinations of The orbital periods are approximately 100 min. [8] A triaxial fluxgate magnetometer is onboard each of those DMSP satellites, which measures vector magnetic fields with a time resolution of 1 s and a 1 bit resolution of 2 nt[rich et al., 1985]. The particle instrument package (SSJ4) uses curved plate electrostatic analyzers to measure ions and electrons from 32 to 30 kev in logarithmically spaced steps [Hardy et al., 1984]. One complete 19 point electron and ion spectrum is obtained each second. The satellites are three axis stabilized, and the detector apertures always point toward local zenith. Since the magnetic field is approximately vertical in the polar region, measured particles are precipitating into the ionosphere. [9] The low energy (30 ev to 1 kev) ion detector onboard some DMSP satellites suffered from degradation immediately or sometime after their launch, which affects the identification of two ion precipitation boundaries, b1i (the zero energy ion precipitation boundary; see section 5.1) and b5i (the poleward boundary of the ion auroral oval; see section 5.5). For those two boundaries we use data obtained only from DMSP F7 for its entire interval, F12 and F14 data throughout 1998 and 2001, respectively, for which the effect of degradation, if at all, is minimal. Another ion boundary, b2i (the point of the maximum flux of ions above 3 kev; see section 5.2), is determined only by measurements made by the high energy ion detector, and therefore, it is not affected by the degradation of the low energy ion detector. [10] We use FAC boundaries identified for each DMSP auroral oval crossing by an automated procedure developed by Higuchi and Ohtani [2000a, 2000b], which has been used for various statistical studies [Ohtani and Higuchi, 2000; Ohtani et al., 2000; 2005a, 2005b, 2009; Ueno et al., 2007; Nakano et al., 2009; Wing et al., 2010]. This procedure is basically the automation of the conventional way we visually examine a plot of satellite magnetic field data. It calculates the maximum variance component of horizontal magnetic perturbations for each crossing of auroral oval and fits line segments to its variation. The structure of magnetic variations is quantified by several parameters such as the ratio of the maximum to minimumvariance component (a) and the ratio of the residual of the line segment fitting to the magnetic change associated with the most intense FAC (R fit ); for more details of these parameters, see Higuchi and Ohtani [2000a, 2000b]. For selecting DMSP orbits, we use the same criteria for those parameters as Ohtani et al. [2009] used. For example, a =2 and R fit = 8%. [11] Figure 1 shows as a function of MLT the number of DMSP orbits for which both the structure of FACs and particle precipitation boundaries, at least electron boundaries, were identified. Here the MLT of each orbit is defined as that at the demarcation between the most equatorward and second equatorward FACs. The positive and negative numbers indicate those of orbits with the most equatorward current flowing into and out of the ionosphere, respectively. The different colors indicate different numbers of largescale FACs crossed along an orbit: red, blue, and green for 2of13

3 two, three, and four sheet structures, respectively. The automated procedure occasionally identifies only one sheet or more than four sheets along an orbit, but the number of such orbits is even smaller than that of four sheet structures. [12] Three features are clear. First, the MLT distribution of orbits is far from uniform. There are 16,000 and 19,000 orbits at 18 MLT < 20 and 20 MLT < 22, respectively, whereas there are only 1300 and 1100 orbits at 00 MLT < 02 and 02 MLT < 04, respectively. This is because the orbit of each DMSP satellite is Sun synchronous, and it covers only a limited range of MLT. Second, the polarity of the most equatorward current is predominantly downward (into the ionosphere) at 18 MLT < 22, whereas it is upward (out of the ionosphere) at 00 MLT < 06. The transition appears to take place in the premidnight sector, 22 MLT < 24. Finally, for the overwhelming majority (83%) of orbits, the FAC has two sheet structures. At 22 MLT < 24, three sheet structures are often observed, but they are still a minority (33%). The last two features are consistent with the statistical spatial distribution of R2 and R1 currents [Iijima and Potemra, 1976], and in the rest of this paper, we refer to the most equatorward and second equatorward FACs as R2 and R1 currents, respectively. 3. Example Events [13] In this section we examine two auroral oval crossings, one in the dusk to midnight sector and another in the midnight to dawn sector and address how electron and ion precipitation differs between the two events. As for precipitation boundaries, we will return to these example events when we examine their statistical distributions in section An Oval Crossing in the Dusk to Midnight Sector [14] Figures 2a shows magnetic field and precipitation measurements made by DMSP F7 on 12 March The satellite crossed the auroral oval from poleward to equatorward at MLT = The top panel shows the two horizontal magnetic components, the minimum variance (B L : blue) and maximum variance (B A : red) components as determined by the principal axis analysis (section 2). The former can be regarded as the latitudinal component (positive in the direction of the satellite velocity, therefore equatorward for this event) and the latter as the azimuthal component (positive in the direction of the vector product between the vertically downward axis and the satellite velocity, and therefore westward for this event). The variations of B L are significantly smaller than those of B A, indicating that the FAC distribution is locally well approximated by sheet structures. The red dashed line represents the result of the line segment fit (section 2), which traces the actual measurements very well. The satellite crossed two large scale FAC sheets. The FAC on the poleward side flows upward, whereas the one on the equatorward side, which is weaker and is distributed more widely in latitude, flows downward. The former can be identified as a R1 current and the latter as a R2 current. [15] The lower half of Figure 2a shows the energy flux and average energy of precipitating electrons (black dots, labeled on the left hand side) and ions (red dots, labeled on the right hand side) and the E t diagrams of precipitating electrons and ions. The vertical axis of the ion E t diagram is inverted. At first glance the characteristics of precipitation are drastically different between the R2 and R1 currents. Electron acceleration is most prominent in the R1 current, whereas in the R2 current, there is no clear sign of the fieldaligned acceleration of electrons, and the average energy as well as the energy flux is noticeably lower than in the R1 current. The electron precipitation terminates rather abruptly in the middle of the R2 current and so is the ion precipitation. Note, however, that the flux of the lowest energy channels extends slightly equatorward for both electrons and ions. The intense precipitation of energetic ( 10 kev) ions is roughly confined between this termination and the R2/R1 demarcation. In the R1 current, in contrast, the ion precipitation is significantly weaker An Oval Crossing in the Midnight to Dawn Sector [16] Figures 2b show an auroral oval crossing on 24 July 1997 in the same format as Figure 2a. The DMSP F13 satellite crossed the auroral oval from equatorward to poleward in the Southern Hemisphere slightly before 0400 MLT. For this event B A and B L are positive westward and poleward, respectively. The change of B L is far less significant than that of B A, indicating the sheet structure of FACs. The polarities of the lower (R2) and higher latitude (R1) FACs are upward and downward, respectively, as expected for the dawnside R2 R1 pair. [17] In the R1 current, electron precipitation is noticeably weaker than in the R2 current, whereas the precipitation of energetic ions is as intense as in the R2 current. In fact, the ion precipitation appears to be continuous, but with the center energy becoming higher, from the poleward edge of the R1 current till about three fourth of the R2 current width, where the overall energy of ion precipitation suddenly becomes lower. The electron precipitation is energetic (>1 kev) but not very structured in the R2 current except near its poleward boundary. The energetic electron precipitation extends far equatorward of the R2 current and fades out. A close inspection of the B A plot, however, suggests that a very weak FAC extends farther equatorward of the R2 current (as identified by the line segment fit), which well overlaps the region of this energetic electron precipitation. Energetic but unstructured electron precipitation characterizes the midnight to dawn sector, which presumably corresponds to diffuse aurora. [18] It should be useful to compare the precipitation signatures of the present event with those of the 12 March 1984 event. First, the characteristics of both ion and electron precipitation in the R1 current are significantly different between the two events. In the downward R1 current of the present event, the ion precipitation was intense and energetic, whereas the electron precipitation was rather weak. In the 12 March 1984 event, in contrast, the upward R1 current was collocated with mono energetic electron precipitation along with very weak ion precipitation. These differences can be explained at least qualitatively in terms of fieldaligned electric fields with the same polarity as the R1 current. The electron precipitation was much more intense and energetic in the upward R2 current of the present event than in the downward R2 current of the 12 March 1984 event. However, the electron precipitation in the present event is not mono energetic, and therefore, it is not likely 3of13

4 Figure 2. (a) An auroral oval crossing in the dusk to midnight sector observed by DMSP F7 on 12 March (b) An auroral oval crossing in the midnight to dawn sector observed by DMSP F13 on 29 July See text for details. 4of13

5 that the field aligned potential difference is the primary cause of the difference between the two events. The ion precipitation in the R2 current is intense and energetic in both events. It is likely that both ion and electron precipitation in the R2 current largely reflects the magnetospheric source population. [19] In closing of this section, we would like to note that a weak FAC often extends farther equatorward of the R2 current as we found for the 24 July Therefore, the equatorward boundary of the R2 current identified by the line segment fit for such events may be regarded as the most poleward possible location of this boundary. In contrast, the identification of the poleward boundary of the R1 current and the demarcation of the R2 and R1 currents is far less unambiguous in general. 4. MLT Dependence [20] In the present study, unless we specify, we focus on events in which DMSP observed a two sheet structure, a pair of R2 and R1 currents, with their polarities consistent with their statistical pattern [Iijima and Potemra, 1976]. That is, for an auroral oval crossing to be selected, we require that the polarities of the observed R2 and R1 currents are downward and upward, respectively, before midnight, and upward and downward after midnight. The MLT of the demarcation between the morning side and evening side R2 and R1 currents moves, for example, as a function of the IMF B Y component [Weimer, 2001; Papitashvili et al., 2002; Anderson et al., 2008]. However, the adoption of this simple criterion can be justified for the following two reasons. First, the events excluded by this criterion are a very small fraction of our event set except for the premidnight (22 MLT < 24) sector (see Figure 1). Second, in the premidnight sector, DMSP observed a number of two sheet structure events with the opposite FAC polarities. However, as we will address in section 5.2, some of such events can be described as a three sheet structure without a R2 current rather than a pair of R2 and R1 currents with the opposite polarities, and the result of the present study would become obscured if we include such events. [21] Figure 3a shows the absolute values of the magnetic latitudes, MLat, of the equatorward boundary of R2 currents (blue), the demarcation between R2 and R1 currents (red), and the poleward boundary of R1 currents (green) against MLT. The median of MLat of each FAC boundary is plotted every 2 h in MLT, and the error bars show ranges between the 16 and 84 percentile points (±1 standard deviation from the median value) for giving the idea of the scatter of data points. The expected errors are actually the standard deviations divided by the square root of the number of events, and because of an extremely large number of events (Figure 1), they are significantly smaller than the error bars shown in the figure and are very often not discernible from the plotted lines. [22] The equatorward boundary of R2 currents shifts poleward, but only by 2, from dusk to dawn. The latitude of the poleward boundary of R1 currents is lowest around midnight and becomes higher away from midnight. These features can also be found in the initial study of R2 and R1 currents [Iijima and Potemra, 1976]. One thing noticeable about the R2/R1 demarcation is that its latitude is higher in the postmidnight sector (00 MLT < 02) than in the surrounding sectors, which affects the statistical result of some boundaries as will be shown in section 5. Here we would like to make two points as possible reasons for this anomaly. [23] First, DMSP occasionally observes at postmidnight a pair of FACs that flow upward and downward at higher and lower latitudes, respectively as expected for the R2 and R1 pair after midnight, but the equatorward upward current is collocated with mono energetic electron precipitation as is often observed for the upward R1 current before midnight (section 5.3), and the precipitation of energetic ions, which are presumably of ring current origin, peaks farther equatorward of this upward current. Such events are better interpreted as a three sheet structure on the premidnight side extending across midnight but with a R2 current absent rather than as a conventional R2 R1 pair at postmidnight. Whereas the present study excludes such events if they occur before midnight (because the FAC polarities are the opposite), they are included if they occur after midnight as far as other criteria are met. Since the (apparent) R2/R1 demarcation is higher in latitude in those events, the overall distribution of the R2/R1 demarcation is shifted poleward. Their impact on the distribution of the poleward boundary of the R1 current may be less significant because the third FAC, the current poleward of the R1 current, is usually thinner than the R1 current. It is, however, still unclear why the equatorward boundary of the R2 current is apparently not affected. [24] Second, in the postmidnight sector, the Sunsynchronous orbits of DMSP satellites, especially DMSP F7, 12, and 14 16, whose orbits are oriented in prenoonpremidnight directions, tend to skim the auroral oval. However, the identification of the FAC structure requires that the satellite crosses through the R2 and R1 currents, for which the contracted auroral oval is preferable. It is therefore possible that the location of events at postmidnight are biased poleward. However, this idea alone does not explain why only the R2 R1 demarcation is higher in latitude but not the equatorward boundary of R2 or the poleward boundary of R1. It is also unclear why the poleward shift can be found only for the 00 MLT < 02 sector but not, for example, for the 02 MLT < 04 sector. [25] Interestingly, some polar distributions of FACs reported in the past show that the latitude of the R2 R1 pair (as defined from the FAC polarities) jumps around midnight [Iijima and Potemra, 1976; Weimer, 2001; Papitashvili et al., 2002]. We should therefore keep in mind that there is possibly an unknown bias in our results for the 00 MLT < 02 sector, whether it is real or it originates from our event selection. [26] The other frames of Figure 3 show the median of MLat of different precipitation boundaries. In each of those frames, the FAC boundaries shown in Figure 3a are plotted by gray lines. We will examine those precipitation boundaries in section 5, but for now, we just like to point out that the magnetic latitude of the precipitation boundaries change smoothly with MLT except that the MLat of the 5of13

6 Figure 3. (a) The absolute values of the magnetic latitudes, MLat, of the equatorward boundary of R2 currents (blue), the demarcation between R2 and R1 currents (red), and the poleward boundary of R1 currents (green) against MLT. MLat of the (b) b1e (red) and b1i (blue), (c) b2i, (d) b3a(red) and b3b (blue), (e) b4s, and (f) b5e (red) and b5i (blue) boundaries. In Figures 3b 3f, the FAC boundaries shown in Figure 3a are plotted by gray lines. The error bars show ranges between the 16 and 84 percentile points (one standard deviation from the median value). The plots are slightly shifted horizontally for distinction. b3b boundary is locally higher at 00 MLT < 02 and that in general the latitudes of precipitation boundaries change following the boundaries of large scale FACs. 5. Precipitiation Boundaries Relative to Large Scale FACs [27] In this section we examine the locations of different precipitation boundaries relative to large scale FACs. For the detailed definition of each precipitation boundary, see Newell et al. [1996]. We introduce a normalized latitudinal coordinate l, which is defined as follows. For each DMSP crossing of the auroral oval, the origin of l (l = 0) is set to be at the demarcation between R2 and R1 currents. l increases poleward irrespective of the hemisphere. In the positive (negative) territory, that is, on the poleward (equatorward) side of l = 0, l is normalized by the width of the R1 (R2) current sheet. That is, l =1(l = 1) at the poleward (equatorward) boundary of the R1 (R2) current. This normalization applies whether or not there is another FAC sheet poleward of the R1 current. In the following we examine the occurrence distribution of l for each boundary Zero Energy Precipitation Boundaries: b1e and b1i [28] Zero energy boundaries are defined separately for ions and electrons and are denoted as b1i and b1e, respectively. For both species the boundaries are identified based 6of13

7 Figure 4. The occurrence ratio of l (normalized MLat) of the (a and b) b1e and (c and d) b1i boundaries. In Figures 4a and 4c (Figures 4b and 4d), the red, blue, and green lines plot the distributions for 18 20, 20 22, and MLT (04 06, 02 04, and MLT), respectively. The error bars show ranges between the 16 and 84 percentile points (one standard deviation from the median value) divided by the square root of the number of events in each bin. on the cutoffs of the lowest energy channel (32 47 ev) flux. Figure 4 shows the occurrence ratio of the normalized magnetic latitude of the b1e (left) and b1i (right) boundaries for 3 different 2 h wide MLT sectors before (top) and after (bottom) midnight. The occurrence ratio p is defined as a number of data points in each bin n i, divided by the total number of data points n, that pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi is, p = n i /n. The error bar is calculated for each bin as pð1 pþ=n. Significantly less numbers of orbits are available for the midnight to dawn sector (Figure 1), which is especially the case for the b1i boundary since the ion data set is smaller (section 2), and accordingly the associated error bars are larger. [29] At dusk to midnight the b1e boundary is located inside the R2 current ( 1 l < 0) for the majority of orbits (Figure 4a). Note that in the dusk to midnight events we selected, the R2 current flows into the ionosphere (section 4), which is presumably carried by upward moving electrons. It is therefore not surprising that the R2 current extends in the region where there is no significant electron precipitation. At earlier MLTs, the occurrence distribution is narrower with its peak located more poleward (relative to the R2 system). The b1i boundary is also very often located inside the R2 current (Figure 4b), but it tends to be equatorward of b1e especially at earlier MLTs. These results can also be confirmed in the 12 March 1984 event (Figure 2a). [30] At midnight to dawn the b1e boundary is located noticeably more equatorward than at dusk to midnight and, in most cases, equatorward of the R2 current (Figure 4b). Its dependence on MLT, if at all, is not very clear (Figures 4b). The result is similar for the b1i boundary (Figure 4d). In the 24 July 1997 event, the b1i boundary is 0.6 equatorward of the equatorward boundary of the R2 current, and the b1e boundary is 2.1 farther equatorward (Figure 2b). This event is a good example of the zero energy (b1i and b1e) boundaries located equatorward of the R2 current at dusk to midnight. [31] The overall results may be explained in terms of the accessibility of zero energy particles to the inner magnetosphere. Zero energy charged particles, whether ions or electrons, do not undergo magnetic gradient or curvature drift. Accordingly, their convection paths in the magnetosphere follow equipotential contours and there is a certain area around Earth they cannot reach. For a simple convection pattern with the corotation electric field superposed to a uniform dawn to dusk electric field, this forbidden area is tear drop shaped with pointed and round sides at dusk and dawn, respectively, and it extends farther out in the dusk sector than in the dawn sector [e.g., Chen, 1970]. It is therefore expected that the outer boundary of this forbidden region is magnetically mapped to the ionosphere more poleward at dusk than at other MLTs, which can be confirmed in Figure 3b. In contrast, the equatorward boundary of the R2 system tends to be located more equatorward at early MLTs in the dusk to midnight sector (Figure 3a). As a result, the normalized latitudes (l) of the b1e and b1i boundaries tend to be highest at dusk Maximum Ion Energy Flux: b2i [32] The b2i boundary is defined as the location where the energy flux of ions above 3 kev has a maximum, and it well agrees with the isotropic boundary of 30 kev ions [Newell et al., 1998]. The isotopic boundary marks the most equatorward point where the pitch angle distribution of precipitating ions is isotropic. If the equatorial field line curvature is sufficiently large (the curvature radius is suffi- 7of13

8 Figure 5. The occurrence ratio of the normalized MLat l of the b2i boundary for (a) dusk to midnight and (b) midnight to dawn sectors. ciently small), ions undergo pitch angle scattering when they cross the equatorial plane [Sergeev et al., 1993]; the pitchangle scattering is effective if the curvature radius is less than 8 times the ion gyroradius for the equatorial magnetic field. Roughly speaking, if the pitch angle distribution is isotropic, the energy flux of ion precipitation increases along with the equatorial ion energy density, which increases with decreasing radial distance in the magnetotail. It is therefore expected that the maximum ion energy flux, that is, the b2i boundary, takes place at, or slightly equatorward of, the isotropic boundary. [33] In both 12 March 1984 and 24 July 1997 events, the b2i boundary was located inside the R2 current (Figures 2a and 2b). We found that these events are typical in terms of the b2i location. Figure 5a shows the occurrence ratio of l of the b2i boundary for three different MLT sectors at dusk to midnight. For the predominant majority of orbits, the b2i boundary is located inside the R2 current ( 1 l < 0) and it is rarely found in the R1 current (0 l < 1). The distribution slightly shifts equatorward (relative to large scale FACs) and becomes wider closer to midnight. [34] Figure 5b shows the occurrence distributions of the b2i boundary at midnight to dawn. The b2i boundary is located mostly inside the R2 current ( 1 l <0)at02 MLT < 04 and 04 MLT < 06 as was found for dusk to midnight. At 00 MLT < 02, in contrast, l tends to be more equatorward, and it is < 1 for 48% of the events in this 2 h sector. However, this tendency is probably overemphasized because the R2/R1 demarcation, the reference of l, in this MLT sector is located more poleward than in the surrounding sectors as we found in section 4. [35] Here we briefly examine the occurrence distribution of the b2i boundary for different FAC structures. Figure 6 shows the occurrence distribution of the b2i boundary for two sheet (gray) and three sheet (red) events at 20 MLT < 22 (Figure 6a) and at 22 MLT < 24 (Figure 6b). The distributions for the two sheets were already shown in Figure 5a. In the selected three sheet events as well as in the two sheet events the most equatorward FAC flows downward, that is, in the same direction as the eveningside R2 current. The second equatorward current flows upward, to which we have been referring as R1 current for the twosheet structure, but its identity is less clear for the threesheet structure. For the three sheet structure, there is an additional current farther poleward, at l 1, which flows downward. Note that three sheet structures are often observed in the premidnight sector [Iijima and Potemra, 1976] (see also Figure 1 for the occurrence frequency). [36] The occurrence distribution of the b2i boundary is almost identical for the two sheet and three sheet events for each MLT sector. In the predominant majority of the events, the boundary is located either inside the R2 current ( 1 l < 0) or farther equatorward (l < 1), and it is rarely located poleward of the R2 current (l 0). Since this is also the case for the midnight to dawn sector (Figures 3c and 5), the b2i boundary may be used for identifying the R2 current. The present result also suggests that the middle upward current of the three sheet structure in the dusk tomidnight sector is mapped farther down the tail than the upward R2 current in the midnight to dawn sector although these currents may appear to be a continuum in the polar distribution [e.g., Iijima and Potemra, 1976; Weimer, 2001; Papitashvili et al., 2002; Anderson et al., 2008]. [37] We also examined events with different FAC polarities. In Figure 7 the red lines show for 20 MLT < 22 Figure 6. The occurrence ratio of the normalized MLat l of the b2i boundary for (a) and (b) in MLT for the two sheet (gray) and three sheet (red) events. In both event sets the polarity of the R2 current is downward. 8of13

9 Figure 7. The occurrence ratio of the normalized MLat l of the b2i boundary for (a) and (b) in MLT for the two sheet events with the standard (gray) and opposite (red) FAC polarities. (Figure 7a) and 22 MLT < 24 (Figure 7b) the occurrence distributions of the b2i boundary for the two sheet events with upward and downward FACs at lower and higher latitudes, respectively, that is, with the FAC polarities opposite to what are expected from the R2 and R1 pair in the premidnight sector. The gray lines show, as a reference, the distributions for the two sheet events with the standard R2 and R1 polarities. For each MLT sector the distribution for the opposite polarity events is significantly different from that for the standard polarity events, and the b2i boundary is located more often at l < 1 than at l 1, that is, farther equatorward of the upward FAC. [38] This result can be explained by assuming that the lower latitude upward FAC in the opposite polarity events is actually the R1 current and that the R2 current is missing for some reason. In such a case the b2i boundary, which would be most likely inside or equatorward of the R2 current, is expected to be farther equatorward of this upward FAC as is shown in Figure 7. Note that the normalization of latitude is different for the equatorward and poleward sides of the R2 R1 demarcation, and therefore, even if the R2 current is missing, the distribution for the opposite polarity events should not be like the distribution for the standard polarity events shifted by 1. The present result strengthens the idea that the b2i boundary can be used for identifying the R2 current Most Equatorward and Poleward Electron Acceleration: b3a and b3b [39] The b3a and b3b boundaries are identified as the most equatorward and most poleward electron acceleration events, in which the electron energy spectrum has a monoenergetic peak indicating the existence of upward fieldaligned electric fields. The field aligned acceleration has to be 3 4 times the thermal energy of the source electrons and the energy flux has to be above 0.25 [ergs/cm 2 s]. [40] Figure 8a shows the occurrence distributions of the b3a and b3b boundaries for the two sheet events with the standard R2 and R1 polarities in the dusk to midnight sector. The distribution of the b3a boundary has a sharp peak at l = 0.1 (the adjacently equatorward data point is at l = 0.1), whereas the distribution of the b3b boundary has a sharp peak at l = 0.9. This result suggests that at dusk to midnight, electron acceleration is very often confined in the upward R1 current (0 l < 1). This is reasonable because the upward electric field, which accelerates electrons downward, is associated with upward FACs. [41] The difference in MLT is rather small, and the only noticeable MLT dependence is that closer to midnight the distribution of b3a tends to extend slightly more equatorward; l is < 0 for 22, 27, and 42% of events at 18 MLT < 20, 20 MLT < 22, and 22 MLT < 24, respectively. The fact that l is < 0 for some b3a boundaries indicates that field aligned electron acceleration occasionally takes place inside the downward R2 current, which presumably corresponds to small scale upward FACs embedded. We found that the b3a boundary is more often located in the downward R2 current as geomagnetic activity becomes higher (as measured by Kp; not shown). Electron acceleration can also take place poleward of the upward R1 current as indicated by the occurrence distribution of b3b extending to l 1. The 12 March 1984 event is a good example, in which the b3b boundary is located outside of the horizontal range of Figure 2a, at 73.8 in geomagnetic latitude, corresponding to an isolated mono energetic precipitation in the polar cap, Figure 8. The occurrence ratio of the normalized MLat l of the b3a and b3b boundaries for (a) dusk to midnight and (b) midnight to dawn sectors. 9of13

10 Figure 9. The occurrence ratio of the normalized MLat l of the b4s boundary for (a) dusk to midnight and (b) midnight to dawn sectors. whereas the b3a boundary is located just poleward of the R2/R1 demarcation. [42] Figure 8b shows the occurrence distributions of b3a and b3b for the midnight to dawn sector, where the R2 and R1 currents (as we selected) flow upward and downward, respectively. The overall distributions are significantly different from those in the dusk to midnight sector. The b3a boundary is located mostly inside or equatorward of the R2 current, that is, at l < 0. However, caution needs to be exercised in interpreting the b3a boundary at midnight to dawn in the context of field aligned acceleration because some data points in diffuse electron precipitation apparently satisfy the criteria for mono energetic electron events (as defined by Newell et al. [1996]). In the 24 July 1997 event the b3a boundary was located in the middle of diffuse electron precipitation (Figure 2b). Such events may be quite common as we will discuss in section 6.3. [43] The occurrence of the b3b boundary is peaked around l = 1, i.e., the poleward boundary of the R1 current, as we found for the dusk to dawn sector, but its distribution spreads far more widely. The b3b boundary poleward of the R1 current can be attributed most likely to isolated electron acceleration inside the polar cap. For example, in the 24 July 1997 event the b3b boundary was at 71.7 in geomagnetic latitude (outside of the horizontal range in Figure 2b), far poleward of the R1 current. In contrast, the b3b boundary inside the R1 current probably corresponds to the internal structure of the R1 current. That is, even though the R1 current (as we selected) flows downward at midnight todawn, upward field aligned electric fields may often be formed along with a local upward current in association with its internal structure Structured/Unstructured Boundary: b4s [44] The b4s boundary marks the transition of electron precipitation from a spatially unstructured (on a scale of 5 10 km) region on its equatorward side to a highly structured one on its poleward side. This boundary is defined based on correlation of electron energy spectra of some consecutive data points and is considered to separate CPS from BPS precipitation [Winningham et al., 1975], which used to be visually identified. In both 12 March 1984 and 24 July 1997 events the b4s boundary is located in the middle of the R2 current (Figure 2). [45] Figure 9 shows the occurrence distributions of this boundary in the same format as Figure 5. At dusk to midnight (Figure 9a), the distributions are centered around l = 0, the R2/R1 demarcation, and are almost confined within 1 l < 1. The occurrence distributions at 02 MLT < 04 and 04 MLT < 06 are narrower and centered slightly more equatorward than the duskside counterparts. The distribution at 00 MLT < 02 is significantly different with a tail extending equatorward, which might be again attributed to the complication of the identification of large scale FACs in this MLT sector (section 4). However, the overall result about the b4s boundary suggests that electron precipitation is more structured in the R1 current than in the R2 current Poleward Boundary of the Auroral Oval: b5e and b5i [46] The last boundaries we examine are the b5e and b5i boundaries, which can be regarded as the poleward boundaries of the auroral oval as determined by an abrupt drop (by a factor of at least 4) in the electron and ion energy flux, respectively. Although statistically the b5e and b5i boundaries agree well with the poleward boundary of the R1 current (Figure 3f), in individual events they do not have to agree. For example, downward FACs can be carried by upward moving electrons and may not be collocated with any significant electron precipitation. It is also possible that intense electron precipitation is not accompanied by any noticeable FAC if the collocated upward electron flux is also large. For example, in the 24 July 1997 event, the b5e boundary is located more than 5 poleward of the poleward boundary of the R1 current (Figure 2b), whereas in the 12 March 1984 event, it is located very close to the poleward boundary of the R1 current. [47] Figures 10a and 10b show the occurrence distributions of b5e for the dusk to midnight and midnight to dawn sectors, respectively. The occurrence ratio peaks around, but slightly equatorward of, the poleward boundary of the R1 current (l = 1) on both sides of midnight. However, the occurrence distribution for the dusk to midnight sector has more extending tails on the poleward side. These features are similar to what we found for the b3b boundary (Figure 8), but a careful inspection reveals that the distributions of b5e slightly shift poleward of b3b as expected from the definitions of these boundaries. [48] The occurrence ratio of the b5i boundary is also centered around l = 1 both before and after midnight (Figures 10c and 10d). Whereas at midnight to dawn, the b5e and b5i distributions are similar, the occurrence of the b5i boundary is more widely distributed than that of the b5e boundary at dusk to midnight. Here intriguing questions are 10 of 13

11 Figure 10. The occurrence ratio of the normalized MLat l of the (a and b) b5e and (c and d) b5i boundaries for the dusk to midnight (Figures 10a and 10c) and midnight to dawn (Figures 10b and 10d) sectors. why at dusk to midnight, the b5e boundary is often located inside the upward R1 current, which is carried mainly by downward electrons, and why at midnight to dawn the b5i boundary is often located inside the downward R1 current, which is presumably mapped to the plasma sheet. We will address these questions at section Discussion 6.1. Zero Energy Precipitation Boundaries: b1e and b1i [49] The locations of the b1e and b1i boundaries relative to the R2 current provides an important clue about the overlap between the ring current and the plasmasphere, which is crucial for better understanding the dynamics of each plasma population [e.g., Fok et al., 2005; Goldstein et al., 2005]. If the magnetospheric convection is steady for a sufficient time for refilling the plasmasphere, the locations of the b1e and b1i boundaries should agree with the location of the plasmapause. On the other hand, it is generally accepted that the R2 current originate in the ring current, although its source region may extend to the near Earth plasma sheet outside of geosynchronous orbit [Iijima et al., 1990]. We found in section 5.1 that at dusk to midnight, both b1e and b1i boundaries are often, more often than not, located inside the R2 current, which suggests that the plasmapause is very often located inside the ring current in this MLT sector. At midnight to dawn, in contrast, those boundaries are very often located equatorward of the R2 current and occasionally inside but near the equatorward boundary, of the R2 current. Therefore, the spatial overlap between the plasmasphere and the ring current may be more limited in this MLT sector although caution needs to be exercised because in principle the ring current can extend closer to Earth than the source region of the R2 current Maximum Ion Energy Flux: b2i [50] We found in section 5.2 that both before and after midnight, the b2i boundary is very often located inside the R2 current. Our result even suggests that the b2i boundary can be used as an identifier of the R2 current. On the poleward side of this boundary the pitch angle scattering at the equator is efficient enough to isotropize ions. Thus, a significant poleward portion of the R2 current should be described based on the assumption of isotropic ions rather than, for example, the conservation of the first and second invariants. [51] A similar remark applies to the formation of the Harang discontinuity, which characterizes the ionospheric convection pattern in the premidnight sector [Maynard, 1974]. The convection flow is directed eastward on its poleward side and westward on its equatorward side. The discontinuity originally referred to the reversal of horizontal ground magnetic deviations, but now the term is used interchangeably with the convection, therefore electric field, reversal. The convection electric field converges at the Harang discontinuity, and therefore, the discontinuity is collocated with an upward FAC. It is most reasonable to consider that this upward current consists in the upward R1 current or the middle upward current in case there are three FAC sheets, which may also be considered to be a R1 current based on the characteristics of particle precipitation (section 5.2). [52] The formation of the Harang discontinuity has been explained in terms of the transport of energetic ions to the inner magnetosphere and the subsequent magnetic drift [Erickson et al., 1991]. The b2i boundary is rarely in the R1 current but it is located farther equatorward in the predominant majority of events. It is therefore suggested that in the magnetotail area corresponding to the Harang discontinuity, the equatorial curvature is large enough to isotropize 11 of 13

12 the local ions. For a typical range of the b2i latitude in the midnight sector (>64 ; see Figure 3c), the b2i boundary is mapped to the equatorial plane outside of geosynchronous orbit (Plate 1 of Wing and Newell [1998]), and therefore, the Harang discontinuity is mapped farther outside Other Precipitation Boundaries: b3a, b3b, b4s, b5e, and b5i [53] The results of the other boundaries (i.e., b3a, b3b, b4s, b5e, and b5i) clearly show the dawn dusk asymmetries of FAC and precipitation signatures. At dusk to midnight, the occurrence distributions of the b3a and b3b boundaries have their peaks just poleward of the R2/R1 demarcation and just equatorward of the poleward boundary of the R1 current, respectively, and the b4s boundary is located around the R2/R1 demarcation. It is therefore suggested that the upward R1 current is structured and is characterized by mono energetic electron precipitation. The structuredness of the R1 current may also explain why the b5e boundary is often, as often as not, located inside the R1 current. In general, an upward FAC is collocated with electron precipitation, but this does not mean that the energy flux of electron precipitation is uniform in latitude. If, for example, small scale downward FACs are embedded, the energy flux can drop significantly, which may be identified as the b5e boundary. Our results also suggest that in the R2 current at dusk to midnight, the spectrum of precipitating electrons changes gradually in space, and field aligned electron acceleration is far less common than in the R1 current. The overall characteristics are consistent with what one may expect from the FAC polarities, that is, upward for the R1 current and downward for the R2 current. [54] The correspondence between FACs and precipitation is more complex in the midnight to dawn sector. There are two points to note. First, the occurrence distribution of the b3a boundary is peaked in the equatorward half of the upward R2 current and has extending tails (Figure 8b). However, we need to be cautious in interpreting this result in terms of the occurrence of auroral electron acceleration. The occurrence distribution of the b4s boundary has its peak in the poleward half of the R2 current, and at 02 MLT < 04 and 04 MLT < 06 it is almost confined poleward of the equatorward boundary of the R2 current (Figure 9b). Thus, we can conclude that the b3a boundary is often located equatorward of the b4s boundary, that is, in the middle of unstructured electron precipitation. We can confirm this tendency by comparing the MLat s of those boundaries at midnight to dawn (see Figures 3d and 3e). It is therefore likely that the data point identified as the b3a boundary at midnight to dawn is often embedded in diffuse electron precipitation as we found in the 24 July 1997 event. [55] Second, at midnight to dawn, the b3b boundary is located most frequently around the poleward boundary of the downward R1 current, l = 1 (Figure 8b), although its occurrence distribution has extending tails. That is, monoenergetic electron precipitation often takes place inside the R1 current as well as farther poleward as we saw in the 24 July 1997 event (Figure 2b). Since mono energetic electron precipitation should be collocated with upward FACs, this result suggests that the downward R1 current at midnight to dawn is very often structured with upward FACs embedded in it. [56] The structuredness of the R1 current may also be the reason why the b5i boundary is often located inside the downward R1 current rather than farther poleward at midnight to dawn. Poleward of the b2i boundary, the magnetic field lines are considered to be stretched enough to effectively fill the loss cone at the equator. The latitudinal profile of the precipitating ion energy flux therefore reflects its equatorial distribution unless precipitating ions are blocked or decelerated by field aligned potential differences. If the ion energy flux abruptly drops in the middle of the downward R1 current (as indicated by the b5i boundary), it suggests either that there exists a sharp spatial gradient of the ion energy density in the plasma sheet or an upward field aligned electric field prevents magnetospheric ions from precipitating into the ionosphere. For the former we cannot think of any good candidate. If the latter is the case the spatial scale of the upward electric field should be much smaller than the scale of the downward R1 current, which is consistent with the above discussion about the b3b boundary. 7. Summary [57] In this study we statistically examined the location of various precipitation boundaries relative to R2 and R1 currents on the night side. For each DMSP pass, the precipitation boundaries were identified based on the definitions proposed by Newell et al. [1996] and the FAC structure was identified by the automated procedure developed by Higuchi and Ohtani [2000a, 2000b]. Except in the analysis of the b2i boundary, we focused on two sheet FAC structure events with downward R2 and upward R1 currents at dusk to midnight and with upward R2 and downward R1 currents at midnight to dawn. The following is the summary of the results: [58] 1. The b1e and b1i boundaries (electron and ion zeroenergy boundaries) are located mostly inside the R2 current at dusk to midnight and near the equatorward boundary of the R2 current or farther equatorward at midnight to dawn. This result suggests that the overlap between the ring current and the plasmasphere is more significant at dusk than at dawn. [59] 2. The b2i boundary (the maximum energy flux of ion precipitation) occurs in the R2 current both before and after midnight. This is also the case for three sheet structure events in the premidnight sector. For premidnight events with the opposite R2 and R1 polarities, the b2i boundary tends to occur equatorward of the R2 current. Such twosheet structure events may be interpreted as a poleward part of the three sheet structure with a R2 current missing. These results suggest that the b2i boundary can be used as an identifier of the R2 current. [60] 3. The occurrence distributions of the b3a and b3b boundaries (the most equatorward and poleward electron acceleration events) indicate that mono energetic electron precipitation is mostly confined in the upward R1 current at dusk to midnight, whereas at midnight to dawn, it occurs in a wider latitudinal range including the downward R1 current. [61] 4. The b4s boundary (the transition between structured and unstructured electron precipitation) tends to occur 12 of 13