The source region and its characteristic of pulsating aurora based on ...

JOURNAL OF GEOPHYSICAL RESEARCH, VOL. 116, A03226, doi:10.1029/2010JA015507, 2011

The source region and its characteristic of pulsating aurora based on the Reimei observations Takanori Nishiyama,1 Takeshi Sakanoi,1 Yoshizumi Miyoshi,2 Yuto Katoh,3 Kazushi Asamura,4 Shoichi Okano,1 and Masafumi Hirahara5 Received 29 March 2010; revised 28 December 2010; accepted 13 January 2011; published 26 March 2011.

[1] Using image and particle data sets obtained from observations by the Reimei satellite, we carried out time‐of‐flight (TOF) analysis for 29 pulsating aurora events to understand the precise properties of pulsating auroras and the possible generation process. While the sources identified using a standard TOF model were distributed almost continuously from magnetic latitudes 50° to −20°, the sources identified using a different TOF model that takes into account whistler mode wave propagation were confined to the equatorial region up to about 15°. The latter source distribution agrees with previous statistical studies of whistler mode chorus waves. In addition, the cold plasma density of the source region and the wave frequency can be estimated from the latter TOF analysis. The estimated cold plasma densities and wave frequencies normalized by the equatorial cyclotron frequency ranged in 0.20–21.7 cm−3 and 0.22–0.65, respectively. The estimated wave frequency showed clear dependence on the invariant latitudes of the pulsating aurora source region and increased up to the frequency range of the upper band chorus as the distance from the Earth decreased (up to about 5–6 RE). These results suggest that both lower and upper band chorus wave contribute to the electron scattering of pulsating auroras, which depends on the radial distance of the source region. Citation: Nishiyama, T., T. Sakanoi, Y. Miyoshi, Y. Katoh, K. Asamura, S. Okano, and M. Hirahara (2011), The source region and its characteristic of pulsating aurora based on the Reimei observations, J. Geophys. Res., 116, A03226, doi:10.1029/2010JA015507.

1. Introduction [2] Pulsating aurora is a common phenomenon in which quasiperiodic emissions vary in a diffuse aurora, and its emission is characterized not by a sinusoidal change but by pulsation, and the typical period is from a few seconds to a few tens of seconds [e.g., Oguti et al., 1981; Yamamoto, 1988]. Pulsating auroras usually appear during the substorm recovery phase in the postmidnight and dawn sectors with various and irregular shapes, such as a patch or a band. Optical observations showed that their size in the latitudinal and longitudinal directions is 10–200 km. Precipitating electrons, which generate pulsating auroras, have been observed with quasi 3 Hz modulations from rockets and low‐altitude satellites [Sandahl et al., 1980; Yau et al., 1981; Sato et al., 2004]. Their energy ranges from a few keV to a few tens of keV, which is higher than typical 1 Planetary Plasma and Atmospheric Research Center, Tohoku University, Sendai, Japan. 2 Solar‐Terrestrial Environment Laboratory, Nagoya University, Nagoya, Japan. 3 Graduate School of Science, Tohoku University, Sendai, Japan. 4 Institute of Space and Astronautical Science, Japan Aerospace Exploration Agency, Sagamihara, Japan. 5 Graduate School of Science, University of Tokyo, Tokyo, Japan.

Copyright 2011 by the American Geophysical Union. 0148‐0227/11/2010JA015507

electron energies in diffuse aurora. Nemzek et al. [1995] demonstrated that high‐energy electrons (>20 keV) are supplied by substorm injection and correlate with the appearance of a pulsating aurora. [3] Since pulsating auroras appear in diffuse aurora, the electrons are expected to undergo cyclotron resonance with whistler mode waves in the equatorial region of the magnetosphere and to precipitate into Earth’s upper atmosphere due to pitch angle scattering [Johnstone, 1983; Davidson, 1990]. Moreover, theoretical studies suggested that the nonlinear growth of whistler mode waves, i.e., chorus, in a magnetic flux tube plays an important role in the periodic precipitations that cause pulsating auroras [Demekhov and Trakhtengerts, 1994; Trakhtengerts, 1999]. [4] The idea that pulsating auroras are generated by wave‐ particle interactions at the magnetic equator is widely accepted and supported by a number of indirect observations [e.g., Bryant et al., 1971; Lepine et al., 1980; Tsuruda et al., 1981; Hansen and Scourfield, 1990]. In particular, some studies deduced that the source regions are located near the equator from time‐of‐flight (TOF) analyses using downward electron flux with different energies [Bryant et al., 1971; Yau et al., 1981]. However, these studies had large error bars due to temporal resolutions of particle measurements and have not used realistic magnetic field models. On the other hand, Sato et al. [2004] demonstrated that the source region of pulsating auroras is located earthward, far

A03226

1 of 12

A03226

NISHIYAMA ET AL.: SOURCE AND MECHANISM OF PULSATING AURORA

from the equatorial plane. Many simultaneous observations at geomagnetic conjugate points revealed that conjugate pulsating auroras are often characterized by differences in shape, luminosity, and period, suggesting that the modulation region is located off the equator [Stenbaek‐Nielsen et al., 1973; Watanabe et al., 2007]. Since observational results have different conclusions about the source for the pulsating aurora, the exact source region and generation mechanism of pulsating auroras remain to be identified. [5] Simultaneous ground‐based and in situ (e.g., rocket and satellite) observations have been useful in elucidating the details of aurora phenomena. However, completely conjugating both types of observations is difficult, and observation chances are inherently limited. The study presented in this paper focused on identifying the source region and production mechanism of pulsating auroras using data obtained by the Reimei satellite. This data is particularly useful because Reimei can simultaneously observe an aurora image and the precipitating particle fluxes, and statistical analyses are possible using data for a number of simultaneous image‐particle observations. Moreover, the high spatial and temporal resolution of the data makes it possible to investigate each pulsation patch and energy‐time dispersion signature with a timescale of a few 100 ms in downward electrons. Therefore we can obtain statistical results using a number of simultaneous image‐particle observation data. [6] Recently, a new TOF model that includes the effect of propagating whistler mode waves has been proposed [Miyoshi et al., 2010]. A case study of TOF analysis using this model and data from the Reimei satellite revealed that the modulation region is distributed around the equator. As the application of this TOF analysis, several plasma parameters such as the cold plasma density in the modulation region were estimated. [7] In this study we conduct statistical studies of the energy dispersions associated with pulsating auroras. Two TOF analyses, the standard TOF and TOF of Miyoshi et al. [2010], are applied to the same auroral events, and the ability of each model to describe the phenomena is discussed in this study. We also investigate the plasma density and wave frequency that can be derived from the TOF analysis proposed by Miyoshi et al. [2010]. Section 2 explains the Reimei satellite and the TOF analysis methods. Section 3 describes the statistical results for the auroral events obtained from the TOF analyses. Section 4 discusses the source distributions and mechanisms of pulsating auroras. Finally, section 5 describes the key findings of this study.

2. Instruments and Analysis Methods [8] Reimei is a scientific small satellite developed for investigating mechanisms responsible for auroral fine structures. It was launched from the Baikonur Space Center and inserted into a Sun‐synchronous polar orbit (inclination: 98.6°) at an altitude of approximately 630 km on 23 August 2005. Its orbital meridian is on 0050–1250 local time, and it takes about 98 min to orbit the Earth. The scientific instruments on board the satellite are a three‐channel monochromatic auroral camera (MAC) [Sakanoi et al., 2003; Obuchi et al., 2008], top‐hat electrostatic electron and ion energy spectrum analyzers (ESA/ISA) [Asamura

A03226

et al., 2003, 2009], and current probes (CRM). The local pitch angles of the measured electrons and ions can be deduced by an attitude magnetometer. [9] The MAC captures an image of aurora at three wavelengths: 427.8 (N+2 first negative band), 557.7 (O green line), and 670.0 (N2 first positive band) nm. The field of view is 7.6°, and the temporal and spatial resolutions are 120 ms and 1 km, respectively. The E/ISA is a top hat type electrostatic analyzer that can detect electrons and ions with energy from 10 eV/q to 12 keV/q, and one scan for all 32 energy steps takes no more than 40 ms. Its field of view covers 300° in the polar angle direction, divided into 30 sectors. Three‐axis attitude control of the satellite makes it possible to point the MAC field of view to the ionospheric footprint connected with the satellite position along the field line. This observation mode is called “mode‐S,” and we use mainly mode‐S MAC and ESA data in this study. [10] We first present a typical pulsating aurora event on 18 October 2007 over the Alaska peninsula from 1135:12 to 1135:59 UT observed by Reimei in mode‐S. Note that this event has been analyzed by Miyoshi et al. [2010] as a case study. During this observation, the magnetic footprint of the Reimei satellite moved along the 0.7 magnetic local time (MLT) sector at invariant latitudes ranging from 65.9 to 62.7° (L = 6.00–4.75), corresponding to latitudes between 65.3 and 62.5° and east longitudes between 213.0 and 210.8° in the geographical coordinates. The image‐particle data obtained from the observations of mode‐S show the fine‐ scale relationship between a pulsating aurora and the precipitating electrons. Figure 1 indicates summary plots of the data from the ESA and MAC observations (1135:00– 1136:00 UT). Figures 1a and 1b are energy‐time spectrograms for the downward electron flux with a local pitch angle of 0–30° during the mode‐S observations. Reimei passed over several regions characterized by energy‐time dispersions of precipitating electrons with energies of more than a few keV. The precipitating electrons have 3–4 Hz modulations and the modulations were clearly seen especially in higher‐energy channels (8.7 and 12 keV). Such quasi‐3 Hz modulations have been reported in previous studies [e.g., Sandahl et al., 1980; Sato et al., 2004]. The precipitations were intermittent from 1135:10 to 1135:18 UT. Figures 1c and 1d are the corresponding images which show that the aurora was bright and extended in the north‐south direction. Figure 1e shows no significant aurora emissions although most of the imaged area overlaps that in Figure 1d. This means that MAC data can be used to identify the time variation in a pulsating aurora. Note that Figures 1c and 1d correspond to the on‐phase and Figure 1e corresponds to the off‐phase and that the quasiperiodic emission had a recurrence period of about 3–4 s. Some precipitations between 1135:20 and 1135:50 UT similarly correspond to pulsating auroras (see auxiliary material obtained from Reimei/MAC).1 [11] As indicated by the red arrow in Figure 1b, an energy‐time dispersion can be clearly seen, with a lower energy limit of about 2 keV, at around 1135:14 UT. Since TOF analysis requires the energies and time delays of a particle arriving at the detector, as described below, we 1 Auxiliary materials are available in the HTML. doi:10.1029/ 2010JA015507.

2 of 12

A03226

NISHIYAMA ET AL.: SOURCE AND MECHANISM OF PULSATING AURORA

A03226

Figure 1. Summary plots of Reimei data for 18 October 2007 auroral event, showing energy‐time spectrograms of precipitating electrons for (a) 1135:00 to 1136:00 and (b) 1135:00 to 1135:30 UT, respectively. Black arrows in Figure 1a indicate the start and end timings of Mode‐S observation, and red arrow in Figure 1b around 1135:14 UT indicates energy dispersion analyzed. Also shown are (c–e) successive images plotted in geographical coordinates showing variations in pulsating aurora during on‐phase (Figures 1c and 1d) and off‐phase (Figure 1e). White square in each image represents location of ionospheric footprint of magnetic field line threading Reimei.

carried out a few analyses to estimate the values precisely. Figure 2a shows the normalized count of electrons for each energy channel (12, 8.7, 6.8, 5.3, 4.2, 3.2, and 2.4 keV) as a function of time for the energy‐time dispersion starting at 1135:14 UT, corresponding to the dispersion indicated by the red arrow in Figure 1b. The normalized counts of higher

energies have a better signal‐to‐noise (S/N) ratio because of their higher count rate. To obtain a better S/N ratio for the lower energy electrons, we applied an exponentially weighted moving average to the count data at each energy channel. For a weighted moving average, we used three data intervals, corresponding to 120 ms. This method makes it

3 of 12

A03226

NISHIYAMA ET AL.: SOURCE AND MECHANISM OF PULSATING AURORA

A03226

Figure 2. (a) Normalized count of electrons for each energy channel (12, 8.7, 6.8, 5.3, 4.2, 3.2, and 2.4 keV) as a function of time starting at 1135:14 UT for 18 October 2007 auroral event. (b) Normalized count on exponentially weighted moving average chart for each energy channel. possible to gain the S/N ratio while retaining the original waveforms. The results are shown in Figure 2b. In this analysis, an initial rise of energy‐time dispersion should satisfy the condition that the flux is larger in one order than previous background flux. The initial rise at 12 keV was identified around 1135:14.3 in this case. Next, cross‐ correlation analysis was conducted, and the time delays among the variations in the different energy channels were estimated. The 2.4 keV electrons, for example, were found to have an estimated time delay of 480 ms compared to the 12 keV electrons. [12] We can calculate the distance between the observation location and source position from the TOF analysis assuming that electrons with different energies start precipitating simultaneously from the same source position. With this standard TOF model, source distance L is given by  


1 = ti
tj vi < vj and tj < ti L
1 ¼ v
1 i
vj

ð1Þ

where v and t are electron velocity and detection time, respectively [Hardy et al., 1990; Sato et al., 2004]. The distance to the electron source is obtained by fitting a line to the observed data. The source in the magnetosphere can be mapped by tracing the field line from the observed satellite position using the realistic magnetic field model of Tsyganenko and Sitnov [2005]. The solar wind parameters used for the magnetic field model were obtained from the OMNI database.

[13] We also carried out an alternative TOF analysis, which takes into account whistler mode wave propagation and wave‐particle interactions [Miyoshi et al., 2010]. It is based on the idea that a whistler mode wave packet originating at the magnetic equator propagates toward higher latitudes and resonates with electrons, when the resonance conditions are satisfied. Previous studies revealed that chorus waves are generated at the magnetic equator and propagate away from it [e.g., Santolik et al., 2003; Omura et al., 2008]. Figure 3 shows the schematic sequence of this process adapted from Miyoshi et al. [2010]. At first, whistler mode wave is generated, and it resonates with the electron with the lowest energy of precipitations at the magnetic equator. Then the whistler mode wave propagates to the high latitudes. The propagating whistler mode wave can resonate with the higher‐energy electrons at the high latitudes. This model has two key features. One is that the source region of a pulsating aurora observed in the Southern (Northern) Hemisphere continues northward (southward) along the field line from the equator. This is because an electron interacts via cyclotron resonance w − kkvk = nwce/g, where w < wce (n = 0, ±1, ±2…) with a whistler mode wave that propagates to the opposite directions. The second key feature is that the higher energy electrons arrive at the ionosphere earlier even though the resonance of higher‐energy electrons occurs after that of lower‐energy electrons and the higher‐energy electrons travel a longer path, which reproduces the observed energy dispersions.

4 of 12

A03226

NISHIYAMA ET AL.: SOURCE AND MECHANISM OF PULSATING AURORA

A03226

Figure 3. Schematic diagrams showing (a–d) a sequence of pitch angle scattering of electrons with propagating whistler mode waves and (e) expected energy‐time diagram at ionospheric altitude (adapted from Miyoshi et al. [2010]). [14] In this paper, the two TOF analyses are hereafter referred to as “TOF‐A” (standard) and “TOF‐B” [Miyoshi et al., 2010], respectively. We carried out TOF‐A and TOF‐B analyses by using the energy dispersion signature of the downward electrons causing a pulsating aurora observed by Reimei. We estimated the distance from the source region by linear fitting to the inverse electron velocity as a function of time in the TOF‐A analysis. Figure 4a shows the results of linear fitting using the TOF‐A analysis; the estimated distance to the source was about 30,000 × (1 ± 0.034) km, corresponding to magnetic latitude 25.0°. The TOF‐B model assumes that whistler mode chorus has a typical rising tone of 7.6 kHz/s [Santolik et al., 2008] and finite frequency ranges (0.10 < w/Wce < 0.50, 0.50 < w/Wce < 0.70) [e.g., Santolik et al., 2005b; Bortnik et al., 2007]. Note that Wce means equatorial cyclotron frequency. We assume that whistler mode waves propagate along a field line keeping their initial frequencies w = aWce, where 0.10 < a < 0.70 [Santolik et al., 2003]. Each frequency component of a rising tone starts to propagate from the equator with a time delay corresponding to the frequency drift, and each frequency component should interact with electrons. In the

analysis, an adjustment for thermal plasma density N (0.10– 30.0 cm−3) and whistler mode wave frequency is needed to fit the theoretical dispersion curve to the measured energy dispersion. Figure 4b is the results of the fitting. The red points represent energy of each ESA channel, and black solid lines indicate an energy‐time dispersion of precipitating electrons corresponding to each frequency component of a rising tone. The line that first appeared in the E‐T diagram resulted from wave‐particle interactions with the minimum frequency component launched from the equator, and subsequent energy‐time dispersions can be seen due to the precipitations caused by higher‐frequency component. There may be inconsistency between ESA measurements and precipitations expected from the model at the energies between 2 and 1 keV, probably because of a few count rates at this lower energy. [15] We estimated a plausible combination of N and w/Wce was 9.7 cm−3 and 0.40 < w/Wce < 0.45 by fitting of the energy‐time dispersion measured by ESA. This best combination was identified by a chi‐square test between the data points of ESA measurements and a curve that corresponded to the minimum resonance frequency. Figure 4c shows the

5 of 12

A03226

NISHIYAMA ET AL.: SOURCE AND MECHANISM OF PULSATING AURORA

A03226

Figure 4. (a) Inverse velocity of measured electrons at 1135:14.39 UT for 18 October 2007 auroral event. Dashed line indicates linear fitting of data for seven ESA channels. Calculated distance between source region and Reimei satellite was 29,600 km. (b) Result of best fitting between theoretical dispersion curve and observed energy dispersion. Red points represent energies of ESA channels. Observed dispersion (red points) fits theoretical dispersion curve well assuming N and w/Wce are 9.7 cm−3 and 0.40 < w/Wce < 0.45, respectively. (c) The density and frequency dependence of the calculated dispersion curve. N and w are changed into (9.7 ± 1.5) cm−3 and (0.40 ± 0.02) Wce, respectively. This result indicates that error bars of the density and frequency could be estimated to be about ±1.5 cm−3 and less than 0.02 Wce, since the calculated dispersion curves are included by error bars of energy (not shown) in each ESA channel. density and frequency dependence of the energy‐time dispersion curve. Error bars of the density and frequency can be estimated to be about ±1.5 cm−3 and less than 0.02 Wce. The energy of precipitating electrons depends on the latitudes where the cyclotron resonance occurs, and therefore the range of source latitudes can be estimated from this analysis. In this case, the electron energy ranged from 2.4 to 11 keV. This energy range corresponds to the region between the magnetic equator and magnetic latitude −14.3°.

3. Results [16] The TOF‐A and B analyses were carried out for the 29 auroral events in 11 satellite paths that showed the electron energy dispersion clearly in the E‐T spectrum, and statistical distributions of the source regions for both TOF analyses were obtained. The data sets were obtained from mode‐S observations in the Northern Hemisphere (L = 5.13–6.34) during winter in 2005–2008. Figure 5a shows the source regions estimated by the TOF‐A analyses mapped onto coordinates for which the horizontal axis corresponds to the magnetic equator of the field lines for each event. Though a few sources estimated by TOF‐A analyses were in the Southern Hemisphere, most of the sources were distributed in the Northern Hemisphere (86%). In addition, this statistical result revealed the important fact that the modulation regions were not localized near the Earth but distributed continuously from the equator to the high

latitude regions of about 50°. Figure 5b shows the source regions obtained from the TOF‐B analysis. Each red line corresponds to the estimated interaction regions for the observed energy range. Since the pulsating aurora in this event was observed in the Northern Hemisphere by Reimei, the observed precipitating electrons interact with whistler mode waves that propagate from the magnetic equator to the Southern Hemisphere. Note that the magnetic equator is defined as the location of the minimum magnetic field along a field line in this study. Chorus waves have been known to often propagate from the off‐equator [Santolik et al., 2005a], and the distance from the equator was estimated to be a few thousand kilometers. This distance corresponds to latitudinal distribution of the source region within 2–3°, and such latitudinal distribution on the source region does not cause any significant impact on this analysis because the difference of the local cyclotron frequency along the field line can be negligible around the magnetic equator. Moreover, the maximum growth of chorus waves is observed at the equator in the simulation study [e.g., Katoh and Omura, 2007], and therefore, it is plausible to assume that the chorus waves propagate from the magnetic equator. [17] Figures 6a and 6b show the latitudinal distributions of the sources calculated from the TOF‐A analysis and the off‐ equator boundaries of the source region calculated from the TOF‐B analysis, respectively. Figure 6a shows that the sources are distributed almost continuously from magnetic latitude −20 to 50° and are somewhat concentrated from 15°

6 of 12

A03226

NISHIYAMA ET AL.: SOURCE AND MECHANISM OF PULSATING AURORA

Figure 5. (a) Statistical distribution of 29 source regions identified by TOF‐A analysis mapped onto coordinates in which horizontal axis corresponds to magnetic equator of field lines for each auroral event (not GSM coordinates). Dotted lines indicate magnetic latitudes at 0, 15, 30, 45, 60, 75, and 90°. Red points represent point sources. (b) Statistical distribution of source regions identified by TOF‐B analysis; red solid lines represent the source regions, where wave‐particle interactions occurred between whistler mode waves and electrons. to 25°. Figure 6b shows that the boundaries of the source regions are distributed around −12°. The red broken line shows the distribution of the calculated off‐equator boundaries assuming precipitation of higher‐energy (25 keV) electrons (not measured by Reimei). In this case, these boundaries extend to −15°. [18] The region where Reimei observed the 29 pulsating auroras ranged from 63.8 to 66.6° (L = 5.13–6.34) in the invariant latitude. This region is close to the plasmapause that exists typically at L = 4–6. Since the cold plasma density is an important parameter for wave‐particle interactions, the relative position of the pulsating aurora source to the plasmapause is essential for understanding of the generation process. Figure 7 shows the relationship between the source region and plasmapause location estimated from an empirical model [O’Brien and Moldwin, 2003]. The four dash‐dotted lines indicate the L values of the empirical plasmapause in each MLT sector as a function of the maximum of the AE index for the preceding 36 h. The diamonds represent the L values and AE indices when Reimei observed pulsating aurora in these MLT sectors. As shown in Figure 7, all 29 auroral events occurred outside the empirical plasmapause. The cross‐correlation coefficient between log10(AE)max and L was −0.46. The source regions of the pulsating auroras also tend to move closer to the Earth as the AE index increase, that is, the plasmapause moves earthward. [19] The cold plasma density and whistler mode wave frequency can be estimated from the TOF‐B analysis. As

A03226

well as an example of Figure 4c, error bars of the density and frequency in all events can be estimated to be about ±1.5 cm−3 and less than 0.02 Wce, respectively. Figure 8 shows the distributions of the cold plasma density and whistler mode wave frequency plotted in the GSM coordinates, invariant latitudes, and MLTs. Figures 8a and 8b show the source locations of the pulsating auroras mapped onto each of the magnetic equatorial planes determined by the empirical model of Tsyganenko and Sitnov [2005]. The colors represent the cold plasma densities and whistler mode wave frequencies. Figures 8c and 8d are the same as Figures 8a and 8b except for mapping by the invariant latitudes and MLTs. These positions were calculated from a footprint at altitude of the 110 km of Reimei position using the international geomagnetic reference field (IGRF) model. The color scales of density and frequency (normalized by the equatorial cyclotron frequency) range from 0.10 to 30.0 cm−3 and from 0.10 to 0.70, respectively. [20] The estimated cold plasma densities are consistent with the results of empirical models [Carpenter and Anderson, 1992; Sheeley et al., 2001], except for only one case where the density reached 21.7 cm−3. As shown in Figure 8a, the density tends to increase as the radial distance from the Earth decreased. The density also increased in the midnight region compared to that at later MLTs (Figure 8c) although there were also low density (1.0–2.0 cm−3) regions. The higher‐ frequencies can be obtained in the near the Earth and the midnight region as shown in Figures 8b and 8d, respectively. It should be important to note that the estimated wave frequency covers from the lower to the upper band chorus. [21] To investigate in detail the characteristics of the cold plasma density and wave frequency, we plot the dependences of density and wave frequency against the radial distance for the 29 events. As shown in Figure 9 (top), the estimated densities were lower (5 keV) associated with pulsating auroras but also an inverted‐V structure ( 0.30) can resonate with electrons of 12 keV under the condition of 20.0 cm−3 in the vicinity of the Earth (