Journal of Geophysical Research: Space Physics

Oxygen aurora during the recovery phase of a major geomagnetic storm

Authors


Abstract

[1] We have analyzed ultraviolet spectra measured in the postsunset auroral zone on 16 July 2000, during the recovery phase of the major geomagnetic storm of 14–16 July 2000. We find enhanced oxygen ion and neutral line emissions above 300 km in the postsunset sector of the auroral oval during the initial fast recovery phase of the storm, also called the Bastille Day storm. No comparable emissions are seen in simultaneous measurements of nitrogen and hydrogen emission features, indicating that these features are not related to electron or proton aurora. The enhancements are seen slightly equatorward of all other nitrogen line and band emissions that are presumed to comprise the nominal diffuse electron auroral oval. On the basis of the altitude profile, location, and timing of the enhancements, we interpret these oxygen emissions as a product of precipitating ring current oxygen ions in the auroral atmosphere, possibly after pitch-angle scattering by electromagnetic ion cyclotron waves. These emissions may provide an important measure of the contribution of collisions and charge exchange interactions to the loss of oxygen ions and the recovery of geomagnetic storms and substorms.

1. Introduction

[2] The terrestrial polar caps and auroral ovals are dynamic regions that visually display significant connections between the Sun and Earth's magnetosphere, ionosphere, and upper atmosphere. Measurements of ultraviolet (UV) auroral emissions have provided important pieces to this complex puzzle. The most abundant ions, atoms, and molecules, including hydrogen, helium, oxygen, nitrogen, and nitric oxide, contain transitions at these wavelengths owing to excitation by precipitating energetic electrons and protons. The first observations of UV aurora were taken by an imaging camera on the Moon during the Apollo 16 mission [Carruthers and Page, 1976]. The field flourished when the Dynamics Explorer satellite used its improved vantage point over the pole to provide the first comprehensive views of the auroral oval at UV wavelengths [Frank et al., 1982; Frank and Craven, 1988].

[3] The properties of auroral emissions change significantly during solar and geomagnetic activity. In the UV, bright signatures of atomic oxygen have been seen in the polar cap and the dayside cusp during geomagnetic storms that were anomalous compared with other times [Paresce et al., 1983; Chakrabarti, 1986]. Other auroral features observed equatorward of the diffuse auroral region during a period of enhanced geomagnetic activity were attributed to precipitating keV oxygen ions that were converted into a stream of energetic neutral atoms via charge exchange with ambient oxygen [Ishimoto et al., 1986, 1989]. This interpretation was supported by the first detection of energetic O+ precipitating into middle to high latitudes during several major geomagnetic storms [Shelley et al., 1972; Sharp et al., 1974, 1976a, 1976b]. These keV ion and neutral oxygen particles produce an emission source equatorward of and at higher altitudes than diffuse electron aurora [Ishimoto et al., 1989]. Emission enhancements at low and equatorial latitudes also have been attributed to the precipitation of oxygen neutral atoms after charge exchange in the ring current [Tinsley, 1979; Stephan et al., 2001].

[4] Each observation can be related to a process within the evolution of a geomagnetic storm. The details of how ion outflow and precipitation influences the development of storms and substorms are voluminous and still a subject of ongoing studies. However, the following simplified picture has developed over the past 4 decades of research. The interplanetary magnetic field (IMF) turns southward for some period of time, leading to dayside merging of the terrestrial and solar magnetic fields. Ionospheric O+ is injected into the magnetotail plasma sheet and inner magnetosphere. The terrestrial field lines reconnect in the nightside magnetotail and force the ions in the sunward direction toward Earth. The energy in the ring current grows and becomes asymmetric as these ions gradient drift to the west, resulting in a decreased equatorial magnetic field that is measured by low-latitude magnetometers and reflected in the Dst geomagnetic index. The storm recovery phase begins after the IMF turns northward, reconnection slows, and any energy injected into the ring current is overwhelmed by the energy lost due to transport of ring current ions out the dayside magnetopause, charge exchange conversion to energetic neutral atoms (ENAs), and precipitation into the lower atmosphere. Ion precipitation from the ring current can occur as a result of scattering of the ions into the loss cone by plasma waves [Jordanova et al., 2001] or by nonadiabatic motion where field line curvature is comparable to the ion gyroradius [Sergeev et al., 1983, 1993; Anderson et al., 1997]. These ions provide clues as to the composition and conditions in the magnetotail, particularly during the fast early recovery phase of the storm when oxygen ions can dominate [Jordanova et al., 2001], and produce the aurora that can be seen from ground and satellite sensors.

[5] This paper presents observations of aurora during the initial recovery phase of the 14–16 July 2000 geomagnetic storm. This prominent event, also referred to as the “Bastille Day storm,” has provided important data and expanded our understanding of the interactions within the Sun-Earth system (for example, see the compilation of papers in Solar Physics, 204(1–2), 2001). In this paper we present observations of the middle and far ultraviolet southern aurora that show increased intensity that we have attributed to the precipitation of oxygen ions during the fast initial recovery phase of this storm. We also show images of the northern aurora obtained during the same time, showing the evolution of oxygen ion emission lines that are approximately conjugate to the emissions seen over the southern pole. We discuss the implications of these observations in the context of the connection between the magnetosphere and the upper atmosphere during geomagnetic storms.

2. Instrumentation

[6] The U. S. Air Force Advanced Research and Global Observation Satellite (ARGOS) was launched on 23 February 1999 into a Sun-synchronous, 98° inclination, circular orbit near 840 km with an ascending node crossing time of approximately 1430 LT. Two experiments on ARGOS were the Ionospheric Spectroscopy and Atmospheric Chemistry (ISAAC) spectrograph and the Low Resolution Airglow and Aurora Spectrograph (LORAAS). These spectrographs were part of an instrument package designed to obtain airglow and auroral measurements across the middle, far, and extreme ultraviolet part of the spectrum from approximately 320 nm down to 50 nm, with instantaneous coverage determined by the operating mode of each particular spectrograph. For the work presented here we used data from ISAAC and LORAAS to examine auroral spectra between 210–250 nm and 80–170 nm, respectively, on 16 July 2000 that includes the beginning of the recovery phase of the 14–16 July 2000 geomagnetic storm.

2.1. ISAAC

[7] The ISAAC instrument is a 1/8 m Ebert-Fastie spectrograph with an image-intensified CCD detector. ISAAC covers a selectable 40 nm instantaneous passband within the 180–320 nm total passband with a spectral resolution of approximately 0.4 nm [Wolfram et al., 1999]. The field of view is 0.03° × 1.1°, corresponding to 1.5 km in altitude and 55 km along the horizon. The spectrograph scans the limb of the Earth in the orbital plane aft of the spacecraft. After approximately 30 s of initially viewing a fixed zenith angle corresponding to a tangent altitude near 350 km, the tangent altitudes of the line of sight changed at a uniform rate from ∼350 km down to 50 km, collecting one spectrum per second and completing the scan in 90 s.

[8] Spectra were obtained in the 210–250 nm passband beginning around 0300 UT on 16 July 2000. Although there are several NO and N2 bands and Rayleigh scattering that dominate the spectrum of the daytime sunlit atmosphere, at night there are no significant emission sources. In the aurora the brightest emissions in this passband are the N II 214.3 nm and O II 247.0 nm lines. For this study we have looked at these two emissions from the southern aurora when the satellite is on the nightside and orbiting toward the dayside, passing on the duskside of the pole due to the inclination of the orbit. This means that as the satellite crosses the southern aurora it is looking aft toward the dark nightside and the observation is solely of the foreground auroral processes. This is a significant advantage in contrast to data taken while the satellite is over the northern pole and looking back toward the bright dayside emissions that obscure the auroral features. Calibration of the data was done using preflight measurements, which is sufficient for our qualitative study. Emission intensities for the N II 214.3 nm and O II 247.0 nm lines were calculated after subtracting background noise from each spectral region.

2.2. LORAAS

[9] LORAAS is a 0.25 m, Wadsworth spectrograph covering the fixed 80–170 nm spectral range at 1.7 nm resolution [McCoy et al., 1994]. The detector is a wedge-and-strip microchannel plate with a CsI photocathode for enhanced sensitivity, with the exception of a masked region across the bright O I 130.4 nm line and part of the H I Lyman-α 121.6 nm line that remained bare. The 0.10° × 2.4° field of view [Budzien et al., 2002] projects to ∼3 km altitude and 120 km along the horizon. A scan mirror is rotated to view the aft limb of the Earth at zenith angles between 100° and 127°, corresponding to tangent point altitudes between 750 km and 50 km. The scan rate is 0.28°/s above 300 km and 0.14°/s below, with one scan completed every 90 s. LORAAS does not include a staring period for these observations, resulting in a higher overall scan frequency.

[10] As with the middle ultraviolet, spectral features in the far and extreme ultraviolet are much brighter in the aurora compared with the nightside, although signatures from radiative recombination become measurable in the intertropical arcs and the hydrogen geocorona exists throughout the nightside. Thus with the exception of H I Lyman α, the signatures measured in the southern aurora are purely auroral features. The spectra were calibrated using prelaunch measurements [Thonnard et al., 1999].

[11] We identified eight primary auroral spectral features and isolated them for this study, although two contain multiple, unresolved components: O II 83.4 nm, O I 98.9 nm, N II 108.5 nm, N I 113.4 nm (with some possible contamination by O I 115.2 nm), H I 121.6 nm, O I 130.4 nm, O I and N2 near 135.6 nm, and N I 149.3 nm. LORAAS has an unfortunate operational bug that causes the pixel locations of each line to shift depending upon the total counts measured by the detector. This causes the O II 83.4 nm line to drift off the recorded area of the detector at moderately high count rates, leading to incomplete profiles. It also slightly reduces the spectral resolution of LORAAS during high count rates and blends lines that would otherwise be resolved, including the O I 115.2 nm and N I 113.4 lines. The exact pixel location of each of the emissions was determined manually by comparing a sequential series of spectra. The background was determined using pixels adjacent to the pixels of interest.

3. Observations

[12] The “Bastille Day storm” began on 14 July 2000 with the radiative phase of an X-class solar flare around 1000 UT followed by a coronal mass ejection (CME) and a large solar proton event. As the CME and interplanetary shock reached Earth late on 15 July 2000, the interplanetary magnetic field (IMF) turned strongly southward and a severe geomagnetic storm began. The 3-hour Ap geomagnetic index reached a maximum value of 400 at 1800–2100 UT, and the Dst index dropped to its minimum value of −300 nT in response to the associated substorm injection. A summary of the storm indices and conditions is shown in Figures 1 and 2. Modeling results have estimated that as much as 60% of the ring current energy was being carried by O+ at the peak of the storm [Jordanova et al., 2001]. Dst began to recover around 0300 UT on 16 July as the ring current began to decay. The recovery occurred with an initial fast recovery phase ending around 0600 UT followed by a slower phase gradually restoring the ring current to normal conditions by early on 18 July. This study focuses on the beginning of this recovery phase, 0100–0800 UT on 16 July, covering the time just before the recovery begins to a few hours after the end of what we consider the initial fast recovery.

Figure 1.

Geomagnetic and solar indices for 14–18 July 2000, including the geomagnetic storm that peaked late on 15 July and early 16 July. Top to bottom: Kp, Dst, and F10.7. The three vertical lines correspond to the times of the three orbits in Figures 35.

Figure 2.

Interplanetary magnetic field (IMF) data from ACE for 15 July (day 197) through 17 July (day 199). The three vertical lines correspond to the times of the three orbits in Figures 35.

3.1. Middle Ultraviolet Emissions: Southern Hemisphere

[13] The O+(2P-4S) doublet at 247.10 nm and 247.11 nm is the brightest feature of the middle ultraviolet (MUV) airglow above 200 km [Cleary et al., 1995]. Only a handful of measurements have been made of this emission [Feldman and Takacs, 1976; Cebula and Feldman, 1984; Cleary et al., 1995; Bucsela et al., 1998], but many experiments have observed the associated emission doublet at 732 nm and 733 nm from the O+(2P-2D) transition [e.g., Walker et al., 1975; Rees et al., 1982; Torr et al., 1990; Fennelly et al., 1993; Broadfoot et al., 1997; Semeter, 2003] that shares the same excited state and thus derives from the same excitation processes. In the sunlight, photoionization and excitation of atomic oxygen is the primary source, while electron impact predominates in auroral regions [Rees et al., 1982]. On the basis of published transition probabilities [Zeippen, 1987; Godefroid and Fischer, 1984] the radiative lifetime of the (2P) state is 5.27 s, with 22% transitioning to the (4S) ground state and emitting a 247 nm photon.

[14] The N+(5S-3P) transition also produces a doublet, at 213.97 nm and 214.35 nm, commonly identified as the N II 214.3 nm emission. This feature was first observed in auroral spectra [Sharp and Rees, 1972; Duysinx and Monfils, 1972] and subsequently in the dayglow [Barth and Steele, 1982]. Photodissociative ionization of N2 dominates the dayglow production of this emission. The most probable source of this emission in the aurora is electron impact dissociative ionization and excitation of N2. The 214.3 nm transition is the only known branch, although discrepancies exist between theoretical and experimental evaluations of the excitation cross section, lifetime of the metastable state, and quenching rates [Meier, 1991, and references therein].

[15] Figures 35 show limb scan measurements of the O II 247.0 nm and N II 214.3 nm emissions observed on three successive orbits over the southern auroral region on 16 July 2000. The times of these orbits are also marked in Figures 1 and 2. The data begin around 0420 UT since ISAAC was measuring a different wavelength range prior to that time, making these the first three orbits that include measurements of the O II and N II emission lines. We present the data that cut through the section of the auroral oval on the nightside of the orbit to avoid the bright dayglow described in section 2. All three figures at each wavelength use the same scale, so changes in intensity can be compared directly. Also shown on each plot are the tangent point altitude of the line of sight (solid line), the approximate altitude of the solar terminator at that tangent point position (dashed-dotted line), and the approximate altitude of the solar terminator at the location of the satellite (dashed line). In calculating these guideline boundaries, we assumed the limb to be opaque at 100 km regardless of dependence on the wavelength and specific atmospheric conditions. The auroral emissions are isolated through most of each segment with some dayglow beginning to appear toward the end of each as the sunlight begins to illuminate the upper atmosphere.

Figure 3.

Intensity of N II 214.3 nm and O II 247.0 nm for 0424–0442 UT measured by ISAAC in the southern auroral oval. Time is plotted in decimal hours. The sawtooth-shaped solid line represents the tangent altitude of the line of sight. The approximate location of the solar terminator above the location of ARGOS (dashed) and above the location of the tangent point (dashed-dotted) are also shown, using some simplifying assumptions about the opacity of the atmospheric limb at UV wavelengths.

Figure 4.

Same as Figure 3, for 0606–0623 UT.

Figure 5.

Same as Figure 3, for 0748–0805 UT.

[16] The most interesting auroral features appear in the O II 247.0 nm emission between 0425 UT and 0435 UT (hour 4.45–4.60 in Figure 3). The emission peaks are nearly twice as bright as adjacent scans and are seen at tangent altitudes above 300 km. No comparable emission is seen in the N II 214.3 nm aurora that appears brightest below 200 km as expected from photoelectron impact on N2. The O II emission also appears earlier in the orbit than the N II emission when the satellite is at lower latitudes. By the next orbit around 0615 UT (hour 6.25), the high-altitude emissions in the O II 247.0 nm aurora have disappeared and the remaining aurora appear reduced by more than a factor of 3. The N II 214.3 nm emission is reduced by nearly the same factor. The final orbit shown, just before 0800 UT, shows the same basic features as the previous orbit with a more moderate decrease in auroral intensity. The sequence shows that the features seen in Figure 3 around 0430 UT were unique to the 247.0 nm emission and did not reappear in either of the two successive orbits.

3.2. Far Ultraviolet Emissions: Southern Hemisphere

[17] We have used simultaneous far ultraviolet (FUV) spectra taken by the LORAAS instrument to help identify the source of these unusual high-altitude auroral O II emissions. Since LORAAS is located on the same satellite as ISAAC, the only difference in the data format is the scan mode unique to each instrument. For LORAAS, each scan begins at higher altitudes and occurs with a higher frequency since the staring mode is not included in the scan. The data are more complete than the MUV measurements because LORAAS had a fixed bandpass and thus was continually monitoring the same emissions. We have analyzed auroral emissions at 108.5, 113.4, and 149.3 nm that are produced by the dissociation of N2 by electrons and at 98.9, 130.4, and 135.6 nm that are normally a product of the impact of energetic electrons on O [Meier et al., 1982]. The 135.6 nm data also include a significant contribution from an N2 Lyman-Birge-Hopfield (LBH) band near 135.4 nm that cannot be distinguished with the resolution of the instrument. At lower altitudes the emission is due entirely to the LBH band, and at higher altitudes it can give rise to as much as half of the observed intensity [Meier et al., 1982]. We note that the O II 834 nm emission also was analyzed, but since complete altitude profiles were not obtained we will not present those results other than to state that the data were consistent with the results presented in the rest of this paper.

[18] The precipitation of oxygen ions creates hot O that can affect oxygen dayglow profiles at 98.9 nm [Cotton et al., 1993b], and so we examined this emission for comparison to the MUV data. Figure 6 shows the 98.9 nm emission for five successive orbits from 0100 UT through 0800 UT, each consisting of approximately 10 limb scans over the southern auroral arc. The data from each orbit have been shifted slightly to match up the magnetic latitudes of each measurement as shown in the last panel. The descending part of each orbit is on the nightside and appears toward the left side of each plot. The salient features we have observed in this series of profiles are the bright emissions as the satellite descends through the southern aurora on the first three passes from 0100 to 0500 UT. The remaining two passes afterward show dim auroral emissions centered on the southernmost point of the orbit that we have attributed to auroral electrons colliding with neutral oxygen. While these electron-induced emissions are also significant contributors to the earlier aurora seen in the first three orbits in Figure 6, enhancements at higher altitudes on the duskside of the auroral oval lead to an asymmetry about the pole that suggests a second source.

Figure 6.

O I 98.9 nm intensity measured by LORAAS during five passes over the southern auroral oval. The tangent altitude of the line of sight is also indicated (dashed line). The bottom panel shows the magnetic latitude for each orbit. The five lines top to bottom match up with the five panels top to bottom, with the solid line referring to the 0059–0117 UT orbit and the long-dashed line referring to the 0748–0805 UT orbit. The data have been shifted in time to produce this symmetry between the orbits.

[19] To support this analysis, we created a similar series of profiles of the N I 149.3 nm emission shown in Figure 7. The initial set of profiles is uniformly bright at low altitudes across the auroral band due the main phase of the storm. Nevertheless, this and each of the successive profiles is centered primarily around the southernmost point of each orbit much like the 98.9 nm data in the last two orbits in Figure 6. Any asymmetry that exists is on the ascending part of the orbit as sunlight illumination begins to play a role. This is most noticeable at the end of the third orbit around 0440 UT, although the 98.9 nm emission also shows similar enhancements at this time and may represent an enhanced patch of auroral electron precipitation. No enhancement is seen on the nightside portion of the orbit. The final two orbits in Figures 6 and 7 show the emissions in approximately the same location, indicative of a return to pure electron aurora.

Figure 7.

N I 149.3 nm intensity in the same format as Figure 6.

[20] Figure 8 shows six of the brightest FUV emissions observed during the 0424–0442 UT time matching Figure 3 and as seen in the third panels of Figures 6 and 7. The O I 98.9 and 130.4 nm emissions have profiles that are similar to each other with the exception of the bright dayglow that appears at 130.4 nm near the end of the pass. The close relationship between these two profiles is expected since the O I 130.4 nm emission includes cascade contributions from the excited state that produces the 98.9 nm emission [Cotton et al., 1993a]. These emissions also match the onset of the emissions seen in the MUV data in Figure 3. The N II 108.5 nm and N I 113.4 nm emissions compare favorably to the N I 149.3 nm emission since all are derived from electron impact on N2. A small signal in the 113.4 nm emission early in the orbit is most likely spectral contamination from the nearby O I 115.2 nm emission that was not completely separated due to a reduction in spectral resolution as described in section 2. The 135.6 nm emission is dominated by the LBH emission at 135.4 nm that compares well to the other emissions with nitrogen sources, although a second source from precipitating oxygen ions can be seen in the double peaked profile seen in the early portions of the orbit around 0428 UT (hour 4.47) that matches the other oxygen-related emissions. Although the results are not presented here, analysis of the H I 121.6 nm Ly-α emission line did not show any enhancements during any of these orbits. This allows us to exclude protons as a source of the auroral emissions at the time and location we have observed.

Figure 8.

Intensity of six FUV spectral features observed by LORAAS from 0424 to 0442 UT (hours 4.40–4.70). Top to bottom, O I 98.9 nm, N II 108.5 nm, N I 113.4 nm, O I 130.4 nm, O I and N2 LBH 135.6 nm, and N I 149.3 nm. The time of this orbit is identical to the time in Figure 3 and the third panels of Figures 6 and 7.

3.3. Extreme Ultraviolet Emissions: Northern Hemisphere

[21] Our final piece of evidence was gathered from observations taken by the Extreme Ultraviolet Imager (EUV) on the Imager for Magetopause-to-Aurora Global Exploration (IMAGE) satellite. IMAGE was launched on 25 March 2000, into an elliptical polar orbit with an apogee altitude of 7.2 Earth radii (45,922 km) and a perigee altitude of 1000 km. During the initial phase of the orbit, including the observation presented here, the apogee was located above the northern polar regions. EUV is designed to observe the He+ 30.4 nm emission in the plasmasphere. However, it has residual sensitivity at wavelengths as long as 80 nm [Sandel et al., 2003] and thus the auroral oval is often visible from the O+ 53.9 nm emission [Burch et al., 2001].

[22] Figures 9a9c shows a series of images taken every 10 min by EUV from near apogee over the northern pole. Each frame has been scaled to a uniform range so that the size of Earth appears constant throughout the sequence. The Sun is to the upper right in each frame and the ionospheric dayglow and inner part of the plasmasphere are visible as the backward-C shaped arc encircling Earth. The aurora is visible starting in the first frame at 0126 UT and persists until ∼0500 UT when it becomes significantly dimmer than the plasmasphere. The aurora includes a bright feature in the premidnight quadrant that becomes significantly brighter at 0440 UT and then rapidly dies out. A comparison to Figure 6 shows a qualitative agreement with the timing of the brightness changes seen in the O I 98.9 nm emission. Although the spatial resolution of EUV is low and it is difficult to pinpoint the location of the MUV and FUV emissions along the line of sight, a rough approximation shows that the brightest emissions around 0430 UT originate from conjugate locations on a field line corresponding to L ∼ 3–4.

Figure 9.

Images of the northern aurora and plasmasphere from the EUV instrument on IMAGE. One frame is taken every 10 or 11 min starting at 0132 UT and ending at 0729 UT. The auroral oval is predominantly from the O+ 53.9 nm emission line. The Sun is to the upper right in the images and the bright oval that fades after approximately 0500 UT occurs near dusk in a location approximately conjugate to the oxygen emissions seen in the MUV and FUV.

Figure 9.

(continued)

Figure 9.

(continued)

4. Discussion

[23] We have presented observations showing oxygen aurora equatorward of emissions induced by auroral electrons during the fast recovery phase of a geomagnetic storm. The emissions appear brightest above 300 km and do not appear to be correlated directly with proton precipitation. The evidence supports the conclusion that these features are produced by a source of oxygen ions from the ring current near L ∼ 4 that precipitate to low altitudes and collide with ambient oxygen atoms and ions. The subsequent charge exchange and excitation leads to the emissions we have observed.

[24] It has been known for some time that oxygen ions of ionospheric origin populate the ring current during a geomagnetic storm [Hamilton et al., 1988]. However, the sequence of events that energizes the ions, injects them into the ring current, and then causes their loss has been studied because of the control they have on the evolution of storms and substorms [Daglis et al., 1999; Jordanova et al., 2001; Kozyra and Liemohn, 2003]. Modeling and observations have identified at least four paths energetic oxygen ions can take once they enter the ring current: (1) they can be lost out the dayside magnetosphere [Fuselier et al., 1991], (2) they can charge exchange with geocoronal hydrogen leading to energetic neutral atoms [Williams, 1987; Daglis et al., 1999] and subsequent low-latitude aurora [Tinsley, 1979; Stephan et al., 2001], (3) they can undergo Coulomb interactions leading to stable auroral red (SAR) arcs [Kozyra, 1997], or (4) they can be scattered into the loss cone and precipitate into the atmosphere [Jordanova et al., 2001]. This last process is the one that is presumed to create the spectral features in the data presented here. Observations from the Comprehensive Energetic Particles and Pitch Angle Distribution (CEPPAD) instrument on the Polar satellite have shown isotropic loss cones for Earthward flowing ions near L = 4 at 0300 UT on 16 July in the northern hemisphere that are only partially filled during a subsequent pass over the southern pole at 2100 UT [Jordanova et al., 2001].

[25] We are unable to resolve Doppler shifts that would determine whether the oxygen emissions we have presented are from the energetic ring current ions or the ambient atmospheric oxygen atoms that collide with these precipitating ions. However, enhancements have been seen equatorward of 60° during a geomagnetic storm in N2 LBH and second positive (2P) bands, N I lines, and O I lines [Ishimoto et al., 1994]. In that study, most of the emissions were found to originate from altitudes between 110 and 130 km, with most of the O I line emissions originating from the primary precipitating oxygen after becoming neutralized through collisions above 300 km. In that context we draw the conclusion that the O II emissions we observe are from these charge exchange collisions that leave behind high-altitude oxygen ions. The precipitating energetic ions become neutralized in the excited core configuration (3D) that produces the 98.9 nm emission. The 130.4 nm emission also can be produced through cascade from the (3D) excited state, although the appearance of the 135.6 nm feature suggests that at least some of those emissions come from either the neutralized ring current ion or the ionized atmospheric oxygen atom after a subsequent recombination with ambient electrons. It also seems apparent that some of these ions mirror before they are neutralized. It is these mirroring ions that create low-altitude ENAs that are seen during the main and recovery phases of the Bastille Day storm [Roelof, 1997; C:Son Brandt et al., 2001] and other similar storms [Mitchell et al., 2003].

[26] From Figure 8 it can be seen that the O I emissions (98.9, 130.4, and 135.6 nm) develop earlier in the orbit while the satellite is still equatorward of 60°S latitude. The N I 149.3 nm and N II 108.5 nm lines have similar histories to each other, developing at lower altitudes and later times when the satellite reaches its poleward extent. The 135.6 nm profile also shows a component that matches that due to the N2 LBH band included in the spectral range. N2 LBH emissions could be excited by precipitating O+ [Mitchell et al., 2003] but our measurements do not have sufficient spectral resolution for us to address this possibility. The N I 113.4 nm profile mostly matches those of the other N I lines, although weak emission seen when the O I lines are peaking may be comparable emissions also induced by the precipitating O+ or may be the O I 115.2 nm emission that has shifted enough to be mistaken for the 113.4 nm line and included when calculating emission intensity.

[27] It is difficult to isolate temporal and spatial changes in the aurora because one satellite pass takes approximately 20 min. From Figures 9a9c and associated data it is clear that the aurora change significantly on timescales of minutes or faster. However, the spectral information in our data is simultaneous and thus the spatial separation of the O I features from the others, in both altitude and latitude, appears to be real and significant. The enhanced auroral emissions appear to be present from the onset of the storm and last for approximately 5 hours. This suggests that oxygen ions with the appropriate energies to create the emissions we have presented are depleted from the ring current after this initial phase of storm recovery. However, although the results from our initial analysis are consistent with the initial fast storm recovery being caused by oxygen ions, a definite link cannot be assumed.

[28] The O+ (2P-4S) transition that produces the 247 nm emission has a counterpart emission from the same excited state near 732–733 nm from O+ (2P-2D) that has been found to have an unexplained peak above 300 km as seen from Sondrestrom, Greenland [Semeter, 2003]. The connection, if any, between those auroral rays and our observations is not clear, but it is interesting to note that they appear after times of extended southward IMF (J. Semeter, personal communication, 2003) that is common in the development of geomagnetic storms, including the Bastille Day storm. The auroral emissions in both cases could be a manifestation of processes scattering O+ into the loss cone, leading to precipitation patterns dictated by the convection of the ring current plasma. Our observations also have implications for the interpretation of storm-time aurora and may provide a means to verify predictions and modeling of the initial fast recovery of geomagnetic storms [Jordanova et al., 2001]. Understanding the complete details of the storm process has important implications for the net balance of terrestrial oxygen [Seki et al., 2001].

5. Summary

[29] We have presented observations of the aurora across the UV spectrum that show an enhancement in atomic and ionic oxygen emissions during the initial fast recovery phase of a major geomagnetic storm. The emissions occur above 300 km and appear equatorward of the electron auroral oval for the 14–16 July 2000 storm. In additions, the emissions are seen in the dusk quadrant of the oval near L ∼ 4. Our interpretation of the data suggests a ring current source of oxygen ions that are scattered into the loss cone and precipitate into the lower atmosphere where they interact with ambient oxygen atoms. This process is thus responsible in part for the recovery of the ring current during a geomagnetic storm, the observations of ENAs at low altitudes during storms, and the observations of subauroral emissions. Implicit in these data is the recognition that the precipitation of oxygen ions may be an important auroral process that must be included with electrons and protons as part of a complex auroral phenomenology.

Acknowledgments

[30] This work is supported by the Office of Naval Research. The authors wish to thank B. Sandel and T. Forrester for their assistance in processing the EUV images.

[31] Arthur Richmond thanks Janet U. Kozyra and another reviewer for their assistance in evaluating this paper.