Observing the magnetosphere through global auroral imaging: 2. Observing techniques



In a companion paper four auroral regions were identified. The source of the first three regions is the plasma sheet, whereas the source of the fourth, the region of Alfvenic auroras, is the ionosphere. It is a primary goal of global auroral imaging to identify these source regions. Space-based imaging can be used to obtain ion and electron, mean energy, and energy flux as a basis for such identification. Measurement of direct emission from precipitating ions or their charge exchange products can be used to determine the ion precipitation characteristics. For electrons, it is necessary to use the atmosphere as a spectrometer. Total precipitated energy can be derived from the luminosity of spectral features where the production cross sections are known. The mean energy of precipitation is inferred from the luminosity height profile deduced from (1) collisional quenching of long lifetime emitters, (2) atmospheric composition, (3) degree of O2 absorption in the UV, or (4) the local atmospheric neutral temperature. There are fundamental advantages in viewing the aurora from space; for example, auroras can be observed in the far ultraviolet range where daylight contamination is much less severe. The various approaches to spaceborne auroral imaging depend on the wavelength selection requirements. UV interferometers show promise of improved light collection efficiency and higher spectral resolution.

1 Introduction

Auroral imaging can be regarded as a way of measuring the properties of auroral precipitating particles through remote sensing. One of the more important tasks is to distinguish the wave-generated electron auroras from the other types of electron auroras (see the companion paper). These auroras can be distinguished by sensing both the mean energy and the total energy flux of the precipitating particles. Although ions also produce light by collisions with atmospheric neutrals, exactly like electrons, more information can be gained from observing emissions generated directly by the precipitating primary particles. For ions, especially for protons, the flux and mean energy can be measured from the intensity of various auroral emission features and from the spectral profile of the emissions of the primary particles. Auroral ion flux and ion energy can be measured through remote sensing from the intensity and Doppler profile of the light emitted by the ions or their charge-exchanged derivatives.

For electrons the measurement of the mean energy of precipitation requires the use of the atmosphere as a spectrometer. The total electron precipitated energy can be obtained by measuring the auroral luminosity in spectral regions, where the excitation mechanism is well understood and the cross sections are known. The mean energy of the electrons can be inferred from the altitude profile of the luminosity. In the companion paper, we have outlined the currently outstanding tasks, which could be accomplished by the application of optical remote-sensing techniques. The remote sensing of the aurora can be performed from the ground by using visible light, which is transmitted through the Earth's atmosphere. Alternately, the aurora can be viewed from space. There are fundamental advantages in viewing from space because we can use ultraviolet in a wavelength range with much less daylight contamination. This makes it possible to study dayside auroras located in the sunlit atmosphere. In addition, viewing from space and using a distant platform allow the observation of the entire auroral region in a single image.

Since the IMAGE project ended in 2005, there has not been satellite-based imagers to make global remote sensing of the aurora. This was mitigated somewhat by the development of the Time History of Events and Macroscale Interactions during Substorms (THEMIS) all-sky imager array [Mende et al., 2008], which produce partial global imaging coverage. This system had several advantages, even over previous spaceborne imagers, because it had higher spatial and temporal resolutions. Like all ground-based optical instruments, the THEMIS all-sky imagers are also unreliable in covering specific geophysical events. The coverage is not truly global, being restricted to the western hemisphere. An even greater problem is the reliance on clear skies. For fully reliable global coverage of the aurora, we should use space-based imagers. In fact, the greatest problem with space-based imaging of the aurora is that we are not doing it anymore.

Ground-based auroral photometry has produced some highly important results in the study of the magnetosphere through observing the aurora; however, to keep the scope of this review reasonable I will focus only on auroral imaging techniques and especially those that might have space-based applications for global imaging. For ions, I will discuss instruments which have high spectral resolution to distinguish the light that is emitted by the precipitating ions or their charge-exchanged derivatives and those that can measure the Doppler profile of these emissions. The various ways of using the atmosphere as a spectrometer for measuring the intensity and energy distribution of the precipitating auroral electron fluxes will be discussed.The different techniques for determining auroral precipitating energy require different spectral resolutions, and previously flown space-based global auroral imagers will be reviewed according to wavelength resolution requirements and their performance.

2 Remote Sensing of Ions

Besides the impact excitation of atmospheric constituents the precipitating ions themselves can produce auroral light. This auroral light created by fast-moving particles has high diagnostic value because the light can be used to monitor the energy and pitch angle distribution of the incoming ions. The most frequently observed ions are protons, which need to first undergo charge exchange and become hydrogen atoms before they can emit light. The fast-moving hydrogen atoms often radiate, and the spectral distribution of the resulting light, including the Doppler shift, can be interpreted in terms of particle velocity. On the IMAGE spacecraft Lyman α emissions were measured and the intensity was mapped. However, no Doppler information was preserved. Being able to obtain the mean energy of the precipitating protons would be a great advantage. Unfortunately, the precipitating proton flux is dependent on the pitch angle diffusion rate, and currently, there are no really reliable models to calculate this.

With ions, it is also possible to measure the intensity of the emission and derive the precipitating flux, provided that the cross sections of the various reactions are known. In addition, it is possible to measure the Doppler shift of the emitted photons and derive the line-of-sight velocity of the emitting particles. So it is intrinsically possible to measure the flux and mean kinetic energy of the precipitating ion fluxes.

A large fraction of energy in the magnetosphere is in the form of particle energy carried by the ions. In most models the ion pressure is assumed to be isotropic, and therefore, field-aligned ions should have the same intrinsic pressure as the trapped population. The bulk of the ions, including protons, are stably trapped in the magnetic field of the closed field line region. Proton precipitation involves a mechanism that continually refills the loss cone. On the nightside closer to the Earth in the region of approximately dipolar field lines, pitch angle diffusion and precipitation are minimal. Further outward, the field lines are more taillike and therefore distorted with larger curvature as they cross the equatorial plane. As the particles cross the highly curved portion of the field lines pitch angle scattering occurs. This is because the ion gyroradius becomes comparable to the radius of curvature of the magnetic field lines [Sergeev et al., 1983]. Particles that drift into the loss cone are precipitated. On the dayside, curvature effects associated with the magnetopause and boundary layers can also cause ion precipitation [Sergeev et al., 1997].

It is also possible to obtain the proton flux from simultaneous observations of Lyman α and several other FUV lines [Strickland et al., 1993; Galand et al., 2002]. Obtaining the mean energy of precipitation is not possible from the Lyman α intensity measurement alone. In fact Frey et al. [2003] show that 1 mW/m2 proton precipitation produces about 3 kR Lyman α for virtually all energy above 8 keV, suggesting that at higher proton energies the precipitated energy flux is proportional to the Lyman α production.

The detection of the various ion lines requires some spectral resolutions to separate them from other auroral or dayglow features. To separate the energetic proton-induced emissions from the very bright geocoronal Lyman α a spectral resolution of about 0.1 nm is needed, which was accomplished on the IMAGE project [Mende et al., 2000b]. Observing lines directly emitted by the precipitating particles is advantageous because of the lack of dayglow at or near the wavelength of the emission line. For example, imaging the Doppler-shifted Lyman α on IMAGE resulted in high contrast auroral images day and night. The IMAGE instrument was on a rotating satellite and had a limited time while it could look at the Earth. Its exposure duration was seriously restricted. A stable platform would allow staring at the Earth, and the same instrument as on IMAGE would produce 24 times higher signal-to-noise ratio than IMAGE at the same cadence, and this gain could be utilized to retrieve Doppler spectral information. Few keV protons would produce a Doppler-shifted component at 0.5 to 0.8 nm away from the zero velocity line. Detecting the Doppler shift of precipitating O or O+ would require about 4 times the resolution to obtain the same Doppler resolution. In the discussion of instruments for global imaging of the aurora instruments that are intended to measure ion Doppler shifts would be in the classified as ultrahigh spectral resolution instrument. For example, the measurement of ions at thermal escape velocity would be very difficult requiring resolutions in the milli-Angstrom range.

3 Remote Sensing of Electrons

Electron flux and energy distribution can be measured remotely by using the atmosphere as a spectrometer. The total energy of the precipitating particles can be measured by combining the known luminous efficiency of the atmospheric constituents and the amount of light emitted by the atmosphere. The energy of the precipitating particles can be obtained by measuring the penetration altitude. A figure from Strickland et al. [1989] was adopted (our Figure 1), which represents the N2+ emission rate, which is closely related to the ionization rate as a function of altitude for Gaussian energy distributions with various mean energies. A large fraction of the emitted light comes from secondary electrons produced by the primary ionizing radiation. The majority of the emitted light will appear to come from the altitude where the ionization is dominant. There are several different ways of finding the penetration altitudes from optical signatures. The altitude of the emissions for electron auroras can be obtained by comparing the intensity of two emission features: (1) from the same or from different atmospheric species with different collisional quenching lifetimes, (2) from two different atmospheric species that have different altitude distributions, (3) from the same or different atmospheric species subject to different degrees of O2 absorption in the UV, or (4) from the same molecular band system to obtain the local atmospheric neutral temperature. These methods are discussed individually.

Figure 1.

N2+ emission rates are closely related to total ionization rates relevant to auroral secondary electron production rate as a function of altitude [Strickland et al., 1989].

3.1 Collisional Quenching of Long Lifetime Emitters

When energetic auroral particles collide with atmospheric constituents they produce molecules or atoms in an excited upper state. Depending on the lifetime of the excited state and the time between collisions, the upper state will either emit a photon or will be collisionally deexcited or quenched without emitting a photon. The process is called collisional quenching, and it reduces the intensity of the emission. The rate of collisional quenching strongly depends on the mean time between collisions, which is presented as function of altitude in Figure 2. By measuring the rate of collisional quenching, the altitude of the emission can be estimated. In practice, two auroral emissions are chosen: a forbidden transition with a longer lifetime and a permitted transition with a short (~10−8 s) lifetime. The quenching of the latter is negligible at auroral altitudes. The ratio of the emission intensities between the two should provide the data for deriving the altitude of the emission, and therefore the depth of the electron penetration. Although it is preferable to chose the same atmospheric constituent and compare two spectral features of different quenching lifetimes, often two different atmospheric constituents are chosen and the results may be subject to more errors due to the variability of the altitude distribution of the two constituents.

Figure 2.

Time between collisions in the neutral atmosphere. When the lifetime of the excited state becomes comparable to the time between collisions, then the auroral emission is quenched. The 630.0 nm with a lifetime of 115 s is seriously quenched below 200 km.

The 630.0 and 557.7 nm emissions of OI are both “forbidden” transitions having long lifetimes of about 110 s and 0.7 s, respectively [Vallance-Jones, 1974] compared to permitted transitions such as the first negative band of N2+ at 427.8 nm with a short lifetime of 10−8 s. Measuring the volume emission ratio of these long lifetime emissions and the 427.8 nm provides the local quenching rate for 630.0 or 557.7 nm, and consequently, the altitude of emission can be retrieved. Judge [1972] computed a model from the emission ratios of atomic oxygen lines at 557.7 and 630.0 nm and N2+ bands at 427.8 and 391.4 nm. In these calculations of the emission profiles, a double Maxwellian electron distribution was used, and the mean energy of one Maxwellian was varied while the other was kept to <100 eV. The column integrated emission rates were calculated, and the different column emission ratios were given as a function of energy. It is known that the emission ratio 630.0/557.7 has a value well over 1 for low-energy electrons [Eather and Mende, 1971; McEwen and Sivjee, 1972] and when the ratio is less than 1 the energy is higher than 1 keV. Rees and Luckey [1974] published their model for emission ratios of 630.0/427.8, 557.7/427.8, and 630.0/557.7 as a function of a characteristic energy of the precipiating electrons although it had been in use for several years [Eather and Mende, 1972, 1973], Rees and Luckey used a time-dependent model to calculate the steady state emission products and included the reaction rates and a J71 model atmosphere with 1000 K exospheric temperature [Jones and Rees, 1973; Rees and Jones, 1973]. Vallance-Jones [1975] produced another model for emission ratios (630.0 + 636.4)/427.8 and 557.7/427.8 as a function of 427.8 nm intensity and characteristic electron energy. He used different model atmospheres and found that the result depended on O/N2. Mende and Eather [1975] measured proton and airglow corrected intensity ratios 630.0/427.8 and 557.7/427.8 with an airborne zenith photometer taking one data point every 2 min. Although the 630.0/427.8 ratio cannot be applied to fast-moving or quickly changing auroras, they were able to obtain a measure of the energy well applicable to the various precipitation regions. Eather and Mende [1972] also found that, in general, auroral intensities were closely related to the mean energy of the electrons as if the particle population of the source region (plasma sheet) density was relatively constant and the acceleration mechanism was responsible for producing the higher energy and consequently the higher auroral intensity. So for weak auroras (<1 kR), usually caused by low-energy electrons, which do not penetrate to 557.7 nm quenching altitudes, the emission ratio of 557.7/427.8 was rather constant. High-energy electrons usually produce higher intensity auroras penetrating deeper into the atmosphere, and the collisional quenching of 0(1S) states reduce the ratio 557.7/427.8 at lower altitudes. However, the production of 557.7 nm is quite complex, with several production mechanisms acting at various altitudes, such as dissociative recombination of O2+, direct electron excitation of atomic O, dissociation of O2 molecules, and chemical reaction producing the 1S state; therefore, using the ratio of 557.7 nm emission, say to 427.8 nm, has been out of favor. Nevertheless, this emission ratio has been applied to the energy estimation of auroras that penetrate deeply into the atmosphere, such as pulsating aurorae [Brekke and Hemiksen, 1972 ; Sawchuk and Anger, 1976a] and also type B aurorae [Sawchuk and Anger, 1976b; Boyd et al., 1971]. More commonly, the emission ratio between 630.0 nm and 427.8 is used. This has been modeled as both a function of the electron energy parameter and the O/N2 ratio [Strickland et al., 1989].

An early example of monochromatic imaging in 630.0 nm and 427.8 nm is included in Figure 3, and it demonstrates the variable nature of the ratio between the two emissions. The image shows two east-west aligned arcs. The northern (top) is generated by soft electrons and shows up only at the high altitude as unquenched 630.0 nm (shown as red). The amount of 427.8 nm (shown in white) emitted here is minimal in this arc. The southern arc, however, at the bottom of the picture shows a whitish lower border, denoting that this aurora was emitted at lower altitude where the 427.8 nm (blue-white) was intense and the 630.0.0 nm was quenched.

Figure 3.

Color monochromatic TV image composite of 630.0 nm in red and 427.8 nm shown in white/violet from a monochromatic all-sky imager employing a filter wheel [Mende and Eather, 1976]. North is up, east to the left, and west to the right. The auroral arc at the bottom is near the southern horizon with soft precipitation higher up in red and whitish intensification nearer to the horizon at lower altitudes.

The image in Figure 3 was taken by a monochromatic all-sky imager, which was developed in the early 1970 [Mende et al., 1977]. This instrument adopted the use of telecentric filtering of the image. In an all-sky camera (fish eye) optics the rays have a very large converging angle as they arrive at the input optics. Interference filters, however, require near parallel light to produce an effectively filtered beam. This type of optical system first produces an image in which the central rays at each image point are parallel to the axis. This is called a telecentric image, and all image points can be filtered uniformly by filtering the rays at this point in the optical train. Using this technique, very wide field all-sky cameras can be constructed with relatively narrow wavelength filtering. This is the basis for most monochromatic all-sky cameras used today in aeronomy research.

The effectiveness of the 630.0 to 427.8 ratios method has been thoroughly tested; for example, Kaila [1989] compared the altitude derived from the emission ratios to auroral triangulations from two all-sky cameras and obtained relatively good agreement. There are several shortcomings of the 630.0/427.8 ratio technique; for example, (1) 630.0 nm is produced by several mechanisms in the aurora, and it is also produced by chemical (airglow) reactions; (2) background 630.0 nm contamination caused by high-temperature ionospheric electrons; and (3) the long 630.0 lifetime of 115 s demands the observation of a steady state aurora, thus restricting the technique to “time stationary” auroras.

In the above discussions the first negative band of N2+ either at 391.4 nm or 427.8 nm was the favored spectral feature for the measurement of the precipitated energy. Recent advances in photoconductive solid state detectors greatly increased the efficiency of detectors in the red spectral regions, and today, it is customary to use component lines of the first positive band of N2 in the red spectral region for measuring the precipitated energy.

Another emission used in remote sensing of the energy of the electrons is the O+(2P) emission at 732 nm [e.g., Dahlgren et al., 2008]. The lifetime of the parent state of the emission is about 5 s [Vallance-Jones, 1974]. Theoretically, this would be a good diagnostic for low-energy electrons, permitting accurate measurement in slowly varying aurorae. There are several issues, however, that make this less attractive. It is a relatively faint emission compared to the others mentioned above requiring a higher-performance instrument. Its wavelength is close to an OH airglow line and to one of the components of the fist positive band system of N2. It either has to be separated spectrally or the first positive band intensity of N2 has to be monitored separately [Spry et al., 2014] to determine the intensity and remove the contamination.

Another issue is that the published source model consists of electron impact on atomic oxygen and Semeter [2003] showed strong evidence that O+(2P) may also be excited directly from ambient O+(4S) ions in the high-altitude ionosphere.

3.2 Composition-Altitude Distribution of Two Different Emitting Species

Another way to obtain the mean energy of electron precipitation remotely is to select auroral emissions that are characteristics of a particular atmospheric constituent (N2, O2, O, and O+), whose altitude distributions are significantly different from each other so that the emission ratios can be interpreted as a change in the altitude of the electron penetration depth. This method is fundamentally different from the first method because short emission lifetimes can be chosen for both emissions allowing high time resolution measurements of the auroral energy.

Figure 4 shows the various main constituent profiles in the atmosphere. In diffusive equilibrium the density profile depends mainly on the molecular mass of the constituent. The N2 and O2 profiles are somewhat similar, whereas atomic oxygen has a very different altitude profile, dominating the mixture above 190 km, and becoming a minor constituent below 110 km. For example, at high altitudes the emissions from the direct excitation of atomic O are likely to be more important, whereas at low altitudes molecular excitation of N2 or O2 seems to be key. However, molecular dissociation commonly occurs in aurora, and it is possible to get atomic O emissions even from low altitudes when O2 molecules dissociate.

Figure 4.

MSIS atmospheric density composition model.

The Auroral Structure and Kinetics (ASK) experiment [Dahlgren et al., 2008, 2015] attempts to take advantage of both quenching and atmospheric composition. This ground-based optical instrument consists of three highly sensitive narrow field-of-view (FOV) identical imagers pointed at magnetic zenith. Lanchester et al. [2009] calculated the altitude profile of three wavelengths as a function of precipitating electron energy (Figure 5). They had three wavelength channels: N2 673 nm for low-altitude energetic electrons, 732 nm O+ emission (lifetime of ~5 s) for measurements of softer electrons, and in channel 3 they imaged 777.4 nm emission from O from higher altitudes. The 777.4 nm emission can be produced through dissociative excitation of molecular oxygen in the E region [Hecht et al., 1985; Lanchester et al., 2009] or even low down in the troposphere by lightning [Mende et al., 2005]. As long as the auroral electrons are soft and do not reach down to regions where O2 molecules are dominant, it is a good auroral energy discriminator for low-energy precipitating electrons. Observing the N2 band at 673.0 may also be a good way of monitoring the unwanted 777.4 nm contribution from low-altitude O2 dissociative excitation.

Figure 5.

Calculated altitude distribution of the emissions used in the ASK experiment after Lanchester et al. [2009]. N2 673.0 nm is used as the low-altitude monitor of high-energy electrons.

In Figure 6, an example of imaging soft aurora is shown from the ASK instrument [Dahlgren et al., 2008, 2015]. There are two image sets taken at given times, and in both, the absence of 562.0 nm O2 emission coming from low altitudes is evident. This shows that the electron precipitation is soft. It is interesting to note the large difference in the appearance of the structure of the aurora between the 732 nm and the 777.4 nm channels.

Figure 6.

Simultaneous imaging by the ASK [Dahlgren et al., 2008] of a soft aurora showing the absence of (left) 562.0 O2 emission (left), which would be coming from low altitudes and the presence of emission from (middle) O+ and (right) O from higher altitudes.

3.3 O2 Absorption in the UV at Lower Altitudes

In space-based imaging, the most commonly used optical technique for remote sensing of the mean energy of auroral electrons is the measurement of O2 absorption in a far-ultraviolet Lyman-Birge-Hopfield (LBH) bands of N2. Referring to Figure 4, the O2 density decreases rapidly with altitude. O2 is also an absorber of FUV wavelength radiation, and this process is wavelength-dependent; the absoption maximizes at shorter wavelengths [Strickland et al., 1983]. Thus, if we have two FUV emissions of different wavelengths produced by the same production mechanisms, but differently absorbed by O2, then the ratio between the two intensity measurements provides information about the extent of the atmospheric O2 absorption along the observing path. This can be used to estimate the depth of the auroral primary electron penetration [Germany et al., 1997]. This technique was used extensively in the data analysis of the Ultra Violet Imager (UVI), flown on the NASA Polar Satellite [Torr et al., 1995]. This instrument will be mentioned later and shown in Figure 13. The UVI on Polar had a filter wheel, and it took sequential images of a wavelength range from 140 to 160 nm (called LBHl) and another region from 160 to 180 nm (LBHh). In both wavelength regions, the LBH bands of N2 produce most of the emissions. Comparison of the two images provided data on the O2 absorption between the source and the satellite, which allows computation of the altitude of emission, hence the electron energy [Germany et al., 1997].

This technique was validated by direct comparison with in situ particle measurements. Figure 7, reproduced from Germany et al. [1997], shows the derived energy maps (bottom) from the UVI observations. A bright line shown in the dusk sector represents the ground track of the Defense Meteorological Satellite Program (DMSP). In Figure 7 (top), the solid curves are DMSP measurements and the crosses are UVI pixel intensity values. The DMSP average energy values are shown for fluxes >0.1 mW/m2. The 1 s DMSP data have been averaged to match the 37 s integration period of the UVI images. The vertical lines represent the beginning and end of the UVI image integration times. The mean energy profile composite was derived from taking ratios of the two LBH images and interpreting the O2 absorption and associated penetration energy. The dashed lines represent the one sigma measurement uncertainty in the UV data.

Figure 7.

(top) The DMSP particle measurements (solid line) and plots of energy flux and mean energy of electrons derived from Polar Ultra Violet Imager (UVI) satellite images (crosses) [Germany et al., 1997]. (bottom) Two images represent the incident energy flux (Eflux) and average energy (Eavg) derived from the long wavelength LBH filter (LBHl, 21:43:20 UT) and the short wavelength filter (LBHs, 21:42:06 UT). The ground track of the DMSP FI2 spacecraft is shown on the maps.

The data show good agreement between derived average energy and DMSP data [Germany et al., 1997]. Nevertheless, this technique also suffers from having relatively poor energy resolution near the critical few keV energy region. Another short coming of the technique is that it relies on a model of the O2 concentration in the atmosphere, and during disturbed conditions, characteristic of periods of geophysically interesting magnetic events, the modeled atmospheric constituents may not be reliable.

3.4 Neutral Temperature

The most often used technique for determining mean energy of precipitating electrons is collisional quenching of long lifetime emitters. Unfortunately, the applicability of this technique to rapidly varying auroras, demanding high time resolution, is questionable. High time resolution is becoming particularly relevant because of the interest in Alfvenic auroras that are highly dynamic. These auroras are usually produced by soft (<2 keV) electrons emitted at higher altitudes where the long lifetime emissions are the key emitting features and the extended decay of these emitters prevents the reproduction of the time variation of the precipitating particle fluxes. The author proposes a way to complement the collisional quenching technique by spectral measurement of the rotational temperature of the emitting atmospheric species. In the spectral distribution of a molecular band, the intensity ratio of the various branches is a function of the rotational temperature of the molecule. Thus, measurement of two relatively broad spectral regions in the spectrum of a molecular band and the ratio of the two intensities defines the rotational temperature of the emitting molecules. Depending on the penetration altitude of the auroral electrons, the rotational temperature of the emitting gas varies. The neutral temperature in the atmosphere changes greatly with altitude in the region between 100 and 140 km, see Figure 8a. In principle, measuring the temperature provides the altitude of emissions, thus the mean energy of electron penetration. The range from 100 to 140 km is the key altitude range where few keV auroral energy electrons stop. Turgeon and Shepherd [1962] and then Hilliard and Shepherd [1966] used the atmospheric temperature as an energy estimator, and they obtained the temperature by measuring the atomic line width with an interferometer. Measurement of the molecular rotational temperature via the intensity ratio of two branches of a band is a much less challenging measurement. Furthermore, the molecular bands in the aurora can be permitted emissions with very short lifetimes (10−8 s) intrinsically suitable for documenting highly dynamic auroras. Unfortunately, inverting the temperature measurement into altitude is not straightforward because auroral bombardment causes heating. We expect a slow temperature rise due to auroral heating, and the quiet time profile shown in Figure 8a may not remain applicable. This means that the molecular temperature method alone will not provide accurate altitudes on a one-to-one basis. However, if the proposed rotational temperature technique is used simultaneously with the quenching method by measuring an additional long lifetime emission then from the long-term averaged mean energy derived from the quenching method can be used to correct the long-term averaged mean energy derived from the rotational temperatures. This way, the rotational temperature determined mean energy can be normalized regardless of auroral heating. Thus, by combining the long-term averaged and the dynamically varying molecular temperature method one can monitor auroral mean electron energy variation in highly dynamic auroras.

Figure 8.

(a) MSIS model of the quiet time neutral temperature of the atmosphere. (b) N2+ first negative band intensity as a function of wavelength for various rotational temperatures. The wavelength bands of the two channels are shown by the dashed curves.

It should be noted that as always, the geometry of the rotational temperature observations is important. Measuring the temperatures would be simple if all the emission originated from a region of single temperature, for example, when a single auroral arc is observed from the side provided the line of sight does not include several auroral forms at different range distances and at various altitudes. Rotational temperatures in different auroral forms have been published in several papers [Vallance Jones et al., 1987; Holma et al., 2000]. Using photometers and comparing them to incoherent scatter radar, Holma et al. [2000] validated the technique by showing that when the aurora intensifies the electron energy increases, the emission altitude and the rotational temperatures both become lower.

The first negative band of N2+ at 427.8 nm is a good candidate emission for measuring the rotational temperature, because it has temperature sensitive branches (see Figure 8b), and a two-channel imager with modest wavelength resolution (~1.5 nm) is capable of recording the temperature change with altitude.

Figure 8b shows the model calculations of the intensity of the first negative band of N2+ at 427.8 nm for different rotational temperatures as a function of wavelength with two filter passbands represented with dashed lines. It can be seen from Figure 8b that the intensity ratio of the two bands depends strongly on the rotational temperature, an increase in the rotational temperature, for example, would increase the ratio of the 425 to 427 nm bands. Conversely, the ratio of intensity in the two bands can be used to estimate the rotational temperature.

In Figure 9 data are presented which were taken by a dual-wavelength channel imager that observed an auroral arc viewed from the side with an exposure cadence of 1 s from Poker Flats, Alaska. The spectral profiles of the two imager channels were the same as the dashed curves on Figure 8b. Figure 9 shows vertical intensity profiles of the two channels from images viewing an arc near the horizon. The data were background-corrected and plotted as a function of vertical pixel number; the 427 nm wavelength channel is illustrated in red and 425 nm in green. The curves show that the emission ratios are higher at low altitudes representing the colder atmosphere. As expected at higher altitude, the intensity in the low-wavelength channel significantly intensifies relative to the other channel, showing that the ratios diminish with altitude because the temperature is higher.

Figure 9.

Relative intensity plot of the two wavelength imager as a function of vertical pixel number. The 427 nm channel shown in red and the 425 nm channel in green. At higher altitudes (higher pixel numbers) green has higher intensity than red showing temperature increase with altitude.

These data show that the method for rotational temperature measurement of the first negative bands of N2+ in auroras is fundamentally feasible and 1 s cadence time resolution was achievable for a modest size split field dual imager in an average brightness aurora. However, great deal more development work is required for an operational ground-based imager using the Neutral Temperature method and for the modeling to separate the temperature rise due to auroral heating.

Since the first negative bands of N2+ lie in the near UV/visible, this kind of observation is only possible when the atmosphere is in absolute darkness. Space-based quantitative auroral observations in visible light would have to be corrected for reflected Earth albedo when looking down, even during moonless nighttime. It is usually advantageous to use the FUV wavelength range where the underlying O2 atmosphere absorbs all the background light from below. In the FUV region, the candidate rotational band system would be the LBH bands in the FUV. In fact, rotational temperature of LBH N2 in dayglow will be used to monitor daytime midlatitude thermospheric temperature by the space satellite program called The Global-scale Observations of the Limb and Disk (GOLD) [Eastes et al., 2015]. A figure from Krywonos et al. [2012] was reproduced as Figure 10 showing a part of the LBH spectral band as a function of different rotational temperatures. The proposed use of the LBH rotational temperature for the GOLD program is for looking at atmospheric temperatures in the dayglow. We suggest here that the same technique could be used in the aurora for electron energy discrimination without the interference due to Earth albedo. This proposal has clear advantages for space-based observations at night. Under sunlit conditions detecting the rotational temperature due to the auroral excitation might be more challenging because of the presence of the dayglow. The dayglow is the optical signal due to excitation of the N2 molecules by locally produced photoelectrons. The signal due to dayglow would carry the rotational temperature characteristics of the bulk illuminated atmosphere, and it would mix with the signal produced by the auroral electrons and therefore would not be an appropriate representation of auroral heights.

Figure 10.

LBH spectral profile as a function of rotational temperature [Krywonos et al., 2012].

4 Instruments for Global Imaging of the Aurora

Spaceborne auroral imagers have made substantial impact in the field of auroral physics. They have allowed viewing of the aurora from a global perspective and thus have removed the space-time ambiguity, a major handicap in all in situ type spaceborne observations. There have been many spacecraft-based global-scale auroral observations, and the most successful ones were operating in the FUV wavelength [Frank et al., 1981; Cogger et al., 1991]. The major advantage is that in the FUV, the Sun's intensity is lower by several orders of magnitude, permitting the observation of the aurora even in the daylit atmosphere. Another key feature of FUV observations is that the atmosphere below the aurora is opaque in the FUV. This is because the O2 bands in the lower atmosphere absorb the FUV radiation, and no corrections are needed to account for Earth's albedo contamination. Even restricting our visible light observations to “nighttime,” global viewing of the aurora can still be very challenging from a high-altitude platform, because some part of the Earth under the satellite can be sunlit. This sunlit part will be a high-intensity undesirable light source near the instrument FOV. As a comparison, consider trying to view a luminous object as faint as an aurora located on the dark part of the Moon's disk when the disk is partially sunlit.

For effective global imaging, we want to observe the aurora in all conditions, including daytime when the atmosphere is sunlit. Thus, an ultraviolet instrument has to be highly “solar blind” to suppress the near-ultraviolet and visible sunlight. FUV detectors can be built to be solar blind to a factor 106. However, sunlit background rejection of 108 is needed to avoid significant contamination. To achieve this, we need an additional visible light suppression of >100 to be provided by the optical system preceding the solar blind detector.

In addition to solar blindness, we need to understand spectral selectivity requirement so that we can classify the types of space-based instruments according to their spectral discrimination. Based somewhat on prior flight instruments we can consider the following types:

  1. Coarse: Broadband UV imaging with visible contamination rejection (20–30 nm, e.g., IMAGE Wideband Imaging Camera (WIC)).
  2. Moderate: Some spectral resolution is needed because several spectral regions in the FUV range are measured to quantify O2 absorption and make electron energy measurements. For example, measuring LBH high and low (wavelength discrimination of ~15 nm e.g., POLAR UVI).
  3. Medium high: The instrument is designed to perform quantitative aeronomy and to discriminate between optically thin and thick components of OI at 135.6 and 130.4 nm (2–3 nm, e.g., Ionospheric CONnection (ICON) FUV).
  4. High: Intended to measure Doppler shift of Lyman α or LBH rotational temperature (0.1 or 0.2 nm, e.g., IMAGE Spectrographic Imager (SI), GOLD).
  5. Ultrahigh: Usually, an interferometric instrument to measure Doppler shift of ions, neutral winds, etc. (from 0.1 to 0.0001 nm).

4.1 Coarse Spectral Resolution Spaceborne Imagers (20–30 nm, e.g., iMAGE WIC)

A typical example of this type of instrument was the Wideband Imaging Camera (WIC) on the NASA IMAGE spacecraft [Mende et al., 2000b]. The WIC mirror substrates were originally manufactured for the Viking imager [Anger et al., 1987]. In Canada, however, unlike the Viking model, WIC had multilayer mirrors that had good reflectivity in the FUV but had poor reflectivity in visible light. The transmitted visible light was lost because it was mostly absorbed by the blackened mirror substrate behind the mirrors. There were two mirrors, the primary and the secondary, with visible reflectivity of less than 4% each, giving a total visible reflectivity of <1.6 × 10−3. This was adequate to make good images of the aurora even at midday when the solar contamination was at maximum. Unfortunately, as the instrument aged, the visible rejection was degraded. The degradation was caused by an increase in the photocathode sensitivity to visible light due to a process of building up traps in the photocathode base material, thus reducing solar blindness. Traps are intermediate energy states in the photo cathode material, which can release photoelectrons even if the incoming photons have less energy than the band gap of the photocathode material. Instruments with higher visible/near-UV rejection like Polar UVI and IMAGE SI did not show this effect, suggesting that the problem can be eliminated by improved suppression of the unwanted visible/medium-UV light. This supports the views that the creation of traps in the photocathode is largely due to visible/near-UV light. After 5 years in orbit, WIC started to show images in which visible features like the outlines of the Antarctic continent were detectible. It is likely that greater suppression of the visible and medium-ultraviolet sunlight in WIC like in Polar UVI and IMAGE SI would have inhibited trap building and WIC's solar blindness would have been prolonged.

A cross-sectional schematic of the WIC is shown in Figure 11. In Figure 12, we show a photograph of the instrument. The detector on WIC was an open detector, and it operated in space in open vacuum conditions. On the ground during tests, the detector had to be evacuated. The plumbing associated with the evacuation of the WIC detector is shown in Figure 12.

Figure 11.

Schematic of WIC. WIC was a two-mirror concentric camera of moderate spectral selectivity. WIC had an open detector which had to be under continuous purge with dry nitrogen for storage.

Figure 12.

IMAGE Wideband Imaging Camera (WIC) is shown with detector evacuating plumbing attached.

4.2 Moderate Resolution Far-Ultraviolet Auroral Imagers (15 nm, e.g., Polar UVI)

As the number of reflectors increase, the solar rejection and spectral resolution can be improved. One of the best examples of moderate-resolution ultraviolet imaging is the Ultraviolet Imager (UVI) flown on the Polar satellite [Torr et al., 1995]. This was a reflective camera with reflective UV filters. The filter system was made by using a number of multilayer reflectors at 45° bouncing the light back and forth. This is shown only schematically as two parallel vertical straight lines on Figure 13. UVI had the capability of selecting various broadbands in the ultraviolet, as shown in Figure 14. By increasing the number of reflections with the four multilayer reflectors in its filter, Polar UVI was able to improve solar rejection and increase its spectral resolution. The most useful feature of Polar UVI was its capability to select two spectral regions for the measurement of the N2 LBH bands. It had sufficient spectral resolution for the O2 absorption technique described above. Unfortunately, the state-of-the-art multilayer filter still could not provide clear separation of the highly desirable atmospheric diagnostic feature at 135.6 nm from the optically thick 130.4 nm lines of OI. To accomplish that task, more accurate filtering was needed. For the IMAGE and the ICON application, a grating instrument, the Spectrographic Imager class of instrument was developed, which meets the spectral resolution requirement and retains high-quality aurora and airglow imaging [Mende et al., 2000b].

Figure 13.

Optical diagram of the Polar UVI camera.

Figure 14.

Passbands of the four spectral filters of the UVI instrument.

4.3 Medium-High-Resolution Auroral and Airglow Imagers (2–3 nm, e.g., ICON FUV)

The OI emission line at 135.6 nm is a relatively bright emission line which is usually optically thin. It can be used to invert the volume emission rate distribution of the emitting optical OI species. The 130.4 nm emission is mostly optically thick in aurora and dayglow, and its presence in the 135.6 nm band represents a contamination. Therefore, it is quite important for a 135.6 nm imager to effectively suppress 130.4 nm in the 135.6 nm channel. One technique of suppression is the use of a heated salt filter, which transmits 135.6 nm and strongly absorbs 130.4 nm, as long as its is heated to about 100°C. The other approach is to use grating spectroscopic discrimination. To make the spectral selection, spectral resolution of about 4 nm is needed. An instrument that has this capability is the FUV spectrographic imager, which is being built for the NASA Ionospheric CONnection (ICON) Explorer satellite.

The grating spectrographic imager technique can provide higher spectral resolution than multilayer spectral filtering, and the same principle was applied in the separation of geocoronal Lyman α from the Doppler-shifted auroral Lyman α in the FUV Spectrographic Imager on the NASA IMAGE spacecraft [Mende et al., 2000b]. For that application, a spectral resolution of about 0.1 nm was required.

Figure 15 illustrates a grating spectrometer for spectrally selective imaging. Figure 15 (top) shows an Imaging Spectrometer, which is the “conventional” approach. In this type of instrument, the object (in our case the atmosphere) is imaged on the slit by some telescope optics. For clarity of understanding, we are using lenses familiar to most readers to illustrate the principle, but in most spacecraft optical systems operating in the UV we would use a mirror or a combination of mirrors because there are very few suitable transmissive materials in the FUV, which work well as lenses. Part of the image passes through the slit, and the light is collimated or made parallel by another optical component, illustrated here as a lens. The collimated (parallel) light passes through the dispersive element, shown here as a transmission grating (again, it would be a reflective grating in the FUV). The grating separates and deflects the light according to wavelength, and the images of the entrance slit (illustrated by an upside down arrow) are formed by the third lens in each wavelength passed by the entrance slit. Thus, the final image consist of multiple images of the entrance slit displaced in the left-to-right direction according to the wavelengths that have passed through the spectrometer. The important number to derive is the etendue or photometric efficiency, which can be derived from the F number of the lens combined with the area of the entrance slit. The slit width is determined by the focal length times the angular slit width. The same way, the slit length is the product of the focal length and the angular slit length or the field angle of the lens shown in the figure. In this type of instrument, the spectral selection component, namely, the grating, is effectively at the pupil of the instrument. The focal plane for the imaging and the spectral separation are coupled because they are collocated. Thus, the final image is a convolution of the spatial and spectral imaging properties of the instrument. Usually, in such instrument, a 2-D image is formed on the detector; in one dimension the light is dispersed according to wavelength, while the instrument preserves one-dimensional spatial imaging in the other dimension parallel with the slit. This is the fundamental concept used on many space-based ultraviolet instruments, such as the Global Ultra Violet Imager flown on the NASA Thermosphere Ionosphere Mesosphere Energetics and Dynamics satellite.

Figure 15.

The two commonly used approaches for making high-resolution images of the aurora or the atmosphere. The top illustration is an imaging spectrometer, while the bottom is the spectrographic imager.

Figure 15 (bottom) is the spectrographic imager. In this concept, the entrance slit is the first element. The first optical component after the slit is the collimator lens (again, this is an illustration of the principle using lenses). In this configuration, the collimator optical element is about one focal length distant from the grating. Because of this arrangement, the collimating lens performs a second task by forming a two-dimensional real image of the scene at or near the grating. This way, the two-dimensional imaging is separated from the one-dimensional imaging for spectral separation. The lens following the grating produces an image of the entrance slit at the exit slit plane, forming a conventional spectrometer between the entrance and exit slits. This lens also participates in the two-dimensional image formation on the detector. This is accomplished by reimaging the intermediate image formed on the grating by this lens and the last lens located behind the exit slit. This last lens collects only the light that passes through the exit slit, and therefore, the light is wavelength filtered by the dispersing properties of the spectrometer. The spectrographic imager is therefore a real 2-D imaging camera, which is filtered because it responds only to light in the selected wavelength band of the spectrometer.

The two concepts, imaging spectrometers and spectrographic imagers, can be generalized. It is possible to apply the two approaches to all kind of grating spectrometers, as well as to interferometers. In a spectrometer, the dispersing element is the grating, which operates in combination with a slit. The slit can be regarded as a mask, selecting only certain angles of the field or image. But the same holds true with an interferometer such as a Fabry-Perot, in which the dispersing element is the etalon, again used in conjunction with a mask that allows only certain angles of the field to pass. The main point is that all of these devices can be used either in the imaging spectrograph mode where the wavelength variation and the spatial imaging form a convolution.The spectral dispersion and the imaging are in “quadrature,” i. e. separate and independent of each other, in a spectrographic imager type of instrument.

In Figure 16, we present a schematic illustration of a dual-wavelength channel Spectrographic Imager of the type described above. The diagram provides two illustrations of the raypaths through the same instrument. The top is the raypath shown for spectral selection. Light enters in at the slit and is collimated by the collimator optics that are placed one focal distance behind the entrance slit. The parallel light at the grating is dispersed according to wavelength. By having two separate exit slits it is possible to have two separate wavelength channels with a single instrument. In Figure 16, the red wavelength is illustrated going up and the blue wavelength going down. After passing the grating, the camera lens focuses the parallel blue light of wavelength λ1 into the lower exit slit, while the red light of wavelength λ2 is focused into the upper slit. The blue light and red light reach separate detectors. In terms of the upper diagram, the instrument can be regarded as a conventional spectrograph with two exit slits. The lower illustration of Figure 16 represents the same optical train but shows how imaging takes place. From each distant object point in the scene, parallel light enters the entrance slit. The collimator lens, which is placed one focal distance in front of the grating, focuses the parallel light on the grating, thus forming an intermediate image at the grating. The “camera lens,” following the grating in combination with the small lenses placed behind each exit slit reimage the intermediate image on the detector. This instrument therefore produces 2-D spectrally filtered images of the same scene on two detectors simultaneously. These instruments were introduced in the ultraviolet for space use where narrowband filtering with multilayer filters would have been otherwise problematic.

Figure 16.

Schematic of the optical principle of a medium high spectral resolution spectrographic imager designed to accept two wavelength channels.

An application of the spectrographic imager is the ICON FUV Spectrographic Imager, which was constructed, tested, and integrated on the ICON payload. In all FUV applications, transmissive optics are not usable and all the optical elements must be reflective. It is relatively straightforward to produce two-channel versions of these instruments, and more channels could be incorporated with ingenious layout designs.

In the ICON adaptation, a two-channel spectrographic imager was built to image the 135.6 nm line and a portion of the N2 Lyman Birge Hopfield (LBH) bands in the FUV. Blocking the 130.4 nm emission while transmitting the 135.6 nm was the most stringent spectral requirement. Historically, the first example of the Spectrographic Imager was flown on the NASA IMAGE mission. This was a Wadsworth configuration instrument with a concave grating and a hole in the grating where the entrance slit was placed. This appeared as a blind region in the final image. For ICON it was highly desirable to remove the obscuration needed for a Wadsworth instrument and a Czerny-Turner spectrograph configuration was selected. With the current resolving requirements, this optical arrangement permitted the use of off-axis mirrors and eliminated the requirement for the central obscuration in any of its mirrors or gratings. A disadvantage of the Czerny Turner-based instrument is that it requires an additional reflecting surface with the associated reflective losses. Fortunately, considerable progress was made in producing highly efficient UV mirrors since the IMAGE program and this disadvantage became a less crucial issue [Quijada et al., 2012].

The ICON FUV instrument views the atmospheric limb at an angle of 20° below the local horizontal at the satellite. A “periscope” steering mirror allows pointing the field of view 30° right or left. This mirror is somewhat like a periscope, except that the mirror plane is not at 45 but 55° to the local vertical giving a viewing axis that is directed 20° down from local horizontal. Figure 17a is a photograph, and Figure 17b is a schematic of the instrument. Light enters at the top of the illustration and is reflected by the steering mirror. The object at infinity, which is being viewed, is illustrated by a horizontal arrow. The scan mirror steers the direction of the optic axis by rotating around a vertical (perpendicular to the page) axis. The spectrograph slit and the detectors are stationary in the instrument frame, and the scan mirror will appear to rotate the outside view. Thus, effectively, the spectral slit will be superposed on the outside atmosphere at an angle depending on the position of the steering mirror.

Figure 17.

(a) Photograph and (b) schematic of the ICON FUV imager.

Because of the lack of transmitting materials in the FUV, all optical elements are reflective with the exception of the detector window, which is MgF2. The spectrograph mirrors M1 and M2 are both spherical. As discussed previously, M1 acts as a spectroscopic collimator as well as a camera mirror to create the intermediate image of the outside view at the grating. M2 is the spectrographic “camera mirror,” focusing the image of the input slit on the plane of the output slits. Depending on the wavelength and the location of the output slits, light is selected at the appropriate wavelength band. For the imaging task, one can regard M2 as a collimator, creating near-parallel light for each image point on the grating to be refocused on the detectors by the back imager mirrors CM1 and CM2, one set each, located in the back imagers of the short wavelength and long wavelength (LW) channels. The exit slits, not specifically labeled, are shown in black. In order to allow more room to accommodate the configuration, the LW channel has a turning flat mirror, which allows placing the LW channel out of the way of the other channel. In each channel the light is imaged on the detectors by the combination of mirrors CM1 and CM2. These mirrors are all aspheric fast mirrors.

The ICON FUV detectors consist of image tubes that are fiber-optically coupled to CCDs. The image tubes have MgF2 windows, and FUV photocathodes evaporated directly on to the microchannel plate (MCP). A stack of two MCPs are used, which provides sufficient charge multiplication to overcome any readout or dark current noise in the CCD. The CCDs, used in fast scan mode, read out 8 times a second. One hundred video frames are coadded digitally in memory to produce images of 12 s duration exposure. The data from the 12 s exposures are downlinked to the ground. Because the satellite moves substantially during 12 s, two types of motion compensation schemes are used. In type one, six horizontally coadded vertical altitude profiles are generated for the measurements of the altitude distribution of the emissions. The six profiles are generated by summing pixels horizontally in the direction of satellite motion without smearing in the vertical direction. The second type of motion compensation scheme produces a data stream containing Time Delay Integrated (TDI) images. In this mode, the individual frames are digitally coadded in memory after they have been projected on to an imaginary horizontal plane, which is moving uniformly in the satellite frame of reference. During the coadding process a constant offset is applied to the address for each frame, which is proportional to the speed of the satellite so that it compensates for the motion of the satellite. This method will provide high-resolution images in spite of the motion of the satellite platform during the 12 s exposure. It should be noted that these techniques require sophisticated real-time image manipulations on board the satellite. This is accomplished in field-programmable gate arrays (FPGAs) and firmware because of the high processing speeds needed. The reader should note that a similar TDI motion compensation was used in the FUV instrument on the NASA IMAGE satellite [Mende et al., 2000a].

4.4 High Spectral Resolution to Measure Doppler Shift of Lyman α or LBH Rotational Temperature (0.1 or 0.2 nm, e.g., IMAGE SI, GOLD)

Depending on the quality of the spectrograph of a spectrographic imager, the instrument can be used for narrowband UV filtering. The first adaptation of this technique was the IMAGE FUV Spectrographic Imager, which was a fairly high (~0.1 nm) spectral resolution version [Mende et al., 2000c].

As mentioned earlier, the IMAGE FUV Spectrographic Imager had a Wadsworth configuration spectrograph with a concave grating. The IMAGE Spectrographic Imager also had a “slit farm” consisting of a grill of nine parallel input slits in the Lyman α channel. The main goal was to minimize the transmission of the “cold” geocoronal Lyman α at and very close to the center of the wavelength band at 121.567 nm while maximizing the throughput for Doppler-shifted Lyman α created by fast-moving hydrogen atoms. The Doppler shift amounted to a range of about 0.1 to 0.8 nm, depending on the velocity (energy) of the incoming primary auroral protons. By introducing the grill, and a similar grill at the exit slit of the Lyman α channel, it was possible to adjust the grills so that they were in anticoincidence, blocking the light at 121.567 nm, but transmitting at other nearby wavelengths. The nine parallel slit techniqe increased sensitivity by ninefold. The relatively high spectral resolution requirement demanded an on-axis collimator, and IMAGE SI grating had a large hole in the center where the input slit was located. Unfortunately, this produced an obscuration in the middle of the image on the grating. The IMAGE instrument was on a rotating platform, and the images were exposed while the instrument scanned the view. Motion compensation was applied by shifting the pixel addresses while coadding the incoming light during the exposures, and this effectively filled in the obscured parts of the image.

The IMAGE FUV SI was on a satellite platform rotating at a rate of one revolution every 2 min. The instrument viewed radially outward and with its FOV of about 16°, an image point passed through from one side to the other in 5 s. Because of these considerations, IMAGE SI took a 5 s exposure with a 120 s cadence. A staring imager would therefore have a light gathering advantage of 120/5 = 24 times. If a platform suitable for a staring imager was available, it would be highly recommended to fly a proton imager with a single entrance slit but with multiple exit slits so that multiple images could be made, one for each exit slit. The luminosity data from each image could be fitted to the Doppler profile of the Lyman α line to obtain the Doppler shift that characterizes each image point. Using this configuration would make it possible to build an energy map of the precipitating protons.

4.5 Ultrahigh Spectral Resolution, Usually an Interferometric Instrument to Measure Doppler Shift of Ions, etc. (<0.1 nm)

There are several applications in which high spectral resolution would be desirable. An obvious application is measuring the Doppler shift of ions other than hydrogen. An example of this is O+ ions which can radiate, and they are also subject to charge exchange but measuring them would require 4 times higher velocity resolution than protons at the same energy, since they are 16 times heavier than hydrogen. The measurement of precipitating O+ ions in the few keV range would be a very important diagnostic in the study of ion precipitation from the magnetosphere. Currently, there is also a great interest in ion outflow, for example, in understanding how significant ion outflow is in refilling the magnetosphere. High sensitivity Doppler measurement of ions could be a very important tools in understanding the distributions of outflowing ions and the mechanisms responsible for driving the ions. Helium is also a good candidate but is less abundant in our geospace. Interferometers in visible light have been used to make Doppler shift measurements in the atmosphere to study winds. The requirements and the demonstrated instrument performance for these type of instruments are quite impressive. For example, in order to measure a 5 m/s wind speed, a measurement accuracy of 0.0001 nm is needed at the visible airglow wavelength of 557.7 nm. Such measurements can at least be made in the visible wavelength range [Englert et al., 2015], and a Michelson interferometer, with the required capabilities to measure atmospheric winds, was built as part of the NASA ICON payload.

Interferometers are the only practical solution for very high spectral resolution, because interferometers have a significant light gathering advantage, which is an absolute necessity for very high spectral resolution observations. However, even for relatively low spectral resolution applications, there are significant advantages in using the interferometric approach. Besides the higher light throughput at a given spectral resolution, the interferometer's high dispersion allows miniaturization of the instrument. It is widely accepted that filtered imagers are intrinsically more efficient than grating spectrometers. Filtered imagers usually rely on interference filters, which are fundamentally interferometers often based on the thin-film/Fabry-Perot principle. Transmitting materials get scarcer and more exotic in the ultraviolet. However, it is still possible to build multilayer reflectors in the FUV wavelength range by using a pair of salts that have a highly different refractive index. If a thin MgF2 spacer could be produced with the required high optical flatness then a multilayer coating applied on each side of it would allow the building of a FUV Fabry-Perot (FP) interferometer. It is instructive to consider the light gain of such an instrument and to compare it to a conventional spectrometer.

Figure 18 shows the comparison between the grating spectrometer and the Fabry Perot etalon. The light-gathering power can be best expressed by considering the etendue of the instruments. Etendue is the area of an optical aperture multiplied by the solid angle subtended by the light beam going through the aperture. In an optical system, the effective etendue is constant through the whole system. One can therefore calculate the etendue at any convenient surface along the optical train.

Figure 18.

Comparison of the light gathering of a grating spectrometer and a Fabry Perot interferometer.

For this discussion, I will again consider a simple lens as the focusing element (Figure 18) even though we realize that the UV instrument optics would probably consist of reflective elements. In our hypothetical instruments, the same size aperture and F number are used for both systems. The etendue can be computed for the spectrometer at the slit and for the interferometer at the aperture mask, where the interferometer pattern would be imaged and the wavelength transmission determined. For this calculation, it is unimportant whether the slit or the interferometer mask is at the entrance pupil, as in a spectrographic imager, or at the imaging plane of the system as in an imaging spectrometer. The etendue can be calculated by considering the area of the slit or aperture and the solid angle of the light cone going through the slit. Thus, the etendue at the slit is equal to A × Ω, where A is the area of slit and Ω is the solid angle. If the F number of the system is given then Ω can be obtained as Ω ~ π/(1 + 4 F2). If the same optics is used then the etendue is proportional to the open area of the slit or mask. In both systems the desired spectral resolution sets this area.

For the grating spectrometer case, the slit area As = xy, where x and y are the width and length of the slit respectively. Parameter y can be quite large, and the limit is given by the wide-angle performance α of the lens. It is unlikely that satisfactory optical performance can be achieved over a very wide field. For the sake of this discussion IMAGE SI parameter was used for α, the field angle of the slit.

In order to find x we have to specify dλ, the required spectral resolution. For first-order diffraction the grating spectrometer equation is

display math

where N is the ruling density in lines/mm, ω is the dispersion angle, and the wavelength λ is also in mm. Differentiating it

display math

where dω is the diffraction angle change and dλ (mm) is the wavelength change.

The slit width x = f dω where f is the focal length.

To define the slit width we can set dλ equal to the desired wavelength resolution while realizing that the transmission of the instrument will be a triangular function whose total width is 2 times dλ.

The length of the slit is y = f tan(α) ~ f α and If we can keep ω small then cos(ω) = 1 and

display math

where α is the field angle.

For the FP interferometer case the wavelength resolution and the dispersion are related as follows:

display math

where dλ is the wavelength resolution, λ is the operating wavelength, is the extreme ray angle consistent with the spectral resolution requirement, and μ is the refractive index.

Expressing the angle:

display math

The radius of the aperture of the FP is ~fdθ, where f is the focal length.

The area for the FP is

display math

Taking the ratio between the FP aperture and the spectrograph slit we get

display math
Example Spectrometer
Wavelength resolution

= 0.1 nm (IMAGE SI had 0.12)

Ruling density

= 3600 lines/mm

Slit width (for 0.1 nm)

= 0.108 mm

Slit length

= 36 mm (IMAGE SI = 36 mm)

Focal length

= 300 mm (IMAGE SI = 500)

Optic size D

= 300/3.5–85 mm


= 3.88 mm2

Example for a Fabry Perot:

= 1.0 (Vacuum gap)

Wavelength resolution

0.1 nm (IMAGE SI requirement 0.12 nm)

Focal length

= 300 mm (same as spectrometer)


= 3.5 (16° IMAGE SI optical performance limit)

Optic size D

= 300/3.5–85 mm

Resultant solid angle 2π

= 0.0052


= 471 mm2

Sensitivity ratio

~ 120 to 1 or 0.83%

The above example shows that the light-gathering performance of an FP is much better than that of the spectrometer of the same physical size, for medium-high spectral resolution imaging. It is possible to find a transparent substrate material in the wavelength region above 110 nm; for example, crystalline MgF2 can be polished to be used as a substrate for an interferometer. Multilayer mirrors have been made for FUV imaging [Torr et al., 1995]. It is likely that a double etalon would be needed to accomplish the appropriate wavelength selection and to overcome the free spectral range limitation of the interferometer. For example, if a UV FP with a finesse of 20 could be produced, then a single etalon with 0.1 nm spectral resolution would have 2 nm free spectral range and a double etalon could have a combined free spectral range of 40. This would be adequate for detecting a single spectral feature. Note that as long as each etalon could be produced with more than 10% efficiency then the combination double FP, at least on paper, would have efficiency of 1%, which is still more than the 0.83% efficiency derived for a spectrometer.

The possible configurations when using an FP as the spectral resolving element are shown in Figure 19. In the imaging spectrograph (top) configuration, an image is created on the FP aperture mask, which is reimaged on the detector. Different regions of the image will have different wavelength transmission characteristics, and the resultant image is a convolution of the image intensity and the spectral profile with a radial dependence on wavelength. The bottom configuration shows an FP operating as a spectrographic imager. In this configuration, the image is created on the FP etalon and is reimaged on the detector. In this case the FP fringe mask is located near the reimaging optics and ensures that the light, which passed through the system, is within the desired wavelength band. The resultant image is a 2-D image that is appropriately filtered by the passband of the etalon. Note that in Figure 19 and the calculations following, we assumed that the FP would have a single circular aperture mask and it would use only central first order. However, in an FP, several fringes could be imaged with a suitable mask consisting of annular apertures. Several orders/fringes can be collected, thus increasing the throughput of the interferometer.

Figure 19.

Schematic of FP configurations. (top) Imaging Spectrograph. (bottom) Spectrographic Imager.

5 Summary

In the accompanying paper we have outlined specific near-term objectives for auroral imaging to make progress in understanding magnetospheric physics. One of the objectives was to develop the optical capability to distinguish between auroras of low electron flux sourced from the plasma sheet and high flux sourced from the ionosphere. The latter auroras are generated by Alfven waves arriving from the magnetosphere. Another goal was identified in making optical observations of the ion precipitation and observing the velocity distribution of the ions from the Doppler shift of optical emission lines. To accomplish these objectives special instruments need to be developed. Clearly, there are many reasons why ground based observations are highly valued considering their economy and the speed in which they can produce results. However, spaceborne auroral imagers allow viewing the aurora from a global perspective, and in many cases they can remove the space-time ambiguites. Another major advantage is that spaceborne imagers can operate in the FUV region where the Sun's intensity is much lower permitting the observation of the aurora even in the daylit atmosphere. An addtional advantage of the FUV observations is that the atmosphere below the aurora is opaque and no corrections needed to be introduced to account for Earth's albedo reflection.

Precipitating free electrons do not emit photons. They can only be detected indirectly through the aurora they create by the excitation of atmospheric constituents. The more important tasks for remote sensing of the magnetosphere through auroral imaging is to distinguish the wave-generated electron auroras from the other types. This can be done by remote sensing the mean energy and energy flux of the precipitating electrons. In order to measure the precipitating energy, we need to use the atmosphere as our spectrometer. The total precipitated energy can be obtained by measuring the auroral luminosity in spectral regions, where the excitation mechanism is well understood and the cross sections are known. The mean energy of the electrons can be inferred from the altitude profile of the luminosity. Theoretically, this can be done either by triangulation from two or more stations, or by two or more spectral features, one of which has a height integrated intensity, which is strongly dependent on the altitude distribution of the emitting atmospheric particle. The altitude dependence can be due to (1) collisional quenching of a long lifetime upper state, (2) the variation of the density distribution with altitude, (3) the height variation of an atmospheric constituent that absorbs the optical emission, and (4) the height variation of the atmospheric temperature, as manifested through the rotational spectra of the emitting species. All of these methods are discussed regarding their advantages and feasibility with examples of data. The determination of the rotational temperature is the least explored technique but does offer promise for obtaining the mean energy of the precipitating electrons with high time resolution, which is needed to detect Alfvénic auroras.

Various instruments were discussed, especially those that were used for global imaging as part of spaceflight programs. Depending on the wavelength selection requirements, I distinguished (1) broadband imagers that reject unwanted mid-UV and visible radiations such as IMAGE WIC; (2) imagers that are able to separate LBH high and low for energy discrimination, such as Polar UVI; (3) spectrographic imagers that are able to discriminate between 135.6 and 130.4 nm emission, such as the ICON instrument; (4) higher-resolution spectrographic imagers such as used on the IMAGE spacecraft to distinguish between geocoronal Lyman α and Doppler-shifted auroral hydrogen emissions; and (5) other instruments that would be used for Doppler shift of heavier particles, such as He and O. We also propose that the development of a Fabry Perot interferometer for the FUV wavelength region would be highly advantageous because an interferometer has an intrinsically higher optical throughput efficiency than a grating instrument of similar size.


The author acknowledges many helpful discussions with Harald U. Frey and others at the University of California, Berkeley. The author gratefully acknowledges funding for the preparation of this article by NASA's Explorers Program under the Ionospheric CONnection Explorer (ICON) project contract number NNG12FA45C and by NSF under grant number 1141961 entitled “Auroral Signatures of Solar Wind Magnetospheric Coupling.” This work is essentially a review paper, and there is no specific publicly available data set associated with the material presented. The unpublished digitized image data presented in Figure 9 are in the author's possession, and it can be provided on request.