Icarus 200 (2009) 176–187 Contents lists available at ScienceDirect Icarus www.elsevier.com/locate/icarus Characteristics of Saturn’s polar atmosphere and auroral electrons derived from HST/STIS, FUSE and Cassini/UVIS spectra J. Gustin a,∗ , J.-C. Gérard a , W. Pryor b , P.D. Feldman c , D. Grodent a , G. Holsclaw d a Laboratoire de Physique Atmosphérique et Planétaire, Université de Liège, Allée du 6 Août, 17, 4000 Liège, Belgium Central Arizona College, Coolidge, AZ 85228, USA c Department of Physics and Astronomy, The Johns Hopkins University, Baltimore, MD 21218, USA d Laboratory for Atmospheric and Space Physics, University of Colorado, Boulder, CO 80303, USA b a r t i c l e i n f o a b s t r a c t Article history: Received 24 July 2008 Revised 21 October 2008 Accepted 10 November 2008 Available online 6 December 2008 Ultraviolet (UV) spectra of Saturn’s aurora obtained with the Hubble Space Telescope Imaging Spectrograph (STIS), the Cassini Ultraviolet Imaging Spectrograph (UVIS) and the Far Ultraviolet Spectroscopic Explorer (FUSE) have been analyzed. Comparisons between the observed spectra and synthetic models of electron-excited H2 have been used to determine various auroral characteristics. Far ultraviolet (FUV: 1200–1700 Å) STIS and UVIS spectra exhibit, below 1400 Å, weak absorption due to methane, with a vertical column ranging between 1.4 × 1015 and 1.2 × 1016 cm−2 . Using the low-latitude Moses et al. [Moses, J.I., Bézard, B., Lellouch, E., Feuchtgruber, H., Gladstone, G.R., Allen, M., 2000. Icarus, 143, 244–298] atmospheric model of Saturn and an electron energy–H2 column relationship, these methane columns are converted into the mean energy of the primary precipitating electrons, estimated to lie in the range 10– 18 keV. This result is confirmed by the study of self-absorption with UVIS and FUSE extreme ultraviolet (EUV: 900–1200 Å) spectra. Below 1200 Å, it is seen that transitions connecting to the v < 2 vibrational levels of the H2 electronic ground state are partially self-absorbed by H2 molecules overlying the auroral emission. Because of its low spectral resolution (∼5.5 Å), the UVIS EUV spectrum we analyzed does not allow us to unequivocally determine reasonable ranges of temperatures and H2 columns. On the other hand, the high spectral resolution (∼0.2 Å) of the FUSE LiF1a and LiF2a EUV spectra we examined resolve the H2 rotational lines and makes it possible to determine the H2 temperature. The modeled spectrum best fitting the FUSE LiF1a observation reveals a temperature of 500 K and self-absorption by a H2 vertical column of 3 × 1019 cm−2 . When converted to energy of precipitating electrons, this H2 column corresponds to primary electrons of ∼10 keV. The model that best fits the LiF2a spectrum is characterized by a temperature of 400 K and is not self-absorbed, making this segment ideal to determine the H2 temperature at the altitude of the auroral emission. The latter value is in agreement with temperatures obtained from H+ 3 infrared polar spectra. Self-absorption is detectable in the LiF2a segment for H2 columns exceeding 6 × 1019 cm−2 , which sets the maximum mean energy determined from the FUSE observations to ∼15 keV. The total electron energy range of 10–18 keV deduced from FUV and EUV observations places the auroral emission peak between the 0.1 and 0.3 μbar pressure levels. These values should be seen as an upper limit, since most of the Voyager UVS spectra of Saturn’s aurora examined by Sandel et al. [Sandel, B.R., Shemansky, D.E., Broadfoot, A.L., Holberg, J.B., Smith, G.R., 1982. Science 215, 548] do not exhibit methane absorption. The auroral H2 emission is thus likely located above but close to the methane homopause. The H2 auroral brightness in the 800–1700 Å bandwidth varies from 2.9 kR to 139 kR, comparable to values derived from FUV Faint Object Camera (FOC) and STIS images. © 2008 Elsevier Inc. All rights reserved. Keywords: Aurorae Saturn Spectroscopy Ultraviolet observations 1. Introduction Observations of Saturn’s aurora in the UV were first carried out by the Pioneer 11 and Voyager 1 and 2 spacecraft in 1979 and * Corresponding author. E-mail address: gustin@astro.ulg.ac.be (J. Gustin). 0019-1035/$ – see front matter doi:10.1016/j.icarus.2008.11.013 © 2008 Elsevier Inc. All rights reserved. 1980. The data collected by the Voyager UV spectrometer (UVS) revealed auroral emissions from both polar regions. Saturn’s aurora appeared as a narrow circumpolar region near 80◦ latitude with no apparent emission present in the polar cap, with typical intensities in the range 10 to 15 kilo-Rayleighs (kR) (Broadfoot et al., 1981; Sandel and Broadfoot, 1981). Intensity variations up to a factor of 5 were observed (Sandel et al., 1982), suggesting a solar wind controlled aurora. As for Jupiter, comparisons between observed UV spectroscopy of Saturn’s aurora with STIS, FUSE and UVIS spectra and models showed that the auroral emission is produced by the interaction between the H2 atmosphere and precipitating electrons (Shemansky and Ajello, 1983). Between 800 and 1800 Å, the auroral emission is dominated by atomic H lines from the Lyman series and H2 vibronic lines from the B 1 Σu+ → X 1 Σ g+ , C 1 Πu → X 1 Σ g+ , B 1 Σu+ → X 1 Σ g+ , D 1 Πu → X 1 Σ g+ , B 1 Σu+ → X 1 Σ g+ , D 1 Πu → X 1 Σ g+ system bands. While the Lyman (B → X ) and Werner (C → X ) bands and the Lyman continuum prevail in the FUV spectral region, the EUV bandwidth includes transitions from the B and higher Rydberg electronic states. Analysis of images obtained with STIS onboard the Hubble Space Telescope (HST) revealed a highly dynamic aurora, in the form of a ∼2◦ wide ring, located between 70◦ and 80◦ latitude with brightness ranging from below the STIS threshold of ∼1 up to ∼100 kR (Gérard et al., 2004; Clarke et al., 2005). Significant changes in morphology have also been observed during short intervals, where the auroral region forms a spiral followed by intervals of reinflation of the auroral region (Grodent et al., 2005). Recent STIS images of Saturn combined with simultaneous Cassini measurement of the solar wind (Crary et al., 2005) and Saturn kilometric radio emission (Kurth et al., 2005) confirmed that the morphology of Saturn’s aurora differs from those of both Earth and Jupiter. In particular, brightening of the aurora with a movement of the emissions toward higher latitudes was observed and correlated with Cassini’s solar wind dynamic pressure measurements, suggesting that the aurora is controlled by the solar wind conditions at Saturn. On the other hand, considerable longitudinal structure and time variations over interval of a few hours were also observed, despite the absence of observable external triggers and solar wind activations (Gérard et al., 2006). This suggests that the aurora is also influenced by an intrinsically dynamical magnetosphere. From a theoretical point of view, auroral rings can either be formed by field-aligned currents, associated with the breakdown of corotation of the plasma rotating in the middle magnetosphere (as in the case of Jupiter), or by interactions between the solar wind and the planet’s magnetosphere at the open-closed field line boundary (as for the Earth). Theoretical work (Cowley et al., 2004a), statistical studies of the location of the auroral emission (Badman et al., 2006) and the recent observation of a field-aligned current at the open-closed field line boundary coincident with HST images of the aurora (Bunce et al., 2008) indicate that the latter is more likely in Saturn’s case. In particular, using a simple quantitative model, Cowley et al. (2004b) showed that field-aligned voltages of ∼10 kV are required to accelerate outer magnetospheric electrons in order to produce an auroral oval of a few tens of kR. The FUV and EUV auroral emissions are known to interact with the atmosphere through absorption by hydrocarbons and selfabsorption by H2 . These two processes are used to estimate the altitude of the aurora. First, the altitude of the aurora relative to the hydrocarbon homopause may be derived from the absorption by methane, which attenuates the H2 emission below 1400 Å, leaving the H2 emission above 1400 Å unattenuated. This absorption is measured by the color ratio CR = I (1550–1620 Å)/ I (1230–1300 Å), which relates the attenuation of the H2 auroral emission to the amount of methane overlying the peak of the emitting layer. CR is 1.1 for an unattenuated spectrum expressed in Rayleighs. From six STIS Saturn spectra of southern aurora in the 1150–1750 Å spectral range, Gérard et al. (2004) found that the auroral emission was little absorbed by methane, with a nearly constant vertical CH4 column of 6 × 1015 cm−2 . Using the Saturn model atmosphere of Moses et al. (2000) at 30◦ North relating the H2 and CH4 column density profiles and the Grodent et al. (2001) atmospheric model that links the H2 column density to the mean energy of the precipitated electrons, Gérard et al. (2004) showed that the energy of the 177 precipitated primary electrons was in the range 12 ± 3 keV. Second, a direct measurement of the H2 column overlying the auroral emission can be obtained from EUV spectra. Below 1200 Å, the photons connecting to the v = 0, 1, 2 levels may be partially or totally absorbed by the overlying column of H2 and redistributed to the FUV portion of the spectrum. This self-absorption process provides a measure of the H2 column overlying the UV emission peak and provides information on the population of the groundstate vibrational levels (Gustin et al., 2004a). Self-absorption on Saturn was earlier pointed out by Broadfoot et al. (1981), Sandel et al. (1982) and Shemansky and Ajello (1983), who analyzed one of the auroral spectra at ∼30 Å resolution obtained with Voyager UVS. Sandel et al. (1982) and Shemansky and Ajello (1983) used a modeled spectrum with foreground H2 abundances of 1 × 1020 and 1 × 1016 cm−2 , respectively, to best fit the UVS data. On the other hand, the low resolution of the UVS observations and the high uncertainty of the vibronic branching ratios available at that time did not allow an accurate measure of self-absorption. Also, the modeled absorption by H2 used to fit the UVS spectra was based on single scattering losses only and did not take into account a line by line treatment. We present in this paper comparisons between synthetic H2 spectra, generated by a spectral generator previously described by Dols et al. (2000) and Gustin et al. (2002, 2004a), and UV spectra of Saturn’s aurora obtained with HST/STIS, FUSE, and Cassini UVIS instruments. Spectra obtained with STIS and the FUV channel of UVIS, with spectral resolution ∼12 and ∼5.5 Å, respectively, are potentially attenuated by the hydrocarbons overlying the auroral emission. Although methane (CH4 ), ethane (C2 H6 ) and acetylene (C2 H2 ) are known to be present within the atmosphere, only CH4 has been previously detected from auroral UV spectra in the 1150–1700 Å range (Gérard et al., 2004). FUSE and UVIS spectra in the EUV have spectral resolutions ∼0.20 and ∼5.5 Å, respectively. As mentioned above, lines connecting to ν 2 may be attenuated by the H2 molecules overlying the emission. Several series of synthetic spectra are generated, parameterized by the methane column density in the FUV and by the H2 temperature and H2 column density in the EUV. Each resulting spectrum is compared to the data, and the synthetic spectrum that minimizes the chisquare (χ 2 ) is considered as the best fit, providing an accurate estimate of the auroral characteristics. The pressure level of the energy deposition of the aurora is then derived from the best parameters and yields the energy of the precipitating electrons, using a relationship between the energy of the incident electrons and the H2 column described hereafter. Results obtained with all three instruments are compared and implications for the auroral characteristics are considered. 2. Observations and data reduction 2.1. STIS The STIS observations presented here were already processed and analyzed by Gérard et al. (2004). They consist of six spatially resolved spectra obtained on 7 and 8 December 2000 with the G140L grating, combined with the 24.7 × 0.5 arcsec2 slit, providing spectra at ∼12 Å resolution in the 1150–1750 Å spectral range. They were imaged on the FUV-MAMA photon counting detector providing 1024 pixels in the spatial direction and 1024 pixels in the spectral direction. The methodology used to process these spectra was described in detail by Gérard et al. (2004) and will not be repeated here. Table 2 of Gérard et al. (2004) summarizes the observation characteristics and their Fig. 6 shows the position of the slit, superimposed on a 24.7 × 24.7 arcsec2 STIS image obtained on 7 December 2000. 178 J. Gustin et al. / Icarus 200 (2009) 176–187 Fig. 2. Viewing geometry of the UVIS observation of Saturn’s aurora from 17 February 2005. The long side of the slit represents the 64 spatial pixels of the UVIS aperture, while each small rectangle inside the slit can be seen as a 1024 pixels wide spectrum. The slit was parallel to the planet’s equator, moving from the east side to the west side of the pole. The slit labeled “A” corresponds to the first record of 240 s, starting at 14:59:23 UT. The “B” slit corresponds to the middle exposure (start time: 18:56:23), while the “C” slit represents the last record of the observation (start time: 22:44:13). Fig. 1. The white square illustrates the field of view of the FUSE LWRS aperture, using an unrelated STIS image taken on 7 December 2000. Assuming an auroral oval situated at 75◦ latitude, the viewing angle of the STIS image (which was taken at the same period as the STIS spectra described in this paper) is 53.1◦ , while the FUSE viewing angle is 53.4◦ . 2.2. FUSE Spectra of Saturn’s aurora in the wavelength range 905 to 1180 Å were obtained by FUSE on 15 January 2001, starting at 03:20:06 UT for a total exposure time of 9966 seconds. The low resolution aperture LWRS (30 × 30 ) was used, providing spectra at ∼0.20 Å spectral resolution, determined by the small but finite extent of the aurora in the aperture. The data presented here were processed using the FUSE pipeline, calfuse, version 2.2.1. Fig. 1 shows the field of view of the LWRS aperture, using an HST/STIS image obtained on 7 December 2000. It is seen that FUSE collected a large portion of photons from Saturn’s disc, as well as its rings. Background emissions from Saturn’s airglow and sunlight reflected by the rings are identified in the observed spectra. This aspect will be discussed in Section 4.4. A detailed description of the FUSE spectrometer has been given by Moos et al. (2000) and Sahnow et al. (2000). FUSE consists of four separate spectrographs, two using lithium fluoride coatings (LiF) and two employing silicon carbide (SiC). Each of four separate detectors, denoted 1a, 1b, 2a, 2b simultaneously records a spectrum from LiF and a spectrum from SiC, giving a total of eight separate raw spectra. In this study, we used the LiF1a and LiF 2a channels. They provide the best signal to noise ratio, while spanning a major portion of the total wavelength range available. Unlike STIS or UVIS spectra presented here, FUSE observations do not provide spatial information of the intensity distribution within the aperture. 2.3. UVIS Saturn’s aurora was observed with UVIS on 17 February 2005, from 14:59:23 UT to 22:48:13 UT. The data consist of a set of 48 consecutive records of 240 s each. The geometry of the observation is illustrated in Fig. 2. The characteristics of the UVIS instrument have been discussed in detail by Esposito et al. (2004). In brief, UVIS includes a two-channel imaging spectrograph, from 1115 to 1912 Å in the FUV, and from 563 to 1182 Å in the EUV. Three slits are available: a high resolution slit (75 and 100 μm slit width for the FUV and EUV channel, respectively), a low resolution slit (150 and 200 μm slit width for the FUV and EUV respectively), and an occultation slit of 800 μm width, identical for both channels. The data discussed here were obtained with the low resolution slit for both the EUV and FUV channels, providing spectra at ∼5.5 Å spectral resolution. The detector is a Codacon (CODed Anode array CONverter), consisting of 1024 pixels in the spectral direction and 64 pixels in the spatial direction. Calibration of the EUV and FUV data follows the pre-flight measurements as described in Esposito et al. (2004). Additionally, a background noise level due to the radioisotope thermoelectric generator (RTG) onboard Cassini is removed and a flat-field correction derived from observations of Spica (Steffl et al., 2004) is applied. Examination of the raw data shows that the RTG background is ∼4.5 × 10−4 count per spatial and per spectral pixel, which corresponds to the value determined by other datasets. In the EUV, two wavelength-dependent contaminating signals affect the recorded spectra. The first is due to internal instrument scattering of Ly-α , which is focused just beyond the long-wavelength end of the EUV detector. This contribution to the EUV signal is small below 1170 Å, and is estimated to be less than 7% of the total signal, based on comparisons with a modeled H2 spectrum (Ajello et al., 2005). Our analysis of UVIS in the EUV focuses on the 900– 1170 Å region, where this effect is negligible. The second contamination source is due to a small light leak that allows undispersed interplanetary Ly-α to reach the EUV detector into one-third of the detector, corresponding to wavelengths between 920 and 1020 Å. A recent calibration observation has been performed on 16 April 2007, when the telescope was pointed on the sky. It allowed us to record this background alone, which was scaled and removed from the EUV data presented here. For the FUV channel, in flight calibrations from September 1999 to November 2007 have revealed that above 1550 Å, the FUV detector absolute sensitivity slowly increases with time. This effect has been considered in a timedependent sensitivity curve (the UVIS team, personal communication) and has been included in our pipeline procedure in order to convert the raw counts into brightness units of Rayleigh Å−1 . Each of the 48 raw records has been carefully examined to determine which spatial pixels were measuring auroral emission. For each EUV record, the spectral pixels in the 1045–1167 Å range (which is the brightest portion of the H2 emission in the EUV bandwidth) are summed, and the result is displayed as a function of the spatial pixels. We then arbitrarily set an auroral emission threshold to 6.3 × 10−3 count pixel−2 s−1 and all spatial pixels above this value were assumed to be auroral. The same procedure is used in the FUV, were all pixels above 7 × 10−3 counts pixel−2 s−1 in the 1270–1740 Å range were assumed to be auroral. For both channels, the selected pixels are then summed to create a single EUV and FUV raw spectrum. These resulting spectra are then processed and used in the present analysis. It should be UV spectroscopy of Saturn’s aurora with STIS, FUSE and UVIS 179 Fig. 3. UVIS FUV spectrum (black line) and best synthetic fit, absorbed with a CH4 vertical column of 1.4 × 1015 cm−2 (dash–dot green line). When compared to an unabsorbed synthetic spectrum (dashed blue line), the UVIS spectrum is very slightly absorbed. For comparison, the red curve shows one of the six STIS spectra of Saturn, which is best fitted with a model absorbed by a vertical methane column of 5.4 × 1015 cm−2 . noted that UVIS observed Saturn’s nightside northern pole, while the STIS and FUSE data were acquired from Saturn’s southern dayside aurora. 3. Spectral generator and absorption process 3.1. Spectral generator The generator of H2 spectra simulates the effects of the impact of auroral electrons on the atmospheric H2 molecules. It has been described and used for jovian spectroscopy analysis by Dols et al. (2000), Gustin et al. (2002, 2004a, 2004b) and Gérard et al. (2002), and for Saturn’s spectroscopy by Gérard et al. (2004). The generator includes the X 1 Σ g+ , B 1 Σu+ , E , F 1 Σ g+ , C 1 Πu , B 1 Σu+ , D 1 Πu , B 1 Σu+ , and D 1 Πu electronic states, whose population is controlled by the direct primary and secondary electron impact on ground state H2 ( X 1 Σ g+ ), assumed to be in local thermodynamic equilibrium (LTE). As seen in Fig. 3 of Gustin et al. (2004a), spectra longward of the atomic H Ly-α line can be described by synthetic spectra including the Lyman (B– X transitions) and Werner (C – X transitions) bands only, taking into account the enhanced rate of excitation of the B 1 Σu+ state due to cascading from the E , F 1 Σ g+ state. Shortward of Ly-α , the modeled spectra require the addition of the B – X , D– X , B – X and D – X bands to accurately fit the data. 3.2. Hydrocarbon absorption Methane is the fourth most abundant compound in the atmosphere of giant planets, after H, He and H2 . It is produced in the troposphere and transported to the stratosphere by convection and turbulent diffusion. Through photolysis, methane is the source of other hydrocarbons and produces ethane and acetylene, in that order of importance on Saturn (Moses et al., 2000; Ollivier et al., 2000; Moses and Greathouse, 2005). Methane has a large wavelength-dependent absorption cross-section in the 900–1400 Å domain. Accordingly, it partially absorbs auroral emissions in this wavelength range and leaves the emission above 1400 Å unattenuated (see Fig. 3 of Gérard et al. (2003) for an illustration of this differential absorption). The color ratio provides a measure of the attenuation of the H2 emission by methane overlying the emitting layer. From comparisons between the observed spectra and several series of synthetic spectra attenuated by CH4 , the methane column overlying the auroral emission is first determined. The model atmosphere of Moses et al. (2000) is then used to relate the CH4 column obtained from the best fit to the pressure level and the H2 column overlying the auroral emission peak. The energy deposition of the precipitated electrons can then be estimated from a stopping power table, giving the average path length traveled by monoenergetic electrons into H2 , as they slow down to rest. We have derived a relationship between the primary electron energy and the H2 column from two sources. The first is the ESTAR database (http://physics.nist.gov/PhysRefData/Star/Text/ESTAR.html), which spans the 10 to 106 keV electron energy range. The second is the two-stream electron transport model of Grodent et al. (2001), which relates the mean electron energy of the primary electrons to their penetration depth in Jupiter’s auroral atmosphere. This degradation model is used to determine the electron energy–H2 column relationship for electrons from 200 eV to 10 keV (since H2 is the dominant constituent for Jupiter and Saturn, the penetration depth of electrons is identical for both planets). Results obtained from the Grodent et al. (2001) model have also been compared to values provided by the stopping power table. It has been found that the mean energy of the Maxwellian electronic distributions used in the Grodent et al. (2001) model and the mono-energetic electrons found in the stopping power table provide similar H2 column values, yielding the energy deposition at the same pressure level (D. Grodent, private communication). This method has been used in a slightly different form for jovian auroral studies by Dols et al. (2000), Gustin et al. (2002, 2004a, 2004b, 2006) and Gérard et al. (2002, 2003) in order to determine electron energies from hydrocarbon absorption. It was also used by Gérard et al. (2004) to study the STIS spectra of Saturn also presented here. 3.3. Self-absorption When an auroral photon is emitted from a given transition in the direction of the observer, the photon can undergo an absorption by an H2 molecule, following the same transition. This 180 J. Gustin et al. / Icarus 200 (2009) 176–187 Table 1 Main results. a Spectral range (Å) FWHM (Å) CH4 column (cm−2 )b H2 column (cm−2 )c Pressure (μbar)d Temperature (K)e Mean energy (keV) H2 brightness (kR) STISf UVIS FUV UVIS EUVg FUSE LiF1a FUSE LiF2ah 1230–1620 12 4.2 × 1015 –1.2 × 1016 – 0.17–0.28 – 13.4–17.6 3.3–14.0 1230–1660 5.5 1.4 × 1015 – 0.10 – 10.2 5.8i 910–1170 5.5 – 2.4 × 1020 0.82 700 21.6 2.9i 1030–1080 0.2 – 3.0 × 1019 0.10 500 10.1 139j 1090–1180 0.2 – 6.0 × 1019 0.21 400 14.8 129j a Spectral range used for the fitting procedure, not the full spectral range of the observed spectra. Vertical values, deduced from the FUV spectra, taking into account the angle between the local vertical and the line of sight in planetographic coordinates. The H2 columns deduced from UVIS EUV and FUSE best fits may be used to derive the corresponding vertical methane column, using the Moses et al. (2000) atmospheric model. We find methane columns of 7.6 × 1016 , 1.3 × 1015 , and 6.3 × 1015 cm−2 for UVIS EUV, FUSE LiF1a and FUSE LIF 2a, respectively. c Vertical values, measured from self-absorption for UVIS EUV and FUSE spectra. The methane columns deduced from STIS and UVIS FUV spectra are used determine the corresponding H2 column, using the Moses et al. (2000) model. We find H2 columns from 5.0 × 1019 to 8.2 × 1019 cm−2 for STIS and 3.1 × 1019 cm−2 for UVIS FUV. b Pressure level of the auroral emission, calculated from the hydrostatic law (weight of the H2 column, with g at 75◦ ). Measured from self-absorption for UVIS EUV and FUSE spectra. f Minimum and maximum values obtained from the lesser and most absorbed of the six spectra. g The extreme values deduced from the χ 2 limits are 9.7 × 1018 to 1.3 × 1022 cm−2 for the vertical H2 column, which correspond to pressure from 0.033 to 45.2 μbar and mean electron energies from 4.4 to 342 keV, respectively (see point 4.3 for details). d e h The H2 column deduced from LiF2a spectrum should be considered as an upper limit value (see Section 4.4 for details). Although the EUV and FUV observations are simultaneous, the two slits are not perfectly aligned, which explains the different auroral brightness derived for both channels. j As the aurora does not completely fill the LWRS aperture, the FUSE values must take into account a filling factor, estimated to be ∼330. Since the auroral morphology was unknown during the FUSE observation, this factor is very uncertain and could differ by a factor of ∼5. i absorption is followed by a radiative de-excitation, producing a photon with an energy equal to or lower than the photon initially absorbed. Self-absorption is therefore a fluorescence process changing the intensity distribution of the H2 lines in the auroral spectra. A measure of the attenuation of lines affected by selfabsorption gives information on the rovibrational temperature of the absorbing molecules through the population of the absorbing rovibrational level, controlled by temperature. It also indicates the pressure level of the auroral emission through the total number of absorbing H2 molecules, i.e. the H2 column density. Self-absorption is included in our spectral model in the form of an optical depth τs , determined for each vibronic transition: τs = P X ( v , j) × C H2 × σλ , where P X ( v , j ) is the population of the ground rovibrational absorbing level, controlled by the rovibrational temperature of the absorbing molecules, assumed to equal the rovibrational temperature of the emitting molecules; C H2 is the total number of molecules crossed by the emitted photons (i.e. the column density), directly related to the depth of the emission; and σλ is the absorption cross-section of the transition considered, described by Wolven and Feldman (1998). This model has been validated by an H2 laboratory spectrum obtained at high pressure, simulating self-absorption, and used previously in the analysis of FUSE spectra of jovian aurora by Gustin et al. (2004a). The redistribution of spectral intensity is illustrated in Fig. 6 of Gustin et al. (2004a). Comparisons between synthetic spectra (parameterized by the H2 temperature and H2 column) and observed spectra then allow one to derive the parameters that best fit the observations. The energy of the precipitating electrons is then derived from the stopping power table mentioned above. As for the emitting H2 molecules, it is assumed that the absorbing H2 molecules are in the X 1 Σ g+ electronic ground state and that their vibrational and rotational populations are in LTE and follow a Maxwell–Boltzmann distribution. According to models (Moses et al., 2000) or observations (Smith et al., 1983), rovibrational temperatures in Saturn’s stratosphere are expected to lie in the range 100–450 K. At a temperature of 400 K, the fraction of molecules in the v = 1 and v = 2 states compared to v = 0 is 3 × 10−7 and 2 × 10−13 , respectively. This means that in the case of Saturn, an attenuation of transitions connecting to the v = 0 and 1 levels is expected. Examination of a synthetic spectrum shows that lines connecting to v = 0, 1 are emitted at wavelengths shorter than Ly-α , which means that only lines in the EUV part of the UV spectrum are expected potentially to be absorbed. 4. Analysis Determination of the characteristics of Saturn’s auroral precipitation is based on a comparison between the observed spectra and several series of synthetic spectra. It is assumed that the auroral emission is produced in a single emitting layer, characterized by an average rovibrational H2 temperature. This layer is overlaid by an absorbing layer of CH4 and H2 . In the FUV bandwidth, CH4 is taken as the only parameter, since it is the major absorbing species affecting the auroral emission. In the EUV, as the auroral emission is influenced by self-absorption through the overlying H2 column and the H2 temperature, these variables are used as parameters in the model. Since they are inter-dependent, nested loops with H2 and T as variable parameters are employed. It is supposed here that the primary electrons do not directly participate into the excitation of the H2 molecules producing the aurora. Through an energy degradation process, primary electrons produce secondary electrons that effectively excite H2 and create the aurora (see Grodent et al., 2001, for a thorough study of the electron energy degradation process). Besides, the excitation cross-section curves of the H2 electronic states are very similar for exciting electrons above 30 eV, meaning that the shape of the H2 spectrum is moderately influenced by the electron energy variations above this value. Consequently, we used 100 eV mono-energetic electrons to simulate secondary electrons and generate the modeled H2 spectra. This reasoning has been validated by the work of Liu and Dalgarno (1996), who demonstrated that auroral UV spectra are not a sensitive indicator of the primary electron energy distribution. Each resulting modeled spectrum is compared with the data and the model that minimizes χ 2 is considered as the best fit. All the results are summarized in Table 1. 4.1. STIS As mentioned previously, Saturn’s spectra obtained with STIS have been analyzed by Gérard et al. (2004). Six spectra of the UV spectroscopy of Saturn’s aurora with STIS, FUSE and UVIS southern aurora were compared to synthetic spectra, and columns of methane have been derived. An example of best fit is displayed in Fig. 7 of Gérard et al. (2004), and one of the STIS spectra is reproduced by the red line in Fig. 3. The CR values derived from these observations are nearly constant, with values from 1.31 to 1.48, compared with 1.1 for an unabsorbed spectrum. The attenuation by CH4 is thus clear in this case, with best fits obtained for vertical columns of methane from 4.2 × 1015 to 1.2 × 1016 cm−2 . Using the atmospheric model of Moses et al. (2000), it is seen that these values correspond to vertical H2 columns from 5.0 × 1019 to 8.2 × 1019 cm−2 which places the auroral emission peak at pressure levels between 0.19 and 0.31 μbar. The mean energy of the precipitating electrons then lies in the range 13.4 to 17.6 keV. We also tested models including C2 H2 and C2 H6 , likely important on Saturn. The fits were not improved, confirming the small amount of these hydrocarbons compared to CH4 . Using a synthetic spectrum, the total brightness of the STIS spectra, converted to an unattenuated H2 emission in the 800–1700 Å bandwidth, ranges from 3.3 to 14 kR. 4.2. FUV channel of UVIS The UVIS FUV spectrum we analyzed is shown by the black line in Fig. 3. The observed spectrum is little absorbed, with a best fit obtained for a vertical CH4 column of 1.4 × 1015 cm−2 , shown in green in Fig. 3. Comparison with a model unaffected by methane (dashed blue line) clearly shows that the addition of weak absorption improves the fit. The vertical H2 column density derived from the hydrocarbon absorption is 3.1 × 1019 cm−2 , which corresponds to a pressure level of 0.11 μbar and precipitating electrons with a mean energy of ∼10 keV. The unattenuated H2 emission in the 800–1700 Å range provides a total brightness of 5.8 kR, in accord with those obtained with STIS. 4.3. EUV channel of UVIS The UVIS EUV spectrum has been compared to several series of synthetic spectra, with the H2 column and temperature as free parameters. The vertical methane column was set to 1.4 × 1015 cm−2 , i.e. the value determined from the UVIS FUV channel. Comparison between the observation (black curve in Fig. 4a) and a model without self-absorption (dashed green curve in Fig. 4a) shows that Saturn’s spectrum undergoes strong attenuations below 1060 Å, while the intensity of some observed lines above 1110 Å are increased. Lines with clear attenuation or intensification have been identified with the help of modeled spectra and are listed in Table 2. It is seen that all the attenuated lines (900–1060 Å) connect to v = 0 and v = 1 vibrational levels, with stronger absorption for lines connecting to v = 0. Above 1110 Å, all the lines amplified by fluorescence connect to v = 3 and v = 4. These considerations well illustrate the intensity redistribution caused by self-absorption. The peak around 1100 Å is more complex to interpret. Comparison with the unabsorbed synthetic spectrum shows that this peak, mainly due to W (0–2) transitions, is slightly intensified. While most lines connected to v < 2 are attenuated and v = 3 and 4 connected transitions are intensified, lines connecting to v = 2 can be seen as an intermediate case, since they do not exhibit clear absorption or intensification. An extensive study of self-absorption has been made by Gustin et al. (2004a), who analyzed FUSE spectra of the jovian aurora. Their Fig. 15 shows the effect of self-absorption on lines connecting to v 3, for temperatures of 400 and 900 K, in the 1090–1180 Å window. Panels a, b and d of their Fig. 15 show that for both temperatures, an increase in the total H2 column implies an attenuation of lines connecting to v = 0, 1 levels, while lines connecting to v = 3 are brightened. Panel c of Fig. 15 shows that at 400 K, 181 lines connecting to v = 2 slightly brighten as the H2 column increases. This means that the v = 2 vibrational level is virtually unpopulated at 400 K and thus cannot absorb photons, whatever the amount of H2 attainable. When the rovibrational temperature is in the 400 K range, transitions connecting to v = 2 can only be enhanced through fluorescence. At 900 K, H2 columns below 5 × 1019 cm−2 tend to brighten lines connecting to v = 2, while higher H2 columns start to absorb these lines. The best fit is obtained for T = 700 K and a vertical H2 column of 2.4 × 1020 cm−2 (red line in Fig. 4a) with a χ 2 value of 0.0307. Features near 1027 and 1053 Å, strongly self-absorbed, are well reproduced, as well as brightened lines near 1145 and 1161 Å. Apart from substantial discrepancies below 950 Å, the overall best fit shows major improvements compared to the unabsorbed model. In particular, it exhibits significant improvements near 1118, 1146 and 1161 Å, implying that the brightening due to fluorescence is well modeled. Previous studies showed that a minimum spectral resolution of ∼1 Å is required to derive a rotational temperature (Gustin et al., 2002). At ∼5.5 Å spectral resolution, the UVIS spectrum does not resolve the rotational lines and does not allow one to unequivocally determine a best T –H2 column pair. In order to determine a range of satisfactory values of these parameters, the χ 2 variations with temperature and H2 column have been examined and are displayed in Fig. 5. The best fit is indicated by a star symbol in Fig. 5, as shown in Fig. 4a. Two other models (square symbols in Fig. 5) at 500 K—1.2 × 1020 cm−2 (dashed orange line) and 900 K— 1.2 × 1019 cm−2 (dashed blue line) are compared to the UVIS EUV data in Fig. 4b. A detailed examination shows that both models in Fig. 4b, although still acceptable with χ 2 < 0.034, are less satisfactory than the best model (χ 2 = 0.0307) in red in Fig. 4a. We found that models with χ 2 > 0.034 started to show some major discrepancies. We thus arbitrarily consider T –H2 column pairs with χ 2 0.034 as satisfactory models to the UVIS EUV data. In the lower left corner in Fig. 5 we also show the T –H2 relationship obtained from the Moses et al. (2000) low-latitude atmosphere. It is seen that for a given H2 column, temperatures associated with the Moses et al. (2000) model are significantly lower than the values derived from our regression analysis. This point will be discussed in Section 5 of the paper. The pressure level corresponding to our best fit, derived from the weight of the observed H2 column, is 0.82 μbar. Using our H2 column–electron energy relationship, this corresponds to an electron mean energy of 21.6 keV. If we take into account the wide range of H2 column values limited by models with χ 2 < 0.034, we obtain a range of H2 vertical columns from 9.7 × 1018 to 1.3 × 1022 cm−2 , corresponding to pressure levels from 0.033 to 45.2 μbar. These values correspond to primary auroral electrons with mean energies ranging from 4.4 to 342 keV. 4.4. FUSE This section describes the analysis of two FUSE LiF segments, which provide the highest signal to noise ratio. The first is the LiF1a segment in the 1030–1080 Å spectral window. The spectrum below 1030 Å is ignored to avoid the bright H-Ly β line at 1025.73 Å. The second spectrum is the entire LiF2a segment in the 1090–1180 Å window. The SiC spectra below 1000 Å were found to be too noisy to provide valuable information on the auroral characteristics and won’t be discussed here. For both analyzed spectra, a CH4 column was not determined, because of the small variation of the methane cross-section with wavelength in these spectral windows. The LiF1a spectrum is presented by the black line in Fig. 6. The spectrum is contaminated by solar pumped fluorescence lines, as well as solar lines reflected from Saturn’s disc and rings, identified by vertical green lines. The best fit, shown by the red line in Fig. 6, is characterized by a rovibrational H2 temperature of 500 K, as- 182 J. Gustin et al. / Icarus 200 (2009) 176–187 Fig. 4. (a) UVIS EUV spectrum (black line) and models. When compared to a synthetic spectrum without self-absorption (dashed green line), it is seen that below 1060 Å, the observed lines connecting to v 1 are attenuated, while lines near 1118, 1146 and 1161 Å, connecting to v 2, are brightened. The best fit, in red, is obtained for a temperature of 700 K and a vertical H2 column of 2.4 × 1020 cm−2 . (b) The spectral resolution of UVIS (∼5.5 Å) does not allow us to resolve the rotational lines. As a consequence, many (T , H2 column) combinations can offer a satisfactory fit to the data as shown here. See the text for a more detailed discussion on the fit quality. sociated with a vertical H2 column of 3.0 × 1019 cm−2 . Compared to a model without self-absorption (dashed curve in Fig. 6), the best fit exhibits major improvements, especially near 1052.5 and 1054 Å. The main lines influenced by self-absorption, which correspond to those identified in the UVIS EUV spectrum, are listed in Table 2. The effect of self-absorption at this resolution and for this spectral window is illustrated in Fig. 11 of Gustin et al. (2004a), who analyzed FUSE spectra of the jovian aurora. This figure shows that the lines affected by self-absorption in the LiF1a window connect to v = 0 and v = 1. The modeled spectrum fits the data very well, especially where self-absorption has a major effect, i.e. near 1052.5 and 1054 Å. The major discrepancies are explained by the contamination by the fluorescence and solar lines. Unlike UVIS, FUSE provides spectra at high resolution (∼0.20 Å) that allow us to resolve the rotational lines, which means that spectra at this res- olution are sensitive to temperature variations. The W (0–1) P (3) line at 1058.82 Å is particularly responsive to temperature and is a good indicator of the overall rovibrational H2 temperature. The T –H2 column pair determined by the best fit can be considered as the unique solution of the fitting regression, since small parameter variations could clearly affect the fit and be visually identified. Although not necessary, we tested models where the temperature controlling the absorbing and emitting molecules is different. This did not improve the fit, which suggests that the emission and absorption of auroral photons occur approximately at the same altitude, as expected from the rapid drop of H2 density with altitude. As seen in Fig. 1, the low resolution aperture of FUSE integrates a large portion of the H2 airglow from Saturn’s disc and limb. In order to see if the airglow has a significant influence on the best parameters we obtained, we used a modeled H2 spectrum of Sat- UV spectroscopy of Saturn’s aurora with STIS, FUSE and UVIS 183 Fig. 5. Map of the χ 2 variations with the H2 rovibrational temperature and H2 column obtained from the comparison between the UVIS EUV spectrum and several series of synthetic spectra. The star corresponds to the best fit displayed in Fig. 4a, the two squares correspond to cases shown in Fig. 4b and discussed within the text. Temperatures and H2 column pairs inside the bold line (χ 2 0.034) are considered to provide a good fit to the data. The dashed line shows temperatures and H2 column from the Moses et al. (2000) atmosphere. Fig. 6. FUSE LiF1a spectrum and best fit model at T = 500 K and H2 vertical column of 3 × 1019 cm−2 . Comparisons between the best fit and a synthetic spectrum without self-absorption (dashed curve) clearly demonstrate the effects of self-absorption, principally near 1052.5 and 1054 Å. The main discrepancies at 1031.86 Å (W (1, 1) Q (3)), 1037.15 Å (L (5, 0) R (1) + P (3)) and 1071.62 Å (L (6, 1) P (1)) correspond to H2 lines pumped by the solar emission (Liu and Dalgarno, 1996). It should be noted that the H2 lines pumped by the solar NIII line at 989.79 Å are mislabeled as solar pumped NII in Liu and Dalgarno (1996). 184 J. Gustin et al. / Icarus 200 (2009) 176–187 H2 molecules is different. They did not improve the best fit. The upper limit of the H2 column gives an upper limit for the precipitating electron mean energy of 14.8 keV. Table 2 Main spectral lines affected by self-absorption at Saturn. Attenuated lines Intensified lines Wavelength (Å) Identification Wavelength (Å) Identification 960.449 965.062 967.673 969.088 970.560 976.548 978.217 985.642 986.519 987.767 991.378 996.124 997.638 997.824 1006.340 1008.383 1014.504 1017.001 1019.500 1028.246 1040.058 L (13–0) P (3) W (2–0) R (1) L (12–0) R (3) L (12–0) P (3) W (2–0) P (3) W (2–0) P (5) L (11–0) P (3) W (1–0) R (1) L (11–0) R (5) L (10–0) P (3) W (1–0) P (3) L (10–0) P (5) W (1–0) P (5) L (9–0) P (3) L (9–0) P (5) L (8–0) P (3) W (0–0) P (3) L (8–0) P (5) L (7–0) P (3) L (7–0) P (5) L (6–0) P (5) 1097.911 1098.115 1099.420 1100.521 1102.162 1104.328 1116.407 1145.904 1161.295 1193.246 1206.639 W (0–2) W (0–2) W (0–2) W (0–2) W (0–2) W (0–2) W (1–3) W (0–3) W (1–4) W (0–4) W (1–5) R (1) R (0) Q (1) Q (2) Q (3) Q (4) Q (1) Q (1) Q (1) Q (1) Q (1) At temperatures near 500 K and absorbing H2 columns of 3 × 1019 cm−2 , most of the lines connecting to the v = 0 vibrational level of ground H2 are saturated and drop to intensities 1% of the unabsorbed values. Very strong lines connecting to v = 1 are reduced buy a factor 2 to 4 while lines connecting to v > 1 may be slightly intensified by the fluorescence process of self-absorption. L stands for Lyman and W for Werner H2 lines. urn’s airglow (Hallett et al., 2005), convolved at FUSE resolution. The intensity of the airglow spectrum was scaled to the total intensity of the FUSE auroral spectrum. We then removed various fractions of airglow from the auroral spectrum, and applied our fitting procedure to the spectrum so obtained. We concluded from this technique that our best fit parameters are not significantly influenced by the removal of various fractions of airglow emission. The H2 column obtained from the best fit is 3.0 × 1019 cm−2 , which places the bulk of the energy deposition near 0.1 μbar. This corresponds to precipitating electrons with a mean energy of 10.1 keV. The second FUSE auroral spectrum we analyzed is the LiF2a, shown by the black line in Fig. 7. A few lines due to pumped Ly β and solar NIII and OVI are clearly identified in the spectrum, as considered by Liu and Dalgarno (1996). The best fit obtained from several series of synthetic spectra is shown by the red line in Fig. 7. It is characterized by a rovibrational temperature of 400 K and a slant H2 column of 1 × 1020 cm−2 , corresponding to a vertical H2 column of 6 × 1019 cm−2 . The fit is excellent and the main discrepancies are explained by the fluorescence lines present in the spectrum. Examination of the χ 2 variations show that in the 100–500 K range, the regression procedure is much more sensitive to temperature variations than to H2 column variations. This is confirmed by Fig. 15, panel e of Gustin et al. (2004a), which shows that at 400 K, the brightness of lines connecting to v = 0, 1, 2 and 3 does not vary significantly in the 1090–1180 Å window, when the H2 column does not exceed 1 × 1020 cm−2 . Indeed, line to line comparisons between models at 400 K at various H2 column values (1 × 1020 cm−2 ) did not show substantial deviations. Comparisons between the observation and models started to exhibit discrepancies for columns higher than 1 × 1020 cm−2 , increasing the χ 2 values at the same time. This reasoning demonstrates that H2 models in this bandwidth are not sensitive to H2 columns 1 × 1020 cm−2 . The slant H2 column of 1 × 1020 cm−2 of the best fit should then be considered as an upper limit of the absorbing H2 column. It should be said that this reasoning is valid for temperatures near 400 K. Fig. 15 of Gustin et al. (2004a) clearly shows that at 900 K, the H2 column starts to change the spectrum shape for H2 columns near 1 × 1017 cm−2 . We also considered models for which the temperature of the emitting and absorbing 5. Summary and conclusions Comparisons between modeled and observed H2 spectra are an effective tool which provides qualitative and quantitative information on the auroral characteristics. Two atmospheric processes have been used to derive the energy of the precipitating electrons in Saturn’s atmosphere: absorption by hydrocarbons through FUV spectra, and self-absorption through EUV spectra. The temperature of the absorbing H2 molecules can also be derived from self-absorption, provided the spectral resolution is sufficiently high, which is the case with the FUSE observations. STIS and UVIS FUV spectra were found to be very little absorbed by methane, with vertical CH4 columns less than or equal to 1.2 × 1016 cm−2 . Using the low-latitude atmospheric model of Moses et al. (2000), the CH4 column derived from STIS and UVIS FUV have been used to estimate the H2 column overlying the emission hence the depth of the auroral emission. From a relationship between electron energy and H2 column, we find that the energy of the primary electrons lies in the range ∼10–18 keV. This is consistent with the value of ∼10 keV obtained by Sandel et al. (1982) from a particularly bright H2 spectrum of Saturn’s aurora obtained with Voyager UVS, absorbed by a methane column of 8 × 1015 cm−2 . Besides, the other, weaker UVS spectra they analyzed did not require any CH4 absorption in the modeling process. This suggests that the peak of the UV auroral emission originates above but close to the methane homopause, and primary energies near ∼18 keV should be seen as an upper limit value. The variability of the methane absorption from one case to another can be explained by the intrinsic variability of the aurora, as described in the introduction. In the EUV, FUSE and UVIS spectra clearly exhibit self-absorbed H2 bands. Comparison between the un-self-absorbed model and the UVIS EUV spectrum we analyzed demonstrates strong attenuation of lines connecting to vibrational levels 0 and 1 of X 1 Σ g+ , and an intensification of lines connecting to v > 1. The model best fitting the UVIS EUV spectrum is characterized by a temperature– vertical H2 column pair of 700 K–2.4 × 1020 cm−2 . Because of its relatively low spectral resolution (∼5.5 Å), this spectrum does not resolve the H2 rotational lines and the T –H2 column pair providing the best fit to UVIS EUV is not unique. The χ 2 boundaries used to limit the temperature and H2 column ranges were found to be insufficient to constrain their extent to a satisfactory level and provide valuable quantitative results. Alternatively, the high resolution of the FUSE EUV spectra allows the determination of a H2 rotational temperature. The LiF1a spectrum qualitatively supports conclusions based on UVIS EUV data, with attenuation of lines connecting to v = 0 and 1. It is best fitted with a model at 500 K and 3.0 × 1019 cm−2 of H2 . The LiF2a spectrum does not reveal self-absorbed lines and is best fitted with a model characterized by a H2 temperature of 400 K. Most of the lines within the LiF2a bandwidth connect to v = 2, and the absence of self-absorption indicates that this vibrational level is virtually not populated at 400 K, as expected from the Maxwell–Boltzmann vibrational distribution used in this study. At 400 K, self-absorption is detectable in the LiF2a segment for a H2 column of 6.0 × 1019 cm−2 , which sets the H2 column upper limit to this latter value. The H2 columns deduced from FUSE LiF1a and LiF2a lead to primary auroral electron energies of 10 and 15 keV, respectively. The best fits to the FUSE spectra were obtained assuming LTE, and no evidence of departure from thermodynamical equilibrium has been detected. This result is in agreement with Melin et al. (2007), who fitted H+ 3 infrared (IR) spectra adopting UV spectroscopy of Saturn’s aurora with STIS, FUSE and UVIS 185 Fig. 7. FUSE LiF2a spectrum and best fit model at T = 400 K without self-absorption. A H2 vertical column of 6 × 1019 cm−2 establishes its upper limit, since models near 400 K are not significantly sensitive to H2 vertical column below this value. Pumped lines at 1118.61, 1119.08, 1163.80 and 166.76 Å are identified as L (6, 2) P (1), W (1, 3) Q (3), W (1, 4) Q (3) and L (6.3) P (1), respectively (Liu and Dalgarno, 1996). LTE. We were able to obtain good fits assuming that the temperature of the emitting and absorbing molecules are equal. This suggests that self-absorption occurs at the altitude of the emission. Owing to the absence of self-absorption in the LiF2a spectrum, this segment is ideal to derive the H2 temperature. The 400 K we find is in accord with the 350–500 K obtained from H+ 3 IR spectra, measured at the H+ 3 peak altitude by Melin et al. (2007). These values are significantly higher than the ∼125 K expected from the Moses et al. (2000) equatorial model near 0.15 μbar, which suggests that an equatorial temperature profile is not suited to polar latitudes. This can be explained by additional heat sources expected at the poles, such as Joule heating, particle precipitation heating and ion drag heating, in addition to the solar EUV input. This scenario has been studied by Smith et al. (2005). From a 3-D general circulation model, they were able to reproduce both the exospheric temperature of ∼400 K at 30◦ N derived from Voyager solar occultation measurement (Smith et al., 1983) and the auroral H+ 3 temperature of 350–500 K obtained by Melin et al. (2007), assuming a polar heating input of ∼8 TW peaking at ∼60 nbars. Our analysis shows that methane absorption in the FUV and self-absorption in the EUV, which are independent processes, both provide the same range of primary electron mean energies, from ∼10 to ∼18 keV, corresponding to an auroral emitting layer located between 0.10 and 0.28 μbar as illustrated in Fig. 8. This 186 J. Gustin et al. / Icarus 200 (2009) 176–187 Fig. 8. Equatorial atmospheric model of Moses et al. (2000). The horizontal lines show the pressure level of the auroral emission peak as listed in Table 1, determined from UVIS-FUV, STIS and FUSE spectra. The two dashed lines corresponding to STIS represent the minimum and maximum pressure levels deduced from the six STIS spectra analyzed. Because of the too broad range of H2 columns derived from the UVIS-EUV spectrum, a precise pressure level has not been obtained. agreement suggests that the methane mixing ratio profile from the Moses et al. (2000) equatorial model is, to a first approximation, relevant to describe the polar atmosphere. On the other hand, the auroral temperatures we obtain suggest that the Moses et al. (2000) temperature profile is not appropriate for polar regions. This aspect is important, since higher temperatures not only affect the hydrocarbon chemistry but also the atmospheric structure and homopause level. It is difficult to infer how a realistic polar atmospheric model would modify the primary electron energy derived from methane absorption, since multiple parameters (that may have opposite effects) are involved. This facet of the problem is beyond the scope of this paper, and it should be emphasized that FUV results are tied to a given methane mixing ratio profile, which is not well known in polar regions. An estimate of the total unabsorbed H2 brightness in the 800– 1700 Å bandwidth has been derived from the synthetic spectra that best fit the observations. We obtain spatial and temporal averaged brightnesses from 2.9 to 14 kR for STIS and UVIS, in the range of the 1 to 100 kR determined by Clarke et al. (2005) from STIS FUV images. Higher brightnesses are derived from FUSE spectra (129 and 139 kR for LiF1a and LiF2a, respectively), similar to the 150 kR value obtained by Gérard et al. (1995) using the Faint Object Camera (FOC). When compared to results obtained from the analysis of STIS (Gérard et al., 2003) or FUSE (Gustin et al., 2004a) spectra of jovian aurora, the present study shows that the H2 columns at the auroral peak are one order of magnitude lower for Saturn, with a bulk of the energy deposition situated higher in the atmosphere. Indeed, the energy of the precipitated electrons in the case of Saturn is modest, compared to the 40–210 keV determined for Jupiter. This is consistent with theoretical studies, predicting accelerated magnetospheric electrons of a few keV to a few tens of keV in the case of Saturn (Cowley et al., 2004b, 2004a) and primary electron energies of order 50–100 keV in the case of Jupiter (Cowley and Bunce, 2001; Cowley et al., 2005). It should be noted that our limited set of observations, with a total of 8 spectra from the south dayside aurora and 2 spectra from the north nightside aurora, does not allow us to formulate a statistically significant picture of the auroral characteristics with respect to the location (north/south, day/night) and time of the emission. The variations of the auroral character- istics with time and location is an important point that should be examined in a near future with the help of more recent observations made by the UVIS spectrometer. Acknowledgments We thank the two referees for their valuable comments and suggestions. This work is based on observations with the NASA/ESA Hubble Space Telescope, obtained at the Space Telescope Science Institute (STScI), which is operated by the AURA, Inc., for NASA, observations by the NASA–CNES–CSA FUSE mission operated by the Johns Hopkins University and observations by the Cassini mission. Financial support to US participants was provided by NASA Contract NAS5-32985 and with the Cassini Project. J.-C.G. and D.G. acknowledge support from the Belgian Fund for Scientific Research (FNRS). The PRODEX program of ESA provided financial support for this research. We also thank Julianne Moses for providing us with an electronic version of her model of Saturn’s atmosphere, Janet Hallet for providing us with a synthetic spectrum of H2 airglow and Kristopher Larsen for the help in Fig. 2. References Ajello, J.M., Pryor, W.R., Esposito, L.W., Stewart, A.I.F., McClintock, W.E., Gustin, J., Grodent, D., Gérard, J.-C., Clarke, J.T., 2005. The Cassini Campaign observations of the Jupiter Aurora by the Ultraviolet Imaging Spectrograph and the Space Telescope Imaging Spectrograph. Icarus 78, 327–345. Badman, S.V., Cowley, S.W.H., Gérard, J.-C., Grodent, D., 2006. A statistical analysis of the location and width of Saturn’s southern auroras. Ann. Geophys. 24, 3533– 3545. Broadfoot, A.L., Sandel, B.R., Shemansky, D.E., Holberg, J.B., Smith, G.R., Strobel, D.F., McConnell, J.C., Kumar, S., Hunten, D.M., Atreya, S.K., Donahue, T.M., Moos, H.W., Bertaux, J.L., Blamont, J.E., Pomphrey, R.B., Linik, S., 1981. Extreme ultraviolet observations from Voyager 1 encounter with Saturn. Science 212, 206–211. Bunce, E.J., Arridge, C.S., Clarke, J.T., Cowley, S.W.H., Dougherty, M.K., Gérard, J.-C., Grodent, D., Hansen, K.C., Nichols, J.D., Southwood, D.J., Talboys, D.L., 2008. Origin of Saturn’s aurora: Simultaneous observations by Cassini and the Hubble Space Telescope. J. Geophys. Res. 113, doi:10.1029/2008JA013257. A09209. Clarke, J.T., Gérard, J.-C., Grodent, D., Wannawichian, S., Gustin, J., Connerney, J., Crary, F., Dougherty, M., Kurth, W., Cowley, S.W.H., Bunce, E.J., Hill, T., Kim, J., 2005. Morphological differences between Saturn’s ultraviolet aurorae and those of Earth and Saturn. Nature 433, 717–719. UV spectroscopy of Saturn’s aurora with STIS, FUSE and UVIS Cowley, S.W.H., Bunce, E.J., 2001. Origin of the main auroral oval in Jupiter’s coupled magnetosphere–ionosphere system. Planet. Space Sci. 49, 1067–1088. Cowley, S.W.H., Bunce, E.J., Prangé, R., 2004a. Saturn’s polar ionospheric flows and their relation to the main auroral oval. Ann. Geophys. 22, 1379–1394. Cowley, S.W.H., Bunce, E.J., O’Rourke, J.M., 2004b. A simple quantitative model of plasma flows and currents in Saturn’s polar ionosphere. J. Geophys. Res. 109, doi:10.1029/2003JA010375. A05212. Cowley, S.W.H., Alexeev, I.I., Belenkaya, E.S., Bunce, E.J., Cottis, C.E., Kalegaev, V.V., Nichols, J.D., Prangé, R., Wilson, F.J., 2005. A simple axi-symmetric model of magnetosphere–ionosphere coupling currents in Jupiter’s polar ionosphere. J. Geophys. Res. 110, doi:10.1029/2005JA011237. A11209. Crary, F.J., Clarke, J.T., Dougherty, M.K., Hanlon, P.G., Hansen, K.C., Steinberg, J.T., Barraclough, B.L., Coates, A.J., Gérard, J.-C., Grodent, D., Kurth, W.S., Mitchell, D.G., Rymer, A.M., Young, D.T., 2005. Solar wind dynamic pressure and electric field as the main factors controlling Saturn’s aurorae. Nature 433, 720–722. Dols, V., Gérard, J.-C., Clarke, J.T., Gustin, J., Grodent, D., 2000. Diagnostics of the jovian aurora deduced from ultraviolet spectroscopy: Model and HST/GHRS observation. Icarus 147, 251–266. Esposito, L., and 18 colleagues, 2004. The Cassini Ultraviolet Imaging Spectrograph investigation. Space Sci. Rev. 115, 299–361. Gérard, J.-C., Dols, V., Grodent, D., Waite, J.H., Gladstone, G.R., Prangé, R., 1995. Simultaneous observations of the saturnian aurora and polar haze with the HST/FOC. Geophys. Res. Lett. 22, 2685–2688. Gérard, J.-C., Gustin, J., Grodent, D., Delamere, P., Clarke, J.T., 2002. The excitation of the FUV Io tail on Jupiter: Characterization of the electron precipitation. J. Geophys. Res. 107 (A11), doi:10.1029/2002JA009410. 1394–1400. Gérard, J.-C., Gustin, J., Grodent, D., Clarke, J.T., Grard, A., 2003. Spectral observations of transient features in the FUV jovian polar aurora. J. Geophys. Res. 108 (A8), 1319–1329, doi:10.1029/2003JA009901. Gérard, J.-C., Grodent, D., Gustin, J., Saglam, A., Clarke, J.T., Trauger, J.T., 2004. Characteristics of Saturn’s FUV aurora observed with the Space Telescope Imaging Spectrograph. J. Geophys. Res. 109, doi:10.1029/2004JA010513. A09207. Gérard, J.-C., Grodent, D., Cowley, S.W.H., Mitchell, D.G., Kurth, W.S., Clarke, J.T., Bunce, E.J., Nichols, J.D., Dougherty, M.K., Crary, F.J., Coates, A.J., 2006. Saturn’s auroral morphology and activity during quiet magnetospheric conditions. J. Geophys. Res. 111, doi:10.1029/2006JA011965. A12210. Grodent, D., Waite Jr., J.H., Gérard, J.C., 2001. A self-consistent model of the jovian auroral thermal structure. J. Geophys. Res. 106, 12933–12952. Grodent, D., Gérard, J.-C., Cowley, S.W.H., Bunce, E.J., Clarke, J.T., 2005. The global morphology of Saturn’s southern ultraviolet aurora. J. Geophys. Res. 110, doi:10.1029/2004JA010717. A01207. Gustin, J., Grodent, D., Gérard, J.C., Clarke, J.T., 2002. Spatially resolved far ultraviolet spectroscopy of the jovian aurora. Icarus 156, 91–103. Gustin, J., Feldman, P.D., Gérard, J.-C., Grodent, D., Vidal-Madjar, A., Ben Jaffel, L., Desert, J.-M., Moos, H.W., Sahnow, D.J., Weaver, H.A., Wolven, B.C., Ajello, J.M., Waite, J.H., Roueff, E., Abgrall, H., 2004a. Jovian auroral spectroscopy with FUSE: Analysis of self-absorption and implications for electron precipitation. Icarus 171, 336–355. 187 Gustin, J., Gérard, J.-C., Grodent, D., Cowley, S.W.H., Clarke, J.T., Grard, A., 2004b. Energy–flux relationship in the FUV jovian aurora deduced from HST-STIS spectral observations. J. Geophys. Res. 109, doi:10.1029/2033JA010365. Gustin, J., Cowley, S.W.H., Gérard, J.-C., Gladstone, G.R., Grodent, D., Clarke, J.T., 2006. Characteristics of jovian morning bright FUV aurora from Hubble Space Telescope/Space Telescope Imaging Spectrograph imaging and spectral observations. J. Geophys. Res. 111, doi:10.1029/2006JA011730. A09220. Hallett, J.T., Shemansky, D.E., Liu, X., 2005. A rotational-level hydrogen physical chemistry model for general astrophysical application. Astrophys. J. 624, 448– 461. Liu, W., Dalgarno, A., 1996. The ultraviolet spectra of the jovian aurora. Astrophys. J. 467, 446–453. Kurth, W.S., Gurnett, D.A., Clarke, J.T., Zarka, P., Desch, M.D., Kaiser, M.L., Cecconi, B., Lecacheux, A., Farrell, W.M., Galopeau, P., Gérard, J.-C., Grodent, D., Prangé, R., Dougherty, M.K., Crary, F.J., 2005. An Earth-like correspondence between Saturn’s auroral features and radio emission. Nature 433, 722–725. Melin, H., Miller, S., Stallard, T., Trafton, L.M., Geballe, T.R., 2007. Variability in the H 3+ emission of Saturn: Consequences for ionisation rates and temperature. Icarus 186, 234–241. Moos, H.W., and 56 colleagues, 2000. Overview of the Far Ultraviolet Spectroscopic Explorer Mission. Astrophys. J. 538, 1–6. Moses, J.I., Greathouse, T.K., 2005. Latitudinal and seasonal models of stratospheric photochemistry on Saturn: Comparison with infrared data from IRTF/TEXES. J. Geophys. Res. 110, doi:10.1029/2005JE002450. E09007. Moses, J.I., Bézard, B., Lellouch, E., Feuchtgruber, H., Gladstone, G.R., Allen, M., 2000. Photochemistry of Saturn’s atmosphere. I. Hydrocarbon chemistry and comparisons with ISO observations. Icarus 143, 244–298. Ollivier, J.L., Dobrijevic, M., Parisot, J.P., 2000. New photochemical model of Saturn’s atmosphere. Planet. Space Sci. 48, 699–716. Sandel, B.R., Broadfoot, A.L., 1981. Morphology of Saturn’s aurora. Nature 292, 679– 682. Sandel, B.R., Shemansky, D.E., Broadfoot, A.L., Holberg, J.B., Smith, G.R., 1982. Extreme ultraviolet observations from the Voyager 2 encounter with Saturn. Science 215, 548–553. Sahnow, D.J., and 40 colleagues, 2000. On-orbit performance of the Far Ultraviolet Spectroscopic Explorer Satellite. Astrophys. J. 538, 7–11. Shemansky, D.E., Ajello, J.M., 1983. The Saturn spectrum in the EUV: Electron excited hydrogen. J. Geophys. Res. 88, 459–464. Smith, G.R., Shemansky, D.E., Holberg, J.B., Broadfoot, A.L., Sandel, B.R., McConnell, J.C., 1983. Saturn’s upper atmosphere from the Voyager 2 EUV solar and stellar occultations. J. Geophys. Res. 88, 8667–8678. Smith, C.G.A., Aylward, A., Miller, S., Müller-Wodarg, I.C.F., 2005. Polar heating in Saturn’s thermosphere. Ann. Geophys. 23, 2465–2477. Steffl, A.J., Stewart, A.I.F., Bagenal, F., 2004. Cassini UVIS observations of the Io plasma torus. I. Initial results. Icarus 172, 78–90. Wolven, B.C., Feldman, P.D., 1998. Self-absorption by vibrationally excited H2 in the Astro-2 Hopkins Ultraviolet Telescope spectrum of the jovian aurora. Geophys. Res. Lett. 25, 1537–1540.