Detection of New Auroral Emissions at Io and Implications for Its Interaction with the Plasma Torus

Z. Milby et al. (2024) analyzed similar HIRES spectra of Ganymede in eclipse and found that the 630.0 nm/636.4 nm [O i] emission-line ratio deviated from the optically thin ratio in a way that could not be explained by collisional de-excitation. As a result, they added a 9% systematic uncertainty to all of their auroral brightness measurements. We evaluated the same 630.0 nm/636.4 nm [O i] emission-line ratio in the Io eclipse spectra to determine whether we needed to include any systematic error. However, we found that the ratio was within one standard deviation of the expected ratio of 3.09 (W. L. Wiese et al. 1996) and subsequently did not include any additional systematic uncertainty for any of the auroral brightnesses reported in this study.

All of the brightnesses listed in this study are disk integrated and then averaged across all spectra used for each night. Though each measurement is independent, there is inherent variability in the brightnesses beyond expected Poisson noise due to short-timescale changes in the number density of the electrons exciting the auroral emissions and/or the atmospheric column densities. We therefore chose to calculate weighted brightness averages rather than arithmetic brightness averages so that measurements with relatively larger uncertainties are appropriately downweighted. We calculated the weighed average brightness of n brightness measurements as

where B_i is the ith measurement and is the inverse of the square of its corresponding measured uncertainty σ_i (equivalently the inverse variance of B_i). The propagated uncertainty we report for the weighted average is

We have further expanded the auroral emission model used by K. de Kleer et al. (2023) and Z. Milby et al. (2024) with additional cross sections relevant to Io’s atmospheric composition. For electron impact on SO₂ we have included O i emission cross sections at 130.4 and 135.6 nm (P. Vatti Palle et al. 2004). Because the O i 777.4 and 844.6 nm emissions cascade into these UV lines, J. M. Ajello et al. (2008) indicated that the 777.4 and 844.6 nm emission cross sections could be approximated by scaling the 130.4 and 135.6 nm cross sections using values they provided. We used the excitation cross section for electron impact on SO₂ producing O(¹S) from W. Kedzierski et al. (2000) to calculate emission cross sections for 297.2 and 557.7 nm using their relative emission probabilities (W. L. Wiese et al. 1996). To date, no measurements of excitation cross sections for electron impact on SO₂ producing O(¹D) or similar emission cross sections for 630.0 or 636.4 nm [O i] have been published; however, the W. Kedzierski et al. (2000) 557.7 nm cross section quantifies the cascade contribution, effectively setting a lower limit. No emission or excitation cross sections have been published for electron impact on SO (J. W. McConkey et al. 2008).

We also used ChantiPy, the Python interface to the CHIANTI atomic database v10.1 (K. P. Dere et al. 1997, 2023), to calculate photon emission-line ratios from atomic and ionic columns as a function of electron energy and density. CHIANTI includes many of the emissions listed in Table 2, with the exception of those from the neutral alkali atoms Na i and K i and the electric dipole O i transitions at 777.4 and 844.6 nm and S i transition at 922.3 nm.

Our analysis of the HIRES spectra revealed emission lines from a variety of neutral and singly ionized atoms, including O, O⁺, Na, S, S⁺, and K. This complete set of emissions includes the first Io eclipse detections of the O i lines at 777.4 and 844.6 nm, the Na i lines at 818.3 and 819.5 nm, the [S i] lines at 458.9 and 772.5 nm, the S i triplet at 922.3 nm, the [O ii] lines at 732.0 and 733.0 nm, and the [S ii] lines at 406.9, 407.6, 671.6, and 673.1 nm. Figure 2 shows examples of what these emissions (along with the previously identified emissions) look like in the reduced HIRES detector images. Io’s angular width was nearly that of the slit (see Figure 1), so any emission from an extended corona would appear only in the vertical direction in the detector images.

Zoom In Zoom Out Reset image size

Table 2 lists each of the emission lines for which we attempted to detect auroral emission, along with the average brightness for those found at a signal-to-noise ratio of 2 or greater. For nondetections (negative brightness or signal-to-noise ratio below 2) we have instead reported the 2σ upper limit for the brightness. Brightnesses are in units of rayleighs (R), defined as

When comparing to brightnesses published by others, we occasionally used kilorayleighs (kR), where 1 kR = 1000 R. Significant order overlap prevented us from calculating any emission brightness for wavelengths shorter than about 380 nm. Additionally, for our choice of echelle and cross-disperser angles, the K i emission at 769.9 nm did not fall onto the detector, though it should have been at a detectable brightness given the previously observed brightness ratio between the potassium D lines (C. Schmidt et al. 2023).

Because HIRES cannot operate with both HIRESr and HIRESb at the same time (or even on the same night), we were unable to simultaneously observe the shorter- and longer-wavelength emissions. Additionally, the inherent temporal variability in the total brightnesses precludes meaningful direct comparisons. However, when we compared the relative brightness of the [O i] 557.7 nm emission line and the [S i] electronic equivalent at 772.5 nm, we found that they had a Pearson correlation coefficient of 0.917 with a p-value of 0.0282, which qualifies as statistically significant beyond a standard 95% confidence threshold. We also compared the relative brightness of the [O i] 557.7 nm emission line to the [S ii] 673.1 nm emission line and found that they had a Pearson correlation coefficient of 0.813 with a p-value of 0.0942, statistically significant for a 90% confidence threshold.

This indicates that the absolute brightnesses of the sulfur and oxygen emissions mostly covary, and therefore the brightnesses and uncertainties derived from HIRESr observations can be scaled to approximate their expected values on 2023 August 9. Since we were able to retrieve the 557.7, 630.0, and 636.4 nm [O i] brightnesses with both cross-dispersers, we used the relative weighted average brightness of these three oxygen emissions to calculate scaling factors and uncertainties of 1.1 ± 0.2 for 2022 November 24, 0.92 ± 0.17 for 2023 August 25, 0.83 ± 0.15 for 2024 September 12, 0.73 ± 0.12 for 2024 October 5, and 0.88 ± 0.16 for 2024 October 21.

The brightnesses we retrieved from our HIRES observations vary substantially from those reported by A. H. Bouchez et al. (2000), potentially indicating large secular variability in aurora brightness. They found a 557.7 nm [O i] brightness of 1.3 ± 0.2 kR, a factor of 3.0 ± 0.5 larger than the weighted average brightness of 0.429 ± 0.004 kR averaged over the six nights in our data. We found a similar factor of 3.0 ± 0.4 for the comparison of the 630.0 nm [O i] brightness and a slightly larger factor of 3.2 ± 0.5 for the 636.4 nm [O i] brightness. However, for the sodium doublet, their results were dimmer by a factor of 0.28 ± 0.04 for the 589.0 nm Na i D₁ line and by a factor of 0.29 ± 0.04 for the 589.6 nm Na i D₂ line. These differences point to stochastic variability in atmospheric column density (discussed further in Section 4.5).

The weighted average brightness we found for the O(¹D) doublet was 7.75 ± 0.02 kR, which exceeds the 4.8 kR average reported by C. Schmidt et al. (2023) despite cross-calibration of our two analysis routines (our pipeline and theirs produced comparable brightnesses when reducing the same HIRES data). This suggests that the difference is a real change in brightness and not a systematic difference in our respective flux calibrations. The eclipsed O(¹D) doublet brightness is less than the 10.5 kR average reported in the sunlit measurements by R. J. Oliversen et al. (2001). We found no evidence for time dependence of the O emissions during the eclipse phase, while the Na emissions do show a strong time dependence due to the sudden loss of solar photochemical production pathways (C. Schmidt et al. 2023). Both R. J. Oliversen et al. (2001) and C. Schmidt et al. (2023) found weak correlations between O(¹D) emission and Io’s location relative to the plasma torus, but they observed considerable variance by up to a factor of 3 in the brightness at a given magnetic longitude.

The identification of so many discrete emission features provided the opportunity to better understand Io’s species-dependent auroral morphology by comparing with images of Io taken in eclipse, first done by A. H. Bouchez et al. (2000) when they compared their five detected emission lines to images of Io in eclipse taken by the Solid State Imaging experiment on the Galileo Orbiter spacecraft (P. E. Geissler et al. 1999).

On its way to Saturn, the Narrow Angle Camera (NAC) of Cassini’s Imaging Science Subsystem (ISS; C. C. Porco et al. 2004) took both broadband and narrowband images of Io in eclipse during its flyby of the Jovian system (see Figure 3 for a broadband clear filter example image and Figure 4 for a description of the filter bandpasses). ISS imaged Io’s trailing hemisphere, in contrast to the HIRES eclipse observing geometry, which is restricted to the sub-Jovian hemisphere. Figure 5 of P. Geissler et al. (2004) shows images taken under 15 different filters and filter combinations on 2001 January 5, nine of which (shown in our Figure 5) had bandpasses that contained wavelengths of the auroral emissions we detected in the HIRES spectra. ISS took four sequences of images with exposure times of 12 s each. The sequences took about 14 minutes and 22 s to complete, and they began and ended with clear filter images. We chose to display images taken during the second sequence in Figure 5, due to their high signal-to-noise ratio. Figure 6 shows the filter sequence sorted by the exposure midpoint time of each image relative to eclipse ingress, which occurred at 2001 January 5 11:03 UTC.

Zoom In Zoom Out Reset image size

Zoom In Zoom Out Reset image size

Zoom In Zoom Out Reset image size

Zoom In Zoom Out Reset image size

These images demonstrate that Io’s optical aurora exhibits some combination of two primary morphological features: the sub/anti-Jovian spots (also called the “equatorial glow”) that were first identified in UV imaging (e.g., F. L. Roesler et al. 1999; K. D. Retherford et al. 2000, 2003, 2007; J. Saur et al. 2000; L. Roth et al. 2011, 2014, 2017), and a diffuse limb glow (e.g., K. D. Retherford et al. 2003; C. Moore et al. 2010). Because the HIRES data do not have the spatial resolution to resolve spots versus a diffuse equatorial enhancement distributed across all visible longitudes, we used the term “equatorial glow” throughout our analysis instead of “sub-Jovian spot” when referring to emission features in the HIRES data (we still used the term for the ISS images). Further, despite its name, the sub-Jovian spot actually appears up to 20° west (wakeward) of the sub-Jovian longitude (K. D. Retherford et al. 2007; C. Moore et al. 2010). Regardless, the FWHMs of the bright emissions in Figure 2 are smaller than the angular size of Io’s disk (indicated by the graticules in the lower left corner of each panel), so the observed emissions are spatially isolated and likely to be primarily from the sub-Jovian spot.

The aurora also shows transient thermal and/or thermally excited emissions from discrete volcanic features such as Reiden Patera, the volcano Pele, and a spot near the north pole that they identify as being associated with the Tvashtar Paterae volcanic plume. Reiden Patera and Pele appear in the CLR (clear) filter image (a combination of filters CL1 and CL2) and are increasingly more prominent at longer wavelengths owing to their thermal emission. The Tvashtar Paterae plume appears under most filters. Figure 3 shows an example ISS CL1+CL2 (clear) filter combination image containing each of these emission features, while Figure 4 shows the transmission bandpasses of the nine filters and the locations of the auroral emissions identified in the HIRES spectra.

P. Geissler et al. (2004) reported the brightnesses of specific auroral morphologies like the anti-Jovian equatorial glow (their Tables 1 and 5) and the limb glow (their Table 3). They did not report disk-integrated brightnesses, which make comparisons with the disk-integrated HIRES brightnesses difficult. However, they did report disk-integrated total photon fluxes and uncertainties in their Table 4. They calculated these fluxes over apertures that included all apparent emission—the equatorial spots, the limb glow, and any potential extended corona—and converted them to total photon flux emitted into 4π sr around the entire satellite. We took these flux values (they did not specify units, so we assumed photons s⁻¹), divided by the cross-sectional area of Io (1.0514 × 10¹² m²) and 4π sr, and then converted to kR using Equation (3). We have included these brightnesses in Table 3.

The BL1 filter image brightness is significantly larger than that derived from the HIRES data. P. Geissler et al. (2004) attributed the brightness in the BL1 filter image to electron-impact quasi-continuum fluorescence of SO₂, though they also proposed some contribution from the S₂(B ⟶ X ) band system to account for the larger-than-expected brightness in this filter relative to the UV1 filter. L. M. Trafton et al. (2012) also observed molecular emission bands in this wavelength region from Io in eclipse.

The GRN filter image contains both the 557.7 nm [O i] and 589.0 and 589.6 nm Na i doublet emissions, while the RED+GRN combination filter contains just the Na i doublet emissions. These images were taken only about 1 minute apart (see Figure 6), so the fact that they recorded nearly identical brightnesses suggests that the [O i] emission dominates in the GRN filter.

The CB1 filter image contains just the 630.0 and 636.4 nm [O i] emissions, which shows that the oxygen emission primarily comes from the equatorial and limb glows and does not seem to be correlated with the volcanic plume. This conclusion is further supported by the RED filter image, which contains the 589.0 and 589.6 nm Na i doublet, the 630.0 and 636.4 nm [O i] lines, and the 671.6 and 673.1 nm [S ii] lines. The strong plume emission must come from the Na i doublet and [S ii] lines, suggesting that the plume emission in the GRN filter is probably from the Na i doublet and not from the 557.7 nm [O i] line, though the different lifetimes between the O(¹S) and O(¹D) states could lead to preferential quenching of the 630.0 and 636.4 nm [O i] lines in the high-density plume. The RED+IR1 filter combination image contains the 671.6 and 673.1 nm [S ii] and 732.0 and 733.0 nm [O ii] emissions, but they fall near the edges of the bandpass, accounting for the nearly featureless image.

The limb glow in the IR1 filter is likely from the O i 777.4 and 844.6 nm emissions and perhaps also from the 772.5 nm [S i] line (the sulfur electronic analog to the 557.7 nm [O i] line). P. Geissler et al. (2004) proposed that the largest contribution to the total brightness in the IR1 filter came from the 766.4 nm K i line, which agrees with the HIRES data, where it accounted for about half of the total disk-integrated brightness across the filter’s bandpass. The IR2 and IR3 filters contain the emissions identified in the HIRES spectra at wavelengths greater than 800 nm, but the signal-to-noise ratio is quite low, and only the equatorial glows are significant.

Table 3 summarizes the species we attribute to the observed morphologies in each filter. Overall, the equatorial glow seems to include emission from all detected species, both neutrals and ions. The limb glow seems to be associated with the bright neutral oxygen emissions, though we cannot rule out a contribution from neutral sulfur as well, which may be the species contributing to the extremely faint limb glows in the BL1, IR2, and IR3 filters. The concentration of molecular species at low to midlatitudes (D. F. Strobel & B. C. Wolven 2001) and global atomic coronae (B. C. Wolven et al. 2001) suggest that dissociative electron impact on molecules like SO₂ and SO produces the bright auroral spots while excitation of the atomic S and O coronae produces the limb glow. Emission from the Tvashtar plume is probably primarily from sodium but may also include emission from ionized atomic oxygen and sulfur.

Emission-line ratios are sensitive to different electron energy distributions because of their individual energy-dependent excitation cross sections. In contrast, absolute auroral brightnesses are sensitive to both electron energy distributions and the product of electron number densities and atmospheric column densities. C. Schmidt et al. (2023) evaluated the 630.0 nm/557.7 nm [O i] brightness ratio in observations taken primarily with the ARC Echelle Spectrograph (ARCES) at the Apache Point Observatory and found that it was higher than expected from electron impact on a column of just atomic oxygen. They attributed the larger ratio to electrons losing energy as they precipitated through Io’s SO₂ atmosphere, allowing them to preferentially excite the 630.0 nm red line emission (which has a lower excitation energy than the 557.7 nm emission). Models like those of J. Saur et al. (1999) and V. Dols et al. (2012) show electron energies decreasing from a typical upstream torus energy of 5 eV down to 0.2 eV near Io’s surface. Figure 7 shows this brightness ratio for electron impact on atomic oxygen as a function of electron energy and density. For electron impact on O the emission-line ratio is sensitive to energy but independent of plasma density for the range of values plausible for Io. Though the impact of 3 to 4 eV electrons on a column of atomic oxygen can explain the observed 630.0 nm/557.7 nm [O i] brightness ratios (Figure 7 shows the average ratio of 13.64 ± 0.13), it cannot account for the ratios with the other oxygen emissions (777.4 and 844.6 nm) detected in the HIRES spectra. While we will present one potential explanation in Section 4.2, both quenching and electron energy variation with latitude, with altitude, and within plumes are plausible alternative explanations.

Zoom In Zoom Out Reset image size

In order to explore how the emissions change with other possible atmospheric compositions, we compared brightness calculations from our auroral emission model for three combinations: O alone (simulating emission from just the atomic oxygen column), an atmosphere of O and SO₂, and an atmosphere of O, SO₂, and a third molecular component. Because of the limited cross sections available for SO₂, we were only able to evaluate three of the five detected oxygen lines: the forbidden 557.7 nm emission and allowed 777.4 and 844.6 nm emissions. We used a Maxwellian electron energy distribution centered at 5 eV to determine whether the observed brightness ratios could be explained through impact by the upstream electron population. Table 4 lists the results produced by our aurora model for the five different atmospheric compositions discussed below.

The single-component atmosphere composed of just O alone cannot replicate the observed brightness ratios of the three emission lines: the best fit underpredicts the 777.4 nm brightnesses and overpredicts the 557.7 and 844.6 nm brightnesses. Including SO₂ improves the fit overall but still fails to reproduce the observed brightnesses within their respective uncertainties, resulting in an overprediction of the 844.6 nm emission and an underprediction of the 777.4 nm emission. However, including O₂ results in an excellent fit for all three emission features well within their observed uncertainties. However, this result yields O₂ column densities of approximately 10¹⁹ m⁻² (see Figure 8), which is about 2 orders of magnitude larger than the estimated 1% mixing ratio between O₂ and SO₂ (J. I. Moses et al. 2002). This probably means that the observed auroral emissions are from a combination of electron impact on exospheric columns of O, SO₂, and a molecule that behaves like O₂ when dissociated into excited fragments by electron impact. The most likely candidate molecule is SO, which has a mixing ratio with SO₂ of around 10% (M. A. McGrath et al. 2000) and may be higher during eclipse when the SO₂ atmosphere condenses onto the surface (C. C. C. Tsang et al. 2016). Unfortunately, emission and/or excitation cross sections have not yet been published for SO (J. W. McConkey et al. 2008), so it cannot be explicitly included in our auroral emission model.

Zoom In Zoom Out Reset image size

S and O occupy the same group in the periodic table, the valence electrons in the p orbitals are the same for both molecules, and therefore their overall state distribution is similar. They also have comparable dissociation energies: approximately 5.1 eV for O₂ and 5.4 eV for SO (B. Darwent 1970). Theoretical calculations show similar excitation cross sections between O₂ and SO, though the SO cross sections tend to be larger owing to the physical size difference between the two molecules (J. S. Rajvanshi & K. L. Baluja 2010). We therefore chose to use the O₂ cross sections as a proxy for the SO cross sections. However, while the relative cross sections (and therefore the brightness ratios) between oxygen emission lines will be similar, the absolute cross sections may not be of the same magnitude. Consequently, while the modeled component from O₂ probably includes emission primarily from SO (along with any trace O₂ present), we cannot determine their relative contributions owing to the inherent degeneracy. As a result, while the O and SO₂ column outputs are real, only the relative variability within the O₂ column density is real; the absolute column density magnitude is not physically meaningful and should not be interpreted as representative of either the modeled O₂ or SO column density. The true SO column density is likely smaller than the modeled result if the SO cross sections are systematically larger than the O₂ cross sections (J. S. Rajvanshi & K. L. Baluja 2010).

To ensure that this exceptionally good agreement from the inclusion of the O₂ component was not an artifact from fitting three emission lines with a three-species atmosphere, we fit two additional three-species atmospheres, substituting H₂O and CO₂ (the only other species for which we had 557.7, 777.4, and 844.6 nm emission cross sections) for the O₂ component. The other three-component atmospheres always resulted in a poor fit for at least one emission line. This could mean that the auroral emission is confined to high altitudes where electrons still maintain their 5 eV plasma torus energies. However, the Cassini/ISS images (specifically the CB1 image in Figure 5) seem to show the longest-lifetime 630.0 and 636.4 nm emissions all the way down to Io’s surface. Hence, while the observed oxygen aurora brightness ratios can be explained by electron impact on Io’s O, SO₂, and SO exospheres by 5 eV torus electrons without invoking the collisional cooling necessary for emission from a pure O column proposed by C. Schmidt et al. (2023), electron energies are much lower in the denser parts of the ionosphere near to Io’s surface (J. Saur et al. 1999; V. Dols et al. 2008, 2012, 2024), and therefore there exists a degeneracy between electron energy and composition that cannot be resolved using forbidden oxygen emissions from Io.

Regardless, the disk-integrated 557.7, 777.4, and 844.6 nm brightnesses are consistent with electron impact on an atmosphere composed of O, SO, and SO₂. For the 557.7 nm [O i] line, the fractional contribution to the total brightness from the SO component varies between 6% and 9%.

It remains unclear what fraction of Io’s SO₂ atmosphere comes from volcanic outgassing and what fraction comes from sublimation of surface ice. Some observations show a large decrease in column density during eclipse associated with the rapid drop in surface temperature, suggesting a predominantly sublimation-supported atmosphere where solar insolation maintains the vapor−pressure equilibrium between Io’s SO₂ atmosphere and surface ice (C. C. C. Tsang et al. 2016; I. de Pater et al. 2020). If the atmosphere is primarily sublimation supported, then during the eclipse phase of Io’s orbit a portion of the SO₂ atmosphere would freeze back onto the surface (J. Saur & D. F. Strobel 2004), perhaps even enough to change the near-surface bound atmosphere from collisional to noncollisional (C. C. C. Tsang et al. 2016).

Figure 8 shows the best-fit model column densities during eclipse for the three-species model containing O, O₂ (an unknown multiple of the SO column), and SO₂. While the O and O₂ (SO) column densities exhibit some short-timescale variability, they are relatively constant over the duration of the eclipse. Further, though there are some decreases in SO₂ column density with time (primarily between 30 and 75 minutes after eclipse ingress), most are followed by a proportional increase well before eclipse egress (typically around 2 hr after ingress). Though this could be explained by the collapse and recovery of some fraction of the SO₂ atmosphere (C. C. C. Tsang et al. 2016; I. de Pater et al. 2020), the correlation between proximity to Jupiter’s limb and the relative decrease in brightness suggests that this may instead be a background subtraction artifact.

The average O column densities are on the order of 3 × 10¹⁸ m⁻², about a factor of 4 larger than that in the V. Dols et al. (2024) model and a factor of 2 smaller than the column density derived by C. Schmidt et al. (2023). The maximum of the SO₂ column densities is approximately 2 × 10¹⁹ m⁻², a factor of about 30 smaller than that in the V. Dols et al. (2024) model and around the lower end of typical observed column densities (see Table 19.1 and associated references in M. A. McGrath et al. 2004 and Table 1 in K. de Kleer et al. 2024). Figure 9 shows how the O/SO₂ column mixing ratio changes over the course of the eclipses. The ratio is typically around 10% but varies by an order of magnitude or more and can approach parity owing to the drop in SO₂ column density near the eclipse midpoint.

Zoom In Zoom Out Reset image size

However, the relative stability of the SO₂ column and its density at the low end of previously derived values suggest that the portion of the emission from dissociative excitation of SO₂ may originate from the high-altitude portion of the column, while the near-surface emission comes primarily from direct excitation of the atomic oxygen component. This would permit the near-surface SO₂ atmosphere to collapse while maintaining the emitting column. This could also explain the apparent delay in the drop in SO₂ emission, since the lower atmosphere would have to collapse first before the higher-altitude emitting column. However, as noted above, this does not account for the apparent “recovery” that occurs in most of the observations after 1 hr in eclipse.

In such a scenario, electron energies near Io’s surface may be insufficient to dissociate SO₂ but are able to excite the lower-energy forbidden emissions in the atomic O, and the much higher electron number density within the ionosphere is sufficient to produce measurable emission despite quenching. The presence of the 630.0 and 636.4 nm [O i] emissions at altitudes near Io’s surface in the ISS images (see Section 3.4) indicates that electrons are penetrating into Io’s ionosphere and therefore must be losing energy as they precipitate.

Laboratory experiments show that SO₂ quenches O(¹D) with a rate coefficient of (2.17 ± 0.19) × 10⁻¹⁶ m³ molecule⁻¹ s⁻¹ (Z. Zhao et al. 2010). The O(¹D₂ ⟶  ³P₂) 630.0 nm transition has a lifetime of 178 s (W. L. Wiese et al. 1996), so the critical density at which SO₂ quenches this emission (the density of SO₂ for which the collisional de-excitation rate equals the auroral photon emission rate) is 2.59 × 10¹³ molecules m⁻³. The O(¹D₂ ⟶  ³P₁) 636.4 nm transition has a much longer lifetime of 549 s (W. L. Wiese et al. 1996), so the critical density at which SO₂ quenches this emission is 8.39 × 10¹² molecules m⁻³. For the atmospheric columns in the V. Dols et al. (2024) model, this quenching rate would limit the modeled atomic O column to 6.059 × 10¹⁷ m⁻² for the emission of 630.0 nm photons (83.5% of the total column, corresponding to a minimum altitude for emission of approximately 200 km) and 5.761 × 10¹⁷ m⁻² for the emission of 636.4 nm photons (79.4% of the total column, corresponding to a minimum altitude for emission of approximately 250 km).

No laboratory measurements have been published of rate coefficient for the quenching of O(¹S) by SO₂. However, measurements have been published for O₂ quenching of both O(¹D) (G. E. Streit et al. 1976) and O(¹S) (R. Atkinson & K. H. Welge 1972; T. G. Slanger et al. 1972). At a temperature of 115 K, O₂ quenches O(¹D) with a rate coefficient of 5.2 × 10⁻¹⁷ m³ molecule⁻¹ s⁻¹ and O(¹S) with a rate coefficient of 2.9 × 10⁻²¹ m³ molecule⁻¹ s⁻¹. These rates yield critical O₂ densities of 1.1 × 10¹⁴ molecules m⁻³ for collisional de-excitation of O(¹D) preventing the 630.0 nm transition, 3.5 × 10¹³ molecules m⁻³ for collisional de-excitation of O(¹D) preventing the 636.4 nm transition, and 4.3 × 10²⁰ molecules m⁻³ for collisional de-excitation of O(¹S) preventing the 557.7 nm transition. The O(¹S) quenching rate is different by more than four orders of magnitude compared to the O(¹D) quenching rate; assuming that a similar difference in magnitude exists for SO₂, there should be effectively no collisional de-excitation of O(¹S) along the entire O column.

The sulfur equivalent to the 630.0/557.7 nm [O i] ratio is the 1082.0/772.5 nm [S i] ratio. I. de Pater et al. (2025) calculated a disk-integrated 1082.0 nm [S i] brightness of 4.2 kR in eclipse in 2023 August. When evaluated in conjunction with the 772.5 nm [S i] brightness we observed with HIRES in 2023 August, the ratio is consistent with emission from an atomic S column impacted by electrons with energies between 4 and 5 eV. The near-surface emissions in the ISS images (see Figure 5) suggest little if any quenching of the emission, though the corresponding increase in electron density makes firm conclusions difficult, and altitude estimates for the exobase vary substantially (M. E. Summers & D. F. Strobel 1996). See I. de Pater et al. (2025) for an in depth discussion of the near-infrared [S i] emissions.

The electron number density in the Jovian magnetosphere forms a Gaussian-like vertical profile of the form

where n₀ is the number density at the centrifugal equator, H is the scale height (equal to 0.77R_J at Io’s average orbital distance; F. Bagenal & P. A. Delamere 2011), r is Io’s instantaneous orbital distance from Jupiter, and d is Io’s distance from the centrifugal equator (T. W. Hill & F. C. Michel 1976). Figure 10 shows a fit of Equation (4) to the disk-integrated 630.0 nm [O i] brightnesses, assuming that the brightness is directly proportional to the local electron number density n. The six nights of observations captured Io at a variety of distances from the centrifugal equator, but some large gaps remain unobserved. Though the data between −0.4R_J and −0.75R_J show some correlation between auroral brightness and electron density, the complete data across all distances are characterized effectively as well by a constant (the horizontal line fit in Figure 10). While this could be interpreted as suggesting that there may be no intrinsic connection between the density of the upstream electrons in the torus and the brightness of the auroral emission, it could also point to variability in the density of plasma torus electrons both longitudinally within the torus with System iii longitude and in terms of bulk torus density on long-term secular timescales. The latter interpretation is consistent with the results of A. J. Steffl et al. (2006), D. Coffin et al. (2020), and R. J. Oliversen et al. (2001), the latter of which saw variability in Io’s sunlit 630.0 nm [O i] emission by factors of 2 to 3 as a function of System iii longitude (see their Figure 4). Regardless, the HIRES data do not show a strong correlation between the upstream electron plasma density and auroral brightness.

Zoom In Zoom Out Reset image size

The connection between the ambient upstream plasma density and auroral brightness has been evaluated before at Io, Europa, and Ganymede. L. Roth et al. (2014) fit UV observations of Io’s 130.4 O i, 135.6 nm O i], and 147.9 nm S i emissions and found electron scale heights between 0.8 R_J and 1.0 R_J, though their observations exhibited considerable variance. At Europa, L. Roth et al. (2016) found some evidence of an exponential decrease in the 130.4 nm O i and 135.6 nm O i] brightnesses with distance from the centrifugal equator (see their Figures 5(c) and (d)), but their data also showed a large variance. K. de Kleer et al. (2023) observed a weak correlation between Europa’s disk-integrated 630.0 nm aurora brightness and the plasma scale height at its orbit.

Others have shown a connection between the column densities of electrons in the flux tubes above and below a satellite at different positions within the plasma torus. L. Roth et al. (2016) showed a hemispheric asymmetry in the spatial distribution of the brightness that correlated with Europa’s position above or below the plasma sheet centrifugal equator (see their Figure 10), suggesting either excitation from bounce motion along the flux tube or a strong gradient in plasma density within the magnetosphere between Europa’s northern and southern hemispheres. Z. Milby et al. (2024) showed that the north–south hemispheric brightness ratio on Ganymede was strongly correlated with its position within the plasma sheet, and specifically with the column density of the flux tube above/below each hemisphere using the F. Bagenal & P. A. Delamere (2011) plasma scale height at its orbital distance from Jupiter. Unfortunately, the HIRES Io data are likely dominated by emission from the equatorial glow and do not have sufficient spatial resolution to spatially resolve Io’s north–south hemispheric limb emission, so we cannot say whether an asymmetry is present in the 630.0 nm HIRES data.

K. D. Retherford et al. (2003) attributed an observed north–south asymmetry in Io’s 135.6 nm O i] limb glow to the different electron column densities above and below each hemisphere, which vary with Io’s position relative to the centrifugal equator. Auroral excitation would therefore come from electron bounce motion along the flux tubes above and below Io rather than impact of upstream torus electrons due to Jupiter’s rotating magnetosphere. This would result in a change in the effective electron column density between the northern and southern hemispheres (and therefore a hemispheric brightness asymmetry), but it should not change the total disk-integrated brightness.

Consequently, we cannot constrain the plasma scale height with the limited sampling of System iii magnetic latitudes available through our current set of observations. Additional observations would help to characterize both the random day-to-day variability in the brightnesses and the systematic variability with magnetic longitude.

Auroral photon emission rates are proportional to both the number density of the exciting electrons and the atmospheric column density, and consequently those two quantities are degenerate. Column densities for the major components of Io’s atmosphere are not well constrained and likely variable on both short and long timescales (see Table 19.1 in M. A. McGrath et al. 2004), limiting the ability for our auroral emission model to accurately determine the required electron density necessary to produce the brightnesses we observe. Further, our model assumes only direct excitation of neutral atoms and dissociative excitation of neutral molecules and does not currently incorporate other processes that can lead to auroral emission like dissociative recombination of molecular ions.

However, the ratio between certain ion emission lines can be used to derive electron density without the need to constrain the atmospheric column if those emission lines share the same ground level and have upper levels very close in energy. Two such lines frequently used to study Io’s plasma torus are the forbidden 671.6 nm S⁺(²D) and 673.1 nm S⁺(²D) emissions (e.g., R. A. Brown & D. E. Shemansky 1982; M. Küppers & K. Jockers 1997; M. Küppers & N. M. Schneider 2000; E. G. Nerney et al. 2024). Figure 11 shows how the ratio between these two emissions changes with electron densities between 1 and 10⁶ cm⁻³. At low electron densities (below around 100 cm⁻³) the relative populations of the upper states are determined by their statistical weights 2J + 1, so the emission-line ratio limits to

At larger electron densities, electron collisions begin to de-excite the ions before they have a change to radiate. Because the 671.6 nm state has a longer lifetime than the 673.1 nm state, collisions preferentially de-excite those ions, and the emission-line ratio begins to decrease. Above an electron density of approximately 10⁴ cm⁻³, collisional de-excitation dominates for both states, and emission therefore comes only from the few atoms that survive long enough to radiate. In this case, the emission-line ratio is modulated by the relative lifetimes τ of the states, equal to the inverse of the Einstein A coefficients for those emissions (τ = 1/A):

For the 671.6 nm [S ii] transition, A_{671.6 nm} is the sum of the Einstein A coefficient for electric quadrupole emission 1.88 × 10⁻⁴ s⁻¹ and magnetic dipole emission 1.39 × 10⁻⁵ s⁻¹ (L. I. Podobedova et al. 2009). Similarly, for the 673.1 nm [S ii] transition, A_{673.1 nm} is the sum of the Einstein A coefficient for electric quadrupole emission 1.21 × 10⁻⁴ s⁻¹ and magnetic dipole emission 5.63 × 10⁻⁵ s⁻¹ (L. I. Podobedova et al. 2009). Dashed gray lines in Figure 11 show these limits; the ratio (black line) does not exactly reach these limits, due to additional processes modeled by CHIANTI that produce excited S⁺ like recombination of S²⁺.

Zoom In Zoom Out Reset image size

The region where the emission-line ratio is sensitive to changing electron densities (between 100 and 10⁴ cm⁻³) includes the range of electron densities expected in the upstream plasma torus (1200 to 3800 cm⁻³; F. Bagenal & V. Dols 2020). However, the observed ratios (see Table 2 and Figure 11) are all below the high-density limit at any electron energy. While Io’s Alfvén wings have much higher electron number densities (between 1.53 × 10⁴ cm^–3 and 3.1 × 10⁴ cm⁻³) compared to the surrounding plasma torus (D. R. Buccino et al. 2025), the observed ratio from our data is still well below the predicted ratio even for their higher electron number density estimates.

These results indicate that there must be either collisional quenching or additional sources of excited S⁺ beyond electron impact on an existing ion column like charge exchange or dissociative recombination of SO⁺ and/or SO. We also looked at the modeled ratios of the other S⁺ emissions at 406.9 and 407.6 nm and the O⁺ emissions at 732.0 and 733.0 nm, but neither exhibited notable variability across the range of electron energies and densities plausible for Io’s ionosphere and extended atmosphere; therefore, they were not useful diagnostics of either electron energy or density.