A Stable H₂O Atmosphere on Europa’s Trailing Hemisphere From HST Images

Abstract

Previous studies of the global intensities of the oxygen emissions at 1,356 Å and 1,304 Å revealed molecular oxygen (O₂) in Europa's atmosphere. Here we investigate the relative changes of the two oxygen emissions when Europa emerges from eclipse as well as the radial profiles of the relative emissions across the sunlit disk in Hubble Space Telescope observations taken in 1999, 2012, 2014, and 2015 while the moon was at various orbital positions. The eclipse observation constrains the atomic oxygen (O) column density to or lower. We then find that the OI1356-Å/OI1304-Å ratio systematically decreases towards the disk center on the trailing hemisphere. The observed emission ratio pattern and the persistence of it from 1999 to 2015 imply a stable H₂O abundance in the central sunlit trailing hemisphere with an H₂O/O₂ ratio of 12–22. On the leading hemisphere, the emissions are consistent with a pure O₂ atmosphere everywhere across the moon disk.

Plain Language Summary

Observations by the Hubble Space Telescope in far-ultraviolet light of Jupiter's icy moons were used in the past to detect their oxygen atmospheres. Results of a new analysis of images and spectra of the moon Ganymede have recently shown that the same observations also contain information that water vapor is abundant in the atmosphere in addition to oxygen. We use the same analysis here for Europa and find a water vapor atmosphere as well, but only above the orbital trailing hemisphere of the moon.

1 Introduction

Jupiter's moon Europa possesses a tenuous atmosphere that is thought to be constantly replenished by erosion of its water ice surface (Johnson et al., 2009; McGrath et al., 2009). The main species in the bound atmosphere are expected to be molecular oxygen (O₂), H₂, and H₂O (Shematovich et al., 2005; Smyth & Marconi, 2006). The first evidence for this atmosphere was provided by a far-ultraviolet (FUV) spectrum of Europa taken by the Hubble Space Telescope, which revealed emissions at 1,304 Å and 1,356 Å. These emissions were related to atomic oxygen multiplets around these wavelengths, but the persistently brighter disk-averaged OI1356 Å emission intensity (Hall et al., 1995, 1998) is in agreement only with electron impact dissociative excitation on molecular oxygen as source. The ratio of the intensities of the FUV oxygen multiplets at 1,356 Å and 1,304 Å, I(OI1365 Å)/I(OI1304 Å), has become a standard diagnostic to probe for atmospheric composition at Jupiter's moons (Cunningham et al., 2015; Feldman et al., 2000; Hansen et al., 2005; Molyneux et al., 2018) or comets (Feldman et al., 2015; Galand et al., 2020).

Roth et al. (2016) analyzed the absolute and relative intensities of the OI1304 Å and OI1356 Å emissions in a data set of spectral images taken by HST's Space Telescope Imaging Spectrograph (STIS). They found that in images of the orbital leading hemisphere, the near-surface OI1356 Å/OI1304 Å ratio of is consistent with a pure O₂ bound atmosphere. On the trailing hemisphere, this near-surface ratio was found to be consistently lower with an average of (see Table 2 of Roth et al., 2016). The different emission ratios were interpreted to be due to differing abundance of atomic oxygen. The higher O abundance on the trailing side (O/O₂ ∼ 0.05) could be explained by preferential production of atomic oxygen from increased electron impact dissociation of O₂, as the trailing side is also the plasma upstream side and more exposed to the plasma impacts (Cassidy et al., 2013; Pospieszalska & Johnson, 1989).

In regions further above the limb of Europa's disk in the images, was found to decrease with distance to Europa (or altitude) compared to the near-surface region. This was explained with an increasing O/O₂ ratio with altitude (Roth et al., 2016), in accordance with Cassini observations (Hansen et al., 2005) and simulations (Shematovich et al., 2005; Smyth & Marconi, 2006).

The relative FUV oxygen emissions at neighboring moon Ganymede show similar behavior with decreasing above the limb and the same hemispheric dichotomy, that is, a lower ratio () above the trailing hemisphere (Molyneux et al., 2018). Using an HST observation of Ganymede directly before and in eclipse of Jupiter, Roth et al. (2021) recently showed that resonant scattering by atomic oxygen is negligible, setting an upper limit on the global abundance of O in the atmosphere. This upper limit effectively rules out atomic O as explanation for a lower oxygen emission ratio in bright emission regions on Ganymede's disk. Investigating the radial profiles of the ratio in STIS images Roth et al. (2021), also found that systematically decreases from the region around the limb () towards the center of the observed disk on both hemispheres. The found profile is shown to be in agreement with an H₂O-dominated atmosphere around the sub-solar point and a O₂-dominated atmosphere away from the sub-solar point. The concentration of H₂O around the sub-solar point and the hemispheric dichotomy with an 6-fold higher estimated H₂O abundance on the trailing hemisphere is in agreement with the difference in ice sublimation yield, given the difference in albedo and thus surface temperature at Ganymede (Leblanc et al., 2017).

Observational evidence for the presence of gaseous H₂O at Europa was provided indirectly through localized H1216 Å (H Lyman-) and OI1304 Å FUV emission patches near the south pole (Roth et al., 2014b) and through infrared emissions from H₂O (Paganini et al., 2019). In both observations, the signal related to water vapor was only marginally significant and interpreted to originate from an active plume source. The interpretation was based on the localized nature of the atomic emissions, the comparably high derived abundances, and the low detection rate suggesting an intermittent nature (Paganini et al., 2019; Roth et al., 2014b). Paganini et al. (2019) also mention that their detection was on the leading hemisphere, while charged particle sputtering and sublimation as potential alternative sources for H₂O are expected to be higher on the darker (and thus warmer) trailing (= plasma upstream) hemisphere where no H₂O emissions were found in the same study.

Here, we use the same approach as recently used by Roth et al. (2021) for Ganymede and study in detail the relative brightness of the two oxygen emissions in order to constrain the H₂O and atomic oxygen (O) abundances relative to O₂ in Europa's global atmosphere. First, we compare consecutive exposures taken in and out of eclipse to constrain the O abundance through its possible resonant scattering contribution. Thereafter, we investigate the emission ratio in various observations of Europa and its variation across moon's disk.

2 HST/STIS Observations

We have selected nine sets of images (or “HST visits”) from the HST/STIS datasets presented in Roth et al. (2016) considering primarily the signal-to-noise ratio of the weaker and noisier OI1304 Å emissions. For each visit, we have used only the periods in the exposures when the brightness from the Earth's geocorona along the slit is near or below the level of the statistical noise.

For the eclipse test, we use the visit from 23/24 March 2015 (Visit 17 in Table 1 of Roth et al., 2016), which contains two exposures with low-geocorona exposure time of 1,615 s in eclipse and 1,820 s after egress. This is the only visit that includes an eclipse observations with reasonable signal-to-noise ratio and a consecutive exposure out of eclipse as reference.

From the other visits with Europa at various orbital positions, we selected those with total low-geocorona exposure time of at least 140 min. All nine visits analyzed here are summarized in Table 1, where we use the original consecutive visit numbers as in Table 1 of Roth et al. (2016).

Table 1. Parameters of the HST/ Space Telescope Imaging Spectrograph (STIS) Observations, First Published in Roth et al. (2016)

Visit	Observed hemisphere	Date	Start time (UTC)	End time (UCT)	No. of exp.	Used (total) exp.time (min)	Europa diameter (arcsec)	Spatial resolution (km/pixel)	Europa CML (°)	System-III longitude (°)
17a	Eclipse	2015-03-23	23:14	23:41	1	26.9 (26.9)	0.92	83.0	10–12	146–161
17b	Egress	2015-03-24	00:27	01:12	1	30.3 (45.6)	0.92	83.0	16–18	194–209
1	Trailing	1,999-10-05	08:39	15:32	5	142.2 (156.0)	1.07	71.5	245–274	300–161
2	Trail./Anti-J.	2012-11-08	20:41	03:33	5	155.0 (183.4)	1.04	73.9	209–238	24–244
3	Leading	2012-12-30	18:49	01:39	5	140.1 (164.1)	1.02	74.9	79–108	360–220
4	Lead./Anti-J.	2014-01-22	14:02	20:53	5	143.4 (183.4)	1.01	76.0	117–146	201–61
13	Trailing	2015-02-22	11:00	16:17	4	154.6 (171.2)	0.98	78.3	256–278	132–301
14	Leading	2015-02-24	05:58	12:51	5	195.3 (217.0)	0.98	78.4	77–107	67–288
15	Trail./Sub-J.	2015-03-09	01:03	08:01	5	193.4 (232.0)	0.96	80.2	296–325	189–52
16	Lead./Anti-J.	2015-03-21	16:59	22:21	4	145.0 (183.3)	0.93	82.4	141–163	209–21

• Note. CML refers to the Central Meridian (West) Longitude on Europa's disk, Jupiter's planetocentric longitude facing Europa is given by the System-III longitude.

For the eclipse visit, we separately analyze the exposure in eclipse and the exposure in sunlight. The two exposures are shown in Figure 4 of Roth et al. (2016). For the other eight visits, we combined all low-geocorona exposure time obtaining one superposed spectral image. Some of the STIS spectral image data analyzed here are displayed in Figure S5 of Roth et al. (2014b), Figure 3 of Roth et al. (2014a) and Figures 2 and 3 of Roth et al. (2016). We follow our standard processing pipeline for correcting the detector images for background and surface reflection (absent in the eclipse visit) signals, see Roth, Saur, Retherford, Strobel, and Feldman (2014); Roth et al. (2014a); Roth et al. (2016), and include the small processing updates from Roth et al. (2021). We furthermore replace the assumption of a uniform disk brightness with the reflectance model from Oren and Nayar (1994) and a roughness parameter of  = 0.57, which was found to provide good agreement with FUV images of Europa (Giono et al., 2020; Sparks et al., 2016).

Finally, 72 × 72 pixel images containing the two oxygen emission images centered on the spectral axis at 1,303.5 Å and 1,356.3 Å are extracted from the spectral detector images and converted to units rayleigh (R). The analysis is carried out in the native detector frame and original pixel resolution without smoothing or binning. Errors are calculated and propagated for each pixel in an image considering all statistical and systematic uncertainties of the HST pipeline (including bad pixels) and our processing steps (c.f. Roth et al., 2016). Figure 1 shows the two oxygen images (a and b) from visit 13 of the trailing hemisphere as an example with good signal-to-noise ratio. The emissions reveal the typical morphology that is determined by the interaction with the plasma environment (Roth et al., 2016).

Figure 1

(a) and (b): HST/ Space Telescope Imaging Spectrograph images of the oxygen emission at O I 1,356 Å (a) and O I 1,304 Å (b) above Europa's trailing hemisphere observed during visit 13 on 2015-2-22. The images were smoothed twice with a 3 × 3 pixel boxcar function. All analysis was carried out with the original data, that is, no binning or smoothing is applied. The vectors marked “N” show the direction to Jupiter North. The slightly dispersed locations of Europa's disk at the individual oxygen multiplet lines are shown by dotted circles. Diamonds indicate the disk center and the asterisks the sub-solar point. (c) Radial intensity profile for both oxygen emissions with angular bins of 0.25 derived from the images. (d) Radial profile of the oxygen emission ratio, , derived from the observations in (c). The numbers indicate the derived ratios with uncertainties averaged over in the central, limb and coronal regions, respectively (dotted vertical lines, see also sketch in Figure 2a). The dashed line shows the ratio for a pure O₂ atmosphere.

The focus of this study is on the oxygen emission ratio. To achieve reasonable signal-to-noise levels (c.f. the noisy ratio images in Figures 2–4 of Roth et al., 2016), we calculate average ratios in larger regions and binned profiles across the images, by dividing the averaged intensity of the OI1356  Å by the averaged intensity of the OI1304  Å in all pixels within the respective bins or areas.

In Figures 2b–2d, we show the oxygen emission ratio averaged over three radial regions, as sketched in panel (a), for all analyzed image pairs. Panel (b) shows the ratio in the limb region, which includes all pixels centered at radial distances between 0.75 and 1.25 from Europa's disk center. The ratios derived in the center region (<0.5 ) is shown in panel (c), and the ratio in a coronal region (1.25 ) is shown in panel (d).

Figure 2

(a) Sketch of the three analyzed regions on and around Europa's disk, which is indicated by the dashed circle. If the center of a pixel falls within the defined radial ranges it is counted for the respective region. (b), (c), and (d) Oxygen emission ratios, , for the three regions. The dashed lines in (b) and (c) show the mean of all visits (except the eclipse/egress exposures). In (d), the light purple shaded area shows variance range of the three images near 90°W longitude, the darker purple shaded area show the standard deviation range around the mean of all four images of the trailing side >180°W longitude. (e) Oxygen emission ratios, , on the dawn side (triangles) and dusk (squares) side (all pixels within 1.25 on either half of the circle, see sketch in inlet) for four the trailing hemisphere images. (f) and (g) Theoretical emission ratio as a function of the relative abundance of atomic oxygen, O, and water vapor, H₂O, in an O₂ atmosphere for electron impact excitation.

3 Excitation Model

We consider electron impact excitation as well as resonant scattering of the solar OI1304 Å radiation, using the same emission rates from electron impact (dissociative) excitation of O₂, H₂O and O as in Roth et al. (2016); Roth et al. (2021) based on the cross sections from Doering and Gulcicek (1989), Kanik et al. (2001), Kanik et al. (2003), and Makarov et al. (2004). The focus of this study is on the OI1356 Å/OI1304 Å ratio, which is independent of the density of the exciting electrons and relatively insensitive to the electron temperature (if resonant scattering is neglected). The OI1356 Å/OI1304 Å ratio is yet sensitive to the atmospheric abundance of H₂O and O relative to O₂.

Following the previous studies, we assumed a thermal electron population with 20 eV plus a 5% higher temperature fraction at 250 eV and calculate an effective emission rate for this population. For the emission rate for OI1356 Å from dissociative excitation H₂O, we scaled the rate for OI1304 Å by a factor of 0.2 based on the laboratory measurements of Makarov et al. (2004) for 100 eV electrons. The resulting oxygen emission ratio from electron-impact excitation as a function of the O/O₂ and H₂O/O₂ abundance ratios in the atmosphere are shown in Figures 2f and 2g.

For the estimates of absolute neutral abundances, we assume an electron density of 160 following de Kleer and Brown (2018). Although this is 4× higher than the density of 40 assumed in some previous studies of Europa's far-UV emissions (Hall et al., 1995, 1998; Roth et al., 2016), it is in better agreement with the latest studies of the Europa plasma environment (Bagenal & Dols, 2020; Bagenal et al., 2015).

Resonant scattering of the solar OI1304 Å is calculated in the optically thin limit with a resonant scattering g factor of 1/s (Cunningham et al., 2015). Hence, 1 R of OI1304 Å emission relates to an O column density of .

4 Results and Interpretation

4.1 Eclipse Test

The comparison of the eclipse exposure to the after-eclipse exposure suggests that resonant scattering by atomic oxygen is negligible. The image-averaged OI1304 Å and OI1356 Å intensities in eclipse are 21.9 R and 40.2(±) R, respectively. The intensities in the following exposure taken after egress were 15.8(±) R and 36.3(±) R, corresponding to a decrease by 28(±)% for OI1304 Å and 10(±)% for OI1356 Å. Thus, both emissions appear to be weaker after eclipse, suggesting a lower auroral excitation. This is consistent with the fact that Europa is moving away from the plasma sheet during this eclipse/egress visit (Table 1). The similar decrease at OI1304 Å compared to OI1356 Å indicates that resonant scattering of solar light, which would add to the OI1304 Å emission (only) and thus lead to an increase (or weaker decrease), is negligible.

Based on the change in OI1356 Å, we assume that 10(±)% of the change in OI1304 Å is due to auroral excitation. After this correction for auroral changes, the residual OI1304 Å intensity change and the propagated related uncertainty are R, thus consistent with no change apart from auroral effects within 1.1. Setting an upper limit for the change of +3 R (corresponding to ∼2σ of the measured change of R) for the resonant scattering contribution at OI1304 Å over Europa's disk, we get an upper limit for the vertical O column density averaged of the same area of 6 .

Considering various aspects of the variable plasma environment and the atmospheric distribution, the minimum O₂ column density in Europa's atmosphere is estimated to be (de Kleer & Brown, 2018; Hall et al., 1998; Roth et al., 2016). The upper limit for the O abundance derived above and this minimum O₂ abundance imply a maximum O/O₂ ratio in the atmosphere of 0.03 (red dash-dotted line in Figure 2f).

We note that the change observed in the coronal region (1.25 ) is different. While the OI1356 Å intensity again decreases, from 14.8(±) R (eclipse) to 6.4(±) R (sunlight), the OI1304 Å emission increases from 4.7(±) R (eclipse) to 12.7(±) R. This can also be seen in Figure 2d: the oxygen emission ratio in the corona drops from in the eclipse exposure (off the vertical range) to after egress (near sub-observer longitude °). The increase in OI1304 Å in the corona region is consistent with resonant scattering contributing to the overall faint signal, possibly originating from Europa's extended atomic corona (Hansen et al., 2005; Smith et al., 2019). However, the error ranges in these individual exposure are large and small systematic uncertainties in the background subtraction (not reflected in the error bar) might lead to additional uncertainties for the faint coronal emission. We also note that the OI1304 Å increase of 8.0(±) R in the corona region (neglecting possible changes due to changing auroral excitation) is also consistent (within the propagated uncertainty) with our 3 R upper limit for resonant scattering in the bound atmosphere derived above.

4.2 Oxygen Emission Ratio Profiles in the Images

Next, we investigate the oxygen emission ratios in the three regions across the disk: corona, limb, and central region. The comparison of the derived ratios for all images (Figures 2b–2d) reveals some systematic differences, both between the regions as well as between the observed hemispheres.

In the limb region, the oxygen emission ratio is similar for all visits (i.e., independent of observing geometry) and consistent with their mean value of (dotted line in Figure 2b) within the individual ∼1- uncertainties. There appears to be a marginal trend with somewhat lower ratios on the trailing and somewhat higher ratios on the leading side. These ratios are consistent with a predominantly O₂ atmosphere with a small O mixing ratio near the upper limit derived for the bound atmosphere of O/O₂ = 0.03 (Figure 2f).

In the coronal region, the ratios are systematically lower and again consistent within 1- with their mean of (dotted line in Figure 2d) for almost all analyzed visits. These results agree well with our previous similar analysis of the coronal emission ratio in the same data, see Table 2 and Figure 9b in Roth et al. (2016). The low ratio can be explained by a higher O/O₂ abundance ratio in the corona than in the bound atmosphere (Roth et al., 2016). The upper limit on O/O₂ was derived within 1.25 , that is, for the bright emissions and the denser near-surface atmosphere, but it does not apply for the coronal region. The intensities of both oxygen emissions in the corona are only around 10 R, consistent with column densities of a few ∼10¹³ for O₂ and a few ∼10¹² for O, and a relative abundance of O/O₂ ∼ 0.2.

In the center region, there appears to be a systematic difference between the images and in particular between images of the trailing hemisphere and some of the images of the leading hemisphere (Figure 2c).

First, we look at the leading side images with sub-observer longitude of 90°, that is, where the leading meridian longitude (°) is ≤45°offset from the central meridian longitude in the image. In the three images in this category, the center emission ratio is similar to (slightly higher than) the limb emission ratio with a mean of (dotted line in the °−° range, Figure 2c). This high ratio is consistent with a pure O₂ atmosphere.

For the four trailing side images, the emission ratios in the center are consistently lower and well below 2 with a mean value of (Figure 2c), similar to the corona region. In contrary to the corona, however, this can not be explained by a higher O/O₂ ratio. In the trailing center region, the intensities of both oxygen emissions are ∼50 R or higher (see example intensity profiles in Figure 1c) and the upper limit of O/O₂ derived above for the bound atmosphere applies. Since an O/O₂ ratio of ∼0.2 would be required to explain the the low (see Figure 2f), that is, ∼10 times above the upper limit, atomic oxygen can be ruled out even when considering the possibility of O being concentrated to the central region.

The radial profiles of the oxygen emission ratio in smaller 0.25- bins in the four trailing hemisphere images reveled a general and consistent behavior, shown for an example visit from 2015 in Figure 1d. The emission ratio peaks near the limb, that is, near radius . Away from the limb, the emission ratio systematically decreases both towards higher altitudes above the limb and towards the disk center. This is the same profile as observed for Ganymede (Roth et al., 2021).

We then consider electron-impact excitation of H₂O as the process cause the low oxygen emission ratio, , in the central disk region. The derived value of then is consistent with H₂O/O₂ ratios between 12 and 22 (see Figure 2g) and thus an H₂O-dominated atmosphere in the trailing center atmosphere.

Next we look at the visit, when Europa's anti-Jovian hemisphere was mostly observed (near sub-observer longitude °). The derived ratio of is between the leading and trailing side values. We refrain from interpreting this value further, as there is an additional uncertainty due to the observing geometry. In this observation of the anti-Jovian side of Europa in its tidally locked orbit, the relatively small angular separation from Jupiter allows for possible nearby emissions from the Io torus, which introduce an extra uncertainty in the background subtraction. Because such torus background would be blocked by Europa, the determination of the background brightness away from the moon might lead to an over-subtraction of this background signal on the disk. This affects both the solar reflection (via the albedo) modeling and the emissions directly.

Finally, in the two individual exposures in eclipse and after egress (near sub-observer longitude °) the low ratios of (eclipse) and (after egress) are very similar to the trailing side ratio. Hence, these ratios are similarly consistent with a substantial H₂O abundance in the central region. This would imply that the H₂O atmosphere in the central disk region is present even on the sub-Jovian hemisphere and throughout the eclipse passage. However, the uncertainties prevent firm conclusions.

We searched for further trends in the emision ratio across the images. Directional profiles along the east-west and south-north directions generally resembled the radial profiles with higher near the limb. In some images, there appear to be trends, but these were not consistently seen in all images and seem to be transient and possibly even due to statistical fluctuations.

Figure 2e shows a systematic trend that is possibly present on the trailing hemisphere. Here, the emission ratio averaged over the dawn half disk and the average ratio on the dusk half disk (all pixels within 1.25 on either side) are compared. We found that is lower on the dawn half disk than on the dusk half disk in all four images of the trailing side (while no systematic behavior was found in the leading side images). Interpretations of this possible trend will be further discussed in the final section.

4.3 Estimates of Trailing Hemisphere H₂O Abundance

We estimate absolute abundances for H₂O and O₂ on the trailing hemisphere from the OI1356 Å and OI1304 Å intensities using the values measured during visit 13 (Figure 1), where the emission intensities had an average level (see, e.g., Figure 6 of Roth et al., 2016). We assume a fixed O column density at the upper level derived above. Using the model assumptions for the exciting electron population and described in Section 3 and taking into account resonant scattering by O, we then simultaneously fit the O₂ and H₂O column densities to match the observed intensities of the two oxygen emissions.

The results from the fitting (Table 2) confirm not only that in the center region the abundance of H₂O is about an order of magnitude higher than that of O₂, but also that in the limb region H₂O is not required to explain the measurements. The limb intensities are well in agreement with a dominant O₂ atmosphere with low atomic oxygen abundance with O/O₂ = 0.024. If the global abundance of O is lower, a small amount of H₂O is required in the limb region and a slightly more H₂O is needed in the center.

Table 2. Observed Intensities During Visit 13 and Fitted Column Densities and the Model Emissions for the Three Considered Species

	Center region			Limb region
		OI1356  Å	OI1304  Å		OI1356  Å	OI1304  Å
Observation STIS Visit 13		R	R		R
Species	Column density ()	OI1356  Å	OI1304  Å	Column density ()	OI1356  Å	OI1304  Å
H2O		4.6 R	20.6 R	<	<0.1 R	<0.1 R
O2		55.5 R	24.9 R		56.6 R	25.3 R
O	^a^a The O abundance is set to the derived upper limit.	0.9 R	6.5  ^b^b The OI1304 intensity from O includes a resonant scattering contribution of 3 R.	^a^a The O abundance is set to the derived upper limit.	0.9 R	6.5  ^b^b The OI1304 intensity from O includes a resonant scattering contribution of 3 R.
Total	–	61.0 R	52.0 R		57.5 R	31.8 R

• ^a The O abundance is set to the derived upper limit.
• ^b The OI1304 intensity from O includes a resonant scattering contribution of 3 R.

We emphasize again that the total abundances are sensitive to the electron properties. We omitted to state uncertainty ranges for the fitted abundances in Table 2, which would only reflect the comparably small statistical uncertainties of about 10%. The uncertainty of the electron properties is, however, considerably larger than this. When assuming a homogeneous distribution of the electrons around Europa, the neutral densities scale inversely with the assumed electron density (see Figure 12a in Roth et al., 2016). For example, in the case of a two times lower electron density, twice higher neutral abundances are required. In reality the electrons will cool in the atmosphere due to collisions and the aurora yield will be sensitive to the details of this interaction (Dols et al., 2016; Rubin et al., 2015; Saur et al., 1998).

5 Discussion and Conclusions

The slightly updated treatment of the solar reflection signal modeling described change the results only marginally compared to our previous study on the same data and we found that all numbers are consistent within uncertainties with the results of Roth et al. (2016).

The upper limit of the atomic to molecular oxygen ratio of O/O₂ = 0.03 is similar to the O mixing ratio derived for Europa from Cassini UVIS measurements of 0.02 (Hansen et al., 2005) and to the upper limit of 0.02 for Ganymede found in an HST eclipse test (Roth et al., 2021).

The oxygen emission ratio is systematically and significantly below the value for a pure O₂ atmosphere in the central trailing hemisphere. Based on the derived upper limit, resonant scattering and electron excitation of O can not cause the low oxygen emission ratio in this region. Electron impact dissociative excitation of H₂O is the most likely viable process to produce the reduced ratio. In addition, two different species are a logical explanation for the particular shape of the radial profiles with a peak ratio near the limb (close to pure O₂) and two separate minima above the limb (due to O) and in the disk center (due to H₂O). Contributions from other possible species like OH or are negligible due to the low expected abundances and low emission cross sections (McConkey et al., 2008; McGrath et al., 2004).

The abundance of H₂O relative to O₂ derived for Europa's trailing hemisphere of H₂O/O₂ = 12–22 is similar to the values found for Ganymede's trailing hemisphere (H₂O/O₂ = 12–32). On Europa's leading hemisphere, there is no indication for the presence of H₂O, as the oxygen emission ratios are even higher in the center region than near the limb.

Our estimate for the absolute H₂O abundance of can be compared to the Keck search for infrared H₂O emissions (Paganini et al., 2019) and in particular to their most sensitive measurement of the same hemisphere (February 27 in Table 2 of Paganini et al., 2019). Their slit-average column density upper limit of is 2.3 times lower than our estimated value. However, the slit covers an area of about 3,250 km × 1,500 km and hence, large areas across Europa's disk and slightly outside the disk. Because the H₂O abundance apparently rapidly decreases to negligible abundances outside the central area, that is, between radii and , the disk-average (or slit-average for the Keck data) column density is expected to be lower by a factor of about 2–4 than the density derived here for the disk center only. Hence, our values are consistent with the Keck upper limit.

A key finding of this study is the consistency in the detection of the reduced oxygen emission ratio on the trailing hemisphere disk center and the overall stability of the ratio profiles in all images with similar geometry. In particular, the oxygen emission ratios in center and limbs regions in the four trailing side visits, which were obtained in 1999, 2012, and 2015 and are all consistent within uncertainties. This means, they are diagnostic for persistent atmospheric properties, in stark contrast to the apparent transient nature of detected features that were interpreted to relate to H₂O plumes (Paganini et al., 2019; Roth et al., 2014b).

Persistent sources for H₂O can be sublimation or sputtering by charged particles of the surface ices, which both can produce a hemispheric difference between the leading and trailing day side hemispheres.

The sublimation yield is sensitive to temperature and the ice fraction of the surface material. Both measurements (Rathbun et al., 2010; Spencer et al., 1999; Trumbo et al., 2018) and modeling (Oza et al., 2019) of the surface properties suggest higher temperatures on the visibly darker (McEwen, 1986) trailing hemisphere, possibly leading to higher sublimated H₂O abundance there.

The trailing hemisphere also coincides with the plasma upstream hemisphere, where most of the thermal plasma impinges on Europa's surface according to modeling (Cassidy et al., 2013; Pospieszalska & Johnson, 1989). In addition, the sputtering yield (amount of neutrals ejected per incident charged particle flux) also increases with surface temperature (Famá et al., 2008), further favoring the trailing hemisphere independent of the distribution of the incident flux.

For both sources, however, modeled H₂O abundances are often significantly lower than our derived value on the trailing side (Plainaki et al., 2013; Shematovich et al., 2005; Smyth & Marconi, 2006). The H₂O sublimation flux at Europa's maximum surface temperatures of 130 K is about two orders of magnitude lower (Feistel & Wagner, 2007) than for Ganymede's surface temperature of 145 K (Orton et al., 1996). H₂O yield from sputtering similarly is estimated to be too low to sustain column densities larger than ∼10¹³ for the short-lived H₂O molecules, which freeze and are lost from the atmosphere upon surface contact (Plainaki et al., 2013; Shematovich et al., 2005; Smyth & Marconi, 2006).

The atmosphere modeling study of Teolis et al. (2017) considers secondary sublimation, that is, sublimation of sputtered H₂O molecules that fall back to the surface, as source in addition to sputtering. In their results the H₂O abundance exceeds the O₂ abundance on the dayside, with column densities up to (their Tables 2 and 3). Thus, according to the results of Teolis et al. (2017), the combined effects of sputtering and sublimation of fresh H₂O deposits might be the source for the detected H₂O atmosphere.

The lower on the dawn side compared to the dusk side of the trailing hemisphere (Figure 2e) can be explained by a relatively stronger emission contribution from H₂O or higher H₂O/O₂ abundance ratio in the dawn sector (or lower H₂O/O₂ ratio in dusk sector). O can be ruled out as main cause for the differing oxygen emission ratios with the same arguments as above. This difference in H₂O/O₂ abundance ratios between dawn and dusk is consistent with the surplus of O₂ on the dusk side suggested by a modeling study (Oza et al., 2019). Roth et al. (2016) showed that the absolute OI1356 Å intensities are also systematically higher on the dusk side in nearly all images. The vast majority of the OI1356 Å emissions originate from O₂ (due to the ∼160 higher emission OI1356 Å rate from O₂ compared to H₂O) and are thus insensitive to the abundance of H₂O. Therefore, a surplus of O₂ on the dusk side compared to the dawn side (and assuming the H₂O abundance is symmetric) is a consistent explanation for the dawn-dusk asymmetries both in the absolute OI1356 Å intensity found in Roth et al. (2016) and in the dawn and dusk emission ratios derived here.

Putting the main results in a nutshell, oxygen emission ratios found in HST observations suggest a persistent H₂O atmosphere above Europa's trailing hemisphere, but the source of the water vapor can not unambiguously identified.

Acknowledgments

L. Roth. acknowledges the support from the Swedish National Space Agency (SNSA) through grant number 154/17 and the Swedish Research Council (VR) through grant number 2017-04897.

Open Research

Data Availability Statement

The HST data used in this work were taken within programs HST-GO-8224, HST-GO-13040, HST-GO-13619, HST-GO-13679, and are available on the MAST archive of STScI at https://archive.stsci.edu/proposal_search.php?mission=hst&id=8224, https://archive.stsci.edu/proposal_search.php?mission=hst&id=13040, https://archive.stsci.edu/proposal_search.php?mission=hst&id=13619, and https://archive.stsci.edu/proposal_search.php?mission=hst&id=13679.