VLT/SINFONI OBSERVATIONS OF EUROPA: NEW INSIGHTS INTO THE SURFACE COMPOSITION

A global compositional mapping campaign of Europa was performed between 2011 October and 2012 January during five nights with the ground-based integral field spectrograph SINFONI. The broad H+K spectral range, which provides sufficient spectral resolution (∼1 nm corresponding to a spectral resolving power in excess of R = λ/Δλ = 1500), was chosen so as to detect any narrow signature in its 1.452–2.447 μm wavelength range. The observational mode of SINFONI is rather complex. An "image slicer" segments the 0.8 by 0.8 arcsec field of view into 32 image slitlets of 12.5 milliarcsec (mas) size, which are redirected toward the spectrograph grating to be re-imaged onto the 2048 × 2048 pixel SINFONI detector. Each of the 32 slitlets is imaged onto 64 pixels of the detector, thus yielding 64 × 32 spectra of the imaged region on the sky. The original field of view is reconstructed into a 64 × 64 binned pixel image-cube, with the third dimension corresponding to wavelength. Actual spatial sampling is 12.5 × 25 mas.

Observations were carried out near opposition to maximize the angular resolution, resulting in angular diameters of Europa varying from 0.897 to 1.083 arcsec, i.e., larger than SINFONI's 08 field of view. To cover the entire disk of Europa, a mosaic of five overlapping frames was necessary to cover the entire surface of the satellite with a ∼35 × 70 km spatial bin (Figure 1). The five frames of the mosaic were first acquired and then followed by the acquisition of three sky background frames. Immediately following each observation, the solar-type star HD 18681 was observed at similar airmass (see Table 1 for more details).

Zoom In Zoom Out Reset image size

2.2.1. ESO Pipeline

2.2.2. Telluric Correction

The next step of the data-reduction process consists of correcting each pixel for atmospheric absorptions and calibrating the reflectance spectra. The following formula is applied to each cube:

$$\begin{eqnarray}{\boldsymbol{R}}({\boldsymbol{\lambda }}) & = & \displaystyle \frac{\mathrm{Europa}\;\mathrm{flux}\;({\boldsymbol{\lambda }})}{\mathrm{Star}\;\mathrm{flux}\;({\boldsymbol{\lambda }})}\times {\boldsymbol{C}}\quad \quad \mathrm{with}\\ {\boldsymbol{C}} & = & \displaystyle \frac{{{\boldsymbol{T}}}_{\mathrm{Telluric}}}{{{\boldsymbol{T}}}_{\mathrm{Europa}}}\times \displaystyle \frac{1}{{\rm{\Omega }}\mathrm{pixel}\times {10}^{\tfrac{{{\boldsymbol{M}}}_{\mathrm{Telluric}}-{{\boldsymbol{M}}}_{\mathrm{Sun}}}{2.5}}}.\end{eqnarray} \tag{ 1 }$$

The coefficient ${\boldsymbol{C}}$ depends on (1) the integration times of Europa and the telluric star $({{\boldsymbol{T}}}_{\mathrm{Europa}}$ and ${{\boldsymbol{T}}}_{\mathrm{Telluric}}$, respectively), (2) the angular surface Ω_pixel of a pixel (3.65 × 10⁻¹⁵ sr), (3) the apparent magnitude of the Sun seen from Jupiter $({{\boldsymbol{M}}}_{\mathrm{Sun}}\quad =\quad -23.16)$, and (4) the apparent magnitude of the telluric $({{\boldsymbol{M}}}_{\mathrm{Telluric}}\quad =\quad 9.23)$. To reduce the noise and obtain more accurate reflectance values, the telluric flux is evaluated over a circle of 11 pixels in radius around the center peak of the point-spread function (PSF). The radius value is visually estimated by studying the aspect of the PSF at 2.4 μm and includes the primary and secondary diffraction lobes. Note that the PSF is better defined at large wavelengths than at small ones. This causes a slight undervaluation of the star flux at small wavelengths, which introduces an overestimate of the reflectance and a slope effect in the correction. This slope effect is minimal for a Strehl value of ∼40% but needs to be corrected for lower values (see Section 3.1).

2.2.3. Mosaic Reconstruction

2.2.4. Photometric Corrections

Photometric correction is then applied to obtain the geometrically corrected reflectance. We approximate Europa's surface as a Lambertian surface and correct the data cube of each run by the following factor:

$$\begin{eqnarray}&&\mathrm{cos}(i)=\sqrt{1-{\left(\displaystyle \frac{\mathrm{SSP}}{{\boldsymbol{r}}}\right)}^{2}},\end{eqnarray} \tag{ 2 }$$

where ${\boldsymbol{r}}$ and SSP are, respectively, the radius of Europa (in pixels) at the time of observation and the distance (in pixels) between the pixel and the sub-solar point. The result is visually convincing until angle values typically ranging between 50° and 60° (Figure 2). For larger angles, the reflectance is larger than 1, which is likely because the Lambertian surface hypothesis is no longer valid. We therefore remove the pixels for incidence >50°, 55°, or 60° depending on the night. Finally, a phase angle effect correction was performed using the phase curve derived in the visible wavelength range from ground-based observations without taking into account the opposition effect (Domingue & Verbiscer 1997). This leads to a maximal correction factor of 1.12.

Zoom In Zoom Out Reset image size

2.2.5. 3D Reflectance Cube

The reflectance map at 2 μm is shown in Figure 3 and compared to the Europa USGS map (USGS Map-a-planet: http://astrogeology.usgs.gov/tools/map). The main geological provinces, as mapped by Doggett et al. (2007, 2009), are also outlined with dark lines. Correlations of the 2 μm map with some of these provinces are clearly visible, as described below. The smallest reflectance values at 2 μm (≤0.08) are found in the Pwyll crater (271°W, −25°N), in the Falga Regio (210°W, 30°N), in an unnamed (and bright in the visible) area in the north of the leading hemisphere centered at (120°W, 30°N), as well as in two undefined areas south of the Annwn Regio (320°W, 20°N) and the Dyfed Regio (260°W, 10°N). The Dyfed Regio and Tara Regio borders are barely visible with medium reflectance values (0.12–0.15). Finally, the highest reflectance values (≥0.16) are observed in two different locations: near the antijovian point (180°W, 0°N) and in a small dark chaos (355°W, 0°N) extending westward of Annwn Regio. Note, however, that the antijovian bright spot can be partly attributed to the accumulation of small approximations (mainly resulting from the telluric division and photometric corrections) during the data reduction process. We will see that most of the albedo variations reported here are correlated to variations of material abundances (Section 5).

Zoom In Zoom Out Reset image size

As a preliminary investigation of Europa's surface composition, we map the water-ice signatures. Within the SINFONI wavelength range, the absorption at 2 μm is the most appropriate for such an analysis. For each pixel, we divide the median reflectance value over the 1.995–2.005 μm range by the continuum value at 2 μm extrapolated from the slope between 1.745 and 2.235 μm (median values over 20 wavelengths centered at each value; Figure 4). Performing spectral ratios has the advantage of eliminating most of the artifacts observed in the 2 μm reflectance map, although some are still present because of the complex geometry of SINFONI, as previously discussed (see Section 2.2.1). The very typical "bullseye" feature reported in previous global surface composition studies (McEwen 1986; Carlson et al. 2005; Grundy et al. 2007; Brown & Hand 2013) is also visible here on the orbital trailing hemisphere. This feature is well correlated to the non-icy areas and to the dark trailing hemisphere in the visible wavelength range. Other large-scale variations associated with geomorphological units identified by Galileo are also sometimes observed. In spite of residual instrumental artefacts recognizable by localized straight lines and spurious pixels, some features (at SINFONI spatial resolution) such as the Pwyll crater and the Tara Regio are barely resolved as exhibiting values of the spectral parameter (0.38 and 0.40 on their center) significantly different from their surrounding terrains. Large areas with low visible albedo, including the Annwn Regio and Dyfed Regio, are also distinguishable by their low 2 μm signatures.

Zoom In Zoom Out Reset image size

We produced a similar global map for the absorption at 1.65 μm, which is characteristic of crystalline water ice (Figure 4). This feature is very well defined in the data set due to the high spectral resolution of the SINFONI instrument. A specific spectral criteria index is derived by dividing the median reflectance in the 1.647–1.653 μm range by the corresponding values of the continuum. A straight-line segment from the absorption's left wing (1.61–1.62 μm) and right wing (1.69–1.70 μm) was used to approximate the continuum line. The spectral criteria is lower than 1 for all pixels, indicating that the 1.65 μm feature is present in all of the SINFONI spectra. The spatial variations are well correlated at first order with the distribution of the 2 μm criteria: the stronger the 2 μm water-ice band is, the stronger the crystalline feature is. The surface of the leading hemisphere exhibits much stronger signatures than the trailing one. The smallest values (≤0.85) derived in the northern hemisphere of the leading side presume the presence of crystalline water in large abundance. As for the absorption at 2 μm, the map highlights correlations with some major geomorphological units: the Powys Regio and Tara Regio located in the leading hemisphere have slightly higher values than their surrounding areas, while the center of the Pwyll crater exhibits a smaller value (∼0.90) than its surroundings. The area with the weakest absorption is actually well centered on the trailing apex and has an almost perfect ellipsoidal shape. Of special interest is the dependence of the value on latitude in the trailing hemisphere: for a given longitude, the absorption increases with latitude, which is consistent with the decreasing implantation of exogenous sulfur flux and the precipitation of energetic electrons with latitude (Paranicas et al. 2001; Hendrix et al. 2011; Patterson et al. 2012; Dalton et al. 2013).

3.3.1. Global Comparison

Spectral features significantly depend on location. Examples are shown in Figure 5. Spectra are extracted from four regions of interest and are well representative of the spectral diversity of Europa in the NIR wavelength range: (1) leading hemisphere at 120°W, 30°N in the bright icy area in the visible, (2) Dyfed Regio in the trailing hemisphere, which exhibits a low albedo in the visible wavelengths range, (3) a transition zone at (195°W, 0°N) close to antijovian point, and (4) Tara Regio.

Zoom In Zoom Out Reset image size

Overall, the spectral features are fairly consistent with previous NIMS and ground-based observations (Carlson et al. 1999b, 2005; Dalton 2007; Brown & Hand 2013): the leading hemisphere spectra are dominated by well-defined and strong water-ice absorptions, while the trailing hemisphere spectra exhibit distorted and asymmetric water-ice absorptions. Between these two end-members, the spectral features vary gently from place to place. Of special interest is the weak bump at ∼2 μm leading to an absorption feature at ∼2.07 μm which is visible for the spectrum extracted from Dyfed Regio. This feature was first highlighted by Brown & Hand (2013) and interpreted as an absorption feature from magnesium sulfate MgSO₄.7(H₂O) (epsomite). In order to emphasize this feature, we attempted to perform a spectral ratio by dividing this spectrum by the icy one lacking this feature. However, when making the ratio, this absorption remains invisible because of its weakness compared to water-ice absorptions, which are predominant (Figure 5). Another feature present in all spectra is a shoulder at 1.73–1.74 μm that is sometimes associated with very weak absorption at 1.77–1.78 μm. This feature was also clearly observed in the NIMS and Keck spectra (McCord et al. 1998b; Brown & Hand 2013). Orlando et al. (2005) associated this feature with small cations such as Na⁺ and Mg²⁺ in brines, while Dalton et al. (2005) highlighted its spectral coherence with mirabilite (Na₂SO₄.10(H₂O)) and Mg-sulfate dodecahydrate (MgSO₄.12(H₂O)). Nevertheless, this feature attracted little attention. Given the lack of clear and distinct absorption of a non-icy component, we will see further in this section that these small shoulders are essential to better constrain the nature of the hydrates responsible for the distortion of the water-ice bands.

3.3.2. Crystalline Water Ice

The four representative spectra reveal clear minima at 1.57 and 1.65 μm. The combination of these two strong signatures provides evidence for the presence of crystalline water ice on Europa's surface, as amorphous water ice does not exhibit such strong absorptions (Mastrapa et al. 2008). These absorptions are of special interest because their position are shifted with temperature, and thus can be used as a temperature proxy (Mastrapa et al. 2008). The most reliable value of the surface temperature should be given by the "pure icy" spectrum extracted from the leading hemisphere, which is considered to be less affected by radiation (Pospieszalska & Johnson 1989; Paranicas et al. 2001; Patterson et al. 2012). This radiation could indeed significantly modify the physical properties of the water ice and produce a variation of the 1.65 μm band profile and position (Leto et al. 2005). To increase the S/N, we calculated an average spectrum from a square of 11 × 11 pixels centered on the pixel (120°W; 30°N) representing the icy area in Figure 4, and then resampled the spectrum over 20 wavelengths. The spectral minimum is found to be at 1.650 ± 0.003 μm (Figure 6), which is consistent with the results of Dalton et al. (2012) from low-noise Galileo NIMS observations of the mean temperature. It corresponds to a temperature close to 120 K according to Mastrapa et al. (2008). This temperature is well representative of the mean temperature of the leading hemisphere because more than 97% of its pixels have a spectral minimum in the range λ = 1.649–1.652 μm, which corresponds to T = 110–130 K according to Mastrapa et al. (2008). Moreover, 33% of pixels have a minimum at λ = 1.650 μm, and this value is in the range of daytime temperatures obtained from the Galileo photopolarimeter-radiometer (Spencer et al. 1999; Rathbun et al. 2010).

Zoom In Zoom Out Reset image size

3.3.3. Additional Feature: 1.73 $\mu m$ Slope Break

As mentioned previously, a second subtle feature is present on every spectrum: a slope break at 1.73–1.74 μm, sometimes with a very weak absorption at 1.77–1.78 μm. The mixture of Mg-sulfate salts with water ice and sulfuric acid hydrate typically used in previously published spectral modeling of Europa's surface cannot reproduce this feature (Figure 7). The sodium sulfate mirabilite (Na₂SO₄.10(H₂O)) was also initially considered as a potential candidate to explain this feature, but the absence of other mirabilite signatures that should be seen in the data led to the rejection of this mineral (Dalton 2007; Brown & Hand 2013).

Zoom In Zoom Out Reset image size

We started considering a very large variety of potential minerals that could help in reproducing this feature. The recent study of Hanley et al. (2014) presents cryogenic reflectance spectra of many chlorine salts. Some of them exhibit a clear shoulder at 1.70–1.75 μm that is likely due to the hydrated phases of these salts. This feature is associated with asymmetric and distorted water-ice absorption presenting a shifted spectral minimum at 1.93–1.97 μm (Figure 7). Such spectral characteristics are consistent with the observations in SINFONI spectra of the non-icy area. In this respect, we will further consider this salt species as a potential candidate to reproduce some of the spectral properties of Europa surface (see Section 4.3).

In order to refine the surface composition and its variations at the surface of Europa, SINFONI spectra are resolved into materials via a linear deconvolution technique. Linear deconvolution is based on the principle that the spectrum is a linear combination of the spectra of the different species composing the icy and non-icy surfaces of the satellite in proportion to their abundance. The unmixing technique uses a library of end-member laboratory spectra to perform an iterative best fit of the spectra. Each modeled spectrum ${\boldsymbol{M}}({\boldsymbol{\lambda }})$ is thus calculated as

$$\begin{eqnarray}&&{\boldsymbol{M}}({\boldsymbol{\lambda }})=\sum _{{\boldsymbol{i}}={\boldsymbol{1}}}^{{\boldsymbol{N}}}\;{{\boldsymbol{A}}}_{{\boldsymbol{i}}}\times {{\boldsymbol{R}}}_{{\boldsymbol{i}}}({\boldsymbol{\lambda }})\times {\boldsymbol{S}},\end{eqnarray} \tag{ 3 }$$

where ${\boldsymbol{N}}$ is the number of reference laboratory spectra ${{\boldsymbol{R}}}_{i}({\boldsymbol{\lambda }}$) used and ${\boldsymbol{A}}$_i is their abundance. Since slight geometric and photometric uncertainties may affect the spectral slope, one additional free parameter S is used to adjust the continuum spectral slope. The downhill simplex method we used (Nelder & Mead 1965) proceeds to successive calculations, varying the abundances of each end-member and the slope value until a best fit is achieved. The quality of the fit is evaluated by a visual, qualitative comparison of the model and measured spectra, and is quantified by the chi-squared value ${{\boldsymbol{\chi }}}^{2}$. To calculate ${{\boldsymbol{\chi }}}^{2}$, we excluded the 1.79–1.94 μm wavelength range due to spurious values produced by residual telluric lines. Conversely, we doubled the weight for the wavelength ranges 1.56–1.67 μm and 1.97–2.10 μm to better reproduce the crystalline water-ice absorption feature and the absorption at ∼2.07 μm on the trailing side spectra (Figure 5 and discussed in Section 3.3.1).

Linear unmixing is a first-order approach justified by the fact that the large pixel scale of SINFONI unavoidably results in the spatial (hence linear) mixing of various surface components. The presence of several components within the SINFONI pixel scale is confirmed based on higher-resolution SSI and NIMS observations. However, as we will discuss later, residuals from the model may be due in part to intimate mixing which is not accounted for in a linear approach.

Initial examinations of the spectra show general matches to the laboratory spectra of end-members already used in previous studies, namely, water ice (both crystalline and amorphous water ice) mixed with non-icy components (possible sulfate minerals and/or sulfuric acid). For sulfate salts, the spectra of bloedite, hexahydrite, and epsomite were measured at 120 K by Dalton & Pitman (2012) and sulfuric acid hydrate is the octohydrate measured at ∼140 K originating from Carlson et al. (1999b). As discussed in Section 3.3.3, some subtle features in the 1.68–1.80 μm wavelength range could indicate the presence of chlorine salts. Various phases were considered with the cryogenic spectra at 80 K all coming from Hanley et al. (2014): the chloride MgCl₂.2(H₂O), the chlorate Mg(ClO₃)₂.6(H₂O), and the perchlorate Mg(ClO₄)₂.6(H₂O). The non-icy reference spectra we used as end-members are shown in Figure 7. The temperature of their acquisition and the size of the grains (when specified) are reported in Table 2.

Finally, the reflectance spectra of amorphous and crystalline water ices at 120 K were derived from the Shkuratov radiative transfer model (Shkuratov et al. 1999; Poulet et al. 2002) using the optical constants of both types of water ices provided by Mastrapa et al. (2008) and by the GhoSST online database (http://ghosst.osug.fr). We generated reflectance spectra for grain sizes d = 5 μm, 25 μm, 200 μm, and 1 mm.

We first applied the linear mixture model to the spectra representative of the spectral diversity of Europa (Figure 4). We adopt a tailored, iterative approach to unmixing, starting with a small set of end-members and then increasing to a larger set of end-members to improve the fit (Figure 8). This step is crucial to determine which components will be required for subsequent modeling of the complete data set (see Section 5). Based on previous modeling, the presence of sulfuric acid hydrate mixed with water ice (both amorphous and crystalline) is assumed to be mandatory, so that the following eight end-members, i.e., four crystalline water-ice spectra (5 μm, 25 μm, 200 μm, 1 mm), three amorphous water-ice spectra (25 μm, 200 μm, 1 mm), and one sulfuric acid hydrate spectrum, are by default included in the spectral library.

Zoom In Zoom Out Reset image size

Mixtures of these eight end-members provide satisfactory fits (referred to as model 1 of Figure 8) of the leading icy area and Tara Regio spectra, while some clear discrepancies are visible for the less icy spectra (named "non-icy" and "transition zone" in Figures 5 and 8). These discrepancies are notably strong for "ice-related" absorption features at 1.5 and 1.65 μm, as well as at ∼1.75 μm and the ∼2 μm bump. These discrepancies decrease gradually as spectra become icier, but even for the "pure" icy spectrum, a small but still significant difference given the high S/N of the data persists in the 1.73–1.80 μm range that cannot be attributed to instrument artefacts or residuals from the data-reduction process.

A second iteration in the optimization procedure consists of adding different sulfate salts such as bloedite, hexahydrate, and epsomite (referred to as model 2 of Figure 8). Surprisingly, the addition of these end-members does not improve the fits, especially for the non-icy one which shows a frank overestimation between 1.69 and 1.8 μm. This result challenges the presence of Mg/Na-sulfate salts on Europa's surface. A third test is applied by replacing Mg/Na-sulfate salts by Mg-chlorate (Mg(ClO₃)₂.6(H₂O)) and Mg/Na-perchlorates (Mg(ClO₄)₂.6(H₂O), NaClO₄.2(H₂O)). Modeled spectra (referred to as model 3 of Figure 8) are clearly improved, especially for the non-icy and the transition zone pixels. As shown in Figure 4, water-ice absorption features at 1.50 and 1.65 μm, as well as the small shoulder at 2.07 μm associated with the ∼2 μm bump, are much better reproduced. In addition, reflectance is no longer overestimated in the 1.69–1.80 μm range, which further emphasizes the presence of chlorine species. We then performed additional tests by replacing the perchlorates used in model 3 by chlorides such as NaCl and MgCl₂.2(H₂O) (referred to as model 4 of Figure 8). Best-fit modeled spectra are globally less convincing than those of the previous test, with the exception of the 1.73 μm slope break and the slight absorption at 1.77 μm visible in the non-icy spectra which are well reproduced. This results from the spectral contribution of MgCl₂.2(H₂O), which has an absorption feature in this wavelength range. For this reason, we will keep this end-member for the next modeling. We also evaluated the presence of Na-chloride and Na-perchlorate instead of Mg-bearing ones (not shown here). Once again, the modeled spectra are worse than those obtained with model 3, as confirmed by larger rms and sometimes by a weak but distinct absorption at ∼2.24 μm resulting from the Na-perchlorate. Such a feature is not present in the data, likely ruling out the presence of this mineral on the surface of Europa.

Several other combinations using the different species listed in Table 2 were tested but the best fits (referred to as the final model of Figure 8) were eventually obtained by using only Mg-bearing chlorinated salts (the chloride MgCl₂.2(H₂O), the chlorate Mg(ClO₃)₂.6(H₂O), and the perchlorate Mg(ClO₄)₂.6(H₂O)) in addition to the default end-members listed previously. Fits are actually excellent for the Tara Regio and the leading hemisphere icy area. The non-icy spectrum with its bump at ∼2 μm is very well reproduced. The discrepancies that are still present in the non-icy spectra are the position and the band depth of the 1.50 and 1.94 μm absorptions and the slightly curved shape of the model in the range 2.25–2.45 μm. These discrepancies could result from a limited range of end-members and a degree of hydration or grain size effect. Radiation could also significantly modify these features (Leto et al. 2005). Using water-ice spectra at 80 and 100 K instead of at 120 K produces only minor variations in terms of abundance (<8%). A last best-fit test was performed using the complete library, but the improvements compared to the previous modeling are not significant. We conclude that the surface of Europa as observed by SINFONI can be correctly modeled with the three following families of species: crystalline and amorphous water ices, sulfuric acid hydrate, and Mg-chlorinated salts; meanwhile, sulfate salts are not required.

As expected, the sulfuric acid hydrate is mainly present on the trailing side (Figure 10) with a maximum concentration of ∼65% near the trailing apex (270°W, 0°N). This abundance is lower than the ∼90% reported in Carlson et al. (2005) and Brown & Hand (2013). The largest concentrations are consistent with the trailing apex (0°W, 270°N) and are not correlated with geomorphological units. In the leading hemisphere, the spatial distribution is also uneven and uncorrelated with geomorphology. The range of values goes from 0% to 10% in the bright, icy area to 30%–40% in the eastern part of the Tara Regio and in the Balgatan Regio (30°W, −50°N). In both hemispheres, the concentration of sulfuric acid hydrate decreases with latitude and is anti-correlated with the water-ice band depth maps previously presented (Figure 4). A first attempt at estimating the uncertainty due to the measurement of the abundance is to compare the derived values for regions that were observed twice. The equatorial region of longitude 175°W was observed during the 2011 October 04 and 2011 October 21 runs. Differences of ∼20%, ∼25%, and ∼10% for sulfuric acid hydrate, water ices, and chlorine salts, respectively, were derived between the best fits of the two juxtaposed regions observed during two different nights. As mentioned in the description of the 2 μm reflectance map (see Section 3.1), this discrepancy is due to an accumulation of small approximations before reconstructing the global 3D reflectance cube and leads to an error of ∼25% on the reflectance level. Additional tests that take into account this uncertainty have been performed on the reference spectra presented in Figure 5 and, as expected, we found very similar abundance variations to those mentioned previously in this section. Thus, the difference of 25% is the largest derived for all of the end-members, and it corresponds to the maximal error on the abundance estimate.

Zoom In Zoom Out Reset image size

5.2.1. Amorphous Water Ice

The amorphous water-ice composition map using the three different grain sizes is exclusively dominated by the leading/trailing effects (Figure 11(a)). The lowest concentrations are located near the trailing apex with an average value lower than 5%, while the highest values reach 40% at high latitudes in the south of the leading hemisphere. Geomorphological units are not distinguishable.

Zoom In Zoom Out Reset image size

When analyzing the spatial distribution as a function of the three different grain sizes used in the modeling, we observe that the largest grain size (1 mm) has a very low abundance below the detection threshold, except for a spot of ∼20% in the south–west of the trailing hemisphere (340°W, −40°N). This area is spatially correlated with the area that exhibits the strongest 2 μm signature of the trailing hemisphere (Figure 4). The 200 μm sized grains are present only on the leading side, particularly in the bright, icy area and in the terrains surrounding the Tara Regio with maximum concentrations of 15%–20%. The spatial distribution of the 25 μm grains is actually very similar to the overall amorphous water ice, with concentrations slightly above 35% for some pixels in the southern part of the Powys Regio. This abundance, which is approximately twice as large as the largest abundance of 200 μm, combined with the quasi-absence of 1 mm grain size, therefore indicates a surface that is globally dominated by small grains.

5.2.2. Crystalline Water Ice

As for the sulfuric acid hydrate, chlorine salts (bearing the Mg cation here) are significantly more abundant in the trailing hemisphere than in the leading one (Figure 12(a)). However, unlike the former, their distribution is correlated with several geomorphological units, with regional enhancements of up to 35% corresponding to, e.g., the Dyfed Regio and the small dark chaos extending westward of the Annwn Regio. Significant concentrations of 20%–30% are also found in the Annwn Regio. In the leading hemisphere, abundances as low as 5% are observed in the icy areas, while the Tara Regio and Powys Regio locally exhibit higher abundances, sometimes exceeding 20%.

Zoom In Zoom Out Reset image size

Among this family, the Mg-perchlorate distribution (Figure 12(b)) shows a strong dichotomy between the two hemispheres as the largest abundances (∼17%) are approximately centered on the trailing apex where the surface is most subject to exogenous ionic and electronic implantation (Patterson et al. 2012; Cassidy et al. 2013). The second compound, Mg-chlorate, is low in abundance (≤12%), and no clear link to geomorphology can really be argued. This phase is at the limit of the detection threshold, although its addition to the spectral library still improves the quality of the fit, especially in the trailing hemisphere. Finally, Mg-chloride is the most abundant of the three chlorine salts used in the modeling, and its distribution, as shown in Figure 12(c), is globally similar to the global distribution of the chlorine species. An abundance peak is observed in the small, dark chaos at 350°W where a spot of 5 μm grain of crystalline water ice was also identified, as well as in the Dyfed Regio and Tara Regio. This unexpected spatial distribution linked with the geomorphology challenges the hypothesis on the fully exogenous origin of the non-H₂O species present on the surface of the satellite. In the following section, we address in greater detail the origin and evolution of the species against exogenic and endogenic processes.

The spectral modeling highlights the need for (Mg-bearing) chlorinated species to better reproduce the NIR spectral signatures on a global scale rather than Na/Mg-sulfates (Figure 8). All of the spectral characteristics not associated with H₂O ice or sulfuric acid hydrate absorptions, such as the slope break ∼1.73 μm and the small shoulder at ∼2.07 μm, can be explained by the spectral features of Mg-bearing chlorinated species. The ∼2 μm distorted signature is also very well fit. Nevertheless, the presence of the three chlorine species used in this work cannot be definitely asserted as their spectral effects are subtle and bear similarities to other hydrated chlorinated species. The temperature of 80 K at which the laboratory spectra were acquired is not exactly representative of Europa's surface temperature, and their concentrations (≤35% in total) still remain minor compared to the H₂O ice and sulfuric acid hydrate. Using a spectral library at temperatures more similar to those of Europa could then modify the derived abundances of these species. Thus, additional high spectral resolution campaigns in different wavelength ranges (<1.5 μm) combined with a cryogenic library at temperatures >80 K would be helpful to better constrain the abundances of the Cl-bearing salts and to confirm or disprove their presence on the surface of the satellite.

Most studies have considered that the dark, non-ice component observed principally on the trailing hemisphere is related to Mg-sulfate species. This predominance of sulfates compared to chlorines was first attributed to a very low cosmic abundance ratio Cl/S (Kargel 1991) and supported by theoretical and experimental estimates of Europa's ocean composition based on the aqueous alteration of a CM chondrite meteorite (Kargel et al. 2000; Fanale et al. 2001; Zolotov & Kargel 2009). However, chlorine ions have been detected in the plasma torus (Küppers & Schneider 2000; Feldman et al. 2001), and water-rock cycling at the silicate seafloor could lead to a chloride-rich ocean (Glein & Shock 2010). While some geochemical models predict that Europa's initial chondritic composition should lead to a chlorine-poor ocean, recent theoretical models of the subglacial ocean of Enceladus, the Saturnian icy moon comparable to Europa, predict a Cl-rich ocean (Zolotov 2007; Glein et al. 2015). The chlorine salts hypothesis was also recently proposed to explain the color of the Europan surface in the visible and NIR wavelength ranges (Fischer et al. 2015; Hand & Carlson 2015).

Concerning the magnesium, the absence of magnesium ions in the Ionian plasma torus and the very low contribution from micrometeoroid bombardment suggests an endogenous origin for this chemical element (Carlson et al. 2009). Its predominance on the surface compared to Na/K-bearing species may be explained by a sputtering erosion rate that is much less efficient for Mg than for Na and K (Horst & Brown 2013), as has been experimentally demonstrated for sulfate salts (McCord et al. 2001). Finally, Brown & Hand (2013) also hypothesized that Mg should primarily be in the form of MgCl₂ before being irradiated on the surface.

Our study confirms the high predominance of H₂O ice and sulfuric acid hydrate, as they represent overall more than 80% of the surface abundance on average. These two components are well anti-correlated: while the overall abundance of H₂O ice rises to almost 90% on the bright (visible) area of the leading northern hemisphere, the sulfuric abundance increases gradually toward the trailing hemisphere until reaching approximately 65% on the orbital trailing apex, which corresponds to the highest magnetospheric bombardment intensity and the center of the sulfur flux (Paranicas et al. 2001; Hendrix et al. 2011; Dalton et al. 2013). However, as previously mentioned, the abundance is ∼30% lower than those obtained by Hand & Brown (2013) and Carlson et al. (2005) from nonlinear modeling. This difference could be due to the use of a linear versus a nonlinear approach. It is important to note that the 1.65 μm band is much more pronounced in our high spectral resolution data and cannot be well reproduced by less than 15% of water ice (Figure 8). The spatial blurring induced by our moderate Strehl ratios should not induce contamination of the 1.65 μm band within the sulfuric unit by neighboring H₂O-rich regions, given the large size of the observed unit.

Most of the spectra can be correctly fit only using the 25 and 200 μm grain sizes. This is consistent with many previous studies (Calvin et al. 1995; Hansen & McCord 2004; Dalton et al. 2012), but no clear relationship between the grain sizes and exogenous processes has been identified, contrary to studies based on very localized Galileo NIMS observations (Dalton et al. 2012; Cassidy et al. 2013). The distributions of larger grains, such as 200 μm and 1 mm (Figure 11(c)), do not show any spatial coherence with the apex of the trailing hemisphere, while the 5 μm crystalline ice distribution also does not look to be controlled by the intensity of the bombardment (Figure 11(d)). This difference suggests that exogenous processes are not the only explanation for the H₂O grain size distribution.

According to a study based on the 3.1 μm water absorption, H₂O ice on Europa is presumed to be mainly amorphous (Hansen & McCord 2004). It is intended that the 3.1 μm Fresnel reflection peak provide major constraints on the nature, whether crystalline or amorphous, of the water ice. However, the 1.65 μm absorption band has also been proven capable of discriminating between the two types of water ice (Schmitt et al. 1998; Mastrapa et al. 2008). Thanks to the high spectral resolution of SINFONI, a detailed study of this absorption has been carried out. Figure 6 clearly highlights that the depth and the position of this absorption cannot be reproduced by amorphous ice alone, both for the leading and trailing hemispheres. The spectral modeling on the global scale also shows the slight predominance of crystalline H₂O ice with the ratio of amorphous/crystalline = 0.57 ± 0.25 on average. This result is not inconsistent with the former study based on the 3.1 μm Fresnel. Indeed, it suggests that a positive vertical crystallinity gradient exists in the ice slab covering the surface, as already proposed by Hansen & McCord (2004). This could provide useful hints for reinvestigating the balance between the disruption of crystalline ice by irradiation and crystallization by heating. The surface temperature of Europa being in the range 110–130 K suggests that the crystallization of impure amorphous ice should occur in less than 10 years (Kouchi et al. 1994; Jenniskens et al. 1998; Hansen & McCord 2004; Mastrapa et al. 2013), while the amorphization time is expected to be about 0.6 year for the ion flux Φ = 10⁷ cm² s⁻¹ corresponding to the trailing apex (Fama et al. 2010).

The large predominance of sulfuric acid hydrate on the trailing hemisphere makes interpretations tricky for any process that could control the nature of the water ice in this hemisphere. However, it is worth noting that the crystalline spots detected at high latitude in this hemisphere (Figures 11(b), (c)) may be explained by the lower efficiency of the radiation-based amorphization at these high latitudes. While both types of water ices are present in the leading hemisphere, significant variations of their relative concentration are nevertheless observed. The relatively high abundance of crystalline ice in the triangular bright, icy area is surprising. It could correspond to a relatively warm formation environment, possibly linked to thermal convection in the ice shell. Melt-through events on Europa due to tidal stress are indeed expected to lead to warmer areas (Greenberg et al. 1998; O'Brien et al. 2002). The increase of amorphous ice concentration in the high latitudes of this hemisphere could indicate a specific source of amorphous ice. Condensation of water vapor near the pole has already been proposed (Hansen & McCord 2004). However, better coverage of the very high latitudes is still required before we can confirm this observation. A very specific and unique concentration of small grain sized crystalline ice is also observed in the chaos located close to the meridian. This could again imply a warm environment, possibly linked to an endogenic process as the chloride salts also exhibit an enhancement of their abundance (see the next section).

In agreement with previous studies, the origin of sulfuric acid is proposed to be exogenous, precisely Iogenic, as its distribution is perfectly correlated to the Iogenic sulfur ion implantation flux (Hendrix et al. 2011; Dalton et al. 2013). An exogenous contribution of chlorine is very likely as the ionic form of chlorine (Cl⁺) is present in the plasma torus with an abundance of 1%–3% relative to S⁺ (Küppers & Schneider 2000; Feldman et al. 2001). Conversely, none of the magnesium ions (Mg⁺, Mg²⁺) have been detected so far in the Ionian plasma torus (Carlson et al. 2009), which favors the endogenous source for this element. Mg²⁺ can indeed be produced in significant quantities by aqueous alteration processes on silicate rocks. This questions whether the spatial distribution of Mg-bearing chlorinated species derived from this study is compatible with an exclusively exogenous source of Cl. If so, then these species would be primarily distributed on the trailing hemisphere and centered on its apex, similar to sulfur-bearing acid hydrate. Regions where the salts, and particularly the chloride species, show an excess concentration correspond to three well-defined chaoses: the Annwn Regio and its westward dark chaos, the Dyfed Regio and, to a lesser degree, the Tara Regio (Figures 12(a), (c)). This specific distribution could be the footprint of active tectonics and cryovolcanism provoked by a subsurface ocean (Greeley et al. 1998; Kattenhorn & Prockter 2014; Roth et al. 2014). As mentioned in Section 6.1, some theoretical models of subglacial ocean composition have predicted an alkaline ocean with Cl⁻ as the major anion species (Zolotov 2007; Glein et al. 2015), hence suggesting an additional endogenous contribution for chlorine. The presence of both Cl⁻ and Mg²⁺ in aqueous solution should lead to the formation of MgCl₂ chloride and its hydrated forms, such as MgCl₂.2(H₂O) used in our modeling.

The presence of chloride minerals is a prerequisite to form other chlorinated species such as chlorates and perchlorates. Recent experimental works mimicking the Martian environment show that ${{\mathrm{ClO}}_{4}}^{-}$ (and ${{\mathrm{ClO}}_{3}}^{-}$) can be generated as end-products of photochemical reactions (like UV radiation) or oxidation by oxygen containing species on Cl-minerals without involving any aqueous phase (Kim et al. 2013; Carrier & Kounaves 2015; Jackson et al. 2015). As photochemical reactions do not require thermal activation, they should also occur at lower temperatures, such as those of Europa. Because these reactions mainly occur on the trailing hemisphere, thanks to the Ionian plasma torus, the distribution of ${{\mathrm{ClO}}_{4}}^{-}$ is, as expected, centered on the trailing hemisphere apex (Figure 12(b)).