Carbon dioxide on Ganymede

Transcription

1 JOURNAL OF GEOPHYSICAL RESEARCH, VOL. 108, NO. E5, 5036, doi: /2002je001956, 2003 Carbon dioxide on Ganymede C. A. Hibbitts 1 Department of Earth and Space Sciences, University of Washington, Seattle, Washington, USA R. T. Pappalardo Astrophysical and Planetary Sciences Department and Laboratory for Atmospheric and Space Physics, University of Colorado, Boulder, Colorado, USA G. B. Hansen and T. B. McCord The Planetary Science Institute, NW Division, Seattle, Washington, USA Received 2 July 2002; revised 19 November 2002; accepted 5 February 2003; published 13 May [1] CO 2 is inferred on the surface of Ganymede by the presence of an absorption band at 4.26 mm in reflectance spectra returned by the Near Infrared Mapping Spectrometer (NIMS) aboard the Galileo spacecraft. Detailed studies of NIMS observations of Ganymede show that the CO 2 absorption band on Ganymede is symmetric about mm ± mm and is negligibly different from the CO 2 absorption band on Callisto. In general, bright terrains (sulci) contain less CO 2 than dark terrains (regiones), little or no CO 2 is detected at the poles, and, unlike for Callisto, there does not appear to be any leading/trailing hemisphere asymmetry in the distribution of CO 2 nor do impact craters tend to be CO 2 rich. High spatial resolution observations show that CO 2 is occasionally enriched in terrain containing larger-grained ice in comparison with adjacent terrain of similar morphology and ice abundance, that the dark ejecta of Kittu is depleted in CO 2 with respect to adjacent terrain, and that only one observed impact crater (Mir) is enriched in CO 2. The CO 2 that is detected by NIMS is in the nonice material(s) present, not in the fraction that is ice. The ice in all terrain types where CO 2 is detected is always sufficiently large-grained that its low reflectance in the region of the CO 2 band prevents discrimination of the CO 2 absorption band from the continuum. Areas with relatively fine-grained ice that are sufficiently bright at 4 mm for CO 2 to be observed, such as the polar regions, do not contain detectable CO 2. INDEX TERMS: 6218 Planetology: Solar System Objects: Jovian satellites; 5410 Planetology: Solid Surface Planets: Composition; 5470 Planetology: Solid Surface Planets: Surface materials and properties; KEYWORDS: Ganymede, NIMS, Galileo, infrared spectroscopy, icy satellites, Galilean satellites Citation: Hibbitts, C. A., R. T. Pappalardo, G. B. Hansen, and T. B. McCord, Carbon dioxide on Ganymede, J. Geophys. Res., 108(E5), 5036, doi: /2002je001956, Introduction [2] Carbon dioxide has been discovered on the surface of Ganymede through its absorption band at 4.26 mm [Carlson et al., 1996; McCord et al., 1998] in infrared reflection spectra returned by the Near Infrared Mapping Spectrometer (NIMS) aboard the Galileo spacecraft. McCord et al. [1998] infer from the shape and location of this absorption band that the CO 2 is bound, or trapped, in a host. This host also prevents the CO 2, which has an equilibrium vapor pressure of 10 4 to 10 torr [James et al., 1992] at the 100 to 160 K temperature of Ganymede s surface [Orton et al., 1996], from otherwise quickly escaping into space. McCord 1 Also at HIGP/SOEST, University of Hawaii, Honolulu, Hawaii, USA. Copyright 2003 by the American Geophysical Union /03/2002JE et al. [1998] mapped the distribution of CO 2 over the anti- Jovian hemisphere using observations from the first Galileo orbit. They describe the CO 2 distribution on the surface as mottled, with a depletion at higher latitudes. They ascribe this depletion to fine-grained ice masking CO 2 -bearing terrain. In this paper, we use all available NIMS observations of Ganymede through the ninth orbit of the Galileo spacecraft to investigate the spectral characteristics of the CO 2, to map its spatial distribution, to search for possible relationships between surface morphology and CO 2 abundance, and to look for any other controls on the abundance, distribution, and host material of CO 2 on Ganymede. 2. Background 2.1. Geology of Ganymede [3] The surface of Ganymede is divided into two principal terrain types, relatively old regions of dark terrain ( regiones, e.g., Galileo Regio) and younger crosscutting 2-1

2 2-2 HIBBITTS ET AL.: CO 2 ON GANYMEDE lanes of bright grooved terrain ( sulci, e.g., Nippur Sulcus). Impact structures, including bright craters, craters with dark ejecta rays, and larger low-relief impact scars, are distributed across all terrains. We use high spatial resolution observations from the cameras aboard the Voyager and Galileo spacecraft to characterize the geology of terrain that has also been compositionally mapped with respect to CO 2 using the NIMS Dark Terrain (Regiones) [4] Areas of heavily cratered dark terrain comprise about 35% of Ganymede s surface [Collins et al., 2000]. Based on measured crater densities, dark terrain is generally estimated to be >4 Gyr old [Neukum, 1997; Zahnle et al., 1998] and is commonly transected by arcs of several kilometer-wide troughs, termed furrows [Smith et al., 1979a, 1979b]. These furrows are inferred to be ringed depressions subconcentric to the sites of ancient, large impacts that punched through a relatively thin, brittle lithosphere into more mobile material below [McKinnon and Melosh, 1980] and are found only in regiones, the oldest terrain on Ganymede. [5] Geological investigations using Galileo high-resolution images suggest that the dark material itself is a relatively thin lag deposit above brighter icy material and has been affected by processes of sublimation, mass wasting, ejecta blanketing, and tectonism [Prockter et al., 1998, 2000]. Dark terrain is very heterogeneous in albedo at decameter scales, presumably owing to thermally driven segregation of ice and nonice surface components [Spencer, 1987]. The darkest deposits occur within local topographic lows such as crater and furrow floors, suggesting that downslope movement of the dark nonice component plays an important role in albedo segregation [Oberst et al., 1999]. Thus it appears for regolith in nonpolar regiones, the dark nonice component is segregated from the icy component Bright Terrain (Sulci) [6] Swaths of bright terrain (sulci), typically 100s of kilometers wide, cross cut the older dark terrain and each other to form an intricate patchwork across 2/3 of Ganymede s surface [Collins et al., 2000; Shoemaker et al., 1982]. The sulci are significantly less cratered than dark terrain and are therefore younger, plausibly emplaced sometime between 500 Myr and 1 Gyr ago [Zahnle et al., 1998]. Sulci are heavily transected by subparallel ridges and troughs, or grooves, at a variety of scales and thus are commonly referred to as grooved terrain. These features are interpreted as normal fault blocks formed by lithospheric stretching, potentially as a result of satellite expansion during differentiation and/or mantle convection [Shoemaker et al., 1982; Parmentier et al., 1982; Golombek, 1982; Grimm and Squyres, 1985; Golombek and Banerdt, 1986; McKinnon and Parmentier, 1986; Collins et al., 1998a; Pappalardo et al., 1998a]. Examination of Galileo images shows that extensional strain in grooved terrain lanes can locally exceed 50% [Collins et al., 1998b; Pappalardo and Collins, 1999]. The high extensional strain indicated by Galileo images, along with new laboratory data on the rheology of ice, suggest that extensional lithospheric boudinage is a viable model for initiation of grooved terrain and imply a brittle lithosphere 2 km thick existed at the time of deformation [Collins et al., 1998b; Dombard and McKinnon, 2001]. Similar dominant wavelengths of deformation across Ganymede suggest a similarly thin brittle lithosphere was pervasive during grooved terrain development [Grimm and Squyres, 1985; Patel et al., 1999]. Some bright lanes are topographically subdued, suggesting that icy volcanism may have played an important role in their emplacement [Schenk et al., 2000; Giese et al., 2001], as originally suspected based on Voyager observations [Parmentier et al., 1982; Shoemaker et al., 1982]. Recent Galileo images of an unusual lane of grooved terrain named Arbela Sulcus suggest that it may be a site of complete lithospheric separation and spreading, akin to some bands on Europa, and also reinforce the important role of tectonism even in terrain that is relatively smooth at Voyager resolution [Head et al., 2002] Impact Features [7] Impact craters on Ganymede can have bright impact ejecta (presumably ice-rich) or dark ejecta (presumably less icy). The origin of dark crater ejecta has been the subject of debate. Hartman [1980] suggests that dark ray craters are indicative of subsurface chondritic silicates exposed by impacts. However, Conca [1981] proposes that dark ray craters are indicative of projectile contamination because the craters are a small minority of all ray craters but occur in all terrain types and are of all sizes. He suggests the concentration on the trailing hemisphere is a consequence of sputtering effects preferentially removing the masking water ice. Based on color analyses, Schenk and McKinnon [1991] concur that dark ray craters most likely represent impactor contamination. They also find that Kittu crater has color properties distinct from other dark ray craters, suggesting a different composition for that impactor. The spectrum of Kittu is flat in the visible, consistent with C- type material from a compositionally distinct impactor, rather than showing a red slope like other dark-ray craters. [8] Ganymede s largest impact scars (up to 100s of kilometers in diameter) appear as low relief bright patches, termed palimpsests, some of which show associated basinlike ring structures [Shoemaker et al., 1982]. Palimpsests are most noticeable in dark terrain, where they are in greatest contrast to their surroundings. Analysis of Galileo images indicates that the bright palimpsest material probably represents relatively clean, fluid-rich icy ejecta excavated from depth (K. B. Jones et al., Morphology and origin of palimpsests on Ganymede from Galileo observations, submitted to Icarus, 2003). They infer that these impact structures have formed with little relief, presumably because the weak icy lithosphere could not support significant crater topography during impact Composition of Ganymede s Surface [9] The surface of Ganymede is primarily composed of water ice and a darker nonice material, which itself may be a mixture of several component materials [e.g. Pilcher et al., 1972; Pollack et al., 1978; Clark and McCord, 1980]. Ganymede s icy polar caps [Smith et al., 1979a, 1979b] are interpreted to result from charged-particle bombardment causing damage and/or sputtering of water ice that is then redeposited locally as frost [Johnson, 1985, 1997; Pappalardo et al., 1998b]. Reflectance spectra containing asymmetric water-ice absorption bands are prevalent in some areas [McCord et al., 2001] and may be indicative of hydrated salts like on Europa [McCord et al., 2001] or of other hydrated minerals [Carlson et al., 1999].

3 HIBBITTS ET AL.: CO 2 ON GANYMEDE 2-3 Figure 1. Average spectra from 16 NIMS observations of Ganymede. The spectra are offset from each other for ease of viewing but not scaled. (a) The water-ice absorptions and the 1.5-mm water-ice absorption band modeled fit (dotted curve) are plotted. (b) The 4.26 mm CO 2 absorption band and its modeled fit (dotted curve). The spectra of the icy observations, G8OSIRIS and G8TRANSI, are dashed. [10] Several other less-abundant materials have been inferred to exist in the surface of Ganymede through the presence of absorption bands in the ultraviolet, visible, and near-infrared. Besides CO 2 that is believed responsible for the 4.26-mm absorption band, these include sulfur and molecules containing S = O bonds detected by ultraviolet absorption features [Nelson et al., 1987] and a characteristic mid-infrared absorption band at 4 mm [McCord et al., 1998]. Ozone has been suggested based on a broad ultraviolet absorption feature [Noll et al., 1996; Hendrix et al., 1999], and the presence of molecular oxygen has been inferred by two absorption bands in the visible [Spencer et al., 1995]. Finally, other absorption bands in the 3-mm to 5-mm spectral region in data returned by NIMS aboard the Galileo spacecraft imply that C-H, C N, and S-H are also in the surface [McCord et al., 1998] Characteristics of the NIMS Infrared Reflection Spectra [11] The water-related absorption bands at 1.25, 1.5, and 2.0 mm are present in all the NIMS spectra of Ganymede and imply that bright areas are icier than the darker areas (e.g., McCord et al. [1998] and Figure 1). Although waterrelated absorption features dominate the spectra in the 1- to 2.5-mm region, the OH-stretch vibration absorption due both to water or OH-bearing minerals produces a strong absorption at 2.7 mm in all spectra. A 3.1-mm reflection peak due to water ice is noticeable in spectra of icier terrain (e.g., G2SIPPAR) and less so in less icy terrain (e.g., G7KITTU). Furthermore, the average spectrum of the G8OSIRIS observation has a broad reflection peak at 3.6 mm, consistent with the presence of fine-grained ice [Hansen et al., 1998]. Beyond these wavelengths, only the absorption feature due to CO 2 is commonly present (Figure 1b). However, other absorption bands likely due to SO 2, SH, and CN, at 4, 3.88, and 4.57 mm, respectively, can be detected in averages of some individual observations (Figure 1 and McCord et al. [1998]). [12] There are several instrumental effects of the NIMS data that have been accounted for in this analysis including a drift in the wavelength calibration during the mission, a failure of two detectors reducing the wavelength coverage, and the presence of anomalous amplitude spikes ( noise ) in the data. These characteristics of NIMS data and their

4 2-4 HIBBITTS ET AL.: CO 2 ON GANYMEDE Figure 2. A Mercator perspective SSI/Voyager mosaic with the outlines of the 17 NIMS observations that were used in the analyses. The names that are highlighted in black correspond to observations that have spatial resolutions (geometrically projected) better than 10 km/pixel. treatment are described more fully by McCord et al. [1998, 2001] and Hibbitts et al. [2000, 2002] NIMS Observations of Ganymede [13] There are 31 NIMS observations of Ganymede that were investigated for this study. In 14 of the observations, the spectra are either not of sufficient signal-to-noise for mapping composition on a pixel-by-pixel basis or contain insignificant CO 2 absorption bands. The 17 remaining observations are primarily of the anti-jovian hemisphere (Figure 2 and Table 1). Of these, 11 have spatial resolutions of the geometrically projected pixel that are less than 10 km and thus begin to resolve portions of craters. Their names are highlighted in black in Figure 2 and bold in Table 1. Between 240 and 005, Ganymede has been observed by NIMS only at low resolution, generally >50 km/projected pixel, except for two small, high spatial resolution observations of impact craters and another high-resolution observation in the northern midlatitudes of the trailing hemisphere (not shown in Figure 2). Longitudes from 5 to 50 have effectively not been observed, and longitudes from 50 to 80 were observed only once at low spatial resolution and high phase angle in an observation not suitable for this analysis. The phase angles of the observations from orbits 1, 2, 6, and 9 are all less than 35. Those from the seventh and eighth orbits of Jupiter have phase angles from 60 to 90. The lower phase angle G1GLOBAL and E6GLOBAL observations provide a global context for interpreting the higher resolution images. 3. Procedures 3.1. Mapping the Depth of Absorption Bands [14] The depth of the CO 2 absorption band at 4.26 microns and the depth of the water-ice absoprtion at 1.5 microns coupled with the 0.7-micron albedo are used as a means to map the distributions of CO 2 and water ice over the surface of Ganymede. Moreover the depth of the CO 2 absorption band can be used as a qualitative measure of relative abundance of CO 2 in the optical surface [Hibbitts et al., 2002]. Quantitative Table 1. NIMS Observations of Ganymede Observation Name Orbit Individual pixel field of view (IFOV), km Resolution of geometrically projected pixel, km Average Phase Angle Average CO 2 band depth, percent GLOBAL G to AMON G to MEMPIS G to MIRRAY G to NIPPUR G to ANTUM G to SIPPAR G to TAMMUZ G to GLOBAL E BRITRL G to KITTU G to MELKAR G to OSIRIS G to URUK G to TRANSI G to LIDARK G to DRKLIT C to

5 HIBBITTS ET AL.: CO 2 ON GANYMEDE 2-5 estimates of the abundance of CO 2 cannot be obtained from the depth of the CO 2 band because the state of the CO 2 is not fully understood and the optical properties of the host are not certain. These uncertainties prevent application of radiativetransfer theory for detailed modeling of the absorption band shape and depth. However, because the CO 2 and host material have a very low reflectance in the wavelength region of the CO 2 absorption band, we assume there is little multiple scattering of light and thus there is a relation, though not necessarily linear, between deepening of the CO 2 band and an increase in abundance [Hibbitts et al., 2002]. The depth of the water-ice 1.5-mm absorption band is used to help correlate NIMS surface morphology to SSI features and to help investigate ice grain size variations. The 1.5-mm absorption band depth can not be used by itself as a measure of relative abundance of ice because variations in grain size also affect the band depth. By combining band depth information and albedo information, we can qualify variations in ice grain size within individual observations. [15] Absorption band depth maps are developed following the same procedures used by Hibbitts et al. [2000, 2002]. In summary, CO 2 and 1.5-mm water-ice absorption band depths are calculated using a mathematical approximation of the absorption band shape determined from analyzing Ganymede observations obtained during the first nine orbits of Galileo. Estimates of absorption band depths are the depths of the modeled function as fit to the actual spectra. This technique is more accurate than using ratios of reflectance at a few wavelengths, especially for observations in which full spectral resolution is not available (e.g., NIMS short map or long map modes where only 6 or 12 of the 24 available wavelengths per detector are selected). [16] The effects of noise and residual artifacts in the data can also be quantified by using this technique. The leastsquares minimization used to fit the model spectra to the actual spectra provides a quantitative estimate of the error in proportion to the depth of the absorption band. This error term is called the band depth signal-to-noise (SNR bd ) and enables a quantitative assessment of the validity of each band depth approximation. Estimates that do not meet a sufficient level of accuracy can be discarded or averaged with their neighbors to improve the quality of the approximation, albeit at the sacrifice of effective spatial resolution. Water-ice or CO 2 band depth estimates which are less than 5 times greater than their least-squares error (SNR bd < 5) are averaged over a box of 3 3 pixels or, if necessary, over a box of 5 5 pixels to improve the fit and reduce the least-squares error. If the least-squares error is not reduced to meet this criterion for significantly nonzero band depth estimates, then the estimates are discarded. However, because the least-squares error of shallow band depth approximations will naturally have greater error relative to the depth of the band, band depth estimates which are less than the average for the observation are not discarded even if they do not meet the criterion after averaging. This technique of averaging or discarding select data improves the accuracy of estimates from deep absorption bands to a certainty of at least ±20% (and usually much greater) while making the visual interpretation of the band depth maps easier by not discarding shallow band depth estimates (Figures 3a and 3b). [17] The visual interpretation of the band depth data is further improved by smoothing the band depth map using bilinear interpolation (Figure 3c). This smoothes the pixelto-pixel boundaries, removing the high-frequency variations and making the interpretation of spatially broad features easier. Large variations between individual pixels are also considered high-frequency variations and are smoothed as well. This technique does not affect the utility of the smoothed band depth map for qualitative interpretations of spatially broad features, for example, the three regions of shallow CO 2 bands in the upper right and the deep bands in the lower left portion of Figure 3c. All quantitative analyses using band depth are performed on data unmodified by interpolation Coregistering NIMS and SSI/Voyager Data [18] At the highest NIMS spatial resolutions, typically on the order of 10 km/pixel and no higher than 1 km/pixel, for the most part it is not possible to resolve effects of geological features on CO 2 distribution. In order to explore the relationship between CO 2 abundance and surface features in greater detail, coregistered maps of CO 2 from NIMS observations and albedo maps from SSI and Voyager images are compared. Therefore it is necessary to use a common geometric perspective between the two instruments observations. Photometrically corrected and contrast-stretched SSI or Voyager panchromatic images are first projected to the same perspective as the corresponding NIMS observations of the same area on Ganymede. The NIMS and panchromatic images are then registered together by linear and nonlinear warping. Warping of the images is necessary because the pointing information between the two instruments is not completely consistent, which results in slight differences when projected to what should be the same perspective. These pointing inconsistencies between the SSI/Voyager and NIMS observations are unique in magnitude and direction for each NIMS observation and require that analyses of registered observations be performed on an observation-by-observation basis. [19] Features that are bright in NIMS images at the visual wavelength of 0.7 mm are also bright in Voyager and SSI panchromatic images. This albedo correlation allows for finely tuned registration between the two data sets. The depth of the 1.5-mm absorption band is not used for aligning NIMS observations with panchromatic SSI observations because, unlike for Callisto [Hibbitts et al., 2002] there is not always a correlation between deeper bands in the NIMS images and brighter features in the panchromatic images. The precision of image registration is ultimately limited by the spatial resolution of NIMS. If a contrasting albedo feature imaged by SSI or Voyager is too small in comparison to the FOVof a particular NIMS observation, it will not be resolved from the background. This results in a registration that is typically accurate to 1 projected NIMS pixel where there are multiple contrasting features that are also resolved (usually impact craters and ejecta). Where there are very few contrasting albedo features or they are too small to be resolved, then the registration is less precise but still no worse than 2 or 3 projected NIMS pixels. 4. Water-Related and CO 2 Absorption Bands 4.1. The 1.5-Mm H 2 O-Related Absorption Band Shape [20] The shape model of the 1.5-mm absorption feature on Ganymede is a modification of the shape model used for

6 2-6 HIBBITTS ET AL.: CO 2 ON GANYMEDE Figure 3. Select averaging of spectra improves the quality of the band depth map, while subsequent bilinear interpolation of the resulting CO 2 band depth map smoothes the pixel-to-pixel boundaries without significantly affecting gross interpretations. (a) CO 2 band depth map derived for the G7KITTU observation. (b) CO 2 band depth map with spectra of select pixels with SNR bd < 5 averaged with their neighbors. Note spuriously deep band depths are largely eliminated. This stage of product is consistently used in quantitative analyses. (c) CO 2 band depth map after a bilinear interpolation is applied, issued only for the visual depiction of band depths. Axes labels designate pixel number. Callisto. Differences between the 1.5-mm absorption band on Callisto and Ganymede required a modification of the Callisto 1.5-mm band shape model in order to match the shape of an average icy 1.5-mm absorption band created from several averages of Ganymede observations. The water-ice bands on Ganymede, which are often not strongly saturated, are generally deeper relative to the continuum than water-ice bands on Callisto [Hibbitts et al., 2000]. As Figure 4a demonstrates, the Ganymede absorption band is broader, has a minimum value at a slightly longer wave-

7 HIBBITTS ET AL.: CO 2 ON GANYMEDE 2-7 Figure 4. (a) The 1.5-mm water-ice band shape model for Callisto (dashed) and Ganymede (solid). These are empirical models based on absorption bands of vastly different strengths and do not suggest the band depths on the moons are similar. (b) The best fit of the Ganymede 1.5-mm water-ice band shape model (solid) to the shape of the water-related absorption in the hydrated non-ice material on Ganymede (dot-dashed). length, and possesses a better defined 1.65-mm absorption band. All of these characteristics are consistent with a greater proportion of fine-grained water ice on Ganymede compared with Callisto [Hansen, 2002]. The wavelength of the minimum reflectance in the model is at 1.50 mm, the expected position of the minimum for an unsaturated waterice absorption band [Grundy and Schmidt, 1998]. The 1.5- mm absorption band of larger-grained ice will saturate, shifting the apparent minimum to shorter wavelengths as seen in spectra of Callisto. [21] A more asymmetric 1.5-mm absorption band, similar in shape to the asymmetric absorption bands commonly found in spectra of Europa, is sometimes present in spectra of Ganymede. It is inferred from the asymmetric bands that hydrated salt-like minerals also exist on Ganymede [McCord et al., 2001]. These distorted bands are found to dominate in only a few low-albedo areas, mainly in dark terrain on or near the trailing hemisphere; consequently, the shape model derived here for the 1.5-mm band is only slightly affected by these asymmetric absorptions which rarely affect the averages used to create the shape model. Where they occur, these distorted bands are sufficiently similar to the shape of the water-ice bands that the shape model closely matches the depth (if not the precise shape) of even the most distorted 1.5-mm water-related absorption band. Consequently, the depth of the hydrate band is measured accurately, even though the model assumes it is actually a water-ice absorption (Figure 4b) Water-Ice Mixing Styles and Grain Sizes [22] Water ice is a major constituent of Ganymede s surface. Therefore a brief discussion of water-ice grain sizes and mixing styles is warranted for later comparison to the distribution of CO 2 over the surface. The ice and nonice material are likely segregated from each other through the gradual sublimation of ice from ice/nonice mixtures and subsequent redeposition of the water vapor onto cold traps, which are bright, cold icy areas [Spencer, 1987; Prockter et al., 1998; Moore et al., 1999]. Where this occurs, the ice and nonice materials are distributed as regions of nearly pure areas of ice or nonice material and are not mixed together grain-by-grain in a specific location. This discrete mixing (in contrast to intimate mixing) occurs at resolutions smaller than NIMS resolves but can be seen in the highest-spatial

8 2-8 HIBBITTS ET AL.: CO 2 ON GANYMEDE Figure 5. The effect of grain size on the reflectance of water ice over the spectral range of NIMS. Shown are three water-ice albedo models compared to an average spectrum from the G1GLOBAL observation of bright, icy pixels mostly from southern high latitudes (light gray) and an average of the least icy regions mostly from Galileo Regio (dark gray). Large-grained ice (500 mm) has a steeper negative slope in the 1 to 2.5-mm range because the water-ice bands are saturating and the 3 to 5-mm range of large-grained ice is nearly black. The fine-grained ice (10 mm) is grossly similar to the icy G1GLOBAL average in the 0.7 to 2.5-mm region but does not match in the longer wavelength region, suggesting the presence of other grain sizes or nonice material. Note that the CO 2 absorption at 4.26 mm is almost undetectable in this average spectrum of brighter, icier regions of Ganymede. resolution SSI images. Panchromatic images of terrain at several meters per pixel resolution show that both regiones and sulci in nonpolar regions generally consist of bright highstanding terrain and darker lower-lying terrain [Prockter et al., 1998; Pappalardo et al., 1998a, 1998b]. Unlike equatorial Ganymede that contains large-grained ice, the polar regions of Ganymede and the bright Osiris Crater contain ice as fine-grained as 50 mm or smaller [Hansen et al., 1998]. The mixing style of the fine-grained ice in these areas may also be more complex than in the equatorial regions. However, over time, fine-grained ice that is intimately mixed with darker material will tend to segregate itself by thermally driven sublimation from the nonice material [Spencer, 1987] and/or will metamorphose at grain-to-grain boundaries into larger grains [Clark et al., 1983]. The poles may contain more variations in ice grain sizes and more ice intimately mixed with nonice materials. Segregation and annealing is slower at polar temperatures and the water-ice molecules, which are continually bombarded by magnetospheric ions, can sputter and then redeposit as frost [Johnson, 1985, 1997; Pappalardo et al., 1998b]. [23] Variations in water-ice grain size and mixing styles induce spectral changes that are similar to those caused by varying the abundance of ice. The defensable assumption of linear mixing for nonpolar regions enables areas of finer and larger grained ices to be discriminated using both water-ice band depth and albedo information. Quantitative grain-size or abundance estimates require more extensive modeling and are beyond the objectives of this paper. The effect of grain size on the spectra of pure water ice can be seen in Figure 5, in which the modeled albedos of several ice grain sizes are compared to the albedos of less-icy and more-icy terrain in the G1GLOBAL observation. The albedos of the ices at 0.7 mm are very similar, but the brightness of large-

9 HIBBITTS ET AL.: CO 2 ON GANYMEDE 2-9 Figure 6. The effect of adding a 20% areal mixture of 5, 10, 20, 50, 100, 200, 500, and 1000-mm grain sizes of water ice on the spectra containing (a) hydrated nonice and (b) anhydrous nonice materials. Representative anhydrous and hydrated nonice endmember spectra (thick black lines) are derived from Callisto [Hibbitts et al., 2001] and Ganymede [McCord et al., 2001] observations, respectively. Dashed lines represent ice grain sizes of 5, 10, and 20 mm. (c) The modeled 1.5-mm band depth as a function of fraction of the continuum, ( o is hydrated nonice mix and x is anhydrous nonice mix). An interpolation to the data is also shown. Note the 0.7-mm albedo of both Ganymede and Callisto terrain decreases by 5.5% when ice grain sizes increases from 50 to 1000 mm, while the 1.5-mm band depth decreases by 40% and 57%, respectively, over the same range. Ice/nonice mixtures of nonpolar Ganymede (and all of Callisto) are described by the right-hand portion of the curve; increasing water-ice grain size results in a shallower 1.5-mm band when ice abundance (approximated by 0.7-mm albedo) is constant. gained ice decreases more than fine-grained ice at longer wavelengths. The NIMS average spectrum of icy terrain appears grossly similar to that of finer grained ice in the 1 to 2.5-mm region. However, the NIMS spectrum of ice-poor terrain is not similar to any ice spectrum because there is a significant contribution of reflected light from the nonice material that reduces the slope of the 1 to 2.5-mm continuum, decreases the depths of the water-ice absorption bands, and brightens the spectrum in the 3 to 5-mm region. [24] Figure 6 shows how the grain size of ice that is mixed with either an anhydrous nonice material (Figure 6a) as found on Callisto [Hibbitts et al., 2001] or a hydrated nonice material (Figure 6b) as found on Ganymede by McCord et al. [2001] affects the depth (as a fraction of the continuum between 1.35 and 1.8 mm) of this water-ice absorption band without greatly affecting the 0.7-mm albedo. Figure 6c shows that given constant water-ice abundance the 1.5-micron band depth decreases as the ice grain diameter decreases from 50 microns because the continuum is constant or increases slightly while the band shallows (dashed lines in Figures 6a and 6b). The band depth also decreases as grain size increases for ice larger than 50 mm because the bottom of the band is fixed by the brightness of the nonice material and the continuum decreases (solid thin lines in Figures 6a and 6b). Consequently, the contrast between the bottom of the band and the continuum is reduced, resulting in a shallower absorption band. Because the ice in nonpolar terrain on Ganymede,

10 2-10 HIBBITTS ET AL.: CO 2 ON GANYMEDE Figure 7. The effect of varying ice abundance from 1 to 50% (areal mixture) on spectra of ice/non-ice mixtures for (a) hydrated nonice material, o, from McCord et al. [2001], and (b) anhydrous nonice material, x, from Hibbitts et al. [2001]. Ice grain size is 200 mm in Figures 7a and 7b. Both the 1.5-mm band depth and the albedo increase with greater ice abundance. (c) Dependence of band depth to albedo ratio on ice abundance for ice and nonice mixtures of 200-mm size ice (solid) and 20-mm size ice (dashed). Band depth increases more quickly than albedo for abundances <15 to 20%. At greater abundances, the albedo increases more rapidly. other than Osiris Crater, is likely larger than 50 mm [Hansen et al., 1998], variations in the depth of the 1.5-mm water ice band for areas of constant ice fraction (constant albedo) can be qualitatively interpreted as variations in water-ice grain size. Thus for similar albedo terrain, a qualitative measure of variation in water-grain size is obtained using the ratio of the depth of the 1.5-mm absorption band to the albedo; a larger band depth to albedo larger ratio (e.g., a deeper band but constant albedo) implies finer-grained ice (Figure 6c). [25] If the abundance of ice varies, then both the 0.7-mm albedo and the 1.5-mm band depth change proportionally (Figure 7). As the effect of grain size variations are shown in Figure 6, the effect of changing ice abundance mixed with both hydrated (Figure 7a) and anhydrous (Figure 7b) nonice materials are shown. Unlike varying grain size, which only appreciably affects the band depth, varying the abundance of ice affects both band depth and albedo (Figure 7c) similarly over a large range of grain sizes. If ice abundance remains below 15%, then the change in band depth is greater than the change in albedo, and the slope of the band depth to albedo ratio is positive for both mixtures of anhydrous and hydrous nonice endmembers for grain sizes from 20 to 200 mm. However, if ice abundance is greater than 15%, then albedo changes begin to dominate the ratio and the curve bends, acquiring a negative slope. Therefore the band depth to albedo ratio can provide information on grain size changes even if ice abundance changes, provided the abundance is known to stay below or above 15%. Ice abundance on Ganymede span this range if traversing from dark to bright terrain [Hansen, 2002], and grain size variations cannot be determined between dark and icy terrain using only band depth and albedo information. However, grain size variations can still be inferred between bright terrains of differing albedos because they very likely contain more than 15% water ice. Therefore we use band depth and albedo information with care to infer changes in

11 HIBBITTS ET AL.: CO 2 ON GANYMEDE 2-11 Figure 8. G1NIPPUR observation. Sketch map based on Voyager and SSI imaging. Epigeus is an ancient palimpsest-like impact scar. A designates a crater that has high albedo in Figure 9. Figure 9. The NIMS G1NIPPUR observation superimposed on a Voyager image mosaic, with the outlines of the palimpsest-like scar Epigeus and the boundary of Nippar Sulcus included, and the bright icy crater in Epigeus marked as A. Red represents higher values or deeper band depths. (a) Albedo (scaled from 0.3 to 0.6); (b) 1.5-mm band depths (scaled from 0.2 to 0.6); (c) ratio of the 1.5-mm band depth to albedo (scaled from 0.5 to 1); (d) CO 2 absorption band depths (scaled from 0 to 0.3).

12 2-12 HIBBITTS ET AL.: CO 2 ON GANYMEDE Figure 10. The shape of the CO 2 band on Ganymede (solid black) is very similar to that on Callisto (dashed) but distinct from crystalline CO 2 at 80 K (solid gray). The crystalline CO 2 absorption band is modeled from optical constants provided by Warren [1986] as derived from laboratory work by Yamada and Person [1964]. We assume only absorption (no Fresnel reflection) because the molecules are likely bound individually or in small groups to a host material [McCord et al., 1998]. The model of the CO 2 absorption on Ganymede is the best fit Gaussian curve to the normalized CO 2 absorption band (secondorder polynomial continuum-removed) for 14 observations. The band shape model for the CO 2 absorption on Callisto similarly derived for those observations [Hibbitts et al., 2000]. Symbols group observations by orbit: orbit 1 ( o ); orbit 2 ( x ); orbit 6 ( < ); orbit 7 ( ^ ); orbit 8 ( * ), orbit 9 ( + ). Derived Ganymede band center at mm ( G ) and Callisto-derived band center at mm ( C ) are marked by dashed lines. The small feature at 4.37 mm, also marked with a dashed line, is consistent in position with C 13 O 2 but may instead be a calibration artifact. water-ice grain sizes for comparison to the distribution of CO 2. [26] Band depth to albedo ratios suggesting smaller grain size than most of equatorial Ganymede have been detected in some terrains that were modified by impacts. In the G1NIPPUR, the scar of an ancient palimpsest-forming impact, Epigeus, disrupts the grooved terrain of Nippur Sulcus (Figure 8). The undisturbed portion of the sulcus (containing grooves) has significant 1.5-mm band depths, while the impact-disrupted portion (containing no grooves) has shallower bands, except around the small impact near its center marked as A in Figures 8 and 9. However, the albedos of portions of both the disturbed and undisturbed portions of the sulcus (except again for the fresh impact) are similar (Figure 9a), as is terrain confined to the regiones both east and west of the sulcus. The small 1.5-mm band depth to albedo ratio of the portion of Epigeus that has similar albedo to the sulcus implies that it contains largergrained ice. The higher albedo of the small fresh crater and its ejecta implies a greater abundance of water, and its lower band depth to albedo is consistent with ice abundances >15% in the sulcus and crater, but grain size variations can not be inferred because the albedo varies. Band depth to albedo ratios also imply finer-grains at latitudes above 30 in both sulcus and regio material. Poleward, Jovian magnetospheric ion bombardment brightens the surface by damaging the surface ice to make it scatter more light and/or by sputtering surface ice of all grain sizes which is then

13 HIBBITTS ET AL.: CO 2 ON GANYMEDE 2-13 Figure 11. The dependence of CO 2 band depth on latitude. Average CO 2 band depths from G1GLOBAL (light gray), E6GLOBAL (gray), and G1MEMPIS (black thin line) have been plotted in latitude bins of 10, 10, and 5, respectively, along with other high-resolution data. Vertical lines represent a variation of 1 standard deviation in band depth. For higher spatial resolution observations (thick black lines) the average CO 2 band depth is plotted over the latitude range of the observation (horizontal line). The thick, gray line is the expected variation in CO 2 band depths if they were to decrease according to the sine of the latitude. redeposited nearby as frost [Johnson, 1985, 1997; Pappalardo et al., 1998b]. In the equatorial and midlatitudes, high-energy ions are largely shielded from the surface by Ganymede s magnetosphere [Cooper et al., 2001]. This may explain the finer-grained ice observed at the higher latitudes in the G1NIPPUR observation (Figure 9c) CO 2 Absorption Band [27] The CO 2 absorption band shape model that was derived for the Callisto observations [Hibbitts et al., 2000] is nearly identical to the shape of the CO 2 absorption band model for Ganymede and was used to calculate the depth of the CO 2 band in the Ganymede observations before the shape of the Ganymede CO 2 band model was derived (Figure 10). The CO 2 absorption band in each observation of Ganymede is consistent with a single band shape that has a minimum reflectance at mm, which is the same as for the CO 2 absorption on Callisto. The shape appears slightly broader than the CO 2 absorption on Callisto and slightly more symmetric, but these differences are within the accuracy of the NIMS spectral resolution and calibration. This CO 2 band shape and the wavelength of minimum absorption are also distinct from that expected by the absorption from crystalline CO 2. The second-order continuum-removed, normalized data of the CO 2 absorption from which the shape of the Ganymede CO 2 model was derived also contains a possible absorption near 4.37 mm, which is weaker but similar in position to an absorption in the Callisto observations [McCord et al., 1998; Hibbitts et al., 2000] but is also present in the spectrum of crystalline CO 2 with a shift in wavelength. Its presence could be due to C 13 O 2, but the shallow depth makes this interpretation equivocal. The CO 2 band depths range from 0 in the G8TRANSI observation average (at high latitude) to 30% of the continuum in the equatorial G7KITTU observation (Table 1) Photometric Effects [28] The incidence and emission angles vary by many tens of degrees over some of the individual observations,

14 2-14 HIBBITTS ET AL.: CO 2 ON GANYMEDE Figure 12. The CO 2 band depths of the same 11 observations as used in Figure 11 show that there is no dependence of the CO 2 band depth on longitude. Each pixel from G1GLOBAL, E6GLOBAL, and G1MEMPIS is plotted. The average band depths from each of the other eight high-resolution observations are plotted, with the horizontal line segments representing the longitudinal range of an observation. The vertical lines represent a variation of 1 standard deviation from the mean. especially in the global observations. Since the amount of light reflected from a surface depends on the illumination geometry, the 0.7-mm reflectance (I/F) in NIMS observations is corrected for photometric effects with a Lommel-Seeliger function in order to derive a term, B o. This term is similar to albedo, independent of photometry, and is used to correlate terrain features between SSI and NIMS observations. where B o ¼ I=F* ðm o þ mþ=m o ; m o ¼ cosðþ; i m ¼ cosðþ; e for incidence angle i and emission angle e. The Lommel- Seeliger correction tends to result in large albedo = B o values, even above 1, for small emission angles. [29] Similar CO 2 band depths in high-resolution observations of similar terrain at similar latitudes but at greatly varying illumination geometry suggests the CO 2 absorption band depth has little photometric dependence. The average depth of the CO 2 absorption band in the overlapping portions of the G1MEMPIS and C9DRKLIT observations are 10.8% ± 3.0% and 8.2% ± 3.1%, respectively. The overlapping portion of the G1MEMPIS observation is on the limb and thus has a very high emission but a low phase angle of 21 (Table 1). The overlapping portion of the C9DRKLIT observation was at a much lower emission angle but similarly low phase angle (Table 1). Despite the difference in illumination geometry the depth of the CO 2 band is constant to within one standard deviation. This is in comparison to the depth of the 1.5-mm water-ice band, whose average changes from 37.5% ± 4.4% for G1MEMPIS to 25.6% ± 3.3% for C9DRKLIT. The very weak photometric dependency of the CO 2 absorption band suggests that the path length of light through the CO 2 in the surface is constant, i.e., that

15 HIBBITTS ET AL.: CO 2 ON GANYMEDE 2-15 Figure 13. The NIMS G2SIPPAR observation of portions of Sippar and Erech Sulci and adjacent Marius Regio (see Figure 1 for location). Sulcus-regio boundaries are indicated in white. (a) Voyager 2 image; (b) NIMS 1.5-mm band depth map; and (c) NIMS CO 2 band depth map (maximum band depth is 0.22). there is little multiple scattering, which would otherwise result in band depth changes with illumination angle. Thus the reflected light containing the CO 2 absorption band was probably not scattered to the surface from depth but penetrated the surface only slightly before being reflected back to space. Thus the CO 2 responsible for the absorption is in the very top surface of Ganymede. A similar but more detailed discussion has been made for the CO 2 discovered in the surface of Callisto. Hibbitts et al. [2002] estimate the abundance there on the order of parts per thousand. Similar band depths in spectra from Ganymede and Callisto suggest similar abundances. More spatially extensive photometric relationships can be investigated by comparing the five overlapping high and moderate phase angle global observations from orbits E4, G7, and C9. However, they are currently not sufficiently free of artifacts to enable even qualitative analysis. 5. Distribution of CO 2 [30] The distribution of CO 2 has been mapped in greatest detail on Ganymede s anti-jovian hemisphere because most of the NIMS observations occur there. The distribution of the CO 2 band depth is generally mottled. There are areas of hundreds of square kilometers, mostly at polar latitudes, for which NIMS spectra contain no CO 2 absorption bands. In the equatorial and midlatitudes, greater abundance of CO 2 is generally associated with the less icy regiones. Unlike on Callisto [Hibbitts et al., 2002], high concentrations of CO 2 associated with impact craters are rare, only observed at Mir crater on Ganymede (section ). There is no leading side/trailing side difference in the depth of the CO 2 band detected nor is there a Jovian/ anti-jovian dependence. However, some complex relationships between CO 2 band depth, albedo, and depth of the 1.5-mm absorption band exist for which we offer some explanations Global Patterns in the CO 2 Distribution Latitudinal Dependence [31] High-resolution NIMS observations of Ganymede encompass terrain from the equator to >60 latitude. The global observations are more comprehensive. Both styles of observation show a latitudinal dependence in the depth of the CO 2 absorption band. Figure 11 presents a scatter plot of CO 2 band depths from 11 observations (three global observations, one high-resolution observation on the trailing hemisphere, and seven high-resolution observations on the leading hemisphere) that span more than ±60 from the equator. The CO 2 band is shallower at latitudes above 40 and may smoothly increase approaching the equator. The trend in CO 2 band depth is consistent among observations of significantly different spatial resolutions, illumination conditions, and terrain types, demonstrating this is a global effect with no strong terrain dependence. The average CO 2 band depths at the equator range from 10 to 21% (Table 1). The least icy spectrum, G7KITTU on the equator of the trailing hemisphere, has the deepest average CO 2 absorption band at 21%. Other high spatial resolution observations near the equator in icier terrain on the leading hemisphere have slightly lower average band depths. Thus the greater CO 2 in the Kittu observation may be due to a greater amount of non-ice

16 2-16 HIBBITTS ET AL.: CO 2 ON GANYMEDE material in the field of view and not a leading/trailing side effect. The approximate limit of ±40 latitude for significantly nonzero CO 2 band depths corresponds to where magnetospheric particles are focused onto the surface by the intrinsic Ganymede magnetic field Longitudinal Dependence [32] There is no evidence that the depth of the CO 2 band on Ganymede is dependent on longitude (Figure 12), unlike on Callisto for which there is an obvious association of relatively deep band depths on its trailing hemisphere, presumably related to the effects of corotating Jovian magnetospheric charged particles [Hibbitts et al., 2000]. The absence of a leading hemisphere/trailing hemisphere pattern in the global distribution of CO 2 implies the Jovian magnetosphere does not effect the lower latitudes of Ganymede as significantly as for Callisto, most likely because Ganymede s internal dipole field largely shields its surface from these effects Correlations Between Less Water Ice and Greater CO 2 Abundance [33] For many NIMS observations at latitudes less than 40, deeper CO 2 bands are generally associated with darker terrain, such as regiones or dark areas within sulci. As shown in Figure 13, a portion of Sippar and Erech Sulci, the generally mottled distribution of the CO 2 band depth is deepest in the northwest corner of this observation. This is regio material that has shallow 1.5-mm band depths and lower reflectance. The brighter and icier sulci, except for dark patches within them, exhibit shallower CO 2 band depths CO 2 Associated With Some Areas Containing Larger Water-Ice Grains [34] In two observations at equatorial and mid-latitudes (G7KITTU and G1NIPPUR, respectively) there appears to be a correlation between deep CO 2 absorption bands and small values of the ratio of the 1.5-mm band depth to albedo. These small ratios imply that the ice in this terrain is larger grained than in the adjacent CO 2 -bearing terrain that have larger ratios but similar albedos. As shown for the G1NIPPUR in Figures 9c and 9d observation, the shallow CO 2 bands and the larger 1.5-mm band depth to albedo ratios are within the undisturbed sulcus material and in the regio material at the eastern side of the observation. These areas likely contain finer-grained ice. This relationship between deep CO 2 bands and largergrained ice is also exhibited in scatterplots of CO 2 band depth versus the 1.5-mm band depth to albedo ratio for the G1NIPPUR and G7KITTU observations (Figures 14 and 14b), for which the depth of the CO 2 band and the ratio of the 1.5-mm band depth to albedo are inversely correlated. The CO 2 bands in the G7KITTU are some of the deepest found on Ganymede, but they are not associated with the crater nor with its dark ejecta but with neighboring bright, preexisting, undisturbed sulcus material (Figure 15). However, this correlation between areas of larger-grained ice and greater abundance of CO 2 in any area of Ganymede does not suggest that the CO 2 is in the larger-grained ice. The brightness of large-grained ice at 4 mm is 1% (Figure 5); therefore absorption band depths due to CO 2 Figure 14. The inverse dependence of the CO 2 band depth on the 1.5-mm: albedo ratio is shown for the (a) G1NIPPUR observation and (b) G7KITTU observation. Two other high spatial resolution observations show this relationship but more weakly. The remaining observations do not show a relationship. The large values of the band depth to albedo ratio in the G7KITTU observation compared to the G1NIPPUR observation are a result of the generally very low albedo of the terrain in the G7KITTU observation. cannot be any greater. Consequently, the CO 2 absorption band is primarily due to CO 2 in the nonice material CO 2 Around Impact Craters and Ejecta [35] In contrast to Callisto where impact craters are commonly enriched in CO 2 relative to surrounding terrain, only one CO 2 -rich impact crater has been observed on Ganymede: Mir crater. In general, CO 2 abundance in impact craters is not different from adjacent terrain. However, darkray craters may be an exception CO 2 -Poor Impact Craters and Ejecta [36] The dark ejecta of Kittu Crater is depleted in CO 2 relative to terrain surrounding it. Kittu Crater is located in the NE portion of the NIMS observation (Figure 15). The

17 HIBBITTS ET AL.: CO 2 ON GANYMEDE 2-17 Figure 15. The NIMS G7KITTU observation on the trailing hemisphere, superimposed on a Voyager/ SSI mosaic. Kittu Crater is in the upper right of the colored portion of each image. The boundaries from the Figure 14 sketch map are plotted. Figure 15a is visible image. (a) Albedo map (scaled from 0.11 to 0.28); (b) 1.5-mm band depth map (scaled from 0.15 to 0.6); (c) 1.5-mm band depth to albedo ratio (scaled from 1.2 to 3.2); and (d) CO 2 band depth map (scaled from 0 to 0.3). Figure 16. Geological sketch map of bright terrain (white), dark terrain, and Kittu crater materials (bright ejecta and dark ray material), based on imaging data from Voyager 1 and Galileo (SSI G7KITTU observation) as shown in Figure 15a.

18 2-18 HIBBITTS ET AL.: CO 2 ON GANYMEDE Figure 17. Spectra of Mir Crater from an average of six NIMS pixels in CO 2 -rich icy impact crater material (black) and from an average of six pixels in surrounding, visibly darker, less CO 2 -rich ejecta (gray). (a) The unscaled spectra from the region of Mir Crater compared to a scaled spectrum of dark ejecta from Kittu Crater (dashed). Vertical bars to the left and right of each spectrum represent a formal error of one standard deviation for the short (0.7 to 2.5 mm) and long (3.2 to 5 mm) portion of each spectrum. Some variation is due to real spectral differences between the averaged pixels, especially in the short-wavelength portion. (b) The CO 2 absorption bands from the same lessicy and darker ejecta and the icy, high albedo CO 2 -rich interior of Mir crater. The continuum of the ice-poor region scaled to the continuum of the icy, CO 2 -rich region. The calculated CO 2 band depth of the ice-poor ejecta is about 1/2 that of the icy crater. terrain surrounding the crater is part of Mysia Sulcus, with some outcrops of Nicholson Regio adjacent and to the south of Kittu (Figure 16). Dark ejecta associated with Kittu extends out several crater radii, reaching above and below the boundaries of the NIMS observation. The reflectance of Kittu s dark ejecta is very low for Ganymede, but the spectrum is similar in shape to other dark material on Ganymede (Figure 17). There is some asymmetry in the 2-mm water-ice absorption band, suggesting the presence of some of the hydrated nonice material identified in other areas by McCord et al. [2001]. The dark ejecta to the northeast of Kittu have shallow CO 2 absorption bands relative to the sulcus material (Figure 15d) as well as the greatest measured values of the 1.5-mm band depth to albedo ratios found on Ganymede. In Figure 14b, pixels indicating shallow CO 2 band depths (15%) correspond to the dark, CO 2 -poor ejecta northwest of Kittu. The CO 2 -poor bright crater also has shallower CO 2 bands than the surrounding sulcus material, and where the albedo of the crater is similar to the sulcus, the ice in the crater is finer-grained. The deep CO 2 bands in the southwest portion of the observation (Figure 15d) correlate to small band depth to albedo ratios (Figure 15c and Figure 14b), implying an association with areas of larger-grained ice; consistent with associations inferred from the G1NIPPUR observation CO 2 -Rich Impact Crater [37] Of the 11 observations that have spatial resolutions better than 10 km/pixel and can therefore resolve details in the CO 2 distribution, only Mir crater in the G1MIRRAY observation is CO 2 -rich (Figure 18). This observation, located just south of the equator on the well-imaged anti- Jovian portion of the trailing hemisphere, contains sulcus and regio material. The crater is not located in CO 2 -rich terrain, although it is at the eastern boundary of Nicholson Regio, which has a greater amount of CO 2 than the other regiones. This aspect of Mir s location may explain the slightly higher levels of CO 2 found in the western portion of the observation in an area that has fine-grained ice relative to the ejecta and sulcus (Figures 18c and 18d). The deep CO 2 band depths of Mir are associated with the bright, icy crater which also contains similarly fine-grained ice as in the north-south trending sulcus and southeast portion of the observation that are not CO 2 -rich (Figures 18a and 18c). The CO 2 -rich terrain in the midwestern edge ( A ) is associated with lower values of the water-ice band depth to albedo ratios and therefore larger-grained ice than in similar albedo terrain. The lower albedo ejecta has a lower abundance of this large-grained ice but not obviously high or low concentrations of CO 2. The average spectrum of the six pixels with deep CO 2 bands from the icy center of Mir is shown in Figure 17 (solid black line). Also shown is an average of six pixels from less-icy ejecta adjacent to the crater that have 50% shallower CO 2 bands (gray line). [38] In contrast to Ganymede, many bright impact craters on Callisto show enhanced concentrations of CO 2 relative to surrounding terrain [Hibbitts et al., 2002]. If the source of the CO 2 in impact craters is endogenic, then this distinctive difference between the satellites may indicate that Ganymede s upper crust is more depleted in CO 2 than Callisto s. Differences in crustal CO 2 content could result from differences in initial satellite composition and/or from early loss of Ganymede s CO 2 by enhanced thermal processing relative to Callisto Host Material for the CO 2 [39] Several lines of evidence support the conclusion that the CO 2 detected by NIMS on the surface of Ganymede is in the nonice material. First, the ice in most of the equatorial region on Ganymede is sufficiently large-grained

19 HIBBITTS ET AL.: CO 2 ON GANYMEDE 2-19 Figure 18. The NIMS G1MIRRAY observation superimposed on Voyager 1 image. (a) Albedo map (scaled from 0.3 to 0.5); (b) 1.5-mm band depth map (scaled from 0.15 to 0.6); (c) 1.5-mm band depth to albedo map (scaled from 0.45 to 0.82); and (d) CO 2 band depth map (scaled from 0 to 0.3). Note the relatively large-grained, bright, CO 2 -poor sulcus running vertically through the center of the observation. The icy crater is relatively fine-grained but CO 2 -rich, a relationship that only occurs in this observation. and of low reflectance at 4.26 mm that CO 2 cannot be detected (Figure 5). However, the inability to detect CO 2 in large-grained ice does not preclude its existence in the ice. Second, in general, the equatorial and midlatitudes contain the deepest CO 2 bands, which are associated with less icy regio material. Third, in several observations, CO 2 abundance is greater in ice-poor areas where the ice is larger-grained ice than in adjacent ice-rich terrain that has finer-grained ice. Fine-grained water ice has greater reflection at 4 mm, and CO 2 can be more easily detected in it than in large-grained ice. However, finer-grained ice at latitudes above 30 to 40 does not contain sufficient CO 2 to induce a 4.26-mm absorption band in NIMS spectra (Figures 1 and 11). The lack of a detectable CO 2 band in many areas at these high latitudes also supports the suggestion by McCord et al. [1998] that this fine-grained, CO 2 -poor ice may physically cover and spectrally mask the presence of CO 2. These facts are consistent with the CO 2 detected by NIMS residing in the nonice material on the surface of Ganymede. [40] The host material for the CO 2 may not be of uniform composition. The CO 2 absorption band in spectra of dark ejecta from Kittu contains an asymmetric 2.0-mm water-ice absorption band suggesting that a hydrated nonice material is one host of CO 2 in this dark ejecta. The spectrum of the hydrated nonice material derived by McCord et al. [2001] also contains the CO 2 absorption band that is very similar in shape and position to that in other nonice material (Figure 19). A single shape model appears to be successful in describing the CO 2 absorption band within spectra from all terrains on the surface of Ganymede and in potentially different host materials on Ganymede. Thus the physical state of the CO 2 appears to be similar everywhere even though the nonice host material may not be. Finally, because the CO 2 absorption band shape on Ganymede is similar to that on Callisto, the nature of the bond is also likely similar for the CO 2 trapped in the surfaces of both moons. The similar maximum depths of the CO 2 band on Ganymede with respect to Callisto [Hibbitts et al., 2002] therefore suggest that the CO 2 abundance on the surface of Ganymede does not exceed parts per thousand. 6. Conclusions [41] The distribution of CO 2 on the surface of Ganymede appears to be controlled by geological processes, including impacts, and a combination of effects from the Jovian and Ganymedian magnetospheres. Characteristics of the CO 2 distribution include: (1) deepest CO 2 bands are found in darker, less icy regio materials; (2) CO 2 is not detected in some icy areas, particularly in the polar regions; (3) only one CO 2 -bearing impact crater has been identified over the surface by NIMS in distinct contrast to Callisto, suggesting that Ganymede s upper crust is depleted in CO 2 relative to Callisto s; (4) dark crater ejecta