Radar-bright features near Mercury's poles have been postulated to be deposits of water ice trapped in cold, permanently shadowed interiors of impact craters. From its orbit about Mercury, MESSENGER repeatedly imaged the planet's south polar region over one Mercury solar day, providing a complete view of the terrain near the south pole and enabling the identification of areas of permanent shadow larger in horizontal extent than approximately 4 km. In Mercury's south polar region, all radar-bright features correspond to areas of permanent shadow. Application of previous thermal models suggests that the radar-bright deposits in Mercury's south polar cold traps are in locations consistent with a composition dominated by water ice provided that some manner of insulation, such as a thin layer of regolith, covers many of the deposits.
 Two decades ago, radar-bright material surrounding Mercury's north pole was discovered in radar images obtained with the 70-m antenna in Goldstone, California, and the Very Large Array [Slade et al., 1992; Butler et al., 1993]. The bright north polar feature was quickly confirmed with observations made from the Arecibo Observatory, and radar-bright material was also discovered in Mercury's south polar region [Harmon and Slade, 1992]. The high reflectivity and circular polarization ratio of the observed radar-bright features are distinguishing traits also seen on the icy Galilean satellites of Jupiter and at the Martian south polar ice cap, and they were interpreted to be evidence of ice in Mercury's polar regions. Thermal modeling calculations indicated that water ice could be stable within permanently shadowed craters at Mercury's poles for billions of years [Paige et al., 1992].
 Subsequent Earth-based radar observations greatly improved the spatial resolution of the radar images, and many individual radar-bright spots were identified as lying within specific impact craters [Harmon et al., 1994, 2001, 2011; Slade et al., 2001; Harcke, 2005; Harmon, 2007]. However, many radar-bright features could not be mapped to craters or other geologic features because spacecraft images of Mercury's poles were limited to those obtained during two encounters by Mariner 10 in 1974–75, which provided coverage of slightly less than half of Mercury's polar regions [Davies et al., 1978]. The three flybys of Mercury by the MErcury Surface, Space ENvironment, GEochemistry, and Ranging (MESSENGER) spacecraft in 2008–09 were along near-equatorial trajectories and did not provide views of the poles [Solomon et al., 2007].
 On 18 March 2011, MESSENGER became the first spacecraft to orbit the planet Mercury, and during the probe's first solar day in orbit, MESSENGER's Mercury Dual Imaging System (MDIS) [Hawkins et al., 2007] periodically obtained images of Mercury's south polar region with its wide-angle camera (WAC). From those images we have mapped the terrain and performed analysis to identify areas that are permanently shadowed or regularly sunlit. In this paper, we compare the results of this south polar imaging campaign with the distribution of radar-bright deposits identified in radar images, and we discuss the implications of that comparative study for the hypothesis that Mercury's polar deposits consist predominantly of water ice.
 From 6 April to 27 September 2011, MDIS viewed Mercury's south polar region over one full Mercury solar day (176 Earth days). WAC images of the region were obtained from approximately every fourth spacecraft orbit, corresponding to one image every other Earth day (Figure 1a). The probe's highly eccentric, 12-h, near-polar orbit with apoapsis at high southern latitudes [Solomon et al., 2007] permits the field of view of each WAC image to cover the entirety of the sunlit portion of the planet from the south pole to a latitude of 73°S at a spatial sampling of ∼1.7 km/pixel; many, but not all, images also extended to regions northward of 73°S. As seen in the WAC mosaic in Figure 1b, the largest impact crater near the south pole is Chao Meng-Fu, with a diameter of ∼180 km (see Animation S1 in theauxiliary material).
 The repeated imaging of the south polar region has enabled a map of solar illumination (Figure 2) to be generated by the same methodology as that which has been applied to the lunar poles [Bussey et al., 1999, 2005; Speyerer and Robinson, 2011]. To produce the illumination map, a threshold was applied to create a binary image that separated the surface into sunlit and shadowed areas, and the binary images were averaged to produce a map of the percentage of time that a given area of surface was illuminated during one Mercury solar day. A preferred threshold value was chosen by inspection of the pixel-value histogram. Selection of lower and higher threshold values, equal to approximately ±40% of the preferred value, resulted in only minor changes to the total area calculated to be in permanent shadow and no changes to the general shadow locations. Results presented in the next section are for the preferred threshold value, and the uncertainty in shadowed area is estimated from the results with the low and high threshold values.
3. Illumination Conditions
 Mercury's obliquity is only 2.11 arc minutes (0.035°) [Margot et al., 2007], so some craters and other topographic depressions near the poles do not receive any direct sunlight. The equatorial regions at longitudes 90°E and 270°E are Mercury's “cold poles,” where local noon occurs at aphelion (Figure 1a) because of Mercury's 3:2 spin-orbit resonance, and insolation is substantially less than at the corresponding “hot poles” because of Mercury's eccentric orbit. These longitudes are also illuminated for a lesser percentage of time than other longitudes, as shown inFigure 2.
 No part of Mercury's south polar region is illuminated for 100% of the time at the ∼1.7 km/pixel spatial sampling of the map shown in Figure 2, which resolves regions greater than ∼4 km in horizontal extent. The highest illumination percentage of 94 ± 4% (sunlit in 80 to 87 of 89 total images, depending on the threshold value) is found along part of the southern rim of Chao Meng-Fu crater. Overall, the area of Mercury's south polar region that is illuminated >80% of the time is small at this resolution, ∼60 km2 with the low threshold value and ∼20 km2 with the preferred threshold. For comparison, the total area illuminated >80% time at the lunar poles is estimated to be ∼1–5 km2 [Noda et al., 2008; Mazarico et al., 2011]. Although our results indicate that large areas of permanent illumination are not present at Mercury's south pole, the 1.7 km/pixel spatial sampling of the dataset limits further conclusions on this topic. Recent analysis of images of the Moon taken with the Lunar Reconnaissance Orbiter Camera (LROC) illustrates the need for higher-resolution datasets when mapping illumination conditions of small areas [Speyerer and Robinson, 2011].
 In contrast, large areas near Mercury's south pole are found to remain in shadow over a full Mercury solar day. The total area estimated to be in permanent shadow south of 73°S is 43,000 ± 14,000 km2 (2.6 ± 0.8% of the region), south of 80°S this area is 40,000 ± 11,000 km2 (7 ± 2% of the region), and south of 85°S this area is 28,000 ± 7,000 km2 (20 ± 5% of the region). In comparison, the fractional area in permanent shadow at the lunar poles as summarized by Mazarico et al.  is 4.5% north of 80°N, 5.6% south of 80°S, 6.6% north of 85°N, and 9.7% south of 85°S. Bussey et al.  reported that 10.9% of the lunar surface is in permanent shadow south of 86°S. The slightly larger percentage of Mercury's south polar region in permanent shadow is consistent with Mercury's smaller spin axis tilt (0.035° versus the 1.54° angle that the Moon's spin axis currently makes with the ecliptic [Siegler et al., 2011]), which produces conditions more favorable for permanent shadow. In addition, the presence of the 180-km-diameter crater Chao Meng-Fu near the south pole contributes to the large total area of permanent shadow.
4. Comparison with Radar Data
 We have compared the mosaic in Figure 1b and the illumination map in Figure 2with Earth-based radar images of Mercury's south polar region obtained at both Arecibo Observatory [Harmon, 2007; Harmon et al., 2011] and Goldstone, California [Slade et al., 2001; Harcke, 2005]. All of the radar-bright features in Mercury's south polar region that appear in either the Arecibo or Goldstone radar images correspond to areas of permanent shadow in the MDIS illumination map.
 A detailed comparison of the highest-resolution Earth-based radar image of Mercury's south polar region with the illumination map is shown inFigure 3. The Arecibo image was obtained at S-band (12.6-cm wavelength), is a weighted sum of same- and opposite-sense circular polarization images obtained on 24–25 March 2005, and has a range resolution of 1.5 km [Harmon et al., 2011]. The floor of Chao Meng-Fu crater is heavily shadowed and hosts a large area of high radar backscatter, as noted since the first south polar radar data were acquired [Harmon and Slade, 1992]. Considerably smaller radar-bright features also collocate with smaller areas of permanent shadow; this agreement notably includes two low-latitude features just north of 75°S, at 79°E and 270°E. To quantify the comparison further, an image consisting only of areas of permanent shadow was compared with a median-filtered radar image. Over 85% of radar-bright features collocate with permanently shadowed areas, and all radar-bright features are within a few kilometers of an area of permanent shadow. (More than half of the radar-bright pixels that do not match precisely with areas classified as permanently shadowed are located within Chao Meng-Fu, showing the limitations of mapping the small-scale structures on this crater's floor with the 1.7 km/pixel spatial sampling of this dataset.) Such offsets are within the uncertainties in the co-registration of the two datasets and the maximum error of approximately 2 WAC pixels, or ∼3 kilometers, in pixel registration experienced during uncontrolled mapping of MDIS images near Mercury's south pole, as determined by comparing uncontrolled and controlled MDIS mosaics [Becker et al., 2012].
 Goldstone X-band (3.5-cm wavelength) radar observations have also been acquired of Mercury's south polar region with a range resolution of 6 km [Harcke, 2005]. Though the Goldstone data are lower in resolution than the Arecibo images, they were acquired over a span of 12 nights with the sub-Earth longitude varying from 122°E to 201°E, and they cover a larger combined area of the south polar region with a view from nearly the opposite direction to that for the Arecibo radar image. From the Goldstone data,Harcke confirmed locations of four radar-bright features previously observed byHarmon et al. and listed coordinates of 35 newly identified radar-bright features in Mercury's south polar region. All of the 35 coordinate pairs correspond to areas of permanent shadow on the MDIS illumination map (Figure 4a).
5. Implications for Radar-Bright Materials
 Though all radar-bright features in Mercury's south polar region map to areas of permanent shadow, not all permanently shadowed areas appear bright in Earth-based radar images. South of 73°S, we have identified 268 craters with diameters greater than 10 km that host some areas of permanent shadow as determined from the 1.7 km/pixel illumination map (Figure 4a). Mapping was limited to craters, though ridges and other topographic features are also sources of permanent shadow near Mercury's south pole. Mapping is incomplete south of ∼85°S, where the large fractional area in shadow and limited resolution make it difficult to discern confidently the source of shadows.
 Early thermal models of polar craters on Mercury showed that the temperatures experienced within permanently shadowed areas depend strongly on the crater's size and location [Paige et al., 1992; Ingersoll et al., 1992; Salvail and Fanale, 1994]. More detailed models by Vasavada et al.  yielded estimates for the maximum and average temperatures within shadowed craters of diameter 100, 40, and 10 km on Mercury. For craters 100 km in diameter and larger, the maximum temperature experienced by the coldest point within a permanently shadowed crater is <100 K for all latitudes poleward of ±80° [Vasavada et al., 1999]. Moreover, the timescale to evaporate 1 m of water ice at a temperature of approximately 110 K is ∼1 Gy [Vasavada et al., 1999], so water-ice deposits in these craters would be stable at the surface on geologic timescales.
 In contrast to the situation for the largest craters, Vasavada et al. concluded that the coldest surface locations in permanent shadow within craters 40 km in diameter or smaller exceeds 110 K at latitudes equatorward of about ±82°. Many of the craters that host radar-bright deposits in Mercury's south polar region are located north of 82°S and have diameters <40 km (Figure 4b), a result therefore inconsistent with long-lived surface water-ice deposits. However, if water-ice deposits were covered by a thin (few decimeters thick) regolith layer, then the peak temperature experienced by those deposits would be approximately equal to the average, rather than the maximum, surface temperature [Vasavada et al., 1999], and the migration and surface loss of the water ice also would be affected [Schorghofer and Taylor, 2007]. For craters 40 km in diameter, Vasavada et al. calculated that the average temperature for the coldest point in permanent shadow would be <100 K at all latitudes investigated (poleward of ±79°). Thus, for many of the craters hosting radar-bright deposits plotted inFigure 4b, some insulating layer, such as a thin regolith cover, is required for the deposits to be long-lived water ice. A thin regolith cover has also been suggested to explain small differences in radar reflectivity and radar scattering parameters derived from the Arecibo 12.6-cm wavelength S-band and Goldstone 3.5-cm wavelength X-band data [Butler et al., 1993; Slade et al., 2004; Harcke, 2005; Harmon et al., 2011].
 Interior temperatures vary strongly with crater morphology. The interiors of simple bowl-shaped craters experience considerably warmer temperatures than flatter-floored complex craters, because of increased indirect heating from scattering off crater walls [Vasavada et al., 1999]. On Mercury, the transition with increasing crater size from simple bowl-shaped craters to flat-floored complex craters occurs at a crater diameter of ∼10 km [Pike, 1988]. Thermal models of Vasavada et al. indicate that for bowl-shaped craters 10 km in diameter, the maximum surface temperature for the coldest shadowed surface location in such craters is >110 K at any latitude, and the average surface temperature is <110 K only when such a crater is located within 2° of the pole. This result suggests that radar-bright deposits within any craters smaller than 10 km in diameter north of 88°S cannot consist of long-lived water ice, even ice beneath a thin insulating layer of regolith.
 Because of the resolution of the MDIS illumination map, we chose to limit the mapping in Figure 4to craters ≥10 km in diameter, but small craters containing radar-bright materials can be identified inFigure 3. For example, the crater chain located at ∼82°S and extending from ∼230° to 240°E is composed of craters <10 km in diameter that provide areas of permanent shadow in which radar-bright materials are located.Harmon et al. noted a number of radar-bright features near Mercury's north pole that collocate with craters <10 km in diameter and commented that this result is potentially at variance with the water-ice hypothesis. However, as those workers also noted, if these small craters are shallower or rougher than the idealized bowl shape modeled byVasavada et al. , the thermal environment inside the crater could be more favorable to retention of water ice. As seen in Figure 4b, only one crater (centered at 82.4°S, 260°E) within the diameter range 10–14 km contains an identified radar-bright deposit. That the majority of small craters do not show radar-bright features is generally consistent with thermal modeling predictions for water ice, but the small craters that are the exceptions are worthy of further investigation on a case-by-case basis, using topography measured by MESSENGER rather than idealized crater shapes.
 In addition to crater size and latitude, another variable in the thermal conditions of permanently shadowed areas is longitude, because of Mercury's 3:2 spin-orbit resonance and eccentric orbit. Though the crater thermal model results ofVasavada et al.  for maximum and average temperatures experienced within craters hosting areas of permanent shadow were not reported for different longitude ranges, it has been noted by Harmon et al. [2001, 2011]that radar-bright features at lower latitudes tend to occur along the cold-pole longitudes.Figure 4aalso shows that craters with radar-bright deposits are more prevalent along cold-pole longitudes, a result that may be due to a more favorable thermal environment. However, the viewing geometry of the radar observations also has the potential to contribute to that prevalence, either because a crater was on the far side of the radar horizon or because it was at a longitude near the radar illumination and viewing direction, where the northernmost walls and floors (locations most favorable to permanent shadowing in the south polar region) could be in radar shadow. Arecibo observations were acquired from a sub-Earth longitude of 11°E whereas Goldstone observations were collected from 122°E to 201°E, providing complementary viewing geometries to cover the entire south polar region but with conditions such that one hot-pole longitude often had higher amounts of radar shadow and the other hot-pole longitude was near the radar horizon.
 Are all of Mercury's cold traps occupied by polar deposits? Poleward of 80°S and among craters >30 km in diameter, there are 46 craters that contain some areas of permanent shadow, 34 of which host radar-bright features. Examining the 12 craters that lack radar-bright features onFigure 4a, it can be seen that some contain only a small amount of permanent shadow, others have overlapping craters that disrupt the original crater form, and the largest ones are located near the 0°E hot pole. Overall, the MDIS illumination map does not show large areas in permanent shadow that lack radar-bright features except along the 0°E hot pole, in contrast to the 180°E hot pole. Conversely, the large majority of areas found to contain permanent shadow in regions thermally favorable for water ice deposits with a thin regolith cover do contain radar-bright features.
 Elemental sulfur concentrated in cold traps near Mercury's poles has been proposed as an alternative composition for the radar-bright polar deposits [Sprague et al., 1995]. A 1 Gy timescale for evaporation of 1 m of elemental sulfur corresponds to a surface temperature of ∼220 K [Vasavada et al., 1999], considerably higher than the 110 K temperature for water ice. In the models of Vasavada et al. , areas in permanent shadow within larger flatter-floored craters experienced maximum temperatures <220 K for all latitudes examined (poleward of ±79°), and bowl-shaped craters 10 km in diameter had average temperatures <220 K for latitudes poleward of ±70°. Thus, radar-bright features associated with small craters do not pose the same thermal problem for sulfur that they do for water ice. However, if sulfur were responsible for the radar-bright features on Mercury, sulfur's stability to higher temperatures is predicted to lead to a sulfur polar cap with a width of approximately 1° in latitude, and no such feature has been seen in radar images [Butler, 1997; Vasavada et al., 1999].
Starukhina suggested that Mercury's high-radar-backscatter signals originate from altered dielectric properties of silicates at very low temperatures rather than from cold-trapped volatiles such as water ice or sulfur. Radar-bright deposits at Mercury's south pole are collocated with cold areas of permanent shadow, but this observation alone cannot distinguish between the cold silicate hypothesis ofStarukhina and the cold-trapped volatiles proposals.
 Imaging of the south polar region of Mercury by MESSENGER's wide-angle camera over one Mercury solar day indicates that all radar-bright features at high southern latitudes are located in areas of permanent shadow. But are the radar-bright materials water ice? From the results in this paper, the cold traps located near Mercury's south pole appear to be occupied in a manner consistent with thermal models of the stability of water ice, as long as many of the deposits also are insulated in some manner, such as by a thin covering layer of regolith. Ongoing measurements by the MESSENGER spacecraft of Mercury's north polar region, in particular laser altimetry and neutron spectrometry in addition to high-resolution imaging, promise to provide further insight into the nature of Mercury's radar-bright polar deposits and will enable comparative studies with the Moon.
 The MESSENGER project is supported by the NASA Discovery Program under contracts NAS5-97271 to The Johns Hopkins University Applied Physics Laboratory and NASW-00002 to the Carnegie Institution of Washington. Support from a MESSENGER Participating Scientist grant to D. T. B. is also appreciated. We thank Paul Spudis and an anonymous reviewer for valuable comments.
 The Editor thanks Paul Spudis and an anonymous reviewer for assisting with the evaluation of this paper.