Age constraints of Mercury’s polar deposits suggest recent delivery of ice

Abstract

Surface ice at the poles of Mercury appears as several-m-thick deposits that are composed of nearly pure water. We provide new age estimates of Mercury’s polar deposits from combined analyses of Poisson statistics and direct observations of crater densities within permanently shadowed, radar-bright regions imaged by the MESSENGER spacecraft. These age estimates suggest that ice was delivered to Mercury within the last ~150 Myr. A single, recent impactor is one possible delivery mechanism that is consistent with our new age constraints, as well as the observed distinct reflectance boundaries of the polar deposits and the relative purity of the ice, as suggested by the Earth-based radar observations. In contrast to ice on Mercury, observations of the lunar poles are suggestive of a highly patchy distribution of surface frost. The patchiness of lunar polar deposits is consistent with long exposure times to the space weathering environment. Given enough time, the polar deposits on Mercury may age into a more heterogeneous spatial distribution, similar to that on the Moon.

1. Introduction

The polar terrains of Mercury are characterized by permanently shadowed regions (PSRs) that are thermally stable environments for water ice on geologic timescales (Vasavada et al., 1999; Paige et al., 2013). Observations from both Earth-based radar and the MErcury Surface, Space ENvironment, GEochemisty, and Ranging (MESSENGER) spacecraft have demonstrated that water-ice deposits occupy such PSRs.

For example, radar observations of Mercury’s poles are suggestive of ice deposits that are at least several meters thick (Harmon, 2007; Black et al., 2010), and modeling of the radar data suggests that these deposits are composed of ~95% pure water ice (Butler, 1997). Neutron data acquired by the MESSENGER spacecraft of the north polar region of Mercury is also consistent with a near pure water-ice composition (Lawrence et al., 2013). While some ice deposits closest to the pole are exposed at the surface (Neumann et al., 2013; Chabot et al., 2014; Deutsch et al., 2017), the majority of ice deposits on Mercury are insulated by 10–30 cm (Lawrence et al., 2013) of low-reflectance materials that are interpreted to be composed of volatiles other than water (Paige et al., 2013). Finally, images (Chabot et al., 2014; 2016) and reflectance measurements (Neumann et al., 2013; Deutsch et al., 2017) of Mercury’s PSRs reveal spatially coherent ice deposits with distinct albedos and sharp reflectance boundaries that align with boundaries of permanent shadow.

Understanding the ages of Mercury’s polar deposits is important when considering the possible source(s) and evolution of the ice. An early pre-MESSENGER analysis used regolith gardening models to constrain the age of Mercury’s polar ice (Crider and Killen, 2005). Because Earth-based radar observations suggest that the radar-bright ice deposits on Mercury’s surface have <5% volume fraction of silicates (Butler, 1997), it is likely that the ice deposits were emplaced relatively recently in order to maintain this high degree of purity through time (Butler, 1997; Crider and Killen, 2005). The regolith gardening models suggest that the ice deposits are <50 Myr in age, given that 20 cm of regolith is expected to cover the deposits in this timeframe, which is not observed (Crider and Killen, 2005).

Data acquired by the MErcury Surface, Space ENvironment, GEochemistry, and Ranging (MESSENGER) spacecraft support the hypothesis that the ice on Mercury may be relatively young. For example, images of the north polar region show that the ice deposits are not buried by regolith (Chabot et al., 2014; 2016). Images (Chabot et al., 2014; 2016) and reflectance measurements (Neumann et al., 2013) reveal sharp reflectance boundaries delineating coherent deposits, suggesting that volatiles were delivered relatively recently or are actively restored.

Here we estimate the ages of specific ice deposits on Mercury using crater counting methods and Poisson statistical analyses and discuss possible sources for ice on Mercury. We conclude by discussing the differences between ice at the poles of Mercury and the Moon, and suggest that differences in ages of the ice may be an important factor contributing to the stark differences in surface characteristics between ice on Mercury and the Moon.

3. Methods

3.1. Age estimates for polar deposits on Mercury

During MESSENGER’s low-altitude campaign, high-resolution images (pixel resolution ≤100 m) were acquired of 31 host craters at the north polar region of Mercury (Chabot et al., 2016). These broadband Wide-Angle Camera (WAC) images use sunlight scattered from nearby illuminated peaks to image within PSRs. The permanently shadowed floors where radar-bright ice deposits are located can be clearly resolved in these images for 28 out of 33 craters, and sometimes reveal central crater structures or small impact craters. (Images of the remaining craters are unclear due to unfavorable viewing geometries.) While the majority of small craters observed in the PSRs may have formed before the deposition of the ice (Chabot et al., 2016; Deutsch et al., 2018), some anomalous small impact craters are associated with high-reflectance rings, suggestive of excavated material (Fig. 2). If these anomalous craters associated with high-reflectance material are superposing the ice, then they can be used to estimate the ages of the ice surfaces.

Fig. 2. Brightness variations associated with impact craters suggest these impact craters superpose the ice deposits on Mercury.

Low-altitude WAC broadband images reveal the interiors of (a) Ensor crater (25 km diameter; 82.3° N, 342.5°E), (b) Laxness crater (26 km diameter; 83.3° N, 310.0°E), and (c) Bechet crater (18 km diameter; 83.1° N, 266.3°E). Yellow arrows point to small craters (100–300 m in diameter) in the PSRs associated with high-reflectance material

We analyze all high-resolution MDIS images of the 33 craters at the north polar region of Mercury that were acquired during MESSENGER’s low-altitude campaign (Chabot et al., 2016). The PSRs of each crater (Deutsch et al., 2016) are visually inspected for the presence of small craters that are associated with high-reflectance rings suggestive of excavated material. Larger-scale brightness variations in polar craters correlate with variations in modeled maximum surface temperature, suggesting that multiple volatile species are contained within the surficial lag deposits that insulate the majority of polar deposits on Mercury (Chabot et al., 2016). The brightness variations associated with the small craters selected for this study are smaller in size than the brightness variations that are suggestive of different volatiles, and they are also not correlated with any predicted temperature variations (Chabot et al., 2016). However, it is possible that the thermal models (1 km pixel resolution) cannot predict variations at the spatial scale of these high-reflectance rings associated with small craters (between 100 m and 300 m in diameter).

If small craters associated with high-reflectance rings are identified in ≥3 images (acquired under different viewing geometries), then we interpret these small craters as superposing the ice deposit, and we interpret the the high-reflectance rings as crater ejecta. Thus, even if high-reflectance rings associated with small craters appear to be present in 2 MDIS images, if the remaining images are unclear due to poor lighting conditions or non-ideal viewing geometries, then we do not use the host craters in our analysis. For the majority of host craters analyzed (27 of 33 craters), the available images do not adequately resolve the permanently shadowed floors such that the presence or absence of high-reflectance rings can be confidently identified in association with small craters. If small craters associated with high-reflectance rings are not identified in any images (and at least ≥3 images) that clearly resolve the host crater floor, then we interpret the host crater as having no small craters that superpose the ice deposit.

We identify 3 north polar craters that host small impact craters associated with high-reflectance ejecta rings (Fig. 2; Table 1). For each of these 3 craters, count areas for crater statistics are measured for the regions of the crater floors that could be resolved in ≥3 images. All small impact craters >100 m that are associated with high-reflectance rings are identified.

Table 1:

Small craters identified as superposing ice deposits at the north polar region of Mercury using MDIS images.

Host crater	Ensor	Laxness	Bechet
Diameter (km)	25	26	18
Location	82.3°N, 342.5°E	83.3°N, 310.0°E	83.1°N, 266.3°E
Absolute model ages (Ma) derived from crater counting statistics	210 ± 60* 23 ± 6** 8.4 ± 2***	29 ± 10* 5.1 ± 5** 1.5 ± 0.5***	47 ± 20* 7.9 ± 3** 2.7 ± 1***
Estimated likely ages (Ma) calculated from Poisson statistics	137 ± 15	100 ± 18	81 ± 19
Images analyzed (pixel resolution)	EW1051458815B (37.6 m) EW1067153761B (64.5 m) EW1046946306B (90.3 m)	EW1052529039B (46.1 m) EW1052586891B (48.0 m) EW1067392056B (55.2) EW1067451645B (57.4 m)	EW1068017656B (55.0 m) EW1068077244B (58.0 m) EW1053138269B (71.5 m) EW1052528979B (32.8 m) EW1068107038B (59.2 m) EW1068136833B (60.6 m)
Count area (km2)	109	79	86
Total craters superposing ice	17	8	6
Diameters of superposed craters (km)	0.15 0.15 0.17 0.19 0.19 0.19 0.20 0.20 0.20 0.20 0.20 0.22 0.22 0.22 0.24 0.27 0.27	0.11 0.14 0.14 0.15 0.16 0.22 0.23 0.28	0.17 0.21 0.22 0.23 0.24 0.29

^*Crater diameters used in crater counting statistics are the diameters measured from images.

^**Crater diameters are scaled by ½ to represent possible sizes of craters impacting into pure ice (Croft et al., 1979; Koschny and Grün, 2001).

^***Crater diameters are scaled by 1/3 to represent the minimum possible sizes of craters impacting into pure ice (Croft et al., 1979; Koschny and Grün, 2001).

Using CraterStatsII (Michael and Neukum, 2010), crater size-frequency distributions (CSFDs) are derived for the ice surfaces by fitting models of the crater-derived retention ages to the chronology and production curves for Mercury (Le Feuvre and Wieczorek, 2011). We estimate ages for each ice deposit from the CraterStatsII output age model.

We also describe the relatively likelihoods of possible ages of the ice surfaces using Poisson statistics, which do not require data binning or curve fitting techniques (Michael et al., 2016). These age estimates are not dependent on the presence of craters because the Poisson statistics determine the time-resolved probability of a given observation within the chronology model (Michael et al., 2016). With this technique, uncertainties stem from the predictions of the chronology model. The likelihood function for the set of observed craters, $\mathbb{D}$ as divided into n bins for any given time, t, is expressed by Michael et al. (2016) in their Eq. 8:

$$\mathrm{pr}(\mathbb{D},t)\propto \exp (-A{[C(d,t)]}_{d_{\max }}^{d_{\min }}{\{\text{C}(\text{d}=1,\text{t})\}}^{n_{\mathbb{D}}}$$ (Eq. 1)

where A is the crater accumulation area, C(d,t)is the cumulative form of the production function, and d is crater diameter. Errors are reported as 1-σ of the median to include 68% of the function describing the likelihood that the surface has a particular age (Michael et al., 2016).

3.2. Statistical and observational biases affecting crater counts of ice deposits on Mercury

A variety of statistical and observational biases can affect the estimated absolute model ages (AMAs) of the ice deposits. For example, low image resolution and non-ideal illumination conditions can render crater identification difficult (Williams et al., 2018). Different illumination geometries can influence the apparent reflectance within individual images. To mitigate this issue, all of the small craters included within our counts are associated with high-reflectance deposits in ≥3 individual images.

The surface areas of the polar ice deposits (<150 km²) require counting smaller craters, and production functions are less certain at smaller crater sizes (Williams et al., 2018). Due to the small size of the identified counted craters (~100–300 m), it is possible that some of the small craters are the result of secondary impacts (McEwen and Bierhaus, 2006). However, when excluding obvious secondary chains and clusters, age dating using small-crater statistics has been shown to be within the statistical uncertainty (e.g., Hartmann, 2007; Werner et al., 2009).

It is also possible that the identified craters are subject to modification, given that small craters are more sensitive to variation (Williams et al., 2018). Previous workers investigating the importance of viscous flow at the martian north polar layered deposits found that present-day impact craters are too small and cold for viscous relaxation to play an important role in controlling crater dimensions, and that this effect may be ignored in analyses of crater size-frequency distributions (Sori et al., 2016). Here, we choose to neglect viscous relaxation as an important modification factor given that the craters on Mercury’s polar deposits that we analyze are similarly small and located in crater interiors, away from steep slopes or warmer topographies. Furthermore, analysis of flow conditions within permanently shadowed craters on Mercury suggests that the extremely cold conditions and limited thickness of the ice prevent substantial flow (Fastook et al., 2019).

The crater counts are completed for individual ice deposits, which have limited count areas (Williams et al., 2018). CSFDs can vary across a putative uniform geologic unit when count areas become small (Warner et al., 2015), however, small count areas down to 4 km² have been shown to be consistent with model ages derived from larger (100 km²) count areas, although with lower accuracy (Pasckert et al., 2015). The count areas analyzed here are between 79 and 109 km², and thus we consider errors due to limited count areas negligible.

Finally, chronology and production curves for Mercury are derived for impact craters in regolith (Le Feuvre and Wieczorek, 2011). Given that the AMAs estimated here are derived from impact craters assumed to be in ice, we estimate AMAs after scaling the superposed impact craters to estimate the sizes of these craters as if they were in regolith. Impact experiments suggest that craters in ice are found to have final diameters 2 to 3 times larger than craters in competent rocky materials (e.g., Croft et al., 1979; Koschny and Grün, 2011). We estimate ages of the ice surfaces from crater populations that are 1 (unscaled), 1/2, and 1/3 times the size of the measured crater diameters.

3.3. Stability of ejected high-reflectance material in Mercury’s PSRs

We use crater scaling laws (Eq. 2) to estimate the depth of material excavated by impacts into Ensor, Laxness, and Bechet craters. The identified small craters used in our crater-counting analysis range in size from 100–300 m in diameter, and the average diameter for all superposed craters is 211 m. We estimate the depth of the transient craters (d_transient) from the measured final crater diameter $\left({\text{D}}_{final}\right)$ in order to estimate the depth from which material is excavated $\left(d_{excavation}\right):$

$$\begin{gathered} d_{excavation}=\frac{1}{10}{d}_{transient} \\ {d}_{transient}=0.84*{\text{D}}_{final} \\ {d}_{excavation}=\frac{1}{10}\left(0.84*{\text{D}}_{final}\right) \end{gathered}$$ (Eq. 2)

We estimate that impactors resulting in craters with diameters between 100 m and 300 m would have excavated material from depths of 10 m and 30 m.

We also estimate the depth of material excavated by impacts after scaling the impacts in order to account for target properties. As discussed above, the crater scaling laws (Eq. 2) are derived for impacts into silicate material. Here we scale the diameters of the identified craters by a factor of 1/3 in order to estimate the minimum final crater diameters for equivalent impacts into regolith. This endmember scaling scenario results in a range of crater diameters from 33 m to 100 m and an average scaled diameter of 70 m. From Eq. 2, we estimate that small impact craters between 100 m and 300 m in diameter (scaled to new diameters between 33 m and 100 m) superposing a pure ice deposit would have excavated between ~3 and ~8 m below the ice surface, respectively, with an average excavation depth of ~6 m expected for all identified craters.

The thickness of water-ice deposits on Mercury is estimated to be several m (Harmon, 2007; Black et al., 2010), although possibly extending deeper with maximum thicknesses estimated of ~50 m (Eke et al., 2017; Deutsch et al., 2018). Assuming a water-ice thickness of several (~5) m (Harmon, 2007; Black et al., 2010), then all small impactors resulting in craters >180 m (scaled regolith diameter of 60 m) may have excavated both ice as well as underlying regolith. If the sizes of the impact craters superposing Mercury’s polar deposits do not differ substantially in size from impact craters in Mercury’s regolith (given that the ice deposits may be only several m thick), then it is possible that all impactors resulting in the craters identified here have excavated both ice and regolith.

Exposed surface water ice and typical mercurian regolith are factors of 4 and 2 brighter, respectively, than the average reflectance of low-albedo lag deposits that insulate the majority of ice deposits on Mercury, including all of the ice deposits analyzed here (Neumann et al., 2013). Therefore, both water ice and excavated regolith would appear as anomalously bright ejecta superposing the ice deposits. Pure water ice may sublimate away on top of the low-albedo lag deposits relatively quickly (1 m of water ice evaporates in 1 Gyr at 110 K (Vasavada et al. (1999)), however the ejected regolith would not. From the estimates of excavation depths above, all impact craters that superpose the ice deposits and have a preserved final diameter >180 m should be associated with high-reflectance materials, assuming an average ice deposit thickness of 5 m and that the preserved final diameters are 1/3 the size of craters from similar impacts into regolith. Alternatively, all impact craters that superpose the ice deposits and have a preserved final diameter >180 m should be associated with high-reflectance materials, assuming an average ice deposit thickness of 5 m and that the preserved final diameters are 1/3 the size of craters from similar impacts into regolith.

4. Results: Age estimates for Mercury’s ice deposits

Here, we suggest that brightness variations associated with small impact craters that appear bright in at least three images of different viewing geometries are indicative of excavated material. Under this assumption, we estimate the ages of specific ice deposits on Mercury using impact craters that superpose individual ice deposits (Fig. 2). Although high-resolution images of 33 north polar ice deposits have been acquired, only 3 individual ice deposits, hosted by Ensor, Laxness, and Bechet craters, show clear evidence for high-reflectance material associated with individual small craters. The remaining 30 ice deposits either show no small craters associated with high-reflectance ejecta rings using the aforementioned criteria (3 craters), or there are not at least 3 clear images for these craters (27 craters). The lack of small, superposing craters identified from high-reflectance rings on 3 north polar craters that were imaged multiple times under varying viewing geometries suggests that the polar ice deposits on Mercury are very young, or that the top ice layer has been recently restored.

Absolute model ages (AMAs) (Le Feuvre and Wieczorek, 2011) derived from superposed impact craters on Ensor, Laxness, and Bechet (Table 1) suggest that the ice deposits are geologically young (Fig. 3a), and were delivered in the most recent Kuiperian period (<280 Ma). The average AMAs for the ice deposits hosted by Ensor, Laxness, and Bechet is 95 Ma ± 33 Ma before scaling the impact craters and 4.2 Ma ± 1.2 Ma after scaling the impact craters by 1/3, where the reported uncertainties are 1-sigma standard errors derived from counting statistics alone (Table 1). Poisson analyses are used to describe the relatively likelihoods of possible ages of the surface ice deposits (Michael et al., 2016). These analyses suggest that Ensor, Laxness, and Bechet craters are all <137 ± 15 Myr, and that any ice deposits that lack superposing impact craters are likely to be <10 ± 1 Myr (Fig. 3a). In conclusion, even the most conservative estimates suggest that the ice was delivered within the last ~150 Myr.

Figure 3: Estimated ages of surface water-ice deposits on Mercury are relatively young.

a, Crater size-frequency distributions of ice deposits hosted by Ensor (blue), Laxness (red), and Bechet (green) craters. Crater counts include all craters >100 m that occur within permanently shadowed, radar-bright deposits and that are associated with high-reflectance rings suggestive of ejected material. b, Estimated ages of the ice surfaces hosted by Ensor (blue), Laxness (red), Bechet (green), and of an ice surface with no superposing craters (yellow). The probability distribution is calculated for 0.001 Gyr time intervals, and represents the likelihood (y-axis) that each surface has a particular age (x-axis).

Relatively young ages for the ice are consistent with stratigraphic relationships of host craters and contained polar ice. Global mapping of craters ≥ 40 km in diameter reveal that craters from the Mansurian period (280 Ma–1.7 Ga) are located within both the north and south polar regions (Prockter et al., 2016) and these craters host radar-bright materials indicative of water ice, placing an upper bound on the delivery of surface water ice (Deutsch et al., 2016; Chabot et al., 2018). Large Kuiperian craters (280 Ma–today) have not been mapped in the north polar region, and the only two Kuiperian craters mapped in the south polar region have not been well-imaged by Earth-based radio observations (Harmon et al., 2011).

A relatively recent delivery may explain why ice deposits on Mercury have sharp reflectance boundaries, coherent surfaces (Neumann et al., 2013; Chabot et al., 2014; Chabot et al., 2016), and high radar backscatter (Harmon et al., 2011). The ages estimated here from crater-counting statistics and Poisson analyses are consistent with the ice being delivered by a single, young impactor (e.g., Chabot et al., 2016, Rubanenko et al., 2018). It is also possible that the ice deposits have accumulated over time via micrometeorite delivery, however, the delivery rate must exceed the regolith overturn rate given the sharp reflectance boundaries observed for all ice deposits on Mercury (Neumann et al., 2013; Chabot et al., 2014; Chabot et al., 2016). Relatively young ages (<150 Myr) are inconsistent with volcanic outgassing delivering volatile species to Mercury’s poles given that the bulk of volcanic activity on Mercury ceased >3.5 Ga (Byrne et al., 2016), and recent, localized pyroclastic activity (Jozwiak et al., 2018) is not expected to contribute 10¹⁶–10¹⁸ g of water (Lawrence et al., 2013) in eruptions.

5. Discussion: Comparison to lunar polar ice

Mercury and the Moon show distinct differences in the abundance and purity of ice cold-trapped at the poles, despite both bodies having PSRs that are thermally stable environments for water ice on geologic timescales (Vasavada et al., 1999) due to the small axial tilt of each body. While both Earth-based (Harmon et al., 2011) and orbital (Lawrence et al., 2013; Neumann et al., 2013; Paige et al., 2013; Chabot et al., 2014; Deutsch et al., 2016; Deutsch et al., 2017) observations indicate that there are extensive water-ice deposits within Mercury’s PSRs (Fig. 1), data suggest that water-ice deposits on the Moon (Lawrence et al., 2006; Colaprete et al., 2010; Stacy et al., 1997; Campbell et al., 2006; Spudis et al., 2010; Thomson et al., 2012; Haruyama et al., 2008; Zuber et al., 2012; Hayne et al., 2015; Fisher et al., 2017; Li et al., 2018) are smaller in extent and less concentrated than those on Mercury.

Figure 1: Water-ice deposits located at the north polar region on Mercury.

Distribution of water ice at the north polar region of Mercury from 80°–90°N. Regions of permanent shadow (Deutsch et al., 2016) are shown in blue and radar-bright materials (Harmon et al., 2011) are shown in yellow. Host craters identified as having superposing, small craters associated with high-reflectance rings are circled in red.

For example, neutron spectrometer data of the lunar polar regions are suggestive of concentrations of only ~1.5 % water ice by mass in the upper ~1 m of the lunar regolith (Lawrence et al., 2006). Furthermore, Lunar CRater Observation and Sensing Satellite detected only ~6 wt. % water ice in the ejecta plume resulting from an impact into Cabeus crater (Colaprete et al., 2010). Earth-based radar observations do not show evidence for concentrated water ice on the Moon (e.g., Campbell et al., 2006) and spacecraft radar observations are suggestive of only patches of ice (e.g., Thomson et al., 2012). Finally, imaging (Haruyama et al., 2008) and reflectance (Zuber et al., 2012; Hayne et al., 2015; Fisher et al., 2017; Li et al., 2018) campaigns of the lunar poles have not revealed thick, coherent ice deposits, but instead are suggestive of a spatially heterogeneous, or patchy, distribution of water frost. Thus, ice on Mercury appears to be purer and more extensive than any ice on the Moon. The cause for the differences between ice on Mercury and the Moon is not understood. One possibility is that differences in delivery time may explain the stark differences between surface ice on these two airless bodies.

Impact gardening (Pieters and Noble, 2016) can produce spatial heterogeneities in ice distribution due to the loss and redistribution of volatiles through time (Crider and Vondrak, 2003; Hurley et al., 2012). Impacts introduce heterogeneity into the system because they remove volatiles via vaporization, and also preserve volatiles through the emplacement of ejecta, with a net effect of breaking up and burying the ice through time (Crider and Vondrak, 2003; Hurley et al., 2012). Because these processes take time, heterogeneity is inherently related to the age of the ice.

Hurley et al. (2012) present Monte Carlo simulations of the evolution of ice in PSRs on the Moon. They find that, statistically, individual locations on the Moon experience more burial events than excavation events and estimate an average burial rate of 1 mm/Myr. Using this average burial rate, we extrapolate backwards in time from the present day (0 Myr) into the past (4500 Myr), starting with a present day ice thickness of 1 mm. If an initial ice layer on the Moon was ever similar in thickness to the ice deposits observed at the poles of Mercury (and thus at least several meters thick (Harmon, 2007; Black et al., 2010)), then the ice observed at the surface of the Moon today must have been delivered very early on in lunar history. For example, in order for regolith gardening processes to rework a 4 m-thick layer of ice into a present-day 1-mm thickness on the surface, then the 4 m of ice should have been accumulated by ~4 Ga. Interestingly, the average burial rate of 1 mm/Myr (Hurley et al., 2012) cannot account for an initial thickness of ice in the lunar PSRs that exceeds, on average, 4.5 m, which may be thinner than what is observed on Mercury today; lower estimates for the thickness of Mercury’s polar ice deposits are at least several meters (Harmon, 2007; Black et al., 2010), while maximum estimates approach ~50 m (Eke et al., 2017; Deutsch et al., 2018). Thus, the regolith gardening simulations presented by Hurley et al. (2012) suggest that either the ice on the Moon is relatively ancient, or that less ice has been delivered to the Moon than has been delivered to Mercury through time.

6. Conclusion

Unlike surface water ice on the Moon, surface water ice on Mercury appears as coherent deposits whose distinct reflectance boundaries align directly with regions of permanent shadow (Chabot et al., 2014; Deutsch et al., 2016; Chabot et al., 2016; Chabot et al., 2018). If ice deposits on Mercury are relatively young, then they have not been exposed to extensive space weathering processes that would break up, destroy, or bury the ice. A relatively young age, as estimated here from both superposed impact craters on the ice surfaces and Poisson analyses, is consistent with (1) regolith gardening models that suggest the ice deposits were emplaced <50 Ma (Crider and Killen, 2005), (2) analysis of high-resolution images of the polar deposits that uniformly reveal sharp reflectance boundaries and spatial coherence within PSRs (Chabot et al., 2014; 2016), (3) the presence of ice in micro-cold traps (Rubanenko et al., 2018), and (4) delivery by a relatively young impactor (e.g., Chabot et al., 2016; Rubanenko et al., 2018).

To date, it is not clear why polar ice on the Moon is relatively less pure and extensive than polar ice on Mercury, which has a water-ice cold-trapping efficiency of only 50% of the Moon (Schorghofer et al., 2016). It is possible that these differences may be related to the age of the ice. Surface ice on the Moon is characterized by a substantial degree of spatial heterogeneity within a given PSR (Hayne et al., 2015). If the ice observed at the lunar poles today was delivered early on during the Moon’s history, it has undergone substantial impact bombardment, leading to a spatially heterogeneous surface distribution (Crider and Vondrak, 2003; Hurley et al., 2012). The ice deposits that show the greatest spatial heterogeneity on the surface may have substantial vertical heterogeneity as well, with additional ice buried in the subsurface (Crider and Vondrak, 2003; Hurley et al., 2012).

The same impact bombardment and space weathering processes operate on Mercury and the Moon, and Mercury’s regolith may be overturned even more frequently than the lunar regolith (e.g., Domingue et al., 2014) due to higher impact rates and speeds (e,g., Cintala, 1992; Borin et al., 2009). Thus, it is possible that relatively ancient, degraded ice deposits exist below the extensive, pure deposits observed on Mercury’s surface today. The lunar polar deposits therefore provide an interesting opportunity to inform us about the ultimate fate of mercurian polar deposits, as well as ices on other airless bodies.

This work has provided new age estimates of Mercury’s polar ice deposits, which provide critical insight into the possible delivery mechanism for the ice. We suggest that the ice was delivered to Mercury relatively recently, within the last ~150 Myr, consistent with delivery by a recent comet impact. As BepiColombo prepares to enter its orbit around Mercury, this work will serve as an important test as the first high-resolution images of the south polar PSRs are acquired.

Highlights.

• Age estimates of Mercury’s ice surfaces suggest that the ice is <150 Myr.

• A recent delivery of the ice is consistent with a single, young impactor.

• Differences between ice on Mercury and the Moon may be related to age of the ice.