Icarus 209 (2010) 210–223

Contents lists available at ScienceDirect

Icarus

journal homepage: www.elsevier .com/ locate/ icarus

Exposure of spectrally distinct material by impact craters on Mercury:Implications for global stratigraphy

Carolyn M. Ernst a,*, Scott L. Murchie a, Olivier S. Barnouin a, Mark S. Robinson b, Brett W. Denevi b,David T. Blewett a, James W. Head c, Noam R. Izenberg a, Sean C. Solomon d, James H. Roberts a

a Johns Hopkins University Applied Physics Laboratory, Laurel, MD 20723, USAb School of Earth and Space Exploration, Arizona State University, Tempe, AZ 85287, USAc Department of Geological Sciences, Brown University, Providence, RI 02912, USAd Department of Terrestrial Magnetism, Carnegie Institution of Washington, Washington, DC 20015, USA

a r t i c l e i n f o

Article history:Available online 1 June 2010

Keywords:MercuryMercury, surfaceCrateringGeological processes

0019-1035/$ - see front matter � 2010 Elsevier Inc. Adoi:10.1016/j.icarus.2010.05.022

* Corresponding author. Fax: +1 240 228 8939.E-mail address: carolyn.ernst@jhuapl.edu (C.M. Er

a b s t r a c t

MESSENGER’s Mercury Dual Imaging System (MDIS) obtained multispectral images for more than 80% ofthe surface of Mercury during its first two flybys. Those images have confirmed that the surface of Mer-cury exhibits subtle color variations, some of which can be attributed to compositional differences. Inmany areas, impact craters are associated with material that is spectrally distinct from the surroundingsurface. These deposits can be located on the crater floor, rim, wall, or central peak or in the ejectadeposit, and represent material that originally resided at depth and was subsequently excavated duringthe cratering process. The resulting craters make it possible to investigate the stratigraphy of Mercury’supper crust. Studies of laboratory, terrestrial, and lunar craters provide a means to bound the depth oforigin of spectrally distinct ejecta and central peak structures. Excavated red material (RM), with compar-atively steep (red) spectral slope, and low-reflectance material (LRM) stand out prominently from thesurrounding terrain in enhanced-color images because they are spectral end-members in Mercury’s com-positional continuum. Newly imaged examples of RM were found to be spectrally similar to the relativelyred, high-reflectance plains (HRP), suggesting that they may represent deposits of HRP-like material thatwere subsequently covered by a thin layer (�1 km thick) of intermediate plains. In one area, craters withdiameters ranging from 30 km to 130 km have excavated and incorporated RM into their rims, suggestingthat the underlying RM layer may be several kilometers thick. LRM deposits are useful as stratigraphicmarkers, due to their unique spectral properties. Some RM and LRM were excavated by pre-Tolstojanbasins, indicating a relatively old age (>4.0 Ga) for the original emplacement of these deposits. Detailedexamination of several small areas on Mercury reveals the complex nature of the local stratigraphy,including the possible presence of buried volcanic plains, and supports sequential buildup of most ofthe upper �5 km of crust by volcanic flows with compositions spanning the range of material now visibleon the surface, distributed heterogeneously across the planet. This emerging picture strongly suggeststhat the crust of Mercury is characterized by a much more substantial component of early volcanism thanrepresented by the phase of mare emplacement on Earth’s Moon.

� 2010 Elsevier Inc. All rights reserved.

1. Introduction

Mercury is the least visited of the terrestrial planets, havingbeen encountered by only two spacecraft. More than half of the pla-net remained unseen after three flybys by Mariner 10 in 1974–1975.Mercury then remained unvisited until 2008, when the MErcury Sur-face, Space ENvironment, GEochemistry, and Ranging (MESSENGER)spacecraft completed two of its three planned flybys (the first on 14January 2008 and the second on 6 October 2008) on its way to

ll rights reserved.

nst).

insertion into orbit about the innermost planet (Solomon et al.,2007). MESSENGER’s Mercury Dual Imaging System (MDIS), consist-ing of a monochrome, narrow-angle camera (NAC) and an 11-band(400–1040 nm), wide-angle camera (WAC) (Hawkins et al., 2007),has imaged over 80% of the surface to date, including approximately50% seen for the first time by a spacecraft. The Mariner 10 andMESSENGER images reveal that Mercury has subtle yet distinct colorvariations across its surface, many of which can be attributed tocompositional heterogeneity (Hapke et al., 1975; Dzurisin, 1977;Schultz, 1977; Rava and Hapke, 1987; Robinson and Lucey, 1997;Robinson and Taylor, 2001; Blewett et al., 2007, 2009; Robinsonet al., 2008; Murchie et al., 2008).

Fig. 1. MDIS reflectance spectra of color units seen on Mercury. The spectra havebeen corrected to 30�, 0�, 30� incidence, emission, and phase angle, respectively,using the most up-to-date photometric correction (Domingue et al., 2010; D.L.Domingue, personal communication, 2010). These color units, except for the bright

C.M. Ernst et al. / Icarus 209 (2010) 210–223 211

Fundamental questions concerning the nature of Mercury’s crustremained after the Mariner 10 flybys. To first order, Mercury’s heav-ily cratered surface resembles that of the Moon, though it lacks astrong albedo contrast like that between the lunar maria and high-lands. There was no consensus, however, on whether Mercury’ssmooth plains originated by impact (e.g., Oberbeck, 1975; Wilhelms,1976) or volcanism (e.g., Murray et al., 1974; Trask and Strom, 1976;Strom, 1977; Dzurisin, 1978; Kiefer and Murray, 1987). Color-ratiomaps made from the Mariner 10 ultraviolet and orange color chan-nels (355 and 575 nm) enabled the identification of distinct surfaceunits on the basis of relative color (Hapke et al., 1975, 1980; Dzurisin,1977; Schultz, 1977; Rava and Hapke, 1987). Later recalibration andreanalysis of the Mariner 10 data allowed the correlation of manycolor boundaries with geomorphic boundaries, suggesting that asubstantial number of the smooth plains are volcanic in origin(Robinson and Lucey, 1997; Robinson and Taylor, 2001). Mercury’salbedo variations are subtle compared to those of the Moon but ex-hibit a range more extensive than that of the lunar highlands andsimilar to variations among the maria (Denevi and Robinson, 2008).

The higher resolution, higher signal-to-noise ratio, near-globalcoverage, and expanded wavelength range of the MESSENGERimages permit a more detailed analysis of Mercury’s surfaceproperties. The MDIS images (Robinson et al., 2008), as well asMercury Atmospheric and Surface Composition Spectrometer data(McClintock et al., 2008) are consistent with early ground-basedobservations of relatively featureless (no obvious absorption bands)reflectance spectra having overall positive (‘‘red”) slopes withincreasing wavelength (e.g., McCord and Adams, 1972; Vilas, 1985,1988). Despite its relatively featureless spectra, the surface ofMercury does exhibit subtle spectral variations, which are due pri-marily to changes in albedo and the steepness of the spectral slopeover the MDIS wavelength range (Denevi et al., 2009). These spectraldifferences have been used to identify several major and minor colorunits (Robinson et al., 2008; Murchie et al., 2008; Denevi et al., 2009;Blewett et al., 2009). Although the first-order color variations areassociated with maturity effects, some color variations can beattributed to compositional differences (Robinson and Lucey, 1997;Blewett et al., 2007; Robinson et al., 2008; Murchie et al., 2008).

MESSENGER images confirm the Mariner-10-based hypothesisthat widespread volcanism occurred on the surface of Mercuryand played an important role in the planet’s surface evolution(Head et al., 2008, 2009a,b; Murchie et al., 2008; Denevi et al.,2009). Little is known about the three-dimensional structure ofthe crust, yet knowledge of Mercury’s crustal stratigraphy is neces-sary to understand the history and evolution of the planet. Withoutin situ measurements, there are very few ways to observe what liesbeneath the surface. Impacts excavate material from depth, mak-ing it possible to assess vertical heterogeneities in the crust usingthe resulting craters. Large basins in particular can excavate mate-rial from depths of tens of kilometers.

In this paper, we investigate the stratigraphy of Mercury’s uppercrust by means of craters that excavated spectrally distinct material.Several areas of Mercury are examined in detail using WAC and NACimages. Studies of laboratory, terrestrial, and lunar craters, alongwith theoretical models of the cratering process, allow us toestimate the depths of origin of the spectrally distinct ejecta. Fromthese results we construct stratigraphies of the upper crust in theseareas to assess how Mercury’s crust varies with depth, to make pre-liminary assessments about the distribution of color units, and toprogress toward an understanding of how the crust evolved.

crater-floor deposits (BCFDs), can be considered as parts of a compositionalcontinuum, with LRM and RM representing the known spectral end-members. Allsurface materials on Mercury display relatively low albedo and featureless spectra(no obvious absorption bands) with overall positive (‘‘red”) spectral slopes. Thethree dashed lines represent three unique RM spectra taken from the areas ofinterest in this study. The pyroclastic example is located within the craterLermontov (crater center at 15.2�N, 311.9�E).

2. The colors of Mercury

The major and minor color units of Mercury have been newlyredefined using data from the first two MESSENGER flybys (Robin-

son et al., 2008; Murchie et al., 2008; Denevi et al., 2009; Blewettet al., 2009). Here, we briefly summarize these units using the def-initions of Denevi et al. (2009) and Blewett et al. (2009). We referto units as having relatively ‘‘red” (steeper) or ‘‘blue” (shallowerbut still positive) spectral slopes through the MDIS wavelengthrange. It is important to note that these terms refer to color relativeto Mercury’s widespread intermediate terrain (Robinson et al.,2008; Denevi et al., 2009), and that all materials observed on Mer-cury exhibit positive spectral slopes.

Three major terrain classes are defined on the basis of spectralslope, reflectance, and morphology: smooth plains, spectrallyintermediate terrain (IT), and low-reflectance material (LRM). Thesmooth plains themselves are subdivided into three types: high-reflectance red plains (HRP), intermediate plains (IP), and low-reflectance blue plains (LBP). Representative MDIS spectra of thesecolor units are shown in Fig. 1. The IT generally corresponds toareas with higher crater density than the IP (Denevi et al., 2009).The IT and IP have similar spectral properties, suggesting theymay have formed in a similar manner at different times (Deneviet al., 2009).

HRP represent the brightest and reddest of the smooth plains,with average reflectance values as much as 20% higher than theglobal mean at 750 nm. IP appear spectrally similar to IT but havelower crater densities. These intermediate units exhibit spectralproperties similar to the global mean. The LBP are intermediate be-tween IP and LRM in both reflectance and spectral slope. TypicalLRM deposits have reflectance values as much as 30% below theglobal mean and are bluer than most other materials on Mercury(spectral slopes �5% lower than HRP). The LRM likely constitutesa compositional end-member (Denevi et al., 2009).

The observed color variations are primarily associated with sub-tle changes in the steepness of the spectral slope, which differs byno more than �5% between the reddest and bluest of the major

Fig. 2. Schematic illustration of how craters of different sizes can reveal clues to thesubsurface stratigraphy. In all cases, the black unit represents impact melt and thedashed line indicates the pre-impact ground surface. (a) A crater too small topenetrate the surface layer (Layer 1) can be used to constrain the minimumthickness of this layer. (b) A crater that penetrates through the top layer andexcavates underlying material constrains the maximum thickness of the top layerand reveals underlying layers (in this example, Layer 2). (c) A crater revealing athird stratigraphic layer (Layer 3) in a central structure can additionally constrainthe thickness of the second layer.

212 C.M. Ernst et al. / Icarus 209 (2010) 210–223

color units. The general spectral trend observed on the surface isdecreasing reflectance with the decreasing (bluer) spectral slope.This trend is opposite to that expected from space weathering,where materials darken and redden over time (e.g., Fischer andPieters, 1994). Therefore, the spectral variations among the majorcolor units can be attributed to compositional differences (Robin-son et al., 2008; Blewett et al., 2009). Most likely, the units HRP,IP/IT, LBP, and LRM can be considered to represent a compositionalcontinuum.

There are three additional, areally minor color units found onMercury’s surface (e.g., Robinson et al., 2008). These units standout prominently in color composites and tend to be associatedwith impact craters. One minor unit consists of fresh (immature)crater ejecta, which have been less affected by space weatheringthan typical surface units on Mercury and appear brighter andbluer than their mature counterparts (Robinson et al., 2008). Thealbedo differences between fresh and mature material producethe most extreme color variations seen on Mercury’s surface (Rob-inson et al., 2008). In many cases fresh deposits are not composi-tionally different from the underlying materials, so they will notbe discussed further in this study.

The second minor color unit consists of bright crater-floordeposits (BCFDs) (e.g., Robinson et al., 2008; Blewett et al., 2009).BCFDs are spectral outliers on the surface of Mercury (see Fig. 1),with high reflectance values and relatively blue (shallow) spectralslopes, implying a distinctive composition (Blewett et al., 2010).These units appear to be associated with impact craters, but theorigin and composition of the BCFDs are not yet known.

The last of the minor color units is characterized by a relativelyred spectrum and an elevated albedo, first noted in Mariner 10 data(Rava and Hapke, 1987; Dzurisin, 1977; Schultz, 1977), and subse-quently recognized in MESSENGER data (Robinson et al., 2008;Murchie et al., 2008; Head et al., 2008; Blewett et al., 2009; Kerberet al., 2009). These red units, sometimes referred to as ‘‘red spots,”have either primary or secondary associations with craters and ba-sins. Primary associations are those where the red material (RM,see Fig. 1) is a part of the structure of the crater, having been em-placed during the impact process. Lunar dark-halo craters (DHCs),sites where impact craters excavate mare basalt from beneath ahighland-rich ejecta deposit, are analogous features on the Moon(e.g., Schultz and Spudis, 1979; Hawke and Bell, 1981; Bell andHawke, 1984; Head and Wilson, 1992; Blewett et al., 1995; Anton-enko et al., 1995). Primary occurrences of red spots can be locatedon the crater floor, rim, wall, or central structures, or in the ejectadeposit. Secondary associations are those where the red unit isfound in or around an impact structure but was emplaced as a re-sult of a subsequent process rather than the impact itself. Theseoccurrences are typically associated with rimless depressions lo-cated on the floor or in the walls of craters and basins and arethought to be of pyroclastic origin (Robinson et al., 2008; Murchieet al., 2008; Head et al., 2008, 2009a; Blewett et al., 2009; Kerberet al., 2009). We focus on the primary associations in this study.

3. Determining depth of origin from impact craters

There are varying degrees to which an impact crater can provideinsight into the local stratigraphy, as illustrated in Fig. 2. If an im-pact does not penetrate through the surface layer, the ejecta willhave the same composition as the surface material (Fig. 2a), withmaturity effects dominating any spectral differences between theexcavated material and the surrounding terrain. In such cases,the minimum thickness of the surface layer can be derived. If animpact penetrates through the surface layer and excavates mate-rial of a different composition, the ejecta will be spectrally distinctfrom the surrounding terrain (Fig. 2b), as in a lunar dark-halo cra-

ter. In this case, the thickness of the surface layer can be con-strained and a minimum thickness of the underlying materialcan be derived. The maximum thickness of the second material willremain unconstrained unless the impact also exposes an additionalcolor unit (e.g., in a central peak, Fig. 2c).

During crater excavation, material from shallower depths isthrown the farthest from the crater center. Closer to the craterrim, these ejecta are covered by the deposition of material fromdeeper layers (e.g., Stöffler et al., 1975). The resulting overturnedflap of ejecta at the crater rim exhibits a stratigraphy inverted fromthat of the pre-impact subsurface, with the most deeply excavatedmaterials visible at the surface; therefore, spectrally distinct mate-rial located on a crater rim should have been derived from depthsapproaching the maximum excavation depth of the crater.

3.1. Maximum excavation depth of ejecta

Ejecta generated by an impact event can originate from nodeeper than the maximum depth of excavation. This depth isapproximately equal to one-tenth the transient crater diameter(Dtc, measured from the pre-impact surface) for vertical impactsand is shallower for oblique ones (Gault et al., 1968; Melosh,

Fig. 3. Rim-to-rim diameter as a function of transient crater diameter calculatedusing the methods of Holsapple (1993), Melosh (1989), and Croft (1985). Themaximum excavation depth is taken to be one-tenth of the transient craterdiameter.

C.M. Ernst et al. / Icarus 209 (2010) 210–223 213

1989). The excavation depth achieved by a vertical impact is there-fore a maximum value.

The transient crater represents the maximum extent of thegrowing cavity at the end of the excavation stage, before any col-lapse has occurred. Because Dtc is not observable in post-impactimages, it is necessary to relate Dtc to the observable rim-to-rimdiameter (Dr). Crater size greatly affects such a relationship. Bothcomplex and simple craters are larger in diameter and shallowerthan their associated transient craters. Due to extensive modifica-tion (e.g., rim failure, slumping) the final Dr of a complex crater issubstantially enlarged relative to the equivalent-sized simple cra-ter. Target-specific conditions also must be considered, includingsurface gravitational acceleration, average impact speed, and tar-get strength (e.g., Cintala et al., 1977; Pike, 1980; Stewart and Val-iant, 2006). These factors affect the diameter at which simplecraters transition to complex ones. This transition diameter (D�)is an important parameter used in deriving the relationship be-tween the observed rim-to-rim diameter and the transient craterdiameter. We use 10.3 ± 4 km as D� for Mercury, as determinedfrom Mariner 10 data by Pike (1988). This variation in the valueof D� results in a ±4% variation in the calculated transient craterdiameter, which is not significant, considering our rule-of-thumbestimate that the maximum depth of excavation is equal to one-tenth of Dtc.

Three analytical methods relating Dr to Dtc were derived fromlunar and terrestrial observations (Croft, 1985; Melosh, 1989;Holsapple, 1993). The most comprehensive of these studies (Hols-apple, 1993) also incorporated experimental observations and im-pact scaling rules. All three methods can be written in the form of apower law:

Dtc ffi CDkr : ð1Þ

The values of constants C and k are given in Table 1. The rela-tionships of Dr to both Dtc and the maximum excavation depthfor all three methods are shown in Fig. 3. The Croft and Holsapplerelationships incorporate transition diameter into the constant C,making these methods more suitable for analyzing Mercury cratersas they inherently account for the differences in complex and sim-ple crater collapse and for some planet-specific variables such asgravity and target strength. The Melosh relationship was derivedonly for lunar and terrestrial craters and requires separate con-stants for simple and complex craters and for each planetary body.

The maximum excavation depths reported in the following sec-tions are calculated with the Holsapple method for vertical im-pacts. The equivalent values using the Croft or Melosh methodsare provided in Table 2. The relative depths and stratigraphy inan area will remain the same with any of the three methods; onlythe absolute depth values are affected. Because all three methodswere derived in large part from lunar and terrestrial data, modifi-cations may be necessary to translate the relationships to Mer-cury’s higher-impact-speed environment.

There is a large spread in the expected impact speed at Mercury,with mean values typically quoted as being between �20 and40 km/s (e.g., Hartmann et al., 1981; Horedt and Neukum, 1984;Schultz, 1988; Cintala, 1992; Marchi et al., 2005; Le Feuvre andWieczorek, 2008). Despite the range in absolute values, the relative

Table 1Constants for relating Dr to Dtc.

C k Source

D015�

0.85 Croft (1985)

0:758D0:079� 0.921 Holsapple (1993)

0.84 1 Melosh (1989) – simple0.5–0.65 1 Melosh (1989) – complex

mean impact speed at Mercury is approximately a factor of twohigher than that at the Moon. For this study, we take the mean im-pact speed on Mercury to be 42.5 km/s, compared to 19.4 km/s forthe Moon and 20.4 km/s for the Earth, on the basis of recent work(Le Feuvre and Wieczorek, 2008). Low-speed (Barnouin-Jha et al.,2007) and preliminary higher-speed (Barnouin-Jha et al., 2009)experimental investigations indicate that increases in impactvelocity may lead to decreases in the depth of the transient crater,and thus the maximum depth of excavation, with an effect of up to�20% for a velocity difference such as that between the mean im-pact speeds at the Moon and Mercury. This area warrants furtherstudy, but the values calculated for lunar impact speeds still pro-vide maximum values for depth of excavation that would hold ina 40 km/s environment, and a broad range in the speed of projec-tiles striking Mercury will lead to impacts in the speed range ofthose occurring on the Moon.

3.2. Central structures

Central peaks and peak rings can be used to constrain furtherthe vertical extent of subsurface material. Central structures areformed by uplift from beneath the zone of excavation and repre-sent the deepest material exposed by an impact. Because meltedrock has little to no strength, it cannot form a stable positive topo-graphic feature such as a central peak or peak ring; therefore, thematerial making up central peaks cannot have been melted duringthe impact process (Cintala and Grieve, 1998a). Consequently, dur-ing rebound, melts flow off the rising central peaks to fill the craterfloor. The minimum depth of origin for the top of the central peakscan thus be estimated as the maximum melting depth for a givenimpact (Grieve and Cintala, 1992; Cintala and Grieve, 1998a).

The peak pressures generated during an impact control the re-gions of a target that melt or vaporize (e.g., O’Keefe and Ahrens,1977; Cintala, 1992; Pierazzo et al., 1997; Cintala and Grieve,1998a). Taking into account the rapid attenuation of pressure nearthe surface of a target, vaporization and melt zones generated inthe target are nearly spherical in shape (e.g., Pierazzo et al.,1997; Cintala and Grieve, 1998a). The maximum depth of the melt-ing zone provides an estimate of the minimum depth from whichcentral peaks can be derived (Cintala and Grieve, 1998a). The

Table 2Measured craters on Mercury.

Region Crater name Diameter (km) Minimum depthof central peakorigin (km)

Maximum depth of excavation (km)

Holsapplemethod

Croftmethod

Melosh method(complex)

Rudaki plainsCalvino 68 6 4.4 5.1 3.4–4.4R1 16 – 1.2 1.5 0.8–1.0

Titian basinTitian 121 12 7.6 8.4 6.1–7.9T1 4 – 0.3 0.5 0.2–0.3T2 9 – 0.7 0.9 0.5–0.6T3 7 – 0.6 0.7 0.4–0.5T4 8 – 0.6 0.8 0.4–0.5

Homer basinDominici 20 – 1.4 1.8 1.0–1.3H1 11 – 0.8 1.1 0.6–0.7H2 38 – 2.6 3.1 1.9–2.5H3 15 – 1.1 1.4 0.8–1.0H4 15 – 1.1 1.4 0.8–1.0H5 50 4 3.3 4.0 2.5–3.3

Hemingway regionHemingway 130 13 8.1 8.9 6.5–8.5de Graft 65 6 4.3 4.9 3.3–4.2He1 43 – 2.9 3.5 2.2–2.8He2 30 – 2.1 2.6 1.5–2.0

Caloris basinAtget 100 10 6.3 7.1 5.0–6.5Apollodorus 41 4 2.8 3.3 2.1–2.7Cunningham 37 3 2.5 3.1 1.9–2.4

214 C.M. Ernst et al. / Icarus 209 (2010) 210–223

minimum depth of origin of peak rings can be determined follow-ing the same logic as for central peaks. Peak rings occur away fromthe crater center, and the maximum melting depth beneath thepeak ring location corresponds to the minimum depth of originof the peak ring.

Three methods are used to calculate the relationship betweenDtc and the maximum melting depth for craters on the Moon,Earth, and Mercury. In all cases, either dense basalt or dunite is as-sumed to compose both the projectile and target and the impactsare assumed to be vertical. Because these methods determine thevolume or depth of melt relative to impactor diameter, a, we calcu-late the corresponding Dtc on a given planetary body using theSchmidt and Housen (1987) scaling relation:

Dtc ¼ 1:16qp

qt

� �13

a0:78v0:44g�0:22; ð2Þ

where qp and qt are the densities of the projectile and target,respectively, v is the impact velocity, and g is the surface gravita-tional acceleration of the body. The relationship between projectilesize and transient crater size is the only place the gravitationalacceleration of the body appears in these calculations.

The first method follows the calculations of Cintala (1992) andCintala and Grieve (1998a,b), which were derived analytically. Thepressure, P, at a given location is determined by

P � Pmaxxa

� ��2cos2b h ð3Þ

This equation is used to calculate the shape of the melt zone, wherePmax is the maximum pressure occurring during the impact, x is thedistance of the pressure zone from the impact point in cylindricalcoordinates, b is the initial ratio of target compression to projectilecompression, and h is the angle measured from the center line be-neath the impactor to the point of interest in the target (Cintala,1992). The phase changes are determined by calculating the entro-

py along the shock Hugoniot using a Murnaghan equation of state[which relates pressure to density non-linearly but does not incor-porate a dependence on internal energy, see Appendix II.4 of Me-losh (1989)] (Cintala, 1992). Solving this equation for x whenh = 0� for the pressure at which complete melting of the targetmaterial occurs gives a measure of the maximum depth of melting.

In order to match the results of Cintala (1992) and Cintala andGrieve (1998a,b), we add a penetration term to the calculateddepth to account for the initial penetration of the projectile intothe target before the projectile is fully coupled and transfers its en-ergy to the target. This value was calculated through impedancematching following Gault and Heitowit (1963). These calculationsare in reasonable agreement with those of other more complexnumerical models [see Fig. 3 of Cintala (1992), O’Keefe and Ahrens(1975, 1977), and Austin et al. (1980)].

The second method uses the melt calculations of Pierazzo et al.(1997), derived from numerical hydrocode calculations of verticalimpacts. These calculations incorporated an Analytical Equationof State (ANEOS), which derives thermodynamic functions (e.g.,pressure) from the Helmholtz free energy (Thompson and Lauson,1972). As in the case of Cintala and Grieve (1998a,b), the equationof state is used to compute the entropy along the shock Hugoniotat which phase changes in the target occur. The volume of melt ex-pected for a given impact is calculated using the dunite-specific fitsto the numerical calculations relating the radius and depth of thecomplete melting region to the impact velocity from Table III ofPierazzo et al. (1997).

Both Cintala and Grieve (1998a, their Fig. 8) and Pierazzo et al.(1997, their Fig. 12) compare their derived impact melt volumeswith terrestrial data, indicating good agreement with the observa-tional data. These comparisons are reproduced in Fig. 4. The Cintalaand Grieve values tend to overestimate the melt, whereas the Pie-razzo et al. values are more consistent with the terrestrial observa-tions; however, the amounts of melt measured in terrestrial craterswill tend to be minimum estimates, as erosion over time will erase

Fig. 4. Comparison of the calculated volume of impact melt for impacts on Earth toterrestrial data (solid diamonds) for the three methods detailed in the text: Cintala(1992) and Cintala and Grieve (1998a,b) (solid line); Pierazzo et al. (1997) (dashedline); and a modification of W.A. Watters et al. (2009) (dotted line). The terrestrialdata are from Grieve and Cintala (1992). The two terrestrial points that lie off thetrend line are those of the Ries and Logoisk craters, at which thick sedimentarylayers overlie the crystalline targets (Grieve and Cintala, 1992; Pierazzo et al.,1997).

Fig. 5. Relationship between the maximum depth of melting (corresponding to theminimum depth-of-origin for a central peak) and the transient crater diameter forMercury, calculated by means of three methods: Cintala (1992) and Cintala andGrieve (1998a,b) (solid line); Pierazzo et al. (1997) (dashed line); and a modificationof W.A. Watters et al. (2009) (dotted line).

C.M. Ernst et al. / Icarus 209 (2010) 210–223 215

evidence of portions of the melt (Grieve and Cintala, 1992). There-fore, the Pierazzo et al. values likely underestimate the amount ofmelt generated during an impact.

Here, we incorporate a third method of calculating the volumeand depth of melt produced during an impact. This method isbased on the W.A. Watters et al. (2009) ‘‘foundering shock” methodfor calculating shock heating. W.A. Watters et al. take the geometryof the shock pressure field as determined by the decay law pro-posed by Ahrens and O’Keefe (1977), with the best-fit decay con-stants to the numerical calculations of Pierazzo et al. (1997), anduse the waste heat or shock-Huogiont method (Gault and Heitowit,1963) to determine the amount of energy released during decom-pression of the shock phase along the Hugoniot. W.A. Watters et al.assumed that all waste heat increases the temperature of the tar-get; this premise does not take into account the latent heat ofthe melted material, which consumes a substantial amount ofthe waste heat. Therefore, as a modification to this method, thewaste heat is assumed here to contribute to an increase in temper-ature up to the melting point of the target material at a given pres-sure, after which additional temperature increase occurs only withpartial melting. The points at which the temperature is equal to orexceeds the melting temperature of the target material determinethe volume and maximum depth of melting. Constant and conduc-tive crustal temperature profiles were examined, and both yieldedsimilar results for the crater sizes of interest for this study.

The melt volumes calculated by this method are also shown inFig. 4. The amounts of melt generated by this modified W.A. Wat-ters et al. (2009) model for terrestrial impacts are greater than forthe results of Pierazzo et al. (1997), but less than those of Cintalaand Grieve (1998a). This model appears to provide the best fit tothe amount of melt seen at terrestrial craters, only slightly overes-timating the amount of terrestrial melt observed.

The relationship between Dtc and the maximum melting depthfor craters on Mercury is shown in Fig. 5 using all three methodsdescribed here for vertical impacts at 42.5 km/s into dunite targets.The assumed shape of the melt volume for the modified W.A.Watters et al. (2009) method is shallower and broader than for

the other two methods (where the volume is roughly spherical)due to the nature of the incorporated pre-existing solidus. Whenthese methods are used to calculate melt depths for high velocities(>25 km/s), the modified W.A. Watters et al. method yields theshallowest values for the maximum melting depth. The spread ofvalues across the three methods for Mercury is less than a factorof two and is low enough to provide reasonable confidence in theabsolute depth values calculated here for Mercury. This range alsoreflects the broad range of impact velocities expected on Mercury(Le Feuvre and Wieczorek, 2008). For all calculations below weuse the maximum depth of melting calculated with the modifiedW.A. Watters et al. method.

4. Observations

We use MESSENGER WAC multispectral data to identify occur-rences of spectrally distinct material in primary association withimpact craters. Material is defined as spectrally distinct if it isfound to contrast with the pre-impact surface on the basis of spec-tral shape, band ratios, and principal component analysis. Spectralcontrasts due solely to maturity effects are excluded from thisstudy. The enhanced-color images featured in this paper containthe inverse of the second principal component (-PC2), the firstprincipal component (PC1), and the 430-nm/1000-nm ratio in thered, green, and blue image planes, respectively. Here, PC2 high-lights spectral (compositional) variations, PC1 is dominated by al-bedo variations (e.g., Robinson et al., 2008), and the 430-nm/1000-nm color ratio emphasizes relative slope variations. Excavated RMand LRM stand out conspicuously from the surrounding terrain inthese color composites, because they are the spectral end-mem-bers in Mercury’s compositional continuum. Therefore, these unitsform the focus of this study.

Once areas of interest are identified, WAC data are overlaid onNAC mosaics in order to examine the regions at higher spatial res-olution. Table 2 lists the location, diameter, maximum excavationdepth, and minimum origin depth of central peaks (if applicable)

Fig. 6. Color composite of high-resolution WAC images, showing -PC2, PC1, and relative visible color (430-nm/1000-nm ratio) in the red (R), green (G), and blue (B) imageplanes, respectively. The scene is approximately 1150 km across and spans from the Rudaki plains in the west to Homer basin in the east. The image is in equirectangularprojection and is centered at 1.3�S, 313.3�E.

Fig. 7. Color composite of the Rudaki plains, centered on Calvino crater (68 km indiameter, 3.9�S, 304�E). The Rudaki plains are IP that have infilled a degraded,�370-km-diameter basin. Calvino, superposed on the IP, has excavated RM in itsrim and LRM in a portion of its central peaks. The inset provides a higher-resolutionview of Calvino crater.

216 C.M. Ernst et al. / Icarus 209 (2010) 210–223

for each crater discussed in this study. The specific areas of interestthat we have examined are described below in Sections 4.1–4.4.

4.1. Second flyby high-resolution color data

The first area that was closely examined is located in the MES-SENGER second flyby departure hemisphere, within the region ofMercury that has been imaged at the highest MDIS spatial resolu-tion to date (�460 m/pixel for the WAC, �120 m/pixel for theNAC). Fig. 6 is a �1150-km-wide WAC mosaic centered at 1.3�S,313.3�E and ranging from the plains just west of Rudaki crater(the Rudaki plains) eastward to Homer basin. The reddest areasin this image represent occurrences of RM, and the darkest blueareas represent LRM. There are several occurrences of excavated,spectrally distinct materials in this area. We focus separately onthe western (Rudaki plains, Fig. 7), central (Titian basin, Fig. 8),and eastern (Homer basin, Fig. 9) portions of the image in orderto examine each area in greater detail.

4.1.1. Rudaki plainsThe Rudaki plains (Fig. 7) have been classified as smooth

volcanic IP material that has filled an unnamed, degraded,�370-km-diameter basin (Denevi et al., 2009). This area containsa prominent example of spectrally distinct, excavated material:the 68-km-diameter complex crater Calvino, located near the cen-ter of the Rudaki plains at 3.9�S, 304�E. Calvino is superposed onthe IP terrain. The crater rim is formed by RM, whereas LRM makesup a portion of its central peak structure. This region was also de-scribed by Denevi et al. (2009).

The multiple stratigraphic units exposed by Calvino constrainthe thicknesses of the surface IP layer and the underlying RMand define the minimum depth of origin for the LRM material ex-posed in the central peaks. Small, post-flooding craters up to 16 kmin diameter (e.g., crater R1 in Fig. 7) do not penetrate through thesurface IP. From the maximum excavation depth of R1 (the largestnon-penetrating crater), the IP layer at the center of the Rudaki

plains must be at least 1.2 km thick. There are no post-flooding cra-ters intermediate in size between R1 and Calvino in the central Ru-daki plains; therefore, the maximum excavation depth of Calvino,�4.4 km, provides the only depth constraint for the location ofthe RM layer. Because Calvino excavates RM, the top of this layermust lie at a depth between 1.2 km (maximum excavation depthof R1) and 4.4 km (maximum excavation depth of Calvino).

C.M. Ernst et al. / Icarus 209 (2010) 210–223 217

The dual composition of Calvino’s central peaks implies eitherthat the lateral distribution of LRM is uneven beneath the crater,or that the peaks are derived from a depth range containing theLRM interface. Under the latter scenario, the LRM unit begins at adepth of�6 km, constraining the thickness of the RM layer beneathCalvino to between �1 and 5 km. No constraint can be placed onthe vertical extent of the LRM in this area.

4.1.2. Titian basinThe impactor that formed Titian basin (121 km in diameter,

3.6�S, 317.9�E, Fig. 8) struck smooth plains (Denevi et al., 2009),which exhibit IP spectral characteristics. The Titian ejecta depositcontains LRM, which conceals the spectral signature of much ofthe underlying plains unit (Denevi et al., 2009). From the maxi-mum excavation depth of Titian, the ejected LRM originated fromno deeper than 7.6 km. The basin is filled with younger smoothplains (Denevi et al., 2009), also of the IP spectral type, and no cen-tral peaks are visible.

Two craters (T1 and T2), 4 and 9 km in diameter, superposed onthe northern floor of Titian excavated LRM from beneath the youn-ger IP (Denevi et al., 2009). Therefore, the IP infill beneath thesecraters must be <0.3 km thick (maximum excavation depth ofT1). A 7-km-diameter crater in the northeastern portion of Titian(T3) does not expose LRM, suggesting an uneven infill of IP, whichmust be >0.6 km beneath T3.

A prominent small, 8-km-diameter crater north of Titian (T4 inFig. 8) excavated through the LRM ejecta deposit to expose theunderlying, older IP unit. This crater is analogous to lunar dark-ha-loed craters (e.g., Schultz and Spudis, 1979; Bell and Hawke, 1984;

Fig. 8. Color composite of Titian basin (121 km in diameter, 3.6�S, 317.9�E) and thesurrounding area. Titian basin excavated large amounts of LRM (outlined), whichcan be seen to the north and east of the basin. The basin and its ejecta aresuperposed on older cratered IP, and its interior was flooded with younger IP.Subsequent impacts (T1, 4 km in diameter, and T2, 9 km in diameter) haveexcavated LRM from beneath the newer IP infill, whereas T3, a 7-km-diametercrater, does not expose LRM. Crater T4 (8 km in diameter) has excavated older IPmaterial from beneath the LRM ejecta deposit, creating a bright-halo crater.

Head and Wilson, 1992), which excavate dark basaltic mare mate-rial from beneath higher-albedo, highland-rich ejecta deposits. TheT4 case exhibits an inverted color relationship relative to the caseof lunar DHCs: it exposes relatively high-reflectance, likely volca-nic IP from beneath dark LRM ejecta. The maximum excavationdepth of this crater limits the thickness of the LRM in this portionof the ejecta deposit to <0.6 km.

4.1.3. Homer basinHomer basin (Fig. 9) is a �310-km-diameter basin that has been

flooded with IP (Denevi et al., 2009). Homer was classified as atwo-ring basin on the basis of Mariner 10 images (e.g., Spudisand Guest, 1988; Pike, 1988). The northern and western portionsof the basin’s outer rim appear to have been modified. The areain and around Homer basin contains several occurrences of thered unit in primary association with craters. There is an 11-km-diameter crater, designated H1 in Fig. 9, located near the centerof Homer that does not appear to have excavated through the IP.The maximum excavation depth of H1 constrains the thicknessof the IP beneath this crater to >0.8 km. There are several possibleexplanations for this observation, including: (1) the red material

Fig. 9. Color composite of Homer basin (�310 km in diameter, centered at 1.2�S,323.8�E). RM is seen on the floor of a 36-km-diameter crater (H2), in the ejecta andon the floor of two 15-km-diameter craters (H3 and H4), and in the ejecta ofDominici, a 20-km-diameter crater. A nearby 11-km-diameter crater (H1) has notexcavated any detectable red material. RM also is present in the central peak of the50-km-diameter crater (H5) to the south of the basin. The dashed and dotted circlesdelineate Homer’s outer and inner rings, respectively. The basin rim has beensubstantially modified. The solid rectangles indicate occurrences of RM in Homerbasin’s peak ring.

218 C.M. Ernst et al. / Icarus 209 (2010) 210–223

was covered by >0.8 km of IP; (2) the impact was oblique, makingthe excavation depth shallower than �1/10 the crater diameter(Gault et al., 1968); or (3) the red unit does not extend beneath thiscrater or is present only at greater depth.

The craters in and around Homer basin that expose RM includethe following: H2 on the floor of a �38-km-diameter crater (max-imum excavation depth �2.6 km); H3 and H4 on the floor and inthe near-crater ejecta of two �15-km-diameter craters (maximumexcavation depth �1.1 km); and the ejecta of Dominici, a 20-km-diameter crater (maximum excavation depth �1.4 km). Thestill-visible rays of Dominici indicate that it is one of the youngestfeatures located within the area depicted in Fig. 6. Some ejecta andthe floor of Dominici appear cyan in the enhanced-color image andare spectrally consistent with BCFD material. The planform of theejecta suggests that the impact may have been oblique, whichwould decrease the maximum excavation depth. NAC images ofDominici show evidence of slumping, perhaps influenced by thecreation of this crater on or close to Homer’s peak ring. Althoughit is difficult to fully interpret the morphology of this crater withthe currently available images, these complications should notmarkedly affect the interpretation of the RM ejecta and maximumexcavation depth given that there are other, similarly sized cratersin the area that have excavated RM (H3 and H4).

Crater H4 has RM only on the western side of its floor and ejectadeposit. Examination of WAC color data (Fig. 10a) together withthe high-resolution (�120 m/pixel) NAC images (Fig. 10b) suggeststhat Homer’s peak ring blocked the crater ejecta in the eastwarddirection. The excess material lying on the eastern half of the craterfloor would then represent the deposition of the blocked ejecta andslumping from the over-steepened eastern crater wall.

From the maximum excavation depths of craters H2–H4 andDominici and from the near-rim location of the RM, an RM layerextends at least between 1.1 and 2.6 km depth (layer >1.5 kmthick). These craters all lie in close proximity to Homer’s peak ring,and there are additional exposures of RM in the southern portion ofthe peak ring (see Fig. 9). Without the special set of high-resolutionimages from MESSENGER’s second flyby, these RM exposureswould not be detectable.

Because it is incorporated into Homer’s peak ring, the RM musthave been present in the pre-basin stratigraphy. The formation ofHomer would have brought this RM up from depth, exposing itat the surface at least in the area of the peak ring. The basin floorwas subsequently infilled by IP, covering the northern portion ofthe peak ring, and leaving RM exposed only in the southern portionof the peak ring. The RM was then easily excavated by the subse-quent H2–H4 and Dominici impacts, which were large enough topenetrate the IP layer. No LRM or other underlying material is ob-

Fig. 10. (a) WAC � 460 m/pixel color composite and (b) NAC � 120-m/pixel imagecentered on a 15-km-diameter crater (H4 from Fig. 9) containing RM at 1.6�S,325.1�E. Arrows indicate the presence of a ridge (Homer’s peak ring) that acted asan obstruction to the ejecta. Some of this blocked ejecta and material from aslumping wall were deposited onto the eastern floor of the crater.

served in the craters on Homer’s floor, so no maximum depth canbe placed on the extent of the RM.

Just south of Homer basin in the highly cratered, intermediateterrain is a 50-km-diameter crater, H5, which exhibits a centralpeak spectrally consistent with the other RM in the scene(Fig. 9). The maximum excavation depth of H5 is 3.3 km, and theminimum depth of origin of this peak is �4 km; therefore, thetop of the RM layer outside of Homer basin must exist betweenthese depths. No RM is seen in the ejecta of H5 or in nearby smallercraters. If this RM represents the same unit as that exposed insideof Homer basin, no large basin-forming impact helped to bring it tothe surface, so the RM outside of Homer remained too deep to betapped by smaller craters.

4.2. Hemingway region

The region surrounding the crater Hemingway (17.5�N, 357.1�E,imaged during MESSENGER’s second flyby) is shown in a �1000-km-wide WAC image (Fig. 11). The spatial resolution of the datain this area (WAC � 2400 m/pixel, NAC � 580 m/pixel) is approxi-mately five times lower than that for the areas described in Sec-tion 4.1. The cyan-colored, linear features running through thearea are part of the extensive ray system of the 95-km-diametercrater Hokusai, to the north and east of the scene (57.8�N,16.8�E). Hokusai was first identified in Earth-based radar images(Slade et al., 1992; Butler et al., 1993; Harmon, 2007).

Four large craters in Fig. 11 are easily identifiable as exposingRM. Hemingway, a 130-km-diameter crater (maximum excavationdepth �8.1 km), has RM exposed on its floor. Hemingway also hasLRM material exposed in its central peak, which has a minimumdepth of origin of 13 km. De Graft, a 65-km-diameter crater (max-imum excavation depth �4.3 km), has a rim of RM. BCFD materialis present inside the crater, and a crater ray is draped around it(Blewett et al., 2010), complicating spectral observations of its cen-tral peaks. A 43-km-diameter crater, labeled He1 in Fig. 11 (maxi-mum excavation depth �2.9 km), and a 30-km-diameter crater,He2 (maximum excavation depth �2.1 km), also have RM rims.

Fig. 11. Color composite of the �1000-km-wide region surrounding Hemingway(17.5�N, 357.1�E), which excavated RM and exhibits central peaks composed ofLRM. Also containing RM in their rims are de Graft (65 km in diameter), He1 (43 kmin diameter), and He2 (30 km in diameter). The incorporation of RM into the rims ofcraters with a wide range of diameters suggests that the underlying RM layer maybe several kilometers thick.

C.M. Ernst et al. / Icarus 209 (2010) 210–223 219

Spectra of the central peaks of craters He1 and He2 cannot be ob-served due to the limited spatial resolution of the WAC images.

Because these craters with a wide range of diameters have exca-vated and incorporated RM into their rims, this material must bepresent over a large depth interval, at minimum ranging from �2to 8 km. If this region has a relatively uniform subsurface, theRM layer would be >6 km thick. Smaller bright-red areas can beseen in Fig. 11; however, the resolution of the data precludesdetermining whether they are associated with craters or other geo-logical structures, such as pyroclastic vents. It is possible that RMexists closer to the surface than 2 km; higher-resolution imagesfrom the MESSENGER mission orbital phase will help to settle thisissue. The LRM in Hemingway’s central peak originates at a mini-mum depth of �13 km, providing a limit to the maximum depthof the RM layer.

4.3. Caloris basin

At 1550 km in diameter, Caloris basin is the largest well-pre-served impact basin on Mercury (Murchie et al., 2008). The redspots located along the periphery of Caloris (Fig. 12) have a second-ary association with the basin. The red spots surround rimlessdepressions and have been interpreted as pyroclastic deposits(Murchie et al., 2008; Head et al., 2008; Kerber et al., 2009; Blewettet al., 2009). The interior of Caloris contains smooth plains thathave been classified as HRP (Robinson et al., 2008; T.R. Watterset al., 2009). Several craters within the basin interior (Nawahi,Munch, Sander, and Poe) are embayed and infilled by HRP (seeFig. 12), indicating that they predate the latest emplacement of thisplains material (Murchie et al., 2008). The still-visible rims of thesecraters provide clues to the nature of the material beneath the HRP,which is spectrally consistent with LRM (Robinson et al., 2008). Ifthis LRM represents the original floor of Caloris basin, it was likelybrought to the surface from great depth by the large basin-formingimpact.

Fig. 12. Color composite of Caloris basin (1550 km in diameter, centered at 30.5�N,170.2�E). The interior of the basin has been infilled by HRP. Several craters withinthe basin interior (Nawahi, Munch, Sander, and Poe) are embayed and infilled bythe HRP, indicating that they predate HRP emplacement. The 100-km-diametercrater Atget exposes LRM in both its ejecta and its central peaks, as does the 41-km-diameter crater Apollodorus. The similarly sized Cunningham (37 km in diameter)exposes LRM only in its central peaks. This variation in LRM exposure suggests anuneven infill of HRP in the basin interior.

Many impacts are superposed on the smooth plains, and someof them were sufficiently large to excavate through the HRP layerand expose the underlying LRM (Murchie et al., 2008), as indicatedin Fig. 12. The largest of these superposed craters is Atget (100 kmin diameter, 25.7�N, 166.1�E). This crater exposed substantialamounts of LRM, both in its ejecta and in its central peaks. Themaximum excavation depth of ejecta from Atget implies thatLRM is present <6.3 km below the surface, and the minimum depthof origin of the central peaks indicates that LRM is present at adepth of 10 km. If these materials were derived from the sameLRM layer, the subsurface unit could be at least 3.7 km in thickness.

Apollodorus (30.6�N, 163.0�E), the 41-km-diameter crater cen-tral to the Pantheon Fossae graben complex, exposes LRM in itsfloor, rim, and central peaks (Murchie et al., 2008), though severalsimilarly sized craters (e.g., Cunningham, 37 km in diameter) ex-pose LRM only in their central peaks. The maximum depth of exca-vation and minimum depth of central peak origin for Apollodorusare 2.8 and 4 km, respectively. The differences in excavated mate-rial between similarly sized craters suggest an uneven depth toHRP inside Caloris. The fact that some 37–40-km-diameter cratersexpose LRM in their central peaks but not in their floors or ejectalimits the thickness of the HRP infill near the center of Caloris tobetween 2.5 and 4 km. Volcanic flooding of lunar impact basinsby mare deposits shows similar trends and thicknesses (e.g., DeHon, 1979; Head, 1982; Head and Wilson, 1992; Budney andLucey, 1998; Thomson et al., 2009).

4.4. LRM centers

LRM covers at least 15% of Mercury’s surface and typicallyoccurs as diffuse regional deposits or in concentrated centers asso-ciated with large craters and basins (Robinson et al., 2008; Blewettet al., 2009; Denevi et al., 2009). Its morphological characteristicsand association with large craters and basins are consistent withexcavation from depth by impacts (Denevi et al., 2009). LRM cen-ters can be seen in several locations across the planet; however,not all large craters and basins have excavated LRM, indicating aheterogeneous distribution in the crust (Denevi et al., 2009). Thelarge (P80-km diameter) impact structures that are associatedwith LRM centers have maximum depths of excavation rangingfrom 5 km to in excess of 22 km (or possibly exceeding 75 km inthe case of Caloris, if these scaling relationships are valid for a ba-sin of that size). Many of the LRM-excavating basins are pre-Tolsto-jan in age, indicating that the time of original emplacement of LRMinto the subsurface crust must be >4.0 Ga.

In some areas, LRM is seen in the ejecta of smaller craters(4–40 km in diameter). Many of these occurrences are seen withinbasins (e.g., Caloris, Titian), where the LRM was originally broughtto the surface or relatively shallow subsurface from greater depthby the basin-forming event. Even if this LRM were subsequentlyburied by volcanism, it thereafter resided sufficiently close to thesurface to be excavated by smaller craters.

5. Discussion

We use the depth bounds calculated in Section 4 to constructschematic cross-sections depicting the overall stratigraphy of thefive analyzed areas (Fig. 13). We make the simplifying assumptionthat subsurface structure is laterally uniform over each area con-sidered and that only the layers observed are present. Comparingthe cross-sections allows a preliminary assessment of color unitsand their vertical and horizontal distributions. Dotted lines repre-sent the pre-basin position of units: the pre-Titian subsurface loca-tion of the LRM, from the maximum excavation depth of Titian;

Fig. 13. Schematic cross-sections depicting the local stratigraphy of Mercury in the five areas detailed in this study. In the Titian and Homer basin cases, the dashed linesrepresent the possible original depth of a given material before modification by a large basin.

Fig. 14. A comparison of RM spectra to pyroclastic deposits and HRP. The spectrahave been corrected to 30�, 0�, 30� incidence, emission, and phase angle,respectively, using the most up-to-date photometric correction (Domingue et al.,2010; D.L. Domingue, personal communication, 2010). The three dashed linesrepresent three unique RM spectra taken from the areas of interest in this study.The two dotted lines represent the areas identified as red spots (RS 1,2) by Blewettet al. (2009) that may have primary associations with impact craters Moody andNarvi. The HRP is from Caloris. The circum-Caloris pyroclastic example (gray line) isa location identified as RS 3 by Blewett et al. (2009). The pyroclastic-Lermontovexample is located within the crater Lermontov (crater center at 15.2�N, 311.9�E).

220 C.M. Ernst et al. / Icarus 209 (2010) 210–223

and the pre-H5 location of the RM exposed in the central peak,unaffected by the formation of Homer basin.

The smooth plains units analyzed in this study all infill largecraters or basins. The three IP units are 61 km thick, and the HRPunit that fills the much larger Caloris basin is at least 2.7 km thickat the center of the basin. Because these plains are believed to bevolcanic in origin and the crater floors may be topographically un-even, the thickness of volcanic plains across each host crater or ba-sin is likely to be variable. These thicknesses are consistent withthose inferred by Head et al. (2009a), which ranged from manyhundreds of meters to several kilometers, and are comparable tothose of many lunar maria (e.g., De Hon, 1979; Head, 1982; Headand Wilson, 1992; Budney and Lucey, 1998; Thomson et al., 2009).

In three of the areas examined (Rudaki, Homer, and Heming-way), there is evidence of RM originating at depths between 1and 4 km. The Homer and Rudaki areas are �850 km apart onthe surface; however, it is unknown whether the RM observed inthe two areas is related. In the Homer and Hemingway regions,the presence of RM in the rims of craters with a range of diameterssuggests that the underlying RM layer may be several kilometersthick. Radar studies of lunar pyroclastic deposits have yieldedthicknesses ranging from 5 to 20 m (Zisk et al., 1974; Hawkeet al., 2009). If these thicknesses are typical of large pyroclasticdeposits, it is unlikely that pyroclastic deposits of several-kilome-ter thickness were emplaced on Mercury and subsequently buried,despite their spectral similarity to the ‘‘red spot” pyroclastic depos-its elsewhere on the planet’s surface.

Of the four ‘‘red spots” identified by Blewett et al. (2009), two(contained within the craters Moody and Navoi) may have primaryassociations with impact craters, like the examples of RM detailed inthis study. The spectra of these two occurrences of RM are similar tothose of the HRP, suggesting that they may represent deposits ofHRP-like material (Blewett et al., 2009). The limited spatial resolu-tion of the current images does not permit the determination ofwhether these deposits represent post-crater flooding or pre-cratervolcanic units that were covered or intrusive and subsequentlyexcavated from depth by impacts. The deposits in Moody and Navoiwarrant further investigation after the acquisition of higher-resolu-tion images from the MESSENGER mission orbital phase.

The spectra of the Blewett et al. (2009) RM as well as the areasof RM described in this study are shown in Fig. 14, along with anexample of HRP from the interior of Caloris basin and examplesof pyroclastic deposits. Although there are slight variations amongthe RM spectra, they all have spectral characteristics (albedo andslope) that are similar to the HRP and distinct from the pyroclasticexamples. If these subsurface RM layers do represent buried or

C.M. Ernst et al. / Icarus 209 (2010) 210–223 221

intrusive volcanic material, they are evidence for large-scaledeposits. If they were deposited at the surface, these layers wouldrepresent an older generation of smooth plains.

Because the observed RM units occur as small deposits thathave been emplaced during subsequent impact events, it is diffi-cult to constrain their original ages. In at least two of the mainexamples, RM has been excavated from beneath subsequent IP fill,limiting the RM to ages older than those of these plains. We canplace additional constraints on the age of the RM in Homer, sinceits incorporation into the peak ring requires that it predate the ba-sin. Homer is highly degraded and was classified as pre-Tolstojanin age from Mariner 10 mapping (Spudis and Guest, 1988), indicat-ing that the RM in its peak ring, whether intrusive or extrusive,must be >4.0 Ga.

In the Rudaki, Titian, Hemingway, and Caloris areas examinedabove, LRM represents the deepest material observed. The CalorisLRM in particular is excavated by craters of many sizes, indicatingthat it is present over a depth range of several kilometers. Theseobservations suggest some possible scenarios: (a) any non-LRMlayers intermixed with the subsurface LRM are very thin and arenot detectable at the resolutions available; (b) non-LRM layersare present but exist far enough below LRM deposits that theywere not tapped by impacts; or (c) non-LRM layers are not presentin these stratigraphic columns. If this last option is true, it couldmean the LRM represents basement material uncovered by ba-sin-forming impacts. The higher-resolution images acquired dur-ing the MESSENGER orbital mission phase will permit thedetailed search for spectrally distinct craters in LRM centers anda closer examination of craters exposing LRM in order to furtherinvestigate these scenarios.

The derived stratigraphies imply that large parts of Mercury’ssurface have been built up over time by a series of volcanic flows,in agreement with several recent studies (e.g., Head et al., 2008,2009b; Murchie et al., 2008; Denevi et al., 2009). In some areas,accumulations of volcanic material reach thicknesses of severalkilometers, at least to a depth of �5 km. The subsurface RM mayrepresent an older generation of HRP-like smooth plains, at leastsome of which are pre-Tolstojan in age. The craters uncoveringRM might, therefore, expose some of the earliest volcanism onMercury, analogous to lunar cryptomare material, ancient marebasalts that are hidden beneath highland-rich ejecta (e.g., Headand Wilson, 1992; Blewett et al., 1995; Antonenko et al., 1995).In this way, these craters are analogous to the Moon’s dark-halocraters (e.g., Schultz and Spudis, 1979; Hawke and Bell, 1981; Belland Hawke, 1984), although most lunar DHCs are smaller thanthese examples on Mercury (Head et al., 2009a). The size differencemay be due to the difference in the overlying material (ejecta de-posit for the DHCs, volcanic plains for the RM), but the MESSENGERflyby images are not of sufficient resolution or coverage to rule outoccurrences on the scale of lunar DHCs. Another analogy to lunarDHCs is the previously mentioned T4 crater near Titian, whichuncovers older IP from beneath an LRM ejecta deposit, producinga bright-halo crater. This is a more direct analogy, since the oldervolcanic material is excavated from beneath ejecta, not later volca-nic flows.

Knowledge of Mercury’s crustal stratigraphy and volcanic his-tory is important for interpreting several types of geological andgeophysical observations that will be made during MESSENGER’sorbital phase. On the most basic level, understanding the thick-nesses of individual volcanic units (both surface and buried) helpsto define the extent and role of volcanism in Mercury’s crustalevolution, which in turn relates to the planet’s thermal history.Additionally, these thicknesses will be valuable inputs to litho-spheric loading models designed to test ideas for the currenttopography within major basins and for the formation of tectonicfeatures. Older generations of volcanism can be catalogued and

examined, in order to search for temporal changes in magma com-position (for example, are these ancient volcanic materials all RMin composition, or do they also span the range of compositionsseen on the surface of Mercury?).

If, as proposed by Robinson and Lucey (1997) and Denevi et al.(2009), LRM has a relatively high proportion of opaque mineralphases such as Fe- or Ti-bearing oxides, it will be denser thanthe average Mercury crustal material. Therefore, the LRM centerscould be locations of positive gravity anomalies. Likewise, an unex-plained positive gravity anomaly, for example one not correlatedwith an impact basin, as with lunar mascons (e.g., Muller and Sjo-gren, 1968), might be a candidate location for an unexposed sub-surface LRM deposit. If the LRM darkening agent is capable ofretaining remanent magnetization, buried LRM bodies, particularlyif emplaced over limited timescales, may also produce measurablecrustal magnetic anomalies.

6. Conclusions

The craters of Mercury provide critical windows into the sub-surface, through which the stratigraphy of the upper crust can beexplored. Craters excavating material that is spectrally distinctfrom the surrounding surface provide a means to bound the depthof origin of the spectrally distinct ejecta and central peak struc-tures. The resulting stratigraphic knowledge provides insights intothe origin and evolution of Mercury’s crust.

The examination of several small areas on Mercury reveals thecomplex nature of the local stratigraphy. The surface smooth plainsdeposits examined range from less than 1 km to �2.7 km in thick-ness and are likely volcanic in origin. Examples of RM, some ofwhich may be several kilometers in thickness, were found to bespectrally similar to the HRP, suggesting that they may representan older generation of volcanic plains that were subsequently cov-ered by intermediate plains. Some of the RM and LRM depositswere excavated by pre-Tolstojan basins, indicating relatively oldcrustal emplacement ages (>4.0 Ga). Exposures of RM by impactsmay be uncovering Mercury’s equivalent to lunar cryptomaria,although most of these Mercury instances are excavated from be-neath subsequent volcanic flows rather than basin ejecta. All ofthese observations highlight the importance of volcanism on Mer-cury and support the origin of large regions of the upper crust bythe buildup of volcanic units (e.g., Denevi et al., 2009), at least toa depth of �5 km. This conclusion points to Mercury as being char-acterized by a volcanic crustal component that is more substantialin thickness, volume, and areal distribution than the mare basaltphase of crustal formation on the Moon (e.g., Head and Wilson,1992).

The five areas examined in this study provide a preliminaryview of Mercury’s crustal stratigraphy. Further refinements arecontingent on global high-resolution MDIS coverage of the planet,along with targeted spectra and topography, which will be ac-quired during MESSENGER’s orbital mission phase beginning in2011. Improvements in coverage and resolution will directly en-hance the level of detail to which we can determine Mercury’scrustal stratigraphy.

Acknowledgments

The MESSENGER project is supported by the NASA DiscoveryProgram under contracts NASW-00002 to the Carnegie Institutionof Washington and NAS5-97271 to the Johns Hopkins UniversityApplied Physics Laboratory. Support from the NASA MESSENGERParticipating Scientist Program is also greatly appreciated(#NNX08AN29G, principal investigator D.T.B.). The authors thank

222 C.M. Ernst et al. / Icarus 209 (2010) 210–223

two anonymous reviewers for their detailed and constructive com-ments, which have improved the quality of this manuscript.

References

Ahrens, T.J., O’Keefe, J.D., 1977. Equations of state and impact-induced shock-waveattenuation on the Moon. In: Roddy, D.J., Pepin, R.O., Merrill, R.B. (Eds.), Impactand Explosion Cratering. Pergamon Press, New York, pp. 639–656.

Antonenko, I., Head, J.W., Mustard, J.F., Hawke, B.R., 1995. Criteria for the detectionof lunar cryptomaria. Earth Moon Planets 69, 141–172.

Austin, M.G., Thomsen, J.M., Ruhl, S.F., Orphal, D.L., Schultz, P.H., 1980. Calculationalinvestigation of impact cratering dynamics: Material motions during the cratergrowth period. Proc. Lunar Planet. Sci. Conf. 11, 2325–2345.

Barnouin-Jha, O.S., Yamamoto, S., Toriumi, T., Sugita, S., Matsui, T., 2007. Non-intrusive measurements of crater growth. Icarus 188, 506–521.

Barnouin-Jha, O.S., Ernst, C.M., Heineck, J.T., Sugita, S., Satoru, Y., 2009. Cratergrowth measured in the laboratory. GSA Abstracts with Programs 41, 364.

Bell, J.F., Hawke, B.R., 1984. Lunar dark-haloed impact craters – Origin andimplications for early mare volcanism. J. Geophys. Res. 89, 6899–6910.

Blewett, D.T., Hawke, B.R., Lucey, P.G., Taylor, G.J., Jaumann, R., Spudis, P.D., 1995.Remote sensing and geologic studies of the Schiller–Schickard region of theMoon. J. Geophys. Res. 100, 16959–16977.

Blewett, D.T., Hawke, B.R., Lucey, P.G., Robinson, M.S., 2007. A Mariner 10 colorstudy of mercurian craters. J. Geophys. Res. 112, E02005. doi:10.1029/2006JE002713.

Blewett, D.T., Robinson, M.S., Denevi, B.W., Gillis-Davis, J.J., Head, J.W., Solomon,S.C., Holsclaw, G.M., McClintock, W.E., 2009. Multispectral images of Mercuryfrom the first MESSENGER flyby: Analysis of global and regional color trends.Earth Planet. Sci. Lett. 285, 272–282.

Blewett, D.T., Denevi, B.W., Robinson, M.S., Ernst, C.M., Purucker, M.E., Solomon, S.C.,2010. The apparent lack of lunar-like swirls on Mercury: Implications for theformation of lunar swirls and for the agent of space weathering. Icarus 209,239–246.

Budney, C.J., Lucey, P.G., 1998. Basalt thickness in Mare Humorum: The craterexcavation method. J. Geophys. Res. 103, 16855–16870.

Butler, B.J., Muhleman, D.O., Slade, M.A., 1993. Mercury: Full-disk radar images andthe detection and stability of ice at the north pole. J. Geophys. Res. 98, 15003–15023.

Cintala, M.J., 1992. Impact-induced thermal effects in the lunar and mercurianregoliths. J. Geophys. Res. 97, 947–973.

Cintala, M.J., Grieve, R.A.F., 1998a. Scaling impact melting and crater dimensions:Implications for the lunar cratering record. Meteorit. Planet. Sci. 33, 889–912.

Cintala, M.J., Grieve, R.A.F., 1998b. Erratum: Scaling impact-melting and craterdimensions: Implications for the lunar cratering record. Meteorit. Planet. Sci.33, 1343.

Cintala, M.J., Wood, C.A., Head, J.W., 1977. The effects of target characteristics onfresh crater morphology: Preliminary results for the Moon and Mercury. Proc.Lunar. Sci. Conf. 8, 3409–3425.

Croft, S.K., 1985. The scaling of complex craters. J. Geophys. Res. 90, C828–C842.De Hon, R.A., 1979. Thickness of western mare basalts. Proc. Lunar Planet. Sci. Conf.

10, 2935–2955.Denevi, B.W., Robinson, M.S., 2008. Mercury’s albedo from Mariner 10: Implications

for the presence of ferrous iron. Icarus 197, 239–246.Denevi, B.W., and 10 colleagues, 2009. The evolution of Mercury’s crust: A global

perspective from MESSENGER. Science 324, 613–618.Domingue, D.L., Murchie, S.L., Chabot, N.L., Denevi, B.W., Izenberg, N.R., Holsclaw,

G.M., 2010. Analysis of the dedicated spectral photometric observations fromMESSENGER’s third Mercury flyby. Lunar Planet. Sci., 41. Abstract 2455.

Dzurisin, D., 1977. Mercurian bright patches: Evidence for physio-chemicalalteration of surface material? Geophys. Res. Lett. 4, 383–386.

Dzurisin, D., 1978. The tectonic and volcanic history of Mercury as inferred fromstudies of scarps, ridges, troughs, and other lineaments. J. Geophys. Res. 83,4883–4906.

Fischer, E.M., Pieters, C.M., 1994. Remote determination of exposure degree and ironconcentration of lunar soil using VISNIR spectroscopic methods. Icarus 111,475–488.

Gault, D.E., Heitowit, E.D., 1963. The partitioning of energy for hypervelocity impactcraters formed in rock. Proc. Sixth Hypervelocity Impact Symp., vol. 2. FirestoneTire and Rubber Co., Akron, OH, pp. 419–456.

Gault, D.E., Quaide, W.L., Oberbeck, V.R., 1968. Impact cratering mechanics andstructures. In: French, B.M., Short, N.M. (Eds.), Shock Metamorphism of NaturalMaterials. Mono Book Corp., Baltimore, MD, pp. 87–99.

Grieve, R.A.F., Cintala, M.J., 1992. An analysis of differential impact melt-craterscaling and implications for the terrestrial impact record. Meteoritics 27, 526–538.

Hapke, B., Danielson Jr., G.E., Klaasen, K., Wilson, L., 1975. Photometric observationsof Mercury from Mariner 10. J. Geophys. Res. 80, 2431–2443.

Hapke, B., Christman, C., Rava, B., Mosher, J., 1980. A color-ratio map of Mercury.Proc. Lunar Planet. Sci. Conf. 11, 817–821.

Harmon, J.K., 2007. Radar imaging of Mercury. Space Sci. Rev. 132, 307–349.Hartmann, W.K., and 10 colleagues, 1981. Chronology of planetary volcanism by

comparative studies of planetary cratering. In: Basaltic Volcanism Study Project,Basaltic Volcanism on the Terrestrial Planets. Pergamon Press, New York, pp.1048–1127.

Hawke, B.R., Bell, J.F., 1981. Remote sensing studies of lunar dark-halo impactcraters: Preliminary results and implications for early volcanism. Proc. LunarPlanet. Sci. Conf. 12, 665–678.

Hawke, B.R., and 10 colleagues, 2009. Remote sensing studies of pyroclastic depositsin the Mare Humorum region. Lunar Planet. Sci., 40. Abstract 1146.

Hawkins III, S.E., and 24 colleagues, 2007. The Mercury Dual Imaging System on theMESSENGER spacecraft. Space Sci. Rev. 131, 247–338.

Head, J.W., 1982. Lava flooding of ancient planetary crusts: Geometry, thickness,and volumes of flooded lunar impact basins. Moon Planets 26, 61–88.

Head, J.W., Wilson, L., 1992. Lunar mare volcanism: Stratigraphy, eruptionconditions, and the evolution of secondary crusts. Geochim. Cosmochim. Acta56, 2155–2175.

Head, J.W., and 10 colleagues, 2008. Volcanism on Mercury: Evidence from the firstMESSENGER flyby. Science 321, 69–72.

Head, J.W., and 12 colleagues, 2009a. Volcanism on Mercury: Evidence from the firstMESSENGER flyby for extrusive and explosive activity and the volcanic origin ofplains. Earth Planet. Sci. Lett. 285, 227–242.

Head, J.W., and 12 colleagues, 2009b. Evidence for intrusive activity on Mercuryfrom the first MESSENGER flyby. Earth Planet. Sci. Lett. 285, 251–262.

Holsapple, K.A., 1993. The scaling of impact processes in planetary sciences. Annu.Rev. Earth Planet. Sci. 21, 333–373.

Horedt, G.P., Neukum, G., 1984. Planetocentric versus heliocentric impacts in theJovian and Saturnian satellite system. J. Geophys. Res. 89, 10405–10410.

Kerber, L., Head, J.W., Solomon, S.C., Murchie, S.L., Blewett, D.T., Wilson, L., 2009.Explosive volcanic eruptions on Mercury: Eruption conditions, magma volatilecontent, and implications for mantle volatile abundances. Earth Planet. Sci. Lett.285, 263–271.

Kiefer, W.S., Murray, B.C., 1987. The formation of Mercury’s smooth plains. Icarus72, 477–491.

Le Feuvre, M., Wieczorek, M.A., 2008. Nonuniform cratering of the terrestrialplanets. Icarus 197, 291–306.

Marchi, S., Morbidelli, A., Cremonese, G., 2005. Flux of meteoroid impacts onMercury. Astron. Astrophys. 431, 1123–1127.

McClintock, W.E., and 12 colleagues, 2008. Spectroscopic observations of Mercury’ssurface reflectance during MESSENGER’s first Mercury flyby. Science 321, 62–65.

McCord, T.B., Adams, J.B., 1972. Mercury: Interpretation of optical observations.Icarus 17, 585–588.

Melosh, H.J., 1989. Impact Cratering: A Geologic Process. Oxford University Press,New York. 245pp.

Muller, P.M., Sjogren, W.L., 1968. Mascons: Lunar mass concentrations. Science 161,680–684.

Murchie, S.M., and 10 colleagues, 2008. Geology of the Caloris basin, Mercury: Aview from MESSENGER. Science 321, 73–76.

Murray, B.C., Belton, M.J.S., Danielson, G.E., Davies, M.E., Gault, D.E., Hapke, B.,O’Leary, B., Strom, R.G., Suomi, V., Trask, N., 1974. Mercury’s surface:Preliminary description and interpretation from Mariner 10 pictures. Science185, 169–179.

Oberbeck, V.R., 1975. The role of ballistic erosion and sedimentation in lunarstratigraphy. Rev. Geophys. Space Phys. 13, 337–362.

O’Keefe, J.D., Ahrens, T.J., 1975. Shock effects from a large impact on the Moon. Proc.Lunar Sci. Conf. 6, 2831–2844.

O’Keefe, J.D., Ahrens, T.J., 1977. Impact-induced energy partitioning, melting, andvaporization on terrestrial planets. Proc. Lunar. Sci. Conf. 8, 3357–3374.

Pierazzo, E., Vickery, A.M., Melosh, H.J., 1997. A reevaluation of impact meltproduction. Icarus 127, 408–423.

Pike, R.J., 1980. Formation of complex impact craters: Evidence from Mars and otherplanets. Icarus 43, 1–19.

Pike, R.J., 1988. Geomorphology of impact craters on Mercury. In: Vilas, F.,Chapman, C.R., Mathews, M.S. (Eds.), Mercury. University of Arizona Press,Tucson, pp. 165–273.

Rava, B., Hapke, B., 1987. An analysis of the Mariner 10 color ratio map of Mercury.Icarus 71, 397–429.

Robinson, M.S., Lucey, P.G., 1997. Recalibrated Mariner 10 color mosaics:Implications for mercurian volcanism. Science 275, 197–200.

Robinson, M.S., Taylor, G.J., 2001. Ferrous oxide in Mercury’s crust and mantle.Meteorit. Planet. Sci. 36, 841–847.

Robinson, M.S., and 12 colleagues, 2008. Reflectance and color variations onMercury: Regolith processes and compositional heterogeneity. Science 321,66–69.

Schmidt, R.M., Housen, K.R., 1987. Some recent advances in the scaling of impactand explosion cratering. Int. J. Impact Eng. 5, 543–560.

Schultz, P.H., 1977. Endogenic modification of impact craters on Mercury. Phys.Earth Planet. Inter. 15, 202–219.

Schultz, P.H., 1988. Cratering on Mercury: A relook. In: Vilas, F., Chapman, C.R.,Mathews, M.S. (Eds.), Mercury. University of Arizona Press, Tucson, pp. 274–335.

Schultz, P.H., Spudis, P.D., 1979. Evidence for ancient mare volcanism. Proc. LunarPlanet. Sci. Conf. 10, 2899–2918.

Slade, M.A., Butler, B.J., Muhleman, D.O., 1992. Mercury radar imaging: Evidence forpolar ice. Science 258, 635–640.

Solomon, S.C., McNutt Jr., R.L., Gold, R.E., Domingue, D.L., 2007. MESSENGER missionoverview. Space Sci. Rev. 131, 3–39.

Spudis, P.D., Guest, J.E., 1988. Stratigraphy and geologic history of Mercury. In: Vilas,F., Chapman, C.R., Mathews, M.S. (Eds.), Mercury. University of Arizona Press,Tucson, pp. 118–164.

C.M. Ernst et al. / Icarus 209 (2010) 210–223 223

Stewart, S.T., Valiant, G.J., 2006. Martian subsurface properties and crater formationprocesses inferred from fresh impact crater geometries. Meteorit. Planet. Sci. 41,1509–1537.

Stöffler, D., Gault, D.E., Wedekind, J., Polkowski, G., 1975. Experimentalhypervelocity impact into quartz sand: Distribution and shock metamorphismof ejecta. J. Geophys. Res. 80, 4062–4077.

Strom, R.G., 1977. Origin and relative age of lunar and mercurian intercrater plains.Phys. Earth Planet. Inter. 15, 156–172.

Thompson, S.L., Lauson, H.S., 1972. Improvements in the CHART D radiation-hydrodynamic code III: Revised analytic equations of state. Technical Report SC-RR-71 0714. Sandia National Laboratories, Albuquerque, NM.

Thomson, B.J., Grosfils, E.B., Bussey, D.B.J., Spudis, P.D., 2009. A new technique forestimating the thickness of mare basalts in Imbrium basin. Geophys. Res. Lett.36, L12201. doi:10.1029/2009GL037600.

Trask, N.J., Strom, R.G., 1976. Additional evidence of mercurian volcanism. Icarus 28,559–563.

Vilas, F., 1985. Mercury: Absence of crystalline Fe2+ in the regolith. Icarus 64, 133–138.

Vilas, F., 1988. Surface composition of Mercury from reflectance spectrophotometry.In: Vilas, F., Chapman, C.R., Mathews, M.S. (Eds.), Mercury. University of ArizonaPress, Tucson, pp. 59–76.

Watters, T.R., Murchie, S.L., Robinson, M.S., Solomon, S.C., Denevi, B.W., André, S.L.,Head, J.W., 2009. Emplacement and tectonic deformation of smooth plains inthe Caloris basin, Mercury. Earth Planet. Sci. Lett. 285, 309–319.

Watters, W.A., Zuber, M.T., Hager, B.H., 2009. Thermal perturbations caused by largeimpacts and consequences for mantle convection. J. Geophys. Res. 114, E02001.doi:10.1029/2007JE002964.

Wilhelms, D.E., 1976. Mercurian volcanism questioned. Icarus 28, 551–558.Zisk, S.H., Pettengill, G.H., Catuna, G.W., 1974. High-resolution radar maps of the

lunar surface at 3. 8-cm wavelength. Moon 10, 17–50.