Icarus 223 (2013) 749765

Contents lists available at SciVerse ScienceDirect

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

Geology and petrology of enormous volumes of impact melt on the Moon:

A case study of the Orientale basin impact melt sea
William M. Vaughan a,, James W. Head a, Lionel Wilson b, Paul C. Hess a
a
Department of Geological Sciences, Brown University, Providence, RI 02912, USA
b
Lancaster Environment Centre, Lancaster University, Lancaster LA1 4YQ, UK

a r t i c l e i n f o a b s t r a c t

Article history: Lunar basin-forming impacts produce enormous volumes (>105 km3) of impact melt. All known basin-
Received 6 August 2012 forming impacts combined may produce 108 km3 of impact melt, 1/20th the volume of the lunar crust.
Revised 11 January 2013 Despite their volumetric importance, the geology and petrology of massive deposits of impact melt on the
Accepted 16 January 2013
Moon have been little studied, in part because most basin impact melt deposits are old and have been
Available online 30 January 2013
obscured or buried by subsequent impact cratering and mare inll. We investigate the geology and model
the petrology of fresh massive impact melt deposits in the relatively young 930 km diameter Orientale
Keywords:
basin. Models of impact melt production combined with geologic analyses based on new LOLA topo-
Moon
Cratering
graphic data suggest that most of the impact melt (2/3) produced by the Orientale-forming impact
Impact processes occurs in a 15 km thick impact melt sheet (better described as an impact melt sea) 350 km in diameter
with a volume of 106 km3. We anticipate that the Orientale melt sea has undergone large-scale igneous
differentiation, since terrestrial impact melt sheets (such as Manicouagan, Sudbury, and Morokweng) less
than a tenth of the thickness and a hundredth of the volume of the Orientale melt sea have differentiated.
We develop a model for the cumulate stratigraphy of the solidied Orientale impact melt sea. A modeled
cumulate stratigraphy (occurring below a quench crust and anorthositic fallback breccia) with an 8 km
thick layer of norite overlying a 4 km layer of pyroxenite and a basal 2 km thick layer of dunite pro-
duced by equilibrium crystallization of a homogenized melt sea, consistent with vigorous convection in
that melt sea, is supported by remotely-sensed norite excavated by the central peak of Maunder crater
from 4 km depth. Generally, we predict that very large basin-forming impacts, including the South
Pole-Aitken (SPA) basin-forming impact, produce melt seas with a cumulate stratigraphy similar to that
of the Orientale melt sea. Impact melt differentiation may explain apparently anomalous lithologies exca-
vated in the SPA basin interior. We note that impact melt differentiates are slow-cooled (the Orientale
melt sea took on the order of 105 years to solidify) and, if meteoritic siderophiles fractionate into metal
or sulde layers, may not be siderophile-enriched; therefore, impact melt differentiates may pass for
pristine highland plutonic rocks in the lunar sample suite. These predictions can be tested with current
and future mission data.
2013 Elsevier Inc. All rights reserved.

1. Introduction ren et al., 1996) to produce glasses or cryptocrystalline lithologies

Simple and complex lunar craters <150 km in diameter contain
relatively small volumes (104 km3) of impact melt (Cintala and
Grieve, 1998). The geology of the small impact melt deposits in these
craters is relatively well understood: impact melt oods crater oors
and drapes crater central peaks, walls, and rims (Bray et al., 2010;
Howard et al., 1974; Howard and Wilshire, 1973, 1975; Plescia
and Cintala, 2012; Spray and Thompson, 2008); melt distribution
is controlled by impact angle, post-impact ejection related to slump-
ing, and pre-existing topography (Hawke and Head, 1977). Thin
sheets of impact melt occurring in these craters freeze quickly (War-
Corresponding author. Fax: +1 401 863 3978.
E-mail address: William_Vaughan@brown.edu (W.M. Vaughan).
chemically similar to target rocks.
Field and experimental evidence shows that the volume of shock
melt produced by an impact scales approximately as the fourth
power of impact crater transient cavity diameter (Abramov et al.,
2012; Cintala and Grieve, 1998; Grieve and Cintala, 1992, 1997).
So although craters <150 km in diameter contain relatively small
volumes (104 km3) of impact melt, basin-forming impacts creat-
ing basins >300 km in diameter are predicted by these analytic scal-
ing relationships (Abramov et al., 2012; Cintala and Grieve, 1998;
Grieve and Cintala, 1992) as well as hydrocode models (Collins
et al., 2002; Ivanov, 2005) to produce enormous volumes
(>105 km3) of impact melt. The volumes of impact melt produced
by basin-forming impacts are well in excess of the volumes of many
terrestrial bodies of water and igneous intrusions (Fig. 1). In fact, all

0019-1035/$ - see front matter 2013 Elsevier Inc. All rights reserved.
http://dx.doi.org/10.1016/j.icarus.2013.01.017
750 W.M. Vaughan et al. / Icarus 223 (2013) 749765
Fig. 1. The volume of impact melt in a 30 km lunar crater, estimated using the
scaling laws of Cintala and Grieve (1998), and the volume of the Orientale impact
melt sea, calculated in Section 2, compared to selected terrestrial igneous intrusions
(Bonini, 1982; Nielsen, 2004) and bodies of water.
basin-forming impacts combined are modeled to produce on the or-
der of 108 km3 of impact melt (see Kring et al. (2012) and our Sec-
tion 4.4), so that as much as 1/20th of the lunar crust may be
impact melt.
Despite their volumetric importance, the geology and petrology
of massive impact melt deposits produced by basin-forming impacts
are poorly understood. This is in part because basin interiors where
massive impact melt deposits occur (Head, 1974; Melosh, 1989;
Spudis, 1993) have been obscured and modied by impact cratering
and mare basalt inll (Head and Wilson, 1992). For example, no im-
pact melt deposits are exposed in the mare basalt-ooded Imbrium,
Serenitatis, and Crisium basin interiors (Head et al., 1993). More-
over, although the thick (>1 km) sheets of impact melt predicted
to occur in lunar basins (Cintala and Grieve, 1998) cool slowly and
may undergo igneous differentiation (Cintala and Grieve, 1998;
Grieve et al., 1991), the cumulate stratigraphy of these thick impact
melt sheets has never been modeled, although several authors have
carefully considered some particular petrologic consequences of im-
pact melt differentiation (Morrison, 1998; Warren et al., 1996).
Where on the Moon should we look to improve our understand-
ing of massive impact melt deposits? The relatively young (3.7
3.8 Ga) (Baldwin, 1974; Whitten et al., 2011) Orientale basin (Head,
1974; Howard et al., 1974; McCauley, 1977), which has relatively
few impact craters and little mare basalt inll (Greeley et al.,
1993; Whitten et al., 2010, 2011), is an ideal location to investigate
the distribution, thickness, and volume of massive impact melt
deposits. Moreover, the relatively well-constrained structure of
the farside crust (Wieczorek et al., 2006) and upper mantle (Khan
et al., 2006) probably melted by the Orientale impact provides nec-
essary compositional data (Shearer et al., 2006) to model the cumu-
late stratigraphy of an Orientale impact melt sheet and to consider
the petrologic implications of impact melt differentiation.
In this paper, we (1) investigate the distribution, thickness, and vol-
ume of impact melt deposits produced by the Orientale-forming im-
pact using models of impact melt production combined with recent
topographic data and (2) model the cumulate stratigraphy of a solidi-
ed, differentiated Orientale melt sheet, considering the general impli-
cations for the lunar highlands crust and the lunar sample suite.
2. Distribution, thickness, and volume of massive impact melt
deposits in Orientale
2.1. General geology of the Orientale basin
The deposits associated with the 930 km diameter Orientale
impact basin (Fig. 2) have been mapped and subdivided into three
units (Fassett et al., 2011; Head, 1974; Howard et al., 1974; McCau-
ley, 1977): the radially-textured Hevelius Formation, lying outside
the 930 km diameter Cordillera Ring; the Montes Rook Forma-
tion, comprising knobby terrain between the Cordillera Ring and
the 620 km diameter Outer Rook Ring; and the Maunder Forma-
tion, which occurs inside the Outer Rook Ring, around the 480 km
diameter Inner Rook Ring, and on the oor of the 350 km diam-
eter central depression. These three major rings (the Inner Rook
Ring, the Outer Rook Ring, and the Cordillera Ring) have each been
interpreted as the Orientale basin crater rim crest (see Spudis
(1993) for a review). We interpret the Orientale basin as a modied
peak-ring basin (Head, 2010), taking the Inner Rook Ring to be the
basin peak ring (Baker et al., 2011a, 2011b, 2012) and the Outer
Rook Ring to be the basin rim crest (Head, 1974, 2010; Head
et al., 1993). In the following, we take the current diameter of
the Outer Rook Ring (620 km) as the best estimate of the Orientale
transient cavity diameter, since there is no clear evidence for ter-
racing within this rim that would have substantially increased its
diameter (Head, 2010; Head, 2012).
2.2. Geology of impact melt-related facies in Orientale
The facies of the Maunder Formation (Fig. 3), named for the
prominent 55 km diameter complex crater Maunder superposed
on the north edge of the central depression of the Orientale basin,
have historically been interpreted as impact melt-related (Head,
1974; Head et al., 1993; Howard et al., 1974; Spudis, 1993). The
central smooth facies (Fig. 3, left), occurring within the inner
depression of the Orientale basin exposed through thin mare ll
(Head, 2010; Whitten et al., 2010, 2011), has been interpreted as
pure impact melt (Head, 1974; Howard et al., 1974; McCauley,
1977). Wrinkle ridges (Fig. 3), fractures, and polygonal cracks are
apparent in the smooth facies and under the overlying mare. At
an average distance of 175 km from the Orientale basin center
(Fig. 2), the topography abruptly rises 1.75 km along a series of
marginal normal faults to a corrugated and ssured rough facies
(Fig. 3, right). This outer corrugated, ssured facies has been inter-
preted as clast-rich impact melt formed from mixing of pure im-
pact melt with brecciated debris from the Orientale impact
(Head, 1974; Head et al., 1993; Howard et al., 1974; McCauley,
1977).
2.3. How much impact melt was produced by the Orientale-forming
impact?
Impact melt volume scaling relationships adjusted for lunar
gravity derived by Cintala and Grieve (1998) predict that a chon-
dritic projectile with an asteroidal velocity of 10 km/s and a trajec-
tory normal to the lunar surface giving rise to an Orientale-sized
transient cavity 620 km in diameter produces 6  106 km3 of
impact melt. This volume is probably an overestimate: many
authors have suggested that the Orientale-forming impact was ob-
lique (Schultz and Papamarcos, 2010; Scott et al., 1977; Wilhelms
et al., 1987) and oblique impacts are known to produce less melt
than vertical impacts of the same size (Pierazzo and Melosh,
2000). Transverse widening of the Cordillera Ring of Orientale
(Schultz et al., 2012) may indicate that the Orientale impactor
had an impact angle of 1530. The volume of melt produced by
an impact with an impact angle of 1530 is 1050% of the volume
of melt produced by a vertical impact (Pierazzo and Melosh, 2000);
taking 22.5, the center of this impact angle range, to be the most
likely impact angle of the Orientale-forming projectile in the
absence of additional constraints, we estimate that the
Orientale-forming impact produced 1.5  106 km3 (25 vol.% of
6  106 km3) of impact melt. We emphasize that the composition
and velocity of the Orientale-forming projectile are not well
 W.M. Vaughan et al. / Icarus 223 (2013) 749765 751

Fig. 2. Lunar Orbiter Laser Altimeter (LOLA)-derived topographic map of the Orientale basin indicating major basin rings and geologic units.

Fig. 3. The smooth (left) and rough (right) facies of Orientales Maunder Formation from an LROC WAC monochrome mosaic of the Orientale basin.

constrained: variations in these parameters can change the impact
melt volume by a factor of two in the Cintala and Grieve (1998)
scaling model.
An independent estimate of the volume of impact melt pro-
duced by the Orientale-forming impact can be made using the im-
pact melt volume scaling relationship for oblique impacts derived
by Abramov et al. (2012). This relationship, which explicitly in-
cludes impact angle as a variable, predicts that a chondritic projec-
tile with an impact angle of 22.5 giving rise to an Orientale-sized
transient cavity 620 km in diameter produces about
2.8  106 km3 of melt, about twice the volume estimated above
by coupling the models of Cintala and Grieve (1998) and Pierazzo
and Melosh (2000). Discrepancies between impact melt volumes
predicted by the scaling relationships of Cintala and Grieve
(1998) and Abramov et al. (2012) have previously been noted by
Abramov et al. (2012). In order to assess the Cintala and Grieve
(1998) and Abramov et al. (2012) estimates of the volume of im-
pact melt produced by the Orientale-forming impact, we later
compare melt deposit thicknesses determined from these calcu-
lated melt volumes (Section 2.4) to an independent determination
of melt deposit thickness (Sections 2.52.7). For simplicity, in the
following section we consider only the melt volume of
1.5  106 km3 calculated from the scaling model of Cintala and
Grieve (1998).
2.4. Accounting for the 1.5  106 km3 of impact melt produced by the
Orientale-forming impact
Where does the estimated 1.5  106 km3 of impact melt pro-
duced by the Orientale-forming impact end up? A priori, because
the surface area of the Moon is only 3.8  107 km2, and since
there is no evidence that the Moon is covered in a 40 m (i.e.,
752 W.M. Vaughan et al. / Icarus 223 (2013) 749765

1.5  106 km3/3.8  107 km2) thick layer of Orientale impact melt,
most of the impact melt produced by the Orientale-forming impact
must occur in or near the Orientale basin.
This is in accordance with what is known about the process of
impact melt generation (Fig. 4). Impact melt forms in a melt cavity
interior to (although, in multi-ring basins, likely extending below)
what will become the crater transient cavity (Fig. 4a) (Cintala and
Grieve, 1998). Some of the generated impact melt is excavated in
forming the transient cavity; additional melt is ejected from the
crater interior during modication (Fig. 4b) (Hawke and Head,
1977; Osinski et al., 2011). Studies of large terrestrial craters such
as Sudbury demonstrate that a relatively small proportion of melt
(<10% by volume) is ejected from the crater interior during modi-
cation (Grieve and Cintala, 1992). Excavation ow modeling sug-
gests that up to 7080 vol.% of impact melt remains inside the
transient cavity in large lunar craters (Cintala and Grieve, 1998).
Based on these models, we estimate that 25 vol.% of impact melt
was ejected from the Orientale basin (perhaps incorporated in
Hevelius Formation ejecta). This implies that 1.1  106 km3
(75 vol.% of 1.5  106 km3) of impact melt produced in the Orien-
tale-forming impact remains inside the Orientale transient cavity
rim, here taken to be the Outer Rook Ring (Head, 1974, 2010,
2012).
Occurring inside the Outer Rook Ring are the impact melt-re-
lated rough and smooth facies (Fig. 3) of the Maunder Formation
described in Section 2.2 above. About 1.1  106 km3 of impact melt
must be apportioned between these two facies. How much impact
melt is contained in each facies?
The rough facies mantles topography similar to the peaks of the
Inner Rook Ring (Head, 1974), which stand up to 23 km above
surrounding terrain (Figs. 2 and 3). This constrains the volume of
impact melt contained in this facies: the relief of the mantled
topography is <1 km (Fig. 5), but the original relief was likely no
greater than the relief of the nearby peaks of the Inner Rook Ring,
23 km; therefore, the mantling deposit is 2 km thick at most.
The area of the rough faciesapproximately the area of the annu-
lus between the edge of the central depression (175 km from the
basin center) and the Outer Rook Ring (310 km from the basin cen-
ter)is 2  105 km2. If the clast-rich rough facies is 2 km thick
(which for 50 vol.% clasts is the equivalent of a 1 km thick layer
of pure impact melt) it contains 1  105 km3 (50 vol.% pure
melt  2 km  2  105 km2) of impact melt. The remaining
1.0  106 km3 (1.1  106 km3  1  105 km3) of impact melt
must be contained in the central smooth facies, thought to be a
pure impact melt sheet (Section 2.2). This smooth facies is
175 km in radius (Fig. 5) or 105 km2 in area, so the impact melt

Fig. 4. The material melted by the Orientale-forming impact (a), the Orientale melt cavity and transient cavity geometry immediately following hypervelocity impact (b), and
the nal crater geometry of Orientale (c) (Andrews-Hanna and Stewart, 2011; Hikida and Wieczorek, 2007) superposed on the lunar highlands crust and upper mantle (Khan
et al., 2006; Wieczorek et al., 2006; Zuber et al., 1994). Melt lls a truncated sphere with radius/truncation height taken as 0.9/1.1, following Barr and Citron, (2011), and a
total volume of 1.5  106 km3 (Cintala and Grieve, 1998), following our discussion in Sections 2.32.7. The basin transient cavity depth-diameter ratio is taken to be 1/10
(Wieczorek and Phillips, 1999) and the transient cavity diameter has been decreased by a factor of for clarity. The excavation ow curve (in blue) is given by the Maxwell Z-
model taking Z = 3 (Melosh, 1989). Note that (1) the transient cavity is gravitationally unstable, so the geometry shown in (b) is never fully realized at any time and (2) the
nal basin diameter (Fig. 2) is much larger than the transient cavity diameter in (b). (For interpretation of the references to color in this gure legend, the reader is referred to
the web version of this article.)
 W.M. Vaughan et al. / Icarus 223 (2013) 749765
Fig. 5. Average radial topographic prole of Maunder Formation facies. Radial proles crossing superposed Maunder and Kopff craters were excluded. See Baker et al. (2011a,
2011b, 2012) for methodology.
sheet must be on the order of 10 km thick to account for the
remaining 1.0  106 km3 of impact melt.
A 10 km thick impact melt sheet is not necessarily unrealistic:
the Sudbury Igneous Complex, a terrestrial impact melt sheet
occurring in a 250 km diameter crater (rim diameter 140 km),
is 2.53 km thick (Thierrault et al., 2002). More compellingly,
there is independent geologic evidence for a 10 km thick impact
melt sheet occurring in the Orientale basins central depression.
2.5. Geologic evidence for a 10 km thick melt sheet
As noted in Section 2.2, the surface of the smooth facies is on
average offset 1.75 km (Fig. 5) below the rough facies along mar-
ginal normal faults. This offset has been previously explained as
vertical subsidence caused by thermal stresses resulting from im-
pact-generated heat and uplift of crustal isotherms (Bratt et al.,
1985). This model predicts a 2 km drop over 100 km of radial dis-
tance (Fig. 14 from Bratt et al. (1985)). However, new, high-resolu-
tion LOLA altimetry (Smith et al., 2010) shows that the vertical
subsidence of the central depression is abrupt: the topography
drops 2 km over a radial distance of 20 km along the west edge
of the depression (Fig. 6), ve times more steeply than predicted by
the model of Bratt et al. (1985).
Wilson and Head (2011) suggest that the vertical subsidence of
the smooth facies, an impact melt sheet, is largely the result of ver-
tical shrinkage upon solidication and cooling of impact melt.
There is evidence for lateral shrinkage due to the same process:
fractures on the smooth facies resemble deformations on the sur-
Fig. 6. Four average radial topographic proles of Maunder Formation facies over different azimuthal ranges. Radial proles crossing superposed Maunder and Kopff craters
were excluded. See Baker et al. (2011a, 2011b, 2012) for methodology.
753
faces of terrestrial lava lakes (Barberi and Varet, 1970) known to
form due to lateral contraction upon solidication and cooling.
We now calculate the amount of vertical subsidence that results
from solidication and cooling of impact melt sheets on the Moon
in order to constrain the thickness and volume of the Orientale
melt sheet.
2.6. Estimating melt sheet thickness from vertical subsidence
The vertical shrinkage of a melt sheet upon solidication and
cooling is related to its total thickness. We now determine the nat-
ure of this relationship in order to calculate the thickness of the
Orientale melt sheet from its observed vertical subsidence. The
derivation follows Wilson and Head (2011).
A melt sheet emplaced at a temperature Te that solidies and
cools to the ambient temperature Tc will decrease in volume both
during solidication and while cooling to Tc. What is the magni-
tude of the total proportional decrease in volume? Volume is inver-
sely proportional to density, so dV/V = dq/q. A simple rule of
thumb is that the density of silicate melt increases (and volume
of melt decreases) by 10% upon solidication. This rule may be
misleading for melts of the lunar crust, since the density of anor-
thite melt increases only 5.5% upon solidication (based on cal-
culations of silicate liquid density following Nelson and
Carmichael (1979)); however, as will be shown in Section 3.3,
the Orientale melt sheet contains abundant mac material, so we
continue to use the 10% estimate. The proportional change in vol-
ume of a material at constant pressure, dV/V, due to subsequent
754 W.M. Vaughan et al. / Icarus 223 (2013) 749765
cooling is equal to adT, where a is the volumetric coefcient of
thermal expansion, 3  105 K1, for silicate rock (Turcotte and
Schubert, 2002). The dT is 1328 K, the difference between Te (taken
to be 1568 K, the average of the lava liquidus, 1713 K, and solidus,
1423 K, temperatures suggested by Williams et al. (2000)) and Tc
(taken to be the average of the lunar day and night temperatures,
240 K); therefore, the volume of the melt sheet must decrease
by 4% upon cooling. A 10% volume decrease compounded with
a 4% volume decrease results in approximately a 14% volume
decrease, since these percentages are both small.
What portion of this 14% volume decrease results in vertical
subsidence as opposed to lateral shrinkage? If the melt sheet con-
tracts equally in all directions, the percentage thickness decrease
will be one-third of the percentage volume decrease, about
4.7%. However, this is physically and geologically unrealistic:
the melt sheet does not cool at an equal rate at each of its bound-
aries, but cools primarily through radiation at its top, and will
therefore not contract equally in all directions; moreover, the con-
traction features (cracks and depressed polygons) present in the S
and SW portion of the smooth facies occupy only 12% of the basin
oor and are not consistent with a 9.3% decrease in surface area. A
more detailed treatment of lava lake solidication and cooling
(Worster et al., 1990, 1993) gives a better estimate of vertical sub-
sidence. The upper surface of a lava lake cools by radiation and
loses heat much faster than the other boundaries of the lava lake.
Initially, lava exposed to the surface solidies to form plates; these
plates must undergo mainly vertical contraction, since lateral con-
traction would cause lava to well up and the solidied plates to
founder into the liquid beneath, settling toward the base of the
pond and probably being partially remelted. Eventually a stable
crust will form; evidence from drilling of the Kilauea Iki lava lake
(Hardee, 1980) conrms that the stable crust forms well before
the mean temperature of the pond has approached the solidus.
The partitioning between vertical shrinkage and lateral shrinkage
from the 4% volume decrease due to the cooling of this solid crust
can be determined from the area of depressions and fractures on
the pond surface; for simplicity, we assume that between one-
third and all of the volume decrease upon cooling results in vertical
subsidence. The melt beneath the stable crust freezes progres-
sively; therefore, almost all of its 10% volume decrease upon solid-
ication will result in vertical contraction.
Fig. 7. Vertical subsidence in meters as a function of impact melt sheet thickness
for the 11.3% and 14% limits derived in this paper and in Wilson and Head (2011).
In short, we anticipate that nearly all the 10% volume decrease
upon solidication results in vertical subsidence and that between
one-third and all of the 4% volume decrease upon cooling results
in vertical subsidence. Therefore, we predict that the vertical
shrinkage of a lunar melt sheet is between 11.3% and 14% of
its depth. Fig. 7 shows the subsidence predicted as a function of
melt sheet depth for these 11.3% and 14% limits.
2.7. Thickness and volume of the Orientale melt sheet
The Orientale melt sheet has an observed average vertical relief
of 1.75 km (Fig. 5). As stated above, the observed vertical subsi-
dence is a factor of ve greater than predicted by Bratt et al.
(1985), so subsidence due to solidication and cooling of melt
must dwarf subsidence due to other thermal stresses. Therefore,
we treat all the observed vertical subsidence as the result of melt
solidication and cooling. The 11.3% and 14% limits due to cooling
and contraction derived above imply that the melt sheet was ini-
tially between 12.5 and 15.5 km thick (1.75 km/14% or 11.3%).
Lateral contraction features are observed in the smooth facies, so,
anticipating that much of the 4% volume decrease upon solidi-
cation resulted in lateral shrinkage, we choose the 11.3% bound
as more realistic and use the corresponding thickness of 15.5 km
as the thickness of the Orientale melt sheet in all following
calculations.
The diameter (350 km) and thickness (15.5 km) of the melt
sheet can be used to calculate the volume of the melt sheet. (Note
that the thickness of the solidied, cooled melt sheet is not
15.5 km, but 14 km i.e., 15.5 km  (10011.3)%; its volume is
84% of the original melt sheet volume.) Unfortunately, the geome-
try of the melt sheet is poorly constrained due to the difculty in
predicting the events following the formation of the transient cav-
ity (Fig. 4) leading up to cavity collapse (Head, 2010). The approx-
imation that the melt sheet is shaped like a spherical cap gives a
melt sheet volume of 8  105 km3. Inward movement and up-
ward translation during transient cavity collapse could potentially
produce a cylindrical melt sheet (Head et al., 1993) with a volume
of 1.5  106 km3. The true melt sheet volume is probably some-
where between the extremes of 8  105 km3 and
1.5  106 km3. These bounds are in striking agreement with the
melt sheet volume 106 km3 estimated from the melt production
model of Cintala and Grieve (1998) but are inconsistent with a
melt sheet volume of 1.9  106 km3 estimated from the melt pro-
duction model of Abramov et al. (2012) and our calculations in Sec-
tion 2.4. Therefore, we take 1.5  106 km3 as the volume of
impact melt produced by the Orientale-forming impact, since this
volume determined from the melt scaling model of Cintala and
Grieve (1998) is most consistent with our estimates of melt sheet
volume. We hereafter describe this 15.5 km thick melt sheet with a
volume of 106 km3 as a melt sea to emphasize its enormous
dimensions (Fig. 1).
In summary, most impact melt (2/3) produced by the Orien-
tale-forming impact occurs in a 15.5 km thick (now 14 km
thick) impact melt sea 350 km in diameter with a volume of
106 km3. We now model the petrology of this impact melt sea.
3. Cumulate stratigraphy of the Orientale melt sea
3.1. Did the Orientale melt sea differentiate?
Mounting petrologic evidence suggests that impact melt sheets
associated with the large terrestrial craters Manicouagan
(OConnell-Cooper and Spray, 2011), Sudbury (Grieve et al., 1991;
Thierrault et al., 2002; Zieg and Marsh, 2005), and Morokweng (Hart
et al., 2002) have undergone large-scale igneous differentiation
 W.M. Vaughan et al. / Icarus 223 (2013) 749765
resulting in cumulate stratigraphies similar to those of well-known
differentiated plutons such as the Skaergaard and the Stillwater. The
Sudbury Igneous Complex, the largest of these impact melt sheets,
has a thickness of 2.53 km and a volume of 104 km3 (Thierrault
et al., 2002). On the basis of our analysis in Section 2, the solidied
Orientale melt sea has a thickness of 14 km and a volume of
106 km3. Terrestrial impact melt sheets less than a tenth of the thick-
ness of the Orientale melt sea and a hundredth of its volume have under-
gone large-scale igneous differentiation. We therefore anticipate that
the Orientale melt sea too has undergone large-scale igneous
differentiation.
The idea that lunar impact melt sheets have undergone igneous
differentiation is by no means new: see, e.g., Basilevsky and Neu-
kum (2010), Cintala and Grieve (1998), Grieve et al. (1991), and
Morrison (1998). However, this idea deserves to be revisited now
more than ever, as recent and ongoing missions (the Moon Miner-
alogy Mapper (M3) instrument on Indias Chandrayaan-1 mission
and the Lunar Reconnaissance Orbiter (LRO) and Gravity Recovery
and Interior Laboratory (GRAIL) missions) collect high-resolution
compositional and geophysical data that make it possible to inves-
tigate the lunar highlands crust (and massive impact melt sheets)
in unprecedented detail.
3.2. Modeling the cumulate stratigraphy of the Orientale melt sea
In the following sections, we develop a model for the cumulate
stratigraphy of the Orientale melt sea (indeed, for any impact melt
sea formed from melting of the lunar highlands crust and/or upper
mantle: see Section 4.2). This model makes predictions about
observables: namely, the lithologies present at depth in a solidied
melt sea, which may be sampled by central peaks and walls of
superposed craters, and the density structure of this solidied melt
sea, which may be determined by seismic and/or gravity
measurements.
A general model of cumulate stratigraphy can be subdivided
into three component models: (1) a mixing model to recover the
composition of the differentiating liquid, if unknown; (2) a thermo-
dynamic model to determine the crystallization sequence of this li-
quid composition; and (3) a uid mechanical model to convert this
crystallization sequence into a stratigraphic sequence. Our model
of the cumulate stratigraphy of the Orientale melt sea (1) deter-
mines the composition of the differentiating liquid by mixing lunar
crust and upper mantle compositions weighted by their mass frac-
tions in a modeled melt cavity, (2) nds the crystallization se-
quence of this liquid on the FoAnQz pseudoternary phase
diagram, and (3) sinks or oats crystals according to crystalliquid
density contrasts in order to convert the crystallization sequence
into a stratigraphic sequence.
3.3. Composition of liquid(s) in the Orientale melt sea (bulk
composition)
We calculate the bulk composition of the Orientale melt sea by
mixing lunar crust and upper mantle compositions weighted by
their mass fractions in a modeled melt cavity (Fig. 4a).
The geometry of the melt cavity (Fig. 4a) has been approxi-
mated as a truncated sphere (Barr and Citron, 2011; Pierazzo
et al., 1997). For simplicity, we treat the melt cavity as a sphere
tangent to the lunar surface. The volume of this spherical melt cav-
ity must be the volume of impact melt produced, which was calcu-
lated in Section 2.3 to be to be 1.5  106 km3. Therefore, the
radius of the melt cavity is 70 km (i.e., the cube root of
1.5  106 km3/(4p/3)), rounded to an even 70 km for the following
calculations. The diameter of the melt cavity is approximately
twice the average crustal thickness of the lunar farside (Wieczorek
et al., 2006; Zuber et al., 1994), implying that the Orientale impact
755
melted the lunar upper mantle. The Orientale impact apparently
did not excavate the lunar upper mantleno mac mantle material
has been detected in Orientale basin deposits (Head, 2010; Head
et al., 2010a; Pieters et al., 2009; Yamamoto et al., 2010)but this
is as expected, since in Orientale-size lunar basins with a transient
cavity 620 km in diameter the depth of melting is modeled to ex-
ceed the depth of excavation by a factor of two (see Fig. 22 of Cint-
ala and Grieve (1998)).
We model the lunar highlands crust and upper mantle as com-
posed of entirely anorthite, enstatite, and forsterite in varying pro-
portions (Shearer et al., 2006). This is a good approximation for the
crustlunar highland rocks generally have modal plagio-
clase + pyroxene + olivine > 99 (Shearer et al., 2006; Wieczorek
et al., 2006)and a reasonable approximation for the upper mantle
from the base of the crust to 200 km, which has been modeled as
orthopyroxenite by thermodynamic models matching measured
and modeled seismic velocities (Khan et al., 2006; Kuskov, 1997).
Ti-rich phases are neglected, but these are probably more impor-
tant in the nearside crust and upper mantle (Wieczorek et al.,
2006). We take the farside crustal thickness to be 70 km (Zuber
et al., 1994). This is somewhat thicker than suggested by recent
modeling (Wieczorek et al., 2006), but using the number of
70 km affords two conveniences: rst, in this case the modeled
depth of melting is twice the thickness of the lunar crust, so that
exactly half the volume of the melt cavity can be taken to be crustal
material; second, approximating the melt cavity geometry as a
sphere rather than as a truncated sphere (Barr and Citron, 2011)
somewhat decreases the volume of crustal material in the melt
sea, which is counteracted by increasing the modeled crustal thick-
ness. The lunar crust is compositionally stratied: the upper crust
is anorthosite and the lower crust, exposed in basin ejecta, is prob-
ably anorthositic norite (Pieters et al., 2009; Wieczorek et al.,
2006). Therefore, we separate the crust into two planar layers of
equal 35 km thickness; the thicknesses of these crustal layers are
equal, as is approximately the case in the dual-layered crustal
thickness model of Wieczorek et al. (2006). The densities and com-
positions of these layers are based on the dual-layered crustal
thickness model of Wieczorek et al. (2006) with the additional con-
straint that enstatite and forsterite are present in the proportion
3:1 in both these layers, justied by the modal mineralogy of anor-
thosite and anorthositic norite samples (some lunar sample modes
are tabulated in Wieczorek et al. (2006)). The upper mantle is taken
to have a density of 3.3 g/cm3 and a composition of 75 wt.% ensta-
tite and 25 wt.% forsterite (Khan et al., 2006). All bulk composition
model parameters are presented in Table 1.
These parameters (Table 1) capture in the AnFoQz system the
salient features of the lunar crust and upper mantlean anortho-
site upper crust, an anorthositic norite lower crust, and a pyroxe-
nite upper mantle. Minor changes in these parameters do not
change the crystallization sequences in Section 3.5.
The mass fraction of each layer melted by a 140 km diameter
spherical melt cavity tangent to the lunar surface is calculated from
the volume of the spherical melt cavity intersecting the layer and the
density of the layer (Table 2). The bulk composition of the melt sea is
then given by the compositions of each layer weighted by their mass
fractions in the melt (Table 2). The resulting melt sea is fairly mac,
with an olivine noritic bulk composition (Stfer et al., 1980). This
modeled mac bulk composition is not necessarily inconsistent with
Table 1
Bulk composition model parameters.
Layer Thickness (km) Density (g/cm3) An (wt.%) En (wt.%) Fo (wt.%)
Upper crust 35 2.8 86 10.5 3.5
Lower crust 35 3.0 60 30 10
Upper mantle >70 3.3 0 75 25
756 W.M. Vaughan et al. / Icarus 223 (2013) 749765
Table 2
Bulk composition of the Orientale melt sea.
Layer (vol.%) (wt.%)
Proportion of melt sea
Upper crust 15.6 14.0
Lower crust 34.4 33.1
Upper mantle 50.0 52.9
An (wt.%) En (wt.%) Fo (wt.%)
Bulk composition of melt sea
31.9 51.1 17.0
remote sensing observations suggesting that surface or near-surface
portions of the melt sea are anorthositic or noritic (see Section 4.1),
since melt sea differentiates need not have the bulk composition of
the melt sea (see the discussion of South Pole-Aitken basin in
Section 4.2.2).
3.4. Initial composition of liquid(s) in the Orientale melt sea (melt
mixing)
We now investigate how the bulk composition of the Orientale
melt sea is related to the initial composition of differentiating li-
quid(s) in the sea.
Vigorous convection, driven by the thermal gradient between
hot melt and a cold free surface, is anticipated in a large melt sea
(Morrison, 1998). Convection may mix and homogenize the melt
sea so that the composition of any small volume of melt is identical
to the bulk composition of the melt. In this case, the initial compo-
sition of the differentiating liquid in the Orientale melt sea is sim-
ply the bulk composition of the Orientale melt sea given in Table 2
above.
However, even a well-mixed melt sea may not be homogenous:
Zieg and Marsh (2005) suggest that the relatively homogenous
granophyre and norite layers of the Sudbury Igneous Complex, a
terrestrial impact melt sheet, formed due to physical liquid immis-
cibility. In their scenario, mac and felsic liquids (melts of different
target lithologies) were prevented from rapidly mixing by shear
due to their order-of-magnitude-different viscosities. Before these
liquids could mix by diffusion, they separated according to their
densities, a process taking only 2 weeks in Sudbury (Zieg and
Marsh, 2005), giving rise to a density-stratied melt sea with a
light granophyre layer overlying a dense norite layer.
Similarly, anorthosite and mac liquids could have separated in
the Orientale melt sea, since plagioclase-rich crustal melts and ma-
c upper mantle melts have very different viscosities and densities.
The density contrast between anorthosite (2.6 g/cm3) and pyrox-
enite liquids (2.9 g/cm3) (Nelson and Carmichael, 1979) is nearly
three times greater than the density contrast between the grano-
phyre (2.7 g/cm3) and norite (2.8 g/cm3) layers of the Sudbury
Igneous Complex (Zieg and Marsh, 2005). The threefold greater
density contrast of lunar felsic and mac liquids compared with
Sudbury liquids could partially offset 1/6th lunar gravity so that
density separation of immiscible liquids could be almost as effec-
tive on the Moon as in the Sudbury Igneous Complex. Therefore,
we also model the relative volume of anorthositic and mac liquids
in a density-stratied melt sea (Table 3).
Table 3
Density-stratied melt sea parameters. Layer thicknesses are for a cylindrical melt sea 15.5 km thick.
Layer Density (g/cm3)
Anorthositic liquid (100 wt.% An) 2.61
Mac liquid (75 wt.% En, 25 wt.% Fo) 2.92
One-third of the melt produced by the Orientale impact may
have been ejected from the Orientale melt sea (Section 2.4) to be
incorporated in the rough facies of the Maunder Formation and
Hevelius Formation ejecta. Separation of physically immiscible liq-
uids takes weeks to months (Zieg and Marsh, 2005) after impact,
much longer than the processes that drive melt ejection, transient
cavity excavation and subsequent crater modication (Melosh,
1989; Hawke and Head, 1977). On this basis, we take the melt to
have been homogeneous during melt cavity collapse: in this case,
even though ejection decreases the volume of the melt sea by
one-third, it does not affect its bulk composition.
3.5. Crystallization sequences in the Orientale melt sea
We now model the crystallization sequences of homogenous
and density-stratied melt seas. Both equilibrium crystallization
(crystals always in equilibrium with liquid) and perfect fractional
crystallization (crystals only instantaneously in equilibrium with
liquid) are modeled in order to bound melt sea lithologies,
although vigorous convection is likely to have kept crystals in equi-
librium with liquid in the Orientale melt sea.
Crystallization sequences are determined using the AnFoQz
pseudoternary phase diagram (Fig. 8a). The 1 atm phase diagram
is used since pressures are <1 kbar even at the bottom of the melt
sea, 15.5 km deep in the crust (Mizutani and Osako, 1974). The
melt has an initial temperature >2000 C (Melosh, 1996), well
above the liquidus of the homogenous melt sea at 1550 C. (The
liquidus of the mac liquid in the density-stratied melt sea is sim-
ilar, 1650 C.)
3.5.1. Equilibrium crystallization of a homogenous melt sea
Olivine-saturated liquid crystallizes olivine until the liquid be-
comes sufciently silicic to react with previously-crystallized olivine
to form orthopyroxene. The liquid evolves along the olivineorthopy-
roxene boundary curve to the peritectic, where plagioclase comes on
the liquidus; the liquid is consumed before all olivine has been re-
sorbed. The nal crystallization sequence is olivine ? orthopyrox-
ene  olivine ? plagioclase + orthopyroxene  olivine (Fig. 8b).
3.5.2. Fractional crystallization of a homogenous melt sea
Olivine-saturated liquid crystallizes olivine until the liquid be-
comes sufciently silicic to crystallize orthopyroxene. No olivine
solids are in equilibrium with the liquid, which leaves the oliv-
ineorthopyroxene boundary curve, crystallizing orthopyroxene
until the liquid becomes plagioclase-saturated. The liquid evolves
along the plagioclase-orthopyroxene boundary curve to the eutec-
tic, where plagioclase, orthopyroxene, and quartz co-crystallize
until the liquid is consumed. The nal crystallization sequence is
olivine ? orthopyroxene ? plagioclase + orthopyroxene ? plagio-
clase + orthopyroxene + quartz (Fig. 8c).
3.5.3. Equilibrium crystallization of a density-stratied melt sea
The pure plagioclase-composition liquid does not undergo dif-
ferentiation. The crystallization sequence of the mac liquid (a
Fo + En mixture) follows from the well-known FoQz binary phase
diagram: olivine-saturated liquid crystallizes olivine until the li-
quid evolves to the peritectic, where olivine reacts with liquid to
form orthopyroxene; the liquid is consumed before all olivine
Proportion of melt sea (wt.%)|(vol.%) Thickness (km)
31.9 34.4 5.3
68.1 65.6 10.2
 W.M. Vaughan et al. / Icarus 223 (2013) 749765
Fig. 8. One-atmosphere AnFoQz ternary (a) phase diagram after Winter (2010) with (b) equilibrium crystallization and (c) fractional crystallization sequences for
composition X, the Orientale melt sea.
has been resorbed. The nal crystallization sequence of the mac
liquid is olivine ? orthopyroxene  olivine.
3.5.4. Fractional crystallization of a density-stratied melt sea
The pure plagioclase-composition liquid does not undergo dif-
ferentiation. The crystallization sequence of the mac liquid (a
Fo + En mixture) follows from the well-known (see, e.g., Winter
(2010)) FoQz binary phase diagram: olivine-saturated liquid crys-
tallizes olivine until the liquid evolves to the peritectic. All olivine
has been removed from the liquid, so the liquid composition
leaves the peritectic, crystallizing orthopyroxene and nally ortho-
pyroxene and quartz at the eutectic. The nal crystallization
sequence of the mac liquid is olivine ? orthopyroxene ?
orthopyroxene + quartz.
The crystallization sequences above are probably a good rst
approximation to the true crystallization sequence of the Orien-
tale melt sea. However, certain phases are not modeled, such as
clinopyroxene and Ti-bearing phases. Also, the composition of
solids is assumed to be constant, whereas in reality plagioclase,
olivine, and orthopyroxene are all solid solutions. The Mg or An
number of the rst olivine/orthopyroxene or plagioclase solids is
almost certainly greater than the Mg or An number of the last
olivine/orthopyroxene or plagioclase solids. Since the appearance
of clinopyroxene or changes in Mg or An number may provide
important constraints on melt sheet differentiation, we plan to
use more sophisticated thermodynamic models to investigate
the crystallization sequences of lunar impact melt sheets in
greater detail.
Fig. 9. Anorthite crystal sink-oat diagram for forsterite Mg number = 40 (a) superposed on AnFoQz ternary phase diagram (b).
757
3.6. Crystalliquid density contrasts in the Orientale melt sea
We compare crystal and liquid densities to determine whether
crystallizing solids will sink or oat in order to convert the crystal-
lization sequences in Section 3.5 to stratigraphic sequences.
AnFoQz liquids are substantially less dense (always >0.1 g/
cm3, calculated from Nelson and Carmichael (1979)) than anorthite
crystals (q = 2.75 g/cm3). Even in an iron-rich liquid with a con-
stant Mg number of 40 (Fig. 9a), more ferroan than even the inter-
stitial mac minerals in ferroan anorthosites (Shearer et al., 2006),
anorthite never oats in a plagioclase-saturated liquid (Fig. 9b).
Our conclusion is that anorthite crystals do not oat in (Ti-poor)
impact melts of the crust and upper mantle. The critical implica-
tions of this idea for the cumulate stratigraphies and density struc-
tures in impact melts of the crust will be considered in
Section 4.2.1. Since plagioclase crystals are substantially less dense
than orthopyroxene and olivine crystals (q  3.2 g/cm3), we con-
clude that all crystallizing solids sink in their coexisting liquids.
Therefore, the crystallization sequences in Section 3.5 above are
effectively stratigraphic sequences: for example, co-crystallizing
plagioclase and orthopyroxene both sink to form norite.
3.7. Normative calculations and stratigraphic columns for the
Orientale melt sea
Finally, we calculate the weight proportions and thicknesses of
each co-crystallizing assemblage to create stratigraphic columns
for solidied melt seas. (Although crystal settling is likely to be
758 W.M. Vaughan et al. / Icarus 223 (2013) 749765

kinetically incompatible with perfect equilibrium crystallization, proportions (Table 4) of all three layers are given simply by the ini-
we nonetheless convert equilibrium crystallization sequences to tial wt.% An, Fo, and En of the liquid.
stratigraphic sequences in order to bound melt sea stratigraphies.
Real stratigraphic sequences may approach sequences based on
3.7.4. Fractional crystallization of a density-stratied melt sea
equilibrium crystallization.) The following sections briey describe
The weight proportions of the three crystallizing lithologies
our normative calculations, which are based on the AnFoQz and
monomineralic olivine (dunite), monomineralic orthopyroxene
FoQz phase diagrams.
(pyroxenite), and orthopyroxene + quartz (quartz pyroxenite)
are constrained by three equations, the rst two relating the com-
3.7.1. Equilibrium crystallization of a homogenous melt sea position of each lithology (measured from the FoQz binary) to the
The liquid crystallizes to give monomineralic olivine (dunite) bulk composition and the third relating the relative proportions of
which then reacts with liquid to form an overlying layer of pyrox- dunite and pyroxenite as determined from olivine and orthopyrox-
enite and nally norite as plagioclase comes on the liquidus at the ene crystallization path length on the FoQz binary. The resulting
peritectic. The weight proportion of the dunite layer is given sim- system of linear equations can be written as the following matrix
ply by the initial wt.% Fo of the liquid; the norite layer has an An/En equation:
ratio of 3/2 as measured from the phase diagram and, accord-
0 10 1 0 1
ingly, its weight proportion is given by the sum of the initial 1 0:7 0:55 X Du 0:78
wt.% An and 2/3 that wt.% of En; the wt.% of the pyroxenite layer B CB C B C
@ 0 0:3 0:45 A@ X Px A @ 0:22 A 2
is equal to the remaining wt.% En (Table 4).
1 3:3 0 X Qz 0

3.7.2. Fractional crystallization of a homogenous melt sea We solve this matrix equation for the weight proportions of these
The weight proportions of the four crystallizing lithologies lithologies (Table 4).
monomineralic olivine (dunite), monomineralic orthopyroxene
(pyroxenite), orthopyroxene + anorthite (norite), and orthopyrox-
ene + anorthite + quartz (quartz diorite)are constrained by four
equations, the rst three relating the compositions of each lithol-
ogy (measured from the AnFoQz ternary) to the bulk composi-
tion and the fourth relating the relative proportions of dunite
and pyroxenite as determined from olivine and orthopyroxene
crystallization path length on the AnFoQz ternary (Fig. 8c). The
resulting system of linear equations can be written as the following
matrix equation:
0 10 1 0 1
1 0:7 0:32 0:16 X Du 0:53
B 0 0:55 0:51 C
0 B C B C
B CB X Px C B 0:32 C
B CB CB C 1
@ 0 0:3 0:13 0:33 A@ X Nr A @ 0:15 A
1 1:5 0 0 X Qz 0

We solve this matrix equation for the weight proportions of these

lithologies (Table 4).

3.7.3. Equilibrium crystallization of a density-stratied melt sea

The overlying anorthositic liquid crystallizes to anorthosite. The
mac liquid crystallizes to form monomineralic olivine (dunite) Fig. 10. Model Orientale cumulate stratigraphies and density proles (to scale)
and monomineralic orthopyroxene (pyroxenite) layers. The weight produced by equilibrium and fractional crystallization of homogenous and density-
stratied melt seas.

Table 4
Layer compositions and thicknesses in solidied cylindrical melt seas 14 km thick.

Layer Composition Density Proportion of melt Thickness

(g/cm3) sea (wt.%) (vol.%) (km)
Equilibrium crystallization, homogenous melt sea Norite 61.6 wt.% An, 38.4 wt.% En 2.95 51.8 54.0 7.6
Pyroxenite 100 wt.% En 3.20 31.2 30.0 4.2
Dunite 100 wt.% Fo 3.25 17.0 16.0 2.2
Fractional crystallization, homogenous melt sea Quartz diorite 51.0 wt.% An, 26.1 wt.% 2.85 13.7 14.6 2.0
Qz, 22.9 wt.% En
Anorthositic norite 55.0 wt.% An, 45.0 wt.% En 2.95 45.3 46.7 6.5
Pyroxenite 100 wt.% En 3.20 16.4 15.6 2.2
Dunite 100 wt.% Fo 3.25 24.6 23.0 3.2
Equilibrium crystallization, density-stratied melt sea Anorthosite 100 wt.% An 2.75 31.9 35.4 5.0
Pyroxenite 100 wt.% En 3.20 51.1 48.7 6.8
Dunite 100 wt.% Fo 3.25 17.0 15.9 2.2
Fractional crystallization, density-stratied melt sea Anorthosite 100 wt.% An, 78.6 wt.% En 2.75 31.9 35.1 4.9
Quartz pyroxenite 21.4 wt.% Qz 3.10 27.8 27.2 3.8
Pyroxenite 100 wt.% En 3.20 9.4 8.9 1.2
Dunite 100 wt.% Fo 3.25 30.9 28.8 4.0
 W.M. Vaughan et al. / Icarus 223 (2013) 749765
Finally, the weight proportions are converted to volume propor-
tions based on layer density (a weighted average of component
mineral density); these volume proportions are directly propor-
tional to the thicknesses in a cylindrical melt sea (Table 4). We
choose to model melt seas as cylindrical since this geometry has
been predicted as the consequence of inward movement and up-
ward translation during transient cavity collapse (Head, 2010,
2012; Head et al., 1993). (The effect of a melt sea with a curved
base, which has more of its volume near the surface, is to increase
the thickness of deep layers and decrease the thickness of shallow
layers.) The cumulate stratigraphies and density structures of these
modeled melt seas are shown in Fig. 10.
4. Discussion
We now assess our modeled cumulate stratigraphies (Fig. 10)
and consider more generally the implications of impact melt differ-
entiation for the lunar highlands crust.
4.1. Constraints on Orientale melt sea cumulate stratigraphy:
implications for processes once operating in the Orientale melt sea
The modeled cumulate stratigraphies (Fig. 10) produced by
melt homogenization/density stratication and equilibrium crys-
tallization/fractional crystallization have observable differences.
Constraints on the lithologies present at depth in the solidied Ori-
entale melt sea make it possible to rule out some of these modeled
stratigraphies and determine what processes operated in the mix-
ing and crystallization of the Orientale melt sea.
There are currently two important constraints on the cumulate
stratigraphy of the Orientale melt sea: (1) The top 1 km of the
solidied impact melt sea is anorthosite. M3 spectral observations
of the Maunder Formation (Head, 2010; Head et al., 2010a; Pieters
et al., 2009) indicate that the upper surface of the smooth facies is
anorthosite. Moreover, small (10 km diameter) craters superpos-
ing the impact melt sheet, which depth/diameter ratios predict
excavate material from a depth of up to 1 km, uncover more
anorthosite. (2) Norite occurs at 4 km depth. The central peak of
Maunder, a 55 km diameter crater superposed on the Orientale
melt sheet, is predicted by depth/diameter ratios to have sampled
the melt sea at a depth of 4 km (Whitten et al., 2011). This peak
has a noritic composition (Whitten et al., 2011). These two con-
straints can themselves be interpreted as evidence for the differen-
tiation of the Orientale melt sea: near-surface lithologies are not
the same as those present at depth.
Which modeled cumulate stratigraphies (Fig. 10) are consistent
with these constraints? Fractional crystallization of a homogenous
melt sea produces a 2 km thick layer of quartz diorite at the top
of the melt sea, clearly inconsistent with constraint (1). Density
stratication produces a 5 km thick layer of pure anorthosite at
the top of the melt sea, inconsistent with constraint (2). Equilib-
rium crystallization of a homogenous melt sea produces a
7.5 km thick layer of norite at top of the melt sea, consistent with
constraint (2). This is not necessarily inconsistent with constraint
(1): the anorthositic roof of the Orientale melt sheet could be an
early-formed quench crust; a fallback breccia, as in Sudbury (Thi-
errault et al., 2002); or the result of modal separation of plagioclase
from pyroxene in late-crystallizing norite.
Therefore, we conclude that the Orientale melt sea was homog-
enized by melt mixing and underwent equilibrium crystallization.
Indeed, both homogenization and equilibrium crystallization sug-
gest vigorous convection, consistent with a cooling lava lake (Wor-
ster et al., 1990) and particularly with a cooling impact melt sea
(Morrison, 1998).
759
It is possible that high-resolution gravity measurements from
the ongoing GRAIL mission could provide additional constraints
on Orientale melt sea cumulate stratigraphy. Certainly, gravity
measurements from the GRAIL mission can provide constraints
on melt sea thickness. We note in passing that, in light of our evi-
dence for a thick sea of impact melt at the center of the Orientale
basin, crustal thickness models for the Orientale basin need to be
reassessed (Fig. 11). We predict that the crust occurring at the
center of the Orientale basin is actually a solidied melt sea with
a density of 3.05 g/cm3, not anorthosite or brecciated anorthosite
with a density of 2.75 g/cm3, as has been traditionally assumed
(Andrews-Hanna and Stewart, 2011; Hikida and Wieczorek,
2007). The 10 km thick anorthositic crust modeled to occur at
the center of the Orientale basin (Andrews-Hanna and Stewart,
2011; Hikida and Wieczorek, 2007) becomes a 20 km thick im-
pact melt sheet (Fig. 11) using this revised density, roughly consis-
tent with (although 1.3 times greater than) our melt sheet
thickness estimates in Section 2.
4.1.1. Is crystal settling an operative differentiation mechanism in
impact melt seas?
In this paper, we interpret the differentiation of the Orientale
impact melt sea (and other lunar impact melt seas; see Section 4.2)
in terms of crystal settling: we transform crystallization sequences
into stratigraphic sequences under the assumption that gravity
efciently segregates crystals from melt to form igneous strata.
We are by no means the rst authors to make this assumption:
Morrison (1998) and Nakamura et al. (2009) used crystal settling
to interpret the differentiation of the South Pole-Aitken Basin im-
pact melt sheet. However, eld studies and theoretical analyses
of terrestrial impact melt sheets and magma chambers (Marsh,
1989; Marsh, 1996; Warren et al., 1996; Zieg and Marsh, 2005)
may indicate that crystal settling does not operate in the differen-
tiation of impact melt. Field studies of the Sudbury Igneous Com-
plex, a terrestrial impact melt sheet, show that massive
granophyre and norite layers in the Sudbury Igneous Complex
are homogenous; crystal settling did not take place in their parent
magmas (Zieg and Marsh, 2005). This is perhaps unsurprising:
crystal settling operates on isolated phenocrysts in a magma body,
not the dense mat of crystals along the cooling margin of a magma
body (Marsh, 1989; Marsh, 1996); impact melts, however, are
superheated (>2000 C) (Melosh, 1996), and contain no pheno-
crysts to settle. Marsh (1996) writes, . . .initially crystal-free mag-
mas [such as impact melts] cannot undergo strong chemical
differentiation by fractional crystallization of singly settling
crystals.
What are the implications for solidied melt sea stratigraphy if
crystal settling indeed does not operate in the differentiation of im-
pact melt seas? Impact melt bodies are then either (1) homogenous
or (2) heterogeneous but differentiated by another mechanism,
such as viscous emulsion differentiation (Zieg and Marsh, 2005).
However, the constraints on Orientale melt sea cumulate stratigra-
phy described in Section 4.1 are inconsistent with large-scale
homogeneity and not entirely consistent with viscous emulsion
differentiation (Fig. 10, lower left). Perhaps our model of viscous
emulsion differentiation is colored by our assumption that the cra-
ter excavation ow does not change the bulk composition of the
impact melt sea: if the excavation ow removed only impact melt
of upper crustal composition, an uppermost anorthosite layer
(Fig. 10, lower left) would be thinned, bringing noritic lithologies
towards the surface. In this case, the noritic central peak of Maun-
der represents recrystallized lower crustal material.
If viscous emulsion differentiation is indeed the operative dif-
ferentiation mechanism in impact melt seas, solidied impact melt
seas recapitulate the structure of the lunar crust and upper mantle,
just as the Sudbury impact melt sheet recapitulates the preexisting
760 W.M. Vaughan et al. / Icarus 223 (2013) 749765

Fig. 11. Traditional (a) crustal thickness models for the center of the Orientale basin, following Andrews-Hanna and Stewart (2011) and Hikida and Wieczorek (2007), and the
revised model (b) described here.

crustal structure of the Canadian shield (Zieg and Marsh, 2005).

Moreover, since impact melt seas have a very different aspect ratio
from the initial melt cavity (Fig. 4), solidied impact melt bodies
may bring lower crust and upper mantle materials to the near sur-
face where they can be excavated by complex craters of moderate
size (such as Maunder in Orientale). Further investigation of lunar
impact melt sea stratigraphy will clarify the differentiation of lunar
impact melts and the roles of crystal settling and viscous emulsion
differentiation.

4.2. Cumulate stratigraphy of other lunar impact melt seas

There are at least 30 lunar basins (craters >300 km diameter)

besides Orientale, as determined from the database of Head et al.
(2010b). Impact melt scaling relations (Cintala and Grieve, 1998)
predict that the impacts that formed these basins each produced
>105 km3 of melt. Very probably, these basins, like Orientale, con-
tain solidied impact melt seas. What is the cumulate stratigraphy
of these melt seas and how might it differ from the cumulate stra-
tigraphy of Orientale? Since target crust/upper mantle structure is
not thought to change substantially across the farside highlands
(Wieczorek et al., 2006), the important variable controlling melt
sea bulk composition is depth of melting, which is a function of ba-
sin size. The compositional endmembers are (1) relatively small
basin-forming impacts that melt only the lunar crust and (2) giant
basin-forming impacts that melt mainly the lunar mantle (Fig. 12).
The compositional watershed is the plagioclase-olivine cotectic
(Fig. 12): small basin-forming impacts form melt seas that become
plagioclase-saturated early, whereas large basin-forming impacts
form melt seas that become plagioclase-saturated late in crystalli-
zation. Having considered the cumulate stratigraphy of a melt sea
in a moderately-sized basin, Orientale, we now proceed to model
the cumulate stratigraphy of melt seas occurring in a small lunar
basin and the very large South-Pole Aitken basin.
4.2.1. Cumulate stratigraphy of a melt sea in a small lunar basin
We consider a hypothetical basin-forming impact that melts the
entire thickness of the lunar crust and none of the upper mantle,
Fig. 12. Bulk composition of an impact melt sea as a function of basin size (km).
corresponding to a spherical melt cavity 70 km in diameter, which
we take to be the thickness of the farside crust (Wieczorek et al.,
2006; Zuber et al., 1994). A spherical melt cavity 70 km in diameter
contains 1.7  105 km3 of melt, which, according to the scaling
relations of Cintala and Grieve (1998), corresponds to a basin tran-
sient cavity 250 km in diameter and a nal basin diameter of
415 km (Croft, 1985), almost exactly the diameter of the basin
Korolev (Baker et al., 2011a). We reuse our modeled bulk composi-
tion parameters (Table 1) to calculate the bulk composition of this
crustal melt sea: 72.5 wt.% An, 20.5 wt.% En, and 7.0 wt.% Fo. We
model equilibrium crystallization of a homogenized crustal melt
sea since these processes best reproduced the observed cumulate
stratigraphy of the Orientale melt sea (Section 4.1).
This crustal melt is initially plagioclase-saturated (Fig. 13a). Pla-
gioclase crystallizes until the liquid becomes olivine-saturated; the
liquid then evolves along the plagioclase-olivine cotectic to the
 W.M. Vaughan et al. / Icarus 223 (2013) 749765
Fig. 13. One-atmosphere AnFoQz ternary (a) phase diagram after Winter (2010) with (b) the equilibrium crystallization sequence for composition X0 , a crustal melt.
peritectic, where orthopyroxene and plagioclase co-crystallize; the
liquid is consumed before all olivine has been resorbed. The crys-
tallization sequence of this liquid (Fig. 13b) is plagioclase ? plagio-
clase + olivine ? plagioclase + orthopyroxene  olivine. Plagio-
clase sinks in impact melts of the crust (Fig. 9), so this
crystallization sequence is effectively the cumulate stratigraphy.
We calculate the weight proportions of these co-crystallizing
assemblages using methods similar to those in presented in Sec-
tion 3.7. The cumulate stratigraphy of a crustal melt sea occurring
in a Korolev-sized basin is compared with the likely cumulate stra-
tigraphy of the solidied Orientale melt sea in Fig. 14.
The lithologies exposed at or near the surface or in the central
peaks of complex craters are modeled to be norite in both small
and large melt seas. The critical observable difference between
these two melt seas is their density structure. The crustal melt
sea has an inverted density prolethe result of early-crystallizing
plagioclase solids sinking to form an anorthosite layer beneath
denser troctolite and norite layerswhereas the Orientale and
other mantle melt seas have a regular density prole with a dense
dunite layer underlying less dense pyroxenite and norite. Future
seismic measurements on solidied melt sea surfaces could poten-
tially resolve these different density structures.
4.2.2. South Pole-Aitken basin melt sea cumulate stratigraphy
The enormous 2500 km diameter South Pole-Aitken (SPA) basin
(Wilhelms et al., 1979) almost certainly melts huge volumes of
ultramac mantle material (Morrison, 1998). However, the SPA
oor, possibly the surface of a melt sea, is noritic, not ultramac,
and craters superposing this oor excavate pyroxene, not mantle
olivine (Borst et al., 2012; Moriarty et al., 2011, 2012; Nakamura
et al., 2009). Morrison (1998) suggested that differentiation of
the SPA melt sheet could reconcile these apparently conicting
observations: a basal ultramac layer could sequester mantle oliv-
ine and an upper noritic layer could concentrate crustal plagio-
clase. We estimate the thickness and model the cumulate
stratigraphy of the South Pole-Aitken basin melt sea to investigate
this idea in greater detail.
How thick is the SPA melt sea? We take the SPA transient cavity
diameter to be 2099 km (Wieczorek and Phillips, 1999). Using the
impact melt volume scaling relationship derived by Abramov et al.
(2012) and the model parameters described in Kring et al. (2012),
we calculate that the South Pole-Aitken basin-forming impact pro-
duced 6.8  107 km3 of impact melt. About 25 vol.% of this melt
was excavated from the basin transient cavity, leaving
5.1  107 km3 of impact melt inside the inner basin ring to form
761
a melt sea. The thickness of this melt sea is constrained by its areal
extent. Morrison (1998) calculated the areal extent of the SPA melt
sea by taking the inner ring of SPA to be a circle 1800 km in diam-
eter (Wilhelms et al., 1979). We calculate the areal extent of the
SPA melt sea using the recent measurements by Garrick-Bethell
and Zuber (2009) suggesting that the inner ring of SPA is an ellipse
with a semimajor axis of 970 km and a semiminor axis of 720 km.
A prismatic melt sea with this areal extent containing
5.1  107 km3 of impact melt is 25 km thick. However, the Ori-
entale melt sea probably occurs in a central depression with a
diameter 70% of that of the Outer Rook Ring (350 km vs.
480 km). If this ratio holds for SPA, so that the semimajor and
semiminor axes of the surface of the SPA melt sea are 680 km
(70% of 970 km) and 500 km (70% of 970 km), then the SPA melt
sea may be 50 km thick.
What is the cumulate stratigraphy of a hypothetical solidied
SPA melt sea? A spherical melt cavity containing the estimated
6.8  107 km3 SPA melt volume has a radius of 250 km, corre-
sponding to a depth of melting of 500 km. The bulk composition
of the SPA melt sea, reusing the bulk composition model parame-
ters in Table 1, is nearly that of the upper mantle, with 4 wt.%
An, 72 wt.% En, and 24 wt.% Fo. The crystallization sequence of this
melt and the modal calculations are the same as for the Orientale
melt lake described in Section 3. The solidied South Pole-Aitken
basin melt sea, if 25 km thick, is thus predicted to consist of a thin
2.6 km layer of norite overlaying a massive 16.5 km layer of
pyroxenite and a basal 5.9 km thick dunite cumulate (Fig. 14).
What are the implications of this cumulate stratigraphy? First,
the noritic oor of SPA could very well be the roof of a massive dif-
ferentiated impact melt sea. Second, observations of crater central
peaks that determine the depth at which norite transitions to
pyroxenite can be used to constrain the thickness of the SPA melt
sea. Third, Nakamura et al. (2009) argue that the SPA impact melt
sea has not undergone differentiation on the basis that all central
peaks observed in their study excavate pyroxenite. We suggest in-
stead that the SPA impact melt sea has undergone differentiation
and that all central peaks observed in their study (which excavate
material from depths of 610 km) excavate a 16.5 km thick
homogenous pyroxenite cumulate (Fig. 14).
4.3. Impact melt differentiates in the lunar sample suite?
Huge volumes of impact melt are produced by basin-forming
impacts. Much of this impact melt may have differentiated: the
Orientale-forming impact alone may have given rise to 106 km3
762 W.M. Vaughan et al. / Icarus 223 (2013) 749765
Fig. 14. Model cumulate stratigraphies and density proles for melt seas in basins of three different sizes: a crustal melt sea occurring in a small basin approximately the
size of Korolev, the Orientale melt sea, and the SPA melt sea. Layers in each melt sea are to scale, although melt seas are not to the same scale.
of impact melt differentiates. For comparison, the total volume of
lunar mare basalts is estimated to be 107 km3 (Head and Wilson,
1992). Could impact melt differentiates excavated from massive
impact melt seas be represented in the lunar sample suite?
4.3.1. Can impact melt differentiates pass as pristine?
Pristine lunar rocks (Warren, 1993), which have coarse-grained
textures and low siderophile element concentrations, are thought
to represent endogenous lunar igneous rocks largely unaltered by
impact processes. We suggest that impact melt differentiates
formed in massive impact melt seas also have coarse-grained tex-
tures and low siderophile element concentrations and so may be
able to pass for pristine lunar rocks (Basilevsky and Neukum,
2010).
Massive impact melt seas cool slowly. The diameter of the Ori-
entale melt sea (175 km) exceeds its probable depth (15.5 km)
by an order of magnitude, so we approximate a (non-convecting)
melt sea as a cooling slab similar to an intruded sill. The solidica-
tion time t of a sill cooling by conduction can be approximated by
the formula
 L 2
2
t 0:694
K
where L is the half-thickness of the sill and K is its thermal diffusiv-
ity, 102 cm2/s for rock (Jaeger, 1968). From (3), we estimate that
the Orientale melt sea took on the order of 105 years to solidify, long
enough for impact melt differentiates formed in that sea to develop
coarse-grained textures (Zieg and Marsh, 2002). These calculations
are supported by the observation that impact melt differentiates
formed at depth in the 3 km thick terrestrial Sudbury impact melt
sheet are also coarse-grained (Zieg and Marsh, 2002) and were long
mistaken for endogenous plutonic rocks (Thierrault et al., 2002).
Also, although impact melt seas may have a relatively high bulk
concentration of meteoritic siderophiles, these siderophile ele-
ments may fractionate into basal deposits of metal or suldes un-
mixed from metal- or sulfur-saturated impact melt, so that
silicates from these impact melt seas may have nearly lunar, not
meteoritic, siderophile element concentrations.
4.3.2. Excavating impact melt differentiates
Most impact melt differentiates form at depth. Certain impact
melt differentiates, such as dunites and troctolites, may form at
depths below 10 km. For these lithologies to be exposed at the sur-
face and introduced into the lunar sample suite, they must be exca-
vated by the impact cratering process. Very deep lithologies can
only be excavated by very large craters; we can see that deep dif-
ferentiates cannot possibly have been excavated at Orientale (no
craters >65 km superpose the Orientale melt sheet, so dunite pos-
sibly occurring below 12 km in the Orientale melt sea has never
been excavated and cannot occur in the lunar sample suite). At
other basin melt sheet locations, shallow impact melt differenti-
ates, probably noritic in composition, are the most likely to be
excavated by impacts and introduced into the lunar sample suite.
Old basins (Fig. 15) with many superposed large craters and basins,
such as the South Pole-Aitken basin (Head, 2010; Head et al., 1993,
2010a; Pieters et al., 1997) are the most likely source regions for
deep impact melt differentiates, and certain norites or pyroxenites
could come from SPA.
Sections 4.3.1 and 4.3.2 suggest that impact melt differentiates
may exist in the lunar sample suite and may be able to pass for
highland plutonic rocks. We now consider three highland plutonic
rock suitesMg-suite lithologies, young ferroan anorthosites, and
Mg-spinel lithologiesto determine whether these puzzling lithol-
ogies could potentially be impact melt differentiates.
4.3.3. Mg-suite lithologies
Mg-suite rocksdunites, norites, and troctolites characterized
by high Mg numbers (7095) and high rare earth element (REE)
content (up to 400 times chondritic) (Hess, 1994; Shearer et al.,
3 2006)are traditionally thought to have formed from intrusions
of mac magmas into ferroan anorthosite crust, since dunites, nor-
ites, and troctolites can be formed from fractional crystallization of
a Fo-rich liquid assimilating plagioclase (see Fig. 8a for reference).
The Mg-suite chemical signature is harder to explain: simple mix-
ing of a high Mg number liquid with a low Mg number, REE-en-
riched (KREEP) liquid produces a liquid with insufciently high
Mg number to be the Mg-suite parent magma; Hess (1994) argued
that the Mg-suite parent magma is not a mixture of melts, but a
melt of a mixture: a mixture of KREEP, olivine-rich dunite, and
anorthosite produces a partial melt with high Mg number and
REE content. The process giving rise to this mixture is unclear; pro-
cesses such as cumulate overturn have been invoked (Shearer
et al., 2006).
Could differentiation of impact melt seas produce Mg-suite
lithologies? Certainly, dunites, troctolites, and norites can form in
impact melt seas: the Orientale melt sea contains norite and prob-
ably dunite, and smaller impacts that melt mainly the lunar crust
(Section 4.2) will produce troctolite. But these lithologies are mix-
tures of melts and for the reasons described above cannot have the
full Mg-suite chemical signature. Also, thermobarometry of Mg-
suite rocks suggests that these rocks formed 4050 km deep in
the crust (McCallum and Schwartz, 2001). It is possible that remo-
tely-sensed Mg-suite lithologies are really unrelated Mg-suite-like
dunites, troctolites, and norites lacking the Mg-suite chemical sig-
nature. These Mg-suite-like lithologies could be impact melt differ-
entiates; however, until Mg-suite-like lithologies are identied in
 W.M. Vaughan et al. / Icarus 223 (2013) 749765
Fig. 15. Lunar basins greater than 300 km in diameter from Head et al. (2010b).
the lunar sample suite, the existence of these counterfeits must be
regarded as speculation.
4.3.4. Young ferroan anorthosites
Certain ferroan anorthosites with ages <4.4 Ga (Borg et al.,
2011) are much too young to have crystallized in the lunar magma
ocean. These ferroan anorthosites are thought to be cumulates
formed during early plutonism (Shearer et al., 2006). Since impact
melting generally resets geochronometers, young ferroan anortho-
sites could have formed in massive impact melt sheets (formation
in thin, ash-frozen impact melt sheets is probably incompatible
with the large grain sizes of these anorthosites). One difculty with
this idea is that anorthite is negatively buoyant in impact melts of
the crust and mantle (Fig. 9), so anorthosite generally lines the bot-
toms of melt seas and may be difcult to excavate in quantity. If
the anorthositic roof of the Orientale melt sheet indeed formed
due to igneous differentiation, young anorthosites from impact
melt sheets may be more common, and easier to excavate, than
has been imagined.
4.3.5. Mg-spinel lithologies
Recently discovered Mg-spinel lithologies (Pieters et al., 2011),
rocks with 30 wt.% spinel containing <5 wt.% Fe, have been ex-
plained as the product of melt-wallrock reactions of Mg-suite mag-
mas with anorthosite crust (Prissel et al., 2012). Potentially, spinel
could form from reaction of anorthite liquids formed by impact
melting with mac minerals or melts. In density-stratied melt
seas, anorthosite liquids overlie mac liquids; large volumes of spi-
nel could potentially be produced at this interface. One difculty
with this idea is that impact melts of the ferroan crust and mantle
are relatively ferroan and could produce ferroan spinels.
4.4. Volumetric importance of impact melt in the lunar highlands crust
How much impact melt is incorporated in the lunar highlands
crust? The database of lunar craters P20 km in diameter compiled
by Head et al. (2010b) includes 30 basins >300 km in diameter.
Using the impact melt volume scaling laws of Cintala and Grieve
(1998), we estimate that collectively these 30 basins contain on
the order of 108 km3 of impact melt, 10 times the total volume of
mare basalt (Head and Wilson, 1992) and about 1/20th the volume
of the lunar crust. These basins, and the massive impact melt
sheets that they contain, are distributed widely across the Moon
763
(Fig. 15). Understanding the geology and the petrology of the
108 km3 of massive impact melt deposits occurring in these ba-
sinswork we have just begun to undertake hereis critically
important for understanding the lunar highlands crust.
5. Conclusions
Basin-forming impacts forming craters >300 km in diameter
produce enormous volumes of impact melt (>105 km3). We have
investigated the geology and petrology of these huge volumes of
impact melt, focusing especially on the relatively young lunar ba-
sin Orientale, and conclude that:
1. Models of impact melt production combined with recent geo-
logic and topographic data suggest that most impact melt
(2/3) produced by the Orientale-forming impact occurs in a
15.5 km thick (solidied to 14 km thick) impact melt sea
350 km in diameter with a volume of 106 km3.
2. The Orientale basin impact melt sea has probably undergone
large-scale igneous differentiation, since terrestrial impact melt
sheets (such as Manicouagan, Sudbury, and Morokweng) a
tenth of its thickness and a hundredth of its volume are known
to have differentiated.
3. A modeled cumulate stratigraphy (occurring below a quench
crust and anorthositic fallback breccia, with an 8 km thick
layer of norite overlying a 4 km thick pyroxenite with a basal
2 km thick layer of dunite) produced by equilibrium crystalli-
zation of a homogenized melt sea is consistent with remotely-
sensed norite excavated from 4 km depth in the central peak
of Maunder crater.
4. Small basins melting mainly the lunar crust produce initially
plagioclase-saturated melt seas that rst crystallize negatively
buoyant anorthosite and therefore develop inverted density
structures. The noritic oor of the South-Pole Aitken basin
may be the roof of a differentiated impact melt lake; crater cen-
tral peaks superposing SPA may excavate a massive layer of
pyroxenite.
5. Thirty lunar basins >300 km in diameter together contain on the
order of 108 km3 of impact melt, one-twentieth the total volume
of the lunar crust. Impact melt differentiates excavated from
these solidied impact melt seas may well be represented in
the lunar sample suite, mimicking pristine highland plutonic
lithologies.
764 W.M. Vaughan et al. / Icarus 223 (2013) 749765

6. High-resolution compositional and gravity data collected by the Greeley, R. et al., 1993. Galileo imaging observations of lunar maria and related
deposits. J. Geophys. Res. 98, 1718317205. http://dx.doi.org/10.1029/
M3 instrument on Indias Chandrayaan-1 mission and the ongo-
93JE01000.
ing LRO and GRAIL missions make it possible to probe the struc- Grieve, R.A.F., Cintala, M.J., 1992. An analysis of differential impact melt-crater
ture of the lunar highlands crust in unprecedented detail; scaling and implications for the terrestrial cratering record. Meteoritics 27,
analyses based on these datasets, such as constraining the 526538.
Grieve, R.A.F., Cintala, M.J., 1997. Planetary differences in impact melting. Adv.
cumulate stratigraphy of melt seas from careful compositional Space Res. 20, 15511560. http://dx.doi.org/10.1016/S0273-1177(97)00877-6.
analyses of crater central peaks, will further our understanding Grieve, R.A.F., Stfer, D., Deutsch, A., 1991. The Sudbury structure: Controversial or
of the geology and petrology of massive impact melts. misunderstood? J. Geophys. Res. 96, 2275322764. http://dx.doi.org/10.1029/
91JE02513.
Hardee, H.C., 1980. Solidication in Kilauea Iki lava lake. J. Volcanol. Geoth. Res. 7,
211223. http://dx.doi.org/10.1016/0377-0273(80)90030-X.
Hart, R.J., Cloete, M., McDonald, I., Carlson, R.W., Andreoli, M.A.G., 2002. Siderophile-
rich inclusions from the Morokweng impact melt sheet, South Africa: Possible
Acknowledgments fragments of a chondritic meteorite. Earth Planet. Sci. Lett. 198, 4962. http://
dx.doi.org/10.1016/S0012-821X(02)00497-1.
Thoughtful reviews by John Spray and Bruce Marsh improved Hawke, B.R., Head, J.W., 1977. Impact melt on lunar crater rims. In: Roddy, D.J.,
Pepin, R.O., Merrill, R.B. (Eds.), Impact and Explosion Cratering: Planetary and
this manuscript. This work was supported under NASA Lunar Sci- Terrestrial Implications. Pergamon Press, New York, NY, pp. 815841.
ence Institute (Brown-MIT) grant NNA09DB34A, which is grate- Head, J.W., 1974. Orientale multi-ringed basin interior and implications for the
fully acknowledged. petrogenesis of lunar highland samples. Moon 11, 327356. http://dx.doi.org/
10.1007/BF00589168.
Head, J.W., 2010. Transition from complex craters to multi-ringed basins on
terrestrial planetary bodies: Scale-dependent role of the expanding melt cavity
References and progressive interaction with the displaced zone. Geophys. Res. Lett. 37,
L02203. http://dx.doi.org/10.1029/2009GL041790.
Head, J.W., 2012. Lunar Orientale Basin: Characterization and Insights into Multi-
Abramov, O., Wong, S.M., Kring, D.A., 2012. Differential melt scaling for oblique
Ringed Basin Formation. Moscow Sol. Syst. P. 3.
impacts on terrestrial planets. Icarus 218, 905916. http://dx.doi.org/10.1016/
Head, J.W., Wilson, L., 1992. Lunar mare volcanism: Stratigraphy, eruption
j.icarus.2011.12.022.
conditions, and the evolution of secondary crusts. Geochim. Cosmochim. Acta
Andrews-Hanna, J.C., Stewart, S.T., 2011. The crustal structure of Orientale and
56, 21552175. http://dx.doi.org/10.1016/0016-7037(92)90183-J.
implications for basin formation. Lunar Planet. Sci. XLII, 2194 (abstract).
Head, J.W. et al., 1993. Lunar impact basins: New data for the western limb and far
Baker, D.M.H. et al., 2011a. The transition from complex crater to peak-ring basin on
side (Orientale and South Pole-Aitken basins) from the rst Galileo yby. J.
the Moon: New observations from the Lunar Orbiter Laser Altimeter (LOLA)
Geophys. Res. 98 (1714917181), 1993. http://dx.doi.org/10.1029/93JE01278.
instrument. Icarus 214, 377393. http://dx.doi.org/10.1016/j.icarus.2011.
Head, J.W. et al., 2010a. The lunar Orientale basin: Structure and crustal mineralogy
05.030.
from Chandrayaan-1 Moon Mineralogy Mapper (M3) data. Lunar Planet. Sci. XLI,
Baker, D.M.H. et al., 2011b. The transition from complex crater to peak-ring basin on
1030 (abstract).
Mercury: New observations from MESSENGER yby data and constraints on
Head, J.W. et al., 2010b. Global distribution of large lunar craters: Implications for
basin-formation models. Planet. Space Sci. 59, 19321948. http://dx.doi.org/
resurfacing and impactor populations. Science 329, 15041507. http://
10.1016/j.pss.2011.05.010.
dx.doi.org/10.1126/science.1195050.
Baker, D.M.H., Head, J.W., Neumann, G.A., Smith, D.E., Zuber, M.T., 2012. The
Hess, P.C., 1994. Petrogenesis of lunar troctolites. J. Geophys. Res. 99, 1908319093.
transition from complex craters to multi-ring basins on the Moon: Quantitative
http://dx.doi.org/10.1029/94JE01869.
geometric properties from Lunar Reconnaissance Orbiter Laser Altimeter (LOLA)
Hikida, H., Wieczorek, M.A., 2007. Crustal thickness of the Moon: New constraints
data. J. Geophys. Res. 117, E00H16. http://dx.doi.org/10.1029/2011JE004021.
from gravity inversions using polyhedral shape models. Icarus 192, 150166.
Baldwin, R.B., 1974. Was there a terminal lunar cataclysm 3.94.0  109 years
http://dx.doi.org/10.1016/j.icarus.2007.06.015.
ago? Icarus 23, 157166. http://dx.doi.org/10.1016/0019-1035(74)90003-7.
Howard, K.A., Wilshire, H.G., 1973. Flows of impact melt at lunar craters. Lunar
Barberi, F., Varet, J., 1970. The Erta Ale volcanic range (Danakil depression, northern
Planet. Sci. VI, 28052829 (abstract).
Afar, Ethiopia). Bull. Volcan. 34, 848917.
Howard, K.A., Wilshire, H.G., 1975. Flows of impact melt at lunar craters. J. Res. U.S.
Barr, A.C., Citron, R.I., 2011. Scaling of melt production in hypervelocity impacts
Geol. Surv. 3, 237251.
from high-resolution numerical simulations. Icarus 211, 913916. http://
Howard, K.A., Wilhelms, D.E., Scott, D.H., 1974. Lunar basin formation and highland
dx.doi.org/10.1016/j.icarus.2010.10.022.
stratigraphy. Rev. Geophys. 12, 309327. http://dx.doi.org/10.1029/
Basilevsky, A.T., Neukum, G., 2010. Are All Lunar Highland Pristine Rocks Really
RG012i003p00309.
Pristine? Nrdlingen 2010: The Ries Crater, the Moon, and the Future of Human
Ivanov, B.A., 2005. Numerical modeling of the largest terrestrial meteorite craters.
Space Exploration, 7015 (abstract).
Solar Syst. Res. 39, 381409. http://dx.doi.org/10.1007/s11208-005-0051-0.
Bonini, W.E., 1982. The size of the Stillwater complex: An estimate from gravity
Jaeger, J.C., 1968. Cooling and solidication of igneous rocks. In: Hess, H.H.,
data. In: Walker, D., McCallum, I.S. (Eds.), Workshop on Magmatic Processes of
Poldervaart, A. (Eds.), Basalts. Interscience, New York, NY, pp. 503536.
Early Planetary Crusts: Magma Oceans and Stratiform Layered Intrusions. Lunar
Khan, A., Connolly, J.A.D., Maclennan, J., Mosegaard, K., 2006. Joint inversion of
and Planetary Institute, Houston, TX, pp. 5355.
seismic and gravity data for lunar composition and thermal state. Geophys. J.
Borg, L.E., Connelly, J.N., Boyet, M., Carlson, R.W., 2011. Chronological evidence that
Int. 168, 243258. http://dx.doi.org/10.1111/j.1365-246X.2006.03200.x.
the Moon is either young or did not have a global magma ocean. Nature 477,
Kring, D.A., Abramov, O., Marchi, S., 2012. Impact melt production during the basin-
7072. http://dx.doi.org/10.1038/nature10328.
forming epoch. Lunar Planet. Sci. XLIII, 1615 (abstract).
Borst, A.M., Foing, B.H., Davies, G.R., van Westrenen, W., 2012. Surface mineralogy
Kuskov, O.L., 1997. Constitution of the Moon: 4. Composition of the mantle from
and stratigraphy of the lunar South Pole-Aitken basin determined from
seismic data. Phys. Earth Planet Inter. 102, 239257. http://dx.doi.org/10.1016/
Clementine UV/VIS and NIR data. Planet. Space. Sci. 68, 7685. http://
S0031-9201(96)03259-1.
dx.doi.org/10.1016/j.pss.2011.07.020.
Marsh, B.D., 1989. Magma chambers. Annu. Rev. Earth Planet. Sci. 17, 439474.
Bratt, S.R., Solomon, S.C., Head, J.W., 1985. The evolution of impact basins: Cooling,
http://dx.doi.org/10.1146/annurev.ea.17.050189.002255.
subsidence, and thermal stress. J. Geophys. Res. 90, 1241512433. http://
Marsh, B.D., 1996. Solidication fronts and magmatic evolution. Mineral. Mag. 60,
dx.doi.org/10.1029/JB090iB14p12415.
540. http://dx.doi.org/10.1180/minmag.1996.060.398.03.
Bray, V.J. et al., 2010. New insight into lunar impact melt mobility from
McCallum, I.S., Schwartz, J.M., 2001. Lunar Mg suite: Thermobarometry and
the LRO camera. Geophys. Res. Lett. 37, L21202. http://dx.doi.org/10.1029/
petrogenesis of parental magmas. J. Geophys. Res. 106, 2796927983. http://
2010GL044666.
dx.doi.org/10.1029/2000JE001397.
Cintala, M.J., Grieve, R.A.F., 1998. Scaling impact melting and crater dimensions:
McCauley, J.F., 1977. Orientale and Caloris. Phys. Earth Planet. Inter. 15, 220250.
Implications for the lunar cratering record. Meteorit. Planet. Sci. 33, 889912.
http://dx.doi.org/10.1016/0031-9201(77)90033-4.
http://dx.doi.org/10.1111/j.1945-5100.1998.tb01695.x.
Melosh, H.J., 1989. Impact Cratering: A Geologic Process. Oxford University Press,
Collins, G.S., Melosh, H.J., Morgan, J.V., Warner, M.R., 2002. Hydrocode simulations
London.
of Chicxulub crater collapse and peak-ring formation. Icarus 157, 2433. http://
Mizutani, H., Osako, M., 1974. Elastic-wave velocities and thermal diffusivities of
dx.doi.org/10.1006/icar.2002.6822.
Apollo 17 rocks and their geophysical implications. Lunar Planet. Sci. V, 2891
Croft, S.K., 1985. The scaling of complex craters. J. Geophys. Res., Suppl. 90, C828
2901 (abstract).
C842.
Moriarty, D.P. et al., 2011. Finsen and alder: A compositional study of lunar central
Fassett, C.I., Head, J.W., Smith, D.E., Zuber, M.T., Neumann, G.A., 2011. Thickness of
peak craters in the South Pole-Aitken basin. Lunar Planet. Sci. XLII, 2564
proximal ejecta from the Orientale basin from Lunar Orbiter Laser Altimeter
(abstract).
(LOLA) data: Implications for multi-ring basin formation. Geophys. Res. Lett. 38,
Moriarty, D.P., Pieters, C.M., Isaacson, P.J., 2012. Compositional heterogeneity within
L17201. http://dx.doi.org/10.1029/2011GL048502.
lunar central peaks. Lunar Planet. Sci. XLIII, 2399 (abstract).
Garrick-Bethell, I., Zuber, M.T., 2009. Elliptical structure of the lunar South
Morrison, D.A., 1998. Did a thick South Pole-Aitken basin melt sheet differentiate to
Pole-Aitken basin. Icarus 204, 399408. http://dx.doi.org/10.1016/
form cumulates? Lunar Planet Sci. XXIX, 1657 (abstract).
j.icarus.2009.05.032.
 W.M. Vaughan et al. / Icarus 223 (2013) 749765
Nakamura, R. et al., 2009. Ultramac impact melt sheet beneath the South Pole-
Aitken basin on the Moon. Geophys. Res. Lett. 36, L22202. http://dx.doi.org/
10.1029/2009GL040765.
Nelson, S.A., Carmichael, I.S.E., 1979. Partial molar volumes of oxide components in
silicate liquids. Contrib. Mineral. Petrol. 71, 117124. http://dx.doi.org/10.1007/
BF00375427.
Nielsen, T.F.D., 2004. The shape and volume of the Skaergaard intrusion, Greenland:
Implications for mass balance and bulk composition. J. Petrol. 45, 507530.
http://dx.doi.org/10.1093/petrology/egg092.
OConnell-Cooper, C.D., Spray, J.G., 2011. Geochemistry of the impact-generated
melt sheet at Manicouagan: Evidence for fractional crystallization. J. Geophys.
Res. 116, B06204. http://dx.doi.org/10.1029/2010JB008084.
Osinski, G.R., Tornabene, L.L., Grieve, R.A.F., 2011. Impact ejecta emplacement on
terrestrial planets. Earth Planet. Sci. Lett. 310, 167181. http://dx.doi.org/
10.1016/j.epsl.2011.08.012.
Pierazzo, E., Melosh, H.J., 2000. Understanding oblique impacts from experiments,
observations, and modeling. Annu. Rev. Earth Planet. Sci. 28, 141167. http://
dx.doi.org/10.1146/annurev.earth.28.1.141.
Pierazzo, E., Vickery, A.M., Melosh, H.J., 1997. A reevaluation of impact melt
production. Icarus 127, 408423. http://dx.doi.org/10.1006/icar.1997.5713.
Pieters, C.M. et al., 1997. Mineralogy of the mac anomaly in the South Pole-Aitken
basin: Implications for excavation of the lunar mantle. Geophys. Res. Lett. 24,
19031906. http://dx.doi.org/10.1029/97GL01718.
Pieters, C.M. et al., 2009. Mineralogy of the lunar crust in spatial context: First
results from the Moon Mineralogy Mapper (M3). Lunar Planet. Sci. XL, 2052
(abstract).
Pieters, C.M. et al., 2011. Mg-spinel lithology: A new rock type on the lunar
farside. J. Geophys. Res. 116, E00G08. http://dx.doi.org/10.1029/2010JE00
3727.
Plescia, J.B., Cintala, M.J., 2012. Impact melt in small lunar highland craters. J.
Geophys. Res. 117, E00H12. http://dx.doi.org/10.1029/2011JE003941.
Prissel, T.C. et al., 2012. Meltwallrock reactions on the Moon: Experimental
constraints on the formation of newly discovered Mg-spinel anorthosites. Lunar
Planet. Sci. XLIII, 2743 (abstract).
Schultz, P.H., Papamarcos, S., 2010. Evolving owelds from the Imbrium and
Orientale impacts. Lunar Planet. Sci. XLI, 2480 (abstract).
Schultz, P.H., Stickle, A.M., Crawford, D.A., 2012. Effect of asteroid decapitation on
craters and basins. Lunar Planet. Sci. XLIII, 2428 (abstract).
Scott, D.H., McCauley, J.F., West, M.N., 1977. Geologic Map of the West Side of the
Moon. U.S. Geol. Survey Misc. Invest. Map I-1034.
Shearer, C.K. et al., 2006. Thermal and magmatic evolution of the Moon. Rev.
Mineral. Geochem. 60, 365518. http://dx.doi.org/10.2138/rmg.2006.60.4.
Smith, D.E. et al., 2010. Initial observations from the Lunar Orbiter Laser Altimeter
(LOLA). Geophys. Res. Lett. 37, L18204. http://dx.doi.org/10.1029/2010GL0
43751.
Spray, J.G., Thompson, L.M., 2008. Constraints on central uplift structure from the
Manicouagan impact crater. Meteorit. Planet. Sci. 43, 20492057. http://
dx.doi.org/10.1111/j.1945-5100.2008.tb00660.x.
Spudis, P.D., 1993. The Geology of Multi-Ring Impact Basins: The Moon and Other
Planets. Cambridge University Press, Cambridge.
Stfer, D., Knoll, H.-D., Marvin, U.B., Simonds, C.H., Warren, P.H., 1980.
Recommended classication and nomenclature of lunar highland rocks A
committee report. In: Papike, J.J., Merrill, R.B. (Eds.), Proceedings of the
765
Conference on the Lunar Highlands Crust. Pergamon Press, New York, NY, pp.
5170.
Thierrault, A.M., Fowler, A.D., Grieve, R.A.F., 2002. The Sudbury Igneous Complex: A
differentiated impact melt sheet. Econ. Geol. 97, 15211540. http://dx.doi.org/
10.2113/gsecongeo.97.7.1521.
Turcotte, D.L., Schubert, G., 2002. Geodynamics. Cambridge University Press,
Cambridge.
Warren, P.H., 1993. A concise compilation of petrologic information on possibly
pristine nonmare Moon rocks. Am. Mineral. 78, 360376.
Warren, P.H., Claeys, P., Esteban, C.-P., 1996. Mega-impact melt petrology
(Chicxulub, Sudbury, and the Moon): Effects of scale and other factors on
potential for fractional crystallization and development of cumulates. Geol. Soc.
Am. Spec. Pap. 307, 105124.
Whitten, J.S. et al., 2010. Characteristics, afnities and ages of volcanic deposits
associated with the Orientale basin from Chandrayaan-1 Moon Mineralogy
Mapper (M3) data: Mare stratigraphy. Lunar Planet. Sci. XLI, 1841 (abstract).
Whitten, J.S. et al., 2011. Lunar mare deposits associated with the Orientale impact
basin: New insights into mineralogy, history, mode of emplacement, and
relation to Orientale basin evolution from Moon Mineralogy Mapper (M3) data
from Chandrayaan-1. J. Geophys. Res. 116, E00G09. http://dx.doi.org/10.1029/
2010JE003736.
Wieczorek, M.A., Phillips, R.J., 1999. Lunar multiring basins and the cratering
process. Icarus 139, 246259. http://dx.doi.org/10.1006/icar.1999.6102.
Wieczorek, M.A. et al., 2006. The constitution and structure of the lunar interior.
Rev. Mineral. Geochem. 60, 221364. http://dx.doi.org/10.2138/rmg.2006.60.3.
Wilhelms, D.E., Howard, K.A., Wilshire, H.G., 1979. Geologic Map of the South Pole
of the Moon. U.S. Geol. Survey. Map I-1162.
Wilhelms, D.E., McCauley, J.F., Trask, N.J., 1987. The Geologic History of the Moon.
U.S. Geol. Survey. Professional Paper 1348.
Williams, D.A., Fagents, S.A., Greeley, R., 2000. A reassessment of the emplacement
and erosional potential of turbulent, low-viscosity lavas on the Moon. J.
Geophys. Res. 105, 2018920205. http://dx.doi.org/10.1029/1999JE001220.
Wilson, L., Head, J.W., 2011. Impact melt sheets in lunar basins: Estimating
thickness from cooling behavior. Lunar Planet. Sci. XLII, 1345 (abstract).
Winter, J.D., 2010. An Introduction to Igneous and Metamorphic Petrology. Prentice
Hall, New Jersey.
Worster, M.G., Huppert, H.E., Sparks, R.S.J., 1990. Convection and crystallization in
magma cooled from above. Earth Planet. Sci. Lett. 101, 789. http://dx.doi.org/
10.1016/0012-821X(90)90126-I.
Worster, M.G., Huppert, H.E., Sparks, R.S.J., 1993. The crystallization of lava lakes. J.
Geophys. Res. 98, 1589115901. http://dx.doi.org/10.1029/93JB01428.
Yamamoto, S. et al., 2010. Possible mantle origin of olivine around lunar impact
basins detected by SELENE. Nat. Geosci. 3, 533536. http://dx.doi.org/10.1038/
ngeo897.
Zieg, M.J., Marsh, B.D., 2002. Crystal size distributions and scaling laws in the
quantication of igneous textures. J. Petrol. 43, 85101. http://dx.doi.org/
10.1093/petrology/43.1.85.
Zieg, M.J., Marsh, B.D., 2005. The Sudbury Igneous Complex: Viscous emulsion
differentiation of a superheated impact melt sheet. Geol. Soc. Am. Bull. 117,
14271450. http://dx.doi.org/10.1130/B25579.1.
Zuber, M.T., Smith, D.E., Lemoine, F.G., Neumann, G.A., 1994. The shape and internal
structure of the Moon from the Clementine mission. Science 266, 18391843.
http://dx.doi.org/10.1126/science.266.5192.1839.