The Deep Structure of Lunar Basins' Implications for Basin Formation and Modification

Home , Mare Orientale

JOURNAL OF GEOPHYSICAL RESEARCH, VOL. 90, NO. B4, PAGES 3049-3064, MARCH 10, 1985

STEVEN R. BRATT AND SEAN C. SOLOMON

Departmentof Earth, Atmospheric,and Planetary Sciences,Massachusetts Institute of Technology,Cambridge

JAMES W. HEAD

Departmentof GeologicalSciences, Brown University, Providence,Rhode Island

CLIFFORD H. THURBER

Departmentof Earth and SpaceSciences, State University of New York at Stony Brook

We presentmodels for the structure of the crust and upper mantle beneath lunar impact basinsfrom an inversionof gravity and topographic data from the nearsideof the moon. All basin models display a thinner crust and an elevated Moho beneath the central basin region compared to surrounding areas, a signature of the processesof basin excavation and mantle uplift during collapse of the transient cavity. There is a general decreasein the magnitude of apparent uplift of mantle material with increasingbasin age; we attribute this relation primarily to enhanced rates of ductile flow of crustal material early in lunar history when crustal temperatureswere relatively high and the effectiveelastic lithosphere was thin. The more relaxed topographic and Moho relief associatedwith older basinson the central nearsidemay, in particular, be at least partly a consequenceof the extensivesubsurface heating associatedwith the formation of the large Procellarum basin. The deep structure of the youngest basins constrains the geometry of the cavity of excavation and the amount of crustal material ejected beyond the basin rim. From the volumes of the topographic basin, of mare basalt fill, and of uplifted mantle material, the volume of crustal material ejectedbeyond the basin rim for an Orientale-sizedevent was of the order of 107 km3. A near-constantthickness of nonmarecrustal material beneath the centralregions of young basinsof various diametersand preimpact crustal thicknessessuggests that the transient cavity excavated to at least the base of the crust for the largest basins; significant excavation into the mantle may have been impeded by an abrupt increasein strength at the lunar Moho.

INTRODUCTION data [Spudis, 1982, 1983; Spudis et al., 1984], but such esti- The formation of multiring impact basins has played a mates depend critically on the accurate identification of pri- major role in the geologicalevolution of the moon. During the mary ejecta and on assumptionsabout chemical layering of first billion years of lunar history the impact of large projec- the lunar crust. In this paper, we apply gravity and topo- tiles onto the lunar surface resulted in the excavation of basin graphic data to infer the three-dimensionalstructure of the cavities hundreds of kilometers in diameter [Wood and Head, crust and upper mantle beneath impact basins on the lunar 1976] and the implantation of large quantitiesof heat into the nearside. With the derived structural models we constrain sev- lunar interior [O'Keefe and Ahrens, 1977]. Impact basins also eral of the important geometrical and physical parameters re- became the focus for volcanic and tectonic activity over a lated to the processesof basin formation and modification, considerabletime period following the basin formation events and we assesstheir variations with basin age and size. [e.g., Head, 1976; Solomonand Head, 1979; Solomon et al., Muller and Sjoqren [1968] were the first to recognize that 1982]. Important constraintson the processesof basin forma- the youngestnearside mare basins are characterizedby posi- tion and modification are provided by the presentvolumes of tive gravity anomalies, which they attributed to "mascons." the topographicbasin, of material ejectedduring basin forma- Since that discovery, a number of efforts have been made to tion, and of mare basalt fill as well as by the degreeof involve- model the gravity anomalies over mascon basins with contri- ment of the mantle in isostatic compensationof basin relief. butions to the anomalous mass placed at the surface IConel The topographic volumes of the basins are reasonably well and Holstrom, 1968; Baldwin, 1968; Booker et al., 1970], at the known. Estimates for the volumes of mare basalt and basin lunar Moho [e.g., Wise and Yates, 1970], or at both locations ejectadeposits have also been obtained from photogeological [e.g., Hulme, 1972; Wood, 1972; Bowin et al., 1975; Sjoqren studies[e.g., Moore et al., 1974; Head et al., 1975; DeHon and and Smith, 1976]. At least some portion of the anomalous Waskom,1976; Head, 1982], but such techniquesare generally mass contributiong to mascon anomalies residesnear the sur- limited in their ability to resolvethe thicknessof the deposits. face [Phillips et al., 1972], but models where mare basalt fill is The depth of excavation of several basin-forming events on the sole source of anomalous mass require an unreasonable the moon has also been estimated from the chemistry and thickness of mare basalt to fit the measured gravity field mineralogy of ejecta deposits inferred from remote sensing [Thurber and Solomon, 1978]. While both mare fill and an elevated Moho likely contribute to the observedgravity, the solution for the distribution of anomalous mass between the Copyright 1985 by the American GeophysicalUnion. two locations given only gravity and topographic data re- Paper number 4B5151. quires additional assumptions. 0148-0227/85/004B- 5151$05.00 Structural models consistentwith gravity and topographic

3049 3050 BRATTET AL.' DEEPSTRUCTURE OF LUNARBASINS

NEARSIDE BASINS

30* N -

20* -

I0' -

O* --

30' S [AA ,I-"'• • I I IO0*W 80* 60*

Fig. 1. Outlines of the lunar basins consideredin this study. The labeled lines indicate the locations of cross sections discussed in text.

data have been constructed for a number of individual basins beneath the largest lunar basins,we derive new bounds on the [e.g., Bowin et al., 1975; Sjogrenand Smith, 1976; Phillips and volume of material ejectedfrom each basin, and we evaluate Dvorak, 1981; Janle, 1981a, b]. Most of these models,how- the implicationsof structuraldifferences among basins for the ever, were developedunder dissimilar setsof assumptionsand processesof basin formation and modification as functionsof constraints,thus hinderinga comparisonof the infe..rredstruc- time on the moon. tures beneathdifferent basins.Also, the interpretationof gravity data over a singlebasin requiresthat the investigatormake PROCEDURE subjective judgements about regional trends in the gravity To determine the crustal structure in the vicinity of lunar data arising from structures outside the area of interest or basins,we perform a simultaneousinversion of gravity and occurring over wavelengthsgreater than the scaleof the basin. topographyfor the low-latitude portion of the lunar nearside, In contrast, global models for lunar crustal structure [Wood, following a proceduresimilar to that of Thurber and Solomon 1973; Bills and Ferrari, 1977a; Thurber and Solomon, 1978] [1978]. The moon is divided into a grid of blocks, each calculatedunder uniform setsof constraintsand assumptions 5øx 5ø in horizontal extent (Figure 2). These block dimen- permit an internally consistentassessment of structural varia- sions represent the approximate limits of resolution of the bility on a regional scale.These global modelswere developed available gravity data, as discussedfurther below. The dis- to addresscrustal structure at scalesgreater than the dimen- turbing gravitational potential at any point over the nearside sionsof most lunar basins,however, and with the exceptionof can be approximated as the sum of the potentials due to the the study of Thurber and Solomon[1978], none considered distribution of anomalous mass within each block. In this mare basalt as a significantcontributor to the observedgrav- section we describethe adopted procedure for calculation of ity field. gravity anomalies,as a forward problem, given a distribution In this paper we determine modelsfor the crustal structure of topography or anomalousmass that is uniform within each in the vicinity of nine impact basins on the lunar nearside block. We then discussthe constrainingassumptions that we (Figure 1). The models are derived as part of a simultaneous have chosento make the inverseproblem well posed.Finally, inversion of nearside gravity and topographic data, using a we describean iterative, linearized inversionprocedure to de- proceduresimilar to that of Thurber and Solomon[1978]. The termine the crustal structure within each block from gravity modelsinclude a low-densitynonmare crustal layer and a and topography,subject to the adoptedconstraints. mare basalt layer, both of variable thickness. The contri- butionsof Moho relief and mare basalt to the observedgrav- Computationof Gravity Anomalies ity anomaliesare separatedwith the assumptionsthat basin In previousanalyses of the gravity anomaly fieldsof planets topography was isostatically compensated prior to mare using spherical shell segments[Morrison, 1976; Thurber and basaltfill and that crustalsubsidence in responseto loading Solomon,1978], contributionsto the anomalousmass within by mare basaltsmay be neglected.These assumptionslead to each segment have been approximated by uniform surface minimum valuesfor mare basalt thicknesses[Thurber and Sol- masses located at a constant radius from the center of the omon, 1978], but becauseof a similarity in density of mare planet. While this method simplifiesthe computation of the basaltsand mantle material, the thicknessof the low-density gravitational potential, it may provide a poor approximation nonmare crust may be reliably estimated.An important con- to the actual potential when the thickness of the block is straint is that the thickness of the nonmare crust in the vicin- similar in magnitude to the distance between the block and ity of the Apollo 12 and 14 landing sitesis known from seis- the observationpoint. mic refraction measurements[Toksoz et al., 1974]. On the As an improved approximation, we representcontributions basisof the nearsidecrustal model we comparethe structure to the anomalous mass within a spherical shell segment as BRATT ET AL.' DEEP STRUCTURE OF LUNAR BASINS 3051

U[r + Ar(r/r)] -- U(r) g(r) = (5) Ar

Equation (3) is used to compute the contributions from blocks beneath and near the observation point. For a block more 5 ø than 900 km distant from the observation point the distribution of anomalousmass in the right-hand side of (1) for that MOHO RELIEF block is representedby four point masses;for a block more CONTRIBUTION than 1200 km distant, a single point mass is used. These Apr. simple approximations provide accuracy comparable to that of (3) with a considerablesavings of computation time. MARE BASALT Constraints and Assumptions CONTRIBUTION APb _ Following Thurber and Solomon [1978], we assume that Airy compensationdominates on the moon and that there are TOPOGRAPHI• three principal contributions from each block to the anoma- CONTRIBUTION Pc lous gravity field (Figure 2): (1) surface topography, measured relative to a datum Dr and contributing to gravity in proportion to the density Pc of the upper crust, (2) Moho relief, measuredrelative to a datum D•t and contributing to gravity in proportion to the density contrast Apm-- Pm- Pc between the densitiesPc and Pmof the crust and mantle, respectively, and (3) mare basalt fill with a density contrast Apo = po- Pc. We adopt the valuesPc = 2.9 g/cm3, Pm= 3.4 g/cm3, and po- 3.4 g/cm3, estimatesthat are uncertainby about 0.1 g/cm3 [Solomon,1975]. Thus Apb= Apm = 0.5 g/cm3 = Ap in this problem. The contribution of topography is evaluated using (3)-(5) and is subtractedfrom the free air anomaly field to yield the Bouguer anomaly field. Fig. 2. Schematicrepresentation of the block model used for cal- culating contributionsto the free air gravity anomaly from topogra- To invert formally the Bouguer anomaly field for crustal phy (relative to the datum DT), Moho relief (relative to the datum structure,it is necessaryto impose two additional constraints D•4), and mare basaltfill. [Thurber and Solomon, 1978]. The first constraint is that by assumption, (1) topography in mare areas was isostatically compensatedby an Airy mechanismbefore the emplacement arisingfrom a finite volumeof thicknessbr and of latitudinal of mare basalts and (2) no compensation of mare units has andlongitudinal extent bo and b,, respectively.In general,the occurredsince their emplacement.These two assumptions,for disturbingpotential at a point r abovethe lunar surfacedue to which we use the term "premare isostasyconstraint" below, a blockof anomalousmass within a sphericalshell segment is permit a unique decompositionof the gravity anomaly into contributions from Moho relief and mare fill. The first as- u(r)=f f ; FdO' d•b'dr' (1) sumption receivessupport from the argument that the crust volume surroundinga recentlyformed impact basin was likely to have of block been hot and incapable of supporting the large deviatoric where stressesassociated with a deep basin depression[e.g., Bratt et (r')2 COS0' al., 1981]. The secondassumption is not strictly correct; mare = (2) ridges and linear rilles provide evidence that the lunar litho- spherehas subsidedin responseto the mare basalt load in the and G is the gravitationalconstant, p is the density(or density mascon maria [Solomon and Head, 1979, 1980]. This second contrast),0 is the latitude, •p is the longitude,and primed assumption, however, has little influence on the estimated variables denote coordinates of anomalous mass within the thickness of nonmare crustal material. The effects of both as- block. All radiusvectors are with respectto the lunar centerof sumptionson the derived structuralmodels are discussedfur- mass.If F is expandedto secondorder in a Taylor series ther below. Under the premare isostasyconstraint, the basalt about the centerof the block, equation(1) simplifiesto thicknesstj and the depthto the Moho dj within any mare u(r)= bob•,br[F+ 7!,½(Foobo 2 + F•,•,b•, 2 + F•b•2)] (3) block j are related by the equation wherethe doublesubscripts designate second derivatives of F (tj + hfpc = (D, -- dj)(pm - Pc) (6) with respectto the primedcoordinates evaluated at the center whereh• is the topography.Both t• and d• are measuredrela- of each block. tive to the surface of a standard section which is assumed to The gravitational potential due to the anomalousmass be in isostatic equilibrium and which has a known depth to within all M blocksin the grid is obtainedby summation: Moho D•. This relationship is illustrated in Figure 3.

M The second constraint on the derived crustal models follows U(r)- • %(r) (4) from the fact that the crust in the general vicinity of the j=l Apollo 12 and 14 landing sites(--•3øS, 20øW) is known from and the free air gravity anomaly field g(r) is obtained by the seismicrefraction measurementsto be 55 km thick [Toksoz et simplefinite differencerelation: al., 1974]. Becausethe Apollo 12 and 14 sites are located in 3052 BRATT ET AL.: DEEP STRUCTURE OF LUNAR BASINS

anomalousmass by the relationship M •]k calc AQk= • Abj (9)

D x ,N• t Ag = P Ab (10) where Ag is an N x 1 column vector of residual gravity anomalies,P is an N x M matrix of partial derivatives(P• = -G Ap A• coso•kj/r•2), and Ab is an M x 1 vectorof thickness corrections to the anomalous mass. For N > M, (10) can be treated as a least squaresproblem • MAREBASALT • CRUST and inverted by any number of methods [-e.g.,Lawson and ,,-• MANTLE Hanson, 1974-] to obtain the corrections to the thickness of anomalous mass in each block. These corrections are then Fig. 3. Schematiccross sectionof the crust and upper mantle, added to the mass thicknesses in the crustal model from the illustratinghow the assumptionof premareisostasy is usedto constrain the distribution of anomalous mass in lunar mare regions.We prior iteration, subjectto the constraint of premare isostasy, use equation (6) and the observedtopography h to constrainthe imposedwhere applicable,to separatethe contributionsfrom thickness t of mare basalt and the depth d to the Moho in mare Moho relief and mare fill. After each iteration the global blocks.The seismicallydetermined crustal thickness in the vicinity of datum D M for Moho relief is adjustedto meet the constraint the Apollo 12 and 14 landingsites serves as a referencesection. on crustal thickness inferred from seismic measurements for the region of the Apollo 12 and 14 landing sites.Finally, a new Bougueranomaly field •calcis computedfrom the adjusted regionsof ancientpremare structure and thin mare fill in the model using (3)-(5), the rms residualgravity anomaly is deter- OceanusProcellarum-Mare Cognitum area, it is reasonable mined from (7), and the inversion processis repeateduntil the to assumethat the degreeof isostasyin the region is nearly rms residual convergesto a minimum. complete.We thereforeadopt a 55-km-thicklayer of nonmare We impose one additional constraint on blocks in mare crustal material of density Pc as the standard sectionfor (6). regions. If the Bouguer anomaly over a mare block is suf- With this assumptionthe model for the nearsidecrustal struc- ficiently small, the premare isostasyconstraint may not be ture developedin this studyautomatically matches the crustal consistent with a finite thickness of basalt. Therefore, if the thicknessin the vicinity of the Apollo 12 and 14 landing sites. adjustmentto the basalt thicknesswithin a mare block forces the total mare thickness to fall below 250 m, the constraint of Inversionfor Crustal Structure premareisostasy for that block is relaxedand the basalt thicknessis held constant at 250 m for the remaining iterations. To determine the Moho configuration and thicknessof A substantialsavings in computation time and core storage mare basalt on the lunar nearside, we invert the Bouguer was achievedby filling the matrix P only with those partial anomaly data, subjectto the adopted assumptionsand con- derivativesassociated with the block directly below the obser- straints. The inversion schemeiteratively improves on an ini- vation point and the eight surroundingblocks. This approxi- tial estimate of the crustal structure. The criterion used to mation is equivalent to assumingthat the residual anomaly determine a best fit model is the minimization of the root- above a given block is causedonly by errors in the adopted mean-squareresidual gravity anomaly: thicknesses of the anomalous mass in the nine nearest blocks. The resultingpartial derivative matrix is then sparseand can rmsresidual = y,.N(•/no• _ qncalc)2/N (7) be easily manipulated into a diagonally dominant form. We k=l solve for the vector Ab of thicknesscorrections by converting

obs where N is the number of gravity anomaly observations,•/n the matrix P to upper triangular form by using a series of is thekth observedgravity anomaly, and •/n•a• is thekth grav- Householder transformationsapplied to sequentiallyaccumu- ity anomalycalculated from the block modelusing the nonlin- lated rows [Lawson and Hanson, 1974, pp. 207-311-l. ear formulations given in (3)-(5). GRAVITY AND TOPOGRAPHIC DATA At each iteration in the inversion scheme, (3)-(5) are approximated by a linear relation betweenthe Bouguergravity The inversion procedure described above to determine anomalyand the thicknessof anomalousmass within a block: crustaland upper mantle structurewould be straightforwardif the free air anomaly and topography were known everywhere •]kcalc = --GAp •M Ajbjcos 2 Ot•j k=1, 2, ." , N (8) on the moon within a well-prescribeduncertainty. Such, un- j: 1 rkj fortunately, is not the case.Information on the lunar gravity wherebj is the totalthickness of the anomalousmass (Moho field is derived from measurementsof Doppler shifts in the uplift plus mare fill) in thejth block, Ap is the densitycontrast frequencyof radio transmissionsalong a line-of-sight(LOS) betweenthe block and the surrounding crust (0.5 g/cm3), Aj is direction between the earth and a satellite in orbit about the the surfacearea of thejth block,rkj is the vectorbetween the moon. High-frequencyvariations in these Doppler observa- kth observationpoint and the centerof massof all anomalous tions contain a signature of gravity anomalies arising from massin thejth block,and •kj is the angleat the observation near-surfaceheterogeneities in density. To extract this signa- pointbetween rkj andthe downward vertical. From the linear ture, the effect of orbital parameterson the LOS Doppler data relation(8) between•]kcalc and bj it followsthat Agn = gnca•cmust be removed or modeled. The estimation of a repre- _ qnobscan be linked to perturbationsin the thicknessof sentationof the free air gravity field over a significantarea of BRATT ET AL.' DEEP STRUCTURE OF LUNAR BASINS 3053

FREE AIR GRAVITY ANOMALY ALTITUDE = I00 km CONTOUR INTERVAL = 25 mgels

I I I I I I e '1 I I I I I 30øN 20"

0 o 7,50 • -

i0 • 75

20 •

30•S

I00 ø W 80 ø 60 ø 40 ø 20 ø 0 ø •0 ø 40 ø 60 ø 80 ø IOOøE Fig. 4. Free air gravity anomaliesover the lunar nearsideat 1• km elevation (1836 km from the lunar center of mass). Gravity anomalies were computed from the disk mass model of Wong et al. [1975] superimposedon the triaxial gravity model of Liu and Laing [1971]. The contour interval is 25 mgal. The peak anomaliesover severalbasins are also indicated.

the lunar surfacerequires the simultaneousinversion of Dop- within + 7 mgal. For instance, the LOS acceleration residual pler data from a large number of orbits for the parameters of for the Apollo 15 subsatellite at 100-120 km altitude over both the spacecraftorbits and the gravity field. The first such Serenitatis,Crisium, and Smythii for orbit 1516 are 6, 4, and 6 representationfor a large fraction of the lunar nearside was mgal, respectively. While these LOS acceleration residuals describedby Wong et al. [1971]. They presented both point over the central nearside (e.g., Serenitatis) provide a measure mass and disk mass representationsof lunar gravity from of the error in the derived free air gravity anomaly, the re- Lunar orbiter and early Apollo tracking data as well as a full sidualsat greater longitudes(e.g., Smythii) reflect more nearly discussionof the proceduresused in data inversion. horizontal spacecraftaccelerations and are lesssimply related For this study we employ an improved representationof the to errors in the gravity field. For the purposeof estimatingthe low-latitude nearside gravity field obtained by Wong et al. error in the derived crustal structure in the next section of this [1975] from an inversion of tracking data from low-altitude paper, we estimate that the free air anomalies at 100 km eleva- Apollo spacecraft.Their representationconsists of 350 near- tion shownin Figure 4 have an associatederror of + 10 mgal surface disk masses distributed between +30 ø latitude and over the central nearsideand +20 mgal at the corners of the + 100ø longitude and superimposedon the triaxial model for area representedby the disk mass solution. lunar gravity of Liu and Laing [1971]. The location and size of Information on lunar topography comes from a variety of each disk were assigneda priori, either on the basisof geologi- sources,including Apollo laser altimetry [Robersonand Kaula, cal information or in an otherwiseregular spacing,generally 1972; Wollenhaupt and Sjogren, 1972; Wollenhaupt et al., 5ø in latitude and longitude. This spacingis a measure of the 1974; Kaula et al., 1972, 1973, 1974], landmark tracking [Wol- horizontal resolution of the gravity field model. Disk radii lenhauptet al., 1974-[,limb profiling [-Watts, 1963-[,and photo- range from 75 to 300 km; the largest disks represent major grammetry [-Hopmann,1967; Mills and Sudbury, 1968; Arthur impact basins. The mass of each disk was derived from the and Bates, 1968-[.Bills and Ferrari [1977b] synthesizeda large simultaneousinversion of a large quantity of Doppler tracking number of these topographic measurementsfrom all sources data from Apollo 14, 15, and 16 spacecraft.Orbits used in the and placed them in a consistent center-of-mass coordinate inversion ranged from 15 to 200 km in altitude and +30 ø in system. latitude. In this paper we use averages,in blocks of 5ø latitude by 5ø The free air gravity anomaly at an altitude of 100 km (or a longitude, of 15,887 observationsof nearsidetopography com- radius of 1836 km from the lunar center of mass) may be piled by B. G. Bills (personalcommunication, 1982). There are readily calculated from the disk model of Wong et al. [1975] a total of 672 blocks spanning the latitude range +40 ø and and is shown in Figure 4 (a similar figure is given by Sjogren the longitude range + 105ø. In regions where the topography [1974]). Unfortunately, the error in this free air anomaly field is poorly resolved,values from the harmonic representationof is not as readily determined. Wong et al. [1975] did not deter- Bills and Ferrari [1977b] supplementthe block averages.The mine formal errors in the mass of each disk. An estimate of the topography of the eastern half of the Orientale basin, the typical uncertainty in the free air gravity field may be ob- youngestof the nearside basins,has recently been reevaluated tained, however, from the time derivatives of the differences by Head et al. [1981• using earth-based telescopicmeasure- between the LOS Doppler observationsand the predictions of ments of limb heights [Watts, 1963•. Apollo laser altimetry the disk mass model of Wong et al. [1975]. For five orbits over the northern edge of Orientale suggeststhat while these over the central nearsidewith Doppler residual data displayed limb height measurementsprovide an accuraterepresentation by Wong et al. [1975] and with spacecraftaltitudes between of the basin relief, they require a downward adjustment of 80 and 160 km, the residual LOS acceleration is typically about 2 km to fit smoothly with the other topographic data 3054 BRATT ET AL.' DEEP STRUCTURE OF LUNAR BASINS

DATUM = 1730 km TOPOGRAPHY CONTOUR INTERVAL = I km

20øI0 o 6/1

0 o

I0 o

20 ø

I00 o W BO ø 60 ø 40 ø 20 ø 0 o 20 ø 40 ø 60 ø BO o I0 Fig. 5. Topographyof the lunarnearside, relative to a datumat 1730km radius.The map is obtainedfrom 5ø x 5ø averagesof topographycompiled by Bills and Ferrari [1977b] and alsoincorporates limb heightobservations of the Orientalebasin summarizedby Head et al. [1981]. The contourinterval is 1 km. The elevationminima within several basins are also indicated.

referenced to the lunar center of mass. The Orientale limb mark tracking, earth-basedphotogrammetry, and limb profil- height measurements,so adjusted, complete the topographic ing range from 0.4 to 0.9 km. data set. A contour map of the topographyis shownin Figure The error in the average topography over a 5ø x 5ø block 5. can be estimated from the uncertainties in the individual de- The accuracyof the block-averagedtopography depends on terminations within that block and from the assumptionthat both the technique employed in making individual measure- all individual measurementsare independent.The estimated ments and the number of measurements in each block. For the error in the average topography over Orientale, Humorum, individual determinations of topographic height from Apollo and Nubium for some blocks approachesthe estimatesfor laser altimetry and orbital photogrammetry, Bills and Ferrari individual observationsbecause these basins are poorly cov- [1977b] have estimated an error of _+0.3km. The majority of ered by spacecrafttracking data. As a result, errors in topog- these measurements were made between 15øS and 40øN lati- raphy over thesebasins represent a large sourceof uncertainty tude on the nearsideof the moon. Error estimatesfor topogra- in the analysisof associatedbasin structure.For example,the phy measuredover the remainder of the nearsideusing land- contribution to the Bouguer correction from an 0.5-km error

BOUGUER GRAVITY ANOMALY ALTITUDE: IOOkm CONTOUR INTERVAL =50 mgals

30ø N 20(•0 -

20 ø

10 ø

0 o

I0 ø

20 ø

30øS

1 I O0 oW 80 ø 60 ø 40 ø 20 ø 0 ø 20 ø 4 O* 60 ø 80 ø O0 ø E Fig. 6. Bouguergravity anomaly for the lunar nearsideat 100 km elevation.A datum D r at a radius of 1736 km was used to make the topographic correction. The contour interval is 50 mgal. Local maxima over several basins are also indicated. BRATTET AL.' DEEPSTRUCTURE OF LUNAR BASINS 3055

CRUSTAL THICKNESS CONTOUR INTERVAL= I0 km

30øN

20 ø

I0 o

0 o

I0 o

20"

30'S

I O0 øW 80 ø 60 ø 40 ø 2.0ø 0 ø 2.0ø 40 ø 60 ø 80 ø I O0 ø E Fig. 7. Crustalthickness (not including mare basalt fill) on the lunar nearside. This model is derived from the Bouguer anomalydata of Figure6, the constraintthat the crustbe 55 km thick beneaththe Apollo 12 and 14 sites,and the assumptionsthat in mareregions the premaretopography was completely compensated by an Airy mechanismand no furthercompensation followed the emplacementof mareunits (see text). Contour interval is 10 km.

in topographywithin a 5ø x 5ø block of crustof density2.9 the kth block due to uncertaintyin both the gravity and topo- g/cm3 near the lunar equatoris about20 mgal at 100 km graphicdata setsmay be estimatedfrom

altitude,or up to twicethe estimateduncertainty in the free M air anomalyderived from the massdisk gravity model. Be- (ag)•,2 = (aj.)•,2 + • (a,)j•,2 k = 1, 2, ..., N (11) causethe total error in the Bouguercorrection is a function of j=l errors in topography everywhere,a baselineerror in the to- where(as)•, is the errorin the freeair gravityanomaly over the pographyover a larger regionwould producea still greater kth block,and (at)j•,is the errorin the Bouguercorrection for error in the calculatedgravity anomaly. the kth block due to the uncertaintyin mean topographyfor Fortunately, the large quantity of topographic data col- the jth block. The area of most reliable Bouguergravity lies lected by orbiting spacecraftfrom the central nearsideresults beneaththe orbital tracks of the Apollo spacecraft.This area in greatlyimproved estimates of the averagetopography for includesthe Serenitatis,Tranquillitatis, Crisium, Fecunditatis, 5ø x 5ø blocksin that region.Uncertainty in the meantopog- Nectaris, and Smythii basins(Figure 1). The Bouguergravity raphy of blockscovered by Apollo orbital tracking,some of anomalies over Orientale, Humorum, and Nubium are less which are sampledby as many as 135 observations,may be accuratelyrepresented, but thesebasins have nonethelessbeen lessthan 0.1 km. Thus the contributionof uncertaintyin to- includedin our study for completeness.Signatures of the Im- pographyto errors in the Bougueranomaly over the central nearside is small. brium and Australebasins are visiblein the gravity and topographic data, but the structuresdetermined in this study for The "observed"Bouguer anomalies are calculatedabove the theseregions, located at the edgeof the area of good coverage, center of each block at two altitudes, 100 and 200 km, which shouldbe regardedas highly uncertain. bracket the altitudesof the majority of Apollo 14, 15, and 16

orbits used to derive the disk gravity model [Wong et al., BASIN STRUCTURAL MODELS 1975]. We found that using 1344, rather than 672, gravity "observations"greatly improvesthe stability of the inversion We have inverted the Bouguer gravity anomaly field of and actsto smooththe resultingstructural model. Of course, Figure 6, following the inversionprocedure described above, the 1344 "observations"are not fully independent.The mass to obtain values for the thicknesses of the nonmare and mare diskmodel is describedby 1400parameters, but many of these basalt components of the crust in each of the 672 blocks on parametersare stronglycorrelated [Wong et al., 1975]. the lunar nearside. Maps of these quantities are shown in The Bouguergravity anomaly field obtained from the appli- Figures 7 and 8, respectively.The crustal thicknessesin blocks cation of (3)-(5) to the observedtopography (Figure 5) and containingthe Apollo 12 and 14 landing sitesare within 1.5 freeair gravityanomalies (Figure 4) is shownin Figure6. The km of the seismicallymeasured value of 55 km. After the final topographicdatum Dr assumedin calculatingthe topographic iteration,the rms residualBouguer anomaly was 6.5 mgal. correction is a sphere of radius of 1736 km, selectedto be The uncertainty(ac)•, in the thicknesstc of nonmarecrustal representativeof the typical elevation of central nearside re- material in the kth block can be estimatedapproximately with gions.The masconbasins Orientale, Humorum, Imbrium, Ser- the assumptionthat it is primarily the resultof the uncertainty enitatis, Nectaris, Crisium, and Smythii display the most in the Bouguer anomaly for that block. Under this assump- prominent Bougueranomalies on the figure. Orientale, Ser- tion, enitatis,and Crisiumare sitesof the largestanomalies (> 200 mgal) relativeto surroundingregions. (ac)k= Sk2(%)k (12) The estimatederror (ag)•, in theBouguer anomaly field over GA•, Ap 3056 BRATT ET AL..' DEEP STRUCTUREOF LUNAR BASINS

MARE BASALT THICKNESS CONTOUR INTERVAL = I km

30 ø N -- ii

[/ t•-- ,,, .....,) ( •\\ 20 o _ iJ J J ] ] • I ]•v•/•x I J/l/J ]/• • "-x] ] J x• •.• // // / / \I /-' - .. -- i0 o -- ', , ' ' ', tY :_:,_. ".... --_ • _..,/' /// '"-• "' / • ".... - 0 o • / ") \',, _. I II,, I(•,,' -

• /"x [---- x I0 o _ (.,.-_,,• '-" ' / ', / x"-• /"•-" \ I j/ \ i - \\ 4.5- I,'.•-, ', ', ,', : ..'/,' ,

., ., :3.8 i

30 øS I• i + ,' ,•

, ,... ,...,', .. , , , , , , 20øiO0øW• "•'•l80 ø 60" 40 ø 20 ø 0 ø 20 ø 40" 60" 80" I0( Fig. 8. Mare basaltthickness on the lunar nearside.See Figure 7 (and text) for model assumptions.The dashedline indicatesthe boundary of mare regions.Contour interval is 1 km.

where sk is the distancefrom the observationpoint (100 km resultingstructure similar to that of Figures 7 and 8; even in elevation) to a point located at the centroid of the block of blocksat the edge of the grid the crustal thicknessesdiffer by nonmare crustal material. We use sk= 127.5 km for all k. The only a few percentfor differentvalues of Dr. Another source estimated uncertaintiesin the values of t½shown in Figure 7 of error in the computed structure is the uncertainty (_+0.1 are displayedin map form in Figure 9. The estimatederror in g/cm3) in the assigneddensities. An underestimateof the den- t½over most of the nearsideranges from about 2 to 3 km in sity contrastbetween crustal and mare or mantle material of areas beneath the tracks of Apollo spacecraft(between 15øS 20%, for instance,results in an overestimateof both the Moho and 40øN latitude) to about 5 km at high latitudes. Errors in relief and mare basalt thicknessby about the samepercentage. t½derived for the westernlimb, includingthe Orientale basin, It is important to recognizethat relaxation of the p•remare are generallylarger (up to 12 km) becauseof the uncertainty isostasyconstraint would not alter significantlythe valuesof in the correct baseline for the local topography. Near the crustal thicknessdepicted in Figure 7. If isostaticcompensa- edgesof the grid area, additional errors may arisefrom neglect tion of the premare basin topographywas lessthan complete of structuralvariations beyond the grid and from the assump- or if some percentageof the mare fill has been compensated, tion of a uniform topographic datum. We repeatedthe inver- the actual thicknessof mare basalt would be greater than in sion usingBouguer anomaly data obtainedwith Dr larger by our model by some amount. Becausemantle material and up to 3 km than the value assumed above and found the mare basalt are similar in density, however, the Moho relief

UNCERTAINTY IN CRUSTAL THICKNESS CONTOUR INTERVAL: 2 km

20 ø

I0 o

0 o

I0 o

ZOo

30 ø S

I00 ø W 80* 60* 40* 20* O* 20* 40* 60* 80' I00' E Fig. 9. Estimatederror in crustal thickness(exclusive of mare basalt fill) resultingfrom uncertaintiesin gravity and topographicdata as calculatedusing equation (12). The contourinterval is 2 km. BRATT ET AL.' DEEP STRUCTURE OF LUNAR BASINS 3057

RADIAL DISTANCE,km

400 200 0 200 400 600 400 200 0 200 400 400 200 0 200 400 600 I I I I I I I I I I I I I I I I I I I I I I I I I I I I I I I I I I I I A -IO B B' C C• - • • - BASEOF o (b) (c) MAREUNITS IO - ORIENTALE SERENITATIS CRISIUM 2o 30-

40- 50- i 60- 70-

80- i

D D / E E / F F • -I0 -

- (d) (e) (f) E• E •o - HUMORUM NECTARIS SMYTHII -'020

I - •- õO- r• 60- 70- 80-

G G' H -I0

o (g) (h) (i) IO NUBIUM FEC U N D I TATIS TRANQUI LLITATIS 20

60 i

8O5O f• '"-• Fig. 10. Cross sectionsof crustal structure for nine basins on the lunar nearside. Dashed lines denote the Moho as determinedfrom the block model along the profile lines shownin Figure 1. The solid curvesdepict smoothedprofiles of the lunar Moho and the basin topographyat the baseof the mare basalt (seetext). Depths are relative to a datum at 1730 km radius.Basins are illustratedin order of increasingage [Wilhelms, 1981]: (a) Orientale,(b) Serenitatis,(c) Crisium,(d) Humorurn,(e) Nectaris,(f) Smythii,(g) Nubium, (h) Fecunditatis,and (i) Tranquillitatis.

would be less by an approximately equal amount. This ap- formation of the transientcavity and collapseof the cavity by proximately even trade-off of anomalous massesoccurs be- some combination of inward and upward flow of crustal and cause the Bouguer anomaly, when measured 100-200 km mantle material [e.g., Melosh and McKinnon, 1978]. By this above the sources of anomalous mass, is much more sensitive interpretation, the Moho relief beneath a basin region is a to the total anomalous mass than to the distribution of that measure of the amount of uplift of mantle material during mass between the surfaceand Moho. If high-densitybasaltic cavity collapse,an important parameter in describingthe ther- to gabbroicintrusions (dikes, sills) permeate the crust beneath mal evolution of a basin region [Bratt et al., 1981]. Of course, the basalt fill in large basins[e.g., Janle, 1981a, b], then lesser the crustal model of Figure 7 reflectsonly the presentMoho amountsof Moho relief and mare fill than shownin Figures 7 relief beneath basins; the relief immediately following basin and 8 would be necessaryto match the observedgravity. The formation may have been substantiallymodified by the effects equivalentthickness of low-densitycrustal material, however, of ductile flow in the crust. would not differ greatly from the values shown in Figure 7. Cross sections through the basin structural models are Though some systemof magma conduits clearly must have shown in Figure 10. The dashed lines show the Moho depth existed at depth to feed the extensivemare units within most accordingto the block model along the profile lines shown in basins,the volume of such structuresis not known. Figure 1. The solid lines show smoothedcross sections of both The crustal thicknessmodel of Figure 7 indicatesthat the the Moho and the baseof mare basalt deposits.The smoothed crust is thinner beneatheach of the major nearsidebasins than profiles were obtained from the contour maps in Figures 7 in immediately surrounding areas. Such a result, of course, and 8 and represent azimuthally averaged cross sections. would follow at least qualitatively from the assumption of From the smoothed Moho profiles and the assumption of premare isostasy.We interpret this thinner crust to be prin- cylindrical symmetry, we have estimated the magnitude of cipally the result of the basin formation process,including Moho relief and the apparent volume V, of uplifted mantle 3058 BRATT ET AL.: DEEP STRUCTUREOF LUNAR BASINS

TABLE 1. Estimates of Moho Relief, Apparent Volume of Mantle Uplift, Basin Volume, and Ejecta Volume for Major Basinson the Lunar Nearside

Radius, Moho Ambient Crustal Vu, V0, Ve, Basin km Relief,km Thickness,km 106 km3 106 km 3 106 km3

Orientale 310 66 85 6 0.7 7 Serenitatis 305 30 60 8 1 9 Crisium 225 31 55 6 1 7 Humorurn 205 27 60 3 0.5 4 Nectaris 300 28 60 4 0.7 5 Smythii 300 33 65 4 0.6 5 Nubium 340 10 60 2 0.3 2 Fecunditatis 345 14 60 2 0.2 2 Tranquillitatis 340 12 60 2 0.2 2

Basinsare listed in order of increasingage [Wilhelms, 1981]. The indicatedradius correspondsto the secondring [Head, 1977; Wilhelrns,1981].

beneath each basin. These quantities are listed in Table 1, in 1.3 km, while the younger circular mare basins (Orientale, orderof increasingbasin age [Wilhelms, 1981]. Serenitatis, Crisium, Humorum, Nectaris, and Smythii) con- Table 1 and Figure 10 suggesta general decreasein the tain as much as 3.5-4.5 km of basalt fill within the central apparent volume of uplifted mantle with increasingbasin age. depressionregion. That these thicknessesin the central por- The Moho beneath the central mare units of Orientale (Figure tions of the circular basinsare greaterby up to 2 km than the 10a) is raised by 66 km relative to the ambient depth of about valuesderived by Thurber and Solomon[1978] under a similar 80 km. Serenitatis(Figure 10b), a basin about the same diam- set of assumptionscan be attributed to the improvedaccuracy eter as Orientale and of somewhatgreater age, is surrounded of the topographicdata usedhere and to the method adopted by crust with an average thicknessof 55 km; mantle relief to evaluatethe gravitational potential due to anomalousmass beneath Serenitatis is 30 km, but the uplifted volume is (equation (3)). It should be noted, however, that the region of broaderand thus somewhatlarger (7.5 x 106 km3 versus6.0 thickestmare fill in each of the Orientale and Smythii basins x 106 km3) than that beneathOrientale. The volumeof ap- constitutesonly a singleblock in the model depictedin Figure parent mantle uplift beneath Orientale is also more uncertain, 8. In someirregular mare regions(e.g., Oceanus Procellarum) however,because of poor data coverageover the westernpor- the constraintof premareisostasy was relaxedduring the in- tion of the basin. The difference in the magnitude of Moho version because the basalt thickness within one or more relief between the two basins nonetheless exceeds the sum of blocks dropped below the 250-m limit after an iteration. For the errors in these two quantities (Figure 9). The oldest basin such regions,either the surficial mare basalt unit is very thin examined is Tranquillitatis (Figure 10i). Though slightly larger or the crust is somewhatthicker than would be predictedfrom in diameter than the nearby Serenitatis basin, Moho relief completeisostatic compensation of premare relief. beneath Tranquillitatis is only 12 km, also a significantdiffer- As noted earlier, the mare basalt thicknessesdepicted in ence in relation to the errors shown in Figure 9. We interpret Figure 8 should probably be regarded only as lower bounds the decrease in the apparent mantle uplift with increasing on the actual thicknesseswithin the central regions of most basin age (Table 1 and Figure 10) to be the result of an age basins[Thurber and Solomon,1978]. If the lunar lithosphere dependenceof the basin formation or modification processes. maintained finite strengthduring the formation of at least the We consider the implications of this inferencein a later sec- younger basins, then the premare basin would have had in- tion. completelycompensated relief and a greater thicknessof mare Figure 7 shows a generally thinner crust over the central basalt would be required to match the Bouguergravity field. nearsideof the moon than near the limb regions,a result also Further, some portion of the mare basalt load in the mascon noted in several earlier studiesof lunar crustal structure [e.g., maria has likely been compensatedby lithospheric flexure Kaula et al., 1974]. This large region of thinner crust coincides [e.g., Solomonand Head, 1979, 1980], which also would lead approximately with the location of the ancient Procellarum to mare units somewhat thicker and Moho relief somewhat basin [Whitaker, 1981], a 3200-km-diameter structure cen- lessthan indicated by the structural model presentedhere. It tered near 26øN and 15øW. This coincidencesuggests that is interesting to note that mare thicknesses estimated by some record of mantle uplift during basin formation is still DeHon and Waskom [1976] and DeHon [1978] from rim apparent even for the largest and one of the oldest of the heightsof craters buried or partially buried by mare deposits preservednearside basins [Wilhelms, 1981]. are generally comparable to the values in Figure 8 for many The thicknessesof mare basalt, shown in map view in mare regions.This similarity between the two determinations Figure 8, are deserving of several comments. The inversion is probably coincidental, however, because the rim height procedure preservedthe constraint of premare isostasywithin technique tends to underestimatemare thicknessby as much all major basins for the complete set of iterations. Equiva- as a factor of 2 [Head, 1982]. lently, the Bouguer gravity anomaly over these basinsis suf- It is also possiblethat the mare basalt thicknessesshown in ficiently large so as to require both an elevated Moho and Figure 8 for some regions may be overestimatesbecause of a more than 250 m of high density mare basalt. Basalt thick- significant contribution to the gravity field from subsurface nesses beneath the centers of the irregular mare basins intrusions of high-density mare basaltic material. The thick- (Nubium, Fecunditatis,and Tranquillitatis) range from 0.5 to ness of mare basalt within the central block of the Orientale BRATT ET AL.: DEEP STRUCTURE OF LUNAR BASINS 3059 basin, for instance,is 3-4 km in Figure 8. On photogeological al., 1984]. These estimates have considerable uncertainty, grounds,however, the basalt thicknesshas been estimated to however, associatedwith the reality of the assumed simple be no greater than 1 km [Head, 1974]. A significant excess model of crustal layering and with the ability to separatepri- mass within the submare crust is a possible explanation for mary from secondaryejecta. this discrepancy.The large uncertainty in the topography and The structural modelsfor lunar nearsidebasins depicted in gravity in the Orientale region may be at least an equal con- Figure 10 provide a straightforward and internally consistent tributor, however. framework with which to addresscavity and ejecta volumes. Let V0 be the volume of the present basin, including the vol- GEOLOGICAL IMPLICATIONS umes of the topographic depressionand of mare fill. We make The structural models derived in this paper have a number the assumption that the crustal thickness of the preimpact of implications for the processesthat accompaniedthe forma- basin region was similar to the present thicknessat 2-3 basin tion and evolution of large impact basins on the moon. We radii from the basin center (excluding other nearby basins). divide the discussioninto basin formation processes,including Then the quantity Ve --- Vb -}- Vu,where V• is the volume of transient cavity collapseand emplacementof ejecta deposits, apparently uplifted mantle, provides a measure of preimpact and basin modification processes,including volcanism and crustal material now external to the basin rim. lithosphericdeformation on time scaleslonger than those nor- The quantityVe should be somewhatless than Vp.This is mally associatedwith the basin formation event.We recognize because inward transport of crustal material during basin that this division is somewhat arbitrary and that the present modification as well as deposition of ejecta from younger characteristicsof each basin may be the product of a combi- basins or of nonmare volcanic material will act to reduce one nation of both typesof processes. or both of V0 and V•. Therefore Ve can be regarded as the difference between the volume of crustal material ejected Early Cavity Geometry beyond the basin rim and the volume of crust returned to the The volume of material ejectedfrom a basin during impact basin region by these later processes.For the youngestnear- is an important quantity, for several reasons.Its relation to side basins, which may have formed at a time when near- basin diameter provides a strong constraint on the cratering surfacetemperatures were comparatively low and the elastic processfor large impacts,particularly on the early cavity ge- lithosphere relatively thick (e.g., Orientale, Serenitatis, and ometry. The ejecta volume also provides information on the Crisium),Ve may approach •. sampling depths of large impacts and thus on the interpreta- Values for Ve derived from the smoothedcross sections of tion of returned lunar samplesfor crustal composition.Head Figure 10 are given in Table 1. The values range from 2 to et al. [1975]have defined the volume V• of "primaryejecta" as 9 x 106km 3. It shouldbe repeatedthat thesevalues are insen- the volume of material excavated from a crater and thrown sitive to the assumptionof premare isostasy,because to first beyondthe craterrim. The quantity• is thusapproximately order the Bouguer anomaly data yield the total anomalous equal to the volume of the early transient cavity minus the mass(mantle uplift plus mare basalts),and it is the sum of the volume of material which falls back in to the collapsingcavity. two componentsthat contributesto Ve.By analogousreason- This equality ignores the effects of density changesdue to ing, thesevalues of Veare also insensitiveto the presenceof compressionof target material, brecciationof fallback materi- high-densityintrusions within the low-density crust underly- al, or other inelastic effects. The volume of primary ejecta ing the basin. The uncertaintiesin gravity and topography residingoutside a major basin, however,is a difficult quantity and in the assumeddensities and model simplificationsresult to determine from photogeologicobservations of ejecta de- in an uncertainty in Ve of about 30% for Orientale and 20% posits,among other reasonsbecause of the difficulty in separ- for the other basinsstudied. The values of Vein Table 1 fall ating primary ejectafrom the locally derived,secondary ejecta within the rangeof estimatesfor • for basinsthe size of surroundingthe basin [Head et al., 1975]. From a variety of Orientale or Serenitatis[Head et al., 1975; Spudiset al., 1984]. observations,models, and scalingrelations, Head et al. [1975] The averagevalue for Vefor the three youngestbasins studied estimatedthat • for a basinthe sizeof Orientalelies between (Orientale,Serenitatis, and Crisium)is 8 x 106 km3 and falls 0.4 x 106 and 12 x 106 km3. This wide range of possible near the upperend of this range.An ejectavolume of 8 x 106 ejecta volumes is equivalent to a large uncertainty in the km3, if spreadevenly over the lunar surface,would constitute thicknessof ejecta depositsexpected from an Orientale-sized a layer 200 m thick. Consideringthe 30 or more impact basins impact. For instance,if the ejecta volume were spread evenly preservedon the moon [e.g., Wood and Head, 1976], such an over the lunar surface, the thickness could be as little as 10 m averagethickness for ejectedmaterial from a singlebasin sug- or as large as 300 m. geststhat the accumulatedthickness of basin ejecta deposits More recently,Spudis [1982, 1983] and Spudiset al. [1984] on the lunar surfacemay, on average,be severalkilometers. have estimated the depth of excavation for lunar basins from An interestingrelationship between apparent mantle uplift remote geochemicalobservations and scaling relations under and crustal thicknessis suggestedby the data from the six the assumption that the lunar crust is layered, with an an- youngestbasins in Table 1. The thicknesstc of the nonmare orthosite layer overlying a more noritic layer [Ryder and crust beneath the central region of each of these six basins is Wood, 1977]. Spudis [1982, 1983] and Spudis et al. [1984] remarkably uniform, 20-30 km, despite differencesin basin concluded that the chemistry and mineralogy of the ejecta size,age, and preimpact crustal thickness.The Moho relief for deposits of five nearside basins provide support for an exca- five basins in crust 55-65 thick is nearly constant at about 30 vation model in which the ratio of the depth of excavationto km, despitea variation of 200 km in basin diameter. While the basin diameter is about 0.1 for basin-sizedimpacts. On the Moho relief beneath Orientale is greater than that of the other basis of such a model the excavation depth for the younger five basins by 30-40 km, the surrounding crust is also thicker nearsidebasins ranged from 40 to 80 km [Spudis,1983], and by 20-30 km. thevolume Vp of primaryejecta excavated during the Orienta- We suggest that this near uniformity of nonmare crustal le basinimpact is estimatedto be 5-8 x 106 km3 [Spudiset thicknessbeneath the younger basinscan be understoodonly 3060 BRATT ET AL.: DEEP STRUCTURE OF LUNAR BASINS

EXCAVATED CAVITY BASIN (a)

(b)

(c)

'""'"'""-.'"":...... :•...... ?•EJECTAAND IMPACT MELT J•J CRUST '•;• MANTLE

BASE OF ELASTIC LITHOSPHERE

Fig. 11. Three scenariosfor the formationof youngimpact basinsformed when the elasticlithosphere is comparable or greaterin thicknessthan the crust.In all cases,uplift of the sublithosphericmantle dominatesthe late-stagecollapse of the transientcavity. (a) The excavatedcavity is confinedto the crust.(b) With a larger ratio of crater diameter to crustal thicknessthe excavatedcavity is preventedfrom extendingsignificantly deeper than the Moho by the relativelygreater strengthof mantle material. (c) With a still larger ratio of crater diameterto crustal thicknessthe transientcavity may extend into the uppermostmantle. If the preservednonmare crust beneaththe basin centeris dominatedby fallback of ejectedcrustal material in the scenariosdepicted in Figures 1lb and 1lc, it may be similar in thicknessfor the two cases. Discoveryof mantle material among the returnedlunar sampleswould strengthenthe possibilityof the scenarioin Figure 11c.

if excavation of the transient cavity for basins400-600 km in ence in strength can be extrapolated to the high strain rates diameter extendedto the base of the crust or to the uppermost thought to accompany cavity formation [e.g., O'Keefe and mantle. If the excavated volumes did not extend at least to Ahrens, 1977], the Moho may have acted more or less as a near the base of the crust for these basins (Figure 11a), then barrier to the deepeningof cavity excavation(Figure 1lb). The the value of tc for a basin of a given diameter should be 20- to 30-km thickness of nonmare crust beneath the central greater for a thicker preimpact crust. The quantity tc should regions of large young impact basins, by this hypothesis, also decreasewith increasing diameter for basins formed in would be a combinationof ejectafallback and crustalmaterial crust of similar thickness.Neither relation is indicated by the transportedlaterally during cavity collapse.These two compo- valuesin Table 1. Of course,the youngestsix basinsin Table 1 nentswould be difficult to distinguish. may have been affected to varying degreesby long-term vis- An alternative explanation for the uniformity of tc is that cous relaxation and other modification processes,but it is the transientcavity for the largest basinsexcavated through hard to imagine how such processeswould have led coin- the crust and into the mantle (Figure 11c). If the equivalent cidentally to the similar presentvalues of tc. As noted above, thicknessof preimpact crustal material that fell back into the the conclusionthat excavationof the youngestnearside basins central basin region was only weakly dependenton basin di- extended to the lower crust has been suggestedon indepen- ameter, then the indicated value of tc would be similar for dent geochemical and photogeological grounds by Spudis basins that had excavated to different mantle depths. Of [1982, 1983] and Spudiset al. [1984]. course,the actual ejecta would have consistedof a mixture of If we accept that the excavated volumes for the basins in crustal and mantle material, but the inversion procedure we Table 1 extended in depth at least to the lowermost crust, we have followedcannot distinguisha thin nonmarecrust from a still must seeka mechanismfor the near constancyof tc. One thicker layer with the samevolume of low-densitycrustal rock explanation is motivated by the observation that mantle ma- admixed with mantle material. The paucity of candidatesin terial is significantlystronger than crustal material at labora- the Apollo sample collection for material from the lunar tory strain rates [e.g., Brace and Kohlstedt, 1980]. If this differ- mantle [see Head et al., 1975], however, suggeststhat exca- BRATT ET AL.' DEEP STRUCTUREOF LUNAR BASINS 3061

! x ...... j/ \•.,......

EXCAVATED CAVITY

SHALLOW BASIN NO FURTHER MODIFIOATION

•(b)

':'"':::.'."•!:.....•EJECTAAND IMPACT MELT [----] CRUST ;?• MANTLE

BASE OF ELASTIC LITHOSPHERE DEEP BASIN VISCOUS RELAXATION

Fig. 12. Two possiblescenarios to accountfor the structureof olderlunar basins (e.g., Figures 10g-10i). (a) Becauseof the high crustaltemperatures and thin elasticlithosphere at the time of impact,flow of both the crust and the mantle accompanytransient cavity collapse, leading to a structuresimilar to that presentlyobserved. (b) The collapseof the transientcavity at leftis accomplishedprimarily by upliftof mantlematerial contemporaneously with the return of impact ejectaand melt to the basin,yielding an initialstructure similar to that of youngerbasins. Because of highnear-surface temperaturesthis initial structure undergoes long-term viscous relaxation of bothtopographic and Moho relief[Solomon et al., 1982].

vation depthsrarely exceededthe Moho depth by any signifi- This effect probably constitutes only a minor contribution, cant amount.We thereforefavor the first explanation. however, to the topographic modification of most lunar basins. The difference in the amount of crustal thinning be- Basin Modification tween the youngestnearside basins (e.g., Orientale, Serenitatis, One of the principal resultsof this paper is that the extent Crisium) and older basins of comparable diameter (Nubium, of apparent mantle uplift or crustal thinning preservedbe- Fecunditatis, Tranquillitatis) is several tens of kilometers. The neath large lunar basinsgenerally decreases with increasing accumulated thickness of younger basin ejecta deposits in basinage. The crosssections shown in Figure 10 illustratethis these older basins is not likely to exceed a few kilometers trend. We have attributed this variation to more extensive [McGetchin et al., 1973]. modificationof older basinsthan youngerones. The time scale Ductileflow. Viscousrelaxation of an impact basin by duc- for this modificationis highly uncertainand may range from tile flow of crustal material can reduce both topographic and the time scale of cavity collapseto the time interval between Moho relief [Solomonet al., 1982]. Becauseof the strong de- basin formation and one or more major episodesof mare pendenceof effective viscosity on temperature, viscousrelax- basalt volcanism. In this section we consider the modification ation was likely to have been a more important modification processesthat can act to reducebasin topographicand Moho process for older basins formed when near-surface temper- relief, and we evaluate the implications of our structural atures on the moon were relatively high and the elastic litho- models for the variation of these processeswith time on the sphere was thin. In particular, the premare topography and moon. the crustal thickness in the Tranquillitatis region are both A significantelement of all endogenicbasin modification consistent with viscous relaxation from an initial structure processesis the additional heatingat depth that accompanies similar to the present structure of Orientale [Solomon et al., basin formation [Bratt et al., 1981]. Both ductile flow and 1982]. It is, of course, uncertain whether the oldest basins magmagenesis can be expectedto be enhancedby this type of actually had initial structuressimilar to that of Orientale or, deep heating. Becausethe outer portions of the moon have perhaps because of a different thermal state at the time of generallycooled with time, the effectsof such temperature- basin formation, never displayed the topographic relief and dependentprocesses might be enhancedfor basinsformed at extent of crustal thinning now observed for young lunar an early stage in lunar history. Older basins are also more basins;the structural modelsderived from presentgravity and likely to sensethe lingeringeffects of the formation of large, topography cannot distinguish between these alternatives preexistingbasins in the sameregion. (Figure 12). We considerfour principal basin modificationprocesses for The formation of the Nubium, Fecunditatis, and Tranquil- youngand old lunar basins:(1) ejectafrom youngerbasins, (2) litatis basins,not resolvably different in age [Wilhelrns, 1981], ductileflow of the crust,(3) nonmarevolcanic activity, and (4) may have occurred over a sufficientlyshort interval of time more volcanism. following the formation of the 3200-km-diameter Procellarum Basin ejecta. The topographic relief of an older basin is basin [Whitaker, 1981] so that the thermal effects of the Pro- reducedby the infill of ejecta depositsfrom youngerbasins. cellarum impact on the creep rates in the crust were still im- 3062 BRATT ET AL..' DEEP STRUCTURE OF LUNAR BASINS

portant [Wilhelms, 1983]. Nubium and Tranquillitatis lie just inferred beneath the Apollo 12 and 14 landing sites from seis- within, and Fecunditatis just outside, the third ring of the mic observations. Procellarum basin [Whitaker, 1981] and are thus within a The basin structural models are characterized by thinned radial range where significantthermal effectscan be expected crust and an elevated Moho beneath the central basin regions, [Bratt et al., 1981]. As noted earlier, the Procellarum basin compared to surrounding areas, presumably the preservedef- may have contributed to the generally lower elevations and fectsof basin excavation and mantle uplift during collapseof thinner crust on the central nearside than elsewhere on the the transient cavity. Orientale exhibits over 60 km of apparent moon. That the Procellarum basin did not itself experience mantle uplift, while the Moho relief beneath Serenitatis,Cris- completeviscous relaxation of surfaceand Moho topography ium, Humorurn, Nectaris, and Smythii is about 30 km. The may be ascribed to the insensitivityof the controlling time three oldest basins included in this study, Nubium, Fecundi- constantto spatial scale[Solomon et al., 1982] and perhapsas tatis, and Tranquillitatis, display only a modest (• 10 km) well to the constraints on lateral flow imposed by spherical amount of preservedMoho relief. geometrywhen the basin diameter approachesthe diameterof The general decreasein the amount of preserved mantle the moon. uplift with increasingbasin age is attributed primarily to en- Nonmare volcanism. An additional modification process hanced rates of lateral flow of crustal material early in lunar that can act to reduce basin relief and that would be enhanced history when crustal temperatureswere relatively high and the by higher temperaturesis nonmare volcanism,i.e., deposition elastic lithosphere comparatively thin. Nonmare volcanic ac- of nonmare basalts with a density similar to the typical den- tivity may also have contributedto the greaterextent of modi- sity of the crust and therefore not resolvable from gravity fication of older basins.The relaxed topographic and Moho anomalies.The volumetric significanceof nonmare volcanism relief for older basins on the central nearside,in particular, for lunar crustal evolution in general and for old lunar basins may be at least partly a consequenceof the extensivecrustal in particular is not well known. To the extent that the crust, as and subcrustalheating associatedwith the formation of the opposedto the upper mantle, is the sourceof magmasfor this large Procellarum basin. volcanic activity, this process involves lateral transport of If we make the assumptionthat the preimpact crustal thick- crustal material to fill the basin depressionand is not readily nessbeneath the basin region was similar to the presentthick- distinguishablefrom ductile flow (Figure 12). If temperatures nessbeneath surrounding regions,then the sum of the volume exceed the solidus in crustal material adjacent to a newly of uplifted mantle, the volume of mare basalt fill, and the formed basin, in fact, the effective viscosity might be suf- volume of the topographic basin provides a lower bound to ficiently low so that bulk flow would dominate magma trans- the volume of crustal material ejected beyond the basin rim port as a mechanismfor reducingrelief. during basin formation. This bound may be similar to the true Mare volcanism. The source regions of mare basalt ejectedvolume for the youngestbasins formed when the elas- magmasare thought to lie at depthsin excessof 100 km [e.g., tic lithosphere may have been significant in thickness and Wyllie et al., 1981]. At such depths the depositionof kinetic when any modification subsequentto cavity collapse may energy as heat, the release of hydrostatic pressure,and the have been relatively minor. On this basis, we estimate the uplift of mantle isothermsduring basin formation [Bratt et al., volume of crustal material ejected from basins the size of 1981] are likely to have had only a small effect on magma Orientale to be of the order of 107 km 3. genesisand then only if the impact was very large [Hubbard The thickness of nonmare crustal material beneath the cen- and Andre, 1983] and the ambient temperature was already tral regionsof the six youngestbasins considered in this study near the solidus.That the thicknessof mare basaltsappears to is remarkably uniform (20-30 km) despitelarge differencesin be less within the older basins (Nubium, Fecunditatis, Tran- basin size and in the preimpact thicknessof the crust. These quillitatis) than nearby youngerbasins (Figure 8) suggeststhat resultssuggest the hypothesisthat basin excavationextended mare volcanismprobably has not contributedsubstantially to in depth at least to the lowermostcrust for theseimpacts and the age dependenceof preservedbasin structure. that significantdeepening of cavity excavationfor the largest Of the basin modification processesconsidered above, later- basins was most likely impeded by an abrupt increase in al flow of lunar crustal material, perhaps augmented by non- strengthat the crust-mantleboundary. The preservedlayer of mare volcanism,is the most likely explanation of the general nonmare crust beneath the basin centers,by this view, may be decreaseof preserved mantle uplift with increasingbasin age. somecombination of fallback and crustalmaterial transported Becauseboth of these processesinvolve the radially inward laterally during cavity collapse. movement of crustal material, the two are difficult to separate The structural models describedin this paper provide im- on the basisof gravity and topographicdata alone. The poor portant constraints on the thermal budget of the basin- degree of preservation of relief of the older basins on the forming processand on the subsequentthermal and tectonic central nearsidemay be a consequenceof both the generally evolution of the basin region [Bratt et al., 1981]. While these high temperaturesin the early lunar crust and the extensive modelsare of necessitysimple, they provide a basisfor testing additional crustal and subcrustalheating associatedwith the the ideas presentedhere with data of higher spatial resolution formation of the ancient Procellarum basin. in individual basin regions on the moon and the other terrestrial planets. SUMMARY AND CONCLUSIONS

We have presentedmodels for the crustal and upper mantle Acknowled.qrnents.We thank Bruce Bills for providing us with a structure beneath major impact basins on the nearside of the tabulated version of his compilaton of lunar topographic data, Bill moon. The models are derived from an inversion of nearside Sjogren for several constructivediscussions on lunar gravity, Paul gravity and topography, subject to the assumptionsthat pre- Spudisfor a copy of his paper prior to publication, and Jan Nattier- Barbaro for help with manuscript preparation. Critical reviewsby J. mare topography was isostatically compensatedby an Airy Dvorak, an additional reviewer, and an Associate Editor resulted in mechanismand that compensationof mare basalt units may significantimprovements to this paper. This researchwas supported be neglected.An additional constraint is the crustal thickness by NASA grants NSG-7081 and NSG-7297. BRATT ET AL,: DEEP STRUCTUREOF LUNAR BASINS 3063

REFERENCES Morrison, F., Algorithms for computing the geopotential using a Arthur, D. W. G., and P. Bates,The Tucson selenodetictriangulation, simpledensity layer, J. Geophys.Res., 81, 4933-4936, 1976. Commun.7, pp. 313-361, Lunar and Planet. Lab., Univ. of Ariz., Muller, P.M., and W. L. Sjogren, Mascons: Lunar mass con- Tucson, 1968. centrations, Science,161, 680-684, 1968. Baldwin, R. B., Lunar mascons: Another interpretation, Science, 162, O'Keefe, J. D., and T. J. Ahrens, Impact-induced energy partitioning, 1407-1408, 1968. melting, and vaporization on terrestrial planets, Proc. Lunar Sci. Bills, B. G., and A. J. Ferrari, A lunar density model consistentwith Conf., 8th, 3357-3374, 1977. topography,gravitational, librational, and seismicdata, d. Geophys. Phillips, R. J., and J. Dvorak, The origin of lunar mascons:Analysis Res., 82, 1306-1314, 1977a. of the Bouguer gravity associated with Grimaldi, Proc. Lunar Bills, B. G., and A. J. Ferrari, A harmonic analysis of lunar topogra- Planet. Sci., 12,,t, 91-104, 1981. phy, Icarus, 31,244-259, 1977b. Phillips, R. J., J. E. Conel, E. A. Abbott, W. L. Sjogren, and J. B. Booker,J. R., R. L. Kovach, and L. Lu, Mascons and lunar gravity, J. Morton, Mascons: Progress toward a unique solution for mass Geophys.Res., 75, 6558-6564, 1970. distribution, d. Geophys.Res., 77, 7106-7114, 1972. Bowin, C., B. Simon, and W. R. Wollenhaupt, Mascons: A two body Roberson, F. I., and W. M. Kaula, Apollo 15 laser altimeter, Apollo solution,d. Geophys.Res., 80, 4947-4955, 1975. 15 Preliminary ScienceReport, NASA Spec. Publ., SP-289, 25-48- Brace, W. F., and D. L. Kohlstedt, Limits on lithospheric stressim- 25-50, 1972. posedby laboratory experiments,d. Geophys.Res., 85, 6248-6252, Ryder, G., and J. A. Wood, Serenitatisand Imbrium impact melts: 1980. Implications for large-scalelayering in the lunar crust, Proc. Lunar Bratt, S. R., S.C. Solomon, and J. W. Head, The evolution of multi- Sci. Conf., 8th, 655-668, 1977. ringed basins: Cooling, subsidence,and thermal stress (abstract), Sjogren, W. L., Lunar gravity at 100 km altitude, Proc. Lunar Sci. Lunar Planet Sci., 12, 109-111, 1981. Conf.,5th., frontispiece,1974. Conel, J. E., and G. B. Holstrom, Lunar mascons: A near-surface Sjogren, W. L., and J. C. Smith, Quantitative mass distribution interpretation, Science,162, 1403-1405, 1968. models for Mare Orientale, Proc. Lunar Sci. Conf., ?th, 2639-2648, DeHon, R. A., Maximum thickness of materials in the western mare 1976. basins(abstract), Lunar Planet. Sci., 9, 229-231, 1978. Solomon, S.C., Mare volcanism and lunar crustal structure, Proc. DeHon, R. A., and J. D. Waskom, Geologic structure of the eastern Lunar Sci. Conf.,6th, 1021-1042, 1975. mare basins,Proc. Lunar Sci. Conf., ?th, 2729-2746, 1976. Solomon, S.C., and J. W. Head, Vertical movement in mare basins: Head, J. W., Orientale multi-ringed basin interior and implications Relation to mare emplacement,basin tectonics,and lunar thermal for the petrogenesisof lunar highland samples,Moon, 11, 327-356, history, J. Geophys.Res., 84, 1667-1682, 1979. 1974. Solomon, S.C., and J. W. Head, Lunar mascon basins: Lava filling, Head, J. W., Lunar volcanismin spaceand time, Rev. Geophys.Space tectonics, and evolution of the lithosphere, Rev. Geophys.Space Phys., 12, 265-300, 1976. Phys., 18, 107-141, 1980. Head, J. W., Origin of outer rings in lunar multi-ringed basins: Evi- Solomon, S.C., R. P. Comer, and J. W. Head, The evolution of dencefrom morphology and ring spacing,in Impact and Explosion impact basins: Viscous relaxation of topographic relief, J. Geophys. Cratering, edited by D. J. Roddy, R. O. Pepin, and R. B. Merrill, Res., 87, 3975-3992, 1982. pp. 563-573, Pergamon,New York, 1977. Spudis, P. D., Orientale basin ejecta: Depths of derivation and impli- Head, J. W., Lava flooding of ancient planetary crusts: Geometry, cations for the basin-forming process(abstract), Lunar Planet Sci., thickness,and volumes of flooded lunar impact basins,Moon Plan- 13, 760-761, 1982. ets, 26, 61-88, 1982. Spudis,P. D., The excavationof lunar multi-ringed basins:Additional Head, J. W., M. Settle, and R. S. Stein, Volume of material ejected results for four nearside basins (abstract), Lunar Planet Sci., 14, from major lunar basins and implications for the depth of exca- 735-736, 1983. vation of lunar samples,Proc. Lunar Sci. Conf., 6th, 2805-2829, Spudis,P. D., B. R. Hawke, and P. Lucey, Composition of Orientale 1975. basin depositsand implicationsfor the lunar basin-formingprocess, Head, J. W., E. Robinson, and R. J. Phillips, Topography of the Proc. Lunar Planet. Sci. Conf. 15th, part 1, J. Geophys.Res., 89, Orientale basin (abstract), Lunar Planet. Sci., 12, 421-423, 1981. suppl., C 197-C210, 1984. Hopmann, J., General-Katalog absoluterHohen auf dem Mond, Ann. Thurber, C. H., and S.C. Solomon, An assessment of crustal thick- Univ. Sternw. Wien, 26, 7 Heft, 1967. ness variations on the lunar near side: Models, uncertainties, and Hubbard, N., and C. G. Andre, Magma genesisin a battered moon: implications for crustal differentiation, Proc. Lunar Planet. Sci. Effectsof basin-formingimpacts, Moon Planets,29, 15-37, 1983. Conf.,9th, 3481-3497, 1978. Hulme, G., Mascons and isostasy,Nature, 238, 448-450, 1972. Toksoz, M. N., A.M. Dainty, S.C. Solomon, and K. R. Anderson, Janle, P., A crustal gravity model of the Mare Serenitatis-Mare Cris- Structure of the moon, Rev. Geophys.Space Phys., 12, 539-567, ium area of the moon, d. Geophys.,49, 57-65, 1981a. 1974. Janle, P., A lunar crustal gravity model across the Triesnecker- Watts, C. B., The marginal zone of the moon, Astron.Pap. Am. Ephem- Hyginus area, Mare Tranquillitatis, and Mare Fecunditatis, Moon eris Nautical Aim., 17, Nautical Alm. Office,Nay. Observ.,Wash- Planets, 24, 441-451, 1981b. ington, D.C., 1963. Kaula, W. M., G. Schubert,R. E. Lingenfelter,W. L. Sjogren, and W. Whitaker, E. A., The lunar Procellarum basin, Proc. Lunar Planet. R. Wollenhaupt, Analysis and interpretation of lunar laser alti- Sci., 12A, 105-111, 1981. metry, Proc. Lunar Sci. Conf.,3rd, 2189-2204, 1972. Wilhelms, D. E., Relative ages of lunar basins, II, Serenitatis (ab- Kaula, W. M., G. Schubert,R. E. Lingenfelter,W. L. Sjogren,and W. stract), Reports of Planetary Geology Program--1981, NASA R. Wollenhaupt, Lunar topography from Apollo 15 and 16 laser Tech. Memo., TM-84211, 405-407, 1981. altimetry,Proc. Lunar Sci. Conf.,4th, 2811-2819, 1973. Wilhelms, D. E., Effects of the Procellarum basin on lunar geology, Kaula, W. M., G. Schubert,R. E. Lingenfelter,W. L. Sjogren,and W. petrology, and tectonism(abstract), Lunar Planet. Sci., 14, 845-846, R. Wollenhaupt, Apollo laser altimetry and inferencesas to lunar 1983. structure,Proc. Lunar Sci. Conf.,5th, 3049-3058, 1974. Wise, D. U., and M. Y. Yates, Mascons as structural relief on a lunar Lawson, C. L., and R. J. Hanson, SolvingLeast SquaresProblems, 340 "Moho", J. Geophys.Res., 75, 261-268, 1970. pp., Prentice-Hall, Englewood Cliffs, N.J., 1974. Wollenhaupt, W. R., and W. L. Sjogren, Apollo 16 laser altimeter, Liu, A. S., and P. A. Laing, Lunar gravity analysis from long-term Apollo 16 Preliminary ScienceReport, NASA Spec.Publ., SP-315, effects,Science, 173, 1017-1020, 1971. 30-1-30-5, 1972. McGetchin, T. R., M. Settle, and J. W. Head, Radial thickness vari- Wollenhaupt, W. R., W. L. Sjogren, R. E. Lingenfelter, G. Schubert, ation in impact crater ejecta: Implicationsfor lunar basin deposits, and W. M. Kaula, Apollo 17 laser altimeter, Apollo 17 Preliminary Earth Planet. Sci. Lett., 20, 226-236, 1973. ScienceReport, NASA Spec. Publ., SP-330, 33-41-33-42, 1974. Melosh, H. J., and W. B. McKinnon, The mechanics of basin forma- Wong, L., G. Buechler, W. Downs, W. Sjogren, P. Muller, and P. tion, Geophys.Res. Lett., 5, 985-988, 1978. Gottlieb, A surface-layerrepresentation of the lunar gravitational Mills, G. A., and P. V. Sudbury,Absolute coordinates of lunar fea- field, J. Geophys.Res., 76, 6220-6236, 1971. tures, Icarus, 9, 538-561, 1968. Wong, L., W. Downs, W. Sjogren, and P. Muller, A lunar mass Moore, H. J., C. A. Hodges, and D. H. Scott, Multi-ringed basins-- distribution from Apollo 14, 15, and 16 CSM-LM and subsatellite Illustrated by Orientale and associatedfeatures, Proc. Lunar Sci. radio Doppler data, report, Jet Propul. Lab., Pasadena,Califi, 1975. Conf.,5th, 71-100, 1974. Wood, C. A., and J. W. Head, Comparison of impact basins on Mer- 3064 BRATT ET AL.: DEEP STRUCTUREOF LUNAR BASINS

cury, Mars, and the moon, Proc. Lunar Sci. Conf., 7th, 3629-3651, 1976. S. R. Bratt and S.C. Solomon,Department of Earth, Atmospheric, and Planetary Sciences,Massachusetts Institute of Technology, Wood, J. A., Fragmentsof terra rock in the Apollo 12 soil samples Room 54-520, Cambridge, MA 02139. and a structuralmodel of the moon, Icarus, 16, 462-501, 1972. J. W. Head, Departmentof GeologicalSciences, Brown University, Wood, J. A., Bombardmentas a causeof lunar asymmetry,Moon, 8, Providence RI 02912. 73-103, 1973. C. H. Thurber, Department of Earth and SpaceSciences, State Wyllie, P. J., C. H. Donaldson, A. J. Irving, S. E. Kesson,R. B. Universityof New York at StonyBrook, StonyBrook, NY 11794. Merrill, D.C. Presnall, E. M. Stolper, T. M. Usselman,and D. Walker, Experimentalpetrology of basaltsand their sourcerocks, (ReceivedJune 11, 1984; in BasalticVolcanism on the TerrestrialPlanets, pp. 493-630, Per- revisedSeptember 25, 1984; gamon, New York, 1981. acceptedNovember 12, 1984.)