martian plate tectonics - quarg: hometjb/het618/nathan/94je00216.pdf · sleep: martian plate...

JOURNAL OF GEOPHYSICAL RESEARCH, VOL. 99, NO. E3, PAGES 5639-5655, MARCH 25, 1994

Martian plate tectonics

Norman H. Sleep Department of Geophysics, Stanford University, Stanford, California

Abstract. The northern lowlands of Mars may have been produced by plate tectonics. Preexisting old thick highland crust was subducted, while seafloor spreading produced thin lowland crust during Late Noachian and Early Hesperian time. In the preferred reconstruction, a breakup margin extended north of Cimmeria Terra between Daedalia Planum and Isidis Planitia where the highland-lowland transition is relatively simple. South dipping subduction occurred beneath Arabia Terra and east dipping subduction beneath Thatsis Montes and Tempe Terra. Lineations associated with Gordii Dotsum are attributed to ridge-parallel structures, while Phelegra Montes and Scandia Colles are interpreted as transform-parallel structures or ridge-fault-fault triple junction tracks. Other than for these few features, there is little topographic roughness in the lowlands. Seafloor spreading, if it occurred, must have been relatively rapid. Quantitative estimates of spreading rate are obtained by considering the physics of seafloor spreading in the lower C 0.4 g) gravity of Mars, the absence of vertical scarps from age differences across fracture zones, and the smooth axial topography. To the first order, the height of vertical scarps across fracture zones does not involve gravity. Crustal thickness at a given potential temperature in the mantle source region scales inversely with gravity. Thus, the velocity of the rough-smooth transition for axial topography also scales inversely with gravity. Plate reorganizations where young crust becomes difficult to subduct are another constraint on spreading age. Possible plate reorganizations, for example, the end of spreading through Alba Patera, occur when the ridge axis is far from the trench. That is, rapid plate motions are inferred to have placed young oceanic crust far from the ridge axis. The preferred full spreading rate 900 from the plate pole is 80 mm yr -1. Plate tectonics, if it occurred, dominated the thermal and stress history of the planet. A geochemical implication is that the lower gravity of Mars allows deeper hydrothermal circulation through cracks and hence more hydration of oceanic crust so that more water is easily subducted than on the Earth. Age and structural relationships from photogeology as well as median wavelength gravity anomalies across the now dead breakup and subduction margins are the data most likely to test and modify hypotheses about Mars plate tectonics.

Introduction

One third of the surface of Mars is topographically low young plains mainly in the northern hemisphere and the other two thirds is old heavily cratered highlands mainly in the southern hemisphere. Although this crustal dichotomy has been known from the time of the earliest Mars probes, there has been little agreement on its origin and even its age of formation [McGill and Dimitrou, 1990; Zimbelman et al., 1991; McGill and Squyres, 1991; Strom et al., 1992]. Both internal tectonic and external impact hypotheses involving a one-plate planet have been proposed for the origin of the dichotomy. For example, impact hypotheses involve either a single large impact basin [Wilhelms and Squyres, 1984; Strom et al., 1992] or several impacts [Frey and Schultz, 1988], while McGill and Dimitrou [1990] favor thinning of intact crust by internal processes.

I proposed informally at a meeting a few years ago that the northern lowlands are the result of oceanic crust produced by seafloor spreading. This possibility has been briefly discussed as a personal communication from me since then by Tanaka [1990a] and by Zirnbelman et al. [1991]. An advantage of this

Copyright 1994 by the American Geophysical Union.

Paper number 94JE00216. 0148-0227/94/94 JE-00216505.00

hypothesis is that plate motions are known to occur on the Earth. That is, knowledge of the Earth leads to quantitative predictions about the physics and kinematics of Martian plates. In contrast, I cannot envision an internal way of thinning crust over a 8000-kin-wide region that does not involve significant horizontal crustal motion and thus cannot begin to study this hypothesis.

The purpose of this paper is to investigate plate tectonics as a mechanism for producing the northern lowlands of Mars. I con- centrate on the physics of the process so that I can extrapolate our knowledge of the Earth to Mars. Reinterpretation of photogeology is beyond the scope of the paper. Thus, I cite recent reviews of photogeological results where possible.

Geography and Constraints

A brief summary of the geography and geology of the crustal dichotomy is given here to put the subsequent discussion of plate tectonics in context. The reader is reminded that Martian geological time is divided into three periods, Noachian, Hesperian, and Amazonian, from old to young. The Noachian and Amazonian periods are further divided into early, middle, and late epochs, while the Hesperian period is divided into early and late epochs. There is, however, some difficulty in reconcil- ing age assignments and crater counts from different groups of

5639

5640 SLEEP: MARTIAN PLATE TECTONICS

Table 1. Absolute Age Estimates

Absolute age range, Ga

Epoch HT NW Late Amazonian 0.25-0.00 0.70-0.00 Middle Amazonian 0.70-0.25 2.50-0.70

Early Amazonian 1.80-0.70 3.55-2.50 Late Hesperian 3.10-1.80 3.70-3.55 Early Hesperian 3.50-3.10 3.80-3.70 Late Noachian 3.85-3.50 4.30-3.80 Middle Noachian 3.92-3.85 4.50-4.30

Early Noachian 4.60-3.92 4.60-4.50

Condensed from Table II of Tanaka et al. [1992]. HT denotes Hartman-Tanaka scale and NW denotes Neukum-Wise scale.

investigators [e.g., Frey eta/., 1988]. Absolute ages are obtained with much uncertainty from models of impactor popu- lations; two estimates are given in Table 1.

Physiology. The northern lowlands are conveniently viewed in polar projection (Figure 1). The position of the highland- lowland transition as mapped by Scott and Tanaka [1986] and by Greeley and Guest [1987] is shown with various geographic features for reference. For convenience and with some

forethought to discussion of plate tectonics, parts of the transition zone are informally referred to here as "margins" of the terrae south of them because, perhaps surprisingly, much of the transition-zone escarpment has not been blessed with geographic names [Batson et al., 1979].

Working west, the Memnonia margin appears the freshest. Lineations parallel to the northeast facing segments of this mar-

180"

30"

O

_20 ø 0 o

Figure 1. Index map of Mars plotted in polar stereographic projection, which locally preserves angles [Wessel and Smith, 1992]. The map extends from the north pole to 20øS latitude. Geological features are compiled from the maps of Scott and Tanaka [1986], Tanaka and Scott [1987], and Greeley and Guest [1987]. The dichotomy boundary where mapped is shown as a thick gray line. Volcanic edifices include Uranius Patera, Alba Patera, Tharsis Montes, Elysium Mons within Elysium Planitia, and Olympus Mons. Daedalia Planurn is a volcanic region near the lowland-highland transition. Chryse Planitia and Isidis Planitia are old impact basins. Linear features include Tempe Fosse within Tempe Terra, Valles Marineris, Gordii Dorsum, Phelegra Montes, Scandia Colles, and fault zones cutting Alba Patera. The Acidalia Planitia and Utopia Planitia regions of the lowlands and the Lunae Planurn, Arabia Terra, Tyrrhena Terra, Cimmeria Terra, and Memnonia Terra regions of the highlands are shown for reference.

SLEEP: MARTIAN PLATE TECTONICS 5641

gin, including Gordii Dorsum, are evident in the nearby lowlands. The Cimmeria and Tyrrhena margins are marked by a well-developed escarpment. The Isidis Planitia marks a change in strike of the transition zone between the Tyrrhena and the Arabia margins. It is likely to be an impact structure formed in Early to Middle Noachian time. The position of the Arabia margin is less evident from topography and the margin is somewhat irregular in detail. The Arabia margin is further divided by a kink around 335øW. Chryse Planitia and Acidalia Planitia form an embayment of the lowlands between Tempe Terra and western Arabia Terra. Chryse is probably an old impact structure like Isidis [Tanaka et al., 1992], but the total embayment is obviously not circular. The boundary is obscured by later volcanic rocks between Tempe Terra and Daedalia Planurn. It is most likely between Alba Patera and Olympus Mons on one side and Tharsis Montes on the other [e.g., Tanaka, 1990b].

In additional to the lineations near Gordii Dorsum, Phelegra Montes and Scandia Colles are prominent on the planetary geological maps [Greeley and Guest, 1987; Tanaka and Scott, 1987]. Otherwise, the northern lowlands are basically plains except where affected by younger features.

Extensive ridged plains, which occur in the southern highlands, are by stratigraphic assignment Lower Hesperian. These surfaces, including Lunae Planurn east of Tharsis Montes, are widely enough distributed that they define a latest limit for the time that the southern highlands were essentially an intact plate. Early limits can be obtained at least locally from intact rings of large impact basins.

Crustal thickness variations. There are no useful con-

straints on the absolute crustal thickness of Mars. Various esti-

mates have been given, but these are dependent on tenuous assumptions about the mode of compensation [see Schubert et al., 1992]. The relative crustal thickness variation across the

transition zone is much better constrained as this boundary is isostatically compensated [e.g., Smith et al., 1993]. The estimate by McGill and Dimitrou [1990] of 21 km (based on an average elevation difference of 3 km and lower crustal and mantle densities of 3000 kgm -3 and 3500 kgm -3, respectively) is used here.

Relative age of lowlands and transition. It is necessary to distinguish between the age of the crustal surface in the lowlands and the age of topographic dichotomy and to consider that portions of these features may have formed at various times. Conceivably both can be quite old; for example, a giant impact late in the primary accretion of Mars may have produced a large basin [Strom et al., 1992]. In this case, the dichotomy would predate the old cratered surfaces in the southern highlands. In the single impact hypothesis of Wilhelms and Squyres [1984] and the multiple impact hypothesis of Frey and Schultz [1988], both the northern lowland crust and the dichotomy are coeval with the southern highlands. Alternatively, the northern lowland crust may be old, while its subsidence by tectonic processes and therefore its margins are younger than the southern highlands [McGill and Dimitrou, 1990]. A plate tectonic origin of the northern lowlands by seafloor spreading of thin crust and subduction of thicker crust would imply that both the lowland crust and its margins are younger than the southern highlands.

McGill and Dimitrou [1990] prefer a Late Noachian to Early Hesperian age for formation of the transition zone and hence in their model for the subsidence of older crust in the northern

lowlands. The younger limit for the existence of a topographic escarpment is well constrained by Hesperian lava flows and fluvial channels that cascaded over it. (The widespread plains deposits so formed are by stratigraphic assignment, Upper

Hesperian.) The existence of old crust in the northern lowlands is less obvious. The main evidence cited by McGill [1989] and McGill and Dimitrou [1990] is knobby terrain that is interpreted as the relics of large old craters and basins in Utopia Planitia and Elysium Planitia that are interpreted to be the result of large impacts. Craddock et al. [1990] interpret Gordii Dorsum and associated lineations to be the result of a large impact basin in Daedalia Planurn.

Origin of the northern lowlands as relatively young oceanic crust implies the highland crust which occupied present lowland areas near Isidis and Chryse is elsewhere and possibly subducted. That is, large basins themselves or ring structures from the Isidis, Chryse, or less exposed highland impacts within the northem lowlands would imply an Early to Middle Noachian age for its crust comparable to the age of the southern highlands and thus preclude relatively recent plate tectonics. The poor exposure in the northern lowlands has left room for both young and old interpretations. Tanaka et al. [1992], for example, consider the Utopia basin to be controversial. The only lowland features which Wichman and Schultz [1989] relate to the Isidis impact are a few younger volcanoes. Frey et al. [1988] recog- nize no very old surface in the northern lowlands. A probably Upper Noachian surface is recognized by this group in the southern highlands with some relicts in the transition zone. In basic agreement with the McGill papers, their most widespread surface in the transition zone is uppermost Noachian to lowermost Hesperian, that is somewhat older than or comparable to the Lunae Planurn surface.

Trial Plate Tectonic History of Mars

A trial plate tectonic history of the northern lowlands is given here to illustrate kinematic features of plate motions and their relationship to plate dynamics. It is based partly on the geometry of the dichotomy and partly on the structures within the northern lowlands. The risk of the reconstructions being illusory is outweighted by having specific features to discuss with regard to methods that are generally applicable to the dynamics of features that are likely to occur in any plate tectonic reconstruction of Mars. Terrestrial terms, like seafloor

spreading, are used although an ocean in the terrestrial sense may not have existed on Mars at the times reconstructed. It is presumed unless otherwise stated that plate tectonics did occur to avoid the verbiage of having to qualify each sentence.

A minimum requirement for a plate tectonic history of the northern lowlands is that older, probably thicker crust, is replaced by thinner younger crust. A simple two-plate geometry for doing this is shown on a plane in Figure 2. Seafloor spreading begins on one side of a (active) plate and subduction at the other. The rest of the plane is another (passive) plate. At the time that all the original crust of the active plate has been subducted, the ridge has migrated half way to the trench. The ridge and trench approach and eventually collide if plate movements continue. The collision restores the plane to being a single plate. Twice the original area of the active plate is subducted during its existence. Certain basic age relationships result: the breakup margin formed at the start of seafloor spreading is older than the new crust formed by seafloor spreading; and the subduction zone in contrast remains active throughout the process.

Breakup. A more complex version of the above two-plate history is required by the real spherical geography of Mars. Plate tectonics in the northern lowlands begins with the formation of Boreal and Austral plates at the expense of a one-plate


planet or an unknown previous plate arrangement (Plate 1). Breakup is considered to occur somewhat before the formation of compressional ridges on the plains in Lunae Planum, but perhaps contemporary with the plains lavas which may indicate an episode of widespread lithospheric tension.

Break-up

Nearly resurfaced

%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%

%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%

•%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%

•%%%%%%%%%%%%%%•%%%%%%%%%%%%%%%%

Ridge-trench collision

•%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%!%%%'•.i I•

Figure 2. History of a two-plate planet with an active rectangu- lar plate shown schematically in map view. Plate boundaries are a ridge on the left, a trench on the right, and transform faults above and below. The area of the active plate decreases from the time of breakup (above) to ridge-trench collision (just after bottom panel). The original lithosphere in the active plate is subducted when the ridge has moved halfway to the trench Oust after middle panel).

The Tyrrhena, Cimmeria, and Memnonia margins are considered here to be the passive margin because their topographic escarpment is much simpler than that of the Arabia margin. The stair-step map pattern of the Memnonia margin suggests ridge and transform fault segments. The Tyrrhena and Cimmeria margins are nearly a great circle. A 180ø-diameter small circle defining a transform fault through this great circle would imply a pole of rotation northeast of Alba Patera and the Boreal plate rotating without much subduction to provide a driving force. The altemative with the northwest striking escarpments along Memnonia margin defining the ridge segments is thus used here. The pole of rotation is selected by eye to keep the Tyrrhena and Cimmeria margins nearly a great circle and to fit the stair-step pattern in Memnonia.

The belt of volcanism and tectonism extending through Tharsis Montes from Tempe to Daedalia is an obvious candidate for an island are [Tanaka, 1990a]. However, only the northern part of this zone has active subduerion at the time of Plate 1. Geometrically, this is a consequence of the selected transform fault direction on the Memnonia margin. An additional subdue- tion zone is needed somewhere to conserve area. It is placed along the Arabia margin in Plate 1. Ridged plain units of volcanic origin which are mapped on the highlands to the south of the Arabia margin by Scott and Tanaka [1986] and by Greeley and Guest [1987] are interpreted as a product of an are. A more complicated alternative would be to continue the Arabia margin in the direction of Lunae Planum and to have the region between the Tharsis Montes margin and Lunae Planurn acerere to the highlands by a suture through Lunae Plantun at some stage of the process.

Seafloor spreading. The plate geometry at a later time when the ridge axis has rotated 50 ø of are about the plate pole is shown in Plate 2. Seafloor between the breakup margin and the ridge axis has been generated by this time and the Boreal plate is considerably reduced from its initial area. Phelegra Montes is beginning to form as a fracture zone associated with a ridge- ridge transform fault. The ridge segment west of Tharsis Montes is approaching the trench.

In analogy with the Earth, relatively fast seafloor spreading is required to produce the new lowland crust in this diagram because rough topography is produced by slowly spreading ridge axes. That is, the smooth topography of the lowlands is much of the evidence about plate processes during this period. One real topographic feature and the lack of topographic expres- sion for another feature in the reconstruction provide some constraints: northwest striking linearions between Gordii Dotsum and the Memnonia margin are considered to be abyssal hills formed at the ridge axis rather than fracture zones associated with transform faults as discussed by Zimbelman etal. [1991] because the several linearions more closely resemble terrestrial abyssal hills than widely spaced transform faults. There are no analogous ridge lineations near Isidis where relatively slow spreading is also expected from Plate 2. The two transform faults farthest from the plate pole represent features associated with real steps in the Memnonia margin. There are no obvious linearions associated with them even near the Memnonia mar-

gin. A large age offset is thus unlikely. Plate reorganization. All the subduction zones at the time

of breakup (Plate 1) subdueted older highland crust. By the time of Plate 2, the northern Tharsis Montes trench was sub-

ducting mainly younger lowland lithosphere while the Arabia trench was subdueting older highland crust. This difference is considered to cause a reorganization of plate motions analogous


180

Plate 1. Plate geometry at the time of breakup overlayed on the index map of Figure 1. Trenches are shown by a tooth pattern, ridges by double lines, and transform faults by single lines. The Boreal-Austral plate pole (23øN, 287øW is indicated by a large circle, which for convenience, is on the plate boundary. The Boreal plate is shown in yellow. Starting outward from the plate pole along the ridge boundary, the first transform fault is drawn to align with Phelegra Montes in the next reconstruction; the second is not associated with a geological feature; and the third and fourth are fit to the stair-step pattern in the transition zone. Lowland features resting on crust that did not exist at this time are plotted in grayer tones.

to when subduction on the Earth slows or ceases when young hot lithosphere approaches a trench. That is, cool highland lithosphere in the Arabia slab produced a greater driving force than the hotter Tharsis Montes slab.

For simplicity in drawing, plate reorganization is assumed to occur instantaneously. Plate boundaries during the reorganization are shown in Figure 3 and boundaries after the reorganization in Plate 3. The Boreal plate breaks into an Acidalia plate subducting beneath Arabia and a Ulysses plate subducting beneath Tempe Terra and Tharsis Montes. The reorganization is accomplished geometrically as follows: The transform fault between the Tempe and Arabia trenches extends to the Acidalia-Austral ridge axis. A new series of ridge and transform segments forms south of this fault so that a new breakup margin cuts lowland crust near Gordii Dorsum and highland crust near Daedalia Planum. This arrangement allows

the Tharsis Montes subduction zone to extend south to Daedalia

Planurn. The highland crust subducted beneath southern Tharsis Montes provides a driving force for subduction of the Ulysses plate.

The reorganization is interpreted to explain Gordii Dorsum as a breakup ridge axis and Phelegra Montes and Scandia Colles as transform fault tracks. Scandia Colles is associated with the

ridge-fault-fault triple junction of the Ulysses, Acidalia, and Austral plates. A somewhat irregular trace of the triple junction is expected as the two transform fault directions are not quite parallel. Rough topography in Phelegra is interpreted as being associated with a triple junction during an earlier stage of the reorganization that is not discussed further. There are no obvious features associated with the plotted extinct Austral-Boreal ridge axis which ends up on the Austral plate between the Scan- dia transform and the Ulysses-Austral plate boundary.


O-

',70 -20 0

18O 1

o***

/ /

Plate 2. Plate positions just before plate reorganization plotted as in Plate 1. Phelegra Montes is an active transform trace. The position of the breakup boundary on the Boreal plate is indicated by the blue line with ticks. Part of this boundary has already been subducted near Uranius Patera; its extrapolated position is shown by a thinner line.

Demise of plate tectonics. After the reorganization the Ulysses and Acidalia plates behave independently. Initially the spreading rate of the Austral-Acidalia ridge is faster than that of the Ulysses-Austral ridge. The Acidalia ridge axis thus moves to the left of the Ulysses axis. A final reconstruction (Plate 4) applies to the time where both ridges have approached their trenches and rapid plate motions have ceased (not necessary at the same time for each plate). Alba Patera and Olympus Mons are interpreted as being associated with abandoned ridge axes that remained volcanically active after rapid spreading ended. (The two-plate arrangement causes the fractures on Alba Patera to be parallel to those nearby Tempe Terra; the ridge axis may have been much further away than now when some of the Tempe fractures formed.) The ridge segment south of Olympus Mons is drawn simply to connect with the southern end of the subduction zone as this region is hidden beneath younger volcanic rocks. No obvious lineations are associated with extinct

Acidalia-Boreal spreading centers; several topographically high exposures mapped by Scott and Tanaka [ 1986], by Tanaka and

Scott [1987], and by Greeley and Guest [1987] as Noachian or Lower Hesperian (Ns in Plate 4) are in the position of the plotted extinct ridge axes in Acidalia Planitia.

Physical Aspects of Ridge Axes on Mars

The trial plate tectonic history presented above has several features involving physical processes: (1) lack of prominent abyssal-hill lineations, (2) lack of prominent fracture zones, and (3) reorganization of plate motions when young lithosphere is subducted. In addition, a plate tectonic history should explain (4) the 3-loth topographic difference between lowlands and highlands; the (5) later volcanism in Alba Patera, Tharsis Montes, and Olympus Mons; and (6) the stress history of the intact lithosphere as inferred from fracture patterns.

The first four items listed above, and to some extent the last

two items, involve rescaling what is known about the Earth to the lower gravity environment of Mars. Such rescaling yields quantitative estimates, for example, for the minimum rate of


180 ø

•0 o

30 ø

o

•0 ø _20 ø

0 o

Figure 3. New and old plate positions at the time of the reorganization plotted as in Plate 2. The Acidalia- Austral plate pole (circle) has the same position as the Boreal-Austral plate pole. The Ulysses-Austral pole (tri- angle) is located nearer to the north pole (60øN, 282øW).

seafloor spreading that produces smooth crust and the young age limit below which lithosphere becomes difficult to subduct. The small size of Mars relative to the Earth is mainly involved in considering the thermal history of its interior.

Fracture zone scarps. The lack of lineations in the lowlands associated with transform offsets of the produced breakup margin in Memnonia allows the spreading rate to be constrained. The process is well understood on the Earth and does not to the first order depend on the magnitude of gravity. Very young oceanic lithosphere is joined with older oceanic lithosphere on the outside corners of transform faults. The younger lithosphere is initially hotter and more elevated than the older lithosphere. The height of this elevation change and its evolution with time are obtained froln the square-root-of-age law for the depth of oceanic crust

D = Do + y•'c (1)

where Do is the depth of the ridge axis, T is a constant which implies 350 m of subsidence of i m.y. crust, and tc is crustal age. The constant T for the Earth includes a factor of 3.3/2.3

for the effect of water in the ocean on isostasy [Turcotte and Schubert, 1982, equation 4-202]; the value of the constant for a ridge on land without this factor implies 250 m of subsidence in the first million years.

The predicted elevation difference across a fracture zone with age difference At decreases with age:

zxt = + (2)

where tc is the age of the younger side of the fracture zone. This elevation difference applies to flat seafloor some distance from the fracture zone. Near the fracture zone the behavior is

more complicated. The initial escarpment having the initial depth offset D is frozen in the upper crust so that the total height of the escarpment does not change. The lithosphere around the fracture zone deforms by flexure with this constraint. Old fracture zones consist of a downwarp of D/2 on the old side and an upwarp of D/2 on the young side [Wessel and Haxby, i990]. Transform valleys and ridges created at the active transform offset may also exist. The mechanics of creat-


180

,%

Plate 3. The new plate geometry just after reorganization plotted as in Figure 3. The Ulysses plate is filled with green and the Acidalia plate is filled with yellow. Starting from the north, the first transform fault on the Ulysses-Austral boundary connects the ridge to the trench. This feature and the nearby Acidalia-Austral transform fault represent a more complicated Ulysses-Acidalia boundary along which a ridge-fault-fault triple junction moved. The second transform fault and the first two ridge segments align with Alba Patera in the next reconstruction. The northern part of the third ridge segment aligns with Olympus Mons in the next reconstruction. A transform fault extending from the ridge to the trench is drawn to indicate that the breakup may not have been synchronous.

ing these features, which are similar to those for creating abyssal hill lineations discussed below, do scale with planetary gravity.

The longest transform fault along the Memnonia margin in the breakup reconstruction (Plate 1) is 12 ø of arc long about the plate pole and located at-26 ø latitude with respect to the pole. It is 640 km long. The upwarp on its young side would be observable if large enough, while the downwarp would probably be filled with supracrustal rocks. The maximum height that might escape detection by being covered or obscured by later geological features is not obvious; 500 m is used here, which implies a 1000-m initial scarp and a maximum 16-m.y. age offset. The minimum (full) spreading rate is 80 mm yr -• for a ridge on land. A minimum (full) spreading rate of 90 mmyr -I is implied 90 ø from the plate pole. These rates are independent

of magnitude of gravity but would be larger if the ridge was covered with water when its was active through the value of qt.

Melting beneath ridge axes. Upwelling mantle material beneath ridge axes is essentially adiabatic if the spreading rate is sufficiently fast. The mantle at great depths is solid as melting temperature increases with pressure and hence depth. Melting begins when the solid adiabat intersects the melting curve. The depth at which this occurs is greater for hot ascending material than for cooler material. The total amount of melt produced determines the thickness of oceanic crust.

Models of this process for realistic compositions have been published by McKenzie [1984], McKenzie and Bickle [1988], and Forsyth [1993] and the petrological implications have been extensively discussed by Klein and Langmuir [1987] and McKenzie and O'nions [1991]. A eutectic material, however, is


180 1

ß / I

-- 30

/

-20 0

Plate 4. Plate geometry at the thne that plate motions cease plotted as in Plate 3. Various exposures in Aci- dalia Planitia mapped as Noachian or Lower Hesperian by Scott and Tanaka [1986], Tanaka and Scott [1987], and G•eeley and Guest [1987] are indicated by Ns. This grouping has been somewhat obscured, as it falls on the boundary of the three Mars maps.

simpler to explain as the purpose here is to rescale their work for the low gravity of Mars and suffices because the actual melting relations of the Martian mantle are unknown. Consider a piece of material that is brought without melting from the depth z,,, where melting begins to a shallower depth z. The solid adiabat of such material is

where Ts is the adiabatic temperature of material that as a solid has the potential temperature Tp at the surface and (3T/3P)s is the adiabatic gradient with respect to pressure P. Assuming that the melting temperature increases linearly with depth implies that the melting temperature is given by

r m '- r 0 q- [-•j P (4) where To is the melting temperature at the surface and

(dTm/dP) is the melting point gradient with respect to pressure. The pressure is related to the lithostatic overburden

P = pgz (5)

where p is density, g is the acceleration of gravity. As the potential temperature and the adiabatic temperature are equal at the depth Zm where melting begins, this depth is related to potential temperature by

Pg -•-•- - s

= LB pg J (6) where ATp is (excess) potential temperature referenced to the hottest adiabat that produces no melting. The second formula- tion is convenient for comparing the Earth and Mars as the


effects of their different compositions are accounted for to the first order by referencing temperatures to the melting point at the surface. The effects of composition on viscosity are also partially accounted for as formulations for viscosity are commonly referenced to the melting temperature [Turcotte and Schubert, 1982, p. 307].

The amount of melting at depth z is obtained by artificially considering that superheated solid material rises adiabatically and then melts at depth z. The specific heat from the difference between the solid adiabatic gradient extrapolated into the melting zone and the melting temperature is balanced by the latent heat of melting:

where Cv is heat capacity, L is latent heat per mass, and Fm is the fraction of melting. Note that the first term on the left is the pressure difference between depths Zm and Z. The total amount of melting in a column extending from Zm to the surface is

pgzm2BCv M- I Fm dZ = 2L (8)

or in terms of excess potential temperature

2u = 2pgLB

(9)

That is, the thickness of oceanic crust generated at a given excess potential temperature and the change of crustal thickness with potential temperature scale inversely with gravity, as noted by Warren [1993].

Martian gravity is 0.3795 that of the Earth [Kiqffkr et al., 1992]. As the Martian mantle is likely to be more iron-rich and denser than the Earth's mantle [Longhi et al., 1992], the ratio of pg is around 0.4. This rounded number is convenient for quick calculations. For example, presently forming normal terrestrial oceanic crust is about 6 km thick. (White et al. [1992] give a 7 km worldwide average which I reduce somewhat to correct for the failure to fully remove crust affected by hotspots from the compilation and for cooling of the Earth's interior with time which causes older oceanic crust to be slightly thicker.) This would imply under a similar excess potential temperature that Martian crust is 15 km thick.

The temperature change needed for the lowland crust to be 21 km thinner than the highland crust is easily estimated from this reasoning. More sophisticated calculations using a noneu- tectic material indicate that 8 km of additional melting is produced by an increase of 100øC above normal potential temperature and 12 km of melt is produced by an additional 100øC increase [McKenzie and Bickle, 1988]. Cooling of about 100øC from the current normal temperature would preclude significant melting. For example, if the lowland crust had similar excess potential temperature as normal terrestrial mantle, then a 105øC increase would produce the 21-km thicker crust beneath the highlands. I make no attempt to estimate crustal thickness variations within the lowlands even though such variations would be useful for monitoring potential temperature variations during the time the lowland formed.

Axial topography. The lack of obvious axial topography constrains the rate of Martian seafloor spreading to be fast. Basically, ridges that are underlain by a significant axial magma chamber have smooth topography while rough ridges lack a

chamber. (Actually, most of the magma chamber is filled with mostly crystalline mush [e.g., Sinton and Detrick, 1992], but the mush apparently behaves as a weak material.) The mechanical processes involve resistance to spreading both from the cool crustal lid and the underlying upwelling mantle [Sleep and Rosendahl, 1979; Chen and Morgan, 1990a, 1990b; Chen and Phipps Morgan, 1993; Phipps Morgan and Chen, 1993]. I qualitatively discuss the essence of this reasoning. Ridge axes are passive in the sense that forces are transmitted to theln by the oceanic plates. At slow ridges both the uppermost mantle and the crust are cool and act as strong regions. A significant driving force is thus needed for spreading. Rough topography forms when the cool lid fails by faulting. At fast ridges, the necessary driving force is much smaller because the strong regions are replaced by hot material and the cool lid is quite thin.

Lateral heat conduction is necessary to remove heat from the axial region and create a cool axial lid. The scaling between the crustal thickness needed for smooth topography and the spreading rate is easily obtained dimensionally. The scale time for crust to cool vertically to depth z is dimensionally

tr = z2/r (10)

where •c is the thermal diffusivity [Turcotte and Schubert, 1982, equation 4-164]. For lateral conduction to be as hnportant as vertical conduction, the lid needs to be deeper than wide. That is, the time for the base of the crust to spread a distance x equal to the lid thickness L needs to be about the time for it to cool

to depth L. The spreading time tL then equals the cooling time

tL = L /u = L 2/•c (11)

where u is the half spreading rate. This implies that the cool lid thickness varies inversely with spreading rate

L = •u (12)

A significant magma chamber can be considered absent when the lid thickness is equal to the crustal thickness. The critical spreading rate for smooth topography defined in this manner scales as

Uc = rdC (13)

where C is crustal thickness. When seafloor spreading is sufficiently rapid that crustal thickness is determined by melt production M, the critical spreading rate scales inversely to gravity at a given excess potential temperature.

Hydrothermal circulation in addition to conduction removes significant amounts of heat from the ridge axis. For ridges near the rough-smooth transition, 60 and 80 mm yr -I full rate [Small and Sandwell, 1992], the magma chamber is about 2 km deep rather than the 1 km that would be expected from pure conduction. (A seismically detectable magma chamber is absent below a full spreading rate of about 45 mm yr -I for 6-km-thick oceanic crust [Phipps Morgan and Chen, 1993].) Hydrothermal circulation is also expected on ancient Mars since liquid water was likely to be present. The maximum depth of vigorous hydrothermal circulation is likely to be controlled by the presence of open cracks which close from the difference between water and rock pressure. As the increase of this pressure difference with depth scales with gravity, the thickness of the region of vigorous hydrothermal circulation scales inversely with gravity. As this is the same scalir•g as for conduction, there is no need to the first order to consider hydrothermal circulation separately


when extrapolating from the Earth to Mars. If anything, the low water or even atmospheric pressures at a Martian ridge axis allowed pyroclastic and phreatic eruptions like in Iceland. The highly porous material in the uppermost Martian crust would enhance hydrothermal circulation to some extent. There is also no need to consider heat transfer from a subaerial ridge axis through the Martian atmosphere. As on the Earth, heat flow from the interior does not affect surface temperature because the internal heat fluxes are much less than solar fluxes.

Published thermal models of terrestrial ridge axes (as well as interpreted seismic sections) can be rescaled for the factor of 2.5 ratio of pg between Earth and Mars by increasing the length scale by a factor of 2.5 and decreasing the spreading rate by a factor of 2.5. This scaling was checked numerically by using simple code that includes gravity and latent heat. If hydrothermal circulations are represented in the models as heat sinks, then the total heat sink in a volume of crust during its history remains the same. The time that a block of material remains at

the ridge axis is increased by the lower spreading rate so heat sinks expressed as power per volume are decreased by a factor of 2.5. The total hydrothermal heat flux per area of crust is increased by 2.5 because the length scale over which circulation occurs is increased. There is thus little need to compute ridge thermal models specifically for Mars.

Small and Sandwell [1992] cite two examples which aid in appraising the effects of crustal thickness at constant spreading rate on topographic roughness. (1) The Reykjanes ridge near the Iceland hotspot spreads at 20 mmyr -1 but has unusually thick crust as would be expected from high mantle temperatures. The roughness is much reduced from normal ridges but is somewhat higher than the roughness of fast ridges. Martian ridges with similar crustal thickness would have cooler mantle and thus

be somewhat rougher. (2) The Australia-Antarctic discordance zone spreads at 76 mm yr -1 but has thin crust and cooler upwelling mantle. It is significantly rougher than normal crust formed at the same spreading rate. This indicates that rough topography will form at any spreading rate if the mantle is cool enough.

Scaling for gravity alone, the rough-smooth transition on Mars should be between 24 and 32 mm yr -1. The actual transition is probably somewhat faster as very porous upper crust rocks on Mars should enhance hydrothermal circulation. The lack of abyssal hill lineations over most of the northern lowlands implies a spreading rate greater than the lower limit of this range. The spreading rate would be expected to increase from this limit near the plate pole to significantly more rapid rates further away. This conclusion is basically compatible with that obtained above from the lack of fracture zone scarps. Both conclusions apply to the time between Plates 1 and 2 when mostly old highland crust was subducted. For example, a full spreading rate of 90 mmyr -1 90 ø from the plate pole would imply 66 m.y. between the two reconstructions, while a rate 24 mm yr -1 would imply 246 m.y. These durations are not in conflict with the absolute time estimates in Table 1.

Ridge-trench interaction. Young oceanic crust is difficult to subduct and spreading slows or ceases as a ridge approaches a trench. This occurs because oceanic crust at low pressures is less dense than the mantle. Therefore, cooling must penetrate a significant distance into the mantle to provide a driving force for subduction. The minimum age for easy subduction scales to the crustal thickness squared from (11) or inversely with gravity squared [Warren, 1993].

Ridge-trench interaction has been extensively studied in the

Pacific Ocean off North America [Atwater, 1989; Atwater and Severinghaus, 1989]. During the Tertiary Period, the East Pacific Rise approached the subduction zone along North Amer- ica. The initially large Farallon plate broke into two major plates when the first transform-ridge intersection collided with the trench. Various small plates were created and subduction of these plates ceased when only very young crust remained. These data indicate that a terrestrial plate with most of the crust younger than 10 m.y. is somewhat difficult to subduct, while a terrestrial plate with most of the crust younger than 5 m.y. is quite difficult to subduct. Rescaling by the inverse of gravity squared to Mars yields ages of 62 m.y. and 31 m.y., respectively.

The part of the Boreal plate subducting beneath Tharsis Montes and Tempe Terra at the time of the reorganization varies from 4200 km (71 ø about the plate pole, 0 ø latitude with respect to the plate pole) on the north to 475 km (12 ø about the plate pole, -48 ø latitude with respect to the plate pole) on the south. If all the plate was younger than 62 m.y. at the time of the reorganization, the full spreading rate (90 ø fi'om the plate pole) would be 135 mm yr -•. If the closest point on the south was 62 m.y. old, then the spreading rate (90 ø from the plate pole) was 23 mm yr -1. The actual rate was likely to be somewhere in between, within the range of estimates obtained by considering the lack of ridge and transform lineations.

An estimate of the spreading rate at Alba Patera before rapid plate motions ceased can be obtained by noting that the northern end of fractures associated with Alba Patera is about 700 km

from the Tharsis Montes trench. If all the plate was younger than 31 m.y., a full spreading rate of 45 mm yr -1 is obtained. Similar reasoning applied to the 1300-km-long transform north of Olympus Mons yields a spreading rate of 84 mm yr -•, again within the range discussed above.

Minimum rate of seafloor spreading. A minimum spreading rate exists below which lateral heat conduction precludes the hot mantle upwelling zone that is required at the ridge axis for it to be weaker than the surrounding lithosphere. This process is similar to the rough-smooth transition except that it occurs at a lower spreading rate because much more drastic cooling of the mantle is required. This rate can be monitored by its effects on crustal thickness. If lateral conduction causes the mantle to

ascend at temperatures below its adiabat, it melts much less at depth than normal mantle and any melts that are produced tend to freeze at normally subcrustal depths. The amount of melt available for producing crust is thus reduced and the thinner crust exposes more mantle to cooling by conduction.

Petrological studies [Klein and Langmuir, 1987; Niu and Batiza, 1993] and seismic studies [White et al., 1992] indicate that this minimum limit where oceanic crust is significantly thinner is below the rates commonly observed on the Earth. Lateral conduction is important only near long transform faults and breakup margins. The Arctic spreading center as it approaches the Asia-North America plate pole is the one useful terrestrial example. The minimum full rate of ridge-ridge transforms of 13.5 mmyr -• is a reasonable estimate of the minimum full spreading rate [Stoddard, 1992]. Rescaling for the lower gravity of Mars using (13) yields a minimum spreading rate of 5.4 mm yr -•.

Time to form northern lowlands. I have obtained estimates

of the spreading rate which formed the bulk of the smooth northern lowlands of around 80 mm yr -• and the minhnum full rate that produces oceanic crust of 5.4 nun yr -•. The northern lowlands are over 8000 km across. Spreading at the faster rate


for 200 m.y. would bring the ridge from the breakup margin to the trench. Sluggish spreading within Alba Patera to produce an upper limit of 500 km of rough crust at the low rate is geometrically permitted for around 100 m.y. These durations are short enough to be compatible with published absolute age scales (Table 1), but long enough that age differences might be detected from crater counts. Thermal topography associated with these age differences has now vanished. It is conceivable that paleotopography when the ridge crest was high relative to flanks might be detected from studies of channel or lava flow patterns.

Stress and Planetary Thermal Evolution

Plate tectonics tends to dominate the thermal budget and lithospheric stress state of the Earth, the one planet where it clearly occurs. Reexamination of Martian thermal history and planetwide stresses is thus in order because the assumption of a qne-plate planet has permeated previous discussions including those involving the author.

Early thermal models of Mars focussed on widespread extensional features on the surface of the planet. The planet was assumed to have accreted cool. Its interior gradually heated up from radioactivity [e.g., Solomon and Chaiken, 1976]. Core formation occurred quickly once the mantle was warm enough to flow. Movement of dense iron to the core heated the planet by an additional 300øC [Solomon, 1979]. Extensional features occurred as the one-plate planet heated up. Compressional features formed as the planet cooled thereafter.

It is now known that this thermal history is untenable. Lead isotopic data from Martian associated meteorites indicate that the Martian core formed quite early in the planet's history, implying that the planetary interior was hot right after accretion [e.g., Chen and Wasserburg, 1986]. More recent studies of photogeological data are not compatible with the implied sequence of events. That is, the tectonic history of Mars is not simply an evolution from early extension to late compression. Continual gradual compression of a one-plate planet that accreted hot and then cooled is also unsatisfactory [Branerdt et al., 1992; Watters, 1993].

Thus there is motivation for investigating the more complicated effects of plate tectonics on stress and thermal history. I present topics in an order that is increasingly speculative, beginning with thermal history, which in part can be kinematically related to plate tectonics. The evolution of the lowland- highland passive margin which is related to the thermal age of the lithosphere is considered next. Stress is then related to plate tectonics in a more complicated way. Finally, a plausible expla- nation is presented for late voluminous volcanism in Tharsis Montes, Alba Patera, and Olympus Mons.

Thermal history. The thermal history of the interior of Mars can be addressed by considering the demise of plate tectonics along with the probable hot interior just after accretion and core formation. The former constraint may be relatively simple. Seafloor spreading becomes difficult if the mantle is too cold to melt extensively. Mantle cooling may have halted plate tectonics on Venus at 500 Ma [Turcotte, 1993]. Low mantle temperatures are starting to affect terrestrial seafloor spreading within the Australia-Antarctic discordance zone. That is, the excess potential temperature at the time the lowlands formed needs to be relatively cool (comparable to the present Earth as assumed in the discussion of spreading rates) so that further cooling ended rapid plate motions.

The earliest thermal history of Mars is more difficult to address. Presumably, the earliest Earth and Mars cooled by conduction through a thin lid of frozen crust and by eruption and cooling of magma at the surface. The latter mechanism is quite efficient when upwelling material is hot enough to almost totally melt [Davies, 1990]. The heat transfer rate is limited only by the rate that solid material can upwell into the zone where nearly complete melting occurs. Lava erupted to the surface cools essentially to the surface temperature. In contrast, plate tectonics requires the lithosphere to cool by conduction. Cooling by erupting magma may have prevented a large molten magma ocean on the Earth [Davies, 1990].

The scaling for gravity and the inability of plate tectonics to rapidly remove heat from a hot planet are addressed following the treatment by Turcotte [1993] for Venus by considering the balance between heat of the planetary interior, radioactive heat generation, and surface heat flow

M C•, -•- = M QR -A qavg (14) where M is planetary mass, QR is radioactive heat generation per mass, A is surface area, and q avg is average surface heat flow. For a given heat flux, Mars cools much faster than the Earth because its surface to mass ratio is greater. The scaling for cooling rate depends inversely on gravity as can be seen from rewriting the inverse square law

GM 4rcGM 4rcGM = = (15) g = r 2 4/1;r 2 A

where G is the gravitational constant. Assuming properties similar to the Earth, a surface heat flow unbalanced by internal radioactive heat generation of 40 mWm -2 would cool the interior of Mars by 250øC in 1 b.y. compared with 100øC for the Earth.

There is a maximum rate that plate tectonics can remove heat because as noted above the mantle beneath the crust needs to

cool before it is subducted [Davies, 1990; Warren, 1993]. Quantitatively, the surface heat flow from plate tectonics depends inversely on the square root of age

q = Cq (tc) -'/• (16)

where Cq implies 500 mWm -2 heat flow through 1-m.y. crust. The average heat flow, assuming that all crust subducts at the same age ts, is

ts

qavg = ts :• I Cq (tc)-'/2dtc = 2C• (ts) -'/2 (17) The crustal thickness scales inversely with gravity from (8), the cooling time scales linearly with gravity, and from (11) cooling time scales to this thickness squared. The maximum average heat flow is linearly dependent on gravity:

qavg = = 2Cq (Cg g-2)-'•, = 2Cq g (Cg)-'A (18)

where Cg is another constant. The cooling rate at this heat flow (if unbalanced by radioactivity) is independent of gravity

aT A q avg = (19)

at M C•, from (15).

However, plate tectonics cannot efficiently cool a planet with an excess potential temperature of several hundred degrees just after accretion [Davies, 1990; Warren, 1993]. For example, consider 36-kin-thick crust in the southern highlands with an


excess potential temperature 105øC hotter than 15-krn-thick lowland crust. Such highland crust is difficult to subduct when it is younger than 360 m.y. (scaling from 10 m.y. for 6-km crust on the Earth). Much thicker crust would preclude plate tectonics altogether on Mars, just like on the Moon [Warren, 1993]. The maximum cooling rate is 330øC in a billion years if unbalanced by radioactivity. Actual radioactive heat generation early in the planet's history might balance this cooling rate.

Although the details are not known, it is likely that the dense crust formed by very extensive melting of the mantle early in terrestrial history would be easily subducted once plate tectonics started [Davies, 1990]. The excess potential temperature of the Martian mantle when plate tectonics became the dominant heat transfer mechanism is harder to estimate. Overcooling of the mantle to an average temperature well below that where extensive melting occurs is conceivable because (similar to plate tectonics) the cooled material sinks rapidly in large blocks to the base of the mantle, while hot material ascends elsewhere to

shallow regions. In an extreme limit, the entire mantle of Mars could erupt over a period of time, cool to the surface temperature, and sink if no heat was transferred by conduction. A likely maximum estimate is the temperature of the Archean mantle of the Earth.

As mentioned above, such early oceanic lithosphere would take hundreds of million years to become subductable on Mars. The surviving crust in the southern highlands might include both this immediate secondary crust and later crust generated by plate tectonics when the excess potential temperature of the Martian mantle was about 100øC hotter than that when the lowlands formed.

Mars is small enough that subduction of thick old lithosphere once started significantly cools the interior of the planet. For example, consider subduction of 400-m.y.-old lithosphere. The heat removed from the lithosphere over this interval is equivalent to cooling a region of thickness

Zt_ 4}Ctc (20)

where }c is thermal diffusivity from the interior temperature to the surface temperature [Turcotte and Schubert, 1982, equation 4-202]. The plate thickness defined by the depth where the cooling is 1/10 of the difference between surface and interior temperature is

Zt = 2.32 (rtc) •/2 (21)

[Turcotte and Schubert, 1982, equation 4-126]. Letting }c = 0.7 x 10 -6 m 2 s -1, an equivalent 106-kan layer at surface temperature is subducted. This is 1/11 of the volume of the planet and approximately 1/11 of the heat capacity. Subduction of a global 400-m.y. lithosphere would cool the planetary interior by about 140øC. The subducted plate thickness from (21) is 218 km, which is 2/11 of the planetary volume. Thus plate tectonics might quickly turn the mantle of Mars inside out cooling its interior.

To summarize, two mechanisms of heat transfer were dis-

cussed above: plate tectonics is moderately efficient at removing heat if the mantle is not so hot that the oceanic crust is too thick

to quickly subduct or so cold that ridge axes are too strong; this range is mach narrower for Mars than the Earth. Nearly complete partial melting and eruption to the surface can remove vast amounts of heat, but only from very hot ascending regions. A third mechanism, secondary convection at the base of the litho-

sphere, is inefficient at the current excess potential temperature in the Earth removing only about 1% of the heat [Davies, 1988]. It is apparently at best moderately efficient above hot mantle plumes in the Earth [Davies, 1992]. (The heat flow from secondary convection scales to the half-space viscosity at the base of the boundary layer to the-1/3 power and to the temperature contrast across the actively flowing boundary layer, which actually drives conv•tion, to the 4/3 power. The relevant temperature contrast scales to the temperature decrease needed to increase viscosity by a factor of e [Davaille and Jau- part, 1993a, 1993b], while the much larger temperature contrast across the lithosphere drives plate tectonics.) I propose a con- ceptual thermal history of Mars with these mechanisms in mind:

1. During accretion and core formation, heat is rapidly removed by melting and eruption. The deeper mantle is cooled as the frozen erupted material sinks in large blocks. This stage predates the geological record.

2. Primary crust from nearly complete melting is subducted and thick oceanic crusts covers the planet. This lithosphere cools and thickens. Excess potential temperature just below the lithosphere is ~100øC above lowland excess potential temperature. At this stage most of the mantle is stably stratified relative to the cold material which sank to great depths. The lithosphere and the region just above the core are unstable boundary layers. Some of the early oceanic crust may still exist as lowermost Noachian surface.

3. Plate tectonics replaces much of the lithosphere from the earlier stage. Formation of the northern lowlands is the last event in this process. The excess potential temperature at the base of the lithosphere becomes too cold for rapid plate tectonics. New lithosphere may underlie some Noachian surfaces in the southern highlands. The lowlands are Upper Noachian to Lower Hesperian.

4. Any subsequent plate motions are too sluggish to affect the interior temperature. The thermal evolution of the planet i• thus effectively one plate from the Late Hesperian Epoch to the present. At first, the lithosphere cools down and the interior heats up until the interior viscosity is low enough for'secondary convection at the base of the lithosphere to balance radioactivity. The interior then slowly cools as radioactivity is used up.

Plate tectonics was inferred to have started earlier in the

sequence of events when the inside of the planet was hot and and perhaps later. The main physical difficulty in starting plate tectonics is that slabs provide a large driving force but rupture of older lithosphere is needed to start a new subduction zone. On the Earth, new subduction zone typically nucleate from other presumably weak boundaries such as transform faults [Mueller and Phillips, 1991]. I make no attempt to develop a quantitative theory for Mars. Rather, I note that initial zones of weak- ness to form subduction zones are likely to have existed on the hot young planet and also during episodes where plate tectonics became sluggish during the formation of the highlands.

Evolution of the passive margin escarpment. The thermal age of subducted highland crust was likely several hundred million years. For example, 36-km thick crust would be difficult to subduct if it was younger than 360 m.y. The elevation of such crust would be 1.7 km below the elevation of a ridge axis forming lowland crust (using subsidence rates beneath air). This difference is compatible with the inference that old highland crust was preferentially subducted relative to young lowland crust because to the first order the buoyancy of the lithosphere is proportional to its elevation.


Thus the lowlands were initially higher than the adjacent highlands and the sense of the lowland-highland escarpment reversed a significant time after breakup. If the highlands had a thermal age of 360 m.y. at the time of breakup, both the highlands and the adjacent oceanic crust would be 2.25 km lower than the ridge axis 81 m.y. after breakup. The throw on the escarpment would gradually increase thereafter, for example, to 0.75 km 240 m.y. after breakup.

This sequence of events has simliar geological implications to the hypothesis of McGill [1989] and McGill and Dirnitrou [1990] that the lowlands subsided relative to the highlands by internal processes as the older surface associated with breakup would be present on both sides of the transition. This surface might be well developed because lava flows, ash flows, and even fluvial sediments would have been able to spread out from the ridge axis over adjacent highlands. It would have been broadly warped by flexure as moderately thick lithosphere existed at the time of slope reversal. A younger surface would correspond to the time of escarpment reserval when extensive water bodies ponded in the lowlands rater than the highlands.

Stress history. Crosscutting relationships of extensional grabens and presumably compressional wrinkle ridges allow photo- geologists to construct histories of the lithospheric stresses on Mars [e.g., Tanaka et al., 1992; Watters, 1993]. Such histories have usually been interpreted Mars in terms of a one-plate planet. The hypothesis that the crust within the Tharsis region, including Alba Patera and Olympus Mons, may have been decoupled like a blister from the underlying strong lnantle [Tanaka et al., 1991] is an exception. Plates tectonics is expected to strongly affect stresses in the lithosphere. In particular, it may cause rapid changes back and forth between lithos-

pheric extension and lithospheric compression. I discuss this implication below in terms of membrane stresses, that is, the differences between the horizontal stress and lithostatic stress.

This is convenient, as photogeological features occur at the surface, where the lithostatic pressure is zero.

I begin by reviewing mechanisms that produce widespread membrane stresses in the lithosphere. Both isostatic compensation and flexurally supported long-wavelength loads are likely to be important on Mars [e.g., Banerdt et al., 1992]. The stresses from isostasy do not relax until great amounts of flow have removed less dense roots to produce a uniform lithosphere and are present both on one-plate and multiplate planets. Planetwide flexure is a feature of one-plate planets only, as it is relaxed by small horizontal strains, several kilometers of extension (or compression) across a 1-radian arc on Mars. In addition, a planetwide bias in membrane stresses may exist; it is conceivable for the membrane stresses over the surface of the planet to be all compressional or all tensional.

The early one-plate thermal histories of Mars were based on such biases. Graben may be cut by perpendicular wrinkle ridges (or sometimes vice versa on the real planet). This indicates a change from a tensile bias to a compressional bias without changing the difference between the horizontal principal stresses. The stress history when interpreted in terms of a one- plate planet involves thermal expansion and contraction of the sublithospheric interior which behaves as a fluid (Figure 4). The lithosphere is already cool and the vertical stress pgz is determined by the overburden. Expansion of the interior, for example, stretches the lithosphere, making •nembrane stresses more tensile [Watters, 1993]. If the lithosphere (or even the uppermost crust) fails under this stress, grabens may form.

Earth

unpinned ridge axis

Midplate compression

One plate planet

Pinned ridge axis

Midplate tension

Global stress bias

Figure 4. Schematic representation of possible mechanisms that may have influenced membrane stress on Mars. (Top, left) Ridge axes are weak on the Earth. This condition may have caused episodes of midplate compression when rapid plate motions occurred on Mars. (Bottom, left) Ridge axes are strong when spreading is sluggish. This condition way have caused widespread extensional episodes on Mars. In contrast, the interior of a one-plate planet heats up (or cools) slowly. (Right) The lithosphere is stretched (or compressed) globally. Changes between tensional and compressional bias occur slowly with this mechanism, which probably applies to latest Hesperian and Amazonian Mars.


However, only small strains, which would be vastly overwhelmed by plate tectonics, can be produced in this way. For example, a 100øC temperature change within Mars would increase (or decrease) the circumference by 22 km [e.g., Ban- erdt etal., 1992]. In addition, any change from planetwide extension to compression is slow because the interior temperature within a one-plate planet changes slowly.

Plate tectonics involves large horizontal strains and also allows the bias in membrane stress to change rapidly back and forth between tensile and compressional. I illustrate this feature using the ridged plains as an example. That is, the eruption of volcanic rocks on the plains may reflect a tensile bias to let the lava out through rifts. The ridges themselves reflect the most obvious compressional episode in the planet's history [Watters, 1993]. The physics of a mechanism that may change the stress bias is illustrated by considering the Earth (Figure 4). Isostatic compensation tends to produce membrane tension over regions with elevated topography and less dense roots like continents and ridge axes. The axes of midoceanic ridges, however, are weak and the principal membrane stress perpendicular to them is small. The membrane stress in oceanic crust becomes more

compressive away from the ridge axis. Because the axial membrane stress is small, the stress bias in the ocean basins is

compressive [Stein and Pelayo, 1991; Richardson, 1992]. Con- versely, if ridge axes were strong and subduction zones weaker, more of the force from the downgoing slab would be transmitted to the horizontal plate which then would have membrane tension.

The post-Noachian history of Mars may involve such transi- tions. The eruption of lava on the ridged plains is attributed to planetwide tension when a previous period of seafloor spreading ceased but remnants of slabs (in unknown positions) still produced a driving force. Once breakup occurred and a fast ridge started to spread, the weak ridge axis and its elevated topography caused a planetwide compressional bias. This compressional bias declined as the length of active ridge decreased, and ceased once rapid spreading ceased. Since only a few tens of kilometers of spreading and subduction would overwhelm one- plate effects of planetwide cooling and stress, sluggish failed attempts at plate tectonics may later have modified the planetwide stress bias. Extension across canyon systems west of Tharsis Montes, including Valles Marineris, during Early Hesperian is attributed to stress changes associated with plate reorganizations. Late Hesperian extension across these features is attributed to stress changes associated with sluggish plate movements. Eventually, the lithosphere became thick enough to behave as a one-plate planet. Flexural compensation of widespread volcanic loads within the Tharsis region, including Alba Patera and Olympus Mons, increased geoid anomalies to the large values observed today.

Tharsis volcanism. As just noted, volcanism in the Tharsis region continued well after the time proposed for plate tectonics. In the reconstructions, Tharsis Montes is an extinct island arc,

while Alba Patera and Olympus Mons grew on top of extinct ridge axes. A mechanism for producing late voluminous volcanism thus needs to be found. I present one involving subducted water in oceanic crust.

Subduction zone volcanism on the Earth is related to the

dehydration reactions within subducted oceanic crust. The water thus released lowers the melting temperature in the asthenospheric mantle above the slab. The process should have also occurred on Mars, but with the difference noted above that

the extensively altered zone of the oceanic crust is 5 km thick

rather than about 2 lcm on the Earth for fast ridges [Pelayo et al., 1994] because of the effect of lower Martian gravity on open cracks. The deeper part of the altered zone on Mars may have remained cool enough that it did not dehydrate but sank with the slab. Eventually, plate tectonics would have generated great volumes of hydrous material within dead slabs with the deep mantle.

This process was probably efficient at removing surface water from Mars, as can be seen from a simple mass balance. Typical greenstones formed at ridge axes have about 3% water by weight or 10% by volume. Alteration of 5 km of basalt thus removes 500 m of water. An equivalent 1-km-thick global layer of water would give Mars the same ocean per mass as the Earth. Subduction of the surface area of Mars plus hydrothermal alteration of the new crust covering the planet would remove half the ocean to the deep mantle and the other half to the oceanic crust. Subducted lithosphere has an ample supply of water for later volcanism. For example, lithosphere subducted beneath Tharsis Montes (1/3 of the area of the planet: about half the total area of the lowlands times the factor of two

ratio between subducted area and plate area in Figure 2) accounted for an equivalent 130-m-thick water layer. For com- parison, the total equivalent water layer for all mapped Upper Noachian and younger lavas on the planet is about 40 m [Gree- ley, 1987].

The history of late volcanism within Tharsis as defined here started when rapid plate tectonics ceased. The lithosphere was then thin beneath the dying spreading centers as well as the Tharsis Montes arc. Detachment of the slab as well as sluggish plate movements helped maintain local areas of thin lithosphere into Late Hesperian time. Hot upwelling mantle tended to pond beneath areas of thin lithosphere. The hydrous nature of the upwelling material and the thin lithosphere increased the fraction of partial melting from that at the base of more normal lithosphere. The process gradually decreased as the interior temperature of Mars decreased, vel3, hydrous regions were used up by partial melting or dispersed by mantle flow, and the lithosphere in the active regions thickened.

Conclusions

Plate tectonics is now a viable mechanism for producing the northern lowlands of Mars. It does better in explaining available data than a single impact hypothesis which conflicts with the probable Late Noachian to Early Hesperian age of the transition zone escarpment. The multiple large impact hypothesis has this difficulty as well as the unlikelihood that most of the large impacts would land within 1/3 of the planet. There are relatively few Upper Noachian to Lower Hesperian features within the lowlands. Obviously, the few observed lineations can be attributed either to plate tectonics or rings and radial fractures from a few poorly documented impact basins. Plate tectonics, unlike other proposed mechanisms, directly explains the grossest observed feature: the low elevation of the northern plains.

The geological features and sequence of events discussed here are intended to be testable by photogeology and at some time by geophysics and even smnpling. At present, photogeology is most promising, as it yields relative age relationships as well as finds linear features. Interpretation of the surface of Mars in the time transgressive and mobile framework implied by plate tectonics has just begun. I expect that currently available photographs will thus be useful. Intermediate wavelength gravity from satellite tracking is most likely to be useful, as the


longest wavelengths are dominated by (later) membrane flexural compensation of Tharsis, while short wavelengths are not meas- ured. For example, bending flexural compensation of passive margins, dead subduerion zones, and oceanic fracture zones might all be detected. Thermal subsidence of oceanic crust and passive margins might be detected from changes in channel slopes from accurate altimeter data combined with geomorphol- ogy. As with gravity, intermediate wavelengths are most likely to be useful because long wavelengths are dominated by the later history of Tharsis.

The well-defined kinematics of plate tectonics and its simple physics allow conclusions to be drawn about plate tectonics in the lowlands, once one presumes that plate motions actually occurred. In particular, the lowlands are smooth and lack evident transform and ridge fabrics. The preferred spreading rate is thus fast, around 80 mmyr -• 90 ø from the plate pole. This constraint holds even if my reconstructions are incorrect. Inferred plate reorganizations as the ridge approached the trench indicate young oceanic crust extended far from the ridge and hence the spreading rate was fast.

A useful scaling relationship is that oceanic c•a•stal thickness at a given excess potential temperature scales inversely with planetary gravity. A 100øC temperature decrease in the excess potential temperature of the Martian interior is needed for the lowland crust to be 3 km lower and 21 km thinner than highland crust. Other scaling relationships are obtained from this inference by considering lateral heat transfer. The spreading rate for the rough-smooth topography transition scales inversely with planetary gravity, as does the much slower spreading rate where well-defined spreading ceases. The age were oceanic crust becomes difficult to subduct scales inversely with gravity squared.

Plate tectonics also changes many of the constraints inferred from physical models of a one-plate planet. The strains associated with plate tectonics greatly overwhelm planetwide extension (compression) associated with heating (cooling)of the planetary interior. Subduction of old lithosphere would cool the interior of Mars much more rapidly than cooling though a thick lithosphere. Subduction when it occurs is also quite effective at removing surface water to the deep mantle since the thickness of hydrated oceanic crust varies inversely with gravity. Some of this deeply subducted hydrous mantle may have later upwelled and produced late volcanism in the Tharsis region.

Finally, the study of Mars will aid us in understanding the Earth. In particular, confirmation of plate tectonics on Mars gives another example that differs from the Earth due to lower gravity. Crust produced by spreading when the mantle was hotter may still exist in the southern highlands, while Archcan crust produced with similar excess potential temperatures on the Earth has been subducted.

Acknowledgments. MEVTV conferences supported by the Lunar and Planetary Institute helped me develop my ideas. Conversations with Don Turcotte were especially helpful. I thank Don Turcotte and Matt Golombek for reviewing the paper. Questions by Walter Kiefer at the Fall 1993 AGU meeting help me realize that the sense of the passive margin escarpment probably reversed during subsidence. Mike Ravine and Jason Phiptx• Morgan independently came up with an alternative plate arrangement. I thank Ginger Barth for critically reading the paper and John Hole for helping me get started with GMT plotting. This research was supported in part by the National Science Foundation grant EAR-9204708

References

Atwater, T., Plate tectonic history of the northeast Pacific and western North America, in The Eastern Pacific Ocean and Hawaii, The Geol- ogy of North America, vol. N., edited by E. L. Winterer, E. L., D. M. Hussong, and R. W. Decker, pp. 21-72, Geological Society of Amer- ica, Boulder Colo., 1989.

Atwater, T. and J. Severinghaus, Tectomc matx• of the northest Pacific, in The Eastern Pacific Ocean and Hawaii, The Geology of North America, vol. N., edited by E. L. Winterer, E. L., D. M. Hussong, and R. W. Decker, pp. 15-20, Geological Society of America, Boulder Colo., 1989.

Banerdt, W. B., M.P. Golombek, and K. L. Tanaka, Stress and tectomcs on Mars, in Mars, edited by H. H. Kieffer, B. M. Jakosky, C. W. Snyder, and M. S. Matthews, pp. 249-297, University of Arizona Press, Tucson, 1992.

Bak•on, R. M., P.M. Bridges, and J. L. Inge, At•v of Mars: The 1:5,000,000 Map Series, 146 pp., NASA, Washington, D.C., 1979.

Chen, J. C., and G. J. Wasserburg, Formation ages and evolution of Shergotty and its parent planet from U-Th-Pb systematics, Geochim. Cosmochim. Acta, 50, 955-968, 1986.

Chen, Y., and W. J. Morgan, Rift valley/no rift valley transition at xnid- ocean ridges, J. Geophys. Res., 95, 17,571-17,582, 1990a.

Chen, Y., and W. J. Morgan, A nonlinear fileology model for mid-ocean ridge topography, J. Geophys. Res., 95, 17,583-17,604, 1990b.

Chen, Y. J., and J. Phipps Morgan, The genesis of oceanic crust: Magma injection, hydrothermal circulation, and crust flow, J. Geo- phys. Res., 98, 6283-6297, 1993.

Craddock, R. A., R. Greeley, and P. R. Christensen, Evidence for an ancient impact basin in Daedalia Planurn, Mars, J. Geophys. Res., 95, 10,729-10,741, 1990.

Davaille, A., and C. Jaupart, Thermal convection in lava lakes, Geophys. Res. Lett., 20, 1827-1830, 1993a.

Davaille, A., and C. Jaupart, Transient high-Rayleigh-number thermal convection with large viscosity variations, J. Fluid Mech., 253, 141- 166, 1993b.

Davies, G. F., Ocean bathymetry and mantle convection, 2, Sxnall-scale flow, .I. Geophys. Res., 93, 10481-10488, 1988.

Davies, G. F., Heat and mass transport in the early Earth, in Origin of the Earth, edited by H. E. Newsom, and J. H. Jones, pp. 175-194, Oxford University Press, New York, 1990.

Davies, G. F., Temporal variation of the Hawaiian plume flux, Earth Planet. Sci. Lett., 113, 277-286, 1992.

Forsyth, D. W., Crustal thickness and the average depth and degree of melting in fractional melting inodels of passive flow beneath midoceanic ridges, J. Geophys. Res., 98, 16073-16079, 1993.

Frey, H., and R. A. Schultz, Large impact basins and the inega-iinpact origin for the crustal dichotoiny on Mars, Geophys. Res. Lett., 15, 229-232, 1988.

Frey, H., A.M. Semeniuk, J. A. Seineniuk, and S. Tokarcik, A widespread common age resurfacing event in the higlfland-lowland transition zone in eastern Mars, Proc. Lunar Planet. Sci. Conf., I8th., 679-699, 1988.

Greeley, R., Release of juvenile water on Mars: Estimated amounts and tiining associated with volcanism, Science, 236, 1653-1654, 1987.

Greeley, R. and J. E. Guest, Geological •nap of the eastern equatorial region of Mars, U.S. Geol. Surv., Misc. Invest. Set., Map 1-1802-B, 1987.

Kieffer, H. H., B. M. Jakosky, and C. W. Snyder, The planet Mars: From antiquity to present, in Mat:v, edited by H. H. Kieffer, B. M. Jakosky, C. W. Snyder, and M. S. Matthews, pp. 1-33, University of Arizona Press, Tucson, 1992.

Klein, E. M., and C. H. Langmuir, Global correlations of oceanic ridge basalt chemistry with axial depth and crustal tlfickness, .I. Geophys. Res., 92, 8089-8115, 1987.

Longhi, J., E. Knittie, J. R. Holloway, and H. Wlinke, The bulk composition, mineralogy and internal structure of Mars, in Mars, edited by


H. H. Kieffer, B. M. Jakosky, C. W. Snyder, and M. S. Matthews, pp. 184-208, University of Arizona Press, Tucson, 1992.

McGill, G. E., Buried topography of Utopia, Mars: Persistence of a giant impact depression, J. Geophys. Res., 94, 2753-2759, 1989.

McGill, G. E., and A.M. Dimitrou, Origin of the Martian global dichotomy by crustal thinning in the Late Noachian or Early Hesperian, J. Geophys. Res., 95, 12,595-12,605, 1990.

McGill, G. E. and S. W. Squyres, Origin of the crustal dichotomy: Evaluating hypotheses, Icarus, 93, 386-393, 1991.

McKenzie, D., The generation and compaction of partial molten rock, J. Petrol., 25, 713-765, 1984.

McKenzie, D. and M. J. B ickle, The volume and co•nposition of melt generated by extension of the lithosphere, .jr. Petrol., 29, 625-679, 1988.

McKenzie, D., and R. K. O'Nions, Partial melt distributions from inver- sion of rare earth element concentrations, J. Petrol., 32, 1021-1091, 1991.

Mueller, S., and R. J. Phillips, On the initiation of subruction, J. Geo- phys. Res., 96, 651-665, 1991.

Niu, Y., and R. Batiza, Chemical variation trends at fast and slow spreading •nid-oceanic ridges, J. Geophys. Res., 98, 7887-7902, 1993.

Pelayo, A.M., S. Stein, and C. A. Stein, Estimation of oceanic hydrother•nal heat flux frown the depths of •nidoceanic ridge seisnficity and magma chainbets and heat flow data, Geophys. Res. Lett., in press, 1994.

Phipps Morgan, J., and Y. J. Chen, The generation of oceanic crust: Magma injection, hydrother•nal circulation, and crustal flow, .jr. Geo- phys. Res., 98, 6283-6297, 1993.

Richardson, R. M., Ridge force, absolute plate motions, and the intra- plate stress field, J. Geophys. Res., 97, 11,739-11,748, 1992.

Sclmbert, G., S.C. Solo•non, D. L. Turcotte, M. J. Drake, and N.H. Sleep, Origin and ther•nal evolution of Mars, in Mars, edited by H. H. Kieffer, B. M. Jakosky, C. W. Snyder, and M. S. Matthews, pp. 147- 183, University of Arizona Press, Tucson, 1992.

Scott, D. H., and K. L. Tanaka, Geological map of the western equatorial region of Mars, U.$. Geol. $urv., Misc. Invest. Set., Map 1- 1802-A, 1986.

Sinton, J. M., and R. S. Detrick, Mid-ocean ridge •nag•na chambers, .jr. Geophys. Res., 97, 197-216, 1992.

Sleep, N.H., and B. R. Rosendahl, Topography and tectonicq of •nid- ocean ridge axes, .jr. Geophys. Res., 84, 3831-3839, 1979.

S•nall, C., and D. T. Sandwell, An analysis of ridge axis gravity roughness and spreading rate, .jr. Geophys. Res., 97, 3235-3245, 1992.

Smith, D. E., F. J. Lerch, R. R. Nere•n, M. T. Zuber, G. B. Patel, S. K. Fricke, and F. G. Lemoine, An i•nproved gravity model for Mars: Goddard Mars model 1, J. Geophys. Res., 98, 20871-20889.

Solo•non, S. C., For•nation, history and energetics of cores in the terrestrial planets, Phys. Earth Planet. Inter., 19, 168-182, 1979.

Solo•non, S.C., and J. Chaiken, Ther•nal expansion and thermal stress in the Moon and terrestrial planets: Clue to early thermal history, Proc. Lunar $ci. Conf, 7, 3229-3243, 1976.

Stein, S., and A. Pelayo, Seis•nological constraints on stress in the oceanic lithosphere, Philos. Trans. R. Soc. London, $er. A, 337, 53-72, 1991.

Stoddard, P. R., On the relation between transfor•n fault resistance and plate motion, J. Geophys. Res., 97, 17,637-17,650, 1992.

Strom, R. G., S. T. Croft, and N. G. Barlow, The Martian impact cratering record, in Mars, edited by Kieffer, H. H., B. M. Jakosky, C. W. Snyder, and M. S. Matthews, pp. 383-423, University of Arizona Press, Tucson, 1992.

Tanaka, K. L., Martian geologic "revolutions": A tale of two processes (abstract), Lunar Planetary Sci., XXI, 1237-1238, 1990a.

Tanaka, K., The tectonic history of the Alba Patera-Ceraunius Fosse region of Mars, Proc. Lunar Planet. Sci. Conf, 20th., 515-523, 1990b.

Tanaka, K. L., and D. H. Scott, Geological map of the polar regions of Mars, U. $. Geol. Sum., Misc. Invest. $er., Map 1-1802-C, 1987.

Tanaka, K. L., M.P. Golombek, and W. B. Banerdt, Reconciliation of stress and structural histories of the Tharsis region of Mars, J. Geo- phys. Res., 96, 15616-15633, 1991.

Tanaka, K. L., D. H. Scott, and R. Greeley, Global stratigraphy, in Mars, edited by H. H. Kieffer, B. M. Jakosky, C. W. Snyder, and M. S. Matthews, pp. 345-382, University of Arizona Press, Tucson, 1992.

Turcotte, D. L., An episodic hypothesis for VenusJan tectonics, J. Geo- phys. Res., 98, 17,061-17,068, 1993.

Turcotte, D. L., and G. Schubert, Geodynamics: Applications of Contin- uum Physics to Geological Problems, 448 pp., John Wiley, New York, 1982.

Warren, P. H., The magma ocean as an impediment to lunar plate tectonics, J. Geophys. Res., 98, 5335-5345, 1993.

Watters, T. R., Compressional tectonics on Mars, J. Geophys. Res., 98, 17,049-17,060, 1993.

Wessel, P., and W. F. Haxby, Thermal stresses, differential subsidence, and flexure at oceanic fracture zones, .jr. Geophys. Res., 95, 375-391, 1990.

Wessel, P., and W. H. F. Smith, The GMT-SYSTEM v. 2.1 Technical Reference & Cookbook, SOEST/SIO, 1992.

White, R. S., D. McKenzie, and R. K. O'Nions, Oceanic crustal thick- hess from seismic measurements and rare earth element versions, J.

Geophys. Res., 97, 19,683-19,715, 1992. Wichman, R. W., and P. H. Sclmltz, Sequence and mechanisms of

deformation around the Hellas and Isidis impact basins of Mars, J. Geophys. Res., 94, 17,333-17,357, 1989.

Wilhelms, D. E., and S. W. Squyres, The Martian he•nispheric dichotomy may be due to a giant impact, Nature, 309, 138-140, 1984.

Zi•nbelman, J. R., S.C. Solomon, and V. L. Sharpton, The evolution of volcanism, tectonics, and volatiles on Mars: An overview of recent progress, Proc. Lunar Planet. $ci. Conf., 21th., 613-626, 1991.

N.H. Sleep, Department of Geophysics, Stanford University, Stan- ford, CA 94305. (email: [email protected])

(Received November 8; 1993; accepted January 24, 1994.)

martian plate tectonics - quarg: hometjb/het618/nathan/94je00216.pdf · sleep: martian plate...

Documents