Survey
* Your assessment is very important for improving the workof artificial intelligence, which forms the content of this project
* Your assessment is very important for improving the workof artificial intelligence, which forms the content of this project
JOURNAL OF GEOPHYSICAL RESEARCH, VOL. 92, NO. B12, PAGES 12,823-12,836, NOVEMBER 10, 1987 Mechanismsfor the Origin of Mid-OceanRidgeAxial Topography: Implicationsfor the ThermalandMechanicalStructureof AccretingPlateBoundaries JASON PHIPPS MORGAN Departmentof Earth, Atmospheric,and Planetary Sciences,Massachusetts Instituteof Technology,Cambridge E.M. PARMENTIER AND J. LIN Departmentof Geological Sciences,Brown University,Providence,Rhode Island To better understandthe implicationsof ridge axis morphologyfor the thermal and mechanicalstmcture of accretingplate boundaries,we examinemechanismsfor the creationof stress-supported axial topography and its dependenceon parameterssuch as spreadingrate, asthenosphereviscosity, and plate thicknessnear the ridge axis. Three basic mechanismsare considered:(1) asthenospheric upwelling in a narrow, steady-stateconduit beneatha ridge axis; (2) mantle flow stressesdue to plate spreading, extending ridge axis lithosphere, or melt segregationand associatedmantle compaction; and (3) topographysupportedby momentsdue to horizontal extensionalstressesin a thickeninglithosphere.If mid-oceanridge topographyis due to asthenosphere rising in a narrow conduit at the ridge axis, then the conduit structurecannot be thermally controlled, as the associatedthermal structurehas relatively broad, flat isothermsinsteadof a conduitlikeshape.Mantle flow associatedwith plate spreadingalone inducesno stress-supported topography. Thus the intuitive notion that divergingflow at the ridge axis due solely to plate spreadingproducesthe axial valley is incorrect. Vertical mantle flow at the baseof a stretching,accretinglithosphereand mantle flow associatedwith melt segregationcan, however, lead to stress-supported axial topography.However, for thesemantle flow related processesto producesignificant amounts of axial relief, the mantle viscosity beneath the ridge axis must be on the order of 1020Pas (1021P (poise)), significantly higherthangenerally accepted valuesof 1018Pas (1019P). Horizontal extensional stressesin a strong, brittle lithosphere that thickens with distance from the ridge axis can produceaxial topographyof the form and amplitudeobservedon slow spreadingridges. A continuumidealizationof deformationdue to faulting is used to examine this mechanism. This model predictsthe axial valley width is controlledby the plate thicknessnear the ridge axis. An 8-km-thick plate near the ridge axis would producethe 30-km-wide axial valley observedat slow spreadingridges. An increasedlithospherethicknessof only several kilometers over the half-width of the axial valley can producekilometer-scalerelief. To examine whether a plate 8-10 km thick at the ridge axis that thickensby severalkilometerswithin 15 km of the axis is plausibleat a slow spreadingridge, thermal modelsthat includeboth the heat of magmaticcrustalaccretionon the ridge axis and upwellingmantle flow beneaththe axis are formulated. Hydrothermalcoolingthat increasesthe effective conductivityby about a factor of 10 is required for an 8-km-thick plate to exist at a slow spreadingridge axis. However, even with this enhancedhydrothermalcooling, the plate thicknessat a fast spreadingridge axis and the resultinghorizontalforce in the plate are too small to result in appreciabletopography. While mid-oceanridge axial topographyis stronglyinfluencedby spreadingrate, other factors also shapeaxial structure. For example,the anomalouslyshallow slow spreadingReykjanesRidge has an axial structuremorelike that of a fast spreadingridge, and the anomalouslydeepintermediatespreading Australian-Antarctic Discordancehas a medianvalley axial structuretypical of a slower spreadingridge. In addition, within a given spreadingsegmentthere is often a systematicdeepeningof the ridge axis toward transform offsets. These non-spreading-rate-dependent axial topographicvariations may be producedby variationsin magmaticheat input associatedwith crustalemplacementwhich shapesthe thermalstructureand lithospherethicknessnear the ridge axis. INTRODUCHON slow spreadingrates but vanishesat faster spreadingrates. Since seafloor bathymetry is the most frequently acquired The well known difference in the axial topography of and widely available geophysicaldata, it is also important slow and fast spreadingmid-oceanridges is shownin Figure to understandthe possibleimplicationsof ridge axis topog1 by typical topographicprofiles across the axes of slow, raphy for the thermal and mechanicalstructureof accreting medium, and fast spreadingridges. The characteristic1-kinplate boundaries. deep axial valley and flanking crestalmountainsthat form at In the few areas where data of sufficient quality and slow spreadingrates do not exist at faster spreadingrates, resolutionhave been obtained,gravity anomaliesprovide an where instead relatively low relief axial topography is obimportant observationalconstrainton the origin of the axserved. Although this spreadingrate dependenceof ridge ial valley and particularly the possiblerole of crustal thickaxis morphology is widely recognized, there still are no ness variations in producing the axial valley topography. conclusive explanations of why an axial valley exists at Gravity data for a segmentalong the Mid-Atlantic Ridge from about 13ø to 15øN has been reportedby Collette et al. Copyright1987 by the AmericanGeophysicalUnion. [1980]. The amplitudeof the observedgravity anomaly showsthat the ridge axis topographycannotbe isostatically Paper number7B5032. 0148-0227/87/007B-5032505.00 compensatedat a depth as shallow as 6-8 km and so cannot 12,823 12,824 PHIPPSMORGANMr AL: THE ORIGINOFMOR AXIAL STRUCRJRE PB VV PB VV i PB FAST ( EPR3 øS) i PB INTERMEDIATE ( EPR 21øN ) PB I v MAR 37 ø N ) VE ,.-, 4x AXIS I , 2:0 I I0 , I , I 0 KM , I0 I , 20 I 30 Macdonald (1982) Fig. 1. Axial topographyalong profilesperpendicularto typical slow (5-25 mm/yr), intermediate(25-45 mm/yr), and fast (>45 mm/yr) spreadingridge axesfrom Macdonald [1982]. A 1- to 2-km-deep,30-kin-wide axial valley with flanking crestalhighs is presentat slow spreadingridges.At fasterspreadingratesthe axial topographyis much more subdued.The symbolsV and PB on the profiles delineatethe width of the regionof active volcanismalong the spreadingcenterand the boundariesof the region of active faulting, respectively. EPR is East Pacific Rise; and MAR, Mid-Atlantic Ridge. The profileshave a vertical exaggeration(VE) of 4. be explained by crustal thicknessvariations [Collette et al., 1980; Parmentier and Forsyth, 1985]. Furthermore,seismic refraction studies along the Mid-Atlantic Ridge [Purdy and Detrick, 1986; Louden et al., 1987] do not show abnormally thin crust beneath the axial valley. Thus, both gravity and seismic refraction studies indicate that the crust within the prescribedso that the horizontally moving plate thickens rapidly with distancefrom the ridge axis, forming a conduit at the axis with a width that is comparableto the width of an axial valley. Flow in a mantle half-space of uniform viscosity occurs due to both the horizontal motion and accretion of the plate. The steady-state thermal structure based on these velocities is shown by isothermsat specified fractionsof the deep mantle temperatureT,n. The shapeof the bottom of the plate clearly differs greatly from that of the isotherms.While the localized upwelling at a ridge axis due to the presenceof a conduit transportsheat into the region beneath the ridge axis, cooling within the rigid plate is due primarily to vertical heat conduction,resulting in the shallow, genfiy sloping isothermsshown in Figure 2. The release of latent heat beneath the ridge axis would raise the temperatureand elevate isothermsto shallower depths. The axial valley is essentially the same thickness as that of older seafloor. Therefore, the axial valley must be a dynamically maintained, stress-supportedtopographicfeature. In this paper we examine simple models for the origin of ridge axis topography. While a number of previousmodels have been presented,these models do not generally provide a clear indication of the origin of vertical forces responsible for the large-amplitude topography at slow spreadingridge axes. To provide a framework and motivation for our discussion of simple models, we begin by reviewing previous models for slow spreadingridge structure. isotherms would, however, maintain their gently sloping character since this is controlled by vertical heat conducPREVIOUS MODEI3 FOR AN AXIAL VAT J.FY tion. Sleep [1969] and Lachenbruch [1973, 1976] have suggestedthat the structureof a slow spreadingridge axis consists of a vertical conduit through which asthenosphererises to form a thick plate near the ridge axis. This is presently the only model which explicitly addressesthe origin of vertical forces producing the axial valley morphology. In this model, the viscous pressure drop associatedwith vertical flow in the conduit creates the axial valley depression. Vertical forces resulting from shear stresseson the conduit walls support the elevated flanking crestal mountains. The principal difficulty of the steady-state conduit model is to explain the existenceof a conduit in terms of thermal structure and reasonable rheological behavior. Specifically, if the lithosphere forms by conductive cooling and the strength or viscosity of mantle rock is primarily a function of temperature, then a conduit of the type envisioned by these studies should not form. This is shown by the example in Figure 2. The methodsused to obtain this result will be discussed later . In this example,the plate thicknessis The above results indicate that no conduit would be presentif the viscosity of mantle rock dependsprimarily on temperature. It has frequently been suggestedthat a conduit could form due to a weakeningeffect of partial melt or water. However, studiesof magma migration suggestthat even small amounts of melt forming a connected network can rapidly migrate out of a solid matrix. Studies of the strength of olivine containing basalt melt [of. Cooper and Kohlstedt, 1986] suggestthat the principal effect of small amountsof melt will be to enhance deformation related to grain boundary diffusion. For deformation that occurs by a combinationof dislocationcreep and grain,boundary diffusion, the presence of a few percentmelt enhancesthe deformationrate only by a factor of 2-5 [Cooper and Kohlstedt, 1986]. A recent study [Karato, 1986] suggeststhat the presenceof a small amount of melt may actually strengthen partially molten rock becausewater, which weakens mineral grains, is partitioned into the melt. Therefore while the effect of partial melting on the strength of rock is not fully resolved, our PttIP•MORGAN ETAL:TIlEORIGIN OFMORAXIAL STRUCTURE .3 .5 .7 .9Tm 50 km Fig. 2. Thermal and mechanicalstructureof a slow spreadingmidocean ridge with a prescribedconduitlikemechanicalstructure. The lithosphere (dot shaded) moves horizontally and thickens rapidly near the ridge axis forming a conduit. Streamlinesshow viscous flow beneaththe plate due to plate motion and accretion assumingthe mantle is a uniform viscosity half-space. The thermal structure corresponding to this velocity field is shown by isotherms at specifiedfractionsof the deepmantletemperatureTm. The bottomof the lithosphereclearly does not correspondto an isotherm, thus a steady-stateridge axis conduit cannot be present if the strength or viscosity is solely a function of temperature. current understandingindicatesthat small amountsof partial melt will not reduce the strength by the several orders of magnitudethat would be neededto form a ridge axis conduit. Compositional change resulting from melt extraction will raise the solidus temperatureand therefore the viscosity of the residualmantle rock. The viscositydue to thermally ac- 12,825 sated by elastic flexure of the lithosphere producing the flanking uplift of the crestal mountains. This is essentially the Vening Meinesz [1950] model for rifts. In its application to mid-ocean ridges, the model does not explain the amplitude and width of the ridge axis depressionin terms of other physical variables, such as extension rate and plate thickness. Furthermore,a variety of evidencesuggeststhat the lithosphere deforms by faulting near the ridge axis. Studies of seafloor structure [Macdonald and Luyendyk, 1977; Macdonald and Atwater, 1978; Laughton and Searle, 1979; Macdonald, 1982] have concludedthat axial valley relief is a consequenceof faulting. Studies along the MidAtlantic Ridge south of the Kane fracture zone [Toomey et al., 1985] show that seismicityextendsto depthsof about 8 km and that microearthquakesoccur beneath both the axial valley floor and the crestal mountains. Deformation due to faulting is not well describedby simple elastic flexure. The numerical models of Sleep and RosendaM [1979] were the first to include the effects of a viscoplasticplate and quantitatively predict the formation of an axial valley. However, the geologiclike complexity of the assumed thermal and theological structure combined with limited numericalresolutiondoes not permit a simple understanding of the mechanism(s)causing the calculated axial valley morphology. Our approachhas been to examine simple models which isolate potentially important physical mechanismscausing seafloor topography near a ridge axis. We will show that horizontal extensionalstressesin a strong continuouslydeforming plate near the ridge axis can produce an axial valley-like ridge axis morphology. Stressesdue to mantle flow and melt segregationbeneath an extendingplate can produce a similar morphology,but a mantle viscosity significantly higher than generally acceptedvalues would be required to explain the observed axial valley relief. TOPOGRAPHY DUE TO MANTLE FLOW STRESSES tivatedcreepis proportional to exp(ATm/T ) whereA ~ 40 If lithosphericstretchingis occurringnear the ridge axis, and T m is the melting temperature[e.g., Brace and Kohlstedt,1980]. Thus,if meltingincreases Tm by 100øC then either vertical mantle flow or magma intrusion into the as suggestedfrom petrologic studies,this would increase the viscosity by approximatelytwo ordersof magnitude. While this effect has not been previously considered,it is potentially important and could result in the formation of the strong lithosphere forming conduit walls [cf. Lachenbruch, 1973]. Alternatively, if the compositionchanges affect the solid fraction of conduit manfie and the presence of melt does not significantly reduce the viscosity, this effect could actually increasethe viscosity of material in the conduit. It is also important to question whether a conduitlike thermal structure may result from hydrothermal cooling. This questionwill be explored in more detail later, but hydrothermal cooling that is simply describedby an enhanced thermal diffusivity will produceeven flatter isothermsin the mantle beneath the ridge axis. The simple example in Figure 2 does not prove that a steady-stateconduit cannot exist, but shows that special assumptionsabout mantle theology or hydrothermalcooling are required. Tapponier and Francheteau [1978] first suggestedsteadystate necking as a conceptualmodel for slow spreadingridge axes. Their model consists of two parts: first, that extension of the plate creates a seafloor depressionat the ridge axis and, second,that this depressionis regionally compen- stretching lithosphere is required to balance lithospheric thinning due to stretching and thus maintain a steady-state lithosphere thickness distribution. Vertical manfie flow at the base of a stretchingand accretinglithosphereis capable of producingstressesthat can supportridge axis topography [Phipps Morgan and Parmentier, 1984]. However, in the following analysiswe will show that mantle flow can create a significant axial valley depressiononly if mantle viscosities beneath the ridge axis are at least two orders of magni- tudelarger(1020Pas or 1021P) thancommonly accepted values(1018Pa s or 1019P). Mantleflow will be induced by both vertical and horizontal motion at the bottom of the lithosphere and by melt segregation. We will first consider mantle flow induced by motion at the bottom of the lithosphere and discussthe effects of melt segregationon mantle flow in the following section. Considerflow in a mantle half-spaceof uniform viscosity beneath a planar, horizontal lithosphere-asthenosphere boundary. Flow in the half-space induced •'Oyvertical and horizontal flow along the top, z=0, can be representedas a superposition of Fourier harmonics in the horizontal x coordinate. For each Fourier component of horizontal or vertical velocity on the top of the mantle half-space, u or w 12,826 PttIPPS MORGAN ETAL:THEORIGIN OFMORAXIAL STRUCTURE ' U I ' • )•I I•_•+ I u(x dx) Velocities: u/U 1 I L,J w 0 0 x/L I 2 I I 4 Topography: 3 •UH/ApgL2 Fig. 3. Axial topography dueto verticalmantleflow at the baseof extending lithosphere.(Top)Modelin which extending lithosphere thickens slowlywithdistance fromtheridgeaxis.Thethickness variation shownis exaggerated for clarity.(Middle)Verticalmantlefloww across theapproximately planarbaseof thelithosphere (dashed line)canbe found by considering flow into the baseof a stretching columnof lithosphere with horizontal velocityu(x) thatis neededfor steaidy-state materialconservation. The assumed errorfunctionplatespreading distribution, corresponding to a Gaussian upwellingdistribution, produces the axial topography shown.(Bottom).Here •5 is the topography, •t is themantle viscosity, U is the half-spreading rate,H is the platethickness in the stretching region,L is the half-widthof the stretching region,Ap= Pm-Pw is thedifference between mantleandwaterdensities, andg is theacceleration of gravity. respectively, the induced normal and shear stresseson the viscosity [x=1021 P (1020Pa s), an axialvalleyhalf-width top of the half-spaceare •Yxz= -2•tku and•Yzz= -21.tkw,where k is the wavenumber.The stressesresultingfrom any prescribeddistributionof horizontal or vertical velocity can be obtainedby the superpositionof Fourier components.Thus any horizontal distributionof velocity (w=0) on the top of the mantle half-space will produce flow-induced shear L=15 kin, a ridge axis lithospherethicknessH=8 km, and a half spreadingrate U=20 mm/yr (20 km/m.y.), this mechanism predictsa 1-kin-deepstress-supported axial valley. Smaller mantle viscosity, smaller lithosphericthickness near the ridge axis, and smaller amountsof lithospheric stretching wouldall reducethe amplitudeof the topography. Note that in calculatingthe seafloortopography,mantle flow inducednormalstressesare assumedto be supported simple result contradictsthe intuition that the diverging only by buoyancyforces arising from topographyat the flow due to plate spreadingwill producean axial valley hole crust-waterboundarywith no contributionfrom the strength at the ridge axis [cf. Nelson, 1981]. of the lithosphere.This will overestimate the topography if For a lithosphere-asthenosphere boundarythat is nearly someof the verticalforceat the bottomof the plateis supbut not exactlypl•anar,mantleflow into a stretchingsteady- ported by stresseswithin the plate. To consider lithosphere thicknessvariations in the more statelithospherecan be approximatedas shownin Figure3. The net upwar d flow is simplythe difference in thehorizon- general case of a highly nonplanar lithosphere-asthenotal flow enteringand leaving a thin vertical slice of height sphereboundaryandto assess the accuracy of the aboveapH, sowI z=0= HOu/Ox, whereH is thelithosphere thicknessproximationof a nearly planarboundary,boundaryelement stresses at the bottom of the lithosphere but no vertical normal stressesthat can support seafloor topography.This outsidethe axial region of stretchingand accretion. With a prescribed spreading velocity u=Uerf(x/L), the calculated normal-stress-supported topographyis shown in Figure 3. Stressesresultingfrom the prescribedvertical velocity were calculatedusinga standardfast Fouriertransform(FFT) algorithm with 256 uniformly distributedpoints. For a mantle solutions for the mantle flow induced normal stress distribu- tion were obtainedfor the severalplate thicknessdistributionsshownin Figures4a-4d. For this problem,the boundary elementmethod(BEM) has two basic advantages over other methods:(1) for arbitrarily shapedtwo-dimensional regions it is a simpler task to discretize a one-dimensional PftIPPSMORGAN ET AL: • ORIGIN OF MOR AXIAL S'IRUCTURE 12,827 1 Spreading Velocity: u/U 1 1 Topography: •UH/A0gL 2 •UH/A0gL 2 -3 -3 Flowlines: Fig. 4a Spreading Velocity: Fig. 4b u/U u/U 1 lO Topography: ]JUH/A0gL2 -30 -3 Flowlines: Fig. 4c Fig. 4d Fig.4. Mantleflowandaxialtopography for several prescribed distributions of lithosphere (dotshaded) thickness and horizontal extension rate.Spreading velocity distribution andcorresponding topography areshown at thetopof each figure. Streamlines of themantleflowareshownat thebottomof eachfigure. Symbols andnotation arethesameasin Figure 3. (a) Constant thickness lithosphere thatis emplaced entirely at theridgeaxisandspreads witha uniform velocity. (b) Constant thickness, stretching andaccreting lithosphere equivalent to themodelof Figure 3. (c) Slowly thickening stretching andaccreting lithosphere. (d)Ridgeaxisconduit withcorresponding thermalstructure shownin Figure 2. 12,828 PHIPPSMORGAN ET AL: THE ORIGIN OF MOR AXIAL STRU• boundary than the whole two-dimensional region, and (2) there are no numerical complicationsneeded to enforce the incompressibilitycondition. The BEM is an extremely convenient technique for solving problems in a piecewise homogeneousmedium where an exact Green's function solution is known. In the BEM, this singular solution is integrated to find the solution for a particular Green's function distribution over a line segment; for example a constant, linear, or quadratic distribution could be employed. The boundaryis discretizedinto a f'mite numberof boundaryline segments or elements, and influence coefficients are calculated that relate the contribution to the stresses and veloci- ties on one element from given stressesor velocitiesacting along another element. In the present application of the BEM, each boundaryelement is a line segmentwith a uniform distribution of normal and tangential force acting along it. This results in unknown normal and tangential force magnitudesfor each element. The velocity or stress distribution along the boundary is prescribed,and with velocities evaluated at the center of each element, the distribu- tion of boundaryforces which will producethis prescribed boundary velocity and stress distribution at each element midpoint is calculatedby solving a linear systemof equations for the unknownmagnitudesof the boundaryforce distributionfor each boundaryelement. The known boundary force distribution can then be used to calculate stretchingw(x)= H(x)3u/3x. (2) Alongthe ridgeaxis(x=0) a symmetry condition is assumed(u=0, Xxz=O). Computationally, elementson the right-hand side of the ridge axis which mirror thoseon the left-handside of the ridge axis are constructed. Becauseof this symmetry only unknownsfor the right-handside of the ridge axis need be computed,and no boundaryelementsare neededon the vertical boundary directly beneaththe axis. (3) Along the bottom of the box, stress-freeboundariesare assumed. (4) At the right-hand side of the box (x=L) the horizontal and vertical velocities are prescribedas the solutionfor a planar half-spacewith a uniform horizontal velocity along the top which is at a depth H(L). At least 50 elementsalongthe top and 40 elementsalong the fight side and bottomwere used. Boundary solutions will H*=lw(x)H(x)a/lw(x)a be least accurate near the boundaries defined. Thecalculated ridge axis topographyfor a slowly varying thickness H=H(x) and w=HOu/Ox in Figure4c is almostidenticalin both form and amplitude to that calculated with a constant lithosphere thicknessH=H* andw=H*Bu/•x (Figure3 or 4b). The componentof mantle deformationthat producesaxial valley-like topographycan be seen in the mantle streamlines of Figure 4. The streamlinesbeneaththe ridge axis convergeas they near the base of the lithosphere,indicating a vertical stretchingof the mantle beneath the ridge axis. The corresponding tensional vertical deviatoric normal stresssupportsthe axial valley depression. For large variations in lithosphere thickness near the ridge axis, as in the case of a ridge axis conduit, the approximations of a nearly planar lithosphere-asthenosphere boundarybreak down. Boundary element solutionscan still be obtainedfor thesecases.The examplein Figure 4d shows that the presenceof a conduitlocalizingupwardmantleflow at the ridge axis greatly increasesthe predictedridge axis topography.The increaseover that in Figure 4c is by about a factor of 10 for this particular lithospherethicknessdistribution. the stresses or velocities at any point within the region. The program (TWOFS) used in the presentstudy is describedby Crouch and Starfield [1983], who also presenta more complete developmentof the BEM. The following boundary conditions are prescribed. (1) Along the lithosphere-asthenosphereboundary, velocities are set to the horizontal plate velocity u(x) and vertical velocity due to element sphere-asthenosphere boundary. To comparethe two results, an average lithospherethicknessin the stretchingregion TOPOG•HY DUE TO 1VIELTSF_.I3REGATION The steady state segregation of melt fi'om mantle matrix and the resulting matrix compactiondue to the loss of melt volume will create an effective volume sink within the man- tle. The mantle flow to fill in this mass deficit in the segregating and compactingregion can produceaxial topography. Melt segregationand associatedmantle compactionis idealized as occurring at a single line sink beneath the ridge axis. The line sink may be regardedas a small compacting region of cylindrical cross section. A more realistic segregating and compactingregion can be constructedby superimposing•these line sink solutions. The normal-stress-supported topographydue to a line sink at a depth D beneath the surface of a viscous halfspace, in this case the crustmantle boundaryat the ridge axis, can be found by addinga line sink solution at a depth D below the surface to an image sink a distanceD above the surface. The resulting normal-stress-supported topographyis shown in Figure 5. For a melt removal rate equal to the crustalproductionrate 2UH at the ridge axis, where H is the crustal thicknessand U is the half-spreadingrate, the maximum axial valley depression is approximately gUH/ApgD 2, where Ap--Prn-Pw, }xisthe of the region, where the assumeddiscontinuousforce distribution is an imperfect approximationto continuousvelocity or stressboundary conditions. However, accordingto Saint Venant's principle, solutionswill become increasinglyaccurate away from the boundaries of the region. From the agreement between boundary element and semi-analytical FFT solutionsfor the planar half-spaceproblem (Figures 3 and 4b), stressesat the element midpointsalong the lithosphere-asthenosphere boundaryare expectedto be accurateto better than 5% for all the lithospherethicknessdistributions shown in Figure 4. viscosity of the mantle matrix, g is the acceleration of gravity, and D is the depth to the segregatingregion. This solution, which is formed by superimposingtwo line sinks [cf. Batchelor, 1967, p. 89], produces the correct normal stressbut satisfies a shear stress-freeboundary condition instead of a no horizontal velocity boundary condition at the surfaceof the half-space.To solve for the contribution to mantle flow from melt segregation beneath a spreading center, a complementary solution to cancel the horizontal flow generated at the crust-mantle boundary by Figure 4c showsthe mantle flow and associatedaxial to- This complementary solution can be obtained by the complex-variable techniques developed, for example, by Muskhelishvili [1953, p. 584]. Since it involves only horizontal flow at the surface of the half-space, this comple- pography due to stretching and accretion of a variable thicknesslithosphere.This is almost identical with the earlier result of Figure 3 which assumesa nearly planar litho- the two line sink solution must be added to this solution. PHIPPS MORGAN ETAL:TIlEORKEN OFMORAXIALSTRUCRJRE Topography: x/L 12,829 tered 15 km beneath the crust-mantle boundary would produce a 30-km-wide axial valley. For a mantle viscosity of 1020 Pa s, a 1-kin-deep axial valleywill resultfrom an along-axis crustalproduction rate rn = 2UH = 230m2/yr correspondingto U = 20 mm/yr and H = 6 km. Melt segregation occurringat greater depthswill producea broader and shallower axial valley for otherwise identical model parameters. For a similar mantle viscosity, depth of melt segregation, and crustal thickness,ridge axis topographyassociated with melt segregationshould be greater on faster spreading ridges. If crustal thickness and mantle viscosity are approximatelythe same on fast and slow spreadingridges, a larger-amplitudetopographyat slower spreadingrates could be explained by shallower or more localized melt segregation. However, melt segregation can contribute significantly to axial topography only if there is a strong localization of melt segregation and the asthenosphereis moreviscous(1020Pa s) thanit is usuallythought to be (1018Pa s). Streamlines in Figure5 showtheflow in a mantle half-space due to melt segregationin a small cylindrical region 30 km beneaththe crust-mantleboundarycombined with flow due to plate spreading. The plate spreading flow is described by vertical and horizontal velocities u= 1 2 2 2U[tan(x/z)-xz/(x +z2 )]/n andw=-2Uz2 /[n(x2+z )], respec- tively, with streamlinesshownin Figure 4a. Note that flow induced by melt segregation and associated mantle comFig. 5. (Top) Axial topographydue to mantle flow into a small paction tends to localize the region of mantle upwelling becylindrical region (line sink) of melt segregation.Here•5 is the neath the ridge axis as shown by comparing the streamlines topography,I• is the mantleviscosity,U is the half-spreading rate,H in Figure 5 with thosein Figure 4a. is the crustalthickness, L is the depthto the segregating region, andAp=pm-Pw. More complicated andrealisticdistributions of melt RtDGE AXiS TOPOGIUa•e DUE TO PIATE EXTENSIONAL STRESSES segregation could be constructedby superimposingline sink solutions. (Bottom) Streamlines of mantle flow due to plate spreading and melt segregation with associatedmantlematrixcomRidge axis topographycan also be producedby horizontal paction. Analyticalsolutionfor manfieflow due to line sinkmelt extensional stresses in a strong plate that thickens with segregationis given in text. Note, by comparingwith Figure 4a, that melt segregationlocalizesupwellingbeneaththe ridge axis. distancefrom a ridge axis [Parmentieret al., 1985;Lin and Parmentier, 1986; Parmentier, 1987]. A perfectly plastic material is considered as a continuum description of mentarysolutiondoesnot contributeto the surfacenormal deformationdue to slip on many closelyspacedfaults. The stress. The mantle flow due to melt segregationin a line strengthof the plate in this case arisesfrom the frictional sink beneatha spreadingcenteris found by addingthe two resistanceto slip on the faults. Simple models to examine line sink and complementarysolutions u+iw=•{2t.Z 2+ +Z2 + •'2+iz[z2 + +•'{•X 2+(Z-Zl) =+X2 +(Z+Zl) 2] Zl + +Z+Z l } +•{X 2+Z(z-z)2 m 1 x x where u and w are velocities in the x (horizontal) and z (vertical) directions, respectively, and rn is the melt the form and amplitude of ridge axis topographywill consider a prescribedplate thicknessvariation. With the bottom of the strongplate definedby a prescribedtemperature, the plate thicknessvariation will be controlledby thickening of the plate due to coolingaway from the ridge axis and by localizedthinningof the plate near the ridge axis due to plate extension. As discussedlater, plate extensionof the magnitudeinferred from seafloor faulting and seismicmoment release will make a smaller contribution than cooling to plate thickeningaway from the axis. Thermal modelsto examine plate thicknessvariation are describedin a later segregationrate into a line sink a depth zI beneath the section. The mechanismproducingtopographyas a variable thicksurface of the half-space or crust-mantleboundary. The complexvariables•=x+iz and •= x-iz, wherex is the real nessplate extendscan be understoodqualitativelyfrom Figand z is the imaginarypart of the complex variable, have ure 6a. Because the plate thicknessvaries, the horizontal force F within the plate acts at different depths on the two vertical sides of a material column, thus creating a moment melt compactionmay be substantialif melt removaloccurs M. This moment, distributed horizontally over the region at shallowdepthbeneaththe ridge axis and if ridge axis as- where the thickness varies, bends the plate. The vertical been used to conciselyexpressthe solution. The normalstress-supported axial topographyproducedby thenosphere viscosities areontheorderof 1020Pas (1021 force due to the resulting seafloor topographymust balance P). Since, as shownin Figure 5, the depth to the segregat- the appliedmomentand thushas the form shown:a central with flankinguplifts. Sincethereis no net vering region is the half width of the resultingtopographic depression feature, melt removal occurringin a cylindrical region cen- tical force acting on the plate, only moments, the topogra- 12,830 PItlP• MORGAN ETAL:THEORIGINOFMORAXIALSTRUC'RJRE (a) (b) O(x) O(x) - dh Oxz-Oxxd-• Fig. 6. (a) Force(F) and moment(M) distributioncan produceaxial topography by flexuraldeformationof a strong viscousor plasticplateof variablethickness thatis beingpulledapartby externalforces.Platethickness variations result in changes in depthof the horizontalforce,thuscreatingmoments withinthe platethatmustbe balanced by moments due to topographic forcesresultingfrom flexuraldeformation.(b) For a strongplatewith slowlyvaryingthickness, stresses due to changesin plate thicknesscan be represented by an effectiveshearstress(•xz actingon the baseof the plate. Considerthe horizontalforce balanceon the darkly shadedtriangularmaterialelement: the horizontalforce Oxxdh acting on the vertical side of this triangle must be balancedby the force (•xz dx acting along the horizontal side. Becausethe mantle is assumedto be a very weak fluid, no horizontalforcesact on the hypotenuseof the triangle. This force balance showsthat the inducedshearstress(•xz is givenby the productof the meantensilestress(•xx and the rate of changein platethickness i•h/i•x. The shearstress Oxz produces internaldeformation shownby the schematic streamline.Seetextfor further discussion. phy averagedacrossthe ridge axis must vanish. Deforma- to be small comparedto the plate thicknessso that the intion of the plate in responseto the applied momentscould ducedshearstresscan be appliedat the bottomof a layer of be approximated as flexure of a viscous or plastic plate. uniform thickness H. This approximation has been freHowever, since the half width of an axial valley is not large quently applied in theoreticalstudiesof folding and boudicomparedto the plate thickness,a more accuratedescription nage. The strengthof the plate and thereforeC•xxis alsoasof the topographycan be obtained by consideringtwo di- sumed to be independentof depth within the plate. The mensional deformation or flow within the plate rather than boundary conditionsat the bottom of the layer (z=H) are (Sxz=(Sxx •h/•x and(Szz=0,andthe boundaryconditionsat the flexural approximations. and As shown in Figure 6b, considera strongplate of per- top of the layer (z=0) are (Sxz=(Sxx•J/•x,(Szz=Apg•J, fectly plasticmaterial that restson a very weak fluid mantle w=•(u•J)/•x. The last equationstatesthat the surfacez=•J(x) substrate. The presenceof a horizontal applied stress C•xx is a particle path with material at the top of the plate movthat causesextensionof the plate resultsin an induced shear ing along this surface. The perturbationflow due to the stressC•xzthat is requiredto maintain equilibrium of the shear stress{Jxz at tl•.•::bottom of the layer can be represmall, darkly shadedmaterial element at the bottom of the sented as a superpositionof Fourier harmonicsin the x dilayer. This inducedshearstresscausesflow within the layer rection. The velocity distribution associatedwith any horiwhich is superimposedon the flow due to horizontalexten- zontal wavenumber is given by solution of equations(10) sion and which results in a central depressionand flanking and (11) of Zuber and Parrnentier [1986]. Figure 7 showsthe topographyresultingfrom a prescribed uplifts. When the plastic plate is kept at yield by the apvariation of theformh exp[-(x/L) 2] where plied stressC•xx,stressesdue to gravitational attraction on platethickness the topography at the top surface of the layer also cause the width of the plate thicknessvariation L = 0.5H, H, and flow which opposesthat of the inducedshearstress. For a 2H. The surfacetopographyand velocity distributionwithin prescribedplate thicknessvariationh(x), the equilibriumto- the layer which satisfiesthe above boundaryconditionswere pographywill be that for which the verticalvelocity at the calculatedusing a standardFFT algorithm with 256 points uniformly distributed over the length of the examples upper surfacevanishes. Flow within the plate is describedby the equilibrium shown. The amplitude of the relief is proportional to the equationsand the incompressibilitycondition. To obtain magnitude of the layer thicknessvariation, h, and depends parameterS=(Sxx/ApgH,where is analyticalsolutions,the thicknessvariation h(x) is assumed on a single dimensionless PHIP• MORGAN ET AL: TIlE ORIGIN OF MOR AXIAL STRUCIIJRE 12,831 Fig. 7. Axial topography due to defomaation resultingfrom horizontalextensional stresses in a lithosphere that thickens with distancefrom a ridge axis. The prescribed thicknessvariationis shownat the bottomof eachfigure and the resultingtopography alongthe top. All casesconsidera perfectlyplasticlithosphere with the samevalueof the strength parameterS definedin the text.Differentlithosphere thickness variationsare considered by varyingthe half-widthof the thickeningregion L =0.5H, H, and 2H. the mean horizontal applied stress,Ap = P,n- Pw,g is the to producethe 1 km of seafloorrelief typically observedon accelerationof gravity, and H is the lithospherethickness slow spreadingridge axes. near the ridge axis. Similarresultsfor a rangeof S values showsthat the amplitudeof the relief is approximatelyproRIDGE AXIS THERMAL MODELS portionalto S and inverselyproportionalto L. If the width As an extending plate thickens with distance from the of the plate thicknessvariation is comparableto or smaller of the thickeningof than the plate thickness,the predicted width of the axial ridge axis, the dynamic consequences valley, measuredfrom flanking high to flanking high, is 3- the deformingplate may createor at least contributesignifi4 times the plate thickness. A wider plate thicknessvaria- cantly to the axial morphologyof slow spreadingridges. It tion producesa broaderaxial valley, but the predictedwidth is important to emphasizethat the total amount of extenof the axial valley is not less than about four plate thick- sion that accumulatesin a thin vertical section of plate as it nesses. With a brittle plate thicknessof about 8 kin, as in- moves through the axial valley is relatively small. Based ferred from the depth of microseismicityon the axis of the on observeddistributionsand heights of fault scarps,MacMid-Atlantic Ridge southof the Kane fracturezone [Toomey donald and Luyendyk [1977] and Macdonald and Atwater et al., 1985], the predictedflank to flank width of the axial [1978] estimate only 10-20% extension. Similarly, the avvalley would be about 30 kin, which agreeswell with the erage seismic mo•aent release along slow spreadingridges, observedaxial valley width along this segmentof the ridge. as reported by Solomon et al. [1986], can be interpreted to Frictional sliding on faults will result in a yield stress indicate that only 5-10% of the total spreadingvelocity is that increaseslinearly with depth [cf. Brace and Kohlstedt, due to seismicfault slip. As seen at the seafloor, most of 1980]. For faults containinghydrostaticwater pressureand the plate spreadingvelocity developswithin a 1- to 2-kinon which frictional sliding occurs with a laboratory-mea- wide neovolcanic zone at the ridge axis where spreadingis sured friction coefficientof 0.85 [Byerlee, 1978], the depth- accommodatedby dike intrusion. The spreading velocity averagedvalue of Oxxis S=0.4. The absenceof water pres- increasesonly slightly away from the axis due to horizontal sure resultsin the larger value S=0.7. For S=0.4 resultslike extensionof the lithosphere. While the amplitude of the those in Figure 7 predict seafloortopographywith a relief topography depends on the magnitude of plate thickness of about 30% of the plate thicknessvariation. Therefore a variation near the ridge axis, the plate thickness variation plate thicknessvariation less than about 3 km is sufficient cannotbe solely a result of plate extension. Extensionwill 12,832 PHIPPS MORCaAN Er AL:TIIEORIGIN OFMORAXIALSTRUC'ruRE tend to thin the plate, but if the above estimatesof exten- to 3-kin-deep seismicreflectionsinterpretedto be the top of a magma chamber. In the following models, to examine the effects of hydrothermalcooling in a simple way, heat transfer due to hydrothermalconvectionis representedas an increasedthermal conductivity. The ordinary thermal conductivityis enhanced by a factorNu, the Nusseltnumberfor hydrothermalconvecwhich the thermal structure for a prescribedvelocity field tion in a permeable layer. Rock at a temperature greater was determined[e.g. Sleep, 1969, 1975; Kuznir, 1980;Reid than 400øC or at a depth greater than 6 km is assumedto be and Jackson,1981]. The modelsof Sleep [1969, 1975] as- impermeablebecauseit cannotsupportopen fractures. Since sumepurelyhorizontalvelocitiesequalto the half spreading water as hot as 400øC dischargesfrom vents on the seafloor, rate and include heat resulting from the solidification and this a minimum value of the maximum temperatureof rock sion are correct, this effect must be relatively small. If the bottomof the strong,brittle plate is definedby a prescribed temperature, the thermalstructuredue to mantleflow beneath the ridge axis and crustalaccretionon the axis will control the plate thicknessvariation. A numberof ridge thermalmodelshave been presentedin cooling of magmatically emplacedcrust on the ridge axis. throughwhich water must circulate. Studiesof the Samall Ophiolite [Gregory and Taylor, 1981] indicate that subsolidus oxygen isotope exchange with seawater occurred viscosityhalf-spaceas shownin Figure 4a. This modelthus primarily within the upper 5-6 km of the ophiolite, thus includes the effect of vertical mantle flow beneath the ridge giving an estimateof the maximum depth of water penetraaxis but ignoresthe heat resultingfrom crustal emplacement tion. on the ridge axis. Since both effects are likely to be The thermal models in Figures 8c and 8d were calculated important,the two modelshave simplybeencombined.The for Nu=3 and 10, respectively,with the same slow spreading velocity field consideredis shown by the streamlinesin rate as in Figure 8b. With Nu=3 the lithosphere,defined by Figure 8a. Crustemplacedon the ridge axis moveshorizon- the 600øC isotherm, at the ridge axis remains thinner than tally at the half spreadingrate, and underlyingmantle flows the crust. However, with Nu=10, the lithosphere is clearly like a uniform viscosity half-spacebeneaththe spreading thicker than the crust and generally consistentwith depth of crust. Heat associated with solidification and cooling of seismicity observed at the Mid-Atlantic Ridge. The plate crust on the axis is treated as a concentrated axial heat thickensby about 3 km over a distanceof 10 km from the source distributed over the thickness of the crust, in a manridge axis. This thickeningof the plate with distancefrom ner analogous to that of Sleep [1975]. The crustalthickness the ridge axis is sufficient for plate extensionto contribute (6 km) anddeepmantletemperature(1200øC) are prescribed. significantly to ridge axis topographyon slow spreading The temperature of magmaticallyemplacedmaterialis taken ridges. For a faster half spreadingrate of 50 mm/yr and to be greaterthan the deep mantle temperatureby an amount Nu=10, the lithosphere at the ridge axis remains much (320øC)representing the heat of fusionof moltenbasalt(80 thinner than the crust as shown in Figure 8e. Reid and Jackson [1981] calculated the temperaturedistribution for mantle flow due to plate spreadingon a uniform cal/g). The steadystate thermalstructurewas calculatedusingfinite differenceapproximations with the distributionof grid points shownin Figure 8a. The isothermsof the resulting thermal structurefor a half spreadingrate of 10 mm/yr are shownin Figure 8b. As inferredfrom the depthof oceanic intraplate seismicity [Bergmanand Solomon, 1984; Weins and Stein, 1984], the bottom of the plate shouldcorrespond to a 600ø-700øCisotherm. Even at this slow spreadingrate, the plate thicknesson the ridge axis is significantly less than the crustal thickness. Seismicity at 8-kin depths beneath the ridge axis [Tooracyet al., 1985] would not be expectedon the basisof this model. Hydrothermalconvectionof seawaterthroughpermeable oceanic crust is widely recognized as an important mechanism of heat transfer, particularly in young oceanic crust. Hydrothermalcooling is one obviousexplanationfor a narrower magmachamberon fast spreadingridges[Lewis and Garmany, 1982;McClain et al., 1985; Detrick et al., 1986] than would be expectedbased on thermal models [Sleep, 1975; Kuznir, 1980]. Several studies of hydrothermalconvection in a permeable crustal layer have been reported [Ribando et al., 1976; Fehn and Cathles, 1979; Fehn et al. 1983]. However,a recentstudyby Morton and Sleep[1985] appearsto be the only previousstudy which addressesthe effect of hydrothermalcooling on the large-scale thermal structureof a spreadingcenter. Representing hydrothermal cooling as a prescribeddistributionof heat sinks within the upper crust, the heat sink distribution was constrainedby requiring that the 1185øC isothermin the modelsmatch 2- Based on studies of the relationship between Rayleigh number and Nusselt number (cf. Combarnous, 1978, Figure 4), a Nusselt number of 10 for hydrothermalconvectionin a 6 kmthicklayerrequires a permeability of about 4x10 -16 m2, withothervalues of thermodynamic andtransport properties appropriatefor water and water-saturatedrock in the oceaniccrust. Permeabilityhas been inferred or measuredin several Deep Sea Drilling Project (DSDP) holes [Anderson and Zoback, 1982; Hickman et al., 1984]. However, the only in situ measurementof permeability at a depth greater than about 600 m in the oceanic crust was obtained at DSDP hole 504B in crust 6.5 m.y. old, severalhundredkilometers away from the axis of the Costa Rica Rift on the Galapagos SpreadingCenter [Andersonand Zoback, 1982]. The mea- sured permeability decreases withdepthto a valueof 10-17 m2 at depths greater than500m. Thislowvaluewouldnot account for hydrothermal convective heat transfer with a Nusselt number as large as 10. However, near the ridge axis where active faulting and cracking associatedwith rapid cooling occur [Lister, 1974], it is not unreasonableto expect larger permeabilities. Further from the axis, cooling cracks and inactive faults are likely to have been filled by hydrothermally depositedminerals. Thus, if hydrothermalcooling can accountfor a relatively thick plate near the axis of a slow spreadingridge, the results of the present study suggest that horizontal extensional stressesin a strong plate near the axis of a slow spreadingridge may explain the formation of the axial valley. The transport of magma through the strong mantle layer is a questionof obvious importancethat is not directly (b) (c) (d) (e) (f) Fig. 8. Mid-oceanridgethermalstructurein a verticalcrosssectionperpendicular to the spreading axis includingthe effectof hydrothermalcooling. Streamlinesof the flow fiold (Figure 8a) correspond to horizontalmotionof crust eraplaced on theridgeaxisandmantleflow dueto platespreading, like thatin Figure4a, beneath the crust. Grid spacing usedin finite differencecalculations is shownby tick marksalongthe sidesof the region. Isotherms(Figures8b-8f) are at temperatureintervalsof 100øC. Magmaticaccretionof the cruston the ridgeaxis is represented by a heat sourcethat extends to the bottom of the crust. (The base of the crust is shown by the solid horizontal line.) Lithosphere, corresponding to materialwitha temperature below600øC,is shaded.Hydrothermal coolingis represented asthe ordinary thermal conductivityenhancedby a constantfactorNu representingthe Nusseltnumberfor hydrothermalconvective cooling. Isothermsare shownfor a 10-mm/yrhalf-spreading rate ridgewith Nu=l (Figure8b), 3 (Figure8c), and 10 (Figure 8d) and a crustalthickness of 6 km. Isothermsare shownfor a 10-mm/yrhalf-spreading rate ridgewith Nu=10 (Figure8e) anda crustalthickness of 12 kin; andfor a 50-mm/yrhalf-spreading rateridgewith Nu=10 (Figure8f) anda crustal thickness of 6 km. See text for further discussion. 12,834 PHIPPS MORGAN ETAt.:TltEORIt'aNOFMOR AXIALSTRUCq'LIRE addressedby the simple models presentedabove. This may explain episodicity of the spreading process and possibly the large difference in frequency of episodicity inferred for fast and slow spreading ridges [cf. Macdonald, 1982]. Magma generatedin the upwelling mantle beneaththe ridge axis may episodicallypenetrate the strongplate at the ridge axis in cracklike bodies resulting in periods of reduced extensional stress in the plate near the ridge axis. As shown in Figure 1, volcanism at a ridge axis is restricted to a 1- to ity halfspace is approximately 4n•t/ApgL, where L is the widthof theaxialvalley. WithL=30 km andI.t=1020Pas (1021P), asrequired formantle flowstresses toproduce a 1kin-deep axialvalley,therelaxation timeis 6x104years. Clearly, the mechanismsproducing the episodiccharacter of spreadingrequire further study. However, axial valley formation by horizontal extensional stressesin a strong plate that deforms by faulting is consistentwith episodic spreadingand with the preservationof axial topographyon 2-km-wide neovolcanic zone that is much narrower than the extinct ridges. The axial structureof mid-oceanridges dependsstrongly width of the axial valley and the region of plate extension. If plate extension is to create the stress-supportedaxial val- on spreadingrate. However, axial structureis also clearly ley, it is sufficient for a continuous strong plate to exist influenced by factors other than the spreadingrate. In the most of the time. During times that magma penetration vicinity of hot spots where ridge axes are much shallower weakens the plate in a narrow zone at the ridge axis, hori- than normal, an axial valley may not be present even at zontal extension of the plate outside this region will stop. slow spreadingrates. The Reykjanes Ridge is an often cited If the plate behaved as a simple viscous material, the axial example. In contrast, at the Australian-Antarctic Discorvalley topography would tend to relax during such times of dance, a deeper than normal section of the intermediate reduced horizontal extension or stress. However, because spreadingrate Australian-Antarctic Ridge, the ridge has a the axial valley is produced by faulting, if stressesdue to well-defined axial valley that is not presenton adjacentsecaxial valley topography alone are insufficient to cause slip tions which are at a more normal depth [Weisseland Hayes, on faults then the axial valley topography will remain 1974; Forsyth et al., 1987]. Ridge axes also deepen sysfrozen-in during these periods of reduced horizontal stress. tematically approachinga transform fault, and the axial valIn our idealizedmodels,stressesdue to axial topographyare ley becomescorrespondinglymore pronounced. This correinsufficient to cause yielding of the plastic plate. This ex- lation of ridge axis structurewith seafloor depth may be expected freezing-in of topography is consistentwith the ob- plained by the relative rates of magmatic heat input associservation that extinct slow spreadingridges in the Labrador ated with crustal formation. In areas where the crustal proSea [Kristoffersen and Talwani, 1977], the northeasternAt- duction rate is low, the seafloor is deep, the magmatic heat lantic Ocean [Talwani and Eldholm, 1977], and the Coral Sea input at the ridge axis is low, and the plate is thicker than [Watts, 1982] preserve their axial valley morphology and normal for that spreading rate. The thicker, and therefore associated gravity anomalies. After extension stopped, the stronger,plate will result in a more pronouncedaxial valley. frozen-in topography should be supported by elastic In contrast, areas of high crustal production will be shallower and will have higher magmatic heat input and therestrengthof the plate as suggestedby Watts [1982]. An axial valley that is created by viscous flow in a ridge fore thinner ridge axis lithosphere with an axial structure axis conduit should disappear after spreadingstops. The re- characteristicof a faster spreadingrate. This is shown by laxation time required for seafloor elevation at the top of a the example in Figure 8f, where the crustalthicknessis douridgeaxisconduit to adjust to changes in pressure withinthe ble that in Figure 8d while all other parametersof the therconduit is 121xH/Apgd•, where H is the conduitheight, d is mal model remain the same. The lithosphere thickness at the conduit width, Ap is seafloor rock-water density differ- the ridge axis was approximately halved by doubling the crustal thickness. For a lithosphere at brittle failure with a ence, and I.t is the asthenosphere viscosity [cf. Phipps Morgan and Parmentier, 1985]. The depth of the axial valley at frictional strength proportional to depth, halving the lithoa halfspreading rateU is õ--12•tUH2/Apgd 3 [cf.Parmentiersphere thickness would decrease the horizontal force and and Forsyth, 1985] so that the relaxation time becomes moment by a factor of 4. For a normal crustal thicknessat simply õd/HU. For a conduitwith H=30 km and d=10 km slow spreadingrates, as shown in the example of Figure 8d, (Lachenbruch, 1973), an initially 1-km-deep axial depres- a significant part of the ridge axis lithospherewill consist sion at a half spreadingrate of 10 mm/yr will relax in ap- of mantle peridotiterather than crustalgabbro. The lower proximately3x104years.The axialvalleydueto sucha ductile flow strength of gabbro relative to peridotite [Chen ridge axis conduit would be well preservedbetweenspreading and Morgan, 1987] would increasethe ridge axis weakening effect causedby thickeningthe crust. episodes occurring each104years. However, preservation of the axial valley after spreadingstops would require cooling and strengtheningof mantle material in the ridge axis SUMMARY conduit.The timeH2/4•cfor conductive coolingto penetrate to a depthH is approximately 8xl 06 yearsforH=30 km Axial topography and associatedgravity anomalies proand•c=l mm2/s. Evenif theupper5-6 km of the oceanicvide an important constrainton the thermal and mechanical lithosphere near the ridge axis were rapidly cooled by hydrothermal convection, this cooling time would remain longer than the relaxation time. It therefore appears unlikely that topography produced by a ridge axis conduit would be preserved after spreading stops. If the plate thickness increases more gradually with distance from the ridge axis, as in Figure 3c, the relaxation time for topography supportedby mantle flow stressesin a uniform viscos- structure of spreadingcenters. Observed gravity anomalies indicate that the axial topography at slow spreadingridges cannot be isostatically compensatedat depths as shallow as 6 km and therefore cannot be simply explained by crustal thicknessvariations. In this study, several mechanismsfor the creation of axial topographyat mid-ocean ridges that is supportedby stresseswithin a strong plate near the ridge axis or the flowing mantle beneath it have been examined. MOROANET AL: TIlE ORI{IN OFMOR AXIAL STR• The relative importanceof mantle and lithospheredeformation on axial topographyhas been explored using elementary solutionsfor viscousflow problemsand simple thermal models. The relatively low relief of fast spreadingridges and the pronouncedaxial valley on slow spreadingridges are the principalfeaturesthat need to be explainedby suchfirst- 12,83$ slow spreadingridges. The same mechanismwill operate at fast spreading ridges, but the plate thickness at the ridge axis, as predicted by our models, should be small enough that the horizontal force in the plate is too small to produce appreciableridge axis relief. order models. A narrow, vertical steady-stateconduit through which viscous asthenosphere rises on the ridge axis is an appealingly simple and frequently cited explanationfor the presenceof an axial valley on slow spreadingridges. However, the temperaturefield associatedwith flow in a narrow conduit does not correspondto an assumedsteep-walledconduit, but instead has the form of a thermal boundary layer, with nearly horizontal isotherms, that thickens gradually with distancefrom the ridge axis. Thus if a steady-stateridge axis conduit does exist, the strength difference between asthenosphereand conduitwalls must be attributedto a significant strengthreduction due to the presenceof partial melt in the rising asthenosphere,by an increasedstrengthof depleted residual mantle forming the conduit walls, or by some other, nonthermally controlled mechanism. Mantle flow due to plate accretion and melt segregation can produceaxial-valley-like topography. However, a mantle viscosity beneath the spreadingcenter on the order of 1020Pas (1021P) is required for thesemechanisms to contribute significantly to the kilometer-scalerelief present on slow spreadingridges. In the absenceof plate accretionand extension near the ridge axis, mantle flow due to plate spreadingalone makes no contributionto the topography. This suggeststhat several previously consideredkinematic models based on this mechanismdo not adequately explain the axial valley. Horizontal extensional stressesin a strong, brittle lithospherethat thickens with distancefrom the ridge axis can produce an axial-valley-like topographywith relief comparable to that observed. Using a continuum idealization of deformationdue to faulting, the axial valley width is controlled by the plate thickness. A 10-km-thick brittle plate near the ridge axis would producea 40-km-wide axial valley as observedon slow spreadingridges. Microseismicity,im- plying brittle plate deformation, has been observed to depthsof 8 km at a slow spreadingridge. Basedon accepted values for sliding friction on faults, a plate that thickensby as little as several kilometers within the width of the axial Acknowledgments. This researchwas supported by National Science Foundation grant OCE-8613946. Final stages of the research and most of the writing were completed at the Irrstitute of Geophysicsand Planetary Physics, ScrippsInstitution of Oceanography, where both J.P.M. and E.M.P. were supportedby Ida and Cecil Green Fellowships. Special thanks are due to John Orcutt, who helped make possible the 1 year stay of J.P.M. and the 2 month summervisit of E.M.P. Norman Sleep and an anonymous reviewer provided helpful commentson an earlier version of the manuscript. REFERENCES Anderson,R. N., and M.D. Zoback, Permeabilityunder pressures and convection in the oceanic crust near the Costa Rica Rift, eastern equatorial Pacific, jr. Geophys. Res., 87, 2860-2868, 1982. Batchelor,G.K., An Introductionto Fluid Dynamics,615 pp., Cambridge University Press,New York, 1967. Bergman, E.A., and S.C. Solomon, Source mechanisms of earthquakesnear mid-oceanridges from body waveform inversion:Implications for the early evolution of oceanic lithosphere,jr. Geophys. Res., 89, 11,415-11,441, 1984. Brace, W.F., and D.L. Kohlstedt, Limits on lithosphere stressimposed by laboratory experiments,J. Geophys.Res., 85, 62486252, 1980. Byedee, J. D., Friction of rocks, Pure Appl. Geophys., 116, 615626, 1978. Chen, Y., and W.J. Morgan, A mechanic model for the contrastbetween the broad rift valleys on slow spreadingcenters and no valleys on fast spreadingones (abstract), Eos Trans. AGU, 68, 407, 1987. Collette, B. J., J. Verhoet, and A. F. J. de Mulder, Gravity and a model of the median valley, jr. Geophys., 47, 91-98, 1980. Combarnous,M., Natural convectionin porousmedia and geothermal systems,Int. Heat Transfer Conf., 6th, 45-59, 1978. Cooper, R. F., and D.L. Kohlstedt, Rheology and structure of olivine-basalt partial melts, jr. Geophys.Res., 91, 9315-9323, 1986. Crouch, S. L., and A.M. Staffield, Boundary Element Methods in Solid Mechanics, 322 pp., Allen and Unwin, Boston, Mass., 1983. Derrick, R.S., P. Buhl, J. Mutter, J. Orcutt, T. Brocker, and J. Madsen, Multichannel seismic imaging of the axial magma chamber along the East Pacific Rise between 9 and 13 N (abstract),Eos Trans. AGU, 67, 360, 1986. valley can producekilometer-scalerelief. Fehn, U., and L. M. Cathles, Hydrothermal convection at slowFor a transition from strong brittle to weaker ductile despreadingmid-oceanridges, Tectonophysics,55, 239-260, 1979. formation that occurs at a relatively constant temperature, Fehn, U., K. C. Green, R. P. Von Herzen, and L. M. Cathles, Numerical models for the hydrothermalfield at the Galapagos the thicknessof the plate near a ridge axis is thermally controlled. Thermal models which consider horizontal SpreadingCenter, jr. Geophys.Res., 88, 1033-1048, 1983. Forsyth, D.W., R.L. Ehrenbard,and S. Chapin, Anomalousupper spreadingof magmaticallyaccretedcrust on the ridge axis mantle beneath the Australian-Antarctic Discordance, Earth Planet. and upwelling mantle flow beneath the axis due to plate Sci. Lett., in press, 1987. spreadingpredicta plate thicknessnear the ridge axis that is Gregory, R., and H. P. Taylor, An oxygen isotopeprofile in a sec- tion of Cretaceousoceanic crust, Samail Ophiolite, Oman: Evithinner than the crust even at slow spreadingrates. Howdence for/j18 0 buffering oftheoceans bydeep (>5km)seawaterever, including the effect of hydrothermal cooling, reprehydrothermal circulation at mid-ocean ridges, jr. Geophys. Res., sentedsimply as an enhancedthermal conductivity,a plate 86, 2737-2755, 1981. thicknessof 8 km can be reasonablypredicted. These mod- Hickman, S. H., M. G. Langseth, J. F. Svitek, In situ permeability and pore pressuremeasurementsnear the Mid-Atlantic Ridge, els show that it is reasonableto expect the plate to thicken DSDP Site 395, Initial Rep. Deep Sea Drill. Proj., 78B, 699-708, by at least several kilometersover a horizontal distanceof 1984. 10 km from the ridge axis. If so, horizontal extensional Karato, S., Does partial melting reducethe creep strengthof the upstressesin lithospherenear the ridge axis can explain both per mantle?,Nature, 319, 309-31O, 1986. the form and the amplitude of axial valley topography at Kristoffersen, Y., and M. Talwani, Extinct triple junction south of 12,836 PHIPPS MORejAN ETAL:TIlEORIflNOFMORAXIALSTR• Greenland and the Tertiary motion of Greenlandrelative to the Due to mantle deformation?(abstract),Eos Trans. AGU, 65, 1088, North America, Geol. Soc. Am. Bull., 88, 1037-1049, 1977. 1984. Kuznir, N.J., Thermal evolutionof the oceaniccrest; its dependence PhippsMorgan, J., arid E. M. Parmentier,Causesand rate-limiting on spreadingrate and effect on crustal structure,Geophys.J. R. mechanismsof ridge propagation:A fracture mechanicsmodel, J. Astron. Soc., 61, 167-181, 1980. Geophys. Res., 90, 8603-8612, 1985. Lachenbruch,A. H., A simplemechanicalmodel for oceanicspread- Purdy, G.M., and R.S. Detrick, The crustal structure of the Miding centers,J. Geophys.Res., 78, 3395-3417, 1973. Atlantic Ridge at 23N from seismicrefractionstudies,J. Geophys. Res., 91, 3739-3762, 1986. ' Lachenbruch,A. H., Dynamics of a passive spreadingcenter, J. Geophys. Res., 81, 1883-1902, 1976. Reid, I., and H. R. Jackson, Oceanic spreadingrate and crustal Laughton, A.S., and R.C. Searle, Tectonic processes on slow thickness,Mar. Geophys.Res., 5, 165-172, 1981. spreadingridges,in Deep Drilling Resultsin the Atlantic Ocean: Ribando, R.J., K.E. Torrance, and D.L. Turcotte, Numerical models Ocean Crust, Maurice Ewing Ser., vol. 2, edited by M. Talwani, for hydrothermal circulation in the oceanic crust, J. Geophys. Res., 81, 3007-3012, 1976. C.G. Harrison, and D.E. Hays, pp. 15-32, AGU, Washington, D.C., 1979. Sleep, N.H., Sensitivityof heat flow and gravity to the mechanism Lewis, B.T.R., and J.D. Garmany, Constraintson the structureof the of sea-floorspreading,J. Geophys.Res., 74, 542-549, 1969. East Pacific Rise from seismic refraction data, J. Geophys.Res., Sleep, N.H., Formation of the oceanic crust: Some thermal con87, 8417-8425, 1982. straints,J. Geophys.Res., 80, 4037-4042, 1975. Lin, J., and E. M. Parmentier,Implications of gravity and topogra- Sleep, N.H., and B. R. Rosendahl, Topographyand tectonicsof mid-oceanic ridge axes, J. Geophys.Res., 84, 6831-6839, 1979. phy for the transient thermal and mechanicalstructureof midSolomon, S.C., P.Y. Huang, and L. Meinke, The seismic moment ocean ridges (abstract),Eos Trans. AGU, 67, 362, 1986. budgetof slow spreadingridges (abstract),Eos Trans. AGU, 67, Lister, C. R. B., On the penetrationof water into hot rocks, Geo1231, 1986. phys. J. R. Astron. Soc., 39, 465-509, 1974. Louden, K. E., R. S. White, C. G. Potts, and D. W. Forsyth, Struc- Talwani, M., and O. Eldholm, Evolution of the Norwegian-Greenland ture and seismotectonics of the Verna Fracture Zone, in The Tec- Sea, Geol. Soc. Am. Bull., 88,969-999, 1977. tonics of Fracture Zones, editedby P.J. Fox, in press,1987. Tapponier, P., and J. Francheteau,Necking of the lithosphereand Macdonald, K. C., Mid-ocean ridges: Fine scale tectonic, volcanic, the mechanicsof slowly accretingplate boundaries,J. Geophys. Res., 83, 3955-3970, 1978. and hydrothermal processes within the plate boundary zone, Annu. Rev. Earth Planet. Sci., 10, 155-190, 1982. Toomey, D. R., S.C. Solomon, G. M. Purdy, and M. H. Murray, Macdonald, K. C., and T. M. Atwater, Evolution of rifted ocean Microearthquakesbeneaththe median valley of the Mid-Atlantic ridges, Earth Planet. Sci. Lett., 39, 319-327, 1978. Ridge near 23N: Hypocentersand focal mechanisms,J. Geophys. Res., 90, 5443-5458, 1985. Macdonald,K.C., and B. Luyendyk,Deep-Tow studiesof the structure of the Mid-Afiantic Ridge crest near 37N (FAMOUS), Geol. Soc. Vening Meinesz, F. A., Les grabenafricainsresultatde compression Am. Bull., 88, 621-636, 1977. ou de tension dans la croute terrestre,Inst. R. Colon. Belg. Bull. McClain, J.S., J.A. Orcutt, and M. Burnett, The East Pacific Rise in Seances, 21, 539-552, 1950. Watts, A.B., Gravity anomalies over oceanic rifts, in Oceanic and 8639, 1985. Continental Riffs, Geodyn. Ser., vol. 8, edited by G. Palmason, Morton, J.L., and N.H. Sleep, A mid-oceanridge thermal model: pp. 99-106, AGU, Washington,D.C., 1982. Constraints on the volume of axial hydrothermal flux, J. GeoWeissel, J.K., and D.E. Hayes, The Australian-AntarcticDiscordance: phys. Res., 90, 11,345-11,353, 1985. New results and implications,J. Geophys.Res., 79, 2579-2587, 1974. Muskhelishvili, N.I., Some Basic Problems of the Mathematical Theory of Elasticity, 704 pp., Nordhoff, Leyden, Netherlands, Wiens, D.A., and S. Stein, Intraplate seismicity and stressesin 1953. young oceanic lithosphere,J. Geophys.Res., 89, 11,442-11,464, 1984. Nelson, K. D., A simple thermal-mechanicalmodel for mid-ocean ridge topographicvariation, Geophys.J. R. Astron. Soc., 65, 19- Zuber, M. T., and E. M. Parmentier,Lithosphericnecking:A dy30, 1981. namic model for rift morphology,Earth Planet. Sci. Lett., 77, 373-383, 1986. Parmentier,E. M., Dynamic topographyin rift zones:Implications for lithospheric heating, Philos. Trans. R. Soc. London, Ser. A, 321, 23-25, 1987. J. PhippsMorgan,Departmentof Earth, Atmospheric and PlaneParmentier, E.M., and D.W. Forsyth, Three-dimensionalflow be- tary Sciences,Rm. 54-824, Massachusetts Institute of Technology, neath a slow-spreadingridge axis: A dynamic contributionto the Cambridge,MA, 02139 deepeningof the median valley towardsfracture zones,J. GeoE.M. Parmentierand J. Lin, Departmentof GeologicalSciences, phys. Res, 90, 678-684, 1985. Box 1847, Brown University, Providence,RI, 02912 cross section: A seismic model, J. Geophys. Res., 90, 8627- Parmentier,E. M., J. PhippsMorgan,and J. Lin, Axial morphology of slow spreadingmid-ocean ridges: steady state necking of a ductile lithosphere (abstract), Eos Trans. AGU, 66, 1091, 1985. PhippsMorgan, J., and E. M. Parmentier,Ridge axis morphology: (received March 2, 1987; revisedMay 9, 1987; acceptedJune 13, 1987.)