melt topology and seismic anisotropy in mantle peridotites of the oman ophiolite

ELSEVIER Earth and Planetary Science Letters 164 (1998) 553–568

Melt topology and seismic anisotropy in mantle peridotitesof the Oman ophiolite

David Jousselin Ł, David Mainprice

Laboratoire de Tectonophysique, CNRS UMR 5568, Universite Montpellier II, Montpellier, France

Received 3 June 1998; revised version received 14 September 1998; accepted 1 October 1998

Abstract

This paper presents shape measurements of plagioclase and clinopyroxene inclusions, assumed to reflect melt topology,in peridotites of the uppermost mantle section of the Oman ophiolite. The plagioclase and clinopyroxene grains are devoidof any intracrystalline deformation in all samples. In contrast, the olivine in these rocks has recorded a high temperatureplastic deformation, with different strengths of the crystallographic preferred orientation (CPO) of the olivine grains.Individual ‘melt pockets’ are first described by ellipses in two dimensions. They are more elongated when they havea larger area, and they are preferentially oriented parallel to the lineation (the X structural axis) of the sample, with abetter defined preferred orientation for samples that have a stronger CPO. In a second step, an average melt phase shapeis defined in three dimensions for each sample, using the image autocorrelation technique. The average shape is nearlyspherical for the samples with weak CPOs, and it is ellipsoidal, with a long axis parallel to X and the short axis parallelto Z (normal to the foliation) for samples with strong CPOs. The long axis of the ellipsoid is 3 times as long as the shortaxis for the sample with the strongest CPO. We use an anisotropic differential effective medium method to estimate theseismic properties of partially molten upper mantle peridotites. The melt pockets were modelled as basalt filled inclusionswith the average shape and orientation given by the image analysis. The CPO of the olivine crystals was used to calculatethe elastic properties of the anisotropic background medium. The calculated P-wave seismic anisotropies ranged from 5 to15% with the anisotropy increasing with the CPO strength and melt fraction. The maximum P-wave velocities are foundalong X with velocities above 8 km=s at 0% melt and an average 0.5 km=s reduction for 10% of melt. The minimumP-wave velocities are found along Z with velocities generally below 7.5 km=s at 0% melt and an average reduction of 0.8km=s for 10% of melt. 1998 Elsevier Science B.V. All rights reserved.

Keywords: mantle; melts; anisotropy

1. Introduction

The melt distribution at present day mid Oceanridges is mainly derived from the interpretation of

Ł Corresponding author. Fax: C33 46714 3603; E-mail:[email protected]. Present address: University of Ore-gon, 1272 Geol. Sciences, Eugene OR 97403, USA.

seismic velocities (e.g. [1]). A key to the seismic in-terpretation is the relationship between melt fractionand seismic velocity. To date most studies addressingthe problem of seismic velocity and melt distributionin the upper mantle have assumed that the solidbackground medium (peridotite) has isotropic prop-erties (e.g. [2–4]) and the melt filled inclusions havea random shape preferred orientation (SPO). We will

0012-821X/98/$ – see front matter 1998 Elsevier Science B.V. All rights reserved.PII: S 0 0 1 2 - 8 2 1 X ( 9 8 ) 0 0 2 3 5 - 0

554 D. Jousselin, D. Mainprice / Earth and Planetary Science Letters 164 (1998) 553–568

present observations on selected samples from an ex-tensive field mapping of the Oman ophiolite, knownto be a paleo mid ocean ridge environment (e.g. [5]),which reveal that the olivine crystals have moder-ate to strong crystallographic preferred orientation(CPO) and the melt phase has a weak to strongSPO. The preferred orientation of olivine results inan anisotropic elastic background medium. A secondelastic anisotropy is introduced by the SPO of themelt inclusions. To model such a complex system,Mainprice [6] introduced a general tensorial differ-ential effective medium approach which specificallyallows the role of the fluid phase to be taken intoaccount by using Gassmann’s poro-elastic theory asin previous isotropic methods [3,4,7]. Using thismethod and the measured CPO of olivine, and SPOof melt inclusions for samples from the uppermostmantle section of the Oman ophiolite we will presenta set of predicted P-wave velocities as a function ofmelt fraction.

Waff and co-workers established that surface ten-sion between crystalline grains and the melt de-termine the melt distribution [8,9]. In the case ofan isotropic crystalline medium, the predicted meltdistribution is an interconnected network of tubulesalong grain edge intersections [10]. However, themelt distribution in experimentally produced partialmelts differs substantially from the tubule geome-try [11,12]. Melt inclusions of the micrometer scaleare often elongated along grain boundaries, form-ing pockets wetting several grains. In the partiallymolten peridotites deformed at a differential stressof 40 MPa, melt wets most grain boundaries withno preferred orientation [13]. At differential stressesgreater than 100 MPa, Bussod and Christie [14]found a network of melt channels oriented at anangle of 30º to the compression axis .¦1/; at simi-lar conditions, Daines and Kohlstedt [15] also foundmelt pockets subparallel to ¦1, but noted that an-nealing can redistribute melt perpendicularly to ¦1.Little is known about the shape and distribution ofmelt inclusions in mantle peridotites which experi-

Fig. 1. Drawing of the melt phase in the X Z thin sections of 4 samples of this study. Black areas represent the clinopyroxenes andplagioclases beads, considered to be the fossilized melt phase, white areas represent olivine (60 to 80%), serpentine (20 to 40%), and theorthopyroxene (<6%), grey is spinel (<5%). An example photograph is shown at a smaller scale for the samples 95OD21 and 91OA133,they are located by a box in the corresponding thin section; visible halos around plagioclase clots are highly serpentinized olivine.

enced natural deformation, and no model exists topredict the distribution of fluids in an anisotropicpolycrystalline aggregate. In the absence of suchdata, this paper presents measurements of plagio-clases and clinopyroxenes (Cpx) inclusions assumedto have fossilized melt topology in dunites and de-pleted harzburgites from the uppermost mantle sec-tion in the Oman ophiolite. Following Faul et al.[12] and Daines and Kohlstedt [15] individual mea-surements of the equivalent ellipse of the inclusionsin a two dimensional section are presented. Averageaspect ratios and orientations are measured usingautocorrelated images of the melt phase in petrolog-ical thin sections. Influence of the strength of theolivine crystallographic fabric on the melt distribu-tion is discussed and the results are used to calculateanisotropic seismic velocities.

2. Melt topology fossilization

Olivine and orthopyroxene (Opx) crystals in pla-gioclase and Cpx impregnated peridotites dredged onthe EPR [16–18] or outcropping in the Oman ophi-olite ([19], this study), exhibit sharp sub-boundariesindicative of high temperature intracrystalline defor-mation, and strong crystallographic fabrics producedby large asthenospheric plastic strain (Fig. 2); in con-trast, the plagioclase and Cpx crystals (Fig. 1) are de-void of any intracrystalline deformation. Plagioclaseand Cpx are therefore unlikely to have crystallizedduring the deformation of the peridotite. The fact thatthey form clots with a shape which is related to theolivine slip direction suggests that, during deforma-tion, melt can form pockets, filled by plagioclase andCpx when the peridotite enters the thermal conduc-tive boundary layer and asthenospheric deformationceases. Some authors interpret these pockets as theresult of crystallization of a trapped melt which waspercolating in the shallow mantle [16–19]. There-fore, plagioclase and Cpx clots will be called ‘meltpockets’. However, we must make some reservations

D. Jousselin, D. Mainprice / Earth and Planetary Science Letters 164 (1998) 553–568 555


about this designation. First, it is possible that themelt also precipitated olivine crystals which are notdistinguishable from the olivine matrix. Second, geo-chemical data suggest that the plagioclase and Cpxcrystals formed via open system processes in whichthe liquid was entering and leaving the rock; thesecrystals would only represent 5 to 25% of the liquidwhich was percolating in the peridotites ([20]; Kele-men, pers. comm.). This leads to a scenario wherethe plagioclase and Cpx do not represent the meltitself but fossilized the initial shape of the ‘meltpockets’ by progressively filling them and by crys-tallizing from a small amount of a circulating melt.In this model, necessary connections between the‘melt pockets’ would not be preserved or would betoo thin to be detected in our samples. In the follow-ing the plagioclase and Cpx fraction will be called‘melt fraction’ as it is the only directly measurablequantity, although it cannot be used as an estimateof the ‘true’ melt fraction which was certainly highlyvariable in space and time.

3. Location and brief description of the samples

The Maqsad area of the Oman ophiolite is be-lieved to be representative of a paleospreading-centerbelow a fast spreading ridge [5,21,22]. It is charac-terized by a mantle diapir with vertical mantle flowrotating to horizontal within a 100 to 500 metersthick Moho transition zone sandwiched between themantle harzburgites and the crustal gabbros [19].The transition zone is composed of harzburgites,‘dry’ and ‘melt impregnated’ dunites with inter-layered gabbro lenses. Samples generally exhibit afoliation plane corresponding to the plane of mineralflattening, i.e. the XY plane of the deformation el-lipsoid [23] and a lineation parallel to X . Becauseof the large strain experienced by the peridotites, thefoliation and the lineation only have a slight obliq-uity with respectively the flow plane and the flowline. Our investigation is based on measurements ofsix samples from the Moho transition zone in theMaqsad area, which have been chosen as function oftheir crystallographic fabric, modal composition andstructural position. All samples have a low degreeof serpentinization (<30%) which is essential forquantitative image analysis. In all samples, olivine

and Opx grain size ranges from 0.5 to 4 mm, with anaverage around 2 mm.

Sample 91OA133: depleted harzburgite and duniteinterlayered at centimeter scale from a zone of hor-izontal flow. The texture is High-Temperature (HT)porphyroclastic [24]. The specimen records the tran-sition from harzburgite to dunite through incongruentdissolution of Opx in a migrating melt. Melt is nowrepresented by concordant beads of plagioclase. Thecrystallographic and shape fabrics are very strong.

Sample 95OD159: dunite from a zone of horizon-tal flow with a HT porphyroclastic texture. The meltis represented by concordant beads of plagioclaseand Cpx. The crystallographic and shape fabrics arevery strong.

Sample 95OD21: depleted harzburgite–dunitefrom a zone of vertical flow with a HT porphy-roclastic texture. The specimen records the transitionfrom harzburgite to dunite, in a similar manner to91OA133. Melt is represented by concordant beadsof Cpx. The crystallographic and shape fabrics arestrong.

Sample 95OD89: dunite from a zone of horizontalflow with a HT porphyroclastic texture. The melt isrepresented by concordant beads of plagioclase andCpx. The crystallographic and shape fabrics have amoderate strength.

Sample 95OD121: dunite from the zone of rota-tion between vertical and horizontal flow at the top ofthe mantle diapir with a HT porphyroclastic texture.The melt is represented by approximately concordantbeads of plagioclase and Cpx. The crystallographicand shape fabrics have a moderate strength.

Sample 92OA2: dunite from a zone of strong meltimpregnation at the top of the diapir with a HTgranular texture [25]. The melt is represented byplagioclase pockets. The crystallographic and shapefabrics are weak comparatively to other samples,with the foliation and the lineation being difficultto define. This type of sample is rather uncommonin the Maqsad area where a HT plastic strain isrecorded everywhere. The sample site may be inter-preted as a place where a large melt fraction wasintroduced in the last stages of the ridge accretion,destroying the peridotite solid framework. This melthas not been compacted and transposed before theridge was sampled to become an ophiolite. One cannote the similarity of melt topology in this sample


(Fig. 1) with, for example, the one shown in thefig. 5b of Bussod and Christie [14] or the fig. 1bof Daines and Kohlstedt [15] although they are at ascale one order of magnitude smaller.

We note that the concordant orientations of pla-gioclase and Cpx clots with the foliation is a com-mon characteristic that we find at all scales. In par-ticular, the interlayered gabbro sills of the transitionzone are lens shaped and have magmatic lineationsparallel to those of the enclosing dunite. This suggestthat this system did not suffer much change stressuntil being totally frozen, and that such distributiondoes not correspond to a post-deformation anneal-ing which would be difficult to detect in the typicalasthenospheric HT textures we study.

4. Crystallographic fabrics

Crystallographic fabrics have been measured us-ing an optical petrological microscope equipped witha 5-axis universal stage. Following Mainprice andSilver [26], we calculated the J index, defined byBunge [27] as the strength of a crystallographic fab-ric, to compare the data of our 6 samples. The Jindex of olivine fabrics has been shown to increasewith finite strain in axial compression [26]. Resultsare presented on Fig. 2. In the following description,we present the fabric data as pole figures. The righthanded orthogonal reference frame is the lineation(X), the normal to the foliation plane (Z ) and thenormal to these two directions (Y ). All olivine fab-rics have a (100) pole figure with a single point max-imum which is adjacent to the grain shape lineation(X); it may also be true for the sample 92OA2, as thedefinition of lineation and foliation is uncertain, if Xis N–S on the pole figure as seems likely. The fourfabrics of the top of Fig. 1 also have a [010] singlepoint maximum perpendicular to the foliation (paral-lel to Z ). With decreasing fabric strength, the fabricsbecome more diffuse and the maxima less defined. Aplot of the fabric strength versus the ‘melt fraction’(Fig. 3) shows no clear relationship, although thelowest ‘melt fraction’ correlates with highest fabricstrength.

5. Image acquisition and processing

The contours of melt pockets were obtained fromprecise hand tracing of pictures of thin sections cutalong the X Z and XY planes of each sample (Fig. 2).Between two and four bitmap images were extractedfrom each scanned drawing of a thin section for pro-cessing and measurement. Each image consists of512 ð 512 pixels with a resolution of 550 pixels perinch (216 pixels=cm). Analysis was carried out byimage processing software (Optilab Pro 2.5 and NIHImage).

To quantify the melt distribution, two types ofprocessing have been adopted. First we used themethod described by Faul et al. [12] where the fol-lowing parameters are defined; the area .A/ is thesum of all contiguous pixels; the perimeter .P/ isthe length of the feature’s boundary and the length.L/ is the longest chord connecting any two pointson the perimeter, these were measured for each in-clusion. These basic measurements were used tocalculate the aspect ratio .Þ/ defined as the ratio ofwidth .w/ to fiber length (fl) of a melt pocket. Thewidth is an approximation of the minor dimensionof the feature and is calculated as w D .4A/=.³L/.The fiber length, an approximation of the length ofthe feature along its medial axis is calculated asfl D 0:5P � 2A=P . An area weighted mean aspectratio is derived for each section. The orientation ofthe long axis of the inclusions is also measured. Us-ing this method, it is difficult to calculate the averageequivalent ellipse of the melt phase in two dimen-sions or ellipsoid in three dimensions, required forcalculating anisotropic seismic velocities of the sam-ple. The average ellipse should take into account thefractional area for all melt inclusions and their orien-tations. A population of ellipsoids with random ori-entations in an isotropic background medium wouldresult in isotropic seismic properties and should berepresented by an average spherical shape. Ratherthan calculating an average shape by weighting eachinclusion as a function of its area and orientation,we have chosen to use the autocorrelation function(ACF). The use of the ACF, first developed in geo-logical field by Panozzo Heilbronner [28], yields theaverage axial ratio and orientation of the inclusionshape fabric. Further details of image processingconcerning the ACF are developed in mathematical


Fig. 3. Plot of the fabric strength versus the melt fraction. Error bars represent the range of melt fractions measured in the different thinsections used for each sample.

terms and explained with clear examples by PanozzoHeilbronner [28]. In this paper we will just explainthe principle of the visualization of the ACF. Let usimagine two copies of an image which consists ofone circle cut on an opaque background. Holdingthe two copies against a light source and superpos-ing them, such that the circles coincide, a state ofmaximum possible transmission of light is obtained.Starting at this position and keeping the first copyfixed, the second may be displaced in any direction.As a consequence, the total amount of transmittedlight decreases. We may regard the amount of lightthat passes through both copies as a measure of thevalue of the ACF at that given displacement. Forcircles, the rate of decrease of the ACF values awayfrom the origin is the same in all directions, thusthe contour lines are circular. In cases of ellipses or

Fig. 2. Crystallographic fabric of the olivine for all samples of this study. Projections on the lower hemisphere, contours at 1% intervals,J is the corresponding calculated fabric strength index, N is the number of olivine grains measured. Lineation direction, X , perpendicularto foliation plane, Z , and mutually perpendicular Y axis directions are shown at upper left.

anisotropic shapes, the ACF contours become elon-gated in the direction parallel to the preferred shapeorientation, with an axial ratio which is identical tothat of the average feature. The contour level whichis considered to be the most representative of theinclusions size is at 0.5 of central peak height.

6. Results of melt pockets measurements

6.1. Individual aspect ratio

Melt pockets with small surface areas .A/ tendto have spherical shapes with aspect ratios .Þ/ closeto unity, whereas large ones have elliptical shapeswith low aspect ratios (Fig. 4). As in experimentalresults [12,15], there is a continuous decrease of ax-


Fig. 4. Plots of the area of individual melt pockets versus their aspect ratio. X Z thin sections stacked in the top graph, XY thin sectionsstacked in the bottom graph, melt pockets of the different samples are distinguished by different symbols (see included boxes).

ial ratio with increasing surface area for all samplesin both the X Z and XY sections despite the varia-tions of modal composition and the fabric strength(Fig. 4). The aspect ratio ranges between 0.1 and 1.0.Weighted mean aspect ratios are in the same range

as those found by Daines and Kohlstedt [15] (Ta-ble 1). There is no obvious correlation between theweighted mean aspect ratio and the fabric strength(Table 1), nor between the individual aspect ratio ofmelt pockets and their orientations within a given


Table 1Compilation of some measurements for X Z and XY thin sections of each sample

V pX , VpY and Vp Z are P waves velocities calculated for the ‘melt fraction’ measured in the sample (melt %). Degree of seismicanisotropy determined is A.%/ D 200.Vp X � Vp Z/=.Vp X C Vp Z/.

thin section (Fig. 5). One can note that the meanaspect ratio is slightly lower in the X Z thin sectionthan in the XY thin section of each sample. Exceptfor the case 92OA2 (a high ‘melt fraction’=low fab-ric strength sample), all samples show a preferredorientation of melt pockets, with long axis subpar-allel to the structural axis X (Fig. 6 and standarddeviations in Table 1). The orientation relationship ismore marked in the X Z thin sections than in the XYones, and it tends to be better defined with increas-ing J index. Such characteristics were expected bysimple inspection of the thin sections (examples inFig. 1), where the X Z section shows well orientedmelt pockets.

6.2. Average shape fabric: results of the ACFprocessing

The shape preferred orientation given by twodimensional autocorrelated images can be analysedin terms of ellipses on X Z and XY sections, withtheir major axes .a; b; c/ parallel to structural axesX , Y , and Z within š15º respectively (Fig. 7).From the aspect ratios c=a and b=a, we can estimatec=b. The maximum axial length .a/ is systematicallyalong X , the intermediate .b/ along Y , and the

minimum .c/ along Z . By normalizing the c axiallength to 1.0, we find that the length along X rangesfrom 1.0 for the least deformed sample (specimen92OA2, with the minimum fabric strength, J D 5:7),to 3.0 for the most deformed sample (specimen91OA133, with the maximum fabric strength, J D19:8), leading to aspect ratio .c=a/ between 0.3 and1.0; similarly, the b axis of the ellipsoid ranges from1.0 to 1.6, leading to aspect ratios .c=b/ between 0.6and 1.0 (Table 1). Although more data are necessary,it seems that the aspect ratio in the X Z section .c=a/decreases with increasing fabric strength.

Comparison of the average shape fabric, given bythe ACF, with the mean aspect ratios, derived fromthe measurements of individual melt pockets, showsthe importance of the shape preferred orientationwhen the fabric strength is high. This is clear in thetwo extreme samples: 92OA2 has the largest melt in-clusions (in area) and thus the minimum mean aspectratios, but its low shape preferred orientation (relatedto a weak crystallographic fabric strength) results inspherical ACF contours. On the contrary, 91OA133has the smallest melt inclusions and thus the maxi-mum mean aspect ratio, but its high shape preferredorientation (related to a high crystallographic fabricstrength) results in the most elongated ACF contours.


Fig. 5. Plot of the orientation of the melt pockets, versus their aspect ratio, for the sample 95OD21.

7. Error range in the measurements

Comparison between individual melt inclusionmeasurements or ACF contours, done for the 2 to 4pictures that result from the drawing of an entire thinsection, give the same results with variations of lessthan 5%. For example, the drawing of the X Z thinsection of sample 92OA2 is divided in two pictureswhich give a weighted mean aspect ratio of 0.130for picture 1, and 0.136 for picture 2, with an overallaverage of 0.131; the aspect ratio a=c of the ACFcontours varied from 1 to 1.05. Measurements of the‘melt fraction’ were more variable from one pictureto another and also from the X Z thin sections to theXY thin sections, with variations of up to 30%. Tocomplete the investigation in the error ranges twoother X Z thin sections were cut in sample 95OD21,and a complementary YZ thin section was cut for3 samples (92OA2, 95OD21, 91OA133). All gavethe same variations, showing that the weighted meanaspect ratios or the elongation of the ACF contours

are measured with relatively good precision (5%)whereas the ‘melt fraction’ is more difficult to mea-sure. Another potential error is that our techniquedoes not allow to resolve linear dimensions belowfew hundreds micrometers. It therefore cannot detectsmall features such as possibly existing thin filmsalong grain boundaries. This problem has been ex-plored by artificially introducing in our images aframe of micrometric high aspect ratio inclusionsmimicking melt films or cross sections of triple junc-tion tubules. We found that if such features existed,they would not change our results because of theirlow volume compared to the ‘melt pockets’.

8. Calculation of seismic velocities frompetrofabrics and average aspect ratio of meltpockets

The low frequencies (e.g. 20 Hz), used in marineseismic experiments, correspond to a regime where


Fig. 6. Histograms of the orientations of melt pockets in the X Z thin section of each sample. The X axis location is mentioned on theabscise within an error range estimated to be š2 degrees. The Z axis is at 90 degrees to X .

the pore fluid has enough time to reach an equilib-rium pore fluid pressure throughout the aggregate.In such conditions, the pore fluid system can beconsidered mechanically connected. This is modeled

by using a poro-elastic relationship between elasticmoduli of porous medium with fluid-filled pores andthe moduli of the same medium with empty poresderived by Gassmann [7], extended to anisotropic


Fig. 7. Graphs of the ACF contours in the X Z thin section of each sample. The contour level which is 0.5 of central peak height ishighlighted by a thicker contour; it is considered as the most representative of the melt pocket size, though neighboring contours aremeasured and give approximately the same result (contour levels are 12, 16, 20, 24, 28, 32, 39, 47, (50), 55, 63, 71, 78, 94% of peakheight).


media by Brown and Korringa [29]. In terms ofcompliances the anisotropic elastic properties of theporous homogeneous rock containing a fluid can beformulated as

Ssati jkl D Sdry

i jkl

�"

.Sdryi jkk � Ssolid

i jkk /.Sdryiikl � Ssolid

iikl /

.Sdryiikk � Ssolid

iikk /C .1=K f � Ssolidiikk /�

#where the superscripts indicate the following, ‘sat’is the saturated rock, ‘dry’ the dry porous medium,‘solid’ the solid mineral components, K f is the bulkmodulus of the pore fluid and f the porosity orpore fluid volume fraction. The differential effectivemedium method was used to calculate the elasticconstants of the ‘dry’ porous medium (see [6] fordetails). The properties of the solid matrix have beencalculated from the single crystal elastic constantsand crystal preferred orientation data using tech-niques recently reviewed by Mainprice and Humbert[30].

Using this model and the sample data presentedabove, we have calculated the P wave seismic ve-locities in our samples along the structural directionsX , Y and Z for different melt percentages (Fig. 8and Table 1). The single crystal elastic constantsof olivine [31] which represent most of the solidcomponents of the rocks, have been calculated forhypersolidus conditions of 1200ºC and pressure of200 MPa. The elastic constants for molten basaltat 1200ºC have been calculated from the P-Wavevelocities measured by Murase and McBirney, [32].The Opx component was neglected because of itslow percentage (Table 1). In all samples the maxi-mum velocity .Vp X/ is parallel to the X axis, andthe minimum .Vp X/, along the Z axis. Consideringthat c=a seems to decrease with increasing fabricstrength and do not correlate with the ‘melt per-centage’ we have assumed that for a given fabricstrength, the average melt shape does not vary withthe melt percentage. At zero percent melt the ve-locity difference between the X , Y and Z directionsis a measure of the seismic anisotropy caused bythe background medium crystallographic preferredorientation, which increases with increasing textureindex J . This anisotropy ranges from 4% for 92OA2to 14% for 91OA133. The increase in melt fractioncauses a reduction of seismic velocity and an in-

crease of anisotropy due the shape fabric. With 0%of melt, Vpx varies from 7.9 km=s for 92OA2 to8.5 km=s for 91OA133, whereas Vpz varies from 7.2km=s for 95OD159 to 7.6 km=s for 92OA2. For 10%of melt, Vpx is decreased in the range of 7.3 to 8km=s (6–7% reduction), and Vpz ranges from 6.4 to7.1 km=s (6–10% reduction), leading to anisotropiesfrom 4% (92OA2) to 19% (91OA133). With 20%of melt the anisotropy for 91OA133 reaches 28%.At melt percentages higher than 40% the anisotropystarts to reduce as the melt which is an isotropiccomponent becomes predominant.

9. Application to oceanic fast spreading ridges

In the case of a refraction profile investigatingthe structure near the Moho level below fast spread-ing ridges, P wave rays are horizontal, thus theyshould travel along either the X or Y structural axeswithin the Moho transition zone where the foliationis horizontal, or along the Y or Z axes where theycross zones of upward flow such as in the 100 km2

mantle diapir documented in the Maqsad area [21].We deduce that the 12% reduction in upper mantlevelocities (a decrease from 8.2 to 7.2 km=s) reportedbelow the East Pacific Rise [33] might be interpretedeither in terms of upward flow with less than 10%melt or in terms of horizontal flow with more than10% melt.

10. Conclusion

Measurements of plagioclase and Cpx inclusionsin naturally deformed, melt impregnated peridotitesof the Oman ophiolite compare well with data fromsynthetic samples, suggesting that these crystals doimage melt topology. Our approach has the meritto allow measurements in strongly and naturallydeformed peridotites. We point out the importanceof the preferred shape orientation of melt pockets,which can be approximated by a single ellipsoid witha long axis .a/ parallel to X , the lineation direction,and a short axis .c/ parallel to Z , perpendicular tothe foliation plane. The average aspect ratio c=a ishigher than 0.3 and might decrease with increasingfinite strain. Using these data and a new technique

566D

.Jousselin,D

.M

ainprice/E

arthand

Planetary

ScienceL

etters164

(1998)553–568

Fig. 8. graphs of calculated anisotropic seismic velocities as a function of the melt fraction for each sample. Vp along X , Y , Z are represented respectively by squares,diamonds and circles. The grey bars represent the range of melt fraction which is actually observed in the samples. The J index and the average melt phase shape given bythe ACF contours is recalled for each sample in the right upper corner of the corresponding graph.


to calculate seismic velocities, we find a decreaseof P waves velocities with increasing melt fraction,and important seismic anisotropies, increasing withmelt fraction until an unrealistic 40% melt fraction isreached. The maximum seismic velocities are foundalong X with velocities above 8 km=s at 0% meltand an average 0.5 km=s reduction with 10% ofmelt; the minimum velocities are found along Z withvelocities generally below 7.5 km=s with 0% of meltand an average reduction of 0.8 km=s with 10% ofmelt. The velocity along the X axis is significantlyhigher than velocities found in the literature. In com-parison, using an oblate spheroid inclusions model(aspect ratio of 0.05) which does not take into ac-count the preferred orientations, other studies lead toa calculated velocity of 6.72 km=s with 10% of melt[4,12]. Experimental studies, which do not consideranisotropy [34], have results close to the curve givenfor the velocity along Z , with Vp from 6.5 to 7 km=swith 10% of melt. On this basis, considering thatseismic waves may travel along the fast axis, inter-pretations of seismic velocities in the mantle mayunderestimate melt fractions in mantle rocks.

Acknowledgements

We thank D. Blackman, U. Faul and an anony-mous referee for helpful reviews. This work hasbenefited from discussions with F. Boudier, A. Nico-las, H. Waff, J.L. Bodinier, P. Kelemen, B. Ildefonseand G. Lamoureux. Technical assistance for the thinsections has been provided by C. Nevado. This studywas possible thanks to the hospitality of M. Kas-sim and H. Al Azri from the Ministry of Minesand Petroleum of Oman and financial support fromCNRS-INSU and a governmental doctoral fellow-ship. [AC]

References

[1] S.C. Solomon, D.R. Toomey, The structure of mid-oceanridges, Annu. Rev. Earth Planet. Sci. 20 (1992) 329–364.

[2] G.M. Mavko, A. Nur, Melt squirt in the asthenosphere, J.Geophys. Res. 80 (1975) 1444–1448.

[3] G.M. Mavko, Velocity and attenuation in partially moltenrocks, J. Geophys. Res. 85 (1980) 5173–5189.

[4] H. Schmeling, Numerical models on the influence of partial

melt on elastic, anelastic and electric properties of rocks.Part I: elasticity and anelasticity, Physics Earth Planet. Inter.41 (1985) 34–57.

[5] A. Nicolas, F. Boudier, Mapping oceanic ridge segments inOman ophiolites, J. Geophys. Res. 100 (B4) (1995) 6179–6197.

[6] D. Mainprice, Modelling the anisotropic seismic proper-ties of partially molten rocks found at mid-ocean ridges,Tectonophysics 279 (1997) 161–179.

[7] F. Gassmann, Uber die elastizitat poroser medien, Vier.Natur Ges. 96 (1951) 1–23.

[8] J. Bulau, H.S. Waf, Mechanical and thermodynamic con-straints on fluid distribution in partial melts, J. Geophys.Res. 84 (B11) (1979) 6102–6108.

[9] H.S. Waff, J.R. Bulau, Experimental studies of near-equi-librium textures in partially molten silicates at high pres-sure, in: High Pressure Research in Geophysics, Adv. EarthPlanet. Sci. 12 (1982).

[10] N. Von Bargen, H.S. Waff, Permeabilities, interfacial ar-eas and curvature of partially molten systems: results ofnumerical computations of equilibrium microstructures, J.Geophys. Res. 91 (1986) 9261–9276.

[11] H.S. Waff, U.H. Faul, Effects of crystalline anisotropy onfluid distribution in ultramafic partial melts, J. Geophys.Res. 97 (1992) 9003–9014.

[12] U.H. Faul, D.R. Toomey, H.S. Waff, Intergranular basalticmelt is distributed in thin, elongated inclusions, Geophys.Res. Lett. 21 (1) (1994) 29–32.

[13] Z.M. Jin, H.W. Green, Y. Zhou, Melt topology in partiallymolten mantle peridotite during ductile deformation, Nature372 (6502) (1994) 164–167.

[14] G.Y. Bussod, Christie, Textural development and melttopology in spinel lherzolite experimentally deformed athypersolidus conditions, J. Petrol. (Special Lherzolite Is-sue) (1991) 17–39.

[15] M.J. Daines, D.L. Kohlstedt, Influence of deformation onmelt topology in peridotites, J. Geophys. Res. 102 (1997)10257–10271.

[16] M. Cannat, D. Bideau, R. Hebert, Plastic deformation andmagmatic impregnation in serpentinized ultramafic rocksfrom the Garret transform fault (East Pacific Rise), EarthPlanet. Sci. Lett. 101 (1990) 216–232.

[17] R. Hekinian, D. Bideau, J. Francheteau, J.L. Cheminee, R.Armijo, P. Lonsdale, N. Blum, Petrology of the East PacificRise crust and upper mantle exposed in Hess deep (EasternEquatorial Pacific), J. Geophys. Res. 98 (B5) (1993) 8069–8094.

[18] J. Girardeau, J. Francheteau, Plagioclase-wherlites and peri-dotites on the East Pacific Rise (Hess Deep) and the Mid-Atlantic Ridge (DSDP site 334): evidence for magma per-colation in the oceanic upper mantle, Earth Planet. Sci.Lett. 115 (1993) 137–149.

[19] F. Boudier, A. Nicolas, Nature of the Moho transition zonein the Oman ophiolite, J. Petrol. 36 (3) (1995) 777–796.

[20] J.L. Bodinier, Geochemistry and petrogenesis of the Lanzoperidotite body, western Alps, Tectonophysics 149 (1988)67–88.


[21] D. Jousselin, A. Nicolas, F. Boudier, Detailed mapping of amantle diapir below a paleo-spreading center in the Omanophiolite, J. Geophys. Res. in press.

[22] G. Ceuleneer, A. Nicolas, F. Boudier, Mantle flow patternsat an oceanic spreading centre: the Oman peridotites record,Tectonophysics 151 (1988) 1–26.

[23] A. Nicolas, J.P. Poirier, Crystalline Plasticity and SolidState Flow in Metamorphic Rocks, Wiley, London, 1976,444 pp.

[24] J.C. Mercier, A. Nicolas, Textures and fabrics of uppermantle peridotites as illustrated by basalt xenoliths, J.Petrol. 16 (1975) 454–487.

[25] A.M. Boulier, A. Nicolas, Classification of textures and fab-rics of peridotites xenoliths from south african kimberlites,Phys. Chim. Earth (1975) 467–475.

[26] D. Mainprice, P.G. Silver, Interpretation of SKS-wavesusing samples from the subcontinental lithosphere, Phys.Earth Planet. Inter. 78 (1993) 257–280.

[27] H.J. Bunge, Texture Analysis in Materials Science, Butter-worths, London, 1982, 593 pp.

[28] R. Panozzo Heilbronner, The autocorrelation function: animage processing tool for fabric analysis, Tectonophysics

212 (1992) 351–370.[29] R.J.S. Brown, J. Korringa, On the dependence of the elastic

properties of a porous rock on the compressibility of thepore fluid, Geophysics 40 (1975) 608–616.

[30] D. Mainprice, M. Humbert, Methods of calculating petro-physical properties from lattice preferred orientation data,Surv. Geophys. 15 (1994) 575–592.

[31] M. Kumazawa, O.L. Anderson, Elastic moduli, pressurederivatives, and temperature derivatives of single-crystalolivine and single-crystal fosterite, J. Geophys. Res. 74(1969) 5961–5972.

[32] T. Murase, A.R. McBirney, Properties of some commonigneous rocks and their melts at high temperatures, Geol.Soc. Am. Bull. 84 (1973) 3563–3592.

[33] R.A. Dunn, D.R. Toomey, Seismological evidence for three-dimensional melt migration beneath the East-Pacific Rise,Nature 388 (1997) 259–262.

[34] H. Sato, I.S. Sacks, T. Murase, The use of laboratoryvelocity data for estimating temperature and partial meltfraction in the low-velocity zone: comparison with heatflow and electrical conductivity studies, J. Geophys. Res.94 (B5) (1989) 5689–5704.

melt topology and seismic anisotropy in mantle peridotites of the oman ophiolite

Documents