Microstructures and iron partitioning in (Mg,Fe)SiO3 perovskite-(Mg,Fe)O magnesiowüstite assemblages: An analytical transmission electron microscopy study

See discussions, stats, and author profiles for this publication at: https://www.researchgate.net/publication/248797662

Microstructures and iron partitioning in (Mg,Fe)SiO3 perovskite-(Mg,Fe)O magnesiowüstite assemblages: An analytical transmission electron microscopy study ARTICLE in JOURNAL OF GEOPHYSICAL RESEARCH ATMOSPHERES · JANUARY 1997 Impact Factor: 3.43 · DOI: 10.1029/96JB03188

CITATIONS

READS

59

12

5 AUTHORS, INCLUDING: Isabelle Martinez

Yanbin Wang

Paris Diderot University

University of Chicago

77 PUBLICATIONS 1,333 CITATIONS

291 PUBLICATIONS 4,657 CITATIONS

SEE PROFILE

SEE PROFILE

François Guyot

Robert C Liebermann

French National Centre for Scientific Research

Stony Brook University

280 PUBLICATIONS 5,792 CITATIONS

212 PUBLICATIONS 5,671 CITATIONS

SEE PROFILE

SEE PROFILE

Available from: Yanbin Wang Retrieved on: 03 February 2016

JOURNAL

OF GEOPHYSICAL

RESEARCH, VOL. 102, NO. B3, PAGES 5265-5280, MARCH

10, 1997

Microstructuresand iron partitioning in (Mg,Fe)SiO3 perovskite(Mg,Fe)O magnesiowfistiteassemblages: An analytical transmissionelectron microscopystudy Isabelle Martinez,• Yanbin Wang,2 FrancoisGuyot,3 and Robert C. Liebermann Center for High PressureResearchand Mineral PhysicsInstitute, State Universityof New York at StonyBrook

Jean-Claude

Doukhan

LaboratoireStructureet Propri6t6sde l'Etat Solide,URA 234, Universit6Scienceset Technologiesde Lille Villeneuve d'Ascq,France

Abstract. San Carlosolivine and its syntheticringwooditepolymorphhave been transformedto (Mg,Fe)SiO3-perovskite and magnesiowtistite at a pressureof 26 GPa in a 2000-t uniaxialsplit-sphereapparatus(USSA-2000) for temperaturesrangingfrom 700øC to 1600øCand run durationsat peak temperaturesof 0 min to 19 hours.The recovered sampleswere studiedby analyticaltransmissionelectronmicroscopyto determinethe evolutionof the microstructuresand the crystallographical relationshipsand iron partitioningbetweenthe coexistingphasesin theseassemblages. At 700øC,metastable olivine remaineduntransformedeven after 19 hours.In runs performed at 1000øCand 1200øC,ringwoodite,in a topotacticrelation with olivine,was identified even though olivine was used as startingmaterial. Our resultsindicate that ringwooditeis an intermediatephasein the olivine -• (Mg,Fe)SiO3-perovskite+ magnesiowtistite transformationin this temperaturerange.At or above1300øCthe transformationof olivineor ringwoodite(usedas the startingmaterial)into (Mg,Fe)SiO3-perovskite + magnesiowtistite was completein lessthan 10 min. The first microstructures that appear are eutectoid-likeas already describedby previousauthors.For longer run durationsthe microstructureconsistedmostlyof cylindricalmagnesiowtistite crystalsembeddedwithin large, twinned(Mg,Fe)SiO3-perovskite crystals.These observations suggestthat magnesiowtistite grainsare very unlikely to be interconnectedfor a wide range of possible bulk mantle compositions; magnesiowtistite will therefore play a relativelyminor role in determiningthe transportpropertiesof Earth's lower mantle.Analysesof (Mg,Fe)SiO3perovskiteand magnesiowtistite crystalsformed in the first stepsof the transformation showthat most of the iron-magnesiumpartitioningis completedwithin the first minutesof the reaction and that subsequentlyonly isochemicalgrain growth of the two phasesoccurs. The iron-magnesium distributionbetween(Mg,Fe)SiO3-perovskite (pv) and

magnesiowtistite (mw),characterized byKd_Fe= (Fe/Mg)mw/(Fe/Mg)p v wasprecisely measuredby analyticaltransmissionelectronmicroscopyin equilibratedruns and found to be Kd_Fe= 3.8(3) at 1300øCandKd_Fe= 4.3(4) at 1600øC. Introduction

The microstructuresof (Mg,Fe)SiO3-perovskite(pv) (Mg,Fe)O magnesiowiistite (mw) phase assemblages are of major importance in determining the transport properties (e.g., rheology,electricalconductivity)in the lower mantle of the Earth. Although numerousdetailed studiesof the microstructuresdevelopedduringthe olivine(a phase)to wadsleyite (/3 phase)or ringwoodite(3/phase)transformations havebeen •Now at Laboratoirede G6ochimiedes IsotopesStables,IPGP, Institut de Physiquedu Globe de Paris.

2Nowat Consortium for AdvanceRadiationSources, Universityof Chicago,Chicago,Illinois.

3Nowat Laboratoirede Mineralogie-Cristallographie and Institut de Physiquedu Globe de Paris. Copyright1997 by the American GeophysicalUnion. Paper number96JB03188. 0148-0227/97/96JB- 03188509.00

reported [e.g.,Brearleyand Rubie, 1994;Brearleyet al., 1992; Guyotet al., 1991;Burnleyand Green,1989],very few studies have been devoted to microstructuresdevelopedduring the breakdownof (Mg,Fe)2SiO4into (Mg,Fe)SiO3-perovskite and magnesiowiistite [e.g.,Poirieret al., 1986;Madon et al., 1989; Ito and Sato, 1991; Wang, 1991]. Moreover, the previousexperimental and theoretical studiesdid not addressthe influenceof suchmicrostructureson the transportpropertiesof this dominantphaseassemblagein the lower mantle. The iron contents of the two main phasesin the lower mantle assemblagealsostronglyinfluencethe transportproperties. The partitioning of iron and magnesiumbetween the coexisting(Mg,Fe)SiO3-perovskiteand magnesiowiistiteis equallyimportantfor constrainingphaserelationsin pressuretemperature-compositionspace. Earlier partitioning studies were mostlybasedon the specificvolumesof (Mg,Fe)SiO3perovskiteand magnesiowiistite measuredwith X ray diffraction [e.g.,Yagiet al., 1979;Ito and Yamada,1982].Both Bell et al. [1979] and Ito and Yamada [1982] have reported large

5265

5266

Table

MARTINEZ

1.

ET AL.:

PEROVSKITE-MAGNESIOWUSTITE

ASSEMBLAGES

Run, Temperature,Run Duration, StartingMaterial, and Productsfor Each Experiment

Run

Temperature,* øC

2051

700

Total Time Above 700øC

SoakTime at Peak T

Products

Starting Material

0 min

10 min

O1 - one singlecrystal+ P

2144 2272 2292

700 1000 1000

0 min 12 min 1.5 hours

19 hours 10 min 1.5 hours

O1O1O1 -

O1 - one singlecrystal

1579

1200

7 min

0 min

2180

1200

7 min

0 min

O1-

21517

1300

15 min

10 min

P?

1351 2173 2079 2090 1052

1300 1300 1300 1600 1600

15 min 1.5 hours 14 hours 10 min 1.75 hours

2 min 1.5 hours 14 hours 2 min 1.5 hours

O1 O1 O1 O1 O1

-

X + P X + P X + P

deformed deformed

olivine olivine

X + P

olivine + •/phase olivine + •/phase + pv + mw •/phase + pv + mw •/phase

X X X X X

pv + pv+ pv+ pv + pv+ pv+

+ + + + +

P P P P P

mw mw mw mw mw mw

All runswere performedat P = 26 GPa. O1, San Carlosolivine;X, crystals(-100-200 •m); P, powder(-10-20 •m); pv, perovskite;mw, magnesiowfistite. *Nominal temperature_+75øC(also see Figure 2). ?Startingmaterial is •/phase synthesized from the sameSan Carlosolivine.

gd_Fevaluesbetween12 and 16, with gd_Fe -- (Fe/Mg)mw/ (Fe/Mg)pv. Electron microprobeanalyseswere employedin later studies[e.g.,Ito et al., 1984;Ito and Takahashi,1989;Fei et al., 1991]. Ito et al. [1984] and Ito and Takahashi[1989] concludedthat gd_Fe increasessignificantlywith the total Fe content in the olivine startingmaterial. However, in a laserheated diamondanvil cell study,Fei et al. [1991] did not observesuchchangesin gd_Fefor significantvariationsof the Fe content of the olivine startingmaterial. As the breakdownreaction of olivine to (Mg,Fe)SiO3perovskite+ magnesiowfistite producesa fine-grainedtexture, it is difficultto determinethe compositions of individualgrains (generallyabout 1 •m in size) using electron microprobe. Guyot et al. [1988] studiedthis reaction in a laser-heated, diamondanvil cell and analyzedthe recoveredsamplesusing analyticaltransmission electronmicroscopy(ATEM); typical values for gd_Fe around 3 to 6 were obtained, which are in agreementwith the studyby Kessonand Fitz Gerald [1991]. More recently,Kanamotoet al. [1992]studiedthispartitioning behaviorusingATEM and suggested that gd_Femay decrease with increasingtemperature(from 4 to 3.2 between1000and 1300øC).The ATEM techniqueallowsanalysison a scalewell below 1 •m, thereby minimizing uncertaintiesdue to grain overlap.However, one of the most fundamentalquestionsin studyingelemental partitioning behavior is whether the run producthas reachedthermodynamicequilibrium.This question had not been addressedin any of the previousstudies.As we discussbelow,the recentlypublishedstudyof Katsuraand Ito [1996]has addressedthis issuewith the use of catalyst. In this paper we report our ATEM studieson specimensof (Mg,Fe)SiO3-perovskite and magnesiowfistitesynthesized from San Carlosolivineand its ringwooditepolymorphat high pressuresand temperaturesin a multianvil apparatus.These studiesreveal important changesin the microstructuresas a functionof temperatureand run duration.The effectsof temperature and kineticson the Fe partitioninghave also been investigated.

was used as the startingmaterial in most of the experiments (seerun detailsin Table 1). Thisbimodaldistributionwasused to investigatethe influenceof grain size on the phasetransformation mechanisms.In examiningthe run products,however, it was difficult to detect differencesbetween regionsof contrastinggrain sizes.For two runs (runs 2051 and 1579), singlecrystalsof San Carlosolivinewere used,in order to allow structuralobservations at the scaleof the opticalmicroscope. The chemical compositionsof the olivine starting materials were determinedby electronmicroprobe.For runs 1052 and 1351, an averagecompositionof ((Mgo.o•Feo.o9)28iO4) was used (compositionA), whereasall the other runs were performedwith an averagecomposition of ((Mgo.aoFeo.•)2SiO4) (compositionB)). Thesetwo olivinescontained0.41 and 0.38 wt % NiO respectively.In one experiment(run 2151), the startingmaterial was ringwoodite(---10-20 •m grain size), previouslysynthesized in a separaterun at 21 GPa and 1300øC in a 10/4mm cell assembly[Gasparik,1990]usingthe sameSan Carlosolivine(compositionB). All of theseexperimentswere conductedin a 2000-tuniaxial split sphereapparatus(USSA 2000) in the StonyBrook High PressureLaboratoryusinga 7-mm MgO octahedralcell assembly (Figure 1) describedby Wanget al. [1992]. The room temperaturepressurecalibrationis givenby Liebermannand Wang[1992](typeA systemin their Figure7). The influenceof temperatureon this pressurecalibrationwas checkedby reversingthe ilmenite-perovskitephasetransitionin MgSiO3 at 1000øCand comparingwith the phase diagram of Gasparik [1990],who useda differentcell assembly.We found that the temperaturehad only a negligibleeffecton the pressurecalibration and concludethat the pressureestimatesof 26 GPa usedin this studyhave uncertaintiesof 5-10% (see alsodiscussions byLiebermann and Wang[1992]andGasparik[1990]). No thermocouplewas usedin theseexperiments;temperatureswere determinedfrom power-temperaturerelationspreviouslyestablishedin runsperformedin this samecell under the sameoil pressureusingW3%Re/W25%Re thermocouples (see differentsymbolsin Figure 2). From the scatterof the temperaturemeasurementsshownin Figure 2, we infer that Experimental Procedure the actual temperature and that estimated from the powerHigh-Pressure Experiments temperaturerelation do not differ by more than _+75øC.This 150øCscattercharacterizesthe reproducibilityof the temperNatural San Carlos olivine, ground into powder (•10-20 /•m in size)and mixedwith coarsecrystals(•200/•m in size), ature-powercalibrationdata.

MARTINEZ

ET AL.: PEROVSKITE-MAGNESIOWUSTITE

ASSEMBLAGES

5267

TEMPERATURE 4O0

600

1200

(K)

1600

2000

2400

Sample

Rhenium

10

I mm

MgO

LaCrO 3

Pyrophy!lite

WC

Figure 1. Schematiccrosssectionof the cell assemblyused in theseexperiments.The pressuremedium is a semisintered MgO octahedron(7 mm edgelength),containinga cylindrical rhenium heater surroundedby a lanthanumchromite (LaCrO3) sleeve.Electricalcontactis ensuredby tungstencarbide (WC) end plugssurroundedby a pyrophyllitesleeve.The rhenium heater also servesas a capsulefor the startingmaterial.

2o

25

• 700øC 1000øC

0 min

Pv +Mw 1300øC

1600øC

Sampleswere first compressedto 26 GPa at room temper10rain 10rain 2 rain 2 rain 19h 1.5h 1.5h 1.5h ature, and then heated to the desiredtemperature.We did not 3o 14 h sinterthe olivineat low pressureand hightemperaturebecause that procedurewould affectboth the pressureefficiencyof the Figure 3. Phasediagramof (Mgo.9Feo. 1)Si204 afterAkaogiet cell assemblyand the validity of the pressurecalibration.This al. [1989].The P-T path (arrow) and the experimentalconditions for each run are superimposedon this diagram.All the

experiments above1000øC arewellwithinthestability fieldof 2500

2000

1500

1000

(Mg,Fe)SiO3-perovskite+ magnesiowtistite. Experimentsat 700øCmay be closeto the boundary.

wasan additionalmotivationfor usinga startingmaterialwith a bimodal grain size distributionin most experiments.About 10 to 30 min were requiredto reach the desiredtemperature; soakingtimesat peakT (run duration)variedfrom 0 min to 19 hours,after which the sampleswere quenchedby shuttingoff the electricalpower. Generally, temperature decreasedvery rapidlyto below500øCwithin a few seconds.Pressurewasthen releasedslowlyover a period of 35-40 hours.The experimental run conditionsare summarizedin Table 1 and Figure 3. Becauseof its relevanceto kinetics,the total time spentin each run above700øCis also givenin Table 1. Analytical Transmission Electron Microscopy

500

Standard 30-/am-thick thin sectionswere made from the recoveredsamples,usinga room temperatureepoxyto avoid amorphizationdue to sampleheating.Thesethin sectionswere then transferredonto a coppergrid and ion-milled under lowenergybeam conditions:2.5 keV and 0.25 mA [Wanget al., 0 100 200 300 400 1992].This requiresvery long thinningdurationsof about 150 Power (Watts) to 200 hours.Transmission electronmicroscopy (TEM) observationswere performedat StonyBrook (New York) usinga Figure 2. Thermocouple reading as a function of electric JEOL TM200 CX transmission electronmicroscope operatingat powerat 26 GPa, usingthe cell assemblydescribedin Figure 1, 200 kV under low current and with rapid operation. in sevendifferentexperiments(eachrepresentedby a different X ray microanalyseswere performed in the TEM mode on symbol).The relationderivedfromthisfigureis T (øC) = 19 + 4.0145W+ 2.5811 x 10-3 W2 (W, powerin watts)witha the JEOL TM200 CX, equippedwith a Si(Li) energydispersive correlation coefficient r2 = 1. The temperature-power cali- system(EDS) and analyticalsoftwareDTSA (DesktopSpecbrationis reproduciblewithin 150øC,asshownby the scatterof trum Analyzer) from National Institute for Standardsand the data points. Technology(NIST). Analysisof elementslighterthan fluorine

5268

MARTINEZ

!

ET AL.: PEROVSKITE-MAGNESIOWUSTITE

i

San

Carlos

olivine

0.8

0.6

0.4

0.2

0

50

100

150

200

Fe+Mg+Si/acquisition

250

time #1052

1600C,1.5h 1.2

-- •- - Magnesiowustites -- c> - Perovskites

1.0

0.8

0.6 0.4

0.2

0.0

• ....

0

I ....

50

I ....

100

I ....

150

I,,,,I,,,,I,,,,I

200

250

Fe+Mg+Si/acquisition

Figure 4.

300

350

time

Fe/Mg count number ratio as a function of a

parameter scaled tosample thickness (ranging between 300• and4000•) for(a) a standard SanCarlos olivine: thezero-

ASSEMBLAGES

Cappellen[1990]. This is illustratedfor olivine on Figure 4a, whichshowsa positivecorrelationbetweenthe Fe/Mg ratio (in count number) and this thicknessindex. The procedurefor analyzing(Mg,Fe)SiO3-perovskite andmagnesiowtistites in the samplesis shownin Figure4b for the specificcaseof run 1052. With such an analyticalprocedure,it is necessaryto collect numerousdata points for the two phasesin order to obtain individualphasecompositionsand thuspartition coefficients. Note that the thicknesseffects and therefore the analytical errorsare more important in the magnesiowOstites, which are moreiron-rich,than in the (Mg,Fe)SiO3-perovskites. Analyseswith quantitative oxygendetermination. In order to improve analyticalprecision,a systemcapableof oxygen detectionwas used. Most of these analyseswere performed with a 3-nm probe size, which representsa compromisebetweengoodspatialresolutionandstabilityof the (Mg,Fe)SiO3perovskite and magnesiowtistitegrains under the electron beam.The absenceof volatilizationwasverifiedby performing a seriesof analysesfor a givenpoint and checkingthe constancyof elementalratioswith time. The main differencewith the analysesdescribedpreviouslyis that oxygencouldbe detectedand thusanalysed.The kx/si factorsusedwere experimentally determined using syntheticsilicate or silicide thin specimensof known and homogeneouscompositionby the procedureusedin the previoussection[Van Cappellen,1990]. For the absorption correction, we used a method recently developedby Van CappellenandDoukhan[1994]basedon the electroneutralityconditionin the analysedphase:the oxygen atomicconcentrationis equal to the sum of the atomic concentrations of the carlonsmultipliedby theirrespective valencies. As a first approximation,we assumedthat all of the iron was divalent in these samples.Deviations from this assumption

thicknessvalueof Fe/Mg allowsusto calibratethe K factor and couldcomefrom the nonnegligible amountsof Fe3+ which (b) run 1052;thicknesseffectsare more importantin magnemightbe presentin (Mg,Fe)SiO3-perovskite and perhapsalso siowtistitethan in (Mg,Fe)SiO3-perovskite, as expectedtheomagnesiowtistite [e.g., McCammon, 1993, 1996; Fei et al., retically. 1996].The samplethicknesswasadjusteduntil electroneutrality wasreached,and a quantitativeanalysiswasthusobtained. Preciseanalysesof crystalsof (Mg,Fe)SiO3-perovskiteand in direct contactwere thuspossible,whereas was impossiblewith this systemdue to the berylliumwindow magnesiowtistite protectingthe Si(Li) crystal.In orderto crosscheck the results, numerousdata pointswere requiredto obtainone composition withoutoxygendetection(e.g.,Figure4b). The we alsoperformedanalysesin the scanningTEM mode on a with analyses PhilipsTMCM30 in Lille (France) operatingat 300 kV. EDS superiorityof this latter methoddemonstratesthe importance analyses were performedusinga Ge detectorwith an ultrathin of quantitativeoxygendeterminationsfor preciselyanalyzing mylar window and a TracorTMVoyager system.Quantitative oxide materials usingATEM. analysesof oxygenwere possiblewith this latter system,allowing improvedanalyticalprecisionas discussed below. Analyses without quantitative Three

standard materials

oxygen determination.

were used to calibrate

Microstructural

Observations

the K factors

usedin the energy-dispersive microanalyses: diopside,magnesiowtistitewith Fe/(Fe + Mg) of 0.17, and the San Carlos olivine used as the startingmaterial in theseexperiments.All of the standardswere analyzedby electron microprobeand were chemically homogeneous.The elemental ratios were measuredfor differentthicknesses and were then extrapolated to zero thicknessin order to extractthe experimentalK factors. Instead of performing direct thicknessmeasurements,which suffer from rather large uncertaintieswhen using common procedures(especiallyon suchsamplesasthosestudiedhere),

The 700øC Experiments

Two experimentswere performedat 700øC(runs2051 and 2144). The startingmaterial for run 2051 consistedof one singlecrystal,surroundedby powderedolivine. No transformation was observed in either run even after 19 hours, indi-

catingthat the transformationkineticsare slow at 700øC.In sample2051 (10 min), the largesinglecrystalshoweda mosaic domain structure,with variable high densitiesof dislocations within the domains.The 19 hours sampleexhibiteda rather differentmicrostructure,probablybecauseof the differencein we used an indirect thickness index. This index is the total startingmaterials(Table 1). The averagegrain size was recount number in the spectrum(i.e., cumulativebackground- ducedafter the experimentto about200-400 nm, mostlikely Therefore we concorrectedpeak intensitiesfor each spectrum)dividedby the due to grain crushingduring compression. live acquisitiontime, which is a continuouslyincreasingfunc- cluded that at 26 GPa, no transformation of olivine could be tion of thickness,accordingto the methoddevelopedby Van activatedat 700øCwithin the durationof our experiments.

MARTINEZ

ß

ET AL.'

PEROVSKITE-MAGNESIOWUSTITE

ASSEMBLAGES

5269

.

.... ========================== 4: ............... •:...:'..:•w}-: ...... ,,

:-•:.;'..

.•:.. ;. .,

..

.

:. ß ß

i'

..

.

....

:'-':'•:•,:'ii' -!..•.i ......... ..... '... .

.. .... .

..:

•..:;; .....O...-.•..:;:...}11 • ...... , ,.... ß ;;;.;: '•%:...-.. . .....-. .... . .... "-':' 5;•;•;.:'.....}--'" ..'•":::"'":'•:• ":""•-•'•.. 2•............... x................ • '"".."-. ' ' . . .

.......::::•: ::::•%.:::?: .....

....• . .?:}--

...... .:

}.::..:....

a

Figure 5. (a) TEM micrographof sample2272. Thin lamellae of ringwooditedevelopingparallel to the olivine{ 100} plane.(b) The electrondiffractionpattern,showingthat a and •/phasesare in topotacticrelation

with (010)aII(•-2)• and(100)aII(•)•;

olivineis indexedusingthe Pbnmspacegroup.

The 1000øCExperiments

Sample2272 (10 min) consists primarilyof fine intergrowths of ringwooditeand olivine,with well-definedtopotaxialrelations (Figure 5). Lamellaeof ringwooditedevelopparallel to the olivine{ 100} plane (hkl in olivineindexedwith the Pbnm

mellaeare alsoobserved(Figures6a and 6b). The sametopotactic relationsbetween olivine and ringwooditeare observed as thosein sample2272. Topotacticrelationshave alsobeen observedbetween ringwoodite,(Mg,Fe)SiOa-perovskiteand

magnesiowiistite crystals: (100)vII(100)mw, (010)-V II(010)mw,

II(110)cpv, and(010)mw II(1-10)cpv (Figure 6c)[Wang spacegroup)withthefollowingorientation relations: (010)aII (100)mw sys(11-2)%(100)aII(***)v and(001)aII(-**0)V (Figure5b).The et al., 1995, 1996],where cpv standsfor the pseudo-cubic

remainingolivineis highlydeformed,asdemonstratedin zones tem of indexingthe orthorhombicperovskitestructure. with high densitiesof dislocations.The ringwooditelamellae exhibitpervasive(111) twinning;the averagesize of the twin The 1200øCExperiments domains is of the order of 50-100 nm. Such defects have Run 1579 (7 min) was performedusinga singlecrystalof alreadybeen reported in severalspinel-structurecompounds San Carlos olivine as the startingmaterial. An optical micrographof a thin sectionmadefrom the sampleshowsa striking (e.g., Vaughanand Kohlstedt[1981]in Mg2GeO4). on In sample2292 (1.5 hours),ringwooditeis the major phase, lameIlar texture (Plates la and b). Raman spectroscopy but residual olivine and some embryos of (Mg,Fe)SiO3- theselamellae indicate that they consistalternativelyof ringperovskiteand magnesiowiistite in very thin intercalatedla- woodite (dark brown lamellaeA in Plate lb) and (Mg,Fe)

5270

MARTINEZ

a

ET AL.: PEROVSKITE-MAGNESIOWUSTITE

ASSEMBLAGES

b

Figure 6. (a) TEM micrographof sample2292. Embryosof fine intergrowthsof (Mg,Fe)SiO3-perovskite and magnesiowiistite (L) are seenwithin ringwooditegrains,whichdisplaynumerousmoir6 patterns(not to be confusedwith the lamellaeL). (b) Enlargedviewof the (Mg,Fe)SiO3-perovskite-magnesiowiistite lamellae. (c) Diffractionpattern of coexistingringwooditeand magnesiowiistite. Additional spotsfrom neighboring ringwoodite,magnesiowiistite, and (Mg,Fe)SiO3-perovskite grainsare alsopresent.

SiO3-perovskite+ magnesiowiistite assemblages (light brown related to the short run duration. Indeed, the TEM observaand white lamellaeB in Plate lb); the Raman spectraof ring- tions show that the (Mg,Fe)SiO3 perovskitegrainsare very wooditeand (Mg,Fe)SiO3-perovskite are shownin Figures7a small(lessthan 0.1/.rm) in this sample[Wanget al., 1996]. and 7b, respectively. The spectrumof (Mg,Fe)SiO3-perovskite Run 2180wasperformedunderthe sameconditionsbutwith showsanomalouslybroad bands,when comparedwith "nor- a mixtureof crystalsand powderas a startingmaterialinstead mal" MgSiO3or (Mg,Fe)SiO3-perovskite Raman spectra[e.g., of a singlecrystal(Table 1). The TEM observations on this Durben and Wolf, 1992]; this might be due either to some sampleshowthat it consists mainlyof largegrainsof ringwooddisorder or to the small grain size, both of which may be ite, whichcontainnumerousstackingfaultson { 110} planesas

MARTINEZ

ET AL.: PEROVSKITE-MAGNESIOWUSTITE

ASSEMBLAGES

A

g

...

B

Plate 1. (a) Opticalmicrographof sample1579,observedin naturallight showinga lamellartexture.(b) a detailedviewof thistextureshowingintercalationof darkbrownlamellaecontainingringwoodite(A) andlight brownand white lamellae(B) containing(Mg,Fe)SiO3-perovskite and magnesiowiistite.

5271

5272

MARTINEZ

ET AL.: PEROVSKITE-MAGNESIOWUSTITE

Ringwoodite 700

-

600

5OO

400

300

2OO

2OO

a

400

600

800

1000

1200

1400

Frequency cm' Perovskite 700 650

•

600

'7,

•

550

•-•

500

ASSEMBLAGES

2079), grain growthoccurs.Figure 10a showsa typicalmicrostructurein sample2079,in whichthe (Mg,Fe)SiO3-perovskite grains typicallyexceed50 /•m. Magnesiowfistitegrains are completelyembeddedwithinthe (Mg,Fe)SiO3-perovskite. The (Mg,Fe)SiO3-perovskite grainscontaina ratherhighdensityof twin domains,whosecrystallographic relationsare identicalto those observedby Wanget al. [1992]. Amorphizationof the (Mg,Fe)SiO3-perovskiteusually starts at the domain walls whichremainvisibleas"ghosts"after completeamorphization. The magnesiowfistite grainsare predominantlyellipsoidalin shape;no preferredorientationof these ellipsoidshas been observed.These featuresare interpreted as being crosssections of roughlycylindricalmagnesiowiistites (with aspectratioscloseto 1), whichdevelopasthe resultof coarsening of the microstructureobservedin the short run (Figure 9). In some cases,the magnesiowiistite grainsare locatedat the intersection of orthogonal(Mg,Fe)SiO3-perovskite twinplanes(Figure 10b).This microstructure mightbe due to an elasticmismatch betweenthe two phases,with twin boundariesproducedon decompression. In sample2079 the (Mg,Fe)SiO3-perovskite phasewasmore stableunderthe electronbeamwhichallowed us to performmore detailedstudies;this is particularlyinterestingin view of the well-knowntendencyof silicateperovskites recoveredfrom high pressuresto amorphiseunder the electronbeam [e.g., Wanget al., 1992]. The followingorientation relationswere observedbetween(Mg,Fe)SiO3-perovskite

andmagnesiowiistite: (100)m wII(110)cpv, (010)mw II(1-10)cpv and(001)n•w II(001)cpv' Theserelations areidentical to those

450

observedin sample2292. In order to verify the effectof oversteppingthe equilibrium phaseboundaries,one of the runswas performedusingringFrequency cm' woodite as the startingmaterial (run 2151). After 10 min the Figure 7. (a) Raman spectrumof the brownishlamellae(la- microstructureresembledthat from the longer runswith olithat (1) the final microstructuredoesnot debeledA in Plate lb), showingthat they consistof ringwoodite. vine, suggesting (b) Raman spectrumof the white lamellae(labeledB) which pend on the startingmaterial,and (2) transformationkinetics are composedof (Mg,Fe)SiO3-perovskite and magnesiowfis- appear to be faster than with the olivine startingmaterial. tite; magnesiowiistite has no first-orderRaman spectrumbut Theseobservations arenot surprisingsincewe notethat during was detectedby electronmicroprobeand ATEM. heatingabove700øCat pressuresof 26 GPa, the olivinestarting materialfirst transformsto metastableringwoodite,as obshownin Figure 8a (e.g., Vaughanand Kohlstedt[1981] in servedin the 1000øCand 1200øCsamples.Similarconclusions Mg2GeO4).The selectedarea diffractionpattern(insetin Fig- have been reached in an earlier diamond anvil cell study ure 8b) showsstreakingalongthe [110]* direction.This streak- [Madonet al., 1989]. ing is more pronouncedon the 111 diffractionspots. 400

b

200250300350400450500550600

The 1600øCExperiments

The 1300øCExperiments

Kineticsbecomeso much faster at this temperatureso that The transformationof olivine into (Mg,Fe)SiO3-perovs- within 2 min at peak temperature(sample2090), the reaction very similarto the 14 kite + magnesiowiistiteis complete at these conditions.No producthas acquireda microstructure trace of remnant olivine, or metastableringwoodite,is de- hours sampleat 1300øC(Figure 1l a). Longer run durations tected,evenin sample1351(2 min). The TEM micrographs of (1.5 hours,sample1052)producelargergrainsizeassemblages this sample show that the magnesiowiistitegrains are very but without modifications of the microstructures.However, in and elongated(Figure 9a). In someareas,they appearas discsor someareas,equantgrainsof both (Mg,Fe)SiO3-perovskite are observedwith 120ø grain boundariesbedistortedellipsoids,whichcan be interpretedas cross-sections magnesiowiistite whichappears of these cylindricalmagnesiowiistitegrains (Figure 9b); the tweenthem (Figure 11b). This microstructure, averagediameterof thesemagnesiowiistite "cylinders"is quite to be well-equilibrated,is not representativeof the entiresamto (Mg,Fe)SiO3small, of the order of 200 nm, and they are elongatedalong ple as the volumeratio of magnesiowiistite highin suchareas.We thusconclude differentdirectionssuchas [111] and [100]. In other areas,the perovskiteis anomalously cross sectionsare rather rectangular, the faces being most that equilibriummicrostructureof the run product is better often {100}. The (Mg,Fe)SiO3-perovskite crystalshavebeen representedby that shownin Figures10a or 11a. amorphised by the electronbeamanddo not showanycontrast on the image.Thismicrostructure is similarto that producedin Analytical Results the laser-heateddiamondanvil cell [Poirieret al., 1986;Madon et al., 1989]. The analyticalresultsare reportedin Tables2 and3 in terms In When the run duration is increasedup to 14 hours (run of Fe/Mg ratiosand Ni/Mg and Mn/Mg ratios,respectively.

MARTINEZ

ET AL.' PEROVSKITE-MAGNESIOWUSTITE

ASSEMBLAGES

5273

a

b

Figure 8. TEM micrograph of sample2180showing a grainof ringwoodite withnumerous stacking faultson {110} planes.(a) Three familiesof planardefectson equivalent{110} planesat approximately 120ø are distinguishable on the image.(b) Streakingalongthe [110]* directionis clearlyvisiblein the diffraction pattern,especially on the 111 diffractionspots(inset).

the procedureusingquantitativeoxygendetermination,analysesof individualgrainscouldbe usedwhen grain sizeswere large enough;compositions of (Mg,Fe)SiO3-perovskite and magnesiowtistite crystals in directcontactaregivenin Tables2 and 3, along with their analyticalstandarddeviationsfrom whichwe estimateerror bars on the calculatedK d values.

Iron-Magnesium Partitioning

Valuesof gd_Fededucedfrom the analyses of the (Mg,Fe) SiO3-perovskites and magnesiowiistites in eachsampleare reportedin Table 2. Analysesperformedwith oxygendeterminationare from grainsin directcontact.However,when the

5274

MARTINEZ

ET AL.: PEROVSKITE-MAGNESIOWUSTITE

ASSEMBLAGES

ix-•-•':'•-•:•.9 0.0005(5) >9 0.0005(5) >8 na 0.0005(5) >8 0.0006(3) >9 na 0.0006(3) >8

(Mn/Mg)p v

0.0022(2) na 0.0022(2) 0.0022(3) 0.0022(2) 0.0022(2) 0.0022(2) 0.0022(2)

0.0022(5) na na 0.0021(4) 0.0016(4) 0.0027(4) 0.0021(4) 0.0027(4)

0.0014(3) 0.0019(2) 0.0017(3) 0.0022(3) 0.0019(3)

0.0017(4) 0.0017(4) 0.0021(4) 0.0021(4) 0.0021(4)

>8*

0.0089(4) 0.0085(3) 0.0081(4) 0.0084(6) 0.0091(4)

Sample1052 (1600øC,1.5 hours) 0.0011(3) >6 0.0011(3) >6 na 0.0011(4) >5 0.0011(4) >6 >5*

gd_Mn 1.1(4) 1.1(4) O.8(2) 1.1(3) 0.8(2) 1.1(3)*

0.9(4) 1.2(4) 0.9(4) 1.1(4) 1.0(4) 1.0(8)*

Ratiosin atomicpercent.Abbreviationsare pv, perovskite;mw, magnesiowfistite; na, not analyzed.All analysesare obtainedusingoxygen determination.Numbersin parenthesesrepresentstandarddeviationson the analysesand uncertaintieson the K a. *Ka valuesobtainedby averagingresultsfrom individualmw-pvcouples.

eningof the materials.The first (Mg,Fe)SiO3-perovskite mag- likelyto be interconnected underequilibriumconditionsin the nesiowfistiteassemblages producedare alsovery fine-grained lowermantle.This conclusionis of greatimportancefor transwhich could induce superplasticbehavior in the subducted port propertiessuchas electricalconductivityand creepsince lithosphericplate. Since the coarseningof the grains occurs they dependstronglyon the spatialdistributionof the phases very rapidly, this weakeningcould be localizedin a relatively in the compositeassemblage. Our experiments stronglysuggest narrowphasetransitionarea, therebycausinglocal thickening that (Mg,Fe)SiO3-perovskite isthe interconnected phasein the of the plate. The geodynamicalimplicationsof such a rheo- lowermantlephaseassemblage andthereforeshoulddominate logical behavior of subductingslabshave been discussedre- both the creep and electricalpropertiesof the lower mantle. centlyby Karato [1995] and Karatoet al. [1995]. At 26 GPa and 1300-1600øC,iron-magnesium partitioning At 1300øCthe transformationof olivine into (Mg,Fe)SiO3- between magnesiowfistiteand (Mg,Fe)SiO3-perovskitehas perovskite + magnesiowfistiteis complete, even after very been observedto be closeto 4 (3.8(3) at 1300øC,4.3(4) at shortrun durations:no trace of remnantolivineor ringwoodite 1600øC)by preciseanalyticalTEM techniques.The range of is detected,evenafter a 2 min run (run 1351).We cannottell valuesreportedhere is comparablewith that reportedby Keswhetherthis ringwooditeformed metastablyas in the experi- sonand Fitz Gerald [1991,Table 2], by Kanamotoet al. [1992], mentsat 1000øCand 1200øC.It seemsindeedequallyplausible by Malavergneet al. [1995], and in the recentmultianvilexperthat direct transformation occurred from olivine to perovsiments of Katsuraand Ito [1996] using B203 catalyst.The kite + magnesiowfistiteat the higher temperature.The fabric averagevalues are lower than those measuredby Ito and of the (Mg,Fe)SiO3-perovskite-magnesiowiistite assemblages Takahashi[1989],gd_Fe = 6, and Fei et al. [1991],Kd_Fe= 8. which appearinitially, is alwaysstronglyanisotropicwith very Such discrepanciesmight reflect differencesin equilibrium elongatedgrains(e.g.,Figure9) and is similarto that observed conditions(multianvilversusdiamondanvil cells), analytical in diamondanvilcell experiments[Poirieret al., 1986;Madon et techniques(e.g., electronmicroprobeor X ray diffractionveral., 1989].After a few hours,however,the microstructure is less anisotropicand the magnesiowiistitecrystalsare completely susATEM) or the correctionsinvolvedin reducingthe anaembeddedwithin the (Mg,Fe)SiO3-perovskitegrains (Figure lyticaldata. Although precisedeterminationshavebeen made 10a). Microstructuresobservedin somepreviousmulti-anvil at two temperaturesin the presentstudy,no significanttempressexperiments[Ito and Sato, 1991] appear to be interme- perature effect could be detected.Experimentsat more extreme temperatures(e.g., 1900øC)are neededin order to dediate between those from our shorter and longer runs. Theoreticalstudieshave shownthat in an isotropictwo- termine if there is a temperature effect on this partition phaseassemblage, the minimumvolumefractionfor a phaseto coefficient.The role of minor elementssuchasA1 may alsobe be interconnectedis 20% [Cahn, 1966].Although the volume very important. The valuesof Kd_Feare measuredwith lessprecisionfrom fraction of magnesiowfistite in our samples(about 30%) is significantlyhigher than this theoreticalthreshold,we never the shortrunsbecauseof the smallergrain size,which results observedindicationsof interconnectedmagnesiowfistite grains in some Mg-devolatilizationunder the electron beam, as resurroundingperovskitecrystals[Cahn, 1966].As the composi- portedearlier by Kessonand Fitz Gerald[1991].However,the tion of the lower mantle is generallybelievedto be between valuesof Kd-Fe= 3.3(4) at 1300øC(run 1351,2 min) and3.9(3) differentfrom pyrolite [e.g., Ringwood, 1975] and pure (Mg,Fe)SiO3- at 1600øC(run 2090,2 min) are not significantly perovskite[e.g., Stixrudeet al., 1992], the volume fraction of those measuredin long runs. The time dependenceof the magnesiowfistite is expectedto be in the range of 0 to 20%. partition coefficientis thus very small: during the first few Thereforewe concludethat the magnesiowfistite phaseis un- minutesof the reaction,most of the partitioningoccursand

MARTINEZ

ET AL.: PEROVSKITE-MAGNESIOWUSTITE

subsequently, onlyisochemical graingrowthoccursin the (Mg, Fe)SiO3-perovskite-magnesiowfistite assemblage. Nickel is stronglypartitionedin magnesiowfistite (Ka_•i > 8 at 1300øCandKa_•i > 5 at 1600øC)in generalagreementwith the studiesby Kessonand Fitz Gerald[1991]andMalavergneet al. [1995].Manganeseis basicallyequipartitionedbetweenthe two phases.

ASSEMBLAGES

5279

shownin thispaper.Very usefuland carefulreviewswere made by the three reviewers: P. Burnley, A. Brearley, and Associate Editor D. Isaak.The high-pressure experimentswere conductedin the Stony Brook High PressureLaboratory,which is jointly supportedby the StateUniversityof New York at StonyBrookandthe NSF Scienceand TechnologyCenter for High PressureResearch(EAR 89-20239) (CHiPR is a National ScienceFoundationCenter for Scienceand Technology).This researchwasalsosupportedby a researchgrantto RCL (EAR 93-04502).This is MPI contribution182.

Conclusions

On the basisof the experimentaland analyticalresultsreportedin this study,we make to the followingconclusions: 1. At pressures 15 GPa aboveits equilibriumstabilityfield,

References

Akaogi, M., E. Ito, and A. Navrotsky,Olivine-modifiedspinel-spinel transitionsin the systemMg2SiO4-Fe2SiO4: calorimetricmeasurements,thermochemical calculation,and geophysical application,J. olivine remains untransformed at 700øC even after 19 hours Geophys.Res.,94, 15,671-15,685,1989. but transformswithin 2 min to the (Mg,Fe)SiO3-perovskite+ Bell, P.M., T. Yagi, and H.-K. Mao, Iron-magnesiumdistribution coefficientbetweenspinel [(Mg,Fe)2SiO4],magnesiowfistite [(Mgmagnesiowfistite phaseassemblage above1200øC. ,Fe)O], and perovskite[(Mg,Fe)SiO3], Year Book CarnegieInst. 2. At temperatures between1000øCand1200øC,ringwoodWashington,78, 618-621, 1979. ite is alwaysobservedas a metastable,intermediatephasein Brearley,A. J., and D.C. Rubie, Transformationmechanismsof San Carlosolivineto (Mg,Fe)2SiO4 D-phaseundersubduction zoneconthe transformation. Topotacticrelationsare systematically obditions,Phys.Earth Planet.Inter., 86, 45-67, 1994. served between olivine and ringwoodite and correspondto Brearley,A. J., D.C. Rubie, and E. Ito, Mechanismsof the transforthosepredictedby Poirier[1981]for a martensitictransformamationsbetweenthe a,/3 and 3/polymorphsof Mg2SiO4 at 15 GPa, tion mechanism. This is the first time that such relations have Phys.Chem.Miner., 18, 343-358, 1992. been clearlyobservedwith natural olivine as the startingma- Burnley,P. C., The effectof non-hydrostatic stresson the olivine-spinel transformationin Mg2GeO4, Ph.D. dissertation,Univ. of Calif., terial. It is interestingto note that under suchnonequilibrium

conditions,the martensitic mechanismoccurseven under the

Davis, 1990.

Burnley,P. C., The fate of olivine in subductingslabs:A reconnaisrelativelylow deviatoricstressconditionsof our experiments. sancestudy,Am. Mineral.,80, 1293-1301,1995. 3. The (Mg,Fe)SiO3-perovskite + magnesiowfistite assem- Burnley,P. C., andA. Brearley,High resolutiontransmission electron blagesform initially highly anisotropic(oriented-eutectoid- microscopyof martensiticallyproducedspinellamellae in olivine, like) microstructures and systematic crystallographic relations Eos Trans.AGU, 75(44), Fall Meet. Suppl.,696, 1994. existbetweenthe differentphases[seealso Wanget al., 1995, Burnley,P. C., and H. W. Green, Stressdependenceof the mechanism of the olivine-spineltransformation,Nature, 338, 735-756, 1989. 1996].The moststrikingobservation is that the denseplanesin Cahn, J. W., A model for connectivityin multiphasestructures,Acta Metall., 14, 477-480, 1966. the oxygenfcc sublattices are in continuitybetweenringwoodite and magnesiowfistite. Most of the iron-magnesium parti- Durben, D. J., and G. H. Wolf, High-temperaturebehaviorof metastableMgSiO3perovskite:A Raman spectroscopic study,Am. Mintioningbetweenthe two phasesoccursduringthe first stepsof eral., 77, 890-893, 1992. the transformation.The existenceof well-defined crystallo- Fei, Y., H.-K. Mao, and B. Mysen, Experimentaldeterminationof graphicalorientationrelationsaswell asthe extremereduction elementpartitioningand calculationof phaserelationsin the MgOFeO-SiO2systemat highpressureandhightemperature,J. Geophys. in grainsizesmighthaveimportantconsequences for the theRes., 96, 2157-2169, 1991. ologyof subductedslabsdeep in the mantle. Fei, Y., Y. Wang, and L. Finger, Maximumsolubilityof FeO in (Mg4. Upon further annealing,isochemicalgrain growth oc,Fe)SiO3perovskiteas a functionof temperatureat 26 GPa: Implicursand the (Mg,Fe)SiO3-perovskite + magnesiowfistite ascationsfor FeO contentin the lower mantle,J. Geophys.Res.,101, 11,525-11,530, 1996. semblages evolvetowarda microstructure consistingof cylindricalor ellipsoidalmagnesiowfistite grainsembeddedin large, Gasparik,T., Phaserelationsin the transitionzone,J. Geophys.Res., twinned(Mg,Fe)SiO3-perovskite crystals.These observations 95, 15,751-15,769, 1990. Green, H. W. II, T. E. Young, D. Walker, and C. H. Scholz,The effect suggestthat magnesiowfistite is unlikely to be an interconof non-hydrostatic stresson the a b and a g olivinephasetransfornectedphasein the lower mantle,and thusplaysa relatively mations.In High-Pressure Research: Applicationto Earth and PlanetarySciences, Geophys. Monogr.Set.,vol. 67, editedby Y. Syonoand minorrole in controllingtransportproperties(e.g.,theological M. H. Manghnani,pp. 229-235. AGU, Washington,D.C., 1992. and electricalproperties)in the deepEarth. F., M. Madon, J. Peyronneau,and J.P. Poirier, X-ray micro5. Iron-magnesiumpartitioningbetweenthe two phases Guyot, analysisof highpressure/high temperaturephasessynthesized from has been accuratelymeasuredin equilibratedsampleswith a natural olivine in a diamond-anvil cell, Earth Planet. Sci. Lett., 90, 52-64, 1988. relativelynew ATEM techniquebasedon quantitativeoxygen determination[Van Cappelenand Doukhan,1994].Values of Guyot,F., G. D. Gwanmesia,andR. C. Liebermann,An olivineto beta Res.Lett., iron-magnesiumpartitioningcoefficientsbetweenmagnesio- phasetransformationmechanismin Mg2SiO4,Geophys. 18, 89-92, 1991. wfistite and (Mg,Fe)SiO3-perovskite are Ka_• of 3.8(3) at Hyde, B. G., T. J. White, M. O'Keeffe,and A. W. S. Johnson,Struc1300øCand 4.3(4) at 1600øC.Within our analyticalprecision, tures related to those of spinel and the b-phase,and a possible mechanismfor the transformationolivine spinel,Z. Kristallogr., neither significanttemperaturenor time dependenceof this 160, 53-62, 1982. partitioningcoefficienthasbeenobserved.Finally,it hasbeen Ito, E., and H. Sato,Aseismicityin the lowermantleby superplasticity confirmedthat nickelis more stronglypartitionedthan iron in of the descendingslab,Nature, 351, 140-141, 1991. the magnesiow•istite phase,whereasmanganeseis basically Ito, E., and E. Takahashi,Postspineltransformationsin the system equipartitionedbetweenthe two phases. Mg2SiO4-Fe2SiO 4 and some geophysicalimplications,J. Geophys. Res., 94, 10,637-10,646, 1989.

Ito, E., and H. Yamada,Stabilityrelationsof silicatespinels,ilmenites, and perovskites,in High-Pressure Research,edited by M. H. MangAcknowledgments.We wishto thankR. J. Reeder for histechnical hnani and Y. Syono,pp. 405-419, Terrapub,Tokyo, 1982. adviceand discussions of the analyticalTEM results.PhilippeGillet is warmlyacknowledged for his help in obtainingthe Raman spectra Ito, E., E. Takahashji,andY. Matsui,The mineralogyandchemistryof

5280

MARTINEZ

ET AL.: PEROVSKITE-MAGNESIOWUSTITE

the lower mantle: An implicationof the ultrahigh-pressure phase relationsin the systemMgO-FeO-SiO2, Earth Planet. Sci. Lett., 67, 238-248, 1984.

Kanamoto,M., T. Irifune, K. Fujino, T. Yagi, and H. Yusa, Determination of element partitioning among high-pressurephases for (Mg,Fe)2SiO4olivinewith an analyticalelectronmicroscope, paper presentedat the 33rd High PressureConferenceof Japan, Kumamoto, Japan, Nov. 18-20, 1992. Karato, S., Interactionof chemicallystratifiedsubductedoceaniclithospherewith 660km discontinuity, Proc.Jpn.Acad.B, 71(7), 203-207, 1995.

ASSEMBLAGES

Stixrude,L., R. J. Hemley,Y. Fei, and H. K. Mao, Thermoelasticity of silicate perovskiteand magnesiowtistiteand stratificationof the Earth's mantle, Science,257, 1099-1101, 1992.

Sung,C. M., and R. G. Burns,Kineticsof highpressurephasetransformations:Implicationto the evolutionof the olivine-spinel transition in the downgoinglithosphereand its consequences on the dynamicsof the mantle, Tectonophysics, 31, 1-32, 1976. Van Cappellen,E., The parameterless correctionmethodin X-ray microanalysis, Microsc.Microanal. Microstruct.,1, 1-22, 1990. Van Cappellen, E., and J. C. Doukhan, Quantitative transmission X-ray microanalysis of ionic compounds,Ultramicroscopy, 53, 343-

349, 1994. Karato, S., S. Zhang, and H. Wenk, Superplasticityin Earth lower mantle:Evidencefrom seismicanisotropyand rockphysics,Science, Vaughan,P. J., and D. L. Kohlstedt,Cationstackingfaultsin magne270, 458-461, 1995. siumgermanatespinel,Phys.Chem.Miner., 7, 241-245, 1981. Katsura,T., and E. Ito, Determinationof Fe-Mg partitioningbetween Wang, Y., Electronmicroscopyand X-ray diffractionstudieson strucperovskiteand magnesiowtistite, Geophys.Res.Lett., 23, 2005-2008, tural phasetransitionsin (Mg,Fe)SiO3-perovskite, Ph.D. thesis,252 1996. pp., Univ. of N.Y. at Stony Brook, 1991. Kerschhofer,L., T. G. Sharp,and D.C. Rubie, Intracrystallinetrans- Wang, Y., F. Guyot, and R. C. Liebermann,Electronmicroscopy of formationof olivineto wadsleyiteand ringwooditeunder subduction (Mg,Fe)SiO3-perovskite: Evidencefor structuralphasetransitions zone conditions, Science,274, 79-81, 1996. and implicationsfor the lower mantle,J. Geophys. Res.,97, 12,32712,347, 1992. Kesson,S. E., and J. D. Fitz Gerald, Partitioningof MgO, FeO, NiO, MnO and Cr203 betweenmagnesiansilicateperovskiteand magne- Wang, Y., I. Martinez, F. Guyot, and R. C. Liebermann,The breaksiowtistite:Implicationsfor the origin of inclusionsin diamondand downof olivineto perovskite+ magnesiowtistite, Eos Trans.AGU, the compositionof the lower mantle, Earth Planet. Sci. Lett., 111, 76(46), Fall Meet. Suppl.,F618, 1995. 229-240, 1991. Wang, Y., I. Martinez, F. Guyot, and R. C. Liebermann, The breakLiebermann,R. C., and Y. Wang, Characterization of sampleenvirondownof olivineto perovskite+ magnesiowtistite, Science,in press, 1996. ment in a uniaxialsplit-sphereapparatus,in High Pressure Research: Applicationto Earth and PlanetarySciences, Geophys. Monogr.Ser., Yagi, T., H. K. Mao, and P.M. Bell, Latticeparametersand specific vol. 67, editedby Y. Syonoand M. H. Manghnani,pp. 19-31, AGU, volume for the (Mg,Fe)SiO3-perovskite phase of orthopyroxene Washington,D.C., 1992. composition,Year Book CarnegieInst. Washington,78, 612-614, 1979. Madon, M., F. Guyot, J. Peyronneau,and J.P. Poirier, Electron microscopyof high pressurephasessynthesizedfrom natural olivine in diamondanvil cell, Phys.Chem.Miner., 16, 320-330, 1989. J.-C. Doukhan,LaboratoireStructureet Propri6t6sde l'Etat Solide Malavergne,V., F. Guyot, J. Peyronneau,and J.P. Poirier, Partition- (URA 234),Universit6Sciences et Technologies de Lille, 59655Villeing of iron, cobalt and nickel betweenmineralsof the Earth's lower neuved'AscqCedex,France. mantle at high pressureand high temperature(in French), C. R. F. Guyot,Laboratoirede Mineralogie-Cristallographie, placeJus-

Acad. Sci., Ser. IIa, 320, 455-462, 1995.

sieu, 75252 Paris cedex 05, France.

McCammon,C. A., Effect of pressureon the compositionof the lower R. C. Liebermann,Center for High PressureResearch,State Unimantle endmemberFexO, Science,259, 66-68, 1993. versityof New York at StonyBrook, StonyBrook, NY 11794-2100. McCammon,C. A., Crystalchemistryof iron-containingperovskites, I. Martinez,Laboratoirede G6ochimiedesIsotopesStables,IPGP, Phase Transitions,58, 1-26, 1996. 4 place Jussieu,75252 Paris cedex05, France. (e-mail: martinez@ Poirier,J.P., Martensiticolivine-spinel transformation andplasticityof ipgp.jussieu.fr) the mantle transitionzone, in Anelasticityin theEarth, Geodyn.Ser., Y. Wang,Consortiumfor AdvanceRadiationSources, Universityof vol. 4, editedby F. D. Stacey,M. S. Paterson,and A. Nicolas,pp. Chicago,5640 SouthEllis Avenue,Chicago,IL 60637. 113-117, AGU, Washington,D.C., 1981. Poirier, J.P., J. Peyronneau,M. Madon, F. Guyot, and A. Revcolevschi, Eutectoidphasetransformationof olivineand spinelinto perovskite and rock salt structures,Nature, 321,603-605, 1986.

Ringwood,A. E., Compositionand Petrologyof the Earth's Mantle, McGraw-Hill, New York, 1975.

(ReceivedApril 24, 1996;revisedAugust16, 1996; acceptedOctober11, 1996.)