JOURNAL OF GEOPHYSICAL RESEARCH, VOL. 98, NO. Bll, PAGES 20,031-20,037, NOVEMBER 10, 1993
ElasticProperties of HydrogrossularGarnet and Implications for Water in the Upper Mantle
BRII)GET O%•EII J •
Departmentof Geologyand Geophysics,University of California,Berkeley
JAY D. BASS
Departmentof Geology,University of Illinois,Urbana
GEORGE R. ROSSMAN
Divisionof Geologicaland Planetary Sciences, California Institute of Technology,Pasadena
The single-crystalelastic properties of a hydroussilicate gamet, hibschite (Ca3A12(SiO4)l.72(H404)l.28), were measuredusing Brillouin spectroscopic techniques. The adiabaticbulk modulusof hibschite,K s = 99.8+1.0GPa, andthe shearmodulus, •t = 64.3_--/-0.5GPa, are 40% lowerthan the bulk and shearmoduli for anhydrousgrossular gamet Ca3A12(SiO4) 3. This increasedcompressibility of hydrogarnetis attributedto increasedhydrogen bonding with pressurein the H404 tetrahedron.Density considerations indicate that hydrogametis likely to be stablerelative to an assemblagewith H20 asa separatephase throughout the upper mantleand probably the transitionzone. Assuminggamet to be the solerepository for mantlewater, the seismicwave velocitiesof a "wet" eclogiticlayer are 6-8% lower thanthose of dry eclogite. A hydrated eclogiticlayer severaltimes thicker than the oceaniccrust would probably be requiredfor a water-rich regionof themantle to be seismologicallydetectable. Lesser quantities of mantlewater than those implied by the abovescenario may be invisibleto seismictechniques.
INTRODUCTION pyralspites,while in ugranditesit may rangeup to 20 wt % H20 [Rossmanand Aines,1991; Lager et al., 1989]. Some Throughoutthe evolution of the Earth it is possiblethat natural hydrogarnetsare estimatedto have equilibratedat water has been stored in the interior so that at present there pressuresof as much as •-6 GPa (•-180-km depth) and may be, inside our planet, severaltimes the amountof water as temperaturesto-•1300øC. Additionally, hydrous garnetshave is in the hydrosphere[Ahrens, 1989]. Where and in what form been synthesizedat temperaturesof 800-1200øC and pressures this water may be storedin the mantle are questionswhich may of 2-5 GPa [Ackermannet al., 19,83;Geiger et al., 1991]. be answeredby examininghydrous phases transported from the Therefore,because garnet is a substantialfraction of the upper mantle [Bell and Rossman, 1992a] and by considering the mantle, the potential of this phase as a water-bearing phaserelations and propertiesof high-pressurehydrous phases component of the Earth's interior warrants investigation. [Thompson,1992]. Among the candidate"wet" mantlephases Garnet is also relevantbecause of the possibilitythat the OH syntheticphase B has attracted attention becauseof its high containedin a variety of mantle silicatesis incorporatedvia density and the fact that it is stableto pressurescharacteristic the hydrogarnetsubstitution, H404 = SiO4. of the transitionzone [Ringwoodand Major, 1967;Akaogi and It haspreviously been arguedthat becausethe molar volume Akimoto, 1980]. However, as pointed out by Finger et al. increases with hydrogarnet substitution (-•20% at zero [1989], phase B has not yet been demonstratedto be a stable pressure),garnet would not be a favorablephase to hold water phasefor a likely bulk mantle composition. In additionto the at high pressures[Martin and Donnay, 1972]. Aines and high-pressurehydrous phases which have been synthesized, Rossman [1984] suggested,however, that if hydrogarnetis somenaturally occurring upper mantle minerals such as olivine more compressiblethan pure silicate (anhydrous) garnet, it and pyroxene are able to incorporateOH and are potential could exist as a hydrous mantle component.To addressthe storagesites for water [Bell and Rossman,1992a]. Serpentine question of hydrogarnet compressibility we have used has been shown to remain hydratedup to 25 GPa and 900 K Brillouin spectroscopyto determinethe elastic propertiesof a [Meade and Jeanloz, 1991]. Another mineral which is both single crystal of hydrogrossulargarnet at ambient conditions. petrologicallyrelevant to theupper mantle and transition Although pyrope, the magnesium end-member, maybe a more zone,and which has a hydratedform, is garnet. Aines and abundantgarnet component inthe upper mantle, we chose to Rossman[1984] and Bell andRossman [1992b] have examine a Ca-rich garnet because much more OH can be investigatedtheOH content ofmantle-derived garnetsand have incorporated intocalcic garnets: there is, in fact,complete foundhydrous components of 50-100 ppm (as H20) in solid solutionbetween the anhydrousend-member Ca3A12Si3012 (grossular) and a hydrous silica-free end- member, Ca3A12(H404)3(synthetic phase). By using samples Copyright1993 by the AmericanGeophysical Union. in this system, it should be possibleto clearly identify the Papernumber 93JB02005. effect of the H40 4 = SiO4 substitutionon the elasticproperties 0148-0227/93193JB-02005505.00 of garnets.
20,031 20,032 O'NEILL ET AL.: HYDROUSGARNET ELASTICITYAND WATER IN THE MANTLE
The hydrogrossularseries is describedby the formula Seven of the natural growth faces of the hibschitewere used Ca3A12(SiO4)3_x(H404)x. Hibschite is the mineralogicalname for the Brillouin measurements.Optical goniometry was used for garnetswith 0
RESULTS TABLE 1. ChemicalAnalysis of Hibschite1457
Element Atomic Ratio The hibschiteelastic moduli determinedfrom the velocity data set are given in Table 4. The uncertaintiesin Table 4 are 1 N a 0.00 standarddeviation calculated from the root-mean-squareerror in Mg 0.04 velocity of the entire data set for eachmodulus. Becauseof the A1 1.87 Si 1.72 inhomogeneityof the hydrous componentin the sample, Ca 2.84 which affectsthe moduli throughan uncertaintyin density,the Mn 0.00 accuracy of the moduli is estimated to be at the 2 standard Fe 0.03 deviationlevel. The adiabaticbulk modulusKs and the shear O* 12.00 H•' 5.14 modulus•t, also given in Table 4, were calculatedusing both Voigt-Reuss-Hill averaging and Hashin-Shtrikman bounds *Oxygen was analyzeddirectly. [Watt et al., 1976], with the results from the two methods •'Calculatedassuming (Si+H/4)=3.0. being in complete agreement. The elastic constantsfor the anhydrous end-member grossular [Bass, 1989] are given for an average of the three analyses. The lattice parameter, comparisonin Table 4, as well as the isothermalbulk modulus measuredby single-crystalXray diffractionusing a SyntexP21 K 0 of pure syntheticend-member katoite recentlymeasured by four-circle diffractometer,is given in Table 2, with the hydrous static compression [Olijnyk et al., 1991]. In comparing (grossular) and anhydrous (katoite) end-member values for hibschitewith grossular,a dramatic increaseis seen in all the comparison. Additionally, to interpret the Brillouin elastic moduli, such that shearingis affectedsimilarly to bulk measurementsit is necessaryto know the index of refraction. compression(see Figure 2). It shouldbe pointedout that the Values for nD and n514,measured by immersionoil techniques, elastic properties of silicate garnet have previously been are also given in Table 2. The katoite end-membervalues were observedto be remarkablyconstant with respectto changesin measuredusing a syntheticsample obtained from GeorgeLager composition[Bass, 1986]. To emphasizethe magnitudeof the of the University of Louisville. Although lattice parameter increasedcompressibility due to hydration,it can be compared, valuesfor the hydrogrossularseries (see Figure la) showsome for reference, with the difference in moduli between the low- scatter and the reported indices of refraction (Figure lb) and high-pressurepolymorphs of Mg2SiO4, olivine (Ks = 129 indicate a nonlinear trend (Figure lb), our measurementsare GPa, la = 81 GPa), and¾-spinel (K S = 184 GPa, la = 119 GPa). compatible with the average OH content of the sample,x = The changeof phase from olivine to ),-spineleffects a 30% 1.28. difference in moduli, whereas in hibschite less than 50%
TABLE 2. HydrogrossularCa3A12(SiO4)3_,•(H404),•Properties
Grossular* Hibschite Katoite
Latticeparameter, ]i 11.851 12.183(3) 12.561(6) Density,g/cm 3 3.594 3.029 2.54 Refractive index nD 1.734 1.661(3) 1.611(3) n514.5 1.670(3) 1.619(3)
The numbersin parenthesisare the uncertaintyin the lastdigit. *The propertiesof grossular(x = 0) arefrom Skinner [1956];properties of hibschite(0 < x < 1.5) andkatoite (1.5 < x < 3.0) arefrom this study (see text). O'NEILL ET AL.: HYDROUSGARNET EI,ASTICITYAND WATERIN THE MANTLE 20,033
a) HydrogrossularLattice Parameter Passaglia, 1985; Lager et al., 1987a, 1989; Armbruster and
12.6 Lager, 1989]. The structureof garnet is shown in Figure 3, with the numberingscheme of Novak and Gibbs [1971] given in Figure3a. Figures3b and 3c showtwo consecutivelayers of polyhedra in the unit cell separated by a/8 in the [001] 12.4 direction. The unit cell is constructedof eight layers similar to these which are alternated and rotated, with each plane of hibschite 1457 cations separatedby a/8. In hydrogarneta substitutionis made in the tetrahedronwhere 4H + replaceSi n+, with the hydrogens 12.2 being located outside and slightly above the face of the tetrahedron[Lager et al., 1989; Armbrusterand Lager, 1989]. With the SiOn= H40 4 substitutionthe O(1)-O(2) and O(1)-O(3) bonds lengthen substantially,from 2.56 Jt and 2.74 Jt, 12.0 respectively,in grossular[Hazen and Finger, 1978]to 3.07 Jt and 3.25 Jt in fully hydrated(x = 3.00) katoite[Lager et al., 1987a; Cohen-Addadet al., 1967; Bartl, 1969]. A studyof the vibrational spectraof hibschiteunder pressureby Knittle et al. 11.8 [1992] indicates that this enlarged tetrahedronis responsible 0.0 0.5 1.0 1.5 2.0 2.5 3.0 for the increased compressibility of hydrous garnets. The Grossular x OH Katoite Raman spectra of hibschite and grossular show that with pressureH40 4 tetrahedraare reducedin size relatively more than SiO 4 tetrahedra. This may be due to an observedincrease b) HydrogrossularIndex of Refraction with pressureof hydrogenbonding between hydroxyl groups and neighboring oxygens, which would help to compressthe 1.74 H40 4 tetrahedronin hibschite. Additionally, the resultsof Knittle et al. [1992] show that the cation polyhedra in 1.72 hibschite, including those tetrahedra containing silicon, have a similar response to pressure as is seen in grossular and 1.70 therefore do not contribute to the increasedcompressibility of hydrous garnet. 1.68 We can examine what effect the high compressibilityof this hydrated garnet phase would have on properties which are measurablein the mantle, such as density and soundvelocity, 1.66 and investigatethe possibleconsequences of water structurally bound in mantle silicates. First, to consider density, we use 1.64 the reaction hibschite 1457 1.62 Ca3A12(SiO4)3+2xH20= Ca3A12(SiO4)3_x[(OH)4]x+xSiO 2 (1)
1.60 with x = 1.28 for the hibschite in this case. The Birch- 0.0 0.5 1.0 1.5 2.0 2.5 3.0 Murnaghanequation of state was used to calculatethe 300 K Grossular x OH Katoite isothermal densities through the upper mantle for the assemblageson either side of the equation,grossular plus H20 Fig. 1. (a) Lattice parametersfor the hydrogrossularseries. Data are (ice VII) and hibschiteplus stishovite(see Figure 4). Olijnyk from Flint et al. [1941], Cohen-Addad et al. [1967], Hazen and Finger [1978], Bassoet al. [ 1983],Passaglia and Rinaldi [1984], Sacerdotiand et al. [1991] measured the pressurederivatives of the bulk Passaglia [1985], Lager et al. [1987a],Lager et al. [1987b],Bass [1989], moduli K 0' for grossularand the hydrousend-member karoire, Rinaldi and Passaglia[1989], Olijnyk et al. [1991],Knittie et al. [1992], as 6.1 and 4.1, respectively. These values were usedto bracket this study;with data of Pabst [1942], Tilley [1957], Frankel [1959], the density of hibschiteplus stishovite,with K0' for hibschite Nalivkina [1960], and Bloxam [1964] from tables of Passaglia and assumed to be between the end-member values of 6.1 and 4.1. Rinaldi [1984]. (b) Indexof refractionn D for the hydrogrossularseries. Data are from Flint et al. [ 1941], Skinner[ 1956], Passagliaand Rinaldi As seen in Figure 4, the assemblagewith structurallybound [1984] and this study;with data of Belyankin and Petrov [1941], Pabst H20 (hibschiteplus stishovite,shown bracketedby dashed [1942], Hutton [1943], Frankel [1959], Nalivkina [1960], and Gross lines) is denserthan the assemblagewith free H20 as a separate [1977] from tablesof Passagliaand Rinaldi [1984]. phaseto at least the top of the transitionzone (-q3 GPa). In the temperaturerange of the uppermantle (-q 600-1750 K from 3-13 GPa) the relative densities of the mineral phases hydrationof grossularcauses a 40% changein the moduli. A 60% difference in moduli is seen between the completely wouldbe affectedby a trade-offbetween thermal expansion and increased compressibility. Perhaps the most dramatic hydratedkato ite and grossular. differencewould be shownby H20, which would be a liquid DISCUSSION phaseat mantle temperatures.The densityof water at 1273 K has been calculatedto 15 GPa [Hill, 1990], where it is --2% less The high compressibilityof hibschitecan be understoodin densethan our calculateddensity for ice VII at 300 K and the terms of the hydrogarnet crystal structure, which has been samepressure. Thereforewe do not expectthe outcomeof our studiedin detail by neutronand xray diffraction[Sacerdoti and calculationsto be significantly altered if extrapolatedto high O'NEILL ET AL.' HYDROUS GARNET ELASTICITY AND WATER IN THE MANTLE
TABLE 3. Velocities Measured From Hibschite 1457
Wave Normal* VelocitiesObserved t, km/s Velocities Calculated, km/s
Na Nb Nc Vp Vsl Vs2 Vp Vsl Vs2
0.678 -0.554 0.484 7.86 7.82 0.679 -0.502 0.536 4.60 4.61 0.809 -0.400 0.431 7.91 4.71 7.82 4.61 0.920 -0.257 0.296 7.79 4.64 7.84 4.60 -0.623 -0.503 0.600 7.81 4.66 7.82 4.62 -0.214 -0.651 0.728 7.84 4.68 7.82 4.63 -0.054 -0.044 0.998 7.84 4.46 7.85 4.59 0.475 0.062 0.878 4.48 4.59 0.048 0.017 0.999 7.79 4.60 7.85 4.59 0.470 0.059 0.881 7.81 4.58 7.83 4.59 0.439 0.043 0.898 7.85 4.60 4.68 7.83 4.59 4.62 0.481 0.085 0.873 7.68 4.57 7.83 4.59 0.057 -0.053 0.997 7.79 4.58 7.85 4.59 0.720 -0.526 0.453 7.87 7.82 0.747 -0.421 0.514 7.84 7.82 0.138 -0.693 0.707 7.88 7.82 0.100 -0.673 0.733 7.91 7.82 -0.004 -0.273 0.962 7.81 7.84
The rms deviation is 0.06 km/s. *The acousticwave normalis givenin termsof directioncosines with respectto thethree crystallographic axes. tShearwave polarizations areknown. temperature,provided the stability field of hydrogrossularis et al., 1987a; Lager et al., 1989; Aines and Rossman, 1984; not exceeded. Ackermann et al., 1983]. Therefore eclogitic regions of the Although hibschiteand katoite have been shownto be stable mantle, which contain a higher proportion of Ca-bearing at pressuresencompassing those of the entire upper mantle, at garnet, such as the pressure-transformedbasaltic portion of room temperature [Knittle et al., 1992; Olijnyk et al., 1991], subducted slabs, are likely locations for a significant similar to phaseB, the extent of the stabilityof hydrousgarnet componentof hydrogrossular. under complete mantle conditionsis an open question. It may A logical question to examine is how the properties of be argued, however, from the estimates of natural hydrous eclogitewith water-bearinggarnet compare with thoseof a dry garnet equilibration at depths as great as 180 km, along with composition. Because seismic waves are the most direct the resultsshown in Figure 4, that hydrogrossularcould act as a methodof probing the mantle, we have calculatedthe seismic vehicle for water transportto greatdepths and for water storage wave velocities for a model eclogite. Starting with an in the mantle, particularly in areasof subductingoceanic crust. assemblageof 50% omphacite and 50% grossularby volume, When the hydrogarnetsubstitution is made(H40 4 = SiO4), it the longitudinaland shear seismic velocities Vp andVs were has been observed that one of the ways the expanded calculated along a 300 K adiabat, using the bulk and shear tetrahedron is accommodatedis by a decreasein the shared moduli measuredby Bhagat and Bass [1992] for omphaciteand octahedral edge length, O(1)-O(4) (see Figure 3a)and an by Bass [1989] for grossular. Pressure derivatives of the increasein the lengthof the unsharededge, O(1)-O(5) [Lager et moduli for both phasesare from Duffy and Anderson [1989]. al., 1987a]. This octahedral distortion increases with the The calculatedseismic velocities for dry eclogiteare compared amount of hydrous component. In grossular the shared in Figures 5a and 5b with those calculated for wet eclogite, octahedraledge is longer than the unsharededge [Hazen and where hibschite substitutes completely for the grossular Finger. 1978], and the octahedral distortion is easily component. The seismicvelocities in the wet eclogite are -•6- accommodated. However, in pyrope Mg3A12Si3012 the 8% slower than in dry eclogite. The value of K 0' used for octahedral shared edge is shorter than the unsharededge, so hibschite was determinedby linear interpolationbetween the that a hydrous component may not be easily incorporated values of K0' for grossular,6.1, and katoite, 4.1. We note, [Sacerdoti and Passaglia, 1985; Lager et al., 1987a]. This is however,the values of K 0 measuredacross the hydrogrossular supportedwith evidence from natural and synthetic pyropes series demonstratea nonlinear relation (Figure 2). Mineral which typically have < 1 wt % H20 [Geiger et al., 1991; Lager garnets,which have substitutionon the Ca or A1 site, do not
TABLE 4. ElasticProperties
C11* C12 C44 K i•
Hibschite 186.5(11) 56.5(14) 63.9(5) 99.8(10) 64.3(5) Grossular t 321.7(8) 91.4(9) 104.6(4) 168.4(7) 108.9(4) Katoite• 66(4)
The numbersin parenthesisare the uncertainty in thelast digit. * All elasticmoduli are givenin unitsof gigapascals. t FromBass [1989]. • FromOlijnyk et al. [1991]. O'NEILL ET AL.: HYDROUSGARNET ELASTICITYAND WATERIN THE MAN'I2.Z 20,035
HydrogrossularBulk andShear Moduli an invisible oceanof water could be storedin the uppermantle. 5.2 Generally,tomographic studies of the uppermantle which
5.0 Aln (K)/z• (density)= 3.8 / a) 4.8 /// grossular / 0 (Ca(1
4.4 / hibs katoite
4.2 o
0.9 1.0 1.1 1.2 In(density)
Fig. 2. Measuredbulk and shearmoduli K and p. for hydrogrossular seriesCa3A12(SiO4)3.x(H404) x (see Table 4). Dataare from Olijnyket al. [1991] for katoite (x = 3.0), from Bass [1989] for grossular(x=O.O), and from this studyfor hibschite(x = 1.28). A nonlinearrelation is seen betweenK andH20 content(x = 3.0 is completelyhydrated); however, between grossularand hibschite, Aln(K)/Aln(p) = 3.8, indicatingthe difference in compressibilitymay largely be a volume effect (see b) discussion). Also, note the decreaseof K due to H20 substitutionis similar to the decreaseof •. provide systematicsto determine whether K 0' behaves nonlinearlyalso, and the valuecan only be bracketedby the end-members. However, if the value of K0' for hibschiteis closer to that of katoite, a greater difference in velocities betweenthe wet and dry eclogitemodels would result. The limit for seismologicalresolution is, conservatively,a traveltime delay of 0.5 s in V•,. Sucha traveltime delay can be the result of a seismic wave traveling for a short distance through an area with a large velocity decreaseor, conversely, following a long path through an area with a small velocity decrease.Given the velocity differencebetween the wet anddry eclogite models, a P wave traveling through 45 km of wet eclogite would be 0.5 s slower than the same wave passing through dry eclogite. In other words, for any thicknessof eclogiteless than 45 km the P wave travel time delay wouldbe below the limit of resolution, and it would not be possibleto c) differentiate between wet or dry eclogite. On the basis of oceanic crustal thickness, a subductionzone eclogite layer might be on the order of 10 km thick rather than 45 km thick (although it may be in the form of a sheet with broader dimensions),and the amountof grossular(or hibschite)in the eclogitemight rangeup to only slightlymore than 50% of the entiregarnet component [Dawson, 1980] ratherthan the 100% of the model. If water is carried into the mantle by subducted crustal layers [Peacock, 1990; yon Huene and Scholl, 199l; Meade and Jeanloz, 1991; Moore and Vrolijk, 1992], these factorscombine to reduce the observabilityof areasthat may be saturatedwith hydrousminerals. Thereforeit is possiblefor a considerable amount of water to be stored, undetected, within the mantle. For example, a massof H20 equivalentto that of Fig. 3. (a) Four polyhedra in the garnet structure,showing the the hydrosphere(-1.4 x 1021kg H20 ) would be contained numberingscheme of Novakand Gibbs[ 1971]. The solidlines indicate within --8.5 x 109 km3 of our wet model eclogite(1.35 wt % the unit cell edges.(b)-(c) Two layersof the garnetstructure projected in the 001 direction. The planeof cationsin (b) are locateda distance H20, 4.78 mol % H20). If thisamount of eclogite,--3vol % of a/8 abovethose in (c). The locationsof Ca and A1 are shownin (a). In the upper mantle, were distributedthroughout the mantle in grossular,Si occupiesthe middle of eachtetrahedron. In hydrogrossular sucha way that the dimensionswere less than 45 km in any an Si atomis replacedwith four H atomslocated outside the tetrahedron, given direction (that is, below the limit of seismicresolution) one slightlyabove each face. 20,036 O'NEILL ET AL.: HYDROUS GARNET ELASTICITY AND WATER IN THE MANTI_Z
DensityComparison composition,a large volume of hydratedgarnet is required 3.9 before a substantialdelay in seismicwave travel timeswould be observed. Additionally, unless there are differencesin 3.8 hibschite (K'--4.1) seismicwave attenuation,a slownessin velocitydue to a plus stishovite warmerregion of the mantlemay be alternativelyinterpreted as 3.7 hydration [Bonatti, 1990]. Since it is anticipatedthat other hydrous phases will show a similar slowness in seismic hibschite(K'=6.1)• velocitycompared to their anhydrousequivalents [Tyburczy et plus stishovite al., 1991], it is plausible that a considerableamount of water may be stored,unseen or unrecognized,in the Earth'smantle.
3.4 Acknowledgments. The authorswould like to thankR. Jeanloz,B. Romanowicz,and D.R. Bell for theirinsightful discussion and comments, 3.3 and R. J. Hemley, M. Vaughan,and an anonymousreviewer for grossularplus H 20 additionalhelpful comments. This researchwas supportedby NSF grants EAR-9018676 and EAR-8908461 and the NSF Research 3.2 Experiencesfor Undergraduatesprogram.
3.1 0 5 10 15 20 25 EclogiteSeismic Shear Velocity Pressure(GPa) 6.0 Fig. 4. The 300 K isothermaldensities for two chemicallyequivalent compositions.Solid circles indicate grossular plus H20, andopen circles indicatehibschite plus stishovite.The assemblagewith boundwater is denserthrough the uppermantle (pressure_< 22 GPa). Isothermalbulk Dry moduli Kr0 of grossularand hibschiteare estimatedfrom Ksomeasured 5.5 by Bass [ 1989] (grossular)and this study(hibschite) assuming a thermal expansioncoefficient, a=20x 10-6 K -1, and Griineisenparameter T= 1.5. K 0' for hibschiteis assumedto be betweenthe katoite-grossularseries 300K end-membervalues [Bass, 1989; Olijnyk et al., 1991] which are usedto bracketthe high-pressure density of hibschiteplus stishovite. Kso and K 0' 5.0 for stishoviteare from Duffy and Anderson[1989] with KT0 estimated Wet usingassumed values of a and•/of 15x10-6 K -1 and 1.5, respectively. The propertiesof ice VII, the stablephase of H20 at the conditionsof the calculation,are from Hemley et al. [ 1987].
4.5 make any inferences about lateral variations in seismic 0 5 10 15 20 25 velocities correlatethese variations to temperaturedifferences. Pressure (GPa) Figure 5 illustrates the effect of temperatureon the seismic velocities of the model eclogite in comparisonwith various EclogiteSeismic Longitudinal Velocity degrees of hydration. For example, based upon the seismic 11.0 velocities it would not be possible to distinguishbetween a 10% hydrationof the garnetcomponent and a 400ø increasein 10.5 the eclogite temperature (from 300 K to 700 K). The results shown in Figure 5 demonstratethat decouplingthe effect of temperature and H20 content on seismicvelocity may prove 10.0 difficult, and indicate that hydrous minerals may be responsible for some of the seismically observed heterogeneityin the upper mantle.
Wet CONCLUSIONS
Hibschite, a hydrogarnetwith approximately42% H40 4 8.5 substitutionfor SiO4, is found to have -•40% greaterbulk and shear compressibilities than grossular, the anhydrous 8.0 counterpart. The crystal chemical basis for the increased 0 5 10 15 20 25 compressibility of hydrogrossular is probably the larger Pressure (GPa) volume of the H40 4 tetrahedronas comparedwith the smaller, more rigid SiO 4 tetrahedronin silicategarnets. A resultof this Fig. 5. Seismic velocities for a model eclogite consistingof 50% omphaciteand 50% grossularcalculated along a 300 K adiabat(dashed increasedcompliance is that through the pressurerange of the line labeled"dry"). Solidlines, alsocalculated along 300 K adiabats,are upper mantle structurallybound H20 in garnethas a greater for eclogite,with the grossularcomponent replaced by hydrogrossular density than garnet coexisting with a separate solid H2¸ Ca3A12(SiO4)3.x(H404)x , where x = 0.32, 0.64, 0.96, and 1.28for 10%, phase. Therefore hydrogarnet is likely to be 20%, 30%, and 40% hydration,respectively. The curvelabeled "wet" thermodynamicallypreferred at high pressure. Although the representsa model eclogite consistingof 50% omphaciteand 50% hibschiteGRR 1457, specifically. Other dashedlines are for seismic seismic velocities of hibschite are -16% slower than those of velocitiesfor "dry" eclogite calculatedalong 700 K, 1000 K, 1500 K, õrossular,when consideredin the context of an eclogitic rock and 2000 K adiabats. O'NEILL ET AL.: HYDROUSGARNET E1,ASTICITYAND WATER IN THE MANTLE 20,037
REFERENCES H. Manghnani,pp. 297-304,AG!J, Washington,D.C., 1992. Lager,G. A., T. Armbruster,and J. Faber,Neutron and x-ray diffraction Ackermann,L., L. Cemic, and K. Langer,Hydrogarnet substitution in studyof hydrogarnetCa3A12(O4H4)3, Am. Mineral.,72, 756-765, pyrope:A possiblelocation for "water"in the mantle,Earth Planet. 1987a. Sci. Lett., 62,208-214, 1983. Lager, G. A., G. R. Rossman,F. J. Rotella, and A. J. Schultz,Neutron- Ahrens,T. J., Water storagein the mantle,Nature, 342, 122-123, 1989. diffractionstructure of a low-watergrossular at 20 K, Amer. Miner., Aines, R. D., and G. R. Rossman,Water contentof mantle garnets, 72,766-768, 1987b. Geology,12,720-723, 1984. Lager, G. A., T. Armbruster,F. J. Rotella, and G. R. Rossman,OH Akaogi,M., andS. J. Akimoto,High-pressure stability of a densehydrous substitutionin garnets:X-ray and neutrondiffraction, infrared, and magnesiansilicate Mg23Si8042H 6 and somegeophysical implications, geometric-modelingstudies, Am. Mineral., 74, 840-851, 1989. J. Geophys.Res., 85, 6944-6948, 1980. Martin,R. F., andG. Donnay,Hydroxyl in themantle, Am. Mineral.,57, Armbruster,T., and G. A. Lager, Oxygen disorderand the hydrogen 554-570, 1972. positionin garnet-hydrogametsolid solutions,Eur. J. Mineral., 1,363- Meade,C., and R. Jeanloz,Deep-focus earthquakes and recyclingof 369, 1989. water into the earth'smantle, Science,252, 68-72, 1991. Bartl,H., R6ntgen-Einkristalluntersuchungenan 3CaO.A1203.6H20 und Moore, J. C., and P. Vrolijk, Fluids in accretionaryprisms, Rev. an 12CaO.7A1203.H20: Neuer Vorschalg zur CaO.7A1203 Stmktur, Geophys.,30, 113-135, 1992. NeuesJahrb. Mineral. Monatsh., 404-413, 1969. Nalivkina, E. B., Hibschitefrom near the Bug River, Zapiski Vses. Bass,J. D., Elasticity of uvaroviteand andraditegarnets, J. Geophys. Mineralog. Obshch.(Mem. All-Union Min. Soc.),89, 717-718 (in Res., 91, 7575-7516, 1986. Russian)(Mineral Abs., 16, 454), 1960. Bass,J. D., Elasticityof grossularand spessartitegarnets by Brillouin Novak,G. A., andG. V. Gibbs,The crystalchemistry of silicategarnets, spectroscopy,J. Geophys.Res., 94, 7621-7628, 1989. Am. Mineral., 56, 791-825, 1971. Basso,R., A. Della Giusta,and L. Zefiro, Crystalstructure refinement of Olijnyk,H., E. Paris,C. A. Geiger,and G. A. Lager,Compressional study plazolite: A highly hydrated natural hydrogrossular,Neues Jahrb. of katoite[Ca3A12(O4H4)3] and grossular garnet, J. Geophys.Res., 96, Mineral. Monatsh., 6, 251-258, 1983. 14,313-14,318, 1991. Bell, R. D., and G. R. Rossman, Water in Earth's mantle: The role of Pabst,A., The crystalstructure of plazolite,Amer. Mineral., 22, 861-868, nominallyanhydrous minerals, Science, 255, 1391-1397, 1992a. 1937. Bell, R. D., and G. R. Rossman,The distributionof hydroxylin garnets Passaglia, E., and R. Rinaldi, Katoite, a new member of the from the subcontinentalmantle of southernAfrica, Contrib. Mineral. Ca3A12(SiO4)3-Ca3A12(OH)l2 series and a new nomenclaturefor the Petrol., 111,161-178, 1992b. hydrogrossulargroup of minerals,Bull. Mineral., 107, 605-618, 1984. Belyankin,D. S., and V. P. Petrov,The grossularoidgroup (hibschite, Peacock,S. M., Fluid processesin subductionzones, Science, 248, 329- plazolite),Amer. Mineral., 26, 450-453, 1941. 337, 1990. Bhagat,S., andJ. D. Bass,Single-crystal elastic properties of omphacite- Rinaldi, R., and E. Passaglia,Hibschite topotype: Crystal chemical C2/c by Brillouin spectroscopy,J. Geophys.Res., 97, 6843-6848, characterization,Eur. J. Mineral., 1,639-644, 1989. 1992. Ringwood, A. E., and A. Major, High-pressurereconnaissance Bloxam,T. W., Hydrogrossularfrom the Grivan-Ballantraecomplex, investigationsin the systemMg2SiO4-MgO-H20 , Earth Planet. Sci. Ayrshire,Mineral. Mag., 33, 814-815, 1964. Lett.,2, 130-133, 1967. Bonatti,E., Not so hot "hot spots"in the oceanicmantle, Science, 250, Rossman,G. R., andR. D. Aines,The hydrouscomponents of garnets: 107-111, 1990. Grossular-hydrogrossular,Am. Mineral., 76, 1153-1164, 1991. Cohen-Addad,C., P. Ducros, and E. F. Bettaut, Etude de las substitution Sacerdoti,M. and E. Passaglia,The crystal structureof katoite and du groupementSiO 4 par (OH)4 dansles compos6sA12Ca3(OH)l 2 et implicationswithin the hydrogrossulargroup of minerals,Bull. A12Ca3(SiO4)2.16(OH)3.36de type grenat, Acta Crystallogr.,23, 220- Mineral., 108, 1-8, 1985. 230, 1967. Sandercock,J. R., Trendsin Brillouinscattering: Studies of opaque Dawson,J. B., Kimberlitesand TheirXenoliths, pp. 168-172,Springer- materials,supported films, and centralmodes, in Light Scatteringin Verlag, New York, 1980. SolidsIII, Top. Appl. Phys., vol. 15, edited by M. Cardonaand G. Duffy, T. S., and D. L. Anderson,Seismic velocities in mantleminerals Gunterhodt,pp. 173-206, Springer-Verlag,New York, 1982. and the mineralogyof the upper mantle,J. Geophys.Res., 94, 1895- Skinner,B. J., Physicalproperties of end-membersof the garnetgroup, 1912, 1989. Am. Mineral., 41,428-436, 1956. Finger,L. W., J. Ko, R. M. Hazen, T. Gasparik,R. J. Hemley, C. T. Thompson,A. B., Water in the Earth'supper mantle, Nature, 358, 295- Prewitt, and D. J. Weidner, Crystal chemistryof phase B and an 302, 1992. anhydrousanalogue: Implications for water storagein the upper Tilley, C. E., On the replacementof anorthositeby hydrogrossularin the mantle, Nature, 341, 140-142, 1989. Transvaal,Trans. Geol. Soc.South Africa, 60, 15, 1957. Flint, E. P., H. F. McMurdie, and L. S. Wells, Hydrothermaland x-ray Tyburczy,J. A., T. S. Duffy, T. J. Ahrensand M. A. Lange,Shock wave studiesof the garnet-hydrogarnetseries and the relationshipof the equationof state of serpentineto 150 GPa: Implicationsfor the seriesto hydration productsof portland cement,J. Res. Natl. Bur. occurrenceof water in the earth'slower mantle,J. Geophys.Res., 96, Stand.,26, 13-33, 1941. 18,011-18,027, 1991. Foshag,W. F., Plazolite,a new mineral,Amer. Miner., 5, 183-185, 1920. Vaughan,M. T., Elasticity and crystalstructure in aluminosilicatesand Frankel,J. J., Uvarovitegarnet and SouthAfrican jade (hydrogrossular) pyroxenes,Ph.D. thesis,State Univ. of N.Y., at StonyBrook, 1979. from the Bushveldcomplex, Transvaal, Amer. Mineral., 44, 565-591, Vaughan,M. T., and J. D. Bass,Single crystalelastic properties of 1959. protoenstatite,Phys. Chem.Miner, 10, 62-68, 1983. Geiger, C. A., K. Langer, D. R. Bell, G. R. Rossman,and B. Winklet, von Huene,R.,and D. W. Scholl,Observations at convergentmargins The hydroxidecomponent in syntheticpyrope, Am. Mineral., 76, 49- concerningsediment subduction, subduction erosion, and the growthof 59, 1991. continentalcrust, Rev. Geophys.,29, 279-316, 1991. Gross,S., The mineralogyof the Hatrurimformation, Israel, Geol. Surv. Watt, J.P., G. F. Davies,and R. J. O'Connell,The elasticproperties of Israel, Bull. No. 70, 1-80, 1977. compositematerials, Rev. Geophys.,14, 541-563, 1976. Hazen,R. M., and L. W. Finger,Crystal structures and compressibilities Weidner, D. J., and H. R. Carleton,Elasticity of coesite,J. Geophys. of pyropeand grossular to 60 kbar,Am. Mineral., 63, 297-303, 1978. Res.. 82, 1334-1346, 1977. Hemley,R. J., A. P. Jephcoat,H. K. Mao, C. S. Zha, L. W. Finger,and D. E. Cox, Static compressionof H20-ice to 128 GPa (1.28 Mbar), J. D. Bass,Department of Geology,University of Illinois,1301 West Nature, 330, 737-740, 1987. Green Street, Urbana, IL 61801. Hill, P. G., A unified fundamentalequation for the thermodynamic B. O'Neill, Departmentof Geologyand Geophysics,University of propertiesof H20, J. Phys.Chem. Ref. Data, 19, 1233-1274,1990. California,301 McConeHall, Berkeley,CA 94720. Hutton,C. O., Hydrogrossular,a new mineralof the garnet-hydrogamet G. R. Rossman,Division of Geologicaland PlanetarySciences, series,Trans. Roy. Soc.New Zealand, 73, 174-180, 1943. CaliforniaInstitute of Technology,Pasadena, CA 91125. Knittle, E., A. Hathorne,M. Davis, and Q. Williams, A spectroscopic studyof the high-pressurebehavior of the O4H4 substitutionin garnet, (Received November 16, 1992; in High-Pressure Research: Application to Earth and Planetary revisedJuly 8, 1993; Sciences,Geophys. Monogr. Ser., vol. 67, editedby Y. Syonoand M. acceptedJuly 20, 1993.)