JOURNAL OF GEOPHYSICAL RESEARCH, VOL. 101, NO.

B1, PAGES 661-672, JANUARY 10, 1996

Thermal equation of state of CaSiO 3 perovskite

Yanbin Wang, Donald J. Weidner, and FrangoisGuyot•
Center for High-PressureResearchand Department of Earth and SpaceSciences,State Universityof New York
at Stony Brook

Abstract. A comprehensive pressure-volume-temperature data set hasbeen obtained

for CaSiO3 perovskiteup to 13 GPa and 1600 K, usingsynchrotronX ray diffraction
with a cubic-anvil,DIA-6 type apparatus(SAM-85). For eachvolumemeasurement,
nonhydrostaticstressis determinedfrom the relative shift in the diffractionlines of NaC1,
within which the samplewas [Link] to above973 K greatly reducedthe
strengthof NaC1 (to below 0.05 GPa), makingthe measurements [Link] room
temperature the cubic perovskitestructureremainsmetastableat pressuresas low as 1
GPa, below which the sampletransformsinto an amorphousphaseas indicatedby a large
background,a marked decreasein diffraction signals,and an anomalousvolume decrease
of the [Link] experimentaluncertaintiesare significantly
smallerthan thosein previousmeasurements,the new data provide a tighter constrainton
the zero pressurebulk modulusfor CaSiO3 perovskite.A new set of room temperature
equation of state parametersare identified so that both our data and the diamond cell
data of Mao et al. [1989] are compatible[Kr0 = 232(8) GPa, K½0= 4.8(3), and
V0- 45.58(4)•3]. Volumemeasurements
alongseveral
isotherms
underbothstable
and metastablepressureconditionsallow isochoricand isobaric interpolations within
the range of experimental pressure and temperature [Link] using various
approaches
yieldedconsistent
7
results
1 1
for (OKr/OT)e
of -0.033(8) GPaK-•, and
(Oa/OP)r of -6.3 X 10- GPa- K- , with a zero-pressurethermal expansiona 0 of
3.0 x 10-s K-•. The thermalpressure
is foundto be virtuallyindependent
of volume,
and thus the Anderson-Grtineisen parameter •r- K¬ = 4.8. These results are used
to predict the bulk modulus and density of CaSiO3 perovskite under lower mantle
[Link] an adiabat with the foot temperature of 2000 K, the density of the
perovskiteagreeswith that of the preliminary reference Earth model (PREM) within
1% throughout the lower mantle. The bulk modulus showsa smaller pressure
dependencealong the adiabat; it matches that of PREM at the top of the lower
mantle but is about 10% too low near the core-mantle boundary.

Introduction ZaM and Madon, 1991]. However, thesephaseswere synthe-

sizedwithout the presenceof the major lower mantle phases
Mineral physicsconstraintson the compositionof the ((Mg,Fe)SiO3 perovskiteand magnesiowustite) and therefore
Earth's lower mantle rely on knowledgeof thermal equations their existencein the lower mantle is questionableon petro-
of state of candidate minerals. Until recently, pressure- logicalgrounds.A recentphaseequilibriumstudyusinga more
volume-temperature(P-V-T) data have been obtainedonly representativecompositionof the mantle (pyrolite)showsthat
for (Mg,Fe)SiO3perovskite[Mao et al., 1991;Wanget al., 1994; A1is mostlyaccommodated in (Mg,Fe)SiO3perovskitewith no
Utsumiet al., 1995;Funamoriet al., 1995]and magnesiowustite separatealuminousphaseobserved[Irifune,1994].
[Fei et al., 1992].Phasescontainingother major elementssuch On the other hand, a numberof other experimentalstudies
as Ca and A1 have not yet been well [Link] indicate that the most likely calcium-bearingphaseis CaSiO3
current estimatesof the compositionof the Earth, an MgO- perovskite[Ringwoodand Major, 1971; Liu and Ringwood,
FeO-SiO2 systemwould accountfor about 90% of the mantle 1975, Gasparik, 1989, 1971]. Thus the lower mantle may be
volume,whereasincludingCaO and m1203componentsmay composedmainly of aluminous(Mg,Fe)SiO3 perovskite,Ca-
improvethe coverageto 99% [e.g.,Anderson,1989]. SiO3 perovskite,and magnesiowustite, as well as 1 or 2% of
Severalphaseshavebeen proposedto be the possiblehosts other phases.
of aluminumin the lower mantle, includingNaA1SiO4[Liu, Equation of state measurementsof CaSiO3 perovskiteare
1978], MgA1204 [Irifune et al., 1991], and A12SiO5 [Ahmed- complicatedby the instabilityof this material. Thus it is not
possibleto synthesizethe samplein one experimentand con-
duct high-pressure,high-temperaturemeasurementsin an-
other. Rather, the synthesisexperiment must be capable of
•Alsoat Institutde Physique
du Globede ParisandLaboratoire
de givingequationof state [Link] of state of Ca-
Mineralogie Cristallographie,Universite de Paris 7, Paris
SiO3perovskitehasbeenmeasuredat ambienttemperatureby
Copyright1996 by the American GeophysicalUnion. severalgroupsin a diamond anvil cell [e.g.,Mao et al., 1989;
Paper number 95JB03254, Tarrida and Richet, 1989; Tamai and Yagi, 1989]. Wang and
0148-0227/96/95 JB-03254 $05.00 Weidner[1994] reportedthe first in situ determinationof the
661
662 WANG ET AL.: THERMOELASTICITY OF CASIO 3 PEROVSKITE

The apparatuswas installedat the superconducting beam line

(X17B1) of the National Synchrotron Light Source at
BrookhavenNational Laboratory,Upton, New York. Details
zirconia
fill •'%•,•'•i boron
] of the apparatusare describedby Weidneret al. [1992].
Figure 1 showsthe sampleassemblyusedin this [Link]
,'ree",SøsXu"
l starting material was synthetic,high-purity wollastonite,the
samematerialusedin the studyof Wangand Weidner[1994].A
1.0-mm-diameter,2.0-mm-longsamplechamberwasfilledwith
one layer of wollastonitein the middle and two layers of a
mixtureof NaC1and BN (weightratio 4:1) on top andbot.•om
of the [Link] the sampleand the NaC1-BN mixturewere
driedat 570K formorethan10h beforetheexperiment.
The
irain
• •; ••boro"nitride sampleassemblywas compressed
about 13-14
at room temperatureto
GPa and heated to above 1400 K in Order to
• tubPng½carb;n
Saheater/,
• ], mple
chamber
synthesizeCaSiO3 [Link] cell volume of the perovs-

• Pth•aterlead
Møcurf
antdIsk
kite was then measuredat various pressureand temperature
[Link] determinedby the thermal equation
sam le ; of stateof NaC1[Decker,1971],andtemperaturewasmeasured
with a W/Re24%-W/Re6% [Link] temperature
was generallystablewithin 1 K up to 900 K and within 5 K to
(b) 'x Diffraction
vect•:?r
d 1600K. The X ray beamwascollimatedto 100/am and 200/am
in the vertical and horizontaldimensions,respectively,and the
temperaturevariationwithin the diffractingvolume is lessthan
5K.

2O The diffraction systemwas carefully calibrated using the

Latticff
planes
withsl•acing
d method describedby Wanget al. [1994]. For CaSiO3 perovs-
kite,upto 11poeaks
wereobserved
whose
d spacings
range
from 2.6 to 0.8 A, makingit possibleto determinepreciselythe
unit cell [Link] unit cell volume refinementsgener-
allyhaveuncertainties
of 0.02•3 (_ +0.05%).An example
of
usedin thisstudy.(a) Cqtawayview the diffractionpatternsobtainedat high pressureand temper-
Figure 1. Cell assembly
of theassembly,
withsample
sandwiched twolayers ature is shownin Figure 2a.
in [Link]
of NaC1/BN mixture and thermocouplejunction in contact
with sample. (b) Approximatestressgeometryand relation
withdiffraction geometry. With a fixed20 of 7.5ø,crystalplanes
that giveriseto diffractionsignalsare approximately parallelto
o
the [Link] suchthat the vertical
stresscomponent(r• is different from the horizontal compo-
nents

phase boundary between CaSiO3 perovskite and the low-

pressurephase assemblageCa2SiO4 + [Link] also
obtainedvolumemeasurements
alongseveralnea.
rly isobaric
paths up to 1600 K, within the stability field. However, the
limited
pressure
range
didnotallowa complete
de(ermination
of the thermal equationof state. B

In this paper we extend the work by Wang and Weidner

[1994] to examinea more completethermal equationof state.
Withbetterunderstanding
of thenonhydrostatic
siress
in the
systemand more preciseP-I/-T measurements, We have im-
provedthe performanceof the cell assemblyso that it allows
pressureto be varied in a large range, makingit possibleto
measure isothermal compressioncurves at various tempera-
[Link],in combination withthoseobtainedby Wang I I I I I I

and Weidner[1994],allowa directexamination of severalther- 400 600 800 1000 1200 1400

modynamicrelationsand a determinationof the thermalequa- Channels

tion of state.
Figure 2. Examplesof the diffractionspectraof CaSiO3 per-
ovskite.(a) Spectrumobtainedat 11.7GPa and 1172K. A total
of 11 diffraction lines are observed. Asterisks indicate diffrac-
Experimental
Procedure tion lines from the lower-pressurephaseassemblageCa2SiO4
Thehighpressure,hightemperature
experiments
werecar- + CaSi2Os.(b) Pattern collectedat 0.59 GPa duringdecom-
riedout.• a DIA-6typecubic-anvil
apparatus
(SAM-85)
with pressionat room [Link] the drop in peak intensity
sintered
•i½iamond
asanvil
material size).and the enhancedbackgrounddue to partial amorphization.
(4x 4mmtruncation
 WANG ET AL.' THERMOELASTICITY OF CASIO3 PEROVSKITE 663

As the specimenis in a solid-medium

environment, nonhy- 1400 -
drostaticstresses
are present,whichmaysignificantly
affectthe
[Link] al. [1994] havediscussed the 1200 -
characteristics of the stress field in the SAM-85 and the meth-
odsto measurethe nonhydrostatic stress(alsoseeFigure 1).
Briefly,the macroscopic stressfield may be approximatedas 000-
beingcylindricalin symmetry,with the verticalstresscompo-
800 -
nent(0.•)differentfromthe girdlestresses (0.2= ½•).The true
pressureis definedasthe averageof the threeprincipalstress
600 -
components: P = (1/3)(0.• + 20.3).
Severalrelatedproblemsarisein determiningP-V-T rela-
400 -
tionsin the presenceof nonhydrostatic stresses.(1) The dif-
fractionvectoris approximatelyparallel to 0.•; only the crys-
tallographicplanesthat are perpendicularto 0.• will be seen. 0 2 4 6 8 10 12

The truepressure isdifferentfrom thatmeasured fromdiffrac- Pressure, GPa

tion of NaC1by approximately(K/I•)A0., whereK and/• are
Figure 3. Pressure-temperaturepath of run 13. Each point
the bulk and shearmoduliof NaC1and A0.(= 0.• - 0.3)is the
representsthe conditionwherethe unit cellvolumeof CaSiO3
magnitudeof the nonhydrostatic stress[Menget al., 1993].(2) perovskiteis [Link] heavygrayline representsthe phase
Similarly,measurements of samplevolumeare either overes- boundaryof CaSiO3perovskitedeterminedin situ by Wang
timatedor underestimated dependingon whether0.•is smaller and Weidner[1994].
or greaterthan rr3 for a givenmean pressure.(3) Under a
boundarycondition0.• -• 0.3, each diffractionline will be
shiftedrelativeto the othersowingto elasticanisotropy,giving in order to minimizenonhydrostaticstresses
which may build
rise to inconsistentcell parameters,resultingin large uncer- up duringram load [Link] P-T path is shownin
tainties in determination of the cell volume. At this stage,we Figure3. At a constantram load,pressurechangesastemper-
have not found a satisfactoryway to correct the effectsof atureis varied,mostlydue to thermalpressureeffects,asillus-
nonhydrostatic stresses
in the volumemeasurement. Thuswe tratedby Figure3. About 60% of the dataare obtainedwithin
chooseto discardthe data that are contaminatedby nonhy- the stabilityfield of the perovskite,while 40% are at lower
drostatic stresses. pressures. However,no signof back transformationof the
We determinethe macroscopicnonhydrostatic stressesin perovskitestructureto a more stablephaseis observedabove
NaC1by measuringthe relativeshiftin the diffractionlinesof 4 GPa(at•:any temperature investigated).
Amorphization is
[Link] algorithmis outlinedby Weidneret al. [1994].At detectedat pressures below1 GPa (seebelow).The volume
room temperatureand pressures above2 GPa the nonhydro- data are summarized in Table 1. Also included in Table 1 are
staticstressin the NaC1pressuremediumis saturated,indicat- the dataobtainedby Wangand Weidner[1994]between10 and
ing that the stresslevel hasreachedthe yield point of NaC1 13GPaandupto 1600K. Thewholedatasetcovers a pressure
[Weidneret al., 1994].As the sampleis embeddedin the NaC1 rangeup to 13 GPa and temperatureto 1600K, with about
medium, the stressfield thus determined is assumedto repre- 80% of the data pointswithin the stabilityfield.
sent the deviatoricstressboundaryconditionfor the sample.
Each volume measurement is thus associatedwith three pa- Amorphizationand AnomalousVolume-PressureBehavior
at Low Pressure
rameters:pressure,temperature,and nonhydrostatic stress,
(o'1- 0.3).At roomtemperaturethe magnitudeof the nonhy- It is a well-known observationthat CaSiO3 perovskite is
drostaticstresscan amountto asmuchas0.3 GPa (roughlythe nonquenchhble andamorphizes atlowpressures [e.g.,
Ring-
yield strengthof NaC1).We foundthat the mosteffectiveway woodand Major, 1971;Tamai and Yagi,1989;Kanzakiet al.,
to minimizenonhydrostatic stressis heating,asthe strengths of 1991;etc.).However,the effectsof the amorphization process
NaC1 and BN decreasedramaticallywith increasingtempera- on volumehave not been addressedpreviouslyin connection
ture. Above 600 K, or immediatelyafter coolingfrom above with equationof statemeasurements. In a high-temperature
X
thistemperature,the measurednonhydrostatic stressis gener- ray diffractionstudyon (Mg,Fe)SiO3perovskite,Wanget al.
ally below0.05 GPa, whichis about the detectabilityof the [1994]observeda significantdecreasein volumeas the per-
[Link] P-V-T measurements were carried ovskitesampleswere heatedat ambientpressure,associated
out only duringand after heating. withamorphization of the(Mg,Fe)SiO3 perovskite.
Afterheat-
ing,the sampleswerefoundto givea cellvolumesignificantly
smallerthan the startingvalue. Followingtheir observations,
Results we establishedthe followingcriteria of amorphizationin Ca-
We obtain volume data for the perovskitealong several SiO3 perovskite:(1) a marked decreasein intensityof the
isothermsat 300, 573, 773, and 973 K, at pressuresbetween4 diffraction
peaks,(2) a significant
increasein background,
and
and 13 GPa. The procedureis as follows:(1) Compressthe (3) a changein the pressure-volume relationassociated
with
startingmaterialat room temperatureto the maximumpres- criteria 1 and 2.
sure,theseheat to 1400K to completelyconvertwollastonite All threephenomena wereobserved in CaSiO3perovskiteat
into the perovskitestructure.(2) Coolfrom 1400K to measure roomtemperature,as the pressurewasloweredfrom above1
both volumeand pressureat 973, 773, 573, and 300 K. (3) GPa. Figure2b showsa diffractionspectrumof the sample
Lower the ram load at room temperatureto the next pressure obtainedat room temperatureand 0.59 GPa. A comparison
and repeat heatingand coolingat the sametemperaturesat with Figure2a (11.7 GPa and 1172K) showsa dramaticdrop
Volume data are collectedon coolingonly, in diffractionintensityand an increasein the background,
severalpressures.
664 WANG ET AL.' THERMOELASTICITY OF CASIO3 PEROVSKITE

Table 1. Summaryof the P-V-T Data for CaSiO3 Table 1. (continued)

PerovskiteObtained in This Study
Pressure, Temperature, Volume, Ao-,
Pressure, Temperature, Volume, Ao-, GPa K •3 GPa*
GPa K •3 GPa*
Run 5õ (continued)
Run 13 12.01 1469 44.86(1) 0.062
12.04 1368 44.76(1) 0.001 11.69 1594 45.08(5) 0.043
11.69 1172 44.54(2) 0.006
0.007 *Differential stress((r• - (r3) determinedfrom NaC1 peak shifts.
11.35 970 44.41(3)
-0.018 Compressionis negative.
11.06 977 44.45(2)
0.008 ?Not used in equationof statefit due to high nonhydrostaticstress.
10.49 980 44.53(3)
-0.021 5Amorphization occurredbetween 0.6 and 1 GPa.
9.37 976 44.71(3)
-0.021 {}Databy Wangand Weidner[1994].
7.95 974 44.92(2)
5.51 980 45.44(2) 0.019
10.92 769 44.23(3) 0.018
10.58 769 44.28(2) 0.008
9.98 770 44.38(3) 0.038
8.97 774 44.51(2) -0.035
-0.015
indicativeof formation of an [Link] room
7.52 772 44.78(2)
5.04 771 45.25(1) -0.030 temperaturevolume-pressurebehavioris shownin Figure 4a,
10.58 572 44.07(2) -0.013 which also comparesour data with the zero-pressuredatum
10.15 572 44.12(1) 0.023 measuredby Kanzaki et al. [1991] on a mostly amorphized
9.50 575 44.22(2) 0.038
sample.A marked decreasein volumeis observedas pressure
8.48 570 44.37(4) 0.002
is lowered from 1 GPa to 0.59 GPa. Such an anomalous volume
7.09 572 44.63(1) -0.007
4.68 570 45.12(1) 0.011 decreaseis closelyrelated with the amorphizationprocess,
10.07 306 43.83(2) -0.017 although the detailed mechanismremains to be clarified.
9.54 303 43.87(3) -0.004
Clearly,the effectof amorphizationcannotbe ignoredin equa-
9.63 304 43.93(3) -0.004
0.023
tion of [Link] in the equa-
8.97 302 44.01(3)
8.98 302 43.99(3) 0.003 tion of state analysiswould result in an overestimateof the
0.031 bulk modulus.
7.97 301 44.08(4)
7.94 303 44.14(2) -0.001 Durbenand Wolf [1992]have studiedthe amorphizationof
6.54 303 44.38(4) -0.006
0.080
MgSiO3 perovskiteat elevatedtemperaturesby Raman spec-
4.95 301 44.59(4)
4.15 303 44.87(1) -0.001 troscopyand found that a temperatureof 350 K is sufficientto
4.12 303 44.87(1) -0.026 partially amorphize the perovskite crystals,whose Raman
2.66 301 45.03(1) 0.092 peaksexhibita residualshift after beingcooledbackto ambi-
1.137 301 45.39(6) 0.132
ent [Link] directionof the shift is consistentwith a
0.595 301 45.22(4) 0.145
reduced volume.

Run 3õ At this stagethe processof amorphizationremainsunclear.

12.00 1074 44.39(1) 0.000 Durbenand Wolf [1992] interpret the residualshift as internal
12.08 1145 44.52(1) 0.047
stress,since the amorphousmaterial formed after heating is
12.16 1210 44.59(1) 0.051
0.062
about 30% lessdensethan the perovskiteand must expand.
12.23 1250 44.62(1)
11.60 1006 44.48(1) 0.118 Our X ray data on both perovskitesindicatea changein peak
11.21 904 44.39(1) 0.064 shapeduring amorphization:diffractionpeaks of perovskite
11.07 837 44.35(1) 0.054 becomebroader and may even exhibitsomefine featuresnear
10.90 761 44.29(1) 0.071
their [Link] we are inclinedto interpret the amor-
10.67 672 44.22(1) 0.076
10.46 577 44.15(1) 0.093 phizationprocessas creatingan increasingnumberof defects
12.04 1207 44.60(1) 0.043 in the formerly [Link] it may appear
12.04 1205 44.60(1) 0.006 that some "perovskite"peaks can be observedduring amor-
11.24 1182 44.73(2) 0.043
phization,the P-V-T relationsmeasuredare not representativeof
10.67 1167 44.81(1) 0.030
0.003
the metastable or stablebehaviorof the perovskite,
asvolumeis
10.26 1159 44.88(2)
no longera [Link] to identifythe breakdownof
Run 4õ the samplewill lead to an incorrectequationof state.
11.61 1177 44.82(1) 0.088
11.85 1261 44.82(1) 0.013
-0.022
Room Temperature Compression and Bulk Modulus
11.77 1339 44.95(2)
11.74 1403 45.02(2) -0.021 Table 1 lists the measured differential stress for each volume
11.53 1301 44.90(1) 0.018
[Link] can be seenthat generallythe nonhydrostatic
11.43 1223 44.82(1) 0.017
0.023
stressis below 0.05 GPa after heating, except for the data
11.25 1137 44.76(2)
11.12 1074 44.68(2) 0.029 below 4 GPa, where no heatingwas conductedand the non-
10.91 978 44.60(2) 0.011 hydrostaticstressbuilt up graduallyon decompression. Espe-
10.73 894 44.52(1) 0.008 cially for the data below 2 GPa, nonhydrostaticstressesare
10.51 798 44.46(2) 0.012
0.019
above0.1 GPa and therefore are discardedin our equationof
11.31 1125 44.73(1) state fits.
Run 5õ We fit our volumedatawith the Birch-Murnaghanequations
12.16 1128 44.53(2) 0.046 of state
12.25 1318 44.70(1) 0.021
3 2/3
12.23 1381 44.75(1) 0.047
P-• 3Kro[r•7/3
_ r•s/3][1
+ 7(K¬- 4)(,! - 1)], (1)
 WANG ET AL.' THERMOELASTICITY OF CASIO 3 PEROVSKITE 665

46
46.0 A
44

2.< 42
E
'" 40
44.5 >

44.0

43.0 • • i • • • I I
0 2 4 6 8 10 12 14 0 20 40 60 80 100 120

Pressure, GPa Pressure, GPa

Figure 4. Room temperaturecompressiondata of CaSiO3 perovskiteobtained in run 13. (a) Unit cell
volume of perovskiteunder stableor metastableconditions(solid circles).Amorphizationis observedat
pressuresbelow 1 GPa; the measuredunit cell volume exhibitsa significantdecreaseas pressureis further
lowered(open square).For comparison,the room pressuredatumfrom Kanzakiet al. [1991]is shown(open
triangle),whichwasobtainedfrom a [Link] plottedare the diamondanvilcell data
of Mao et al. [1989]in a similarpressurerange(open circles).Notice that in their data at pressures
below1
GPa, the cellvolumealsodecreases with decreasing
[Link] dashedcurve(in both Figure4a and Figure
4b)isthefit byMaoetal. [1989],
withKTO= 281GPaandl/o = 45.37A3,withanassumed
K•ro= 4.0.
Clearly, the fit deviatesfrom our data. The solid curveis the fit to our data which givesKro = 232 GPa and
l/o• 45.58(4)
•3,with
K•ro
= [Link]
dotted
curve
(inboth
Figure
4aand
Figure
4b)isour
fittoMao
etal.s data[Kro= 244(12)GPaandl/o = 45.71(12)•3, K•ro= 4.8(3)]. (b) Comparison
of theroom
temperaturevolumewith pressuredata obtainedin this study(solid circles)and previouslyin the diamond
anvilcell up to 134 GPa (opencircles[Mao et al., 1989]).The dottedcurveis consistent
with both data sets.

whereKro and K•o are the ambientisothermalbulk modulus to changeits [Link] addition,as the perovskitestruc-
and its pressurederivative, and r• = l/o/l/. With fixed K•ro ture becomesunstablenear zero pressure,amorphizationmay
rangingfrom 3 to 5, the resultsfrom fitting equation(1) yield occur, and as in the observationsin our samples,the unit cell
a rangeof Kro from 238(7) to 227(7); the corresponding zero- volumebecomessmalleraspressureis decreasedfrom above1
pressure volume rangesfrom45.58(4) to45.60(4) •3, respec-GPa. On the basisof a leastsquaresfit of the entire data set to
tively. This narrow range indicatesthat our P-I/data, which the Birch-Murnaghanequationof state,Mao et al. [1989] ob-
are nonhydrostaticstressfree, define a robust trend toward tainedKro - 281(4) GPa with an assumed K•ro = 4.0. This
zero pressure. is the dashed curve shown in Figure 4a. Clearly, this fit is
This range of Kro value is significantlysmaller than the incompatiblewith our data, as the fit underestimatesthe unit
previousvaluesof 275(15) [Tartidaand Richet,1989], 281(4) cell volume below 6 GPa and overestimates it above.
[Maoetal., 1989],and325(20)GPa [Tamaiand Yagi,1989].All With all of the Mao et al. [1989] data weighted by their
thesepreviousdiamondanvilcell studieswere conductedwith- respective pressure uncertainties (their Table 1), a least
out pressuremediumbecauseof the nonquenchablenature of squaresfit to the Birch-Murnaghanequation of state gives
CaSiO3 perovskite;the level of nonhydrostaticstressis thus gro = 282(8)GPa,K•ro= 4.0(2), andI/o = 45.3(5)•3.
[Link] have found that nonhydrostaticstresscan sig- When the lower-volumedata below 1 GPa (most likely con-
nificantly affect the resultson Kro. For example, the room taminatedby amorphization)are excluded,the fit usingequa-
temperaturedata reported by Wangand Weidner[1994] were tion (1)yields Kro = 268(8) GPa, K•ro = 4.3(2), and I/o =
measuredon a CaSiO3perovskitesampleon decreasingpres- 45.47(6) •3, withthesame weighting scheme. ThislatterKro
sure. A careful examination on these data reveals that a dif- is significantlysmallerthan the fittingresultbasedon the entire
ferential stressof about 0.2-0.3 GPa built up as pressurewas data [Link] is not surprising,asKro correspondsto the slope
decreasedto below 6 GPa; the minimum compressionalstress of the P-I/curve at zero pressure.
was in the vertical direction,resultingin an underestimatein Although the data of Mao et al. [1989] below 12 GPa appear
pressureand therefore an overestimateof the bulk modulus. to have smaller pressureuncertainties,they are most likely
In a careful diamond anvil cell study, Mao et al. [1989] contaminatedby nonhydrostaticstressand thus shouldnot be
measuredthe cell volume of CaSiO3 perovskiteto 134 GPa. weightedmore than the higher-pressuredata,whichhavebeen
Their data are compared with ours in Figures 4a and 4b. [Link] we weightthe remainingdata equally,thisleads
Clearly, our data have much smalleruncertaintiesand define a tOKro = 244(12) GPa, K•ro = 4.8(3), and Vo = 45.71(12)
clearer trend in the I/-P [Link] the Mao et al. study,at •3. Thisis thedottedcurvein Figure4a. In Figure4b we
pressuresgreater than 12 GPa the sample was laser heated comparethis fit with that of Mao et al. [1989] over the entire
when the pressurewas changed,so that effectsof nonhydro- pressurerange. At pressuresabove 15 GPa the two fits are
static stress were minimized. Below 12 GPa, however, laser virtually identical; at lower pressuresthe latter fit is more
heatingwas not appliedbecauseof the unstablenature of the consistentwith the slope defined by our data (Figure 4a).
material. Thus although the low-pressuredata may appear These latter equation of state parametersdescribethe mea-
better characterizedin terms of pressureuncertainty,nonhy- surements from both the diamond anvil cell and the more
drostaticstressmay havebeen built up, causingthe P-I/curve accuratelarge-volumepressdata. This analysisindicatesthat
666 WANG ET AL.: THERMOELASTICITY OF CASIO 3 PEROVSKITE

16

14 A vo•. B
12 45.2 ¾ 8

0 8 44.4 •
• 7
I• 6 44.0•'
2

0 • 6
300 400 500 600 700 800 900 1000 1100 43 44 45 46 47
Temperature,
K Volume
Figure 5. Isochoricdataof CaSiO3perovskiteinterpolatedwithinthepressureandtemperaturerangeof the
measurement. (a) Isochorcs
in theP-T spaceshowinga nearlylinearrelationbetweenpressureandtemper-
ature for eachconstantunit cell volume.(b) Slopeof the isochores
in Figure5a as a functionof unit cell
volume.
A linearfit to thedatagives(OKr/OT)v= -0.36(8) x 10-2 GPaK-•.

high-precisiondata on the low-pressure sideof the P-l/mea- expansion,i.e., a = a + b T, the coefficientb is found very
surementsare critical in constrainingKro and that a combi- small(of theorderof 10-• K-2), withanuncertainty
several
nationwith high-pressure data helpsto constrainK•ro. Unfor- ordersof magnitudegreaterthanthevalueitself(about10-9).
tunately,there is a small but noticeablesystematicshift in Thuswe concludethat at eachpressure,the thermalexpansion
volumesbetweenour data and thoseof Mao et al. [1989] (see coefficientis bestexpressedasa constant,independentof tem-
Figure4), preventingusfromincorporating the twodatasetsin peraturewithin of experimentalconditionsand the uncertain-
the equationof statefit. We thusadoptK•ro = 4.8 from the ties. Of course,as temperatureis decreasedto closeto 0 K,
above analysison Mao et al. data in fitting our data with thermalexpansionis expectedto vanish;our data do not pro-
equation•1) to obtainKto = 232(8) GPa and I/o = vide any constrainttoward this end.
45.58(4) •3 for ourdata(thesolidcurvein Figure4a). Wangand Weidner[1994] observedthat the thermal expan-
sion coefficientof CaSiO3 perovskiteis also independentof
Thermal Equation of State Analysis temperaturewithin the stabilityfield (10-12 GPa) up to 1600
Isochoric and isobaric interpolation. We interpolate the K. A similarbehaviorwasalsoobservedin MgSiO3perovskite
volumedatawithin the pressureand temperaturerangeof the by Wanget al. [1994],Utsumiet al. [1995],andFunamoriet al.
measurementsto obtain isochoricdata (OP/OT)v. At each [1995]. These studiesshowthat under pressurethe thermal
volume,pressureis linear with temperature(Figure 5a), and expansioncoefficientis essentiallyconstantover a wide range
the slope represents(OP/OT)v. In Figure 5b the (OP/OT)v of temperaturesup to about 1800K at pressures from 4 to 30
valuesare plottedagainstthe unit cellvolumeand againshow GPa. Thusit appearsthat pressuresuppresses the temperature
linear variationswithin the [Link] slope dependence of thermalexpansionin both MgSiO3 andCaSiO3
of the (OP/OT)vversusV plotis 7.9(2)x 10-5 GPaK-•. perovskite,makingthe thermal expansioncoefficientvirtually
Accordingto the thermodynamicidentity(OP/OT)v = aK r, independentof temperaturefrom 300 to well above1000K.
the slopein Figure5b is O(aKr)/OV. In addition,aoKro can The isobaricthermal expansiondata are plotted in Figure 6
bemeasured
directly
fromthecurve
at V = Vo= 45.58•3 to illustrate the pressuredependencein a. The Wangand
to be 7.2 x 10-3 GPaK-•. As no temperature
dependence
is Weidner[1994]data are alsoincludedin Figure6 for compar-
detectedin the pressureand temperaturerangeof the data,we [Link] isobaricthermal expansiondata obtainedfrom sev-
assumethat the slopeis approximately[ 0 (aKr)/0 V] r, which eral runs agree very well. A linear fit to all of the thermal
expansion
is equal to -(OKr/OT)v/V. Therefore at zero pressure,we datayields
(Oa/OP)r= -5.24(19)x 10-7 GPa-• K-l,
have(OKr/OT)v = -0.36 x 10-2 GPa K-•. From the withao = 3.23(2)x 10-5 K-•. Fromthermodynamic identity
thermodynamicidentity (Oa/OP)r= (OKT/OT)pK•
-2,weobtain(ogr/OT)e = -0.028GPa
K -• at ambient condition.
(OKr/OT)v = (OKr/OT)•,+ a,Xn,(OKr/Oe)r (2)
Thermal pressure equation of state. O. L. Anderson
weobtain(OKr/OT)e= -0.031 GPaK-l, withaoKro= 7.2 [Anderson,1980,1984;Andersonet al., 1989]proposeda ther-
x 10-3 GPaK-1, Vo = 45.58•3, and(OKr/OP)r= 4.8. mal equationof statebasedon the Mie-Gruneisentheory,by
ForKro = 232 GPa,aoisabout3.1x 10-s K-•. We alsonote computingthermalpressures from thermodynamic parameters
that fitting (0P/0T)v as a linear functionof In V yieldsvery measurablein the [Link] any temperatureT and for a
similar results. givenvolumeV, the pressurecan be expressed asthe sumof
We theninterpolatethe datato obtainvolumesasa function the static pressureat a reference temperature To and the
of temperatureat [Link] give thermalpressureAPth [e.g.,Jacksonand Rigden,1996].
thermal expansiondirectlyat each pressure,where the loga-
P(V, r)- P(V, to) q- APth- [Link](•, r) (3a)
rithm of volume is then fitted by a linear function in temper-
ature. Attempts have been made to fit theseisobaricdata to where
obtainthe temperaturedependenceof [Link]
all cases, however, even with a linear term in the thermal X 1- T•7/3--T•5/3 b• = 3Kr(Vo, T0)/2
 WANG ET AL.: THERMOELASTICITY OF CASIO3 PEROVSKITE 667

9 t
X 2 = Xi(T•2/3-- 1) b2= •Kr(V(,,To)[Kro(Vo,
To)- 4]
P(V, To): b,X, + b2X2 (3b)
X3 = (T - T0), b.• =

X4 = -(r- To) In r/ b4: (or•/or)•

Xs = (r- To)2 b5: [(02p/or2)v]/2

From these coefficientssome relevant thermoelasticparame-
ters can be calculated,for example [seeJacksonand Rigden,
1996],
200 400 600 800 1000 1200 1400 1600
(orr(v, r)/or)e = (art(v, r)or)v- art(v, T)K•(V, T)
Temperature, K
(Oa/or)v = (02p/OT2)v/Kr(V,T) Figure 7. Thermal pressureof CaSiO3 perovskiteup to 13
GPa and 1600 K. All of the data listed in Table 1 are used.

(Oa/or)e = (Oa/or)v- a(OKr/OT)e/Kr(V, T)

ao = 3.04(5)x 10-s K-•. Thesimilarity
in thevalues
indicates
Figure 7 showsour data obtainedin all four runs,in the form that the resultsare quite robust.
of thermal pressure,which is the differencebetweenthe mea- The fact that (OKr/OT)v is virtually nonresolvablefrom
suredpressureand the calculatedpressurebasedon the room zero suggeststhat there is no volume dependencein thermal
temperature(T o = 300 K) Birch-Murnaghanequation of pressure.
Indeed,whenth,e thermalpressure
isplottedagainst
state at the same volume. It is clear that the data fall in a
volume(Figure 8), measurements alongall the isothermsare
narrow trend, varying linearly with temperature. A fit using bestfit by a constantthermalpressure
[Link](OKv/OT)v=
(3a) with the first three terms (but with a fixedK•ro - 4.8) 0, equation (2) gives(OKr/OT)p= -0.035 GPa-• K-•, for
yieldsVo = 45.61(4)A3,Kro : 231(7) GPa,andao = Kro = 232 GPa,c•o = 3.2 x 10-sK -1 andK•ro= 4 8 , ß o

3.33(6)x 10-s K-•. Whenthefourthtermis included, b4 is As an interestingoutcomefrom Figure8, the thermalpressure

foundto be -5.27 x 10-3 GPaK-•, buttheuncertainty is0.02 beingindependentof volumeindicatesthat [ 0(art)/0 T) r =
GPaK-•; clearly,b4 cannotbe resolvedfromthisfit. If we fix (&r - Kr) aKr/V = 0 [Andersonet al., 1992]. Thus the
b4 at -3.6 x 10-3 GPaK-•, according
to theisochoric results, Anderson-Gruneisenparameter &r = - (0Kr/0T)v/aKr =
thenweobtainVo = 45.61(4)•3, Krø = 230(7)GPa,and K•r = 4.8.
ao = 3.20(6)x 10-s K-•. Resultsfromvariousfitsaresum- High-temperature Birch-Murnaghan equation of state.
marized in Table 2.
We now assumethat the Birch-Murnaghanequation of state
To examinethe effectsof systematicerrors in the four runs, canbe appliedto anygiventemperatureand try to analyzethe
we have analyzed the data from run 13 alone. A similar whole data setwith suchan equationof [Link] useequation
weightedfit using the first three terms in (3a) gives Vo = (1), with r/ = VR/V; nowKro, VR, andK•ro are parametersat
45.61(2),•3,Kro = 229(4)GPa,andao = 3.04(5)x 10-s zero pressureand at anyconstanttemperatureT; K•ro is again
K- [Link],b4 cannotbe resolved;
fixingb4 at - 3.6 x 10-3 assumedconstant(= 4.8). V• = V(0, T) is relatedto Vo via
GPaK-• yields
Vo = 45.61(2),•3,Kro - 227(4)GPa,and zero-pressurethermal expansion•(0, T), which is expressed
asa + bT - cT2 wherea b andc areall positive
constants.
It turns out that the resultsof the fit usingthis equation of
state vary noticeablywith various weightingfunctionsused,
4.0 when both Kro and Vo are set free in the fit. A closecorrela-
tion is observedin thesefits: where Kro is low, Vo is high and
I(OKr/OT)v is low,[Link] areforcedto fix
3.5 -
KroaswellasVo,at232GPaand45.58•3, respectively.
Even
in this situation,the coefficientsb and c in the zero-pressure
thermalexpansionstill cannotbe resolvedin the weightedfits.
The coefficientb is barely resolvable in an unweighted fit
o (Table 2). With both b and c set to zero, the unweightedfit
2.5- gives(OKr/OT)v= -0.022(10)GPaK-•, andso= 3.06(14)x
10-s K-• (shownin Figure9).
Summary of the equation of state analysis. Table 2 gives
2.0
the thermoelasticpropertiesof CaSiO3 perovskiteusingvari-
0 2 4 6 8 10 12 14
ous [Link] each parameter the uncertaintiesin pa-
Pressure, OPa renthesesrepresentthe standarddeviationobtainedfrom the
fit. Other parametersmay be derivedfrom equationssuchas
Figure 6. Isobaricthermal expansionobtained as a function
of [Link] circlesare obtained by fitting the interpo- (3c); for theseparameters we do not giveuncertainties
in the
latedcellvolumes(e.g.,Figure5a). The slopeof thisplot gives table, as they can be obtainedby standarderror propagation.
a Oa/OPof -5.24(19)x 10-7 GPa-• K-•. Opencircles arethe We prefer the parameterset obtainedfrom the thermal pres-
isobaricthermalexpansiondata obtainedby Wangand Weidner sureequation of statefit witha weightingfunction1/Ap2(p)
[1994], agreeingwell with the presentmeasurement. + 1/Ap2(v), whereAP(P) = - is the pressure
668 WANG ET AL.: THERMOELASTICITY OF CASIO3 PEROVSKITE

Table 2. Summaryof ThermoelasticParametersfor CaSiO3PerovskiteObtained From Various Approaches

Isothermal Thermal PressureEquation of State High-TemperatureBirch-Murnaghan
Room
Parameter Temperature Isobaric Isochoric Weighted,1 Weighted,2 Unweighted Weighted Unweighted,1 Unweighted,2

Krro, GPa 232(8) ...... 233(7) 229(4) 233(9) [2321 [2321 [2321
K}-o [4.8] [4.8] [4.81 [4.81 I4.81 [4.8] [4.81 [4.81 [4.81
Vo,,•,3 45.58(4) ...... 45.60(5) 45.61(2) 45.58(4) [45.58] [45.58] [45.58]
ao,10-s K • '" 3.23(2) 3.1 3.23 3.01 3.06 3.55(18) 3.06(14) 3.19(48)
(aP/OT)v , ... 7.2 7.2(3) 7.5(5) 6.9(1) 7.1(1) 8.2 7.1 7.4
10 -• GPa K -•
(OKr/OT)v, '" 0.008 -0.004 0.005(22) [0] [0] 0.004 0.012 0.001
GPa K •
(oK•/or)•, ... -0.028 -0.031 -0.031 -0.033 -0.034 -0.036(8) -0.022(10) -0.035(15)
GPa K-'
(Oa/OP)•., "' -5.24(19) -5.8 -5.7 -6.3 -6.4 -6.7 -4.1 -6.5
10-7 (GPaK) '
(Oa/or)v , ... [0] ... 0.36 0.43 0.44 [0] [0] 0.45(40)
10-8 K-2
(02p/OT2)v
' ... [0] [01 [01 [01 [01 [01
GPa 2 K-2

Squarebracketsindicatefixedparametersin the fit. Uncertainties

givenin parentheses
representthe standarddeviationobtainedfrom the fit.

uncertaintydue to deviatoricstressand AP(V) = Kro AV/V 1/o= (2[Si-O])3Z, (5)

is the pressureuncertaintydue to volumeerror AV. The pre-
ferredvalues,(OKr/OT)/, = -0.033 GPaK- • ando•o= 3.06x where Z is the numberof chemicalformulaeper unit cell. The
10-5 K-•, giveanAnderson-Gruneisen parameter8r = 4.65, tilt angle for the orthorhombicMgSiO3 perovskite(Pbnm)
can be estimated from unit cell dimensions via
which is about equal to K•.- 4.8(3).
(:I)= cos-'( x/•a2/bc). (6)
Discussion
Thus pressure- and temperature-inducedvolume change
Comparison of Thermoelastic Behavior BetweenCaSiO3 may be expressedas a combinationof changesin bond length
and MgSiO3 Perovskites and tilt [Link] thermalexpansion(•) and compressibility
Weidnerand Zhao [1992] have proposed a formalism in (/3), differentiating(4) with respectto temperatureand pres-
definingthe equationof stateof perovskitewhich allowseval- sure leads to
uation of the effectsof structural [Link]
•v = •v,, + •v, = 30 In [Si-O]/OT + 20 In cos tP/OT (7)
the SiO6 octahedraare regular, the unit cell volume of the
orthorhombicperovskitemay be written as 13v: 13v,,
+ 13v,,,
= -30 In [Si-O]/OP- 20 ln cos(I)/0P. (8)
V =abc = Viicos2 (I), (4)
It has been recognizedthat the tilt angle and bond length
where (P is the tilt angle about the "threefold" axes of the calculatedusing (4)-(6) show a systematicoffset compared
octahedrawhich describesthe degree of distortionof the per- with those derived from atomic positionsdetermined from
ovskite structure and V(, is the volume of an idealized cubic structural [Link] example, in MgSiO3 perovskite,the
perovskitethat dependsonly on the octahedronbond length ambient [Si-O]bondlength(1.778•) is underestimated by
[si-o]:

46.O
KN%1200 X,•
660
K%
K
0¸ '

oo o
o•oo
7oo
•-e ß
973 K

773 K
ß
c•
45.5
•5•73•
300••K
• 45.0

E • 44.5
.673 K
• 2

300 k 44.0

43.5 44.0 44.5 45.0 45.5 0 2 4 6 8 10 12 14 16

Volume, Pressure, (3Pa

Figure 8. Thermal pressureas a function of volume. Solid Figure 9. High-temperatureBirch-Murnaghanequation of

circles indicate data from run 13. Within the experimental statefit (unweightedfit I in Table 2) to all CaSiO3perovskite
errors,no volumedependencein thermal pressureis observed. data listed in Table 1.
 WANG ET AL.' THERMOELASTICITY OF CASIO3 PEROVSKITE 669

0.0153, (-1%), compared withthesingle-crystal

X raydata
(1.793
• [Horiuchie!al.,1987]).
A similar
situation
isobserved
for neighborite,NaMgF3 [Zhao et al., 1993, 1994] and is at-
tributed to slightdistortionin the [Link] pressureand
1.80
I
1.79

"• 1.78
temperaturedependence(i.e., the partial derivativesin (7) and
(8)), however,is rather consistentwith structuraldata [Zhao et
al., 1993]. In caseof CaSiO3,insofaras the structureis cubic
(Z = 1), the tilt anglemustbe zero,and the volumeis directly
• .77-1.76

related with [Si-O] bond distancevia equation (5). Table 3 1.75 -

o

comparesselectedstructuraland thermoelasticparametersin
1.74 -
responseto pressureand temperaturefor the two perovskites.
Figures 10 and 11 show the bond length and tilt angle of
20
MgSiO3 perovskiteas a functionof pressureand temperature,
respectively,from the data reportedby Wanget al. [1994]. In
the pressureplot (Figure 10), averagelinearthermalexpansion
coefficients have been used to correct the data to room tem-
• 19
perature. These linear thermal expansiondata are obtained
from fitting the individualcell edge data in separateruns in
•_ 18
Table 4 of Wanget al. [1994].In the temperatureplot (Figure
11), the cell parametersare correctedto zero pressureusing
the linearcompressibilitydata of Mao et al. [1991]obtainedat 17

room temperatureup to 30 GPa (/3, = 1.29,/3b = 1.05, and

/3c = 1.33TPa-•). Superimposed
onthebondlengthdataare 0 2 4 6 8 10 12
the [Si-O]lengthof CaSiO3 perovskitemeasuredin the present Pressure, GPa
study. Similar to the MgSiO3 perovskite, the bond length
[Si-O], definedas a/2 for CaSiO3,is alsocorrectedusinga Figure 10. Calculated[Si-O] bond length and tilt angle of
linear compressibility(/3, = 1/3Kro ) in Figure 10 and an MgSiO3 perovskiteas a functionof pressure,basedon data of
averagelinear thermal expansionof a, = av./3 in Figure 11. Wanget al. [1994]to 11 GPa and 1300K (opencircles).[Si-O]
distance is also calculated from the cell edge length a of
During compressionthe tilt angle for MgSiO3 increases
CaSiO3 perovskitefor comparison(solidcircles).Linear ther-
(O(I)/OP= 2.72 x 10-2 deg GPa-•), whereasthe bond mal expansioncoefficients,obtainedfrom fitting the cell edge
lengthdecreases (O[Si-O]/OP= -1.979 x 10-3 fk GPa-•) data of Wanget al. [1994]are usedto correctthe data to 300 K.
(Figure 10). The contributionsto the compressibility
are dif- For CaSiO3,a linear thermal expansionof a •./3 is usedfor the
ferentbyoneorderof magnitude, with13•.,,
= 3.3 x 10-3 [Link] that [Si-O] bond distancesin both materials
GPa-• and/3v-. = 3.2 x 10--4GPa-•, respectively. Onthe respondsimilarlyto pressure,but for MgSiO3 the tilt angle
other hand,the magnitudeof the changein bond lengthisvery changesvery little. The total volumecompressionis dominated
similar to that of CaSiO3, where O[Si-O]/OP= -2.107 x by shrinkingin bond length.
10-3fkGPa-• and/3
•.,,= 3.6x 10-3GPa-•.Thesimilarity
in compressibilities of both perovskitesindicatesthat the com-
pressionmechanismof MgSiO3 is similar to that of CaSiO3. fkK-• which
,
gives
anaverage
av0about
1.32x 10-s K-•
Thus compressionof MgSiO3 is overwhelminglydominatedby This valueis lessthan half of the bondlengthexpansion(1.833
changein the [Si-O] bond distance. x 10-s f• K-•) for CaSiO3
(Figure11).Clearly,
the[Si-O]
However, it is evident from Figure 11 that upon heating
bond distancerespondsto heatingquite differentlyin the two
MgSiO3perovskiteexpandsboth by changingbond lengthand perovskites.
Thus in MgSiO3 orthorhombicperovskitethe mechanisms
by adjustingthe tilt [Link] changein the tilt angleis about
-7.563x 10-4 degK-• correspondingtoameanav-.of0.88 of thermal expansionand compressionare quite distinct,with
x 10-s K-•, overtheentiretemperature
rangeof 300to 1300 the former dominatedby changein the tilt angle and the latter
K. Themeanslopeof thebonddistancechange is7.81x 10-s dominated by change in [Si-O] bond [Link] if
pressureand temperature are increasedin such a way as to
maintain a constant volume, both the octahedron size and the
Table 3. Some StructuralParametersof MgSiO3 and tilt anglewill be [Link] eachconstantvolume the various
CaSiO3Perovskitesin Respondingto Pressureand degreesof the changesin bond length and tilt angle will de-
Temperature pend on the pressureand [Link] this,
MgSiO3 CaSiO3
one can directlyinfer that frequencies(v) of at leastsomeof
the normal modes of vibration will show a strong intrinsic
Si-O 1.780 (1.793), 1.786 temperaturedependence,that is, (0 v/OT)v differentfrom 0.
Distance,* Such a behavior is characteristicof intrinsic anharmonicity
(I)*, deg 18.2 0.0
3.3 3.5 [e.g., Mammone and Sharma, 1979; Gillet et al., 1991], and
/3[.0,TPa-1
/3[.., TPa-• 3.2 0.0 mantle perovskitemight be a stronglyanharmonicmaterial. In
av-0, 10-5K-1 1.32 3.08 describingthe P-V-T behavior of materials,one of the most
av,, 10-5 K- 0.88 0.0
commonlyused assumptionsis that phonon eigenfrequencies
*Calculatedfrom equations(5) and (6). depend only on volume, which expressesthe net effect of the
?Datum in parenthesesis from single-crystalstructuralrefinement pressureand temperature state of the material. In view of the
[Horiuchiet al., 1987]. distinctivepressureand temperatureeffectson the structureof
670 WANG ET AL.: THERMOELASTICITY OF CASIO 3 PEROVSKITE

1.84 -
700 I • • • • • 1OhO
K

650
-A
600
-
550 -
•e//•ø
••
K 0K

1.80- a. 500 - -

• 450- -
""' 400- -
350 - -
1.76-
300 -

5.5
BI I I I I ]
20-

-••••o •øø• øø
o o -•) g o
4.5 -
OOogo O0

16- I I I I 1 I I
0 20 40 60 80 100 120 140 160

Pressure, GPa
14

400 600 800 1000 1200 1400 1600 Figure 12. Thermoelasticparametersof CaSiO3 perovskite
Temperature,K
(solidcurves)extrapolatedto lowermantleconditionsascom-
paredwith PREM (solidsymbols).(a) Adiabaticbulk modulus.
Figure 11. Calculated[Si-O] bond distanceand tilt angle of Data for CaSiO3 are plotted along severaladiabatswith the
MgSiO3 perovskiteusingequations(6)-(8) as a function of foot temperaturesof 1000, 2000, and 3000 K. (b) Similar plot
temperature,basedon data of Wanget al. [1994]to 11 GPa and for [Link] CaSiO3perovskiteis slightlyhigherthan
1300 K (open circles).[Si-O] distanceis alsocalculatedfrom that of PREM.
the cell edge length a of CaSiO3 perovskitefor comparison
(solid circles).Note that the slopea[Si-O]/OTfor CaSiO3is
about 2 times that for MgSiO3. Roughly about 30% of the
volume expansionof MgSiO3 is due to [Si-O] bond increase the ambientAnderson-Gruneisenparameter(St0 = 4.8) is
with temperature;the dominant changein volume is due to applicablefor the pressureand temperatureconditionsof the
changein tilt angle. lower mantle. Thermal expansionat any pressureis thus ap-
proximated
bya •o•V/Vo) [Link] Gruneisenparameter is
estimated
byg•ng andWeidner
[1994]fromthefittingresults
MgSi03 perovskite,however, this simple picture may be in- at 10-12 GPa (7 = 1.60 - 1.75), with 7/7o = ([//Vo) q. For
valid.
q -- 1, the estimated70 rangesfrom 1.66 to 1.82 [Wangand
Weidner,1994].
Implications for Lower Mantle Composition Models
With these thermoelasticparameterswe calculatethe den-
The new data on CaSiO3 perovskiteallow us to examinethe sityand Ks for CaSiO3 perovskitein the lower [Link]
effectsof this material on constraintson compositionof the 12ashowsthe adiabaticbulk modulusof the perovskitealong
[Link] 4 liststhermoelasticparametersof CaSiO3 three adiabats1000,2000, and 3000 K (the foot temperatures
perovskiteused for this [Link] extrapolatingthe thermal are at zero pressure),in comparisonwith the preliminaryref-
equationof state to lower mantle conditions,we assumethat erence Earth model (PREM) [Dziewonskiand Anderson,
1981].Along the 2000 K adiabat,the K, of the perovskiteis
very closeto that of PREM at the top of the lower mantle but
Table 4. ThermoelasticParametersfor MgSiO3 and gradually [Link] the core-mantleboundarythe bulk
CaSiO3 Perovskites modulus is about 10% lower than that of PREM. Given that

MgSiO•* CaSiO3 CaSiO3perovskitcis most likely below 10 vol % in the lower

mantle,sucha differencein K, would be insignificantin affect-
Vo, cm3 mol-I 24.46 27.45 ing previous analyses of composition based on data on
P0,gcm 3 4.108 4.231
ao,10-5 K-l 1.7 3.01 (Mg,Fe)SiO3 pcrovskiteand magnesiowustite alone.
Kro , GPa 261 232 Figure 12b shows density profiles of CaSiO3 perovskite
K•r0 4 4.8 [Link] PREM, CaSiO3 may be
(OKr/OT)/,, GPa/K -0.023 -0.033 slightlydenser,but onlyby about 1%. Thuswe concludethat in
19o, K 1030 11007
8To 4.9 4.8
modelingthe compositionof the lower mantle, CaSiO3 per-
7o 1.4 1.77 ovskitecan be regarded as invisible.
q½ 1 1 However,the bulk modulusof CaSiO3 perovskiteis signifi-
cantlyhigherthan that of the TransitionZone. As this perovs-
*Data from Wanget al. [1994].
?From Wangand Weidner[1994].
kite maybeginto form at pressuresaslow as17 GPa [Gasparik,
$Assumedpressuredependencein Gruneisen parameter through 1989, 1990], it will contributeto the velocitygradientsin this
volume. region of the Earth.
 WANG ET AL.: THERMOELASTICITY OF CASIO 3 PEROVSKITE 671

Conclusions stable MgSiO3 perovskite:A Raman spectroscopicstudy,/tin. Min-

eral., 77, 890-893, 1992.
Using synchrotronX radiation and a DIA-type, cubic anvil Dziewonski,A.M., and D. L. Anderson,PreliminaryreferenceEarth
apparatus,P-I/-T measurementson CaSiO3 havebeen carried model, [Link] Planet. Inter., 25, 297-356, 1981.
out at pressuresbetween0 and 13 GPa and temperaturesup to Fei, Y., H. K. Mao, J. Shu, and J. Hu, P-I/-T equation of state of
1600 K, under quasi-hydrostaticconditionswith nonhydro- magnesiowustite(Mgo.•,F%.4)O , Phys. Chem. Miner., 18, 416-422,
1992.
static stressescarefully determined. At room temperature, Funamori, N., T. Yagi, W. Utsumi, T. Kondo, and T. Uchida, Ther-
amorphizationis observedat pressures below 1 GPa. Although moelasticpropertiesof MgSiO3 perovskitedetermined by in situ X
some of the "perovskite" diffraction lines seem to remain at ray observationsup to 30 GPa and 2000 K,J. [Link].,in press,
1995.
theselow pressures,the unit cell volumeexhibitsa significant
decreasecomparedwith [Link] behavior Gasparik,T., Transformationsof enstatite-jadeite-diopsidepyroxenes
to garnet, [Link] Petrol., 102, 389-405, 1989.
is comparedwith that of MgSiO3 perovskite,where on heating Gasparik, T., Phaserelationsin the transitionzone, J. [Link].,
to above400 K at zero pressure,a similar amorphizationpro- 95, 15,751-15,769, 1990.
cesshas been observed[Wanget al., 1994]. Once amorphiza- Gillet, P., P. Richer, F. Guyot, and G. Fiquet, High-temperaturether-
tion occurs,volumeis no longera statefunctionand cannotbe modynamic properties of forsterire, J. [Link]., 96, 11,805-
used in determiningequation of state parameters. 11,816, 1991.
Horiuchi, H., E. Ito, and D. J. Weidner, Perovskite-typeMgSiO3:
A completethermal equationof state has been determined
Single crystal X-ray diffraction study,,4m. Mineral., 72, 357-360,
for [Link] datawere analyzedin variousways, 1987.
and the results are consistent. We have identified a set of Irifune, T., Absenceof an aluminousphase in the upper part of the
thermoelasticparametersthat are compatiblewith both our Earth's lower mantle, Nature, 370, 131-133, 1994.
data to 1600 K and previousdiamondanvil cell measurements Irifune, T., K. Fujino, and E. Ohtani, A new high-pressureform of
MgA1204,Nature, 349, 409-411, 1991.
at room temperatureto 134 GPa. The ambientisothermalbulk
Jackson,I., and S. Rigden, Analysisof P-l/- T data: Constraintson the
modulus is 232(8) GPa, with K•ro = 4.8(3) and l/o = thermoelasticproperties of high-pressureminerals, Phys. Earth
45.58(4)•3. Theobserved
P-I/-T behavior
of thecubic
Ca- [Link]., in press,1996.
SiO3 perovskiteis compared with that of the orthorhombic Kanzaki, M., J. F. Stebbins,and X. Xue, Characterizationof quenched
perovskiteMgSiO3 to showthat crystalstructureplaysan im- highpressure
phasesin CaSiO3systemby X-ray and 29SiNMR,
portant role in equationsof [Link] MgSiO3, as Si-O-Si tilt [Link]., 18, 463-466, 1991.
Kudoh, Y., On the symmetryof the CaSiO3 perovskitequenchedfrom
angleand [Si-O] bond distancerespondto pressureand tem- 15 GPa and 1500øC,paper presentedat 14th AIRAPT Conference,
peraturerather differently,care mustbe taken in extrapolating Assoc. Int. for Res. and Adv. of High-PressureSci. and Technol.,
the laboratory-measured P-l/-T relationsto deepmantlecon- Colorado Springs,Colo., 1993.
[Link] the cubic CaSiO3 perovskite,the situationis less Liu, L. G., High-pressurephasetransformationsof albite, jadeite and
nepheline,Earth [Link]. Lett., 37, 438-444, 1978.
[Link] measurementsindicatethat under lower mantle
Liu, L. G., and A. E. Ringwood,Synthesisof a perovskite-typepoly-
conditionsof pressureand temperature,the bulk modulusand morph of CaSiO3,Earth Planet. Sci. Lett., 14, 1079-1082, 1975.
densityof CaSiO3 perovskitewouldbe nearlyidenticalto those Mammone, J. F., and S. K. Sharma,Pressureand temperaturedepen-
of PREM. The role of CaSiO3 perovskitein the Transition dence of the Raman spectraof rutile-structuredoxides,Year Book
Zone awaitsfurther investigation. CarnegieInst. Washington,78, 369-373, 1979.
Mao, H. K., L. C. Chen, R. J. Hemley, A. P. Jephcoat,and Y. Wu,
Stability and equationof state of CaSiO3 perovskiteto 134 GPa, J.
Acknowledgments. We wish to thank D. H. Lindsley for starting
[Link]., 94, 17,889-17,894, 1989.
materialandM. T. Vaughan,R. C. Liebermann,J. Zhang, I. Martinez, Mao, H. K., R. J. Hemley, J. Shu,L. Chen, A. P. Jephcoat,Y. Wu, and
W. A. Bassett,Effect of pressure,temperature, and compositionon
and Y. Sinelnikovfor their assistanceduringthe [Link] also
thank D. Chapman,N. Lazarz, and W. Thomlinsonof the X17 beam- the lattice parametersand densityof (Mg,Fe)SiO3 perovskiteto 30
GPa, J. [Link]., 96, 8069-8079, 1991.
line at NSLS for their technicalassistance.
Commentsmade by Pascal
Richer and an anonymousreviewer have significantlyimproved the Meng, Y., D. J. Weidner, and Y. Fei, Deviatoric stressin a quasi-
[Link] SAM-85 project was supportedby NSF grant EAR hydrostaticdiamondanvil cell:Effect on the volume-basedpressure
calibration,[Link]., 20, 1147-1150, 1993.
89-20239to the Centerfor High PressureResearch,StateUniversityof
New York at StonyBrook. Mineral PhysicsInstitute, State University Ringwood,A. E., and A. Major, Synthesisof majorite and other high
of New York at Stony Brook, contribution 144. pressuregarnets and perovskites,Earth Planet. Sci. Lett., 12, 411-
418, 1971.
Tamai, H., and T. Yagi, High-pressureand high-temperaturephase
References relationsin CaSiO3and CaMgSi206 and elasticityof perovskite-type
CaSiO3,[Link] [Link]., 54, 370-377, 1989.
Ahmed-ZaYd,I., and M. Madon, A high-pressureform of A12SiO5as a Tarrida, M., and P. Richet, Equation of state of CaSiO3perovskiteto
possiblehost of aluminum in the lower mantle, Nature, 353, 426- 96 GPa, [Link]. Lett., 16, 1351-1354, 1989.
428, 1991. Utsumi, W., N. Funamori, T. Yagi, E. Ito, T. Kikegawa, and O. Shi-
Anderson, D. L., Compositionof the Earth, Science,243, 367-370, momura, Thermal expansivityof MgSiO3 perovskite under high
1989. pressuresup to 20 GPa, [Link]., 22, 1005-1008, 1995.
Anderson,O. L., An experimentalhigh-temperaturethermalequation Wang, Y., and D. J. Weidner, Thermoelasticityof CaSiO3 perovskite
of state bypassingthe Gruneisen parameter, [Link] Planet. and implicationsfor the lower mantle, [Link]., 21,895-
Inter., 22, 173-183, 1980. 898, 1994.
Anderson,O. L., A universalequationof state,J. Geodyn.,1,185-214, Wang, Y., D. J. Weidner, and R. C. Liebermann,Thermal equationof
1984. state of (Mg,Fe)SiO3 perovskiteand constraintson compositionof
Anderson,O. L., D. G. Isaak, and S. Yamamoto, Anharmonicityand the lower mantle, [Link] [Link]., 83, 13-40, 1994.
the equationof state of gold,J. ,4ppl. Phys.,65, 1534-1543, 1989. Weidner, D. J., and Y. Zhao, Elasticity and equation of state of
Anderson, O. L., D. Isaak, and H. Oda, High-temperature elastic perovskite:Implicationsfor the Earth's lower mantle, in High Pres-
constantdata on mineralsrelevantto geophysics, Rev. Geophys.,30, sureResearch: ,4pplicationto Earth and PlanetarySciences,Geophys.
57-90, 1992. Monogr. Ser.,vol. 67, editedby Y. Syonoand M. H. Manghnani,pp.
Decker, D. L., High-pressureequation of state for NaC1, KC1, and 191-196, AGU, Washington, D.C., 1992.
CsC1,J. ,4ppl. Phys.,42, 3239-3244, 1971. Weidner, D. J., M. T. Vaughan, J. Ko, Y. Wang, X. Liu, A. Yeganeh-
Durben, D. J., and G. H. Wolf, High-temperaturebehavior of meta- Haeri, R. E. Pocalo,and Y. Zhao, Characterizationof stress,pres-
672 WANG ET AL.: THERMOELASTICITY OF CASIO 3 PEROVSKITE

sure, and temperaturein SAM85, a DIA type high pressureappa- fraction studyusingmonochromaticsynchrotronX-radiation, Am.
ratus,in High Pressure Research: Applicationto Earth and Planetary Mineral., 79, 615-621, 1994.
Sciences,[Link]. Set., vol. 67, edited by Y. Syono and
M. H. Manghnani,pp. 13-17, AGU, Washington,D.C., 1992. F. Guyot, Y. Wang, and D. J. Weidner, Center for High-Pressure
Weidner, D. J., Y. Wang, and M. T. Vaughan,Yield strengthat high Researchand Departmentof Earth and SpaceSciences, StateUniver-
pressureand [Link]. [Link]., 21,753-756, 1994. sity of New York at Stony Brook, Stony Brook, NY 11794-2100.
Zhao, Y., D. J. Weidner, J. B. Parise,and D. E. Cox, Thermal expan- (e-mail:ywang@[Link])
sion and structuraldistortionof perovskite--Data for NaMgF 3 per-
ovskite,I, [Link] [Link]., 76, 1-16, 1993.
Zhao, Y., J. B. Parise, Y. Wang, K. Kusaba,M. T. Vaughan, D. J.
Weidner, T. Kikegawaa,J. Chen, and O. Shimomura,High-pressure (ReceivedMarch 24, 1995;revisedSeptember8, 1995;
crystalchemistryof neighborite,NaMgF3: An angle-dispersive dif- acceptedOctober 17, 1995.)