American Mineralogist, Volume 79, pages 497-503, 1994
Melting of pyrope,MgrAlrSirOrr, at7-16 GPa
J. Zn.lNc*
Departmentof Earth and EnvironmentalSciences,
The City University of New York, New York 10036,U.S.A.
C. Hnnznpnc
Department of Geological Sciences,Rutgers University, New Brunswick, New Jersey08903, U.S.A.
and Center for High PressureResearchand Department of Earth and SpaceSciences,The State University of New York at
Stony Brook, Stony Brook, New York 11794,U.S.A.
Ansrucr
Melting experiments on pyrope have been performed from 7 to l6 GPa using the multianvil press.The results show that the melting curve of pyrope is unusual relative to those
for pyroxene and forsterite becauseeither it is linear in T-P spaceor it has a break in the
slopeat around l0 GPa. At 3-10 GPa the melting curve appearsto be normal in that it
can be calculatedfrom reasonablethermodynamic and elastic parametersusing the method of minimizing the Gibbs free energyof melting. However, at pressures> 10 GPa, the
observedmelting temperaturesare higher than predicted. This can be explained either by
cation disordering in crystalline pyrope or by a stiffening of liquid MgrAlrSirO,, associated
with pressure-induced
coordinationchangesinvolving Al3*. In either case,pyrope is stabilized relative to liquid, and the slope dZldP of the melting curve is higher than that for
enstatiteand forsterite at pressuresin excessof l0 GPa. The effectofpressure is, therefore,
to increasethe thermal stability of pyrope at the expenseof olivine and to place garnet on
the liquidus for komatiite, peridotite, and chondrite compositions.
Ir.rrnonucrroN
The liquidus phasefor komatite, peridotite,and chondrite compositions changesfrom olivine to garnet at high
pressures(Herzberg, 1983; Takahashi, 1986; Ohtani et
al., I 986; Ito and Takahashi,1987;-Herzberget al., I 990;
Zhangand Herzberg,1993),which increasesopportunities for garnet fractionation to occur in nature. Garnet
fractionation has been demonstratedto be important in
understandingthe geochemistryof some komatiites
(Herzberg, 1992), and it may have occurred during an
early diflerentiation event in the Earth (Ohtani and Sawamoto, 1987;Herzbergand Gasparik,l99l). The melting temperaturesof pyrope, MgrAlrSirO,r, at high pressureare thereforeimportant for understandingthe stability
of garnet in multicomponent systems.
Early studieshave shown that pyrope melts incongruently to liquid + spinel or liquid + aluminous enstatite
up to 3.5 GPa (Boyd and England, 1962),and it melts
congruentlyat higherpressures(Boyd and England,1962;
Ohtani et al., 1981;Irifune and Ohtani, 1986).Of particular interest in these studies is the slope of the pyrope
melting curve, which was reported to changefrom about
82 "C/GPa at 3.5 GPa to 0 'ClGPa at pressuresabove l0
GPa (Irifune and Ohtani, 1986). This differs from the
melting curve of forsterite,which maintains a positive
* Present address: Center for High PressureResearch, Department of Earth and SpaceSciences,The State University of
New York at Stony Brook, Stony Brook, New York I 1794,U.S.A.
0003-{04xl94l0506-0497$02.00
slopeat all pressures(Ohtaniand Kumazawa,1981;Presnall and Walter, 1993). The effect of pressureinferred
from theseend-membersystemsis to increasethe stability of olivine at the expenseof garnet,just the opposite
of what is observedin multicomponentsystems.A major
objectiveofthis work is to resolvethis contradictionby
reinvestigatingthe melting curve of pyrope.
Exprnrprnx'rAt,
METHoDS AND RESULTS
All experimentswerecarriedout usingthe 2000-t splitspheremultianvil apparatus(USSA-2000)locatedat Stony
Brook. A detailed description of the press,the sample
assembly,and the techniqueshasbeengiven in numerous
papers(Gasparik,1989,1990;Herzberget al., 1990;Liebermann and Wang, 1992) arrd is not repeatedhere.
Two typesof startingmaterialwereused.One wascrystalline pyrope synthesizedfrom a stoichiometricmixture
of MgO, Al,O., and SiO, at 5 GPa and 1300"C for 5 h;
optical examination and X-ray diffraction showed that
the recovered starting material consisted largely of pyrope but containedminor amounts of unreactedoxides.
The second type of starting material was a pyrope-free
mixture of oxides,and this was usedto examinehow the
experimental results might be afected by the nature of
the startingmaterial.
The startingmaterialswere loadedinto Re containers,
and assembliesof 10 mm were used. These were then
fired at 1000 "C in an Ar atmospherefor I h prior to the
experiment to expel all HrO. The time to reach the target
temperaturewas l5-30 min, and experimentdurations
49'.l
498
ZHANG AND HERZBERG: MELTING OF PYROPE
Trale 1. Experimentalresultson the meltingof pyrope
P(GPa)
70
8.0
8.0
en
on
9.0
v5
1 00
10.0
11.5
115
1 30
13.0
130
13.0
14.5
145
1 45
1 60
16.0
16.0
1 60
f fC)
1980
1980
2010
2040
2035
2070
2060
2110
2050
2150
2190
2100
2200
2190
2300
2280
2300
2365
2200
2300
2380
2s00
t (min)
J.5
3.0
3.0
4.O
J.C
4.O
40
30
3.0
3.0
4.O
0.5
1.0
3.0
3.0
1.0
5h
40
3.0
3.0
1.0
30
Starting
materials
Expt
products
mixture
mixture
mixture
mixture
mixture
mixture
mixture
pyrope
pyrope
pyrope
mixture
pyrope
pyrope
pyrope
pyrope
pyrope
mixture
mixture
pyrope
pyrope
pyrope
pyrope
Py.,L-Py
Py,L
Py,L
Py
Py
Py
Py' L
Py
Py
Py
Py
Py
Py
Py,L
Py
Py
Py
Py
Py
Py
Py,L
'Py : pyrope.
'. L: liquidquenchedto glass + extremelyfine-grained
(<1 pm) intergrowth of crystallinephases
at the target temperatureswere typically 3 min before the
experimentswere terminated by shutting off the power.
A reversalexperimentinvolving garnet at 2185 'C and
l4 GPa demonstratedthat equilibrium was reachedwithin 3 min in the presenceof a melt phase(Herzberget al.,
1990).The experiment products were preparedas thin
sectionsfor optical examinationand electronprobe analysis (Table 1).
The temperaturesgiven in Table I are those recorded
by the thermocouple,which was located0.5 + 0.2 mm
from the hot spot (Fig. l) and which reads 50 "C lower
than the hot-spot temperature of 1200 "C (Gasparik,
1989).The temperaturegradient varies as a function of
distancealongthe heater,beingflat near the hot spot and
steepat the cold end, and is approximatelysymmetrical
on each side of the hot spot. The temperature difference
betweenthe hot spot and the cold end of the capsuleis
200 "C in the range1400-1700'C (Gasparik,1989)and
about 300 "C in the range 2000-2200 "C (Presnall and
Gasparik,1990).In thoseexperimentsfor which a quench
liquid phase is reported in Table l, the interface with
crystallinepyrope is sharp (Fig. l). The temperatureof
this interfaceis the melting temperatureof pyrope, and
its dispositionin all experimentsis within 0.7 mm of the
hot spot; this resultsin a thermocoupletemperaturethat
is very similar to the melting temperatureof pyrope.
Precisionin the measurementof temperatureis typically +30 'C (Gasparik, 1990; Presnall and Gasparik,
1990).Temperatureaccuracyis more problematicalbecausethe effectof pressureon thermocoupleemf has not
been determinedand is not included in Table l. However, it has beenestimatedto be about - 5 'ClGPa (Zhang
and Herzberg,1993),on the basisof the melting temper-
Thermocouple
Junction
(
ttot Spot
Rhenium
Pyrope + Corundum
s---to,;-------at l3 GPaand2300'C showFig. l. Sketchof an experiment
oxides.
ingthe distributionofpyrope,liquid,andunreacted
atureof MgSiO, perovskitedeterminedby a diamond cell
heatedby a CO, laser (2727 + 50 'C at 22 GPa: Zen and
Boehler, 1993),compared with a multianvil estimation
(2600 'C and 22.5 GPa: Gasparik, I 990). If this pressure
effecton thermocoupleemf is realistic,then the temperatures given in Table I are too low by 35 "C at 7 GPa
and by 80 .C at 16 GPa. Uncertainties stemming from
precisionand accuracyshould thereforebe within +50'C.
Irifune and Ohtani (1986) reported that the liquid phase
quenchesto glassand aluminous enstatiteat pressures
lessthan about 7 GPa, and to garnetat higher pressures.
The liquid phase in the experiments reported here
quenchedto an extremely fine intergrowth of glassand
quenchcrystallinephasesthat were too small to identify
by electronprobe analysis.In experimentslike these,the
liquid phaseis separatedfrom crystallinepyrope down
the temperaturegradientby an interfacethat is invariably
sharp (Fig. l). Pyrope grows with two kinds of morphology, depending on the distance away from the liquid.
Within a narrow band 50-100 pm in width next to the
liquid (Fig. l), which translatesto about 20 "C down the
temperature gradient, the pyrope crystals are the largest
in size (10-20 pm), euhedral,and inclusionfree. Farther
down the temperature gradient, the pyrope crystals are
smaller and contain inclusionsof corundum. Inclusions
of unreactedoxidesof SiO2,Al2Or, and MgO occupythe
coldestportions of all experimentalsamplesfor both types
of startingmaterial,crystallinepyrope and mixed oxides,
but are completely absentwhere a liquid phaseexisted
(Fig. l). Only those parts of the samplesthat contain in-
499
ZHANG AND HERZBERG: MELTING OF PYROPE
clusion-freepyrope bands next to liquid can be considered as stable becausea liquid phase is invariably required as a fluxing agent for equilibrium to occur
(Gasparik,1989).
The experimentalresultslisted in Table I are shown
in Figure 2. Pyrope is interpreted to melt congruently
between7 and 16 GPa, and this is consistentwith previous studies between 3.5 and l0 GPa (Boyd and England, 1962; Irifune and Ohtani, 1986).There are many
observationsin support ofthis conclusion.Pyrope is the
only crystallinephasethat is found togetherwith liquid,
and the two phasesare separatedby a very sharp interface
(Fig. l). Electron microprobe analysesof inclusion-free
pyrope crystalsand rastor microbeamanalysesof the adjacent liquid yielded stoichiometricpyrope within analytical error. No other components,including Re from
the container,are containedin pyrope. The bands, 1020 pm in width, of inclusion-freepyrope next to liquid
are not an artifact of incongruentmelting,but ratherthey
are most reasonablyinterpreted as arising from the crystallization and melting that occurred from a temperature
fluctuationof about + l0 'C, which is typical of multianvil experiments in the 2000 'C range.
Although we have not explored in detail the effect of
starting material preparation on the experimental results
by conducting duplicate Z-P experimentson both the
mixed oxidesand the crystallinepyrope, the melting curve
shown in Figure 2 describesthe results for both starting
materialsequallywell. Similarly,our experimentalresults
at 7-8 GPa usingmixed oxidesas a startingmaterial are
in excellentagreementwith the resultsof Irifune and Ohtani (1986),who usedcrystallinepyrope. The major differenceis in the abundanceof mixed oxidesthat are preserved. In experimentsusing mixed oxides, large and
inclusion-freepyrope crystalsinvariably grow in bands
50-100 pm wide adjacentto liquid, but the crystalscontain a much greaterabundanceofunreactedoxidesin the
melt-freecold parts of the charges.
Shown in Figure 2 are out melting temperaturescompared with those reportedby Irifune and Ohtani (1986).
At 7 and 8 GPa there is excellentagreement,but at 9 and
10 GPa our melting temperaturesare about 50'C higher.
This difference may be due to ambiguities arising from
the rapid transformationof the graphite heatersto diamond in the experimentsreportedby Irifune and Ohtani
(1986) above at 2000 'C and 9-10 GPa. At pressures
above 10 GPa, our melting temperaturesare 100-350'C
higherthan the Kraut-Kennedyextrapolatedcurve of Irifune and Ohtani (1986). But this extrapolatedcurve is
problematicalbecauseit is heavily weightedby their ambiguousmeasurementsat 9-10 GPa.
p esoo
o
)
(E
o
CL
E
p zooo
2
6
10
14
18
PressureGPa
resultsand a linearmeltingcurve
Frg.2. The experimental
: this
for pyrope.Circles: IrifuneandOhtani(1986);squares
study.Opensymbols- solid;solidsymbols: quenchliquid.
The dash-dotted
curveis thedataoflrifune andOhtani(1986)
equation.
fittedby themto a Kraut-Kennedy
Z - ( ' C ) : 1 5 5 5+ 5 8 . 4 P ( G P a )
(l)
to within +40 "C, which is about the magnitudeof the
error in the measurementof temperature.But from a
thermodynamicpoint of view, a fusion curve that is linear is difficult to understandbecausemelting involves
important changesin volume, and this is usuallymanifest
in melting curves that are indeed curved in I-P space.
An attempt was made to model the melting of pyrope by
minimizing the Gibbs free energyof melting (Fei et al.,
1990),and the resultsare shown in Figures3 and 4. Although the data are not of sufficientaccuracyto determine whether the melting of pyrope is linear or curved,
the calculationsthat follow are useful for the purposeof
developingan understandingof our experimentalobservatrons.
The Gibbs free energyof melting has been minimized
using standardthermodynamicequationsfor Gibbs energy, enthalpy, and entropy (e.g.,Fei et al., 1990).The
heat capacityin theseequations,given by Cr: a 1- bT
-t c/72 -r e/73 + g/T (e.g.,Fei et al., 1990), and the
coefficients in Table 2 arc a description of calorimetric
measurements(T6qui et al., 1991; Richet and Bottinga,
1984).The volume changeof melting [i.e.,AZ(I,P)]has
been calculated using a third-order Birch-Murnaghan
equation of state with several modifications. Although it
is common to treat the bulk modulus (Kr) as a linear
function of temperature [e.g., K, : Krn" + dK/dT(I' 298):Fei et al., 19901,we usedKr: l/(Po+ ltT + P2T2
+ 0rT'), where B, are the coemcientsof compressibility.
DrscussroN
This resultsin better internal consistencyamongheat caThe simplest way of describing the experimental data pacity,thermal expansion,and the bulk modulusfor data
is with a meltingcurve that is actuallylinear in 7"-Pspace, systemizationwith the thermodynamicconstraint C" :
and this is shown in Figure 2. The experimental data can Cv + IPKTVT (Saxena,1988; Saxenaand Zhang, 1989
be describedby:
Saxenaand Shen,1992).The term V(1,298)/V(P,298)tr
ZHANG AND HERZBERG: MELTING OF PYROPE
500
TABLE2. Thermodynamicpropertiesof Mg"AlrSi"O,,pyrope and
liquid
Pyrope
Parameters
p zsoo
I5
6
o
CL
E
p zooo
d (K
1) :
ao t
a6(10 n)
o, (10 3)
d2
2
Liquid
-5697713
-5456713
1005.5
11 5 9 . 6
Heat capacity C" U/(mol K)l : a + bT + clT' + elT3 + glT
a
519.70
630.00
b
0.0139
0.035
0
.
5
1
6
1
c(x 10?)
-0.5074
e(x 10s)
-0.4194
g1x 105)
113.37
136.50
%,s (cmimol)
AHgs?o(J/mol)
ASg5,oU/(mol.K)l
6
pressJlecpa
14
18
Fig. 3. The experimentalresultsand calculatedmelting curves
for pyrope. Melting curves were calculated for completely ordered pyrope using the thermodynamic and elastic parameters
listed in Table 2. The three melting curves show how changesin
the pressurederivative of the bulk modulus K', for the liquid
can afect the results. Note that higher values of K'. stabilize
pyrope relative to liquid. The symbols are as for Fig. 2.
the original Birch-Murnaghan
by V(l,T)/ V(P, 7), where
K,(GPa) :1l,i
0 o( 1 0 u )
B, (10 ,o)
A,(0-'o)
0.(10'u)
K',
a,T I
a"lT2
0.2428
0.565
0.2578
-o 4421
fl: (Po+ PjT + A2T2+ fuTsl
0.5501
0.8643
0.4798
0.1398
4.93
5.4054
6 5 (3-10 GPa)
Notei pyrope: AH?5?0,
Charlu et al. (1975) and T6qui et al. (1991);
AS?s,0,Haseltonand Westrum(1980)and Tequi et al. (1991);C", Tequi
et al. (1991); %,s, Robieet al. (1978);a, Skinner(1966);B, O'Neillet al.
(1991); ,(',, Webb (1989).Liquid:AH!.,0 and AS?s7o,
T6qui et al. (1991);
Cp, Richetand Bottinga(1984);o and P, Lange and Carmichael(1987);
V2saandK',, this study; AH957o
and AS9u,o: the standardenthalpyand
entropyat 1570 K; G: heat capacity; y3e€: molarvolumeat 298 K; c
: thermal expansion; Kr and P : isothermalbulk modulusand compressibility, respectively; K'r: pressure derivativeof the bulk modulus.
equation has been replaced
v(t,r):o,,[.-r(J:
" .4]
Zr* is the molar volume at 298.15 K and I atm, and a
is the coefficient of isobaric thermal expansion as given
by o : ao * a,T * ar/72. Heat capacity,thermal expansion, and bulk modulus data are listed in Table 2. The
high-temperaturebulk modulus of pyrope calculatedfrom
the Co - Cu relation is in good agreementwith the data
of Anderson et al. (1991) within the temperaturerange
of their measurements.
The molar volume of MgrAlrSirO,, liquid from Lange
and Carmichael(1987) yields a calculatedmelting curve
for pyrope that is 100-250 "C higher than what we observein the range3-10 GPa. A more successfulsimulation is obtainedwith a molar volume that is 3.4ololower
(Table 2), and the resultsare shown in Figure 3. At pressures <10 GPa, we calculatea melting curve that is in
excellent agreementwith our experimental data for the
following MgrAlrSi.O,, liquid parameters:tr/!n,: 136.5
+ 4 . 0 c m 3 / m o lK, . : 1 8 . 5+ 1 . 0G P a ,a n d K ' r : 6 . 5 +
1.5, where Z!n, is the molar volume at 298 K, K, is the
isothermal bulk modulus, and Ki is the pressurederivative of the bulk modulus. The errors associatedwith
theseestimatesare basedon errors reported for the entropy and enthalpy of melting (T6qui et al., l99l), uncertaintiesin temperature(above),and uncertaintiesin
the molar volume of crystallinepyrope (Skinner, 1956).
But at pressures> l0 GPa, our preferredvolume parametersyield melting temperaturesthat are too low. By increasingK', from 6.5 to 10, it is possibleto calculatea
melting temperatureat l6 GPa that matchesour exper-
imental observations,but it yields melting temperatures
that are too high in the range 3-10 GPa (Fie.3). We
cannot produce a melting curve from a singleset of elastic
constantsthat fits the entire data basein the range3-16
GPa.
Another possibleway of describingthe melting curve
is shown in Figure 4. It has a slope that gradually decreasesfrom 82'ClGPa at 3.5 GPa to 35 "C/GPa at 9l0 GPa (Boyd and England, 1962; Irifune and Ohtani,
1986); at pressuresabove l0 GPa, our data require a
slope that is about 65 "C/GPa, indicating the possible
existenceof a breakingin the slopeof the melting curve.
The increasedstabilizationof pyrope relativeto liquid in
the range 10-16 GPa indicatesthe occurrenceof structural changesin either pyrope or liquid. The most obvious possibility is a phase transformation in crystalline
pyrope.To testthis idea,a synthesisexperimentwas conducted aI 14.5GPa and 2200"C for 5 h. X-ray diffraction
of the recoveredsample yielded a standardJCPDS pyrope pattern, with minor peaks attributed to unreacted
corundum, indicating that the garnet is somewhatdeficient in AlrOr. As discussedabove, equilibrium was
probably not reachedbecauseof the absenceof a liquid
phase in this experiment. However, the results suggest
that no phase transformation occurred in crystalline pyropic garnet,unlessit is fundamentally nonquenchablein
nature.
If cation disordering occulred in the crystallographic
sites of pyrope, the configurational entropy would lower
its Gibbs energy,pyrope would becomestabilized relative
to liquid, and this could changethe T-P slopeof the melting curve. In addition, some disorderingin pyrope may
be anticipated becausedisordering of Mg and Si has been
ZHANG AND HERZBERG: MELTING OF PYROPE
501
observedin the octahedralsitesin MgSiO, majorite garnet synthesized
at 2000'C (Phillipset al., 1992).From
Liquid
the thermodynamicparametersin Table 2, we calculate
that only 9.8 and 21.3 J/(mol.K) are neededto raisethe
melting temperaturesfrom a completelyorderedpyrope
o
BO
melting curve (seeFig. 3; K'r: 6.5) to temperatureswe
determinedat l3 and 16 GPa, respectively.
I=
o
o o
Pyrope, with a structural formula tsll\{g.lel[lrt+1$irQ,r, E
tr tr
o
E
-+'
has Si in tetrahedralsites,Al in octahedralsites,and Mg
lrilune and
CL
otfani (1986)
E
in dodecahedralsitesin its completelyorderedstate.The
o 2000
coo
a
spectroscopic
studiesof McMillan et al. (1989)on pyrope
oo-o
Pyrope
synthesizedat 900-1000'C showedno disordering,but
the site occupanciesof interesthereare thoseabove2 100
"C. The configurationalentropy of cation disorderingis
calculatedfrom the relation S. : -nR[> Xlln X1], where
I 700
2
6
14
18
n is the number of 7 sitesper formula unit, R is the gas
pressrl9e
epa
constant, and Xji the cation site occupancyof an i atom
(l : Mg, Al, Si) on a 7 crystallographic site (7 : the doresultsand a meltingcurvefor pyFig. 4. The experimental
with a pressure
decahedralsite preferred by Mg, the octahedral site pre- rope.The curvefrom 3 ro 10GPawascalculated
ferred by Al, and the tetrahedral site preferred by Si). As derivativeof the bulk modulusK;: 6.5,from Fig. 3. At l0curveis thedata
no crystallographicor spectroscopic
information on high- l6 GPathecurveis empirical.Thedash-dotted
temperaturepyrope exists, we consider a few possible oflrifune andOhtani(1986)fittedby themto a Kraut-Kennedy
equation.The brokencurveindicatesthe pressurerangewhere
cases.For the total mixing of Al and Si in two octahedral
a breakin the meltingcurveis possible.The symbolsareas for
sites and three tetrahedralsites,the configurationalen- Fig.2.
tropy is 28 J/(mol'K). In the unlikely eventthat Al should
mix with Mg in the unusualdodecahedralsite, the total
mixing of Mg and Al in three dodecahedralsitesand two volves how thesestructuralchangesaflectthe volume of
octahedralsitesis also 28 J/(mol.K). The maximum pos- pyrope liquid.
sible configurational entropy is 37.4 J/(mol'K) from the
For pyrope that is assumedto remain fully orderedat
mixing of Mg and AI in three dodecahedralsitesand one all conditions, the only way that melting temperatures
octahedralsite, in addition to the mixing of Al and Si in can be raisedis by increasingK', for MgrAlrSi.O,, liquid
one octahedralsite and three tetrahedralsites.Whatever from 6.5 at P < l0 GPa to about 10 at 16 GPa, as inthe casemay be, theseconfigurationalentropy valuesare dicatedin Figure 3. The effectthis has on the molar volsignificantlyhigher than the 9.8 ard 21.3 J/(mol.K) of ume of Mg.AlrSi,O'' liquid is shown in Figure 5 and is
extra entropy that are neededto raise the melting tem- calculatedfrom the melting curve (Fig. 4) using thermoperature of ordered pyrope to the temperaturesthat we chemicalparametersfor a fully orderedpyrope(Table2).
experimentallyobservedat l3 and l6 GPa, respectively, It is demonstrated that there is a tendency for
indicating that disorder may be partial and temperature Mg.Al,Si,O,, liquid to stiffenat > 10 GPa and 2000 'C,
dependent.Another sourceofextra entropy in pyrope is an arbitrarily selectedtemperature.Rigden et al. (1988)
from positional disorder of the small Mg cation at high also observedthis kind ofbehavior from shock-waveextemperaturesif thermal expansionis accommodatedby perimentson molten Ano,uDiouo,a compositionthat cona substantialincreasein the sizeofthe largedodecahedral tains 15.4olo
AlrOr. The volume in their experimentswas
site. In any case,it is concludedthat cation disordering observed to changegradually over the pressureinterval
in pyrope at temperaturesabove 2100 'C is a possible from I atm to 25 GPa, but negligiblyat higherpressures.
way of explainingour experimentaldata and may explain Rigden et al. (1988) interpretedtheir resultsto indicatea
why the melting curve has the break in slopeat l0 GPa gradualtransformationof Al and Si from dominantly tet(Fig. a).
rahedralto octahedralcoordinationin the range I atm to
We concludethis analysiswith an examinationof pos- 25 GPa; at higher pressuresthey suggestedthat these
sible changesin the properties of pyrope liquid. The most transformationsmay be essentiallycomplete,and there
obvious considerationinvolves possibleAl3+ coordina- may be a similarity in the compressibilitiesof the liquid
tion changesin the liquid from fourfold to sixfold coor- and an increasein the melting slopeof MgSiO. perovskite
dination at around l0 GPa, where there may be a break at around 60 GPa, but this break has not been verified
in the melting slope.The evidencefor coordination in- by Zen andBoehler(1993).This interpretation,however,
creasesin Al3* at about this pressurewas from'z?Alsolid- is ambiguous becauseP-V breaks were not observed in
state NMR measurementson silicateglasses(Ohtani et shock-wave experiments for molten anorthite (36.60/o
al., 1985;Xue et a1.,1991),moleculardynamicssimu- AlrOr: Rigdenet al., 1989)and for molten komatiite(8.20lo
lations (Angell et al., 1982, 1987),and the observationof AlrO.: Miller et al., l99l) over a comparablepressure
garnetquenchedfrom liquid pyropeat pressures>7 GPa range. Furthermore, the melting curve of jadeite (Na(Irifune and Ohtani, 1986). The essentialquestion inAlSi,O6) from2.4 to 16.5GPa showsno I-Pbreaks (Fig.
502
ZHANG AND HERZBERG: MELTING OF PYROPE
160
f = 2000'C
-g
o
E 14o
o
E
C)
o
E 12o
5
o
100
o
5
10
20
15
Pressure GPa
Fig. 5. PossibleP- Zrelationsof liquid MgrAl,SirO,, at 2000
"C derived from the melting of ordered pyrope along the melting
curve in Fig. 4. The solid curve was calculatedfrom the elastic
and thermodynamic parameterslisted in Table 2, with K,, raised
from 6.5 to 10.2at pressuresin the range 10-16 GPa. The broken curvewascalculatedwith K;: 6.5 (Fig. 3). Errorsstemming
from uncertainties in temperature, enthalpy, and entropy of
melting and from the molar volume of crystalline pyrope are +4
cm3/mol.
6) that could be attributed to structural changes that are
known to occur in albite glass (Ohtani et al., 1985). The
evidence is therefore somewhat contradictory, but the bulk
of it indicates that disordering in pyrope may be more
important than structural changes in liquid MgrAlrSirO,r.
Spectroscopic studies on pyrope synthesized at around
2700
Py
2200
o
f
6
o
CL
E
1700
Solid
1200
0
Acxxowr,nocMENTs
This researchwas partially supportedby grants from the National ScienceFoundation to Claude Herzberg(EAR-89 I 6836 and EAR-9 I I 7 l 84).
The high-pressureexperimentsreported in this paper were performed in
the Stony Brook High PressureLaboratory, which is jointly supportedby
the National ScienceFoundation (EAR-8917563) and the State University of New York at Stony Brook. This laboratory is now a part of the
NSF Scienceand Technology Center for High PressureResearch(EAR8920239)establishedat Stony Brook in conjunction with Princeton University and the Geophysical Laboratory of the Carnegie Institution of
Washington Thanks are extended to J. Bass, G. Harlow, G. Miller, E.
Ohtani, and L. Stixrude for their helpful discussionsand comments. Special thanks are extendedto Tibor Gasparik for technical assistance.This
is Mineral PhysicsInstitute Publication no. 78.
Rnrnnrxcns crrED
Liquid
o
o
2500 'C could be important in constraining these
possibilities.
Figure 6 summarizesthe melting curve of pyrope in
relation to the melting curves of enstatite(Boyd et al.,
1964;Presnalland Gasparik, 1990),forsterite(Davis and
England,1964; Ohtani and Kumazawa,l98l; Presnall
and Walter, 1993), and jadeite (Litvin and Gasparik,
1993).The very high melting temperaturesobservedfor
pyrope at around l5 GPa show that pressurestabilizes
garnetat the expenseof forsterite,in agreementwith phase
equilibrium studies in multicomponent systems(Takahashi,I 986;Ohtaniet al., I 986;Ito and Takahashi,1987;
Herzberget al., 1990; Zhang and Herzberg, 1993). We
emphasizethat the experimentalmethod utilized in this
study is identical to that used for the enstatite and forHerzsterite fusion curves (Presnalland Gasparik, 19901,
berg et al., 1990; Presnalland Walter, 1993),and it is
also identical to the method usedin calibratingthe solidus curve in peridotite systems(Herzberget al., 1990).
Therefore,the resultspresentedhere cannot be an artifact
of the experimentalmethod.
6
12
18
PressureGPa
Fig. 6. The melting curve of pyrope compared with those for
enstatite (Presnall and Gasparik, 1990), forsterite (Presnall and
Walter, 1993),and jadeite (Litvin and Gasparik, 1993).
Anderson,O.L., Isaak,D.L., and Oda, H. (1991)Thermoelasticparameters for six minerals at high temperature.Journal ofGeophysical Research,96, 18037-l 8046.
Angell, A.C., Cheesman,P., and Tamaddon, S. (1982) Pressureenhancement of ion mobilities in liquid silicates from computer simulation
studiesto 800 kilobars.Science,218,885-887.
Angell, A C., Cheesman,P., and Kadiyala, R.R. (1987) Diffusivity and
thermodynamic properties of diopside and jadeite by computer simulation studies.ChemicalGeology,62,83-92.
Boyd, F.R., and England, l.L (1962) Mantle minerals. CarnegieInstitution ofWashingtonYear Book, 61,107-l12
Boyd, F.R, England,J.L., and Davis, B.T C. (1964) Effectsof melting
and polymorphism of enstatite, MgSiO.. Journal of Geophysical Research,69, 210l-2109.
Charlu, T.V., Newton, R.C., and Kleppa, O.J. (1975)Enthalpiesof formation at 970 K of compounds in the system MgO-Al,O,-SiO, from
high temperature solution calorimetry. Geochimica et Cosmochimica
Acta.39. 1487-1497.
Davis, B.T.C., and England, J.L. (1964) The melting of forsterite up to
50 kilobars JournalofGeophysicalResearch,69, I I l3-l I l6
Fei, Y., Saxena,S K , and Navrotsky, A. (1990) Intemally consistentdata
and equilibrium phase relations for compounds in the system MgOSiO. at high pressureand high temperature.Journal of Geophysical
Research.95. 69 | 5-6928.
Gasparik, T. (l 989) Transformation of enstatite-diopside-jadeitepyroxenesto garnet. Contributions to Mineralogy and Petrology, 102, 389405.
ZHANG AND HERZBERG: MELTING OF PYROPE
-
(l 990) Phaserelations in the transition zone Journal of Geophysical Research,95, 1575 1-1 5769
Haselton,H T , and Westrum, E.F. (1980)Thermodynamicsof grossularpyrope garnetsand their stabilities at high temperaturesand high pressures.Journal of GeophysicalResearch,85, 6973-6982
Herzberg, C. (1983) Solidus and liquidus temperaturesand mineralogies
for anhydrous garnetJherzolite to 15 GPa. Physics of the Earth and
PlanetaryInteriors, 32, 193-202
-(1992)
Depth and degreeof melting of komatiites. Journal of Geophysical Research,97, 4521-4540.
Herzberg,C., and Gasparik, T. (199 l) Garnet and pyroxenesin the mantle: A test of the majorite fractionation hypothesis. Journal of GeophysicalResearch,96, | 6263- | 6274
Herzberg, C., Gasparik, T, and Sawamoto, H (1990) Origrn of mantle
peridotite: Constraintsfrom melting experimentsup to 16.5GPa Journal ofGeophysicalResearch,95, 15779-15803.
Irifune, T., and Ohtani, E. (1986)Melting of pyropeMg,Al,Si,O,. up to
l0 GPa: Possibility ofa pressure-inducedstructural changein pyrope
melt. Joumal of GeophysicalResearch,9 l, 935l -9366.
Ito, E., and Takahashi,E. (1987)Melting ofperidotite under the lower
mantle condition Nature, 328, 514-517
Knittle, E , and Jeanloz,R (1989) Melting cuwe of (Mg,Fe)SiO,perovskite to 96 GPa, evidence for a structural transition in lower mantle
melts.GeophysicalResearchl-etters,16, 421-424.
Iange, R A, and Carmichael,I S.E.(1987)Densitiesof Na,O-K.O-CaOMgO-FeO-Fe.O.-Al.O.TiO,-SiO. liquids: New measurementsand derived partial molar properties Geochimica et CosmochimicaActa, 5 l,
2931-2946.
Liebermann, R.C , and Wang, Y. (1992) Characterizationof sample enyironment in a uniaxial split-sphereapparatus.In Y. Syono and M.H.
Manghnani, Eds., High-pressureresearch:Application to Earth and
planetary sciences,p. 19-31. Terra Scientific, Tokyo, and American
GeophysicalUnion, Washington, DC
Litvin, Y.A., and Gasparik,T (1993)Melting ofjadeite to 16.5GPa and
melting relations on the enstatite-jadeitejoin Geochimica et Cosmochimica Acta. 57. 203f-204O.
McMillan, P., Akaogi, M., Ohtani, E, Williams, Q., Nieman, R., and
Sato, R. (1989) Cation disorder in garnetsalong the MgrAl'Si.O,,MgSi.O,, join: An infrared, Raman and NMR study. Physics and
Chemistryof Minerals, 16,428-435.
Miller, G.H., Stolper,E.M , and Ahrens,T.J. (1991)The equationof state
of a molten komatiite. I. Shock wave compressionto 36 GPa Joumal
ofGeophysicalResearch,96, I l83l-l 1848.
Ohtani, E , and Kumazawa, M. ( 198I ) Melting of forsterite Mg.SiO, up
to I 5 GPa. Physicsof the Earth and PlanetaryInteriors, Z7, 32-38.
Ohtani, E., and Sawamoto,H (1987)Melting experimentson a model
chondritic mantle composition at 25 GPa GeophysicalResearchLett e r s .1 4 , 7 3 3 - 7 3 6 .
Ohtani, E , Irifune, T , and Fujino, K. (1981) Fusion of pyrope at high
pressuresand rapid crystal growth from the pyrope melt. Nature, 294,
62-64
Ohtani, E., Taulelle,F, and Angell, C.A. (1985)AIt coordinationchanges
under pressure.Nature, 314, 78-81.
in liquid aluminosilicates
Ohtani, E., Kato, T., and Sawamoto,H. (1986)Melting of a model chondritic mantle to 20 GPa Nature, 322, 352-353
O'Neill, B., Bass,J D, Rossman,G.R., Geiger,C.A., and Langer,K.
( I 99 l) Elastic propertiesofpyrope. Physicsand Chemistry ofMinerals,
t7 - 617-621
503
Phillips, B P, Howell, D.A., Kirkpatrick, R.J', and Gasparik, T. (1992)
rTAl
Investigation ofcation order in MgSiO.-rich garnet using'nSi and
MAS NMR spectroscopyAmerican Mineralogist, 77, 704-712Presnall, D.C., and Gasparik, T. (1990) Melting of enstatite from l0 to
eutecticat 16.5GPa, and implica16.5GPa, the beta phase-majorite
tions for the origin ofthe mantle. Joumal ofGeophysical Research,95,
1577l-15777 .
Presnall. D.C., and Walter, M.J (1993) Melting of forsterite, Mg.SiO"'
from 9.7 to 16.5 GPa. Joumal ofGeophysicalResearch,9S'1977719783
Richet,P., and Bottinga,Y. (1984)Anonhite, andesine,wollastonite,diopside,cordierite, and pyrope: Thermodynamicsofmelting, glasstransitions, and properties of the amorphous phases.Eanh and Planetary
ScienceLetters, 67, 4 15-422.
Rigden,S.M , Ahrens,T.J., and Stolper,E.M (1988)Shockcompression
of molten silicate:Resultsfor a model basalticcompositlon. Journal of
GeophysicalResearch,93, 367-382.
-(1989)
High-pressureequation of state of molten anorthite and
diopside Journalof GeophysicalResearch,94, 9508-9522.
Robie, R.A., Hemingway,B.S., and Fisher,J. (1978) Thermodynamic
properties of minerals and related substancesat 298.15 K and I bar
(105Pascals)and at high temperatures U.S. GeologicalSurvey Bulletin,
1 4 5 2 , 4 5 2p
Saxena,S K. (1988) Assessmentofthermal expansion,bulk modulus, and
heat capacity ofenstatite and forsterite Physicsand Chemistry ofSolids.49. 1233-1235.
Saxena,S.K., and Shen,G. (1992) Assessmentof thermophysicaland
thermochemical data in some oxides and silicates. Journal of GeophysicalResearch,97, 198I 3-19825
Saxena,S.K., and Zhang, J. (1989) Assessedhigh-temperaturethermochemical data on some solids. Physics and Chemistry of Solids, 50'
723-727
Skinner, B.J. (1956) Physical properties of end-members of the gamet
group American Mineralogist, 41, 428-436
-(1966)
Thermal expansion. In Geological Society of America
Memoir.97.75-95
Takahashi,E (1986)Melting of a dry pendotite KLB-I up to 14 GPa:
Implications on the origin ofperidotitic upper mantle. Journal ofGeophysical Research,9 l, 9367-9382.
T6qui, C., Robie,R.A , Hemingway,B.S , Neuville, D.R., and Richet,P.
(1991)Melting and thermodynamicpropertiesof pyropeMg,Al'Si.O'.
Geochimicaet CosmochimicaActa, 55, 1005-1010.
Webb, S L (1989)The elasticity of the upper mantle orthosilicatesolivine
and garnet to 3 GPa. Physicsand Chemistry ofMinerals, 16' 684-692.
Xue. X., Stebbins,J F, Kanzaki,M , McMillan, P.F.,and Poe' B. (1991)
Pressure-inducedsilicon coordinatlon and tetrahedral structural changes
in alkali oxide-silica melts up to 12 GPa: NMR, Raman, and infrared
spectroscopyAmerican Mineralogist, 76, 8-26 to
Zen, A, and Boehter,R. (1993) Melting of (Mg,Fe)SiO,-perovskite
625 kilobars: Indication of a high melting temperature in the lower
mantle Science.262. 553-555
Zhang, J, and Herzberg, C (1993) Melting experiments on peridotite
KLB-l to 22.5 GPa (abs). American GeophvsicalUnion Spring Meeting,14,345.
Mev 10, 1993
Meruscnrm REcEIvED
Meu-rscnrrr AcCEPTED
Jnm;env 19, 1994