REVIEWS OF GEOPHYSICS AND SPACE PHYSICS, VOL. 20, NO.

4, PAGES 827-849, NOVEMBER 1982

Thermodynamics and Lattice Vibrations of Minerals'

Applications to Phase Equilibria, Isotopic Fractionation,
and High-Pressure Thermodynamic Properties
SUSAN WERNER KIEFFER

U.S. Geological Survey, Flagstaff, Arizona 86001

In previous papersin this series, a model that usesavailable elastic, structural, and spectroscopic
data on minerals has been used to predict the thermodynamicfunctions Cv (heat capacity), S
(entropy), E (internal energy), and F (Helmholtz free energy). In this paper, four applicationsto
problemsof currentgeochemicaland geophysicalinterestare presented:(1) interpretationof complex
trends of calorimetric data; (2) calculation of phase equilibria; (3) calculation of oxygen-isotopic
fractionation factors; and (4) estimation of the effect of pressure on thermodynamic functions. The
model demonstrates that trends in high-temperature thermodynamic properties of silicates are
determined by the position and relative numbersof high-frequencymodes, generally antisymmetric
(Si, AI)-O stretching modes. The position of these modes varies systematically with degree of
polymerization of tetrahedra, and therefore high-temperaturecalorimetric behavior is relatively
systematicas a function of crystal structure and mineral composition.Trends at low frequency are
much more complex because the low-frequency optic modes that most strongly influence the low-
temperature thermodynamic functions dependin a complex way on the size, coordination, and massof
cations and various polyhedra in the minerals. The heat capacity curves of kyanite, andalusite, and
sillimaniteand of quartz, coesite, and stishoviteshow crossoversthat cannot be explainedby Debye
theory, which accountsonly for acousticmode behavior, but can be explainedby the model spectra
proposedbecauseproper accountis taken of the changinglow- and high-frequencyoptic modesupon
polymorphic transformations.The proposedmodel is sutficientlyaccurate that phase equilibrium
problems can be addressed:the quartz-coesite-stishoviteequilibrium curves, the kyanite-andalusite-
sillimanite triple point, and the breakdown of albite to jadeite-plus-quartz are cited as specific
examples. For each example, predicted slopes of equilibrium curves agree moderately well to
excellently with slopesdeterminedexperimentally.The calculatedslopesare sensitiveto spectrosco-
pic parameters, particularly to the distribution of optic modes in the far infrared; this sensitivity is
discussedin detail for the albite breakdownreaction. The model can be usedfor predictionof isotopic
fractionationfactors if spectraof the isotopicforms of the mineral are known or postulated.A simple
set of 'rules' for generatinghypotheticalspectraof •80 mineralsfrom measuredspectraof the •60
forms is given. Reduced partition functions are calculated for 13 minerals. At 298øK the model values
of reduced partition function, 1000 In a, of these minerals decreasein the order quartz > calcite •>
albite > muscovite > clinoenstatite • anorthite > diopside > pyrope > grossular> zircon > forsterite
> andradite > rutile, in good agreement with experimental data. At 1000øKthe first six minerals show
small crossovers so that the order becomes calcite, muscovite • albite, quartz, anorthite, and
clinoenstatite; the differences in 1000 In a at high temperature for these minerals are so small that the
model probably cannot addressthe deviationsfrom experimental trends. The model clearly definesthe
region in which the fractionationfactorsdo not follow a 1/T2 trend and shouldbe useful for
extrapolation of experimental data to low temperatures. Finally, a modified Griineisen parameter
model is proposedfor shift of the lattice vibrational frequenciesunder compression,and thermody-
namic properties to 1000 kbar, 1000øK,are given for nine minerals. At 1 Mbar, the predicted decrease
in entropy at 298øK rangesfrom 54% (of the l-bar value) for periclaseto 25% for stishovite.

CONTENTS Concluding comments..................................... 847

Notation ................................................. 847
Introduction ............................................. 827
Interpretationand predictionof thermodynamicfunctions..... 828
Phaseequilibria .......................................... 830 1. INTRODUCTION
SiO2 polymorphs(quartz, coesite, stishovite).............. 831
AI2SiO5equilibria....................................... 832 A modelfor estimatingthermodynamicpropertiesof min-
Stability of albite ....................................... 833 erals from elastic, structural, and spectroscopicdata was
Isotopic fractionation ..................................... 833 proposedin the first four papers of this series [Kieffer,
Frequency shifts upon isotopic substitution:A simple
set of rules for silicates ................................ 836 1979a,b, c, 1980](hereinafterreferred to as papers 1, 2, 3,
Results ................................................ 839 and4). The purposeof this paperis to illustrate,in a general
Effect of pressureon the thermodynamicfunctions........... 841 way, application of the model to geologic problems of
Griineisenparameters.................. : ................ 841 interest.
Calculation of thermodynamic properties at high
pressure............................................. 844
Four applicationsof the modelare given in this paper: (1)
Results ................................................ 845 interpretation of complex trends of calorimetric data and
prediction of thermodynamicproperties of minerals for
which calorimetry cannot be performed; (2) calculationof
This paper is not subject to U.S. copyright. Publishedin 1982 by
the American Geophysical Union.
phaseequilibria;(3) predictionof oxygenisotopicfraction-
ation factors;and (4) estimationof the effectof pressureon
Paper number 2R1228. thermodynamic functions.
827
828 KIEFFER: THERMODYNAMICS AND LATTICE VIBRATIONS OF MINERALS, 5

For detailed applicationsin problemsof current research,
knowledgeof mineral properties beyond the scope of the
particular thermodynamic study of papers 1-4 is often
required: for example, phase equilibria calculations may
require inclusionof order-disorderphenomena,and magnet-
ic or configurationaleffects; isotopic fractionation calcula-
tions may require considerationof the detailedbehaviorof
compoundscontaininghydrogen ions or water molecules;
andpressuredependencecalculationsrequiredetailedevalu-
ation of the behavior of the thermal-expansionand GrQnei-
senparameters,both of which are difficultand controversial
subjects.Therefore most of this discussionis meant to be
illustrative, usingcommon, well-studiedmineralsand phase
equilibria, in order to demonstrateboth the potentialand the
limitations of the model.
2. INTERPRETATION AND PREDICTION
OF THERMODYNAMIC FUNCTIONS
As was mentioned in papers 1-4, the model can be used
either to predict heat capacitiesindependentlyof calorimet-
ric data or to extrapolateheat capacity data to higherand/or
lower temperaturesthan thoseattainedby calorimetrictech-
niques; for example, heat capacitiesfor coesite and stisho-
vite could be predicted above 300øK at which temperature
experimental data cannot be obtained because the phases
becomemetastable, and heat capacitiescan be estimatedfor
many minerals where data below 54øK are not available. For
example, Gaffney [1982] has used this model to predict the
heat capacitiesof water ices for which calorimetric data are
not available. For data extrapolation, a modified model that
finds the best fit between any available calorimetric data and
a simple vibrational spectrum was suggestedfor obtaining
greater accuracy (paper 4).
The model can also be used to indicate possible inconsis-
tencies among various data sets: for example, calorimetric
data on stishovitecould not be reproducedusingelastic data
and reasonablespectroscopicmodels (paper 3); calorimetric
and low-frequency spectroscopicdata on enstatite seem to
be inconsistent(paper 4); and acousticvelocities estimated
for brucite from shock wave data give model heat capacities
much higher than measuredheat capacities(without includ-
ing any contribution from optic modes) (paper 3). If the
measuredbrucite heat capacity data are correct, the infer-
ence would be that the estimated wave velocities are too

I I I I I ! I I I

(a)

O.Si,AI=2 COES
ALBT/MICR

CRIS

MuSC(e•o-1000•)
ENST
1.6 1150' 1.6

1.4 •usC•
-•••-•
DIOP
(e•=o-lO2O o)
1.4

0:Si,A1=2.5-3 o,o
.... 1.2
1.0

-- t8 08

212)
0-6 SILL (ec•,..-,12oo •)
06
0.4 ) O.4
O:Si,AI=4 z,.o
OROS (eoo.--..'K)oo'")
/ •/
KYAN (e•-.-,lOOO •) '"

1.6 1.6
1.4 1.4

1.2
OTHERS 1.2
1.0
.•;;•'•'CRND
(e•""945ø)...... J 1.0
0.8
0.6

0%
I
20
I
40 60
I I
80
I
100
I
120
I
140 160
I I
180 200
JpERI
300
½e•ø•'?:>O
) ---'••
0.8 I
400
i
500 600
i
0.6

0.4

TEMPERATURE (K)
Fig. 1. Curves of 0cal(T)/0elversus T for (a) low-temperatureheat capacity data and (b) high-temperaturedata
(abscissascale of Figure lb is different from that of Figure la). See Figure 4, paper 1, for notes and references.
 KIEFFER: THERMODYNAMICSAND LATTICE VIBRATIONS OF MINERALS, 5 829

2 I I I I I I I I I I I i 2
• [[(MINERAL
T3• -ANDALUSITE
) K=KYANITE
S= ] [I (MINERAL
SILLIMANITE -QUARTZ
) S=
C:
COESITE
STISHOVITE
e, I t /

-/ I I I I I I I I I i
b
--'
0 I• 200 3• 400 5• 6• 7• 8OO O I00 2• 300 400 500

TEMPERATURE (K) TEMPERATURE (K)

Fig. 2. (a) Heat capacity deviations for AI:SiO• polymo•hs; kyanite and andalusitedata are from R. Robie and B.
Hemingway (personalcommunication,1981),and sillimanitedata are from Todd [ 1950].Model valuesare from paper 4.
(b) Heat capacity deviations for SiOl polymo•hs. Data for qua•z are from Lord and Morrow [1957]; data for coesite
and stishoviteare experimental data of Holm et al. [ 1967]to 350øK. Values beyond this are from model. For stishovite,
model 2 of paper 3 was used, and, for coesite, model 1 of paper 3 because these models gave the best match to
experimental Ce* data at 300ø-350øK.

low. In model calculations for albite and sillimanite, calori-
metric data (presumed accurate) could not be matched by
model calculationsusing spectral data (infrared only) avail-
able at the time (papers 3 and 4). Additional low-frequency
vibrationswere requiredto fit the heatcapacitydata,and so
we obtained Raman spectra for these minerals. These spec-
tra showedlow-frequency Ramanlinesat 83 cm-I for albite
and at 115 cm-• for sillimanite(J. Delany, Universityof functions
withtemperature
California, Los Angeles, unpublisheddata, 1978), exactly in
positions that brought model values into good agreement
with measured values of Cv. One should always keep in
mind that one possible cause of apparent inconsistencies
between data sets indicated by the model can be that the
assumed form of the model spectrum does not adequately
resemblethe real lattice vibrational spectrum of a mineral.
However, the general agreement between predicted and
measuredthermodynamic functions for 30 minerals exam-
ined suggeststhat use of the model provides a reasonable
way to attempt to integrate elastic, spectroscopic, and
thermodynamic data sets and that substantial inconsisten-
cies call for reexamination of experimental data sets used.
The model provides insight into observed calorimetric
trends given in Figure 4 of paper 1, reproducedas Figure 1 of
this paper. The simplified vibrational spectra of the model
reveal systematic variations in various groups of modes that
appreciablyinfluencethe low- and high-temperaturethermo-
dynamic functions:
1. Trends at high temperatures (Figure lb) are deter-
mined by the position and relative numbersof the highest-
frequency modes (antisymmetric (Si, AI)-O stretching modes
for most silicates and AI-O, O-H stretchingor Si-O modes
for some other minerals studied). The position of these
modes varies systematically with degree of polymerization
of tetrahedra, and therefore high-temperature calorimetric
behavior is relatively systematic as a function of crystal
structure and mineral composition.
2. Trends at low temperatures (Figure l a) are deter-
mined primarily by the position of the lowest-frequency
optic modes and secondarily by the magnitude and relative
proportion of acoustic modes and their relationship to the
optic mode frequencies. Becausethe position of the lowest-
frequency optic modes is sensitive to the size and coordina-
tion of cations and various polyhedra in the minerals, low-
temperature calorimetric trends are not nearly as systematic
as high-temperature trends and must be more cautiously
interpreted and predicted in terms of individual mineral
properties.
The modelpredictsabsolutevaluesof the thermodynamic
functionsof mineralsbetter than a Debye or Einstein model
and can be usedto explainvariationsof the thermodynamic
thattheseone-parameter
models
cannot explain. An example discussedin paper 4 was the
relation between heat capacity (Cv) and entropy (S) of
pyrope and grossular. Measured data show that the heat
capacityof pyrope at 298øK is less than that of grossular,
whereasthe entropyis greater.This anomalyoccursbecause
of relationsin the low-frequencypart of the spectrumand is
nearlyreproducedby the theoreticalmodel(whichpredicts
that the heat capacities will be about equal instead of
reversedfrom the entropies). Comparisonof pyrope and
grossularmodel spectra(paper 4) showsthat pyrope has a
lower-frequencyoptic modethangrossular,but alsoa slight-
ly lower averagedensity of optic modes.These two effects
opposeeach other in contributingto the low-temperature
heat capacity,with the net resultof the nearequalityof Cv
for pyrope and grossularat low temperature. However,
whenC•,/T(• Cv/T at low temperature)is integratedto give
entropy,the entropyof pyropeis greaterthanthe entropyof
grossularbecausethe 1/T weightingfactor in entropy en-
hancesthe effectof the lower-frequencymodesof pyrope.
Considertwo additionalexamplesof thermodynamicbe-
havior that cannotbe explainedby a simpleDebye model:
crossoversin the heat capacitiesof kyanite, andalusite,and
sillimaniteand similar crossoversin the heat capacitiesof
quartz, coesite, and stishoviteillustrated in Figures 2a and
2b. In thesefigures,deviationcurvesshowingmeasuredand
calculatedheat capacitydifferencesbetweenmineralpairs
areshown.
Evenin thelowesttemperature
rangewherethe
model is least accurate, these curves reproducewell the
trends of the deviation curves and, reasonablywell, their
magnitudes. Therefore interpretation of the cause of this
behaviorin terms of modelparametersappearsjustified.
Consider first the AI2SiO5 polymorphs(Figure 2a). At
temperaturesbetween54øKand 350øK,their heat capacities
increase in the order kyanite, andalusite, sillimanite. The
heat capacity curves cross over each other between 350øand
830 KIEFFER: THERMODYNAMICS AND LATTICE VIBRATIONS OF MINERALS, 5

800-- 2x
'G' - Quartz
5 6 Coesite Stishovite

ua
•
600[ -

:•400[ --
i

200(0,/(C,(0,11
(S)
//
/

00 20 40 60 80 100 120 140

PRESSURE (KBARS)

Fig. 3. A partial compilationof data on SiO2phaseequilibria.Other data can be foundin the work by Weaveret Hi.
[ 1979]. The curves are from the following references: (curve 1) Dachi!!e and Roy [ 1959], (curve 2) MacDonald [ 1956],
(curve 3) Takahashi [1963], (curve 4) Kitahara and Kennedy [1964], (curve 5) Boyd and England [1960] extrapolated to
lower temperatures,(curve 6) Boyd et Hi. [1966] extrapolationof revisedcurve, (curve 7) Griggs and Kennedy [1956],
(curve 8) Ostrovsky [1967], (curve 9) Akimoto and Syono [1969], (curve 10) Stishov [1964]; and (curve 11) Bohlen and
Boettcher [1982]. The two crosseson the quartz-coesitecurve are the data of Boettcher and Wy!!ie [1968]. HKW refers
to the curvescalculatedby Holm et Hi. [ 1967]. Yagi and Akimoto [ 1976]have additionaldata on coesite-stishovite; from
their dataone may infer a slopeof about11 barsdeg-• from the coesite-stishovite
equilibrium[Weavereta!., 1979].
Jeanlozand Thompson[1982]prefera slopeof 21 barsdeg-•'

400øK, and at high temperaturesthe heat capacitiesincrease are lower than for quartz, and thus at relatively low tempera-
in the order sillimanite,andalusite,kyanite. Elastic Debye tures the heat capacity of coesite rises above that of quartz.
temperatures of the polymorphs increase in the order anda- At low temperatures, stishovite has a lower heat capacity
lusite (761øK), sillimanite (782øK), kyanite (916øK), There- than either quartz or coesite because it has a very high
fore, accordingto a Debye model, at all temperaturesthe elastic Debye temperature (hence high acoustic mode fre-
relative order of heat capacitiesshouldincreasefrom kyanite quencies) due to its high density, and also becauseits lowest
to sillimanite to andalusite; it does not. The model calcula- optic modes are at relatively high frequencies. However, the
tions show that the influenceof low-frequencyoptic modes heat capacity of stishovite rises more rapidly as temperature
is sufficientlygreat at 20øKto upsetthe relationshipsexpect- rises than that of quartz or coesite becausethe Si-O stretch-
ed from acousticvelocitiesalone. Becausethe frequencyof ing modes of stishoviteare at lower frequencies(Si in sixfold
the lowest optic modes increasesin the order sillimanite(115 coordination)than the stretchingmodesof quartz and coe-
cm-• at K = 0), andalusite
(156 cm-•), and kyanite(237 site (Si in fourfold coordination). Thus at about 300øK,
cm-•) and becausethe positionof thesemodesdominates where these stretching modes become effective, the heat
heat capacity relationships at low temperatures, the ob- capacity of stishovite rises above that of either quartz or
served order of increasingheat capacity is from kyanite to coesite.
andalusite to sillimanite. However, the optic modes of In general, crossoversin heat capacities and reversals in
kyanite are compressedinto a smallerfrequencyrange than relative entropyvaluesshouldbe expectedfor mineralpairs
thoseof andalusite andsillimanite(uppercutoff720cm- • for if relative magnitudesof elasticDebye temperatures,and/or
kYanite,780cm-• for andalusite,and955cm-• for silliman- frequenciesof lowestoptic modesor of highestoptic modes,
ite), so that relationshipsin heat capacityare reversedas the show opposingtrends. Such opposingtrends will be com-
influenceof higher-frequencyoptic modesbecomesimpor- mon, for example, acrosspolymorphicphasechangeswhen
tant at about room temperature. the influence of the density change is to drive acoustic
The heat capacities of quartz, coesite, and stishovite also frequencies upward and the influence of the coordination
show crossovers, as illustrated by the deviation curves of changeis to drive optic frequenciesin the opposingdirec-
Figure 2b. Crossoversin quartz-coesiteand quartz-stisho- tion, downward. Navrotsky [1980] recognizedthat this be-
vite are indicated by the Holm et al. [1967] data; crossovers havior could occur across polymorphic phase changesin
between coesite and stishovite can be inferred from the oxidesand could give rise to negativepressure-temperature
trends at the 350øK limit of their data and from my model slopesfor phase changesin the upper mantle.
values. The heat capacity of coesite is much more similar to
that of quartz than is the heat capacity of stishovite,because 3. PHASE EQUILIBRIA
the coordination, and hence the vibrational spectrum, is Calculation of phase equilibria provides a more stringent
more similar. The heat capacity of coesiteis lessthan that of test of the model than calculation of thermodynamicfunc-
quartz at low temperaturesbecausethe Debye temperature tionsalone, becausesmalldifferencesin the thermodynamic
is higher; however, the frequenciesof the optic continuum functions,rather than their absolutemagnitude,determine
 KIEFFER: THERMODYNAMICS AND LATTICE VIBRATIONS OF MINERALS, 5 831

TABLE 1. Entropies and Volumes of SiO2 Phases

Property a Quartz /3 Quartz Coesite Stishovite
Volume, cm3 mol-1
at 1 bar 22.69 23.72 20.64 14.02
at 30 kbar 21.33 22.29 20.02 13.88
at 70 kbar 19.97 20.87 19.40 13.74
at 90 kbar 19.51 20.40 19.20 13.73
Entropy, cal mo1-1
oK- 1 (from
model)
at 298øK 10.08 ßßß model 1 10.05 7.77 a
model 2 9.00 7.65 a
at 600øK 19.29 model 1 19.29 17.19 a
model 2 17.86 16.89 a
at 1000øK -.. 27.82 model 1 27.51 25.30 a
model 2 25.98 25.14 a
Entropy,cal mo1-1
oK- 1 [from
Robie et al.,
1978]
at 298øK 9.91 ßßß 9.65 6.64
at 600øK 19.08 ßßß 18.68 15.74
at 1000øK ßßß 27.61 26.88 23.88

aAnharmonic entropy correction uncertain.

the phase relations. Therefore errors resulting from aver- /3quartzwasobtainedby takingthe modelvaluefor a quartz
aging assumptions in the model become quantitatively at 1000øK (22.69cal mol-• øK-l) andadding0.13cal mol-•
important. The following equilibria are examined: quartz- øK-• for anapproximate entropyof transition.
coesite-stishovite,kyanite-andalusite-sillimanite, and albite- Consider,first, the quartz-coesiteequilibrium.Experi-
jadeite-quartz. mentallydeterminedcurvesshowa wide rangein slopes(2
to 37barsdeg-• havebeenproposed);
someof themeasured
Si02 Polymorphs (Quartz, Coesite, Stishovite) equilibrium curves are shown in Figure 3. However,
Experiments have suggested a wide range of stability Boettcherand Wyllie [1968] and Weaver et al. [1979] have
fields for quartz, coesite, and stishovite(Figure 3). Isobaric critically discussedexperimentaldata, pointingout that the
temperature uncertainties are of the order of 500øK for the curves having the highest slopes (curves 1, 2, and 7) are
univariant equilibrium curves. The quartz-coesite equilibri- generallyconsideredunreliable.Weaver et al. [1979]pre-
um is difficult to determine experimentally because of the ferreda slopeof 13barsdeg-•;BohlenandBoettcher[1982]
notorious sluggishnessof the quartz-coesite reaction. The prefer-- 8 barsdeg-•. The two modelsproposed for the
coesite-stishovite equilibrium is difficult to determine be- vibrationalspectrumof coesitein paper3 give substantially
cause the high pressures at which it occurs are difficult to differentvaluesof entropyand, consequently,
quitedifferent
obtain and to calibrate. The metastable quartz-stishovite valuesof slopeof the phaseboundary.The entropydiffer-
transition, of considerable interest to shock wave geophysi- encebetweenquartz and coesiteis so smallthat the large
cists and astrogeologists(see Kieffer et al. [1976], for exam- uncertaintiesin the thermal expansion, reflected in anhar-
ple), has only been determined by calculation from calori- monicentropy,appreciably
affectthecalculated
slope.The
metric data. The data of Figure 3 show, however, that slopes slopesobtained•frommodels1 and2 are6 and34barsdeg-l,
of the equilibrium curves are less certain than their position respectively.The latter value is probably not to be consid-
in P-T space. The slopes of the equilibrium curves are eredseriouslybecausethe modelvaluesof heatcapacityare
related to entropy and volume through the Clapeyron rela- so far from measuredvalueson coesite(on the other hand,
tion: Weaver et al. [1979] showedthat someproblemsmay also
dP/dT = AS/AV (1) exist with the coesite calorimetry). Although the model
valueof 6 barsdeg-•is slightly
lowerthanthatpreferred
by
Becausemy modelpredictsonly entropiesof the phases,not experimentalists
for thisboundary,it is in remarkableagree-
free energies
of transition,i will focusonlyon calculated mentgivenuncertaintyin thermalexpansions.
slopesof equilibriumcurvesin this sectionto comparewith ConsideLnext, the coesite-stishovite
equilibrium.Two
the slopesdeterminedexperimentally. different vibrational models were examined for stishovite in
I have calculatedthe slopesfor the quartz-coesite-stisho- paper 3. However, these models differed only in their
vite equilibria from entropiescalculatedat 1000øK, 1 bar, treatmentof the higher-frequencyoptic modeswhich do not
and from volumesappropriateto the approximatepressures have a strong influence on the entropy. Therefore little
of the transitions: 30 kbar for quartz-coesite, 70 kbar for differencein entropyat temperaturesabove298øKis predict-
metastable quartz-stishovite, and 90 kbar for coesite-stisho- ed by the models, and, for simplicity, only model 2 is
vite. (Although a method of obtaining the thermodynamic discussedhere. The calculatedslopefor this transitionis 17
functionsat high pressureis given in section5, not enough barsdeg-•. Experimental valuesrangefrom 11to 36 bars:
data exist on the thermal expansionsof coesite and stisho- Weaver et al. [1979] prefer a value between 11 and 23 bars;
vite to allow it to be usedfor this calculation.)The entropies Jeanloz and Thompson [1982] prefer a value of 21 bars
and volumesusedare listed in Table 1. A model entropyfor deg-•.Giventhewidescatter in theexperimental
curvesand
832 KIEFFER: THERMODYNAMICS
AND LATTICE VIBRATIONSOF MINERALS,5

othertwo polymorphs,extendingout to 1200cm-1, nearly

6
-HK O'
• 5
(MS)-
• 4
KYANITE SILLIMANITE
o / ANDALUSITE
/ sillimanite-andalusite,
/ kyanite;dP/dT = 13.0 bars øK-l; and sillimanite-kyanite,
o I I I I dP/dT = 20.1 barsøK-1. For comparison, the slopespre-
4oo ,500 600 700 800 900
TEMPERATURE (K)
Fig. 4. Calculated slopes of aluminosilicate phase boundaries
from model entropies. The intercepts at P = 0 were determined from In order to compare a triple point calculated from model
the work of Holdaway [1971],as describedin the text. The shaded data with triple points of other authors, it is necessary to
triangular area is bounded by the three phase boundaries;because know the free enthalpies of reaction or, alternately, inter-
the slopesof the curvesare determinedindependentlyfrom the
spectroscopic data, they do not necessarily intercept at a point. cepts of the equilibrium curves at some pressure and tem-
Triple points as determined by other investigatorsare shown: H,
Holdaway [1971]; HK, Holm and Kleppa [1966]; R, Richardson et
al. [1969]; A, Althaus [1967]; (MS), Mueller and Saxena [1977], a
'reasonable estimate' from survey of literature.
the small number of data on which they are based,the model
valueof 17barsdeg-1is probablyto be considered
satisfac-
tory.
The calculated slope of the metastablequartz-stishovite and a linear equation for a phase boundary becomes
equilibriumis 15barsdeg-1, lessthanthe slopeof 22 bars
deg-1 thatcanbe inferredfromthecalorimeterdataof Holm
et al. [1967], as extrapolated by Robie et al. [1978]. This
differenceoccursbecausemy modelentropiesfor stishovite
are appreciably greater than those obtained from calorime-
try. It is worth noting here that in a recent analysis of
existingphaseequilibria data on quartz, coesite,and stisho-
vite, Weaver et al. [1979] suggestedthat the entropy of
stishovite
is 7.7 calmo1-1øK-l, in excellentagreement
with
the valuesof 7.65 to 7.80 cal mo1-1øK-1 givenby these
models in paper 4. This value is about 1 cal mol-• øK-1
greater than the calorimetric values, and would give a
quartz-stishoviteslope close to that predicted above.
AI2Si05 Equilibria
The aluminosilicate
phaserelations,the subjectof much
controversy (for example, see Zen [1972] and Day and
Kumin [1980]), provide a very sensitive test of the model
becausefree-energydifferencesbetweenthe phasesare very
small.Suchsmalldifferencesin free energies(inferredfrom
shapesof stabilityfields)are somewhatsurprisingin view of
the rather large apparent differencesbetween the vibrational
spectrum of sillimanite and the spectra of kyanite and
andalusite(paper2). The Si-O stretchingbandsof sillimanite
cover a much broader range of frequenciesthan those of the
the highest frequencies found for the asymmetric Si-O
stretching modes in any silicate. At lower frequencies,
spectra of the polymorphs show systematic, although not
? easilyinterpretable,differences.The lowestkyaniteband is
at 237 cm-1 the lowest andalusite band is at 165 cm-1 and
R thelowestsillimanitebandis at 115cm-•. In effect,spectra
are stretched to both lower and higher frequencies as the
aluminum coordinationprogressesfrom sixfold in kyanite to
fourfold in sillimanite. The thermodynamicfunctionsof the
polymorphs remain similar in spite of the spectral changes
because the stretching in frequency at one end of the
spectrumis counteracted by the stretchingat the other end.
Calculated entropies for these polymorphsat 298øKare as
follows' kyanite, 20.80; andalusite, 23.12; and sillimanite,
23.60 cal mo1-1øK-1. Molar volumesof the three poly-
morphsare (in the sameorder)44.09,51.53,and49.90cm3.
Therefore the following slopes at 298øK are predicted:
dP/dT = -12.3 barsøK-l; andalusite-
I000 I100
ferred by Holdaway [1971] at the same temperature are
sillimanite-andalusite, -14.0; andalusite-kyanite, 14.2; and
sillimanite-kyanite,
19.5barsøK-1.
perature. I have adopted intercepts of the phase boundaries
at P = 0 from Holdaway [1971] (Figure 4). Adoption of the
P - 0 intercepts avoids explicit considerationof AI-Si order-
disorder and of curvature of phase boundaries. AS is as-
sumedconstant over the range of pressureand temperature
considered.
The heat of reaction(AH)ør,at the intercept
temperature Ti is then implicitly given by
Ti(AS)øT,
= (AH)øT, (2)
[TAS - AH ø] [T- T•]AS
P = = (3)
AV AV
Holdaway's intercepts are 1043øK for the andalusite-silli-
manite(A-S) boundary,473øKfor the kyanite-andalusite(K-
A) boundary, and 599øKfor the sillimanite-kyanitebound-
ary. These intercepts,combinedwith modelentropies,give
rise to the followingphaseboundaryequations:
K-A
P (bars) = 13.0[T(øK) - 473]
A-S
P (bars) = - 12.3[T(øK) - 1043]
S-K
P (bars) = 20.1[T(øK) - 599]
When solved simultaneously,the equilibrium curves inter-
sect each other as follows: A-S and A-K at 750øK = 477øC
(P = 3.6kbars),S-K'andA-K at 830øK= 557øC
(P = 4.6
kbars),and S-K and A-S at 769øK= 496øC(P = 3.4 kbars).
As shownin Figure4, a triangularregionsurrounding the
Holdaway triple point is definedby the three model curves.
Because
theslopes
oi:allthreecurves
obtained
frommy
model are independent,such good convergencetoward a
 KIEFFER: THERMODYNAMICS AND LATTICE VIBRATIONS OF MINERALS, 5 833

single point is actually quite remarkable. Note that the ==Birch

8•LeComte
[1960]
/ '
I Newton
&Smith
[1967]
_
elongation of the 'triple point' is actually caused by near-
parallelism of the kyanite-andalusiteand kyanite-sillimanite • Uollond
D9O0]
25-,, Essene
eta1.[1972]/
curves. Thus this example shows that calculation of phase A Hays
andBell . ß
boundary slopesusing only spectroscopicand acoustic data a: 20 -
and calculation of the phase boundary position using some rn Jd+Qz
measure of heats of reaction gives mineral stability fields
tu 15-
that are in good agreement with those obtained by other
/
methods, even for polymorphs such as those of A12SiOswith
relatively small free-energy differences.

Stability of Albite
Finally, consider use of the model for computing condi-
5 /• j ALBITE
tions for the reaction
02
O0 I
400 I I
GO0 800 I
I000 I
1200 1400
NaA1Si308 = NaA1Si206 + SiO: TEMPERATURE (K)

Experimental determinations of this phase boundary [Birch
and LeComte, 1960; Newton and Smith, 1967; Hlabse and
Kleppa, 1968; Boettcher and Wyllie, 1969; Johannes et al.,
1971; Essene et al., 1972; Hays and Bell, 1973; Holland,
1980] suggestisobaric differencesin equilibrium temperature
of the order of 300øK at the higher pressures, or isothermal
differences of the order of 7 kbar (Figure 5). Curves calculat-
ed from calorimetry have an uncertainty of about 1 kbar,
largelydueto an uncertainty
of about0.5 cal mol-• øK-• in
albite and jadeite entropies [Openshaw, 1974]. The phase
equilibria are complicated by both the degree of ordering in
albite (the low-high transition from ordered to disordered
probably occurs at -< ---1000øK) and the low-high quartz
transition (848øK). For this calculation, entropies of low
albite and low quartz are used.
The sensitivity of the equilibrium curve to small entropy
variations can be seen by consideration of Figure 5. The
solid curves represent three different equilibrium lines, each
based on the same value of AH øfor the reaction, but different
values of entropy for the phases. An equilibrium line calcu-
lated with calorimetric data obtained for albite by Kelley et
al. [1953] and calorimetric data for quartz and jadeite by
Robie and Waldbaum [1968] is given as curve 1. An equilib-
rium line with the more recent data on albite by Openshaw
[1974] and the same data for quartz and jadeite is given as
curve 2; and an equilibrium line calculated with the Open-
shaw [ 1974] albite data and my model jadeite entropy values
is shown as curve 3. The difference between curves 2 and 3 is
causedby the fact that my modeljadeite entropy at 298øKis
0.6 cal mol-• øK-• lower than the calorimetric value listed in
the work by Robie and Waldbaum [!968]- A!thn•.ghthe spreadof val•o• nf oq•ilihri•m implied by tho
This equilibrium calculation can be used to illustrate how
experimental data in Figure 5 could be due to a number of
uncertaintiesin vibrationalspectraaffectthe thermodynamic factors not consideredhere (variationsin jadeite properties,
applicationsof the modelto phaseequilibria.The dangerof experimentalconditions),it is possiblethat they are due to
not havingadequateor accuratespectrais revealedby such
an attempt. The two dashedcurves, 4 and 5, in Figure 5 are
equilibriumlines computedusingmodelvaluesof entropy
for all three minerals. The top one (5) is calculatedwith the
lowestopticalmodeof albiteat 45 cm-1, as suggested
in a
far-infraredspectrumpublishedby Kovachet al. [1975,p.
249]. The bottom dashed curve (4) is one with the lowest
opticalmodetakenat 92 cm-•, as reportedby liishi et al.
[1971] (this line is the strongestlow-frequencyline in the
infraredspectrumof albite shownin Figure l d of paper2).
Modelvaluesof entropyat 298øKvary by 6 cal mol-• øK-•
when these two values of lowest optical mode are used as
lower limits of the optic continuum,and the slopeof the
Fig. 5. Measured and calculated phase boundaries for the
breakdown of albite to jadeRe + quartz. Curves numbered 1-5 are
discussed in text.
equilibrium line changesby a factor of 2, accordingly. The
lack of agreement between curve 4 and the experimental
datasuggested
thatthe92 cm-• lineobserved
by liishi et al.
[1971] was not the lowest-frequency mode. On the other
hand,curve5, basedon a line at 45 cm-• appearsto lie
considerably higher than most of the reported data. I sus-
pected that there might actually be an optical mode in the
vicinityof 80 cm-• that wouldproducean albite model
entropy such that the calculated phase boundary would
agree with those boundaries determined from the calorimet-
ric data of either Kelley et al. [1953] or Openshaw[1974]
(curves 1 and 2). The infrared spectrumof albite (paper 2,
Figure ld) hasa slightband on the low-frequencyshoulderof
the92 cm-• band.Onthebasisof theabovehypothesis
and
this suggestionof the weak band in the infrared spectrum,
we obtained a Raman spectrum on albite and found that
there is, in fact, a pronouncedRamanmodeat 83 cm-1 (J.
Delany, unpublished data, 1978). With this mode taken as
the lowest optical frequency at K = 0 and reduced to 63
cm-• at K = Kmax,themodelentropyat 298øKiswithin0.26
cal mol-• øK-• of the valueof Openshaw[1974],andthe
calculated equilibrium line is then nearly coincident with
curve 1. The experimentaldata shown, except thoseof Birch
and LeComte [1960], are consistentwith a lowest optical
modefrequencyat K = 0 betweenabout85 and 65 cm-•.
actual variations in the albite structure that influence the
vibrational spectrum and, therefore, the thermodynamic
functions.For example, order-disordereffects, if reflectedin
variations in low-frequency spectral characteristics, could
account for the scatter in experimental values.
4. ISOTOPIC FRACTIoNATION
Many interestingproblems in isotopic thermometry await
a theoreticalmodel that can predict the energiesof isotopic
forms of mineralswith sufficientaccuracyto allow accurate
calculation of isotopic fractionation factors. The energy
changesin isotopic exchangereactions are of the order of a
few calories per mole and are thus too small to be measured
834 KIEFFER: THERMODYNAMICSAND LATTICE VIBRATIONS OF MINERALS, 5

calorimetrically [Clayton, 1981]. Direct measurement by each primitive unit cell vibrates with the same normal modes
exchange reactions is difficult, particularly at low tempera- (see paper 3). Thus a 'reduced' partition function ratio f can
tures, because the driving force for isotopic exchange reac- be defined for the unit cell; it is equivalent to the partition
tions is small. Therefore experiments must frequently be functionoriginally definedfor an ideal gasmoleculeby Urey
performed at high temperatures,and the data obtainedmust [1947]:
be extrapolated to lower temperatures. In the range of
temperaturesof geologicinterest, extrapolationlaws are not
well understood. Thus even if a theoretical
predict fractionation factors to high accuracy, it might be
model cannot
Q*'
_Q*(m)
3r/2
f= Q,- Q •-• (10)
where Q*' and Q' refer to the partition functions for a unit
useful in providing a basisfor extrapolation of existing data.
cell, Q* and Q refer to the partition functionsfor a crystalline
In this section I explore the application and limitations of
solid, rn and m* are the massesof exchangingisotopes,r is
this model to prediction of oxygen isotopic fractionation
factors.
the number of atoms being exchangedin the unit cell, and
the product over 3N in equation (9) for Q*/Q becomes a
Consider [after Urey, 1947] an isotopic exchangereaction
product over 3s, the number of degrees of freedom in the
aA + bB* = aA* + bB (4) primitive cell [Becker, 1971]. The reducedpartition function
ratio corresponds to an equilibrium constant for the ex-
where a and b are stoichiometric coefficients relative to the
change reaction between the unit cell and its separated
exchange reaction, A and B are molecules containing a
atoms. For convenience, the reduced partition function ratio
common element being exchanged, and the asterisk indi-
is usually written in terms of natural logarithms as
cates the molecule containing the heavy isotope. For exam-
(1/r)1000In f. Bigeleisenand Mayer [1947] showedthat In f
ple, in the following reaction between enstatite (px) and
o•(1//)2 as T--> o•;that is, thereis no fractionation
at high
olivine (ol),
temperatures. At low temperatures, In f or 1/T [e.g., Hul-
(px) (ol) (px) (ol) ston, 1978, p. 212].
The Helmholtz free energy of a material is related to the
-}MgSi1603
+ • Mg2Si1804
--• -•MgSi1803
+ • Mg2Si1604 partition function by
the equilibrium constant is [Clayton, 1981] F = kTln Q (11)

K •
(180/160)
A • (180/160)px If the moleculesA and B contain isotopesX and X*, the
(180/160)B (180/160)ol fractionationfactor a of the isotopesis

As was shown by Urey [1947], the equilibrium constant is ([X*/XI)ax•-AX:

a = (12)
related to the free-energy change in the isotopic exchange ([X*/XI)•Xm--•Xm*
reaction. From statistical mechanics, the equilibrium con-
stant K for complete exchange of the isotopes is where [X*/X] denotes the ratio of the isotopesin the com-
poundssubscripted.If the X isotopesare randomly distribut-
K = (QA*/QA)
'• + (QB*/QB)
b (5) ed in the unit cells, Epstein [1959] has shown that
Q is the partition function for the molecule and is given by ot= K l/r (13)
where r is the number of atoms of the element undergoing
Q = • e-e'/kt (6) exchange in (4). From (11),
i

Q* F* - F
where Ei is the energy of the ith state relative to a rest state In • = (14)
with no zero-point energy, k is the Boltzmann constant, and Q -kT
T is the absolutetemperature. The summationis taken over
Thus the partition function ratio or fractionation factor can
all energy levels of the molecule, i.
be calculated directly from values of F and F*'
For a crystalline solid, in the harmonic approximation,
F* - F 3r rn
Ei -' (n + «)hvi n = 0, 1, 2,''' (7) In f = + -- In (15)
-kT 2 m*
where h is Planck's constant and vi is mode frequency.
Therefore From papers 1 and 3,

3N-6 e-Xi/2
Q= H (8) F = 3,qNAl3+ kr In {1 - [exp (-gto/kT)]}g(to)dto (16)
i=1
(1 - e -x')
where
where xi = hvi/kT, as in paper 3, and N is the number of 3NAn
atoms in the solid.
3nNAl3
= - Vo- « • gto (17)
In exchangereactions, it is ratios of the partition functions r=l
that are of interest:
In theseequationsI have switchedfrom v to toto preserve
Q, 3N-6=3N e-X,*/2 (1 -- e-x') the identitiesfrom paper 3. V0 is the potentialenergyof the
Q = i=1
I-[ (1- e-x'*)e-x'/2 (9) equilibriumconfigurationof the crystal, the summationterm
is the zero-pointenergyterm, tois the angularfrequency,NA
In the lattice dynamics approximation for crystalline solids, is Avogadro'snumber,g = h/2rr, n is the numberof atomsin
 KIEFFER: THERMODYNAMICS AND LATTICE VIBRATIONS OF MINERALS, 5
the chemicalformula of the compound,g(to)is the frequency
distribution of normal modes, and toi is the highest lattice
vibrational frequency. A similar equation holds for F* with
the spectrumg(to)beingreplacedby g*(to)for the isotopical-
ly heavy compound.
The equationsabove, and the methodsdescribedbelow,
could be appliedto fractionationof many differentisotopes
such as those of oxygen, sulfur, and silicon. Further discus-
sion, however, is restricted to oxygen-isotopesubstitution.
The problem facing the theoretician is to specify the
frequency distributionsg(to) and g*(to) for isotopicallylight
and heavy minerals. The most common approach to this
problem has been one in which detailed lattice dynamics
models are formulated, and force constants for the isotopi-
cally light compoundsare evaluatedfrom spectroscopicand
elastic-wavevelocity data. The models are used to predict
vibrational frequencies of isotopically heavy species (for
example, seeBottinga [1968, 1969],Lambert [1975],Becker
[1971], Kawabe [1978], Elcorobe [1974], Elcorobe and Hul-
ston [1975], and Hulston [1978]). Such methods allow de-
tailed treatment of individual minerals but have been limited
in application because of their complexity. Taylor and
Epstein [1962] showedthat complexsilicatesshowtrendsin
fractionation that can be explained from knowledgeof the
typesof oxygenbonds,andtheywereableto estimate
•80
values by linear combination of bond types. Sakai and
Honma [ 1969] were able to calculate mineral-water fraction-
ation factors for 10 minerals simply from examinationof the
Si-O modes without detailed knowledge of the spectra.
In the approach that follows, I have adopted a model
intermediate between detailed spectroscopic calculations
and less detailed Si-O bond models. I assume that lattice
vibrationalspectrag(to)of isotopicallylight mineralscan be
approximatedfrom spectroscopicand elastic data as de-
scribedin papers 1-4. A hypotheticalspectrumg*(to) is then
generatedfor the isotopicallyheavy mineral from a set of
rules (givenbelow) describingexpectedfrequencyshiftsfor
different types of vibrational modesupon substitutionof a
heavy isotope into the mineral structure. The required
integrationsand normalizationsto (15) and (16) can then be
performedto give the reducedpartition functionratio (1000
In f) or, with an assumedpartition function for water, a
mineral-water fractionation factor, (1000 In a). The method
assumes that vibrational modes are describable in terms of
fairly simpleatomic motions.While this hasbeen a common
practice in spectroscopy,it is certainly not a universally
acceptedpractice,and as spectroscopic knowledgeof indi-
vidual minerals accumulates, mode assignments tend to
become more complex.
Rather than approximatingthe spectrumof an isotopically
light mineral by a seriesof Einstein oscillators,as has been
common in previous models cited above, I have used a
refinementof the model proposedin papers 3 and 4, subdi-
vidingthe spectrumof eachmineralconsideredinto various
groupsof vibrationalmodescharacteristicof the minerals,as
shown schematically in Figure 6. In general, I have used
more detailedapproximationsto the mineralspectrafor the
purposeof calculatingisotopicfractionationfactors than I
did for calculation of thermodynamic variables and phase
equilibria.This greaterdetailis necessaryin orderto identify
and isolatemodesthat might have commonfrequencyshifts
upon isotopic substitution. The heavy line in Figure 6
representsthe spectrumof an isotopicallylight mineral.The
835
0)
(S•, AI)-O anh-
symmetric stretch
(z/i,t/3)
(2)
(S•, AI)-O- S•
syrnrnetrlc
stretch
•coushc
ß
S•- 0 def.
(3)
Si, AI- 0 bending
modes ß c•hon-O del. (V2, t/4)
polyhedrol def.,
I
Fig. 6. Schematicvibrational spectra used for the isotopically
light mineral (heavy bottom line) and the isotopicallyheavy mineral
(light upper line).
major groupsof vibrations identified for each silicate mineral
are (in approximate order of decreasing frequency) (1)
tetrahedral antisymmetric stretching modes, denoted as
and t,3 in the vibrations of an isolated tetrahedron, (2)
symmetric stretching modes, often termed 'bridging defor-
mations,' (3) tetrahedral bending modes, (4) polyhedral, Si-
O, and cation-oxygen 'deformations,' and (5) acoustic
modes. Acoustic modes are representedas in papers 3 and 4
by one longitudinal and two shear modes which are assumed
to obey a simple sine dispersion law between wave vector
and frequency. Optic modesare representedin two ways: (1)
where justified and possible, by Einstein oscillators, for
example, for Si-O stretching modes, internal carbonate
modes, and hydroxyl stretching modes, and (2) by an optic
continuum or segmentedcontinua as shown in Figure 6. The
number of modes of each type and their frequency or
frequency range must be specifiedfrom data in the spectros-
copy literature, or, too often, from interpretation or extrapo-
lation of these data. As will be demonstratedbelow, applica-
tion of the model is often limited by lack of spectroscopic
data, or by lack of mode assignmentsto existing data, and
much more data and interpretation are needed.
Once these groups of modes have been defined and
enumerated for the isotopically light mineral, a frequency
shift for each group of modes must be determined to give a
hypothetical spectrum of the isotopically heavy mineral.
Such a shifted spectrumis shown schematicallyin the upper
part of Figure 6 (light line). Note that different modesmay be
shifted by different amounts: for substitution of oxygen
isotopes, the frequency may not shift if the modes do not
involve displacementof oxygen atoms, whereasthe frequen-
cy maydecrease
asmuchas94.3%[= (16/18)u2]if themodes
involve only motions of oxygen atoms.
Because normal mode frequencies and deformation types
depend in detail on the geometry of the vibrating unit or
cluster, a detailed calculation of the effect of isotopic substi-
tution in principle requires a detailed model of force con-
stantsfor lattice vibrations. However, as shown in papers 1-
4, a useful approximation is that the vibrations of individual
clusters of atoms can be enumerated and characterized.
Herzberg [1945, pp. 227-238] discussed models for the
vibrations of isolated molecules of the form XY2, XYZ3, and
XY4. These models give solutionsfor frequency that can be
836 KIEFFER.' THERMODYNAMICS AND LATTICE VIBRATIONS OF MINERALS, 5

TABLE 2. Frequency
Decreases
Assumed
for Silicates
When160Is Replaced
by •80
Antisymmetric Symmetric Bending,
(Si, A1)-O ,(Si, A1)-O-Si Distortion,
Stretch Stretch Other Acoustic
Mineral Class (Rule 1) (Rules 2, 5) (Rules 3, 5) (Rule 4)
Framework 0.96 0.99 for those HT limit a (M/M*) •/2t'
identified; HT
limit for those at
low frequency
(see text)
Sheet 0.96 0.99 HT limit (M/M*) •/2
Chain 0.96 0.99 HT limit (M/M*) •/2
Orthosilicates v• - 0.943 not present bending (M/M*) •/2
v3 - 0.977 modes,0.95
other: HT limit

aHT limit (high-temperaturelimit)' seeequations(20) and (21) and discussion.

t'M is the formulaweight;the asteriskdenotesthe isotopicallyheavymineral.

used to estimate the dependenceof frequency on atomic the shift of v• is very simple [Herzberg, 1945]'
mass of various atoms in the cluster. Although complete
solutionsfor frequency shifts of clustersof all compositions
imaginable in minerals are currently beyond reach, Herz-
berg's calculationsand considerationsof end-membertypes
of vibrations can be used to provide an estimate of the
__=[t16m)
/•1
/•1* 1/2
18m = 0.942809 --• 0.943 (18)

expectedshiftsin thosecaseswhere the vibrationsare fairly
well isolated in frequency, for example, for Si-O antisym-
metric stretching modes and for internal modes of the
carbonate ion.
The shift of v3is more complex, dependingnot only on the
relative massesof oxygen isotopes,but also on the relative
shift of the v4 (bending) mode. Herzberg's [1945, p. 182]
valence-force model can be used to derive a set of simulta-

For vibrational modes not well approximated by the neous equations for Vl/Vl*, v2/v2*, v3/v3*, and vn/v4*.These
vibrations of an isolated cluster, various considerationswere equationscan be solvedfor v3/v3*if the ratios Vl/V2and v3/v4
used to construct the rules presented below for frequency of isotopically light minerals are known. These two ratios
shifts. The advantage of these rules is that they can be vary in minerals becausethey dependon cation massesand
force constants (note that this fact shows the weakness in the
applied uniformly to a wide variety of minerals having
different structural and compositional characteristics.The assumptionthat molecular clusters can be treated as isolated
set of partition functionsthus obtained is internally consis- clusters). Values of Vl = 819, v2 = 340, v3 = 956, and v4 =
tent, but values will changeas more is learned about the shift 527cm-1 [fromWhite,1975,p. 328]aretypical,although
the
factors or as spectra are treated with higher resolutionor spreadof proposedvaluesmay be as muchas 200cm-1 for
more detail. The disadvantageof these rules is that they are any one mode. With these values, the calculatedratio v3/v3*
not based on detailed consideration of the atomic structure. is 0.977, and this value is used throughoutthis paper. For
orthosilicatesI assumedthat the ratio of the numberof v3to
Vl modes was 3' 1, as it would be for an isolated tetrahedron.
Frequency Shifts Upon Isotopic Substitution:
For silicates in which the tetrahedra are partially to
A Simple Set of Rules for Silicates
completely polymerized, separate identification of the anti-
Groups of modestreated separatelyare listed in Table 2. symmetricvibrational modesas Vl and v3is not possible,or
Althoughvariousgroupsof modescan overlapin frequency perhaps, even meaningful. Therefore a general rule for
(as discussed,for example, in paper 4), such overlaps calculation of the shift of Vl and v3 antisymmetric mode
generallyinvolve only a few modes and are largely ignored frequencies resulting from180substitution cannotbe given
here for simplicity;one exceptionis Si-O symmetricstretch- for polymerized silicates. However, a few data do exist. A
ing modes discussedbelow. For reasons that will become shift of 0.96 was observedexperimentallyfor thesemodesin
clear at the end of the section, the modes are discussed thin films of vitreous silica [Ligenza and Spitzer, 1960].
below in the order (1) antisymmetricstretchingmodes, (2) Kawabe [1978] gave a detailed model for oxygen-isotope
symmetricstretchingmodes, (3) bendingmodes,(4) acoustic fractionation in quartz, and the weighted average of his
modes, and (5) other deformational modes. calculated shifts of these modes is 0.9575. Therefore it is
probably reasonable to assume a shift of 0.96 for the
Rule 1 (antisymmetric(Si, Al)-O stretchingmodes): If v•
and v3 modes can be separately identified and enumerated,
antisymmetricstretching modesof SiO4-4 tetrahedrafor the
frameworksilicates.It is assumedto be the samefor AlO4-5
their frequencies are assumedto decrease by 0.943 and
tetrahedra. No data exist to test the validity of this assumed
0.977,respectively,
uponsubstitution
of •80 for •60. If they
shift for silicatesof intermediate polymerization. However,
cannot be separately enumerated, an average shift of 0.96 is
the fact that the weighted averagesof the Vl and v3shifts for
used for the antisymmetric modes.
isolatedtetrahedraand polymerizedtetrahedraare so similar
The (Si, Al)-O antisymmetric stretchingmodes are those suggeststhat values near 0.96 should apply to the Si-O-Si
called v• and v3 when they can be identified in isolated antisymmetric stretching modes rather independently of
moleculesor orthosilicates.In relatively isolatedmolecules, degree of polymerization of tetrahedra.
 KIEFFER: THERMODYNAMICSAND LATTICE VIBRATIONS OF MINEttALS, 5 837

Rule2 (symmetric
(Si, Al)-O-Sistretch): The frequency present in vitreous silica, and the spectral data of Ligenza
shift of symmetric (Si, AI)-O-Si stretching modes upon and Spitzer [1960] show a frequency shift of only 0.9868
substitutionof •80 for •60 is taken to be 0.990, for those upon substitutionof oxygen isotopes. I have rounded this to
modes involving mainly Si motion; some modes, by acci- 0.990 in my model.
dents of mineral symmetry, become deformational modes The sixth Si-O-Si deformational mode in quartz becomes
and must be treated differently. an O-Si-O deformation or Si-O bending mode at an anoma-
louslylow frequency,466 cm-• [Moenke,1974,p. 366],
In minerals containing bridging oxygens, there are some because motion of Si is not permitted in the A• species to
Si-O-Si stretching modes whose deformations involve pri- which this mode belongs. Kawabe [1978, p. 616] calculated a
shift of 0.945 for this mode; I placed this mode in the
marily motion of Si; these are usually called symmetric Si-O-
Si or Si-Si stretchingmodesin spectroscopic assignments. continuum with other distortional modes (between 90 and
The method of enumerating these modes was given in paper 550cm-•; Table3) anddetermined the shiftfor thisgroup
4; I assume that one symmetric Si-O-Si stretch exists for from the high-temperature product rule (discussedbelow).
each Si-O-Si bond. Theaverage
shiftfor thisgroupis 0.949917
-• 0.950,which
In the quartz spectrum, five of the six expected modes gives a shift for this one mode in good agreement with that
havebeenfoundbetween697 and 807 cm-• [Lippincottet calculated by Kawabe.
al., 1958; Becker, 1971]: at 807, 795, 778, and 697 (doubly In the feldspars, 16 (Si, Al)-O-(Si, Al) symmetric stretch-
degenerate)cm-•. These modesshouldshow very little ing modes would be ,expected. According to Lazarev [1972,
frequency shift upon substitution of oxygen isotopes. Al- p. 182], eight should be infrared active, and he finds these
though no data are available on quartz, the same modes are eightto be between500and800cm-•; thesefrequencies,
as

TABLE 3. Rangeof Modes,q, and FrequencyShiftAssumedfor Calculations

of OxygenIsotopeFractionation
VsSi-O-Si or,
for Orthosilicates, Si-O
VasSi-O Bending Modes Lattice Distortion
Mineral Range q Shift Range q Shift Range q Shift Acoustic
Quartz 1200 0.074 697-809 0.1852 0.99 90a-550ø 0.4817 0.949917
1162 0.037 0.96 0.968246
1117 0.074
1080 0.037
Albite 1010 0.051 0.96 720-780 0.1025 0.99 500-650 0.1025 0.961462
1123 0.154 0.9• 0.1025 0.99 and 0.970797
63-476 0.5515 0.961462
Anorthite 960 0.051 0.96 666-770 0.1025 0.99 460-620 0.1025 0.961367
1070 0.154 0.96 and 0.972429
85-410 0.5515 0.961367
Muscovite 3633 0.032 0.996 700-900 0.095 0.99 91-620 0.6902 0.963671
|000 0.159 0.960 0.971170
Clinoenstatite 1000 0.100 0.960 650-750 0.067 0.99 170-563 0.683 0.964254
900 0.100 0.960 0.971398
Diopside 1000 0.100 0.96 610-710 0.067 0.99 128-557 0.683 0.963853
900 0.100 0.96 0.973328
Pyrope 975 0.15 0.977 450-630 0.25 0.95 115-450 0.5375 0.971167 0.971503
875 0.05 0.943
Grossular 950 0.15 0.977 450-619 0.25 0.95 152-450 0.5375 0.971100 0.974379
850 0.05 0.943
Andradite 925 0.15 0.977 450-589 0.25 0.95 116-450 0.5375 0.971035
825 0.05 0.943 0.977192
Zircon 974 0.055 0.943 400-608 0.275 0.95 185-400 0.4217 0.962066 0.971503
1000 0.!00 0.977
885 0.055 0.977
Forsterite 930 0.142 0.977 420-650 0.238 0.95 128-420 0.5363 0.973624 0.972710
837 0.048 0.943
Rutile 824 0.167 0.96 450-500 0.222 0.990 80-450 0.389 0.940418 0.975871
610 0.055 0.96

Internal Carbonate Modes External Modes External Modes

Mineral Range q Shift Range q Shift Range q Shift Acoustic

Calcitec 1460 0.133 0.9837 300-387 0.100 0.970 200-325 0.200 0.982260 0.971311
1070 0.067 0.9427 and
881 0.067 0.9888 77-200
712 0.133 0.9476 0.200 0.943

aLowerlimit of opticcontinuumcitedis thatat K = K.... Thiscorresponds

to thelowestIR or Ramanmodeat K = 0 modifiedby an as-
sumeddispersioncurve acrossthe Brillouin zone as discussedin paper 3.
bKma
x from Elcorobe[1967].
CDatafor calcite partially from Bottinga [1968].
838 KIEFFER:THERMODYNAMICS
AND LATTICEVIBRATIONSOF MINERALS,5

well as the antisymmetric (Si, A1)-O stretching frequencies, •60 is determined

fromthe high-temperature
productrule
are lower than the corresponding frequencies in quartz that requires that the fractionation go to zero at high
because the substitution of AI for Si into the structure lowers temperatures.
its rigidity and, hence, vibrational frequencies. Iiishi et al.
[1971] further subdivide these modes into Si-Si or Si-AI(Si)
The groupsdiscussedabove accountfor roughly 40-50%
of the modesof typical silicates.The remainingmodesare of
stretching between720and780cm-• andO-Si(AI)-Obend-
ingbetween500and650cm-•. The diagrams shownby Iiishi several types: lattice deformationsand distortions, polyhe-
dral deformations, tetrahedral bending modes, and, in some
et al. for vibrational modes suggestthat the symmetric (Si,
minerals, OH-bending or librational modes. The shift to be
AI)-O-(Si, AI) vibrations of feldspars resemble those of
quartz:The onesbetween720and780cm-• involvemainly applied to this remaining group of modes is not arbitrary.
Becausefractionation is required to approach zero at high
motion of silicon, whereas those at the lower frequency
involve considerable motion of oxygen and, in fact, merge
temperature, frequency shifts must obey the high-tempera-
ture productrule [Becker, 1971],derived from (9) and (10) by
into other types of deformational modes. Accordingly, I
taking the high-temperaturelimit:
have taken half of the (Si, AI)-O-(Si, AI) symmetric stretch-
ing modes of feldspars to lie in the higher-frequency group
(720-780cm-• for albite;666-770cm-• for anorthite)with a
shift of 0.99. I have taken the other half at lower frequencies 3s
Xi(••__)3r/2
.= xi*
(500-650cm-• for albite;460-620cm-• for anorthite)com- where s is the number of atoms per unit cell, r is the number
(19)

bined with other deformational modes and have determined

their frequency shift by the high-temperatureproduct rule
discussed below.
Rule 3 (Si-O bending modes in orthosilicates): If the Si-
O bending modes in orthosilicates can be identified and
enumerated,
their frequencyshiftuponsubstitution
of 180
for •60 is taken to be 0.95.
In orthosilicates
the Si-O bendingmodesof isolatedtetra-
hedra can be identified and enumerated. The isolated tetra-
hedron has 1.25 times as many bending modes (v2 + v4) as
antisymmetric stretching modes (//1 •t_//3). The same relative
proportionsare used here. Considerationof deformationof
an isolated tetrahedron indicates that •*/• = 0.943 and v4*/
v4 - 0.949 [Herzberg, 1945]. The weighted average shift for
the modes would be 0.947. A value of 0.95 was adopted for
this model. This shift is very large because the bending
modes for an isolated tetrahedron involve principally the
motions of oxygen (remember that the maximum possible
shift is 0.943 for modes involving only motions of oxygens).
If the motions of oxygen are hindered either by interactions
with the cations or by polymerization of the bonds, the shifts
should be less; indeed, the only data available on a frame-
work silicate, vitreous silica [Ligenza and Spitzer, 1960],
give a value of 0.966 for the bending modes.
Rule 4 (acoustic modes): The ratio of frequenciesupon
substitution
of •80 for 160is takento beproportional
to the
square root of the ratio of the mean molecular masses.
The lowest-frequencymodes,the acousticmodes,are the
last group of modes for which a frequency shift can be
estimatedon empiricalor theoreticalconsiderations. At long
wavelengthsthe fundamentalvibratingunit is the primitive
unit cell, and, by analogywith the vibrationsof a monatomic
chain(paper 1, equation(20)), the ratio of acousticfrequen-
cies of isotopicallyheavy and light mineralsis w*/w = (M/
M*) •/2,whereM denotes
thetotalmassof theunitcellor,in
the ratio of masses,can be taken to be the formulaweight.
The mineral molar volume, which is requiredfor calculation
of the acousticmodefrequencies(equations(8), paper 3), is
assumedto decreaselinearly with the molecular-massratio.
Rule 5 (remainingundifferentiatedmodes): The frequen-
cy shiftof the remainingmodesuponsubstitutionof •80 for
of atoms being exchangedin the isotopicreaction, m is the
mass of the exchanging atom, and the asterisk denotes the
heavyisotope.This rule can be rewritten as
3s 3r
(20)
• in(xi/xi*)
i=1
- -•-In(m*/m)
If the spectrumis broken into parts containinga, b, c, d, ß ß ß
modes, each part having a common frequency shift (i.e., xi/
xi* = const Ka, Ko, Kc, Ka, '''), then this equation
becomes
alnKa+ blnKo+ c ln Kc + d ln Ka + '''
3r
= -- In (m*/m) (21)
2
Thus if all weighting factors a, b, c, d, ß ß ßare known and if
all shiftsbut one (e.g., Ka) are known, the unknownshiftcan
be obtainedfrom the above equation,This approachwas
used by Becker [1971] to determine the relation between the
shift assumed for the acoustic modes and that of the lowest-
frequency optic modes for quartz. It is used here to deter-
mine the shift of the modesremainingafter identificationof
all others by the proceduresdescribedabove.
The spectral shifts used for the minerals are summarized
inTable3. Theshiftshownto sixdecimal
places(otherthan
the acousticmode shift) is the shift determinedby the high-
temperature rule; this degree of accuracy is necessaryto
ensure convergence in the numerical results.
Calcitewas includedin this studyin orderto comparethis
modelwith the previouscalculationsof Bottinga [1968].For
thiscomparison,the shiftsof all modesexcept20% between
200 and 325 cm-• were takento be similarto Bottinga's
shifts;the shift of the modesbetween 200 and 325 cm-1 was
determinedby the high-temperatureproduct rule. As Bot-
tingaderived shiftsfor thesemodesin a differentway, this
comparisonprovidesa test of the use of the high-tempera-
ture product rule for representing average shifts. Four
internal modes of the carbonate ion were identified and
considered
asEinsteinoscillators://•
at 1070cm-• //3at 1460
cm-1 v2 at 881 cm-• and //4 at 712 cm-• From the
Herzberg [1945, p. 177] model for the carbonate ion and the
observed frequencies given above, one can calculate a
theoretical shift for the internal mode frequenciesof an
 KIEFFER: THERMODYNAMICS AND LATTICE VIBRATIONS OF MINERALS, 5 839

TEMPERATURE (K)
o

gg
0a• o
o o
o o
o o
o o
•o o
o o
u•
IOO
I I I I I

8o

,.•., 6o Rutl

• 4o

2O

I I I I
o 2 4 6 8 I0

106/T2 (K-2)

Fig. 7. Reducedpartitionfunctionversus1/T2 (bottomaxis)or T (topaxis);r is the numberof exchangeable

oxygens in theunitcellof themineral.TheheavycurvelabeledH20(/) is thereducedpartitionfunctionof 'liquid'water
fromBecker[1971].Thedashedcurveshowsa T-2 extrapolation of thehigh-temperature quartzcurveto emphasize the
curvatureof the curvesat low temperature.The mineralabbreviationsareasfollows:Qtz (quartz),Calc(calcite),Albt
(albite),Musc (muscovite),Enst (clinoenstatite),
Anor (anorthite),Diop (diopside),Pyrp (pyrope),Gros(grossular),
Zrcn (zircon), Fors (forsterite),Andr (andradite),and Rutl (rutile).

isolatedcarbonateion upon substitution

of •80: v•*/v• = Results
0.942809, v2*/v2 = 0.988826, v3*/v3 = 0.99063, and v4*/v4 =
0.941093. Bottinga, using a more detailed model, obtained The results of the model are shown in several ways in
very similar values: v•*/v• = 0.942682, v2*/v2 - 0.988797, Figures 7 and 8 and in Table 4. In Figure 7 the reduced
v3*/v3 - 0.983664, and v4*/v4 - 0.941093. Therefore, to partitionfunction,1/r 1000In f, is plottedagainst1/T2. In
simplify comparison of my model with Bottinga's work, I Figure 8 the fractionation between quartz and the other
used the latter values, because the major differences be- mineralsstudiedisplottedagainst1/T2.In Table4, valuesof
tween our two models are in the remaining part of the the mineral-water fractionations at 298øand 1000øKare given
spectrum. This remaining part of the calcite spectrumwas and compared with experimental data where available. Val-
dividedinto threeregions.From 99 cm-• (at K = 0) to 200 ues of the reduced partition function of water used for this
cm-•, modesinvolveonlymotionsof theoxygensaboutthe calculation are those of Becker [1971, p. 52]: at 298.16øK,
central carbon atoms, which remain stationary [see White, 1000 In f is 70.81, and at 1000øK it is 11.55.
1974, p. 231]. I therefore assumedthat the frequency shift is The model predicts correctly that the reduced partition
(16/18) u2= 0.943;Bottingafounda similarlylargeshiftfor function of the minerals changes systematically with struc-
most of these modes. Near 300 cm-• (300-387 cm-• if ture, as first explained by Taylor and Epstein [1962]. For
longitudinal and transverse frequencies are included), three anhydrous silicates, it decreases in the order framework,
modes v6and v8 (doubly degenerate)exist [White, 1974, p. chain, and orthosilicates as shown in Figure 7. This order
233], and all involve opposingmotions of the calcium atoms reflects the decreasing frequency of the antisymmetric Si-O
and carbonate ions, in either compression or shear. I as- stretching modes with this structural progression. Calcite
sumed that the frequencies of such modes might scale with has a high reduced partition function because of the high
the square root of reduced mass of the calcium atom and the frequencies of the internal carbonate ion; rutile has a low
carbonate ion; this assumption would give a shift of 0.982. partition function because of the low frequencies of the Ti-O
Bottinga's average shift for these three modes was 0.970. modes.
The latter value was used. The shift of the remaining modes In detail, the relative order of the reduced partition
between200 and 325 cm-• was determinedfrom the high- functions is slightly different from those orders determined
temperature product rule, equation (21) above, to be by experiment, but the differences only occur where differ-
0.982260. ences in the reduced partition function are of the order of
 840 KIEFFER:
THERMODYNAMICS
ANDLATTICE
VIBRATIONS
OFMINERALS,
5

TEMPERATURE (K)
oo
o o o o o o o o

•o•o o o o
,,-, o •

I I

z•0
I
N
I'-- 8

0
0
0 4

2 4 6 8 I0

106/T2 ( K-2)
Fig. 8. Fractionation
factorsbetween
quartzandtheminerals
indicated
(abbreviations
arethesameasin Figure7)
versus 1/T2.

1%o.For example,the orderpredictedby the modelat 298øK phasechangewouldraisethe reducedpartitionfunctionof

is quartz > calcite • albite > muscovite > enstatite quartzin relationto the other minerals[Kawabe, 1978],
anorthite,whereasTaylor[1974,p. 846]givesquartz> albite bringingthe predictedvaluesinto better agreementwith
> calcitefor the firstthree.At 800øK,just belowthe a-/3 Taylor's sequence.
quartz transition,the model predictsthe order quartz It has been demonstrated that crossovers in fractionation
calcite> albite-- muscovite> anorthite> enstatite,where- factorsin chemicalcompoundscan occur [Stern et al.,
as at high temperatures,Taylor gives quartz > calcite 1968],but they generally do not occur in setsof mineralsof
albite > muscovite --- anorthite. One difference is the value comparablecomposition(e.g., anhydroussilicatescarbon-
of quartzrelativeto calcite,and this couldbe accountedfor ate•)because of the regularityin behavior of vibrational
by the a-/3 phasechange,not consideredin the model. This modeswith degreeof polymerizationof the Si-O bondsand

TABLE 4. Calculated
andExperimental
Mineral-Water
Fractionation
Factorsat 298øand1000øK
298øK 1000OK

Mineral Model Experimental" Model Experimental

Quartz 30.9 30-34 NAt' NAt'
Calcite 28.1 28 -0.2 -0.2
Albite 27.2 29-30 -0.7 -0.5 -• - 1.0
Muscovite 23.1 17-23 -0.3 - 1.0-• - 1.5
Clinoenstatite 18.3 - 2.2
Anorthite 17.9 19-21 - 1.9 - 1.7-• -2.0
Diopside 14.1 - 2.6
Pyrope 13.8 -2.9
Grossular 12.7 -3.0 . . .'c
Zircon 11.4 - 3.0
Forsteritc 9.1 - 3.4
Andradite 8.2 - 3.5 ....
Rutile - 12.0 -5.6 -3.1
aDatafromeitherFriedman
andO'Neil[1977], Taylor[1974],
or Taylor
andColeman[1968].
t'NAisnotapplicable
because
of a-/3transition.
CFriedmanandO'Neil[1977]give1000In a for grossular-H20
andandradite-H20
as -1.6 and
-3.28, respectively.
Thismodelgives-3.6 and-4.2, respectively.
 KIEFFER2 THERMODYNAMICS AND LATTICE VIBRATIONS OF MINERALS, 5 841

the monotonicdependenceof the partition functionson this
behavior. This model suggeststhat crossoversin fraction-
ation factors of small magnitude can occur between pairs of
structurally dissimilar minerals: for example, the small
crossover between enstatite and anorthite around 325øK,
between muscovite and albite around 800øK, and between
calcite and muscovite at about 1100øK. The temperatures at
which thesecrossoversoccur cannot be accurately predicted
because the curves intersect nearly tangentially. Kyser
[1979] has reported a crossoverbetween olivine and pyrox-
ene at --•1200øC;there is no indication of such a crossover in
the results of the model used here, and effects other than
tion might be possible. At present, this requires more
information than is generally available for many minerals.
Griineisen Parameters
The harmonic thermodynamic properties change as a
function of pressurebecauselattice vibrational frequencies
shift as a result of compression.A frequently used theory is
that of Gr•ineisen [1912, 1926], in which the vibrational
frequenciesvi are assumedto shift as volume accordingto
,/i = -0 In l•i/C3
In V (22a)
or

fractionation resulting from the harmonic properties of lat-

v/vo= (Vo/V) • (22b)
tice vibrations need to be considered, such as anharmonicity
or nonequilibrium effects. The quantity •,• is referred to as the 'mode Griineisen
The model also shows that at high temperatures the parameter' of the lattice vibration at frequency v•. Many
reducedpartitionfunctionis proportional to 1/T2 asdemon- other definitionsof the Grtineisenparametercan be found in
stratedby Bottinga and Javoy [ 1973]. However, below about the literature, and one problem with application of the
700øK the reduced partition function for all minerals is less thermodynamic model of these papers to high-pressure
thanwouldbe predicted by the 1/T2trend.Accurateextrap- problems is a proper choice of a Grtineisen parameter or
olation of high-temperaturedata toward ambient conditions parametersto be appliedto the assumedlattice vibrational
(273øK) must account for this curvature. For an example of spectra.
the magnitude of the curvature effect, a difference of 20%o If the lattice vibrational spectrum of a mineral were well
from the calculated value at 298øK would be obtained by a approximatedby a Debye spectrum, then Debye theory
strict 1/T2 extrapolationof the high-temperature quartz would allow a simplecalculationof the effect of pressureon
values. the thermodynamicproperties,becausea singleGriineisen
As can be seenfrom Table 4, the quantitative agreementof parameter, •'a½,would be appropriate; effects of pressure,
the predicted values with data is good, except for anorthite temperature,and wave vector on • would be ignored, and
at low temperatureand rutile at high temperature. Quantita- compressionaland shearwave velocitieswould be averaged
tive uncertaintiescannot be placed on a model of this type, into a mean velocity. The changein the characteristicDebye
which is subjectto a large amount of data interpretation. My frequency vo resulting from volume change would be as-
impressionis that relative fractionationvalues for minerals sumed to be given by
that are structurally and chemically very similar (such as the
albite-anorthite relations, the garnet relations) are reliable 0 In vo 0 In 0o
•a½= -- (23)
becausethe spectral trends are systematic,but that a larger 0 In V 0 In V
uncertainty is associatedwith structurally dissimilar materi-
als because of the difficulties in assigningmode frequencies The Gr•neisen parameter appropriate here is an acoustic
and numbersto different types of minerals. On the basis of modeGr•neisen parameterobtainedfrom spatialaveragesof
Table 4, I would suggestthat at 298øK, differencesin pairs of the individual ,/• for the acousticbranches(denoted '/s and
mineral-water fractionations are accurate to -+ 1%ofor struc- '/v) [e.g., Anderson et al., 1968, p. 498]. With this assump-
turally similar minerals and _+2%o
for structurally dissimilar tion, the characteristic Debye frequency or Debye tempera-
minerals. ture changeswith volume according to

5. EFFECT OF PRESSURE ON THE THERMODYNAMIC

FUNCTIONS
•0-•-] (24)
A Murnaghanequationof state is commonlyused to relate
The influenceof pressureon thermodynamicpropertiesof the volume change to the pressure change:
solids becomes important at mantle and core pressures.
Predictionof phaseboundariesat thesepressures,reduction P = • - 1 (25)
of shockwave Hugoniotsfor isothermaland adiabaticphysi- n

cal parameters, and prediction of many physical properties

at high pressuresare examples of problems that require
estimation of the effect of pressure on such properties as
heat capacity and free energy.
The following discussion is restricted to the harmonic
approximation of the thermodynamic functions. At low
pressures, anharmonic effects are small, but anharmonic
behavior is expectedto be more significantat high pressure
because interatomic distances are reduced. With treatment
of total free energy of an atomic systemas a perturbation
expansionof the harmonicenergy, suchas that presentedby
Wallace [1972, chapter4], supplementationof the harmonic
results discussed here with an additive anharmonic correc-
where Br = - V(OP/OV)r and n = (OB•/OP)z.(This temporary
use of the symbol n as a derivative in the Murnaghan
equation should not cause confusion with its general use
throughout these papers as the number of atoms in the
chemical formula.) For example, for periclase, Bz = 1600
kbar, n = 4.5, and at 1000 kbar, Vo/V = 1.35. These values
give0o/0o
ø = 1.56;thusif 0oø= 950øK,0o= 1430øK
at 1000
kbar. The value of Cv* at 298øK and 1000 kbar is 0.59 of that
at 1 bar. For corundum, the decreasewould be to 0.85 of the
1-bar value and, for (metastable) quartz, to 0.86 of the 1-bar
value.
As was discussedin papers 1-4, a Debye spectrum does
842 KIEFFER: THERMODYNAMICSAND LATTICE VIBRATIONS OF MINERALS, 5

not representthe lattice vibrationalspectraof mostminerals (The reader is reminded of the discussionin paper 2 of the
with sufficient accuracy to be useful in problems of current vibrational spectrumof rutile which has at least one highly
interest. On the other hand, it is not generally possible to temperature-dependentmode.) $haner [1973] discussedthe
specify the real lattice vibrational spectrum, including the temperature dependence of •/i for rutile; the temperature
dependence of normal mode frequency on temperature, dependenceof •/i has also been discussedby a number of
pressure,and wave vector. Analogously,althougha single authors cited above.
Grfineisen parameter is not likely to describe the volume Slater [1939] has demonstrated that
dependenceof the vibrational spectrum,it is not practicalto
'• = '}/thermal
= eVBT/Cv (27)
specifya Griineisenparameterfor each of the 3s modesand
to specify its dependenceon temperature and wave vector. where e is the thermal expansionand B:r is the isothermal
Therefore it is necessaryto find a description of the way that bulk modulus. The proof by Slater does not require that the
lattice frequencies change with compressionthat is more individual mode Grfineisen parameters Yi all be equal, as
accurate than permitted by Debye theory and the elastic frequently required by other authors for this equality. If
Grfineisenparameter discussedabove, yet is feasiblewithin adequatedata on changesof elastic constantsand infrared
the context of limited data on the pressure dependenceof and Raman spectral lines with volume existed, the equality
vibrational spectra. of •/with •/thin (27) and (28) could be checked. Although
The work describedin thesepapershas demonstratedthat such data do not exist for most minerals, data for quartz are
vibrationalspectrahave at least three distinctparts--acous- sufficient to allow this equality to be checked (discussed
tic modes at low frequencies,optical continuummodesat below).
intermediatefrequencies,and stretchingmodesat high fre- If (26) is rewritten explicitly in terms of the 3s degreesof
quencies-that representvery differentdeformationalstyles. freedom used to characterizeprimitive unit cells of minerals,
One might reasonablyexpect that thesedifferentgroupsof it yields
modes might responddifferently to compression,that is,
1
might be characterizedby different Griineisenparameters.
For example, as a crystal is initially compressed,the 'soft' '•-'•VV
{('¾1sCv1s
q-'¾2sCv2s
q-'¾3pCv3P)
q-('y4Cv4
q-'¾5Cv5
(distortion)modesmight changefrequencyquite drastically
+''' + YxCvx)+ (Yx+•Cvx+•+ ''' + Y3sCv3s)} (28a)
(•/i large), whereasthe 'hard' (stretching)modesmight be
relatively unaffected(7i • 0). In fact, as will be discussed where the subscripts 1S and 2S indicate the two acoustic
below,thisbehaviorhasbeenobservedin spectraof quartz, shear modes; 3P indicates the longitudinal mode; 4, 5, ß ß ß,
olivine, orthopyroxene,and garnet taken at high pressure. x indicate the modes of the optic continuum; and x q- 1, ß ß ß,
Thus I have attempted to formulate a theory that will 3s indicate the modes of the Einstein oscillator(s). In each
describethe responseof the three parts of the vibrational case, the heat capacity is that of the mode at the temperature
spectrum--acoustic,low-frequencyoptic, and high-frequen- T at which •/is evaluated.
cy optic--to compression.The methodof extensionof this Equation (28a) can be written in terms of three compo-
treatment to more types of modes is obvious. nents that indicate the cumulative contribution of the acous-
In order to formulate the parameters of such a theory, tic, optic continuum,and Einstein modes,respectively,to •/:
assume,after Griineisen [1926], that a set of parameters, 3'i,
describes the changes of mode frequencies with volume • = '•ac+ '•op+ '•E (28b)
accordingto (22a) and (22b) above. The calculationof these where
mode Grtineisenparametersand their averageshas been a
subjectof much controversy(Born [1923], Barron [1955, 1
1957], Schueleand Smith [1964], Anderson [1968], Knopoff •/ac
= •VV
('}/1sCv1s
q-'}/2sCv2s
q-'}/3pCv3P)
(28c)
and Shapiro[1969], Shaner[1973],Demarest[1974],Ander-
son [1974], Mulargia and Boschi [1978], Demarest [1980], 1
Mulargia [1980], and many others). In the quasi-harmonic •op = • ('Y4Cv4q- 'y5Cv5• ''' q- %cCvx
) (28d)
approximationadoptedhere (in whichnormalmodefrequen- Cv
cies dependon volume, but not explicitly on temperature),
individual mode frequenciescan be averagedaccordingto 1

1 3s
•/=•vv
'••l3•,Cv, (26)Note that C/a
Grfineisenparametersand are not necessarilyequalto any of
Each mode has a characteristic•/i, and, in principle, the •/i
are functions of the magnitude and direction of the wave
vector K. Because this dependence is not known and be-
cause the model does not invite such intricacy, the depen-
dence of Yi on K is ignored. Wallace [1972, sections32, 34]
discussedthis dependencebriefly and gave calculatedvalues
of •/i as a function of K for sodium and potassium;no
experimental data exist to compare with calculated curves.
For consistencywith previous assumptions,I assumehere
that the mode Grfineisenparameters3•i,as well as the lattice
vibrational spectrumg(to), are independentof temperature.
•E= •VV
('}/x+
1Cvx+
1q-' ' ' q-'}/3sCv3s)
(28e)
c, C/o
p, and c/Eare weightedaveragesof mode
the •/i, and note the temperature dependenceof the relative
contributions•/ac,•/op,and •/Eimpliedin the Cv weighting
factors.
Acoustic mode Griineisen parameters, •/•s, 'Y2s, 'Y3P.
Several schemes have been devised for relating acoustic
mode Grtineisen parameters (•/•s, •/2s, •/3P)to elastic con-
stantsof the minerals. Although these schemesare based on
considerationsfor simple, isotropic minerals, they are fre-
quently applied to anisotropic minerals [Anderson et al.,
1968]; this practice will be followed here, because I have
demonstratedin papers 1-4 that acousticmodesaccountfor
 KIEFFER: THERMODYNAMICS AND LATTICE VIBRATIONS OF MINERALS, 5 843

so few of the total degreesof freedom for most mineralsthat
slighterrors due to averagingtechniquesfor wave velocities
will not appreciablyinfluencethe calculatedthermodynamic
properties.
The most common approximation [Barron, 1955, 1957] is
calculations by uncertainty in compression, the calculated
value of •/= 0.73 _ 0.09 is remarkably close to ¾th(0.703),as
predicted by (27).
The relative contributions of acoustic and optic modesto •/
can be obtained from evaluation of the three sets of paren-
theses in (28c)-(28e):

7i-'• Vi•OV/T
=•-+• ' (29a) •/ac= 0.065 (31a)

where V is volume, vi is an acoustic velocity (i = 1, 2 for •/op= 0.66 (3lb)

shear modes; i = 3 for the compressionalmode), and Kr is •/E '• 0 (31C)
the isothermal bulk modulus. Another form of this equation
was given by Schuele and Smith [1964] as (Values given are the mean of calculations for two values of
V/Vo as discussed in footnotes to Table 5.) The acoustic
modes and high-frequency Si-O stretching modes do not
= + -
contribute appreciably to •. The modes of the optic continu-
um are by far the most significant, and, in fact, as can be
In this equation, M• is the appropriate elastic modulus seen from Table 5, the two lowest-frequency optic modes
co•esponding to the shearor longitudinalvelocities. In (29a) alone account for much of •/and, therefore, 3'th.The relative
the vi are to be taken as directionally averaged sound speeds. contributions of the modes are determined (1) by the values
However, as was discussedin paper 1, t•e directionally of the mode 3'i, (2) by the values of the heat capacity
averaged velocities are rarely available. Furthermore, their weighting factors, and (3) by the relative numbers of modes.
pressure derivatives are also not available. Therefore, for Because the modes having the largest 3'iare low-frequency
implementation of (29a) the velocities are commonly re- modes, at room temperature their heat capacity weighting
placed by Vogt-Reuss-Hill (VRH) averagedvelocities. That factors are relatively large, and they dominate •/and 3%.
convention is followed in this paper to calculate acoustic y• The contribution of Si-O antisymmetric stretching modes
because the derivatives Ovi/OPfor directionally averaged to the averaged Gr•neisen parameter of quartz at 298øK is
modes are not known. Therefore, even though the two shear negligible for two reasons: (1) the heat capacity weighting
modes are taken with differing velocities in the model, the factors are small, and (2) the mode Gr6neisen parameters are
VRH shear velocity is used in conjunction with an averaged also typically small. The mode Gr•neisen parameters are
velocity pressurederivative to calculate Yi in (29a). For the small becausethe strong Si-O bond is relatively incompress-
exampleof qua•z to be considered here, V = 22.69cm3, ible (for example, see discussionsin the work by Lazarev
vva•,s= 4.05kms-•, vw•,e = 6.02kms-• (paper1),(OVs/ [1972]) so that at relatively low pressures, compression is
OP)= -3.4 x 10-3 km s-• kbar-•, and (Ove/OP) = 10.4 x taken up by weaker bonds in the mineral structures. There-
10-3 km s-• kbar-• [Andersonet al., 1968].Equation(29a) fore the frequencies of the lower-frequency modes associat-
gives Y•s = Y:s = 0.023 and Y3e = 1.18 in agreementwith ed with weaker bonds are most affected by overall compres-
Anderson et al. [1968]. sion. Similar behavior is observed in the few other silicates
Optic mode Graneisen parameters, Y4''' Y3s. If infra- for which high-pressurespectral data are available.
red and Raman spectra of minerals have been measuredas a Dietrich and Arndt [1982a, b] have measured the mid-
function of volume, then optic mode Grfineisen parameters infrared absorption spectrumof olivine, orthopyroxene, and
can be calculated from
garnet to 50-kbar pressure and have calculated microscopic
Yi = -0 In v•/OIn V (30) Gr•neisen parameters for the observed modes. These mea-
surementsshow that the ¾iof the vibrations associatedwith
Spectroscopicdata [e.g., Asell and Nicol, 1968]demonstrate the SiO4-4 tetrahedronare smallandcontributelittle to the
that the y• are generallydifferentfor the differentmodesand, macroscopicGr•neisen parameters 3'th.However, because
indeed, may be positive, negative, or zero. The volume the measurements do not include all infrared modes or
changein (30) must be measureddirectly or infe•ed from Raman modes, the equality of •/and ¾thfor these minerals
measuredpressurechangesthroughequation-of-statedata. cannot yet be tested through (28) and (27). Calculations of
Asell and Nicol [ 1968]havepublishedfrequencyshiftsfor deformation of individual polyhedra in minerals, such as
a su•cient number of Raman active lines of the a qua•z those of Hazen and Prewitt [1977] and Hazen and Finger
spectrumto allow detailed examinationof the 3 of optic [1979], when combined with spectroscopic observations,
modes and testing of (27) and (28). Their stated highest such as those of Asell and Nicol [1968], may eventually
pressurewas 'about 40 kbar,' later co•ected to 32 kbar (M. provide more accurate values of the mode Gr•neisen param-
Nicol, personalcommunication,1975).Given the pressure eters and allow more detailed treatment of the pressure
unce•ainty and adoptingseveralpossibleequationsof state, dependenceof the thermodynamicfunctions than is possible
the relativevolume (V• at the highestpressurecouldhave with current available data.
been 1.067-1.086. The compressionis probably uncertain Grt;ineisenparameters used in model. The simplestas-
within this range, and therefore the calculations below are sumption that might be used is that all mode Gr•neisen
given for the two values; the averageof the two values is parametersare equaland, from (27), are equalto 3'th;this will
taken as the most probable value. Infe•ed values of the be referredto as model1. This modelis directlycomparable
Graneisenparameterare very sensitiveto volumechanges to a Debye model, in that there is a single Gr6neisen
and thereforeto unce•ainties in volumechanges.The calcu- parameter,but the form of the lattice vibrationalspectrumis
lations of • from the mode Graneisenparametersy• are changedfrom a Debye spectrumto the more realistic form
shown in Table 5. Within unce•ainties introduced into the proposedin paper 3.
 844 KIEFFER:
THERMODYNAMICS
ANDLATTICE
VIBRATIONS
OFMINERALS,
5

TABLE 5. Valuesof 3'ifor QuartzandCalculations

to Demonstrate
theEquivalence
of • and3'th
Mode
Vo, v32
kbar,
a v0,
b Degen- Cv*(Xoi)
d 3'ifor 3/iCvi
cm-• cm-• cm-• eracy Xoic 3R Vo/V= 1.067e 3R
Optic Continuum
128 147 128 2 0.62 0.036 2.13 0.077
207 262 _ 5 207 1 1.00 0.034 f 3.63 0.123
265 ('-• 280)g 264 2 1.28 0.032 0.85 0.027
356 358 356 1 1.72 0.029 0.09 0.000
365 1 1.76 0.029 (0.00) (0.000)
394 395 397 2 1.92 0.027 0.04 0.001
401 407 + 5 ...n
450 471 _ 5 452 2 2.18 0.025 0.70 0.018
464 495 _+4 466 1 2.25 0.025 1.00 0.025
509 ß ßß 520 1 2.51 0.022 0.5i 0.011
697 720 _ 6 695 2 3.36 0.016 0.5 0.008
795 820 780 1 3.77 0.013 0.48 0.006
807 835 800 2 3.86 0.012 0.53 0.006
Si-O Stretch
1055 1 5.10 0.006 0.01 0.000
1072 1073 1064 2 5.14 0.006 0.01 0.000
1085 1088 1082 1 5.22 0.006 0.04 0.000
1162 1155 1160 2 5.60 0.004 - 0.09 - 0.000
1231 1240 _ 10 ...n
1235 1250 . . .n

Acoustic, From Model

S1 102 0.493 0.036 0.023 0.000
$2 122 0.589 0.036 0.023 0.000
P3 164 0.792 0.035 1.18 0.041

aOpticmodefrequencies
at ambient
pressure,
Vo,andat 32kbar.Thepressure
wasoriginally
reported
as'about40 kbar'byAselland
Nicol [1968],and was correctedto 32 kbar (M. Nicol, personalcommunication,
1975).
•'Frequencies
andmodedegeneracies
atambient
pressure,
fromLordandMorrow
[1957],
given
forcomparison.
CEqualto hvi/kT, T = 298øK.
'. dThe
heat
capacity
summations
required
fortheweighting
factors
in(27)
are•i=4
3sCv?/3R
(for
alloptical
modes)
=0.480;
•i=4
3sCvi*/3R
(foracoustic
modes)= 0.107;Z Cv? (total)= 3.50calmol*-l øK-1in goodagreement withtheexperimental value,3.54calmol*-l øK-1.
eCalculated
from3'= -0 In v/OIn V = In V/Vo
+ In Vo/V.If Vo/V= 1.067(asin thetable),In Vo/V= 0.06485.
If Vo/V= 1.087,In Vo/V=
0.0834 and all optical mode 3'ishownin this columnwould be reducedto 77% of the listed value.
fForVo/V= 1.067,
9 = Z3'iCvJZCvi
= (0.439
+ 0.0426)/0.588
= 0.82.ForVo/V
= 1.087,
9 = 0.65.Therefore
thebest
estimated
value
of9,
givenuncertainties
in compression,
is the mean,0.73 -+ 0.09. For comparison,
3'th= 0.703[Anderson
et al., 1968].
gEstimatedby extrapolation of Asell and Nicol [ 1968] data at 20 kbar.
nNotseen,assumed
spurious.
/Assumed.

A model that more nearly agreeswith the spectroscopic obtained from

datadiscussed abovecanbe obtainedby estimatingseparate 3 3s

Grfineisenparametersfor acousticandopticmodes;thiswill '•op--3/th--3/iZ Cvi q- Z Cvi (33)

be referred to as model 2. If elastic constants and thermal- i=1 i=1

expansiondata are available, then (27) and (28) can be used The individual3/opto be appliedto the opticcontinuumis
to allow specificationof acousticand optic modeGrfineisen thendetermined from(32).As an example,consider quartz:
parameters. 3/th-- 0.703, 3/• = 3/2= 0.023, 3/3= 1.18, and •ac = 0.07.
Considerthe three parts of (28), givenas (28c), (28d), and
Therefore9op= 0.63.At 298øKthe ratioof specificheatsof
(28e). The first part, (28c), contains the acoustic mode quartzrequired in (32)is 0.77,so3/op - 0.63+ 0.77-- 0.82.
Grfineisenparameterswhich can be specifiedfrom (29a) or In the calculations that follow, the valuesof 3/usedhave
(29b). The secondpart, (28d), containsGriineisenparame- beenevaluatedwith heat capacitydata (for (28)) and ther-
ters appropriateto the optic continuum;theseare generally mal-expansion data(for 3/th)takenat 298øKonly.The error
unknown. Theywillbeassumed uniformanddenoted as3/op, introduced by not considering the variationof the 3/with
where
temperatureabove 298øK is probably less than that intro-
x 3s x 3s
ducedby the harmonicapproximations used.
'•op--Z Cvi3/op
"--Z Cvi--3/opZ Cvi+ Z CVi (32) Calculationof Thermodynamic
Properties
i=4 i= 1 i--4 i= 1 at High Pressure

Equation(28e)contains theGrfineisenparametersappropri- For the model calculations,pressureand volume are

ate to the high-frequencystretchingmodes;the summation assumedto be relatedthrougha Murnaghanequationof
over this set of parametersis assumedto be zero, either state'
because
theheat
capacities
ofth•modes
at298øK
arenearly
zero or becausethe modeGrfindisenparametersthemselves P = •n - 1 (34a)
areapproximately
zero.Then,if 9 = 3/th
asin(27),'•op
canbe
 KIEFFER:THERMODYNAMICS
AND LATTICEVIBRATIONS
OF MINERALS,5 845

structural
types'halite,periclase,corundum, spinel,quartz,
V = V0
[ii•TT
]-1/n
+ 1

A = (V/Vo- 1)
(34b) stishovite,rutile, forsterite, and pyrope. The Griineisen
parameters
usedinthethreemodels
(34c) bulkmodulus
pressure
derivatives
discussed
aboveandthe
usedfor the calculations
are givenin Table 6. Becausemanyof theseminerals
whereBr is the isothermal bulk modulus,n is the bulk undergo phasechanges at highpressures, thehigh-pressure
modulus pressure derivative, andA is the dilatation. The calculationsmust De regardedas applyingto metastable
Murnaghan equation of stateis quiteaccurate at relatively phasesfor most minerals.
low pressures (say,belowthe pressure equalto the bulk To test the effect of usinga singleGriineisenparameterto
modulus) but is lessaccurateat highpressure anddoesnot represent the shiftof the opticcontinuum ratherthanindi-
accountfor phasechanges. Morecomplicated andaccurate vidual mode Graneisenparameters,•t comparisonof the
equationsofstate,andbetterandmoredetailed valuesofthe modelvaluesof heatcapacityfor quartzwasmadewith heat
Griineisenparameters (including perhaps temperature and capacities obtained byconsidering theindividualmode'shifts
volumedependences), couldbeusedfor minerals forwhich measured at 32kbar,byAsellandN•Col[1968].Following
availabledataon physicalparameters justifysuchtreatment the method of Lord and Morrow [ 1957], the 0ptic"modesof
and the logic of the methodpresentedhere wouldnot quartzwererepresented by 24 Einsteinoscillatorsat the
change. frequencies givenin the third columnof Table 5 for a
All modelparameters mustbe specified at the pressure pressure of 1barandin thesecond column fora pressureof
underconsideration.
Densitychangesare impliedby (34b). 32 kbar,. The acoustic modes were representedby three
Acoustic velocities are assumedto change according to
Debye
distributionS.
Theharmonic
molar
heatcapacities
at
[Kieffer, 1976] 300øK..thus
obtained
were10.53calmol-• øK-• at 1barand
Oi = OOi
[1 - A(Ti-- -})] (35) 10.30
calmol-• øK-•at32kbar.Withmodel 2 o•thispaper,
the harmonicheat capacityat 300øK,1 bar, is 10.59cal
andall characteristic
opticmodefrequencies
areassumed
to mol-• øK-l,andthatat 300øK,
32kba,
r, is 10.25
½almol-•
changeaccordingto oK- 2. The < 1% differencein valuesat 1 bar is causedby the
12
i = 12oi(
Vo/V)•/i (36) verydifferent
natureofthevibrational
spectra assumed.The
importantpointis thatin bothmodelsthe heatcapacityat
The spectrum
of vibrational
frequencies
thusimpliedis 300øKchanges by nearlythesameamount: it is97to 98%of
appropriate
to thepressure
P andvolumeV, andcalculationthe 1-bar value at 32 kbar. Therefore, in at least this one
of thethermodynamicfunctions
followsfromtheequations case,representation of the frequencyshift of the optic
of paper 3. continuumby an averagedGraneisenparameterdoesnot
substantially
changethecalculated
pressure
depen•lence
of
Results the heatcapacities.
The agreement
is comparably
goodat
Because
datarequired
for calculation andoptic higher temperatures.
of acoustic
parameters
arelimited,calculations
arepresented
herefor Results of the calculationsfor other minerals are given in
onlynineofthe32minerals were Table7 (valuesof the heat capa•city
of thisstudy.Thefollowing at 298øKfor the two
selectedto representdifferentchemicalcompositions and modelsand for a Debye model •at1 bar and100,500,and

TABLE 6. GriineisenParametersand Bulk ModulusDerivative Used for High-PressureCalculations

M•del1 Model2
Tth(From Tth(Model dKr
Literature) Parameters) Tl = T2 % •op Top dP
Halite 1.55 1.50 0.14a 2.64a 1.05 2.19 6.00b
Periclase
c 1.54 1.56 1.40 1.71 0.68 1.62 4.52
Corundum c 1.32 1.48 1.20 1.53 1.15 1.35 3.99
Spinel
c 1.13 1.47 0.49 1.33 1.0,6 1.19 4.19
Quartz
c 0.70 . . .d 0.02 1.18 0.63 0.82 6.40
Stishovite 1.00
e 2.50
r 0.59
r 2.35
f 0.71 0.95 5.00
e
Rutile 1.50
g 1.69 0.70
g 0.70
g 1.35 1.73 6.80h
Fosterite
c,/ 1.3j 1.25 0.88 1.65 1.24 1.41 4.90
Pyrope
c,i 1.21 1.12 1.13 1.95 1.18 1.30 5.45
aSchuele and Smith [1964].
bBirch[ 1966].
CAnderson [ 1974].
dThesemodelvaluesare very uncertainbecauseof behaviorof thermalexpansion.
eKalinin and Pan'kov [1973].
fEstimated
fromdKr/dP= 5.6anddKr/dT- 6.9andtheequations
of Anderson
et al. [1968]byJ. Delaney
(personal
communication,
1979).
gShaner [1973].
hAnderson[ 1974].
/Dataof DietrichandArndt[1982a,b], available
onlyaftercompletion
of thiswork,allowcalculation
of Tifor twobandsin theoptic
continuum of olivineandthreeforpyrope.Thesebands arisefrombendingmotionsof SiO4
-4 tetrahedra,
andtheTiaresmaller thanthe
valueadopted hereforthewholeopticcontinuum. However, theSiO4-4 bending
modes constitute
onlya relatively
small fraction
ofthe
modes in thecontinuum (seetextsection
onisotopic
fractionation),
sothesedataarenotinconsistent
withthevaluesof •/op
adoptedhere.
JChung[ 1971].
846 KIEFFER: THERMODYNAMICSAND LATTICE VIBRATIONS OF MINERALS, 5

TABLE 7. Cv* and Shar*(in cal mol*-• øK-i) at 298øK

Halite Periclase Corundum Spinel Quartz Stishovite Rutile Forsterite Pyrope
Pressure Cv* S* Cv* S* Cv* S* Cv* S* Cv* S* Cv* S* Cv* S* Cv* S* Cv* S*
Model 1'•
1 bar 5.66 8.39 4.25 3.15 3.78 2.47 3.89 2.71 3.53 3.48 3.56 2.54 4.26 3.97 3.98 3.35 3.88 3.24
100 kbar 5.40 6.46 4.00 2.74 3.62 2.28 3.74 2.50 3.28 3.06 3.48 2.42 4.11 3.70 3.77 3.02 3.72 3.00
500 kbar 4.95 4.84 3.33 1.91 3.15 1.77 3.28 1.97 2.97 2.59 3.22 2.11 3.73 3.09 3.26 2.34 3.32 2.45
1000 kbar 4.66 4.19 2.84 1.46 2.74 1.41 2.90 1.62 2.82 2.38 2.99 1.87 3.45 2.71 2.89 1.93 3.04 2.11

Model 2
1 bar 5.66 8.39 4.25 3.15 3.78 2.47 3.89 2.71 3.53 3.48 3.56 2.54 4.26 3.97 3.98 3.35 3.88 3.24
100 kbar 5.27 6.42 4.00 2.74 3.62 2.27 3.73 2.49 3.31 3.05 3.48 2.42 4.09 3.69 3.81 3.02 3.76 3.00
500 kbar 4.53 4.79 3.30 1.91 3.13 1.76 3.24 1.96 3.00 2.57 3.22 2.12 3.67 3.07 3.36 2.33 3.42 2.44
1000 kbar 4.04 4.18 2.80 1.46 2.71 1.40 2.85 1.60 2.85 2.35 3.00 1.89 3.36 2.69 3.02 1.92 3.16 2.09

Debye Theory
1 bar 5.65 3.79 3.52 4.00 5.01 3.86 4.33 4.43 4.29
100 kbar 5.40 3.53 3.36 3.84 4.81 3.77 4.16 4.23 4.11
500 kbar 4.91 2.79 2.85 3.36 4.50 3.50 3.71 3.65 3.64
1000 kbar 4.56 2.22 2.41 2.97 4.34 3.26 3.36 3.21 3.29

For Shar*, only the harmonic contribution to the entropy is given because the effect of pressure on the thermal expansion is not well
known. Hence these numbers differ by a few percent from those in paper 4, Table 5.
aFormodel1, all 3'= Tth.For model2, T•sand Yes= Ts;T3z,= %,;Topfor theopticcontinuumis describedin the text. The parameter3'for
Einstein oscillators = 0.

1000 kbar) and in Figure 9 (a plot of Cv* versus T for the model spectra rather closely resemblea Debye spectrum
same pressures). Consider first the values of Cv* and $* (see paper 3). Debye theory does not predict accurateheat
shown in Table 7. The results of models 1 and 2 at 298øK are capacitiesfor the remainingminerals;thusonly the predict-
very similar to each other and differ substantially from ed relative decreaseof heat capacity with pressuremight
Debye theory, except for halite, for which the actual and have any significance.Comparisonof the resultsof model 1
with the Debye theoryshowsthe effectof differingvibration-
6 al spectrabecauseboth model 1 and Debye theory use the
sameGriineisenparameter, 7th,for all modes.
4 With the exceptionof halite, models1 and 2 do not give
significantlydifferent results for Cv* and S* at 298øK, an
2 initially surprisingresult because of the apparent large
differencesin Griineisenparametersusedin the two models.
o
However, in a generalway, this is explainedby the fact that
valuesof 7thusedfor all modesin model 1 are rather closeto
4
the valuesof %0 appliedto the opticcontinuumin model2
(Table 6). Because the optic continuum contains >50% of
2
the modes, the major spectral shifts are similar in the two
models,and therefore,sois the pressuredependenceof Cv*
• 0
and S*. The differences between the two models occur in
two spectral regions not strongly reflected in the heat
• 4
capacityor entropyat 298øK:in the low-frequencyacoustic
modes and in the high-frequencyEinstein oscillators.At
298øK, the temperature for which the values are listed in
Table 7, the acousticmodesof mostmineralsare nearlyfully
active and contributetheir maximumheat capacity,regard-
less of the shift of their frequencieswith pressure.On the
otherhand, at this temperature,the high-frequencyEinstein
oscillatorscontributevery little to the heat capacity,and so
2 PYROPE
differencesin their spectralshiftsare not stronglysensedin
the thermodynamic functions.
Halite is the only mineralfor which the two modelsgive
substantiallydifferent pressuredependencesof Cv*. The
differences arise because neither the acoustic mode nor the
opticmodeGriineisenparametersare closeto 7th:71= 3/2<<
0•" • I I I I I I THEIEND!I
0 200 400 parameter are applied to the spectrum in model 2, the shear
600 800 0 200 400 600 800 I000
7,th,73>> 7th,and •/op> 7th-When thesevaluesof Griineisen

TEMPERATURE (K) TEMPERATURE (K) acousticmodesremain at lower frequenciesthan in model 1,

Fig. 9. Model heat capacities of various minerals (or their andthe compressional modegoesto higherfrequencies(this
metastablelow-pressurephases)at pressuresof 1 bar and 100, 500, is analogousto the way a Debye spectrumsplitwhenshear
and 1000 kbar. and acousticmodeswere consideredseparatelyas discussed
 KIEFFER:
THERMODYNAMICS
ANDLATTICE
VIBRATIONS
OFMINERALS,
5 847

in paper1).Theopticcontinuum
of model2 is shifted
to m, m :g massesof exchangingisotopes,(10);the asterisk
substantially
higher
frequencies
because
ofthehighvalueof denotesisotopically
heavysubstance not to be
confusedwith the asteriskas usedin Cv* (cal
•op.The net effectis thatmodel2 predicts
a stronger mol*-• øK-1) where the asterisk denotesper
decrease
in heatcapacitywith pressure
thanmodel1. The
atomic unit.
entropy
islesssensitive
to thesemodeldifferences
because
muchof theentropyof haliteis determined
by theacoustic n bulk moduluspressurederivative,(25).
modeswhich have opposingand generallycompensatory partitionfunction,(5).
effects. reducedpartitionfunction,(10).
For periclase,
corundum,
andpyrope,theacoustic and number of atoms being exchangedin exchange
reaction, (10).
opticmodeGr0neisen
parameters
areall similarto 3•th,
SO
models1 and 2 are similar.For spinel,quartz, stishovite, fractionation factor, (13).
rutile, and forsteritc,the shearGr0neisenparametersare dilatation, (34c).
low,asin thecaseof NaCI,buttheacoustic modesaccount weightedaverageGr0neisen
parameter,(26).
for sofew of the totaldegreesof freedomthatthe effecton acousticmode Gr0neisenparameter,(23).
%pthrough
(33)andonthepressure
dependence
oftheheat averageGr0neisenparameterfor the Einstein
oscillators continuum, (3 lc).
capacityis small.
As shownin Figure 9, the relativedecreasein heat mode Gr0neisenparameter, (22a).
capacityat a givenpressureis a functionof temperature, averageGr0neisenparameter
for the opticcon-
beinglargeonthesteeppartoftheCv*curveandapproach- tinuum, (3 lb).
ingtheDulong-Petitlimitat hightemperatures at all pres- •t3p acousticlongitudinalmodeGrtineisenparameter,
sures. Anharmoniceffects would changethese high-tem- (28).
acousticshearmodeGrOneisenparameters,(28).
perature limits.Becausethechanges in heatcapacity
are ')/IS, •t2S

largeatrelativelylowtemperatures,
it istobeexpectedthat Yth thermal Gr0neisenparameter,(27).
changes intheentropy withpressure
should besubstantial,
as is documented in Table 7. Acknowledgments. It hasbeena greatpleasure
to workwiththe
editors
andstaffofReviewsofGeophysicshndSpace Physics
during
6. CONCLUDING COMMENTS
preparation
of thesepapers,especiallywithassociate-editor
Art
Boettcherandthe innumerablereviewers.Helpfulreviewcomments
The modelspresented of on
in this paperfor calculations thispaperwereprovided by J. Delany,R. Reeber,R. Robinson
and V. Wall. The author bears full responsibilityfor remaining
pressureeffectson the thermodynamic functions,phase problems,andis wellawareof howmanywereeliminated bytheir
equilibria,
andisotopicfractionationcannotbe correctin efforts.
detail becauseof the many simplifications
that had to be
REFERENCES
madein therepresentation of physicalparameters,suchas
thethermalexpansions andtheGr0neisen parameters, and Akimoto, S.-I., and Y. Syono, Coesite-stishovite transition,J.
in the formulationof simplifiedandevenhypothetical spec- Geophys.Res., 74, 1653-1659,1969.
tra.Thesimplifications weregenerallynecessary because of Althaus,
E., Thetriplepointandalusite-sillimanite-kyanite,
Mineral. Petrol., 16, 29-44, 1967.
Contrib.
a lackof experimental data,notbecause of a lackof theory Anderson, O. L., Someremarksonthevolumedependence of the
to incorporatetheeffectsintothemodel.Forexample, the Griineisenparameter, J. Geophys. Res.,73,5187-5193,1968.
temperature and volumedependences of the Griineisen Anderson, O. L., Thedetermination of thevolumedependence of
parameters
caneasilybeincluded
inthemodel theGriineisen
aspresented,Anderson, parameter,
J. Geophys.
Res.,79, 1153-1155,
1974.
but available data on most minerals do not allow such a O. L., E. Schreiber,
R. C. Leibermann,andN. Soga,
Some elasticconstantdata on mineralsrelevantto geophysics,
refinement;
thesimplified
spectra
canbedivided
andsubdi- Rev. Geophys.SpacePhys.,6, 491-524,1968.
vided into smallercontinuaand more Einsteinoscillatorsas Ascii,J. F., andM. Nicol, Ramanspectrum
of a-quartzat high
ourdatabaseandunderstanding spectra pressures,
oflatticevibrational J. Chern.Phys.,49, 5395-5399,19.68.
increase.The work incorporated
in the fivepapersherehas Barron,T. H. K., On the thermalexpansion of solidsat low
temperatures,
Philos.Mag.,Ser.7, 46,720-734,1955.
spannedmorethana decade, during whichthedataavailable Barron,
T. H. K., Grii,neisen
parameters
fortheequation ofstateof
haveincreased
sorapidly
thatnotallofit could
beincoporat- solids,Ann. Phys., 1, 77-99, 1957.
in dataleadsto hope Becker,R. H., Carbonandoxygenisotoperatiosin ironformation
edintothemodels.Thisrapidincrease
that the current decadewill be one in which continuing and associatedrocks from the Hamersley Range of Western
Australiaand their implications,Ph.D. thesis,Dep. of Chem.,
progress
will be madein measurement of the physical Univ. of Chicago,Chicago,Ill., 1971.
properties
of minerals,
in understanding
of theirvibrationalBigeleisen,
J., andM. G. Mayer,Calculation of equilibrium
con-
spectra,
andin integrationof the understandingof their stantsfor isotopicexchangereaction,J. Chem.Phys.,15, 261-
physical,
chemical,thermodynamic, structural,andspec- 267, 1947.
Birch,F. C., Compressibility;
elasticconstants,
in Handbook of
troscopicproperties. PhysicalConstants,editedby S. P. Clark,Mem.Geol.Soc.Am.,
NOTATION
vol.97,97-173,GeologicalSocietyof America,NewYork, 1966.
Birch,F. C., and P. LeComte,Temperature-pressure planefor
The reader is referred to the notation lists at the ends of albitecomposition,
Am.J. Sci.,258,209-217,
1960.
Boettcher,A. L., and P. J. Wyllie,The quartz-coesite
,transition
papers1-4. Additional
symbols
defined
belowareusedin measuredin thepresence
of a silicateliquidandcalibration
of
thispaper(equationnumberfollowsdefinition). piston-cylinder
apparatus,
Contrib.Mineral.Petrol.,17,224-232,
1968.
Br isothermalbulk modulus,(25). Boettcher,
A. L., andP. J. Wyllie,Phaserelationships
in thesystem
Ei molecularenergylevel, (6). NaAISiO4-SiO2-H20 to 35 kilobarspressure, Am. J. Sci., 267,
•f reducedpartitionfunctionratio, (10). 875-909, 1969.
K equilibriumconstant,(5). Bohlen,S. R., andA. L. Boettcher,
Thequartz-coesite
transforma-
848 KIEFFER: THERMODYNAMICSAND LATTICE VIBRATIONS OF MINERALS, 5

tion: A precisedeterminationand the effectsof other components, Hazen, R. M., and C. T. Prewitt, Effects of temperature and
J. Geophys. Res., 87, 7073-7078, 1982. pressureon interatomic distancesin oxygen-basedminerals,Am.
Born, M., Atomtheorie des festen Zustandes, in Encyklopadie der Mineral., 62, 309-315, 1977,
Mathematischen Wissenschaften,Physik, vol. 3, Teubner, Leip- Herzberg, G., Infrared and Raman Spectra of Polyatomic Mole-
zig, 1923. cules, 632 pp., Van Nostrand, New York, 1945.
Bottinga, Y., Calculation of fractionation factors for carbon and Hlabse, T., and O. J. Kleppa, The thermochemistryof jadeite, Am.
oxygen isotopic exchangein the systemcalcite-carbondioxide- Mineral., 53, 1281-1292, 1968.
water, J. Phys. Chem., 72, 800-808, 1968. Holdaway, J. M., Stability of andalusiteand the aluminum silicate
Bottinga, Y., Carbon isotope fractionation between graphite, dia- phase diagram, Am. J. $ci., 271, 97-131, 1971.
mond and carbon dioxide, Earth Planet. $ci. Lett., 5, 301-307, Holland, T. J. B., The reaction albite = jadeite + quartz determined
1969. experimentally in the range 600-1200øC, Am. Mineral., 65, 129-
Bottinga, Y., and M. Javoy, Commentson oxygenisotopegeother- 134, 1980.
mometry, Earth Planet. $ci. Lett., 20, 250-265, 1973. Holm, J. L., and O. J. Kleppa, The thermodynamicpropertiesof the
Boyd, F. R., and J. L. England, The quart.z-coesitetransition, J. aluminum silicates, Am. Mineral., 51, 1608-1622, 1966.
Geophys. Res., 65, 749-756, 1960. Holm, J. L., O. L. Kleppa, and E. F. Westrum, Jr., Thermodynam-
Boyd, F. R., P.M. Bell, J. L. England, and M. C. Gilbert, Pressure ics of polymorphictransformationsin silica, Thermal properties
measurementin single-stageapparatus, Year Book Carnegie Inst. from 5 to 1070 K and pressure-temperaturestability fields for
Washington, 65, 410, 1966. coesite and stishovite, Geochim. CosmoChim.Acta, 31, 2289-
Chung,D. H., Elasticityandequations
of stateof olivinesfor the 2307, 1967.
Mg2SiO4-Fe2SiO
4 system,Geophys.J. R. Astron. $oc., 25, 511- Hulston, J. R., Methods of calculating isotopic fractionation in
538, 1971. minerals, in Stable Isotopes in the Earth Sciences, edited by P.
Clayton, R. N., Isotopic thermometry, in Thermodynamics of W. Robinson, pp. 211-219, New Zealand Department of Science
Minerals and Melts, Adv. Phys. Geochem., vol. 1, edited by R. C. and Industrial Research, Science Information Division, Welling-
Newton, A. Navrotsky, and B. J. Wood, pp. 85-109, Springer- ton, N. Z., 1978.
Verlag, New York, 1981. Iiishi, K., T. Tomisaka, T. Kato, and Y. Umegaki, Isomorphous
Dachille, F. D., and R. Roy, High pressure region of the silica substitution and infrared and far infrared spectra of the feldspar
isotopes, Z. Krista!!ogr., 111, 451, 1959. group, Neues Jahrb. Mineral. Abh., 115, 98-119, 1971.
Day, H. W., and H. J. Kumin, Thermodynamic analysis of the Jeanloz, R., and A. B. Thompson, Phase transitions and mantle
aluminumsilicatetriple point,Am. J. $ci., 280, 265-287,1980. discontinuities,Rev. Geophys. Space Phys., in press, 1982.
Demarest, H., Lattice model calculation of Hugoniot curves and the Johannes, W., et al., An interlaboratory comparison of piston
Griienisen parameter at high pressurefor alkali halides,J. Phys. pressurecalibrationusingthe albite breakdownreaction, Contrib.
Chem. Solids, 35, 1393-1404, 1974. Mineral. Petrol., 32, 24-38, 1971.
Demarest, H., Comment on 'On the theoretical evaluation of the Kalinin, V. A., and V. L. Pan'kov, Equations of state of stishovite,
Griineisenfunction in the harmonicapproximation'by Francesco coesite and quartz, Izv. Akad. Nauk $$SR Fiz. Zemli, 8, 3-16,
Mulargia,J. Geophys.Res., 85, 1511-1512,1980. 1973. (Izv. Acad. $ci. U$$R Phys. Solid Earth, Engl. Transl., 8,
Dietrich, P., and J. Arndt, Effects of pressure and temperature on 495-503, 1973.)
the physical behavior of mantle-relevant olivine, orthopyroxene Kawabe, I., Calculation of oxygen isotope fractionation in quartz-
and garnet, 1, Compressibility, therma.l properties and macro- water system with special reference to the low temperature
scopicGriineisen parameters,Proc. Lunar Planet. $ci. Conf., in fractionation, Geochim. Cosmochim. Acta, 42, 613-621, 1978.
press, 1982a. Kelley, K. K., et al., Thermodynamicproperties of sodium-alumi-
Dietrich,P., and J. Arndt, Effectsof pressureandtemperatureon num and potassium-aluminumsilicates, U.S. Bur. Mines Rep.
the physical behavior of mantle-relevant olivine, orthopyroxene Invest., 4955, 1953.
and garnet, 2, Infrared absorption and microscopic Griineisen Kieffer, S. W., Lattice thermal conductivity within the earth and
parameters, Proc. Lunar Planet. $ci. Conf., in press, 1982b. considerationsof a relationshipbetweenthe pressuredependence
Elcorobe, M. M., Some aspects of the lattice dynamics of quartz, of the thermal diffusivityand the volume dependenceof the
Proc. Phys. $oc. London, 91, 947, 1967. GrClneisenparameters,J. Geophys.Res., 81, 3025-3030, 1976.
Elcombe,M. M., Determinationof atomiceigenvectormagnitudes Kieffer, S. W., Thermodynamicsand lattice vibrations of minerals,
by isotopic fractionation between sphalerite and galena using 1, Mineral heat capacitiesand their relationshipsto simplelattice
latticedynamics,Earth Planet. $ci. Lett., 28• 172'-180,1974. vibrational models, Rev. Geophys. Space Phys., 17, 1-19, 1979a.
Elcombe,M. M., andJ. R. Hulston,Calculationof sulphurisotope Kieffer, S. W., Thermodynamicsand lattice vibrationsof minerals,
fractionationbetweensphaleriteand galena4singlattice dynam- 2, Vibrational characteristicsof silicates,Rev. Geophys. Space
ics, Earth Planet. $ci. Lett., 28, 172-180, 1975. Phys., 17, 20-34, 1979b.
Epstein,S., The variationsof the O•8/O•6ratioin natureandsome Kieffer, S. W., Thermodynamics and lattice vibrations of minerals,
geologic applications, in Researches in Geochemistry, edited by 3, Lattice dynamics and an approximation for minerals with
p. H. Abelson, pp. 217-240, John Wiley, New York, 1959. application to simple substancesand framework silicates, Rev.
Essene, E., A. L. Boettcher, and G. A. Furst, Indirect measure- Geophys. Space Phys., 17, 35-59, 1979c.
ments of AG for quartz + corundum = kyanite (abstract), Eos Kieffer, S. W., Thermodynamicsand lattice vibrations of minerals,
Trans. AGU, 53, 554, 1972. 4, Application to chain and sheet silicatesand orthosilicates,Rev.
Friedman, I., and J. R. O'Neil, Compilation of stable isotope Geophys. Space Phys., 18, 862-886, 1980.
fractionation factors of geochemical interest, U.S. Geol. $urv. Kieffer, S. W., P. P. Phakey, and J. L. Christie, Shockprocessesin
Prof. Pap., 440-KK, KKI-KK12, 1977. porousquartzite: Transmissionelectron microscopeobservations
Gaffney, E., Specific heat models of ices and clathrate hydrates, and theory, Contrib. Mineral. Petrol., 59, 41-93, 1976.
paper presented at 6th International Symposium on the Physics Kitahara, S., and G. C. Kennedy, The quartz-coesitetransition, J.
andChemistryof Ice, Am. Phys.Soc., Univ. of Mo., Rolla,Mo., Geophys. Res., 69, 5395-5400, 1964.
Aug. 2-6, 1982. Knopoff, L., and J. N. Shapiro, Commentson the interrelationships
Griggs, D. R., and G. C. Kennedy, A simple apparatusfor high between Griineisen'sparameter and shock and isothermalequa-
pressures and temperatures, Am. J. $ci., 254, 72, 1956. tions of state, J. Geophys. Res., 74, 1439-1450, 1969.
Griineisen, E., Theorie des festen ZustandeeinatomigerElemente, Kovach, J. J., A. L. Hiser, and C. Karr, Jr., Far-infrared spectros-
Ann. Phys. Berlin, 39, 257-306, 1912. copy of minerals, in Infrared and Raman Spectroscopyof Lunar
Griineisen, E., Die Zustandsfinderungen des festen K•rpers, and Terrestrial Minerals, edited by C. Karr, Jr., pp. 231-254,
Handb. Phys., 10, 1, 1926. Academic, New York, 1975.
Hays, J. F., and P.M. Bell, Albite-jadeite-quartz equilibrium: A Kyser, T. K., The temperature dependence of oxygen isotope
hydrostaticdetermination,YearBook CarnegieInst. Washington, distributions and the origin of basalts and ultramafic nodules,
72, 706-708, 1973. Abstr. Programs, Geol. $oc. Am., 11, 462, 1979.
Hazen, R. M., and L. W. Finger, Bulk modulus-volumerelationship Lambert, S., Stable isotope studiesof some active hydrothermal
for'cation-anion polyhedra, J. Geophys. Res., 84, 6723-6728, systems, Ph.D. thesis, Geol. Sci. Div., Calif. Inst. of Technol.,
1979. Pasadena, Calif., 1975.
 KIEFFER:THERMODYNAMICS
ANDLATTICEVIBRATIONS
OF MINERALS,
5 849

Lazarev,A. N., VibrationalSpectraand Structureof Silicates, Shaner,J. W., Griineisen,/of rutile(TiO2),Phys.Rev.B., 7, 5008-
translated
fromRussian,302pp., Consultants
Bureau,New York, 5010, 1973.
1972. Slater,J. C., Introduction
to ChemicalPhysics,521pp., McGraw-
Ligenza,J. R., andW. G. Spitzer,The mechanisms
for silicon Hill, New York, 1939.
oxidationin steamand oxygen,J. Phys. Chem.Solids,14, 131- Stem,M., W. Spindel, andE. U. Monse,Temperature dependences
136, 1960. of isotopeeffects,J. Chem.Phys.,48, 2908-2919,1968.
Lippincott,E. R., A. V. Valkenburg,C. E. Weir, and E. N. Stishov,S. M., Equilibriumbetweencoesiteand the rutile-like
Bunting,
Infraredstudiesof polymorphs
of silicaandgermana,
J. modificationof silica, Dokl. Akad. Nauk SSSR, 148, 153-155,
Res. Natl. Bur. Stand., 61, 61-70, 1958. 1964.
Lord,R. C., andJ. C. Morrow,Calculation of Takahashi,T., Factorsinfluencingpressurein multi-anvildevices,
of theheatcapacity
a-quartzandvitreoussilicafromspectroscopic data,J. Chem. in High-Pressure
Measurement, editedby A. A. GiardiniandE.
Phys., 26, 230-232, 1957. C. Lloyd, p. 240, Butterworth,Woburn,Mass., 1963.
MacDonald,G. J. F., Quartz-coesite stabilityrelationsat high Taylor, H. P., The application of oxygenand hydrogen isotope
temperatures andpressures,
Am. J. Sci.,254,713, 1956. studiesto problemsof hyrothermal alterationandoredeposition,
Moenke,H. H. W., Silica,thethree-dimensional borosili- Econ. Geol., 69, 843-883, 1974.
silicates,
cates,andberylliumsilicates,in TheInfraredSpectraof Miner- Taylor,H. P., andR. G. Coleman, O•8/O•6 ratiosof coexisting
als., Mineral. Soc.Monogr.,vol. 4, editedby V. C. Farmer,pp. mineralsin glaucophane-bearing metamorphic rocks,Geol.Soc.
365-382,MineralogicalSociety,London, 1974. Am. Bull., 79, 1727-1756, 1968.
Mueller,R. F., and S. K. Saxena,ChemicalPetrology,394 pp., Taylor,H. P. andS. Epstein,
Relation
between
O•8/O
16ratioin
Springer-Verlag,New York, 1977. coexisting
mineralsof igneousandmetamorphic rocks,2, Appli-
Mulargia,F., Reply,J. Geophys.
Res.,85, 1513-1514,1980. cationsto petrologicproblems,Geol. Soc. Am. Bull., 73, 675,
Mulargia,
F., andE. Boschi, of theMie-Grfinei- 1962.
Thegeneralization
senequation J. R. Astron.Soc.,55, 263-267. Todd,S.S., Heat capacities
of state,Geophys. at low temperatures
andentropies
at
1978. 298.16K of andalusite,kyanite and sillimanite,J. Am. Chem.
Navrotsky,
A., Lowermantlephasetransitions
maygenerally
have Soc., 72, 4742-4743, 1950.
negativepressure-temperature
slopes,Geophys.Res. Lett., 7, Urey,H. C., Thethermodynamicproperties
of isotopicsubstances,
709-711, 1980. J. Chem. Soc., 1, 562, 1947.
Newton,R. C., and J. V. Smith,Investigations
concerning
the Wallace,D.C., Thermodynamics
of Crystals,p. 484,JohnWiley,
breakdownof albiteat depthin the earth,J. Geol., 75, 268-286, New York, 1972.
1967. Weaver,J. S., D. W. Chipman,and T. Takahashi,Comparison
Openshaw,
R., Thelowtemperature
heatcapacities
of analbite,
low betweenthermochemicaland phasestabilitydata for the quartz-
albite,microcline,andsanidine,
Ph.D. thesis,312pp., Princeton coesite-stishovitetransformations,Am. Mineral., 64, 604-614,
Univ., Princeton, N.J., 1974. 1979.
Ostrovsky,
I. A., On somesources of errorsin phase-equilibriaWhite,W. B., The carbonate
minerals,in TheInfraredSpectraof
investigations
at ultra-highpressure;
phasediagramof silica, Minerals,Mineral. Soc.Monogr.,vol. 4, editedby V. C. Farmer,
Geol. J., 5, 321-328, 1967. p. 227, MineralogicalSociety,London,1974.
Richardson,S. W., M. C. Gilbert,andP.M. Bell, Experimental White, W. B., Structuralinterpretationof lunar and terrestrial
determinationof kyanite-andalusite and andalusite-sillimanite mineralsby Ramanspectroscopy,in InfraredandRamanSpec-
equilibria;
the aluminumsilicatetriplepoint,Am. J. Sci., 267, troscopyof Lunarand TerrestrialMinerals,editedby C. Karr,
259-272, 1969. Jr., pp. 325-358,Academic,New York, 1975.
Robie,R. A., andD. R. Waldbaum,
Thermodynamic
properties
of Yagi,T., andS. Akimoto,Directdetermination
of coesite-stishovite
minerals and related substancesat 298.15øK (25.0 C) and one transitionby in situ x-ray measurements,
Tectonophysics,
35,
atmosphere
(1.013 bars) pressureand at highertemperatures, 259-270, 1976.
U.S. Geol. Surv. Bull., 1259, 256 pp., 1968. Zen, E., Gibbsfree energy,enthalpy,and entropyof ten rock-
andJ. R. Fisher,Thermodynamic formingminerals:Calculations,
Robie,R. A., B. S. Hemingway, discrepancies,
implications,
Am.
properties
of minerals
andrelatedsubstances
at 298.15K and1 Mineral., 57, 524-553, 1972.
bar(105pascals)
pressure
andat highertemperatures,
U.S.Geol.
Surv. Bull., 1452, 456 pp., 1978.
Sakai,H., and H. Honma,Oxygenisotopechemistryin rock-
formingprocesses (in Japanese),
Geol.Soc.Jpn.,5, 9-26, 1969. (Received January 2, 1980'
Schuele,D. E., andC. S. Smith,Low temperature
thermalexpan- revised July 20, 1982;
sionof RbI, J. Phys.Chem.Solids,25, 801-814, 1964. accepted July 27, 1982.)