You are on page 1of 23

Canadian Mineralogist

Yol.29, pp. 833-855 (1991)

THERMOBAROMETRY CALCULATIONS:
USINGMULTI-EOUILIBRIUM
A NEW TECHNIOUE,WITH PETROLOGICAL
APPLICATIONS-

ROBERT G. BERMAN
Geological Survey of Canada, 601 Booth Street,
Ottowa, Ontario KIA 0E8

ABSTRACT impliquant ces pdles. Les rsultats sont present6s


graphiquementet interpr6tescomme 6tant les meilleurs
estims,tels que d6termin6sp:u une moyennepond6r6e
An "internally consistent method" (Powell 1985) for
desrdsultatsgraphiques.Plusieursexemplesmontrent que
performing thermobarometriccalculationsis describedin
la convergencede tous les 6quilibres dans un secteur
which all possibleequilibria implied by a given mineral unique de P et T concorde bien avec les indicateurs
assemblageare computed using an internally consistent texturauxet chimiquesde l'6quilibre. La techniquefacilite
set of thermodynamicdata for end membersand mixing donc l'6valuation de l'6tat d'6quilibre pour un 6chantillon
propeilies. Resultsare presentedgraphically and as best
dansun systCmeoil lespropri6t6sthermodynamiquessont
estimates determined by weighted averaging of the
bien d6crites. Des segmentsde trac6s d'6volution en
graphicalresults,Severalexamplesillustrate that conver-
de P et T peuventainsi Otrereconstruitspour des
genceof all equilibria in a single P-T region correlates termes
chantillonsqui manifestentdes texturesde ddsquilibre
well with textural and chemicalindicationsof equilibrium,
mais qui r6ussissentquand m8me ce test de I'existence
suggestingthat this technique makes it possible to
the stateof equilibration of samplesfor d'un 6quilibre local. L'exactitude desdonn6esthermody-
reasonablyassess
namiques, particuferemenl pour les solutions solides'
systemsin which the thermodynamicpropertiesappearto
be well described,One implication is that portions of P-T limite prdsentement I'applicabilit6 gdn6rale de cette
ir mesure
paths may be reconstructed from rocks that display technique;son applicabilit6ira en s'amdliorant
pass que ces propri6t6s deviennent mieux connues. Cette
disequilibrium textures but that this test for local
processusde I'affinement de donnes:
equilibrium. The generalapplicability of this techniqueis approchefacilite le
phasesmin6ralessont le plus
presentlylimited by the accuracyof thermodynamicdata, l) elle indique quellesdes
particularlyfor solid solutions,but shouldexpandasthese incompatiblesdanschaquecascalcul,et ainsi en mesure
properties are refined in the future. This method d'6tre affin6es; 2) elle illustre la sensibilitddes r6sultats
offre ainsi
facilitatesthe refinementprocessby (1) highlightingwhich au choix des diffdrents modblesde solution et
avec
minerals are most incompatible in each calculation and la chanced'6tablir une corrdlationde cesdiffdrences
permet la
thus may be in need of refinement, (2) illustrating the les observationsp6trographiques,et 3) elle
qui peuvent i
sensitivity of results to different solution-modelsand s6lectiond'dchantillons "bien 6quilibr6s"
leur tour servir dans un affinement des propri6ts
offering the chance to correlate these differenceswith
petrographicobservations,and (3) providing a meansto thermodynamiquesen I'absencede donn6esexprimen-
select "well-equilibrated" samplesthat can be used to tales.
refine thermodynamic properties in the absence of
appropriateexperimentaldata. (Traduit par la R6daction)

Keywords: thermobarometry, metamorphic, pressure, Mots-clds;thermobarom6trie,m6tamorphique,pression,


temperature,P-T paths, thermodynamicproperties, temp6rature,6volution P-T, donnesthermodynami-
software. ques,logiciel.

SoruNaarns INTRODUCTION

Ce travail porte sur une m6thode d compatibilit One of the primary goals of petrology is to
interne (Powell 1985)d'effectuer descalculsthermobaro- quantify accurately petrogenetic conditions and
m6triques;tous les 6quilibrespossiblesimpliqu6s par un relate them to geological processes. For metamor-
assemblagede min6raux sont calcul6sau moyen d'une phic petrologists, a great deal of attention has been
banque de donn6esthermodynamiquesd compatibilit6 focused on determination of peak conditions of
interne pour les p6les et les propri6t6s de m6lange metamorphism, and much experimental and
theoretical work has been expended on the
"Geological Survey of Canada contribution number important task of calibration of equilibria that can
19091. be used for the purpose of thermobarometry. Spear
833
834 THE CANADIAN MINERALOCIST

(1989) provided a recent comprehensivereview. pl = pP = pf =.- (l)


Most recently, petrologistshave been increasingly
adopting a more dynamic perspective,in which the For a balanced chemical reaction among ,? end
focus is on the changingP-T conditionsthat a rock membersof each phase(1.e.,phasecomponents)
experiencedover time, as it is this information that (f;), givenby:
is important in formulating models of crustal
evolution linking metamorphismand plutonism to
tectonic processes(England & Richardson 1977, Jl,'d,
. tYl Q)
Ll
England & Thompson 1984).Significant advances i
have beenmade recentlyin relating compositional
zonation in mineralsto pressure(P) - temperature where zi is the stoichiometricreaction coefficient,
(T) changes(Selverstone & Spear1985,Spear1989); the thermodynamicrelation that must be satisfied
these calculations rely on traditional ther- for chemicalequilibrium is:
mobarometric techniques (described below) to
provide the absolute P-T conditions. For high- A!c' = -RTlnK (3)
grade rocks that do not preserve compositional
zonation, other methodsof determiningP-T paths A,G- is the standard state Gibbs free energy of
are required. reaction at P and T, R is the gas constant, and K
As one consequence of the recognitionthat rocks is the equilibrium constant,given by:
may trap some record of their path of changing
P-T conditions, petrologistsmust apply the most n
stringent tests of equilibrium in order to obtain ..
* = T-T ,,,
Ilui, @)
reliable thermobarometric data. Most commonly, i
petrographic observations,data on chemical par-
titioning, and intra- and intersample consistency with a1equalto the activity of phasecomponenti.
are usedto arguefor the attainmentof equilibrium The Gibbs free energy of reaction is computed
(e.9., Zen 1963,Vernon 1976,1977,Spear 1989). from:
This paperpresentsa new techniquefor performing
n
thermobarometriccalculationsin which the equi f-!
libration state of a sample may be assessedby A .' C { = | , v , L ' 'C ! (5)
H
comparisonof the positionsof all equilibria implied i
by a given assemblage.Although limitations are
imposedby the dependencyof resultson the quality whereAiG', the Gibbs free energyof formation of
of thermodynamicdata input, examplespresented any phasecomponentat P and T, is given by:
below suggest that these data are of sufficient
quality in some mineral systemsto allow this test ArCn = AfHo +
to be applied usefully; some implications for
determining portions of P-T paths are explored. TTP
For other systems,involving minerals with less lcpar - rlso + Jf1cpzr)drl
+ fvdp (6)
well-definedthermodlmamicpropertiesat present, J ' J
298 298 l
this techniquecan assistin refining theseproperties
by (a) providing a method to select "well-equi-
AlHo and So are the standard state enthalpy of
librated" samples that can be used to place
formation and entropy at I bar and 298 K, Cp is
constraintson thermodynamicproperties that are
poorly constrainedby experiment,(b) highlighting the heat capacity, and V the volume.
Substitutionof eqns(5) and (6) into (3), omitting
those mineral componentsthat are most incom-
patible in each set of calculations and whose Cp and volume-derivativeterms, yields the stand-
ard form for a geothermometeror geobarometer
thermodynamic properties may be in need of
(A, refers to a reaction property):
refinement, and (c) helping to evaluate solution
modelsof mineralsthrough comparisonsof results (arHo + ArVo(P-l))
obtained with alternativemodels. T_ (7a)
(A.So - RlnK)

THERMoDYNAMIC RELATIoNS (-RTlnK - ArHo + TArSo)


o _ t1r
vJJ
ArVo
For a system at chemical equilibrium, the
chemicalpotentialsof all systemcomponents(e.g., In a systemwith c systemcomponentsand n end
i) are equal in all phases(A, B, C): membersof the phases,the maximum number of
THERMOBAROMETRYUSING MULTI.EQUILIBRIUM CALCULATIONS 835

possible reactions that can be used as ther- to assess this assumption, support for ther-
mobarometersis given by: mobarometric results must come from intra- and
intersampleconsistencyand independentgeological
n!/l(c+l)!(r-(c+l))!l (8) evidence.

Owing to compositionaldegeneracies, the actual 3. Ambiguity: A final problem is that of deciding


which formulation of each particular ther-
number of unique reactionsmay be some number
less. Of these, there are only n - c independent mobarometer one will use. The garnet - biotite
reactioris.Any of the other possiblereactionscan equilibrium, for example, has at least seven
be generated from linear combinations of the different calibrations(Spear1989,p. l3)' among
independentreactions, although the set of inde- which it is difficult for the lay-petrologistand the
pendentreactionsis not unique. expert to decideobjectively. This situation has led
miny geologists to use several, many, or all
different formulations, and rely on other geological
PNOSTTN4
Rgvmw oF THETHERMOBAROMETRY evidence, intersample consistency, or simply
geologicalintuition (!) to selectpreferred results.
The most common method of thermobarometric In many cases, the process of having to select
analysis[seeSpear(1989)for a review] is through among the various calibrationsdefeatsthe original
application of a minimum of two independently goal of obtaining a quantitative or independent
calibrated equations (7), whether they are P-T estimate.
calibrated directly using experimental results or Powell (1985) referred to the traditional ther-
indirectly from empirical observations of lnK mobarometricapproachas the "directly calibrated
variations in natural assemblages.Ideally, these method", and introduced an improved "internally
equilibria have slopes that are independent of consistentmethod" that hasbeenfurther developed
pressure (geothermometer) and temperature by Powell& Holland (1988).With this method,the
(geobarometer),so that uncertaintiesin determina- consistencyproblem is avoided by computing an
tion of either P or T are not translated into average pressure for a given sample from an
uncertaintiesin the other variable. This technique independent set of equilibria, all of which are
introducesthree main problems: computedusing one set of thermodynamicdata. A
discussionof the techniqueof derivation of the
1. Internal consistency:A major problem with the data is beyond the scopeof this paper (see,for
classicaltechniqueis the lack of internal consistency example, Berman et al. 1986, Holland & Powell
in the calculations. As one reflection of this 1990),but the calibration of thermodynamicdata
problem, three separately calibrated equilibria, will, in the ideal case, account for each of the
only two of which are independent,will in general different sets of experiments and constraints
not intersect at the invariant point required by imposed by natural assemblagesused in the
thermodynamicconsistency.An important conse- different calibrationsof eachthermobarometer.An
quence of this inconsistencyis that effects of additional advantageof the "internally consistent
disequilibrium may not be recognized from the method" is that new experimental data for
thermobarometric results, as discussedbelow. In thermobarometerA, for example,have the poten-
addition, inspection of equation (7) indicatesthat tial of improving thermobarometric results using
two independently calibrated thermobaromelers other equilibria if the experimentsare usedto revise
will in generalimply somethingdifferent about the the thermodynamicdata-base.An ambiguity prob-
thermodynamicproperties(both standardstateand Iem similar to that described above will arise,
solution) of any phasesor phasecomponentsthat however,when alternativesetsof standardstateor
participate in both equilibria. Since both sets of solution data are available. In this case, the
propertiesfor a given phasecannot be correct, the comparisonscan help to elucidate the effects of
P and T defined through intersection of both different data-setsor models, thus leading to the
equilibria also cannot be correct in detail. continuous improvement of these data and the
thermobarometricapplicationsutilizing them.
2. Verification: The most severeproblem is that
thermobarometrybasedon the intersectionof two PROPOSED METUON FOR THERMOBAROMETRY
equilibria offers no way to checkthe validity of the
results, in particular the validity of the critical
assumption that all minerals that comprise ther- General description
mobarometers equilibrated at the same pressure
and temperature (here referred to as the "equ! In order to address the problems described
librium assumption"). Without a quantitative way above, this paper presentsan alternative "internally
836 THE CANADIAN MINERALOGIST

consistent" thermobarometric technique to that circular reasoning,it being impossibleto conclude


describedby Powell (1985)and Powell & Holland in any one calculation whether, for example,
(1988). With the presentmethod, referred to as inaccurate thermodynamic data, compositional
TWEEQU (ThermobarometryWirh Estimation of data, or disequilibriummay havecauseddiscrepant
EQUilibration state), the positions of all possible results.Someof the examplesdiscussed in the next
equilibria implied by a given mineral assemblage section, however, illustrate that, when results for
are computed in P - T - X(CO,) - activity space severalor many rocks are comparedto eachother
by the TWQ program, plotted for visual inspection and to pelrographic observations, these three
and interpretation, and utilized to compute an sources of potential error can commonly be
averagepressureand temperature(or other inten- discriminated with some degree of confidence.
sive variable) simultaneouslyfrom all intersections Even where the source of error cannot be
(INTERSX program). The TWEEQU calculations, pinpointed, it is important to recognizethat one or
described in more detail in Appendix I, are more of the basic assumptionshave been com-
performed utilizing a set of internally consistent promised,and weight the thermobarometricresults
thermodynamicdata for end membersand solution accordingly. As further experimental work and
propertiesso that thermobarometricconsistencyis analysis of data on natural assemblageslead to
guaranteed. If three conditions are met: (l) improved thermodynamic data in the future, one
thermodynamicdata areperfect,(2) compositional can expect to be able to resolve the source of
data are perfect, and (3) all minerals last equi- discrepant thermobarometric results with greater
librated at the samepressureand temperature,all confidence.It should also be noted that, because
equilibria will intersectat a single P-T point (Fig. of its dependenceon thermodynamic data, this
l), or, statedanotherway, a singleP and T satisfies techniquemust to someextentcontinueto rely on,
eqn. 3 for all equilibria. This is one consequence and indeed can be complementedby, individually
of the usual definition of equilibrium in terms of calibrated thermobarometers that may involve
the equivalenceof chemicalpotentialsof all system mineralsor accountfor data not incorporatedin
components(eqn. 1). Significant deviations from the thermodynamiccalibrations.
this ideal caselead one to question the validity of
the resultsand analyzethe geological(and possibly A major difference between the technique
describedhereand that of Powell& Holland (1988)
thermodynamic) input data with the aim of
pres,iuresbased
discoveringwhich of thesethreeassumptionsis in is that the latter calculatesaverage
error. There is unfortunately much room for on on independentset of equilibria, rather than
pressuresand temperaturesfrom a// equilibria. As
different results are produced with each different
independent set that forms the basis of the
calculations,the choice of independentset affects
the final numerical result. The equilibria that are
vt best suited to form the independentset are those
Perfect Result with the smallestuncertaintiesand which are least
h
-thermodynamic data sensitiveto errors in input data (those with the
cg
-compositional data largest {S, A,V, and smallest lnK). Because
-equilibrium uncertainties in activities comprise the largest
sourceof error (Powell& Holland 1988,Holland
(D & Powell 1990,Kohn & Spearl99l) and because
tr
these errors are magnified to different extentsin
eachequilibrium by virtue of the different reaction-
(t) coefficients (eqn. 4), it is difficult to derive a
(a generalstrategyfor selectionof the optimal set of
0) independentequilibria. Moreover, once considera-
h tion is given to errors caused by geological
A processesthat may have operated to perturb the
tst
equilibrium composition of any phase (e.g-,
r00
re-equilibration during cooling, interaction with
200 300 400 500 600 ^ ?00
fluids), an optimal set becomes impossible to
Temperature ("C) predict a priori. This can be illustrated by
Frc. 1. P-T diagram showing ideal thermobarometric consideringa hypothetical ideal casein which all
resultsin a casewhere the thermodynamicdata and equilibria intersectat a singlepoint, P,-.-T,,u. (Fig.
compositional data are perfect, and all minerals l). In this example construed with the equilibria
equilibratedat the sameP-T conditions. listed in Table I, there are three independent
THERMOBAROMETRYUSING MULTI-EQUILIBRIUM CALCULATIONS 837

TABLE 1. TWEEQU RESIJLTS FOR CALC-SIUCATE RP13 (IIOLLAND & POWELL 1990)
Mineralogy: Chl-Bt-Ms{rt-Pt-Ep-Cc-Qtz
Assemblageusedin TWEEQU calculations:Chl-Bt-MsCrt-Pl-EpQu
kevious P-T Esults: P=7.5 t 0.5 kbar @ assumed520 oC (Powell & Holland (1988)
P=7.2 I 0.3 kbar @ assumed530 oC Stolland & Powell (190)
Avemged TWEEQU results$: P=7.2 t 0.3 kbar / T=533 * 15 oC (3 indePendentreactions)
Outliss: Equilibrria 1,9, l0 (end-mernbenin comrnon:annite' almandine)
Equilib,rium 6 controls P-T detsrmination (excluding annite)
AveragedTWEEQU resuld: k 6.8 * 0.6 kbar / T=467 t 32 oC (3 indepondentrcactions)
Outlen: Equilibrium 9 (containsreliability level 3 end-memben: annite, clinozoisite)
Equilibria plottod in Figue 3 4s 4v
1: Alm + Phl = Prp + Ann -73.7 -0.31
2:5 Prp + 24Czo + 3 Qtz = 33 An + 3 Cln + 5 Grs -842.7 -67.23
3:Prp+Ms+Gn=3An+Phl -tt9.2 -7.26
4: 3 Qtz + 5 Aln + 24 Czo + 5 Phl = 5 Grs + 3 Cln + 33 An + 5 Ann -911.1 -68.77
5:Ms+Gn+Alrn=Ann+3An -132.9 -7.57
6: 6Prp + 1l Ms + 16 Grs +3 Cln = 3 Qtz+ 24Czo + 11Phl -469.0 -12.68
7: SPhl + 24 &o+ 3 Qtz = 18 An + 3 Cln + 10 GIs + 5 Ms -%t6.5 -30.91
8:3Qv,+?ACm+5Ms+10Prp=5Prrl+3Cln+48An -1438'9 -103'55
9: llMs+16Gn+3Cln+11 Alm=3Qtz+ 1lAnn+24Czo+5Prp -619.6 -16.08
1 0 :l 1 M s + 1 6 G r s + 3 C l n + 6 A h n = 3 Q t z + 6 A n n + 2 4 C z o + 5 P h l - 5 5 1 . 1 ' 1 4 . 5 3
11:3Qtz+5Aln+24Czo+5Ms+5Prp=3Cln+48Ar+5Ann -1507'3 -105'09
1 2 : 3 Q t z + 1 0 A l m + 2 4 C z o + 5 M s + 5 P h l= 3 C l n + 4 8 A n + l 0 A n n -1575.8 -106.64
$ Resultsusing Holland & Powell (1990) themodynamic data and activities; # Resuhsusing
thermodynamicdata of Berman (1988; 1990) with ideal mixing in biotite; t for TWEEQU results
is 1o imprecisiou A$ in J.Kil 4V in J'barl: abheviuions from Kretz (1983); Ch =
Clinochlore

reactionsamongten phasecomponents.Uncertain- correlation betweenW and the equilibria that most


ties in activities, whether due to analytical errors frequentlycompriseone of the three closestto Ptrue.
or retrograded mineral compositions, have been Thus, although there is no one set of three
simulated using random perturbations between 0 equilibria that is best suitedto retrieval of the true
arLd *.2t10 in the mole fractions of cations that p-t in all cases,the good correlationof decreasing
contribute to the activities of each solid-solution difference in pressureaveraged for all 500 per-
component.For each of 500 sets of perturbed turbed samples(Ap,uu) with increasingW meansthat
compositions, all equilibria were computed and the Ap," are roughly equal for independentsets
then processedwith the INTERSX program to formed-from any of the sevenequilibria (Fig. 2)
determine average pressuresas well as the dif- that are least sensitiveto compositional perturba-
ference(Ap) betweenPr,o"and the position of each tions.
equilibrium 4t Tt,u". Inspection of equation (7b) The simulation results indicate that Ap.uucom-
indicatesthat pressuresshould vary leastfrom P,,u. puted from the weighted averageof all equilibria
for those equilibria that have the smallestchanges (O.OStbar) is about half the magnitude of Aq,uv
in lnK and the largestchangesin A,V. Accordingly' basedon an unweightedaverageof all equilibria;
averagepressureswere computed for each sample (0.12 kbar), but is equal to Ap,uubased on the
using a weighting factor, W, for each equilibrium averageof the set of independentequilibria (#3, 5'
given by 8 of Table 1) with the smallest A".uu. For
A,v (9) nonoptimum independent sets formed from the
w =
lnK' sevenequilibria least sensitiveto input errors (Fig.
2\, Ao * is not significantly worse(0.07 verszs0.05
where K' differs from the equilibrium constant in kbarj,'-but the Ap calculatedfor each of the 500
that the activities of equation 4 are raised to the perturbedsamplesis higherthan Ap calculatedfrom
absolute values of the reaction coefficients. This the weighted averageof all equilibria for about
term has the effect of giving the most weight to 70t/o of the samples.The weighted averagebased
those equilibria that have the smallest reaction- on all equilibria, therefore?appearsto provide the
coefficientsand site multiplicities and that involve most stable solution unless the optimum set of
minerals closestto end-membercompositions. independentequilibria can be chosenunambiguous-
The results of a Monte Carlo analysis of this ly. For this reason,the techniqueproposedhere is
system (Fig. 2) show that there is an imprecise to compute the average P-T using a// possible
838 THE CANADIAN MINERALOGIST

250
MonteCarloAnalvsis

200

# tinesequilibrium
is
150 I of 3 closest
to P,ro,

100

50

5 t0 l5 io t{ 30
Ar V/lnK'
Frc. 2. Resultsof a Monte Carlo simulationof calculationsof averagepressure
using mineral compositions perturbed from perfect case (Fig. l). Average
deviation in pressure(Ap.uu;squares)is well correlatedwith weighting factor
(eqn. 9) for each equilibrium. A less precisecorrelation is observedwith the
number of timesa given reactionis one of the threeclosestto the "true" pressure
(triangles;seetext).

equilibria implied by a given mineral assemblage. contain phase componentsfor which one of the
The INTERSX program computesaveragesand th.e three basic assumptions have been most com-
amount of scatter (expressed as a standard promised. Indeed, one of the main reasons for
deviation) of the diagram variables from all computing all equilibria, as opposed to an
intersections(including indifferent ones)among all independentset, is to determineif any of the phase
equilibria, both stable and metastable. The components used in a particular calculation are
averagesare weightedtoward the best-constrained significantly at odds with thesethree assumptions
intersections by using a weighting factor, and to omit them from further calculations.unless
l6.Wr.\1, wittr reassessment of all input data is permissiveof other
A' Sl 'A' vl combinations of "equilibrium" compositions of
wr = (lo)
lnKi minerals. There are several ',ryaysthe TWEEQU
analysis fulfills this task. The first is by visual
where B is the acute angle of intersectionbetween inspection of graphical thermobarometricresuhs,
equilibriai and j, and K' is definedabove. which is the easiestway of recognizing outlying
It is important to appreciatethat equilibria that equilibria. A second way is with the INTERSX
occur furthest from the true P-T of equilibration program, which repeatsthe averagingcalculations
839
THERMOBAROMETRYUSING MULTI.EQUILIBRIUM CALCU LATIONS

excludingone phasecomponentand then onephase of uncertaintyin comparisonto estimatesbasedon


at a time. Powell & Holland (1988)also usedthis the compatibility of different geobarometers'For
technique to filter the results of their averaging example, Powell & Holland (1988) computed
calculations. INTERSX results are hiehlighted ou.tull uncertainties(2o) of about 3 kbar for the
when the scatterin the resultsdecreases significant- GASP equilibrium:
ly (lessthan one-third) or when the averageis more
than 2o outside of the average, considering all Grst2KY+Qtz=3An
equilibria. Another effectivemethod performed by
the INTERSX program is finding those phase yet calculations of GASP pressures repeatedly
componentsin common to any outlying equilibria' produce results that place rocks with one
operationallydefinedas those equilibria that differ aluminosilicatepolymorph in its appropriate field
from the averageP-T by more than 1.5o. A
of stability and rocks with two aluminosilicates
warning is given when the excluded phase com- thought to be at equilibrium within about 0'5 kbar
ponent or phase componentscommon to .the
of thle appropriate phase-boundary(e.g.' Fig' 13
outlying equilibria have less reliable ther- of Bermin 1990).Such comparisonssuggestthat a
modynamic data, which alerts the user to the more reasonableestimateof overall uncertainty is
possibility that the aberrant resultswere produced generally around 1 kbar (2o), except for- as-
for this reason. Thermodynamic data have been
iemblaeBs with Ab-rich plagioclase, for which
assignedthree levelsof reliability, similar to those activity coefficientsare poorly known. The un^c-er-
proposedby Holland & Powell(1990):(1) datathat
tainties calculatedby Hodges & McKenna (1987)
are well constrained and consistentwith a large and McKenna & Hodges(1988)also appearoverly
body of experimental data, and are expectedto pessimistic,at least in part because.they consider
change little in the future, (2) data that are bnly the uncertaintiesassociatedwith the phase-
constrained by few experiments, among which equilibrium data they used to determine the
some contradictionsmay exist, or for which some position of the GASP equilibrium. Theseuncertain-
estimated thermodynamic parameters are more iies arereducedsubstantiallythrough consideration
uncertain, and (3) data that are considered of other phase-equilibriumand calorimetric data
preliminary becauseof large inconsistencies among
that constiain the thermodynamicpropertiesof the
experimental studies or because their retrieval phasesin this equilibrium. The above estimatesof
involved major assumptionsabout mixing proper-
uncertainty based on barometer compatibility are
ties. supportedby the recent analysisof Kohn^& S-pear
(tSgf), whiih suggestsoverall errors of l'6-2'4
Uncertainties kbar for the GASP equilibrium, with about I kbar
stemming from the uncertarnty in estimated
Although standard deviations computed by the temperature.
INTERSX program give a measureof imprecision The method favoredherefocusseson what is the
that allows an assessmentof the overall consistency largest source of systematic error for most
of the results in terms of the three critical eqnilibtia, that stemming from uncertainlies in
to thermobarometry' it is important solution models (Holland & Powell 1988,Holland
assumptions
to note that these standard deviations do not & Powell 1990, Kohn & Spearl99l). As this type
in results,
representstatistically valid uncertainties. Overall of error producessystematicdifferences
assess their effects are best discovered and explored by
uncertainties are exceedingly difficult to
with
quantitatively becauseerrors in standardstate and performing different sets of calculations
that are
solution data are systematic'dependingon which alternative proposed solution-models
model and data set is adopted' and cannot be compatible with standard state thermodynamic
thesedifferent
accommodatedby standard statistical techniques. properties.Through comparisonof
possibleto gain
Additional problemsare encounteredin the defini- ieti of calculations,it should be
helpto refine
tion of uncertainties in thermodynamic data important insightsthat will eventually
This processwill
retrievedfrom phase-equilibriumdata (seediscus- und imptoue the solution models.
sion of Bermanet al. 1986),which are characterized be further aided by the selection of "well-equi-
by non-Gaussian probability distributions librated" samplesusing the TWEEQU approach'
(Demarest& Haselton 1981, Kolassa 1990).The Within each set of calculationsusing different
"overall" uncertarnties calculated by Powell & solution-models, the random errors related to
Holland (1988)by propagation of systematic errors analytical data are best quantified using Monte
Lieberman &
in thermodynamicdata and random errors due to Carlo techniques, described by
(Powell & Holland 1988' p. Petrakakis (ltgl). A similar technique can be
analyticaluncertainties
derive overall thermobarometric uncer-
177) appearto lead to overly pessimisticestimates applied to
840 THE CANADIAN MINERALOGIST

tainties when uncertaintiesin standard state and Second, uncertainties in thermodynamic data,
solution properties are better defined (Engi et al. particularly activity - composition relationships,
1990). are larger than many expecteddifferencesin T.,
precludingthe generalinterpretation at the present
Assessmentof equilibrium time of thermobarometricdifferencesin terms of
T6 differences.As theseuncertaintiesare reduced,
this type of interpretation may prove feasibleand
Poorly defined uncertaintiesin thermodynamic may allow geological rates of cooling to be
data, particularly solution properties, seriously determined from detailed study of ther-
compromisethe ability to quantitatively assessthe mobarometric differences. Third, and perhaps
state of equilibration of sampleson an individual most convincing, is the empirical observationthat
basis. In addition, becauseeach equilibrium has thermobarometricresultssuggestingattainment of
different uncertainties, and different reaction- equilibrium are well correlated in a variety of
propertiesthat translate into different sensitivities
mineral systemswith textural and chemicalcriteria
to errors in input data, a single limit on pressure for equilibrium. A number of these results are
and temperature for the amount of intersection discussedin detail below.
scatter cannot be establishedas the criterion to
assessequilibration state. Ideally, equilibration
state should be determined by comparison of APPLICATIoNS
calculated scatter in the intersection with thar
stemming solely from calibration and analytical Thermobarometricresultsfor RP13, a calc-sili-
uncertainties.Until the calibration uncertaintiesare caterock consideredby Holland & Powell (1990)
better defined, the strategy recommendedhere is to be "remarkably well equilibrated", offer a good
that confidencecan be gained using the TWEEQU opportunity to illustrate the TWEEQU technique.
techniquewhen the thermodynamicdata passthe This softwarefacilitatescalculationsusingdifferent
first-order test of producing results for specific thermodynamic data-bases;Table I and Figure 3
systemsand phasecomponentswith (a) little scatter show TWEEQU results for this rock using
when classic"equilibrium" textures(e.g., Vernon thermodynamic data and activities of phasecom-
7976,1977,Spear1989)are observed,and (b) good ponents from Holland & Powell (1990). The
consistencyamong different domains of single calculations are based on non-fluid-phase com-
samplesand among samplesthought to have the ponents that are also given in the TWEEQU
same metamorphic and delormational history. data-base.Immediatelyobvious from the inconsis-
Calculationsfor a set of samplesshould define the tency in the position of invariant point A with the
criteria usedto define the "equilibrium" case,and other equilibria shown in Figure 3 is that one of
TWEEQU resultsfor samplesof similar bulk-com- the three basic thermobarometricassumptionshas
position with ambiguoustexturescan be taken with been violated. The sameconclusionis reflected in
some degree of confidence. Examples discussed the largelo scatter(1.5kbar, 65oC)calculatedwith
below indicate that reasonablecriteria for satisfac- the INTERSX program when all equilibria are
tion of the equilibrium assumptioncorrespondto included. This program proceedsto calculatethat
INTERSX 1o standarddeviationsof +0-40"C and threeequilibria(#l, 9, l0 of Table 1) aremorethan
0.2-0.5kbar. l.5o outside of the averageP-T, and that these
It should be noted thar rhe expectationthat all three equilibria have almandine and annite in
equilibria intersect at a single pressure and common. The position of invariant point A is
temperature(Fig. l) is unjustified in view of the controlled by the garnet - biotite Fe-Mg exchange
different closure-temperature (T6) and resistanceto equilibrium, which givestemperaturesmuch lower
retrogression of each equilibrium (Lasaga 1983, than that (520'C) preferred by Powell & Holland
Ghent et al. 1988). For example, equilibria that (1988) on the basis of calcite - dolomite ther-
depend on the Al content of orthopyroxenemay mometry. The INTERSX program focussesatten-
record higher pressuresand temperatures than tion on annite, becauseit has a reliability level of
equilibria that involve only the Fe and Mg end three, suggestingthat the problemsmay be related
members (e.g., Aranovich & Podlesskii 1989) to this phase component. Calculations repeated
becauseof the largerkinetic barriersto re-equilibra- without this phase component give the P-T of
tion of Al distribution. Nevertheless,therearethree invariant point B, but the program warns that an
reasonswhy the kind of equilibrium test described independentP or T estimateis advisedsincethere
aboveshouldbe reasonable.First, closuretempera- is no meansof checkingthis P-T determinedfrom
tures depend on grain size and cooling rate, and the intersectionof only two independentequilibria.
thus may be operationally quite similar for many An important differencewith the THERMOCALC
samplesat common geological rates of cooling. procedures(Powell & Holland 1988) is that the
THERMOBAROMETRYUSING MULTI-EQUILIBRIUM CALCULATIONS 841

10000
Cal-Dol
9000

(t)
Lr 7000
ctt

6000
<t)
r- 5000
(A
Ea 4000
<t)
tr
tr.( 3ooo

RP-13
3 rR - H&P(1990

Temperature ('C)
Frc. 3. TWEEQU resultsfor calc-silicale RP-13 (Holland& Powell 1990)'Box in
lower left corner indicatesthere are three independentequilibria (3 lR)' and
that calculationsweremadeusing the thermodynamicdata of Holland & Powell
(1990). Arrow shows the preferred temperatureof Powell & Holland (1988)
(520"C) based on calcite-dolomite thermometry. I bar shows the 2o average
bracket associatedwith pressure(Holland & Powell 1990).Numbered curves
refer to equilibria identified in Table l. See text for discussionof invariant
pointsA and B. InsetshowsTWEEQU resultsin T-X(CO2)sectionat 7.2kbar,
omitting annite from the calculations.

TWEEQU resultssuggestthal annite is a probable with the TWEEQU thermodynamicdata, depend-


bad actor in RPl3, and shouldnot be includedfor ing on whether nonideal (McMullin et al. l99l) ot
the most accuratethermobarometriccalculations. ideal mixing is assumedfor biotite. This suggests
The TWEEQU calculationsalone do not make it that much of the problem illustrated in Figure 3 is
clear, however, whether it is a bad actor because likely a calibration problem based on different
of compositional data or thermodynamic data. statesof oxidation in retrievedthermodynamicdata
Calculationsusing the TWEEQU thermodynamic for annite and in thesethermobarometricapplica-
data (Berman 1988, 1990, McMullin et al. l99l) tions. In the examplesthat follow, all Fe is treated
and the same activities as given by Holland & asFe2+becausethermodynamicdata wereretrieved
Powell (1990)yield low garnet - biotite tempera- with this sameassumption(Berman 1988,Berman
tures as well (about 430'C at 7 kbar), which 1990. McMullh et ol. l99l). When annite is
suggeststhat the problem lies with the activity of excludedfrom the analysisof RPl3, TWEEQU
annite. Holland & Powell (1990) calculated the calculationsin a T-X(CO) sectionat 7.2 kbar yield
activity of annite taking accountof a wet-chemical a very preciseestimateof X(CO) of 0.18 * 0.03
analysisgiving Fe3+.If all Fe is calculatedas Fd*, (Fig. 3, inset).
the activity of anniteis raised,and a garnet- biotite It is interesting to note that results using the
temperaturebetween480 and 5l0oC is produced TWEEQU thermodynamicdata produce a similar
842 THE CANADIAN MINERALOCIST

discrepancy between the positions of invariant Geothermobarometry


points A and B, but with invariant point A at lower
pressurethan invariant point B. This difference, Much attention has been focussedrecently on
highlighted by the TWEEQU technique(Table l), the calibration of geobarometersthat are com-
is a direct consequenceof the different properties patible with one another. In granulites, for
for clinozoisite in the two data-bases.Holland & example, Newton & Perkins (1982) presented
Powell (1990) recommendedthat clinozoisite is 4 calibrations for the equilibria:
kJmol-r more stable than indicated by the ex-
perimentalresultsof Jenkinset al. (1983),a change Grs + 2Prp + 3 Qtz : 3 An + 6En (El)
that would improve the overall consistencyof the
TWEEQU calculation for RPl3. As suggestedby
Holland & Powell (1990), feedback from other Prp + 2Grs * 3 Qtz = 3 An + 3 Di (E2)
thermobarometricapplicationsis neededto further
evaluatethis change.In the meantime,this example and recommendeda + 2 kbar correctionbe applied
serves as an important reminder that ther- to the Cpx-bearingequilibrium in order 1o obtain
mobarometric calculations with alternative inter- similar pressureswith the Opx-bearingequilibrium.
nally consistentdata-basesoffer one of the best Powell & Holland (1988) showed that their
methods to discover systematicerrors in calibra- calibration of thermodynamic data for these
tion. This step is particularly important for minerals leads to quite reasonableagreementof
verification of the stateof equilibration of samples thesetwo barometersin applicationsto granulites
used for the purpose of refining any ther- from southern India (Harris et al. 1982),assumed
modynamicproperties. to have equilibratedat 750'C. Equilibria El and

12000 FuruaComplex
# c-311.1
(1980)
Coolen
I r000
U)
h
s
10000

<t)
l< 9000
(t)
(t)
ct)
h
8000

7000
2 IR - TWBBQU/Ian91
6000J-
300 350 450 550 600 650 700 750 800
Temperature('C)
Ftc. 4. TWEEQU results in the system CaO-MgO-Al2O3-SiO2 (CMAS) for
first-generationphasesin granuliteC-3 I I . I from the Furua Complex,computed
with the January, 1991TWEEQU thermodynamicdata. Two of five equilibria
shown are independent(2 IR).
THERMOBAROMETRYUSING MULTI.EQUILIBRIUM CALCULATIONS 843

E2 are two of five equilibria that intersect in an suchcalculationsdependon the generallyunknown


invariant point (Fig. 4), the temperatureof which activitiesof the fluid species.
is definedlargely by the position of the equilibrium Coolen (19S0) reported chemical data for
two-pyroxene granulites of the Furua Complex,
texturally distinct
6En+Grs=3Di+Prp ( E 3 ) Tarzania that contain two
p:uageneses,distinguished by first- (Ml) and
second-generation(M2) garnet and pyroxene.For
the
Owing to the very small {V of reaction (0.16 first-generationphases,all equilibria involving
position of this equilibrium is highly assemblage Cpx-Opx-Grt-Pl-Qtz except E3, the
Jbar-r), the
equilibrium with the smallest A,V of the II
uncertain [2o between40 and 50 kbar based on
(1988, Table equilibria (3 independent), intersect in a very tight
calculationsof Powell & Holland their
pressureand temperature
6)1. Nevertheless,Figure 4 servesto illustrate that clusternot only defining
the problem of achieving compatibility of simultaneously, but also adding a degree of
confidence to the result (Fig. 5a, Table 2). This
barometersis in fact a problem of being able to
retrievean accurateestimateof the temperature of confidence is what is lacking from traditional
of the common
equilibration, the temperature at which the thermobarometricmethodsbecause
use of only two equilibria. On the other hand, the
barometers intersect. For some mineral systems,
5a
like that illustrated in Figure 4, which involve clustering of intersections shown in Figure
allows one to conclude that the three basic
extremelysmall A"V or A.S equilibria, the stringen-
to be satis-
cy of this test of thermodynamic data is too thermobarometricassumptionsappear
demandingto be of practical value. However, many fied in this case. This conclusion is further
supported by the similar, but less perfect,results
other systems,in which the most temperature-de-
pendent equilibria are not as sensitiveto errors in produced for different samples from the Furua
and 1
input data (larger valuesof A"V or A,S), do afford Complex,which show scatterup to .r40'C
pressure and kbar in the positions of these intersections, except
a reasonableopportunity to determine
results are
temperaturesimultaneously.Efforts to assurethe those with equilibrium E3. That such
of barometers defined for Fe and Mg obtained from samples of the Furua Complex and
compatibility
with a range in bulk
end members(e.g., Moecheret al. 1988)will always from other granulite terranes
fail at some level until they test the ability of composition (0.49 < X,;,^ <0.65) suggeststhat
thermodynamic data to retrieve an accurate the thermodynamic data are not seriously in error
temperature,as well as a pressure,of equilibration. for these phase components in this compositional
range and that the consistencyin the positions of
The main application of the TWEEQU software
all equilibria results from the attainment of
lies in performing general thermobarometric cal-
chemicalequilibrium among thesephases.Results
culationsin which temperatureas well as pressure
including biotite producea largeamount of scatter,
are determinedsimultaneouslywhereverpossible.
likely attributedto retrogressionof the biotite'
The necessarycriteria are that at least two of the most
modally the least abundant Phase.
equilibria implied by a given mineral assemblage
intersect at a reasonably high angle (optimally Although chemicalequilibrium is impossibleto
90'C), and arenot overly sensitiveto errorsin input prove, a number of tests must be met to be
data. It is not necessaryto havetwo equilibria with consistent with it. One important test is the
the ideal P- and T-independentslopes.One of the counterexamplethat phasesin apparent textural
largest classesof equilibria that are useful for disequilibrium display results with substantial
temperature estimation are those that involve scatter.Figure 5b showsresultsfor the samesample
Fe-Mg exchange, many of which have been as Figure 5a, but for the case where "second-
calibrated independentlyas geothermometers.Be- generation" (Coolen 1980) garnet and ortho-
causethe A,V of theseequilibria is generallymuch pyroxene are included in the calculation
smallerthan thal of other equilibria, calculationof together with "first-generation" clinopyroxene.
thesegeothermometers is an extremelysensitivetest Assuming little error in compositional or ther-
of the quality of the thermodynamicdata, although modynamicdata, the scatterin resultsof +150"C
a test not independentof input mineral composi and 4 kbar illustrates the effects of chemical
tions and the assumption of equilibrium among disequilibrium that are implied by Coolen's (1980)
these minerals. Becauseof this sensitivity, the textural observations.The ability of the TWEEQU
greatestconfidencein resultsis obtainedwhenthere techniqueto producethe range in resultsshown in
are two or more independentequilibria that both Figures5a and 5b is a necessarycriterion to allow
yield a consistenttemperature.Equilibria involving one to discriminate between equilibrium and
fluid specieshave not been used to estimateP-T disequilibrium.Although trivial in this casebecause
conditionsin the examplesdiscussedbelow because of their obvious petrographic differences(Coolen
844 THE CANADIAN MINERALOGIST

3 IR - TWEEQU/Ian91
FuruaComplex
# c-3 1 1 .1
Coolen(1980)
u)
t<
6J
,a

<D
Lr

an
(a)
<L)
tsr

200 300 400 500 600 700


Temperature('C)

FuruaConplex
# c-3 1 1 .1
Coolen(1980)

9n
h
cl'
,a

<D
k

v)
aa) ^
ct)
!r
tr<

3 IR - TWEEQU/Ian91
300 400 500 600 700 800

Temperature('C)
THERMOBAROMETRYUSING MULTI-EQUILIBRIUM CALCULATIONS 845

16000
c
3 IR - TWEEQU/Jan91
r5000

14000

u) r3000
Lr
clt
,a 12000

) 11000 Moecher
et al.
Lr (1e
88 )
(t) 10000
v) (1990)
Berman
)
L 9000
tr<
8000
FuruaConplex
7000 # c- 311.1
Coolen(1980)
6000
500 550 600 650 700 750 800 850 900 950 1000
Temperature('c)
Ftc. 5. a. TWEEQU resultsin the CMFAS system(F: FeO) for first-generation
phasesin granulite C-3ll.l from the Furua Complex, computedwith the
January, 1991 TWEEQU thermodynamic data. Numbered equilibria (3
independent)are listed in Table 2. b. TWEEQU resultsin the CMFAS system
for first-generation clinopyroxene and second-generationgarnet and or-
thopyroxenein granuliteC-311.1from the Furua Complex,computedwith the
January, 1991 TWEEQU thermodynamic data. Numbered equilibria (3
independent)are listed in Table 2. Note the wide scatterin the positions of all
equilibria, indicative of the disequilibrium assemblage.c. TWEEQU resultsin
the CMFAS systemfor first-generation phasesin granuliteC-311.1 from the
Furua Complex. Curves were computed with the January, l99l TWEEQU
thermodynamicdata (including garnet model of Berman 1990)'Open polygons
show the range of intersectionsfor the same equilibria using the garnet
calibrationsof Hodges& Spear(1982)and Moecheret al. (1988).Small filled
polygon shows range of intersectionsfor mixed first- and second-generation
phasescomputed with the garnet calibration of Moecher et 4/. (comparewith
Fig. sb).

1980), the importance of this ability becomes difficult to analyze rigorously because of the
paramount in the most common situation systematicdifferencesamong the data with which
petrologists are faced with, that of ambiguous they are calibrated. The TWEEQU method has
petrographic textures. Particularly critical is the been set up to facilitate comparisonsof the effect
ability to test which mineral compositionsmight be of different solution models on all equilibria'
in equilibrium where zoned minerals are present. Figure 5c compares results for first-generation
A major problem in interpreting results like phasesin the sameFurua Complex rock, using the
those shown in Figures5a and 5b is the sensitivity garnet mixing propertiesof Moecher et ol. (1988)'
of results to thermodynamic data, particularly Hodges& Spear(1982),and Berman (1990,Fig.
alternative solution-models. This sensitivity is 5a). Results with the first two calibrations are
846 THE CANADIAN MINERALOCIST

FOR MAFIC GRANULIIE C.311. I most samplesshowing scatter in intersectionsof


Mirenlogr Opx{bx{n-Pl"QE-Mc-Scp-Hbl-BlApMag"tld
AMblage uled t! TWEEQU calculsions: OpxCpiCn-PIQE
lessthan *15oC and 0.30 kbar (Fig. 6, Table 3).
Odgilal P-T qfffi: F7.8 kbq/T-771836 oC (C@len 1980) Intersample consistencyalso suggeststhat equi-
Avqsge TWBBQU F ults: F10, Lbd r .05 / T-795 r 3 oC (3 indepedenr @rions/
Outli@: Bqdltbtiu l0
librium was maintained in these samples. The
Equilibli! plouid in Fig@s 4-5 ArS
overall accuracyof the thermobarometricresultsis
4V
liQa+DirAl$=Atr+E!+3Fs -28.0 -2.59 supported by the general agreementbetweenthe
2: 3 Qtz icm 12 Ald = 3 An r 6Fe
3: 3 Qtr r 3 Di +4 AIB = 3 An + 12Fs + Itp
-109.0
-1m.5
-7.32
-8.07
calculatedandalusite- sillimanite boundary (Ber-
4:Alm+3Di.Gn{3Fsi3En 25.0 4.45 man 1988)and the distribution of the andalusite-
-53.0 -2.t4
5:QETCB+2E!-Air2Di
6:Alrnr3En.Fp+3Fs -16.5 4.29
zone (light triangles) and sillimanite-zone (bold
7:Qtr+PrprDi-Anr4En "1r.5 -2.30 triangles)samples(Fig. 6).
8: 2Aln r 3 Di -Pqr a Cn + 6Fg E.5 4.74
913 Qa f Prpi2CB - 3 A! + 3 Di -117.5 -6.58 Two samplesdisplaylesspreciseresults(20-40'C
10:Prp+3Di=6En+Cn 41.5 -0.16
+2PlDrGa-3As+6En
and 0.44-0.76 kbar scatter). The results for
t fo TWEEQUBula is 10imp@i$ioqA1sin J.Ki; 4V in J.brli abbpviationsfron sillimanite-bearingsample #1104-l fall largely in
t<Ee 0983) the stability field of kyanite. This sample has the
most sodic plagioclase(Ans.:), strongly suggesting
that activity coefficientsgiven by the Fuhrman &
summarizedby showing the boxesthat enclosethe Lindsley(1988)calibrationare too low for Ab-rich
rangeof all intersections,excludingequilibrium E3. plagioclase. In contrast, plagioclase-independent
Not only is a different averageP-T producedwith pressuresfor this samesamplebasedon equilibrium
each different calibration, but the conclusion of 5 of Table3 (Qtz-Grt-Aluminosilicate-Ms-Bt,1.e.,
equilibrium would not be supported by the SGAM: McMullin et ql. l99l) fall in the sillimanite
calculations utilizing the Moecher et al. (1988) field. The sameexplanationcannotapply to sample
calibration. This calibration incorporatesthe main 674-A, whichcontainsa plagioclase of composition
featuresof those of Ganguly & Saxena(1984)and Ano2. Although slightly more magnesian(within
Anovitz & Essene(1987).As discussedin detail by Xp" = 0.03), the overall similarity of its mineral
Berman(1990)and Berman& Koziol (1991),these chemistryto that of the other samplessuggeststhat
calibrations are incompatible with recent reversed calibration errors should not be significantly
phase-equilibriumdata in the Prp-Alm, Grs-Alm, different for this sample and that the discrepant
and Grs-Prp-Alm systems(Koziol & Bohlen 1990, resultsare likely due to disequilibrium.
Lee & Ganguly 1988, Koziol 1990, Koziol & Pelitic rocks from the MonasheeMountains,
Newton 1989).Similarly,the calibrationof Hodges British Columbia, studied in detail by Journeay
& Spear (1982) is inconsistentwith experimental (1986),haveassemblages characteristicofthe St-Sil
data on the Grs-Alm join (Koziol 1990). It is zones of the Barrovian facies series.Ms-Bt-Grt-
instructiveto note that TWEEQU resultsusing the Pl-Ky-Qtz assemblagesoccur stratigraphically
Moecher et al. (1988)calibration give a very tight about I km below the MonasheeD6collement,a
clusterof intersectionswhen the samecombination major crustal scaleshear-zone.Deformation con-
of first- and second-generationphases is used tinued after the growth of kyanite and garnet
(black box in Fig. 5c) as in the calculationsshown porphyroblasts,but petrographicobservationscon-
in Figure 5b. Clearly, one needs to establish cerning the state of equilibration are ambiguous.
confidence in thermodynamic data before con- Figure 7 summarizes TWEEQU results using
clusions regarding the state of equilibration of a unpublished analytical data for some of these
sample can be regarded with anything but skep- samplesand the sameequilibria shown in Figure 6
ticism. and Table 3. Although all sampleswere collected
Ferry (1980) has described chlorite- to sil- within a l-km area, they record temperaturesfrom
limanite-zonepelites of the Waterville Formation 650 to 950"C. The fact that the bulk composition
that have undergone one metamorphic event of these rocks is very similar to those used to
postdating most of the deformation. The sub- calibrate thermodynamic data for biotite (Mc-
assemblageMs-Bt-Grt-Pl-And/Sil-Qtz was as- Mullin et al. l99l) and to the Waterville Formation
sumed by Ferry (1980)to be in equilibrium when peliteswith plagioclasegreater than An15suggests
petrographicallyobservedwithin a 2.5 cm diameter that the problem is not one related to calibration
area without evidenceof textural disequilibrium, of thermodynamicdata. To the contrary, disequi-
such as inclusions of one mineral in another or librium is suggestedby the wide discrepancy(4-10
mantling relationships(Ferry 1976).Resultsusing kbar) in the positionsof the two main barometers,
the January, l99l TWEEQU thermodynamicdata equilibria 1 (GASP = Crt-Ky-Qtz-Pl) and 5
(Berman1988,Berman 1990,McMullir. et al. l99l) (SGAM) of Table 3. The position of the GASP
show excellentintrasampleconsistencybetweenthe equilibrium, generally close to the kyanite -
six possible equilibria (three independent), with sillimanite boundary (Fig. 7), is broadly consistent
THERMOBAROMETRYUSING MULTI.EQUILIBRIUM CALCULATIONS 847

6000
Waterville (Feny 1980)
Formation
5500

5000
a)
L
cl! 4500
,Q

4000
(l)
k

;3500
an
g)

:3ooo

2500
{ Sittinanite-bearing
( Andalusite-bearing
2000
460 480 500 520 540 560 580 600
Temperature('C)
Frc. 6. Summaryof TWEEQU resultsin the KCMFASH system(K: K2O' H: H2O)
for 18samplesof WatervilleFormation pelite, computedwith the January, 1991
TWEEQU thermodynamic data. Light and dark triangles, defined by the
intersectionsof equilibria l-3 of Table 3, showthe rangeof intersectionsamong
all six equilibria (3 independent)for andalusite-and sillimanite-zonesamples'
respectively.

TABLE 3. TWEEQU RES(r'LTSFOR WATERVILLE FORMATION with the other mineral phasesin theserocks. This
PELITES GERRY I98O)
may be the resultof retrogression of biotite caused
Minonlo$/: Bt+Qtz t Cln, Cn, Pl, St, Crd" And, Sil, Mc, Cc, Ep, Cm, Ap,
Gr, Ilm, Po
by interaction with late fluids introduced through
Assmblage usedin TWEEQU calculadons:Ms-Bt-Gn-Pl-Qtz-Aod/Sil aicess along the MonasheeD6collement.Alterna-
Original P-T csdmate: P= 3.5 * 0.2 kbr / T= 46G550 oC (Ferry I 980) tively, the disequilibrium assemblagemay have
Avemge TWEEQU rcsults (l 8 mples): P= 2.6 - 4.2 kbu /'f = 490-575oC
(3 independentMclions) beenproducedby mechanicalrepositioningduring
post-metamorPhicshearing.
Equilibda ploued in Figms 6, 7, l0 ArS drV
1:2Sil+Qtz+Gn=3An -lll.1 -5.45 Pigage (1976) eavedetailed compositional data
2: Prp + Ms = Phl + Qtz + 2 Sil -8.2 -1.81 for St-Sit zone pelitesof the Settler Schist,British
-t3.7
3: Phl + Alm = Ann + Prp
-132.9
-0.31
Columbia, that contain the assemblageGrt-Ms-
4:Ms+Gn+Alm=Ann+3An
5: Alm + Ms = 2 Sil + Qtz + Ann "21.8 Bt-Pl-Qtz-Rt-Ilm + St * KvlSil. Pigage(1976)
6:PrD+Ms+Cs=3An+Phl -119.2 -7.26 reported that syn- to post-tectonic metamorphic
from Krctz (1983)
A"Sin J.K't; d"V in J.brl; abbrcviations reirystallization produced sharp crystal margins
with little alteration,but that "textural evidencefor
equilibrium is inconclusive". Chemicalhomogene-
with the occurrence of these kyanite-bearing ity on a thin-sectionscaleand uniform distribution-
samplesjust downgradeof the sillimanitezone.The coefficientsbetweensamplesare suggestive,how-
likely explanation for the discrepant SGAM ever, of attainment of equilibrium. TWEEQU
pressures and the extreme spread in Grt-Bt calculationsbasedon 2l equilibria (4 independent)
temperaturesis that biotite is out of equilibrium support this conclusion(Fig. 8, Table 4). For each
848 THE CANADIAN MINERALOGIST

14000
Monashee
Mountains
Pelites
tr GASPpre$sure
12000
o SGAMpressurc

at)
t<
10000
@
cl!

8000
<D
t<
6000
at)
(t)
q)
t<
4000

2000

400 500 600 700 800 900 1000


Temperature('C)
Ftc. 7. Summaryof TWEEQU resullsin the KCMFASH systemfor Monashee
Mountainpelites,computedwith the January,l99l TWEEQU thermodynamic
data. The wide rangein Grt-Bt temperaturesand extremediscrepancybetween
pressurescomputed with the GASP (squares)and SGAM (triangles)equilibria
suggestdisequilibrium. Note the differenceof scalewith Figure 6.

of six samplesthat contain Ky or Sil, the computed areaand by uniform chemicalpartitioningamong


scatteris lessthan *.25oC and 0.45 kbar. Average phases(Tracy 1978). Zoned garnet, modally less
valuesof P-T display excellentintersampleagree- than lVo, is the only mineral that showssignificant
ment (inset in Fig. 8), and indicate an inverted inhomogeneityon a thin-section scale. TWEEQU
metamorphicgradient,with a temperatureincrease, results based on 2l equilibria (Fig. 9, thin-ruled
and pressure decreasefrom south to north polygons) using rim compositions of the garnet
paralleling the distribution of metamorphic show a similar degreeof scatteras the Settlerschist
isograds.Support for the overall accuracycomes samples(Fig. 8) for only one sample(#595C).The
from the fact that TWEEQU results all plot much larger scatterfor the other rocks is reflected
narrowly in the kyanite field, in agreementwith the in the position of the GRAIL (Grt-Rt-Sil-Ilm-Qrz)
occurrenceof kyanite in all samples except #7, equilibrium 2-4 kbar higher in pressurethan GASP
which containssillimanite. or SGAM. Use of core compositionsof the garnet
Tracy (1978)studiedmigmatitic peliteswithin the producesa somewhatlarger scatterat higher P-T
Bronson Hill anticlinorium in west-centralMas- conditions. The INTERSX program finds that six
sachusetts,four of which contain the same e q u i l i b r i a( # 6 , 7 , 8 , 1 3 , 1 5 , l 8 o f T a b l e4 ) a r e
subassemblage, Grt-Ms-Bt-Pl-Qtz-Rt-Ilm-Sil, as significant outliers. That all six involve rutile and
the Settler schist, in addition to pyrrhotite and ilmenite suggeststhat one or both of theseminerals
graphite .r chalcopyrite. Equilibrium is suggested in low modal abundanceis out of equilibrium with
by the mutual contactof all mineralswithin a small the other minerals.Eliminating thesemineralsfrom
THERMOBAROMETRYUSING MULTI.EQUILIBRIUM CALCULATIONS 849

10000

9000

8000
ut2
L
cll 7000
,Q

6000
J
,<

; 5000 Settler Schist


(n
(D Pigage (L976)
!.. 4000 4IR
tr( TWEEQU/Ian91
3000

2000
300 350 400 450 500 550 600 650 700 750 800
re ('c)
Temperatu
Frc. 8. TWEEQU resultsin the KCMFASTH system(T: TiO2)for sample5 of the
SettlerSchist,computedwith the January,1991TwEEQU thermodynamic data.
All equilibria(4 independent)are listed in Table 3. Inset showsthe rangeof
intersectionsfor all six samplesthat contain Ky-Sil.

PEUTES(PIGACE 1976)
TABLB 4. TWEEQURESUL]SFORSrrrl-m' SCHTST the calculationsleads to much less scatter in the
MireElogJrM&BtCn-PlQtr-R![m (! Ky, Sil, Sr Chl)
As@blsee usedi! TmEQU @lculdioN: M!-BtCd-Pl-Ky/Sil{bRlIln
resultsbasedon threeindependentequilibria (Table
Previo$ P-TqdmE: P - G8 kbu I T = 550-7moCeigage, 1t6) 3) for eachsample(Fig. 9, bold triangles).Much
A@ge TWIEQU Bula (6 mpls)i P = 5.7-7.I kbs / T = 554-6180C
(4 indepetrdetrt
improved consistency among all equilibria is
@ti@)
obtained with a reducedactivity of rutile (inset in
Equilibris ploaed itr Fig@ 8 4s Fig. 9), which might be attributed to dilution by a
1: Als + Ihl s Itp + Am -13.7 -0.31
2:6Rt+2Prp+Cn+2Am=3Anr6llnr2Phlr3Qu "119.4 ,8.89 smallamount of nonstoichiometriccomponents(no
3 : 6 R r i P r p + 2 A d = 6 [ m + M s + P h+
l 3 Qtz 0.1
4:3 Rt+P!p+Am= 3 lln iPhl t2 QtrrSil 5.5 -1J7 analytical data are given for rutile by Tracy 1978).
5:6RraGni2Alr!-3A!+6ILn+3Qu lK.1
-n.5
-9.5l Similar improvement is obtained with increased
6:2Aln+ltrlr6Rt=3QtriPrp+Ms+6116
7:Aln+Am+6Rt-3Qtr+Ms+6[rn -13.8 -1.93 activity of ilmenite, possibly reflecting continued
-,9.2 -2.08
8:Ala+3Rr=Sir2Qtrr3nn
-119.2 -7.26
retrograde exchange of ilmenite with sulfide
9i Prp+ Ms + GN - 3 An + Phl
10:cc+6nm12Msr3Qtr=6Rtr3An+2Am -119.1 -5.& minerals.
1l:3Silr3Rt+Iip+2Gs+Am=6Anr3llniPhl -2n,2 "t2.47
l2:2 Sil+Qu+GE = 3 A! "108.3 ').JJ

13:Prp+2M3+3llrn=Am+Phl+3tu+3Sil -16.3 -2,06


14:Ep r Ms = Phl + QE + 2 Sil -10.9 -1.9t Cqlculation of P-T paths
l5:3lln+MsrQe=Silr3RrfAs -5.4 -0.14
16:Msr C6 + Al$ = Am + 3 Ar -132.9 :7.57
-235.9
17:3Silr3Rr12Gn+Aln-6Aa+3llm
18:rIlrtr+2Mir2Prp=3Sil+3Rti2Phl+Alm -2.6
"12.7E
-1./) Becauseof their tectonic significance(England
1 9 : 2 M s + 3[ m + A l m= 2 A m + 3 R t t 3 S i l -30.0
-u.6
& Richardson1977,England& Thompson1984),
20: Ms + Aln = Am + QE + 2 sil
2l:3Sil+3tutG6tAm=3Anr3lintMs -102.9 -5.21 a great deal of effort has been focussed on
ArSin J.Kl; ArV in J.brl; sbbRvidiorsftm I(nE (1983) determiningP-T paths from zonedminerals(Spear
850 THE CANADIAN MINERALOGIST

Tracy(1978)

a)
t<
6
'Q
LogActivityRutile
e)
t<

q)
(t)
(D
Lr
-

Temperature('C)
Ftc. 9. Summaryof TWEEQU resultsin the KCMFASTH systemfor four pelitic
samples with the assemblageGrt-Ms-Bt-Pl-Qtz-Rt-Ilm-Sil (Tracy 1978),
computed with the January, 1991TWEEQU thermodynamicdata. Thin-ruled
polygons show the range of intersectionsamong 2l equilibria (4 independent)
listed in Table 4. Bold triangles show the range of intersectionsamong six
equilibria (3 independent)listed in Table 3 that do not involve ilmenite and
rutile. Inset shows P-Activity(rutile) diagram computed at 500"C for sample
534r (seetext for discussion).

& Selverstone1983,Thompson& England 1984, mometers than net-transfer barometric reactions


Selverstone & Spear 1985, St. Onge 1987, Spear (e.9., Ghent et al. 1988, Frost & Chacko 1989).
1989). In high-grade rocks, however, prograde Frost & Chacko (1989)showed,for example,that
compositional zonation has been erased by dif- re-equilibration of only the Fe-Mg exchange
fusional equilibration. This has led some inves- equilibria producesapparentnear-isobariccooling
tigators(e.g.,Bohlen1987,Anovitz & Essene1990, paths evenif the P-sensitiveequilibria undergo no
Indares& Martignole 1990)to infer the retrograde re-equilibration.
portion of P-T paths from thermobarometry Given the expectation that some temperature-
applied to garnet cores and thin rims that show and pressure-sensitiveequilibria may close at
retrogradecompositionalzonation. A major prob- different temperatures,it becomesnecessaryfor
lem with this type of analysisis the relianceon the petrologiststo decipherP-T paths from the study
assumption that all minerals involved in the of inclusionsthat may remain (e.g., Selverstone &
thermobarometric equilibria maintained equi Chamberlain1990)or to establishthat both setsof
librium through the retrogradeevent.This assump- equilibria reachedequilibrium at the samepressure
tion is in fact unlikely, given the very different and temperature in order for the results to be
kinetic mechanisms and hence lower expected trusted (no matter how reasonablethe calculated
closuretemperaturesfor Fe-Mg exchangegeother- retrogradepaths appear). The techniquedescribed
THERMOBAROMETRYUSING MULTI-EQUILIBRIUM CALCULATIONS 85I

in this paper lends itself to this problem becauseit equilibria listed in Table 3 is significantly greater
offers a general way to test this assumption of than obtained for other "well-equilibrated" suites
equilibrium. In addition, given confidencein the of samples(e.g., Figs. 6 and 8), TWEEQU results
thermodynamic data used, one can look for (nig. lb) appear to indicate that many of their
textural signsof disequilibriumto try to reconstruct samples did re-equilibrate fully at temperatures
P-T paths from samplesor portions of samples between 475 and 600'C. Although it may be
that have locally re-equilibratedat different times tempting to conclude that the distribution of
along a prograde or retrogradepath. samples parallel to the slope of the GASP
St-Onge(1987)deriveda P-T path from analysis equilibrium is an artifact of Fe-Mg exchangeto
of mineral inclusionsin a zonedgarnet, but the full lower temperaturewithout GASP re-equilibration'
advantageof the TWEEQU technique cannot be considerationof the effects of Fe-Mg exchangeon
gained in this example because only two inde- the SGAM equilibrium in the manner describedby
pendent equilibria can be written for the Grt-Bt- Frost & Chacko (1989)showsthat this conclusion
Ky-Pl-Qtz assemblagepresent in these rocks. is unjustified. The extent of Fe-Mg exchange
Hodges & Royden (1984) have calculated a P-T necessaryto reducethe temperatureof the 725'C
path for a portion of the northern Scandinavian sampleof Hodges& Royden(1984)by the observed
Caledonidesfrom pelitic samplesthat apparently 250;C (increasingXp" in Grt or decreasingXo" in
underwent different histories of retrogression. Bt by about 0.14) producesa shift in the SGAM
Although the scatter for three rocks using the barometer between4 and 8 kbar' The magnitude

13m
3 IR - TWEEQU/Ian9I
umO
(1984)
& Royden
Hodges
llm

a lmo
t-
GI
'a 9m

<u 8m0
t-r

7t) 7m0
(t>
)
r{ m00
-
5m0

4m0

30m
4m 450 5m 550 6m 650 7m ?50 8m 850 9m
Temperature('c)
Frc. 10. TWEEQU resultsin the KCMFASH svstemfor pelitesfrom th1l9I49T
Scandinaviari Caledonides, computed with the January, 1991 TWEEQU
thermodynamicdata. Triangles, defined by the intersectionsof equilibria 1-3
of Table 3, show the rangeof intersectionsof all equilibria (3 independent)'See
text for discussion.
8s2 THE CANADIAN MINERALOGIST

of this effect should clearly be discerniblewithin cheque form, payable to "Receiver General of
reasonableexpectationsof uncertaintiesin calibra- Canada", in order to coverhandlingcosts.
tion. The TWEEQU results,therefore,support the
notion that thesesampleshave frozen in a record ACKNOWLEDGEMENTS
of a portion of the retrogradeP-T path, although
the exact P-T trajectory derived here is somewhat I am grateful for the efforts of L. Aranovich,
lesssteepthan that derived by Hodges & Royden B.R. Frost, J. Lieberman,J.A. Percival,M.R. St.
(1e84). Onge,and M.L. Williams, who offerred perceptive
and very useful reviewsof this paper. I would also
like to use the opportunity of this specialvolume
CONCLUSIoNS
honoring H.J. Greenwoodto thank Hugh for the
years of generous guidance and suppofi, and
The thermobarometrictechniqueproposed in oftentimesinspiration he has given me.
this paper offers a test of the critical assumption
that all minerals used in a calculation lasr REFERENCES
equilibratedat the samepressureand temperature.
The correlation of convergentTWEEQU results Arlovrrz,L.M. & EssENE, E.J. (1987):Compatibility
with independent indications of equilibrium in of geobarometersin the systemCaO-FeO-Al2O3-
SiO2-TiO2 (CFAST): implications for garnet
someof the examplesdiscussedabovesuggeststhat mixing models.J. Geol. 95, 633-645.
this techniquemakesit possibleto reasonablyassess
the stateof equilibration of samplesfor systemsin & _ (re90):Thermobarometry and
which the thermodynamicpropertiesappear to be pressure - temperature paths in the Grenville
well described.The number of situationsin which Province of Ontario. ,I. Petrol. 31. 197-241.
independentevidencefor equilibrium is ambiguous
underscoresthe importance of being able to apply AnaNovrcu, L.Ye.. & Poolrssrrr, K.K. (1989):
this thermobarometric test in a general way. In Geothermobarometry of high-grade metapelites:
addition, the technique offers the means to simultaneously operating reactions. .|le Evolution
of Metamorphic Belts (J.S. Daly, R.A. Cliff &
reconstructportions of P-T pathsfrom appropriate
B.W. Yardley, eds.). Geol. Soc. London, Spec.
samples. Publ. 43,45-61.
Although this analysis is hindered by poorly
describeduncertaintiesin thermodynamicdata, as Brnuer, R.G. (1988): Internally-consistent ther-
theseuncertaintiesbecomebetter known, they can modynamic data for minerals in the system
be incorporated into the analysisin the sameway Na2O-K2O-CaO-MgO-FeO-Fq03 -Al2O3-Si02*
as Lieberman & Petrakakis (1991) describefor TiO2-H2O-CO2. J. Petrol. 29, M5-522.
compositionalerrors.At present,this techniquewill
(1990): Mixing properties of Ca-Mg-Fe-Mn
prove most useful in systemsthat involve minerals
garnets.Am. Mineral. 75,328-3M.
with the smallestuncertaintiesin calibration. but
its general applicability should expand a. th.r- & BnowN, T.H. (1984): A thermodynamic
modynamicpropertiescontinueto be refined in the model for multicomponent melts, with application
future. This method facilitates the refinemenr to the system CaO-Al2O3-SiO2. Geochim. Cos-
processby (l) highlighting which mineralsare most mochim. Acta 48, 661-678.
incompatible in each calculation and may be in
need of refinement, (2) illustrating rhe sensitivity - & PenrrNs,E.H. (1987):GE@-CALC:
of resultsto different solution-modelsand offering software for calculation and display of pressure -
the chance to correlate these differences with temperature - composition phase diagrams. Am.
Mineral. 72,861-862.
petrographic observations, and (3) providing a
meansto select"well-equilibrated" samplesthat can ENcr, M., GnreNwooo,H.J. & BnowN, T.H.
be usedto refine thermodynamicpropertiesin the (1986): Derivation of internally-consistent ther-
absenceof appropriate experimental data. It is modynamic data by the technique of mathematical
hoped that feedback from users of this software programming, a review with application to the
will increase the pace of improvements in the system MgO-SiO2-H2O. J. Petrol. 27, 133l-1364.
thermodynamicdata, which in turn will provide an
increasingability to gain insights into petrological & Kozror, A.M. (1991): Ternary excess
problemsand to reconstructgeologicalhistories. properties of grossular - pyrope - almandine
garnetsand their influence in geothermobarometry.
The TWEEQU software may be obtained from Am. Mineral. 76, 1223-1231.
the author by sending a self-addressedenvelope,
5-1/4" or 3-l/2" floppy disc,and $20Canadianin BoHLEN,S.R. (1987): Pressure- temperature- time
THERMOBAROMETRYUSING MULTI-EQUILIBRIUM CALCULATIONS 853

paths and a tectonic model for the evolution of and experimental data, and applications to geother-
granulites. J. Geol. 95, 617-632. mo-barometry. Am. Mineral. 69, 88-97.

Bnowr, T.H., BrnveN, R.G. & PEmtNs,E.H. (1988): GurNr, E.D., Srour, M.Z. & PeRRtsH,R.R' (1988):
GEO-CALC: software package for calculation and Determination of metamorphic pressure- tempera-
display of pressure - -
temperature composition ture - time (P-T-0: paths. ,Ir Heat, Metamor-
phase diagrams using an IBM or compatible phism, and Tectonics (E.G. Nisbet & C-R.M.
personal computer. Comput &. Geosci, 14, 279- Fowler, eds.).Mineral. Assoc. Can., ShortCourse
289. Handbook 14, 155-188.

CHerrcn:m, N.D. & Fnorsr, E. (1975): A ther- Hennts,N.B.W., Holr, R.W. & Dnunv, S.A. (1982):
modynamic study of the pseudobinary join mus- Geobarometry, geothermometry, and late Archean
covite - paragonite in the system KAlSi3O6-NaAl geothermsfrom the granulite faciesterrain of south
Si3O8-A12O3-SiO2-H2O. Am. Mineral. 60, 985- India. "/. Geol. 90, 505-527.
993.
HoDcrs, K.V. & McKrNNe, L.W. (1987): Realistic
CooleN, J.J.M. (1980): Chemical petrology of the propagation of uncertainties in geologic ther-
Furua granulite complex, southern Tanzania, GUA mobarometry. Am. Mineral. 72, 671-680.
(Amsterdam), Pap. 13, l-258.
& RovopN, L. (1984): Geologic ther-
Drvenesr, H.H., Jn. & HesELtoN,H.T., Jn. (1981): mobarometry of retrograded metamorphic rocks:
Error analysis for bracketed phase equilibrium an indication of the uplift trajectory of a portion
data. Geochim. Cosmochim. Acta 45,217-224. of the northern Scandinavian Caledonides. ./.
Geophys. Res. E9, 7077-7090.
ENcr, M., LtmnnveN, J. & Brnrr,raN,R.G. (1990):
Uncertainties in thermodynamic data for mineral & Spsen, F.S. (1982): Geothermometry'
geobarometry and the Al2SiOs triple point at Mt.
systems. Geol. Assoc, Can, - Mineral. Assoc,
Can., Program Abstr. 15, 436. Moosilauke, New Hampshire. Am. Mineral. 61,
I I 1 8 - l1 3 4 .
ENclano, P.C. & RtcH,rnnsoN, S.W. (1977): The
HollaNr, T.J.B. & Powell, R. (1990): An enlarged
influence of erosion upon the mineral facies of
and updated internally consistent thermodynamic
rocks from different metamorphic environments.
dataset with uncertainties and correlations: the
J. Geol. Soc. London 134,201-213.
system KrO-Na2O-CaO-MgO-MnO-FeO-FqO3-
Al2o3-TiO2-SiO2-c-H2-O2. J. Metamorph. Geol.
& THovesoN,A.B. (1984):Pressure- tempera- 8.89-124.
ture - time paths of regional metamorphism. I.
Heat transfer during the evolution of regions of INoenrs. A. & MerrlcNor-e' J. (1990): Metamorphic
thickened continental crust. ,f. Petrol. 25,894-928.
constraintson the evolution of the gneissesfrom
the parautochthonous and allochthonous polycyclic
FEnnv, J.M. (1976): Metamorphism of calcareous belts. Grenville Province, westernQuebec.Can. J.
sediments in the Waterville - Vassalboro area, Earth. Sci. 27. 357-370.
south-central Maine: mineral reactions and graphi-
cal analysis.Am. J. 9ci.276,841-882. JENrrNs,D.M., NrwroN, R.C. & Gor-ostnltrH, J.R.
(1983): Fe-free clinozoisite stability relative to
(1980): A comparative study of geother- zoisite. Nature 304. 622-623.
mometers and geobarometersin pelitic schistsfrom
south-central M.aine. A m. M iner al. 65,'720-732. JounNeey, J.M. (1986): Stratigraphy, Internal Strain'
ond Thermotectonic Evolution of Northern
-& Seran,F.S. (1978):Experimentalcalibration Frenchman Cap Dome: on Exhumed Duplex
of partitioning of Fe and Mg between biotite and Structure, Omineca Hinterland, S.E. Canadian
garnet. Contrib. Mineral. Petrol. 66, 113-117. Cordillera. Ph.D. thesis, Queen's Univ', Kingston'
Ontario.
Fnosr, B.R. & CHacro, T. (1989): The granulite
uncertainty principle: limitations on ther- KonN, M.J. & Sernn, F.S. (1991):Error propagation
mobarometry in granulites.J. Geol. 97,435-450. for barometers. 2. Application to natural rocks.
Am. Mineral. 76, 138-147.
FuunlreN, M.L. & LrNosl-ev,D.H. (1988): Ternary-
feldspar modeling and thermometry, Am. Mineral. KoLesse, J.E. (1990): Confidence intervals for ther-
73.20t-2r5. modynamic constants. IBM Res. Rep. 15769.

GeNcur-v,J. & SexrNa,S.K. (1984):Mixing properties KozroL,A.M. (1990):Activity - compositionrelation-


of aluminosilcate garnets: constraints from natural ships of binary Ca-Fe and Ca-Mn garnets
854 THE CANADIAN MINERALOCIST

determined by reversed, displaced phase equi- Schist,southwestof Yale, British Columbia. Can.
librium experiments. Am. Mineral. 75, 319-327. J. Earth Sci. 13, 405-421.

& BoHrcN, S.R. (1990): Garnet (almandine - Powgr-r-, R. (1985): Geothermometry and
pyrope) geobarometry. Geol. Soc. Am., Abstr. geobarometry: a discussion. J. Geol. Soc. London
Program 22, A7l. r42.29-38.
& NewroN,R.C. (1989):Grossularactivity- & HolmNo, T.J.B. (1988): An internally
composition relationships in ternary garnets deter- consistent dataset with uncertainties and correla-
mined by reversed displaced-equilibrium experi- tions. 3. Applications to geobarometry, worked
ments. Contrib. Mineral. Petrol. 103, 423-433. examplesand a computer program. J. Metamorph,
Geol. 6, 173-204.
Knerz, R. (1983):Symbolsfor rock-forming minerals.
Am. Mineral. 68, 271-279. Sr-ONcr, M.R. (1987):Zoned poikiloblastic garnets:
P-T paths and syn-metamorphic uplift through 30
Lasece, A.C. (1983): Geospeedometry:an extension km of structural depth, Wopmay Orogen, Canada.
of geothermometry.In Kinetics and Equilibrium in J. Petrol.28. l-21.
Mineral Reactions (S.K. Saxena, ed.). Springer-
Verlag, New York (81-114). SrLvrnsroNr, J. & CHeNrsrRLAlN, C.P. (1990): Ap-
parent isobaric cooling paths from granulites: two
Lee, H.Y. & GeNculv, J. (1988): Equilibrium counterexamples from British Columbia and New
compositions of coexisting garnet and or- Hampshire. Geology 18, 307-310.
thopyroxene: experimental determinations in the
system FeO-MgO-Al2O3-SiO2, and applications. -& Srnen, F.S. (1985):Metamorphic P-T paths
J. Petrol. 29,93-113. from pelitic schists and greenstones from the
south-west Tauern Window, Eastern Alps. "/.
LrerrnuaN, J. & PrrncKarrs, K. (1991): TWEEQU Metamorph. Geol. 3, 439-465.
thermobarometry: analysis of uncertainties and
applications to granulites from western Alaska and Seran, F.S. (1989): Petrologic determination of
Austria. Can. Mineral. 29, 857-887. metamorphicpressure- temperature- time paths.
/r Metamorphic Pressure - Temperature - Time
MAorn, U.K. & BrnlreN, R.G. (1991):A high pressure Paths (F.S. Spear & S.M. Peacock, eds.). Am.
equation of state for carbon dioxide consistentwith Geophys. Union, Short Course in Geology 7.
phase equilibrium and P-V-T data. Am. Mineral.
76,1547-t559. & SrlvensroNr,J. (1983):QuantitativeP-T
paths from zoned minerals: theory and tectonic
McKrNNa, L.W. & Hoocts, K.V. (1988): Accuracy applications. Contrib. Mineral. Petrol. 83, 348-
versus precision in locating reaction boundaries: 357.
implications for the garnet - plagioclase -
aluminum silicate - quartz geobarometer. Am, TuovrsoN, A.B. & ENcLnNo,P.C. (1984):Pressure-
MineraL 73. 1205-1208. temperature - time paths of regional metamor-
phism. II. Their inference and interpretation using
McMullm, D.W., BEnruaN, R.G. & GnrrNwoon, H.J, mineral assemblages in metamorphic rocks. ./.
(1991):Calibration of the SGAM thermobarometer Petrol. 25.929-955.
for pelitic rocks using data from phase-equilibrium
experiments and natural assemblages. Can. Tnecv, R.J. (1978):High grademetamorphicreactions
Mineral. 29, 889-908. and partial melting in pelitic schist, west-central
Massachusetts. Am. J. Sci. 218, 150-178.
MorcHrn, D.P., EssrNr,E,J. & ANovrrz, L.M. (1988):
Calculation and application of clinopyroxene - VrnNoN, R.H. (1976): Metamorphic Processes:Reac-
garnet - plagioclase - qtaftz geobarometers. tions and Microstructure Development, George
Contrib. Minerol. Petrol. 100. 92-106. Allen and Unwin, London.

NEwrolr, R.C. (1983): Geobarometry of high-grade - (1977):Relationships


betweenmicrostructures
metamorphic rocks. Am. "/. Sci. 283-A, l-28. and metamorphicassemblages, Tectonophys.39,
439-452,
& PnnriNs, D., III (1982): Thermodynamic
calibration of geobarometers based on the as- Zru, E-AN (1963): Components,phases,and criteria
semblages garnet - plagioclase - onhoplroxene of chemical equilibrium in rocks. Am. J. Sci. 261,
(clinopyroxene) - quartz. Am. Mineral. 67,203- 929-942.
222.
Received January 14, 1991, revised manuscript ac-
PrcAcE, L.C. (1976): Metamorphism of the Settler cepted October 2, 1991.
THERMOBAROMETRYUSING MULTI-EQUILIBRIUM CALCULATIONS 855

APPeNPtx: involves the incorporation of generalizedMargules-type


DESCRrprroN oF rHE TWEEQU Sorrwane, solution equations (Berman & Brown 1984)to describe
activity coefficientsof multisiteminerals.Their incorpora-
tion allows accuratedisplacementof all equilibria using
The general expression for equilibrium (3) is a P-T-dependentmodelsthat describethe nonideal mixing
nonlinear equation in P and T that defines a locus of along with mineral compositions read from a user-
points in P-T space that must be solved by iterative specified file. These mixing models are defined in-an
techniques. The thermobarometric method advocated external file that can be modified to add other
here presentsall solutions, both graphically, in the form solid-solution phasesor alternativemixing models for a
of P-T-X(CO)-activity diagrams,and analytically,based given solution phase.At run time, the educateduser can
on the graphical results. All calculationsare performed usefully select between alternative solution-models in
by a computer program (TWQ) modified as described order to explore the ramifications on the ther-
below from the GEO-CALC software package for mobarometric results. Details of these proceduresare
computation of phase diagrams (Berman et al. 1987, given in a documentation file that accompaniesthis
Brown el a/. 1988).Detailsofthe variousalgorithmsused software.
by this software can be found elsewhere(Brown el a/. The January, 1991 TWEEQU data-base includes
1988),and only a brief summary is given here. standard state thermodynamic properties of Berman
Operationof this softwareis facilitatedby an extremely (1988) with modifications and additions of Berman
user-friendlyinterface.The only information required of (t$o), Vaa.r & Berman (1991),and McMullin er a/'
the useris the choiceof type of diagram(any combination
41991).Nonideal mixing is accountedfor using the
of P, T, X(CO), activity), the desired range of these calibrationsof Berman(1990)for garnet' McMullin e/ 4/.
variablesin the diagram, the chemicalsystemand phase (1991) for biotite, Furhman & Lindsley (1988) for
components to include in the calculations. For most plagioclase,and Chatterjee&Froese(1975)for muscovite.
applications,all componentsof phasesthought to have ideal two-site mixing is assumed for ortho- and
equilibratedat the sameP and T are selected.The TWQ clinopyroxene(Newton 1983). In order to facilitate
progrnm generatesall possible reaction-stoichiometries comparisonswith alternativedata-bases, the program can
using linear algebraic methods, and, after reading an accommodateother data-bases with a variety of equations
accompanl'rngfile containingall thermodynamicdata and of state (e.g., Holland & Powell 1990).This fealure is
mineral compositions, follows standard root-finding particularly useful to explore the sensitivity of ther-
procedureswithin the selectedP-T spaceto locatepoints mobarometric results to uncertaintiesin standard state
on eachequilibrium curve in turn. Resultsare written to thermodynamicdata. One caveat,however,is that some
an ASCII file that is read by separateprogramsthat allow standard state and mixing properties are tied together
the graphicalresultsto be viewed on a video monitor or from their method of calibration, and this internal
hard-copy device (dot matrix or laser printer), or to be consistencymust be preserved. For example, Berman
analyzedby the INTERSX program. (1990)derivedannite propeniesfrom the garnet- biotite
Whereas the GEO-CALC software implements a exchange data of Ferry& Spear(1978),assuminga newly
strategy for calculating metastability and removing derivedset of garnet mixing parametersand ideal Fe-Mg
metastableequilibria and extensions,the presentmethod mixing in biotite. Use of a different set of mixing
is simplified by not having to test for stability sinceevety propeities for garnet requiresuse of different properties
-
equilibrium curve is equally valid for thermobarometric ior annite in order to maintain consistencywith the
purposes regardless of its stability level. The main exDerimentaldata.
modification of the GE@-CALC software, however,

You might also like