American Mineralogist.

Volume78, pages68 I-693, 1993

in pelitic schists:A rapidly eyolyingfieldx

Geothermobarometry
M. J. Hor,o,q.wav, Brsw.l;rr MuxrroploHyAy
Department of Geological Sciences,Southern Methodist University, Dallas, Texas 75275, U.S.A.

Ansrnrcr
The study of pelitic rocks for the purpose of deciphering pressure-temperature(P-Z)
information is an important area for collaboration betweenmineralogistsand petrologists.
Accurate geothermobarometryis especiallyimportant for studiesof pressure-temperature-
time paths of terranesduring tectonism and for studies of the movement of metamorphic
fluids. During the past decade, new developments have included thermodynamic data
bases,sophisticatedcrystal structure determinations and associatedsite assignments,and
analysesfor Fe3+ and light elements (especially H and Li) for many minerals within a
petrogeneticcontext. It is now necessaryto apply thesenew findings to geothermobarome-
try in several important ways: (l) The stoichiometric basis for hydrous minerals should
be revised in light of highly variable H and variable Fe3*, which can now be estimated
with considerableaccuracyas a function of gradeor mineral assemblage.(2) Mole fraction
(activity) models should be based on the best crystal chemical considerations.For some
minerals, H may be omitted from the model if all the substitutionsinvolving H are coupled
substitutions. (3) Thermodynamic data should be basedon careful analysisof all available
experiments and secondarycomparison with natural assemblages.(4) The possibility of
nonideal solid solution should be considered,as ideality is merely a specialcaseof nonide-
ality. It is better to estimate binary interaction parameters than to assumethat they are
zero. However, it is difficult to determine ternary interaction parameters.In such cases,
little error is likely to result from assumingthat strictly ternary interaction parametersare
zero, while evaluating all the binary terms. Our understanding of garnet and biotite solu-
tions has improved, and the garnet-biotite geothermometerhas undergonenumerous re-
finements,but additional information on theseminerals and improvements in the garnet-
biotite thermometer are still necessary.
As the P and ?"determinations become more precise,we can better evaluate Xnro, X"o,
and X.ro in the fluid phase. In all probability, patterns will be discovered in the compo-
sitions of pelitic metamorphic fluids, especially when graphite is present. The hematite-
ilmenite solid solution system is potentially useful for determining .fo2,oree the solvus is
better understood. In graphitic rocks at known P and T, knowledge of /o, from coexisting
hemoilmenite and rutile or hemoilmenite and magnetite enables one to determine the
mole fraction of all fluid species.
Careful application of the present state of knowledge,combined with modest improve-
ments, should allow accuracyapproaching +250 bars and +25 "C in rocks formed at low
to moderate pressures.It is very important to continue to revise thermochemical data,
activity models, Margules parameters,and stoichiometric information, and it is equally
important for petrologistsand tectonicists to make use of these revisions as they become
available.

INrnooucrroN if we are not continually watching for them and correcting

In recent years, we have become concernedabout the
simplified view taken by some petrologists and tectoni-
cists regardingquantitative estimation of the pressureand
temperature of formation of metamorphic rocks. There
are many possiblesourcesof error in this procedure,and,
t This report is a revised, expanded version of the Mineral-
ogical Society of America Presidential Address given by M.J.
Holdaway at the Geological Societyof America Meeting in Cin-
cinnati, Ohio, on October 27, 1992.
0003-004x/93l0708-o68I $02.00
them wherever possible, we will not get the best preci-
sion, especiallyin the P-T ratge outside of the range of
calibration. This is an ideal area for collaboration be-
tween mineralogists and petrologists. Mineralogists pro-
vide valuable information on mineral structure and site
assignments,and petrologistsapply the results to activity
models. Both mineralogists and petrologists are making
important contributions to careful, complete chemical
analysis of common metamorphic minerals.
Interesting work is being done on pressure-tempera-
681
682 HOLDAWAY AND MUKHOPADHYAY: GEOTHERMOBAROMETRY IN PELITES

P difference between these two estimates is 450/oof the

lower value.
Here we review the processof P, T, and X"ro estima-
tion, focusing on five important areas:(l) determination
of mineral stoichiometry, (2) mineral mole fraction (ac-
o tivity) models, using the mixing-on-sites model, (3) non-
E
o ideality in mineral solid solutions, (4) derivation and use
U)
-6 of thermodynamic data, and (5) fluid composition and
c
o its use to test results. If petrologists carefully consider
E
l what is known in eachofthese areasand continue to heed
z
developmentsas they becameavailable, we will converge
on a consistent P-T-Xtro grid for all metamorphic ter-
ranes and grades.In our analysis, we will focus on me-
dium-grade pelitic metamorphism, but some of what we
have to say is equally applicable to other gradesor bulk
compositions. The major silicate minerals we will be dis-
Wt.%Water
cussing are muscovite, biotite, garnet, staurolite, cordi-
Fig. 1. Histogramof H.O contentof all muscoviteand bio-
erite, aluminum silicates, and chlorite; the Fe-Ti oxides
tite samplesanalyzed by M.D. Dyar.After Dyar et al. (1993),
reprintedwith the kind permissionof PergamonPress. ilmenite, hematite, magnetite, and rutile, as well as
graphite, will also come up for discussion.We will con-
sider mainly the dominant mineral end-membersand will
ture-time paths (e.9.,Spearand Selverstone,1983) and not be consideringsulfide equilibria. The effectsofsulflde
on fluid composition and movement in metamorphic ter- on Xrro in medium-grade reduced pelites can be ignored
ranes(e.9.,Ferry and Dipple, l99l). There are also sig- ifthe sulfideis pyrrohotite and not pyrite (Shi, 1992).
nificant controversies over how much P-Z information
MrNsnAl, SToICHIoMETRY
one can learn from zoned minerals (e.g.,Frost and Tracy,
1991) and how much fluid actually passedthrough the Extensive and, commonly, exclusiveuse of the electron
rocks (e.g.,Gray et al., l99l). The quality of the results microprobe for mineral analysis has left us short on im-
of many of these studies may well be only as good as the portant data for many minerals, especiallyhydrous ones.
quality of the geothermobarometryused. Only when sig- In many studies, mineral chemistry is based on micro-
nificant improvementsare made to the accuracyof P-T probe data alone. The resultant shortcomings include a
determination will we possibly be able to answer these lack of data for Fe3+/(Fe2++ Fe'*) and a lack of light-
questions. That alone is an important justification for elementanalyses,such as H, Li, B, and Be, on minerals
continuing our efforts to improve geothermobarometry. as diverse as micas, chlorite, staurolite, garnet, and tour-
To illustrate the seriousnessof the problem, we com- maline. In this list all may contain significant amounts of
pare two recent studiesinvolving geothermobarometryin Fe3+,all but garnethave variable H, biotite and staurolite
pelitic rocks. In Acadian M3 in Maine and in the Dela- may have significant Li, and tourmaline has variable B
merian of South Australia. near Adelaide. isobaric ther- (Dyar et a1.,1992,1993;Holdawayet al., 1986a,1986b;
mal metamorphism produces assemblagescontaining Dutrow et al., 1986). The assumption of ideal or zero
staurolite * andalusite followed by sillimanite at higher content of light elementsand zero Fe3* may have pro-
grade. In both areasthe appearanceof sillimanite corre- duced inaccuratestoichiometry,and it has precludedthe
sponds approximately to the disappearanceof both an- possibility of several important substitutions that are
dalusite and staurolite. In both areas,cordierite is locally coupledwith thoseof major elements.Dependingon the
present,probably becauseofan earlier, lower-P subevent, mineral, the H content of hydrous minerals may be a
a local decreasein P, or a differencein bulk composition. function of metamorphic grade,of mineral assemblage,
All petrologic evidenceindicates that thesetwo localities or both. With light-elementand Fe3* analyticaldata for
have enjoyed about the same P-T conditions. Average representativeminerals,we can make appropriatecorrec-
metamorphic conditions for the staurolite-bearingrocks tions to the stoichiometricbasis.Here we summarizewhat
p":+1 and light elements,
deducedby Holdaway et al. (1988) for Maine M3 are P we know about Fe3*/(Fe2+*
: 3.1 kbar, T : 531 'C; conditionsdeducedby Dymoke how they affect stoichiometry,and possiblesite assign-
and Sandiford (1992\ for the Delamerian near Adelaide mentsfor Fe3*.
are P : 4.5 kbar, T : 594 oC, corresponding to an ap-
parent P difference of 1.4 kbar or a depth difference of Muscovite
nearly 5 km. Whereaspetrologists who work in rocks at Muscovite suffers from two compositional problems.
higher P and I might not consider these to be serious (l) Preliminary Mdssbauerstudy (M. D. Dyar, unpub-
differences, we believe it should be possible to obtain lished data) shows that half to two-thirds of the Fe in
much greater accuracyin low- to moderate-P rocks than muscovite is Fe3+.Except for highly oxidized, Iow-grade,
is implied by this comparison. It is worth noting that the or high-P rocks, this efect can be ignored becauseit has
 HOLDAWAY AND MUKHOPADHYAY: GEOTHERMOBAROMETRY IN PELITES 683

TABLE1. HrO content of Maine pelitic micas as a function of Trale 2. Stoichiometryof micas,specimen143'
grade-
Muscovite.* Biotitel
Muscovite No H, Fe3* No H, Fe3*
Average Average correction corrected correction corrected
No. (%) No. ('/") 2.696
rrlSi 3 034 3 076 2.666
Garnet 8 3.54 0 86 16 4.52 0.30 rrlAl 0.966 0.924 1.334 1.201
Staurolite 10 3.78 0.39 10 4.16 0.47 rrlFe3r 0.000 0.000 0.000 0 103
Sillimanite 26 4.06 0.35 29 3.83 0 35 rMrAl 1.866 1.947 0 390 0.542
Ksp-Sil-Mus 4 3.84 020 5 3.21 0.30 rMtTi 0.038 0.039 0 173 0.175
Atl 48 3.90 0.49 60 4.02 0.51 rMtFe2+ 0.060 0.031 1.277 1.136
rMtFes+ 0.000 0.030 0.000 0.052
. Data of Dyar et al. (1993 and unpublisheddata) lutMg 0.05s 0.056 0.952 0.963
tMtMn 0 001 0 001 0.011 0.011
rM1! 0.980 0.896 0197 0.121
ratK 0.917 0.930 0.971 0 982
IAINa 0.064 0.065 0 018 0.018
a trivial effect on stoichiometry and activity. (2) Dyar et OH 2.000 1.666 2 000 1.602
al. (1993and unpublisheddata)have determinedthe HrO Mole fraction+ 0 781 0.785 0.066 0.044
content of 48 pelitic muscovitesamplesfrom Maine that 'Holdaway et al. (1988) F neglected.
.* Assumesall IMrAIis on M2, all IMrnis on M1i tor correctedvalues,half
show a range from 2.5 Lo 4.8o/oHrO. Data for all the
biotite and muscovite samplesthey have analyzedare of Fe is Fe3*
t Assumingdisorderof all specieson M1 and M2, Guidottiand Dyar
summarizedin Figure l; the two minerals have similar ( 1 9 9 1 )F e 3 + .
rangesof HrO content. There is a generalpositive rela- t See Table 4 for mole fraction models. Biotite mole fractions are for
annite
tionship betweenHrO content and gradefor muscovite,
with a slight decreaseat the highestgrade(Table 1).Note
however, that most pelitic muscovite has less than the
stoichiometricvalue of 4.5o/o HrO. This is also important titatively significant amounts of Li, as shown by Dutrow
becauseit may affectthe nature of ionic substitutionsin et al. (1986)and Dyar etal. (1992and unpublisheddata).
muscovite. For a muscovite in the potassiumfeldspar- This effect can commonly be ignored but may be the
sillimanite-muscovite zone, estimating3.80/oHrO from cause of P-I discrepanciesin individual specimens.(3)
the resultsofDyar et al. (1993and unpublisheddata)has According to the work of Dyar et al. (1993 and unpub-
a small effecton the stoichiometry(Table 2). For a mi- lished data) on 60 specimens,pelitic biotite rangesbe-
croprobe analysis,with O substituting for OH , the O tween 2.9 and 5.2o/oin HrO, whereasthe ideal value is
basisbecomesI 1.15 insteadof I 1, slightly increasingthe 4o/o.As with muscovite,that affectsthe nature of ionic
total cationic content. The H* deficiencymay compen- substitutions.Biotite shows a negative relationship be-
sate for excesschargein the octahedrallayer over that tween HrO content and grade (Table l). For a biotite
necessaryto chargebalancethe tetrahedrallayer. F con- samplein the potassiumfeldspar-sillimanite-muscovite
tents of muscoviteare low and insufficientto explain the zone, Dyar's averageHrO content of 3.2o/o increasesthe
H* variations (Dyar et al., 1993 and unpublisheddata). O basisto I 1.2, increasesthe content ofcations, and re-
Correctingfor H and Fe3+increasesactivity slightly (Ta- ducesthe vacancycontent of the octahedralsites(Table
ble 2). 2). The deficiencyof H in many pelitic biotite samples
cannot be explainedby F (Dyar et al., 1992 and unpub-
Biotite lished data) but may compensatefor increasedchargeof
Metamorphic biotite hasthree problems associatedwith Fe3*, Al, and Ti in the octahedralsitesabove that bal-
composition. (l) An pelitic biotite that grew under re- anced by tetrahedralsubstitution (Dyar et al., 1993).In
duced conditions with graphite and ilmenite has about this case,note that making correctionsfor H and Fe3*
l2oloof Fe as Fe3*,as shown by wet-chemicaland Moss- reducesactivity substantially,mainly becauseof the effect
bauer studies.Guidotti and Dyar (1991) interpretedthe of Fe3*.
Miissbauer spectraon reducedpelitic biotite from Maine
to indicate on average80/oof the ps 4s t+lps:+and 4oloas Staurolite
I6lFe3+.Both the amount and the siteassignments for Fe3+ Work on staurolite shows that determination of stau-
in these pelitic biotite samples (Dyar, 1990) have been rolite stoichiometry suffers from the same problems as
questionedby Rancourt etal. (1992) in a study ofbiotite biotite. (l) Staurolite that grew under reducing conditions
of different composition and origin. However, Dyar (1993) containson average3.50/o of the Fe as Fe3* (Dyar et al.,
pointed out that detection oft4lFe3+requires random ori- l99l; Holdaway et al., 1991),probably in the tarFe(T2)
entation in micas, I41Fe3+ has often been identified in bi- sites.(2) Most staurolite containsLi (Dutrow et al., 1986),
otite and other phyllosilicates,and the Fel;f has been con- and in the absenceof a Li analysis or evidence of high
firmed by many independent wet-chemical studies, such Li, it is best to assumethe averagevalue of 0.2 Li ions
as that of Williams and Grambling (1990). Guidotti's to avoid systematicerrors(Holdawayet al., l99l). Better
(1984) review of chemistry of pelitic biotite also showed yet, representative specimensfrom a telTane should be
substantial Fe3*. (2) A few biotite samplescontain quan- analyzed for Li with the ion microprobe or proton-in-
684 HOLDAWAY AND MUKHOPADHYAY: GEOTHERMOBAROMETRY IN PELITES

Trele 3. Stoichiometryof natural staurolite samples at ex- Moln FRAcrroN MoDELS

tremes of H content.
The proper choice of a mole fraction (activity) model
3-3 71-627 is crucial for all subsequentcalculations. The best possi-
rMltAl 7.70 7.70 ble choice of models is always a problem, and it is im-
tM1lR2+ 030 0.30 portant to be as correct as possible. Also, petrologists
IMrtAl 779 7.79
{Mrr(Rr+ ,Ti) 0.21
should acceptthe fact that as our understandingof crystal
0.21
rM3tAl 187 1.59 structure, site assignments,and composition improves,
r M 3 t ( Rt ,s , n ) 0.13 0.41 we must continue to allow activity models to evolve. A
lM4lFe2+ 002 051
{M4r(Zn,Li,!) 1.98 149 number of important points are as follows: (l) It is better
rrltSi 763 to make an incorrect decision on disorder or even site
rrltAl 0.37 041
tr2tFe2+
assignmentthan to make an error that renders the activ-
3.01 218
r r 4( R r + ,R 3 + , L i , ! ) * * 0.99 192 ity model inconsistent. For example, both the mole frac-
H 2.68 4.56 tion basis of an ion and its exponentmust be the same
si +Al -%Li+2/sTi+Fe3+ 25.53 25.38
number. The equation for mole fraction must flt the
. Analyticaldata from Holdaway
et al (1986b)and Dyar et al. (1991), model. (2) Care should be taken to avoid accounting for
site assignmentsfrom a model based on Hawthorneet al. (1993a)and
Holdawayet al (1991),modifiedby Holdawayet al. (in preparation).
two substitutions when the ions are individually coupled
. ' E x c l u d i n gF e ' z + . by charge balance constraints. In staurolite, H substitu-
tion is alwayscoupledto other major substitutions.(3) If
a given model was necessaryfor retrieval of thermody-
duced 7 emission.(3) Staurolitehas variable HrO con- namic data, then that model, if reasonable,should also
tent. That which occurs with biotite or garnet has about be used for geothermobarometry. (4) Any mineral that
3 H+ pfu, whereasthat which occurs with aluminum sil- has two or more independent sites also has the potential
icate and no biotite or garnet has about 4 H+ pfu (Hol- for site interaction energies.When we ignore this possi-
daway et al., 1986a;Dyar et al., 1991).This observation bility, we are assuming these energiesare zero. Even if
can be used to determine the stoichiometry of staurolite mixing on the independent sites is ideal, this cross-site
that occurs with quartz in the absenceof a HrO analysis, interaction can have an efect on the mole fraction cal-
but in our opinion it is better to take advantage of the culation. (5) In most cases,it is not yet possibleto test
constancyof Si + Al and normalize the stoichiometry to alternative mole fraction models, although more such ef-
Si + Al - VtLi + 2/tTi + Fer+ : 25.55 (Holdawayet al., forts should be attempted. With thesepoints in mind, we
1991; Table 3, this report). This approachpreservesthe present possible models for some of the common min-
constancy of Si and Al while correcting for the minor erals (Table 4).
substitutions in natural staurolite that involve Si and Al. Activity models for micas are still very much open to
With this approach, subtraction of the total cation charge question; the ones presentedhere are simply reasonable
from 96 estimatesH+ within about I pfu. Either of these approximations. We assume octahedral sites in biotite
approachesis significantly better than the previously used are disordered. This is-supportedby the lack ofcontrary
method of assumingeither 2 or 4 H pfu. X-ray diffraction evidenceand the factthat, at metamor-
Comprehensivecrystal structure studiesby Hawthorne phic T, disorder is even more likely than at room Z. We
et al. (1993a,1993b,1993c)enableone to apportionthe assumetatAlavoidancein micas, as advocatedby Circone
ions of a staurolite analysis into all the sites (Table 3). and Navrotsky (1992), and make the simplifying as-
These two Fe-rich staurolite samplesrepresentextremes sumption that alternate tetrahedral sites are occupied by
ofthe known rangeofH content. Note that the Fe content Si and show no solid solution. We suggestthat the re-
of the T2 site is considerablylessin the high-H staurolite maining two tetrahedral sites, occupied by one Al and
becauseof the vacancy content of about one ion. This one Si in end-member micas, and by Al, Si, and possibly
approachof assigningions to their proper sitesusing crys- Fe3* in natural micas, should be incorporated into the
tal structure determinations as a model considerably im- mole fraction model. Ackermann et al. (1993) presented
proves the quality of stoichiometry and the resulting mole spectroscopicevidence that OH- replaces some apical
fraction calculations for stoichiometric staurolite end- I4lO2-in synthetic germanium muscovite. This observa-
members. tion and compositional data suggestthat at least some
torAl is charge compensatedby adjacent OH replacing
Chlorite O'-. The t4rAlin micas is probably not chargecoupled to
We only wish to mention chlorite briefly. The work of octahedralsubstitutions on an ion by ion basis.However,
Dyar and Guidotti (1992) showed that the stoichiometry Essene(1989) suggested that the tetrahedralsitesshould
of chlorite in pelites is affectedby two main variables: (1) be omitted from mica models. The l2-fold site vacancies
Chlorite contains Fe3+ (about I lolo of Fe). (2) Average in many micas might be occupied by some form of H
HrO content is 100/ocompared with the ideal value of (e.g.,Miller et al., l99l). H* deficiencyin the OH po-
12010.These observations should be taken into account sitions is almost certainly H+ vacanciesrather than OH
when determining chlorite stoichiometry and activity. vacancies and is probably coupled to cation charge in
 HOLDAWAY AND MUKHOPADHYAY: GEOTHERMOBAROMETRY IN PELITES 685

TABLE4. Possiblemole fractionmodelsfor some peliticmineralsin the FASH system

End-member Composition Activity model-

Muscovite KAlrtSisAlOlo(OH), 4.I^lX<IM2lXlt IM1lXtrtrlxAr ttlx", X3"**

Annite KFe3Si3AlOlo(OH), 4 tArXKIMrXPe.trlXA,trlx",.XSx**
prxPer6rxer
r4rx3
Almandine Fe3AlrSi3Of
2H staurolite HrFe4!rAllsSisO4s Ir?lxneIMatX2tM1lxl.lM2lxil tM3lxir rrrlxg,l
{M4lX2e IM3rXrAr,rrrlXBl
IMItX€r,rM2lXh
6H staurolite H6!lFerAl,sSioO4s
0H cordierite Fe,Al4SisOls 0HrO I"rX?. (1 - &.o)
2H cordierite FerAl4SisOls lHrO IU]X?. X,,O

'All Fe refersto Fe,*, except where noted

" For mica models,IrrAlavoidanceis assumed,rrrxs,: (Si 2)l2,rtx^t:rtAV2, XBH: (OHlwl*, where w: OH + F Essene(1989)suggestedthat
the tetrahedral mica factors should be omitted. Circone and Navrotsky (1992) suggested tMlAlis mixed on a single site in phlogopite-eastonite-
t F o r s t a u r o l i t e m o d e l s , t P l x p . : ( t r 2 t F e l n ) . , w h e r e n : r r , l ( F e + M g + Z n + M+n A
+ lL i+ F e 3 * ) .

octahedral layers above that of the Tschermak substitu- cordierite with garnet using the Ganguly and Saxena
tion (Dyar et al., 1993). (1984) garnet mixing model. Much of this nonideality
The staurolitemole fraction models and end-member disappears if more recent gamet mixing and cordierite
formulas (Table 4) result in significantpart from the ef- hydration data are incorporated into the approach. Fe-
fortsofHawthorneet al. (1993a,1993b,1993c).The first Mg end-membersof biotite may also show some nonide-
three positionsin the formula are coupledas demonstrat- ality, which is not yet well understood.Hoisch (1991)has
ed by their work. The best way to model a 2H staurolite shown from natural occuffencesthat Ti and Al substi-
end-memberis with the mole fraction of vacancyon M4 tutions in biotite are nonideal solid solutions, and Cir-
(formerly U), with an additional correction for dilution coneand Navrotsky (1992)have confirmedHoisch'scon-
ef tarFsz+by other ions on T2, and the best way to model clusion for phlogopite-eastonite.Examples of probable
a 6H stauroliteend-memberis with the mole fraction of nonideality betweenmajor end-membersthat have com-
t6lFe2+ on M4. Even simple FASH staurolitesamplesre- monly been overlooked include iron-magnesium stauro-
sulting from experimentalstudiescontain nonstoichiom- lite and hematite-ilmenite.
etric amountsof Fe2+-H+solid solution (Holdawayet al., Even though there is some question regardingthe strict
l 993). applicability of the Margules approach, it seems to be
For cordierite, there has been a problem concerningthe useful in most cases,and it is far better than assuming
correct HrO content for the fully hydrous end-member. ideality. There appear to be severalinconsistenciesin the
Carey and Navrotsky (1992) have shown that the cordi- literature regarding the nature of Margules equations for
erite-hydrous cordierite solid solution fits an ideal one- multicomponent systems.Mukhopadhyay et al. (1993)
site model (Table 4). have derived all the equations for an excessparameter
These suggestionsshould make it clear that mole frac- such as Gxs and for RZ ln 7 in symmetrical and asym-
tion models for many of these minerals will continue to metrical n -component systemsfrom first principles using
evolve as additional information comes to light. a Taylor seriesexpansion. From a strictly mathematical
standpoint, ifthe constituent binaries in a ternary system
NoNrrn.l,lrrY rN MINERALS are symmetrical, there are no ternary interaction param-
There is a tendencyfor petrologiststo believe that many eters. Both Helffrich and Wood (1989) and Mukhopa-
minerals may be modeled as ideal solid solutions, when dhyay et al. (1993)showedthat, for a ternaryasymmetric
in fact this is commonly not the case.In our view, ideality solution, if it becomesnecessaryto assume that ternary
is merely a special caseof nonideality, and some degree interaction parameters are zero for want of better infor-
of nonideality is common in most solid solutions. When mation, we must be careful to use formulations that in-
one seesindications of a solvus or irregular partitioning clude all of the binary interactions. For an asymmetric
of major ions betweenmineral pairs, it is likely that there regular ternary solution, G*s is given by
is nonideality. In our opinion, when there are no exper-
imental data, it is better to estimate the nonideality from G": X,Xr(XrWn + XtWr,) -l XI3(X3WB + XtW3t)
natural data than to assume ideality. A petrochemical
analysis by Guidotti et al. (in preparation) may provide + x2x3(x3wn + x2wrr) -f xtx2x|cD3
for the best muscovite-paragonitesolvus information and where
shows clearly that the solvus is affectedby phengite con-
tent; the Margules parametersfor the 12-fold site in mus- Cr.r: t/z(Wr,+ W, + WB + W, + Wn + Wi2)
covite should be corrected for the new muscovite-parag- - wrrr.
onite data and for the effect of phengite component.
Bhattacharyaet al. (1988) showedthat Fe-Mg mixing in It is far better to assumeWrz.:0 than to assumeC,r. :
cordierite is nonideal with computations basedon natural 0 becauseC,r, is significantly basedon the binary param-
686 HOLDAWAY AND MUKHOPADHYAY: GEOTHERMOBAROMETRY IN PELITES

TreLe5, Recommended Margulesparameters

for garnet,per than most Zbased on more sophisticatedthermodynam-
threecations ic data bases.This is becausethe thermometry is little
lnteraction WH(kJl ws (J/K) W, (Jlbatl'
affected by other intensive variables, and nonideality is
Source.*
corrected.
Mg-Fe 6350 0 0.020 1
Fe-Mg 2080 0 -0.030 1
Mg-Ca 15000 eo Gnon-q,norrnrnns
o.'t77 2
Ca-Mg 56s20 2.1 0.0s0 Severalgeobarometershave met with considerablesuc-
Fe-Ca 7770 0 0.099 3
Ca-Fe -3270 0 0.064 3 cess(Table 6). Most have the advantage of being reac-
Mg-Mn 0 0 0? 4,5 tions that involve no fluid component, which means that
Fe-Mn 0 0 o? 4,6 P rnay be estimated independently of XH2o.Almandine
Ca-Mn 0 0 0 3,4
Mg-Fe-Ca o? 0? 0 182 * muscovite + biotite * aluminum silicate has potential
- Values determined becauseit involves only major components,but it sufers
by statistical analysis of synthetic garnet molar
volumesusingdata from Haseltonand Newton(1980),Geigeret al. (1987), from low A,V and AS (Essene,1989).Almandine * cor-
Wood (1988),Kozioland Newton(1989),and Koziol(1990).
.'Sources tot WH,
dierite + sillimanite also has potential, mainly for high-
and l\ls: 1 : Hackler and Wood (1989); 2 : Wood grade rocks, but does involve HrO. Ultimately, both P
( 1 9 8 8 ) ; 3 : K o z i o(l 1 9 9 0 ) ; 4 : B e r m a n( 1 9 9 0 ) ; 5 : W o o d( 1 9 9 1 )a; n d 6
: o'Neiil et ar. fi989). and l'can be determined with this assemblage,providing
something is known about Xg,o. We are in the processof
revising this geobarometer.The presenceof garnet in all
of thesegeobarometersemphasizesthe importance of re-
eters. Comparison shows that although the equations of solving garnet Margules parameters and the problem of
various authors appear to be quite different, most are the obtaining P from garnet-absentpelites. Even with signif-
sameand are. with lew exceptions,correct. icant improvements in thermodynamic data, mole frac-
The garnetsrepresentan example of a four-component tion models, and Margules parameters, some of these
system that has been extensively studied for the past de- geobarometersmay never be more accuratethan +0.5
cade.The data used are from both experimental and nat- kbar.
ural occurrences.We have carefully reviewed the avail-
able resultsand arrived at what is probably the best overall TrmnrronvuAMlc DATA AND DATA BASES
set of Margules parametersfor determining Gxs and t'xs The most important singleaspectof geothermobarom-
(Table 5). One of the problemsthat arisesin this type of etry is the quality of the thermodynamic data on which
analysis is the failure to adequately account for error in P-7" results are based. These data come from a combi-
the data. This is important, as sometimeserrors are great- nation ofthree sources:(1) calorimetricdata, (2) experi-
er in magnitude than the mean parameter because (l) mental data, and (3) natural occurrences.Ideally, we would
experimental data were collected without proper design Iike to have agreementbetween these sources,but that is
in the statistical sense,and (2) the number of data points not common. The following cautions should be noted for
is not statistically large enoughto put tight constraints on deriving thermodynamic data: (l) Calorimetric enthalpy
the amounts of the errors.As a result, in some instances (II) data derived from solution calorimetry tend to have
more complicated expressionshave been advocatedthan errors larger than what can be obtained from the analysis
are warranted by the data. of good experimental data. On the other hand, calori-
metric entropy (,S),specific heat (C"), and X-ray diffrac-
Excn.c.Ncp GnorunnnroMETERS tion molar volume (V) data are likely to be more precise
Exchange geothermometers are especially useful for than H. (2) Methods of experimental studies should be
medium-grade pelites becausethey have low sensitivity checked carefully to be sure that the experiments repre-
to P and no sensitivity to XHro. However, at granulite sent true reversalsof stable alternative assemblagesand
grade they are easily reset and may have serious difficul- that experimental starting materials and products have
ties. Although it still has some problems, the garnet-bi- been characterizedas well as possible. (3) Each reversal
otite geothermometer,corrected for at least garnet non- should be independently judged, and conditions should
ideality, is by far the bestexchangegeothermometer.Ferry be extended away from equilibrium to the limit of ex-
and Spear(1978)emphasizedthe kinds ofrocks bestsuit- perimental error. We have found this to be especially
ed and the error limitations of the method. For best re- important with the aluminum silicates,as seenbelow. (4)
sults, graphite-bearingrocks should be used, and Fe3+in Comparison with natural assemblagesshould be a sec-
both minerals should be accounted for, as done by Wil- ondary processand should be done with extreme caution.
liams and Grambling (1990). Where garnet is present in (5) If natural assemblagesare to be used for calibration,
medium-grade pelites, garnet-biotite geothermometry correctionsshould be made for mineral composition, Fe3+,
approachesan error of +25 "C under reducing conditions and fluid composition, where appropriate. The presence
and a high biotite/garnet volume ratio. At high grades, of graphite reduces the X",o in fluid to 900/oor less, de-
or under conditions ofresetting during cooling, accuracy pending on P-T conditions. It is even possible that fluid
is considerably worse. In appropriate rocks, I based on in micaceouspelitic schistsescapesmore rapidly than in
garnet-biotite geothermometryis probably more accurate quartzites,and so P,".in micaceousrocks would be slight-
 HOLDAWAY AND MUKHOPADHYAY: GEOTHERMOBAROMETRY IN PELITES 687

TABLE6. Some geobarometersfor peliticrocks.

Geobarometer(from high- to low-P) Reference Comment

Almandine-rutile-ilmenite-aluminum silicate Bohlenet al. (1983) very good

Almandine-grossular-rutile-ilmenite-anorthite Bohlenand Uotta (1986) very good
Grossular-anorthite-aluminum silicate Newtonand Haselton(1981)-' very good
Pyrope-grossular-muscovite-anorthite-phlogopite Ghentand Stout (1981)f gooo
Almandine-grossular-muscovite-anorthite-annite Ghentand Stout (1981)t good
Almandine-muscovite-annite-aluminum silicate McMullinet al. (1991) good
Almandine-cordierite-aluminumsilicate-H"O Holdawayand Lee (1977) needswork
'Listed by end-membercomponentsfor use in quartz-bearingrocks.
-- Also Kozioland Newton(1988).
t Also Hoisch(1990).

ly lower than that in adjacent quartzites, as suggestedby
Holdaway and Goodge(1990).
There are a number of possiblenaturally occulTingpe-
litic assemblageswhose presenceor absenceshould serve
as a test of the quality of thermodynamic data bases,
provided correctionsare made for mineral and fluid com-
position. These include (l) biotite * andalusite + mus-
covite + quartz, a common assemblagethat should not
occur, according to calculations based on the published
version of the Holland and Powell (1990) data base (see
also Powell and Holland, 1990); (2) cordierite + stauro-
lite + muscovite + quartz, which does not occur in na-
ture except in apparent disequilibrium or perhaps when
stabilized by high Zn orLi in staurolite; (3) cordierite +
garnet + muscovite + quartz, which can only occur if
the garnet is stabilized by high Mn; (4) chloritoid + sil-
limanite + qvarlz, which has a narrow stability range in
graphite-free, Fe-Al-rich rocks over a range of fo,; and
(5) paragonite + sillimanite * quartz, which has a naffow
stability range in Na-Al-rich rocks. A necessarybut not
sufficient test of a thermodynamic data base is whether
these conditions are met.
A problem with thermodynamic data basesin general
is that, by their mathematical nature, they are self-con-
sistent. Therefore, they tend to produce self-consistency
among various reactions in a rock or group of related
able reversalsand half-reversalsare totally consistent for
eachofthe three phaseboundaries(e.g.,Fig. 2) and con-
sistentwith a triple point at 504 + 20 oC,3.75 + 0.25
kbar. We also evaluated all the available molar volume
data; our best estimates are given in Table 7. Kyanite,
with only two determinations, needs more carefully de-
termined unit-cell data on low-Fe3* samples.Using these
results and the calorimetric data of Robie and Heming-
way (1984)and Hemingway etal. (1991),we have refined
H and S (Table 7). These values are recommended for
aluminum silicate thermodynamic data.
The iron cordierite stability results of Holdaway and
l-ee (1977),Weisbrod (1973),and Richardson(1968) have
been criticized on theoretical grounds. We now tealize
that in our work, unstable hercynite in the synthesized
cordierite and almandine favored growth of metastable
Sillimanite
6
5
(I)

rocks. This self-consistencydoes not necessarily mean

32.
that the data base is accurate. It may very well be dis- c)
placedin P and Ifrom the correctposition or be incon- L

sistent with P and T determined from other bulk com-

positions and reactions. 1.0
Andalusite
An advantageof data basesis that retrieving ,FIand S
of minerals from experiments and from natural occur-
rencestends to correct for disorder and nonidealities that 0.0
may not be easily correctedwith direct calorimetric mea- 450 650 750 850
surementson stoichiometric end-members.Examplesare Temperature( oC)
end-member disorder, minor amounts of solid solution
such as l2-fold-site H or vacancy in micas, and partial Frg.2. Pressure-temperaturephase diagram for the andalu-
site-sillimanite reaction showing all valid reversalsand half-re-
replacementof Si by Al in staurolite.
versalsadjustedto their error limits away from the phasebound-
Many of the important pelitic reactions used in data
ary. Closed symbols : andalusite-stable half-reversals; open
basesdependcritically on aluminum silicate and, to some symbols : sillimanite-stable half-reversals.This boundary and
extent, on cordierite equilibria. Using studies since 1966, similarly determined boundaries for kyanite * andalusite and
Holdaway and Mukhopadhyay (1993) have carefully kyanite + sillimanite are consistent with the thermochemical
evaluated all experimental work on aluminum silicates. data given in Table 7. After Holdaway and Mukhopadhyay
If each reversal is judged independently, all the accept- (l 993).
688 HOLDAWAY AND MUKHOPADHYAY: GEOTHERMOBAROMETRY IN PELITES
TABLE7. Thermodynamicdata for aluminumsilicates.
Molar volume Entropy Enthalpy
(J/ba4 (J/mol K) (kJ/mot)
-2593.70(300)
Kyanite 4.408(21 82.86(50)
Andalusite 5.146(2) s1 60(s2) -258s.66(300)
Sillimanite 4.984(2) 9s.08(52) 2586.37(300)
'Holdaway
and Mukhopadhyay(1993),based on analysisof experi-
mental reversaldata since 1966 and on calorimetricdata of Robie and
Hemingway(1984)and Hemingwayet at (1991).
cordierite, especially at lower Z. Mukhopadhyay et al.
(1991, in preparation)have recently redeterminedthis
equilibrium (Fig. 3) and found it to have a positive slope.
At 600 "C, equilibrium is 1.7 kbar lower than in previous
determinations. This now will form the basis of a recal-
ibrated cordierite-almandine-sillimanite geobarometer.
In a further effort to constrain thermodynamic data for
pelitic minerals, we have retrieved thermodynamic data
for 2H and 6H stoichiometric iron staurolite end-mem-
bers from experimental reversalresults. One major prob-
lem with staurolite is that neither natural nor synthetic
staurolite is ever stoichiometric (Holdaway et al., 1991,
1993). Even simple FASH staurolite has solid solution
on as many as six independent sites. Thermodynamic
data reported previously have no real validity for the cal-
culation of staurolite stability relations becausethey fail
to take into account the Fe2+-H+ substitution (Haw-
thorne et al., 1993a),somewhatanalogousto that of hy-
drogarnet. Our first step was to conduct multiple linear
regressionson good quality unit-cell data of completely
analyzednatural staurolite to estimate molar volumes for
the two FASH end-members(Holdaway et al., 1993).Two
models, one holding Si and Al constant at averagevalues
and usingan eight-component(Fe,*, Mg, Zn,Li, Mn, Ti,
Fe3*, tr) fictive end-member approach (Fig. 4), and one
using all ten elements(Si, Al, Fe2*, Mg, Zn, Li, Mn, Ti,
Fe3+,H) agreeand allow estimation of molar volumes
for stoichiometric end-members:HrFe.,Al,8Si8Oo8 : 44.61
+ 0.03 J/bar,HuFe,Al,sSi8O18 :44.17 + 0.03J/bar.We
have fitted experimental data in the FASH system to es-
timate thermodynamic data for 2H and 6H staurolite,
allowing H+ content of staurolite to vary as a function of
P and ?"(Fig. 5). The P-Zfield of staurolite + quartz may
be contoured on the basis of the H* content of staurolite,
dependingon whether staurolite coexistswith almandine
or aluminum silicate.Points wherethesetwo setsof curves
intersect for a given H+ content define the limits of stau-
rolite + quartz stability. This approach does not explain
the failure of agreementbetween experimental and nat-
ural staurolite P-T estimates (e.g., Pigage and Green-
wood, 1982;Holdaway et al., I 988).A future projectwill
attempt to show whether or not staurolite T2 sitesexhibit
strongly nonideal behavior, as suggestedby Mukhopa-
dhyay et al. (1990).
There are four main thermodynamic data bases: (l)
Helgesonet al. ( I 978),(2) Berman ( I 988, I 99 I ) and Brown
et al. (1988) (Ge0-Calc,most recently updated in Janu-
Weisbrod(1973)
\
.v Holdaway and lre
(r977)
I
o-<a
{\
Mukhopadhyayand Holdaway, ms,
550 600 650 700 1so 800 850
Temperature, degrees C
Fig. 3. Experimentaliron cordieritestabilitydiagramdeter-
minedby Mukhopadhyay. Closedsymbols: iron cordieritesta-
ble; opensymbols: almandine,sillimanite,quartz,and H.O
stable.Previouslydeterminediron cordieritephaseboundaries
aregivenfor comparison. AfterMukhopadhyay et al. (1991).
ary, 1992),(3) Spearand Cheney(1989),and (4) Powell
and Holland (1988) and Holland and Powell (1990)
(Thermocalc,most recentlyupdatedin September,1992).
Uncritical use of thesedata basesmay lead to substantial
errors for the following reasons:(l) Various reactions oc-
curring in a single rock will tend to plot in a self-consis-
tent fashion but may be incorrect becauseof systematic
errors in the thermodynamic data. (2) Many P-Z slopes
are positive, and so intersections may have large errors.
(3) X"r" may have been different than that assumed,shift-
ing dehydration curves to various extents. (4) Failure to
properly account for nonideality may have produced er-
rors. (5) Added weight may not have been given to the
best-known reactions or thermodynamic data. We rec-
ommend extreme caution in using these data bases.The
recently developedTWEEQU method (Berman, 1991),
which draws a P-T line for every possible end-member
reaction occurring in a rock, can also be misleading if
equal weight is given to each line and ifthe sole criterion
for the quality of thermodynamic data is a tight cluster
of intersections.
We have also observedthat in some studies that apply
data bases,not enough attention is paid to field relations,
petrology, and geologichistory. In such studies,the treat-
ment of the geologic environment is so brief as to be
misleading.Two examplesfrom Maine illustrate our con-
cern: A recent P-f study of samples from an area near
Augusta (Berman, 1991, p. 846) stated that the region
suffereda single metamorphic event, when in fact papers
on the area (Novak and Holdaway, 1981; Holdaway et
al., 1982, 1988) demonstrated that there have beert at
least two events. The analysis shows that the high-grade
rocks formed at higher P and appears to show that P
increaseswith grade. The medium-grade metamorphic
 HOLDAWAY AND MUKHOPADHYAY: GEOTHERMOBAROMETRY IN PELITES 689

Kyanite
+
741
,-t
' /-
..-'
y^ ,,--
a Almmdine
,.,-- +
'tt't
/ Quartz
"",

E740
6
t--t--1
6
o ?39

7381(
734 ?39 740 741 742
Measured Volume

Fig. 4. Plot of measured molarvolumefor 22

vs. calculated
staurolitespecimens modelin
usinga multiplelinearregression Sillimanite
which compositionalvariablesare fictive end-members Fez*,
to have Almandine
Mg,Zn, Li, Mn, Ti, Fer+,andfl. Si andAl areassumed
+
average values.AfterHoldawayet al. (1993).
Quartz

4s0 500 t:o 600 6so ?oo

events (producing andalusite and sillimanite) are Acadi-
an, and the high-grademetamorphism (locally producing
kyanite) is late Paleozoic,and the region was more deeply
buried at that time. In an analysis of rocks by Holdaway
et al. (1988) in an area between Augusta and Rangeley
(McMullin et a1.,199I , p. 901),later plutonism and post-
metamorphic folding are advocated as causesfor P vari-
ation. In fact, the area suffered posttectonic metamor-
phism, and the observed sharp reentrants in isogradsare
the result of overlapping nonsynchronous adjacent gra-
nitic heat sources(Holdaway et al., 1982, 1988).All the
metamorphism is synchronous with plutonic intrusion,
and intrusion occurredover at least 75 m.y., during which
time P increasedsignificantly. In both of the casescited
above, P increasedfrom north to south and with decreas-
ing age of the metamorphism.
The application of a data baseto only a very few spec-
imens from an areais of concernto us. Thesespecimens
may not be representativein their degreesof equilibra-
tion, or they might representlimiting values of the P and
T rangeor mineral compositions.
As an example of the kinds of problems that may oc-
cur, we use a contact metamorphic assemblagefrom
western Maine at the north end of the Lexington batho-
lith, studiedby Dickersonand Holdaway (1989).The as-
femperature (oC)
Fig. 5. Phasediagramfor FASH staurolitewith variableH
from 2.8 to 4.4 coexistingwith quartz.Heavylines showalu-
minum silicatestabilityrelationsandthelimit of stauolite-quartz
stability.SolidcontoursshowH contentof staurolitecoexisting
with aluminum silicate,dashedcontoursshow H contentof
staurolitecoexistingwith almandine.Limit of staurolite-quartz
stabilityoccurswherecontoursmeetfor a givenH content.After
Holdawayet al. (in preparation).
semblage,seenin several specimens,is andalusite + sil-
limanite + muscovite * potassium feldspar * cordierite
+ biotite + quartz * graphite, lying on Carmichael's
(197S)bathograd.The first occurrenceof sillimanite and
the first occuffence of sillimanite * potassium feldspar
consistentlyare found in the sameplacein the field. Dick-
ersonand Holdaway (1989)estimatedmetamorphiccon-
ditions of 2.35 kbar, 600 "C, using three independentex-
perimental P-I reactions.One problem with this locality
is the total absenceof garnet, the result of low P. The
Berman data base(1988, updated in 1992)gave similar
P-I conditions (Table 8). Berman's unpublished cordi-
erite data do not account for HrO and are provisional.
His iron cordierite data arebasedon experimental results

Trele 8. Estimatesof P-f conditionsof formationfor a rock from Maine"

P
(kba4 rfc) Details

235 600 Dickersonand Holdaway(1989) three experimentalreactions,activity-corrected

2.45 600 Berman(1988,updated1992) includingiron cordierite,excluding magnesiumcordierite
2.25 595 Berman(1988,updated1992) excludingall cordierite,aluminumsilicate' muscoviteonly
3.00 635 Hollandand Powell(1990,updated1992) includingiron cordierite and magnesiumcordierite
. Mineralassemblage: andalusite+ sillimanite+ muscovite+ potassium feldspar + cordierite + biotite + quartz + graphite' Locality: north
lobe of Lexingtonbatholith,Maine(Dickersonand Holdaway,1989)
 690 HOLDAWAY AND MUKHOPADHYAY:
FeO Fe3Oa F er O 3
Fig.6. FeO-FerO.-TiO, diagramshowingpossiblehematite-
ilmenitephaserelationsat about 600 .C. The dashedline is a
schematicO, isobar.
ofMukhopadhyay et al. (1991) and can be expectedto
be most accurate close to the experimental P-T condi-
tions. The Holland and Powell data base(1990, updated
in 1992) gave conditions ar higher P and T (Table 8).
Cordierite data in this data baseaccount for HrO in cor-
dierite but have not been adjusted to the new iron cor-
dierite experimental results. In our opinion, the main
problem with the Holland and Powell data base is that
their thermodynamic data for aluminum silicare are nor
based on a careful analysis of all available experimental
data (e.g.,seeHoldaway and Mukhopadhyay, 1993).
For the same assemblagein the Ballachulish aureole,
Pattison (1992) deduced condirions of 3 kbar, 645 "C,
using, in part, thermodynamic data for cordierite. Two
differences between the Lexington and Ballachulish lo-
calities are that in most specimensthe Ballachulish as-
semblagedoes not contain graphite and that in several
specimensandalusite appearsto have been stabilized by
high Fe3+.Pattison assumedthe fluid was pure HrO. We
do not believe theseconditions can be defendedas a point
on the andalusite-sillimanite P-Zboundary, as advocated
by Pattison(1992).
In summary, much more careful work is neededbefore
petrologistscan dependon the useof thermodynamic data
basesfor P and T estimation. Eventually one should be
able simply to use the data bases,with appropriate mole
fraction models and corrections for nonideality, to eval-
uate P from the various geobarometersgiven in Table 6.
Until that time, thesedata basesshould not be usedwith-
out careful critical examination of the results.
Fr,urn coMposrrroN
In sulfide-absentmedium-grade pelitic rocks, the pre-
dominant fluid speciesare H.O and COr, and, where
graphite is present,CHo as well. Qualitative argumentsto
indicate the approximate composition of pelitic fluids in-
GEOTHERMOBAROMETRY IN PELITES
clude the following: (l) In many medium-grade pelitic
terranes,coexisting thin metacarbonateunits commonly
exhibit minerals like grossularor epidote and sometimes
contain no carbonate, indicating high Xsro even in the
metacarbonateunits. (2) Many metamorphic terraneshave
granitic heat sources,which are by nature HrO-rich. (3)
The dominant prograde HrO-consuming mass-transfer
reaction in most graphitic pelites is the reaction of HrO
with graphite C + HrO : CO, + CH4 and when this
reaction, combined with simple dehydration reactions,
controls fluid composition, it controls it near 800/oHrO,
100/oCOr, 100/oCHo, and at,fo, values near the QFM
buffer, depending on P and T, as shown by Ohmoto and
Kerrick (1977). (a) This observation is confirmed by the
highly reducing nature of most graphitic schists,with no
hematite component in the ilmenite. We can take advan-
tage of this by assuming that the graphite * HrO equi-
librium dominates, with either a closed system or one
partially open to fluid movement, which allows X.o, to
be equal to or somewhat greater than X."r. Using this
method, Holdaway et al. (1988)assumedX.o, was up to
ten times X.r". This approach enablesus to estimate the
composition of pelitic fluids in graphitic rocks. It will not
work well if the graphitic rocks are at all oxidizing, with
hematite component in the ilmenite, or if the pelites are
deepand undergoinganatexis.The accuracyofthe results
also dependson the accuracyof the thermodynamic data
and mixing models for fluids. Great strides have recently
beenmade concerningthermodynamic data for fluids and
fluid solutions, especiallyby Shi and Saxena(1992) and
by Shi ( I 992).At presentthe above approachmay be the
bestway to estimatefluid compositionin appropriatepe-
lites.
This approach can be strengthenedby an accurate,in-
dependentdetermination of /o, becauseall fluid compo-
sitional variables in graphite schists are uniquely deter-
mined by knowing P, T, and fo,, and the assumption of
subequal Xro, and X.ro becomes unnecessary.We are
hoping to determine the hematite-ilmenite solvus exper-
imentally. The schematic plot of Figure 6 incorporates
what we know from petrologic observationsat about 600
oC and illustrates that the system is nonideal and has a
solvus. The dashed line is a schematic O, isobar, which
shows that for a constant composition of ilmenite, /o,
varies with the nature of the coexisting phase. Experi-
mental determination of this system should enable one
to determine fo, for the ilmenite * rutile equilibrium,
2FerOr..,"u- + 4TiO, : 4FeTiOr.. + Or, or the ilmenite
+ magnetiteequilibrium, 6FerO,ssinirm: 4FerOo* Or,
in rocks, and to estimate a limited rangeof fo, for ilmen-
ite-only assemblagesby constraining f, from the ilmen-
ite + rutile and ilmenite + magnetite equilibria.
The alternative approach for determining fluid com-
position is to use geothermometersfor ?nestimates,use
fluid-absent geobarometersfor P estimates,and, with the
estimated T and P, use dehydration reactionsto estimate
Xrro from the compositions of coexistingphases.In some
casesthe results of such efforts indicate Xn,o values far
 HOLDAWAY AND MUKHOPADHYAY:
from the predicted rangeof70-900/oin sulfide-absentgra-
phitic pelites. That might be taken to indicate that such
metamorphic rocks have a wide range of fluid composi-
tions. We believe it commonly indicates a compounding
of errors in T, P, thermodynamic data, mole fraction
model, and corrections for nonideality.
In our opinion, when we reach the point where various
dehydration equilibria in graphitic schistsindependently
and consistently give us reasonablevalues of Xr,o, then
we will be making real progressin geothermobarometry
and homing in on the best thermodynamic data for all
the minerals. Knowledge of fluid compositions can then
serve as a check on the quality of the geothermobarome-
try and the thermodynamic data. At that stage,we will
be able to say much more about the nature of fluids in
pelitic rocks and perhaps test models for fluid behavior
and movement.
Suunrlnv AND RECOMMENDATIONS
In summary, we recommend the following: (l) We
should use all information available to determine the best
possiblestoichiometryand site assignments.(2) We should
carefully think out mole fraction models and continue to
experiment with and test them. (3) Nonideality should
be evaluatedwheneverpossible.(4) Thermodynamic data
basesshould be continually reevaluated,should be treat-
ed with some degreeof skepticism,and should be applied
with care. (5) Fluid reactions should be used to test and
refine our methods.(6) Consistentwith the times, we must
be willing to accept change.
As we home in on the best activity models, values for
thermodynamic data, and corrections for nonideality, it
'C,
should be possible to approach accuraciesof +25
+250 bars, and + l0o/oXr.oin low- to medium-P rocks.
AcxNowr-nucMENTs
We acknowledgewith thanks the following people for allowing use of
M. Darby Dyar, Charles
their unpublisheddata and for helpful drscussions:
Steven
Guidotti, Frank Hawthorne, Barbara Dutrow, Klaus Langer, and
Guggenheim. M Darby Dyar provided a thoughtful review. We thank
Rob Berman for making the latest version of his Ge0-Calc data base
(Jan92.RGB)available to us, and Roger Powell for running the lrxington
research
batholith data on Thermocalc. This researchis supported by
funds from the National ScienceFoundation, grants EAR-89-04777 and
EAR-90-r9277.
This work is dedicated to the memory of Patricia Piercy Holdaway,
t937-199t.
RnrnnnNcss crrED
Ackermann, L., Langer, K., and Rieder, M. (1993) Germanium musco-
vites with excesshydroxyl water, KAl,[Ge3 ,Al'*-,O'o ,(OH),(OH),]
and the questionofexcessOH in natural muscovites.EuropeanJournal
of Mineralogy, in press.
Berman, R.G (1988) Internally-consistentthermodynamic data for min-
erals in the systemNa'O-K,O-CaO-MgO-FeO-FerO,-Al'O,-SiOr-TiO,-
H,O-COr. Journal of Petrology, 29, 445-522
-(1990) Mixing properties of Ca-Mg-Fe-Mn garnets. Amencan
Mineralogist.75. 328-344.
- (1991) Thermobarometry using multi-equilibrium calculations:A
new technique,with petrologrcalapplications. Canadian Mineralogist,
29,833-855.
Bhattacharya, A, Mazumdar,A.C., and Sen,S.K. (1988)Fe-Mg mixing
GEOTHERMOBAROMETRY IN PELITES 691
in cordierite: Constraints from natural data and implications for cor-
dierite-garnet geothermometry in granulites. American Mineralogist,
73,338-344
Bohlen.S.R.,and Liotta, J.J.(1986)A barometerfor gametamphibolites
and gamet granulites.Journal of Petrology, 27, 1025- 1056.
Bohlen,S R., Wall, v J., and Boettcher,A.L. (1983)Experimentalinves-
tigations and geologicalapplications of equilibna in the system FeO-
TiO,-Al.O.-SiO,-H.O. AmericanMineralogrst,68, I 049- 1058.
Brown.T.H . Berman,R.G., and Perkins,E H' (1988)Ge0-Calc:Software
packagefor calculation and display of pressure-temperature-composr-
tion phasediagramsusing an IBM or compatible computer. Computers
and Geosciences, 14, 279-289.
Carey,J.W., and Navrotsky, A. (1992)The molar enthalpy of dehydration
of cordierite. American Mineralogist, 77, 930-936.
Carmichael. D.M (1978) Metamorphic bathozones and bathograds: A
measure of the depth of post-metamorphic uplift and erosion on a
regional scale.American Journal of Science,278, 769-7 97.
Circone, S., and Navrotsky, A. (1992) Substitution of164rAlin phlogopite:
Hrgh-temperaturesolution calorimetry, heat capacities,and thermo-
dynamic propertiesof the phlogopite-eastonitejoin. American Miner-
alogist,77, I 19 l- 1205.
Dickerson, R P., and Holdaway, M.J (1989) Acadian metamorphism as-
sociated with the Lexington batholith, Bingham, Maine American
Journal of Science,289, 9 45-974
Dutrow, B.L , Holdaway,M.J., and Hinton, R.W (1986)Lithium in stau-
rolite and its petrologic significance'Contributions to Mineralogy and
Petrology,94,496-506
Dyar, M.D. (1990) Mdssbauerspectraof biotite from metapelites.Amer-
ican Mineralogist,'l 5, 656-666.
- ( 1993)M6ssbauerspectroscopyof tetrahedralFer* in trioctahedral
micas- DiscussionAmericanMineralogist,78, 665-668.
Dyar, M D , and Guidotti, C.V. ( 1992)Controls on ferric iron in chlorite
Geological Society of America Abstracts with Programs,24, 130'
Dyar, M D, Perry,C.L, Rebbert,C R, Dutrow, B.L, Holdaway,M'J',
and l-ang, H.M. (1991) Mijssbauer spectroscopyof synthetic and nat-
urally occurring staurolite American Mineralogst, 76' 27-41'
Dyar, M.D, Guidotti, C V, Meadows,E', and Robertson'J.D (1992)
PIGE analysesof F and Li in muscovite and biotite from metapelites
and associatedgranites of W. Maine Abstracts to AGU Fall Meeting,
Supplementto Eos,618
Dyar, MD., Guidotti, C.V., Holdaway, M.J., and Colucci, M (1993)
Nonstoichiometnc hydrogen contents in common rock-forming hy-
droxyl silicates.Geochimica et Cosmochimica Acta,57, in press.
Dymoke, P, and Sandiford, M (1992) Phase relationships in Buchan
facies seriespelitic assemblages:Calculations with application to an-
dalusite-stauroliteparageneses in the Mount Lofty Ranges,South Aus-
tralia. Contributions to Mineralogy and Petrology, ll0, lzl-132'
Essene,E.J (1989) The current status of thermobarometry in metamor-
pbic rocks. In J.S. Daly, R A. Clitr, and B W.D Yardley, Eds., Evolu-
tion of metamorphic belts, p. 1-44 Geological Society Special Publi-
cation, Cambridge,England.
Ferry, J.M., and Dipple, G.M (1991)Fluid flow, mineral reactions,and
metasomatismGeology,19, 2ll-214.
Ferry,J.M, and Spear,FS. (1978)Experimentalcalibrationofthe par-
titioning of Fe and Mg between biotite and garnet. Contributions to
Mineralogyand Petrology,66, I l3-l 17.
Frost, B R., and Tracy, R J (l 99 l) P-f paths from zoned garnets:Some
minimum criteria American Journal of Science,29 1' 9 17-939.
Ganguly, J., and Saxena,S.K. (1984)Mixing propertiesof aluminosilicate
gamets:Constraints from natural and experimental data, and applica-
tions to geothermobarometry.American Mineralogrst, 69' 88-97.
Geiger,C.A., Newton, R.C , and Kleppa,O.J. (198?)Enthalpvof mixing
of synthetic almandine-grossularand almandine-pyropegamets from
high-temperaturesolution calorimetry. Geochimica et Cosmochimica
Acta,51, 1755-1763.
Ghent. E.D.. and Stout, M.Z. (1981) Geobarometryand geothermometry
of plagioclase-biotite-gamet-muscovite assemblages.Contributions to
Mineralogy and Petrology, 76, 92-97.
Gray,D.R., Gregory,R.T., and Durney,D.W. (1991)Rock-buffered fluid-
rock interaction in deformed quartz-rich turbidite sequences,eastern
Australia. Journal of GeophysicalResearch' 96' 19681-19704.
 692 HOLDAWAY AND MUKHOPADHYAY
Guidotti, C.V. (1984) Micas in metamorphic rocks. In Mineralogical So-
ciety of America Reviews in Mineralogy, 13, 357-467.
Guidotti, C.V., and Dyar, M.D. (1991)Ferriciron in metamorphrcbiotite
and its petrologrcand crystallochemicalimplications. American Min-
e r a l o g l S t/ ,t r , l O l - l / )
Hackler, R.T., and Wood, B.J (1989) Experimental determination of Fe
and Mg exchangebetweengarnet and olivine and estimation of Fe-Mg
mixing propertiesin gamet. American Mineralogist, j4,994-999.
Haselton,H.T., and Newton, R.C (1980)Thermodynamicsof pyrope-
grossular garnets and their stabilities at high temperaturesand high
pressures.Journal of GeophysicalResearch, 85, 69'73-6982.
Hawthorne, F.C., Ungaretti, L, Oberti, R , Caucia, F., and Callegari,A.
( I 993a)The crystal chemistry of staurolite I. Crystal structure and site
occupanciesCanadian Mineralogist, 31, in press.
- (1993b) The crystal chemistry ofstaurolite. II. Order-disorderand
the monoclinic ' orthorhombic phase transition. Canadian Mineral-
ogist,31, in press
-(1993c) The crystal chemistry of staurolite. III. Local order and
chemical composition. Canadian Mineralogist, 31, in press.
Helffrich, G., and Wood, B.J. (1989) Subregularmodel for multicompo-
nent solutions.AmericanMineralogist,74, l0l6-1022.
Helgeson,H.C , Delany,J.M., Nesbit, H W., and Bird, D K. 097g) Sum-
mary and critique of the thermodynamic properties of rock-forming
minerals. American Joumal of Science,278A, 229 p.
Hemingway,B.S, Robie, R.A., Evans,H.T., and Kenick. D.M_ (l991)
Heat capacitiesand entropies of sillimanite, fibrolite, andalusite,kya-
nrte, and quartz and the AlrSiO, phasediagram American Mineralo-
g i s t ,7 6 , 1 5 9 7 - 1 6 1 3 .
Hoisch, T.D (1990) Empirical calibration of six geobarometersfor the
mineral assemblageqlartz + muscovite + biotite + plagioclase *
garnet Contnbutions to Mineralogy and petrology, 104,225-234.
-(1991) Equilibria wrthin the mineral assemblage quanz + mus-
covite + biotite + gamet + plagioclase,and implications for the mix-
ing properties of octahedrally-coordinatedcations in muscovite and
biotite. Contributions to Mineralogy and petrology, 108, 43-54
Holdaway,M.J, and Goodge,J W. (t990) Rock pressurevs. fluid pres-
sure as a controlling influence on mineral stability: An example from
New Mexico. AmericanMineralogist,75, 1043-1058.
Holdaway, M J., and I-ee,S M. (1977) Fe-Mg cordierite stability in high-
grade pelitic rocks basedon experimental,theoretical, and natural ob-
seryations.Contributions to Mineralogy and petrology, 63,1:-5-l9g
Holdaway,M.J., and Mukhopadhyay,B (1993) A reevaluationof the
stability relations ofandalusite: Thermochemical data and phasedia-
gram for the aluminum sllicates.AmericanMineralogist,jg,Z9B-315.
Holdaway,M.J , Guidotti, C.V, Novak, J.M, and Henry, W E (1982)
Polymetamorphism in medium- to high-grade pelitic metamorphic
rocks, west-centralMaine Geological Society of America Bulletin, 93,
572-584
Holdaway,M.J., Dutrow, B.L-, Borthwick,J., Shore,P.J.,Harmon, R.S.,
and Hinton, R.W (1986a)H contentofstaurolite as determinedby H
extractionline and ion microprobe.AmericanMineralogist,71, I 135-
I l4l.
Holdaway,M.J , Dutrow, B.L., and Shore,P.J. (1986b)A model for the
crystalchemistryof staurolite.AmericanMineralogist,jl,ll4}-1159
Holdaway,M.J, Dutrow, B L., and Hinton, R W. (1988)Devonian and
Carboniferousmetamorphism in west-centralMaine: The muscoylte-
almandine geobarometerand the staurolite problem revisited. Arneri-
c a nM i n e r a l o g i s 7 t ,3 , 2 0 - 4 1.
Holdaway, M.J., Mukhopadhyay, B., Dyar, M.D , Durrow, B L., Rumble,
D., III, and Grambling,J.A. (1991)A new perspectiveon staurolite
crystal chemistry: Use of stoichiometric and chemical end-members
for a mole fraction model. American Mineralogist, 76, l9l0-lgl9.
Holdaway,M.J., Gunst,R.F, Mukhopadhyay,B , and Dyar, M D. (1993)
Stauroliteend-membermolar volumes determined from unit-cell mea-
surementsof natural specimens.American Mineralogist, 78, 56-67.
Holland, T.J.B, and Powell, R. (1990)An enlargedand updatedinter-
nally consistentthermodynamic datasetwith uncertaintiesand corre-
lations: The system K,O-NarO-CaO-MgO-MnO-FeO-Fe,O.-Al,O.-
TiO,-SiO,-C-Hr-O, Journalof MetamorphicGeology,8, 89-124.
Koziol, A.M (l 990) Activity-composition relationships of binary Ca-Fe
GEOTHERMOBAROMETRY IN PELITES
and Ca-Mn garnetsdetermined by reversed,displacedequilibrium ex-
perients American Mineralogist, 75, 319-327.
Koziol, A.M., and Newton, R.C. ( 1988)Redetermination of the anorthite
breakdownreactionand improvement of the plagioclase-garnet-Alrsio5-
quartz geobarometer.American Mineralogist, 73, 216-223.
- (l 989) Grossularactivity-composition relationshipsin ternary gar-
nets determined by reverseddisplaced-equilibrium experiments Con-
tributions to Mineralogy and Petrology, lO3,423-433.
McMullin, D W A, Berman,R.G., and Greenwood,H.J. (1991)Calibra-
tion of the SGAM thermobarometer for pelitic rocks using data from
phase-equilibrium experiments and natural assemblages.Canadian
Mineralogist,29, 889-908
Miller, A.K., Guggenheim,S., and Koster van Groos, A.F. (1991) The
incorporation of "water" in a high-pressure2:l layer silicate: A high
pressuredifferential thermal analysisof the l0 A phase.American Mrn-
eralogist,76, 106-112.
Mukhopadhyay,B , Holdaway,M.J., Gunst,R.F., and Dyar, M.D. (1990)
Endmember thermochemical parameters and mixing properties of
3-hydrogenFe-staurolite GeologicalSocietyof America Abstractswith
Programs,22, 349.
Mukhopadhyay,B., Holdaway,M.J., and Dyar, M.D. (1991)New exper-
imental determination of the reaction Fe-cordierite : almandine +
sillimanite + quartz under hydrous condition Geological Sociely of
AmericaAbstractswith Programs,23,391-392.
Mukhopadhyay,B , Basu,S., and Holdaway,M.J. (1993)A discussionof
Margules-typeformulations for multicomponent solutions with a gen-
eralizedapproach Geochimicaet CosmochimicaActa, 57, 277-283.
Newton,R.C.,and Haselton,H T. (1981)Thermodynamicsof the garnet-
plagioclase-AlrSiOr-quartzgeobarometer In Thermodynamics of min-
eralsand melts,advancesin physicalgeochemistry, vol. l, p. 131-148.
Springer-Verlag. New York
Novak, J.M., and Holdaway,M.J. (1981)Metamorphicpetrology,min-
eral equilibria, and polymetamorphism in the Augusta Quadrangle,
south-centralMaine. American Mineralogist, 66, 5 l-69.
Ohmoto, H., and Kerrick, D.M. ( I 977) Devolatilization equilibria in gra-
phitic systems.AmericanJoumal of Science,277, 1013-1044.
O'Neill, H.St.C.,Pownceby,M.J., and Wall, V.J. (1989) Activity-com-
position relationsin FeTiO,-MnTiO, ilmenite solid solutionsfrom EMF
measurements at 1050-1300K. Contributionsto Mineralogyand Pe-
t r o l o g y ,1 0 3 , 2 1 6 - 2 2 2
Pattison,D.R.M. (1992) Stability of andalusiteand sillimanite and the
AlrSiO, tnple point: Constraints from the Ballachulish aureole, Scot-
land Joumal of Geology,100,423-446
Pigage,L C., and Greenwood, H.J. (1982) Intemally consistentestimates
ofpressureand temperature:The stauroliteproblem. American Journal
of Science,282, 943-969.
Powell,R., and Holland, T.J B (1988)An internallyconsistentthermo-
dynamic dataset with uncertainties and correlations. III. Application
methods, worked examplesand a computer program Joumal of Meta-
morphic Geology,6, 173-204.
- (1990)Calculatedmineral equilibria in the pelitic sysremKFMASH
(K,O-FeO-MgO-AlrOj-SiO.-H,O).American Mineralogist,75, 367-380.
Rancourt,D.G., Dang,M.-2., and Lalonde,A.E. (1992)Mdssbauerspec-
troscopy of tetrahedral Fer* in trioctahedral micas. American Miner-
alogist, 77, 34-41.
Richardson, S W (1968) Staurolite stability in a part ofthe system Fe-
Al-Si-O-H Journal of Petrology, 9, 467-488.
Robie, R.A, and Hemingway,B S. (1984)Entropiesofkyanite, andalu-
site, and sillimanite: Additional constraints on the pressureand tem-
perature of the AlrSiO, tnple point. American Mineralogist, 69,298-
306.
Shi, P. (1992) Fluid fugacitiesand phaseequilibria in the Fe-Si-O-H-S
system.American Mineralogist,77, 1050-1066.
Shi, P., and Saxena,S-K (1992)Thermodynamicmodelingof the C-H-
O-S fluid system.AmericanMineralogist,77,1038-1049.
Spear,F S , and Cheney,J.T. (l 989) A petrogeneticgrid for pelitic schisrs
in the systemSiO,-Al,O,-FeO-MgO-K,O-H,O. Contributions to Min-
eralogyand Petrology,l0l, 149-164.
Spear,F.S.,and Selverstone, J. (1983)QuantitativeP-?paths from zoned
 HOLDAWAY AND MUKHOPADHYAY: GEOTHERMOBAROMETRY IN PELITES 693

minerals: Theory and tectonic applications- Contnbutions to Miner- relationshipsforpyrope-grossulargarnets.Ioumalofceology,96,12l-

'129.
alogyand Petrology,83, 348-357.
Weisbrod, A. (l973) Cordierite-gametequilibrium in the systemFe-Mn- - ( l99l) Experimental measurementsof Mn-Mg mixing properties
Al-Si-O-H. Camegie Institution of Washington Year Book, 72, 515- in gamet, olivine and oxide. Geological Society of America Abshacts
523. with Programs,23, 93
Williams, M.L., and Grambling, J.A. (1990) Manganese,ferric iron, and
the equilibrium betweengamet and triotite. American Mineralogist, 75,
886-908 Mnmxcnrpr RECEIvED Mencn 3, 1993
Wood, B.J (1988) Activity measurementsand excessentropy-volume MeNuscnrvr AccEprEDMercn 18, 1993