American Mineralogist, Volume 71, pages 301-316, 1986
Phaserelationshipsof gabbro-tonalite-granite-water
at 15 kbar with
applicationsto differentiation and anatexis
Wuu-LrlNc HuaNc
Exxon Production ResearchCompany, Houston, Texas 77001
Pprnn J. Wrr-r-rn
Division of Geologicaland Planetary Sciences,California Institute of Technology,Pasadena,California 9 1I 25
Ansrnncr
Phase relationships have been determined at 15 kbar with variable HrO contents for
three rocks-a gabbro,a tonalite, and a muscovite granite-representing the compositional
trend of the calc-alkalinerock series.Experimentswereconductedin 0.5-in. piston-cylinder
apparatususing Ag-Pd or Pt capsules.The results obtained are acceptablerepresentations
ofthe phaserelationships for elucidation ofpetrological processes.Selectedruns with 5olo
HrO were completed using Au capsulesto overcome the problem of iron loss, up to the
temperature limit of Au, for reliable electron microprobe analysesof glass,garnet, clinopyroxene, amphibole, and plagioclase.No glassescould be analyzedin the gabbro. The
data provide Ko values as a function of temperaturefor pairs of coexisting minerals and
liquid. The chemical variations are usedto evaluate(l) the fractionation trends in hydrous
magmasat 55 km in thickened continental crust or uppermost mantle and(2) the products
ofanatexis ofthe thickened crust. At this depth, fractionation ofhydrous basalt,or anatexis
of gabbro and tonalite or their metamorphosedequivalents, appearsto produce liquids
diverging from the calc-alkalinetrend. Anatexis ofgranite producesa liquid less siliceous
than itself, with some chemical characteristicsof syenite.
INrnooucrroN
ExpnnrunNTAL
One of the major chemical trends in igneouspetrology
is given by the calc-alkaline rock seriesgabbro-tonalitegranodiorite-granite(basalt-andesite-rhyolite).The phase
relationships of specific rocks in this series have been
studied under a variety of conditions by many investigators, commonly with applications to particular petrogeneticproblems related to the specificrocks. The experimental resultsfor three rocks at l5 kbar are presentedin
this paper in T-Xrro phase diagrams,together with a set
of reliable analysesof minerals and glasses.
The more general construction of the phase relationships for the complete rock series,representinga compositional line through the componentsof igneousrocks,
hasbeenaddressedby Piwinskii and Wyllie (1968, 1970)
and Piwinskii (1968, 1973a, 1973b, 1975) with excess
HrO at 2 and l0 kbar, by Green and Ringwood (1968)
for the anhydrous seriesat 18,27, and 36 kbar [seealso
Green (1972), Green and Ringwood (1972) for the effect
of HrO on the liquidus at high pressuresl,by Stern and
Wyllie (1973,1978) at 30 kbar with HrO contentsfrom
5oloto excess,and by Green ( I 982) at I 5 and 30 kbar from
dry to excessHrO. An important petrogeneticquestion is
to what extent the liquid compositions within the crystallization intervals of gabbro and tonalite parallel the
calc-alkalinecomposition trend, for a variety of pressures
and water contents.
0003{04xi8610304{30I $02.00
METHoDs
Starting materials
Natural olivine tholeiitic basalt (gabbro), tonalite, and muscovite granite were usedas startingmaterials.The chemicalcompositions,clrw norms, and approximatemodesarelisted in Table
l. The tholeiite is from a fino-grained basaltic sill in Scotland
(Lambert and Wyllie,1972). The tonalite from the SierraNevada
batholith was kindly provided by P. C. Batemanand F. C. Dodge
(tonalite 101, or M 127 in Piwinskii, 1973b). Field and petrographic descriptionsofthe tonalite have been given by Bateman
et al. (1963). The granite collected from Harney Peak, South
Dakota, was kindly provided by T. T. Norton and R. T. McLaughlin of the U.S. GeologicalSurvey(Huangand Wyllie, 1973,
1981). The chemistry of major elements in these rocks corresponds closely to the calc-alkaline trend (although the granite is
of S-type).
Apparatus and procedures
Experimentswere performed in a single-stagepiston-cylinder
apparatussimilar to that describedby Boyd and England(1960),
using a 0.5-in.-diameterpiston and chamber with hardenedsteel
liner. The apparatus was calibrated for pressureand temperature
asdescribedby Boettcherand Wyllie (1968).Temperatures,measuredwith chromel-alumel(below 850€) and Pt-PtroRh,o(above
850"C) thermocouples, were precise to +5"C and accurate to
+ l5'C. Furnaceassemblieswere constructedfrom graphite and
talc for mns up to 850'C, but for higher temperatures,the talc
rods were replaced by pyrex glass to avoid talc dehydration and
301
HUANG AND WYLLIE: GABBRO-TONALITE-GRANITE-WATER
AT l5 kbar
302
Table 2. Results of quenchingexperiments
for gabbro (DW-l) at l5-kbar pressure
Table 1. Chemical analyses,crpw nonns and approximate
modes for whole rocks in the present study
cabbro,
Dw-1
Tonafite
I01
cranite
L26
(1) Cheni@L a@L!6ea
45-91
o.94
17,19
2.33
Ti02
A 1 2 03
F e 2 O3
Fe0
ho
ugo
o.22
'7
.4A
r3. 54
1.63
0.14
1.78
Na2O
Kzo
H zo +
Pzos
cQz
0.04
14-66
0-03
15.55
0.17
o.42
0.07
0.02
o.42
4-29
3.08
0.66
0,04
0.15
o.02
59. 14
o.79
18.23
2.32
0.I1
2.50
5.92
3.81
2. r9
0.82
0.04
0,30
0.01
( 2) CfPW Mome
Quartz
orthoclase
AIbite
horthite
Diopside
0.83
13.78
39.14
22.59
5-95
8.55
3. 39
0.78
0.10
olivine
Ilmenite
Apatite
Corudw
1I. 7
12.9
3 2- 2
26.2
0.9
9.4
18.20
36.24
o.98
3.4
1.5
o-7
0.06
4.80
800
825
a25
850
4A4
451
480
4s4
453
452
456
455
451
463
461
46a
470
46I
469
472
476
47r
459
462
30
4-4
30
30
30
23
22
1l
30
40
20
25
25
3014L
4013L
5
1022L
1523L
3072L
458
1200
514L
s
488
IOOO
4a6
llOO
485
1150
30
Lpx,
rc,
I
L
Cpx,
Hb,
Hb,
I
Cpx'
Hb'
Gt(?)
12
12
Lr
1l
I0
cPx, b,
I
cpx, ft, I
cpx, Hb, I
CPx, ft, I
Cpx, ft, I
Cpx, Hb(tr),
cpx, Hb(tr),
Cpx(tr) + L
+ L
cpx(tr)
CPx(tr) + L
14
cpx+r
23-5
34
23
23.5
30
5
900
925
975
1000
rO25
1050
1050
1075
I075
I075
1075
1075
1100
1100
1100
1100
Cpx,
Cpx(tr),
5
5
5
1r
3
I
1
+ L
L
I
L
L
CPr' Hb' Ga, L
cpx, Hb(?), L
Cpx(tr), L
4811200511
(3)
App"aeimate
l4ade6
Quartz
a1 Lal i
F-l Ac^:r
PIagj-oclase
Augite
4'7
41
Horhblende
Opaques
krnet
Apatite
473
]
I "t.t
12.5
9
Biotite
Olivine
I3
4.5
59
3
3
2
0.1
466
475
411
I
l
)
47a
0.9
479
Muscovite
414
465
embrittlement. All runs were brought to final pressure with the
"piston-out" proceduresas describedby Boyd et al. (1967).Pressures listed are nominal, incorporate no corrections for friction,
and are consideredaccurateto +590.
Crystalline rock sampleswere used in the experiments. Oxygen
fugacity was not controlled. The oxidation state of the sample
during the mn conditions should be close to the Ni-NiO buffer
(Allen et al., 1975). Water and (dry) powdered rock sample were
weighed into capsules in the required proportions, sealed, and
run at 15 kbar and selectedtemperatures.The run tableslist the
HrO weighed into the capsules.The actual HrO present in a
capsule during a mn was thus the total of HrO stored in the
hydrous minerals of the rock, plus the HrO added. Two types of
runs were made to determine the phaseboundaries:(l) "singlestage" runs in which the crystalline rock samples were brought
directly to the desired temperatures and (2) "two-stage" runs in
which the crystalline samples were held above the liquidus until
completely melted and then the temperature was reduced to the
desired value. The first type was prevalently used in the present
study while the secondwas made as a test for equilibrium.
Electron-microprobe analyses were performed on quenched
charges at two intervals during the present study. Early reconnaissanceanalyseswere made on an ARr-/EMxelectron-microprobe analyzer at 15 kV,0.02 pA sample current, for l0 s per
element. The beam size was about 2 pm. The microprobe analyzer was then automated. Most of the analysesfor the present
study were performed on the computerized .ornr-/er'.rx
analper,
using standard procedures for the J. V. Smith laboratory at the
University ofChicago. The accuracyofanalyses is -r2% ofthe
amount of major elementspresent.
1150
1020
1150
1040
1200
1050
1150
r050
rr50
1050
l]so
1050
1200
1075
r200
40
40
40
40
5
5
15
15
20
20
30
30
5
5
'7
22
7
25
9
25
2r
24
6
Cpx, S(?),
cpx,
L
L
cpx, L
cpx, Eb(?), L
cPx,
L
9
24L
11
23L
'l
Identification of phases
Phasespresent in quenched runs were identified using standard
optical and X-ray powder-diffraction techniques, described in
previous papers on these rocks. Phases encountered include in
the gabbro-watersystem,amphibole, clinopyroxene,garnet,liquid, and vapor deposits; in the tonalite-water system, garnet,
clinopyroxene, plagioclase, quartz, biotite, liquid, and vapor deposits; in the granite-water system, quartz, plagioclase, alkalifeldspar, muscovite, kyanite, sillimanite, corundum, garnet, liquid, and vapor deposits. Opaque minerals, although of interest,
were too sparsefor confrdent identification.
Expnnrlvrmlur,
REsULTS AT 15 KBAR
One major uncertainty in the interpretation of experimental
results is the absorption of iron from samplesby the noble-metal
capsules(Merrill and Wyllie, 1973;Sternand Wyllie, 1975;Nehru and Wyllie, 1975). Experimentsfor the gabbro-watersystem
were first performed using Pt capsules.In order to test the effect
of iron loss on the liquidus boundary, some additional runs with
5% HrO were performed, using AgrrPd25capsulesto determine
the liquidus temperature. Most runs for the tonalite-water system
capsules.Runs for the low-iron granwere made using Ag3oPd?o
ite-water system were performed using Pt capsules. When the
phase diagrams had been determined, selected runs lvere com-
AT 15 kbar
HUANG AND WYLLIE: GABBRO-TONALITE-GRANITE-WATER
303
I
I
t
I
I
I
I
I
ll
LI
r200
v
i Hb-ouT
t
I
CJ
cc
=
1000
ro
I
i\o
lr
r\
O
CL
=
Hb
+
cpx
+
lGal
+
t
Hb+Cpx+02+lcal+l+v
Hb+Gpx+Pl+(L+lGal
Hb+CpxaPl+(lz+lGal+U
WATER
WT.%
Fig. |. Phaserelations of gabbro Oasal| at l5 kbar with varylng HrO contents. For definitive runs and abbreviations, seeTable
2. Squaresindicate runs oflambert and Wyllie (1972). Seetext for other data sources.HrO bound in the original rock is indicated
by the arrow. Only for 090 HrO is mineral assemblagedevoid of Hb + Bi. (Ga) indicates garnet, which should be stable, but failed
to nucleateexceptin Au capsules(seetext).
pleted in Au capsulesfor analysis of minerals and glasseswith
minimum iron loss, up to the temperature limit of Au.
Phase relationships for gabbrewater
Figure I shows the experimental results at 15 kbar for the
melting relations of gabbro with 5-400/0HrO added. The H,O
structurally bound in the original rock is shown by the arrow.
Near-solidus runs with excessHrO are from I-ambert and Wyllie
(1972). The definitive mns are listed in Table 2. The liquidus
temperature of an anhydrous rock of similar composition was
determined by Green and fungwood (1968). The solubility of
HrO in gabbro liquid at 15 kbar and 1080"C,which should be
given by the change of the slope of liquidus boundary, is not
constrained.
The subsolidus mineral assemblageat this pressurewith excess
water is Ga + Hb + Cpx + Pl + Qz + V according to the results
of Lambert and Wyllie (1972). There are three solidus temperatures, depending upon the amount of water present. For a dehydrated rock, the solidus temperature is about 1200€, esti-
mated from Green and Ringwood (1968). For a vapor-absent
rock, with all HrO stored in amphibole, the solidus corresponds
to the reaction temperature of amphibole in this assemblage,
extrapolated from runs with HrO added. This temperature was
not determined, but it is probably between 1000 and ll00"C.
For a rock with a free vapor phase, the solidus is at 635qC.
Plagioclaseand quartz disappear, respectively, at I 5 and 40€
above the solidus in the presence of excess water. The upper
stability boundaries for plagioclase and quartz are estimated by
interpolation between the anhydrous melting relations of similar
rock compositions (Green and Ringwood, 1968; Ito and Kennedy, 1968)and the melting boundariesin the presenceof excess
H,O. The boundaries fall steeply with increasing HrO in the
vapor-absent region. The runs do not define the position of the
dashed boundary separating the phase fields with vapor present
or absent.
Amphibole is present from subsotdus temperature to nearliquidus temperature (1070"C) with excess HrO. In the vaporabsent region, lve expect the upper stability temperature of
amphibole to increaseslighfly with decreasingHrO content (Hol-
HUANG AND WYLLIE: GABBRO-TONALITE-GRANITE-WATER
304
AT 15 kbar
1200
s0u0us
oo
lllllll
llt
G
.,'
G
EI
G
=
H
Hb+Cpx+Ky+[+v
800
f' Q r\<*,
r,"
(Ga
+ l[ , , ' - - -
H b +c l x+ ( G a ) +
B i a K Y +t + v
D
f i l * C o+
, 0 r * ( G .*] B i+ x y + [ + V [
l Bi
H b + G p x+ P l + 0 z + ( G a +
tr
H b + C p x + P l + 0 z + ( G a ) +B i + V
WATER
WT %
Fig. 2. Phaserelations of tonalite at 15 kbar with varying HrO contents. For definitive runs and abbreviations, see Table 3.
Squaresindicate runs from Lambert and Wyllie (1974). Seetext for subsoliduschanges,other data sources,and seelegendfor Figure 1.
loway, 1973). No garnet was observedin run products using h
and Ag-Pd capsules,except in a single run (#R488) at 1000t
with 5% HrO (Ag^Pd.). Yet garnets occur in runs from Au
capsuleswith 50/oHrO between 700 and 1000qC(Table 5). The
failure ofgarnet to nucleateunder theseconditions is well known
and much reviewed (Lambert and Wyllie, 1972, 1974; Green,
1972). The nucleation ofgarnet in Au capsulesmay be due to
retention of iron in the sample. (Ga) indicates regions where
garnet should be stable, despite its failure to nucleate in Pt and
Ag-Pd capsules.
Clinopyroxene and garnet are the near-liquidus phasesfor all
HrO contents (separate boundaries not distinguished). They are
joined by amphibole just below the liquidus with excessHrO.
Through more than 350{, liquid coexistswith amphibole, clinopyroxene,and garnet for HrO contents at least down to 5olo.
Phase relationships for tonalite-water
Figure 2 shows the experimental results at 15 kbar for the
melting relations of tonalite with 5-35 wt% HrO added.The HrO
structurally bound in the rock is shown by the arrow. Near-solidus
runs with excessHrO are from Lambert and Wyllie (1974). The
definitive runs are listed in Table 3. The liquidus temperature
for anhydrous tonalite is near l250qC as estimatedfrom Green's
(1972) study of a rock of similar composition. The liquidus temperature drops to about 1 140"C for the vapor-absent rock (with
HrO releasedfrom amphibole and biotite), and down to 940qC
with excessH,O. The solubility of H,O in tonalite liquid at 15
kbar and 950qC,which should be grven by the changein slope of
the liquidus boundary, is not constrained. Clinopyroxene and
amphibole are liquidus or nearJiquidus phaseswith excessHrO.
Clinopyroxene occurs only in minor amounts except near the
liquidus. In the vapor-absent region, the upper stability temperature of amphibole changes little with decreasing HrO content
and thereforedivergesfrom the liquidus. Amphibole is the dominant phase down to the solidus region. The biotite-out curve
rises to somewhat higher temperatures in the vapor-absent region. Precisedetermination of rhe amphibole-out and biotite-out
curves at low HrO contents is highly desirable, but experimentally
difficult.
The solidus at 15 kbar is within the garnet stability field and
in the region of the reaction of plagioclase to jadeitic pyroxene.
305
15kbar
HUANG AND WYLLIE: GABBRO-TONALITE-GRANITE-WATERAT
According to Lambert and Wyllie (1974), the subsolidusassemblage with excessHrO is Pl + Qz + Hb + Bi, u.ith a trace of
Cpx and possible Ky, and with Ga, which failed to nucleate.
Plagioclase should be replaced completely by clinopyroxene and
garnet within a short temperature interval below the solidus. This
is not indicated in Figure 2, and no nrns were completed to
establish this, but the generalpicture is clear (Lambert and Wyllie,
1974, Fig. l). There are three different solidus reactions and
temperatures, depending on HrO content, as discussed for gabbro-HrO.
Plagioclasedisappearswithin 5'C above the solidus with excess
HrO, but is stable to considerably higher temperature, 950eC,
with 50/oHrO added, and is stable to almost 1350€ in a dehydrated rock (Green, 1972).For a dehydratedandesite,quartz is
stable to about I 150'C, 50'C below the liquidus (Green, 1972),
but addition of 50/oHrO is sufrcient to lower its upper stability
temperatureto about 775'C. Quartz disappearsabout 50€ above
the solidus with excessHrO. The dashed boundary separating
vapor-absent from vapor-present fields is only estimated. However, its estimated position constrains the low-temperature ends
and the curvaturesofthe plagioclase-outand quartz-out boundaries. The latter boundary has the normal concave-upward curvature (seeFigs. I and 3). The unusual, retrograde shape for the
plagioclase-outcurve is a consequenceof the plagioclase-clinopyroxene reaction near the solidus, which can be modeled in the
systemNaAlSiO4-SiOr-HrO.
In the presenceofexcess vapor, trace amounts ofgarnet were
observed at 15 kbar between 725 all.d 875"tCby Lambert and
Wyllie (1972). In the presentexperimentswith Ag-Pd capsules,
garnetswere observeddown to 850qCwith 5% HrO added. The
amount of garnet increaseswith temperature, and near the liquidus, garnet and clinopyroxene become the dominant phases.
However, garnet appearsin runs down to 700qCin Au capsules
with 50/oHrO (Table 5). The failure of gamet to nucleateat lower
temperatures where it should be stable, denoted in Figure 2 by
(Ga), was alreadydiscussedin connectionwith the gabbro-water
systemand by I-ambert and Wyllie (1972).
Phase relationships for granite-water
Figure 3 shows the experimental results at 15 kbar for the
melting relationships of muscovite-granite with and without water
added, and the results are extrapolated toward the anhydrous
rock composition (Huang and Wyllie, 1973, l98l). The HrO
structurally bound in the muscovite granite is shown by the arrow
at0.66o/o.
The definitive runs arelisted in Table 4. The subsolidus
assemblageat this pressure corresponds to that of the natural
rock, two feldspars,quartz, and muscovite, with just a trace of
garnet.
The phase-boundarybetweenvapor-absentand vapor-present
fields is determined by runs in Figure 3, in contrast to Figures 1
and 2, and this gives a better estimate of HrO solubility in the
liquid at about 77 5C. ln the vapor-absentregion, upper boundaries for quartz, feldspars, and garnet decreaseconsiderably with
increasing HrO content. The muscovite-out temperature increasessomewhat with decreasingHrO content, as anticipated
from phase relationships in synthetic mica systems (Yoder and
Kushiro, 1969;Huang and Wyllie, 1974).The vapor-absentsolidus for the beginning of reaction of Na-bearing muscovite in this
assemblagehas been measured(Huang and Wyllie, 1981).
Quartz was the liquidus phase for the biotite-granite studied
by Stern and Wyllie (1973, l98l), and the correspondingboundary is shown by the heavy line in Figure 3. For the aluminous
granite used in the present study, traces ofcorundum persisted
Table 3. Results of quenchingexperimentsfor tonalite (l0l)
at l5-kbar Pressure
Rm
----:-------
444
441**
4r9
442
4I3
44r
403
422
420
429
404
423
4I2
4L6
421
4L1
402
4r4
445*r
435
410
40a
424
4ra
431
407
405
800
425
850
475
925
925
925
940
94A
950
950
950
975
975
975
1000
1000
1025
rO25
102s
to50
rloo
Lr25
1150
1150
1150
1200
430
rooo
915
rO50
925
1050
950
1050
96s
lO50
975
1100
1000
1100
1025
43A
42a
425
433
432
42r
gonet;
Hzo
Results*
Bio,
Cpx, PL
cpx, P], Bio'
cpx, Ca(tr),
cpx, @, PI'
Pl,
cpx, G,
cpx, Ga, Pl,
KY'
cPx(tr),
KYr L
L
Pr, Bio(?)'
KY(?), r
L
KY(tr),
L
Ky(tr),
5
5
5
5
5
5
33
32.7
f5
9,6
24
9
12
3
13
1l
Hb,
&,
b,
S,
&,
6,
Eb,
I
50
20
20
28
5
L4 7
1
20
2l
18.5
1
2I-S
L
6(tr),
CPx, KY, I
b, cpx, KY(tr), L
L
ft(?). cpx' c, I
cpx, I
L2I4L
33
5
5
s
5
25
3
3
L
cpx, L
cpx,@,l
cpx, Ga, a
0
05L
0
5lal
5
33
33
339
33
235
23lAL
204
2o,4
13
13
52r
5
5
L
18
A
5
I7
s(tr),
22
Cpx, L
25
L
L
cpx,l
t2
10
35
11
24
KY(?),
cpx, r
cpx(?), r
Cpx' L
cPx, r
514L
Addi-tional abb?eoiatians:
see Table 2
For ahbpetiatian
PL = plagioc1'ase; Bio = biotitej
4i = kAmite.
0a =
to higher temperatures. Metastable corundum forms from the
breakdown ofmuscovite at 15 kbar, instead ofthe anticipated
AlrSiO, polymorph. Corundum, once formed metastably, persists indefinitely in this system, as well as in parts of the synthetic
muscovite system (Huang and Wyllie, 1974).
Phase relationships for gabbretonalite-granite-water
Figure 4 comparesand combines the phase relationships for
the rock seriesat 15 kbar with 5qoHrO. The phaseboundaries
have been drawn through points transferred from Figures 1, 2,
and 3. The data of Green and fungwood (1972) on a rhyodacite
with 690loSiO, were included in constnrction ofthe diagram. The
diagram shows two liquidus piercing points bounding the garnet
liquidus between about 60 arrd 700/oSiOr, with lowest liquidus
temperaturenear 700/oSiOr. The effectof pressureand HrO content on the liquidus profile for the calc-alkaline rock series has
been discussedby Green and Ringwood (1968) and Stern and
Wyllie (1973,1978).
Test of equilibrium
The phase diagrams in Figures l, 2, and 3 are reaction diagrams, with phase boundaries drawn on the basis of single-stage
runs using capsulesofPt and AgrrPd' for gabbro, AgroPdrofor
tonalite, and Pt for granite. Tests for continuation or completion
of reaction and for the attainment of equilibrium include increasingrun duration and making two-stagenrns (also necessarily
oflonger duration) to reversethe reactions.The longer durations
306
HUANG AND WYLLIE: GABBRO-TONALITE-GRANITE-WATER
AT 15 kbar
I
t
I
I
CJ
o
I
I
I
I
I
I
I
I
I
1000
trt
CE
oc
\l
ltt
4
=
ll,l
e\
\%
tu
o
0z+Ms+L
0z+Ms+[+V
Pl+02+Ga+Ms+L+V
Pl*0r+02+Ga+Ms+[*V
Pl+0r+02+Ga+Ms+V
30
WT %
WATER
Fig. 3. Phase relations ofgranite at 15 kbar with varying HrO contents. For definitive runs and abbreviations, see Table 4.
Modified after Huang and Wyllie (1973).
for enhancing the degree of reaction and testing reversibility introduce other problems, such as increased iron loss from the
sarnple and contamination of thermocouple junctions, causing
drift of temperature.
The liquidus boundaries for gabbro-water were determined in
three groups of runs: (l) ft capsulesin single-stage,(2) AgrrPd*
capsules in single-stage, and (3) Pt capsules in two-stage runs.
The definitive runs are listed in Table 2. The liquidus boundary
determined by the first group is at 1080.,Cwith excesswater.
With 5oloH,O, the liquidus boundary is I l3OC, identical to that
from the second group of runs. However, the liquidus boundary
determined by two-stageruns is 55"C lower with 590HrO and
30"C lower with excessHrO. For the tonalite-water system, the
liquidus boundaries for clinopyroxene were reversed successfrrlly
in two-stage runs within a 25 and 50"C bracket, respectively, for
33 and 50/o
watercontents;butno amphibole orgarnetcrystallized
in these two runs in contrast with their presence in the singlestageruns. The diference between liquidus temperatures determined by single-stage and two-stage runs for both gabbro and
tonalite is tentatively attributed to the nucleation problem, but
also to the iron loss ofbulk composition from charge to capsules.
For the granite-water system, the upper boundary ofquartz was
experimentally determined with a 25"C bracket at 15 kbar with
excesswater (runs 240 and242 in Table 4) and with 9.5% water
addedat 25kbar (runs2l5 and2I7; HuangandWyllie, l98l).
The granite-water system is notorious for its sluggish reaction
rates. dthough the results in Figures 1, 2, and 3 must be regarded
as provisional, for the reasons stated above (similar statements
could be made for most published rock-HrO studies), the results
with Au capsules (50/oHrO) are broadly consistent (Table 5).
Therefore, the diagrams are considered to be acceptable representationsof the phase relationships,with respectto major assemblagesand geometricalsequences.
The etrect of iron loss from the sample in changing the phase
assemblage,in addition to changingthe compositions of phases,
can be seen by comparison of the results listed in Table 4 for
selectedruns in Au capsules,with the results plotted in Figures
I, 2, and 3. For gabbro with 5% H,O, the use of Au capsules
between 700 and l00OC produced garnet where none nucleated
with Pt capsules;garnet was also present in the 1000t run using
Ag-Pd. A similar difference occurred with tonalite-5% HrO. With
Ag-Pd capsules,garnetformed only at 850€ and above,whereas
in Au, it occurred down to 700C, the lowest-temperature run.
No quartz occurred in the 700qCrun with Au, whereas with AgPd capsules it should be present. No kyanite occurred in runs
\4'ith Au, whereasit was presentup to 925qCin Ag-Pd capsules.
307
AT I 5 kbar
HUANG AND WYLLIE: GABBRO-TONALITE-GRANITE-WATER
Table 4. Resultsof quenchingexperimentsfor muscovite
granite (L26) at 15-kbar pressure
TemDerature
Ru
H,0
added
*at
226
227
600
610
351
660
21r
374
307
3r0
240 I
i
23I
242 t
|
2a2
299
3a6
301
229
395
293
337
660
700
700
700
425
7oo
725
a25
125
750
750
750
750
150
750
715
715
2r9
303
257
2A3
223
302
306
715
195
800
800
aOO
820
520
2ro
292
38a
2AO
230
390
3€2
289 |
{
2ea
219
284
354
425
425
850
850
475
875
475
94O
87s
900
900
925
950
241
269
363
266
265
50r
502
263
296 |
+
247
2 5 4|
+
249
1000
1000
1000
1050
1100
1125
1175
1200
1315
12OO
1250
1315
r24O
1300
Time
hous
29
3a
9
12
9
9
20
16
22
l6
1
5
LO
36.3
36. 3
24
25
40
43
20
46
5
0
l5
5
9
o
45
15
3.2
l4
0
2.7
5
9.5
9.5
9.5
9.5
0
r.6
r.6
o
o
0
0
0
0
0
4I
21
6
16
26
6
ls
20
24
l2
22
24
I2
22
22
21
2!
20
23
20
2T
20
22
18
20
25
ro.5
l0
9
11
11
20
9
3
43
5
42
47
23
Qz,
92,
Qz,
Qz,
Qz,
Qz,
Qz,
Qz,
Qz,
Qz,
M,
S,
Ab,
Ab,
Ab,
re,
re'
M,
Ab,
w,
10000c
950.C
900'c
800'c
?oo'c
Or, Ms, Ga
Or(tr), Ms, e, L
re, Co(?), Ga{tr), L
re, Ga(tr) , t
re, L
I
I
Or, re, G, I
b,
Ga, I
L
5 h
15 h
36 h
3 days
53.5h
A93oPd7o
9250C
900'c
A930Pd70
Au
Au
700.c
40
7OO'c
63h
Ar = orthocLase;
eoL;'d soLution;
Amphibole-out occurredbetween900 and 950"Cin Au capsules,
whereasa trace of amphibole may remain in the 975'C run with
an Ag-Pd capsule. In the granite-water system, results in Au
capsulesbetween650 and 1000{ were the same as those in Pt
capsules,probably becauseofthe smaller Fe content ofthe granrte.
RUNS
Several runs made in Pt and Ag-Pd capsules were originally selected for microprobe analyses. Figure 5A shows
the percentage ofiron loss (forglass) quenched from above
the liquidus as a function of time. Figure 58 shows the
difference in Mg/(Mg + 2Fe) values for coexisting garnet,
clinopyroxene, and glass using Ag.oPdro and AgrrPd* capsules under the same conditions. Diferent phases lose iron
G+cpx+6+L
G+Cpx+b+L
Ga+CPx+&+i
Ga+cPx+6+L
6+cPx+b+L
lAj
445
Ga + CFx + L
4I2
G+cpx+&(?)
609
G+CPx+L
+L
44I
G+S+cpx+Pl+Ky(t!)
606
fr+&+Cpx+P1+L
442
4a9
+L
6+S+cpx+P1+(Y(?J
Ga(tr) +S+cpx+Pl+Bio(?)
444
602
6+Cp*+P1+Bio+Ky+l
&+ft+cPx+PI+Blo+L
au
603
e+e+CPx+P1+Bio+L
au
610
Ga+b+cPx+P]+Bio+L
Au
Au
506
Qz+Co+L
509
Qz+Co+L
25 h
505
48 h
Au
Au
507
Qz+co+L
re+Pl+Qz+Co+L
700"c
90 h
Au
508
Ms+PI+Qz+
Ga+L
Au
5to
b+P]+Qz+
Ga+L
5 days + 21 h
+L
A93oPd7o
Au
t h
7 h
650'c
L
Dw-1
493
494
4A9
49o
492
Tanalitu
900"c
800"c
9500c
Co(tr?) ' I
Q z( t ! )
RESULTS FRoM GoLD cApsuLE
15
30
10000c
Qz, Co, L
Q2, Co, L
co, L
si(?), co, L
Qz, si(?), co(tr), L
Qz, FsP, Co, G, L
Qz, Fsp, Co, I
Qz, co, L
92, Fsp, Ky(tr), Co, L
Qz, Kyl?), Co(tr), I
Qz, Co, L
Co, L
Qz, si, L
si(tr),
Co(tr),
Co(tr), I
aoo.c
a000c
kbbro
A9!0Pd?0
AgroPdTo
a50ec
Ga, L(tr)
u
u
u
u
u
975.C
9 5 00 C
e750c
Qz, re(tr), !
Qz, re, Co(tr), I
Qz, Ms, Co(tr), I
92, re, Co, I
co, f,
Qz, b(tr),
Ky(tr), Co, L
Ky(tr), Co, L
Qz, Co, L
Qz{tr), Co, L
Ky, Co, L
Ky(tr), Co, L
02, Ab, re, Co, I
Qz, Ab, Or, Ms, k
Qz, co(tr), L
Co(tr), L
Qz, Ab, re(?), co, L
Qz, Co, L
Qz, I$t Ot, re(tr) ' Co(tr),
Co(tr), L
Qz, Co, L
L
Qz, t\b, Co, e{?),
co, I
qz, Fsp, Co, Ga, L
e, L
Qz, Fsp,6,
Qz, Co, L
A
A
A
A
A
AqTsPd2s
I025.C
QZ, re, L
Qz, MS, L
Abb\eoiation,:
Qz = qu@tz; Ab = albite;
lsp = alkaLi-feldepd"
Ga = gffiet;
Me = Mcaotte;
Si = siLlimite'
KA = kAnite;
Co = cowdn;
AN.ql-yrrcAr-
Table 5. Selectedruns for microanalysesat l5-kbar pressure
and 5oloH"O
+Kv+L
at different rates. Therefore,theseanalyseswere discarded.
Runs using Au capsuleswith 50/oHrO added are listed
in Table 5 for gabbro, tonalite, and granite compositions
at temperaturesof 650, 700, 800, 900, 950, 1000, and
1025'C. The phase results are broadly consistent with
those shown in Figures 1,2, and 3, with the differences
described in the preceding section. Electron-microprobe
analyseswere performed for l0 elements(Si, Ti, Al, Cr,
Fe, Mn, Mg, Ca, Na, and K) for garnet, clinopyroxene,
amphibole, plagioclase,muscovite, and glass. For most
runs, there is considerablescatter of analysesfrom one
mineral grain to another, with less variation from center
to margin of individual grains. Rangesof values and averagecompositions are plotted in Figures 6 through 10.
Glass compositions plotted are averagesof 4 to l0 analyses.We have eight tablesof analysesthat, unfortunately,
cannot be fitted into this paper. Spacelimitations also
precludedetailed comparison of our analyseswith related
resultsby Green and Ringwood ( 1972),Allen and Boettcher
(1978, 1983),Allen et al. (1975), and Stern and Wyllie
(l 978).
Garnet
Garnet analysesare plotted in Figure 64' Note the range
ofresults at eachtemperature. Garnet synthesizedat 950qC
contains less gtossular component in the center than in
the margin. Garnet synthesizedat 800€ exhibits lesszoning. The averagevalues in Figure 68 define a trend showing significant decreasein pyrope component with decreasingtemperaturesfor both gabbro and tonalite. There
is also a decreaseof grossular components in garnet as
rock composition becomesmore silicic from gabbro to
tonalite to rhyodacite. Data for rhyodacite with 50/oHrO
at 13.5kbar werereportedby Greenand Ringwood(1972),
with garnet compositions located on the upper right-hand
side of the tonalite curve in Figure 68.
HUANG AND WYLLIE: GABBRO-TONALITE-GRANITE-WATER
AT 15kbar
308
1400
rSKb 5WT.%r{20
t200
GRANITE
TONATITE
GABBRO
t
t
I
1000
qt
E
CE
u
IL
=
F 8oo
--t---
,
t,
600
Hb+ Cpx
Pl+ oz
Ga*V
40
r,
Fsp+ 0z
H b + c p r + F s p + o z + c a + M i c av+
IrJ,
!
+{i
I
50
60
70
80
slucAwT %
Fig.4. Phaserelations of the calc-alkalinerock seriesas a function of SiO, at 15 kbar and with 50/oHrO. Rectanglesindicate the
temperaturesofphase boundaries from Figures I,2, ald 3. The vapor-out boundary is situated a few degreesabove the solidus.
For abbreviations, seeTables 2, 3, and,4.
0
1025c
15Kb,3h
5 W lY o H 2 0
1234
H()URS
TIME
Ago
As$ Pd70
As75Pd25 A9100
% As lN AsPdCAPSUTES
Fig. 5. Absorption ofiron by AgroPdroand AgrrPd- capsules.(A) Percentageofiron lossin glassquenchedfrom liquid oftonalite
composition as function of time. (B) Comparison of Mgl(Mg + Fe) values for coexistinggarnet, clinopyroxene,and glassquenched
from two runs differing only in capsules,AgroPdroand AgrrPdrr.
HUANG AND WYLLIE: GABBRO-TONALITE-GRANITE-WATER
309
AT 15 kbar
GARNET
Alm 60
15Kb 5%H,O
Alm 20
Py 80
Alm 60
Gr 40;
ALMANDINE
{ + SPESSARTINE)
Alm 6 0
\ P y 40
An 6O
PLAGIOCLASE
(TONALITE}
Alm
Grt
MOLE%
GROSSULAR
20
o
Py 80
u
PYROPE
AsPd\\Au
P soo
Fig. 6. Garnetcompositions
from gabbroand tonalite,partially meltedat 15kbarin Au capsules,
with 50/o
H,O added.(A)
Rangeof analyses
in eachsample.(B) Trendlinesasa function
of temperature
shownby average
analyses.
15Kb
5 YoH.O
u
o-
=
u
F
7oo
RIM
CORE
I-----l
600
Clinopyroxene and amphibole
Clinopyroxene compositions for each run with gabbro
are plotted in Figure 7. There is considerableoverlap of
resultsfrom 700 to 1000"c. The averagecompositions for
each temperature are closely clustered, with no defined
variation as a function of temperature.The averagecomposition of clinopyroxene in tonalite at 1025'C has a
somewhatlower Mg/(Mg + ZFe) value than that in gabbro.
Average amphibole analysesfor gabbro and tonalite,
plotted in Figure 7, are very similar. According to the
classificationof kake (1978), the amphiboles lie closeto
the boundary between magnesio-hornblendeand tschermakitic hornblende.
EXCESSWATER
30
60
40
50
MOLE% AnCONTENT
70
Fig. 8. Plagioclasecompositions from partly melted tonalite
and granite with 50/oHrO at 15 kbar. (A) Individual analysesfrom
runs in Au capsules.(B) Average analysesfrom tonalite as a
function of temperature, compared with the zoned plagioclasein
original tonalite. Results of Piwinskii (1968) on the same rock
at 2 kbar with excessHrO are shown for comparison.
Plagioclase and muscovite
Plagioclase analyses for granite and tonalite runs are
plotted in Figure 8A. Results for granite at 650 and 700"C
remain similar to the 50/oAn in the original plagioclase
Ca2Q
MsSO
MOLE PERCENT
Ca2O
'M920
Fe60
Fig. 7. Clinopyroxene and amphibole compositions in partly melted gabbro and tonalite in Au capsules,with 5% HrO at 15
kbar. The closedsymbols indicate the composition of amphibole in gabbro.
310
AT l5 kbar
HUANG AND WYLLIE: GABBRO-TONALITE-GRANITE-WATER
Fe0
r TONAIITE
1sKb5%H,O
A GABNET
O AMPHIBOIE
. GLASS
O PTAGIOCTASE
1/4Si02
t\\
ailo"l
s00''i
I
I
I
O GLASS
(AsPd)
9500. i
t
t2-
.."4
I
I
I
a-"
gsoo
I
.{o
(!'r*'o
"
'
8000
o
o
MgO+ FeO
Ga0
B
WETGHTPERCENT
Fig. 9. Compositions of glass and crystals quenched from partially melted tonalite at l5 kbar with 50/oHrO added. All runs rvere
in Au capsules, except for the open circles for glassesfrom Ag-Pd capsules. The dashed lines show the average calc-alkaline rock
trend.
Na20+ l(20
A
(Huang and Wyllie, 1973). For the tonalite, the average
plagioclasecompositions (closeto clusteredanalysesdespite broad ranges)show increaseofAn between900 and
800'C, and decreaseof An between 800 and 700"C (Fig.
8B). This is probably related to changesresulting from
proximity to the albite-jadeite reaction and the retrograde
curvature ofplagioclase-out in Figure 2. Piwinskii (1968)
determined the composition of plagioclasein the same
tonalite at 2 kbar with excesswater and found the expected
increase of An content with increasing temperature, as
shown in Figure 8B. There is a decreaseof l-2o/o An
component in plagioclasefrom tonalite run in Ag-Pd capsules,compared with Au capsules.
Muscovite was analyzedin two granite nrns. The paragonite componentsat 700 and 650'C were,respectively5.9
and3.7 molo/0,in contrast to 6.7 molo/oparagonite
in muscovite of the original granite (Huang and Wyllie, 1973).
Glass in tonalite
The composition of glassesquenchedfrom the melting
interval for the tonalite were measuredat 950, 900, and
800'C, with resultscomparedwith mineral analysesin the
AMF and CMS diagramsof Figure 9. The glassesat 800900'C contain 7 1.6-72o/oSiO2, 3.2-4.290NarO, and 2.83.10/oKrO (cf. Table l). The runs in Au capsulesfollow
the calc-alkalinecomposition trend quite well in the AMF
diagram, but deviate somewhat from this trend in the
CMS diagram. The loss of iron from samplesrun in AgPd capsulesis obvious in Figure 9A.
Ifthe differencebetweenthe total ofeach analysisand
1000/o
is equatedto the dissolved HrO in the glasses,and
thus to that dissolved in the original HrO-undersaturated
liquid, the results indicate l2-l5o/o HrO between800 and
950"C, compared with an estimated saturation value at
15 kbar of l9o/o(poorly constrained in Fig. 2).
r GRANITE
15 Kb 5% H,O
. GLASS
O MICA
O PLAGIOCLASE
Na20+ K20
Mgo Ga0
WEIGHT PERCENT
Mso
Fig. 10. Compositions of glass and crystals quenchedfrom partially melted granite in Au capsulesat 15 kbar with 5% HrO
added. The dashed lines show the average calc-alkaline rock trend.
AT l5 kbar
HUANG AND WYLLIE: GABBRO-TONALITE-GRANITE-WATER
r GABBBO
3lI
1/4.SiO2
1sKb 5% H,o
A GARNET
O CI.INOPYROXENE
O AMPHIBOTE
MgO+ FeO
WEIGHTPER CENT
B
Fig. 11. Compositions of crystals coexisting with liquid in partially melted gabbro in Au capsuleswith 5% HrO at 15 kbar.
Gamet compositions at 1000 and 700'C are extrapolated from data measured at other temperatures (Fig. 6). The dashed lines show
the average calc-alkaline rock trend.
Glass in granite
The compositions of glassesquenchedfrom the melting
interval for the granite were measuredat six temperatures.
The analyseswerepublishedbyHuangandWyllie (1981),
and results are plotted in Figure 10. The glasscompositions change fairly regularly with increasing temperature
(although the 800'C glassdeparts from the main trend).
AlrO,
At 650 and 700"C,SiO, is in the range61.4-65.50/0,
in the range 22.6-26.70/o(considerablyhigher than in the
original granite; Table l), KrO is about twice as high as
in the original granite, and NarO is somewhatlower. The
glasscomposition has similarities with syenite. Between
700 and 900'C, SiO, increasesto the 750lolevel, and AlrO,
and KrO reachgranitic levels.The NarO content increases
to the granitic level between 800 and 900"C, where the
plagioclasedissolves. If the difference between the total
analysesand 1000/ois equated to the HrO content of the
glasses,and thus to that of the original HrO-undersaturated liquids, the results indicate that dissolved HrO
between650 and 800oC,with
changedfrom about I 1-130/o
BULKMItrERAL
coMx)stTtoN
l[61[ * KrO
A
MgO
CaO
WEIGHT PER CENT
Fig. 12. The dashed lines show the average calc-alkaline rock trend. (A) Ifliquids are assumed to lie on the dashed trend line,
the bulk mineral composition must be as shown. (B) From partial modal analyses of partially melted gabbro in Au capsulesat 15
kbar with 50/oadded HrO, the analyzed minerals from Figure 11 yield by calculation the bulk mineral compositions plotted as a
function of temperature. The plotted liquid compositions in equilibrium with these calculated mineral assemblagesdiverge significantly from the calc-alkaline trend.
3t2
AT 15kbar
HUANG AND WYLLIE: GABBRO-TONALITE-GRANITE-WATER
15Kblr
1000
5o/oH.O
c- i - - $-TONAL
.q?19
o
o
uJ
PLAGIOCLASE
5 eoo
ul
kE
kcc
=
u
E 8oo
uJ
o-
u
F
700
o
\. AMPHIBOLE
TONALITE
1.5
1.O
o.5
Log Na/K (ATOMICRATIO)
2.O
1OO
Mg/Mg+ Fe (ATOM;C
RATTO)
Fig. 14. Variationof Na/K for coexisting
liquidsandcrystals
HrO. The dashedand solid
at 15 kbar with 50/o
Fig. I 3. Variationof Mil (Me + Fe)for coexistingliquidsand in Au capsules
Arresultsfor gabbroand tonalite,respectively.
crystalsin Au capsules
at l5 kbarwith 5olo
HrO.Thedashedlines linesrepresent
for
rock
rows
show
log(Na/K)
the
compositions.
and solidlinesrepresent
resultsfor gabbroand tonalite,respectively.ArrowsshowMg/Mg + Fe)for the rock compositions.
used to calculatethe averagemodal percentageof crystals
through the melting interval. Finally, these modal percentagesof the crystals and their compositions (Fig. I l)
were used to calculatethe liquid compositions plotted in
Figure l2B. The results diverge from the calc-alkaline
trend in Figure l28. The maximum liquid variation that
Liquid in gabbro
could be produced from the calculated mineral assemThe presenceof quenchcrystalsprecludedreliable mea- blageslies betweenthe pair of arrows. Similaily, if it was
surementof liquid composition for the gabbro.However, assumedthat the liquids lay on the calc-alkalinetrend line
a mass-balancecalculation was made to determine if the in Figure I 28, the correspondingcalculation showedthat
liquid compositions were close to the averagecalc-alka- the liquids could not lie on the trend in Figure l2A,. Thus,
line-rock trend line. The compositions of averagegarnets, althouglr the liquid compositions are not known, it is
clinopyroxenes,and amphiboles transferredfrom Figures demonstratedthat the liquid path doesnot follow the calc6 and 7 to Figure I I were used in the calculations. The alkaline trend.
first assumptionis that the liquid compositions follow the
DrsrnrnurloN
oF ELEMENTSAS FUNcrroN oF
calc-alkalinetrend in terms of the (NarO + KrOlMgOTEMPERATURE
FeO, as illustrated in Figure 12A; then, a seriesof calparameters
culations were made to determine whether the same liqTwo chemical
of significancein magmatic
uids would also follow the trend in terms ofCaO{MgO +
history are the Mg-number, representedin Figure 13 as
FeOfSiO, in Figure l2B.
the atomic ratio Mg/(Mg * Fe), and Na/K, represented
The percentagesof liquid producedthrough the melting in Figure 14 as the log of the atomic ratio. Figures 13 and
interval were point counted petrographically from the run
14 compare the values of these parametersfor the bulk
products. The results are 25, 40, 52, 60, and 800/oby rock compositions of gabbro and tonalite with the values
volume with 5olouncertainty, respectively, at 700, 800, derived from the analysesof minerals and glasses(plotted
'7,
900, 950, and 1000"C.The percentages
ofgarnet, clino- in Figs. 6,
8, and 9) through the melting intervals of
pyroxene, and amphiboles could only be roughly esti- gabbro and tonalite.
mated from the polished sections.The percentagesof the
The Mg-numbersofgarnet (Fig. l3) arelower than those
minerals in Figure I I required to keep the liquids on the of coexisting amphibole, clinopyroxene, and liquid, but
calc-alkaline trend in Figure l2A, were calculated. The at high temperaturesthey are similar to or slightly lower
bulk crystal composition required to yield the liquids fol- than those of the rocks; they decreaserapidly with delowing the trend is indicated by the open circle in Figure creasing temperature in both gabbro and tonalite. The
l2A; the shaded area indicates the limited range of the variation of Mg-numbers of clinopyroxene with temperbulk crystal composition when the maximum variation ature is small, but the Mg-numbers are much higher than
ofthe trend is taken into account,giving the rangebetween the coexisting garnet and amphibole, and higher than those
the pair of arrows. The bulk crystal composition was then ofthe rocks.
a decreaseto about 7-80loassociatedwith the increased
melting where plagioclaseand muscovite disappear between 800 and 900"C. According to Figure 3, aboul 22o/o
HrO is required for saturation of the liquid at 15 kbkar.
AT l5 kbar
HUANG AND WYLLIE: GABBRO-TONALITE-GRANITE-WATER
1100
10
OC
TEMPERATURE
900
900
2.5
AC
TEMPERATURE
800
30
2.O
5
4
3
JIJ
3.0
20
1.5
t'o
tn*o
KD2
2.O
L nK p
KP
0.5
1
0.5
o
-.--- GABBRO
o.o
TONALITE
" " " " R & c ( 1 9 7 4 ) -0.5
G R E E (N1 9 7 7 } - . - " - - E & c ( 1 9 7 9 )
7.O
8.O
10.O
9.0
11.0
1fi oKx10a
Fig. 15. The partition coefrcientsof Mg/Febetweencoexistingmineralsand glassas a functionof temperature
with 50/o
HrO at 15 kbar.The dashedand solidlinesarefor gabbroand
tonalitecompositions,
respectively.
Thedottedanddash{ouble
dot lines grve Ko valuesbetweenclinopyroxeneand garnetin
anhydroustholeiiteat 15 kbar, interpolatedfrom Raheimand
Green(1974)andEllisandGreen(1979).Thedash-dotlinegives
Ko valuesbetweenglassand garnetfrom a pelitecomposition
(Green,1977).
The compositions of amphibole in both gabbro and
tonalite are similar in terms of Mg-number, which is close
to the value for the gabbro, but higher than that for the
tonalite (Fie. l3). Figures13 and 14 also show that amphibole variations in Mg-number and Na/K with temperature are small. The Mg-number of the amphibole appears to be nearly independent of bulk composition, as
well as temperature. Na/K ratio, on the contrary, is significantly controlled by bulk composition. The Na/K for
amphibole in tonalite is slightly lower than that of the
tonalite, whereas amphibole in gabbro changes from
slightly lower to slightly higher than that of the gabbro
between800 and 700'C. The Na/K ratio of plagioclasein
gabbro, considerably higher than that in tonalite, shows
little variation with temperature.
The Mg-numbers of successiveliquids are lower than
the coexisting amphiboles but higher than the coexisting
garnet. The Na/K ratio of liquids increasesslightly with
degreeof partial melting.
Partition coefficientsamong coexisting phases
Figure l5 showsthe partition coefficientsof Mg/Fe ratio
as a function of temperaturefor gabbro and tonalite compositions. For gabbro, the partition coefficientsof Mg/Fe
betweencoexistingclinopyroxene and garnet,K": (M9
Fe)""./(MglFe)o",and betweenclinopyroxene and amphibole decreasewith increasingtemperature. For tonalite,
the KD values betweenamphibole and garnet, amphibole
and glass,and glassand garnet decreasewith increasing
temperature.
Raheim and Green (1974) experimentally determined
the temperature and pressuredependenceof the Fe-Mg
1.0
0
9.0
9.5
1/T oKxlOo
10.0
10.5
Fig. 16. The partition coefrcieRts
of Na/K and CalNa bemineralsandglassasa functionof temperature
tweencoexisting
with 50/oHrO at 15 kbar. The dashedand solid lines are for
gabbroandtonalite,respectively.
partition coefrciont (K") for coexisting garnet and clinopyroxene. Our results on basaltic composition for Ko vs.
temperatureare in good agreementwith theirs, as shown
in Figure 15. Note that the definition of Ko used in our
study, (Mg/Fe).o./(Mg/Fe)"",is different from theirs, (Fe/
Mg)""/(Fe/Mg)"o.,but the actualK" value is the same.Ellis
and Green (1979) redetermined the coefrcient by taking
the effect of Ca on the garnet-clinopyroxeneFe-Mg equilibrium into consideration.They derived a new empirical
equation by incorporating the effectofthe Ca component
in garnet. When the equation was applied to calculateKo
by using the Ca component in garnetsat 800, 900, and
950"C from our experiments, the Ko values are significantly higher than our experimentally determined data.
This suggeststhat some other parameter that was not
identified by Ellis and Green (1979) may significantly affect the Ko value. These unknown parameterscausethe
inconsistencybetweenour data and results calculated from
Ellis and Green( I 979). However, we are unableto explain
why our data are consistent with those of Raheim and
Green(1974).
The partition coefficients of Mg/Fe between hydrous
glassand garnet of a model pelite composition at l5 kbar
(Green, 1977)also show closeagreementwith our tonalite
data, in spite of different bulk composition. However, the
same coefrcient for Mt. Hood andesite determined by
Allen and Boettcher(1983) at 900"Cand 19 kbar is significantlylower (0.8)than ours(1.4 at 900C1andl5 kbar).
Figure 16 showsthe partition coefficientsof Na/K and
CalNa ratios as a function of temperature. The different
partitioning of Na/K and CalNa betweencoexisting plagioclaseand liquid is evident. The Na/K ratio in plagroclase is much higher than that in glass,but the CalNa
ratio is lower in plagioclasethan in glass.The K" of Na/K
betweenplagioclaseand glassand that of CalNa between
glass and plagioclasedecreasewith increasing temperature. In contrast to plagioclase,amphibole in tonalite has
314
HUANG AND WYLLIE: GABBRO-TONALITE-GRANITE-WATER
AT 15 kbar
lower Na/K and higher CalNa than the coexistingliquid.
The K" of Na/K betweenamphibole and glassdecreases
with increasing temperature whereas the Ko of CalNa
between amphibole and glass increaseswith increasing
temperature.
compositional range,with overall higher Mg-numbers than
ours (Fig. l3), and on this basis they concludedthat amphibole fractionation could cause iron-enrichment in
magmas. However, this contradiction now appears less
pronounced, since the new data of Allen and Boettcher
(1983) on the samesamples,with lessiron loss,gives Mg/
Pnrnor,ocrcAr, AppLrcATroNS
(Mg + Fe) values for amphibole closer to our results in
The generationof magmasby differentiation of a parent Figure 13.
or by partial fusion is complex and probably intermediate
The variety of amphibole compositions determined in
betweenequilibrium and fractional conditions. However, experimental studies, and especially those of Allen and
equilibrium experimental results are required as a starting Boettcher(1978, 1983)under a variety of experimental
point. Our experimentally determined phase diagrams, conditions, implies that many factors including pressure,
although not representingperfect equilibrium, neverthe- water content, noble-metal capsule,and oxygenfugacity,
lessdefinethe phaserelationshipswell enoughto elucidate as well as liquid composition and temperature,may affect
processes.The resultson liquid compositionsand the dis- amphibole composition. Similarly, many factors will intribution of elementsbetweencoexistingcrystalsand liq- fluence amphibole composition in natural magmas. In
uids are reliable, having beenmeasuredfor runs complete order to evaluate the actual role of amphibole fractionin Au capsules.They provide some basis for prediction ation during magma differentiation, a detailed and sysof the differentiation trends of hydrous basaltic and an- tematic experimental program for determination of amdesitic magmas in the deep crust or uppermost mantle phibole composition as a function of the variables is
and for the products of anatexis of deep crustal rocks required.
composedof gabbro, tonalite, and granite or their metamorphic equivalents.A pressureof l5 kbar is attained at Magma differentiation at 55-km depth
a depth of about 55 km in thickened crust of orogenic
Although fractionation of amphibole from basaltic
belts.
magmamay be capableofproducing differentiatedliquids
with Mg/Fe correspondingto those of the calc-alkaline
Amphibole fractionation at 55-km depth
trend at leastduring the early stagesofdifferentiation, our
It has been suggestedthat amphibole fractionation ex- results indicate that it is not the dominant process in
erts a major control on the formation and differentiation hydrous basalts at 55-km depth. Fractionation of clinoof the calc-alkalinerock suite (e.g.,Boettcher, 1973;C,aw- pyroxene and garnet togetheris the main processcausing
thorn and O'Hara, 1976).Figure l3 showsthat Mg-num- changesin composition ofhydrous basalt, becausethese
bers of amphibole in gabbro (basalt)and tonalite (andes- two minerals are the liquidus phases,with subordinate
ite) lie in a small range from 56 to 60 and are nearly amphibole joining them below the liquidus; these three
independent of temperature and bulk composition. The minerals are coprecipitated through several hundred deamphiboles have Mg-numbers close to those of basalt, grees (Fig. l). The analysesand calculations plotted in
but much higher than those of andesite,as well as of the Figures I I and I 2 confirm that the differentiating hydrous
liquids in the crystallization interval of andesite. These basalticliquid doesnot follow the calc-alkalinetrend. The
Mg-numbers are in good agreementwith the values be- crystallization of clinopyroxene with high Mg-numbers
tween 59 and 62 reported by Holloway and Burnham averaging70 tends to produce a liquid with high Fe/Mg,
(1972) for amphibole in basalt of similar composition at but this effect is offset at least at the lower temperatures
5 and 8 kbar.
by the low Mg-numbers of garnet(Fig. l3). Crystallization
Ringwood (1974) usedthe data of Holloway and Burn- of amphibole controls Na/K in liquids until the low temham (1972) to support his proposal that fractionation of peratureswhere plagioclasebegins to crystallize.
amphibole from basaltic magmas may not cause ironGarnet and subordinateclinopyroxeneare also the liqenrichment in diferentiating basalt.Our resultsfor gabbro uidus minerals in crystallization of hydrous andesite(Fig.
are consistentwith this conclusionasillustrated in Figures 2), joined within a small temperatureinterval by amphiI lA and 13. However, our data demonstratein addition bole and plagioclase,with biotite and quartz appearingat
that precipitation of amphibole from hydrous andesitic lower temperatures.Precipitation of clinopyroxene and
magma at this depth may contribute to iron-enrichment amphibole, with Mg-numbers much higher than that of
of liquids (Fig. l3). The latter conclusion is contradicted the liquids, tends to causeiron-enrichment of the liquids,
by the data reported by Green (1972) on amphiboles in but garnetoperatesin the other direction (Fig. 9), asshown
andesite(Mg-number: 48) with 2o/otolDo/oHrObetween by values of K" for Hb/glass and glass/Gain Figure 15.
5 and 15 kbar. He reported Mg-numbers for amphibole The glassanalysesplotted in Figure 9 indicate that liquid
lower than thoseplotted in Figure 13. Allen et al. (1975) compositions in the middle-temperature range approxiand Allen and Boettcher (1978) also reported amphibole mate the calc-alkalinetrend, with slight divergencetoward
analysesfrom basaltsand andesitesthrough pressureranges higher CaOl(MgO + FeO) with decreasingtemperature.
including 15 kbar with H,O-CO, mixtures giving varia- The Na/K is mainly controlled by the relative amount of
tions in X"ro. Their amphibole analysescovered a wide amphibole and plagioclaseprecipitating. Becauseof the
AT 15kbar
HUANG AND WYLLIE: GABBRO-TONALITE-GRANITE-WATER
very high K" for PVglass(Fig. 16), we expectthe effect of
plagioclaseto dominate, causingdecreaseof Na/K in the
liquid with decreasingtemperature, which is consistent
with the glassanalysesin Figure 14.
We do not expect to find granite or rhyolite magma
differentiating at a depth of 55 km, but Figures 3 and l0
illustrate the controls on suchbehavior. Hydrous rhyolite
liquid would precipitatequartz alone (excludingthe metastablecorundum) through 100"Cor more, causingdepletion of the liquid in SiOr, as illustrated by the glasscompositions plotted in Figure l0B. With subsequent
coprecipitation of plagioclase,the depletion of the liquid
in SiO, continues,with the formation of liquid similar in
composition to syenite (Huang and Wyllie, 198l).
Anatexis of lower crust, thickened to 60 km
It is generallyassumedthat the baseof the continental
crust is composeddominantly of the metamorphic equivalentsofgabbro and tonalite, hydrated to greateror lesser
extent. At compressiveplate boundaries with thickened
crust, or during the Archean when localized subduction
might have been more prevalent, we consider also the
prospect that granitic rocks could be carried as deep as
for theserocks at 5555-60 km. The mineral assemblages
km depth are given in the subsolidusregions ofFigures
1,2,3, and, in combination,in Figure 4. Wyllie (1977)
reviewed the conditions for anatexisof the samerangeof
rock compositions at l0 kbar, correspondingto crustal
thickness of 35-40 km, concluding that HrO-undersaturated granite or rhyolite magma was the normal product
of progressive metamorphism of many crustal rocks at
moderate depths and that the generationof magmasless
siliceous required the addition of heat or magmatic material or both from the underlying mantle.
The mineralogy in a thickened crust differs from that
in the normal crust by the addition of the high-pressure
minerals garnet and clinopyroxene in the gabbro and tonalite, but the muscovite granite retains essentially the
same mineral assemblage.Wyllie (1977) reviewed the
conditions for partial fusion of the rocks both with pore
fluid and without. In this paper, we consider only the
anatexisofthe rocks in the presenceofaqueouspore fluid,
using compositions of glassesmeasuredin runs with 50/o
HrO as a possible guide to liquid paths. Unfortunately,
we have no glasscompositions for the gabbro, and it was
difficult to obtain reliable glass measurementsnear the
solidus with small fractions of melting.
Anatexis ofgabbro and tonalite under theseconditions
involves similar minerals, with a quartz field-boundary
controlling the composition of the first liquid. The first
liquids developedfrom either rock are thereforesiliceous,
and probably similar in composition. Successiveliquids,
however, will follow different paths becausequartz and
plagioclase(with biotite) remain among the residual minerals through a wider temperature in the progressivefusion oftonalite than ofgabbro. The analysesand calculations based on Figures I I and 12 indicate that partial
fusion of hydrous gabbro at this depth does not yield a
315
seriesof liquids with chemical variation correspondingto
the calc-alkalinerock series.The glassanalysesfrom partially melted tonalite indicate a liquid trend with CaO/
(MgO + FeO) somewhat higher than the calc-alkaline
trend. At 900'C, with a generousallowance of 50/oHrO,
the residual mineral assemblageHb + Ga + Cpx + Pl
NarO, and
coexistswith liquid containingT2o/oSiOr, 4.2010
3.10/oKrO. Despite its high SiO, content, this does not
correspondto a granite becausegranite under these conditions would precipitate only quartz through a wide temperature interval below its hydrous liquidus (seeFig. 3).
The evidence indicates that if magmas are generatedby
partial fusion of gabbro and tonalite in thickened continental crust, they do not correspond to the rocks composing the calc-alkaline batholiths, nor do they give rise
to the batholiths by fractional crystallization.
The partial melting of deep granite proceedsalong a
path quite different from that for tonalite and gabbro.The
alkali feldspar dissolves first, yielding a potassic liquid,
and subsequentfusion is controlled essentially by the
quartz-plagioclaseboundary. The quartz volume expands
considerablyat high water pressures,and the liquid compositions coexistingwith quartz and feldspar arethus forced
to lower SiO, contents (Huang and Wyllie, 1975).This is
indicated in Figure l0B. The relatively low-SiOr, highKrO liquids at 650 and 700'C have some characteristics
of syenite.
Trondhjemite is an important component of Archean
igneous activity and of some more recent calc-alkaline
batholiths. The data in this paperdo not provide evidence
for the conditions ofgeneration oftrondhjemites, but they
do demonstratesome conditions unsuitable for their formation. Na/K ratios as high as 2-4 are characteristicof
trondhjemites. None of the hydrous glassesanalyzed in
the tonalite (Fig. la) or granite approachedthis value, and
extrapolation from the analyzedglassesthrough other parts
of the determined phaserelationships (Figs. 2 and 3) suggest that trondhjemites are not simply related to either
granites or tonalites at 15 kbar. We have no analysesof
glassesfrom the gabbro,but within the phaserelationships
at 15 kbar (Fig. l), there appearsto be a prospect that
liquids of appropriate composition might be found at
moderate HrO contents near the plagioclase-out and
quartz-out boundaries. Holloway and Burnham (1972)
and Helz (1976) analyzedglassesfrom partially melted
hydrous basaltsat lower pressures,reporting trondhjemitic to tonalitic compositions with l5-300/omelting, 850
to 1000'C.
AcxNowr,nocMENTS
This researchwas supportedby the Earth ScienceSectionof
the National ScienceFoundation,NSF grant EAR 84-07115.
Companycontributedto manuscript
ExxonProductionResearch
preparationcosts.
RnrnnrNcns
Allen, J.C.,and Boettcher,A.L. (1978)Amphibolesin andesite
andbasalt:II. Stabilityasa functionof P-T-f',o-for. American
Mineralogist,63, 1074-1087.
316
-
HUANG AND WYLLIE: GABBRO-TONALITE-GRANITE-WATERAT
15kbar
(1983) The stability of amphibole in andesiteand basalt
at high pressures.American Mineralogist, 68, 307-314.
Allen, J.C.,Boettcher,A.L., and Marland,G. (1975)Amphiboles
in andesite and basalt: I. Stability as a function of P-T-fo,.
American Mineralogist,60, 1069-1085.
Bateman,P.C., Clark, L., Huber, N.K., Moore, J.G., and Rhinehart, C.D. (1963) The Sierra Nevada batholith, a synthesisof
recent work across the central part. U.S. Geological Survey
ProfessionalPaper,414-D, Dl-D46.
Boettcher,A.L. (1973)Volcanism and orogenicbelts-The origin
of andesites.Tectonophysics,17, 204-233.
Boettcher,A.L., and Wyllie, P.J. (1968) The quartz-coesitetransition measuredin the presenceof a silicate liquid and calibration of piston-cylinder apparatus. Contributions to Mineralogy and Petrology, 17, 224-232.
Boyd, F.R., and England,J.L. (1960) Apparatus for phase-equilibrium measurementsat pressuresup to 50 kilobars and temperaturesup to 1750"rC.Journal ofGeophysical Research,65,
741-748.
Boyd, F.R., Bell, P.M., England,J.L., and Gilbert, M.C. (1967).
Pressuremeasurementin single-stageapparatus.CarnegieInstitute of Washington Year Book 65, 410-414.
Cawthorn, R.G., and O'Hara, M.J. (1976) Amphibole fractionation in calc-alkaline magma genesis.American Journal of
Science,276,309-329.
Ellis, D.J., and Green, D.H. (1979) An experimentalstudy of the
etrectof Ca upon garnet-clinopyroxeneFe-Mg exchangeequilibria. Contributions to Mineralogy and Petrology, 71,13-32.
Green,T.H. (1972)Crystallizationofcalc-alkalineandesiteunder
controlled high-pressurehydrous conditions. Contributions to
Mineralogy and Petrology, 34, 150-166.
(1977) Garnet in silicic liquids and its possibleuse as a
P- 7 indicator. Contributions to Mineralogy and Petrology,65,
59-67.
(1982) Anatexis of mafic crust and high pressurecrystallization of andesite.In R.S. Thorpe, Ed. Andesites:Orogenic
andesitesand related rocks, 465-487. John Wiley, Chichester.
Green, T.H., and Ringwood, A.E. (1968) Genesisof the calcalkaline igneousrock suite. Contributions to Mineralogy and
Petrology,18, 105-162.
(1972) Crystallization of garnet-bearingrhyodacite under
high-pressurehydrous conditions. Geological Society of Australia Journal. 19. 203J12.
Helz, R.T. ( 1976)Phaserelationsofbasaltsin their melting ranges
at Prro : 5 kb. ft. II. Melt compositions.Journal of Petrology,
Ito, K., and Kennedy, G.C. (1968) Melting and phaserelations
in the plant tholeiitelherzolite-nepheline basaniteto 40 kilobars with geologicalimplications. Contributions to Mineralogy
and Petrology,19, 177-211.
Lambert, I.B., and Wyllie, P.J. (1972) Melting of gabbro (quartz
eclogite)with excesswater to 35 kilobars, with geologicalapplications. Journal ofGeology, 80, 693-708.
(1974) Melting of tonalite and crystallization of andesite
liquid with excesswater to 30 kilobars. Journal ofGeology,
E2, E8-97.
Leake,B.E. (1978)Nomenclatureofamphiboles. CanadianMineralogist,16, 501-520.
Merrill, R. 8., and Wyllie, P.J.( 1973) Iron absorptionby platinum
capsulesin high pressurerock melting experiments.American
Mineralogist, 58, 16-20.
Nehru, C.E.,and Wyllie, P.J,(1975)Compositionsofglassesfrom
St. Paul's peridotite partially melted at 20 kilobars. Journal of
Geology,83,455-471.
Piwinskii, A.J. (1968) Experimental studies of igneousrock series. Central Sierra Nevada batholith, Califomia. Journal of
Geology,76,548-570,
(1973a)Experimental studiesofgranitoids from the central and southern Coast Ranges,California. TschermaksMineralogischeund PetrographischeMitteilungen, 20, 107-l 30.
(l 973b) Experimentalstudiesofigneous rock series,Central Sierra Nevada batholith, California: Part 2. Neues Jahrbuch ftir Mineralogie Monatshefte, 6, 193-215.
(1975) Experimental studies ofgranitoid rocks near the
San Andreas fault zone in the Coast-TransverseRanges,Mohave Desert, California. Tectonophysics,25, 217-23 L.
Piwinskii, A,J., and Wyllie, P.J. (1968) Experimental studies of
igneousrock series:A zoned pluton in the Wallowa batholith,
Oregon. Journal of Geology, 76, 205-234.
(1970) Experimental studiesofigneous rock series:Felsic
body suite from the Needle Point pluton, Wallowa batholith,
Oregon. Joumal of Geology, 78, 52-76.
Raheim, A., and Green,D. H. ( I 974) Experimentaldetermination
ofthe temperatureand prossuredependenceofthe Fe-Mg partition coefficientfor coexistinggarnetand clinopyroxene.Contributions to Mineralogy and Petrology, 48, 179-203.
Ringwood, A.E. (1974) The petrological evolution of island arc
systems.GeologicalSocietyof London Journal, I 30, I 83-204.
Stern, C.R., and Wyllie, P.J. (1973) Melting relations and basaltandesite-rhyolite-HrO and a pelagic red clay at 30 kilobars.
Contributions to Mineralogy and Petrology,42,313-323.
r7, r39-r93.
(1975) Efect ofiron absorption by noble-metal capsules
Holloway, J.R. (1973) The systempargasite-HrO-COr:A model
on phase boundaries in rock-melting experiments at 30 kilofor melting of a hydrous mineral with a mixed-volatile fluid.
bars. American Mineralogist, 60, 681-689.
I. Experimental resultsto 8 kb. Geochimica et Cosmochimica (1978) Phasecompositions through crystallization interActa,37, 651-666.
vals in basalt-andesite-HrOat 30 kb with implications for
Holloway, J.R., and Burnham, C.W. (1972) Melting relations of
subduction zone magmas. American Mineralogist, 63, 641basaltwith equilibrium water pressurelessthan total pressure.
663.
Journal ofPetrology, 13, l-29.
(1981) Phaserelationshipsof I-type granite with H,O to
Huang, W.L., and Wyllie, P.J. (1973) Melting relations of mus35 kb: The Dinkey Lakesbiotite granitefrom the SierraNevada
covite-graniteto 35 kb asa model for fusion of metamorphosed
batlrolith. Journal of GeophysicalResearch,86, 10412-10422.
subducted oceanic sediments.Contributions to Mineralogy and Wyllie, P.J. (1977) Crustal anatexis: An experimental review.
Petrology,42, l-14.
Tectonophysics,13, 4l-71.
(1974) Melting relations of muscovite with quartz and (1979) Magmasand volatilecomponents.American Minsanidinein the KrO-Al,O-SiOr-HrO systemto 30 kilobars and
eralogist,64, 469-500.
outline of paragonite melting relations. American Journal of
Yoder, H.S., and Kushiro, L (1969) Melting of a hydrous phase:
Science,274,378-395.
Phlogopite. American Journal of Science,267 -A., 5 58-582.
(1975) Melting reactions in the system NaAlSi3OEKAlSi3Os-SiO,to 35 kilobars, dry and with excesswater. Journal of Geology,83, 737-7 48.
(1981) Phaserelationshipsof S-typegranite with HrO to
35 kb: Muscovite granite from Harney Peak, South Dakota. MeNuscnlPT REcETvEDApnrr- 8, 1985
Journal of GeophysicalResearch,86, 10I 5-1029.
MeNuscnrpr AccEp"rEDNovENansn 4, 1985