American Mineralogist, Volume 79, pages353-369, 1994
Chlorine, fluid immiscibility, and degassingin peralkaline magmas
from Pantelleria, Italy
Jlcon
B. LownNsrnRN*
Mineral ResourcesDepartment, Geological Survey ofJapan, l-l-3 Higashi, Tsukuba, Ibaraki 305, Japan
Ansrnlcr
This paper documentsimmiscibility among vapor, highly salineliquid, and silicate melt
during the crystallization of peralkaline rhyolites from Pantelleia, Italy, prior to their
eruption. Experimentsconductedin a mufre furnace and with a high-temperatureheating
stagerevealedthree major types of silicate melt inclusions trapped in quartz phenocrysts.
After entrapment in the host phenocryst, type I inclusions contained silicate melt. Type
II inclusions contained silicate melt + hydrosaline melt (-60-80 wto/oNaCl equivalent),
and type III inclusions contained silicate melt + HrO-CO, vapor. Two inclusions contained all three immiscible fluids: vapor, hydrosaline melt, and silicate melt. Fluid inclusions within outgassedmatrix glass,viewed at room temperature, are interpreted as the
crystallized equivalents of the hydrosaline melts within type II inclusions. These inclusions, 2-10 pm in size, consist ofa bubble typically surrounded by a spherical shell of
halite.
The presenceof both vapor and hydrosaline melt in the magma indicates that the
pantellerite was saturatedwith subcritical NaCl-HrO fluids. At a given temperature and
pressure,the fixed activity of Cl in these two fluids delermines the activity and concentration of Cl in the silicate melt. The high concentrations of Cl in these pantellerites
(-9000 ppm) are thus a function of the low activity coefficient for NaCl in pantellerite
relative to metaluminous silicate liquids. The Cl contentsof Pantellerianrhyolites indicate
equilibration at pressuresbetween 50 and 100 MPa. The high Cl contents of outgassed
pantelleritesmay be due to minimal loss of HCI (not NaCl) during eruption, as compared
with metaluminous rhyolites, which exsolve more HCI-rich vapors.
Discrepanciesbetween the results of heating-stageexperiments and longer muffie-furnace experiments indicate that measurementsof melting and homogenization temperatures of melt inclusions may not be accurateunless sufficient time (> I h) is allowed for
equilibration at magmatic temperatures.
fNtnoouctloN
Experimental studies show that the NaCl-HrO system
is characterizedby immiscibility under a wide range of
pressuresand temperaturesin the shallow crust (Sourirajan and Kennedy, 1962; Bodnar et al., 1985; Chou,
1987). Furtherrnore, researchon the silicate melt-HrOalkali chloride ternary indicatesthat the Cl and HrO contents of many magmasare sufficient to saturatethe melt
with immiscible vapor and liquid (hydrosaline melt)
phases(Shinoharaet al., 1989;Malinin et al., 1989;Metrich and Rutherford, 1992; Webster, 1992a).Evidence
for immiscibility betweensilicate and HrO-NaCl fluids is
widespread in fluid inclusions found in phenocrysts of
intrusive igneous bodies such as granites, syenites,and
porphyry ore deposits (Roedder, 1972, 1984, 1992;
Roedder and Coombs, 1967; Frost and Touret, 1989:
tPresent address:U.S. Geological
Survey, M.S. 910, 345
Middlefield Road, Menlo Park, California 94025, U.S.A.
0003-o04x/94l0304-0353$02.00
Hansteen, 1989; Frezzotti, 1992). Some silicate melts
show evidencefor saturation with both vapor and hydrosalinemelt (e.g.,Frost and Touret, 1989).BecauseNaClHrO fluids are precursorsto ore-forming hydrothermal
solutions,it is important to determinethe factorsthat control their evolution and composition. Volcanic rocks are
ideal for such studiesbecausethey contain quenchedmatrix and glass inclusions that can preserve the concentrations of magmatic volatiles during preeruptivedegassing.
Studiesof fluid inclusions in phenocryst-poorvolcanic
rocks have only rarely been undertaken. This stems, in
part, from the scarcityoffluid inclusionsin volcanic rocks
(Tuttle, 1952), despite the oft-repeated conclusion that
many igneoussystemsare fluid-saturated during crystallization and prior to eruption (Newmanet al., 1988;Andersonet al., 1989;Luhr, 1990;Lowensternet al., l99l;
Lowenstern, 1993).Ofthe handful of studiesof coexisting
fluid and melt inclusionsin volcanic systems,severalhave
focusedon rhyolites from Pantelleia, ltaly. Abstracts by
Clocchiatti et al. (1990) and Solovova et al. (1991) re-
353
354
LOWENSTERN: CHLORINE IN PERALKALINE MAGMA
320
A
300 E
oo.
tr
t.ta
I
2€,o:
E
f
260 E
t
240 c6
J
20
I
11000
200
B
10000
=
a 9000
a
t-a
o
8000
I
'-E
: Tooo
looa
a a
6000
5000'
30
40
45
50
Thorium(ppm)
Fig. l. Trace-element
trendsfor glassy,unalteredPantellerian rhyoliteswith an agpaiticindex >1.75. (A) Relativelyincompatibleelements
Th andLa (distributioncoefficient
between
bulk crystaland melt <0.01 and -0.1, respectively;
Mahood
andStimac,1990)increase
sympathetically
duringfractionation.
(B) Cl concentrations
do not correlatewith indicesofdifferentiation suchasTh. Th and La databy INAA, Cl by specificion
perelectrode(G.A. Mahood,unpublisheddata).All analyses
formedon wholerocksandglassseparates;
Cl contentsmayhave
beenloweredslightlyby syneruptiveoutgassing.
Uncerlainties
(2o)areapproximately
the sizeof the datapoints.
35
ported finding hydrosaline melts within silicate-melt inclusions from this locality. Lowenstern et al. (1991)
showed that COr- and Cu-bearing vapor bubbles were
also presentin some melt inclusions.De Vivo etaI. (1992)
found fluid inclusions of a presumed magmatic vapor
phase in syenite nodules ejectedduring eruption. Kovalenko et al. (1993)observedsaltclusterswithin outgassed
pantellerite glassand concludedthat the erupting pantellerite was saturatedwith both vapor and salt-rich (hydrosaline)melt.
In this report, I corroborate the findings ofthis earlier
research on pantellerites and elaborate upon them by
documentingimmiscibility amonghydrosalinemelts, lowdensity vapors, and silicate melts at high temperatureand
pressure.Seventy-ninesilicate-melt inclusions in 28 phenocrysts were observed during l-atm experiments using
a high-temperatureheating stageto record their behavior
upon heating, consequent pressurization, and melting
(Clocchiatti, 1972, 1975;' Benhamou and Clocchiatti,
1976;Clocchiattiet al., 1990;Solovovaet al., l99l;Frezzotti, 1992). This technique allows collection of photographic evidence and thermometric data about melt inclusions and the fluids trapped within them. As shown
below, one can then evaluate both the effectsof fluid im-
miscibility and subsequentdegassingon the behavior and
distribution of Cl in volcanic systems.Additionally, this
paper attempts to explain the behavior of melt inclusions
during cooling in nature and in the laboratory, permitting
greaterinsight into petrologic and volcanologicprocesses.
PlNrnr,r.BnrrEs, Cl, AND MELT TNCLUSIoNS
Pantelleria is a volcanic island that sits astride a
drowned continental rift in the Strait of Sicily, between
Sicily and Tunisia. The central part of the island is composed predominantly ofpantellerites and associatedtrachytes and pantelleritic trachytes, though basalts are
common on the island periphery (Civetta et al., 1984;
Mahood and Hildreth, 1986).The island is the type locality of pantellerites, which are quartz-normative peralkaline rhyolites having agpaitic indices [A.I.: molar
(Na,O + KrO)/AlrO3l commonly > 1.75(Macdonaldand
Bailey, I 973). Pantelleritesare notable for their unusually
high concentrations of total Fe (i.e., FeO,o,- 80/o)and
trace elements such as Zr, Nb, La, Hf, Zn, Mo, Rb, and
Th. They have among the highest Cl contents of any igneousrocks on Earth, some samplesreachingover I wto/o
(Kovalenko et al., 1988;Fig. I of this study).At Pantelleria, Cl abundancesare relatively constant, unlike incompatible elementsthat increasein concentration during magmatic differentiation (Fig. l). One explanation for
this behavior can be inferred from the experiments of
Metrich and Rutherford (1992), as shown in Figure 2.
Those workers equilibrated pantellerites and other rhyolites with solutions of alkali chloride and HrO and found
that the Cl content of pantelleritesremainedconslanteven
as the Cl content ofthe system increased.This occurred
as long as the composition of the bulk fluid (vapor +
hydrosaline melt) lay within the field of immiscibility for
the system NaCl-HrO (all pressures<160 MPa at 800
'C: Chou, 1987).For example,at 830 "C, the silicatemelt
contained-9500 ppm Cl at 50 MPa and -8500 ppm at
100 MPa. Under those conditions,the Cl content of the
silicate melt must have been buffered at constant composition becauseof the fixed activities of HrO and NaCl
in the two immiscible fluids (Shinoharaetal., 1989;Malinin et al., 1989).Therefore,high and constantCl concentrations in strongly peralkaline magmasare inevitable
when they are saturatedwith immiscible HrO-NaCl fluids (vapor + hydrosaline melt) in near-surfacemagma
reservolrs.
Evidence for shallow (-2-4 km) pantellerite magma
chambersis abundant and includes the formation of calderas during small-volume eruptions, the absenceof exotic lithic fragments in volcanic ejecta, and phase equilibrium constraints (Mahood, 1984). Evidence for
saturation of pantellerite with respectto two NaCl-HrO
fluids is shown below.
The samples for this study come from two, crystalpoor, glassy lava (obsidian) flows approximately ten
thousand years old from Contradas Sciuvechi and Valenza(P32and P104,respectively:Table l) of Pantelleria.
The units have been described previously by Mahood
LOWENSTERN: CHLORINE IN PERALKALINE MAGMA
Trale 1. Major-element
compositionsof wholerocks (WR)and
matrix glasses(G) used in this study
I End-member
lmmiscible
Fluids
Datatrom Metrichand Rutherford(1992)
o 50MPa
q)
A 100 MPa
Sample-
_ SOMpa
.9 r.o
o
I Paffelleribs lrom
J Pantslleria
c
.905
o
s
>
1
10
Wt%NaClin Fluid
355
100
sio,
Tio,
Al2o3
FeO,.
MnO
Mgo
CaO
Na"O
K.o
P"ou
F
H"O
P32WR
P32G
P 1 0 4W R
69.3
0.36
8.18
8.60
0.29
0.10
0.34
6.76
4.45
<0.05
0.24
0.77
o.28
0.55
99.17
1.95
70.3
0.41
7.61
9.15
0.30
0.11
0.36
6.81
4.59
na
na
0.92
0.10.'
0.48
10 0 . 18
2.12
69.6
0.36
8.13
8.71
0.29
0.11
0.35
6.38
4.41
<0.05
O.27
0.80
0.43
0.58
99.31
1.94
P 1 0 4G
71.O
0.42
7. 4 1
8.26
0.30
0.06
0.31
6.73
4.26
<0.05
na
0.88
0.15*
0.47
99.36
2.12
Fig. 2. Cl contentof pantelleritemelt vs. bulk composition - F , C t : O
Total
of NaCl-HrOfluid at 830 "C. The Cl contentsof evolvedpanAgpaitic index
telleritesfrom Pantelleria(7500-9000ppm)areconsistent
with
'Samples P32 WR and P32 G from Mahoodand Stimac(1990).Pl04
their havingequilibratedat 50- 100MPa with subcriticalNaClthis study.
HrO fluids.At 50 and 100MPa, Cl concentrarions
in the melt WR from G.A. Mahood (unpublisheddata) and P104 G from
Analytical methods are describedin Mahood and Stimac (1990).
arefixedaslongasthe coexistingbulk fluid lieswithin the field
'. H,O content of matrix glass from Lowenstern and Mahood (1991).
of immiscibilityfor the systemNaCl-HrO(datafrom Metrich
andRutherford,1992).As longasbothnon-sihcate
fluids(vapor
(at a
andhydrosaline
melt)arepresent,their fixedcompositions
giventemperatureand pressure)
requirethat the actiwity,and
Vor-uvtnrnrc BEHAvToRoF MELT INcLUSToNS
thusconcentration,
of Cl in themeltremainsconstant.
At higher
pressures,
Melt inclusions can be understoodwithin the samethethe fluid is supercritical,
and any increasein Cl content of the systemresultsin increasing
Cl concentration
in the oretical and experimental framework that has been demelt andfluid (for all fluid compositions).
The interpretivecurves veloped for other types of fluid inclusions (Hollister and
arebasedon Shinohara
et al. (1989).The compositions
ofthe Crawford. 1982;Roedder, 1984).Relative to silicatemelt,
vaporandhydrosaline
melt at 50 and 100MPa(at 825oC;from quartz and most other magmatic phenocrystsare incomBodnaret al., 1985)definethe pivot pointsof the curves.The pressible
and inexpansible, and so after entrapment, the
largesolid circle represents
the Cl contentof NaCl-saturated,
volume is nearly constant(Roedder, 1984).This
inclusion
pantellerite
(1.11+ 0.03wto/o;
anhydrous,
seeAppendixl).
means that the melt inclusion approximatesan isochoric
(constant volume) system ldtth dP/dT equal to the ratio
of the coefficient of thermal expansionof the melt (a) to
and Hildreth (1986)and Lowensternand Mahood (1991). its compressibility (d). Becausesilicate melts are relativeLowensternand Mahood (1991)identifiedtwo groupsof ly incompressible, their isochores have relatively steep
silicate melt inclusionsin P32. P104. and other units. slopes (6-7 bars/'C for hydrous pantellerite: estimated
Glassy inclusions had degassedprior to or during erup- from data of Lange and Carmichael, 1990), and the intion, usually along narrow capillaries that connect the ternal pressureof inclusions changesrapidly with relainclusions to the outside of the host phenocryst,but also tively small changesin temperature (Fig. 3). As long as
along cracks(seealso Anderson, l99l). This population the quartz host is strong enoughto withstand the pressure
of melt inclusions had < I Mo/o HrO, ranging down to differential betweeninclusion and intratelluric melt (Tait,
<0.2 wo/o. Other silicate-melt inclusions were micro- 1992), the pressurein the inclusion is a function of its
crystalline massesof quartz, sanidine, alkali amphibole, temperature.When an inclusion-bearingcrystal is heated
and other unidentified phases.When heated in the labo- to Z, (its temperature of original entrapment: Table 2),
ratory at high temperature (825 "C), they melted to hy- the inclusion should also return to its pressureat the time
drous glass(1.8 wto/oHrO for P32 and 2.1 wto/ofor Pl04) of entrapment (provided that compressibility of the host
with a major-element composition equivalent to that of is negligible), even if the outside of the phenocryst host
the degassedmatrix. Evidently, their higher HrO con- is at atmospheric pressure.Therefore, high-temperature
tents, relative to glassy,leakedinclusions,promoted crys- heating experiments can recreatethe conditions of crystallization during slow cooling after eruption. All the tallization in the magma reservoir prior to eruption.
samplesof Lowensternand Mahood (1991)and LowenAs an inclusion cools below I,, it will depressurize
stern et al. (1991) were remeltedin a furnaceat I atm enoughfor a shrinkagebubble to nucleateat Io [compare
and were observedbefore and after the experiments (see with Tait (1992), who considers isothermal decompresalso Skirius et al., 1990).In the presentresearch,all melt sion of melt inclusionsl.Once the compressiblebubble
inclusions were originally microcrystalline when selected has formed, the inclusion cools along a curve more shalfrom phenocryst separates.They were heated in a high- low than the isochore (the melt-vapor curve of Fig. 3)
temperature stageattached to a microscope so that they until the supercooledsilicate melt passesthrough the glass
could be observedand photographed.
transition at f". Additionally, as the inclusion cools to
3s6
LOWENSTERN: CHLORINE IN PERALKALINE MAGMA
TABLE2. Notation for describing characteristicsof melt inclusrons
Description
T,
T
U
Ts
U>
u>
U
(r
o-
TuT^
T
temperatureof entrapmentof silicate melt inclusionin host phenocrysr
temperatureat which bubble nucleatesduring cooling of silicate
melt inclusion
temperatureduring cooling at which silicatemelt undergoes
transition to glassy state
temperatureat which quartz undergoesphase transformation
(-573rcatlatm)
temperature,during heating, at which a microcrystallinesilicate
melt inclusionis converted to silicate melt t vapor
temperature,during heating,ot homogenizationof silicate
melt + vapor to a single phase
Ihr
Tu,z
rs
rF-o
Tbt
TV
T\
Thz
utt'r=ura"orrra
trapped only silicate melt. In such cases,Tn ) T, (Inclusion 2 in Fig. 3).
Fig. 3. Schematicpressure-temperaturetrajectoriesfor three
AN.lLYrrcAl, TECHNTeuES
melt inclusions. Inclusion l, trapping vapor-undersaturatedsilHeating
stage
experiments
icate melt at fr,, cools along Isochore | (A to B) until reaching
vapor saturation at 7b, and a bubble nucleates.The inclusion
Quartz phenocrystsbearing melt inclusions were doucools along the melt-vapor curve (from B to C), and the size of bly polished to provide optimal viewing conditions durthe bubble increasesuntil 573 "C or 7'u,"Q. At C, the quartz
ing high-temperature experiments. Quartz grains were
host undergoes lolo volumetric contraction, and the inclusion
preferred over feldsparbecausethey had significantly less
increasesin pressure(C to D). The inclusion then cools along a
new melt-vapor curve (from D to E) until I" (E), when the tendency to break during sample preparation and inclusilicate melt passesthrough the glasstransition, and the bubble sion homogenization.Also, becausequartz is the last maceasesto grow. Upon hearing, the inclusion retracesthe same jor phase to crystallize in pantellerite magmas,its inclupath (E to D to C lo B\ until bubble and melt homogenizeat B
sions are representative of melt compositions shortly
(?n",),and the inclusion joins Isochore l. An alternative meta- before eruption. Once doubly polished, crystalswere typstable cooling path for Inclusion I occurs if a bubble fails to ically soaked in acetone to remove mounting resin and
nucleate (so that To < Tn). The inclusion then cools along its impurities. All experiments were done at I atm in an
metastableisochore and becomesunderpressured,or vapor-su- HrO-cooled l-nltz 1350 microscope heating stage atpersaturated(B to F), until it reaches?noi,when the bubble nutached on an Ortholux II Pol-MK microscopewith phocleatesand equilibrates with the silicate melt (-Fto G).
tographic capabilities at the Geological Survey ofJapan.
Inclusion 2, which trapped two phases,silicate melt + a primary vapor bubble, atT,,(B), has ?"n> Z, becausethe inclusion Temperature was measuredwith a Pts?Rhr3thermocoumust be overpressuredto dissolve the extra vapor. In the labo- ple and recorded on a chart plotter. The system was cal'C),
ratory, such an inclusion must be heated(from B to H) until the ibrated at the melting temperaturesof KrCrrO? (398
vapor is dissolved at Zn,. Upon further heating, the inclusion Ag (961 "C), Au (1063 "C), and NaCl (800 "C). Temperfollows Isochore 2 (shown as the dotted curve).
ature gradientswithin the samplewere negligiblebecause
room temperature, the internal pressure in the bubble
may change as gases condense to their liquid state. At 25
'C, bubbles composed
of pure HrO should have internal
pressuresequivalent to the vapor pressure ofHro (0.026
atm). Bubbles with relatively noncondensable gases (e.g.,
COr) retain higher internal pressures at room temperature (up to -60 atm if liquid CO, is absent; Angus et al.,
r976).
If a melt inclusion is heatedalong the melt-vapor curve,
its bubble homogenizesinto the melt at In. Ideally, Z"
should be = 7,. However, some inclusions may contain a
vapor bubble that was trapped along with silicate melt
(i.e., two phaseswere trapped). Such inclusions must be
brought to higher pressure,by heating above 71, to dissolve the extra increment of vapor and should have higher homogenization temperatures than inclusions that
of the small size of individual quartz grains (<l mm)
relative to the sample chamber (l cm). All sampleswere
atalyzed one at a time, and the temperature calibration
for each experiment was corroborated by noting the temperature at which vapor bubbles in the inclusions underwent contraction, correspondingto the quartz B to a transition (573 "C). Recorded temperaturesare estimated to
be accurate to within 15 "C. Typically, inclusions were
heated at a rate of 50 'Clmin up to a temperatureof 600
"C. They were then allowed to equilibrate for 10-20 min
before the temperature was raised to 700 'C and they
equilibrated for an additional l0 min. The temperature
was then raised in 50" increments at 800'C (againallowing at least l0 min for equilibration at eachtemperature)
and in 25oincrements thereafter.Maximum cooling rate
for the heating stage was -400 "/min. All experiments
were completedin <3 h.
As discussedbelow, heating rates may have been too
35'r
LOWENSTERN: CHLORINE IN PERALKALINE MAGMA
Trsle 3. Resultsof heating-stageexperimentson Pantellerianmelt inclusions
I n
Sample'
fC)
P32-26
800
P32-28
850
P32-29
840
P32-32.1 <850
P32-32.2 <850
<850
P32-32.3
P32-32.4 <850
P32-32.5 <850
P32-33
930
P32-36
950
P32-38.1
980
P32-38.2
850
P32-38.3
850
P32-38.4
850
P32-41.1
860
P32-41.2
810
P32-41.3
810
P32-41.4
810
P32-46.1
830
P32-46.2
830
P32-49.1
800
P32-49.2
875
P32-49.3
800
P32-49.4
800
P32-49.5
800
P32-51.1
800
P32-51.2
800
P32-51.3
800
P32-s1.4
800
P32-51.5
800
P32-52.1
825
P32-52.2
850
P32-52.3
850
P32-53.1 >1000
<900
P32-53.2
P32-53.3 <900
P32-54.1
900
P32-54.2
8s0
P32-54.3
850
P32-54.4
850
4
fC)
940
870
880
860
860
860
860
860
>930
>950
>980
870
870
850
1200
810
810
11 0 0
880
880
875
>930
8s0
850
875
950
930
>950
850
850
850
880
880
> 1000
900
900
>1080
850
850
850
Type..
I or lll
il (1)
l (2)
I
lll or V
I
il (15) + ill
ill
il (2)
il (1)
il (2)
ill
il (1)
I
il (3)
lll or V
lll or V
lll or V
I
il (3)
il (1)
I
il (1)
lll or V
il0)
I
I
Opaque
crystals?
No
Yes
Yes
Yes
Yes
No
No
No
NO
No
No
Yes
Yes
No
No
NO
NO
No
Yes
Yes
Yes
No
Yes
Yes
Yes
No
No
No
No
NO
No
Yes
Yes
No
No
No
No
Yes
Yes
Yes
Inc size
(pm)t
b5x/5
oSxoJ
60x55
12O x 120
N.R.+
N.R.
N.R.
N.R.
N.R.
N.R.
70x55
40x30
39x20
10x10
90x60
15x15
20x20
15x15
40x30
25x20
105 x 60
135 x 90
30x30
35x35
20x20
75x70
155 x 50
75 x 120
2Ox2O
35x20
105 x 85
20x25
25x10
105 x 200
30x30
30x40
150 x 150
30x30
30x30
30x30
Sample
P32-54.5
P32-54.6
P32-56.1
P32-562
P32-58
P32-59
P32-60
P32-64.1
P32-6s.1
P32-65.2
P32-65.3
P32-65.4
P32-67.1
P32-67.2
P32-67.8
P32-67.9
P32-67.10
P32-73.1
P32-73.2
P32-74:t
P32-74.2
P32-74.3
P32-74.4
P32-74.5
P32-74.6
P32-77.1
P32-77.2
P32-77.3
P32-77.4
P32-77.5
P32-80.1
P32-80.2
P32-80.3
P32-80.4
P32-81
P104-25
P104-26.1
P104-26.2
P104-26.3
In
lh
fC)
fc)
850
850
>980
885
800
850
780
>850
800
850
880
850
800
850
880
850
880
800
850
>900
900
850
900
850
850
775
775
>850
>850
850
<850$
<850$
<850$
<850$
<850$
800
810
810
810
850
850
>980
885
890
890
820
N.R.
880
880
880
880
850
850
880
850
880
870
870
>900
850
850
850
850
850
850
850
>850
>850
850
<850
<850
<850
<850
>950
830
825
825
825
rype
il (3)
I
il (3)
il (1)
I
il (1)
il (2)
I
I
I
I
I
il (2)
il (2)
il (2)
il (1)
il (2)
il (2)
ll + lll
Opaque
crystals?
Inc size
0rm)
Yes
Yes
30 x 30
30 x 30
245 x 100
65 x 20
240 x 50
80 x 105
69 x 20
110 x 35
130 x 100
90 x 50
40 x 35
N.R.
90x90
39x25
35x35
25x20
25x20
90x85
15x 15
24O x 140
100 x 45
30x30
50 x 25
10x 10
15x15
150 x 130
N.R.
N.R.
N.R.
N.R.
60x50
40x40
65x45
80x80
270 x 60
150 x 90
59x40
55x50
200 x 55
Yes
No
No
Yes
Yes
Yes
Yes
Yes
Yes
No
No
No
No
No
No
No
Yes
No
Yes
No
No
No
No
No
No
Yes
No
No
No
No
No
No
No
No
No
. Sample number before decimal identifieshost
crystal. Numbersafter the decimal representmultiple inclusionswithin the same phenocryst.
*" Numbersin parenthesesindicate number of hydrosalinemelt
droplets in type ll inclusions.
I Inclusionsize representsestimate ot the longest two dimensions(if inclusionis a parallelepiped).For all inclusions,the third distance (thickness)
is intermediatebetween the two quoted dimensions.
+ N.R. : not recorded.
$ Inclusionsfirst homogenizedin Yamato muffle furnace.
rapid to ensurecomplete diffusional equilibrium between
silicate melt and coexistingvapor bubbleswithin the time
frame of the heating-stageexperiments. Therefore, homogenization temperatureslisted in Tables 3 and 4 may
be 25-7 5 'C higher than the actual temperaturesof inclusion entrapment. Even so, all experiments were conducted in a similar manner, and the recordeddata reproducibly identify different populations of inclusions (e.g.,
one can differentiate Inclusions I and 2 of Fig. 3).
Muffi e-furnaceexperiments
Homogenization temperatures were also determined
during longer experiments in a Yamato mufle furnace.
Unpolished and doubly polished inclusion-bearingquartz
grains were placed in a Pt crucible and heated at atmospheric pressureto the experiment temperature at a rate
of -1.25 "C/min. The sampleswere quenchedby remov-
ing the crucible from the furnace and cooling it in air.
The crucible was cool enough to touch within 2 min,
indicating quench rates of -400 'Clmin, similar to those
in the heating stage.
Scanning electron microscope(SEM)
Chips of glassygroundmass(- I mm in diameter) were
separated from P32 and soaked in distilled, deionized
HrO to remove any impurities or salts on the glasssurface. They were then rinsed in acetoneand dried. After
the chips were crushedbetweentwo cleanglassslides,the
pantellerite glassshardswere sprinkled onto a brass disk
coated with wet C paint. The mount was then sprayed
with compressedair to remove loose fragments and Au
coated.The sampleswere viewed both with a JEOL 6400
scanning microscope at the Geological Survey of Japan
and a Cambridge Instruments 250 Mark II at the U.S.
358
LOWENSTERN: CHLORINE IN PERALKALINE MAGMA
TABLE
4. Summaryof melt inclusiontypes and their characteristics
rype
L fC)
4 fC).
No. of
examples.*
tl
775-850
77s-850
825-900
825-900
43-44
24
ill
IV
775-850
glassy at 25
>950
not studied
3-10
>1000
>1000
4-8
Phases present at
900 "Ct
Phasesat 25'C after
melting experiment
silicate melt
silicatemelt + hydrosalinemelt
silicatemelt + vapor
not studied
silicateglass + shrinkage bubble
silicateglass + (halite+ smallbubble) + shrinkage bubble
silicateglass + largevapor bubble
degassedsilicateglass + bubble
microcrystallinesilicate
mass + vapor
microcrystallinesilicatemass + vapor
present at
Phases
']i
(inferred)
silicatemelt
silicatemelt + hydrosaline melt
silicatemelt + vapor
silicatemelt t vapor:
degassedduring eruption
silicatemelt a vapor
+ hydrosalinemelt:
degassedduring eruption
'These homogenizationtemperatureswere measuredwith the heating stage. for group I and ll inclusionsthus may be 25-75" higher than
I.
4
Becausetype lll inclusionsrepresent heterogeneousentrapment of magmatic vapor + silicatemelt, 4 for them has no real geologicalsignificance.
" Some inclusionsare classifiedas more than one type (e.9., ll + lll) and others are uncertain(e.9., lll or V)
t All type l, ll, lll, and V inclusionscontainedquartz blebs before, during, and after high-temperatureexperiments.The small blebs (<20 pm) were
apparently refractory daughter crystals that failed to melt during the experiments(see text).
Some inclusions contained small opaque grains that
did not completely dissolve at Z*. Table 3 notes those
inclusions that contained opaque grains, but the Z. list
BnH.lvron oF PANTELLERTAN
TNCLUSToNS
ings do not account for their presence,as they usually
DURING HEATING EXPERIMENTS
made up <<0.I volo/oof the inclusion.In all cases,opaque
Table 3 lists the results of all heating-stageexperiments grains could be dissolved by increasingthe temperature
performed on melt inclusions in quartz. The data are to 950 "C for severalminutes.
summarized in Table 4 and below.
Cooling of type I inclusions.During cooling below ?"n,
vapor bubblesusually did not nucleateuntil the inclusion
Type I: Silicate melt inclusions
reached600-700 "C (Fig. 4E);i.e.,7rowasas much as 275
Melting of type I inclusions. Upon heating, type I in- "C below Zn. In fact, when cooled at ratesof 300-400 "C/
clusions remained microcrystalline until reachinga tem- min, some inclusions failed to nucleate a bubble. Such
perature between 600 and 700 'C (Fig. a). At that point, phenomenawere more common in small inclusions than
the inclusions becamenoticeably lighter in color, chang- large ones, though sphericalinclusions with diameters of
ing from a dark black-blue to green-grayat the inclusion 100 pm usually could be cooled fast enough to produce
periphery. With progressiveheating, the inclusions con- bubble-freeinclusions. Bubble formation and growth was
tinued to melt until they consisted of melt + bubble(s) also affectedby the quartz 0 to a phasetransition. During
+ quartz, at temperaturesranging from 750 to 850 "C. some experiments, bubbles would nucleate at temperaThe temperature at which the inclusion became com- tures just above 600 "C and would then shrink (or be
pletely molten (exceptfor quartz and bubbles) is termed resorbed) when the quartz host contracted at the B to a
Z-. Upon heating above this temperature, the bubble phasetransition (-573 'C).
would continue to shrink as the internal pressurein the
Bubbles commonly nucleated on the inclusion wall or
inclusion increased.All type I inclusions were homoge- on quartz blebs, though many bubbles nucleated within
nized to melt + quartz blebsat temperaturesrangingfrom the melt, away from any obvious surface. If more than
830 to 900'C (Zn). Seventy-fivepercentwere homoge- one bubble formed during the cooling of an inclusion,
nized between 850 and 880 'C. When inclusions were these bubbles always nucleated simultaneously.
homogenized for a secondor third time, Zn was usually
reproducible to within 25 "C, though the location and Type II: Silicate melt inclusions with spherical
number of bubbles produced during cooling varied from globules (hydrosaline rnelts)
experiment to experiment.
Heating experiments. Type II inclusions had similar
The quartz blebs (Fig. 4C) ranged from 5 to 25 pm in T^ and Zn as type I inclusions,but at temperatures> 700
size. Becausethey were also present in all microcrystal- 'C, they contained small colorlesssphericalglobules I -4
line inclusions (unheated) viewed with the SEM or re- pm in size (Fig. 5) of a substancewith high optical relief
flecting microscope, the quartz blebs must have grown compared with the silicate melt. During heating experiduring original cooling of the inclusions after entrapment ments on microcrystalline inclusions, the globules be(also see Skirius et al., 1990). The quartz blebs never came visible as the inclusion beganto melt, between650
made up more than 5oloof the total inclusion volume, and 750 "C. During reheatingof glassy,previously melted
and only could be dissolved during long heating experi- inclusions (more transparentthan the microcrystalline inments (48 h) or at high temperature (> 1000 "C). Quartz clusions), the featureswere visible at room temperature
blebs were not observed in homogenized inclusions in as one or more small (<3 pm) colorlesscubesand an
feldspar (J. B. Lowenstern, unpublished data).
approximately equal volume of bubble. As the inclusion
Geological Survey. Mineral phaseswere identified by using an energy-dispersivedetector.
LOWENSTERN: CHLORINE IN PERALKALINE MAGMA
was heated above 600 oC, the cube and bubbles homogenized to a single phase: high-relief, spherical globules.
Homogenizationwas completeat temperaturesbelow 700
"C. The globules did not changesize during subsequent
heatingabove 700 "C and did not dissolveinto the silicate
melt even after 30 min at 1000'C in the heatingstageor
48 h at 900 "C in the muffie furnace. There was no correlation betweenthe size of an inclusion and the number
of colorlessglobuleswithin it. If there was more than one
inclusion in a phenocryst, a globule might be located in
one inclusion, but none would be present in the others.
During heating, most globules were not located near vapor bubbles; however, some bubbles apparently contained thesesmall globuleswithin them. At fr, thesebubbles would be resorbedinto the silicate melt, leaving only
the globule remaining.
As in some type I inclusions, most type II inclusions
contained micrometer-sizedopaqueminerals that melted
at temperaturesabove 900'C (Table 3 and Fig. 58) or at
lower temperatures (e.g., 800 'C) during longer experiments.
Cooling of type II inclusions.As with type I inclusions,
when type II inclusions were cooled below In, vapor bubbles would nucleatebetween 700 and 600 "C, depending
on cooling rate. Some vapor bubbles nucleatedon spherical globules,though most bubbles formed independently
of these features(Fig. 5A).
All globules in type II inclusions crystallized to one or
more cubesand an approximately equal volume of bubble at 490 + l5 "C. This temperatureis essentiallyidentical to the 500 + l0'C reported by Clocchiatti et al.
(1990)for hydrosalinemelts within melt inclusionsfrom
pantelleritesof Montagna Grande on Pantelleria.Bubbles
formed by this processnever grew >l pm in size (i.e.,
they were much smaller than bubbles formed by shrinkage of the silicate melt), presumably becausethe silicate
melt went through the glass transition close to 490 'C
(Bacon, 1977), preventingthese bubbles from growing.
Crystallization of colorlessglobuleswas rapid (< I s). The
temperature of crystallization was not affected by the
cooling rate ofthe inclusion, and colorlessglobulescould
not be metastablyquenchedwithout crystallization, even
at cooling rates of 400 'Clmin. At room temperature,
remnants of the colorless globules consisted of a small
bubble and a subequalvolume of crystal. The two phases
either constituted a sphere(Fig. 5E), or the sphericalbubble touched the cubic crystal at one of its corners (Fig.
5D). During the months following the heating-stageexperiments, these featureschangedshape,indicating they
were able to equilibrate or recrystallizeat room temperature.As discussedin a later section,their composition,
thermometric behavior, and crystallization is consistent
with that of hydrosaline melts.
Type III: Yapor-rich silicate melt inclusions
This group of inclusions melted to silicate melt + bubbles at similar temperaturesas inclusions of types I and
II but contained more or larger bubbles than the other
types (Fig. 6). Some of theseinclusions did not reach Zn,
359
even at ll00'C. Instead,they consistedof silicatemelt
+ bubbles. Only one bubble was large enoughIo analyze
by FTIR (36 pm), and the analysis showed that the inclusion containedconsiderableCOr. Aines et al. (1990)
also found CO, in several large, Cu-rich bubbles within
pantellerite inclusions and interpreted them to be COrbearing vapors present along with silicate melt in the inclusions (i.e., two phaseswere trapped). After quenching
from 900 "C to room temperature, bubbles in these inclusionswere sufrciently largethat they made up >3 vol0/o
of their host inclusions. Shrinkage bubbles in type I inclusions,evenwhen allowedto equilibrateat 600-700'C
for 20 min, never made lp >2 volo/oof the inclusion. I
interpret type III inclusions as containing one or more
magmatic vapor bubbles, as well as silicate melt. Both
phaseswere trapped together in the inclusion at the time
of quartz crystallization(Lowensternet al., l99l). Such
a conclusionis consistentwith their high homogenization
temperatures,the presenceof COr, and their similar I*
to type I and II inclusions.
Mixed II + III inclusions
Two inclusions contained both hydrosaline melts and
large vapor bubbles that homogenizedwith silicate melt
above 950 'C. Both of theseinclusionscontainedmore
than l5 globules.The globulesin P32-81 crystallizedto
colorlesscubes+ small (- I pm) bubblesat 600 + l0 "C,
about 100 "C higher than globulesin type II inclusions.
P32-41.1 decrepitatedat I 100 'C and thereforecould not
be observedduring cooling.
Type IY: Glassy melt inclusions
The group of glassyinclusions was studied by Lowenstern and Mahood (1991: Fig. la of that paper)and was
shown to have degassedthrough cracks and narrow capillaries. No additional heating experiments were performed on this population of inclusions.
Type V: Leaked microcrystalline inclusions
Some inclusions could not be melted at temperatures
below 1000'C. Someof thesewere locatednear obvious
cracks, though no crack was visible near others. These
inclusions are interpreted as having partially degassed
during or after eruption. Presumably, enough HrO was
left within the inclusion (or cooling was slow enough) to
promote devitrification of inclusion glass,and so this class
of inclusions may be differentiated from type IV inclusions.Some partially devitrified inclusions,with capillaries visible, appear to be an intermediate class of inclusionsbetweentypesIV and V.
Murrr,n-TURNACE ExPERTMENTS
Several experiments were done in a mufle furnace to
assessthe effect of time on melting and the homogenization of melt inclusions (Table 5). In one experiment,
microcrystalline inclusions were heated for 30 h at 750
'C. Three inclusions(out of 16) melted completelyand
homogenizedto a singlemelt phase.The cooling rate was
evidently fast enough to prevent shrinkagebubbles from
LOWENSTERN: CHLORINE IN PERALKALINE MAGMA
LOWENSTERN: CHLORINE IN PERALKALINE MAGMA
361
e-
25 pm
Fig. 4. Transmitted light photographsof type I melt inclusion (P32-52; 105 pm in maximum diameter, trapped in quartz)
during heating-stageexperiment. (A) During heating, the inclusion remained microcrystalline until temperaturesabove 700 "C,
when (B) melting began around the inclusion periphery. (C) At
800 qC, the inclusion had reached I. and consistedof pantellerite melt (m), refractory quartz (q), and vapor bubbles (v). (D)
At 850'C, the inclusion reachedZn, when the bubble was homogenizedinto the silicate melt. (E) During cooling of the inclusion, a vapor bubble nucleatedat 70, which, in this example,
was - 140 qC below Z'.
l)::
nucleating.Becausetheseinclusionshad reachedZn, the
experiment apparently indicates that at least some of the
inclusions were trapped at temperaturesas low as 750 "C,
100 "C lower than the temperatures recorded in most
heating-stageexperiments. Other inclusions, though, remained partially crystalline. Interestingly, a greater proportion of large inclusions (>50 pm) than small inclusionshad reachedZ-.
Another experiment showedthat the sphericalglobules
in type II inclusions were not resorbed into the silicate
melt even after as much as 48 h at 900'C. Two experiments on type III inclusions showed that the number of
bubbles in these inclusions decreasedsignificantly after
five or more hoursat 850'C. The remainingbubblesgrew
larger, though the total volume of bubbles stayed approximately the same. Becausethe bubbles did not appear to move during these experiments, the growth of
largebubblesis likely due to Ostwald ripening rather than
the actualcoalescence
ofbubbles.
An experiment on inclusion P32-8I showed that the
quartz blebs, presentwithin all inclusions, shrank in size
(by about 500/o)after 48 h at 900 'C. Additionally, the
walls of this inclusionhad becomemore facetedand less
rounded. Skirius et al. (1990) discussedfacetingin melt
inclusions from the Bishop Tuffand concludedthat, given sufficient time at high temperature, the walls of melt
inclusions will recrystallizeto form inclusions with neg-
(Fig. 5. Transmined lighr photographs of rype II melr (m)
inclusions trapped in quartz. (A) Inclusion P32-29 (60 pn in
maximum diameter) reached To after cooling from Zn. Three
bubbles simultaneously formed at 700 "C; none of them nucleated on the white globule (hydrosaline melt dropler; h), though
two bubbles nucleatedon a refractory quartz bleb (q). (B) Two
hydrosaline melt droplets (4 pm diameter each) were present
within P32-49.1(105 pm in maximum diameter).During heating, at 875 'C, some opaquecrystals (o) remained unmelted but
were dissolved above 900 'C. (C) During cooling, below 490 +
15 "C, the hydrosalinemelts crystallizedand could not be clearly
viewed except at 1250x magnification (D and E, for left and
right hydrosaline melts, respectively),which showed them to
consist ofhost gJass(m), a vapor (+ liquid?) bubble, and a white
crystal with cubic habit (presumablyhalite). The host crystal was
flipped and rotated before photographingD and E.
Fig. 6. Transmittedlight photographof typeIII melt inclu(135pm in maximumdiameter)at room temsron,P32-49.2
perature.The inclusionconsistsof pantelleriteglass(g),refractory quartz(q),andhvevaporbubbles,two ofwhicharein focus.
Theinclusionhad In >930'C, andthebubbles
did not homogenizeafter6 h in themufle furnaceat 850'C. Thelargestbubble
(-39 pm in diameter)containedCOr, as detectedby infrared
spectroscopy.
ative crystal shapes (Clocchiatti, 1975). I interpret the
quartz blebs in pantellerite melt inclusions in quartz to
be daughter products that form during crystallization of
the silicate melt to the blue microcrystalline mass.During
high-temperatureexperiments,melting initiates at the inclusion-hostborder. Partial dissolutionof the host might
causethe inclusion to become saturatedwith respectto
SiO, before all the inclusion contentsare melted; as such,
no further quartz can be dissolved, and quartz daughter
crystals (blebs) remain. However, given sufficient time,
the quartz blebs may dissolve and be reprecipitated on
the inclusion wall becauseofthe favorable energeticsof
inclusionswith negativecrystalshapes.
AssnssrvrnNT oF EeurLrBRruM IN THE
LABORATORY
AND NATURE
The reliability of f, measurements
Data from this study can be used to constrain the temperature of entrapment of silicate melt inclusions to between 750 and 875 "C. Much of this spreadappearsto be
due to real diferences in the temperature of entrapment
of inclusions.However, severalexperimentsdone in the
mufle furnace indicated lower Zn for the melt inclusions
than experiments using the heating stage.The primary
difference between these types of experiments was the
time allowed for equilibration. This means that temperatures measured during heating-stageexperiments may
not reflect the actual Zn becausethey did not allow sufficient time for diffusion of HrO betweenthe vapor bubble and silicate melt. The Zn values in Table 3 appear to
be between 25 and 75 'C too high, as compared with
results shown in Table 5. Similar heating-stageexperiments (J. B. Lowenstern,unpublishedresults)on bubblebearingmelt inclusionsfrom the Valley of Ten Thousand
Smokes showed In between 25 and 75 'C higher than
LOWENSTERN: CHLORINE IN PERALKALINE MAGMA
502
TABLE
5. Resultsof experimentsin mufilefurnace
Sample
Type
Tt
fC).
(hf'
Descriotionof inclusion
before exoerimentt
Descriotionof inclusion
after exoeriment
Interpretedresult+
il, il, t|
825
850
4
6
P32-492
ill
8s0
6
-70 pm inc: g + largev
90 x 60,30, and 20 pm incs:
a l lw i t h g + h m r
135 x 90 rrm inc: g + -30 v
P32-51.1
P32-54
lll or V
7
825
850
4
6
75 x 70 pmt g + 1 large v
150 x 150 pm inc: g * -30 v
g+llargev
host crystal cracked and inc
vesiculated
P32-59.1
I
750
30
105 x 80 pm inc: g
s
750
750
30
30
35rrminc:g+x+v
60 pm inc, 20 pm inc, 30 pm
inc, 20 pm inc, 15 pm inc.
A l lg + x
2 partially melted incs
(110 x 35 and 20 x 50 rrm)
g+v+coarserx
all incs had g + coarser x
1 large(80 x 110 pm) + 1
small (30 rrm) micro inc
1 large(140 pm) + 1 small(35
pm) micro inc
100 pm inc + 50 pm inc + 40
rm inc + 25 pm inc. All micro rncs
120 x 60 and 50 pm micro
Incs.
270 x 60 rrm inc: g + >20 v
bubbles+hmr+quartz
blebs$
largeinc: g; small
Inc:g+v+x
largeinc: g; small
inc:v+g+x
1 0 0p m i n c : g + v ; o t h ers:g+v+x
hydrosalinemelt stable at
850:Land4>850
(tor both incs)
largeinc: Land 4 < 750;
small inc: I. > 750
large inc: 7- and 7i < 750;
small inc: I- > 750
100 pm inc I. < 750; others: I. > 750
g+v+x
L>750
g+3v+hmr+smaller(by
-50%) quartz blebs
hydrosalinemelts stable at
900 and did not coalesce; large v bubbles
grew: small bubbles
wereresorbed. T" > 900;
quartz blebs dissolve if
given sufficienttime
P32-26
P32-49.1,
3,5
I
P32-59.2
P32-63
P32-64
t,tl
850
P32-68
t,l
750
30
P32-69
t,l
/cu
30
P32-70
t,t,t,l
750
30
P32-71
t,l
750
30
P32-81
ll + lll
900
48
s
g+nmr
g+2v
2 incs with g + x + v: larger
inc has hmr
T <825
hydrosalinemelt stable at
850
bubblescoalescedat 850:
T<T
T- > 825 or inc leaked
Inc leaked
T^ < 750 (?) or silicate
melt metastable;v bubble did not nucleateat
750
L>750
L > 750; v bubblesdo
not nucleateat 750
' T: experimenttemperaturein degrees Celsius.
'* f: length of experiment.Time was apparently sufficientto ensure accurate I. and values.
4
t Dimensionsindicate longest and shortest sides of cylindrical and parallelepipedinclusionsor average diameter of spherical inclusions.Values
rounded off to nearest multiple of five. All observations were made at room temperature. Abbreviations used: micro: microcrystalline;g : glass;
hmr: the products of crysrallizationof the hydrosaline melts (i.e., micrometer-sizedcube + subequal bubble); inc: inclusion; v : vapor bubble;
x : silicateor oxide crystals (not quartz blebs).
+ Unless otherwise stated, ?i was not reachedduring the experiment(i.e., f < 4). Tin degrees Celsius.
$ All inclusions >30 pm in diameter containedsmall quartz blebs. Those in P32-81 were observed in greater detail.
preeruptive temperaturesindicated by iron titanium oxide geothermometry (Hildreth, I 983).
Using solutions provided by Qin et al. (1992) for diffusional exchangebetween a sphereofradius a (bubble)
located within a sphere of radius b (inclusion), one can
calculate the time necessaryfor the attainment of equilibrium. If a/b : 0.01, b : 50 pm, and the diffusion
coemcient for HrO (or other diffusing species)is 10-7
cm2ls,the systemreaches>950/oequilibrium in 1.6 min
(if the melt-vapor partition coefficient for diffusing species is >0.1). Larger bubbles equilibrate faster than this
estimate. A decreaseof I log unit in the diffusion coefrcient increasesthe time necessaryfor equilibration by a
factor of ten. Above 800 'C, HrO probably diffuses fast
enough to attain equilibrium within the time frame of
heating-stageexperiments(Karsten et al., 1982). However, becausesome inclusions appeared to homogenize
at lower temperaturesduring the mufle furnace experiments than in the heating stage,> I h may be necessary
for full equilibration at temperatures below 800 "C. Cl
and COr, slowerdiffusingspecies(Watson,l99l), would
reachequilibrium with the bubblesin severaltens of minutes to several hours, within the time allotted for the
muffie furnace experiments and many of the heating-stage
experiments at temperaturesabove 800 "C.
The major- and trace-elementcompositions of Pantellerian melt inclusions should becomehomogeneouswithin the time scaleof most heating-stageexperiments.Because the phaseswithin microcrystalline inclusions are
very small (< I pm except for quartz blebs) and appear
to have homogeneousdistribution, diffusion paths are
short, and remelted inclusions should become homogeneous within several tens of minutes.
The control of cooling rate on shrinkage bubble
volumes
Besidesits strong control on Zo, the cooling rate also
affectsbubble size.Comparison of the sizesof bubbles in
silicate melt inclusions from volcanic rocks may therefore
be misleading, unless inclusions with a similar host and
LOWENSTERN: CHLORINE IN PERALKALINE MAGMA
t63
similar size, cooling history, and composition are compared. Comparison of bubble volumes in quartz and plagioclasemay be of little value becauseof the strong effect
of the quartz 0 to a transition on the size of bubbles
measured at room temperature. A more reproducible
method for comparing sizesof shrinkageor primary bubbles in melt inclusions would be to measurethem at nearly magmatictemperatures.Even then, careshouldbe taken
to allow sufficient time to eliminate any compositional
gradientsin the inclusion and to allow the bubble to reach
its equilibrium volume.
When cooling rates are very rapid, homogenizedinclusions may not nucleatea bubble. Many authors have noted that melt inclusions from crystals in Plinian eruptive
products tend not to contain bubbles (e.g., Anderson,
1991; Dunbar and Hervig, 1992; Lowenstern, 1993),
whereas those from ignimbrites almost always contain
them. Clocchiatti(1972) concludedthat crystalsfrom the
l9l2 ignimbrite of the Valley of Ten ThousandSmokes
must have had a relatively slow cooling history because
they all contained bubbles. Data from the present study
Fig. 7. (A andB) Transmittedlight photographs
of fluid inindicate that melt inclusions in hydrous peralkaline rhypantelleritematrix glass(g) from sample
clusionsin outgassed
olites should not contain shrinkagebubblesifcooled from
P32.Thesefeatures(both 9 pm across)consistofa parallelepi7"nat rates faster than -300"/min.
ped-shaped
bubble(v) insidea sphericalcrystallineshell(s,presumablyhalite).Severalunidentifiedopaquecrystalsline the
IonNtmIc.luoN
oF corroRlEss GLoBULESAS
inclusionwalls.The inclusionsareinterpretedto be the crystallizedremainsof hydrosaline
melts.(C) Syntheticfluid inclusion
HYDROSALINE
MELTS
(3'l
in
diameter),
similar
to naturalinclusions(A and B),
rrm
The behavior ofthe colorlesssphericalglobulesduring
producedby saturatingpantelleritemelt with a solutionof 800/o
cooling, including their crystallization to a small cube *
NaCl and 2lo/oH2Oat 200MPa and 900 'C (seeAppendixl).
bubble around 500'C, indicates that these features are
neither silicate nor oxide phases.Instead, their behavior
is consistentwith that of hydrosaline melts. They did not
homogenize with the silicate melt, even during a 48-h- temperature of the globules may be more useful toward
long experiment at 900 "C (which is above the liquidus; determining the composition of this phase.During heatseeAppendix 1) and other experimentsat lower temper- ing experiments,the cube * bubble homogenizedto the
atures, indicating that they represent a separatephase. hydrosaline liquid at temperaturesbetween 600 and 700
Their presencein many, though not all, inclusions makes "C, liquidus temperaturesfor solutions with 75-85 wto/o
it likely that they were trapped by the quartz along with
NaCl. Therefore,if this phasewas an NaCl-HrO solution,
the silicate melt (i.e., two phaseswere trapped, which was it contained between 60 and 85 wto/oNaCl. Though the
termed mixed type I-II inclusions by Roedder and hydrosaline melt could have contained KCl, FeClr, or
Coombs, 1967). The obvious difference in behavior of
other salts, the high Na and Cl contents of pantellerite
vapor (shrinkage)bubbles and hydrosaline melts, the fact make it probable that the phase was mostly NaCl and
that thesebubbles did not always nucleateon the hydroHrO. Furthermore, featureswithin outgassedpantellerite
salinemelts,and the observationthat the two phasescould matrix, discussedbelow, appear to corroborate the hytouch each other without coalescingindicate that they pothesis that the cubeswithin melt inclusions were prewere not miscible. Furthermore, when coexistingprimary
dominantly halite.
(trapped) bubbles and hydrosaline melts were observed,
Fr,urn INCLUSToNSAND HALTTECUBESrN
as in P32-81 (a mixed II-III inclusion),the two phases
OUTGASSED MATRIX GLASS
touched each other at temperatures>800 "C and yet did
not mix.
Small (l-10 pm), sphericalfluid inclusionswere idenThe salinity of the hydrosaline melt may be estimated tified in matrix glassof samplesP32 (Fig. 7) and P104.
by the temperatureat which this phasecrystallizesduring In transmitted light, the fluid inclusions of P32 usually
cooling. If the fluid were an NaCl-HrO mixture, the crys- appear as spherical droplets with two main phasesartallization temperatureof 490'C would correspondto the ranged concentrically. The outside ofthe sphereconsists
liquidus for a solution with -60 wt0/oNaCl (Gunter et of a transparent crystalline material that often displays
al., 1983).However,490'C could representa metastable cubic cleavageor habit. Inside the crystalline phase recrystallization temperatureif a phasemore salinethan 60 sides a bubble with a spherical to rectangular shape.
wto/owere supercooled. Therefore, the homogenization Sometimes,though, the bubble touchesthe host glass.In
JO4
LOWENSTERN: CHLORINE IN PERALKALINE MAGMA
to halite + vapor before eruption and extrusion (at pressuresof 300-400 bars for a fluid with 50-85 wto/oNaCl:
Chou, 1987). As long as the magma temperaturewas
<800'C (the melting temperatureof NaCl) during extrusion at the surface,the halite would be solid and nonvolatile.
Not all halite cubes found in the pantellerite matrix
may be related to hydrosaline melts. Becausethe pantellerite melt contained -2o/o HrO prior to eruption, and
matrix glassnow containsonly 0.15 wto/oHrO (Lowenstern and Mahood, l99l), the magma must have vesiculated at some point. The relative scarcity of vesiclesin
glassymatrix means that either degassingoccurred without bubble formation, or, more probably, that bubbles
were interconnected as a permeable foam, so that gas
could escapeto the atmosphere.The consequentloss in
vapor pressurewould have caused most bubbles to be
resorbedor flattened during magma ascent(Eichelberger
et al., 1986),leaving a denseobsidian.The vapor exsolved
during low-pressuredegassingof the magma would contain <0.01 wto/oNaCl, becauseof low solubility of this
speciesin high-temperature low-pressurevapors (Pitzer
and Pabalan, 1986). Therefore,most halite crystals in
matrix glassprobably were not precipitated from exsolving vapor. A few halite crystals, however, were found in
some large vesicles(>50 s.m) in the pantellerite matrix.
Flattening and resorption of such vesicleswould causea
cubic, vapor-precipitated halite crystal to be embedded
entirely in glass.This may have resulted in featuressuch
as A, B, and H in Figure 8. However, most halite crystals
are associatedwith a small bubble (cavity) and are similar
in size and aspect to the cubes formed during crystallization of hydrosaline melts found in type II melt inclusions and to sphericalfluid inclusions found in the matrix
(e.g.,C, D, E, F, of Fig. 8). Such a texture would not be
expected from the collapse of large, halite-bearing vesicles.In someinstances,hydrosalinemelts may have served
as nucleation sitesfor vesiclesof low-salinity vapor. Such
two-phasevesicleswould be halite rich, and their collapse
might result in featuressuch as G of Figure 8.
The halite itself should not be thought of as phenocrystic becauseit would have been unstable at magmatic
temperaturein either the original or the degassedmagma.
Bprrl,vron oF HyDRosALTNEMELTS DURTNG
Phenocrystic halite could only exist if the concentration
DEGASSING AND ERUPTION
of NaCl in the melt exceededthe solubility of halite in
I interpret most of the halite cubes viewed in SEM pantellerite. The Cl content of NaCl-saturated, anhyimages (Fig. 8) as corresponding to fragments (or cross drous, pantelleriteis l.l1 + 0.03 wto/o(Fig. 2 and Apsections)of spherical fluid inclusions observed in trans- pendix 1),whereasthe Cl contentsof melt inclusions(0.87
mitted light images of matrix glass (Fig. 7). The hemi- + 0. l0 wto/o)and matrix glass(0.92 + 0.07 wto/o)of P32
spherical cavities in many SEM images also may corre- and P104 are considerablyless(J. B. Lowenstern,unpubspond to bubbles observed within fluid inclusions such lishedelectronmicroprobeanalyses;also seeFig. l). Any
as those shown in Figure 7A and 78. Becausethese fea- HrO in the system dilutes the molten salt, lowering the
tures are reminiscent of cubesand bubblesformed during activity of NaCl and making halite precipitation even less
crystallization of hydrosaline melts in type II silicate melt likely. Halite, therefore, should be resorbedinto the siliinclusions (Fig. 5D), I interpret them to be a related phe- cate or hydrosaline melt phasesof these magmasat high
nomenon. They are the cooled and dehydrated remains temperatures(>700 "C). I interpret halite crystalsin panof hydrosaline melts present during magma storagein a tellerite matrix to be the crystallized remains of hydroshallow reservoir. The hydrosaline melt would crystallize saline melts present within quenched,metastableglass.
samplePI04, fluid inclusionsare ellipsoidaland elongate
parallel to flow lineations in the glass,so that the bubble
commonly touchesthe sides of the inclusion (the host
glass).Presumably,the bubbles contain liquid as well as
vapor, although that could not be verified optically. In all
inclusions, the bubble has lower relief than both the crystalline host and silicate glass(r : 1.516),whereasthe
crystalline material has higher relief than silicate glass
(consistentwith halite: n: 1.544\.Small. submicrometer, opaque crystals could be seen within the inclusions
but could not be identified. The inclusions are virtually
identical in appearanceto synthetic fluid inclusions producedby saturatingpantelleritemelt with hydrosalinemelt
(80 wto/oNaCl) at high temperatureand pressure(Fig. 7C;
seeAppendix l). The inclusionscould not be homogenized at high temperature becauseheating above 500'C
causedcracking and further degassingof the host matrix
glass.Similar fluid inclusions were not found in phenocrysts from the pantellerites from this study; however,
Solovovaet al. (1991)reportedfinding highly salinefluid
inclusions (>90 wto/oNaCl) in anorthoclasefrom felsic
volcanic rocks ofPantelleria (locality not specified).
In this study, the identification of crystalline material
in fluid inclusions was aided by use of the scanningelectron microscope(SEM). Small clustersof halite, with and
without associatedbubbles, were observed in SEM imagesof crushed matrix glass from sample P32. For example, Figure 8 showsexamplesof the - 100 halite cubes
found in three SEM mounts. Halite was typically found
as one to ten small cubesembeddedin glass.Commonly,
thesecubeswould be next to a small cavity or bubble (C,
D, and F of Fig. 8). Occasionally, no bubble would be
visible,as in A, B, G, and H. Other times,groupsof cubes
would be found in a circular region, with an associated
bubble (e.g.,E). All featureslabeledin Figure 8 were verified to contain NaCl by energy-dispersiveanalysis. No
other salts (e.g.,KCI) were found in the pantellerite matrix, though not all of the very smallest cubeswere analyzed. The abundanceof halite-bearing inclusions is estimated at 0.0 I -0. I volo/oof the rock, meaning that these
featurescontain only 0.6-6.0 wto/oof the total Cl in the
magma.
LOWENSTERN: CHLORINE IN PERALKALINE MAGMA
365
Fig. 8. (A-H) Secondaryelectron images (with SEM) of halite embedded in naturally outgassed,glassymatrix of pantellerite
P32 (an obsidian flow). Many of thesecrystals are interpreted to be crystallized remnants of hydrosaline melts (Figs. 5 and 7). See
text for details.
Iprvrrscrnr,B FLUTDSrN pANTErr.ERrrESAND
of Metrich and Rutherford (1992) show that pantellerites
equilibrated with vapor and hydrosaline melt at pressures
> 100 MPa should have lower Cl contents than the units
Equilibration depth vs. Cl content of pantellerites
consideredhere (compareFigs. I and 2). This should hold
Geologicalconstraintsindicate that pantelleritemagma
as long as the CO, in the system does not strongly affect
chambers may reside at relatively shallow depths (2-6
the
solubility of Cl in silicatemelt.
km) beneath the surface(Mahood, 1984). Information
The
HrO contents of the pantellerite melt inclusions
available from pantellerite melt inclusions is consistent
are
also
consistent with fractionation at relatively low
with this assertion. Becausethe glass in pantellerite inpressures.
Given that the solubility of HrO in peralkaline
(Lowenstern
clusions contains little CO,
and Mahood,
I 99 I ), true shrinkagebubblesshould contain mostly HrO. melts is about l5ologreater than that in metaluminous
H,O measured
Immiscibility betweenthese HrO-rich shrinkagebubbles rhyolites (Webster,1992b),the 1.8-2.10/0
and hydrosaline melts at 800'C requiresthat the pressure in melt inclusions from the pantellerites of this study
in the inclusionbe lower than 160MPa, or the two NaCl- (Lowensternand Mahood, l99l) would be consistentwith
HrO fluids would mix (Chou, 1987).Moreover, the data H,O saturationat 30-40 MPa (Silver et al.. 1990).HowOTHER MAGMATIC
SYSTEMS
366
LOWENSTERN: CHLORINE IN PERALKALINE MAGMA
ever, the presenceofCOr-bearing bubblesin type III inclusions indicates that the magma was saturated with a
mixed H'O-CO, vapor. As such,the HrO contentsof the
melt inclusions would be consistentwith vapor saturation
at higher pressures(e.g.,at 80 MPa, if the vapor contained - 50 molo/oH,O). Lowenstern and Mahood ( I 99 I )
argued that P32 and Pl04 were not HrO saturated becauseHrO contents continued to increasewith differentiation. It thus appearsthat thesepantelleritesequilibrated with a COr-bearing vapor and hydrosaline melt at
pressuresbetween 50 and 100 MPa. If the solubility of
CO, in pantellerites is similar to that in metaluminous
rhyolites (Fogeland Rutherford, 1990),then the CO, contents (<100 ppm) of pantellerites(Lowensternand Mahood, 199l) are largely consistentwith saturation with a
mixed H,O-CO, vapor at 50-100 MPa. The strongeffect
of fluid composition on volatile solubilitiesmakesmore
quantitative estimates of equilibration pressure highly
speculative.
In a similar study of pantellerite melt inclusions from
FantaleVolcano, Ethiopia, Websteret al., (1993) found
that the peralkaline melt contained higher HrO contents
(4.6-4.9 wto/o)and much lower Cl concentrations(0.30.4 wt0/0)than the rocks studied here. If these magmas
were HrO saturated, their HrO contents would be consistentwith fractionation at pressuresof 125-145 MPa
(Silver et al., 1990). At such pressuresand reasonable
magmatic temperatures, equilibrium HrO-NaCl fluids
may be supercritical. The relatively low Cl contents of
the Fantale pantelleritesmight then be the result of equilibration of the melt with a relatively Cl-poor fluid and
the low fluid-melt partition coefficient for chlorides in
peralkalinesystems(Websteret al., 1993).Additionally,
the low Cl contents may be partially related to the lower
FeO,o,contentsand agpaitic indices of the Fantale glasses
relative to the Kenyan pantelleritesused by Metrich and
Rutherford (1992) and the Pantellerian units studied in
this report (Table l). Those two factors strongly influence
the solubility of Cl in silicate melts (Metrich and Rutherford, 1992;Webster,1992b).
than the degree of enrichment of such elements as La,
Nb, Zn, and Zr, which have concentrations ten to 25
times those found in metaluminous rhyolites (compare
Macdonald and Bailey, 1973; Macdonald et al., 1992).
From this perspective,the high Cl concentrationsofpantellerites appear more reasonable,and the cause of enrichment will be understood when workers agreeon the
factors that result in high trace-elementconcentrationsin
theserocks.
Do metaluminousrhyolites contain hydrosaline melts?
The solubility of Cl in metaluminous rhyolites is significantly lower than in peralkaline systems(Metrich and
Rutherford, 1992;Webster,1992b).When saturatedwith
immiscible NaCl-HrO fluids at 100 MPa, thesemagmas
contain about 2700 ppm Cl, as opposedto -8500 ppm
in pantellerites(Metrich and Rutherford, 1992).The lower Cl solubility in metaluminous systemsis due to a larger
activity coefficient for chloride in the silicate melt, plausibly caused by the lesser number of charge-balancing
cations (Na*, K*, Fe2*;Webster, 1992b).Although halitebearing fluid inclusions (e.g.,Figs. 7 and 8) have not been
reported in metaluminous rhyolitic obsidians, their Cl
and HrO concentrationsare typically consistentwith hydrosalinemelt saturation.Layne and Stix (1991) found
that melt inclusions from the Cerro Toledo rhyolites at
Valles calderacontained about 2900 ppm CI and 4.3 u'to/o
HrO and concluded that thesevalues might indicate saturation with respectto immiscible, subcritical NaCl-HrO
fluids. Many of the silicate melt inclusions saturatedwith
hydrosaline melt describedby Frezzotti (1992) also contain similar amounts of Cl. A remarkable number of rhyolitescontain 4-5 wto/oH,O and I 700-2100 ppm Cl [e.g.,
rhyolite from the l9l2 Katmai eruption (Westrich et al.,
1991;Lowenstern,1993),Crater Lake climacticeruption
(Bacon et al., 1992), and the Taupo rhyolite (Dunbar et
al., 1989)1.It is difficult to envision how processessuch
as crystal fractionation and magma mixing could fortuitously result in such similar volatile concentrationsin a
variety of silicic magmas.Possibly, the Cl and HrO concentrationsofthese rhyolites are controlled by interaction
Pantellerites and their high Cl contents
with NaCl-H,O fluids at shallow crustal depths. If so,
From the above discussion,it is clear that the pantel- peralkaline magmatic systemsmay only appear to have
lerite melt had a high Cl concentration becauseit was high Cl contents relative to metaluminous systems bebuffered by the immiscible vapor and hydrosaline melt. causethe Cl is partitioned into the melt phaserather than
However, this begsthe question, how did the Cl content in coexisting fluids. The presenceof immiscible NaClof the system reach levels where the vapor (and silicate HrO fluids in metaluminous porphyry intrusions (Roedrnelt) could be saturated with hydrosaline melt? The Sr der, 1984,chapters14 and l5) indicatesthat thesemagand O isotopic compositions of the pantellerites are in- mas eventually become saturatedwith immiscible NaClconsistent with incorporation of sea water or melting of HrO fluids.
seawater-alteredcrust (Mahood et al., 1990).Apparently, the high Cl abundancesare magmatic in nature and Degassingof Cl during eruptions: HCI vs. NaCl
can be understood by determining the causesof the high
Nicholls and Carmichael(1969) arguedthat the high
concentrations of a variety of trace elements in pantel- and relatively constant Cl contents of pantellerites indilerites. Indeed, the concentrations of Cl in pantellerites cated that Cl was not degassedduring eruption of these
are only about five to ten times those of most non-de- magmas. Because Cl was assumed to be partitioned
gassedmetaluminous rhyolites (Dunbar et al., 1989; stronglyinto exsolvingaqueousfluids, they concludedthat
Westrich er al., l99l; Bacon et al., 1992). This is less pantellerites must have low HrO contents. In actuality,
LOWENSTERN: CHLORINE IN PERALKALINE MAGMA
pantellerites have moderate to high HrO contents (Kovalenko et al., 19881Lowenstern and Mahood. l99l:
Websteret al., 1993),but the observationof Nicholls and
Carmichael ( I 969) is still valid; Cl appearsto be retained
in the silicate melt during pantellerite eruptions. The relative nonvolatility of Cl was demonstratedby Webster et
al. (1993), who showed that Cl contentsof pantellerite
melt inclusions from Fantale, Ethiopia, are very similar
to those of outgassedmatrix. Similar relationships hold
at Pantelleria, where Cl appears to be held in the melt
dunng eruption (Kovalenko et al., 1988, 1993).Unpublished data of Lowenstern show an average of 8700 +
1000ppm Cl in 12 melt inclusionsvs. 9150 + 770 ppm
Cl in matrix glass.This contrastswith metaluminous rhyolites, where matrix glass commonly has lower Cl contents than silicate melt inclusions (e.9., Dunbar et al.,
I 989; Westrich et a1.,199l; Baconet al., 1992).Because
of the low solubility of NaCl in high-temperature HrOvapor at low pressure(Pitzer and Pabalan, 1986), no
magma is likely to lose significant amounts of NaCl during eruptive degassing.However, HCl, a minor component of magmaticvaporsat pressures>50 MPa, increasingly partitions into the vapor (not hydrosaline melt) at
low pressure(Shinoharaet al., 1984; Shinohara,l99l).
Besidespressure,the major factor controlling HCI partitioning between silicate melt and vapor is melt composition. Urabe (1985), in experimentsdone at 350 MPa,
showed that the HCI concentration of magmatic fluid is
inversely proportional to the peralkalinity ofthe coexisting silicate melt. Evidently, metaluminous magma tends
to buffer the vapor toward more acidic compositions.This
may account for the greater loss of Cl during degassing
of metaluminousmagmas(as HCI) and the association
of H* metasomatism(argillic alteration)with shallowcalcalkaline intrusions.
AcxNowr-nocMENTS
Support for this research was provided by the Japanese Agency for
Industrial Scienceand Technology. The data were gatheredat the Geological Survey of Japan (G.S J.); I thank A. Sawaki for offering use of the
I-eitz 1350 homogenization stage,H. Shinohara for help with the internally heatedpressurevessel,and Y. Okuyama for aid with the JEOL 6400
SEM. The manuscript was completedat the U.S. GeologicalSurvey,with
support from the National ResearchCouncil. Photographic equipment
was made available by C.R. Bacon, and R. Oscarsonoperatedthe SEM.
G.A. Mahood of Stanford University allowed me to use some of her
unpublisheddata, shown in Figure I and Table l, and provided the samples used in this study. Initial study of these samplesbeganwhile I was
supponed by N.S.F. gant EAR-8805074 to Mahood. I am grateful for
reviews by C.R. Bacon, H. Belkin, G.A Mahood, E. Roedder, H. Shinohara, and J. Webster Finally, I am indebted to H. Shinohara for his
insi&tful comments, friendship, and generosityduring my stay at G.S.J.
RnrnnnNcns crrED
Aines, R.D, Lowenstern, J.B., and Mahood, G.A. (1990) Evidence for
CO,-rich vapor in pantellerite magma chambers (abs.). Eos, 7l (43),
t699.
Anderson, A.T., Jr. (1991) Hourglassinclusions:Theory and application
to the Bishop Rhyolitic Tuff. American Mineralogrst,76, 530-547.
Anderson, A.T., Jr., Newman, S., Williams, S.N, Druitt, T.H., Skirius,
C., and Stolper,E. (1989) H,O, CO,, Cl and gasin Plinian and ash flow
Bishop rhyolite Geology, 17, 221-225.
367
Angus,S., Armstrong, B., deReuch,K.M., Altunin, V.V., Gadetskii, O.G.,
Chapela,G.A., and Rowlinson, J.S.(1976)Interational thermodynamic
tables of the fluid state, vol 3, Carbon dioxide, 385 p. Pergamon,
Oxford, England.
Bacon, C.R. (1977) Iligh temperatureheat content and heat capacity of
silicate glasses:Experimental determination and a model for calculation. American Mineralogist, 277, lO9-135.
Bacon, C.R., Newman, S., and Stolper, E. (1992) Water, CO,, Cl, and F
in melt inclusions in phenocrystsfrom three Holoceneexplosiveeruptions, Crater lake, Oregon. American Mineralogist, 77, 102l-1030.
Benhamou,G., and Clocchiatti, R. (1976) I.a thermom6trie optique, outil
p6trologique:Essaid'utilisation et d'application I l'6tude des reliquats
magmatiquesdes ph6nocristauxde laves hyperalcalinesde Pantelleria.
Bulletin de la Soci6t6 Frangaise de Min6ralogie et de Crystallographie,
99,ill-116.
Bodnar, R.J., Burnham, C.W., and Sterner, S.M. (1985) Synthetic fluid
inclusions in natural quartz. III. Determination of phaseequilibrium
properties in the system H,O-NaCl to 1000 9C and 1500 bars. Geochimica et CosmochimicaActa.49. l86l-1873.
Chou, I-M. (1987) Phaserelationsin the systemNaCl-KCl-HrO. III. Solubilities ofhalite in vapor-saturatedliquids above 445 qCand redetermination of phaseequilibrium properties in the systemNaCl-H,O to
1000'C and I 500 bars.Geochimicaet CosmochimicaActa. 5 l. 19651975.
Civetta, L., Cornette, Y., Grisci, G., Gillot, P.Y., Orsi, G., and Requejo,
C.S, (1984) Geology, geochronologyand chemical evolution ofthe island of Pantelleria.Geological Magazine, l2l, 541-562.
Clocchiatti, R. (1972) I-es cristaux de quartz des poncesde la Yall6e des
Dix Milles Fumees(Katmai, Alaska).Comptes Rendusde I'Acad6mie
des Sciences
de Paris.274.3037-3040.
-(1975)
Les inclusions vitreusesdes cristaux de quartz: Etude optique, thermo-optique et chimique: Applications g6ologiques.M6moires de la Soci6t6G6ologiquede France, LIV, M6moire 122, 1-96.
Clocchiatti,R., Massare,D., Metrich,N., Joron,J.L.,and Weiss,J. (1990)
Glassinclusions in quartz of pumice of Montagna Grande (Pantelleria,
Island): Appropriate experimentalsystemsfor the study ofthe physicochemical properties of hyperalkaline magmas.Thirteenth Conference
on Earth SciencesAbstracts,Grenoble, 31 (in French).
De Vivo, B , Frezzotti, M.L., and Mahood, G. (1992) Fluid inclusions in
xenoliths yield evidencefor fluid evolution in peralkalinegranitic bodies at Pantelleria (Italy). Journal of Volcanology and Geothermal Research.52. 295-301.
Dunbar, N.W., and Hervig, R.L. (1992) Volatile and trace element composition of melt inclusionsfrom the Lower BandelierTuft Implications
for magma chamber processesand eruptive style Journal of Geophysi c a l R e s e a r c h9.7 . 1 5 1 5 1 - 1 5 1 7 0 .
Dunbar, N.W., Hervig, R.L., and Kyle, P.R. (1989)Determination of preeruptiye H,O, F and Cl contents of silicic magmas using melt inclusions:Examplesfrom Taupo volcanic center,New Zealand.Bulletin of
Volcanology,51, 177-184.
Eichelberger,
J.C, Carrigan,C.R.,Westrich,H.R, and Price,R.H. (1986)
Non-explosive silicic volcanism. Nature, 323, 598-602.
Fogel,R.A , and Rutherford, M.J. (1990)The solubility of carbon dioxide
in rhyolitic melts: A quantitative FTIR study. American Mineralogist,
75, t3tr-1326.
Frczzotti, M.L. (1992) Magmatic immiscibility and fluid phaseevolution
in the Mount Genis granite (southeasternSardinia, Italy). Geochimica
et Cosmochimica ActA, 56, 2l-33
Frost, B.R., and Touret, J.L.R. (19E9) Magmatic CO, and saline melts
from the Sybille Monzosyenite,I-aramie Anorthosite Complex, Wyoming. Contributions to Mineralogy and Petrology, 103, 178-186.
Gunter, W.D., Chou, I.-M., and Girsperger,S. (1983) Phaserelations in
the systemNaCl-KCl-HrO II. Differential thermal analysisof the halite liquidus in the NaCl-H,O binary above 450 lC. Geochimica et
Cosmochimica Acta, 47, 863-873.
Hansteen,T.H. (1989) Fluid and silicate melt inclusions in the EikerenSkrim peralkaline granite complex, the Oslo paleorift, Norway. European Current Researchon Fluid Inclusions Abstracts, X symposium,
I-ondon, 45
Hildreth, W. (1983) The compositionally zoned eruption of 19l2 in the
368
LOWENSTERN: CHLORINE IN PERALKALINE MAGMA
Valley ofTen Thousand Smokes,Katmai National Park, Alaska. Journal ofYolcanology and Geothermal Research,18, l-56.
Hollister, L.S., and Crawford, M.L., Eds. (1982) Fluid inclusions:Applicationsto petrology.Mineralogical AssociationofCanada Short Course
Handbook, 6, 3O4p.
Karsten, J.L., Holloway, J.R., and Delaney, J.R. (1982) Ion microprobe
studies of water in silicate melts: Temperature-dependent water dimrsion in obsidian. Eanh and PlanetaryScienceIrtters, 59,420-428.
Kovalenko, V.I., Hervig, R.L., and Sheridan,M.F. (1988)Ion microprobe
analyses of trace elements in anorthoclase, hedenbergite, aenigmatite,
quartz, apatite and glass in pantellerite: Evidence for high water contents in pantellerite melt. American Mineralogist, 73, 1038-1045.
Kovalenko, Y.I., Naumov, Y.B., and Solovova, I.P. (1993) Behavior of
Cl during differentiation and eruption ofmagmas at Pantelleria.Geochemistry International, 30(7), 105-108.
lange, R.L., and Carmichael,I.S.E. (1990) Thermodynamic propertiesof
silicate liquids with emphasison density, thermal expansionand compressibility. In Mineralogical Societyof America Reviews in Mineralogy,24,25-64.
layne, G.D., and Stix, J. (1991) Volatile and light lithophile element
(LLE) evolution of the Jemez Mountains magmatic system.I. The interval betweencaldera formation at l.5l Ma and 1.14 Ma (abs.).Eos,
72(44),577.
Lowenstern, J.B. (1993) Evidence for a copper-bearingfluid in magma
erupted at the Valley of Ten Thousand Smokes,Alaska. Contributions
to Mineralogy and Petrology, | | 4, 409-421 .
I-owenstem,J.B., and Mahood, G.A. (1991) New data on magmatic H,O
conlents of pantellerites,with implications for petrogenesisand eruptive dynamics at Pantelleria.Bulletin ofVolcanology, 54, 78-83.
Lowenstern,J.B., Mahood, G.A., Rivers, M.L., and Sutton,S.R. (1991)
Evidence for extreme partitioning of copper into a magmatic vapor
phase.Science,252, 1405-1409.
Luhr, J F. (1990) Experimental phaserelations ofwater- and sulfur-saturated arc mag,masand the 1982 eruptions of El Chich6n Volcano.
JournalofPetrology,31, l07l-l I 14.
Macdonald, R., and Bailey, D.K. (1973) The chemistry of the peralkaline
oversaturatedobsidians. U.S Geological Survey ProfessionalPaper,
440-N-1, 37 p.
Macdonald, R., Smith, R.L., and Thomas, J.E. (1992) Chemistry of the
subalkalic silicic obsidians.U.S GeologicalSurvey ProfessionalPaper,
1 5 2 3 , 2 1 4p .
Mahood, G.A. (1984) Plroclastic rocks and calderas associatedwith
strongly peralkaline magmatism.Journal of GeophysicalResearch,89,
8540-8552.
Mahood, G.A., and Hildreth, W. (1986) Geology of the peralkaline volcano at Pantelleria,Strait of Sicily. Bulletin of Volcanology, 48, 143172.
Mahood, G.A., and Stimac, J.A. (1990) Trace-element partitioning in
pantellerites and trachltes. Geochimica et Cosmochimica Acta, 54,
2257-2276.
Mahood, G A., Halliday, A.N., and Hildreth, W. (1990)Isotopic evidence
for the origin ofpantellerites in a rift-related alkalic suite: Pantelleria,
Italy. International Volcanological CongressAbstracts, September38, 1990, Mainz, Germany.
Malinin, S.D., Kravchuk, I.F., and Delbove, F. (19E9) Chloride distribution between phases in hydrated and dry chloride-aluminosilicate
melt systemsas a function of phasecomposition. GeochemistryInternational. 26. 32-38.
Metrich, N., and Rutherford, M.J. (1992) Experimental study of chlorine
in hydrous silicic melts. Geochimica et CosmochimicaActz,56,607616.
Newman, S., Epstein, S., and Stolper, E. (1988) Water, carbon dioxide,
and hydrogen isotopes in glassesfrom the ca. I 340 ,c..o.eruption of the
Mono Craters, California: Constraints on degassingphenomena and
initial volatile content. Journal of Volcanology and Geothermal Research,35, 75-96.
Nicholls, J., and Carmichael, LS.E. (1969) Peralkaline acid liquids: A
petrological study. Contributions to Mineralogy and Petrology, 20,26r.
294.
Pitzer, K.S., and Pabalan,R.T. (1986)Thermodynamicsof NaCl in steam.
Geochimica et CosmochimicaActa. 50. 1445-1454.
Qin, 2., Lu, F., Anderson, A.T., Jr. (1992) Difiirsive reequilibration of
melt and fluid inclusions. American Mineralogist, 77, 565-576.
Roedder, E. (1972) I:boratory studies on inclusions in minerals of Ascension Island granitic blocks, and their petrologic significance. ProceedingsofCOFFI, 5, 129-138.
(1984)Fluid inclusions.MineralogicalSocietyof America Reviews
in Mineralogy, 12,644 p.
(1992) Fluid inclusion evidencefor immiscibility in magmatic differentiation- Geochimica et Cosmochimica Actz, 56, 5-20.
Roedder, 8., and Coombs, D.S. (1967) Immiscibility in gmnitic melts,
indicated by fluid inclusions in ejected granitic blocks fiom Ascension
Island. Journal of Petrology, 8, 417-45 l.
Shinohara, H. (1991) Pressuredependenceof water/rock reaction and
control of HCyNaCl ratio. In Y. Matsuhisa, M. Aoki, and J.W. Hedenquist, Eds., High-temperature acid fluids and associatedalteration
and mineralization: Report of the Geological Survey of Japan, 277,
97-99.
Shinohara, H., Iiyama, J.T., and Matsuo, S. (1984) Comportement du
chlore dans le systeme magma granitique-eau' Cornptes Rendus de
I'Acad6mie des Sciencesde Paris, 298,741-743.
(1989) Partition ofchlorine compoundsbetweensilicate rnelt and
hydrothermal solutions. I. Partition of NaCl-KCl. Geochimica et Cosmochimica Acta. 53. 2617-2630.
Silver, L.A., Ihinger, P.D., and Stolper, E. (1990) The influence ofbulk
composition on the speciationofwater in silicate glass.Contributions
to Mineralogy and Petrology, lO4,142-162.
Skirius, C.M., Peterson,J.W., and Anderson, A.T., Jr. (1990) Homogenizing rhyolitic glassinclusions from the Bishop Tuf' American Mineralogist,75, l38l-1398.
Solovova, I., Naumov, V, Girnis, A., Kovalenko, V., and Guzhova, A.
(1991) High+emperaturefluid heterogeneity:Evidencesfrom microinclusionsin Pantelleria volcanics (abs.).Plinius, 5, 206.
Sourirajan, S., and Kennedy, G.C. (1962) The system HrO-NaCl at elevated temperaturesand pressures.American Journal of Science,260,
ll5-141.
Tait, S. (1992) Selectivepreservationof melt inclusions in igneousphenocrysts.American Mineralogist, 77, 146-155.
Tuttle, O.F. (1952) Origin of the contrastingmineralogy of extrusive and
plutonic salic rocks. Journal ofGeoloey,60, l0'l-124.
Urabe, T. (1985) Aluminous granite as a sourcemagma ofhydrothermal
ore deposits:An experimentalstudy. Economic Geology,E0, 148-157.
Watson, E.B. (1991) Diffirsion of dissolved CO' and Cl in hydrous silicic
to intermediate magmas. Geochimica et Cosmochimica Acta, 55, lE971902.
Webster,J.D. (1992a) Fluid-melt interactions involving Cl-rich granites:
Experimental study from 2 to 8 kbar. Geochimica et Cosmochimica
Acra. 56. 659-678.
-(1992b)
Water solubility and chlorine partilioning in Cl-rich graoC.Geonitic systems:Effectsof melt composition at 2 kbar and 800
Acta,
56,
679-687.
chimica et Cosmochimica
Webster,J.D., Taylor, R.P., and Bean, C. (1993) Pre-eruptivemelt composition and constraints on degassingof a water-rich pantellerite magma, Fantale Volcano, Ethiopia. Contributions to Mineralogy and Petrology, I 14,53-62.
Westrich, H.R., Eichelberger,J.C., and Hervig, R.L. (1991) Degassingof
the 1912 Katmai Magmas. Geophysical ResearchLetters, 18, 15611564.
M,c.NuscnrFrRECEIVED
Juxe 14, 1993
NowbrseR 22, 1993
M,',m-rscnrrr ACCEPTED
AppBNorx
1.
Powderedmatrix glassfrom unit P32 was equilibrated at 200
MPa and 900 "C for 95 h. The /o, was kept closeto the Co-CoO
buffer. The sampleswere quenchedby turning offpower to the
internally heated pressurevessel,after which the sample temperaturedropped to <400 "C within 5 min. Pantelleritewith -2
wto/oadded HrO [that of melt inclusions from P32 (Lowenstern
LOWENSTERN: CHLORINE IN PERALKALINE MAGMA
and Mahood, l99l)l was crystal free after the experiment, indicating that 900 "C is above the liquidus for this sample. The
experiment shown in Figure 7C was equilibrated with a solution
of 80 wto/oNaCl and 20 vlto/oHrO (added to the charge as NaCl
crystals and deionized HrO). The glassyproduct contained approximately 9.5 wt9oNarO, 0.40loK2O,4.4o/oFeO,.,, -40lo HrO,
6500 ppm Cl, and amounts of other elements similar to those
of the starting composition. The loss of K and Fe from the sam-
369
ples was due to the lack of those elementsin the added fluid (as
salts) and the high fluid to glassratio (2.5). The NaCl-saturated
pantellerite (large solid circle in Fig. 2) was synthesizedunder
similar conditions, and the product containedapproximately 9.6
wtVoNarO, 2.0o/oKrO, 7.8o/oFeO,.,, <0.50/oHrO, I1700 ppm
Cl, and amounts of other elementssimilar to those in P32 glass.
NaCl was added to the charge as halite (halite to glass ratio :
0.3).