Phase equilibrium constraints on the viscosity of silicic
magmas 1. Volcanic-plutonic comparison
Bruno Scaillet, Holtz François, Michel Pichavant
To cite this version:
Bruno Scaillet, Holtz François, Michel Pichavant. Phase equilibrium constraints on the
viscosity of silicic magmas 1. Volcanic-plutonic comparison. Journal of Geophysical Research : Solid Earth, American Geophysical Union, 1998, 103 (B11), pp.27,257-27,266.
<10.1029/98JB02469>. <insu-00717688>
HAL Id: insu-00717688
https://hal-insu.archives-ouvertes.fr/insu-00717688
Submitted on 13 Jul 2012
HAL is a multi-disciplinary open access
archive for the deposit and dissemination of scientific research documents, whether they are published or not. The documents may come from
teaching and research institutions in France or
abroad, or from public or private research centers.
L’archive ouverte pluridisciplinaire HAL, est
destinée au dépôt et à la diffusion de documents
scientifiques de niveau recherche, publiés ou non,
émanant des établissements d’enseignement et de
recherche français ou étrangers, des laboratoires
publics ou privés.
JOURNAL
OF GEOPHYSICAL
RESEARCH,
VOL. 103, NO. Bll, PAGES 27,257-27,266, NOVEMBER
10, 1998
Phase equilibrium constraintson the viscosityof silicic
magmas
1. Volcanic-plutonic comparison
BrunoScaillet,Francois
Holtz,• andMichelPichavant
Centrede Recherchessur la Synth•seet Chimie desMin6raux, CNRS, Orl6ans,France
Abstract. By usingrecentlydetermined
experimental
phaseequilibriawe showthatthe viscosity
of granitic
magmas
emplaced
atuppercrustal
levelsisapproximately
constant
at -• 104.5
Pas,
irrespective
of theirtemperature
andlevelof emplacement.
Magmascrystallizingasgranitic
plutonsarenot water-poorandthusnot moreviscousthantheir extrusiveequivalents.Instead,
comparison
betweenpre-eruption
magmaviscosities
of extrusivesilicic-intermediate
andintrusive
graniticmagmasshowsthatthe formerareon averageslightlymoreviscous.Giventhe typical
strainratesin silicicmagmachambers,
magmarheologicalbehavioris expectedto be dominantly
Newtonian,bubbleshavinga minorrheologicalinfluenceat depthalthoughexceptions
canexist.
Thuswhethera silicic-intermediate
magmais eruptedor frozenat depthdependsprimarilyon the
rheologicalpropertiesof surrounding
terranesor on externaltectonicfactors,but not on the
rheologyof themagmaitself.However,preeruptive
viscosities
of extrusivemagmasrarelyexceed
106Pa.s,whichsuggests
thatcrystal-melt
mushes
withhigher
viscosities
cannot
leavethemagma
storageregionsbeneathvolcanoes.The narrowrangeof viscosities
displayedby silicicintermediate
magmasresultsfromboththe strongcontrolthatpressureexertson volatile
solubilitiesin silicatemeltsandthermallimitationsrequiredto produceacidmagmas.
Considerations
of the relationships
betweenmagmacrystallinities,
bulk SiO2,andpreeruptivemelt
H20 contentsshowthat the higherthe meltH20 contentis the higherthe maximumcrystallinity
that a givenmagmawill be while still beingpotentiallyerupted.An empiricalcorrelationis
proposed
thatenablesusto estimatepreeruptivemeltH:O contentsof eruptedmagmasby knowing
theircrystallinity
andbulk SiO:.
1.
Introduction
Silicic magmasare often chosento illustratethe end-member
behavior of viscous fluids on Earth, with inferred viscosities
that spannearly 10 ordersof magnitude(105-1013Pa.s),
whereas
basaltviscosities
clustertightlyaround101-102Pa.s
[e.g.,McBirney and Murase, 1984]. Such contrastedbehavior
is still viewed as a major reason for basalts being easily
erupted at the Earth' surface while silicic magrnaspond at
depth. However, recent studies have shown that from a
rheologicalstandpoint,granitesbehave in muchthe sameway
as basalts,yet they are colder by 300ø-500øC [see Wall et al.,
1987; Clemensand Mawer, 1992]. They are emplacedat high
levelsin the crustthroughfractures,with magmaflow ratesof ~
0.01-1ms-1andwithtemperatures
close
to theliquidus
when
emplaced,filling, in someinstances,upper crustalchambersof
~ 150km3withina fewyears
[e.g.,Clemens
andMawer,1992;
Scaillet et al., 1996]. Granitic magmasare neverthelessstill
widely perceived as failed rhyolites. This view implicitly
•Nowat Institutflir Mineralogie,
Universitat
Hannover,
Hannover,
Germany.
Copyright1998by the AmericanGeophysical
Union.
Papernumber98JB02469.
0148-0227/98/98 JB-02469509.00
assumesthat the water contents of the magmaticprecursorsof
plutonic rocks are much lower than that of their volcanic
counterparts[e.g., Johnsonet al., 1994]. Attemptsto determine
intensive parametersgoverning the evolution of plutonic
rocks through a classical petrological approachgive rise to
estimatesthat are, at best, legacies of near-solidus conditions
[e.g., Whitney, 1988]. Most of the current viscosity
determinations
for graniticmagmasare thus necessarilyfraught
with assumptionsregarding temperature,melt composition,
and crystal content. Early experimental work [Maaloe and
Wyllie 1975], thoughpioneeringthe use of phaseequilibrium
studiesas a workingtool for inferring intensive conditions in
the evolution of plutonic rocks, has sufferedfrom the lack of
analytical methods for measuring the water content of
quenched experimentalcharges and is thus of little use for
constrainingthe rheologicalpropertiesof granitic liquids and
magmas.This is apparentin the compilationof Clemens [1984]
on the water content of silicic-intermediate magrnas,where
most phase equilibrium studies carried out on plutonic rocks
provide only maximumestimatesof melt H20 contents. More
recently, Whitney [1988] reviewed experimentalevidence for
the water content of granitic magmasand concludedthat their
H20 contents were more likely in the range 2-4 wt %,
although for similar reasons(i.e., lack of analytical tools) the
melt H20 contents(not to be confusedwith the magmaH20
content) could not be better determined.In contrast, recently
27,257
27,258
SCAILLET ET AL.: VISCOSITY OF SILICIC MAGMAS, 1
performedphaseequilibrium studies have benefitedfromthe
analytical improvementsmade in the past 10 years in the
determination
of melt H20 contents[e.g.,Newmanet al., 1986;
Canyon Tuff latite, Spor Mountain rhyolite, E1 Chich6n
trachyandesite,Pinatubo dacite, and Mt. Pel•e andesite) or
from the direct measurementof the H20 content of melt
Deloule et al., 1995; Devine et al., 1995]. This enables a more inclusions,and FeTi oxidesfor temperature(BishopTuff, Cerro
accurate characterization of the rheological properties of Toledo, Lower Bandelier, Taupo Rhyolite, Crater Lake, Pine
experimentalphase assemblagesand granitic magmas.In Grove, Katmai, Fantale, Inyo Dome, and Krakatau). For the
parallel, the new analytical tools for measuringthe H20
Bishop Tuff several recent studies have determined its
content of quenched glasses,together with well-calibrated preeruptivemelt H20 content and variation during the main
thermobarometers for the determination of temperature, eruptive cycles [Andersonet al., 1989; Skirius et al., 1990;
pressure,andfO 2, havealso beenextensivelyapplied recently Dunbar and Hervig, 1992a; Wallace et al., 1995]. In this
to volcanicrocks. The two types of approachesnow allow an study we use melt H,•O contents given by Anderson et al.
accuratedeterminationof the major factorscontrolling magma [1989], Skirius et al. [1990] and Wallace et al. [1995], yet to
viscosity (temperature,liquid compositionincluding water avoid redundancy,only those of Skirius et al. are reportedin
content,and crystal content),either, for volcanic systems,in table 1 and in figures 1-4. For caseswhere the determinedH20
the magmastoragezone [e.g., Gardner et al., 1995a] or, for contentsof melt inclusionsanalyzedin a given sampledisplay
plutonicsystems,at the emplacement
level of the magrnas
[e.g., a dispersionmuchlarger than the analytical uncertainty (for
Clemens and Wall, 1981].
instance,Cerro Toledo) [Stix and Layne , 1996], although the
In this paperwe userecentlydeterminedexperimentalphase anhydrousmajorelementcompositionof the glass inclusions
equilibria to infer the viscosities of five crustally derived remainsconstant,only the highest values were used for the
granitesduring their ascentand emplacement
and to compare viscosity calculations.In doing so we assumethat H20-poor
themto the preeruptionmelt viscositiesof silicic-intermediate glass inclusionswere affectedby syneruptiveto posteruptive
extrusive magrnas.We show that melt viscosities cluster H20 loss, an assumptionsupportedby results obtained from
around104-105Pa s irrespective
of the eitherextrusiveor the study of Pinatubomelt inclusions[Westrich and Gerlach,
intrusivenature,temperature,and H20 contentof the magmas.
1992; Rutherford and Devine, 1996; Gerlach et al., 1996;
Scaillet and Evans, 1998]. Apart from Macusani, volcanic
rocks for which preeruptiveH20 contentshave been inferred
only throughmineral-fluid or mineral-meltequilibria were not
consideredhere. The reasonfor this is that such an approach
2. Method
and Rocks Studied
oftenyields a poorly constrainedrange of melt H,•O contents
To constrainthe viscosityof silicic plutonic rocks,we use given the uncertainties in experimental calibrations and
the results obtained from the phase relations recently solution models. These problems, added to the fact that
determinedon severalgranites.The granitic plutons are those reportedtemperaturesmay vary by more than 100øC, make it
of Strathbogie,Watergums(Australia), Manaslu, Gangotri difficult to chosethe appropriateT-H20 .... l•(wt%) values for
(Himalaya)and Jamon(Brazil) [Clemensand Wall, 1981; calculating melt viscosities. The Macusani volcanics are
Clemens et al., 1986; Scaillet et al., 1995; R. Dall'Agnoll et consideredhere becausepre-eruptive melt H,•O content and
al., Evolutionof A-type granitemagmas:an experimentalstudy temperatureare particularlywell constrained[Pichavantet al.,
of the lower proterozoicJamongranite, easternAmazonian 1988a,b] and allow the database to be extended toward
craton,Brazil, submittedto Journal of Petrology, 1998]. Their temperaturesbelow 650øC (Table 1). The compositional
SiO2 contentsrangefrom71 to 75 wt % and the rocksare spectrumcoveredby the volcanic rocks is largerthan that for
either peraluminousor metaluminous.
They are representative the plutonics, with bulk SiO2 contents ranging from
of differenttypesof graniticmagmatism
althoughall areformed trachyandesitic to rhyolitic (57-78 wt %). The data set
by partialmeltingof the continentalcrust.The experimental includes peralkaline, metaluminous, and peraluminous
studieshave constrainedthe T-H20 m melt(wt%) conditions compositions, which are associated with extensional
andcrystalcontentduringthe emplacement
of the magmas,
by (peralkaline) to compressional (peraluminous) regional
comparing
the experimental
phaseequilibriawith the natural tectonic stress regimes. Although examples of explosive
crystallization
sequence
as deducedfrom texturalobservations.eruptionslargelydominatethe dataset, effusivevolcanicrocks
All these studies have invariably confirmedthe crystal-poor are alsorepresented(Inyo Dome). For the explosiveeruptions
nature of the magmasduring emplacement
[see Wall et al., the samplesrepresenta wide range in ejected masses,with
1987], suggestingthat fractionationwas limited during ascent volumes
in denserockequivalent
rangingfrom0.1 km3 (Mt.
and hence that magmacomposition at its emplacementlevel Pe16e)
upto 3000km3 (FishCanyonTuff).Takentogether,
the
(where the magmachamber was built) was close to that plutonic and volcanic rocks representa range of intermediate
producedin the partial melting zone. In other words, these to silicic magmasinvolved in a large variety of tectonic
magrnaswere emplacedin a nearly entirely molten state in settings, which ensures that our conclusions are of general
upper crustal chambers,where crystallization occurred.The validity.
mostimportantaspectof the experimentalwork is that it has
confirmedthe water-rich nature (3-7 wt % H2Oln m•lt)of the
magrnas
parentalto these plutonic rocks. Thesestudies thus
representthe best opportunity to quantify the viscosity of 3. Viscosity Calculations
granites during at least two major steps of their evolution,
For metaluminousfelsic melts that fit the following chemical
ascent,emplacement,and crystallization.
For volcanic rocks we restrict our analysis to rocks whose criteria (wt %), we use the empiricalequation of Hess and
preeruptionconditionshavebeendeterminedeitherfrom phase Dingwell [1996]: SiO2 - 73.2-78.6; CaO+MgO+FeO= 0-2.34;
+ 2Fe3+)= 0.93- 1.02).For
equilibrium experiments(Mount St. Helens dacite, Fish (2Na+ 2K+ Mg+ Ca+ Fe2+)/(2A1
SCAILLET ET AL.: VISCOSITY OF SILICIC MAGMAS, 1
27,259
Table 1. CalculatedMelt and Magma Viscositiesfor Natural Intermediateto Silicic Plutonic and Volcanic Rocks
SiO2
SiO2'
Magma,
Melt,
Gangotri
Manaslu
Jamon
Stratboghie
Watergums
72.9
73.0
71.5
70.4
74.5
72.9
73.0
71.5
70.4
74.5
< 10
<10
<10
< 10
<10
700
750
920
850
900
Fantale
Fantale
Pine Grove
Pinatubo
El Chich6n
Crater Lake
Crater Lake
St. Helens-1980
St. Helens-T
St. Helens-Wn
St. Helens-Bi
St. Helens-Pu
St. Helens-Ye
St. Helens-Yn
St. Helens-Yn
FishCanyon
SpotMountain
Katmai Andesite
Katmai Dacite
KatmaiRhyolite
Inyo Dome
Inyo Dome
TaupoPlinian
Taupo Okaia
Taupo Hapete
BishopPlinian
71.0
71.0
75.1
64.0
57.2
73.0
73.0
62.8
63.2
67.2
64.1
63.3
65.5
65.9
65.8
64.4
72.9
58.6
64.6
77.4
71.5
73.7
74.0
74.0
74.0
77.0
71.0
71.0
76.4
78.3
71.4
73.0
73.0
73.5
70.2
74.8
72.8
74..3
75.6
75.5
74.8
77.1
72.9
74.3
77.4
78.5
71.5
73.7
76.0
76.0
76.0
77.0
14
14
32
47
57
10
10
37
31
27
41
46
40
35
32
40
30
25
10
1
5
5
5
13
5
7
800
800
675
780
800
885
885
930
893
847
913
870
795
768
790
760
735
990
950
900
897
874
850
798
839
710
BishopChidago
BishopMono
Cerro Toledo-15-8
Cerro Toledo-15-9
Cerro Toledo-15-11
Cerro Toledo-6-8
BandelierUpper
Mt. Pe16e
Montserrat
Krakatau Rhyodac.
77.0
77.0
72.0
73.4
73.0
73.0
73.1
61.0
59.0
69.5
77.0
77.0
72.0
73.4
73.0
73.0
73.1
75.0
77.6
72.0
Macusani
73.0
72.0
10
22
<5
<5
<5
<5
<5
50
40
10
45
Rock
wt %
wt %
Crystals,
% volume
T,
H20
logrl
logrla
Melt,
Melt,
Magma,
Pas
Method
b
References
c
4.2
5.0
4.1
4.6
4.7
4.4
5.2
4.2
4.8
4.9
Exp, Sc
Exp, Sc
Exp, Sh
Exp, Sh
Exp, Sh
1
1
2
3
4
4.6
4.9
7.1
3.9
3.7
5.3
4.2
4.0
5.8
IR, Sh
IR, Sh
IR, HD
5
5
6, 7
720
790
813
813
777
697
697
875
830
885
6.4
5.0
3.9
4.7
4.6
4.6
4.8
3.7
4.3
5.6
6.5
6.3
5.0
6.0
1.0
2.8
4.0
3.9
4.3
4.3
5.9
4.3
6.3
6.5
4.2
5.3
6.8
6.8
6.7
6.2
5.5
4.7
4.0
8, 9, 10
11
12
12
13
14
14
14
14
14
14
14
15
16
17, 18
19, 18
19, 18
20, 18
20, 18
21, 22
21, 22
21, 22
23, 18
23, 18
23, 18
24, 25
24, 25
24, 25
24, 25
24, 26
27
28, 29
30
7.2
5.7
7.2
4.9
4.4
4.4
4.3
5.1
5.3
5.7
5.3
5.0
4.7
6.0
5.6
4.7
4.6
4.4
4.5
4.5
5.0
4.6
5.1
5.3
5.2
5.7
4.7
4.3
4.5
5.3
5.5
5.8
5.5
4.6
5.9
Exp, IR, IP, HD
Exp, Sh
IR, Sh
IR, Sh
Exp, VDB, Sh
Exp, Sh
Exp, Sh
Exp, Sh
Exp, Sh
Exp, Sh
Exp, Sh
Exp, Sh
Exp, HD
Exp, HD
IR, IP, Sh
IR, IP, HD
IR, IP, HD
IP, Sh
IP, HD
IP, Sh
IP, Sh
IP, Sh
IR, HD
IR, HD
IR, HD
IP, HD
IP, HD
IP, HD
IP, HD
IP, HD
Exp, IR, Sh
Exp, IR, Sh
VBD, Sh
635
4.5
4.5
4.7
4.2
3.7
3.8
3.8
4.5
4.6
4.5
4.4
4.2
5.2
4.8
4.1
4.5
4.4
4.5
4.5
4.9
4.4
5.0
5.2
5.0
5.2
4.6
4.2
4.4
5.2
5.4
3.9
4.7
4.4
4.9
Calc, Sc
31
øC
wt %
Pas
Plutonics
7.0
5.5
4.5
4.5
3.5
Volcanics
For the SporMountainandMacusanirhyolitesthe H20 contentsincludeF, B, andLi.
aMagma
viscosity
calculated
usingthemeltviscosity
andtheEisnstein-Roscoe
relationship
asmodified
byMarsh[1981] forcrystal
contentsup to 30% volumefrom equation(2) for highercrystalcontents.
bMethod
ofH20determination.
Exp,experimental
phase
equilibria;
IR,infrared
spectroscopy;
IP,ionprobe;
VDB,volatile
by
difference;Calc.,calculatedthroughmineralequilibria.The modelusedto calculatethe melt viscosityis givenas follows:Sc, Scaillet
et al. [1996]; Sh, Shaw [1972]; or HD, Hessand Dingwell [1996].
peraluminousleucograniteones the equation ofScaillet et al.
[1996], experimentally calibrated on a natural strongly
peraluminousmelt, is used. The model of Shaw [1972] was
usedto calculatesilicate melt viscositiesof all remaining melts
because(1) it is the only one that explicitly takes into account
the liquid composition in hydrous systems, and (2) its
experimentalrangeof calibration,in termsof melt H20 content
and temperature,is similar to that relevant for natural granites
and for most silicic erupted magmas,as deduced fi•om the
studies mentioned in section 2. Although recent work has
shown that the Shaw model may either underestimate or
overestimatemelt viscositiesat low water contents [Stein and
Spera, 1993; Schulze et al., 1996; Dingwell et al., 1996;
Baker, 1996], good agreement between experimental and
calculatedmelt viscositiesis observedin the rangeof melt H20
contents of interest for the present study (mainly, 3-7 wt %
H20), differencesbarely exceeding1 log unit. Although low
melt H20 contentsdo occur in the evolutionof magmaticrocks,
they are attained mainly in volcanic contexts,during the final
stagesof degassing.All theseprocessesare not consideredin
this paper (see, for instance,Dingwell et al. [1996]) where only
preeruptionconditionsare taken into account.
It is well establishedthat silicate melt viscosities depend,
among other factors, on the strain rate, viscosities at high
strain rates exhibiting significant departure fi'omNewtonian
behavior [e.g., Webb and Dingwell, 1990; Dingwell and
27,260
SCAILLET ET AL.: VISCOSITY OF SILICIC MAGMAS, 1
Webb, 1990; Bottinga, 1994]. The strain rates required to
viscousstresses[Bagdassarovand Dingwell, 1992, 1993].
observenon-Newtonianbehavior in silicate melts correspond This relaxationtime 'cbis relatedto the melt viscosity and
to timescales3 orders of magnitudelower than the structural bubble radius R:
relaxationtimescale'cfor the melt [Webb and Dingwell, 1990],
'•b= •meltR / IJ
(3)
the latterparameterbeing relatedto the relaxedshearviscosity
rlmeltby the Maxwell relation:
c•beingthe surfacetension.If the strainrate is greaterthan 'cb
-•,
bubbles behave as rigid particles, like crystals, and the
T -' ]•melt/ Goo
(1) viscosityof the bubble-bearing
magmais increased[Steinand
Goo
being the infinite frequencyshearmodulus of the melt [see Spera, 1992]. Inversely,when the strainrate is slow (i.e., in
Dingwell and Webb, 1990]. Given that Goois constantfor most magmachambers),bubbles will deform viscously and the
silicate melts (log Goo= 10+0.5 Pa) [Dingwell and Webb , magmaviscosity is decreased[Bagdassarovand Dingwell,
of R at
1990], variations in the relaxation time arise predominantly 1993]. There are, obviously, no direct measurements
frommelt viscosity. For the range of melt viscosities found in preeruptionconditions(excludingfluid inclusionstrappedin
this paper (see below), the structuralrelaxationtime is of the melt inclusions or crystals).Crystallization experimentsof
with H20-CO 2 (+ H2S) fluidsaimed
orderof 10'5-10'6s. Estimated
convective
ratesin silicicmagma naturalmagmassaturated
chambers
areoftheorderof 10'8s'1 [seeSperaet al., 1988], at definingpreeruptionconditionsgenerallygive bubbleradii
which is largely below the threshold for the onset of nonNewtonian viscosity(3 log units below the log of the relaxed
in therange10-100gm [e.g.,Scailletand Evans, 1998]. These
areprobablyminimumvaluessincethe timescaleof laboratory
strainrate (1/'c= 105 s'l)). In addition,therehave been experimentsis likely to be much shorterthan the residencetime
suggestionsbased both on analogical [e.g., Brandeis and
Marsh, 1989] and numerical [e.g., Brandeis and Jaupart,
1986] modeling that convection in silicic magmachambersis
weak or nonexistent, convective motions virtually ceasing
when crystallizationstarts[Marsh, 1989]. Thesefactssuggest
that the melt phaseobeys dominantly a Newtonian behavior
under pre-eruptionconditions.
For calculating magmaviscosities having less than 30%
of magrnasin the storageregion, during which bubbles can
coalesceand thus increasein size (but note that large
deformations
will have the oppositeeffect;Stein and Spera
[1992]). Takingan averagemelt viscosityof 105Pa s, a surface
tension
of 0.1N m'• [Bagdassarov
andDingwell,1992], and
R = 100gm yields'cb= 100swhose
inverseis 10'2s-•.Thisis
significantly higher than strain rates in silicic magma
chambers,so that bubble-bearingmagrnascan be predicted to
have
lower viscositiesthan bubble-freeones. The experiments
(volume)of crystalswe use the Einstein-Roscoerelationship,
performed
by Bagdassarovand Dingwell [1992] on a viscous
as modifiedby Marsh [1981]. Recentstudieshave shownthat
'7 s-•)
this empirical relation holds well at low crystal contents (•- 109Pa s) rhyoliticmeltat low strainrates(10'5-10
[Pinkertonand Stevenson,1992; Lejeuneand Richet, 1995]. yielded the following relationship:
At highercrystalcontents,however,this relation is no longer
]']magma:
]']melt
[1 / (1 + 24.4f)]
(4)
valid [seePinkerton and Stevenson,1992], and we use the
following equation, as recommendedby Dingwell et al. wheref is the volume fraction of bubbles. Use of (4), together
[1993]:
with the estimateof preemption fluid contents,results in a
maximumdecreaseof the magmaviscosityby a factorof 3. This
]•magma-'
rlme,t{1 + 0.75 [•/f,•) / (1 'f/frei}'(2) lowering effectof bubbles upon magmaviscosity has been
wherel]magma
andlimeItare the viscosities
of the magmaand the shown to hold also in melt+crystal+bubble suspensions
[Bagdassarov et al., 1994], the presence of bubbles
melt, respectively,
f is the volume fractionof crystal,andfm is
counteracting
that of solid particles. For instance, a magma
the concentrationof crystals beyond which the magmahas an
having
16%
bubbles
and 16% sapphire sphereshas a shear
"infinite"viscosity. Following Marsh [1981] and Pinkerton
viscosity
almost
identical
to that of the melt phase alone.
and Stevenson[1992] we have chosen a value of 0.6 for fro.
Strictly, fm is a function of crystal size distribution, which is
unknownfor the magmasconsideredin this report,and thuswe
havetakenthe value for a monodispersesuspension.Equation
(2) gives relaxed shear viscosities, i.e., the viscosity at low
strainrates,in keepingwith the inferenceof low strain rates at
work in silicic magmachambers.
In additionto crystals,magmascan contain bubbles. Recent
studies of silicic- intermediate arc magmasusing either trace
element behavior [Wallace et al., 1995] or sulfur yields
associatedwith explosive eruptions [Scaillet et al., 1998]
yielded preeruption fluid contents that generally did not
exceed5 wt %, with fluids having H20-rich compositions.For
H20 densities in the range 800ø-900øC and 200-300 MPa,
Although strainratesin silicic magmachambersare expectedto
be low, it is instructive to calculate how much magma
viscosity will be affectedby bubbles under high strain rates.
Stein and Spera [1992] have obtainedthe following empirical
equation for calculating the viscosity of magmaticemulsions
witha meltviscosityof 105Pa s, andlow gascontent
(< 7%
vol),at relatively
highstrainrates(0.06-7s'l):
]•magrna
= •melt( 1 + 13.1f)
(5)
wherel]magma
standsfor the viscosity of melt+bubblesandf is
the volumefractionof bubbles.Equation(5) showsthat under
the storageconditionsof silicic magmas,high strain rateswill
rise by a factor of 2 the magmaviscosity. Altogether, these
such an amount translates into a fluid volume fraction of •
considerationssuggestthat as far as preeruptionconditions
10%. As with crystals, the effect of bubbles on magma are considered,bubblesplay a minor role in magmarheology
viscosity dependson size distribution and strain rate, among when compared,for example,to that played by variations in
other parameters(i.e., capillary number and viscosity ratio) crystalcontentsof eruptedmagmas(0-60%) or to the viscosity
variations arising from changes in melt H,_O contents.
[Spera et al., 1988; Stein and Spera, 1992; Bagdassarov and
Dingwell, 1992, 1993]. The presenceof bubbles introducesan Accordingly, unless otherwise stated, magma viscosities
additionalrelaxationtime, relatedto the responseof bubblesto reportedhere do not take into accountthe role of bubbles. In
SCAILLET ET AL.' VISCOSITY OF SILICIC MAGMAS, 1
view of the above calculations and considering the possible
rangeof fluid contentfound in natural magrnas(see above), the
melt viscosities reported in this work can therefore be
considered
to be accurate to within
I meltviscosity
I
a factor of 3.
o
o o.
4. Melt and Magma Viscosities Before Uprise
The calculatedmelt and magmaviscositiesof the rocks used
in this study, together with the main compositional features
necessaryfor the calculations,are listedin Table 1. In Figure 1
the melt H20 contents of granitic magmasare plotted against
the temperature,both variables being taken for conditions
prevailing during magmaemplacementin the upper crust. A
similar plot is constructedfor the preeruptionT-H20 (wt %)
conditionsof the volcanicrocks.The overall patternis a nearly
linear inverse relationship between T and H20, the drier
magrnasbeing the hottest. No difference appears between
plutonic and volcanicrocks.In Figure 2 the melt viscosities of
granitic magrnas are plotted against the emplacement
temperatureof the magma and comparedto the preeruptive.melt
viscosities of the volcanic rocks. Again, both plutonic and
volcanic melt viscositieslie on a single horizontal trend over
more than 350øC, virtually spanning the entire temperature
range coveredby silicic magrnas[Ghiorso and Sack, 1991]. It
is worth mentioning that the Macusani volcanics, whose
estimated preeruptive temperatureand melt H20 content are
635øC and 5 wt %, respectively [Pichavant et al., 1988a,b],
give a liquid viscosityof 7.2 unit log Pa.s; that is, they would
plot well above the general trend. However, if the effectson
melt viscosity ofvolatiles other than H20 (F, Li, and B) are
considered [Dingwell et al., 1992; Baker and Vaillancourt,
1995] and assumingthat these effectsadd up, the preeruptive
melt viscosityfor the Macusanivolcanicsis 4.89 log unit Pa s,
in good agreementwith the calculatedviscositiesfor other less
evolvedfelsicmagmas.A similarrationalecan be appliedto the
Spor Mountain rhyolite in which the additive effectsof H2¸
and F result in a preeruption viscosity similar to other
volcanicrocksas well (Table 1). Theseexamplesillustrate the
fact that the pattern of Figure 2 is not specific to magmasin
which the dominantvolatile is H2¸ but also applies to those
highly enriched in halogens or other volatile elements.
Conversely, it suggeststhat melts enriched in volatiles other
o o
ßo oa
0
600
0
volcanic
ß
plutonic
i
i
700
i
27,261
o
oO
a
Ovolcanic
1
ß
0
600
!
I
700
i
i
,
800
Temperature
plutonic
I
900
lOOO
(øC)
Figure 2. Melt viscosities for granitic magmas(solid circles),
plotted as a function of their eraplacementtemperatureand
comparedto the preeruptionmelt viscosities of volcanic rocks
(open circles) listed in Table 1. For all granites the melt
composition used in the viscosity calculation is that of the
bulk compositionused in the phaseequilibrium studies. See
text and Table 1 for detailson viscositycalculations.
than water will not segregatefrom their source earlier (i.e., at
lower viscosity) than those containing only dissolved water.
The averagevalue of melt viscosityfor the whole data set is log
•melt= 4.5 -t-0.5 (Pa s).
The experimentalwork indicatesthat magmasparentalto the
plutonic rocks were all emplacedat near liquidus conditions,
with crystal contents less than 10 wt % and thus that melt
viscosities were close to magma viscosities. In contrast,
preeruptive crystallinities for volcanic rocks vary from 0 to
60%, displaying a clear inverse relationship with bulk SiO2
content (Figure 3). This relationship could partly reflect the
increasein incubation time for the onset of crystallization in
silicatemelts as SiO2 increases.Numerical modeling indicates,
however, that crystallization in large magmachambersoccurs
under quasi-equilibrium conditions with small undercooling,
exceptnear the very borders at early stages of the intrusion
[Brandeis et al., 1984]. Given that the volcanic rocks used in
this studycomefrom magmachambersof moderateto large sizes
(1 to severalhundredkm3, the volumeof eruptedmagma
representingon average 10-20% of the magmachamber),it i s
likely that the thermal regime in their storage region was
characterized by slow cooling rates and thus that the
relationshipof Figure 3 is not due to kinetic effects.Therefore,
given that both intrusiveand extrusivemagmasdisplay similar
melt viscosities,if the effectsof crystalsare taken into account,
the viscositiesof silicic volcanic rocks are on average slightly
higher than that of their granitic counterparts (Figure 4).
Nevertheless, apart from a single outlier (El Chich6n
trachyandesite),volcanic rock viscosities plot in a restricted
viscosity
range,104-106Pa s. Thissuggests
thatmagmas
more
viscous than 106 Pa s cannot leave their host chamber but
I
I
800
900
1000
Temperature (øC)
Figure 1. Relationship between temperatureand melt H20
content for both plutonic (solid circles) and volcanic (open
circles) rocks. For the plutonic rocks the conditions
correspondto magmaemplacement,
and for volcanic rocks the
conditions correspondto preeruption conditions. Source of
data is given in Table 1.
insteadfreeze at depth. This does not preclude silicic magmas
from reachinghigher viscosities,however, but highly viscous
conditionsare likely to be attained only during the last stages
of magmaascentand degassing [e.g., Dingwell et al., 1996].
Although magma viscosities are less clustered than melt
viscosities, the general pattern with temperaturevariation is
again flat, which indicates that hot magrnasare not erupted
more easily than cold ones. Our analysis therefore suggests
that in many, if not all, instances the preeruption melt and
magma
viscositiesare at, or below, 104.5and 106 Pa s,
respectively.One apparentexceptionto this rule is provided
27,262
SCAILLET ET AL.: VISCOSITY OF SILICIC MAGMAS, 1
[1981] for a series of basaltic-andesiticlavas erupted fromthe
Aleutiansvolcaniccenterof Atka. Both trends are parallel, the
one obtainedhere being displacedtoward higher SiO2 content
(Figure 3). In both cases,volcanic rocks with more than 5 560% crystals are not found, an observation that suggests, as
Marsh noted,that there is a locking point, characterizedby a
critical crystallinity,beyondwhich magma viscosity increases
so dramatically that it behaves essentially as a solid, a
proposal corroboratedby experimentaldata [e.g., Pinkerton
and Stevenson, 1992; Lejeune and Richet, 1995]. In his
(a)
•
60
0
oo
•
40
ß-
20
o
o
o
0
Marsh [1981]
o
0 0 0
0
..............
55
65
60
¸'
70
'
75
80
$i02 (wt %) bulk rock
......................
80IIntermediate
tosilicic
mugmad
I
o•'60| •
•
•
volcanic
rocks
(log
T]magma-5.1
_+
0.7
Pas).
However,
(b) for
although
the
magmatic
H20
contents
were
not
determined,
maximum
crystallinity
.....
L •"'--••^
>, [ •
•
•
.•C••
analysis
of theAleutians
lavas,Marsh deriveda valueof 104'6
Pas for this viscositythreshold.This value is lower by 0.5 log
units than the averagemagma viscosity obtained in this study
nearly
dryconditions
(< 1wt%)wereinferred
fortheAleutian
wt'/oH20 lavas
[Marsh,
1981],
asopposed
to thevolcanic
rocks
•O
- melt
I
presented
here
whose
average
melt
H20
content
is•-4-5wt%.
'E
40••,,••-•••
I The
two
samples
lying
inbetween
the
Aleutian
lavas
trend
and
••20
•
1 4 s 6O-.•,.Q½....•
Katmai,
whose
water
contents
are
among
the
lowest
of
the
data
[ 0•a•rs••:]•
• ••.v,••,.x••• set,
land
2.8
wt
ø/3,
respectively.
Conversely,
the
point
the one obtained here are the andesitc and dacite erupted at
0
55
•
60
65
70
•
75
'\.r• l
80
plotting well abovethe generaltrend corresponds
to the
volcanicrock richestin water,with 7.1 wt % dissolvedH20
(Pine Grove). Thus we confirmthe prediction of Marsh that
water increasesthe critical crystallinity at a given SiO2
Figure3. Relationship
between
magma
crystallinityandbulk content.In otherwords,mugmadrich in water can be eruptedat
rockSiO2for thevolcanicrocksof Table1, including
(a) a plot
a higher crystallinitythan thosepoor in H20. As proposedby
of theraw datatogetherwiththeAleutianlavastrendobtained
Marsh, the relationshipbetweenbulk SiO2 contentand magma
by Marsh [1981]and(b)the samediagramcountoured
with crystallinity can be used in a quantitative way to infer
lines of isomeltH20 contents.All fits are handmadebest
preeruptive melt H20 contents. A generalization of his model
estimates.
has been attemptedwith our data set and is shown in Figure
3b, where the diagramof Figure 3a is contoured with lines of
by the E1 Chich6n magma,whose crystal content is close to constant melt H20 contents. All correlations were assumed to
that at which the magmareachesthe solid behavior field. The
be linear, yet some curvaturesprobably occur at both ends of
calculatedmagmaviscosity is morethan 1 order of magnitude
the trends. Most of the previous analysesignore the possible
higher relative to other mugmadand, at such high crystal role of bubbles. The overall coherent trend obtained does
content, mugmadcan be anticipatedto also have a significant
suggest,however, that in most instancesthe gas phase
yield strength[Pinkerton and Stevenson,1992]. Nevertheless, abundanceis too low to have any appreciableeffecton magma
this magmahas been erupted,and the eruption was of Plinian theologyunderpreeruptionconditions,as previouslyargued.
type. Interestingly, the restoredpreeruptionfluid content of Nevertheless,not all the volcanic rocks conformto the lines of
this magmarangesbetween 8 and 26 wt %, dependingon the
isomeltH20 content,in particularin the high SiO2 range.For
CO2-SO2ratio used [Scaillet et al., 1998], which implies that instance,the Mono unit of the Bishop Tuff is clearly off the
the volume fractionof the fluid was • 20-50% of the scavenged
generaltrend (Table 1). Owing to the lack of preciseprepart of the magmachamber.Consideringthe role of bubbles on eruptivetemperature
constraints,the Lower BandelJerTuffs
magmatheologyas summarizedabove, we tentatively suggest
that the eruptionof E1Chich6nmagmawas madepossible,or at
leastgreatlyfacilitated,by the large volume of the fluid phase.
0 volcanic
Specifically,the experimentalresults of Bagdassarov et al.
ß plutonic
[1994] on the effectof bubblespredict that the viscosity of E1
Chich6n magmais only marginally differentfrom that of the
melt phasealone, falling well within the range displayed by
o o
o
the othermugmad(Table 1). Finally, thoughin someinstancesa
relation between eruption intensity and magmaviscosity has
been clearly established(Mount St. Helens) [Gardner et al.,
1995a], no correlationbetweenmagmaviscosity and eruption
magma
viscosity
I
volume,or any otherparameter,was found.
0
I
I
,
Si02 (wt %) bulk rock
o
600
,Oooo o%
o
.Oo%
8
700
800
Temperature
5. Inferring Preeruptive Melt H20 Contents
From Crystallinity
The relationship between crystallinity and bulk SiO2
(Figure 3) is qualitatively similarto that obtainedby Marsh
¸
900
1000
(øC)
Figure 4. Magma viscosities versus temperature for the
plutonic(solid circles) and volcanic(open circles)rocks listed
in Table 1. Note that plutonic rocks are on average slightly
less viscous than volcanic
ones.
SCAILLET ET AL.: VISCOSITY OF SILICIC MAGMAS, 1
analyzed by Dunbar and Hervig [1992b] could not be used in
the present study, but someof the sampleswould plot in the
samearea as the Mono lobe. Although, as illustrated by E1
Chich6n,a high amountof fluid could explain the off position
of thesehigh-silica rhyolites, the restoredfluid content of the
Mono unit is not abnormally high (<5 wt %) [Wallace et al.,
1995]. This lack of correlation may thus reflect a specific
characteristicof the large-magnitudevolcanic eventsthat gives
rise to ignimbrite deposits, with eruptive dynamics
fundamentally different from those of smaller eruptions. It is
possiblethat the locking point rule does not hold for certain
eruptions of particularly large magnitude. However, not all
ignimbrites plot out off the trends, as exemplifiedby the Fish
CanyonTuff, one of the largestknown ignimbrite deposits that
fits well with other eruptive events. Clearly, future work will
improvethe quality of the fit betweencrystallinity, bulk SiO2,
and melt H20 content, but we stress that such a diagram
(Figure 3) is useful for a first-order estimation of preeruptive
melt H20 contentsof eruptedmagmas,especially those ejected
during eruptionsof small to moderatemagnitude.
27,263
producedin fluid-absent melting experimentscarried out on
crustalprotolithsarein the range103-105Pa s [Rushmet,
1995; Wolf and Wyllie, 1995], which supports this
conclusion.It should be stressedthat the different granites
consideredin this study, althoughoriginating by fluid-absent
melting processes,differ in the nature of the main hydrous
phaseinvolved in the melting reaction [Whitney, 1988]. Yet
they all lie on the sameviscosity-temperaturetrend, which
indicates that the nature of the melting reaction has little
influenceon melt segregationand separationfrom its source
region.In the sameway, the extentof melting,which mayvary
dependingon the natureand abundanceof the hydrousphase
breakingdown (from 10 to 50%), appearsnot to be the critical
factortriggeringupwardmelt migration.The experimentaldata
on plutonic rocks imply that granitic melts should be nearly
perfectly extracted from their source with little entrainment of
restites [see Clemensand Mawer, 1992].The fact that there is
no crustal, restite-poor, granite magmaemplacedwith melt
viscosities
demonstrably
higherthan10s Pa s, suggests
that
this value also representsa viscosity threshold above which
no melt can migratealone (i.e., without resrites),whateverthe
P-T-conditions
and the state of stress in the source.
The overall viscosity pattern clearly indicates that the
volcanic-plutonic dichotomy does not arise from rhyolites
The rather low viscositiesfound in this study are consistent
being systematically less viscous than granites. It
with the recent proposal that felsic magmamigration occurs demonstratesalso that viscosity cannot control the level of
through dikes rather than via diapirs [Clemens and Mawer,
emplacementof silicic magmas.Factors like regional or local
1992; Petford et al., 1993]. Emerman and Marret [1990] have tectonic stress fields [e.g., Scaillet et al., 1995] or the
shownthat the modeof ascentof magmas(as either a dike or a mechanicalpropertiesof the traversedterranes [e.g., Hogan
diapir) can be estimated from the following dimensionless
and Gilbert, 1995] are morelikely causesfor the trapping of
parameter:
silicic magmasin the crust[seeClemensand Mawer, 1992]. In
contrast
to viscosity,densityis probablyone of the key factors
)k= l']magma
C/ d A(•
(6)
that will control the ultimatelevel of emplacement
of magmas.
wherec is the speedof propagation,d is the differencebetween However, there is no consensus about the actual value of the
the depth at which deviatoric stress becomessignificant and partial molar volume of water dissolved in silicate melts, as
the depthof final emplacement,
andAo is the differencebetween well as about the possiblecompositionaldependencyof this
the least compressivestress(horizontal for dike emplacement) parameter[Lange, 1994]. In addition,as emphasizedin section
and the other horizontal principal stresses. This relation 3, most silicic-intermediate magmasare saturated in a fluid
6. Dikes, Diapirs, and Emplacement Levels
evaluates
the time needed to reorientate
a dike relative
to its
phase[e.g., Wallaceet al., 1995], which seriouslycomplicates
rateof propagation.
If X is muchlower than 1, then the magma density calculations. For instance,assumingfor simplicity a
will propagateupward as a dike without sufferingsignificant
reorientation due to the externalstressfield. Taking A(• = 10
crystal+melt
density
of 2.4g cm-3anda fluid(H20) densityof
0.5g cm-3,themagma
density
decreases
between
2.31 and2.02
MPa,d = 10,000rn,c = 1 ms-1,and•lmagma
= 104'5Pas,the g cm-3if its fluid contentincreases
from1 to 5 wt %, i.e.,a
averageviscosity found in this study for plutonic magmas, density variation similar to that arising from a temperature
changeexceeding1000øC[see Lange and Carmichael, 1990].
chanceof endingtheir crustaltravel as diapirs. Note that even This outlinesthe dramaticimportancethat the fluid phasemay
if the speed of propagation is decreasedby 5 orders of have on magma density. The large uncertainties that still
magnitude
(i.e.,c = 104m s-l),themagmawill still ascend
asa surround this last factor make the results of density
dike. The parameterX may approach,however,values close or calculationsfor magmashighly model-dependent,
and for these
higher than 1 at very small ascentdistances(d = 10-100 m), reasonssuchcalculationswere not attemptedin this study.
and the emplacement
level would be close, in fact,to the melt
generation level.
7. Concluding Statements
The diking mechanismis characterizedby rapid flow of the
melt, which preventscrystallizationand contaminationduring
One of the striking featuresof this study is that silicic
uprise [Clemensand Mawer, 1992], and explainsthe crystal- magmasdisplaynot only an upper limit in viscositybut also a
poor characterof the magmasat the end of their crustaltravel. lower one. Although the formerlimit is understandablefrom a
Thus, for magmasthat are producedby partial melting of the rheologicalpoint of view (i.e., the magma becomestoo viscous
to flow), the reasonsfor the latter are less obvious. However,
middle-lowercontinentalcrustand sincepressurevariationsof
with a viscositylowerthan 104Pa s are
less than 1 GPa have a negligible effecton melt viscosity silicic magmas
apparently not produced. This compilation illustrates that
[Burnham, 1964; Schulzeet al., 1996; Scaillet et al., 1996],
the viscosity-temperature
trend of Figure 2 suggeststhat melt besides their anhydrous composition, two main variables can
extractionfromthe partially molten sourceregion also occurs lower the viscosityof silicic magmas,temperatureand volatile
for viscosities
around104-105Pa s. Viscosities
of liquids contents. However, as discussed below, neither temperature
givesX = 10-6'5,whichimplies
thatgranitic
magmas
havelittle
27,264
SCA1LLET ET AL.: VISCOSITY OF SILICIC MAGMAS, 1
nor the volatile content can be expected to vary so as to
decreasethe viscosity of silicic magmasbeyond the lower
viscositythresholddisplayedin Figure 1. In order to decrease
significantly the viscosity of silicic magmas,temperatures
largely in excessof 1000øC would be required (Figure 2).
Nevertheless,at least for moderngeologicalperiods (i.e., postArchaean), the geothermalregime of the continental crust
preventssuch high temperaturesfrom being reached,even in
thinned continental crust. Ponding of hot maficmagmasat the
base of the crust can induce partial melting, but thermal
modeling indicates that the temperature of silicic magmas
generatedin this way hardly exceeds950øC [Huppert and
Sparks, 1988]. In addition, the producedmelts,although hot,
will also be dry owing to the low H20 content of mosthighgrademetamorphicrocks [e.g., Clemensand Vielzeuf,1987]. If
free H20 is present, then melting of crustal lithologies at
temperaturesaround 1000øC will not yield rhyolitic magmas
but, instead, andesitic magmas or even more mafic
compositions. Protracted fractional crystallization of mafic
magmas,on the other hand, can produce silicic derivatives
[e.g., Helz, 1987] but only at temperaturesbelow 1000øC,
especiallyunder H20-rich conditions[e.g., Sissonand Grove,
19931.
An increase in the water content
of the melt to decrease its
magmaswith restored high fluid contents are also those with
high crystal contents,which intuitively makes sense,since the
amount of fluid must come in part from the crystallization of
anhydrous minerals.However, the mechanismsat work in arc
magmasare much more complicatedthan those arising from a
simpleclosedcrystallizationmechanism.
Processes
like mixing
with mafic magmascan seriouslycomplicatethis scenario.This
parameterneeds to be specifically addressedfor each single
volcanicevent, and we encouragethat such type of studiesbe
carriedout in the future. Importantand immediateapplications
of suchdata would be to accuratelyestimatemagmadensities.
For plutonicrocksthere are no, as yet, methodsto estimatethe
potential importanceof fluids at the magmaticstage, though
some were clearly rich in water at early stages of their
evolution [Scaillet et al., 1995]. The diking mode of magma
ascentimplies high ascentrates and, in turn, high strain rates.
Under these conditions,as with eruptive phenomena,bubbles
will increase magma viscosity. However, if we take the
comparison done in this work at face value, there are no
reasonsto believe that plutonic magmasdisplay higher fluid
contentsthan their volcanic counterparts.Instead, the current
consensus is that crustally derived granitic magmas are
generated, and most probably evolved, under fluid-absent
conditions[i.e., Clemensand Vielzeuf,1987].
viscosity is the other possibility to obtain a melt viscosity
significantly
lowerthan104Pas.However,
owingto therange
of pressuresat which mostextrusivemagrnas
are stored (200300 MPa), the maximumwater solubility of a rhyolitic melt
will be in the range 6-7 wt %, and the correspondingmelt
Acknowledgement. This studywas financiatedby grantsfrom the
IDYL researchprogram of the CNRS. Informal reviews done by Jos6
PonsandRay Macdonaldare greatly appreciated.The official reviews
doneby F. Spera,J. CrispandD. Dingwellare gratefullyacknowledged.
viscositywill beagainin therange104-105
Pa s, depending References
on temperature.Clearly, a silicicmelt at 1000øCand with 7 wt
% H20 will have a very low viscosity but, again, there are no
known geological mechanisms able to produce such
conditions, in keeping with the strong inverse correlation of
Figure 1 that demonstrates
that H20 and temperatureare not
independentparametersin magmas.Another parameterto be
consideredis the presenceof volatiles other than H20 that
have a lowering effect upon viscosity (B or F). However,
despite the fact that these elementshave an additive effecton
viscosity (F and H20 )[Baker and Vailtancourt, 1995], the
available evidence(Macusani and Spor Mountain) shows that
there is an inverse correlation
between
the abundance
of water
and that of other fluxing elements,so that the net effect on
viscosity is not expected to be large. A possible major
exception to this rule is provided by strongly peralkaline
rhyolites whose water solubility can be noticeably higher
than 6-7 wt % at 200-300 MPa, as shown experimentally [e.g.,
Linnen et al., 1996; Dingwell et al., 1997], with
correspondinglow viscosities [Dingwell et al., 1998]. Such
magmas
maywell displayviscositieslowerthan 104Pa s
beforeeruption.Thus,apartmaybefrom peralkalinemagmas,the
above lines of evidence suggest that whatever is the
mechanismthat produces a silicic melt, the magma most
inevitably endsin a quite narrow viscosity window, which is
defined both by the strong control that pressure exerts on
volatile solubilities in silicate melts and by the thermal
limitationsrequiredto produceacid magmas.
Consideringthe strainratesin silicic magmachambers,it is
expected that the magmarheology falls in the Newtonian
domain.In general,the availableevidencesuggestthat the role
of bubbleson magmarheology can be neglected,but examples
like the E1 Chich6nmagmashowthat the fluid phasemay be an
important rheological factor. In this respect,we note that
Anderson, A. T., S. Newman, S.N. Williams, T.H. Druitt, C.
Skirius,and E. Stolper,H20, CO2, C1,and gas in Plinian
and ash-flow Bishoprhyolite, Geology,17, 221-225, 1989.
Bacon,C.R., S. Newman,andE. Stolper, Water, CO2, C1, and F
in melt inclusions in phenocrysts from three Holocene
explosiveruptions,Crater Lake, Oregon, Am. Mineral., 77,
1021-1030,
1992.
Bagdassarov, N.S., and D.B. Dingwell,
Rheological
investigationsof vesicular rhyolite, d. Volcanol. Geotherm.
Res., 50, 307-322, 1992.
Bagdassarov,N.S., and D.B. Dingwell, Frequency dependent
rheology of vesicular rhyolite, d. Geophys.Res., 98, 64776487, 1993.
Bagdassarov, N.S., D.B. Dingwell, and S.L. Webb,
Viscoelasticity of crystal- and bubbles-bearing rhyolite
melts, Phys. Earth Planet. Inter., 83, 83-99, 1994.
Baker, D.B., Granitic melt viscosities: Empirical and
configurational entropy models for their calculation, Am.
Mineral., 81, 126-134, 1996.
Baker, D. B., and J. Vaillancourt, The low viscosities of
F+H20-bearing granitic melts and implications for melt
extractionand transport, Earth Planet. Sci. Lett., 132, 199211, 1995.
Barclay, J., M.J. Rutherford, M.R. Carroll, M.D. Murphy, J.D.
Devine, J. Gardner, and R.S.J. Sparks, Experimental phase
equilibria constraintson pre-eruptivestorageconditionsof
the SoufriereHills magma,Geophys.Res. Lett., in press,
1998.
Bottinga, Y., Rheology and rupture of homogeneoussilicate
liquids at magmatictemperatures,d. Geophys. Res., 99,
9415-9422, 1994.
Brandeis, G., and C. Jaupart, On the interaction between
convection and crystallization in cooling magmas
chambers,Earth Planet. Sci. Lett., 77, 345-361, 1986.
Brandeis,G., and B.D. Marsh, The convective liquidus in a
solidifyingmagmachamber:A fluid dynamicinvestigation,
Nature, 339, 613-616, 1989.
SCAILLET ET AL.: VISCOSITY OF SILICIC MAGMAS, 1
27,265
Fire and Mud.' Eruptions and Lahars of Mount Pinatubo,
Brandeis, G., C. Jaupart,and C.J. All•gre, Nucleation, crystal
editedby C.G. Newhall and R. Punongbayan,pp. 415-434,
growth and the thermal regime of cooling magmas,d.
Univ. of Wash. Press, Seattle, 1996.
Geophys.Res., 89, 10161-10177, 1984.
Burnham,C.W., Viscosityof a water rich pegmatitemelt at high Ghiorso, M.S., and R.O. Sack, Fe-Ti oxide geothermometry:
Thermodynamicformulationand the estimation of intensive
pressures,Spec.Pap. Geol. Soc.Am., 76, 26, 1964.
variablesin silicic magmas,Contrib. Mineral. Petrol., ! 08,
Clemens,J.D., Water contentsof silicic to intermediatemagmas,
Lithos, 17, 273-287, 1984.
Clemens,J.D., and C.K. Mawer, Granitic magmatransport by
fracturepropagation,Tectonophysics,
204, 339-360, 1992.
Clemens, J.D., and D. Vielzeuf, Constraints on melting and
magmaproductionin the crust, Earth Planet. Sci. Lett., 86,
485-510, 1991.
Helz, R. T., Differentiation behavior of Kilauea Iki lava lake,
Kilauea volcano, Hawaii: An overview of past and current
work, in Magmatic
Processes.' Physicochemical
Principles, vol. 1, editedby B.O. Mysen, pp. 241-258, The
Geochem. Soc., Pennsylvania,1987.
287-306, 1987.
Clemens,J.D., and V.J. Wall, Crystallizationand origin of some Hervig, R.L., N. Dunbar, H.R. Westrich, and P.R. Kyle, Preeruptivewater contentof rhyolitic magmasas determinedby
peraluminous(S-type) graniticmagmas,Can. Mineral., 19,
ion microprobeanalysesof melt inclusions in phenocrysts,
111-131, 1981.
d. Volcanol. Geotherm. Res., 36, 293-302, 1989.
Clemens,J., J. Holloway, andA.J.R. White, Origin of an A-type
granite: Experimentalconstraints,Am. Mineral., 71, 317- Hess, K.U., and D.B. Dingwell, Viscosities of hydrous
leucogranite melts: A non-arrhenianmodel, Am. Mineral.,
324, 1986.
Deloule, E., O. Paillat, M. Pichavant, and B. Scaillet, Ion
81, 1297-1300, 1996.
microprobe determination of water in silicate glasses: Hogan, J.P., and M. C. Gilbert, The A-type Mount Scott
granite sheets: Importance of crustals magrnastraps, d.
Methods and applications,Chem. Geol., 125, 19-28, 1995.
Geophys.Res., 100, 15779-15792, 1995.
Devine, J.D., J.E. Gardner, H.P. Brack, G.D. Layne, and M.R.
Rutherford, Comparison of microanalytical methods for. Huppert, H.E., and R.S.J. Sparks, The generation of granitic
magmasby intrusion of basalt into continental crust, d.
estimating H20 contents of silicic volcanic glasses,Am.
Mineral., 80, 319-328, 1995.
Petrol., 29, 599-624, 1988.
Dingwell, D.B., and S.L. Webb, Relaxation in silicate melts, Johnson, M.C., and M.C. Rutherford, Experimentally
determinedconditions in the Fish Canyon Tuff, Colorado,
Eur. d. Mineral., 2, 427-449, 1990.
magmachamber,d. Petrol., 30, 711-737, 1989.
Dingwell, D.B., R. Knoche,S.L. Webb, and M. Pichavant,The
effectof B20 3 on the viscosityof haplograniticliquids,Am. Johnson, M.C., A.T. Anderson, and M.J., Rutherford, Preeruptive volatile contents of magmas,in Volatiles in
Mineral., 77, 457-461, 1992.
Magmas, Rev. Mineral., vol. 30, editedby M.R. Carroll and
Dingwell, D.B., N.S. Bagdassarov,J. Bussod,and S.L. Webb,
J.R. Holloway, pp. 281-330,
Mineral. Soc. of Am.,
Magmarheology, in: Short Handbook on Experimentsat
Washington,D.C., 1994.
High Pressureand Applicationsto the Earth's Mantle, vol.
21, edited by R.W. Luth, pp. 131-196, Mineral. Assoc. Keith, J.D., and W.C. Shanks III, Chemical evolution and
volatile fugacitiesof the Pine grove porphyry molybdenum
Canada., Ontario, 1993.
and ash-flow tuff system, southwestern Utah, in Recent
Dingwell, D.B., C. Romano, and K.U. Hess, The effectof water
Advances in the Geology of Granite-Related Mineral
on the viscosity of a haplogranitic melt under P-T-X
Deposits,editedby R.P. Taylor and D.F. strong,Spec. Vol.
conditions relevant to silicic volcanism, Contrib. Mineral.
Petrol., 124, 19-28, 1996.
Dingwell, D.B., F. Holtz, and H. Behrens,The solubility of
H20 in peralkaline and peraluminousgranitic melts,Am.
Mineral., 82, 434-437, 1997.
Dingwell, D.B., K.U. Hess, and C. Romano,Extremelyfluid
behavior of hydrous peralkaline rhyolites, Earth Planet.
Can. Inst. Min. Metall., 39, 402-423, 1988.
Lange,R.A., The effectof H20, CO2 and F on the density and
viscosity of silicate melts, in Volatiles in Magmas, Rev.
Mineral., vol. 30, editedby M.R. Carrolland J.R. Holloway,
pp. 331-369, Mineral. Soc. of Am., Washington, D.C.,
1994.
Lange, R.A., and I.S.E. Carmichael,Thermodynamicproperties
Sci. Lett., 158, 31-38, 1998.
of silicate liquids with emphasis on density, thermal
Dunbar, N.W., and R.L. Herrig, Petrogenesis and volatile
expansion and compressibility, in Modern Methods in
stratigraphy of the Bishop Tuff: Evidence from melt
IgneousPetrology.' Understanding Magmatic Processes,
inclusion analysis, d. Geophys. Res., 97, 15129-15150,
Rev. Mineral., vol. 24, edited by J. Nicholls and J.K.
1992a.
Russell,pp. 25-64, Mineral. Soc.of Am., Washington, D.C.,
Dunbar, N.W., and R.L. Hervig, Volatile and trace elements
1990.
compositionof melt inclusions from the Lower Bandelier
Lejeune,
A., and P. Richet, Rheologyof crystal-bearingsilicate
Tuff: Implicationsfor magmachamberprocesses
anderuptive
melts: An experimental study at high viscosities, d.
style,d. Geophys.Res., 97, 15151-15170, 1992b.
Geophys.Res., 100, 4215-4229, 1995.
Dunbar, N.W., P.R. Kyle, and C.J.N. Wilson, Evidence for
limited zonation in silicic magmasystems,Taupo Volcanic Linnen, R.B., M. Pichavant,and F. Holtz, The combined effects
offO 2 and melt compositionon SnO2 solubility and tin
Zone New Zealand,Geology,17, 234-236, 1989a.
diffusivity in haplogranitic melts, Geochim. Cosmochim.
Dunbar, N.W., R.L. Herrig, and P.R. Kyle, Determination of
Acta, 60, 4965-4976, 1996.
pre-eruptiveH20, F, C1 contentsof silicic magmasusing
Lowenstern,
J.B., Evidencefor a copper-bearing
fluid in magma
melt inclusions:Examplesfrom Taupovolcaniccenter,New
erupted at the Valley of Ten Thousand Smokes,Alaska,
Zealand, Bull. Volcanol., 51, 177-184, 1989b.
Emerman,S.H., and R. Marret, Why dikes?, Geology, 18, 231233, 1990.
Contrib. Mineral. Petrol., i14, 409-421, 1993.
Lowenstern, J.B., Dissolved volatile concentrations in an ore-
formingmagma,Geology,22, 893-896, 1994.
Gardner, J.E., S. Carey, H. Sigurdsson, and M. Rutherford,
Influenceof magmacompositionon the eruptive activity of Luhr, J., Experimentalphase relations of water- and sulfursaturatedarc magmasand the 1982 eruptionsof E1 Chich6n
Mount St. Helens, Washington, Geology, 23, 523-526,
volcano, d. Petrol., 31, 1071-1114, 1990.
1995a.
Gardner, J.E., M. Rutherford, S. Carey, and H. Sigurdsson, Maaloe, S., and P.J. Wyllie, Water content of a granite magma
deduced from the sequenceof crystallization determined
Experimental constraints on pre-eruptive water contents
experimentally with water-undersaturated conditions,
and changingmagmastorageprior to explosiveeruptionsof
Contrib. Mineral. Petrol., 52, 175-191, 1975.
Mount St. Helens volcano, Bull. Volcanol., 57, 1-17,
1995b.
Mandeville,C.W., S. Carey,andH. Sigurdsson,
Magmamixing,
Gerlach,T.M., H.R. Westrich, and R.B. Symonds,Pre-eruption
fractionalcrystallizationand volatile degassingduring the
1883 eruptionof Krakatau volcano, Indonesia,d. Volcanol.
vaporin magmaof the climacticMount Pinatubo eruption:
Geotherm. Res., 74, 243-274, 1996.
Sourceof the giant stratosphericsulfur dioxide cloud, in
27,266
SCAILLET ET AL.: VISCOSITY OF SILICIC MAGMAS, 1
Marsh, B., On the crystallinity, probability of occurenceand
rheologyof lava and magma,Contrib. Mineral. Petrol., 78,
85-98, 1981.
Marsh, B., On the convective style and vigour in sheet-like
magmachambers,d. Petrol., 30, 479-530, 1989.
Martel, C., M. Pichavant, J.L. Bourdier, H. Traineau, F. Holtz,
and B. Scaillet, Magma storage and control of eruption
regime in silicic volcanoes:ExperimentalevidencefromMt.
Pe16e,Earth Planet. Sci. Lett., 156, 89-99, 1998.
Spera,F.J.,A. Borgia,J. Strimple,and M. Feigenson,Rheology
of melts and magmatic suspensions, 1, Design and
calibration of concentric cylinder viscometer with
application to rhyolitic magma,d. Geophys. Res., 93,
10273-10294,
1988.
Stein, D.J., and F.J. Spera, Rheology and microstructureof
magmaticemulsions:Theory and experiments,
d. Volcanol.
Geotherm. Res., 49, 157-174, 1992.
Stein, D.J., and F.J. Spera, Rheometry of a dacitic melt:
Experimental results and tests of empirical models for
viscosity estimation, Geophys. Res. Lett., 20, 1923-1926,
McBirney, A. R., and T. R. Murase, Rheological properties of
magmas,A. Rev. Earth Planet. Sci., 12, 1-10, 1984.
1993.
Newman, S., E.M. Stolper, and S. Epstein, Measurementof
water in rhyolitic glasses: Calibration of an infrared Stix, J., and M.P. Gorton, Variations in sanidine partition
coefficientsin the Cerro Toledo Rhyolite, JemezMountains,
spectroscopictechnique, Am. Mineral., 71, 1527-1541,
1986.
New Mexico: Effects of composition, temperature,and
volatiles, Geochim. Cosmochim. Acta, 54, 2697-2708,
Petford, N., R.C. Kerr, and J.R. Lister, Dike transport of
1990a.
granitoidmagmas,Geology,21, 845-848, 1993.
Stix,
J., and M.P. Gorton, Changesin silicic melt structure
Pichavant,M., D.J. Kontak, J.Valencia Herrera, and A.H. Clark,
betweenthe two Bandelier caldera-formingeruptions,New
The Miocene-Pliocene Macusani Volcanics, SE Peru, I;
Mexico, USA: Evidencefrom zirconiumand light rare earth
Mineralogy and magmatic evolution of a two-micas
elements, J. Petrol., 31, 1261-1283, 1990b.
aluminosilicate-bearing
ignimbrite suite, Contrib. Mineral.
Petrol., 100, 300-324, 1988a.
Stix, J., and G.D. Layne, Gas saturation and evolution of
volatile and light lithophile elements in the Bandelier
Pichavant,M., D.J. Kontak, L. Briqueu, J.ValenciaHerrera, and
magmachamberbetweentwo caldera-formingeruptions,d.
A.H. Clark, The miocene-Pliocene Macusani Volcanics, SE
Geophys.Res., 101, 25181-25196, 1996.
Peru, II; Geochemistryand origin of a felsic peraluminous
magma, Contrib. Mineral. Petrol., 100, 325-338, 1988b.
Wall, V.J., J.C. Clemens,and D.B. Clarke, Models of granitoid
evolution and sourcecompositions,d. Geol., 95, 731-749,
Pinkerton,H., and R.J. Stevenson,Methods of determiningthe
1987.
rheological properties of magrnas at sub-liquidus
temperatures,d. Volcanol.Geotherm.Res., 53, 47-66, 1992. Wallace, P.J., and T.M. Gerlach, Magmatic vapor source for
sulfur dioxide released during volcanic eruptions:
Rushmer,T., An experimentaldeformationstudy of partially
Evidence from Mount Pinatubo, Science, 265, 497-499,
molten amphibolite: Application to low-melt fraction
1994.
segregation,d. Geophys.Res., 100, 15681-15695, 1995.
Wallace,
P.J., A.T. Anderson, and A.M. Davis, Quantification
Rutherford, M.J., and J. Devine, Pre-eruption pressureof pre-eruptiveexsolvedgas contents in silicic magrnas,
temperatureconditions and volatiles in the 1991 dacitic
Nature, 377, 612-616, 1995.
magmaof Mount Pinatubo, in Fire and Mud.' Eruptions
and Lahars of Mount Pinatubo,editedby C. Newhall and Warshaw, C., and R.L. Smith, Pyroxenes and fayalites in the
R. Punongbayan, pp.751-766 , Univ. of Wash. Press,
Bandelier Tuff;New Mexico: Temperaturesand comparison
Seattle, 1996.
with other rhyolites,Am. Mineral., 73, 1025-1037, 1988.
Rutherford,M.J., H. Sigurdsson,S. Carey, and A. Davis, The Webb, S.L., and D.B. Dingwell, Non-newtonian rheology of
igneous melts at high stresses and strain rates:
May 18, 1980, eruption of Mount St. Helens, 1, Melt
Experimental results for rhyolite, andesite, basalt and
compositionand experimentalphaseequilibria,d. Geophys.
Res., 90, 2929-2947, 1985.
nephelinite,J. Geophys.Res., 95, 15695-15701, 1990.
Rutherford,M.H., J.D. Devine, andJ. Barclay, Changing magma Webster,J.D., J.R. Holloway, and R.L. Hervig, Phaseequilibria
conditions and ascent rates during the Soufriere Hills
of a Be, U and F-enrichedvitrophyre fromSpor Mountain,
Utah, Geochim. Cosmochim.Acta, 51, 389-402, 1987.
eruptionson Montserrat,GSA Today, 8, 1-7, 1998.
Scaillet, B., and B.W. Evans, The June 15, 1991 eruption of Webster, J.D., R.P. Taylor, and C. Bean, Pre-eruptive melt
compositionand constraintson degassingof a water-rich
Mount Pinatubo,I, Phase equilibria and pre-eruptionP-TfO2-fH20 conditions of the dacite magma,d. Petrol., in
pantelleritic magma,Fantale volcano, Ethiopia, Contrib.
Mineral. Petrol., 114, 53-62, 1993.
press, 1998.
Scaillet, B., M. Pichavant, and J. Roux, Experimental Westrich,H.R., andT. M. Gerlach,Magmaticgas sourcefor the
crystallizationof leucogranitemagmas,d. Petrol., 36, 664stratospheric
SO2 cloudfrom the June15, 1991, eruption of
706, 1995.
Mount Pinatubo,Geology, 20, 867-870, 1992.
Scaillet,B., F. Holtz, M. Pichavant, and M. Schmidt,Viscosity Westrich,H.R., J.C.Eichelberger,and R.L. Hervig, Degassing
of Himalayan leucogranites:Implicationsfor mechanismsof
ofthe 1912,Katmaimagmas,Geophys.Res. Lett., 18, 15611564, 1991.
graniticmagmaascent,d. Geophys.Res., 101, 27691-27699,
1996.
Whitney, J.A., The origin of granite: The role and sourceof
Scaillet, B., B. Clemente, B.W. Evans, and M. Pichavant, Redox
water in the evolution of granitic magmas,Geol. Soc. Am.
Bull., 100, 1886-1897, 1988.
control of sulfur degassingin silicic magmas,d. Geophys.
Res., in press, 1998.
Wolf, M. B. and P.J. Wyllie, Liquid segregationparameters
Schulze, F., H. Behrens, F. Holtz, J. Roux, and W. Johannes,The
from amphibolite dehydration melting experiments,d.
Geophys.Res., 100, 15611-15621, 1995.
influenceof water on the viscosityof a haplogranite liquid,
Am. Mineral. 81, 1155-1165, 1996.
Shaw, H.R., Viscosities of magmatic silicate liquid: An
empirical method of prediction, Am. d. Sci., 272, 870-893,
1972.
Sisson,T.W., and T.L. Grove, Experimentalinvestigations of
the role of H20 in calc-alkaline differentiation and
subduction zone magmatism,Contrib. Mineral. Petrol.,
F.Holtz,
Institut
filrMineralogie,
Welfengarten
1,Universitltt
Hannover,
W3000Hannover,
Germany.
M. Pithavant
andB. Scaillet,
CRSCM,
CNRS,1AruedelaFerollerie,
45071Orleans
cedex
02,France
([email protected])
113, 143-166, 1993.
Skirius, C.M., J.W. Peterson, and A.T. Anderson,
Homogenizingrhyolitic glass inclusions fromthe Bishop
Tuff, Am. Mineral., 75, 1381-1398, 1990.
(Received
January
26, 1998;revised
June8, 1998;
accepted
July20, 1998.)