The Role of Water in Generating the Calc-alkaline Trend

34 Pages·2010·1.08 MB·English

JOURNALOFPETROLOGY VOLUME0 NUMBER0 PAGES1^34 2010 doi:10.1093/petrology/egq062
Journal of Petrology Advance Access published November 18, 2010
The Role of Water in Generating the
Calc-alkaline Trend: NewVolatile Data for
Aleutian Magmas and a NewTholeiitic Index
MINDY M. ZIMMER1*,TERRY PLANK1y, ERIK H. HAURI2,
GENE M.YOGODZINSKI3, PETER STELLING4, JESSICA LARSEN5,
BRAD SINGER6, BRIANJICHA6, CHARLES MANDEVILLE7 AND
CHRISTOPHERJ. NYE8
D
o
1DEPARTMENTOFEARTHSCIENCES, BOSTONUNIVERSITY, BOSTON, MA 02215, USA w
n
2DEPARTMENTOFTERRESTRIALMAGNETISM, CARNEGIEINSTITUTIONOFWASHINGTON,WASHINGTON, DC 20015, loa
d
e
USA d
3DEPARTMENTOFGEOLOGICALSCIENCES, UNIVERSITYOFSOUTHCAROLINA, COLUMBIA, SC 29208, USA fro
m
4GEOLOGYDEPARTMENT,WESTERN WASHINGTONUNIVERSITY, BELLINGHAM,WA 98225, USA pe
5ALASKAVOLCANO OBSERVATORY, GEOPHYSICALINSTITUTE, UNIVERSITYOFALASKA FAIRBANKS, FAIRBANKS, trolo
g
AK 99775, USA y
.o
6GEOLOGYANDGEOPHYSICS, UNIVERSITYOFWISCONSIN, MADISON,WI 53706, USA xfo
rd
7DEPARTMENTOFEARTH ANDPLANETARYSCIENCES, AMERICANMUSEUMOFNATURALHISTORY, NEW YORK, jo
u
NY10024, USA rn
a
8ALASKAVOLCANO OBSERVATORY, ALASKA DIVISIONOFGEOLOGICAL ANDGEOPHYSICALSURVEYS, FAIRBANKS, ls.o
rg
AK 99775, USA b
y
g
u
e
s
RECEIVEDDECEMBER 9,2010; ACCEPTED SEPTEMBER16,2010 t o
n
N
o
v
e
m
b
e
Theoriginoftholeiitic(TH)versuscalc-alkaline(CA)magmatic in melt inclusions from eight volcanoesfrom theAleutian volcanic r 2
1
trends has long been debated. Part of the problem stems from the arc (Augustine, Emmons, Shishaldin, Akutan, Unalaska, Okmok, , 2
lackofaquantitativemeasureforthewayinwhichamagmaevolves. Seguam, and Korovin). Least degassed H2O contents vary from 010
RecognizingthatthesalientfeatureinmanyTH^CAdiscrimination (cid:3)2wt % (Shishaldin) to47wt % (Augustine), spanning the
diagrams is enrichment in Fe during magma evolution, we have globalrangeinarcmaficmagmas.WithintheAleutiandata,H O
2
developed a quantitative index of Fe enrichment, the Tholeiitic correlates negatively with THI, from strongly calc-alkaline
Index(THI):THI¼Fe /Fe ,whereFe istheaverageFeO* (Augustine, THI¼0·65) to moderately tholeiitic (Shishaldin,
4·0 8·0 4·0
concentration ofsamples with 4(cid:2)1wt % MgO, and Fe is the THI¼1·16).The relationship betweenTHIand magmatic water
8·0
average FeO* at 8(cid:2)1wt % MgO. Magmas withTHI41have ismaintainedwhendataareincludedfromadditionalarcvolcanoes,
enriched in FeO* during differentiation from basalts to andesites back-arc basins, ocean islands, and mid-ocean ridge basalts
and are tholeiitic; magmas withTHI51are calc-alkaline. Most (MORBs),supportingadominantroleofmagmaticwateringener-
subduction zone volcanism is CA, but to varyingextents; theTHI ating CA trends. An effective break betweenTH and CA trends
expresses the continuum of Fe enrichment observed in magmatic occursat (cid:3)2wt % H O. BothpMELTscalculationsand labora-
2
suitesinalltectonicsettings.Totestvariouscontrolsonthedevelop- toryexperimentsdemonstratethattheobservedco-variationofH O
2
mentofCAtrends, wepresent new magmatic water measurements and THI in arcs can be generated by the effect of H O on the
2
*Correspondingauthor.Presentaddress:ExxonMobilResearchand
Engineering Co., Annandale, NJ 08809, USA. Telephone: (908)
442-6602.Fax:[phone].E-mail:[email protected]
yPresent address: Lamont^Doherty Earth Observatory and (cid:2) The Author 2010. Published by Oxford University Press. All
Department of Earth and Environmental Sciences, Columbia rightsreserved.ForPermissions,pleasee-mail:journals.permissions@
University,Palisades,NY10964,USA. oxfordjournals.org
JOURNALOFPETROLOGY VOLUME0 NUMBER0 MONTH2010
suppressionofplagioclaseandtherelativeenhancementofFe-oxides the range of Fe enrichment observed in natural systems.
ontheliquidlineofdescent.ThefullTHI^H Oarrayrequiresan We redefine tholeiitic (TH) and calc-alkaline (CA)
2
increaseinfO withH O,from(cid:4)FMQ(whereFMQisthefaya- magma seriesbasedontheTHIanduse this indextotest
2 2
lite^magnetite^quartz buffer) in MORB to (cid:3)(cid:2)FMQ þ0·5 to modelsfortheoriginofCA fractionation.Asanexample,
þ2 in arcs, consistent with inferences from measured Fe and S we use the volcanoes of the Aleutian arc, which span
speciesinglassesandmeltinclusions.Acurvefittothedata,H2O the global range of subduction-relatedTHand CAevolu-
(wt %(cid:2)1·2)¼exp[(1·26 ^ THI)/0·32], may provide a useful tion and have long served as type examples for different
toolforestimatingtheH2Ocontentofmagmasthatareinaccessible evolutionary models (Fenner, 1926; Bowen, 1928; Kay
tomeltinclusionstudy. etal.,1982;Kay&Kay,1985,1994;Myersetal.,1985;Baker
& Eggler,1987; Gust & Perfit,1987; Brophy,1990; Miller
et al., 1992; Brophy et al., 1999; Kelemen et al., 2003;
KEYWORDS:subduction;water;calc-alkaline;tholeiitic;Aleutian George etal.,2004).Totestthe relationshipbetween mag-
maticwatercontentandTHI,wepresentthefirstextensive
studyof the watercontents in Aleutian magmas, focusing
INTRODUCTION
on melt inclusions in mafic tephra from eight Aleutian
The origin and evolution of magmas remains a central
volcanoes.We introduce Shishaldin and Augustine as end
D
question in igneous petrology. One of the earliest debates member TH and CA volcanoes, and then examine how ow
was spawned by Bowen’s classic theory (Bowen, 1928; n
variations in magmatic watercontent, pressure ofcrystal- lo
Young,1998), which emphasized silica enrichment as the a
lization, oxygen fugacity, and parental magma compos- d
dominant form of magma evolution. Fenner (1931) and ed
others countered with examples of magmas that experi- ition may lead to TH vs CA evolution. A strong fro
anti-correlation between measured magmatic water con- m
esinocne(iWroangeenrr&ichDmeeenr,t,19m3o9)s.tNnootwabitlyisthreecSokganeirzgeadatrhdaitnitrroun- cinenCtrAatiofrnasctaionndaTtiHonI,iamnpdlicmaateys pardovoimdeinaanptrreodleictoivfewatoteorl petro
enrichment during magmatic evolution is ubiquitous in lo
for estimating the water content of magmatic suites g
erivdegreymtecatgomniacesentrtiicnhgesoninEiarrotnh.aIsnditeeedv,olevveesr,yamndid-souccheana worldwide. y.oxfo
trendisreproducedwellbylow-pressurefractionalcrystal- rd
jo
lization of dry magma (Walker et al., 1979; Juster et al., u
rn
1989; Langmuir et al., 1992). Extremely iron-enriched, BACKGROUND als
end-stage liquids resulting from fractional crystallization .o
Terminology: tholeiitic vs calc-alkaline rg
havebeenobservedinlayeredmaficintrusionsfromdiffer- trends by
ent tectonic settings (Wager & Deer,1939; Sparks,1988; g
The term‘calc-alkaline’ has a long and complex history u
Wiebe, 1997). In contrast, subduction zones produce e
s
magmas with a range of evolutionary trends, including (see Arculus, 2003, for a comprehensive review of the t o
pronounced iron depletion (e.g. Fenner,1926; Chesner & etymology). Abrief summary is given here.Two volcanic n N
Rose, 1984; Kay & Kay, 1985; Carr et al., 2003; Grove rock series were defined by Harker (1909): the Pacific ove
Branch (richer in CaO and MgO) and the Atlantic m
etal.,2003). Iron depletion is also observed incontinental b
Branch (richer in alkalis). Holmes (1918) noted that the e
e1rt9if6atl9.m;,I2ab0ga0mr7ra)osala(nTdMruouacne‹oeatzn,a1li.9s,7la109n;9dT9h;biWarslaawnlatgsll(eeett.agal.l..,,M2200u00n‹20o;)z.KGraiernci|¤taz, atrten(cid:3)d6s2SwiOt2%^CaSOiOa2nfdorSi‘Oca2l^c-aallkkaalliis’,(PNaac2iOficþBKra2nOc)h-ctryopses r 21, 20
1
rocks. Peacock (1931) subsequently defined four subsets of 0
Theenrichment vsdepletionof iron in magmatic suites
is generally described as tholeiitic (TH) vs calc-alkaline magma using the SiO2 value at which CaO and alkali
(CA) differentiation, respectively. Use of these terms, trends intersected, including a‘calc-alkalic’group (inter-
however, is varied and vague, with different meanings section at 56^61wt % SiO2). The term ‘calc-alkaline’
to different workers (Arculus, 2003). Certainly, the term stems from this characterization, but has evolved sub-
‘calc-alkaline’ has become something of a colloquialism stantially away from it. In addition, the term ‘tholeiitic’
for ‘arc magma evolution’.The use of multiple discrimin- is not to be confused with ‘tholeiite’, referring to the
ationdiagrams(AFM, orAlkali^FeO*^MgO,andFeO*/ olivine- and hypersthene-normative rock defined on
MgO vs SiO, where FeO* denotes all Fe calculated as the basalt tetrahedron. The definition of TH and CA
2
FeO; Kuno,1968; Irvine & Baragar,1971; Miyashiro,1974) has evolved with the popular AFM diagram (Irvine
contributestotheambiguity.Inaddition,magmasdisplay & Baragar, 1971) and FeO*/MgO^SiO diagram
2
a continuum of Fe enrichment, so assigning a magma to (Miyashiro,1974), which emphasize Fe enrichment (TH)
oneoftwogroupsisarecognizedsimplification. or depletion (CA; Gill, 1981; Arculus, 2003). Here we
Here we develop a quantitative index based on Fe also consider Fe enrichment the fundamental feature of
enrichment, theTholeiitic Index (THI), which describes aTH magmatic suite.
2
ZIMMER etal. WATERANDCALC-ALKALINETREND
D
o
w
n
lo
a
d
e
d
fro
m
p
e
tro
lo
g
y
.o
x
fo
rd
jo
u
rn
a
ls
.o
rg
b
y
g
u
e
s
t o
n
N
o
v
e
m
b
e
r 2
1
, 2
0
1
0
Fig.1. Discriminationdiagramsfortholeiitic(TH)andcalc-alkaline(CA)magmaseries.(a)AFMdiagramwithdividinglinefromIrvine&
Baragar(1971;continuousgrayline)andvariousvolcanicsuites:Thingmuli,Iceland(Carmichael,1964);Shishaldin,Aleutianarc(P.Stelling
& C. J. Nye, unpublished data; Fournelle,1988; this study); Augustine, Aleutian arc (Daley,1986; Johnson etal.,1996; this study); Shasta,
Cascadia(includinghigh-magnesiumandesites;Bakeretal.,1994;Groveetal.,2001).(b)FeO*/MgOvsSiO discriminationdiagramwithdivid-
2
inglinefromMiyashiro(1974),withthesamesuitesasshownin(a)andhigh-Mg-numberandesitesfromtheSetouchivolcanicbelt,Japan
(Shirakietal.,1991),shownasagrayfield.
Existing TH and CA classification
FeO* apex before eventually enriching in the alkalis
schemes and evolving toward the alkali apex (e.g. Shishaldin
Variousgeochemicaldiagramshavebeendevelopedtode- volcano and Thingmuli, Fig. 1a). A CA magmatic suite
scribe TH vs CA evolution. On an Alkali^FeO*^MgO plotsonatrendthroughthemiddleoftheternary,toward
(AFM)diagram,aTHmagmaticsuiteevolvestowardthe the alkali apex, without showing a trend toward FeO*
3
JOURNALOFPETROLOGY VOLUME0 NUMBER0 MONTH2010
Fig.2. DatafromKlyuchevskoyvolcano(graycircles)illustratingdifferinginterpretationspossibleontwodiagrams.TheMgOvsFeO*dia- D
o
gram(a)illustratesthattheliquidlineofdescentprogressesatconstantFeO*,whereastheFeO*/MgO(b)ratioincreases.Thiscreatesthe w
falseimpressionofFeenrichmentontheFeO*/MgOvsSiO2diagram.DespiteconstantFeO*,FeO*/MgOincreasesandthemagmaevolves nlo
fromtheCAtotheTHfield.AfterArculus(2003).DatafromKersting&Arculus(1994). a
d
e
d
fro
m
(e.g. Augustine and Shasta volcanoes, Fig. 1a). Although FeO*/MgOofaliquidisrelatedtoFeO*/MgOofcoexist- p
e
theAFMdiagramiseffectiveinillustratingFeenrichment ingolivinebynearlyaconstantfactor(KD(cid:3)0·3;Roedder trolo
interms of FeO* and MgO components, the inclusion of &Emslie,1970;Fordetal.,1983).Thus,amagmainequilib- g
y
thealkalisisproblematic.Forexample,Augustinevolcano rium with the mantle ((cid:6)Fo olivine) will have FeO*/ .o
89 x
displays a CA trend (Fig. 1a), but its low alkali content MgO(cid:4)0·73. However, it is problematic to use FeO*/ fo
rd
(0·5wt % K O for Augustine vs 2wt % for Shishaldin) MgO to describe magmatic evolution because absolute jo
2 u
places the least evolved compositions well into the TH FeO*andMgOconcentrationsmayevolveindependently. rn
a
field. Gill (1981) noted that K O varies more than other Increasing FeO*/MgO may result from increasing FeO* ls
2 .o
major elements within andesites from different tectonic ordecreasingMgO,orfrommagmaswithnoabsoluteen- rg
b
settings, and such variations are due in part to different richment in FeO*. Figure2 showsanexample ofdecreas- y
g
mantle sources and different extents of partial melting. ing MgO at constant FeO* and the resulting increase in u
e
s
Suchprimary variations canoverwhelmmagmatic evolu- FeO*/MgO.Suchamagmaticseriesmightbeclassifiedas t o
tionary trends and lead to systematics on the AFM dia- TH, despite the fact that the magmas do not enrich in n N
gram that are unrelated to Fe enrichment. In addition, FeO*. Hence FeO*/MgO obscures the recognition of ov
e
theAFMdiagramdoesnotprovideaquantitativemeasure processesaffectingFeO*evolutionalongamagma’sliquid m
b
oftheextenttowhichmagmasareTHorCA.Thereisno lineofdescent. e
r 2
sciagnntilfyicaabnoceveasosrigbneeldowtothaemTaHgm^CatAicdsiuviitdeinpglotlitnineg, osrighnoifwi- diaAgnramaddisittihoenuaslecoofmSpiOle2xiotnytwheitahbstchiessaF.eSOil*ic/aMdgoOm^inSaiOte2s 1, 20
1
steep or shallow the trend is toward FeO*. Although themajorelementbudgetof mostmagmas, but variations 0
the AFM diagram shows Fe enrichment, the variation of in SiO may result from magma generation processes,
2
alkalis and the lack of quantifiable Fe enrichment make independentofhowamagmasubsequentlyevolves.Forex-
the diagram difficult to use to test models for the origin ample, melts of subducted oceanic crust can be andesitic
ofTHandCAtrends. (Tatsumi,1981; Rapp & Watson,1995), as canthereaction
There are similar ambiguities in the use of the FeO*/ products of basaltic melt and upper mantle (Kelemen,
MgO^SiO plot(Fig.1b)ofMiyashiro(1974).Itspopular- 1990; Grove et al., 2003), and both types of magma com-
2
ityresultsinpartfromthesimple,lineardivisionbetween monly originate in the CA field. Other processes that in-
TH and CA regions, for which Miyashiro provided an crease the primary SiO content of a magma include
2
equation: SiO (wt %)¼6·4(cid:5)(FeO*/MgO)þ42·8.This mantle meltingat low pressures (Kushiro,1972; Albare'de,
2
equation was admittedly arbitrary, serving to divide the 1992; Rogers et al.,1995; Lee et al., 2009), equilibration of
magmatic suites discussed in the original study (the melt with harzburgite (Parman & Grove, 2004; Grove
Skaergaard intrusion, Izu^Bonin,Tonga, Kermadec, and etal.,2005), high-pressure fractionation inthe presence of
NE Japan volcanic arcs). The FeO*/MgO ratio is useful garnet (Macpherson et al.,2006), and low-degree melting
when evaluating olivine^liquid equilibrium because of H O-bearing peridotite (Baker et al.,1995; Gaetani &
2
4
ZIMMER etal. WATERANDCALC-ALKALINETREND
Grove, 1998; Hirschmann et al., 1998). Such processes and most natural LLDs have initial FeO*^MgO slopes
are thought to generate several types of silicic primary consistent with olivine-dominated crystallization. Thus
magma, including adakites, high-Mg-number andesites, Fe approximates a magma’s primary FeO* concen-
8·0
high-Mg-number granitoids (sanukitoids), and boninites tration. The choice of FeO* at 4wt % MgO is more
(Kay,1978;Tatsumi & Ishizaka,1981; Kelemen etal.,1990; arbitrary, and reflects a compromise between an MgO
Yogodzinski et al., 1994; Falloon & Danyushevsky, 2000; value that is lowenoughto capture sufficient Fe variation
Martin et al., 2005), and most of these will originate in and one that is high enough to avoid the dominance of
the CA field, independentof how they evolve. Lastly, the factors such as crustal assimilation and magma mixing
silicacontentofanhydrousordegassedcompositionscom- that cause LLDs to converge.The advantage of theTHI
monlyplottedonsuchdiagramsmaybeartificiallyraised is that it isolates the evolution of Fe independently of
if the magma contains several wt % H2O, as a result of primary magma composition (Fe8·0) or other chemical
thedisproportionateeffectonsilicaofclosingmajorelem- components (e.g. silicaoralkalis).TheTHIquantifiesthe
ent analyses to 100%. Thus, although the FeO*/MgO^ mode of magmaevolution inthe early partof its path, as
SiO2 diagram maybe useful in identifying primarysilica magmas evolve from basaltic to low-silica andesitic
variations, this too complicates the isolation of Fe vari- compositions.
ations,andobscuresthedistinctionbetweenmagmagener- There are a number of limitationsto such anapproach
D
ationandevolution. Itisalso difficulttodiscernprocesses to quantify a magmatic LLD. Most of the data inthe lit- ow
affecting Fe enrichment in primitive compositions using n
erature are whole-rock analyses, but a whole-rock is not lo
FeO*/MgO^SiO because changes in absolute concentra- a
2 necessarily a good approximation to a magmatic liquid d
e
tinioSnisOof2.FeO*andMgOcanoccuroververysmallchanges a(ec.cgu.mEuiclhateilobne,rgpearrteitaalls.,eg2r0e0g6a)t.ioPnro,caensdsefsilitnecrlupdreinssgincgrymstaayl d from
Therehavebeensomeattemptstoquantifytheextentto result in divergence of whole-rockcompositions fromtrue p
e
which a magma isTHor CAonthe Miyashiro diagram. liquids (Sisson & Bacon, 1999). Not only have we used tro
Arculus (2003) concluded a recent review of the issues of lo
whole-rocksamplestoapproximatetheLLD,butwehave g
‘CA’nomenclature with a suggestion of delineating fields usually combined all data from a single volcano. Ideally, y.o
x
oSfiOhi.ghM-,omreedreiucmen-t,lya,nHdolroaw-eFteaol.n(2a0p09lo)tporfoFpeoOse*d/MagnOew^ theTHI wouldbecalculatedforliquids fromsingleerup- ford
2 tive episodes, but in mostcases, too fewof such dataexist jo
CA/TH index based on the slope on this diagram. to calculate a meaningfulTHI. Future work may exploit urn
ArelmthaoiungwhitshucinhteqruparenttiinfigcatthieonFeiOs*a/nMigmOprroavtieomaenndt,diescsounes- improvements in the volcanic datasets as more data als.o
becomeavailableonmagmaticliquids(e.g.meltinclusions rg
volvingprimarysilicavariations. b
and aphyric rocks), for multiple eruptions from a single y
g
volcano.Despitethelimitationsofthecurrentglobaldata- u
e
s
THE THOLEIITIC INDEX (THI) set in these ways, there are still large variations inTHI t o
Our goal here is to develop a simple tool for quantifying toexamineregionallyandglobally. n N
theextentof Fe enrichmentthat amagmatic suiteexperi- One advantage of theTHI is that it describes the con- ov
e
ences during crystallization. Many of the previous efforts tinuum of Fe enrichment observed in natural samples, m
b
describedabovefocusedonwheremagmasplotwithincer- eliminating the need for further terminology. However, er 2
tinainhodwiscmriamgimnaatsinegvofilevledsa,nwdhheroewastowequaarentmifyortehientcerryessttaeld- wemebreedcdoegdniizneththealtitetrhaetutreer;mthsu‘sTTHH’ aanndd‘CCAA’aarreeddeefienpeldy 1, 20
1
lization paths they take. Toward this end, we propose a here based on theTHI. Magma suites that are tholeiitic 0
new index of Fe enrichment, theTholeiitic Index (THI). enrichinFeandthushaveTHI41;magmasuitesthatare
TheTHI is based on absolute concentrations of FeO* vs calc-alkaline deplete in Fe and have THI(cid:4)1. Although
MgO: theTHIdoesnot include Bowen’soriginalnotionsregard-
ing silica enrichment, nor subsequent inferencesbased on
THI¼Fe =Fe ð1Þ
4(cid:7)0 8(cid:7)0 therateofsilicaenrichment,Fe encompassesarangeof
4·0
where Fe4·0 is the average FeO* concentration of samples compositions, from 50wt % SiO2 (basalt) to 56wt %
with 4(cid:2)1wt % MgO, and Fe8·0 is the average FeO* SiO2(low-silicaandesite),whichincludesamajorfraction
concentration at 8(cid:2)1wt % MgO along a given liquid ofthemagmaseruptedonEarth.
line of descent (LLD). The THI is simply the factor of We illustrate the utility of the THI by examining its
enrichment or depletion in FeO* as a magma evolves variation in several magmatic suites from different tec-
from parental compositions at 8% MgOto more fractio- tonic settings, including 41 magmatic suites from arcs,
nated compositions at 4% MgO. At (cid:6)8wt % MgO, back-arc basins, mid-ocean ridges (mid-ocean ridge
olivinecrystallizationfrombasalticliquiddrivesrelatively basalt; MORB), ocean islands (ocean islandbasalt; OIB),
little variation in FeO* (Langmuir & Hanson, 1980), and the alkaline volcanoes Etna and Vesuvius (Table 1).
5
JOURNALOFPETROLOGY VOLUME0 NUMBER0 MONTH2010
Table1: Tholeiiticindex(THI)andaveragewatercontent(H O)fordatashowninFig.71
2
Abbrev. Volcano Fe4·0 (cid:2)2s Fe8·0 or (cid:2)2s THI (cid:2)2s H2Oave (cid:2)1s.d. No. Tectonic
SE max2 SE SE (wt%) MI3 setting4
AU Augustine 6·25 0·24 9·62 * 0·48 0·65 0·041 6·35 0·25 19/59 S
EM Emmons 8·13 0·14 8·40 * 0·42 0·97 0·051 2·40 0·13 5/111 S
SH Shishaldin 10·98 0·24 9·50 0·26 1·16 0·040 2·03 0·18 10/33 S
AK Akutan 9·44 0·40 10·07 1·02 0·94 0·103 3·72 0·09 2/39 S
UN Unalaska 8·21 0·51 9·20 0·38 0·89 0·066 3·03 0·42 4/26 S
OK Okmok 9·96 0·48 11·10 0·44 0·90 0·056 2·47 0·15 4/35 S
SE Seguam 7·30 0·38 8·40 0·50 0·87 0·069 3·30 0·33 14/29 S
KO Korovin 8·02 0·22 10·19 0·41 0·79 0·038 3·74 0·63 5/14 S
KL Klyuchevskoy 8·16 0·24 8·39 0·07 0·97 0·030 2·68 0·23 5/29 S
TO Tolbachik 8·84 0·33 9·10 0·48 0·97 0·063 2·69 0·18 8/11 S
KS Ksudach 9·08 0·57 9·55 * 0·95 0·95 0·112 3·35 0·19 4/4 S
PA Parcutin 6·84 0·30 8·46 * 0·85 0·81 0·088 3·29 0·66 5/26 S D
o
FG Fuego 8·65 0·11 11·01 0·51 0·79 0·038 5·24 0·72 5/28 S w
n
CN CerroNegro 9·68 0·21 10·61 0·20 0·91 0·026 5·08 0·72 17/56 S lo
a
AR Arenal 7·22 0·12 9·19 * 0·92 0·79 0·080 3·43 0·48 7/19 S d
e
AIZG AIragzrui´gan 106··0709 00··7118 118··6064 * 10··1173 00··8864 00··100256 42··5696 00··5556 34//811 SS d from
SA Sarigan 7·24 0·41 9·16 0·57 0·79 0·067 5·61 0·74 2/4 S p
e
ST Shasta 4·35 0·14 7·08 0·50 0·61 0·048 8·005 0·80 9/15 S tro
SI SatsumaIwojima 8·97 0·59 9·52 * 0·95 0·94 0·113 2·45 0·31 6/21 S lo
g
SR Stromboli 7·18 0·13 7·72 0·22 0·93 0·031 2·83 0·36 22/23 S y.o
ET Etna 7·99 0·16 9·58 0·12 0·83 0·020 2·63 0·66 74/101 I xfo
VE Vesuvius 6·61 0·54 7·30 0·38 0·91 0·088 1·97 0·34 18/44 I rd
MT MarianaTrough 10·95 0·59 7·82 0·17 1·40 0·081 1·36 0·42 44/66 B jou
LB LauBasin6 11·24 0·25 9·03 0·21 1·24 0·040 1·18 0·73 24/56 B rn
a
CL LauBasin–CLSC7 16·44 0·21 9·14 0·32 1·80 0·067 0·52 0·27 6/24 B ls.o
MB ManusBasin 12·74 1·04 10·15 0·59 1·26 0·125 0·74 0·39 27/52 B rg
SC E.ScotiaRidge 10·15 0·34 8·17 0·22 1·24 0·054 1·05 0·50 6/73 B by
WB WoodlarkBasin 12·52 1·01 8·94 0·33 1·40 0·124 0·32 0·05 14/16 B gu
e
FJ N.FijiBasin 10·98 0·72 9·57 0·12 1·15 0·077 0·81 0·54 14/19 B s
GL Galapagos 16·08 0·72 10·04 0·17 1·60 0·077 0·31 0·21 35/42 M t o
n
ER EastPacificRise 15·68 0·66 9·89 0·27 1·59 0·079 0·25 0·08 26/26 M N
o
IR SEIndianRidge 13·19 0·10 8·58 0·31 1·54 0·057 0 – M v
e
SM Samoa 12·48 0·26 11·40 0·29 1·09 0·036 1·24 0·12 40/94 O m
b
HREK HReeuknlaion 1132··9314 01··7312 1121··5580 00··6150 11··1017 00··126311 10··0776 00··9102 244//429 OO er 21
HKi Hawaii:Kilauea 13·00 0·07 11·22 0·06 1·16 0·009 0·408 0·04 O , 2
0
HMk Hawaii:MaunaKea 10·61 0·10 11·52 0·09 0·92 0·011 0·368 0·04 O 10
HMl Hawaii:MaunaLoa 14·96 0·37 10·80 0·06 1·39 0·035 0·368 0·04 O
Hko Hawaii:Koolau 11·48 1·08 10·56 0·09 1·09 0·102 0·408 0·04 O
HLo Hawaii:Loihi 12·35 0·37 11·44 0·20 1·08 0·038 0·488 0·05 O
TH Thingmuli 13·81 0·66 11·51 * 1·15 1·20 0·133 – – – LM
SK Skaergaard 21·64 2·24 15·09 3·00 1·43 0·321 – – – LM
1THI¼Fe /Fe . Fe is the average FeO* for compositions with 3–5wt % MgO; data are binned in 0·5wt % MgO
4·0 8·0 4·0
intervals, andthe averageofthese four binsistakenasFe (cid:2)2sstandard errorof themean. Fe ¼averageFeO*from
4·0 8·0
7 to 9wt % MgO, calculated in the same manner as Fe . THI error propagated from the 2s standard deviation of the
4·0
mean of Fe and Fe . (See Electronic Appendix A for whole-rock and melt inclusion or glass references, and details
4·0 8·0
on calculations for each suite.)
2For volcanoes where Fe is constrained by one value (or max), error on Fe is assumed to be (cid:2)10%.
8·0 8·0
3First value: number of melt inclusions or glasses used to calculate H O (wt %); second value: total number of melt
2
inclusions or glasses evaluated. MI that are degassed or have compositions falling off the volcano LLD were excluded
from THI and H O calculation (see text for details).
2
4S,subductionzone;M,mid-oceanridge;B,back-arcbasin;O,oceanisland;I,intraplatevolcano;LM,layeredmaficintrusion.
5Water contents for Shasta derived from experimental data only (Grove et al., 2002). Error assumed to be (cid:2)10%.
6Includes data from Eastern Lau Spreading Center (LSC), Intermediate LSC, Mangatolu Triple Junction, Valu Fa.
7Includes data from Central LSC only.
8H O (wt %) from estimated primary water contents based on glass or melt inclusion data (see text): Hauri (2002):
2
Kilauea, Mauna Kea, Mauna Loa, Koolau, Loihi; Davis et al. (2003): Kilauea, Mauna Loa, Loihi; Seaman et al. (2004):
Mauna Kea. Error assumed to be (cid:2)10%.
6
ZIMMER etal. WATERANDCALC-ALKALINETREND
D
o
w
n
lo
a
d
e
d
fro
m
Fig.3. CalculationoftheTHI(TholeiiticIndex)¼Fe /Fe .Fe istheaverageFeO*concentrationofsamplesbetween3and5wt%MgO. p
4·0 8·0 4·0 e
Iftherearesufficientdata(asforShishaldin),Fe8·0istheaverageFeO*concentrationofsamplesbetween7and9wt%MgO;ifnot,then tro
stthreonsagmlypclaelcw-aitlhkathlienehtigrehnedst(MTHgOI5is1)u.sedtoapproximateFe8·0(asforAugustine).Shishaldinhasatholeiitictrend(THI41);Augustinehasa logy
.o
x
fo
rd
We calculate theTHI using mostly whole-rock and some volcanic suites with sparse data46wt % MgO, orby re- jou
rn
melt inclusion data from the literature and this study, gressing the data to calculate Fe4·0 and Fe8·0. For the als
making substantial use of the GEOROC database Aleutians, these other methods resulted in very little .o
(http://georoc.mpch-mainz.gwdg.de/georoc/) and PetDB change in theTHI, with the differences generally falling brg
(http://www.petdb.org/petdbWeb/index.jsp). Details on within the reported error of the THI. Thus we have y g
u
thedatatypeandreferencesforeachvolcanicsuitecanbe chosen to calculate THI using binned averages for the e
s
foundin Electronic Appendix A (available at http://www Aleutians and all other suites examined here. Provided t o
n
.petrology.oxfordjournals.org). When abundant analyses the treatment of the data and error propagation are N
o
areavailable,thedataarebinnedin0·5wt%MgOinter- robust, however, such alternative methods of calculating v
e
vals, and the four bins between 7 and 9wt % MgO are Fe andFe areacceptable. m
4·0 8·0 b
e
averaged to obtain Fe8·0. A similar procedure is used to As an example of the efficacy of theTHI, Fig. 3 illus- r 2
obtainFe .UncertaintiesinTHIarecalculatedbypropa- trates MgO and FeO* variations in Shishaldin and 1
4·0 , 2
gating the 2s standard error of the mean for Fe4·0 and Augustine volcanoes in the Aleutians, which span the 01
Fe onto their ratio (Table 1). Unfortunately, many arc global array of arc TH and CA evolutionary trends 0
8·0
volcanoes in particular do not erupt magmas with47wt (Table1). Shishaldin volcano has the highestTHI of 1·16,
%MgO,orthosewith47wt%MgOmayreflectolivine indicating that FeO* has enrichedby16% from 8 to 4%
and/or clinopyroxene accumulation. In these cases, the MgO. TheTHI clearly underestimates Fe enrichment in
FeO* content of the samples with the highest MgO was thiscase,whichisashighas1·4at(cid:3)5%MgO,whenmag-
used to estimate Fe , provided a suite had at least one netite fractionation starts to affect the LLD. Although it
8·0
samplewith(cid:6)6wt% MgO.Wealso discoveredmanyex- would be preferable to account for such features in each
amplesofreportingerrors,whereFe O hasbeenreported LLD, such inferences become difficult for suites with few
2 3
as FeO, or vice versa. Such issues are pervasive in the data, and sowe have chosenaconsistent measure suchas
global dataset, andwe have contacted original authors in Fe .Despitetheseissues,theTHIforShishaldinclassifies
4·0
somecasestorectifytheissuebutsuspectthatmanycases it as one of the most tholeiitic arc suites, consistent with
have gone undetected. These issues introduce additional its placement on the traditional AFM and FeO*/MgO^
errors in the estimation of THI from the global dataset. SiO diagrams (Fig. 1). At the other extreme, Augustine
2
In addition, other methods of calculating Fe and Fe volcano has a THI of 0·65, indicating 35% depletion
4·0 8·0
were also examined, including extrapolating to Fe for in FeO* at 4% MgO. Augustine is thus strongly CA,
8·0
7
JOURNALOFPETROLOGY VOLUME0 NUMBER0 MONTH2010
D
o
w
n
lo
a
d
e
d
fro
m
p
e
tro
lo
g
y
.o
x
fo
rd
jo
u
rn
a
ls
.o
rg
b
y
g
u
e
s
t o
n
N
o
Fig.4. HistogramsshowingtherangeinTHIformagmaticsuitesfromvarioustectonicsettings.MostdataareinTable1(seereferencesin v
e
ElectronicAppendixA).MORB,mid-oceanridgebasalt;BABB,back-arcbasinbasalt;OIB,oceanislandbasalt. m
b
e
r 2
1
consistent with where most of the data plot on the trad- between magmas from different tectonic settings, withall , 20
1
itional diagrams. Thus, THI discriminates strongly be- MORB andback-arc volcanic suitesbeingTH, and some 0
tween these two volcanoes of the Aleutian arc and OIB and most arc suitesbeing strongly CA.TH magmas
provides a quantitative measure of their Fe evolutionthat are found in all tectonic settings, but most arc volcanoes
canbe tested against intensive variables (e.g. water, pres- are CA.TheTHIthus conforms to the common notions
sure, composition, fO) andpredictions frompetrological andusageofthetermstholeiiticandcalc-alkaline.
2
models. In summary, theTHIquantitatively describes the con-
Figure 4 compares the THI for volcanic suites from tinuum of Fe enrichment in magmas (Miyashiro, 1974)
differenttectonicsettings.FewMORBsuitesevolveto4% andimprovesuponothertermssuchas‘transitional’(Kay
MgO, but those that do have a highTHI (41·4), and are & Kay,1985) and‘high-, medium-, and low-Fe’ (Arculus,
strongly tholeiitic. Eight back-arc spreading segments in- 2003). The THI effectively discriminates some of the
clude compositions with 4% MgO; some overlap in their majormagmaseriesonaglobalbasis,notjustwithinsub-
THI with MORB, but most extend to lower values duction zones, with THI decreasing on average from
(THI¼1·14^1·80). Ocean island and volcanic arc suites MORBs toback-arcbasinbasalts to OIB to arc magmas.
show the largest variations inTHIand range fromTH to Most importantly, the THI allows quantitative tests of
stronglyCAtrends.Thus,theTHIdiscriminateseffectively hypothesesfortheoriginofTHandCAmagmaseries.
8
ZIMMER etal. WATERANDCALC-ALKALINETREND
D
o
w
n
lo
a
d
e
d
Fig.5. Alaska^Aleutianvolcanicarc,resultingfromconvergencebetweenthePacificandNorthAmericanplates.Volcanoesshownastriangles; from
blacktrianglesdenotevolcanoeswithmeltinclusiondatareportedinthisstudy.AU,Augustine;EM,Emmons;SH,Shishaldin;AK,Akutan; p
UN,UnalaskaIsland(includingPakushinandAngela’sCone);OK,Okmok;SE,Seguam;KO,Korovin. etro
lo
g
y
.o
x
Generating calc-alkaline fractionation (Osborn, 1959; Gill, 1981; Sisson & Grove, 1993a, 1993b), fo
rd
trends and amphibole, which requires high H O and Na O jo
2 2 u
Thebenefitofdevelopinganew indexofFeenrichmentis (Cawthorn & O’Hara, 1976; Grove et al., 2003). rna
not only to improve upon the ambiguities with current Assimilation and mixing with low-Fe silicic melts (Grove ls.o
classification schemes for igneous rocks, but to shed light & Baker,1984; McBirney et al.,1987; George et al., 2004) rg
b
ontheirorigins.Tholeiitic magma series, asobservedubi- or mantle^melt reaction (Kelemen,1990) may also create y g
quitously at mid-ocean ridges, are well reproduced with aCAtrend. ue
s
bothlaboratoryandthermodynamicsimulationsofcrystal Of each of these controls, magmatic water content has t o
fractionation of dry magma at low pressure (Grove & long figured prominently in the theories for CA evolu- n N
Bryan,1983; Langmuiretal.,1992). Drybasaltic liquids at tion because it can drive CA differentiation in three in- ov
e
low pressure generally crystallize olivine first, which has dependent ways: by suppressing plagioclase, enhancing m
b
little effect on FeO*. Plagioclase is typically the next magnetite, and permitting amphibole crystallization.Yet er 2
phase on the cotectic, which both retards the rate of until recently, measurements of pre-eruptive water con- 1
Mg depletion and accelerates the rate of Fe enrichment tents have been scarce and inconsistent, largely because , 20
1
per incrementcrystallized.The appearance of plagioclase all volcanic rocks degas as they reach the surface where 0
typicallycoincides withthe onsetof strong Fe enrichment we sample them. Recent developments in microanalysis
andtholeiiticdifferentiation. [Fourier transform infrared spectroscopy (FTIR) and
Conversely,therearemanywaystogenerateaCAtrend secondary ionization mass spectrometry (SIMS)] and
ofFedepletion.Theseinvolve,logically,thesuppressionof successful applicationto quenched glasses and melt inclu-
plagioclasecrystallizationasaresultof(1)highwatercon- sions have produced a rapidly expanding database of
tent (Grove & Baker,1984; Baker & Eggler,1987; Kinzler water measurements in magmas from subduction zones,
& Grove,1992; Sisson & Grove,1993a; Grove etal.,2003), mid-ocean ridges, back-arc basins, and ocean islands.
(2) high pressure of crystallization (Baker & Eggler,1987; For this reason, we focus this study on the role of water
Gust & Perfit,1987; Kinzler & Grove,1992), or (3) bulk in generating the CA trend, and we present new data
composition (variations in CaO/Al O, Na O/Al O, and for the water contents of magmas from Aleutian volca-
2 3 2 2 3
FeO*; Grove & Baker, 1984; Miller et al., 1992). Other noes that possess large variations in THI (Table 1,
long-standing theories for generating CA trends involve Fig. 5). After presentation of the new data, we return to
theearlycrystallizationofFe-richphases,suchasmagnet- assess the quantitative role of water in driving differenti-
ite, which requires high oxygen fugacity and/or H O ation in arc magmas.
2
9
JOURNALOFPETROLOGY VOLUME0 NUMBER0 MONTH2010
SAMPLES AND METHODS for whole-rocks (Electronic Appendix D; http://www
.petrology.oxfordjournals.org); trace element data will be
Melt inclusion(MI)datawerecollectedfromeight volca-
presentedelsewhere.
noes in the Aleutian^Alaskan volcanic arc: Augustine,
Emmons, Shishaldin, Akutan, Makushin, Okmok, Melt inclusions: major elements and
Seguam, and Korovin (Fig. 5). Great effort was made to
volatiles
obtain fresh, small-diameter mafic tephra (52cm lapilli),
To extract MI-bearing phenocrysts, tephra samples were
asthismaterial hascooledrapidlyandpreservestheglas-
crushed with an alumina jaw crusher or steel percussion
siest MIs withthehighest volatile contents. Samples from
mortar and sieved. Olivine-, plagioclase-, and pyroxene-
PakushinandAngela’sCone(flankconesofMakushinvol-
hostedMIswerehand-pickedfromsievedaliquotsranging
cano on Unalaska Island), Akutan volcano, and Pyre
in size from 150 to 1000mm.We were highly selective in
Peak of Seguam Island were collected during dedicated
picking MIs for analysis, selecting only glassy inclusions
field campaigns in 2005 and 2007, with sample numbers
lacking cracks, breaches, daughter crystals or devitrifica-
prefaced by 05MZPK-, 05EHPK-, 05AKTAP-, and
tion textures, and avoiding inclusions in contact with
SEG-07- (see further sample details and locations in
theexteriorofthehostphenocryst.Wedidnotheat,hom-
Electronic Appendices B and C; http://www.petrology
ogenize, or treat the MIs in any way, except for some
.oxfordjournals.org). Many other samples were donated D
phenocrystsfromAngela’sCone.Asaresultofextremeal- o
generously by the Alaska Volcano Observatory and the w
teration, the exteriorof olivine phenocrysts from Angela’s n
Plate Boundary Observatory, as well asby scientists from lo
Cone had an orange tint. After observing contamination a
oAfetrlhtrheerodu,ignshastmitsupatlmeiospnalserse(eafsrtpoemmcioashltliysHtoAor.liocAcaelnndeeerriunspotnaigoaenns(dsweeRer.beeKlopawrye))-.. aonftdheHM2IOsurinfaccreeadsuerinwgitShIMcoSnatninaulyesdis(sep.ug.ttCerOin2gd),ecrsoeamsee ded from
phenocrysts (05MZPK-09_oliv5-10) were soaked in a 2N p
We sought olivine-hosted MIs from mafic scoria because e
these MIs are the most relevant for the part of the LLD HCl bath for 20minto prevent the altered material from tro
smearing onto the exposed surface during polishing. lo
raenleveaanrtlytoathnedTHmIaj(oir.e.p8htaose4.%SoMmgeO)p,lwaghieorcelaoslei-vinaenids Aspluthttoeurignhgc(soenetabmeloinwa)t,iothnewleansgtgheonfertaimllyefaovrotihdeedconbtyamprine-- gy.oxfo
pyroxene-hosted MIs were also analyzed in samples ran- ation signalto disappear was significantly reduced inthe rd
gingfrombasalttoandesiteincomposition. MIs treated with HCl. Melt inclusions and host pheno- journ
crysts from each volcano were polished and mounted in a
Whole-rocks: major elements indium metal, which minimizes background volatile ls.o
rg
Bulk-rock samples were analyzed for major elements at countsintheionprobe(Haurietal.,2002).Meltinclusions b
y
Boston University (BU) and Michigan State University. from Emmons volcano were either mounted in indium g
u
At BU, samples were fused following the techniques of metal or mounted in holes drilled into a brass ring and e
s
Kelley et al. (2003) andWade et al. (2006). Samples were back-filledwithepoxy. t o
n
crushed in an alumina jaw-crusher, hand-picked to avoid ThevolatilesH O,CO ,S,Cl,andFwereanalyzedby N
2 2 o
weathered surfaces, sonicated, driedat51008C, and pow- SIMS (Electronic Appendix E; http://www.petrology ve
m
deredinanaluminaballmill.Formajorelements,powders .oxfordjournals.org) at the Carnegie Institution of b
e
werefusedwithLiBO2thendissolvedinnitricacid,andfi- Washingtonwith a Cameca 6f ion microprobe during six r 2
nally diluted to (cid:3)4300(cid:5)dilution of the original sample. sessions. Inclusions were analyzed following the protocol 1, 2
The solutions were analyzed for 10 major elements and of Hauri et al. (2002), using an 8nAbeam current,30mm 0
1
seven trace elements by inductively coupled plasma spot size, and 5keVaccelerating voltage. On-peak count 0
atomic emission spectrometry (ICP-AES) using a Jobin- times were (cid:3)5 s. Prior to each analytical session, glass
Yvon170C system.Whole-rock samples from Okmok and standards with known volatile concentrations were ana-
Seguam were analyzed by X-ray fluorescence (XRF) at lyzed to generate calibration curves, with no explicit cor-
MichiganStateUniversity followingtheprotocolofVogel rection for Si. Water was measured as 1H or 1H16O.
etal. (2006). Glass disks were fusedusing lithiumtetrabo- Precision during the sessions was 5% 1s for H O and
2
rate flux and ammonium nitrate to enhance oxidation CO, and 6% 1sfor F, S, and Clbasedon replicateana-
2
during fusion. Fused disks were then analyzed for major lysesofourmeltinclusionsandstandardglasses.Precision
elements and three trace elements (Rb, Sr, Zr) on a betweenmultiplesessionswas(cid:3)10%forallvolatilespecies
Bruker S4 Pioneer XRF system. Datawerereducedusing (Cooper,2009).
fundamental parameters in SPECTRAplus software Major element, S, and Cl concentrations were deter-
(BrukerAXS,Germany)ontheS4Pioneer;traceelements mined by electron microprobe analysis (EPMA) on a
weredeterminedusinglinearregressionofexternalstand- JEOLJXA-733 Superprobe atthe Massachusetts Institute
ards. Precision for most elements is 51% RSD, except of Technology (MIT) or on a Cameca SX100 EPMA at
for P O (52%). Only major element data are reported the American Museum of Natural History (AMNH;
2 5
10