Mallik, A. & R. Dasgupta, Reactive Infiltration of MORB-Eclogite-Derived Carbonated Silicate Melt

34 Pages·2013·2.66 MB·English

JOURNALOFPETROLOGY VOLUME54 NUMBER11 PAGES2267^2300 2013 doi:10.1093/petrology/egt047
Reactive Infiltration of MORB-Eclogite-Derived
Carbonated Silicate Melt into Fertile Peridotite at
3 GPa and Genesis of Alkalic Magmas
ANANYA MALLIK1* AND RAJDEEP DASGUPTA1
D
o
w
1RICEUNIVERSITY, DEPARTMENTOFEARTHSCIENCE,6100 MAINSTREET, MS126, HOUSTON,TX 77005, USA nlo
a
d
e
d
fro
RECEIVEDJANUARY 20,2013; ACCEPTEDAUGUST 8,2013 m
h
ADVANCE ACCESS PUBLICATIONSEPTEMBER10,2013 ttp
://p
e
tro
lo
Weperformed experimentsbetween twodifferent carbonated eclog- 415wt %) natural nephelinites in termsof SiO, Al O, FeO*, g
2 2 3 y
ite-derivedmeltsandlherzoliteat13758Cand3GPabyvaryingthe CaO,Na OandCaO/Al O.Notonlycanthesereactedmeltserupt .o
2 2 3 x
reactingmeltfractionfrom8to50wt%.Thetwostartingmeltcom- by themselves, they can also act as metasomatizing agents in the ford
positions were (1) alkalic basalt with 11·7wt % dissolved CO Earth’smantle.Ourstudysuggeststhatacombinationofsubducted, jo
2 u
(ABC), (2) basaltic andesite with 2·6wt % dissolved CO silica-saturated crust^peridotite interaction and the presence of rn
(tiBteAtCo)s.iTmhuelasttearptoinrogumserletascwtivereeimnfiixletdrahtioomnoogfenmeoeultsliynwtihtehpEearridtho’s-2 CraOng2eionftphreimmiatinvtelealskoaulriccebraegsaiolnts.aArelssou,fmfiacinetnltetpootpernotdiaulceteamplaerrag-e aals.org/
mantle.Alltheexperimentsproducedanassemblageofmeltþortho- tures of 1330^13508C appear sufficient to produce high-MgO, t U
pyroxeneþclinopyroxeneþgarnet(cid:2)olivine;olivinewasabsentfor primitive basanite^nephelinite if carbonated eclogite melt and niv
e
a reacting melt fraction of 50wt % forABC and 40wt % for peridotiteinteractionistakenintoaccount. rs
BAC. Basanitic ABC evolved to melilitites (on a CO2-freebasis, ita' d
SiO (cid:3)27^39wt%,TiO (cid:3)2·8^6·3wt%,Al O (cid:3)4·1^9·1wt eg
%,2FeO* (cid:3)11^16wt %,2MgO (cid:3)17^21wt%,2Ca3O (cid:3)13^21wt KEYWORDS:alkalicbasalts;carbonatedsilicatemelt;MORB-eclog- li Stu
%,Na2O(cid:3)4^7wt%,CO2(cid:3)10^25wt%)uponmelt^rockreac- ite;peridotite;reactiveinfiltration di R
tion and the degree of alkalinity of the reacted melts is positively o
m
correlated with melt^rock ratio. On the other hand, reacted melts a
L
dTeirOived (cid:3)fro6m·4^B8A·7Cw(ton%a,COA2l-Ofree b(cid:3)a1s0is·5^S1iO2·23w(cid:3)t42^%53,wFte%O*, INTRODUCTION a Sap
(cid:3)6·52^10·5wt %, MgO (cid:3)7·92^135·4wt %, CaO (cid:3)7·3^10·3wt Intraplate oceanic basalts are one of the key tools to de- ien
z
%,Na O(cid:3)3·4^4wt%,CO (cid:3)6·2^11·7wt%)increaseinalka- ciphermantleprocesses.Thevariabilitydisplayedintheir a o
2 2 n
linity with decreasing melt^rock ratio.We demonstrate that owing trace element and isotopic characteristics indicates the N
o
tothepresenceofonly0·65wt%ofCO2inthebulkmelt^rockmix- presence of several heterogeneousdomains intheir source vem
ture(correspondingto25wt%BACþlherzolitemixture),nepheli- (i.e.theEarth’smantle)(Zindler&Hart,1986;Hofmann, b
e
nitic-basanite melts can be generated by partial reactive 1997). Because recycled, altered oceanic crust (MORB- r 2
0
crystallizationofbasalticandesiteasopposedtobasanitesproduced eclogiteatuppermantleconditions)hasbeenproposedto , 2
0
in volatile-free conditions. Post 20% olivine fractionation, the be a major contributor to mantle heterogeneity (e.g. 13
reacted melts derived from ABC at low to intermediate melt^rock Hofmann,1988,1997; Lassiter & Hauri,1998; Stixrude &
ratiosmatchwith20^40%ofthepopulationofnaturalnephelinites Lithgow-Bertelloni, 2012), it is important to evaluate its
and melilitites in terms of SiO and CaO/Al O, 60^80% in role in contributing to the genesis of oceanic basalts.
2 2 3
termsofTiO, Al O and FeO, and520% intermsof CaOand Basedontheirhigh 206Pb/204Pbratio, it hasbeen invoked
2 2 3
Na O. The reacted melts from BAC, at intermediate melt^rock that HIMU (high-m) basalts [i.e. basalts that bear evi-
2
ratios, are excellent matches for some of the Mg-rich (MgO dence of high, time-integrated 238U/204Pb (m)] contain
(cid:2) The Author 2013. Published by Oxford University Press. All
rightsreserved.ForPermissions,pleasee-mail:journals.permissions@
*Correspondingauthor.E-mail:[email protected] oup.com
JOURNALOFPETROLOGY VOLUME54 NUMBER11 NOVEMBER2013
recycled oceanic crust in their source (e.g. Chase, 1981; do not produce any melt compositions that are close to
Hofmann&White,1982;Hofmann,1997). silica-undersaturated basalts such as melilitites or nephel-
Basalts with HIMU signatures are found to be alkalic inites(LeBas,1989).
and Si-undersaturated (Kogiso et al., 1998; Jackson & Basedonthedegreeofsilica-undersaturationrequiredto
Dasgupta, 2008) and the compositions of such alkalic generatesomeofthealkalicbasaltsingeneralandHIMU
magmascannotsimplybegeneratedbyvaryingthecondi- basalts inparticular, thepresence of CO inthesource of
2
tions of melting of mantle peridotite (Dasgupta et al., these magmas has been proposed by many previous ex-
2010). Furthermore, partial melting of a volatile-free perimental studies (e.g. Wyllie & Huang, 1976; Eggler,
MORB-eclogite (subducted oceanic crust) alone does not 1978; Wyllie, 1978; Spera, 1981; Hirose, 1997; Dasgupta
explainthegenesisofsuchlavas,becausethepartialmelts et al., 2006, 2007, 2010; Jackson & Dasgupta, 2008;
are dacitic to basaltic and do not contain high enough Gerbode & Dasgupta, 2010).The involvement of CO in
2 D
MgO and low enough SiO (Yaxley & Green, 1998; the formation of silica-poor ocean island basalt (OIB) is o
2 w
Pertermann&Hirschmann,2003;Spandleretal.,2008)to also supported by the natural association of carbonatites, nlo
a
serve as primary alkalic basalts. One may argue that be- carbonate minerals, and CO -rich fluids and silica-poor d
2 e
d
cause low-degree partial melting of peridotite produces alkalic basalts and/or mantle xenoliths and has been fro
alkalic magmas, the contribution from recycled MORB- pointed out by many previous researchers (e.g. Dasgupta m
h
eclogite needs to provide only the appropriate chemical etal.,2007,andreferencestherein) ttp
components;thatis,MORB-eclogiteshouldberesponsible In previous experimental studies it has been observed ://p
e
onlyforsupplyingthenecessarychemicalenrichmentthat that the presence of CO2 in the source reduces the SiO2 tro
drives the peridotite partial melts to match plausible pri- and Al2O3 content and enhances the FeO*, CaO and log
y
mary alkali basalt compositions. However, some of those CaO/Al2O3 of peridotite-derived (Hirose,1997; Dasgupta .ox
necessary vectors suchas low SiO, lowAl O, high CaO etal.,2007) andeclogite-derivedpartialmelts (Gerbode & fo
2 2 3 rd
and high FeO* also cannot be derived from MORB- Dasgupta,2010; Kiseeva et al.,2012) when compared with jo
u
eclogitemeltingalone. theirCO -freecounterparts.Althoughpartialmeltsofcar- rn
2 a
Onekeyconsideration, however, isthat partial melts of bonated peridotite, by themselves, are not sufficiently ls.o
MORB-eclogite, entrained in the convective mantle, are enriched in TiO2 to explain most of the alkalic ocean arg/
not likely to segregate and mix with peridotite partial islandbasalts, partialmelts ofcarbonated MORB-eclogite t U
melts or erupt unmodified. Bothvolatile-free andcarbon- failtoexplainthehighMgOandlowAl2O3concentrations niv
e
ate-bearing recycled oceanic crust begin to melt deeper required by most HIMU alkalic basalts, basanites and rs
than volatile-free peridotite (Yasuda et al., 1994; Perter- nephelinites (Dasgupta et al.,2006; Gerbode & Dasgupta, ita' d
mann&Hirschmann,2003;Spandleretal.,2008),thereby 2010;Kiseevaetal.,2012;Fig.1),whichindicatestherequire- eg
creating a depth range over which MORB-eclogite- ment of a peridotitic component (Dasgupta et al., 2007; li S
tu
derivedpartialmeltsmay interact with subsolidusperido- Jackson&Dasgupta,2008).However,althoughlow-degree di R
tite. The siliceous partial melts of MORB-eclogite, not melting of carbonated, silica-excess eclogite (with coesite o
m
being in equilibrium with the surrounding mantle, present inthe residue) yields highly siliceous (andesitic to a L
undergo reactive crystallization (Yaxley & Green, 1998; dacitic)melts,suchmeltsmaycarryamodestconcentration a S
a
Lambartetal.,2012;Mallik&Dasgupta,2012).Therecent ofdissolvedCO atequilibriumwithimmisciblecarbonati- p
2 ie
study of Mallik & Dasgupta (2012) investigated such a tic melts (Hammouda, 2003; Kiseeva et al., 2012). nz
a
melt^rock reaction in a volatile-free system to evaluate Therefore, reactive infiltration of MORB-eclogite-derived on
howanMgO-poor,siliceouspartialmeltofMORB-eclog- partial melt into a peridotite matrix needs to be con- No
v
iteevolvesuponreactionwithasub-solidusgarnetlherzo- strained notonly under volatile-free conditionsbut also in em
lite, in variable ratio, in the uppermost part of the scenarios where such reactive process takes place in the be
cthoantvesicltiicveeoumsaMntOle.RABt-leocwlogmiteel-td^eroricvkedramtioesltistewvaoslvoebdsetorvaeld- epcrleosgenitceepoafrtiCalOm2,elotrisincaortbhoenratwedo.rdIsncwidheenrtealtlyh,earlleapcrteivnig- r 20, 20
1
kalic basalts in the porous flow regime with many major ousstudiesonpartialmeltingofcarbonatedperidotitecon- 3
element characteristics (high TiO and CaO; low SiO sidered a direct flux of CO or carbonates on the partial
2 2 2
and Al O) similar to those of HIMU lavas; however, at meltingbehavior nottaking intoaccountthe possibilityof
2 3
the experimental pressure^temperature conditions of this a‘flux’intheformofdissolved CO inaneclogite-derived
2
study (i.e. at 3GPa and1375^14408C) they were not low silicate melt. The difference in the two processes of CO
2
enoughinSiO andAl O orhighenoughinFeO*,CaO fluxtotheperidotitemayresultindistinctlydifferentequi-
2 2 3
andCaO/Al O toexplainthecompositionsofmanynat- librated melt compositions; however, no systematic studies
2 3
urally occurringbasanites and nephelinites. Furthermore, havebeenperformedonthelatter.
volatile-free eclogite melt^peridotite interactions (Mallik One of the key reactions that drive the composition of
& Dasgupta, 2012) at the base of the oceanic lithosphere dacitic or andesitic eclogitic melt undergoing partial
2268
MALLIK&DASGUPTA MORB-ECLOGITE-DERIVEDMELT
65
ABC Starting melt
60 BAC compositions
Carbonated partial melt
55 compositions
Volatile-free partial melt
)
% compositions
. 50
t
w
(

2
O 45
Si D
o
w
40 n
lo
a
d
e
d
35 fro
m
h
30 ttp
Natural alkalic basalts ://p
e
18 (oceanic and conti- trolo
nental) g
y
16 Average HIMU .oxfo
composition rd
jo
%) 14 urn
a
wt. ls.o
(O23 12 at Urg/
Al 10 niv
e
rs
8 ita' d
e
g
6 li S
tu
d
i R
4 om
a
30 40 50 60 70 80 90 100 L
a
S
Mg# ap
ie
n
dFeigri.v1e.d,Mvgol#ativles-fSrieOe,2M(wOtR%B)-eacnlodgAitel2Opa3r(twiatl%me)ltcocnocmenptorsaittiioonnssi(nPsetratretrimnganmnel&tcoHmirpsochsimtioannsn,A2B0C03a;nSdpBaAndClecroemtpaal.r,e2d0w08i)thaenxdpecrairmbeonntaatleldy za on
MORB-eclogitepartialmeltcompositions(Hammouda,2003;Gerbode&Dasgupta,2010;Kiseevaetal.,2012),naturaloceanicandcontinental N
o
alkalicbasalts(referencesgivenincaptionsofFigs12,14and15)andaverageHIMUcomposition(Jackson&Dasgupta,2008).Allcompositions v
e
areplottedonavolatile-freebasis. m
b
e
r 2
0
, 2
reactive crystallization in a peridotite matrix is the con- The presence of CO may critically affect these reac- 0
2 1
3
sumption of olivine and precipitation of orthopyroxene tions, because it is well known fromthe pioneering study
(Yaxley & Green, 1998; Lambart et al., 2012; Mallik & ofKushiro(1975)anddocumentedbylaterstudiesbyBrey
Dasgupta,2012)asfollows: & Green (1975,1977) that the stabilityof orthopyroxeneis
(cid:2) (cid:3) (cid:2) (cid:3) enhanced at the expense of olivine at the liquidus of
SiO þ Fe,Mg SiO ¼ Fe,Mg Si O
2 2 4 (cid:2) 2 2 6(cid:3) ð1Þ primary basaltic melts. Thus, one may expect that the
ðmeltÞ ðolivineÞ orthopyroxene
presence of CO would cause enhanced crystallization of
2
(cid:2) (cid:3) (cid:2) (cid:3) (cid:2) (cid:3) orthopyroxene at the expense of olivine, driving the
Al O þ Fe,Mg SiO ¼ Fe,Mg AlAlSiO þ Fe,Mg O:
2 3 2 4 (cid:2) (cid:3)6 reacted melt towards a more silica-depleted composition
ðmeltÞ ðolivineÞ orthopyroxene ðmeltÞ
as compared with its CO -free counterpart. Similarly,
2
ð2Þ comparison of 3GPa melting reactions of CO -free
2
2269
JOURNALOFPETROLOGY VOLUME54 NUMBER11 NOVEMBER2013
peridotite partial melting (Walter, 1998) with those of Table1: Compositionsofstartingmaterials
CO -present peridotite partial melting (Dasgupta et al.,
2
2007)suggeststhatgarnetstabilityisenhancedinacarbo-
natedsystem.Ifasimilareffectisapplicableduringreact- Peridotite Startingmelts
ive crystallization of MORB-eclogite melt in lherzolite,
KLB-1ox MixKLB-11 ABC BAC G2PM12
then the reacted melt may achieve a more alumina-
depletedcompositionthroughenhancedstabilityofgarnet
and thus become a better candidate for primary alkalic SiO2 44·82 44·54 44·22 56·46 56·3(8)
basalts.HencetheeffectofCO2duringreactiveinfiltration TiO2 0·15 0·21 2·87 5·66 5·65(8)
of siliceous MORB-melts into fertile peridotite could be Al2O3 3·51 3·70 13·45 15·61 15·6(4)
criticalingeneratingmanyofthemajorelementgeochem- Cr2O3 0·32 0·23 0·05 0·01 0·01(1)
D
ical features of strongly alkalic basalts. Nevertheless, the FeO* 8·19 8·08 15·28 8·17 8·2(3) o
w
phaseequilibriaofreactiveinfiltrationofMORB-eclogite- MnO 0·12 0·14 0·33 0·09 0·08(2) n
lo
derivedcarbonatedpartialmeltintoaperidotiticmedium MgO 39·50 39·30 6·87 2·51 2·5(1) ad
e
areInunthciosnsstturadiynewde. investigatethe fate of MORB-eclogite- CNaaO2O 03··3007 03··2592 142··2318 37··7560 47··052(2(9)) d from
derived,carbonated,partialmeltuponreactionwithvola- K2O 0·02 0·01 0·33 0·21 0·21(1) http
tile-free subsolidus, fertile peridotite in the uppermost Mg# 89·58 89·66 44·50 35·42 35·4(9) ://p
part of the convective mantle (base of the lithosphere). e
CO2 — — 11·7 2·6 — tro
The experiments were performed at the same pressure^ Sum 100·00 100·02 100·00 100·00 99·99 lo
g
temperature conditions as the recent study by Mallik & y
.o
Dasgupta (2012), to understand directly the effect of CO2 Major element compositions of KLB-1ox, ABC and BAC xfo
in such a melt^rock reaction scenario. We demonstrate and CO concentrations based on proportions of oxides rd
that infiltration of MORB-eclogite-derived, andesitic to acnodncceanrtbr2aotnioantessomfiAxeBdCinanthdeBsAtaCrtianrgecroepmoprotesditioonns.aOCxOide- journ
mildly alkalic carbonated basalts into peridotite causes free basis. 2 als
thereactedmeltstoevolvetonephelinitic-basanitetomeli- *All Fe assumed to be FeO. .org
lititic compositions through crystallization of orthopyrox- 1Composition of fertile peridotite as used by Mallik & a/
Dasgupta (2012). t U
ecnome paonsditiognasrnaerte. aWbeetatlesromshaotcwh ftohratnastuucrhalrneeapctheedlinmiteelst, 2MCaolmlikp&osiDtioansguopftaM(O2R01B2-)e.clogite partial melt as used by nivers
nephelinitic basanites, and melilitites from oceanic and ita
continental provinces as compared with partial melts of ' d
(Fig.1).Volatile-freeperidotiteKLB-1oxusedinthis study eg
carbonated eclogite and carbonated peridotite, as well as is similar to the fertile peridotite composition MixKLB-1 li S
melts that are the product of reaction between a partial tu
meltofMORB-eclogiteandsubsolidusperidotiteinavola- used by Mallik & Dasgupta (2012) and, within1serror, di R
similar to the KLB-1peridotite composition of Herzberg o
tile-freesystem. m
et al. (1990) and Davis et al. (2009). All starting compos- a L
itionsusedinthisstudyarereportedinTable1. a S
EXPERIMENTAL TECHNIQUES Allthestartingmaterials(carbonatedmeltsandperido- ap
ie
Starting materials tite)weresynthesizedusinghigh-purityoxidesandcarbon- nz
a
ates from Alfa Aesar. SiO,TiO, Al O and MgO were o
TheMORB-eclogite-derivedcarbonatedmeltsusedinthe 2 2 2 3 n
fired overnight at 10008C, Fe O at 8008C, MnO at N
reaction couple with volatile-free, fertile peridotite are (1) 2 3 2 o
v
4008C,MgCO at2008C,CaCO at2508C,andNa CO e
‘ABC’, which is an alkalicbasalt similar to a 43% partial 3 3 2 3 m
amGneedrltb(o2d)oef‘B&AcaCDr’b,aoswgnhuaiptcethdai(s2Ma01O0b)aRsaaBnl-tdeicccloaonngtidateeisnitseG11w2·7Citwhtua%sesidmCiOlba2yr, Kawna2dsCOKad32.dCeTOdo3mtoaatiAn11Bt0aC8inCatsthoeMmdgeinCsiirOmed3i,zepCraoadpCsooOrrtb3i,oendNaow2faCtMeOrg.3:CCaOOnd22 ber 20, 201
composition to G2PM1 used by Mallik & Dasgupta inthismelt,MgwasaddedpartlyasMgOandtherestas 3
(2012), when normalizedonavolatile-freebasis. BAC cor- MgCO3. Reagent grade MgCO3 often contains variable
responds to 8·9wt % partial melt of natural volatile-free amounts of water, especially in the form of brucite. To
MORB-eclogite G2 (Pertermann & Hirschmann, 2003) ensure that addition of MgCO3 does not introduce water
andcontains2·6wt%ofCO (Table1).TheCO concen- in our ABC melt mix, X-ray diffraction of the fired
2 2
trations mentionedabove arebasedontheproportions in MgCO powderwasperformed,whichyieldednodiscern-
3
whichoxidesandcarbonatesweremixedtosynthesizethe ible peaks for Mg(OH). In the case of BAC, CO was
2 2
starting melts. Both starting melt compositions lie within added only as Na CO. Prior to adding carbonates to
2 3
the range of MORB-eclogite ((cid:2) carbonates) derived par- introduceCO,theoxidesandcarbonatesweregroundto-
2
tial melts produced experimentally in previous studies getherunderethanolinanagatemortarfor30min.Once
2270
MALLIK&DASGUPTA MORB-ECLOGITE-DERIVEDMELT
theethanolevaporated,themixtureswerefiredat10008C (2010) for the graphite^carbonate^oxygen buffer; these
in a CO^CO mixing furnace at log fO (cid:3)QFM -2 variedfromQFM-1·8toQFM-2·3,whichisinagreement
2 2
(where QFM is the quartz^fayalite^magnetite buffer) for withthe predictions of Frost & Wood (1995) and Medard
24hto convert allthe Fe3þ to Fe2þ aswellasto decarbo- et al. (2008) for experiments performed in graphite cap-
nate the powders. Once reduction was complete, carbon- sules.The calculated fO values also lie within the range
2
ateswereaddedinthecaseofABCandBACtointroduce ofoxygenfugacityestimates(QFMþ1·5to(cid:4)3·5)forocea-
CO intothestartingmeltmixes.Thefinalmixtureswere nicandabyssalperidotitescompiledbyFoley(2011).
2
ground in an agate mortar under ethanol for 30min. All ExperimentalconditionsarereportedinTable2.
starting mixtures were stored at1108C to prevent adsorp-
Analysis of run products
tion of moisture at any time.The starting materials com-
posed of 8^50wt % carbonated eclogite-derived melt Analyses of textures and phase compositions were per-
D
(ABC or BAC), mixed homogeneously with peridotite, formed using a Cameca SX50 electron microprobe at o
w
simulatereactiveinfiltrationofeclogitemeltintoperidotite TexasA&MUniversity.High-resolutionimagingwasper- n
lo
via porous flow.The bulk CO2 content ranged from 0·93 formed using an FEI Quanta 400 FEG-SEM at Rice ade
taond5·83frwomt%0f·o2r1extpoer1im·3ewnttsw%ithKforLBe-1xpþeAriBmCenmtsixtwuritehs UGnamivemrsaitTy.ecPhh(aPseGsTw)eereneirdgeyn-tdiifsiepdersuisviengspIemctirxosPcoripnyceatnond d from
KLB-1þBACmixtures. reportedphasecompositionsweredeterminedusingwave- h
ttp
Experimental procedure lwenergetha-ndaislypzeerdsivaet aspneactcrcoeslceoraptyi.ngAvltohlotauggehoafl1l5tkhVe, gplhaassseess ://pe
The experimentswere performedusing end-loadedpiston were analyzed under a beam current and diameter of trolo
g
cylinders at Rice University using a half-inch assembly 10nA and 5^10mm, respectively; all other phases except y
.o
comprising BaCO3 pressure medium, straight graphite quench aggregates in melt pool (olivine, orthopyroxene, xfo
furnace, crushable MgO spacers, and Pt^graphite double clinopyroxene and garnet) were analyzed using a beam rd
jo
capsules. Pressure^temperature calibrations relevant for current and diameter of 20nA and1mm, respectively. In urn
our experimental set-up have been given by Tsuno & alltheexperiments,thereactedmeltquenchedtoahetero- als
Dasgupta (2011).The homogeneous mixtures were loaded geneous assemblage (except for a few experiments where .org
intothickgraphitecapsulessurroundedbyouterplatinum someofthemeltformedglasspools).Alargebeamsizeof a/
capsules. Before the platinum capsules were welded shut, 20^30mm and current of 20nA were used to obtain the t U
n
the loaded capsules were kept at 1108C overnight to average composition of these heterogeneous aggregates ive
rs
removeanymoistureandensuretheleastwatercontamin- andmanysuchpointsacrosstwotothreeexposedsections ita
ationpossible.Thelossinweightofthecapsulesafterweld- were analyzed to obtain a reliable mean composition of ' d
e
g
ing was measured and in no case was the loss observed thereactedmelt.Nofilteringtechniquewasappliedtothe li S
greater than 0·3% relative. The experiments were per- heterogeneous quench aggregate compositions to avoid tu
d
formed at13758C,3GPa; that is, at pressure^temperature any selective bias. Counting times for the elements ana- i R
o
conditions equivalenttothebase of mature oceaniclitho- lysedvariedfrom20to80sonthepeakandhalfthetime m
a
spherewheretheeclogite-derivedcarbonatedmeltswould oneachbackground.Nawasanalyzedfirstonagivenspec- L
a
be above their respective liquidibut the peridotite would trometer and for only 20s on the peak and 10s on each S
a
p
be below its solidus. In all the experiments, the pressure background to limit its loss. Compositional zoning was ie
n
was first raised to 3GPa andthen the systemwas heated observedin some garnets andin suchcases, therimcom- za
to13758Cat1008Cmin(cid:4)1.Temperaturesoftheexperiments positions were analyzedassuming that the rim is in equi- on
N
were controlled and monitored using W Re/W Re librium with the rest of the assemblage. The natural o
95 5 74 26 v
e
(Type C) thermocouples. The run durations varied from mineralandbasalticstandardsusedforphasecomposition m
b
47 to 172h. Experiments were quenched by cutting off analysis are the same as those used by Mallik & er 2
power from the heater after which the runs were decom- Dasgupta (2012). For some quench aggregate analyses, Ca 0
, 2
pressedslowly,thepositionofthecapsuleandthermocouple and Mg concentrations were measured using a dolomite 0
1
3
with respect to the hotspot was checked, and the capsules standard. However, it was found that these analyses were
were recovered.The capsuleswere then mounted in epoxy nodifferentfromthosewherethesameelementsweremea-
and polished using silicon-carbide paper (240^600 grit) suredusinga diopside standard, as inthe studyof Mallik
and polycrystalline diamond powder (diameter varying & Dasgupta (2012), thereby suggesting no particular ad-
from3to0·25mm)onnylonandvelvetcloths.Inallcircum- vantage of using carbonate standards over silicate stand-
stances, polishing was carried out in the absence of water ards for measuring major element cation concentrations
oranylubricatingliquidtopreservequenchcarbonatecrys- incarbonatedsilicatemelts.
talsexpectedtobepresentinthereactedmeltpool. Phase proportions for each experiment were calculated
Oxygenfugacitiesofthemelt^rockreactionexperiments basedonmassbalanceofcomponentsusingtheoptimiza-
were calculated using the calibration of Stagno & Frost tion tool Solver in MS-Excel. Minor elements such as
2271
JOURNALOFPETROLOGY VOLUME54 NUMBER11 NOVEMBER2013
Table2: Summaryofmelt^rockreactionexperimentsperformedinthisstudyincludingexperimentalconditions,phasespre-
sent,andtheirmassproportions
Runno. Starting Reacting T(8C) TBKN(8C) TR00(8C) P(GPa) Duration(h) Mineralmodes(wt%) (cid:2)r2
materialsused meltmass
Ol Opx Cpx Gt Melt
(wt%)
G209* ABCþKLB-1ox 8 1375 1332 1295 3 119 51(1) 13(1) 18(1) 13(0) 4 0·5(2)
G217y ABCþKLB-1ox 15 1375 1350 1298 3 144 40(2) 26(3) 13(3) 13(1) 9(2) 0·6(2)
G239 ABCþKLB-1ox 25 1375 1327 1319 3 96 30(3) 28(5) 13(3) 13(1) 15(4) 0·3(1) D
o
G240 ABCþKLB-1ox 33 1375 1323 1313 3 96 27(2) 23(4) 16(2) 14(1) 20(3) 2·0(1) w
n
G227y ABCþKLB-1ox 40 1375 1328 1293 3 116 22(3) 25(5) 12(2) 20(0) 21(5) 0·20(3) lo
a
d
G229 ABCþKLB-1ox 50 1375 1349 1301 3 93 — 40(2) 15(4) 24(1) 21(2) 0·53(9) e
d
B126* BACþKLB-1ox 8 1375 1334 1276 3 95 42(1) 27(1) 14(1) 12(0) 3 0·7(4) fro
m
B157* BACþKLB-1ox 15 1375 1357 1314 3 100 36(1) 24(1) 17(1) 14(1) 6 0·8(3) h
B222 BACþKLB-1ox 25 1375 1340 1483 3 92 23(1) 43(1) 8(1) 7(1) 18(2) 0·2(2) ttp
B223 BACþKLB-1ox 33 1375 1325 1404 3 139 16(1) 43(2) 12(1) 11(1) 17(2) 0·27(6) ://pe
G231 BACþKLB-1ox 40 1375 1378 1281 3 90 — 52(1) 20(4) 18(1) 11(2) 2(1) trolo
G226 BACþKLB-1ox 50 1375 1399 1308 3 115 — 33(3) 28(5) 15(1) 24(3) 3·7(6) gy
.o
x
fo
*Experiments where melt mass proportions are calculated by the method of linear extrapolation (equations of extrapo- rd
lation are given in the caption of Fig. 3). jo
u
yExperiments re-reported from Dasgupta et al. (2013). rn
The(cid:2)1serrors,basedonreplicateelectronmicroprobeanalyses,aregiveninparenthesesandreportedastheleastdigits als
cited.Forexample,51(1)shouldbereadas51(cid:2)1wt%.ABCandBACare11·6and2·6wt%CO -bearingstartingmelts, .o
respectively and KLB-1ox is the oxide mixture used as fertile peridotite. TBKN (8C), temperature o2btained using the two- arg/
pyroxene thermometer of Brey & Kohler (1990); TR00 (8C), temperature obtained using the garnet–clinopyroxene therm- t U
ometer of Ravna (2000) using average pyroxene and garnet compositions reported in Tables 5, 6 and 7; Ol, olivine; Opx, n
orthopyroxene; Cpx, clinopyroxene; Gt, garnet; (cid:2)r2, sum of residual squares obtained using phase proportions, phase ive
compositions, and the bulk starting composition; —, absence of a phase. rsita
' d
e
g
Cr O,MnOandK O,aswellasCO ,wereneglectedin Because it has been demonstrated that averaging quench li S
2 3 2 2 tu
the mass balance. The phase proportions obtained from aggregates produces reliable estimates of melt compos- di R
massbalancewereverifiedtexturallytoensuretheirvalid- itions by comparison with the compositions of glassy o
m
ity. Based on mass balance of FeO in the melt^rock sys- patches, CO2 estimates obtained‘bydifference’havebeen a L
tems, we concludedthat there was no Fe loss tothe outer used to calculate the volatile-free melt compositions a S
a
Ptcapsule. reportedinTable3. p
ie
CO2concentrationsinthereactedmeltareestimatedin nza
twoways: (1) bydifferencebetween100% andthe micro- o
n
probetotalsoftheaveragedcompositionofthequenchag- EXPERIMENTAL RESULTS N
o
v
gregates(assumingallofthedeficitisattributabletoCO ; Experimental conditions and phase proportions are re- e
2 m
mhceeonrdetraaaflttepiorrnorpteoofrertrieroaencdsteotdofmmaeeslltt‘CbbOyeca2alulboseyttinndogiffsteohrleeidnbcuceal’k)r;bCo(2nO)a2tefcrsooonmr- paprooersitstehidoonwisnnaTrianeblFpeilgo2.t.t2e.PdhMoaotsodamalifcuprnroocgtpriooarpnthioosfnosrfeaatnhcdteinmrguinnmeprearltlodcmouamcstss- ber 20, 20
1
CO -richfluidphasewaspresentaftermelt^rockreaction inFigs3^8andlistedinTables3^7. 3
2
andbecause carbon solubility in nominally C-free mantle
silicates is negligible (Shcheka etal.,2006). Both estimates Textures and phase assemblages
are included in Table 3 and plotted in Supplementary MixKLB-1 at 3GPa and 13758C produced a four-phase
DataFig.S1(availablefordownloadingathttp://www.pet- lherzolite with olivine, orthopyroxene, clinopyroxene and
rology.oxfordjournals.org). Given the large uncertainty in garnet as describedby Mallik & Dasgupta (2012). All ex-
theestimatedCO contentsofthemelts,apositivecorrel- periments with peridotite^basaltic andesite (BAC) and
2
ationbetweenthetwosuggeststhatthetrendinmeltcom- peridotite^alkali basalt (ABC) mixtures produced an as-
position evolution remains unaffected by whichever semblage of orthopyroxeneþclinopyroxeneþgarnetþ
methodofCO estimationinthereactedmeltsisadopted. melt(cid:2)olivine. In the case of 8wt % melt-added
2
2272
MALLIK&DASGUPTA MORB-ECLOGITE-DERIVEDMELT
Table3: Compositionofreactedmelts
Runno.: G217 G239 G240 G227 G229 B222 B223 G231 G231gl G226 G226gl
Startingmelt: ABC ABC ABC ABC ABC BAC BAC BAC BAC BAC BAC
Meltadded(wt%) 15 25 33 40 50 25 33 40 40 50 50
n: 26 23 18 18 50 23 18 21 32 60 60
SiO2 30(4) 39(3) 38(2) 31(5) 27(6) 42(2) 47·3(1) 52(1) 51·5(6) 53(1) 53·2(6)
TiO2 6·3(9) 2·8(7) 3·4(2) 4·6(8) 5(1) 6·4(4) 5·9(2) 7·0(2) 7·0(2) 8·7(3) 8·34(9)
Al2O3 5·2(8) 8·4(5) 9·1(5) 5·3(8) 4·1(8) 10·5(5) 11·9(4) 12·9(3) 12·7(2) 12·3(2) 12·3(1) D
Cr2O3 0·8(6) 0·13(4) 0·10(4) 0·2(3) 0·0(4) 0·2(1) 0·2(2) 0·04(1) 0·04(2) 0·1(1) 0·05(1) ow
FeO* 14(2) 11(1) 12·3(9) 15(2) 16(2) 10(1) 9·9(9) 7·7(4) 7·6(2) 6·5(5) 6·35(3) nlo
a
MnO 0·21(3) 0·15(2) 0·17(2) 0·22(3) 0·23(4) 0·12(3) 0·11(3) 0·05(1) 0·06(1) 0·06(1) 0·064(8) de
d
MgO 21(2) 20·3(8) 17·2(9) 19(1) 20(1) 15·4(7) 12·2(8) 8·3(8) 9·1(5) 7·9(8) 8·6(3) fro
CaO 18(5) 14(2) 13(1) 20(5) 21(3) 10(1) 7·9(3) 8·7(6) 9·0(3) 7·3(4) 7·6(2) m
h
Na2O 4(2) 4(1) 6(2) 4(2) 7(3) 4·0(7) 4·2(4) 3·0(3) 2·9(2) 3·4(3) 3·1(4) ttp
K2O 0·23(1) 0·11(6) 0·21(5) 0·3(2) 0·4(2) 0·4(1) 0·34(5) 0·18(6) 0·14(2) 0·35(9) 0·31(4) ://pe
Sum 100·00 100·00 100·00 100·00 100·00 100·00 100·00 100·00 100·00 100·00 100·00 tro
lo
Mg# 73(9) 73(3) 69(4) 70(6) 69(4) 73(3) 69(2) 66(9) 68(4) 68(9) 71(3) g
y
CO21 22(5) 12(3) 10(3) 23(5) 25(3) 12(6) 4(2) 7(3) 7(1) 6(2) 7·2(3) .ox
TiO22 3·4(6) 2·8(4) 3·2(3) 3·5(5) 4·7(4) 6·0(5) 7·0(4) 12(3) 12(3) 9(1) 9(1) ford
NK2aO2O22 50··25((81)) 40··83(68()9) 05··73(47()5) 06·(318)(8) 70··88(27()9) 04··53(74()4) 05··52(33()2) 05··49((73)) 50··49((73)) 40··02(15()5) 40··021(5(5)) journals
CO22 15(4) 17(3) 19(1) 23(3) 20(2) 4(2) 5·1(6) 5·8(7) 7(1) 4·0(1) 7·2(3) .org
a/
*All Fe assumed to be FeO. t U
1CO2 concentration determined by difference. niv
2TiO2, Na2O, K2O and CO2 concentrations determined from mass balance (refer to text for details). ers
All oxide concentrations are reported in weight per cent and on a CO2-free basis. ita
' d
e
g
experimentsforbothstartingmeltcompositionsand15wt experiments with melt present only at triple junctions, li Stu
% BAC-added experiment, no melt pool, separated as a meltmodalproportionswereestimatedbylinearextrapo- di R
layer,coexistingwiththeresidualfour-phaselherzoliticas- lationofthetrendofmodalproportionsofmeltasafunc- o
m
semblage, was observed. In these experiments, melt was tionofamountof melt addedtotheexperiments. Overall a L
a
found only as pockets in triple-grain junctions (Fig. 2c). the melt mass increased from 4wt % (where melt is con- S
a
For higher melt-added experiments, a distinct melt pool centratedonlyintriple-grainjunctions)to21wt%forex- pie
composed of heterogeneous quench crystals was observed periments with ABC and from 3·1 to 24wt % for nz
a
incoexistencewithlherzoliticphases,exceptforthe40wt experiments with BAC. It is observed that the olivine on
N
% BAC-addedexperiment and 50wt % ABC-and BAC- massfractiondecreasesfrom60wt%inthestartingperi- o
v
added experiments, where olivine was absent. The dotite to 0wt % when 50wt % of ABC and 40wt % of em
b
quenched melt pools are composed of quench clinopyrox- BACareadded,respectively.Olivineisconsumedbecause e
ene-likesilicatesintimatelyassociatedwithCa^Na^Fecar- of pyroxene crystallization. Accordingly, the orthopyrox- r 20
bonates (Fig. 2a, b and d). No textural evidence of a ene mode overall increases from14 to 40wt % asaresult , 20
1
separate CO -rich vapor phase was observed. This is in of addition of 0^50wt % ABC. It increases from 28 to 3
2
keeping with CO solubilities for silica-undersaturated 52wt % until 40wt % BAC is added, followed by a de-
2
melts (similartothose obtainedinthis study) determined creaseto33wt%oforthopyroxenewhen50wt%BACis
by previous studies (e.g. Mysen et al.,1975; Brooker,1998; added.This decrease in orthopyroxene modal proportion
Brookeretal.,2001)beingsimilartoorexceedingtheCO between40and50wt%BAC-addedexperimentsiscom-
2
concentrations determined for thereacted melts fromthis pensatedbyanincreaseinclinopyroxenemodefrom20to
study. 29wt % over the same range of conditions.This reversal
In our experiments, it is observedthat there is abroad in pyroxene stabilization canbe explainedby an increase
linear correlation between reacting melt mass and mass in bulk Ca/Mg with increasing melt added (from 0·14
fraction of melt after melt^rock reaction (Fig.3). For the for 40% BAC-added to 0·18 for 50% BAC-added
2273
JOURNALOFPETROLOGY VOLUME54 NUMBER11 NOVEMBER2013
(a) 200 μm (b) Graphite
Quenched
Ol melt
Gt Quench cc
Opx
D
o
Opx Quench w
n
cpx lo
a
d
Gt Ol ed
fro
m
Cpx h
ttp
Graphite 100 μm ://p
e
tro
lo
g
y
(c) (d) .o
GGrraapphhiittee x
fo
rd
jo
u
rn
Gt a
ls
.o
rg
a/
t U
n
iv
e
rs
ita
' d
e
MeMlt eplto cket OOppxx gli S
pockets tu
d
i R
GGllaassss ppooooll o
m
a
40 μm QQ uu ee nnmmcc eehhllttee dd OO55pp00xx μμmm La Sapien
z
a
o
n
N
Fig.2. Back-scatteredelectronimagesofexperimentswith(a)15wt%ABCþKLB-1mixture(G217).Quenchedmeltpoolisobservedtoco- o
existwitholivine,orthopyroxene,clinopyroxeneandgarnetintheresidue.Thewhiteboxrepresentstheareathatisshownmagnifiedin(b). ve
m
(b)Magnificationoftheareamarkedbywhiteboxin(a).Thequenchaggregatescompriseclinopyroxene-likesilicatesandCa^Fe^Nacarbon- b
ates.(c)Meltquenchedinthetriplejunctionoforthopyroxenecrystalsintheresidueofmelt^rockreaction(G209).Themeltdoesnotforma er 2
separatepoolinthisexperimentunlikein(a).(d)Glasspoolformedalongwithquencheddendriticaggregatein40wt%BACþKLB-1oxex- 0
periment(G231).Thesimilarityinglassandaveragedquenchmeltpoolcompositions(Fig.4)demonstratesthereliabilityofouraveragingtech- , 2
0
niqueusedtomeasurequenchmeltpoolcompositions. 13
experiments). ForABC-addedexperiments, clinopyroxene experimentswithBACfortheinitialmeltmassincreasing
mode displays an overall decrease from19 to15wt % for from0to50wt%(Fig.3).
0^50wt%melt-addedruns.ForBAC-addedexperiments,
theclinopyroxenemode increases sharply from 8 to 28wt Assessment of chemical equilibrium
% for added melt fraction of 25^50wt %. Garnet modal The experiments in this study have not been reversed.
proportions display an overall increase from 10 to 24wt Garnets are zoned in some instances, even though the
% for experiments with ABC and to 15wt % for cores are less than 1% in volume with respect to the
2274
MALLIK&DASGUPTA MORB-ECLOGITE-DERIVEDMELT
60 60
%) %)
on (wt. 50 n (wt. 50
porti 40 ortio 40
odal pro 2300 dal prop 2300
m o
ne 10 x m 10
Olivi 300 Op%) 0 Dow
dal proportion (wt.%) 122505 odal proportion (wt. 112050 http://petronloaded from
o m lo
x m 10 net 5 gy.o
Cp 5 Gar 0 xford
jo
%) 1:1 0 10 20 30 40 50 urn
n (wt. 40 Reacting melt mass (wt.%) als.org
ortio 30 ABC Modes determined from at U/
op BAC mass-balance. niv
al pr 20 ABC Modes determined from ersita
mod BAC extrapolation. ' deg
Melt 10 Modes in MixKLB-1. li Stud
0 Volatile-free melt-peridotite experiments. i Ro
0 10 20 30 40 50 (Mallik and Dasgupta, 2012) ma
L
Reacting melt mass (wt.%) Linear extrapolation for ABC a Sa
p
Linear extrapolation for BAC ien
z
a
o
n
Fig.3. Massfractions(inwt%)ofolivine,orthopyroxene(opx),clinopyroxene(cpx),garnet,andmeltplottedasafunctionoftheinitialmelt N
mass(inwt%)reactedwithperidotiteforexperimentswithtwostartingmeltsçABC(diamonds)andBAC(filledsquares).Alsoplottedfor ov
e
comparison are the fraction of reacted melts fromvolatile-free G2PM1þMixKLB-1mixtures performed at13758C,3GPa from Mallik & m
Dasgupta(2012)(opensquares).Becausenoreactedmeltpoolwasobservedforthe8wt%ABC-addedrun,and8wt%and15wt%BAC- be
addedexperiments,themeltfractionsfortheseexperimentswereestimatedbylinearextrapolationofmeltconsumptiontrendsconstructed r 2
0
forABC-andBAC-addedexperiments.Theextrapolatedmeltproportionsareplottedwithgraysymbolsandtheequationsusedare:(final , 2
meltfraction)¼0·51(cid:5)(reactingmeltfraction)forABCþKLB-1oxexperimentsand(finalmeltfraction)¼0·45(cid:5)(reactingmeltfraction)for 01
BACþKLB-1oxexperiments.The1:1linerepresentsthelocusofallpointsforwhichthemassfractionofmeltaddedandtheresultingmeltfrac- 3
tionsarethesame.Errorbars((cid:2)1swt%)arederivedfrommassbalancecalculationsbypropagatingcompositionaluncertaintiesbasedon
replicatemicroprobeanalyses.
entire volume of garnet, rendering them insignificant for (1)Thesumofresidualsquares((cid:2)r2;Table2)forthecar-
affectingmassbalanceoroverallmineralogyintheexperi- bonated melt^rock reaction experiments varies from 0·2
ments.Thus, even though complete equilibration has not totwo. Giventheuncertainty inanalysesof meltcompos-
been achieved, an approach to equilibrium and mainten- ition from averaging the heterogeneous quench phases,
ance of a closed system during the experiments can be the experiments canbe consideredto havebeenclosedto
demonstratedbythefollowingcharacteristics. materialexchange.
2275
JOURNALOFPETROLOGY VOLUME54 NUMBER11 NOVEMBER2013
60 24
20
50
%) %) 16
(wt.2 40 O (wt. 12
O
Si 30 Mg 8
4
20 0
8 22 D
o
wt.%) 6 wt.%)18 wnloade
O (2 4 aO (14 d from
Ti C h
2 10 ttp
://p
0 6 etro
14 9 log
y
8 .o
wt.%) 1102 wt.%) 67 xfordjou
OAl (23 468 NaO (2 345 arnals.org/
2 t U
2 1 niv
e
16 80 rsita
' d
%) 14 70 eg
wt. 12 #60 li Stu
O* ( 10 Mg50 di Ro
e m
F a
8 40 L
a
S
a
6 30 p
ie
0 20 40 60 80 100 n
z
25 a
Reacting melt mass (wt.%) o
n
20 N
o
%) ABC ve
wt. 15 BAC Reacted melts mbe
(2 10 r 20
O ABC , 2
C 5 Starting melts 01
3
BAC
0
0 20 40 60 80 100 Glass (BAC)
Reacting melt mass (wt.%) Volatile-free melt-peridotite
experiments (Mallik and
Dasgupta, 2012)
Fig.4. AveragecompositionsofreactedmeltsderivedfromABC-added(diamonds)andBAC-added(filledsquares)experimentsplottedasa
functionofreactingmeltmass.PlottedforcomparisonarethestartingmeltcompositionsABCandBAC.Thereactedmeltcompositionsfor
8wt%ABC-added,8wt%and15wt%BAC-addedexperimentscouldnotbedeterminedowingtotheabsenceofaquenchorglasspool
and the fact that the melt pockets at triple-grain junctions (Fig. 2c) were not large enough to be analyzed reliably by microprobe. For
(continued)
2276