Post-collisional magmatism: Crustal growth not identified by zircon Hf–O isotopes

Post-collisional

magmatism:

Crustal

growth

not

identified

by

zircon

Hf–O

isotopes

Simon Couzinié

a,b,

∗

,

Oscar Laurent

c,d

,

Jean-François Moyen

a

,

Armin Zeh

c,e

,

Pierre Bouilhol

f

,

Arnaud Villaros

g,h,i

a_{Univ.Lyon,UJM-Saint-Etienne,UBP,CNRS,IRD,LaboratoireMagmasetVolcansUMR6524,F-42023SaintEtienne,France} b_{UniversityofStellenbosch,DepartmentofEarthSciences,PrivateBagX1,7602Matieland,SouthAfrica}

c_{J.W.GoetheUniversität,InstitutfürGeowissenschaften,Altenhöferallee1,D-60438FrankfurtamMain,Germany} d_{UniversitédeLiège,DépartementdeGéologieB20,QuartierAgora,alléedusixAoût12,B-4000Liège,Belgium}

e_{KarlsruherInstitutfürTechnologie,CampusSüd,InstitutfürAngewandteGeowissenschaften,AbteilungMineralogieundPetrologie,Kaiserstrasse12,} D-76131Karlsruhe,Germany

f_{DepartmentofEarthSciences,DurhamUniversity,ScienceLabs,DurhamDH13LE,UnitedKingdom} g_{Universitéd’Orléans,ISTO,UMR7327,45071,Orléans,France}

h_{CNRS,ISTO,UMR7327,F-45071Orléans,France} i_{BRGM,ISTO,UMR7327,BP36009,45060Orléans,France}

a

r

t

i

c

l

e

i

n

f

o

a

b

s

t

r

a

c

t

Articlehistory: Received9March2016

Receivedinrevisedform19September 2016

Accepted21September2016 Availableonlinexxxx Editor:A.Yin Keywords:

post-collisionalmaficmagmas crustalgrowth

zircon Lu–Hfisotopes Oisotopes

metasomatizedmantle

The combinationofU–Pb,Lu–Hfand Oisotopicanalysesinglobalzircondatabaseshasrecentlybeen used to constrain continental crustal growth and evolution. To identify crust-forming events, these studies rely on the assumptionthat new crust is formed from depleted mantle sources.In contrast, thisworksuggeststhat post-collisionalmaficmagmasand theirderivativesrepresentanon-negligible contributiontocrustalgrowth,despitehavingzirconswith“crust-like”Hf–Oisotopiccharacteristics.We address thisparadoxand itsimplications for crustal evolutiononthe basisofacase studyfromthe VariscanFrenchMassifCentral(FMC).Thelatestagesofcontinentalcollisionsaresystematicallymarked bytheemplacementofpeculiarmaficmagmas,richinbothcompatible(Fe,Mg,Ni,Cr)andincompatible elements(K2O,HFSE,LREE)anddisplayingcrust-liketraceelementpatterns.Thisdualsignatureisbest

explainedby meltingof phlogopite- (and/oramphibole-)bearingperidotite, formed bycontamination ofthe mantle bylimitedamounts (10–20%)of crustal materialduringcontinental subduction shortly preceding collision. Mass balance constraints show that inmelts derived from sucha hybrid source, 62–85%ofthebulkmassisprovidedbythemantlecomponent,whereasincompatibletraceelementsare dominantly crustal inorigin.Thereby,post-collisional maficmagmas representsignificantadditions to thecrust,whilsttheirzirconshave“crustal”isotopesignatures(e.g.₋2<

ε

Hft<−9 and+6.4< δ18O<

+10h intheFMC).Becausepost-collisional mafic magmasare(i) ubiquitoussince the lateArchean; (ii)theparentalmagmasofvoluminousgranitoidsuites;and(iii)selectivelypreservedinthegeological record, zircons crystallizedfromsuchmagmas(and anymaterialderived fromtheirdifferentiationor reworking)biasthecrustalgrowthrecordofglobalzirconHf–Oisotopicdatasetstowardsancientcrust formation and, specifically, may lead to an under-estimation of crustal growth rates since the late Archean.

1. Introduction

Theformationofthecontinentalcrustis amajorconsequence

of planetary differentiation and played a key role in the

evolu-*

Correspondingauthorat:Univ.Lyon,UJM-Saint-Etienne,UBP,CNRS,IRD, Labo-ratoireMagmasetVolcansUMR6524,F-42023SaintEtienne,France.

E-mailaddresses:simon.couzinie@univ-st-etienne.fr, simon.couzinie@ens-lyon.org(S. Couzinié).

tion of climate and life (Campbell and Allen, 2008; Lowe and

Tice, 2004). The mechanisms and rates of continental formation

are therefore fundamental parameters to be addressed and have

long beenamatter ofcontroversy(Arndt, 2013; Hawkesworthet

al., 2010). The formation of new continental crust requires two

fundamental conditions to be fulfilled, regardless of the tectonic

settinginwhichittakesplace:(1)genesis(anddifferentiation)of

mantle-derivedigneous material;and(2)long-termincorporation

and preservation of this material into the pre-existing

Table 1

SummaryofthemaindocumentedoccurrencesofPCMMinthegeologicalrecord.

Orogen AgeofPCMM Specificnomenclature Differentiates Proposedreferences

LateArcheanterranes 2950to2500Ma Sanukitoid Sanukitoidsuite Laurentetal. (2014a) Svecofennian(Scandinavia) 1800Ma Ladogite,Nevoite Post-kinematicgranitoids Rutanenetal. (2011)

Pan-African 620–560Ma Sanukitoid High-Kcalc-alkaline(HKCA)

granitoids

Liégeoisetal. (1998) Ross(Antarctica) 515–500Ma Nospecificnamegiven Nospecificnamegiven Hagen-Peteretal. (2015)

EastKunlun(China) c.450Ma Appinite Notdescribed Xiongetal. (2015)

Caledonian 425–410Ma Appinite HighBa–Srgranites FowlerandRollinson (2012)

CentralAsianOrogenicBelt 425to300Ma (severalpulses)

Sanukitoid Notdescribed Yinetal. (2015)

Variscan 335–300Ma Vaugnerite,Durbachite, Redwitzite,Appinite

Felsicdurbachites;K-richcalc-alkaline granitoids(KCG)

Moyenetal. (inpress); Holub (1997);

von

Raumer etal. (2014)

NorthChinaBlockassembly (Dabie-Suluorogen. . . )

320to115Ma (severalpulses)

Appinite Post-collisionalgranitoids Zhaoetal. (2013) Himalaya 20to3Ma Post-collisional(ultra)potassic

magmas

(Ultra)potassicsilicicmagmas Williamsetal. (2004)

tal volume (Condie etal., 2011; Hawkesworth etal., 2009; Stern

and Scholl, 2010). Radiogenic isotope systems like Rb/Sr, Sm/Nd

andLu/Hfareextensivelyused totrackcrust-forming events,and

itisconsensuallyassumedthattheradiogenicisotopecomposition

of newly-formed crust is identical to that of the depleted

man-tle(DM)(McCullochandWasserburg,1978; Patchett etal., 1982;

VervoortandBlichert-Toft,1999).

In thisrespect, thestudy ofcontinental growthrecently

ben-efitedfromadvancesinanalytical techniquesenabling insitu

iso-topicmeasurements inzircon,a widespreadaccessory mineralof

continental igneous rocks. Because zircon is able to survive

sev-eralmetamorphic,igneousandsedimentarycyclesandcontainsa

wealth of isotopic information (U–Pb, Lu–Hf, O isotopes),it

rep-resents an outstanding archive of crust formation and evolution

(Bouilhol et al., 2013; Bouilhol et al., 2011; Condie et al., 2011;

RobertsandSpencer,2015).The analysesofU–Pb andLu–Hf

iso-tope compositions within the same zircon give access to both

the age and the Hf isotopic signature of the magma in which

it crystallized (Griffin et al., 2002; Woodhead et al., 2004). The

latter is converted into a crustal residence (“model”) age

corre-sponding to the time at which the crustal source of the

zircon-hosting magma would have been extractedfrom the DM (Kemp

et al., 2006). The coupled analysis of O isotopes has been

pro-posedto discriminate betweenzircons crystallized frommagmas

having a sedimentary source (

δ

18O

>

6

.

5

h

; Kemp et al., 2006;

Valley et al., 1998) and thus, meaningless “mixed” model ages

(Arndt and Goldstein, 1987), from those having an ultimately

mantle-derivedsource(

δ

18O

∼

5

.

5

±

1

.

0

h

)andsupposedtoyield

reliablemodel ages(HawkesworthandKemp,2006). Accordingly,

statistical analyses of U–Pb–Hf(–O) isotopic databases from both

igneous and detrital zircons were extensively used to quantify

the timing, amount and mechanisms ofcontinental crust

forma-tion through time (Belousova et al., 2010; Dhuime et al., 2012;

HawkesworthandKemp,2006; Lancasteretal.,2011).

Nevertheless,thereliabilityofthisapproachhasrecentlybeen

questioned,especiallybecauseincorporationinthemantleof

con-tinentalcrust-derivedmaterials(subductedsupracrustallithologies

and/oranyfluid/meltgeneratedfromthem)maysignificantlyblur

the zirconHf–O isotope signatures, a problemthat hasnot been

thoroughlytakenintoaccount incrustalevolution models(Payne

et al., 2016; Roberts et al., 2012; Roberts and Spencer, 2015).

Mafic magmas emplaced shortly after continental collision

rep-resent a typical, yet poorly considered case: although of

unam-biguousmantleorigin(Bonin,2004),theydisplay“crustal” whole-rockradiogenicisotopescompositions(Nelson,1992; Turpinetal., 1988) and zirconHfisotopic signatures(e.g.Heilimoetal.,2013;

Laurent and Zeh, 2015; Liu et al., 2014; Siebel and Chen, 2009;

Zhao et al., 2013). Although such mafic magmas are ubiquitous

amongst orogenicsystemssincethe lateArcheanandmaybe the

parentofvolumetricallyimportantgranitesuites(Fowlerand

Hen-ney,1996; Laurentetal.,2013),littleattentionhasbeenpaidsofar

totheirroleintheformationofnewcrust,aswellastheimpactof

theirambiguousisotopicsignaturesoncontinentalevolution

mod-els.

This studyaims toreview the characteristicsandpetrogenesis

ofpost-collisional mafic magmas;propose aunified modelto

ex-plain theirorigin; quantifyto whatextent theycontribute to the

formation of new crust; and address the potential biases

intro-duced by their zircon Hf–O isotopic signature in global,

zircon-based crustal evolution models. We address these issues using:

(i) a new, comprehensive geochemical and zircon Hf–O isotopic

datasetofpost-collisionalmaficmagmasfromtheVariscanFrench

Massif Central, and(ii) a global compilation ofgeochemical data

fromsimilarrocksovergeologicaltime.

2. Post-Collisional Mafic Magmas (PCMM)

2.1. Geologicalsetting

Hereafter,thepost-collisionalperiodisreferredtoasthestage

of the orogenic cycle immediately following the docking of two

continental masses. Voluminousmagmatism, oftenreferred to as

“syn- to post-collisional”(Bonin, 2004),takes placeat that stage.

Thecorrespondingvolcanic/plutonicrocksbelongtotwo

contrast-ingsuites:(i)aperaluminoussilicicsuite,originatingfrommelting ofthelocalcrust;and(ii)amafic-felsichigh-Kcalc-alkaline asso-ciationinwhichthefelsicmaterialisgenerallyregardedasthe re-sultofdifferentiationofthemaficparentalmelt,withvarying con-tributions(assimilation/mixing)fromtheoldercrust(Bonin,2004;

KüsterandHarms,1998; Laurentetal.,2013; Liégeoisetal.,1998;

Moyenetal.,inpress).Inthiscontribution,wefocusonthe

gen-esisofthemaficparentsofthehigh-Kcalc-alkalinesuitethatwill

bereferredtoasPost-CollisionalMaficMagmas(PCMM)hereafter.

PCMM have beenrecognized in mostorogenic systems since the

late Archean,especiallythose correspondingto theamalgamation

of the main supercontinents in Earth history (Table 1; see also

Murphy, 2013). Because oftheir peculiar characteristics,they are

oftenmentionedusinglocal,non-IUGSnomenclature(Table 1)that

hampersacomprehensiveviewoftheirglobalsignificance.

2.2. Casestudy:theVariscanFrenchMassifCentral

The VariscanbeltofwesternEurope(Fig. 1a)wasformed

dur-ing the assembly of Pangea, as a result of the convergence

be-tween LaurussiaandGondwanafrom420to295 Ma.Inthe

Fig. 1. (a)GenerallocationoftheVariscanbeltofEuropeandinferredsuturezones.Abbreviations:GMCGalicia-MassifCentral,STZSaxo-ThurigianZone,RHZRhenohercynian Zone,MZMoldanubianZone.TheyellowstarhighlightsthelocationoftheVelayComplex.(b)GeologicalmapoftheeasternFrenchMassifCentral(Velayarea)redrawnafter Chantraineetal. (1996).Post-collisionalmaficmagmas(triangles)andtheirderivatives(granitebodiesinblue)arehighlighted.Theinsetshowsthelocationofthestudy areawithintheMassifCentral.(Forinterpretationofthereferencestocolorinthisfigurelegend,thereaderisreferredtothewebversionofthisarticle.)

at c. 360–340 Ma and was followed by orogenic extension and

collapse. The latter stage was associated with significant crustal

meltingresulting in widespread granite magmatism andthe

for-mation of granite–migmatite domes (e.g. Velay Complex) (Ledru

etal.,2001andreferencestherein).PCMMarepresentthroughout

the Variscan belt (von Raumer etal., 2014). In the eastern FMC,

they intruded the crust from c. 335 to 300 Ma (Couzinié et al.,

2014; Laurent et al., 2015), forming meter- to hectometer-sized

(monzo)diorite to monzonite bodies in granites/migmatites; and

consist of an assemblage of plagioclase, amphibole, biotite and

clinopyroxene;withorthopyroxene,K-feldsparandquartzin

subor-dinateandvariousproportions(Sabatier,1991).Accessoryminerals

include apatite, allanite,titanite andzircon. Like in other PCMM

occurrencesworldwide(e.g.Laurentetal.,2014a; Murphy, 2013),

thepredominanceofbiotiteand/oramphiboleasmainmafic

min-erals points to high H2O contents in theoriginal magma.In the

FMC, a suite of K-rich calc-alkaline granitoids demonstrably

re-sultsfromdifferentiationofPCCMwithlittleornoinputfromthe oldercrust (Moyenetal., inpress).In thisarea,PCMM andtheir

derivativesrepresent

≈

10% ofthe exposed surface (Moyenetal.,

inpress).

2.3. Samples,methodsanddatacompilation

Inthisstudy,wepresentnewwhole-rockgeochemicaldataand

a new zircon Lu–Hf and O isotope dataset on PCMM from the

eastern FMC (Fig. 1b). The whole-rock major and trace element

compositions of 27 samples are reported in the Supplementary

Material (Table S1). The Hf isotopic composition of zircons from

12sampleswasmeasuredby laserablation–multicollector –

in-ductivelycoupledplasma–massspectrometry(LA-MC-ICP-MS)at

Goethe Universität Frankfurt, Germany, after imaging of internal

zircon structures (cathodoluminescence and back-scattered

elec-tron images) by scanning electron microscopy. The zircons were

previouslyanalyzedforU–PbisotopesbyLA-ICP-MSforage

deter-mination(Laurentetal.,2015).Outofthe12samplesinvestigated

forHfisotopes, zircons from8sampleswere alsoanalyzed forO

isotopesbySHRIMPattheUniversityofGranada,Spain.Thenature

andagesof theinvestigated samplesare detailedinTable 2;the

SupplementaryMaterialcontainsdetailsaboutanalyticalmethods,

resultsof standardmeasurements (Table S2 for HfandS4 forO)

andthecompletedatasets(Table S3forHf,Table S5forO).

We also present a compilation of data from PCMM

world-wide, available in the Supplementary Materials and including

whole-rock geochemistry (Nsamples

=

1647; Table S6) and zircon

Hf (Nspots

=

2425; Table S7) and O (Nspots

=

633; Table S8)

iso-topes.

3. Whole-rock geochemistry and Hf–O isotope composition of

PCMM

3.1. Whole-rockgeochemistry

PCMMall sharesimilargeochemicalfeatures (Fig. 2).Theyare

mafic to intermediate (44 to 63 wt.% SiO2) and dominantly plot

in the high-K calc-alkaline to shoshonite fields (Fig. 2a). PCMM

typicallyshowadualgeochemicalsignaturecharacterizedby

rich-ness in both compatible and incompatible elements (Fowler and

Rollinson, 2012; Holub, 1997; Laurent et al., 2014b). In PCMM

from the FMC, this is best illustrated by high contents in both

K2O(1.0–7.5 wt.%)andFeOt

+

MgO(7–25 wt.%)(Fig. 2b).Likewise,

bothBa

+

Sr (0.1–0.6 wt.%) andNi

+

Cr contents (50–1000 ppm)

areveryhigh(Fig. 2c).Notably,thereisnocorrelationwhatsoever

betweencompatibleandincompatibleelements(Fig. 2b,c).PCMM

displayhighLREEandHFSEcontents(Fig. 2d),togetherwith

crust-likeincompatible trace elementpatternssystematically

character-izedbynegativeNb–TaandTianomalies(Fig. 2e).Strikingly,these

rocksaremarkedlyenrichedinmostincompatibleelements (LILE,

LREE,HFSE)relativetothebulkcontinentalcrust(Fig. 2e).

3.2. Hfisotopes

Acompilation ofavailable zirconHfisotopic datafromPCMM

invarious orogenic systems ispresented inFig. 3a. Apartfroma

fewexceptions,suchzirconsdisplay(sub-)chondritic

ε

Hf(t),clearly Ta

Fig. 2. Maingeochemicalfeaturesofpost-collisionalmaficmagmaticrockswithspecialemphasisontheeasternFMCexample(bluetriangles).Eachgreyfieldillustratesthe rangeofcompositionforsimilarrocksfromotherorogenicsystems(datafromTable S6).Plotsof(a)K2Ovs.SiO2concentrations(thedifferentseriesareafter

Peccerillo

andTaylor,1976);(b)K2Ovs.MgO+FeOtconcentrations;(c)LargeIonLithophileElements(Ba+Sr)vs.transitionelement(Ni+Cr)concentrationsand(d)sumofLREE (La+Nd+Ce)vs.HFSE(Zr+Nb)normalizedtoYb;(e)Multi-elementdiagramnormalizedtothePrimitivemantle(McDonoughandSun,1995);thefieldscorrespondtothe 1stto3rdquartilerangeofsamplesfromtheeasternFMC(inblue)andeachotherorogenicsystem(ingrey),respectively.BCCstandsfortheBulkContinentalCrust(values from

Rudnick

andGao,2003),OIBforOceanIslandBasalts(valuesfrom

Sun

andMcDonough,1989),N-MORBforNormal-typeOceanRidgeBasalts(valuesfrom

Sun

and McDonough,1989).(Forinterpretationofthereferencestocolorinthisfigurelegend,thereaderisreferredtothewebversionofthisarticle.)

differing fromthat of DMatthe time of magma formation.This

typicalfeatureofPCMMiswellreflectedbyrocksfromtheeastern

FMC(Fig. 3bandTable 2).Zircongrainsfromsuchrocksshow

sim-pleinternal structures,withoutcore–rimrelationships andyielda

singlepopulation ofconcordantU–Pb ages(Couzinié etal.,2014;

Laurentetal.,2015).Thisarguesforasingleeventofzircon

forma-tionduring magmacrystallizationandlimitedornohost-rock

as-similation.Hence,theinitialHfisotopiccomposition(176Hf/177Hft),

canreliablybe consideredasrepresentativeofthatofthemagma

atthetimeofzirconcrystallization.

Thenewly-obtainedzirconHfdatasetforPCMMfromtheFMC

Fig. 3. (a)Compilationofavailablein-situ HfisotopedataforzirconsfromPCMMworldwideasafunctionoftheirintrusionage(notethatthetime-scaleislogarithmic). TotalnumberofzirconHfmeasurements=2425.TheVariscanzircons(box 3)areshownindetail

Fig. 3

b.TheεHf(t)rangeforthedepletedmantlereservoirisbracketed bythemodelsof

Naeraa

etal. (2012)(lowermostvalue)and

Griffin

etal. (2002)(uppermostvalue).DatafromTableS7.(b)MeasuredzirconεHf(t)inpost-collisionalmafic rocksfromtheVariscanbeltofEuropeasafunctionofintrusionage.Additionaldatafrom

Siebel

andChen (2009)and

Villaseca

etal. (2010).

thefouroldestsamplesoftheinvestigateddataset(309–320 Ma).

The zircon 176Hf/177Hft ratios of samples from subset A range

from0

.

282319

±

0

.

000032 to0

.

282480

±

0

.

000028 (2S.E.–

stan-dard error), corresponding to subchondritic

ε

Hf(t) of

−

3.7 to

−

9.4 (Fig. 3b). Subset

B includes

the younger samples, all

hav-ingagesintherange299–310 Ma.Theirzirconshave176_Hf/177_Hf t

ranging between0

.

282441

±

0

.

000026 and 0

.

282562

±

0

.

000026

(2 S.E.) corresponding to subchondritic

ε

Hft of

−

1.1 to

−

5.0

(Fig. 3b). The

ε

Hft variability within a givensample is small

(al-ways

<

3

ε

Hf-units) and in most cases close to the analytical

uncertainties (i.e.

±

1.1

ε

Hf-units), indicating that zircons

crystal-lized froma magma withhomogeneous Hf isotope composition.

Although some overlap occurs between the two groups, zircons

fromsamplesofsubsetAhavemorevariable(thescatterisalways

>

4

ε

Hf-units for a given sample) and generally less radiogenic

176_Hf/177_Hf

t thansamplesfromsubsetB,withmostgrainshaving

ε

Hft lower than

−

5 anddown to

−

9.4 (Fig. 3b). Becauseof this

large dispersion, the average

ε

Hft calculated for each sample of

subsetAisprobablyoflittlegeologicalsignificanceandthescatter

mayrathercorrespondtoheterogeneousHfisotopecompositionof

theoriginalmagma.

3.3. Oxygenisotopes

Fig. 4a summarizes available insitu O isotopes measurements

onPCMMzircons.The

δ

18OSMOWofzirconsfromPCMMarehighly

variable and range from

+

2.5 to

+

10

h

. They are often out of

the range of

δ

18OSMOW expected for zircons crystallizing from

mantle-derivedmagmas,especiallyinthemostrecent(Himalayan,

VariscanandCaledonian)orogens(Fig. 4a).

All zircon data from PCMM of the eastern FMC show high

δ

18OSMOWvalues(Fig. 4b),mostlybetween6

.

39

±

0

.

32 and8

.

92

±

0

.

26

h

(2S.E.–standarderror)withafewspotswithevenhigher

δ

18OSMOWupto9

.

97

±

0

.

24.Thevariabilitypersampleistypically

large,

>

0.6

h

(2 S.D.), i.e.well above the range ofanalytical un-certainties(

<

0.3

h

;seeTable S5).ThisarguesforheterogeneousO

isotopecompositionsofthemagmasatthetime ofzircon

crystal-lization.

4. Discussion

4.1. PetrogenesisofPCMMintheFMC

4.1.1. Sourcevs.emplacementprocesses

The most mafic components of the PCMM suite in the FMC

displaylowSiO2(

<

53%),withhighMgO

+

FeOt(

>

12%)and

transi-tionelement(Cr

>

250 ppm.Ni

>

120 ppm)contents.This

demon-strates that the parental melt cannot be derived from crustal

lithologies, but rather equilibratedwith a(n) (ultra)mafic residue

(i.e.peridotite/pyroxenite)(Holub,1997; Sabatier,1991).Fromthat

perspective, the high incompatible element contents and

“crust-like” Hf–O isotopic signatures of PCMM could result from two

distinct processes: (i) evolution ofan incompatible element-poor,

mantle-derived basaltic melt by fractionation and/or assimilation

ofcrustallithologies(i.e.AFC processes);or(ii)partialmeltingof amantlesourcealreadyenrichedinK2O,H2OandLILE.

Severallines ofevidence indicatethat AFC cannot explain the

geochemicalsignatureofPCMM:

(1) Even starting from an “enriched” basaltic composition (i.e.

OIB-like), considerable amounts of crystallization (

>

50%) are

requiredtodrivetheincompatibleelementcontentstothe

el-evated concentrationsobserved inPCMM, such that thefinal

meltwouldcertainlynotbemaficanymore(Fig. 5a).

(2) Assimilationofcontinentalcrust duringmagmaemplacement

is precluded by the lower incompatible element contents of

themostwidespreadcrustallithologiescomparedwithPCMM,

which holds true forboth the FMC case andthe global

per-spective(insetFig. 5a).

(3) Even assuming that a crustal contaminant with the

ade-quate trace element composition exists in the FMC andthat

onlylimitedfractionation (

≤

20%)took placeso thatthefinal

magma remained mafic, AFC calculations show that driving

theHfisotopiccomposition ofaDM-derivedbasalttothatof

thePCMMwouldrequireacrustalcontaminantwith

ε

Hf310 Ma

of

−

50 (Fig. 6a), corresponding to an early Archean crustal

component(modelage

>

3.6 Ga).Thisisextremelyunlikelyin

thecaseoftheFMC wheretheoldestcrust isearlyPaleozoic

Fig. 4. (a)Compilationofin-situ Oisotopedataforzirconsofpost-collisionalmaficrocksfromworldwideoccurrences.Totalnumberofzirconmeasurements=633.Data fromTableS8.Theδ18_{Oforthemantleisfrom}

_Valley

_{etal. (1998).Arrowsforsupra-crustalweatheredmaterialsandridge-alteredmaterialsaccordingto}

_{Bindeman (2008)}

_. (b)ZirconoxygenisotopecompositionofPCMMfromtheeasternFrenchMassifCentral.Foreachsamplearedisplayedthemeanand2S.D.confidencelevel.Errorbarsare theanalyticalstandarderrorsofeachmeasurementplottedat2σ.

In contrast, PCMM plausibly represent primary melts derived

fromananomalouslyK2O-,LILE- andH2O-richmantle.Indeed,

ex-perimentalmelting of phlogopite- and/or amphibole-bearing (i.e.

K-rich)peridotites,particularlyspinel-lherzolites,yieldsmeltswith

major element compositions matching those of PCMM from the

FMC (Fig. 5b) (Conceição and Green, 2004; Condamine and

Mé-dard, 2014; Green, 2015; Mengel and Green, 1989; Thibault et

al., 1992). Importantly, experiments conducted with mixed

peri-dotite

+

felsic(granitoid)starting materials producemeltsof sim-ilarcompositions(Malliketal., 2015; Prouteauetal.,2001; Rapp

etal., 2010). The two sets of experiments are equivalent from a

petro-geochemicalpoint ofview because,inthesecondapproach,

the mixture yields a phlogopite- and/or amphibole-bearing

peri-dotite/pyroxenite inequilibrium witha mafic K-rich melt.

There-fore, experimental evidence unequivocally shows that

hybridiza-tionbetweenperidotite andafelsic componentisrequiredto

ex-plainthemajor-elementcompositionofPCMM(irrespectiveofthe

hybridizationmechanismandthe exactnature ofthe felsic

com-ponent,whicharediscussedin§4.1.2).

Thesameconclusioncanbedrawnfortraceelements. We

cal-culatedtheREEcontentsofameltgeneratedfromahybridsource

consistingof 75–90% DMand 10–25% ofa LREE-rich component

(having the composition of the bulk continental crust). Fig. 5c

showsthat the resultingmelts (for5 to 20% melting) matchthe

compositionofPCMMfromtheFMC,againwithabetterfitfora

spinel-faciessource.

Therefore,we proposethatPCMMintheFMC resultfrom

par-tialmelting, at

<

2 GPa,of aphlogopite-/amphibole-bearing

peri-dotite (or pyroxenite) corresponding to mantledomains enriched

inK2O,H2Oandotherincompatibleelements.

4.1.2. Mechanismofmantleenrichmentandnatureofthemetasomatic

agent

Considering that they reflect the composition of the hybrid

mantlesource,theobservednegative

ε

Hft signaturesinPCMM

zir-conscanbeexplainedbytwodistinctscenarios:(i)contamination

of themantle (with

ε

Hft

>>

0) by crustal materialwith strongly

negativeHfisotopecomposition(

ε

Hft

<<

0)shortlypriortoPCMM

formation;or(ii)alongtimespanbetweenmantlecontamination

(by any material having low 176Lu/177Hf) and PCMM formation,

wherebytheinitialdecreaseofthemantle176_Lu/177_Hfratioleads

toatime-integrated,non-radiogenicHfisotopecomposition.

Several lines of evidence clearly favor the first hypothesis.

Firstly, the mantle beneath the FMC (sampled as xenoliths in

Cenozoic volcanoes) do not show evidence for any geological

event older than ca. 0.6 Ga (Wittig et al., 2007), which is also

the age of the oldest autochtonous crust (Melleton etal., 2010).

In contrast, the strongly negative

ε

Hft of PCMM zircons in the

FMC would require the involvement of much older mantle, of

at least Mesoproterozoic age (

>

1.1 Ga, considering an unlikely

176_Lu/177_Hf

₌

_{0 for the source). Secondly, prior to the}

emplace-ment of PCMM in the late Carboniferous (330–300 Ma), other

maficmagmaswereemplacedthroughouttheFMCduringthelate

DevonianandearlyCarboniferous(380–360 Ma)andcontrastingly

show positivebulk-rock

ε

Ndt (

+

1to

+

5,upto

+

8) (Pinand

Pa-quette,1997, 2002),indicativeofa depletedmantleorigin. Ifany

FMC, it is difficult to envisage that it would have been sampled

by magmatism at ca. 330–300 Ma, but not at 380–360 Ma,

es-pecially since “enriched” upper mantle domains are particularly

prone to melting (solidus temperatures as low as 1025–1075◦C

at 1 GPa; Conceição and Green, 2004; Condamine and Médard,

2014). Thirdly, PCMM of the FMC only post-date by 50–100 Ma

an episode of Silurian–Devonian tectonic accretion and

subduc-tion of continental/oceanic units belonging to the northern

mar-gin ofGondwana (Faure etal., 2008; Melletonetal., 2010). Such

units,nowexposedasmigmatiticortho- andparagneisses

contain-ing eclogitic relics, were buried at depths

>

100 km and

under-wentHPpartialmelting(Faureetal.,2008; Lardeauxetal., 2001;

Pin andLancelot,1982),thereby providing a straightforwardway

to metasomatize the mantle with crustal-derived K-, H2O- and

LILE-rich material. The samemodel was proposed to explain the

occurrence of PCMM in the Variscan Bohemian Massif(Janoušek

andHolub,2007),theHimalayas(Campbelletal.,2014; Guoetal.,

2014) andtheDabie-Suluorogen(Zhaoetal.,2013).

Importantly, mass balance calculations show that binary

mix-ing between (i) 80–90% DM and (ii) 10–20% of a crustal

com-ponent with

ε

Hft

= −

10 and

δ

18O

= +

20

h

readily explains the

Hf–O isotopiccomposition ofPCMMzircons intheFMC (Fig. 6b).

High

δ

18_{Oistypicallyobservedinsubductedcontinental}

sedimen-tary materials havingexperienced a low-T weatheringcycle, e.g.

the EastIndonesian (

δ

18O

= +

15 to

+

24; Vroon etal., 2001) or Lesser Antilles (

δ

18O

= +

19 to

+

21; Davidson, 1987) sediments.

Moreover, the pre-Variscan crust of the FMC is dominated by

paragneisses(andmeta-granitoidsderivedfrommeltingofthe

lat-ter) corresponding to former sediments of the north Gondwana

margin, which are characterized by Archean (2.5–3.3 Ga),

Paleo-proterozoic(1.8–2.2 Ga)andNeoproterozoic (0.5–0.7 Ga)detritus,

as indicated by age patterns of detrital zircon grains (Albert et

al., 2015; Linnemann et al., 2014). Fig. 7a shows that a mixture

of these three crustal components, including at least 45–85% of

Paleoproterozoic–Archean crust,wouldreadilyaccount forthe

in-ferred Hf isotopic signature (

ε

Hf310 Ma

= −

10; TDM

=

1.8 Ga) of

thecrustal materialinthehybridmantlesource.Infact,thebulk

Hf budget carriedby detrital zircons in thepre-Variscan crust of

the FMC is consistently characterized by an average

ε

Hf310Ma of

ca.

−

10 (Fig. 7b). This observation clearly supports that the

hy-brid source of PCMM in the FMC results from interactions

be-tween mantle peridotite and subducted pre-Variscan continental

shelfsedimentsoftheformernorthGondwanamargin.

Furtherconstraintsonthenatureofthecrust-derivedmaterials

andhow they wereincorporated inthemantle canbe placedby

consideringbulkrockTh/Laratios,whicharenotsignificantly

frac-tionated during mantle melting(Plank, 2005). Indeed,the source

of PCMM samples with the highest Th/La and

δ

18O and lowest

ε

Hft (SubsetA) was likelymetasomatized byfelsic meltsderived

fromterrigenous sediments(Fig. 6c,d), which is consistentwith

the fact that subducted rocks in the FMC underwent meltingat

HP (Faure et al., 2008; Pin and Lancelot, 1982). The other

sam-ples (SubsetB) are better explainedby mixing betweenDM and

abulkcontinentalcrustcomposition(Fig. 6c,d),possiblyresulting

frommechanicalmélangesduringthesubductionofthe

continen-talunits.Importantly,compositionscorrespondingtoaverage,bulk

lower crust (lowTh/La) cannot account forthe observed

compo-sitions,evenifweconsiderthat suchmaterialhadhigh

δ

18Oand

low

ε

Hft (Fig. 6c,d),showingthatsubductionofsupracrustal

ma-terialisrequired.

4.2. AunifiedmodelforPCMMpetrogenesis

As outlined earlier, PCMM worldwide share a common

geo-tectonic, petrographic identity andsimilar geochemicalandHf–O

isotopiccharacteristics,whicharguesforacommonorigin.Thefact

that our samplesfrom the FMC are chemically representative of

thediversityofPCMMworldwide(Fig. 3)suggeststhatthegeneral

featuresofthepetrogeneticmodelproposedabovecanbe

reason-ablyextendedtoallPCMMsuites.

Consistently,many workersadvocatedthat the particular

geo-chemistryofPCMMnecessarilyrequiresanenrichedmantlesource

(FowlerandRollinson,2012; Guoetal.,2014; Laurentetal.,2011;

Laurent etal., 2014a; Laurentetal., 2014b; Liégeois etal., 1998;

Murphy, 2013; Prelevic et al., 2012; von Raumer et al., 2014;

Williamsetal.,2004).Moreover,thepresenceofcontinental

crust-derived material in such a source has often been proposed to

explain the non-DM and “crustal” isotopic signatures of these

rocks (Guo et al., 2014; Laurent et al., 2014b; Nelson, 1992;

Prelevic et al., 2012). Given the generally higher

δ

18Oof PCMM

zircons compared with mantlematerial (Fig. 4a), it is likelythat

thesecrustalmaterialsaremostoftenrepresentedbysupracrustal

rocks. Finally, mantle enrichment shortly prior to PCMM genesis

seemstobetherulebecause,likeintheFMC,PCMMemplacement

systematicallypost-datesby

<

50 Maanoceanicand/orcontinental subductionstage,forinstanceinthelateArcheanterranes(Laurent

etal., 2014a) andinthe Himalayan (Chung etal., 2005),

Caledo-nian (Atherton andGhani, 2002) andSvecofennian (Andersson et

al.,2006) orogens.

Therefore, we propose that PCMM are derived from: (i)

in-teractions between subducted continental material (most often

of supracrustalorigin), or melts/fluids generated fromthem, and

mantle peridotite; and (ii) subsequent melting of the resulting

hybrid source shortly thereafter. Mantle enrichment through the

subductionofcontinentalunitsattheonsetofcollision(Campbell

etal., 2014; Guoetal., 2014; JanoušekandHolub, 2007; Zhaoet

al., 2013) wouldbe themostappropriate scenario toaccount for

theshorttimelapsebetweenthetwosteps;thesystematic

occur-rence of PCMM in collision belts;and themore evolvedisotopic

composition andgreater incompatibleelement contentsofPCMM

compared with classical arcmagmas (related to oceanic

subduc-tion).

4.3. Quantifyingcrustandmantlecontributionsinpost-collisional

maficmagmas

Theaimofthissectionistoquantifytherespectivemass

frac-tionofmantle- vs. crust-derivedelements inthe“hybrid” mantle

sourceofPCMM.Thiswillbe usedinturntodeterminethemass

Fig. 7. (a)Triangulardiagram showingthe modelage andεHf310 Ma (inbrackets)ofavirtualsedimentary crustalmaterialas afunctionoftherelativeproportionof Archean,Eburnean(Paleoproterozoic)andPan-African(Neoproterozoic–EarlyPaleozoic)crustinthelatter,assumingidenticalHfconcentrationsinthethreeend-members. Thehighlightedmodelageof1.8 GacorrespondstothecompositionofthecrustalmaterialrequiredtoexplainthezirconHfisotopiccompositionsofPCMMintheFMC

(εHf310 Maofca.−10;see

Fig. 6

andtextfordiscussion).(b)HistogramshowingthedistributionofzirconεHf310 Mafor>360Ma-oldgrainsfromallgneissunitsinthe

easternFMC,takenasrepresentativeoftheisotopicdistributionofHfinthepre-Variscancrustat thetimeofPCMMformation.TheεHf310 Maarecalculatedusingthe measured176_Hf/177_Hfand176_Lu/177_{Hfratiosofeachzircongrain.Datafrom}

_{Chelle-Michou}

_{etal. (2015).}

fractionofPCMMthatderivesfromthemantle,whichisacritical

parametertounraveltheircontributiontocrustalgrowth.

Asdemonstrated above,the hybridmantlesourceof PCMMis

formedbyinteractionsbetweenmantleperidotiteandcrustal

ma-terials(fluid,meltorsolid).Theseinteractionsmayresultin

differ-ent cases in nature, e.g. a mechanical mixture of peridotite and

meta-sedimentary/meta-igneous crustal rocks; a peridotite

con-taining veins formed by reaction with percolating liquids; or a

metasomatized rock where the added material is distributed in

newly-crystallizedphases(pyroxenes,amphibole,phlogopite).

Irre-spectiveofthenatureofthemetasomaticagent,thefinitemassof

each chemical element within anyof these“hybrid” sources will

be partitioned betweenthe fraction that was already presentin

theoriginal mantlepriortotheinteractions andthatsupplied by

theaddedcrust-derivedmaterial.ThedualgeochemistryofPCMM

showsthatthey sampledbothgroupsofelements, regardlessthe

exactphysicalnatureofthesourceandmeltingprocess.Therefore, foreachchemicalelement,themassfractionof“crust”vs.“mantle”

atomsinthesourceofPCMMcanbeapproximatedusingasimple

binarymixingmodelbasedonmass-balance,assumingreasonable

compositionsforthetwocomponents(respectivelyBulk

Continen-tal Crust and DM). Although simplistic, this calculation has the

advantageofrelying onfewparameters, i.e.themassfractions of

thetwo end-membersandtheir compositions, which inthe case

ofcrust–mantlemixingare notsocriticalgiventhehuge

concen-trationcontrastbetweenthetwoformostelements.

Foreachchemicalelement,wedefinexcrustasthemassfraction

ofatomsthatoriginatesfromcrust-derivedmaterials.Fig. 8shows

that ina hybridmantlecontaining 10 to25% crustal component,

xcrust variesforeachelement,incompatible elementsbeing

domi-nantlycontrolledbythecrustalcomponent(xcrustbeingashighas

99% forRb, Ba, Th,Uasthe initial mantlecontains very little of

these)whilst compatible elements mostly comefrom themantle

(e.g.97% oftheMgO,Cr andNi isofmantleorigin).Importantly,

the fivemost abundant major oxides inthe source (SiO2, Al2O3,

MgO,FeO,CaO;ca.95%ofthebulkmass)aredominantlyofmantle

origin(xcrust isgenerally

<

30%)becauseofthelimited

concentra-tioncontrast(less thanone orderofmagnitude) betweenmantle

andcrust andthelow massfractionofthe crustalcomponentin

the hybridsource(Fig. 8).As adirectandcriticalconsequenceof this,assumingthattheproportionof“crust”vs.“mantle”atomsfor

eachelementdoesnotchangeduringmeltingofthehybridsource,

thebulkmassfractionofPCMMthatderivesfromthecrustal

com-ponent is low. Typically,for three representative PCMM samples

(fromaLate-ArcheanterraneandtheVariscanandHimalayan

oro-gens),itliesbetween15–21%and31–38%,forrespectively10%and

25%ofcrustalmaterialsinthehybridsource(seeFig. 8),meaning that

>

62%ofthebulkmassofthesesamplesareofmantleorigin

andthusrepresentnewadditionstothecrust.

Thisfirst-ordercalculationimplicitlypresumesthat“crust”and

“mantle”atomsofeach elementareequallydistributedover

min-eral phases in the hybrid source so that xcrust keeps constant

during melting. This is likely over-simplistic for trace elements,

especially the dominantly crust-derived elements that would be

preferentially partitionedintothephasesformedvia metasomatic

reactions such aspyroxenes,amphibole andphlogopite(Prouteau

et al., 2001; Rapp etal., 2010), whichin turnpreferentially

con-tributetomeltingreactions(CondamineandMédard,2014; Mallik

etal., 2015). However,formajoroxides, xcrust likelyremains

con-stantthroughouttheprocess.Indeed,metasomaticreactionswould

randomly use both “crust” and “mantle” elements to form new

minerals and, even if not, solid-state diffusion, phase

recrystal-lization and mixing enhanced by high-temperature deformation

(Linckensetal.,2014) wouldsmooth anysignificantdifference of

xcrustfromamineraltoanother.Moreover,evenconsideringan

ar-tificial two-fold increase in the proportion of “crustal” SiO2 and

Al2O3 betweenahighly-enrichedmantlecomposition(25%crustal

component)andthemeltowingtotheirpotentialconcentrationin

newlyformedphlogopite,amphiboleandpyroxene,themass

frac-tionofcrust-derivedelementsinthethreePCMMsamplesusedfor

calculation would not exceed 50–63%.These are definitely

maxi-mumvaluessincethisscenarioissomewhatextreme.Therefore,it

isreasonable to concludethat theproportionofreworked crustal

materialsinPCMMdoesnotexceed63%andismostoften,as

cal-culatedearlier,between15and38%.

Paradoxically,mostPCMM(especiallythethreespecimensused

in the calculations) display markedly negative zircon

ε

Hft. Then,

cor-Fig. 8. Plotofthe“xcrust”parameterformajoroxides(panel(a),sortedasafunctionoftheirabundanceinPCMM)andtraceelements(panel(b),reportedbyincreasing compatibility).xcrustisdefinedasthemassfractionofanyelementi controlledbythecrustalcomponentinahybridmantlesourceconsistingofDepletedMantle(Workman andHart,2005)+10(blackline)to25%(greyline)crustalmaterial(BulkContinentalCrust,

Rudnick

andGao,2003),calculatedasfollows:xi

crust=fcrust[i]crust/(fcrust[i]crust+

(1−fcrust)[i]DM)where[i]depictstheconcentrationofelementi andfcrusttheproportionofcrustalcomponentinthehybridsource.Thebluefieldrepresentsthebulk

fractionofcrust-derivedelements( Xcrust)inmaficmagmasderivedfromsuchasource,calculatedforthreerepresentativesamplesofPCMMsuites:533-2(thisstudy); MAT-60(Laurentetal.,2014b);and08YR05(Liuetal.,2014). Thecalculationhasbeenperformedby(i)weightingtheconcentrationsofallelementsineachsamplebythe correspondingvaluesofxcrust,and(ii)summinguptheresultingvalues.Seetextfordiscussion.Xcrust=ni=1

 xi

crust·[i]PCMM

 .

respondtopure reworking offormercrustal lithologies. This

dis-crepancysimply results from the large dominance of the crustal

componenton the incompatible trace element budget (including

theradiogenictracersSr,Nd,Hf,Pb)inthehybridsource(Fig. 8).

Typically,theHfbudgetinthesourceofPCMMconsistsof72–89%

non-radiogenic“crustal” Hfand only11–28% radiogenic“mantle”

Hf. Therefore, as long as incompatible element tracers are

con-sidered, PCMM wouldbear a “crustal” signature andthe mantle

contributiontotheir origin, althoughdominantintermsofmass,

wouldbeinvisible.

4.4.Consequencesforcrustalgrowth

Asdemonstratedearlier(Table 1), PCMMhavebeen emplaced

duringorattheendofcontinentalcollisionsincethelateArchean

(Bonin, 2004; Fowler andRollinson, 2012; Laurentet al., 2014a),

specificallyateachstageofsupercontinentamalgamation.Theyare

thereforeasystematicmagmaticfeatureoftheorogeniccycle,just

asMORB,arcmagmasorcollisionalgranitesare,andtheir

signifi-canceforglobalcrustalevolutionmustbeaddressed.

Although the surface expression of PCMM is limited, it has

been demonstrated in several settings that they would actually

representtheparental magmas ofmuch morevoluminous

grani-toidsuitesviafractionation(FowlerandHenney,1996; Küsterand

Harms,1998; Laurent et al., 2013; Moyenetal., in press) and/or

limitedinteractionwithcoevalcrustalmelts(Clemensetal.,2009;

Janoušek et al., 2004; Laurent et al., 2014a; Parat et al., 2009).

In terms of mass balance, these two processes are antagonists:

silicic liquids will represent only a fraction of the initial mass

ofmafic magma inthe caseof fractionation,whereas interaction

withcrustalmaterialrepresentsamassadditiontoPCMMtoform

granitoids.Hence,itisreasonabletoconsiderthatinagiven

oro-genicsystem,the initialvolumeofPCMMwas roughly equivalent

tothatoftheir geneticallyrelatedgranitoids.Accordingto Moyen

etal. (inpress),PCMM-derivedgranitoidsrepresentca.10%of

ex-posedrocksintheVariscanbelt;giventhat62to85%ofthebulk

massofPCMMconsistofelementsextractedfromthemantle(see

section4.3),andassumingthatsurfaceexposurereflectsthe repar-titionoflithologiesatdepth,then6.2to8.5%ofthebulkmassof

the Variscan crust represents new additions from the mantle by

PCMM.

At first glance, this estimate appears modest in comparison

witharcsettings,in whichthe large majorityofthe bulk crustal

volumeisderivedfromthedifferentiationofmantle-derivedmafic

magmas (Jagoutz and Kelemen, 2015). However, mantle-derived

magmatic rocks only become long-term additions to the

conti-nental volume ifthey are subsequentlypreserved: while a small

fraction of arc granitoids may be captured in collisional orogens

(Condieetal.,2011),themajorityofthemarereadilyrecycledinto

the mantle shortlyafter their formation owing to sediment

sub-ductionand/orsubductionerosion(SternandScholl,2010).In

con-trast,PCMMandderivativesrepresentsmallvolumesofonlypartly

mantle-derivedmagmas,butbecausetheyareshieldedinthecore

of newly formed continental masses, they have a much greater

preservationpotentialinthegeologicalrecord(Hawkesworthetal.,

2010). Asa result, they may representsignificant contributors to

long-termcrustalgrowth.

Recently,theproblemofcrustalgrowththroughtimehasbeen

extensively addressed via the statistical analysis of globalzircon

Hf (Belousova etal., 2010; Condie etal., 2011) or coupled Hf–O

(Dhuimeetal., 2012) isotopicdatabases: Hfmodelages areused

todeterminethetimingofcrustformation,andOisotopesto

dis-tinguishzirconsformedingranitoidsreworkinga“juvenile”source (

δ

18O

=

5

.

5

±

1

.

0

h

) or a “mixed” sedimentary source (

δ

18O

>

6

.

5

h

) (e.g.HawkesworthandKemp,2006).The reliabilityofthis

approachhasbeenquestionedrecently,especiallybecauseHfand

Oisotopesinzirconreflectinmostcasesthecontributionof

multi-plecomponentsintheoriginoftheirhostmagma(magmamixing,

heterogeneoussources,crustalcontamination)(Payneetal., 2016;

Roberts and Spencer, 2015). The case of PCMM raises an even

moreproblematicissue,showingthatigneousmaterial

represent-inganon-negligiblecontributiontolong-termcrustalgrowthmay

nothaveDM-likeHfisotopecompositionsand“pristinemantle”O

isotope compositions because ofthe small proportions ofcrustal

(possiblysedimentary)materialintheirhybridmantlesource.

Several lines of evidence suggest that, despite their limited

volume, zircons from PCMM andderivatives representa sizeable

proportion of global detrital zircon databases. First, because of

their selective preservation (Hawkesworth et al., 2010), zircons

from collision-related igneous rocks dominatethe detrital

popu-lations insedimentaryrocks wherever recentorancientorogenic

crust ispresentinthecatchmentarea (Spenceretal., 2015).

Sec-ond, because of their high Zr contents (370 ppm in average in

the eastern FMC), rocksfrom the PCMM suitehave a higher

zir-confertilityfactor(“ZFF”ofDickinson,2008)thanothercollisional

Third,magmatic zirconsfromtheserocks showa greater

propor-tion(

∼

80%)ofconcordantU–Pbagescomparedtothoseofpurely

crust-derived (“S-type”) granites (

∼

50%) (Couzinié et al., 2014;

Laurent et al., 2015), such that in a detrital zircon population,

theprobability todiscardzirconsfromPCMMandderivatives

be-causeofU–Pbdiscordance(NemchinandCawood,2005) issmaller

thanforS-typegranites.Toillustratethetwolatterissues,the ero-sionoftheVariscancrust,assumingthatthelatterconsistsof10%

PCMM-derivedgranitoids(ZFF

=

2.5;80%U–Pbconcordance);30%

crust-derivedgranites (ZFF

=

1; 50% U–Pbconcordance); and60%

of metamorphic and sedimentary rocks (ZFF

=

1; 67% U–Pb

con-cordance), would yield a detrital zircon population in which the

final, U–Pb concordant datasetcontains a minimum of 27% data

fromPCMM-derivedmaterial–notwithstandingthelikely

possibil-itythatmetamorphic andsedimentaryrocks maycontain zircons

fromreworked, earlierPCMMandrelatedgranites,exactlyforthe

samereasons.

Therefore, zircons derived from PCMM, their differentiated

productsandanyigneousandsedimentarymaterialderived from

them,wouldsignificantly biasthe crustalgrowth signal ofglobal

detritalzircondatabases,intwodistinctways:

(1) ZirconsfromanyPCMMcrystallizedattimet wouldhavebeen

eitherdiscarded(non-mantle

δ

18_{Oresultingfromthe}

involve-ment of sedimentary components) or attributed to ancient

eventsofcrustformationbecauseofnon-radiogenicHfisotope

compositions(andoldHfmodelages).Althoughthelatter

fea-turecertainlyreflectsreworkingofoldercrusttosomeextent,

thisreworkingisover-estimatedbecausethemantle

contribu-tion at time t cannot be resolved by Hf isotopes alone (see

section4.3).

(2) If PCMM-derived crustal material is reworked by later

mag-maticevents,thenthecalculationofmodelagesbasedonDM

forzirconsformedintheseyoungermagmas isinappropriate

sincetheircrustal sourcehadanon-DMHfisotopic

composi-tion.

Both would have the same global effect, i.e. skewing crustal

growth models towards an over-estimation of ancient crustal

growth withrespect to younger crust formation. Specifically,

be-causePCMMarepresentonEarthsincethelateArchean(Table 1),

post-Archean crustal growth rates may be significantly

under-estimated.

5. Conclusions

1. PCMMareubiquitousinallcontinentalcollisionsettingssince

thelateArcheanandarecharacterizedbyadualgeochemistry,

withhigh contents in both compatible, “mantle-hosted”

ele-ments (FeOt

+

MgO, Ni, Cr) and incompatible, “crust-hosted”

elements(LILE,REE,HFSE).

2. Asillustratedbythecasestudyfromtheeastern FMC,PCMM

originated from low-degree partial melting of a

phlogopite-(and/oramphibole-)bearingperidotite/pyroxenite.IntheFMC,

thenon-radiogenic

ε

Hf(t) (

−

2to

−

9) andelevated

δ

18OSMOW

(

+

6 to

+

10

h

) of PCMM zircons suggest that this mantle

source was contaminated by 10 to 25% of subducted

conti-nentalshelfsedimentsderivedfromPrecambriancrust,shortly

priortomelting.

3. Massbalancecalculationsindicatethattheincompatibletrace

element budget (including radiogenic isotope tracers Sr, Nd,

Hf,Pb)ofthePCMMsourceismainlycontrolledbythecrustal

component. Therefore, resulting magmas display “crust-like”

non-radiogenic isotope compositions even though 62 to 85%

oftheirbulkmassoriginatesfromthemantlecomponentand

representsnewadditiontothecrust.

4. PCMM have contributed to crustal growth since the late

Archean becausethey aretheparent ofvoluminousgranitoid

suitesandhaveahighpreservationpotentialinthegeological record.Owingtotheirpeculiarhybridorigin,thiscontribution

cannotberesolvedinglobalcrustalgrowthmodelsbasedonly

on zircon Hf and Oisotopes, such that post-Archean crustal

growthmaybesignificantlyunder-estimated.

Acknowledgements

We acknowledge financial support from an ERC grant (StG

279828, project MASE to J. van Hunen), two PNP grants from

French CNRS (“Différentiation et évolution de la croûte tardi

orogénique”,2014and“Quantificationdeladuréed’unépisodede

fusionpartielleencontextededésépaississementtardi-orogénique”,

2016), the Deutsche Forschungsgemeinschaft (grant Zeh424/11-2

to A.Z.) and the Deutscher Akademischer Austauschdienst (grant

A/13/70682 to O.L.). A.V. acknowledges funding by the Labex

VOLTAIRE (ANR-10-LABX-100-01). V. Gardien provided two

sam-ples for this study. D. Garcia kindly supplied his database and

guided us in the field. A. Vézinet, M. Benoit and E. Bruand

as-sisted during zircon separation procedures. Helpful discussions

withV. JanoušekandG. Stevensweremuchappreciated.A. Gerdes

assisted during Hf analytical sessions and P. Montero performed

zircon imaging and SHRIMP measurements. This contribution is

IBERSIMS publication N◦ 35. We are grateful to A. Yin for

edi-torial handling. C.L. Farmer and S. Wilde provided thorough and

constructivereviews;R. Rudnick,P. Cawoodandananonymous

re-viewer providedinsightfulcommentsonan earlierversionofthis

manuscript–maytheyallbethanked.

Appendix A. Supplementary material

Supplementarymaterialrelatedtothisarticlecanbefound

on-lineathttp://dx.doi.org/10.1016/j.epsl.2016.09.033.

References

Albert,R.,Arenas,R.,Gerdes,A.,SanchezMartinez,S.,Marko,L.,2015.Provenance oftheHP–HTsubductedmargin intheVariscanbelt(CaboOrtegalComplex, NWIberianMassif).J.Metamorph.Geol. 33,959–979.

Andersson,U.B.,Eklund,O.,Fröjdö,S.,Konopelko,D.,2006.1.8Gamagmatisminthe FennoscandianShield;lateralvariations insubcontinentalmantleenrichment. Lithos 86,110–136.

Arndt,N.T.,2013.Formationandevolutionofthecontinentalcrust.Geochem. Per-spect. 2(3),405–533.

Arndt,N.T.,Goldstein,S.L.,1987.Useandabuseofcrust-formationages.Geology 15, 893–895.

Atherton,M.P.,Ghani,A.A.,2002.Slabbreakoff:amodelforCaledonian,LateGranite syn-collisionalmagmatismintheorthotectonic(metamorphic)zoneofScotland andDonegal,Ireland.Lithos 62.

Belousova,E.A.,Kostitsyn,Y.A.,Griffin,W.L.,Begg,G.C.,O’Reilly,S.Y.,Pearson,N.J., 2010.Thegrowthofthecontinentalcrust:constraintsfromzirconHf-isotope data.Lithos 119,457–466.

Bindeman,I.,2008.Oxygenisotopesinmantleandcrustalmagmasasrevealedby singlecrystalanalysis.Rev.Mineral.Geochem. 69,445–478.

Bonin,B.,2004.Docoevalmaficandfelsicmagmasinpost-collisional to within-plateregimesnecessarilyimplytwocontrasting,mantleandcrustal,sources? A review.Lithos 78,1–24.

Bouilhol,P.,Jagoutz,O.,Hanchar,J.M.,Dudas,F.O.,2013.DatingtheIndia–Eurasia collisionthrougharcmagmaticrecords.EarthPlanet.Sci.Lett. 366,163–175. Bouilhol,P.,Schaltegger,U.,Chiaradia,M.,Ovtcharova,M.,Stracke,A.,Burg,J.-P.,

Da-wood,H.,2011. Timingofjuvenile arccrustformation andevolutioninthe SapatComplex(Kohistan–Pakistan).Chem.Geol. 280,243–256.

Bouvier,A.,Vervoort,J.D.,Patchett,P.J.,2008.TheLu–HfandSm–Ndisotopic com-positionofCHUR:constraintsfromunequilibratedchondritesandimplications forthebulkcompositionofterrestrialplanets.EarthPlanet.Sci.Lett. 273(1–2), 48–57.

Campbell,I.H.,Allen,C.,2008.Formationofsupercontinentslinkedtoincreasesin atmosphericoxygen.Nat.Geosci. 1,554–558.

Chantraine,J.,Autran,A.,Cavelier,C.,1996.CartegéologiquedelaFranceàl’échelle dumillionième,6ème_{édition.BRGM,Orléans.}

Chelle-Michou,C., Laurent, O., Moyen, J.F., Block, S., Gardien, V., Paquette, J.L., Couzinié,S.,2015.NewU–PbandHfzircondatafromtheeasternMassif Cen-tral:fromGondwanatoPangeainanutshell.Géol.Fr. 1,37–38.

Chung,S.-L.,Chu,M.-F.,Zhang,Y.,Xie,Y.,Lo,C.-H.,Lee,T.-Y.,Lan,C.-Y.,Li,X.,Zhang, Q.,Wang,Y.,2005.Tibetantectonicevolutioninferredfromspatialandtemporal variationsinpost-collisionalmagmatism.Earth-Sci.Rev. 68,173–196. Clemens,J.D.,Darbyshire, D.P.F.,Flinders,J., 2009. Sourcesofpost-orogenic

calc-alkalinemagmas:theArrocharandGarabalHill–GlenFynecomplexes,Scotland. Lithos 112,524–542.

Conceição,R., Green,D.H.,2004. Derivationofpotassic (shoshonitic)magmasby decompressionmeltingofphlogopite+pargasitelherzolite.Lithos 72,209–229. Condamine,P.,Médard,E.,2014.Experimentalmeltingofphlogopite-bearing man-tleat1GPa:implicationsforpotassicmagmatism.EarthPlanet.Sci.Lett. 397, 80–92.

Condie,K.C.,Bickford,M.E.,Aster,R.C.,Belousova, E.,Scholl,D.W.,2011. Episodic zirconages, Hfisotopiccomposition,andthepreservationrateofcontinental crust.Geol.Soc.Am.Bull. 123,951–957.

Couzinié,S.,Moyen,J.F.,Villaros,A.,Paquette,J.L.,Scarrow,J.H.,Marignac,C.,2014. TemporalrelationshipsbetweenMg–Kmaficmagmatismandcatastrophic melt-ingoftheVariscancrustinthesouthernpartofVelayComplex(MassifCentral, France).J.Geosci.,69–86.

Davidson,J.,1987.Crustalcontaminationversussubductionzoneenrichment: exam-plesfromtheLesserAntillesandimplications formantlesourcecompositions ofislandarcvolcanicrocks.Geochim.Cosmochim.Acta 51,2185–2198. Dhuime,B.,Hawkesworth,C.J.,Cawood,P.A.,Storey,C.D.,2012.Achangeinthe

geo-dynamicsofcontinentalgrowth3billionyearsago.Science 335,1334–1336. Dickinson,W.R.,2008.Impactofdifferentialzirconfertilityofgranitoidbasement

rocksinNorthAmericaonagepopulationsofdetritalzirconsandimplications forgranitepetrogenesis.EarthPlanet.Sci.Lett. 275,80–92.

Downes,H.,Dupuy, C.,Leyreloup,A.F., 1990.Crustal evolutionofthe Hercynian beltofWesternEurope:evidencefromlower-crustalgranuliticxenoliths(French MassifCentral).Chem.Geol. 83,209–231.

Faure,M.,BéMézème,E.,Cocherie,A.,Rossi,P.,Chemenda,A.,Boutelier,D.,2008. DevoniangeodynamicevolutionoftheVariscanBelt,insightsfromtheFrench MassifCentralandMassifArmoricain.Tectonics 27,TC2005.

Fowler, M., Henney,P.J., 1996.Mixed Caledonianappinite magmas: implications forlamprophyrefractionationandhighBa–Srgranitegenesis.Contrib.Mineral. Petrol. 126,199–215.

Fowler,M.,Rollinson,H.,2012.PhanerozoicsanukitoidsfromCaledonianScotland: implicationsforArcheansubduction.Geology 40,1079–1082.

Gaetani,G.A.,Grove,T.L.,1998.Theinfluenceofwateronmeltingofmantle peri-dotite.Contrib.Mineral.Petrol. 131,323–346.

Green,D.H.,2015. Experimentalpetrologyofperidotites,including effectsof wa-terandcarbononmeltingintheEarth’suppermantle.Phys.Chem.Miner. 42, 95–122.

Grégoire,M.,Moine,B.N.,O’Reilly,S.Y.,Cottin,J.-Y.,Giret,A.,2000.Traceelement residenceand partitioning inmantle xenolithsmetasomatizedbyhighly al-kaline, silicate- and carbonate-rich melts (KerguelenIslands, Indian Ocean). J. Petrol. 41,477–509.

Griffin,W.L.,Wang,X.,Jackson,S.E.,Pearson,N.J.,O’Reilly,S.Y.,Xu,X.-S.,Zhou,X., 2002.Zirconchemistryandmagmamixing,SEChina:in-situanalysisofHf iso-topes,TongluandPingtanigneouscomplexes.Lithos 61,237–269.

Guo,Z.,Wilson,M.,Zhang,L.,Zhang,M.,Cheng,Z.,Liu,J.,2014.Theroleof sub-ductionchannelmélangesandconvergentsubductionsystemsinthe petroge-nesisofpost-collisionalK-richmaficmagmatisminNWTibet.Lithos 198–199, 184–201.

Hagen-Peter,G.,Cottle,J.M.,Tulloch,A.J.,Cox,S.C.,2015.Mixingbetweenenriched lithosphericmantleandcrustalcomponentsinashort-livedsubduction-related magmasystem,DryValleysarea,Anatarctica:insightsfromU–Pb geochronol-ogy,Hfisotopesandwhole-rockgeochemistry.Lithosphere 7(2),174–188. Hawkesworth,C.,Cawood,P.,Kemp,T.,Storey,C.,Dhuime,B., 2009.Amatterof

preservation.Science 323,49–50.

Hawkesworth,C.J., Dhuime, B., Pietranik,A.B.,Cawood, P.A.,Kemp,A.I.S., Storey, C.D.,2010.Thegenerationandevolutionofthecontinentalcrust.J.Geol.Soc. (Lond.) 167,229–248.

Hawkesworth,C.J.,Kemp,A.I.S.,2006.Usinghafniumandoxygenisotopesinzircons tounraveltherecordofcrustalevolution.Chem.Geol. 226,144–162. Heilimo,E.,Halla,J.,Andersen,T.,Huhma,H.,2013.Neoarcheancrustalrecycling

andmantlemetasomatism:Hf–Nd–Pb–Oisotopeevidencefromsanukitoidsof theFennoscandianshield.PrecambrianRes. 228,250–266.

Hermann,J.,Rubatto,D.,2009.Accessoryphasecontrolonthetraceelement signa-tureofsedimentmeltsinsubductionzones.Chem.Geol. 265,512–526. Holub,F.V.,1997.Ultrapotassicplutonicrocksofthedurbachiteseriesinthe

Bo-hemian Massif: petrology, geochemistry and petrogenetic interpretation. Sb. Geol.Vˇed,Ložisk.Geol.Mineral. 31,5–26.

Jagoutz,O.,Kelemen,P.B.,2015.Roleofarcprocessesintheformationofcontinental crust.Annu.Rev.EarthPlanet.Sci. 43,363–404.

Janoušek,V.,Braithwaite,C.J.R.,Bowes,D.R.,Gerdes,A.,2004.Magma-mixinginthe genesisofHercyniancalc-alkaline granitoids:anintegrated petrographicand

geochemicalstudyoftheSázavaintrusion,CentralBohemianPluton,Czech Re-public.Lithos 78,67–99.

Janoušek,V.,Holub,F.V.,2007.ThecausallinkbetweenHP–HTmetamorphismand ultrapotassicmagmatismincollisionalorogens:casestudyfromthe Moldanu-bianZoneoftheBohemianMassif.Proc.Geol.Assoc. 118,75–86.

Kemp,A.I.,Hawkesworth,C.J.,Paterson,B.A.,Kinny,P.D.,2006.Episodicgrowthof theGondwanasupercontinentfromhafniumandoxygenisotopesinzircon. Na-ture 439,580–583.

Küster,D.,Harms,U.,1998.Post-collisionalpotassicgranitoidsfromthesouthern andnorthwesternpartsoftheLateNeoproterozoicEastAfricanOrogen:a re-view.Lithos 45,177–195.

Lancaster,P.J.,Storey,C.D.,Hawkesworth,C.J.,Dhuime,B.,2011.Understandingthe roles ofcrustalgrowthand preservationinthe detrital zirconrecord. Earth Planet.Sci.Lett. 305,405–412.

Lardeaux, J.M.,Ledru, P.,Daniel,I., Duchene,S.,2001. TheVariscanFrench Mas-sifCentral–anewadditiontotheultra-high pressuremetamorphic “club”: exhumation processes and geodynamic consequences. Tectonophysics 332, 143–167.

Laurent,O.,Couzinié,S.,Vanderhaeghe,O.,Zeh,A.,Moyen,J.F.,Villaros,A.,Gardien, V.,2015.U–PbdatingofVariscanigneousrocksfromtheeasternFrench Mas-sifCentral:southwardmigrationofcoevalcrust- andmantle-meltingwitnesses late-orogenicslabretreat.Géol.Fr. 1,82–83.

Laurent,O.,Doucelance,R.,Martin,H.,Moyen,J.-F.,2013.Differentiationofthe late-Archaeansanukitoidseriesandsomeimplicationsforcrustalgrowth:insights fromgeochemicalmodellingontheBulaipluton,CentralLimpopoBelt,South Africa.PrecambrianRes. 227,186–203.

Laurent,O.,Martin,H.,Doucelance,R.,Moyen,J.F.,Paquette,J.L.,2011. Geochem-istry and petrogenesis of high-K“sanukitoids” from the Bulai pluton, Cen-ralLimpopoBelt, South Africa:implications for geodynamic changes at the Archaean–Proterozoicboundary.Lithos 123,73–91.

Laurent,O.,Martin,H.,Moyen,J.F.,Doucelance,R.,2014a.Thediversityand evolu-tionoflate-Archeangranitoids:evidencefortheonsetof“modern-style”plate tectonicsbetween3.0and2.5Ga.Lithos 205,208–235.

Laurent,O.,Rapopo,M.,Stevens,G.,Moyen,J.F.,Martin,H.,Doucelance,R.,Bosq,C., 2014b.ContrastingpetrogenesisofMg–KandFe–Kgranitoidsandimplications forpost-collisionalmagmatism:casestudyfromtheLate-ArcheanMatokpluton (Pietersburgblock,SouthAfrica).Lithos 196–197,131–149.

Laurent,O.,Zeh,A.,2015.AlinearHfisotope-agearraydespitedifferentgranitoid sourcesandcomplexArcheangeodynamics:examplefromthePietersburgblock (SouthAfrica).EarthPlanet.Sci.Lett. 430,326–338.

Ledru,P.,Courrioux,G.,Dallain,C.,Lardeaux,J.M.,Montel,J.M.,Vanderhaeghe,O., Vitel,G.,2001.TheVelaydome(FrenchMassifCentral):meltgenerationand graniteemplacementduringorogenicevolution.Tectonophysics 342,207–237. Lee,C.-T.A.,Harbert, A.,Leeman,W.P.,2007. Extensionoflattice straintheoryto

mineral/mineralrare-earthelementpartitioning:anapproachforassessing dis-equilibriumanddevelopinginternallyconsistentpartitioncoefficientsbetween olivine,orthopyroxene,clinopyroxeneandbasalticmelt.Geochim.Cosmochim. Acta 71,481–496.

Liégeois, J.-P., Navez,J., Hertogen,J., Black,R., 1998. Contrastingorigin of post-collisionalhigh-Kcalc-alkalineandshoshoniticversusalkalineandperalkaline granitoids.Theuseofslidingnormalization.Lithos 45,1–28.

Linckens, J.,Bruijn,R.H.C.,Skemer,P.,2014. Dynamicrecrystallizationand phase mixing in experimentally deformed peridotite. Earth Planet. Sci. Lett. 388, 134–142.

Linnemann,U.,Gerdes,A.,Hofmann,M., Marko,L.,2014. TheCadomianOrogen: NeoproterozoictoEarlyCambriancrustalgrowthandorogeniczoningalongthe peripheryoftheWestAfricanCraton—constraintsfromU–PbzirconagesandHf isotopes(SchwarzburgAntiform,Germany).PrecambrianRes. 244,236–278. Liu,D.,Zhao,Z.,Zhu,D.-C.,Niu,Y.,DePaolo,D.J.,Harrison,T.M.,Mo,X.,Dong,G.,

Zhou,S.,Sun,C.,Zhang,Z.,Liu,J.,2014.Postcollisionalpotassicandultrapotassic rocksinsouthern Tibet:mantleand crustaloriginsinresponsetoIndia–Asia collisionandconvergence.Geochim.Cosmochim.Acta 143,207–231.

Lowe,D.R.,Tice,M.M.,2004.GeologicevidenceforArcheanatmosphericand cli-maticevolution: fluctuatinglevels ofCO2, CH4, and O2 with anoverriding

tectoniccontrol.Geology 32,493–496.

Mallik,A.,Nelson,J.,Dasgupta,R.,2015.Partialmeltingoffertileperidotitefluxed byhydrousrhyoliticmeltat2–3GPa:implicationsformantlewedge hybridiza-tionbysedimentmeltandgenerationofultrapotassicmagmasinconvergent margins.Contrib.Mineral.Petrol. 169(5),1–24.

McCulloch,M.T.,Wasserburg,G.J.,1978.Sm–NdandRb–Srchronologyofcontinental crustformation.Science 200,1003–1011.

McDonough,W.F.,Sun,S.-s.,1995.ThecompositionoftheEarth.Chem.Geol. 120, 223–253.

McKenzie,D.,O’Nions,R.K.,1991.Partialmeltdistributionsfrominversionofrare Earthelementconcentrations.J.Petrol. 32(5),1021–1091.

Melleton,J.,Cocherie,A.,Faure,M.,Rossi,P.,2010.PrecambrianprotolithsandEarly PaleozoicmagmatismintheFrenchMassifCentral:U–PbdataandtheNorth GondwanaconnectionintheWestEuropeanVariscanbelt.GondwanaRes. 17, 13–25.