Coo

Ca

b

c

ARRAA

KTCOPSE

1

cv2fwe2esmmP

0h

Precambrian Research 231 (2013) 206– 217

Contents lists available at SciVerse ScienceDirect

Precambrian Research

journa l h om epa ge : www.elsev ier .com/ locate /precamres

onstraints from experimental melting of amphibolite on the depthf formation of garnet-rich restites, and implications for modelsf Early Archean crustal growth

hao Zhanga,∗, Francois Holtza, Jürgen Koepkea, Paul Eric Wolffa, Changqian Mab, Jean H. Bédardc

Institut für Mineralogie, Leibniz Universität Hannover, Callinstr. 3, Hannover D-30167, GermanyState Key Laboratory of Geological Processes and Mineral Resources, Faculty of Earth Sciences, China University of Geosciences, Wuhan 430074, ChinaGeological Survey of Canada, 490 de la Couronne, Québec, QC G1K 9A9, Canada

a r t i c l e i n f o

rticle history:eceived 15 January 2013eceived in revised form 11 March 2013ccepted 14 March 2013vailable online xxx

eywords:TGontinental crustceanic plateauartial meltingubductionxperimental petrology

a b s t r a c t

The felsic continental crust formed in the early Earth most likely resulted from melting of basaltic pro-toliths, but the geodynamic processes leading to partial melting are still debated. Attempts to reconcilegeochronological data, thermal modeling and experimental results have led to two major alternativescenarios: (1) partial melting of subducted oceanic slabs and (2) dehydration melting at the base of thick(or thickened) oceanic/simatic protocrusts. Existing experimental data on melting of metabasalt suggestthat garnet only becomes an important residual phase (>10 wt%) at depths >50–60 km, which has beenseen as difficulty for model 2. We present results of amphibolite dehydration-melting experiments atpressures of 5–15 kbar and provide constraints on melting reactions of a hydrated metabasalt with SiO2

of 47.5 wt% and Al2O3 of 16.4 wt%. Our experiments demonstrate that felsic melts and complementaryrestites with ∼20 wt% garnet can form at ca. 900 ◦C and 12 kbar, conditions corresponding to the base of a40-km thick basaltic protocrust that might be prevalent in the Early Archean. Based on phase proportionsdetermined experimentally and trace element partitioning data, our modeling suggests that such partial

melts resemble the Early Archean tonalite–trondhjemite–granodiorite (TTG) suites, including high Al2O3

and low MgO contents, and modestly high La/Yb and Sr/Y ratios. The garnet-rich restite is calculated tobe denser than the underlying Early Archean lherzolitic upper mantle and would have the potential todelaminate. Our experimental results and combined geochemical modeling are consistent with modelswhere the initial growth of continental crust on the Archean Earth occurred in non-subduction settingsby anatexis of the base of basaltic plateaus.

. Introduction

The tectonic mechanism by which the Archean continentalrust formed has been debated for decades and is still contro-ersial (Armstrong, 1991; Hawkesworth et al., 2010; Rollinson,006). Oceanic subduction is commonly advocated as a key scenarioor generating the tonalite–trondhjemite–granodiorite (TTG) suiteshich compose the main part of Archean cratonic nuclei (Arth

t al., 1978; Foley et al., 2002; Martin, 1986; Martin and Moyen,002; Moyen and van Hunen, 2012; Rapp et al., 2003; Smithiest al., 2003). The geochemical similarities between Archean TTGuites and modern adakites (Martin et al., 2005) is considered by

any to imply melting of subducting oceanic slabs as the primaryechanism by which TTGs were generated, even though many

hanerozoic adakite suites are now thought to form by remelting

∗ Corresponding author. Tel.: +49 0511 762 5517; fax: +49 0511 762 3045.E-mail address: c.zhang@mineralogie.uni-hannover.de (C. Zhang).

301-9268/$ – see front matter © 2013 Elsevier B.V. All rights reserved.ttp://dx.doi.org/10.1016/j.precamres.2013.03.004

© 2013 Elsevier B.V. All rights reserved.

of thickened lower crust (e.g., Coldwell et al., 2011; Girardi et al.,2012; Petford and Gallagher, 2001). There is increasing evidence fora Middle-Late Archean onset of plate tectonics, as a consequence ofprogressing cooling of the mantle and densification and rigidifica-tion of oceanic lithosphere (Hamilton, 2003; Herzberg et al., 2010;Keller and Schoene, 2012; Korenaga, 2008; Labrosse and Jaupart,2007; Shirey and Richardson, 2011; Sizova et al., 2010; Stern, 2005;Van Kranendonk, 2010; Van Kranendonk et al., 2007; Vlaar et al.,1994). In addition, paired metamorphic belts, blueschists and ophi-olites as characteristic outcomes of subduction are absent in theArchean (Bédard et al., 2013; Brown, 2008; Stern, 2008). Together,these lines of evidence suggest that oceanic subduction is unlikelyto be the appropriate tectonic context for generation of the ear-liest continental crust, which began to from as early as ca. 4.4 Ga(Harrison, 2009; Wilde et al., 2001).

A vertical growth model, which interprets the initial formationof Early Archean continental crust by partial melting at the baseof oceanic plateaus in a non-subduction setting, provides a possi-ble alternative (Bédard, 2006a; Bédard et al., 2013; Condie, 1980,

n Research 231 (2013) 206– 217 207

12vdpnnurot1

taptcba11ttmuSwt2cnm1mmt≥Xe32m2m2

ewO4Tr

2

2

t7Itm3

0

10

20

30

40

50

60

70

Wei

ght p

erce

n t (%

)

Experimental duration (day )

P=12 kbar, T=900°C

40 12 8 16 20 24

AmpCpxGrtPlgMelt

Fig. 1. Evolution of phase proportion (wt%) as a function of experimental dura-tion for an experiment conducted at 12 kbar and 900 ◦C and using powered startingmaterials. Near-equilibrium conditions are reached for experimental runs withdurations of ≥8 days. Weight percents of phases are calculated from volume per-cents (determined by image analysis) and mineral densities: amphibole (3.3 g/cm3),clinopyroxene (3.2 g/cm3), garnet (4.0 g/cm3), plagioclase (2.7 g/cm3), and melt

C. Zhang et al. / Precambria

986; McKenzie, 1984; Smithies et al., 2009; Zegers and van Keken,001). This model involves three sequential processes, i.e., (1) initialertical (volcanic and intrusive) accretion of thick oceanic plateauserived by mantle melting, (2) partial melting at the base of oceaniclateaus resulting in generation of felsic magmas constituting theuclei of cratonic continental crust, and (3) subsequent delami-ation of dense garnet-rich melting residues downward into thepper mantle. The engine of this process is the hot geothermalegime in the Early Archean, which has resulted in the formation ofceanic protocrusts (i.e., plateaus) with greater thickness relativeo modern ones due to more extensive mantle melting (Vlaar et al.,994; Korenaga, 2006; Herzberg et al., 2010).

A direct check on the validity of the vertical growth model iso compare the composition of melts derived from amphibolitest conditions corresponding to the base of early Earth’s oceaniclateaus with that of magmatic rocks of the oldest felsic continen-al crust. The latter are the Early Archean TTG suites, and the formeran be reproduced by partial melting experiments and simulatedy modeling. Dehydration partial melting experiments on variousmphibolitic (metabasaltic) protoliths, performed mainly in the990s (Beard and Lofgren, 1989; Rapp et al., 1991; Wolf and Wyllie,991), provided many first-order physicochemical constraints forhe formation of TTGs. One conclusion which is of particular impor-ance for this study is that pressure exerts a strong control on

elting reactions and on the assemblages and proportions of resid-al phases (Foley et al., 2002; Rapp et al., 2003; Sen and Dunn, 1994;pringer and Seck, 1997; Wyllie and Wolf, 1993; Xiong et al., 2005),hich consequently influences melt compositions especially for

race elements (Drummond and Defant, 1990; Moyen and Stevens,006). In order to provide better constraints on trace elementompositions which were not measurable in those experiments,umerical modeling of trace-element partitioning during partialelting has been applied (Drummond and Defant, 1990; Martin,

987; Xiong, 2006). Several recent studies based on dehydration-elting experiments, thermodynamic modeling and trace elementodeling, have emphasized that the melting of amphibolitic pro-

ocrust resulting in TTG-like melts needs to occur at depths of50 km according to Nair and Chacko (2008), Adam et al. (2012),iong et al. (2005) and Xiong (2006), or ≥60 km according to Moyent al. (2010). Since the thickest existing oceanic plateau is ca.5–38 km thick (Ontong Java; Neal et al., 1997; Richardson et al.,000), proponents of this model have had to invoke high, long-liveagmatic fluxes so as to construct such thickness (e.g., Bédard et al.,

013); or else invoke melting of delaminated eclogitized crustalaterial (Coldwell et al., 2011; Gao et al., 2004; van Thienen et al.,

004a,b; Zhang et al., 2010).In this paper, we present new amphibolite dehydration-melting

xperiments and trace-element models, using a starting materiale believe to be appropriate to early Earth’s basaltic protocrust.ur new data suggest that melting at the base of a stable0 km-thick oceanic plateau can generate substantial amount ofTG-composition melt, which is in equilibrium with a garnet-richestite which has the potential to be delaminated.

. Experimental investigation

.1. Experimental procedures

We conducted dehydration melting experiments on syn-hetic amphibolites in the temperature and pressure range of00–1000 ◦C and 5–15 kbar using a piston-cylinder apparatus at the

nstitute of Mineralogy, University of Hannover. Two experimen-al set-ups were used. In the first set of experiments, the starting

aterials were a plagioclase crystal (cube with side length of ca. mm) surrounded by hornblende powder (<100 �m). This set was

(2.3 g/cm3).

aimed at better detecting melting reactions and constraining thesolidus, because reaction phenomena between the two separatedstarting phases are easy to observe. In the second set of experi-ments, which were aimed at determining the phase proportionsand melt compositions at near-equilibrium conditions, mechan-ically mixed hornblende and plagioclase powders (both < 30 �m)with a weight ratio of 2:1 were used as starting material. All startingmaterials were first loaded into a one-side welded noble metal cap-sule and placed inside a drying oven at 110 ◦C for 2 hours, in orderto diminish the amount of absorbed surface water. For melting runsbelow 900 ◦C, Au capsules were used; for higher temperature runs,Ag75Pd25 capsules were used to minimize iron loss down to 1∼3%relative to Fe in the starting material (see Supplementary Fig. 1).The redox condition was measured by COH sensor, being betweenthe Fe3O4-FeO and Co-CoO buffers. Detailed methodologies cor-responding to the two experimental set-ups have been describedby Johannes (1989) and Johannes and Koepke (2001), respectively.Some time-dependent runs (2–24 days) were conducted (Fig. 1),and the results indicate that near-equilibrium mineral assemblageswere obtained in experiments conducted with powder as startingmaterial if experimental durations were of ≥8 days at temperatures≥800 ◦C. The experimental phases of the second set (using mixedpowder) with duration of ≥8 days were analyzed by electron micro-probe (Cameca SX100). Phase proportions were estimated by 2Danalysis of backscattered electron (BSE) images and constrained bymajor element mass-balance. The composition of the starting mate-rial is listed in Table 1. Experimental conditions and run productsfrom powdered starting materials are listed in Tables 2–3, and aresummarized in Sections 2.3 and 2.4.

2.2. Composition of starting material

The composition of the minerals and the bulk compositionin the experiments conducted with powdered starting mate-rial are given in Table 1. The plagioclase has a compositionAn57Ab41Or2 and the amphibole is a magnesiohornblende with

Mg# [100 × Mg/(Mg+Fetot)] of 54. The compositions of these twophases are typical for amphibolites (e.g., Harlov and Förster, 2002).In contrast to several previous experiments (e.g., Rapp et al., 1991),care was taken to select two mineral phases with compositions that

208 C. Zhang et al. / Precambrian Research 231 (2013) 206– 217

Table 1Composition of the starting materials for dehydration-melting experiments and modeling (wt%).

SiO2 TiO2 Al2O3 FeOtot MnO MgO CaO Na2O K2O H2O Total Mg#

Amphibolea 44.85 1.66 10.20 16.49 0.27 11.00 11.82 1.34 0.46 2.00 100.00 54.3Plagioclaseb 52.90 28.80 0.54 11.70 4.70 0.30 99.00Amphibolite c 47.48 1.11 16.38 11.16 0.18 7.33 11.77 2.46 0.41 98.48 53.9

W&Wd 48.5 0.4 14.4 8.5 0.2 10.8 14.8 1.7 0.2 99.5 69.43VGe 51.46 0.80 14.14 11.18 0.18 7.29 11.06 2.52 0.88 100.00 54.0

Archean greenstone PC-227f 52.47 0.40 17.19 9.44 0.21 11.13 7.45 0.96 0.74 99.99 67.82000–13g 50.72 0.64 8.30 12.07 0.22 13.98 11.01 0.07 1.57 99.92 67.4Avgh 50.36 0.75 10.63 11.70 0.21 11.81 9.56 1.58 0.29 99.88 64.3

N-MORB EPRi 50.16 1.47 15.79 9.51 0.16 7.58 12.19 2.76 0.13 99.75 58.7

E-MORBj EPR 50.17 1.63 15.54 9.65 0.19 7.79 11.45 2.86 0.23 99.50 59.0MAR 49.73 1.46 15.69 9.22 0.16 7.80 11.59 2.70 0.24 98.59 60.1Mac 48.56 1.42 17.71 7.36 0.12 7.46 11.80 3.21 0.87 98.50 66.6APR 49.79 2.21 16.76 9.20 0.17 5.98 9.90 3.36 1.23 99.04 55.6

a Amphibole is derived from the serpentinite-amphibolite association in Crete Island (Seidel et al., 1981).b Plagioclase (An57An41Or2) is derived from Lake County, Oregon (Wenk et al., 1980).c The synthetic amphibolite, made up of mixed amphibole and plagioclase powders (<30 �m) with a weight ratio of 2:1, is used as starting material in this experimental

study.d The starting material used in the experiments of Wolf and Wyllie (1991, 1993, 1994).e The sample 3VG is the MORB-type starting material used in the experiments of Nair and Chacko (2008).f The sample PC-227 is the starting material (4.3 Ga Nuvvuagittuq greenstone) used in the experiments of Adam et al. (2012).g The sample 2000–13 is an Eoarchean tholeiite from the inner arc sequence of the Isua Supracrustal Belt (Polat and Hofmann, 2003) and the starting material used in the

thermodynamic melting-modeling of Nagel et al. (2012).h Average value of Eoarchean metabasalts of the Isua Supracrutal Belt (Polat and Hofmann, 2003; Hoffmann et al., 2011).i An N-MORB reference from northern East Pacific Rise is listed for comparison (Klein, 2003).

ge of 8l al., 19P ntarc

amawtfta

ib(icfss2tca

TS

cp

j Some E-MORB references are listed for comparison. EPR: Seamount lavas (averaavas (average of 93 samples) from 10◦ to 24◦ N North Mid-Atlantic Ridge (Dosso et

acific (Kamenetsky et al., 2000). APR: Basaltic lavas (average of 36 samples) from A

re close to that are expected to be stable at conditions approxi-ating the beginning of dehydration melting at 5–8 kbar, so as to

void disequilibrium melting reactions. In experiments conductedith natural rock powders with a bulk amphibolite composition

hat include low-temperature mineral phases (e.g., greenschistacies), metastable melting reactions may occur at the interface ofhe minerals because these starting minerals are not in equilibriumt temperatures close to the dehydration-melting solidus.

The bulk composition of the powdered starting material is givenn Table 1 and is close to that of modern enriched mid-ocean ridgeasalts (E-MORB), with similar contents of Al2O3 (15–17 wt%), MgO7–8 wt%), CaO (11–12 wt%) and Na2O (2–3 wt%). The SiO2 contents slightly lower by less than 1 to 2.5 wt%, depending on the E-MORBomposition, and the FeO content is higher by 1.5 to 2.0 wt% (exceptor one composition from Macquarie Island). The exact bulk compo-ition of Early Archean metabasalts (i.e., possible protoliths of TTGuites) is still controversial (e.g., Furnes et al., 2007; Smithies et al.,

009). We note that Early Archean basalts derived from the man-le of Early Earth tend to have lower SiO2 but higher FeO and MgOontents than modern normal mid-ocean ridge basalts (N-MORB)s a result of the higher Archean mantle potential temperatures

able 2ummary of experimental runs using powder starting materialsa

Pressure (kbar) Tem. (◦C) Time (days) Phas

Amp

8 900 8 50.010 900 8 58.012 800 8 47.112 850 8 39.212 900 24 33.712 950 8 28.212 1000 16 19.115 900 8 19.6

a Phase proportions are estimated based on the analysis of BSE images, calibrated uponube crystal and hornblende powder were used to constrain the phase diagram; see Figroportions cannot be determined.

0 samples) from 5◦–15◦ N East Pacific Rise (Niu and Batiza, 1997). MAR: Seamount93). Mac: Near-primitive glasses (average of 6 samples) from Macquarie Island, SWtic-Phoenix Ridge (Choe et al., 2007).

(Herzberg et al., 2010); but would likely resemble that of E-MORBand thus the starting composition used in this study. A discus-sion on the possible effect of variations of the bulk compositionsfrom recent published experimental studies on phase assemblagesis given in Section 3.1.

2.3. Solidus of amphibolite

The knowledge of the position of the fluid-absent solidus ofamphibolite is a prerequisite for understanding the melting relationof a metabasaltic crust at any tectonic setting. In the investigatedsystem, the dehydration melting reaction is characterized by a pro-gressive breakdown of amphibole and plagioclase to form silicatemelt, clinopyroxene and/or garnet. A typical run product obtainedusing a plagioclase single crystal and hornblende powder at 12 kbarand 900 ◦C is presented in Supplementary Fig. 2. It is emphasizedthat plagioclase and amphibole are expected to remain stable for an

interval above the dehydration melting solidus because the equi-librium compositions of amphibole and plagioclase change withchanging temperature and pressure (as does the melt fraction).Although numerous melting experiments have been conducted in

e proportions (wt%)

Plg Cpx Grt Melt

31.4 14.8 – 3.9 27.1 4.7 7.4 2.7

24.9 19.0 6.4 2.6 24.2 20.8 10.3 5.5 15.4 21.3 19.0 10.7 11.0 22.2 25.1 13.6 3.6 23.1 32.8 21.4 3.1 18.9 32.0 26.4

mineral densities and mass balance. Other experimental runs using a plagioclase. 1 and Supplementary Table 1 for their P–T conditions and phases, but the phase

C. Zhang et al. / Precambrian Research 231 (2013) 206– 217 209

600 700 800 900 1000 1100 12004

6

8

10

12

14

16

Temperature (°C)

Pre

ssur

e(k

bar)

Grt-in

Amp+Plg

dehydration-meltingsolidus (this study )

VS-2001WW-1993

Amp+Plg+CpxAmp+Plg+Cpx+Grt

Fig. 2. Phase stabilities in an amphibolite (see bulk composition in Table 1). Thegray thick curve is the dehydration melting solidus curve determined in this study.The curves (labeled with WW-1993 and VS-2001) are the classic amphibolite solidiestimated by Wyllie and Wolf (1993) and Vielzeuf and Schmidt (2001), respectively.TbS

tedTwaSsad1adt8amsba

2

mddTprgataefoot

0

10

20

30

40

50

800 850 900 950 1000

Temperature (°C )

Wei

ght p

erce

nt (%

)

P=12 kbar, duration=8 days AmpCpxGrtPlgMelt

Fig. 3. Phase proportion as a function of temperature (at 12 kbar). All experimentaldurations are 8 days. Weight percents of phases are calculated from volume per-

3

of Foley et al. (2002).It has been proposed that 40 km (∼12 kbar) is the most plausible

“equilibrium” thickness of oceanic plateaus at ∼4.0 Ga (Korenaga,

M&M-2002

M&M-2002

10 Ma10 Ma

20 Ma20 Ma

40 kmTP= 1700 °C

3

1

2

20%

10%5%

600 700 800 900 1000 1100 12004

6

8

10

12

14

16

Temperature (°C)

Pre

ssur

e(k

bar)

dry basaltsolidusdry basaltsolidus

X

Y

Z

Grt-in

30%

Fig. 4. Phase diagram showing the garnet proportion in the residue of an amphibo-lite determined in this work (green curves with percentages). The gray thick curveis the dehydration melting solidus determined in this study. The curves (1) and (2)are modeled geothermal gradients of oceanic crust with an age of 10 Ma and 20 Ma,respectively (Vlaar et al., 1994; potential temperature of the mantle (TP) is 1750 ◦C).The curve (3) is the putative Early Archean geotherm at subduction zone from Martinand Moyen (2002). The 40-km dashed line indicates the stable thickness of oceanicplateaus derived from a hot mantle with TP of ca. 1700 ◦C (Korenaga, 2006). With

he garnet-in curve above 800 ◦C was drawn with a weak negative dP/dT slope toe consistent with entropy consideration of this incongruent reaction (Vielzeuf andchmidt, 2001).

he last two decades, few experimental studies have focused on thexact determination of the P–T conditions of the amphibole break-own dehydration melting solidus (Moyen and Stevens, 2006).he S-shaped dehydration melting solidus of amphibolite, whichas revealed by the pioneering study of Wyllie and Wolf (1993)

nd reinforced by further theoretical consideration of Vielzeuf andchmidt (2001), has not been fully verified by further experimentaltudies. The experimental results of this study are shown in Fig. 2nd demonstrate the validity of this S-shaped solidus. The dehy-ration melting curve leading to the formation of garnet above0 kbar has a negative steep P/T slope constrained between 740nd 780 ◦C in the pressure range 10–15 kbar. The position of theehydration melting curve below ca. 9 kbar is related to the forma-ion of clinopyroxene and is nearly independent of pressure at ca.80 ◦C. At all investigated pressures, clinopyroxene belongs to thenhydrous assemblage produced as the result of the dehydrationelting of the plagioclase-amphibole assemblage, while garnet is

table at pressures ≥10 kbar, which is consistent with phase assem-lages observed in previous experimental studies (Patino Doucend Beard, 1995; Rapp et al., 1991).

.4. Phase proportions and compositions

The proportions of phases at supersolidus conditions were esti-ated based on the analysis of BSE images, calibrated upon mineral

ensities and mass balance. The phase proportions obtained by 8-ay runs at 12 kbar are shown in Fig. 3 as a function of temperature.he proportion of garnet in restite increases with increasing tem-erature (Fig. 3) and with increasing pressure (Fig. 4) within theange 800–1000 ◦C and 10–15 kbar. Importantly, the proportion ofarnet is larger than that obtained in the experimental study of Nairnd Chacko (2008). For example, at 12.5 kbar and 850–950 ◦C, lesshan 5 wt% garnet was observed in the residue obtained by Nairnd Chacko (2008), whereas ca. 20 wt% garnet is obtained from ourxperimental dataset for the run at 12 kbar and 900 ◦C. Since the

ormation of garnet is expected to be enhanced by a higher activityf Al2O3, this difference might be explained by the composition ofur starting material, which has a slightly lower SiO2 concentra-ion and higher Al2O3 concentration compared to that of Nair and

cents (determined by image analysis) and mineral densities: amphibole (3.3 g/cm ),clinopyroxene (3.2 g/cm3), garnet (4.0 g/cm3), plagioclase (2.7 g/cm3), and melt(2.3 g/cm3).

Chacko (2008) (Table 1). As a result, for protoliths similar to E-MORBas investigated in this study, pressures of 12 kbar appear sufficientto stabilize 20 wt% garnet in the residue (also see Section 3.1). Inaddition, abundant amphibole remains in the solid assemblage atpressures in the range 10–15 kbar, which is consistent with theexperimental results of Nair and Chacko (2008) and the suggestion

increasing maturation of the oceanic protocrust, the beginning of melting transfersfrom X (∼20 km, 10 Ma) to Y. At the base of a table oceanic plateau (near Z), thepartial melts are geochemically similar to Earth Archean high-Al TTG suites and therestite (including ∼20 wt% garnet) has the potential to delaminate (see the text forexplanation).

2 n Rese

2mib8tc(wctas(ttffobtga1iaai

3

3

aodvttgWwsie(edoa

msococatSmme(

10 C. Zhang et al. / Precambria

006; Vlaar et al., 1994) (see also Section 3.2). Thus, our experi-ental results at this pressure (Table 3) are of particular interest

n interpreting the initial growth of continental crust from aasaltic oceanic protocrust. At 12 kbar and in the temperature range00–1000 ◦C, the An content of plagioclase is slightly higher thanhat of the starting plagioclase (An57) and does not change signifi-antly (An65–68), consistent with the experiments of Sen and Dunn1994). However, the composition of amphibole changes slightlyith increasing temperature, showing higher Ti content, lower Fe

ontent and higher Mg#. The composition of clinopyroxene is nearhe boundary between diopside and augite, and no systematic vari-tion in Mg# (ca. 60–70) with temperature is observed. There is alight compositional difference between the core and rim of garnetincrease in Mg# from core to rim by a value of ca. 2–5), showinghat early formed garnets have not been fully re-equilibrated withinhe adopted experimental duration. With increasing temperaturerom 800 ◦C to 1000 ◦C, the Mg# of garnet rims increases stronglyrom ca. 23 to 43. This Mg# increase cannot be attributed to a lossf Fe in the capsule material (see Supplementary Fig. 1) but has toe related to a rapid growth of garnet crystals in the first phase ofhe experiment. The partial melts are tonalitic in the An–Ab–Or dia-ram and have low Mg# values (ca. 30), consistent with previousmphibolite-melting experiments at pressures ≥8 kbar (Rapp et al.,991; Rushmer, 1991). In summary, we suggest that partial melt-

ng of E-MORB-like Early Archean basalts at conditions of 12 kbarnd ≥900 ◦C can generate restites with more than 20 wt% garnet,nd high-silica tonalitic melts similar to Early Archean TTG suitesn major elements (Condie, 2005).

. Discussion

.1. Starting material and its effect on melting products

Garnet is a residual phase from the incongruent melting ofmphibolite, and its formation and proportion exert a crucial effectn the composition of partial melt. In experimental simulation ofehydration melting of amphibolites, the nucleation of garnet isery sluggish compared to other phases, and the delayed nuclea-ion may cause disequilibrium phase compositions (e.g., melt) inhe beginning of the experiments. Problems with nucleation androwth of garnet are illustrated by the experiments of Wolf andyllie (1991, 1993, 1994). Although the same starting materialas used (Table 1), garnet was not present in experiments using

olid piece of amphibole as starting materials, whereas it formedn experiments using powdered starting materials. The duration ofxperiments may also influence the proportion of garnet that formsWolf and Wyllie, 1993; this study). As a result, a critical consid-ration of the approach to equilibrium conditions in amphiboliteehydration-melting experiments is crucial to assess the amountf garnet that can be produced at a given individual P–T condition,nd consequently, to determine the equilibrium melt composition.

To better understand the effect of composition of the startingaterial on garnet stability in an amphibolite system, we compare

ome recent experimental and modeling studies which are basedn slightly different starting materials (Table 1) to constrain the P–Tonditions of the formation of TTGs. The starting material (3VG)f amphibolite used in the experiments of Nair and Chacko (2008)ontains lower Al2O3 and higher SiO2 contents than modern MORBsnd our starting material, which might explain the lower propor-ions of garnet in their experimental runs at a given P–T condition.imilarly, the starting amphibolite (PC-227) used in the experi-

ents of Adam et al. (2012) has higher SiO2 and MgO contents thanodern MORBs, which would also reduce the amount of garnet at

quilibrium (Green and Ringwood, 1967); although the high Al2O317.19 wt%) of PC-227 would potentially promote garnet growth.

arch 231 (2013) 206– 217

The bulk composition used by Nagel et al. (2012) to model phaseproportions is representative of a very mafic Isua tholeiite, which isextremely poor in Al2O3 (8.4 wt%) and rich in MgO (14.2 wt%). UsingTheriak/Domino (de Capitani and Petrakakis, 2010), Nagel et al.(2012) predicted up to 10 wt% garnet at 10 kbar (ca. 920 ◦C) and con-stant proportions of ∼20 wt% garnet at 14 kbar over 700–1000 ◦C.However, their modeling yielded melt compositions which haveFeO and MgO contents that are slightly too high compared withthat of Early Archean TTGs (Table 3). This inconsistent estimationof the melt composition may in turn affect the phase proportionsin the calculation. A detailed review of the starting materials ofall previous amphibole melting experiments is not the purpose ofthis paper, but from the three examples discussed above, it is clearthat the amounts of garnet can change significantly as a result ofsmall compositional variations. However, our experimental results,together with those of Wolf and Wyllie (1993, 1994), as well as thethermodynamic modeling of Nagel et al. (2012), clearly indicatethat at least 10 wt% of garnet, and possibly more, could be presentin the solid residue of partially melted amphibolites at pressuresof 10–12 kbar and temperatures of 900 ◦C. As shown in Table 1,except for a slightly lower SiO2 and higher FeO content, the start-ing material of synthetic amphibolite (mixed powder) used in thisstudy has a bulk major-element composition similar to that of E-MORB (Table 1), with similar contents of Al2O3 (15–17 wt%), MgO(7–8 wt%), CaO (11–12 wt%) and Na2O (2–3 wt%). If such a compo-sition resembles that of early Earth’s simatic crust (Hadean or EarlyArchean; e.g., Furnes et al., 2007; Hoffmann et al., 2011; Smithieset al., 2009), then our experiments can be applied to simulatethe formation of the early felsic rocks derived from these ancientbasalts.

3.2. Early Earth’s oceanic plateaus

In the Early and Middle Archean, portions of the mantle hadpotential temperatures that were ca. 300–400 ◦C higher than whattypifies modern mid-ocean ridges (ca. 1350 ◦C), as demonstrated bythe existence of Archean high-Mg basaltic and komatiitic magmas(Herzberg et al., 2010), and as implied by the modeling of Earth’sthermal history (Korenaga, 2006). As a result of these higher tem-peratures, more extensive mantle melting is predicted compared tomodern mid-ocean ridges. When the possible action of plume-likeupwelling is also considered, then it is plausible to infer the devel-opment of segments of “oceanic” simatic crust in the Hadean andEarly Archean up to ca. 40 km (Korenaga, 2006; Vlaar et al., 1994).The mantle plume model has been evoked to explain the geody-namic setting for magmatism in several Archean terrains (Campbellet al., 1989; Ernst and Buchan, 2003), such as Yilgarn and Pilbara inWest Australia (Said and Kerrich, 2009), Abitibi in North America(Fan and Kerrich, 1997), South India (Jayananda et al., 2000) andNorth China (Zhao et al., 1998, 1999; Zhao et al., 2001). Possiblemodern analogs of Archean oceanic plateaus include the OntongJava plateau (Neal et al., 1997; Tarduno et al., 1991) and the Icelandplateau (Willbold et al., 2009).

Regardless of specific tectonic settings, the Early ArcheanTTG suites are believed to have been formed by partial meltingof hydrous metabasites at high pressures with abundant gar-net and minor titanate phases in the residue (Condie, 2005).Oxygen and hafnium isotope ratios of Hadean detrital zirconsimply that the hydration process of the Early Archean protocrustshould have involved terrestrial fluids and rapid crustal recycling(Hawkesworth et al., 2010). It has been suggested (Bindemanet al., 2005) that TTGs with these characteristic isotopic signatures

originate when oceanic crust affected by seawater alteration is sub-ducted and melts. Alternatively, as suggested by Bédard (2006a),hydrothermal processes or seawater-rock interaction occurringbetween the emplacements of successive lava flows during volcanic

C. Zhang

et al.

/ Precam

brian R

esearch 231 (2013) 206– 217

211

Table 3Composition of run products at 12 kbar (wt%)a

T (◦C) time (days) N SiO2 � TiO2 � Al2O3 � FeOtot � MnO � MgO � CaO � Na2O � K2O � Total Mg#

Amp 800 8 8 43.54 0.84 1.43 0.13 15.22 1.32 15.11 0.54 0.21 0.07 9.37 0.59 11.50 0.18 1.87 0.12 0.46 0.04 97.72 52.5850 8 8 43.62 1.49 1.64 0.14 14.94 1.37 14.28 0.89 0.19 0.05 10.56 0.91 11.39 0.11 2.35 0.14 0.49 0.05 99.74 56.9900 8 9 43.26 1.05 2.14 0.25 15.90 0.77 11.31 0.59 0.13 0.05 11.24 0.70 11.60 0.41 2.50 0.15 0.47 0.04 98.62 63.9900 24 17 42.08 0.86 2.26 0.43 15.11 0.53 12.52 0.36 0.11 0.05 11.12 0.45 11.06 0.28 2.76 0.11 0.58 0.09 97.77 61.3950 8 8 42.03 0.67 2.74 0.26 15.77 0.57 12.36 0.32 0.09 0.04 10.98 0.43 11.20 0.23 2.98 0.10 0.52 0.02 98.72 61.3

Plg 800 8 8 51.70 0.22 31.01 0.13 0.76 0.12 14.12 0.12 3.70 0.07 0.09 0.03 101.38 67.5850 8 8 52.63 0.88 30.51 0.98 0.60 0.10 13.53 0.57 3.81 0.06 0.11 0.07 101.19 65.8900 8 8 51.00 0.15 30.86 0.24 0.46 0.12 14.08 0.18 3.64 0.09 0.09 0.02 100.13 67.8900 24 8 50.92 1.45 31.60 0.98 0.39 0.08 14.27 1.10 3.73 0.61 0.07 0.02 100.98 67.9950 8 5 51.78 0.27 30.92 0.17 0.49 0.06 13.73 0.15 3.97 0.11 0.10 0.01 100.99 65.3

1000 8 4 50.97 0.19 30.77 0.24 0.50 0.17 13.90 0.08 3.74 0.07 0.09 0.01 100.78 66.91000 16 7 52.09 0.49 30.46 0.26 0.41 0.07 13.55 0.31 4.04 0.17 0.09 0.01 100.64 64.6

Cpx 800 8 4 50.17 0.54 0.74 0.10 7.84 1.71 10.12 0.59 0.28 0.07 8.42 0.51 21.34 0.76 0.77 0.20 0.03 0.02 99.71 59.7850 8 5 49.71 0.58 0.77 0.18 7.85 1.06 8.65 0.43 0.14 0.04 10.57 0.54 21.50 0.57 0.90 0.12 0.04 0.02 100.13 68.5900 8 5 49.44 1.05 1.11 0.19 9.26 1.11 7.48 0.59 0.18 0.10 10.52 0.47 21.89 0.64 0.82 0.08 0.04 0.03 100.74 71.5900 24 8 48.08 0.92 0.88 0.18 8.88 1.34 10.23 1.00 0.22 0.08 10.04 0.67 20.14 0.69 0.85 0.12 0.03 0.04 99.35 63.6950 8 5 49.01 0.64 1.05 0.12 9.33 0.80 9.67 0.85 0.14 0.03 10.12 0.25 20.24 0.32 1.03 0.14 0.02 0.01 100.61 65.1

1000 8 7 49.53 0.78 1.25 0.14 9.18 0.84 8.51 0.88 0.12 0.07 10.96 0.63 20.26 1.13 1.18 0.29 0.06 0.06 101.05 69.71000 16 12 48.26 0.63 1.55 0.19 10.33 0.82 7.66 0.59 0.09 0.04 10.29 0.22 20.90 0.49 1.26 0.05 0.01 0.01 100.35 70.5

Grt 800 8 C/5 38.20 0.60 1.25 0.15 20.88 0.48 20.99 0.45 1.57 0.14 3.13 0.31 14.25 0.50 0.04 0.03 0.01 0.01 100.32 21.0R/5 38.48 0.30 1.16 0.11 21.36 0.21 20.89 0.63 1.05 0.20 3.54 0.10 13.89 0.94 0.04 0.03 0.00 0.00 100.41 23.2

850 8 C/5 38.69 0.39 1.25 0.07 21.11 0.49 21.85 0.44 1.21 0.32 4.52 0.56 12.71 0.33 0.04 0.02 0.01 0.01 101.39 26.9R/8 39.01 0.57 1.12 0.10 21.54 0.64 20.48 0.69 0.68 0.09 5.40 0.28 12.86 0.48 0.16 0.20 0.02 0.02 101.27 32.0

900 8 C/5 38.79 0.42 1.39 0.24 20.62 0.72 21.16 0.74 0.90 0.16 5.77 0.71 12.26 0.55 0.07 0.03 0.01 0.01 100.97 32.7R/8 38.89 0.35 1.02 0.13 21.33 0.20 18.75 0.61 0.68 0.11 7.34 0.18 11.82 0.39 0.06 0.02 0.02 0.01 99.91 41.1

900 24 C/5 38.74 0.37 1.33 0.08 20.74 0.16 22.26 0.46 0.76 0.21 6.02 0.44 10.84 0.49 0.04 0.03 0.01 0.01 100.74 32.5R/8 38.55 0.15 0.77 0.13 21.29 0.17 20.97 0.15 0.44 0.06 6.91 0.13 10.76 0.32 0.06 0.02 0.01 0.01 99.76 37.0

950 8 C/5 38.84 0.29 1.72 0.33 20.58 0.26 22.08 0.75 0.71 0.10 6.48 0.39 10.64 0.52 0.04 0.03 0.01 0.00 101.10 34.4R/8 39.50 0.24 0.82 0.09 21.56 0.28 20.18 0.36 0.43 0.06 7.25 0.21 10.91 0.27 0.09 0.04 0.02 0.02 100.76 39.0

1000 8 C/8 38.34 0.71 1.59 0.16 19.82 0.64 21.62 0.93 0.71 0.07 6.53 0.75 10.19 0.27 0.04 0.04 0.01 0.01 98.85 35.0R/8 39.23 0.64 0.90 0.28 21.13 0.59 20.43 0.50 0.44 0.09 7.55 0.28 10.34 0.39 0.07 0.05 0.01 0.01 100.10 39.7

1000 16 C/8 38.90 0.79 1.43 0.53 20.94 0.90 20.49 1.19 0.63 0.22 7.52 0.71 10.56 0.37 0.09 0.01 0.01 0.02 100.57 39.6R/8 39.64 0.37 0.62 0.18 22.01 0.46 19.17 0.11 0.55 0.09 8.27 0.25 10.27 0.28 0.04 0.01 0.01 0.01 100.58 43.5

Melt 900 24 8 68.63 0.56 0.41 0.05 18.33 0.21 3.11 0.18 0.06 0.05 0.69 0.05 4.52 0.08 2.72 0.44 1.55 0.08 100.00 28.2950 8 8 65.21 0.94 0.64 0.06 19.42 0.28 3.20 0.19 0.04 0.05 0.90 0.08 4.44 0.24 4.65 0.47 1.48 0.10 100.00 33.3

1000 8 6 65.51 1.29 0.72 0.10 20.57 0.59 3.09 0.29 0.04 0.05 0.75 0.34 4.56 0.64 3.27 0.69 1.50 0.05 100.00 30.41000 16 6 61.97 1.15 0.94 0.08 20.23 0.34 4.43 0.88 0.10 0.07 1.17 0.32 4.82 0.54 5.04 0.24 1.29 0.08 100.00 32.1

TTG b 70.4 2.9 0.31 0.14 15.2 1.1 2.51 1.1 0.06 0.02 0.96 0.62 2.74 0.88 4.71 0.61 2.22 0.68 99.49 40.8

W&W c 66.34 0.33 18.95 3.99 0.05 1.36 7.50 1.13 0.40 100.05 37.8Adam d 62.58 0.53 0.56 0.02 18.15 0.10 4.08 0.14 0.05 0.01 3.10 0.11 6.80 0.10 2.51 0.04 2.11 0.03 99.97 57.5Nagel e 64.10 0.14 14.96 7.13 3.86 2.29 7.51 100.00 49.1

The italics show that they are values of standard deviation.a N denotes analytical amount; An% of plagioclase is listed in column of Mg#; H2O content of amphibole is derived by difference method; both core (C) and rim (R) of garnet are analyzed; melt compositions are normalized to

anhydrous 100 wt%.b Average composition of Early Archean (≥3.5 Ga) high-Al TTG suites (Condie, 2005).c Experimental melt of an amphibolite (MgO = 10.8 wt%) at 10 kbar and 900 ◦C (duration of 9 days) (Wolf and Wyllie, 1994).d Experimental melt of a Nuvvuagittuq greenstone (PC-227, see Table 1) at 15 kbar and 1050 ◦C (Adam et al., 2012).e Modeled melt composition derived from a Isua tholeiite (2000–13, see Table 1) with 10% melting at 14 kbar (Nagel et al., 2012).

2 n Rese

cgps(pg

dEtAmt∼nbeacp

3

Tirw(mac(toewab

dcrnpdc(Rcs2CatEeMae1capX

12 C. Zhang et al. / Precambria

onstruction of plateau-type crust would produce identical oxy-en isotopic characteristics, as shown by some Phanerozoic oceaniclateaus (Kerr et al., 1998). Therefore, even in the absence of oceanicubduction, extensive partial melting of hydrated metabasaltsamphibolites) could occur at the base of oceanic plateaus (atressure of ∼12 kbar) and could be an important mechanism forenerating the earliest felsic continental crust.

Reconciling the experimental constraints on phase assemblageserived from our experiments and the thermal modeling of thearly Archean oceanic crust has important geological implica-ions. As indicated by modeled geothermal gradients for the Earlyrchean Earth (Korenaga, 2006; Vlaar et al., 1994), the base of aature 20-Ma old hydrated oceanic protocrust can reach tempera-

ures as high as 900 ◦C and melting at such conditions can generate10 wt% felsic melt and leave a restite containing ∼20 wt% gar-et (melting zone labeled “Z” in Fig. 4). The amount of garnet maye much lower if the pressure is lower, but in any case, garnet isxpected to be present in the residue of dehydration melting of

20-Ma old oceanic protocrust. For a younger (thinner) oceanicrust, however, garnet is not expected to occur in the residue whenartial melting occurs at temperature ≥900 ◦C (Fig. 4).

.3. Generation of Early Archean TTG suites

The oldest continental crustal remnants are the Early ArcheanTG suites (Arth et al., 1978; Jahn et al., 1981), which are character-zed by high Al2O3 and Na2O, low Mg# and strongly fractionatedare earth element (REE) and incompatible element patterns,ith high Sr/Y and La/Yb ratios and negative Nb-Ta anomalies

Bédard, 2006a; Condie, 2005; Foley et al., 2002). In the experi-ents reported here, the partial melts of amphibolite produced

t 12 kbar and 900–1000 ◦C have Na2O contents (2.72–5.04 wt%)omparable to the average composition of Early Archean TTG suites4.71 ± 0.61 wt% Na2O), while the Al2O3 contents of the experimen-al melts (18–20 wt%) are similar to the high-Al2O3 endmemberf the Early Archean TTG suites (14.5–19 wt% Al2O3; Drummondt al., 1996). The Mg# values of Early Archean TTG suites are mainlyithin the range 20–50 (Condie, 2005; Martin et al., 2005), which

re clearly lower than slab-derived adakites (mainly within 50–70)ut cover the Mg# of our experimental melts (28–32).

In order to model the trace-element composition of partial meltserived from the Early Archean protocrust using the experimentalonstraints, we must estimate the composition of the starting mate-ial, i.e., a basaltic rock derived from a primitive mantle which hasot been extensively depleted by previous melt extraction. In com-arison to modern N-MORB, which is thought to be derived from aepleted mantle, Early Archean basalts should contain higher con-entrations of lithophile elements (e.g., Rb and Ba) and light REELREE; e.g., La and Ce) but probably lower concentrations of heavyEE (e.g., Yb and Y) (see discussion in Smithies et al., 2009). Suchompositions occur in the Early Archean Isua greenstone belt inouthwest Greenland (Hoffmann et al., 2011; Polat and Hofmann,003) and in the pre-3.5 Ga lower Warrawoona Group of the Pilbararaton (Smithies et al., 2009). Here we use different starting materi-ls, including the average composition of Eoarchean metabasalts ofhe Isua Supracrutal Belt, the average compositions of N-MORB and-MORB, and E-MORB from different locations, to model the tracelements of partial melts generated by batch melting (Table 4).inor amounts of titanite (0.5 wt%) and rutile (0.5 wt%) were also

ssumed to be present in the residue because these minerals gen-rally occur as common accessory phases in amphibolites (e.g.,2 kbar, 900 ◦C; John et al., 2011). Most of the employed partition

oefficients for trace element modeling are after Bédard (2006a),nd some important calibrations dependent on P–T conditions andhase compositions have been made (Bédard, 2006b; Foley, 2008;iong et al., 2011). It is emphasized that rutile will not affect the

arch 231 (2013) 206– 217

Sr/Y ratios or REE concentrations but mostly the concentrations ofsome high-field-strength elements (such as Ti, Zr, Nb, Hf and Ta).

Fig. 5 shows the calculated trace element concentrations inmelts formed at 12 kbar and 900 ◦C from different starting compo-sitions. The target TTG compositions shown in Fig. 5 correspond tothe Early Archean (>3.5 Ga) TTG suites, which represent the old-est felsic continental crustal relics. Compared to younger TTGs,these have lower Sr but similar Y contents (Condie, 2005; Kellerand Schoene, 2012; Martin et al., 2005). The results of the mod-eling indicate that, using the average trace element compositionof global E-MORB (Sun and McDonough, 1989), E-MORB from EastPacific Rise (Niu and Batiza, 1997) or E-MORB from Mid-AtlanticRidge (Dosso et al., 1993) as starting materials, Early Archean high-Al TTGs can be simulated if melting conditions are ∼12 kbar with20 wt% garnet in the residue. It is notable that starting materials ofE-MORB from the Macquarie Island, SW Pacific (Kamenetsky et al.,2000) and E-MORB from the Antarctic-Phoenix Ridge (Choe et al.,2007) would lead to melt compositions with higher concentrationsof lithophile elements and LREE when compared with average EarlyArchean high-Al TTG (Condie, 2005). In contrast, if an N-MORBis used as a starting material (Sun and McDonough, 1989), thelithophile elements and LREE enrichments observed in TTGs can-not be simulated. This implies that the Archean metabasaltic sourceto the TTGs was significantly less “depleted” than N-MORB. Exceptfor Ba, Nb and Ta, the modeled partial melt of average Eoarcheanmetabasalts of the Isua Supracrutal Belt (Hoffmann et al., 2011;Polat and Hofmann, 2003) can generate TTG-like melts for mosttrace elements. The strong depletions in Nb and Ta of the par-tial melt of the Isua metabasalt are primarily inherited from theprotolith, which could result from recycling of previous crustalmaterials by subduction process (Shirey et al., 2008) or multipleintra-crustal remelting events (Bédard, 2006a).

In conclusion, we demonstrate that the geochemistry of EarlyArchean TTGs can be reproduced at the P–T conditions of ∼12 kbarand ∼900 ◦C of a mature oceanic plateau on the early Earth(∼3.5–4.0 Ga), in agreement with geophysical modeling (Fig. 4;Vlaar et al., 1994; Korenaga, 2006). Using the average E-MORBcomposition (Sun and McDonough, 1989) as the starting material,the experimental residual modes yield model melts with Sr/Y andLa/Yb ratios appropriate for those observed in the Early ArcheanTTGs (Fig. 6). Modeled Sr/Y and La/Yb ratios of partial melts derivedfrom different starting materials discussed above are shown in Sup-plementary Fig. 3 for comparison. The data confirm that the mostoptimal simulation of the formation of Early Archean TTGs fromthe average E-MORB composition is obtained if the phase assem-blage and proportions from the experiment at 12 kbar and 900 ◦Care considered. Of course, pressure higher than 12 kbar could alsolead to garnet proportions ≥20% to model the high Sr/Y and La/Ybratios, but the crucial point proved by this study is that very highpressures (≥15 kbar) and hence very great crustal thickness are notrequired. Assuming a pressure of 12 kbar, a temperature of ∼900 ◦Cis the lower limit to generate 20% garnet with an appropriate meltcomposition.

Although crustal thickening due either to subduction-induced(Nagel et al., 2012) or mantle traction-induced (Bédard et al., 2013)converge can also explain such a moderately thickened crust, thesemechanisms are not required to explain Early Archean TTG gene-sis, since oceanic plateaus with appropriate thickness were presentprior to the initiation of subduction (Korenaga, 2006; Shirey andRichardson, 2011). Furthermore, a high temperature of ca. 900 ◦Cof oceanic crust can hardly be achieved in a subduction settingat 12–15 kbar (see thermo-mechanical models of Syracuse et al.,

2010; van Keken et al., 2002), even in case of a hot subduc-tion geotherm for the Early Archean (Martin and Moyen, 2002);typical Mesoarchean metamorphic rocks which are proposed tohave formed by subduction processes (e.g., Moyen et al., 2006;

C. Zhang

et al.

/ Precam

brian R

esearch 231 (2013) 206– 217

213

Table 4Starting material, partition coefficients and results of trace element modelinga

N-MORB E-MORB Partition coefficients (D)b N-Melt E-Melt TTG

Global EPR APR Grt Cpx Amp c Plg d Ttn Rut e Global EPR APR

Rb 0.56 5.04 3.3 20.19 0.0007 0.01 0.055 0.068 0.5 0.0076 4.10 36.87 24.14 147.71 76Ba 6.3 57 40.27 262.03 0.0004 0.006 0.046 0.2 1.5 0.0043 39.70 359.21 258.59 1682.62 500Th 0.12 0.6 0.44 3.55 0.0075 0.104 0.055 0.095 0.16 0.2 0.76 3.80 2.77 22.36 4.1U 0.047 0.18 0.15 1 0.024 0.032 0.05 0.091 0.14 0.2 0.32 1.24 1.03 6.86 1.2Nb 2.33 8.3 6.47 45.04 0.04 0.007 0.44 0.239 2.2 217.1 1.80 6.41 5.14 35.77 6.1Ta 0.132 0.47 0.39 3.15 0.08 0.028 0.22 0.053 6.55 195.1 0.12 0.42 0.35 2.87 0.41La 2.5 6.3 6.08 24.46 0.028 0.028 0.319 0.358 4.73 0.0057 8.13 20.49 21.49 86.46 22Ce 7.5 15 16.21 50.54 0.08 0.059 0.56 0.339 7.57 0.0065 17.91 35.83 42.63 132.91 40Sr 90 155 152.73 329.62 0.019 0.032 0.389 1.0 2.68 0.036 208.12 358.43 399.53 862.25 362Nd 7.3 9 11.94 24.83 0.222 0.115 1.32 0.289 12.4 0.0082 9.85 12.14 18.13 37.71 18Sm 2.63 2.6 3.62 5.82 1.43 0.259 2.09 0.237 14 0.0954 2.09 2.06 3.22 5.18 2.9Zr 74 73 109.8 213.34 0.537 0.125 0.417 0.078 1.92 3.7 181.41 178.96 286.97 557.58 152Hf 2.05 2.03 2.74 5.11 0.431 0.208 0.781 0.069 2.43 4.97 3.79 3.75 5.52 10.29 3.8Eu 1.02 0.91 1.3 1.96 1.54 0.341 1.79 2.17 13.8 0.00037 0.67 0.60 0.98 1.47 0.82Ti 7600 6000 9761 13,225 2.63 0.473 4.03 1.0 67 45 2729.6 2154.9 3915.4 5304.9 1858Gd 3.68 2.97 4.53 6.46 4.84 0.422 2.53 0.192 11.9 0.0106 1.80 1.45 2.45 3.50 2.2Dy 4.55 3.55 5.11 6.34 11.5 0.57 2.55 0.15 8.27 0.0116 1.40 1.09 1.71 2.12 1.8Y 28 22 30.22 33.99 14.1 0.603 2.47 0.138 5.42 0.0118 7.60 5.97 8.90 10.01 8.5Er 2.97 2.31 3.15 3.77 18.8 0.64 2.22 0.117 5.54 0.0122 0.67 0.52 0.76 0.91 0.77Yb 3.05 2.37 3.04 3.48 23.2 0.635 1.79 0.094 3.02 0.0126 0.60 0.47 0.64 0.74 0.82Sr/Y 3.21 7.05 5.05 9.70 28.6 62.6 44.90 86.16 72 ± 24(La/Yb)N 0.56 1.81 1.36 4.77 9.3 30.1 22.68 79.70 25 ± 15Nb/Ta 17.65 17.66 16.59 14.30 15.4 15.4 14.48 1.99 12 ± 2.8

a N-MORB and global E-MORB (Sun and McDonough, 1989) are used as starting compositions for trace element modeling, and N-Melt and global E-Melt are modeled partial melts from them respectively. Average compositionof Early Archean high-Al TTG (Condie, 2005; Martin et al., 2005) is listed for comparison. La/Yb is normalized to chondrite (McDonough and Sun, 1995). E-MORBs from EPR (East Pacific Rise; Niu and Batiza, 1997) and from APR(Antarctic-Phoenix Ridge; Choe et al., 2007) are also listed, and modeled partial melt compositions derived from them are illustrated in Fig. 6.

b Most partition coefficients are after Bédard (2006a), and some adjustments are made and noted below.c For amphibole, because DNb and DTa vary as a function of Mg# (Foley et al., 2002), their values are after Foley (2008) for amphibole composition with Mg# of 60, which is consistent to our experimental amphibole composition

(Table 3).d For plagioclase, because DBa, DSr and DTi vary as a function of An content (Bédard, 2006b), these values are adjusted to An = 67 (Table 3).e For rutile, because DNb and DTa are dependent on temperature, pressure and melt water content (Xiong et al., 2011), these values are adjusted to T = 900 ◦C, P = 12 kbar and H2O content = 5.34 wt%.

214 C. Zhang et al. / Precambrian Research 231 (2013) 206– 217

0.1

1

10

100

1000

norm

aliz

edt o

prim

itive

man

tle

Early Archean TTG (with 1 ) σ

Partial melt of E-MORB (EPR)

Partial melt of E-MORB ( )MacPartial melt of E-MORB (APR)

Partial melt of E-MORB (MAR)

(a)

(b)

norm

aliz

edto

prim

itive

man

tle

Early Archean TTG (with 1 )σPartial melt of N-MORB (S&M)Partial melt of E-MORB (S&M)Partial melt of Eoarchean metabasalt

0.1

1

10

100

1000

Ba Th U Nb Ta La Ce Sr Nd Sm Zr Hf Eu Ti Gd Dy Y Er YbRb

Fig. 5. Modeled trace element distribution patterns assuming batch melting, based on phase relations of amphibolite dehydration-melting experiment at 12 kbar and 900 ◦C(see Fig. 3) and assuming minor titanite (0.5 wt%) and (0.5 wt%) rutile in the residue. The partition coefficients applied are listed in Table 3. Modeling results using the followingstarting materials are shown: (a) N-MORB and E-MORB (Sun and McDonough, 1989), and average value of the least altered Eoarchean metabasalts of the Isua SupracrutalBelt (Polat and Hofmann, 2003; Hoffmann et al., 2011); (b) E-MORB (MAR) (Seamount lavas from 10◦ to 24◦ N North Mid-Atlantic Ridge; Dosso et al., 1993), E-MORB (EPR)(Seamount lavas from 5◦ to 15◦ N East Pacific Rise; Niu and Batiza, 1997), E-MORB (APR) (Basaltic lavas from Antarctic-Phoenix Ridge; Choe et al., 2007), E-MORB (Mac)( 0). Pric .

vto(s≤iRom

3

ror2c

Near-primitive glasses from Macquarie Island, SW Pacific; Kamenetsky et al., 200omposition of Early Archean high-Al TTGs (Condie, 2005) is shown for comparison

an Hunen and Moyen, 2012) reveal peak metamorphic tempera-ures clearly lower than 800 ◦C at pressures ≤15 kbar. The formationf Early Archean high-Al TTGs from melting of hydrous metabasaltsamphibolites) at the base of mature oceanic plateaus in a non-ubduction setting also accounts for the low MgO contents (mostly1 wt%) of the Early Archean TTGs, which excludes a common

nteraction between felsic melts and mantle ultramafic rocks (e.g.,app et al., 2010), which should be an inevitable outcome ofceanic subduction-induced partial melting and ascent of felsicelt through the mantle wedge.

.4. Delamination of garnet amphibolite

Delamination of dense rocks, which may correspond to theestite of dehydration melting of amphibolites after extraction

f TTG-composition melts, is hypothesized to play a catalyticole in the growth of Early Archean continental crust (Bédard,006a, 2013). As demonstrated by this experimental study, garnet-linopyroxene-bearing amphibolite is likely to represent restite

mitive mantle for normalization is after Sun and McDonough (1989). The average

from metabasalt anatexis (Fig. 3), which however is mineralogi-cally different from eclogitic restites generated at higher pressures(Anderson, 2005). The density contrast between the crustal restiticrocks and the underlying mantle (assumed to have a lherzoliticcomposition) exerts a primary control on whether or not delamina-tion is possible. Based on mineral densities and phase proportions at12 kbar and 900 ◦C, the estimated density of a restitic garnet amphi-bolite is 3339 kg/m3 whereas the density of lherzolite is 3243 kg/m3

(Table 5). Thus, the density of garnet amphibolite similar to ourexperimental restites is slightly higher than that of the subjacentmantle, and so delamination is mechanically possible. Our den-sity estimation is consistent with the thermodynamic modelingof Hacker et al. (2003), which also indicates that metamorphosedMORB (garnet amphibolite) is slightly denser than lherzolite at850–900 ◦C and ∼12 kbar. In addition to this density contrast, the

viscosity of the ambient mantle also affects the rate of delamina-tion. Temperature influences strongly the viscosity of rocks andan increase of 100 ◦C would reduce the mantle viscosity by ca. 1order of magnitude (Karato and Wu, 1993). Considering that the

C. Zhang et al. / Precambrian Research 231 (2013) 206– 217 215

100 150 200 250 300 350 400 450 5000

10

20

30

40

50

60

70

80

90

100

110

20 64 10 8 12 14 16 180

5

10

15

20

25

30

35

40

45

(/

)La

Yb

N

Sr/

Y

Sr (ppm) YbN Sr/Y

(/

)La

Yb

N

0

5

10

15

20

25

30

35

40

45

100 20 30 40 50 60 70 80 90 100 110

800°C850°C

900°C

950°C

1000°C

800°C

850°C

900°C

950°C1000°C

800°C

850°C

900°C

950°C

1000°CEarly Archean TTG (with )1σ

E-MORB (S&M)

Partial melt of E-MORB

(a) (b) (c)

Fig. 6. Selected modeled trace element characteristics of partial melts of amphibolite at 12 kbar and various temperatures. (a) Sr versus Sr/Y. (b) YbN versus (La/Yb)N (N,normalized to average chondrite, McDonough and Sun, 1995). (c) Sr/Y versus (La/Yb)N. Models are based on batch melting using E-MORB (Sun and McDonough, 1989) asstarting material and phase relations determined in this amphibolite dehydration-meltinin the residue. The partition coefficients applied are listed in Table 3. Note that the melt fhigh-Al TTGs (Condie, 2005).

Table 5Density estimation for melt, melting residue (garnet amphibolite) and upper mantle(lherzolite)a

Phases Density (kg/m3)

Melt 2300Olivine 3237Orthopyroxene 3224Clinopyroxene 3289Amphibole 3269Plagioclase 2780Garnetb 3967Lherzolitec 3243Garnet amphibolited 3339

a Except for melt, whose density is estimated upon Hack and Thompson (2011) at12 kbar, all the other densities are estimated upon the calculation sheet of Hackerand Abers (2004) at 12 kbar and 900 ◦C.

b Garnet density is estimated supposing a simple linear relationship of bulk den-sity with endmember proportions (ca. Prp25Alm50Grs25, 12 kbar and 900 ◦C, seeTable 3).

c Lherzolite density is estimated using a supposed mineral assemblage ofO

a

Epc2EptuK

4

wo∼nchcos

l45Opx35Cpx20.d Garnet amphibolite (melting residue) formed at 12 kbar and 900 ◦C has a mineral

ssemblage of Amp38Cpx24Grt21Cpx17 (also see Table 2 and Fig. 3).

arly Archean mantle was much hotter than present-day mantle,ossibly by as much as 350 ◦C, a small density excess may be suffi-ient to trigger extensive and rapid delamination (Jull and Kelemen,001). If crustal restite delamination operated extensively in thearly Archean (Bédard, 2006a, 2013), it would shift the bulk com-osition of continental crust from basaltic to andesitic, introduceerrestrial fluids into the mantle, promote compensatory mantlepwelling, and might induce re-melting of TTG suites to form more2O-rich granites (Moyen, 2011).

. Conclusions

Dehydration-melting of amphibolite at 12 kbar and 900 ◦C,hich are conditions that are appropriate to the base of thick

ceanic plateaus in the Hadean and Early Archean, may generate10 wt% felsic melt and a coexisting restite containing ∼20 wt% gar-et. The major element composition and the modeled trace elementomposition of the felsic partial melts are similar to Early Archean

igh-Al TTG suites, reproducing their high Al2O3 and low MgOontents, and modestly high Sr/Y and La/Yb ratios. After extractionf the partial melts to build the initial felsic continental crust, theolid garnet-rich residue has the potential to delaminate into the

g experiment (see Fig. 3). Minor titanite (0.5 wt%) and rutile (0.5 wt%) are assumedormed at ∼900 ◦C is best consistent with the average composition of Early Archean

underlying lherzolitic upper mantle because of its higher density.These results are permissive of vertical growth/basal anatexismodels for the initial generation of Early Archean continentalcrustal nucleii.

Acknowledgments

This work was funded by the German Research Council (DFG)grants to F. Holtz and J. Koepke. C. Zhang appreciates the financialsupports from the China Scholarship Council (CSC) and the GermanAcademic Exchange Service (DAAD). C. Ma acknowledges NationalNature Science Foundation of China (NSFC Grants 41272079 &90814004) and China Geological Survey (Grant 1212011121270)for supporting related studies. This paper benefited from theinsightful and helpful reviews of Editor-in-chief Guochun Zhao andtwo anonymous reviewers, and Elis Hoffmann’s review on an ear-lier version. This is NRCAN/ESS/GSC contribution no. (20130001).

Appendix A. Supplementary data

Supplementary data associated with this article can be found,in the online version, at http://dx.doi.org/10.1016/j.precamres.2013.03.004.

References

Adam, J., Rushmer, T., O’Neil, J., Francis, D., 2012. Hadean greenstones from theNuvvuagittuq fold belt and the origin of the Earth’s early continental crust.Geology 40, 363–366.

Anderson, D.L., 2005. Large igneous provinces, delamination, and fertile mantle.Elements 1, 271.

Armstrong, R.L., 1991. The persistent myth of crustal growth. Australian Journal ofEarth Sciences 38, 613–630.

Arth, J.G., Barker, F., Peterman, Z.E., Friedman, I., 1978. Geochemistry of thegabbro–diorite–tonalite–trondhjemite suite of southwest Finland and its impli-cations for the origin of tonalitic and trondhjemitic magmas. Journal of Petrology19, 289–316.

Beard, J.S., Lofgren, G.E., 1989. Effect of water on the composition of partial melts ofgreenstone and amphibolite. Science 244, 195–197.

Bédard, J.H., 2006a. A catalytic delamination-driven model for coupled genesis ofArchaean crust and sub-continental lithospheric mantle. Geochimica et Cos-mochimica Acta 70, 1188–1214.

Bédard, J.H., 2006b. Trace element partitioning in plagioclase feldspar. Geochimicaet Cosmochimica Acta 70, 3717–3742.

Bédard, J.H., 2013. How many arcs can dance on the head of a plume? A ‘Comment’

on: a critical assessment of Neoarchean ‘plume only’ geodynamics: evidencefrom the Superior province, by Derek Wyman, Precambrian Research. Precam-brian Research 229, 189–197.

Bédard, J.H., Harris, L.B., Thurston, P.C., 2013. The hunting of the snArc. PrecambrianResearch 229, 20–48.

2 n Rese

B

B

C

C

C

C

C

Cd

D

D

D

E

F

F

F

F

G

G

G

H

H

H

HH

H

H

H

H

J

16 C. Zhang et al. / Precambria

indeman, I.N., Eiler, J.M., Yogodzinski, G.M., Tatsumi, Y., Stern, C.R., Grove, T.L.,Portnyagin, M., Hoernle, K., Danyushevsky, L.V., 2005. Oxygen isotope evidencefor slab melting in modern and ancient subduction zones. Earth and PlanetaryScience Letters 235, 480–496.

rown, M., 2008. Characteristic thermal regimes of plate tectonics and their meta-morphic imprint throughout Earth history: when did Earth first adopt a platetectonics mode of behavior. In: Condie, K.C., Pease, V. (Eds.), When did Plate Tec-tonics Begin on Planet Earth? Geological Society of America Special Papers. , pp.97–128.

ampbell, I., Griffiths, R., Hill, R., 1989. Melting in an Archaean mantle plume: headsit’s basalts, tails it’s komatiites. Nature 339, 697–699.

hoe, W.H., Lee, J.I., Lee, M.J., Hur, S.D., Jin, Y.K., 2007. Origin of E-MORB in a fos-sil spreading center: the Antarctic-Phoenix Ridge, Drake Passage, Antarctica.Geosciences Journal 11, 185–199.

oldwell, B., Clemens, J., Petford, N., 2011. Deep crustal melting in the PeruvianAndes: felsic magma generation during delamination and uplift. Lithos 125,272–286.

ondie, K.C., 1980. Origin and early development of the earth’s crust. PrecambrianResearch 11, 183–197.

ondie, K.C., 1986. Origin and early growth rate of continents. Precambrian Research32, 261–278.

ondie, K.C., 2005. TTGs and adakites: are they both slab melts? Lithos 80, 33–44.e Capitani, C., Petrakakis, K., 2010. The computation of equilibrium assem-

blage diagrams with Theriak/Domino software. American Mineralogist 95,1006–1016.

osso, L., Bougault, H., Joron, J.L., 1993. Geochemical morphology of the North Mid-Atlantic Ridge, 10◦–24◦N: trace element-isotope complementarity. Earth andPlanetary Science Letters 120, 443–462.

rummond, M.S., Defant, M.J., 1990. A model for trondhjemite–tonalite–dacitegenesis and crustal growth via slab melting: archean to modern comparisons.Journal of Geophysical Research 95, 21503–21521.

rummond, M.S., Defant, M.J., Kepezhinskas, P.K., 1996. Petrogenesis of slab-derivedtrondhjemite–tonalite–dacite/adakite magmas. Transactions of the Royal Soci-ety of Edinburgh: Earth Sciences 87, 205–215.

rnst, R.E., Buchan, K.L., 2003. Recognizing mantle plumes in the geological record.Annual Review of Earth and Planetary Sciences 31, 469–523.

an, J., Kerrich, R., 1997. Geochemical characteristics of aluminum depleted andundepleted komatiites and HREE-enriched low-Ti tholeiites, western Abitibigreenstone belt: a heterogeneous mantle plume-convergent margin environ-ment. Geochimica et Cosmochimica Acta 61, 4723–4744.

oley, S., 2008. A trace element perspective on Archean crust formation and onthe presence or absence of Archean subduction. Geological Society of AmericaSpecial Papers 440, 31–50.

oley, S., Tiepolo, M., Vannucci, R., 2002. Growth of early continental crust controlledby melting of amphibolite in subduction zones. Nature 417, 837–840.

urnes, H., de Wit, M., Staudigel, H., Rosing, M., Muehlenbachs, K., 2007. A vestigeof Earth’s oldest ophiolite. Science 315, 1704–1707.

ao, S., Rudnick, R., Yuan, H., Liu, X., Liu, Y., Xu, W., Ling, W., Ayers, J., Wang, X., Wang,Q., 2004. Recycling lower continental crust in the North China craton. Nature432, 892–897.

irardi, J.D., Patchett, P.J., Ducea, M.N., Gehrels, G.E., Cecil, M.R., Rusmore, M.E.,Woodsworth, G.J., Pearson, D.M., Manthei, C., Wetmore, P., 2012. Elemental andisotopic evidence for granitoid genesis from deep-seated sources in the CoastMountains Batholith, British Columbia. Journal of Petrology 53, 1505–1536.

reen, D.H., Ringwood, A.E., 1967. An experimental investigation of the gabbro toeclogite transformation and its petrological applications. Geochimica et Cos-mochimica Acta 31, 767–833.

ack, A.C., Thompson, A.B., 2011. Density and viscosity of hydrous magmas andrelated fluids and their role in subduction zone processes. Journal of Petrology52, 1333–1362.

acker, B.R., Abers, G.A., Peacock, S.M., 2003. Subduction factory 1. Theoreticalmineralogy, densities, seismic wave speeds, and H2O contents. Journal of Geo-physical Research 108, 2029.

acker, B.R., Abers, G.A., 2004. Subduction factory 3: an excel worksheet and macrofor calculating the densities, seismic wave speeds, and H2O contents of mineralsand rocks at pressure and temperature. Geochemistry Geophysics Geosystems5, Q01005.

amilton, W.B., 2003. An alternative Earth. GSA Today 30, 4–12.arlov, D.E., Förster, H.-J., 2002. High-grade fluid metasomatism on both a local and

a regional scale: the Seward Peninsula, Alaska, and the Val Strona di Omegna,Ivrea–Verbano Zone, Northern Italy. Part I: Petrography and Silicate MineralChemistry. Journal of Petrology 43, 769–799.

arrison, T.M., 2009. The hadean crust: evidence from >4 Ga Zircons. Annual Reviewof Earth and Planetary Sciences 37, 479–505.

awkesworth, C.J., Dhuime, B., Pietranik, A.B., Cawood, P.A., Kemp, A.I.S., Storey,C.D., 2010. The generation and evolution of the continental crust. Journal of theGeological Society 167, 229–248.

erzberg, C., Condie, K., Korenaga, J., 2010. Thermal history of the Earth and itspetrological expression. Earth and Planetary Science Letters 292, 79–88.

offmann, J.E., Münker, C., Polat, A., Rosing, M.T., Schulz, T., 2011. The origin ofdecoupled Hf–Nd isotope compositions in Eoarchean rocks from southern West

Greenland. Geochimica et Cosmochimica Acta 75, 6610–6628.

ahn, B.-M., Glikson, A.Y., Peucat, J.J., Hickman, A.H., 1981. REE geochemistry andisotopic data of Archean silicic volcanics and granitoids from the Pilbara Block.Western Australia: implications for the early crustal evolution. Geochimica etCosmochimica Acta 45, 1633–1652.

arch 231 (2013) 206– 217

Jayananda, M., Moyen, J.F., Martin, H., Peucat, J.J., Auvray, B., Mahabaleswar, B., 2000.Late Archaean (2550–2520 Ma) juvenile magmatism in the Eastern Dharwar cra-ton, southern India: constraints from geochronology, Nd–Sr isotopes and wholerock geochemistry. Precambrian Research 99, 225–254.

Johannes, W., 1989. Melting of plagioclase-quartz assemblages at 2 kbar waterpressure. Contributions to Mineralogy and Petrology 103, 270–276.

Johannes, W., Koepke, J., 2001. Incomplete reaction of plagioclase in experimentaldehydration melting of amphibolite. Australian Journal of Earth Sciences 48,581–590.

John, T., Klemd, R., Klemme, S., Pfänder, J., Elis Hoffmann, J., Gao, J., 2011. Nb–Tafractionation by partial melting at the titanite–rutile transition. Contributionsto Mineralogy and Petrology 161, 35–45.

Jull, M., Kelemen, P., 2001. On the conditions for lower crustal convective instability.Journal of Geophysical Research 106, 6423–6446.

Kamenetsky, V.S., Everard, J.L., Crawford, A.J., Varne, R., Eggins, S.M., Lanyon, R.,2000. Enriched end-member of primitive MORB melts: petrology and geo-chemistry of glasses from Macquarie Island (SW Pacific). Journal of Petrology 41,411–430.

Karato, S.-I., Wu, P., 1993. Rheology of the upper mantle: a synthesis. Science 260,771–778.

Keller, C.B., Schoene, B., 2012. Statistical geochemistry reveals disruption in secularlithospheric evolution about 2.5 Gyr ago. Nature 485, 490–493.

Kerr, A.C., Tarney, J., Nivia, A., Marriner, G.F., Saunders, A.D., 1998. The internalstructure of oceanic plateaus: inferences from obducted Cretaceous terranes inwestern Colombia and the Caribbean. Tectonophysics 292, 173–188.

Klein, E.M., 2003. Geochemistry of the igneous oceanic crust. In: Rudnick, R.L. (Ed.),The Crust. Treatise in Geochemistry. Elsevier, Amsterdam, pp. 432–463.

Korenaga, J., 2006. Archean geodynamics and the thermal evolution of Earth. In:Benn, K., Mareschal, J.-C., Condie, K. (Eds.), Archean Geodynamics and Environ-ments. American Geophysical Union Monograph, Washington, DC, pp. 7–32.

Korenaga, J., 2008. Plate tectonics, flood basalts and the evolution of Earth’s oceans.Terra Nova 20, 419–439.

Labrosse, S., Jaupart, C., 2007. Thermal evolution of the Earth: secular changes andfluctuations of plate characteristics. Earth and Planetary Science Letters 260,465–481.

Martin, H., 1986. Effect of steeper Archean geothermal gradient on geochemistry ofsubduction-zone magmas. Geology 14, 753–756.

Martin, H., 1987. Petrogenesis of archaean trondhjemites, tonalites, and granodi-orites from eastern Finland: major and trace element geochemistry. Journal ofPetrology 28, 921–953.

Martin, H., Moyen, J. -F., 2002. Secular changes in tonalite–trondhjemite–granodiorite composition as markers of the progressive cooling of Earth. Geology30, 319–322.

Martin, H., Smithies, R.H., Rapp, R., Moyen, J.F., Champion, D., 2005. An overviewof adakite, tonalite–trondhjemite–granodiorite (TTG), and sanukitoid: relation-ships and some implications for crustal evolution. Lithos 79, 1–24.

McDonough, W.F., Sun, S.s., 1995. The composition of the Earth. Chemical Geology120, 223–253.

McKenzie, D., 1984. A possible mechanism for epeirogenic uplift. Nature 307,616–618.

Moyen, J.-F., 2011. The composite Archaean grey gneisses: petrological significance,and evidence for a non-unique tectonic setting for Archaean crustal growth.Lithos 124, 21–36.

Moyen, J.-F., Champion, D., Smithies, R.H., 2010. The geochemistry of Archaeanplagioclase-rich granites as a marker of source enrichment and depth of melting.Earth and Environmental Science Transactions of the Royal Society of Edinburgh100, 35–50.

Moyen, J.-F, Stevens, G., Kisters, A., 2006. Record of mid-Archaean subduction frommetamorphism in the Barberton terrain, South Africa. Nature 442, 559–562.

Moyen, J.-F., van Hunen, J., 2012. Short-term episodicity of archaean plate tectonics.Geology 40, 451–454.

Moyen, J., Stevens, G., 2006. Experimental constraints on TTG petrogenesis: implica-tions for Archean geodynamics. In: Been, K., Mareschal, J.-C., Condie, K.C. (Eds.),Archean Geodynamics and Environments. American Geophysical Union Mono-graph, Washington, DC, pp. 149–175.

Nagel, T.J., Hoffmann, J.E., Münker, C., 2012. Generation of Eoarcheantonalite–trondhjemite–granodiorite series from thickened mafic arc crust. Geol-ogy 40, 375–378.

Nair, R., Chacko, T., 2008. Role of oceanic plateaus in the initiation of subductionand origin of continental crust. Geology 36, 583–586.

Neal, C.R., Mahoney, J.J., Kroenke, L.W., Duncan, R.A., Petterson, M.G., 1997. Theontong java plateau. In: Mahoney, J.J., Coffin, M.F. (Eds.), Large Igneous Provinces:Continental, Oceanic, and Planetary Flood Volcanism. Geophys. Monogr. Ser.American Geophysical Union Geophysical Monograph, Washington, DC, pp.183–216.

Niu, Y., Batiza, R., 1997. Trace element evidence from seamounts for recycled oceaniccrust in the Eastern Pacific mantle. Earth and Planetary Science Letters 148,471–483.

Patino Douce, A.E., Beard, J., 1995. Dehydration-melting of biotite gneiss and quartzamphibolite from 3 to 15 kbar. Journal of Petrology 36, 707–738.

Petford, N., Gallagher, K., 2001. Partial melting of mafic (amphibolitic) lower crust

by periodic influx of basaltic magma. Earth and Planetary Science Letters 193,483–499.

Polat, A., Hofmann, A.W., 2003. Alteration and geochemical patterns in the3.7–3.8 Ga Isua greenstone belt, West Greenland. Precambrian Research 126,197–218.

n Rese

R

R

R

R

R

R

S

S

S

S

S

S

S

S

S

S

S

S

S

T

v

C. Zhang et al. / Precambria

app, R., Watson, E., Miller, C., 1991. Partial melting of amphibolite/eclogite and theorigin of Archean trondhjemites and tonalites. Precambrian Research 51, 1–25.

app, R.P., Norman, M.D., Laporte, D., Yaxley, G.M., Martin, H., Foley, S.F., 2010.Continent formation in the archean and chemical evolution of the cratonic litho-sphere: melt–rock reaction experiments at 3–4 GPa and petrogenesis of archeanmg-diorites (sanukitoids). Journal of Petrology 51, 1237–1266.

app, R.P., Shimizu, N., Norman, M.D.t, 2003. Growth of early continental crust bypartial melting of eclogite. Nature 425, 605–609.

ichardson, W.P., Okal, E.A., Van der Lee, S., 2000. Rayleigh–wave tomographyof the Ontong–Java Plateau. Physics of the Earth and Planetary Interiors 118,29–51.

ollinson, H.R., 2006. Crustal generation in the Archean. In: Brown, M., Rushmer, T.(Eds.), Evolution and Differentiation of Continetnal Crust. Cambridge UniversityPress, New York, pp. 173–230.

ushmer, T., 1991. Partial melting of two amphibolites: contrasting experimentalresults under fluid-absent conditions. Contributions to Mineralogy and Petro-logy 107, 41–59.

aid, N., Kerrich, R., 2009. Geochemistry of coexisting depleted and enriched ParingaBasalts, in the 2.7 Ga Kalgoorlie Terrane, Yilgarn Craton, Western Australia: evi-dence for a heterogeneous mantle plume event. Precambrian Research 174,287–309.

eidel, E., Okrusch, M., Kreuzer, H., Raschka, H., Harre, W., 1981. Eo-alpine meta-morphism in the uppermost unit of the Cretan nappe system — petrology andgeochronology. Contributions to Mineralogy and Petrology 76, 351–361.

en, C., Dunn, T., 1994. Dehydration melting of a basaltic composition amphibo-lite at 1.5 and 2.0 GPa: implications for the origin of adakites. Contributions toMineralogy and Petrology 117, 394–409.

hirey, S.B., Kamber, B.S., Whitehouse, M.J., Mueller, P.A., Basu, A.R., 2008. A reviewof the isotopic and trace element evidence for mantle and crustal processes inthe Hadean and Archean: implications for the onset of plate tectonic subduction.Geological Society of America Special Papers 440, 1–29.

hirey, S.B., Richardson, S.H., 2011. Start of the wilson cycle at 3 Ga shown bydiamonds from subcontinental mantle. Science 333, 434–436.

izova, E., Gerya, T., Brown, M., Perchuk, L.L., 2010. Subduction styles in the Precam-brian: Insight from numerical experiments. Lithos 116, 209–229.

mithies, R.H., Champion, D.C., Cassidy, K.F., 2003. Formation of Earth’s earlyArchaean continental crust. Precambrian Research 127, 89–101.

mithies, R.H., Champion, D.C., Van Kranendonk, M.J., 2009. Formation of Pale-oarchean continental crust through infracrustal melting of enriched basalt. Earthand Planetary Science Letters 281, 298–306.

pringer, W., Seck, H., 1997. Partial fusion of basic granulites at 5 to 15 kbar: implica-tions for the origin of TTG magmas. Contributions to Mineralogy and Petrology127, 30–45.

tern, R.J., 2005. Evidence from ophiolites, blueschists, and ultrahigh-pressure meta-morphic terranes that the modern episode of subduction tectonics began inNeoproterozoic time. Geology 33, 557–560.

tern, R.J., 2008. Modern-style plate tectonics began in Neoproterozoic time: Analternative interpretation of Earth’s tectonic history. In: Condie, K.C., Pease, V.(Eds.), When did Plate Tectonics Begin on Planet Earth? Geological Society ofAmerica Special Paper. , pp. 265–280.

un, S.S., McDonough, W.F., 1989. Chemical and isotopic systematics of oceanicbasalts: implications for mantle compositions and processes. In: Saunders, A.D.,Norry, M.J. (Eds.), Magmatism in the Ocean Basins. Geological Society of London,pp. 313–345.

yracuse, E.M., van Keken, P.E., Abers, G.A., 2010. The global range of subduc-tion zone thermal models. Physics of the Earth and Planetary Interiors 183,73–90.

arduno, J.A., Mayer, H., Winterer, E.L., Sliter, W.V., Kroenke, L., Mahoney, J.J., Leckie,M., Musgrave, R., Storey, M., 1991. Rapid formation of Ontong Java plateau byaptian mantle plume volcanism. Science 254, 399–403.

an Hunen, J., Moyen, J.-F., 2012. Archean Subduction: Fact or Fiction? AnnualReview of Earth and Planetary Sciences 40, 195–219.

arch 231 (2013) 206– 217 217

van Keken, P.E., Kiefer, B., Peacock, S.M., 2002. High-resolution models of subductionzones: implications for mineral dehydration reactions and the transport of waterinto the deep mantle. Geochemistry Geophysics Geosystems 3, 1056.

Van Kranendonk, M.J., 2010. Two types of Archean continental crust: plume andplate tectonics on early Earth. American Journal of Science 310, 1187–1209.

Van Kranendonk, M.J., Hugh Smithies, R., Hickman, A.H., Champion, D.C., 2007.Review: secular tectonic evolution of Archean continental crust: interplay

between horizontal and vertical processes in the formation of the Pilbara Craton,Australia. Terra Nova 19, 1–38.

van Thienen, P., van den Berg, A., Vlaar, N., 2004a. On the formation of continen-tal silicic melts in thermochemical mantle convection models: implications forearly Earth. Tectonophysics 394, 111–124.

van Thienen, P., van den Berg, A.P., Vlaar, N.J., 2004b. Production and recycling ofoceanic crust in the early Earth. Tectonophysics 386, 41–65.

Vielzeuf, D., Schmidt, M.W., 2001. Melting relations in hydrous systems revisited:application to metapelites, metagreywackes and metabasalts. Contributions toMineralogy and Petrology 141, 251–267.

Vlaar, N.J., van Keken, P.E., van den Berg, A.P., 1994. Cooling of the earth in theArchaean: consequences of pressure-release melting in a hotter mantle. Earthand Planetary Science Letters 121, 1–18.

Wenk, H.R., Joswig, W., Tagai, T., Korekawa, M., Smith, B.K., 1980. The averagestructure of An 62–66 labradorite. American Mineralogist 65, 81–95.

Wilde, S.A., Valley, J.W., Peck, W.H., Graham, C.M., 2001. Evidence from detritalzircons for the existence of continental crust and oceans on the Earth 4.4 Gyrago. Nature 409, 175–178.

Willbold, M., Hegner, E., Stracke, A., Rocholl, A., 2009. Continental geochemical sig-natures in dacites from Iceland and implications for models of early Archaeancrust formation. Earth and Planetary Science Letters 279, 44–52.

Wolf, M., Wyllie, P., 1993. Garnet growth during amphibolite anatexis: implicationsof a garnetiferous restite. The Journal of Geology 101, 357–373.

Wolf, M., Wyllie, P., 1994. Dehydration-melting of amphibolite at 10 kbar: theeffects of temperature and time. Contributions to Mineralogy and Petrology 115,369–383.

Wolf, M.B., Wyllie, P.J., 1991. Dehydration-melting of solid amphibolite at10 kbar—textural development, liquid interconnectivity and applications to thesegregation of magmas. Mineralogy and Petrology 44, 151–179.

Wyllie, P.J., Wolf, M.B., 1993. Amphibolite dehydration-melting: sorting out thesolidus. Geological Society, London, Special Publications 76, 405–416.

Xiong, X., 2006. Trace element evidence for growth of early continental crust bymelting of rutile-bearing hydrous eclogite. Geology 34, 945–948.

Xiong, X., Keppler, H., Audétat, A., Ni, H., Sun, W., Li, Y., 2011. Partitioning of Nband Ta between rutile and felsic melt and the fractionation of Nb/Ta duringpartial melting of hydrous metabasalt. Geochimica et Cosmochimica Acta 75,1673–1692.

Xiong, X.L., Adam, J., Green, T.H., 2005. Rutile stability and rutile/melt HFSE parti-tioning during partial melting of hydrous basalt: implications for TTG genesis.Chemical Geology 218, 339–359.

Zegers, T.E., van Keken, P.E., 2001. Middle archean continent formation by crustaldelamination. Geology 29, 1083–1086.

Zhang, C., Ma, C., Holtz, F., 2010. Origin of high-Mg adakitic magmatic enclaves fromthe Meichuan pluton, southern Dabie orogen (central China): implications fordelamination of the lower continental crust and melt-mantle interaction. Lithos119, 467–484.

Zhao, G., Wilde, S.A., Cawood, P.A., Lu, L., 1998. Thermal evolution of Archean base-ment rocks from the eastern part of the North China Craton and its bearing ontectonic setting. International Geology Review 40, 706–721.

Zhao, G., Wilde, S.A., Cawood, P.A., Lu, L., 1999. Thermal evolution of two textural

types of mafic granulites in the North China craton: evidence for both mantleplume and collisional tectonics. Geological Magazine 136, 223–240.

Zhao, G., Wilde, S.A., Cawood, P.A., Sun, M., 2001. Archean blocks and their bound-aries in the North China Craton: lithological, geochemical, structural and P–Tpath constraints and tectonic evolution. Precambrian Research 107, 45–73.