Geoscience Frontiers 5 (2014) 63e75

Contents lists available at SciVerse ScienceDirect

China University of Geosciences (Beijing)

Geoscience Frontiers

journal homepage: www.elsevier .com/locate/gsf

Research paper

The off-crust origin of granite batholiths

Antonio Castro*

Unidad Asociada de Petrología Experimental, CSIC-Universidad de Huelva. Campus El Carmen, 21071 Huelva, Spain

a r t i c l e i n f o

Article history:Received 23 January 2013Received in revised form20 June 2013Accepted 21 June 2013Available online 9 July 2013

Keywords:BatholithGranodioriteAndesiteRelaminationGranuliteLower crust

* Tel.: þ34 959219828.E-mail address: acdorado@gmail.com.

Peer-review under responsibility of China University

Production and hosting by Els

1674-9871/$ e see front matter � 2013, China Univerhttp://dx.doi.org/10.1016/j.gsf.2013.06.006

a b s t r a c t

Granitod batholiths of I-type features (mostly granodiorites and tonalites), and particularly those formingthe large plutonic associations of active continental margins and intracontinental collisional belts,represent the most outstanding magmatic episodes occurred in the continental crust. The origin ofmagmas, however, remains controversial. The application of principles from phase equilibria is crucial tounderstand the problem of granitoid magma generation. An adequate comparison between rock com-positions and experimental liquids has been addressed by using a projected compositional space in theplane F(Fe þ Mg)eAnorthiteeOrthoclase. Many calc-alkaline granitoid trends can be considered cotecticliquids. Assimilation of country rocks and other not-cotectic processes are identified in the projecteddiagram. The identification of cotectic patterns in batholith implies high temperatures of magmasegregation and fractionation (or partial melting) from an intermediate (andesitic) source. The com-parison of batholiths with lower crust granulites, in terms of major-element geochemistry, yields thatboth represent liquids and solid residues respectively from a common andesitic system. This iscompatible with magmas being formed by melting, and eventual reaction with the peridotite mantle, ofsubducted mélanges that are finally relaminated as magmas to the lower crust. Thus, the off-crustgeneration of granitoids batholiths constitutes a new paradigm in which important geological implica-tions can be satisfactorily explained. Geochemical features of Cordilleran-type batholiths are totallycompatible with this new conception.

� 2013, China University of Geosciences (Beijing) and Peking University. Production and hosting byElsevier B.V. All rights reserved.

Geology, as the science of Earth history, is prone to controversy.The study of history of any kind depends upon documents andrecords. For the history of the Earth’s crust, these documents arethe rocks and their reading and interpretation are often difficultoperations.

H.H. Read (1959) “The Granite Controversy”

The method of science is tried and true. It is not perfect; it’s justthe best we have. And to abandon it with its skeptical protocolsis the pathway to a dark age.

Carl Sagan (1997) “The Demon-HauntedWorld: Science as a Candlein the Dark”

of Geosciences (Beijing)

evier

sity of Geosciences (Beijing) and P

1. Introduction

Granites are among the most enigmatic rocks of the Earth’scontinental crust. They have been enigmatic for long time along thehistory of Geology and still they are in present days. Granite geol-ogy, similar to other problems related to the origin and evolution ofthe Earth, had its proper dark ages, where all kind of conjecturesprevailed. Today, with application of advanced methods of modernEarth Sciences, particularly those provided by Mineral Thermody-namics, Geochemistry, Isotope Geology and Geophysics, we have amore accurate view of the granite problem and its implication inthe origin of the continents (Taylor and McLennan, 1985; Windley,1995). Experimental Petrology was the first light into the darkhistory of the granite controversy. Laboratory experiments are ourparticular “candle in the dark” that opened the new sight on granitemagma generation and accounted for observed field relations andgeochemical trends.

Granitic rocks, in contrast with other rock types forming theEarth’s continental crust, have been subjects of several contro-versies along the recent history of Geology. A vigorous debate wasleaded in the middle half of the past century by Norman L. Bowen

eking University. Production and hosting by Elsevier B.V. All rights reserved.

A. Castro / Geoscience Frontiers 5 (2014) 63e7564

and H.H. Read (see Gilluly, 1948). The debate confronted experi-mental petrology and field geology, crucibles and plutons. Theacceptance of more than one granite generation mechanism, andhence more than one granite type (“granites and granites”; Read,1948, 1957), spread some kind of peace. The experimental deter-mination of fundamental phase relationships in the Ab-Or-Qtz-H2Osystem (Tuttle and Bowen, 1958) opened a new window into thegranite problem. However, this was a short-lived peace. Only veryfew granites can be produced at conditions of water saturation andvery few have the composition of the granite minimum. The exis-tence of different types of granites is an empirical fact. The S-Igranite types (Chappell and White, 1974, 2001) and the anorogenicA-type (Loiselle and Wones, 1979; Bonin, 2007) are broadlyrecognized around the world. Although the recognition of severalgranite types is not a solution to the problem, the granite typeclassification is an important step to approach a global solution. Animportant advantage of the classification scheme is to set thegranite problem at the scale of the whole continental crust as afunction of the relative abundance of each granite type. Interest-ingly, the most enigmatic granites in relation to origin are the mostabundant ones, those belonging to the I-type according to theChappell-White’s classification. This short review is focused onthese I-type granitoids that not only from the Cordilleran-typebatholiths, but also appear forming large post-collisional batho-liths in intracontinental orogenic domains. The anatectic S-typegranites, formed by partial melting of metasediments, and theanorogenic A-type granites are not included in this discussion.However, transitions between S- and I-types have been reportedrecently in large batholithic areas of Central Spain (Diaz-Alvaradoet al., 2011) and the Famatinian magmatic arc in Argentina(Grosse et al., 2011), as well as between A- and I-types in Mesozoicmetamorphic core complexes of large regions of NE Asia (Guo et al.,2012). Whilst S/I transitions are identified as the result of localassimilation of partially molten metasediments at the emplace-ment level of I-type batholiths, the generation of A/I transitionsremains obscure in the same degree of uncertainty than the originof A-type granites.

But, why these apparently simple rocks, mostly composed ofquartz and feldspars, are so enigmatic? First, granites of theCordilleran-type batholiths are not so simple as believed. Second, asolution to the problem of the origin can be given in the context ofthe new paradigm of arc magmatism, which is linked to a newconception of the thermal structure of the mantle in supra-subduction zones. I will show in this short review both facets of theproblem: on one hand, the relative complexity of Cordilleran-type(i.e., I-type) granitic rocks and, on the other hand, the new geneticmechanisms emerging from thermomechanical models, which arepointing to an off-crust generation of batholiths.

2. Models for granitoid (granodiorite-tonalite) magmageneration

In addressing the problem of granitoid (mostly granodiorite andtonalite) magma generation, we find two main handicaps: (1)Petrogenetic models based on experimental phase equilibria (e.g.,Naney, 1983; Patiño Douce, 1995; Castro et al., 2010) require ther-mal conditions (T > 1000 �C) that are not prevailing within thecontinental crust. (2) Hypotheses to get abnormal T gradients in thecrust by advective heating from the mantle by basalt underplating(e.g., Annen et al., 2006) do not receive support from geologicaldata related to lower crust composition (Castro et al., 2013b).Consequently, the origin of granite batholiths remains enigmatic,full of controversial and subject to speculative models. An in-depthdiscussion of the varied models is out of scope of this review, whichis focused on geological data supporting an off-crust generation of

granite batholiths. However, a short discussion on the most clas-sical “on-crust” models may help to better understanding of theproposed off-crust origin.

Anymodel for the generation of granite batholiths must accountfor essential natural observations. Several models have been pro-posed to account for the generation of granodiorite-tonalitemagmas (Fig. 1). However, no one fully satisfies natural observa-tions of batholiths. By contrast, they entail unrealistic and para-doxical implications, which will be discussed below. I will showhere that all paradoxes about granite magma generation are solved,or dissolved, if granite sources are initially rootedwithin themantleand not within the continental crust. Two large model categoriesare distinguished depending on the locus of magma generation.These are (1) on-crust models (Fig. 1aec) and (2) off-crust models(Fig. 1def). On-crust models postulate batholith magma generationfrom lower crust rocks. By contrast, off-crust models propose thegeneration of parental andesite magmas by processes within themantle by melting and/or reaction of subducted materials. Amongthe most relevant on-crust models (Fig. 1) we may mention (1)basaltic underplating and crustal delamination, (2) melting of thelower crust by intrusion of basalts, (3) crustal assimilation by basaltmagmas and (4) magma mixing.

2.1. Basaltic underplating. A two-stage process

Basically, this model proposes that granite batholiths aregenerated in two stages from the mantle. In a first stage, basalts aregenerated by melting of the peridotite mantle and emplaced byunderplating at the lower continental crust. In a second stage, ba-salts solidified and are partially molten to produce silicic melts thatmay form batholiths. Large tonalite intrusions from the CordilleraBlanca batholith are explained by this genetic mechanism (Petfordand Atherton, 1996). The model was refined (Hawkesworth andKemp, 2006) and applied as a general mechanism to generate thecontinental crust. The implication of granite magma generationfrom partial melting of underplated basalts at the lower crust is theformation of large volumes of ultramafic residues (about 70 wt.% ofthe intruding basalt), which are missing in the lower continentalcrust. Two possible solutions can be given to this paradoxical hy-pothesis. First, the ultramafic residues are missing because theyhave been sunk into the mantle by process of crustal delamination.Second, granites are not derived from melting, or incompletecrystallization, of underplated basalts at the lower crust. Delami-nation is an old concept (Bird, 1979) applied by Kay and Kay (1993)to account for a particular magmatism, supposedly unique, of thePuna region of Northern Argentina. A thickened lower crust is anecessary condition to increase the density of a lithosphere thatotherwise is buoyant (Kay and Kay, 1993). The delamination hy-pothesis is supported by seismic evidence in Sierra Nevada, Cali-fornia (Gilbert et al., 2012). Wemay leave the debate on lithospheredelamination aside and ask a central question: are basalts able tofractionate to granite melts at the lower crust? The questionreceived attention in several studies on mechanisms of new crustgeneration (e.g., Hawkesworth and Kemp, 2006; Lee et al., 2006,2007; Castro et al., 2013b). Phase relations are crucial on thispoint. Olivine is not stable at the pressures of the lower crust and itis necessary to increase the silica of the residual melt without largedepletion in Fe and Mg. The stable phase is pyroxene (Px) in a drybasalt system, and pyroxene crystallization slightly modifies thesilica content of the residual liquid with respect to the initialbasaltic composition (see crystallization modeling with MELTScode in Castro et al. (2013b)). It has been proposed, as a variant ofthe model, that water-bearing basalt is more favorable to frac-tionate to silicic (granitic) magmas (Thompson et al., 2002). How-ever, the composition of residual liquids from a wet-basalt is not

Sediments

Subd.erosion

Silicic diapir (partially moltenmélange)

Continental crust

Sediments

Reaction channels

Continental crust

Sediments

Wet picritic basalt

Continental crust

Metasomatized mantle

SedimentsContinental crust

Metasomatized mantle

Lower crustmelting

Delamination ofmafic residues

Fluids

SedimentsContinental crust

Metasomatized mantle

Melting of thelower arc crust

Delamination oflower arc crust

Fluids Hot mantle upwelling

Continental crust

Delamination oflower crust and mantle lithosphere

Hot mantle upwelling

Melting of thelower crust

Continental crust

On-crust models Off-crust models

Basalt magmas (blue) invade the lower crust and provoke melting to give rise to batholith magmas (red) and solid residues.Delamination of ultramafic residues (dark grey) is needed.Large basalt:crust ratios of 2:1 are required.Lower crust rocks are not appropriate to produce granite batholiths.

Lower crust melting assisted by basalt magma intrusion

Lower crust melting assisted by delamination and mantle upwelling

Fractionation from high-Mg andesites (HMA) formed in reaction channels

Fractionation of andesite formed by previous fractionation of wet picritic basalts

Melting of subducted mélanges and magma relamination to the lower crust

Delamination of lower crust and lithosphere allows lower crust heating by upwelling of hot asthenospheric mantle. Melting of the lower crust produce batholiths (red).Thickenning results by additions of magma to the lower crust in arcs or by intracontinental collision.Basalt magmatism, expected to occur as consequence of mantle upwelling, is missing in intracontinen-tal orogens and mature (thick) arcs or active margins.Lower crust rocks are not appropria-te to produce granite batholiths.Possibly working in intraoceanic arcs if lower crust is already andesitic. The andesite source is the result of off-crust processes.

Melts from the slab react with the peroidotite mantle and produce high-Mg andesites (blue).These fractionate within the crust, or in thew way upwards through rhe mantle, to produce silica-rich melts (batholiths) (red).The model accounts for isotopic features and thermomechanical models.HMA and equivalent plutonic (sanukitoids and appinites) are common in relation to batholiths.

Water-rich picritic basalts (green) fractionate within the mantle to andesite melts (blue). These are underplated at the lower crust and fractionate again to silicic magmas of batholiths (red).An extra process of crustal assimilation is needed to account for geochemical and isotopic features of batholiths.High water contents not compatible with undersaturarated character of bahtoliths

Silicic andesitic magmas (blue) are directly generated from subducted materials within the mantle wedge in the form of silicic diapirs (purple). These are fractionated within the crust to form silicic batholiths (red) and residual granulites of the lower crust.Hybrid geochemical features are acquired from the source. Supported by geochemistry, phase equilibria and numerical models.

Fluids

a d

b

c

e

f

Figure 1. Schematic representation of petrogenetic-tectonic models (not at scale) proposed for the generation of batholiths of calc-alkaline, Cordilleran-type affinities (I-type granites). Two large categories are distinguished dependingon the locus of magma generation, on-crust (aec) or off-crust (def). On-crust models postulate batholith magma generation by melting of lower crust rocks. These models need of an extra heat source that can be supplied by mantle-derived basalts (a). The required basalt: crust ratio of 2:1 is based on thermal modeling (Annen et al., 2006). Mantle upwelling due to lower crust delamination (b) has also been proposed in active margins (Kay and Kay, 1993) and inintracontinental orogens (e.g., North China; Zhai et al., 2007). The absence of basaltic magmatism and the low fertility of the refractory lower crust, are the two main handicaps of these on-crust models. Off-crust models propose thegeneration of parental andesite magmas by processes within the mantle by melting and/or reaction of subducted materials. Generation of high-Mg andesites (d) may occur in reaction channels within the mantle (Kelemen, 1995;Kelemen et al., 2003), involving melts and fluids from the subduction zone. The process is supported on recent data from numerical modeling (Vogt et al., 2013) and laboratory experiments (Castro et al., 2013a). Generation ofandesites from wet basalt within the mantle may also produce the required silicic compositions that fractionate within the crust to generate batholiths (Alonso Perez et al., 2009). Finally, the generation of batholiths from melting ofsubducted mélanges (f) has been proposed as the most plausible explanation (see text for further explanations).

A.Castro

/Geoscience

Frontiers5(2014)

63e75

65

A. Castro / Geoscience Frontiers 5 (2014) 63e7566

matching fundamental geochemical features of granite batholiths.Assimilation of crustal rocks is needed in these models to accountfor crustal features, such as high K and incompatible elements thatcharacterize granite batholiths. Furthermore, the water content ofthe silica-rich granite liquid produced by fractionation of a wetbasalt with a minimum initial water content of about 3 wt.% H2O,will be at saturation in the residual melt, which is completely un-realistic and far from the low water content that characterizeCordilleran-type granite batholiths. The application of this model tothe water-rich granitoids of the Adamello massif (Thompson et al.,2002) seems to be a plausible explanation, but its application to thecommon water-undersaturated Cordilleran granites is unlike.

2.2. Lower crust melting

Because melting of the peridotite mantle cannot producegranite magma, partial melting of the lower crust is the most im-mediate explanation for the generation of silicic magmas that areemplaced in the middle and upper crust. It is proposed in thesemodels that granite magmatism is produced at the lower crust byheating from deep asthenospheric mantle that replaces a suppos-edly delaminated lithosphere. Thus, the need for delamination isalso present in these models. Alternatively, heating of the lowercrust can be achieved by advective heat transport of mantle-derived basalt to the lower crust (Annen et al., 2006). Accordingto these thermal models, a basalt-to-crust ratio of 2:1 is needed toreach the high temperatures required to get granodiorite andtonalite melts. Furthermore, it is assumed that lower crust is sup-posedly fertile to produce the required granite magma composi-tions. Nonetheless, the expected high proportions of basaltic rocksare missing in the continental crust, and they never occur in largecrustal regions that were pervasively invaded by granite batholiths(e.g. the active margins of the Americas; Pankhurst et al., 1999;Hervé et al., 2007; Lee et al., 2007; Crawford et al., 2009; Castroet al., 2011b) and the Mesozoic large igneous province of NE Asia(Guo et al., 2012). Are they also delaminated and sank into themantle? I have documented a case ofmelting of lower crust assistedby water- and K-rich fluids released fromwet basalts intruding intotonalitic TTG gneisses from the Northern Highlands is Scotland(Castro, 2004). The process, however, needs of an already graniticsource (TTG complexes) and cannot be applied to the Px-rich, maficgranulites that dominate the relatively more mafic (andesitic) post-Achaean lower continental crust.

2.3. Assimilation and magma mixing

Assimilation of upper crustal rocks by intruding granodiorite-tonalite batholiths is well documented in many granitic areas(Erdmann et al., 2007; Saito et al., 2007; Diaz-Alvarado et al., 2011).The process may be accomplished by means of reactive bulk-assimilation (Beard et al., 2005) without the implication of largeenergetic budget from the intruding magma. However, geologicalevidence is not in favor of assimilation as a mechanism to producethe I-type granodiorite and tonalite rocks. Assimilation of crustalrocks by basalt magmas has been tested experimentally as amechanism able to produce granodiorite melts (Patiño-Douce,1995; Castro et al., 1999). However, temperatures as high as1000 �C are still insufficient to produce tonalite and granodioritemelts matching the composition of batholiths.

Magma mixing and assimilation are normally identified as localphenomena that cannot be applied to account for the generation ofgranodiorite-tonalite rocks at the scale of batholiths (seediscussionsin Castro et al., 2010; Clemens and Stevens, 2012). Both processeshave in common that they are not selective; they affect the wholecompositions of the involved end-members. The implication is that

compositional variations in major elements must be accompaniedbysimilar variations in traceelements and isotopes. For instance, in ageneral diorite-tonalite-granodiorite-granite trend, the terms richerin K (granite) must be the richer in radiogenic 87Sr and poorer inradiogenic 143Nd, in agreementwithpredictions of the Sr-Nd isotopesystematics. By contrast, the general observation is the existence ofsignificant decoupling between major element contents and radio-genic isotopes. An example is the trend displayed in Sr-Nd initialisotope ratios by granitoids of the Patagonian batholith in SouthAmerica (Pankhurst et al., 1999; Hervé et al., 2007) where graniteswith identical composition in terms ofmajor elements plot either inthe field of mantle-like sources or in the field of evolved crust.Geochemical decoupling is a characteristic feature of Cordillerangranite batholiths. The implication is that rocks represent cotecticliquids from heterogeneous sources, being the composition of liq-uids buffered by the co-existing solid assemblage over awide rangeof source compositions (Castro et al., 2010). These observations areessential to constrain potential processes of granite magma gener-ation with two additional implications. First, assimilation andmagma mixing can be ruled out as general mechanisms able toproduce granodiorite-tonalite magmas. Second, they imply thatgeochemical variations may be related to cotectic evolution of liq-uids in equilibriumwith a changing solid assemblage. Evidences forsuch a behavior of I-type granite systems are given below.

3. Granodiorite-tonalite systems: magmas or liquids?

Identification of granodiorite-tonalite magmas as either liquidsformed by cotectic magmatic systems or magma mushes carryingsuspended mafic minerals, is an essential step previous to any dis-cussion about processes of magma generation. Several experimentalstudies have been addressed to seek for conditions of granodiorite-tonalite liquid generation. Composite systems, containing crustaland mantle sources together are preferred in most experimentalstudies in order to satisfy the mantle/crust hybrid geochemical sig-natures, in terms of isotopic ratios, of I-type granitoids (Johnston andWyllie, 1988, 1989; Patiño-Douce, 1995; Castro et al., 1999). In gen-eral, the reportedexperimentalmelts in these studies arepoorer inFeand Mg compared to natural rocks. It was proposed partial entrain-ment of peritectic pyroxene (Castro et al., 1999, pag. 274), in a waysimilar to that proposed lately by Clemens and Stevens (2012), toaccount for the high Fe and Mg content of most granodiorite andtonalite rocks compared to experimental melts. Restite entrainmentis well supported in S-type granitic rocks in which, maficity(expressed as the sum of molar FeþMgþ Ti; Clemens et al., 2011) ispositively correlated with Al saturation index, denoting theentrainment of peritectic minerals such as Grt and/or Cord, whichsupply jointly Fe þ Mg and Al to the system. The application of per-itectic restite entrainment to I-type granodiorites and tonalites is notstraightforward. Simple mass balance yields that about 20 wt.% oforthopyroxene must be added to experimental granodiorite meltsformed at 1000 �C to increase theirmaficity to the normal values of I-type granodiorites (Castro et al., 1999). Nevertheless, the generationof these leucogranodiorite melts in equilibrium with pyroxene andplagioclase, either by breakdown of amphibole or by partial crystal-lization of an andesite system, requires temperatures of about950e1000 �C. That is, granodioritemeltsmay increasehypotheticallytheirmaficity by incorporating restitic pyroxene from the source, buttemperature as high as 1000 �C is required to get pyroxene in equi-librium with melt at the source. According to experiments, thisrequired high temperature increases with increasing pressure.

Addition of water may reduce the liquidus temperature at thesame time that Fe andMg increase their solubility by increasing thewater content of the melts. However, a characteristic feature of I-type granitic magmas is strong water undersaturation, as it is

A. Castro / Geoscience Frontiers 5 (2014) 63e75 67

demonstrated by the crystallization sequence with plagioclase,instead of amphibole, as the liquidus phase. Although some gain ofmaficity can be accomplished by peritectic restite entrainment, theobserved chemical variation trends of batholiths strongly supportthe cotectic co-variation of Ca, Mg and Fe in tonalite-granodioritesystems. Identification of cotectic trends implies that composi-tional variations are controlled by the co-existing mineral assem-blage, and this is in turn controlled by changes in intensivevariables (particularly in T) as a result, for instance, of cooling andcrystallization at the place of emplacement.

3.1. Granodiorites and tonalites as cotectic liquids

As granodiorite and tonalite contain a significant proportion ofthe minimum granite composition, the temperature required formagma generation may range from 650 �C at water saturatedconditions to about 850 �C for water-undersaturated systems (atmid-crustal pressures of 600 MPa) inasmuch as the non-minimumcomponents (NMC), namely Ca, Ti, Fe andMg, are only contained incalcic and mafic minerals and transported as suspensions by themagmas from the source region. However, if these NMC compo-nents are dissolved in the melt, the corresponding temperaturescan be considerably higher, up to >1000 �C for water undersatu-rated conditions (Maaloe and Wyllie, 1975). The implication is thatsuch a high temperature melts are unlikely produced by partialmelting of crustal rocks, but they can be produced from andesiticmagma systems that, after undergoing potential magmatic differ-entiationwithin the lithospheric mantle, can be finally fractionatedinto liquids and solid residues within the continental crust.

It is apparent from textures and phase equilibrium relationships ofwater-undersaturated magmas (Maaloe and Wyllie, 1975; Naney,1983) that hydrous mafic minerals, namely biotite and amphibole,are latephases thatprecipitate fromresidualmelts at the late stages ofmagma evolution. The point at which these hydrous ferromagnesianminerals start to precipitate in themagma depends on compositionalfactors, principally the initial water content, pressure and tempera-ture. Experiments confirm that hydrous ferromagnesian mineralsformasperitectic phases associated to the breakdownof pyroxenes inthecourseof cooling. The implication is that ferromagnesianminerals,either anhydrous or hydrous, are part of the coexisting assemblageover awide crystallization interval from liquidus to solidus in diorite-tonalite-granodiorite systems. Consequently, the composition ofmelts is controlled by phase equilibria. Consideration of granodiorite-tonalite rocks as either liquids or mechanical mixtures of crystals andlow-temperature liquids requires a rigorous comparison of wholerocks and experimental glasses, in terms of major elements, onappropriate diagrams. It is proposed here a projection for calc-akalineandesite-dacite-rhyolite (diorite-granodiorite-granite) systems ontothe plane F(Fe þ Mg)-An-Or of a complex system defined by thecomponents Si, Al, Ti, Fe, Mg, Mn, Ca, Na and K (Fig. 2a). Details of thisprojection are given in Castro (2013). The most important composi-tionalvariations fromtonalites to granites (sensu stricto) are expandedonto this plane projection by two relevant ratios: (1) the K/K þ Camolar ratio (K#) ranging from almost 1.0 in leucogranites to 0.1 intonalites and Qtz-diorites, and (2) the maficity (Fe þ Mg content) inrelation to feldspar components Ca and K (Na is a projection point).The lowest values ofK# correspond tomesocratic andmafic rocks thatdisplay the highest values of MgO and FeO. Consequently, thesecomponents recast to a molecular base in the form of F (FeOþMgO),Orthoclase and Anorthite, are the ones that better describe composi-tional variations in the calc-alkaline magmatic systems and they areselected for the molar projection shown in Fig. 2a. It is interesting tonotice the experimental finding of a buffered melt composition fromheterogeneous sources (Castro et al., 2010) in a so complex multi-component system, validating the assumption that lines defined by

liquid compositions can be treated as cotectic. Although differences inthebulk systemcompositionare affecting thepositionof cotectic linesin this projection, the attitude of the multisaturation lines (i.e. thecurvature and slopewith respect to the basis F-Or of the diagram) andthe position of the minimum in the system are rather related tointensive variables and water content than to differences in the bulkcomposition. Combining these criteria, cotectic lines, traced by inter-polation of the selected experimental data, can be grouped in threemain categories (Fig. 2b): (1) water-rich and saturated systems, (2)high-pressure and undersaturated systems, and (3) low-pressure andundersaturated systems.

The first group is characterized by short-range variations andevolution with decreasing temperature towards the An apex. Manytrondhjemites associated to water-saturated melting of maficsources plot in this region. The field overlaps partially with the areaof adakites. The second group is characterized by multisaturationlines showing depletion in ferromagnesian components (F apex).These cotectic lines are almost parallel to the An-Or join and arerelated to the presence of Grt in the coexisting assemblage. Thesesystems have in common the high pressure (>1.5 GPa) and the lowwater content (ca. 1 wt.% H2O). The third group is characterized by agentler slope, compared with the former one, and by a curvilinearshape concave to the F apex. The common features of these systemsare water undersaturation (1e2.5 wt.% H2O) and low to moderatepressure (1.0e0.3 GPa). These are also characterized by the com-mon presence of Hbl and Pl (�Bt, �Px) along a wide temperatureinterval along the cotectic lines. The diagram is robust to identifycotectic and non-cotectic relations. For instance, assimilation ofcountrymetasediments is easily identifiable by vectors (labeled a inFig. 2b) departing from the cotectic array. Another non-cotecticpattern is formed by restite unmixing (labeled ru in Fig. 2b)commonly displayed by S-type granites in which incomplete sep-aration of restite and melt produce particular patterns that areeasily identified as non-cotectic in the diagram.

Projection of granitoid samples from different calc-alkalinebatholiths strongly suggests that they represent essentially liq-uids fractionated from a common magmatic source of broadlyandesitic (Qtz-dioritic) composition.

4. Records of high temperature and high pressure ingranitoids

Slow cooling is an intrinsic feature of plutonic rocks. Theconsequence is resetting of mineral-mineral and mineral-meltequilibria to near-solidus magmatic conditions of the earlymagmatic stages at the time of segregation from the source orfractionation in a deep-seated magma chamber. However, somerelations survive partially as relics from the early magmatic stages,providing with important information on magmatic temperaturesand initial water contents of the magma at the time of segregation.Depth is preserved as relations of particular trace elements that arepartitioned into mineral phases whose stability field is mostlydependent on pressure. These are the Sr/Y and Ce/Yb ratios, whichare strongly controlled by the presence of garnet in the source re-gion. A detailed review on the meaning of the adakitic signature,emphasizing on cautions in handling these geochemical parame-ters, is given by Moyen (2009). Near-liquidus temperatures may berevealed by the presence of relic minerals that survived peritectictransformations in the course of slow cooling. Crystallizationsequence is revealing initial water content in the magma systems.

4.1. Crystallization sequence

Amphibole (Amp) is a common mafic mineral in I-type granodi-orites and tonalites.WhetherAmp is early or late in the crystallization

Pelites

Pelites

Greywackes

An

OrF

Cpx

Hbl

Grt

Opx

Bt

Greywackes

Primary magmas(Andesitic)

assimilation

restiteentrainment

Adakites

TonalitesGranodiorites

Monzogranites and granites s.s.

1

1 2 3 4

5 6 7

8

p1: Opx+Liq Hbl p2: Hbl+Liq Bt

0.3 GPa2.5 %wt H O2

Opx

Pl

F

1000

950

An

Or

Bte

900

p2p1

Hbl

Hbl

Cpx

Bt

Pl

Opx

TonalitesGranodiorites

Monzogranites and granites s.s.

+Qtz +Als +Il +Ab +FeMg-1 +FeMn-1

+Qtz +Als +Il +Ab +FeMg-1 +FeMn-1

+Qtz +Als +Il +Ab +FeMg-1 +FeMn-1

F

An

Or

Hbl

Cpx

Bt

Pl

Opx

2040

Pelites

SiO : 63-70 % 2

SiO : <63 % 2

SiO : >70 % 2 Assimilation

TonalitesGranodiorites

Monzogranites and granites s.s.

+Qtz +Als +Il +Ab +FeMg-1 +FeMn-1

F

An

Or

Hbl

Cpx

Bt

Pl

Opx

Pl

TonalitesG

ranodioritesM

onzogranites and granites s.s.

a b

c d

7 87

Mafic enclavesTonalite-Monzogranite seriesGranitesContaminated granites

Velasco batholith,Famatinian arc(Argentina)

Gredos batholith,Iberian massif(Spain)

Figure 2. Projection onto the plane F (FeO þ MgO)-Anorthite-Orthoclase (F-Or-An). (a) Projection of cotectic lines interpolated from experiments with andesite systems. Projectionpoints are labeled at the upper left corner. Data sources are following numbers beside each curve refer to data source as follows: 1. Picrite-fractionated liquids (Alonso Perez et al.,2009); 2. Qtz-eclogite (Skjierlie and Patino Douce, 2002); 3. Low-silica tonalite (Carroll and Wyllie, 1990); 4. Basalt-sediment mélange (Castro et al., 2010); 5 and 7, Mélange (Castroet al., 2013a); 6 and 8, Andesite and shshonitic andesites (Sisson et al., 2005). The water contents of these experimental works ranges from the lowest values (ca. 1 wt.% H2O) ofnatural mélanges (0.87 wt.% H2O for a sediment fraction Xs ¼ 0.5), to the highest values of 4 and 8 wt.% H2O in the runs with a picrite-fractionated composition (Alonso Perez et al.,2009). The runs with natural basaltic andesite and shoshonitic basalt (Sisson et al., 2005) contain 1.68 and 2.33 wt.% H2O respectively. Red arrows indicate potential not-cotecticvectors dominated by assimilation of pelitic rocks and restite entrainment. (b) Tentative position of peritectic points and cotectic lines according to assemblages in experiments at0.3 GPa and 2.5 wt.% water with a basalt-sediment mélange (Castro et al., 2013a). Composition of the starting material is represented by large blue star in the field of andesites. (c)Position of cotectic granodiorites from the Velasco batholith (Bellos et al., 2013). Large arrow indicates assimilation in samples departing from the low P cotectic, number 7 in (a).Enclaves may represent the parental magmas of this batholith. Heterogenous granites with S/I transitional features (Grosse et al., 2011) of the same batholith are the result of uppercrust contamination with pelitic host rocks. They may contain up to 40 wt.% of contaminant (red dashed curves and numbers beside in red). (d) Projection onto the same diagram ofcotectic and not-cotectic (assimilation processes) of granodiorite and monzogranite rocks of the Gredos batholith in Central Spain (Diaz-Alvarado et al., 2011). In this case, cotecticnumber 8 is the closest to the rock trend for the high silica rocks. Mineral abbreviations after Kretz (1983).

A. Castro / Geoscience Frontiers 5 (2014) 63e7568

sequence is important for two reasons. First,magmatic crystallizationof Amp, and of any other mafic mineral, implies that the maficcomponents Fe and Mg were dissolved in the silicate melt and werenot transported in mafic restitic minerals entrained from the source.Textural relations are clearly indicating that Amp crystallized fromthe magma. However, crystallization is rarely early in the magmaticsequence, indicating that the initial water content of the magmawaslower than the minimum water required to stabilize Amp. In the

course of crystallization of plagioclase and pyroxene, the remainingliquid is becoming richer inwater leading to amphibole saturation atany intermediate point of the crystallization sequence between liq-uidus and solidus. A rough estimation of the initial water content canbe made by analyzing the crystallization sequence, which can becompared with predictions of experimental phase relations inT � XH2O sections (e.g., Maaloe and Wyllie, 1975; Naney, 1983). Ac-cording to experiments in granodiorite systems, the required water

A. Castro / Geoscience Frontiers 5 (2014) 63e75 69

content to stabilize Amp is of 4wt.% H2O at P¼ 200MPa and 2.5wt.%H2O at P¼ 800MPa (Naney,1983). Amphibole crystals show typicallyanhedral crystalline habits in Cordilleran granodiorites and tonalites.Amphibole is late with respect to plagioclase in the crystallizationsequence and itmay show euhedral habit only in contact with quartzand/or alkali feldspar. Even in the particular cases, in which Amp isapparently euhedral, it shows molding relations against plagioclaseand euhedral habit against quartz (Fig. 3), suggesting that Amp wasnot the liquidus phase of the system and, thus, that the initial watercontent of themagmawas lower than that required to stabilize Amp.It is difficult to assess the liquid fraction of the magma at the time of

Figure 3. Contrasting textural features of two tonalite rocks from the Cordilleran North Patagthe interstitial texture of Hbl in (b) compared with the euhedral habit of Hbl in (c). Note that Hthe crystallization sequence. Euhedral faces are preferred at the Hbl-Qtz contacts contrastingmagma was partially crystallized at the time of Hbl growing and, consequently, subsaturatesaturation is reached at late stages indicating that the initial water content of the magma wa

Amp crystallization. In the reasonable case of 50 wt.% ofplagioclaseþ pyroxene crystallization, the initialwater content is halfof the content required for Amp crystallization. This may range from1.25 wt.% H2O if crystallization proceeds at low pressure(P¼ 200MPa) to 2wt.%H2O for deeper crystallization (P¼ 800Mpa),according to the above mentioned experimental requirements(Naney,1983). These estimations imply high liquidus temperatures ofmore than 1000 �C for Amp-bearing granodiorite and tonalite liquids.Low temperature granodioritic liquids must have higher water con-tent of around 11 wt.% H2O at T ¼ 900 �C and P ¼ 800 MPa (Naney,1983; his Fig. 4), a T value considered as reasonable in anomalously

onian batholith (Bariloche, Argentina) and the Adamello batholith in Northern Italy. Notebl of Adamello tonalite contains abundant Pl inclusions denoting that Pl was previous inwith the more irregular shape of the Hbl-Pl contact. These observations support that thed in water, previous to the formation of Qtz. In the case of the Bariloche tonalite, Hbls very low. Scale bar of micrographs is 1 mm. Mineral abbreviations after Kretz (1983).

300 500 700 900 1100

60

50

40

30

20

10

0.5

1.0

1.5

And Sil

KyUHTW

etpe

lite

solidus

20 ºC/km

Typical Barrovian P-T range

Ultra-high temperature metamorphism

T (ºC)

P (G

Pa)

Z (k

m)

Typical Barrovian P-T-t path

Dehydration melting of amphibolite

2

50 ºC/km

Paso de Indios, Argentina (9)

Kohistan gabbros

Pali Aike, Chile (1)Arabian plate (G-I) (3)Arabian plate (G-II) (3)Nushan, East China (4)Tuoyun, NW China (10)Califiornia (high Mg) (11)

Kohistan granites

E Peninsular Ranges batholithPatagonian batholith

Califiornia (low Mg) (11)

Legend

Low

er c

rust

xeno

liths

BC

0

5

10

15

20

25

40 45 50 55 60 65 70 75 80

SiO2 (wt%)

PB g

ap

MgO (wt%)

02468

101214

46 50 54 58 62 66 70 74

Patagonian batholith

SiO2 (wt%)

Num

ber o

f sam

ples

6

7

2

43

5

8

a

b

91

Melting experiments, 1.5 GPa

Residues (with T in °C)Liquids

(non-reactive mélange)Starting material (mélange)

1000

1050

1100

100010501100

Figure 4. (a) Pressure-temperature diagram showing relevant examples of lower crust granulites (xenoliths and lower crust sections). Temperatures of most granulite xenolithsexceed the buffering effect imposed by the amphibole breakdown reaction of amphibolites, expected to occur in case of a hypothetical amphibolitic lower crust. Data sources forgranulites are labeled with numbers on the PT field of each xenoliths locality of lower crust region in (a). The same number refers to geochemical data analyses in (b). These are:1. Pali Aike, Chile (Selverstone and Stern, 1983); 2. Mecaderes, Colombia (Weber et al., 2002); 3. Arabian plate (Al-Mishwat and Nasir, 2004); 4. Nushan, East China (Huang et al.,2004); 5. Junan, North China (Ying et al., 2010); 6. Patan-Dasu complex, Kohistan; 7. Jijal granulites, Kohistan (Yoshino and Okudaira, 2004); 8. Valle Fértil metagabbros, Argentina

A. Castro / Geoscience Frontiers 5 (2014) 63e7570

A. Castro / Geoscience Frontiers 5 (2014) 63e75 71

heated lower continental crust domains. This water-rich magmawillstabilize Amp at its liquidus, earlier than plagioclase in the sequence.However, so high water contents are unrealistic according to texturalrelations (Fig. 3), which are characteristic of strongly water-undersaturated magma systems. The scarcity of pegmatites in I-type granites is in agreementwith lowwater content in themagmas.

It is very common the presence of polycrystalline aggregates(clots) of Amp in I-type granodiorites and related enclaves. Thedetailed study of textural relations of these Amp clots is sup-porting provenance from pyroxene (Px) crystals that are trans-formed to Amp during magma crystallization (Castro andStephens, 1992). The presence of these clots is telling us that Pxwas in equilibrium with the melt at least at the time of magmasegregation. Whatever the origin of these clots as early pheno-crysts or peritectic restites from the source (Stephens, 2001), theyimply that temperature was high enough to stabilize Px in awater-bearing magma system. According to experiments (Patiño-Douce, 1995; Castro et al., 1999), this temperature may be around1000 �C in andesite magmas.

4.2. Metagabbro inclusions

Other indication for high temperature is given by metagabbroinclusions within tonalites and Qtz-diorties of the Valle Fértilbatholith in Argentina (Castro et al., 2012). The high temperatureassemblages of these inclusions is not reset to low temperatureduring cooling because they are agmatites with an effective sepa-ration of liquid (trondhjemitic) and solid (Opx) residues. Ortho-pyroxene is present in association to hornblende breakdown.Orthopyroxene is partially transformed to pargasitic hornblendeformed in equilibrium with plagioclase during cooling. Thermom-etry results in metagabbros are obtained from those HblePl pairs,yielding values of T ¼ 1000 � 40 �C. New data of UePb SHRIMPzircon age determinations, together with thermobarometric andgeochemical data (Castro et al., 2012), yield that batholith magmaintrusion is the responsible for heating and self-granulitization ofearly gabbro pulses. Metagabbros are not common inclusions inbatholiths. They are more common in deeply-emplaced plutons,where they provide a valuable information about magma earlytemperatures.

5. The granite-granulite coupled association

In an independent way, it is possible to address the origin ofgranite batholiths by addressing experimentally the reverse prob-lem. That is, to determine the nature and composition of the sourceby identification of near-liquidus saturation phases. Reverse prob-lem studies are very scarce in tonalite-granodiorite systems (Naney,1983). In partially crystallized andesite systems, a granodioriteliquid is in equilibrium with assemblages dominated by orthopyr-oxene and plagioclase at temperatures of about 1000e1100 �C forinitial water contents in the magma of about 1.0 wt.% H2O (Patiño-Douce, 1995; Castro et al., 1999, 2010). Interestingly, many granulitexenoliths from the lower crust record peak temperatures close to,or even higher than 1000 �C (Fig. 4a) and fairly reproduce the Px-Pl

(Castro et al., 2012); 9. Paso de Indios xenoliths, Argentina (Castro et al., 2011a). Other fields o1998), Ultrahigh pressure metamorphic area (Kelsey, 2008) and Dehydration melting of amcrust granulites and cordilleran batholiths, and comparison with experimental residues and mof the bulk crust (BC; Rudnick and Gao, 2003) and the arc section of Kohistan (Garrido etxenoliths taken from references (1, 3, 4 and 9) of the upper panel (a), labeled with the same nWest China (Zheng et al., 2006); 11. California (Lee et al., 2007). Data from Patagonian baPeninsular Ranges are from Lee et al. (2007). Average bulk crust (labeled BC) is andesitic and(labeled PB gap) at about 54e58 wt.% SiO2 (histogram in b).

(�Grt, �Spl) solid assemblages predicted by experiments withandesite systems mentioned above.

Textures indicating decompression at 1.0 GPa and 1000 �C(Opx / Olivine reaction textures; see Castro et al. (2011a), are infavor of a system that was magmatic in origin and was fraction-ated and recrystallized during cooling and decompression as asolid rock to produce the observed metamorphic-like textures. Thepossibility that these granulite rocks are the source of granitemagmas is very unlike. First, they are characterized by refractoryassemblages and, in case they are partially molten, the needed Tmay exceed 1100 �C for very low melt fractions. Second, thesemelts will not be granodioritic, but andesitic. That lower crustgranulites were in equilibrium with a granodiorite melt is stronglysupported by experimental phase equilibria studies. The possiblelink between batholiths and lower crust granulites have beenexplored recently (Castro et al., 2013a) with implications onmechanisms of new crust generation in arcs without involvementof crustal delamination. The most outstanding geochemical re-lations between granulites and batholiths are summarized in thissection.

It is a common observation that granite batholiths define lineararrays in geochemical variation diagrams, which resemble cotectic-like relations (Fig. 2). By contrast, diorite and Qtz-diorite rockcompositions (SiO2< 60wt.%) and lower crust granulites (xenolithsand the Kohistan gabbros) are scattered and do not defined lineararrays (Fig. 4b). There is a compositional gap at SiO2 ¼ 54e58 wt.%separating linear and scattered regions in silica variation diagramsof batholiths (Fig. 4b). Granite samples plotting along a commoncotectic trend are not always coeval, with differences in age ofabout 20 Ma (170e150 Ma) in the case of the North Patagoniabatholith (Castro et al., 2011b). These non-coeval samples cannot befractionated from a common magma at the place of emplacement.More likely, they represent cotectic magma pulses extracted atdifferent temperatures from a common partially crystallizedmagma at depth. Looking at the MgO-SiO2 diagram of Fig. 4b, it isapparent that parental magma composition to the Patagonianbatholith is at the silica gap with values of 56e60 wt.% SiO2. Thesame reasoning can be applied to granites and lower crust gabbrosof the Kohistan (Pakistan) arc section (Fig. 4b). Lower crust gran-ulite xenoliths, which are scattered on the low-silica side of thediagrams, left of the silica gap, can be interpreted in the same way,as residues left after granite (batholiths) magma segregation. Thescattered distributions of lower crust granulites report a largecompositional heterogeneity for the lower crust, which sharplycontrast with the more homogeneous upper crust, dominated bylarge homogeneous granodiorite-granite batholiths.

The wide compositional region of lower crust xenoliths plottedin the MgO-SiO2 diagram (Fig. 4b), overlaps the composition ofgabbroic rocks that form the ca. 30 km thick lower crust at theKohistan arc section (Garrido et al., 2006; Dhuime et al., 2009). Allthese mafic xenoliths and the Kohistan gabbros have in common anabnormal composition compared to common magmas: in spite ofhaving silica contents close to basalts (ca. 50 wt.%), they have valuesof MgO < 7.0 wt.% and Mg# < 0.6, which are too low for basalticmagmas equilibrated with the peridotite mantle. Although a spe-cific explanation has been proposed for these low Mg# values

f interest included in the plot are: Typical Barrovian path and PT range (Jamieson et al.,phibolite (López and Castro, 2001). (b) Major compositional relations between lowerelts (Castro et al., 2010) from basalt-sediment mélanges respectively. The compositional., 2006; Dhuime et al., 2009) are shown for reference. Composition of lower crustumbers in the legend of the lower panel (b), and references to labels 10, Tuoyun, Norththolith are from Herve et al. (2007) and Castro et al. (2011b), and data from Easternlies close to the compositional gap displayed by granitoids of the Patagonian batholith

A. Castro / Geoscience Frontiers 5 (2014) 63e7572

(Jagoutz et al., 2006), these values are, in general, much lower thancumulates or residues left either after incomplete crystallization orpartial melting of a basaltic source. Consequently, the possibilitythat lower crust mafic granulites represent underplated basalticmagma can be questioned. Moreover, trace element relations (e.g.,low values of Rb, K and REE, positive Eu anomaly, among others)strongly support that they represent residues left after segregationof a melt. Because any process of fractionation and melt extractionwill produce residues richer inMgO and poorer in SiO2 compared tothe parental magma, the composition of the residual lower crustimplies a parental magma that can be close to an andesite.

That batholiths represent melts and lower crust granulitesrepresent the corresponding solid residues is firmly supported ongeological and petrological data, and at the same time compatiblewith thermal models and experimental phase equilibria. No otherhypothesis on granite magma generation has a so strong support.Interestingly, temperatures of more than 900 �C are recorded bylower crust xenoliths (Fig. 4a). These high temperatures are coin-cident with those of melting experiments needed to generatetonalitic and granodiorite batholiths from partially molten basalt-sediment subduction mélanges (Castro et al., 2010). If lower crustgranulites are the product of thermally induced metamorphism bymantle upwelling and lithosphere extension, mantlemelts (basalts)must be produced pervasively from the ascending hot mantle andintruded into the crust at the time of lower crust metamorphism.However, they are absent. Advective heat transport by mantle-derived basalt to the lower crust has been postulated as apossible cause of lower crust metamorphism and melting (Annenet al., 2006). However, according to these thermal models, abasalt-to-crust ratio of 2:1 is needed to reach the high tempera-tures recorded by granulites. Nevertheless, these expected highproportions of basaltic rocks are missing in the continental crust,and they never occur in relation with batholith magma generation.Several interesting questions emerge from these observations.Where is the heat source for lower crust granulite metamorphism?Are lower crust granulite true metamorphic rocks? Or simply arethey magmatic residues left after granite magma segregation? Isthere a protolith at the lower crust undergoing metamorphism andpartial melting? The hypothesis of magmatic residues seems to bethe most plausible. The high temperature recorded by mineralequilibria is that of granite magma at the time of segregation. Re-lations with batholiths (upper crust) are obvious from both majorand trace elements.

6. Plume assisted relamination: the off-crust origin ofbatholiths

Plume-assisted relamination is a new conception in subduction-related magmatism emerging from varied and independent ap-proaches, namely thermomechanical numerical experiments(Gerya and Yuen, 2003; Gerya et al., 2004; Gerya and Meilick, 2011;Vogt et al., 2013), experimental phase equilibria of batholithmagma generation (Castro et al., 2010, 2013a) and mass balancecalculations (Hacker et al., 2011). Essentially, the new conceptionproposes that magmas of intermediate composition (diorite,andesite) are generated by melting of subducted materials in siliciccomposite plumes (oceanic crust and sediments), which are finallyrelaminated to below the lower crust, where they split by meltsegregation into liquids, which are emplaced at the middle andupper crust (batholiths), and solid residues that remain at the lowercrust (mafic granulites). The relaminated andesite magmas mayreach the lower continental crust at high temperatures of about1000e1100 �C, containing a crystal fraction of about 50% accordingto predictions by phase equilibria experiments. Temperatures ofthe same order (ca. 1000 �C) are recorded by lower crust granulite

xenoliths around the world (Fig. 4a). Furthermore, the character-istic low-water content of calc-alkaline batholiths, with initialwater contents of about 1e2 wt.% H2O, is compatible with hightemperatures of about 1000 �C at the time of magma segregationfrom the composite underplated plumes.

The important role of subducted materials (sediments andaltered oceanic crust) in arc magmatism has been pointed out bymeans of geochemical studies of arc lavas (Plank and Langmuir,1998; Plank, 2005; Hacker et al., 2011). A detailed geochemicalstudy of batholitic rocks with adakitic affinities in Tibet (Gandesebatholith; Gao et al., 2007), yield to the conclusion that subductedsediments, and not crustal assimilation, were responsible forcrustal signatures. The rocks of the Gandese batholith in Tibet aretruly calc-alkaline granodiorites and tonalites in which, a promi-nent adakitic signature is identified by high Sr/Y and La/Yb ratios.The long-lived granodiorite-tonalite magmatism represented bythe large Cordilleran batholiths of the Americas is pointing in thesame off-crust origin for magmas. For instance, the Mesozoic Ca-nadian Coast batholith, extending for more than 1200 km along thecoast of British Columbia (Canada), shows an almost continuousand uniform plutonic activity for about 105 Ma (Crawford et al.,2005); the Coast batholith in northern Chile (Parada et al., 1999)shows an almost continuous plutonic activity for about 200 Masince Carboniferous to Tertiary; the North Patagonian batholith(Pankhurst et al., 1999) is formed by amalgamation of 25 plutonswith uniform granodiorite to tonalite composition, which areemplaced along 105 Ma; and, finally, the South Patagonian batho-lith is characterized by an intense plutonic activity of granodioriteto tonalite composition lasting for 150 Ma along Mesozoic andTertiary. These data strongly suggest that a crustal source, locatedwithin the continental crust, is very unlike, either as crustalcontaminant or as direct source of magmas. Subduction is the onlymechanism able to renew continuously the source of magmaswithout changing the composition of batholiths.

In the new paradigm of mantle-wedge diapirs, emerging fromthe above-mentioned thermomechanical numerical modeling,batholiths are the natural consequence of protracted introductionof fertile subducted materials into hot zones of the sublithosphericmantle. Potential magmatic implications have been checked bylaboratory experiments aimed to assess melt fractions and com-positions and to compare these with natural rocks (Castro et al.,2010, 2013a). Also the geochemical implications in terms ofradiogenic isotopes Sr and Nd have been modeled (Vogt et al.,2013). In summary, thermal and compositional requirements aretotally satisfied by the plume-assisted relamination model (Vogtet al., 2012; Castro et al., 2013b). A detailed analysis of chemicalvariation trends in batholiths, and their comparison with experi-mental phase equilibria, yield that an andesite magmatic precur-sor is favored to account for the generation of both granitebatholiths and lower crust granulites. Consequently, the study oflower crustal rocks, represented by xenoliths transported by ba-salts and exumed sections in continent-continent or arc-continentcollision zones (e.g., the Kohistan arc section; Garrido et al., 2006,2007; Dhuime et al., 2009), is essential to understand the origin ofbatholiths.

Fig. 5 shows a cartoon based on numerical thermomechanicalmodels (Gerya and Meilick, 2011; Vogt et al., 2012, 2013) summa-rizing the essentials of the proposed mechanism of plume-assistedcrustal relamination (Castro et al., 2013b). Details of structuresgenerated in numerical models supporting this general scheme(Vogt et al., 2013) reveal the complexity of processes with variedmagmatic implications. Two main types of diapiric structures areformed in the models, depending on the magnitude of meltweakening effects on the overlying lithosphere (Gerya and Meilick,2011; Vogt et al., 2013). These are: (1) Underplating diapirs (Fig. 5a),

Coupling

Roll-back

Roll-back

Sediments

Underplating silicic diapir

Subd.erosion

Partially moltenmélange

Continental crust

Partial crystallization and fractionation of relaminated magma chambersFormation of translithospheric diapirs

Sediment subduction, crustal erosion and silicic (magmatic) diapir generation

Batholith emplacement and generation of lower crust mafic residuesMafic magmas from reaction channels

Batholiths

Mafic granulites

a

b

c

1100 °C

Translithospheric diapir(reaction channels)

Translithospheric diapir(reaction channels)

Figure 5. Cartoon showing the time evolution of an active continentalmargin as based on thermomechanical numericalmodel predictions and compatibilitywith phase equilibria aboutmelt andmineral compositions. (a) Stageof plate coupling and consequent introductionof sediments and erodedportions of the continentalmargin into the subduction channel, togetherwith altered oceanic crust (mélange). Partial molten mélange is being transported into buoyant structures (silicic diapir), which is emplaced into the hot region of themantlewedge. (b)Melt percolation through the lithosphereprovokesplatedecoupling (GeryaandMeilick, 2011;Vogtet al., 2012). Thewholediapircanbeemplacedat the lowercrust (magma relamination)and fractionate intogranulite residuesandgranodiorie-tonalite liquids that formthebatholiths.Translithosphericdiapirs canbe formedat this stage, giving rise to reactionchannels (c) andnewmagma generation compatible with high-Mg andesites and sanukitoids that accompany to continental arc andesitic volcanism and batholiths respectively.

A. Castro / Geoscience Frontiers 5 (2014) 63e75 73

which spread below the lithosphere for several million years if meltpercolation has little effect on lithosphere strength and (2) trans-lithospheric diapirs (Fig. 5b), which are formed when lithosphere isweakened by melt propagation, allowing silicic diapir to move

upwards. These translithospheric diapirs ascend rapidly throughthe lithosphere mantle and are finally emplaced at crustal levelsresulting in the formation of magma chambers and batholiths (blueplutons in Fig. 5c). The lower parts of the diapiric structure (“diapir

A. Castro / Geoscience Frontiers 5 (2014) 63e7574

tail”) remain attached to the slab, forming a weak zone, whereintense reaction between subducted materials and the surroundingperidotite mantle is expected to occur. These translithospheric di-apirs can be seen as reaction channels, which were previouslyinferred to account for the chemistry of arc lavas, and particularlythe generation of high-Mg andesites (Kelemen, 1995; Kelemenet al., 2003). These inferences were tested experimentally withsatisfactory results (Castro et al., 2013b). By contrast, underplateddiapirs are growing for long periods of time (up to 20Ma) below thelithosphere (Fig. 5a). Melt extraction from the diapir may weakenthe lithosphere and provoke the ascent and emplacement at thelower crust, contributing largely to relamination and crustalgrowing (Fig. 5b). During the ascent and emplacement within thelower crust, granitic melt may segregate to form silicic (tonalite-granodiorite) batholiths at the upper crust (red plutons in Fig. 5c),leaving behind solid residues that form lower crust mafic granu-lites, which are mostly composed of pyroxene and plagioclase.Depending on pressure and temperature during magma segrega-tion of these composed diapirs, garnet may be a stable phase in thesolid residue, which may confer to the melt particular adakiticsignatures (high Sr/Y and Ce/Yb ratios) that are reported inparticular cases of calc-alcaline intrusions. Because garnet is not theliquidus phase in water-bearing andesite systems below 2.3 GPa(Ringwood, 1982), andesite liquids can be devoid of this adakiticsignature, which is not transferred to the respective fractionatedliquids. Finally, we have to mention that mantle peridotite sur-rounding both translithispheric channels and sublithospheric di-apirs, can be affected by fluids released from the water-rich silicicsystems leading to mantle metasomatism and generation of par-gasitic amphibole within the mantle. This metasomatized mantleregion is ready to generate new magma types of shoshonitic andmonzonitic affinities by decompression melting. These potassicmagmas are typically associated to late stages of lithosphereconvergence related to extension and mantle upwelling.

7. Concluding remarks

Fundamental paradoxes emerging from the application of on-crust models for the generation of granite batholiths can besolved by application of off-crust models. These receive supportfrom thermomechanical numerical modeling and laboratory ex-periments. The off-crust generation of granite batholiths consti-tutes a new paradigm inwhich important geological implications ofgranite magmatism can be satisfactorily explained. Geochemicalvariations with time in Cordilleran batholiths are in good agree-ment with this new conception.

That granites are hybrid rocks is a fact widely documented byisotopic ratios. However, theways bywhich these hybrid signaturesare acquired have remained controversial. That they are acquiredwithin the continental crust by modifications imposed to mantle-derived magmas is not supported by natural data.

Phase equilibria relations allow us to explore the possibilitythat batholith rock series represent liquids derived from crystal-lization or partial melting of a hybrid, homogeneous system.Major-element compositional trends follow cotectic trends, whosecomposition is largely dependent on intensive variables. Thecomparison between experiments and batholith chemical trendsproduces an empirical thermodynamic framework useful to un-derstand in a first approach the origin and evolution of magmasforming granite batholiths.

Acknowledgments

I acknowledge with thanks fruitful discussions about granitemagma generation along the last years with Taras Gerya and

Katharina Vogt at ETH-Zurich, Guillermo Corretgé at University ofOviedo, Michel Pichavant at ISTO-CNRS Orléans, Antonio García-Casco at University of Granada and Carlos Fernández and IgnacioMoreno-Ventas at University of Huelva. I thank M. Santosh forencouraging me to write this short review. Constructive criticismsby Taras Gerya and Antonio García-Casco contributed to improve anearlier version of themanuscript. Financial support for this researchcomes from Grants P09-RNM-05378 and CGL2010-22022-C02-01.

References

Al-Mishwat, A.T., Nasir, S.J., 2004. Composition of the lower crust of the ArabianPlate: a xenolith perspective. Lithos 72, 45e72.

Alonso Perez, R., Müntener, O., Ulmer, P., 2009. Igneous garnet and amphibolitefractionation in the roots of island arcs: experimental constraints on andesiteliquids. Contributions to Mineralogy and Petrology 157, 541e558.

Annen, C., Blundy, J.D., Sparks, R.S.J., 2006. The genesis of intermediate and silicicmagmas in deep crustal hot zones. Journal of Petrology 47, 505e539.

Beard, J.S., Ragland, P.C., Crawford, M.L., 2005. Reactive bulk assimilation: a modelfor crust-mantle mixing in silicic magmas. Geology 33, 681e684.

Bellos, L.I., Castro, A., Diaz-Alvarado, J., Toselli, A.J., 2013. Multi-pulse CotecticEvolution and In-situ Fractionation of Calc-alkaline Tonalite-granodiorite Rocks,Velasco Cambro-Ordovician batholith, Famatinian belt, Argentina. GondwanaResearch (submitted for publication).

Bird, P., 1979. Continental delamination and the Colorado Plateau. Journal ofGeophysical Research 84, 7561e7571.

Bonin, B., 2007. A-type granites and related rocks: evolution of a concept, problemsand prospects. Lithos 97, 1e29.

Carroll, M.R., Wyllie, P.J., 1990. The system tonalite-H2O at 15 kbar and the genesisof calc-alkaline magmas. American Mineralogist 75, 345e357.

Castro, A., 2004. The source of granites: inferences from the Lewisian complex.Scottish Journal of Geology 40, 49e65.

Castro, A., Carrasquilla, S., Bellos, L.I., Garcia-Casco, A., Paterson, S., 2013a. Basifi-cation of Cordilleran granitic magmas. Geology (submitted for publication).

Castro, A., Vogt, K., Gerya, T.V., 2013b. Generation of new continental crust bysublithospheric silicic-magma relamination in arcs: a test of Taylor’s andesitemodel. Gondwana Research 23, 1554e1556.

Castro, A., Aragón, E., Diaz-Alvarado, J., Blanco-Quintero, I., García-Casco, A., Vogt, K.,Liu, D.-Y., 2011a. Age and composition of granulite xenoliths from Paso de Indios,Chubutprovince,Argentina. SouthAmerican JournalofEarthSciences32, 567e574.

Castro, A., Moreno-Ventas, I., Fernández, C., Vujovich, G., Gallastegui, G., Heredia, N.,Martino, R.D., Becchio, R., Corretgé, L.G., Díaz-Alvarado, J., García-Arias, M.,Liu, D.-Y., 2011b. Petrology and SHRIMP U-Pb zircon geochronology of Cordil-leran granitoids of the Bariloche area, Argentina. Journal of South AmericanEarth Sciences 32, 508e530.

Castro, A., Diaz-Alvarado, J., Fernandez, C., 2012. Fractionation and incipient self-granulitization during deep-crust emplacement of Lower Ordovician ValleFértil batholith at the Gondwana active margin of South America. GondwanaResearch. http://dx.doi.org/10.1016/j.gr.2012.1008.1011 (in press).

Castro, A., Gerya, T., García-Casco, A., Fernández, C., Díaz Alvarado, J., Moreno-Ventas, I., Loew, I., 2010. Melting relations of MORB-sediment mélanges inunderplated mantle wedge plumes. Implications for the origin of cordilleran-type batholiths. Journal of Petrology 51, 1267e1295.

Castro, A., Patiño Douce, A.E., Corretgé, L.G., De La Rosa, J.D., El-Biad, M., El-Hmidi, H., 1999. Origin of peraluminous granites and granodiorites, iberianmassif, Spain. An experimental test of granite petrogenesis. Contributions toMineralogy and Petrology 135, 255e276.

Castro, A., Stephens, W.E., 1992. Amphibole-rich polycrystalline clots in calc-alkaline granitic rocks and their enclaves. The Canadian Mineralogist 30,1093e1112.

Chappell, B.W., White, A.J.R., 1974. Two contrasting granite types. Pacific Geology 8,173e174.

Chappell, B.W., White, A.J.R., 2001. Two contrasting granite types: 25 years later.Australian Journal of Earth Sciences 48, 489e500.

Clemens, J.D., Stevens, G., 2012. What controls chemical variation in graniticmagmas? Lithos 134e135, 317e329.

Clemens, J.D., Stevens, G., Farina, F., 2011. The enigmatic sources of I-type granites:the peritectic connexion. Lithos 126, 174e181.

Crawford, M.L., Klepeis, K.A., Gehrels, G.E., Lindline, J., 2009. Mid-cretaceous-recentCrustal Evolution in the Central Coast Orogen, British Columbia and South-eastern Alaska. Special Paper of the Geological Society of America, 97e124.

Crawford, M.L., William, A., Crawford, W.M., Lindline, J., 2005. 105 million years ofigneous activity, Wrangell, Alaska, to Prince Rupert, British Columbia. CanadianJournal of Earth Sciences 42, 1097e1116.

Dhuime, B., Bosch, D., Garrido, C.J., Bodinier, J.L., Bruguier, O., Hussain, S.S.,Dawood, H., 2009. Geochemical architecture of the lower- to middle-crustalsection of a Paleo-island arc (Kohistan Complex, Jijal-Kamila area, NorthernPakistan): implications for the evolution of an oceanic subduction zone. Journalof Petrology 50, 531e569.

Diaz-Alvarado, J., Castro, A., Fernandez, C., Moreno-Ventas, I., 2011. Assessing bulkassimilation in Cordierite-bearing granitoids from the central system batholith,

A. Castro / Geoscience Frontiers 5 (2014) 63e75 75

Spain; experimental, geochemical and geochronological constraints. Journal ofPetrology 52, 223e256.

Erdmann, S., London, D., Morgan Vi, G.B., Clarke, D.B., 2007. The contamination ofgranitic magma by metasedimentary country-rock material: an experimentalstudy. Canadian Mineralogist 45, 43e61.

Gao, Y., Hou, Z., Kamber, B.S., Wei, R., Meng, X., Zhao, R., 2007. Adakite-like por-phyries from the southern Tibetan continental collision zones: evidence for slabmelt metasomatism. Contributions to Mineralogy and Petrology 153, 105e120.

Garrido, C.J., Bodinier, J.-L., Burg, J.-P., Zeilinger, G., Hussani, S.S., Dawood, H.,Chaudhry, M.N., Gervilla, F., 2006. Petrogenesis of mafic garnet granulite in thelower crust of the Kohistan paleo-arc complex (Northern Pakistan): implica-tions for intra-crustal differentiation of island arcs and generation of conti-nental crust. Journal of Petrology 47, 1873e1914.

Garrido, C.J., Bodinier, J.L., Dhuime, B., Bosch, D., Chanefo, I., Bruguier, O.,Hussain, S.S., Dawood, H., Burg, J.P., 2007. Origin of the island arc Moho tran-sition zone via melt-rock reaction and its implications for intracrustal differ-entiation of island arcs: evidence from the Jijal complex (Kohistan complex,northern Pakistan). Geology 35, 683e686.

Gerya, T., Yuen, D.A., 2003. RayleigheTaylor instabilities from hydration andmelting propel “cold Plumes” at subduction zones. Earth and Planetary ScienceLetters 212, 47e62.

Gerya, T.V., Meilick, F.I., 2011. Geodynamic regimes of subduction under an activemargin: effects of rheological weakening by fluids and melts. Journal ofMetamorphic Geology 29, 7e31.

Gerya, T.V., Yuen, D.A., Sevre, E.O.D., 2004. Dynamical causes for incipient magmachambers above slabs. Geology 32, 89e92.

Gilbert, H., Yang, Y., Forsyth, D.W., Jones, C.H., Owens, T.J., Zandt, G., Stachnik, J.C.,2012. Imaging lithospheric foundering in the structure of the Sierra Nevada.Geosphere 8, 1310e1330.

Gilluly, J.,1948. Origin of Granite. TheGeological Society of America,memoir 28, p.139.Grosse, P., Bellos, L.I., de los Hoyos, C., Larrovere, M.A., Rossi, J.N., Toselli, A.J., 2011.

Across-arc variation of the Famatinian magmatic arc (NW Argentina) exem-plified by I-, S- and transitional I/S-type Early Ordovician granitoids of the Sierrade Velasco. Journal of South American Earth Sciences 32, 110e126.

Guo, L., Wang, T., Castro, A., Zhang, J.-J., Liu, J., Li, J., 2012. Petrogenesis and evolutionof late Mesozoic granitic magmatism in the Hohhot metamorphic core complex,Daqing Shan, North China. International Geology Review 54, 1885e1905.

Hacker, B.R., Kelemen, P.B., Behn, M.D., 2011. Differentiation of the continental crustby relamination. Earth and Planetary Science Letters 307, 501e516.

Hawkesworth, C.J., Kemp, A.I.S., 2006. Evolution of the continental crust. Nature443, 811e817.

Herve, F., Pankhurst, R.J., Fanning, C.M., Calderon, M., Yaxley, G.M., 2007. The SouthPatagonian batholith: 150 my of granite magmatism on a plate margin. Lithos97, 373e394.

Huang, X.-L., Xu, Y.-G., Liu, D.-Y., 2004. Geochronology, petrology and geochemistryof the granulite xenoliths from Nushan, East China: implication for a hetero-geneous lower crust beneath the Sino-Korean Craton. Geochimica et Cosmo-chimica Acta 68, 127e149.

Jagoutz, O., Müntener, O., Burg, J.-P., Ulmer, P., Jagoutz, E., 2006. Lower continentalcrust formation through focused flow in km-scale melt conduits: the zonedultramafic bodies of the Chilas Complex in the Kohistan island arc (NWPakistan). Earth and Planetary Science Letters 242, 320e342.

Jamieson, R.A., Beaumont, C., Fullsack, P., Lee, B., 1998. Barrovian regional meta-morphism: where’s the heat. In: Treloar, P.J., O’Brien, P.J. (Eds.), What DrivesMetamorphism and Metamorphic Reactions. The Geological Society, London,Special Publication 138, pp. 23e51.

Johnston, A.D., Wyllie, P.J., 1988. Interaction of granitic and basic magmas: experi-mental observations on contamination processes at 10 kbar with H2O. Contri-butions to Mineralogy and Petrology 98, 352e362.

Johnston, A.D., Wyllie, P.J., 1989. The system tonalite-peridotite-H2O at 30 kbar, withapplications to hybridization in subduction zone magmatism. Contributions toMineralogy and Petrology 102, 257e264.

Kay, R.W., Kay, S.M., 1993. Delamination and delamination magmatism. Tectono-physics 219, 177e189.

Kelemen, P.B., 1995. Genesis of high-Mg andesites and the continental crust. Con-tributions to Mineralogy and Petrology 120, 1e19.

Kelemen, P.B., Hanghøj, K., Greene, A., 2003. One view of the geochemistry ofsubduction-related magmatic arcs, with an emphasis on primitive andesite andlower crust. In: Holland, H.D., Turekian, K.K., Rudnick, R.L. (Eds.), The Crust.Treatise on Geochemistry, Vol. 3. Elsevier Pergamon, Oxford, pp. 593e659.

Kelsey, D.E., 2008. On ultrahigh-temperature crustal metamorphism. GondwanaResearch 13, 1e29.

Kretz, R., 1983. Symbols for rock-forming minerals. American Mineralogist 68,277e279.

Lee, C.-T.A., Morton, D.M., Kistler, R.W., Baird, A.K., 2007. Petrology and tectonics ofPhanerozoic continent formation: from island arcs to accretion and continentalarc magmatism. Earth and Planetary Science Letters 263, 370e387.

Lee, C.T.A., Cheng, X., Horodyskyj, U., 2006. The development and refinementof continental arcs by primary basaltic magmatism, garnet pyroxeniteaccumulation, basaltic recharge and delamination: insights from the SierraNevada, California. Contributions to Mineralogy and Petrology 151,222e242.

Loiselle, M.C., Wones, D.R., 1979. Characteristics and origin of anorogenic granites.In: Annual Meetings of the Geological Society of America and Associated So-cieties, San Diego, California, p. 468.

López, S., Castro, A., 2001. Determination of the fluideabsent solidus curve andsupersolidus phase relationships of MORB-derived amphibolites in the range4e14 kbar. American Mineralogist 86, 1396e1403.

Maaloe, S., Wyllie, P.J., 1975. Water content of a granite magma deduced from thesequence of crystallization determined experimentally with water-undersaturated conditions. Contributions to Mineralogy and Petrology 52,175e191.

Moyen, J.F., 2009. High Sr/Y and La/Yb ratios: the meaning of the “adakitic signa-ture”. Lithos 112, 556e574.

Naney, M.T., 1983. Phase equilibria of rock-forming ferromagnesian silicates ingranitic systems. American Journal of Science 283, 993e1033.

Pankhurst, R.J., Weaver, S.D., Herve, F., Larrondo, P., 1999. Mesozoic-cenozoic evo-lution of the North Patagonian batholith in Aysen, southern Chile. Journal of theGeological Society 156, 673e694.

Parada, M.A., Nyström, J.O., Levi, B., 1999. Multiple sources for the Coastal Batholithof central Chile 31e348S/: geochemical and SreNd isotopic evidence and tec-tonic implications. Lithos 46, 504e521.

Patiño-Douce, A.E., 1995. Experimental generation of hybrid silicic melts by reactionof high-Al basalt with metamorphic rocks. Journal of Geophysical Research 100,15623e15639.

Petford, N., Atherton, M., 1996. Na-rich partial melts from newly underplatedbasaltic crust: the Cordillera Blanca Batholith, Peru. Journal of Petrology 37,1491e1521.

Plank, T., 2005. Constraints from thorium/lanthanum on sediment recycling atsubduction zones and the evolution of the continents. Journal of Petrology 46,921e944.

Plank, T., Langmuir, C.H., 1998. The chemical composition of subducted sedimentand its consequences for the crust and mantle. Chemical Geology 145,325e394.

Read, H.H., 1948. Granites and granites. In: Gilluly, J. (Ed.), Origin of Granite.Geological Society of America, pp. 1e19.

Read, H.H., 1957. The Granite Controversy. Thomas Murby & Co., London.Ringwood, A.E., 1982. Phase transformations and differentiation in subducted

lithosphere: implications for mantle dynamics, basalt petrogenesis, and crustalevolution. Journal of Geology 90, 611e643.

Rudnick, R.L., Gao, S., 2003. Composition of the continental crust. In: Rudnick, R.L.(Ed.), The Crust. Elsevier-Pergamon, Oxford, pp. 1e64.

Saito, S., Arima, M., Nakajima, T., 2007. Hybridization of a shallow ‘I-type’ granitoidpluton and its host migmatite by magma-chamber wall collapse: the Tokuwapluton, Central Japan. Journal of Petrology 48, 79e111.

Selverstone, J., Stern, C.R., 1983. Petrochemistry and recrystallization history ofgranulite xenoliths from the Pali-Aike volcanic field, Chile. American Mineral-ogist 68, 1102e1112.

Sisson, T.W., Ratajeski, K., Hankins, W.B., Glazner, A.F., 2005. Voluminous graniticmagmas from common basaltic sources. Contributions to Mineralogy andPetrology 148, 635e661.

Skjierlie, K.P., Patino Douce, A.E., 2002. The fluid-absent partial melting of a zoisite-bearing quartz-eclogite from 1.0 to 3.2 GPa; Implications for melting of thick-ened continental crust and for subduction processes. Journal of Petrology 43,291e314.

Stephens, W.E., 2001. Polycrystalline amphibole aggregates (clots) in granites aspotential I-type restite: an ion microprobe study of rare-earth distributions.Australian Journal of Earth Sciences 48, 591e602.

Taylor, S.R., McLennan, S.M., 1985. The Continental Crust: Its Composition andEvolution. Blackwell, Melbourne.

Thompson, A.B., Matile, L., Ulmer, P., 2002. Some thermal constraints oncrustal assimilation during fractionation of hydrous, mantle-derivedmagmas with examples from Central Alpine Batholiths. Journal ofPetrology 43, 403e422.

Tuttle, O.F., Bowen, N.L., 1958. Origin of Granites in the Light of ExperimentalStudies in the System NaAlSi3O8-KAlSi3O8-SiO2-H2O, Memoir, 74. GeologicalSociety of America, p. 153.

Vogt, K., Castro, A., Gerya, T.V., 2013. Numerical modelling of geochemical variationscaused by crustal relamination (G-cubed). Geochemistry, Geophysics, Geo-systems 14, 470e487.

Vogt, K., Gerya, T.V., Castro, A., 2012. Crustal growth at active continental margins:numerical modeling. Physics of the Earth and Planetary Interiors 192e193,1e20.

Weber, M.B.I., Tarney, J., Kempton, P.D., Kent, R.W., 2002. Crustal make-up of theNorthern Andes: evidence based on deep crustal xenolith suites, Mercaderes,SW Colombia. Tectonophysics 345, 49e82.

Windley, B.F., 1995. The Evolving Continents, third ed. John Wiley & Sons Ltd.,Chichester.

Ying, J.F., Zhang, H.F., Tang, Y.J., 2010. Lower crustal xenoliths from Junan, Shandongprovince and their bearing on the nature of the lower crust beneath the NorthChina Craton. Lithos 119, 363e376.

Yoshino, T., Okudaira, T., 2004. Crustal growth by magmatic accretion constrainedby metamorphic PeT paths and thermal models of the Kohistan Arc, NWHimalayas. Journal of Petrology 45, 2287e2302.

Zhai, M., Fan, Q., Zhang, H., Sui, J., Shao, J.a, 2007. Lower crustal processes leading toMesozoic lithospheric thinning beneath eastern North China: underplating,replacement and delamination. Lithos 96, 36e54.

Zheng, J., Griffin, W.L., O’Reilly, S.Y., Zhang, M., Liou, J.G., Pearson, N., 2006. Granulitexenoliths and their zircons, Tuoyun, NW China: insights into southwesternTianshan lower crust. Precambrian Research 145, 159e181.