Earth-Science Reviews 160 (2016) 333–349

Contents lists available at ScienceDirect

Earth-Science Reviews

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

Invited review

Melt segregation and magma interactions during crustal melting:

Breaking out of the matrix
John D. Clemens ⁎, Gary Stevens
Department of Earth Sciences, University of Stellenbosch, Private Bag X1, Matieland 7602, South Africa

a r t i c l e i n f o a b s t r a c t

Article history: Differentiation of the continental crust begins with its partial melting. The products of crustal melting are silicic,
Received 29 February 2016 hydrous, H2O-undersaturated, granitic liquids that are generated within matrices of residual crystals. Crustal dif-
Received in revised form 25 July 2016 ferentiation requires that felsic magmas form and escape from these solid residua. An important question is
Accepted 25 July 2016
whether granitic magmas collect into large batches, within or near their sources, which then give rise to ascent
Available online 29 July 2016
or, alternatively, bleed out of the sources in smaller streams or pulses. We demonstrate that the physical reality
Keywords:
is closer to the second alternative, and thus question the validity of the concepts of magma segregation and
Crustal melting source fertility, as they are sometimes visualised.
Disequilibrium melting Granitic plutons contain chemically distinct sub-populations formed by source-level entrainment of the
Melt segregation peritectic assemblages into the melts. Rapid evacuation at source levels and high ascent rates protect the magmas
Melt migration from wall-rock interactions during their tenure at source depths and during their ascent through cool upper
Granitic magma crust. The existence of different types of granites, and hence of clearly defined chemically different magmas with-
Magma heterogeneity in plutons, dictates that felsic magmas must separate efficiently from their anatectic sources and must ascend to
Magma chambers
the sites of their emplacement with minimal chemical interaction with crust through which the magma must
move. Exposed deep-crustal sections are typically lithologically diverse, with the more fertile rocks commonly
forming discrete layers surrounded by rocks with contrasting compositions. If the melts were required to segre-
gate and accumulate into large volumes within their hot sources, especially by slow, gravity-driven melt perco-
lation, it is very likely that they would be substantially modified by reaction with diverse source rocks. Thus, the
processes mooted to occur in melting, assimilation, storage and homogenisation (MASH) or deep crustal hot
(DCH) zones would act to erase original compositional heterogeneities and produce larger batches of more ho-
mogeneous magma, perhaps carrying evidence of extensive magma mixing. From various lines of evidence, we
conclude that MASH and DCH zones may not exist and, even if they do, they cannot form the sources of most gra-
nitic magmas. The intrinsic heterogeneity in crustal source rocks, the likely occurrence of peritectic assemblage
entrainment and the inefficiency of magma mixing decree that granitic rocks will retain source-inherited chem-
ical and isotopic heterogeneities. Rapid, semi-continuous and disequilibrium withdrawal of small magma batches
from source rocks means that elements of whole-rock, trace-element chemistry and accessory mineral chemistry
will be decoupled from major-element variations. This also means that a large source with low nominal fertility
(due to low hydrous mineral content, for example) could still produce a substantial granitic pluton, by additions
of individually small increments of magma.
© 2016 Elsevier B.V. All rights reserved.

Contents

1. Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 334
2. Source volumes, fertility, heterogeneity and timescales . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 337
2.1. General considerations . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 337
2.2. Elements of source heterogeneity preserved in granitic rocks . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 338
3. Melt connectivity and melt migration . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 339
3.1. Experimental evidence . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 339

⁎ Corresponding author.
E-mail addresses: jclemens@sun.ac.za (J.D. Clemens), gs@sun.ac.za (G. Stevens).

http://dx.doi.org/10.1016/j.earscirev.2016.07.012
0012-8252/© 2016 Elsevier B.V. All rights reserved.
334 J.D. Clemens, G. Stevens / Earth-Science Reviews 160 (2016) 333–349

3.2. Field evidence . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 340

3.3. A note on early crystallisation in granitic magmas . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 340
3.4. A note on volcanic-plutonic relationships . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 340
4. The extent of chemical equilibration during the formation of granitic magmas . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 341
5. What does a granitic magma's source look like? . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 341
5.1. Theoretical considerations . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 341
5.2. Field observations . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 341
5.3. MASH and deep crustal hot zones . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 342
6. Some magma processes and the likelihood of their occurrence . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 343
6.1. Crystal fractionation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 343
6.2. Magma mixing and mingling . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 343
7. Conclusions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 345
Acknowledgements . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 347
References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 347

1. Introduction that lead to the formation of granitic plutons. Later in this paper we
present a more realistic view of magma formation and extraction,
Compositional and therefore density differentiation of Earth's conti- some reflections on the likelihood of the occurrence of processes, such
nental crust is fundamental to the stability and longevity of the conti- as fractional crystallisation and magma mixing in granitic systems,
nental masses, and always begins with a partial melting event, be it in and the reality of the mooted zones in which such processes might
the mantle or the crust. As pointed out by Sawyer et al. (2011) crustal occur. The perceived distinctions between some of the hypothetical
melting influences global-scale tectonic events and the chemical differ- steps involved in granitic magmatism have even been codified using
entiation of the crust is effectively irreversible. For partial melting of the the formal logic of information theory (e.g. Petford et al., 1997). One
crust, some broad principles apply to all anatectic events that result in model for the behaviour of granitic melt is that it grows in proportion,
significant differentiation of the crust. The products of such crustal melt- within the source, eventually forming a large, gravitationally unstable
ing events are silicic, hydrous, yet strongly H2O-undersaturated body that then ascends (e.g. Wickham, 1987). Although this idea per-
(Clemens, 1984), broadly granitic liquids that are initially contained sists, it seems unlikely in view of the implied diapiric mechanism (e.g.
within their complementary, restitic solid matrices. This solid material Clemens and Mawer, 1992; Petford et al., 1993). A second possibility
is formed from minerals that are residual from the subsolidus history is that magma could aggregate into a collection zone, or deep-crustal
of the source rocks, together with the peritectic crystals that formed staging chamber, outside the magma source, before ascending to final
as products of the melting reactions. To produce liquids that are suffi- emplacement depths. This may be the most common conception of
ciently hot and H2O-undersaturated that they are capable of ascending the process of magma segregation (e.g. Fig. 1a and the centre panel of
to high levels within the crust, and in some cases of erupting, the Fig. 1b). The third alternative is that magma could bleed as a continuous
peritectic fraction must be dominated by high-temperature ferromag- or intermittent stream out of the source, in relatively small batches that
nesian minerals produced by the incongruent melting of biotite- and connect the emplacement site to the source terrane in an upward-fo-
hornblende-bearing mineral assemblages. Studies of granulite-facies cused dendritic network of fractures that open and close episodically.
rocks, in which such partial melting reactions can be demonstrated to If the physical reality is closer to the third of these alternatives, then it
have occurred, confirm that peak metamorphic temperatures do not ex- is pertinent to ask whether there is validity in the concept of magma
ceed the limits of biotite or hornblende stability (e.g. Taylor et al. 2012; segregation, as it is commonly thought of. Similarly, at the opposite
Nicoli et al., 2015). Cases do exist where all available biotite and/or pole of the crustal differentiation process, if magma arrives continuous-
hornblende was consumed by the melting process, but this is restricted ly or semi-continuously over a protracted time period, is there validity
to rare, ultra-high-temperature rocks (e.g. Harley, 1998; Sengupta et al., in the concept of the large, melt-dominated granitoid magma chamber?
1999). The anhydrous, high-temperature mineral assemblages that at- Current models for partial melting of the crust, e.g. in ‘deep crustal
test to fluid-absent partial melting in such granulites are preserved be- hot zones’ (Annen et al., 2006; Solano et al., 2012) use the term ‘segre-
cause a significant fraction of the melt was lost from the system (White gation’ to apply to what is viewed as the stage of the process that follows
and Powell, 2002) and therefore unavailable for back-reaction on the partial melting, just as Petford et al. (1997) formally codified it. The im-
retrograde path. Thus, irrespective of the details of tectonic setting, na- plication here is that melt or magma collects somewhere within its
ture of the heat source and composition of the magma source, the H2O- source until it is physically able to migrate out of the system. See for ex-
undersaturated magmas that effect differentiation of the continental ample Fig. 1 and the model presented by Miller et al. (1988), for exam-
crust have segregated from granulite-facies residua within the temper- ples of this improbable behaviour. This view is at least partially
ature range of 830 to 900 °C, and typically at middle to lower crustal conditioned by long-standing mental pictures of substantial magma
depths. It is axiomatic that, for crustal differentiation to occur, a signifi- masses ascending out of their sources. We suggest that this impression
cant proportion of this segregated magma (with or without some is due to; a. the assumption that the felsic bodies within migmatites rep-
entrained solids) must ascend to shallower crustal levels. An important resent melt segregations, b. familiarity with partial melting experiments
question considered here is whether segregation is a staged or a semi- that map out melt percentage and composition over a temperature
continuous process, leading to steady or pulsed emplacement. range during equilibrium batch melting, and c. experiments that appear
Historically, granitic plutonism has been seen very much as a staged to show that magmas can only form once the threshold of some ‘rheo-
process, with distinct steps of melting, growth of the melt volume in the logically critical melt percentage’ has been crossed (Arzi, 1978;
source, magma segregation, ascent and emplacement, although the ex- Fountain et al., 1989; Vigneresse et al., 1996; Wickham, 1987). We con-
istence of feedbacks between these processes have been acknowledged tend that this staged and melt-accumulative vision of the process hin-
(e.g. Clemens, 1998; Brown, 2001). Fig. 1a, reproduced from Fig. 5 of ders improved understanding of what occurs during crustal melting,
Richards (2003), illustrates many of these supposed steps in the produc- the formation and compositional evolution of granitic magmas and, ul-
tion of a granitic magma. Fig. 1b, from Fig. 1 of Vigneresse (2006), illus- timately, the differentiation of the planet's continental crust. Instead, we
trates how this author saw the development of models for the processes propose a more oscillatory, analogue view, in which large masses of
 J.D. Clemens, G. Stevens / Earth-Science Reviews 160 (2016) 333–349 335

Fig. 1. Diagrams showing common perceptions of and the development of ideas on the processes (of melting, magma segregation and magma ascent) involved in granite genesis. (a)
Slightly modified version of Fig. 5 of Richards (2003), illustrating what appears to remain a common conception of the processes involved in the production of granitic magmas.
Richards regards the migmatitic lower crustal section of the diagram as a kind of MASH zone (after the terminology of Hildreth and Moorbath, 1988) in which melt proportion grows
by percolation until a diatexite is formed. Once a critical melt fraction (commonly considered to lie somewhere between 20 and 50 vol.%; Arzi, 1978; Vigneresse et al., 1996; Renner et
al., 2000) is reached, the diatexite spawns magma diapirs that ascend and eventually coalesce to form large dykes that transport the magma to upper crustal levels, where it may be
emplaced at its level of neutral buoyancy, experience various magma-chamber processes and perhaps release a fraction as volcanic magma. (b) A panel of three diagrams taken from
Fig. 1 of Vigneresse (2006), who conceived the temporal evolution of ideas running from top to bottom. The earliest (MASH) model (Hildreth and Moorbath, 1988) saw the processes
as melting, assimilation, storage and homogenisation. Vigneresse stated that this had been replaced by what he styled the MSAE model, adapted from Petford et al. (1997), with clear
stages of melting, segregation, ascent and emplacement. In the final panel, Vigneresse added a mantle component and distinguished the final stages from that of melting, thus
producing his m(M-SAE) model. The text discusses the various misconceptions in all these models, including the sizes of melt pulses, the role of migmatites (and of diatexites in
particular), the process of melt percolation, the concept of a rheologically critical melt fraction, the functionality of diapirism, the role of neutral buoyancy in magma emplacement and
the existence of MASH zones or of large liquid-dominated magma bodies in the crust. Figs. 8 and 9 illustrate our model for the processes that lead to the formation of granitic plutons.

magma form only at the ultimate emplacement sites of granitic plutons, melt is governed by the ability of a partially molten material to compact
most commonly in the upper crust and distant from their original, par- under gravity. This process has been modelled by McKenzie (1984,
tially melted sources. 1985). As pointed out by Miller et al. (1988), this process is predicted
A growing body of field, petrological, chemical and experimental ev- to result in “unrealistically long extraction times when applied to crustal
idence implies that granitic melts can escape their solid matrices almost melting situations” (op. cit., p. 145). Clemens and Mawer (1992)
as soon as they are formed, in bulk fractions of perhaps b2 vol.% and cer-
tainly b7 vol.% (Bons et al., 2004; Rabinowicz and Vigneresse, 2004;
Rosenberg and Handy, 2005). It may be difficult to conceptualise the ap-
pearance of such small amounts of melt, so Fig. 2 illustrates how a melt
volume of 2% might appear to an observer examining a 2D rock surface.
An informative study by Brown (2007) dealt with the case of fluid-pres-
ent melting, but most of the insights from this work are applicable also
to the more prevalent case of fluid-absent partial melting to produce
high-temperature, H2O-poor granitic melts. Brown found that melt
was lost from the source in a large number of individual, local, accumu-
lation-and-loss events, each triggered by the melt volume reaching the
connectivity transition at between 7 and 10 vol.% melt. Similarly, useful
insight into the nature of the magma segregation and migration process
was provided by Bons et al. (2004) who used numerical modelling to
predict that melt transport is normally episodic and that such systems
will quickly reach a state of self-organised criticality (Ortoleva, 1994)
in which melt may be evacuated without any further change to the me-
chanical configuration of the system. In other words, it is unnecessary
Fig. 2. Illustration of the appearance of two objects (a white band and square) each of
for melt to accumulate further to be withdrawn from its source, and which occupies 2% of the total area of the larger grey square (outlined in black). This 2D
slow, gravity-driven melt percolation is a physically improbable mode view represents how 2 vol.% melt might appear on the exposed surface of a partially
of melt migration in partially molten crust. Upward percolation of melted rock, and serves to calibrate the perception of the observer.
336 J.D. Clemens, G. Stevens / Earth-Science Reviews 160 (2016) 333–349

concluded that “Melt segregation is far more efficient than indicated by temperature crust can control overall crustal dynamics (e.g. Solano et
simple gravity compaction models. It is driven by shearing in the par- al., 2012). The idea that melt volume in the source increases to the
tially molten source regions, creating a mesh of melt-filled grain-edge point where the physical behaviour of the whole source changes funda-
fractures” (op. cit. p. 354). As pointed out by Clemens and Mawer, mentally is not supported by field observations. There are many exam-
magma transport is likely to be through networks of hydrofractures ples in which rocks have undergone partial melting at granulite-facies
that, due to their intrinsic inability to remain open permanently conditions and are well exposed. As noted by Clemens and Mawer
(Weertman, 1971), will result in pulsed magma extraction. Also, as (1992), such granulites do not generally have the textural characteris-
discussed by Hall and Kisters (2012) these fracture networks will com- tics of diatexites (disrupted and comminuted remnants of layers and
monly be self-organised (e.g. Petford and Koenders, 1998), with magma foliae dispersed in granitic-looking matrices). Rather, they observed
draining into fractures that have orientations that depend on the vary- that there is common coherency in features, such a primary composi-
ing melt production rate, existing anisotropies and the regional strains tional layering, over considerable length scales (see also Vernon et al.,
in deforming sources. High melt production rates will favour magma es- 1990; Stevens and van Reenen, 1992). Thus, although many granulites
cape through extensional fractures that shut off when melt production have lost a significant melt fraction, it is unlikely that they represent re-
wanes (Hall and Kisters, 2012). sidua produced by melt loss from diatexites. Nevertheless, it is com-
Likewise, there is now evidence that at least some granitic magmas monly taken as implicit that the sources of granitic magma normally
must escape from their sources within times of b 1 kyr (Villaros et al., pass through a diatexitic stage, even though high-level granitic plutons
2009a). Such time scales do not characterise segregation as it has com- are clearly dyke-fed and we seem to lack any geological evidence for a
monly been envisioned — a process lasting tens to hundreds of thou- transition between diatexitic rocks and metre-scale granitic dykes that
sands of years (e.g. Rabinowicz and Vigneresse, 2004). Rushmer could transport magma significant distances (see e.g. Vanderhaeghe,
(2001) demonstrated that biotite fluid-absent melting reactions com- 2001). Thus, in the literature, there appears to be some confusion over
monly do not result in positive volume changes. Thus, at low melt vol- the role of melt connectivity at low melt fractions and the physical be-
umes, differential melt migration may require deformation. The role of haviour of partially molten systems at high melt fractions. This may par-
shear deformation in the creation of pressure gradients that can lead tially stem from the assumption that migmatites preserve important
to small- to medium-scale melt localisation and transfer has been com- insights into the genesis of mobile granitic magmas (e.g. Brown, 1994;
prehensively explored using theoretical and experimental approaches Weinberg et al., 2013). Clemens and Droop (1998) reviewed the
(e.g. Holtzman et al., 2003, and references therein). Notwithstanding phase-equilibrium reasons why some melts never escape their source
the possible requirement for deformation to localise small melt vol- regions (and become migmatites) while others leave their sources and
umes, the experiments of Rushmer (1995, 2001) and Holyoke and go on to produce granitic plutons at much higher crustal levels. Hall
Rushmer (2001) clearly demonstrate that even the tiniest volumes of and Kisters (2012) examined the question of melt retention or escape
initial melt, generated in amphibole- and mica-bearing rocks, seem to from the point of view of structure and melting rate, concluding that
have been injected into fractures (brittle deformation) created by vol- melt escape depends on high melt production rates and the opening
ume changes attending the melting reactions (Fig. 3). Thus, there is po- of steeply dipping extensional fractures, which can be either layer-par-
tential for melt evacuation even in the absence of shear deformation of allel or cross-cutting (e.g. Fig. 4). The overall implication is that some
the rock body. Even though some published works (Vigneresse et al., migmatites (and especially stromatic migmatites and diatexites that
1996) acknowledge the potential for melt evacuation (by gravity or de- do not include well preserved granulite-facies peritectic minerals)
formation) at very low melt fractions, some (e.g. Vigneresse et al., 1996) may be intrinsically unsuitable as analogues for granite-forming pro-
then go on to discuss the perceived profound importance of rheological cesses. To put it another way, the efficiency of magma extraction
transitions (between bulk solid-like and bulk liquid-like behaviour of a means that it is unlikely that either stromatic or diatexitic migmatites
magma source), with the implication that such transitions (commonly will even be formed in the sources of granitic magmas that are
placed at melt fractions of 35 to 50 vol.%) somehow govern the overall emplaced distant from their points of origin.
process of magma escape from the solid matrix (e.g. Simakin and If it is unnecessary for melts to aggregate to some critical volume,
Talbot, 2001). Indeed, there is a growing body of literature proposing and they instead escape their matrices in a semi-continuous fashion,
that long residency periods of large magma volumes in the high- this has important implications for the rheology of the deep crust, for

Fig. 3. Reproductions of backscattered electron SEM images of the products of fluid-absent partial melting experiments on natural rock materials, (a) a muscovite-bearing quartzite (Fig. 2
of Rushmer, 2001) and (b) a biotite-plagioclase quartz rock (Fig. 8b of Holyoke and Rushmer, 2001). Both images show low-volume partial melt (labelled Liq in (b)) that has been injected
into fractures that penetrate quartz (Qtz) grains. In (b) the label ‘RP’ refers to solid reaction products (in this case orthopyroxene and spinel).
 J.D. Clemens, G. Stevens / Earth-Science Reviews 160 (2016) 333–349 337

Fig. 4. Photograph of an outcrop of an amphibolitic schist of the Khan Formation in the

Central Zone of the Damara Belt of Namibia. Here, in situ melting involved amphibole as
both a reactant and a product, and probably occurred under conditions of relatively high
aH2O. High-angle, ladder-like leucogranite networks illustrate the connectivity between
lens-like melt segregations hosted in extensional fractures connected to and draining
into larger, layer-parallel, vertical granite sheets. These sheets are interpreted to
represent ascent conduits that drained granitic magma to higher structural levels.

the chemistry and mineralogy of felsic magmas, for the interpretation of

migmatitic terranes and for the efficiency of crustal differentiation. The Fig. 5. Outcrop of the syntectonic Donkerhuk batholith, which was intruded along the
present work reviews the evidence, which appears to demand that northern edge of the Southern Zone of the Damara Belt in Namibia. Here the granite has
high-temperature sources must generally bleed off magma continuous- been assembled from literally thousands (and, regionally, tens of thousands) of
ly, and offers an explanation of how this process is reflected in the individual magma sheets derived from heterogeneous source rocks.

chemistry of the granitic magmas that intrude at high crustal levels. A

major focus is on what the existence of S- and I-type granites, and of
separate, chemically distinct sub-varieties of these, within individual
plutons, implies about magma segregation processes and how the in- scenario, a large magma chamber (20 km diameter by 5 km thick) in-
ferred behaviour affects magma mineralogy, chemistry and the degree truded into the middle crust and cooled only be conduction, would so-
of chemical equilibrium attained. lidify in b 1 Myr. More sophisticated modelling of the cooling of a
granite pluton by conduction only, which takes into account the fact
2. Source volumes, fertility, heterogeneity and timescales that wall rocks become more insulating as temperature rises, demon-
strates that a 5 km wide and 2 km thick granitic pluton, instantaneously
2.1. General considerations emplaced at 3 km depth, into country rocks with a 30 °C/km thermal
gradient, will cool to its 680 °C solidus in b60 kyr (Nabelek et al.,
A variety of different types of evidence (e.g. Glazner et al., 2004; 2012). Thus, plutons crystallising over time frames in the Myr range
Glazner, 2014; Memeti et al., 2010; Farina et al., 2012) indicates that must be constructed incrementally, and indeed in many quite small in-
the large, instantaneously emplaced magma chamber (the ‘big-tank’ crements (c.f. Fig. 5). The margins of the Tuolumne intrusive suite are
model) is unrealistic, as are some of the processes that have been envis- characterised by numerous sheeted intrusions (Glazner et al., 2004).
aged to occur within such chambers. As noted by Glazner et al. (2004) Similar structures have been described from the margin of the granitic
the volcanic rock record indicates that felsic magmas that erupt do so to granodioritic Heerenveen batholith in Mpumalanga province, South
typically in relatively small volumes, with thick and laterally extensive Africa (Belcher and Kisters, 2006) and are interpreted to reflect early se-
deposits being built up over protracted periods of time, and that larger quential injections of magma that quenched against the initially cold
eruptions, that form the most voluminous ignimbrite deposits, are country rocks. This mode of intrusion will cease once sufficient heat
caused by caldera collapse and consequent venting of magma from has been transferred into the emplacement site and individual magma
chambers that were themselves constructed over a considerable period. batches do not completely crystallise prior to the arrival of the subse-
Glazner et al. (2004) suggest that plutonic rocks are produced by the quent magma batches (e.g. Westraat et al., 2005; Belcher and Kisters,
same form of incremental construction as the volcanic sequences. 2006). This proposed behaviour is consistent with the estimated
From the mapped structure of large granitic intrusions we know magma flux required to construct the Mount Givens granodiorite of
that plutons are constructed through sheet-like injections of new the Sierra Nevada batholith; this large pluton is chemically and textur-
magma into the growing reservoir of partially crystallised magma. ally relatively homogenous but was constructed over a period of
Fig. 5 shows a spectacular example of the assembly of a batholith from 7 Myr (Frazer et al., 2014), with an average magma flux of only
sheet injections. In the case of the Tuolumne intrusive suite of the Sierra 10−3 km3/yr.
Nevada of California, the suite can be shown to have crystallised over at A pulsed mode of intrusion, with significant cooling between pulses,
least 10 Myr, while one of the lithological units, the Half Dome granodi- appears to be necessary to explain the preservation of inherited zircons
orite, crystallised over a period of 4 Myr (Coleman et al., 2004). This is in zircon-undersaturated leucogranites, as well as to preserve the differ-
highly significant because a liquid magma body the size of the Half ences observed in Hf isotope compositions of magmatic zircons from in-
Dome granodiorite would have crystallised in less than a tenth of this dividual granite samples. For example, Villaros et al. (2009a)
time if it were intruded as a single injection (Coleman et al., 2004). In- documented the existence of detrital zircons from leucogranites of the
deed, as demonstrated by Glazner et al. (2004), even the slowest-case Peninsula pluton in the Cape Granite Suite of South Africa. Here, the
338 J.D. Clemens, G. Stevens / Earth-Science Reviews 160 (2016) 333–349

leucogranitic magmas, which arose through biotite fluid-absent partial accumulate before spawning mobile magmas. Thus, if source rocks can
melting, were strongly zircon-undersaturated, yet contain a significant bleed off magmas at low melt fractions (e.g. 5%), as discussed above,
population of inherited zircon crystals. Indeed, this is the case for a the concept of source fertility will have little meaning for the production
great many granitic batholiths, e.g. in southeastern Australia (e.g. of granitic plutons.
Clemens, 2003). Using a static magma model and the experimentally
determined zircon dissolution rate (Watson, 1996) in granitic melt at 2.2. Elements of source heterogeneity preserved in granitic rocks
850 °C, the minimum reasonable temperature for biotite fluid-absent
melting, Villaros et al. (2009a) calculated b1 kyr zircon residence in Since granitic plutons are derived from sources that are substantially
the melt, prior to significant cooling. This implies that such magmas seg- more voluminous than the magmatic products, observation of expo-
regated from their sources, ascended, intruded the upper crust and sures of feasible source materials illustrates very clearly that the idea
cooled significantly, all within 1 kyr. The above constraints appear to of even a small pluton being derived from a single homogenous source
fit best with the concept of source rocks that yield magma in small must be erroneous, for both common I- and S-type plutons.
batches that can rapidly ascend into the upper crust and merge into a Metasedimentary rocks are typically compositionally layered on the
pluton that is substantially cooler than each newly arrived magma cm to m scales, commonly with alternating layers that represent origi-
batch. Similar observations have been made in the granites and grano- nally more clay-rich and more quartz- and feldspar-rich sediments.
diorites of the Central Iberian Zone (Bea et al., 2007) where Ti-in-zircon The incongruent melting reactions that will occur in such layers are like-
thermometry indicates magma temperatures of around 900 °C, yet the ly to differ in stoichiometry and temperature range of occurrence. As-
zircon in some of these rocks is almost wholly inherited. In such rocks, suming sufficient pressure to stabilize garnet over cordierite, melting
magma formation and transfer to the upper crust must have been in originally clay-rich layers is likely to occur through the model reac-
rapid. The common phenomenon of chemical undersaturation in zircon, tion: Bt + Qz + Pl + Sil = Grt + Ilm + melt, beginning at ~ 820 °C,
in magmas that contain inherited zircons, that were derived from zir- while melting in the originally more sandy layers is likely to proceed
con-bearing sources, also has profound implications for our under- through the reaction: Bt + Qz + Pl = Grt + Opx + Ilm + melt, begin-
standing of the longevity of magma residence in the source regions, as ning at a temperature some 30 °C higher (see e.g. Clemens, 2006). Even
well as the rates of many other magma processes. Such magmas in instances where the stoichiometry of reaction is similar, the Mg/
would readily become zircon-saturated if they were to linger in the (Mg + Fe) ratio of the biotite and the An content of the plagioclase
source for any geologically significant time period (N1 kyr). Conse- may vary from layer to layer, with consequences for melt chemistry.
quently, the very existence of zircon-undersaturated granites that also Similarly, within the intermediate composition metavolcanic and
carry inherited zircons demonstrates that such magmas did not linger metavolcaniclastic sequences proposed to constitute I-type granite
in their sources but were withdrawn essentially as they were formed, sources, the An content of plagioclase, biotite:hornblende ratio and bio-
to begin rapid ascent through the crust to their emplacement sites. tite content are likely to be important source variables that exert signif-
The mode of magma segregation exerts a considerable effect on icant influences on magma chemistry.
source fertility. During fluid-absent melting, a given crustal rock will From the above, it is clear that granitic plutons will inherit varying
be most fertile in the case where magma withdrawal occurs only at degrees of heterogeneity derived from their sources, as distinct from
the stage when the maximum melt volume has been attained. For heterogeneities that they may acquire through subsequent magmatic
most granulite-facies rocks this point in the P-T evolution will be at or processes; see also Bons et al. (2004). Some examples of source-
close to the temperature peak. Magma withdrawal at lower melt frac- inherited heterogeneities in granitic rocks are illustrated in Fig. 6. As
tions will reduce source fertility through reduction in the concentra- discussed by Stevens et al. (2007) and Clemens and Stevens (2012),
tions of incompatible elements in the solid residues (Yakymchuk and the major-element compositions of the liquid parts of granitic magmas
Brown, 2014). In partial melting experiments, all melt in the system is will vary in response to feldspar compositions, feldspar ratios and
retained over the temperature range investigated. Thus, these experi- quartz saturation in their source rocks. These effects control the concen-
ments map out maximum fertility for the bulk compositions used as trations of incompatible, fast-diffusing major elements, such as the alka-
starting materials. Despite this, experiments show that melt volumes li metals, and they determine parameters such as the K/Na and Ca/Na
produced from a source in the temperature range of biotite fluid-absent ratios of the melts. In metasedimentary sources the original clay, quartz
melting (approximately 850 to 900 °C), seldom exceed 30 vol.%, for bulk and plagioclase contents are the main determinants of these features of
compositions modelled on typical metapelites and metagreywackes melt chemistry.
(e.g. Stevens et al., 1997; Montel and Vielzeuf, 1997). Thermodynamic The temperature of melting and the stoichiometry of the melting re-
modelling, based on relevant metasedimentary compositions, predicts action in a magma source also influence factors such as the Fe, Mg and Ti
the formation of similar melt volumes (approximately 30 to 40%) at contents of the melts. However, these are considerations that assume
900 °C, if all the melt is retained until this temperature is reached (e.g. maintenance of chemical equilibrium. As recognised by Brown et al.
Nabelek and Bartlett, 2000; Yakymchuk and Brown, 2014). The sources (1995), if melt and residue are separated efficiently, the concentrations
of I-type granites are a more contentious issue, but the chemistry of I- of these more compatible and slow-diffusing elements will be signifi-
type rock associations dictates that they must arise through partial cantly influenced or even controlled by the rate of magma withdrawal.
melting reactions that consume hornblende and produce peritectic Indeed, Valentini et al. (2007) showed that natural millimetre- to
clinopyroxene (Clemens et al., 2011a; Clemens and Stevens, 2012). metre-scale vein or fracture networks in rocks have geometrical charac-
Such source rocks will generally produce less melt at a given tempera- teristics that equate with highly efficient melt transport and almost
ture than will typical metapelites. Using experimental constraints on guarantee withdrawal of melt before it can reach equilibrium with its
the fertilities of various common crustal rock types (e.g. Clemens, residue. Numerous small channels probably give way rapidly (in space
2006 and references therein) it is reasonable to infer that the source- and time) to fewer and larger channels (isolated hydrofractures or
rock packages from which granitic magmas are derived must generally ephemeral dykes) that focus flow efficiently upward and out of the
occupy at least three times the volumes of the magmatic products (see source regions (e.g. Clemens and Mawer, 1992; Kelemen et al., 1995;
also Clemens, 1988; Vielzeuf et al., 1990). Thus, given the scales of het- Bons et al., 2004; Kisters et al., 2009).
erogeneities observed in crustal rock sequences, it seems certain that To some extent, all the abovementioned factors influence the de-
even moderately sized granitic plutons could not be derived from chem- grees of source-inherited heterogeneity of the magmas and the rocks
ically homogenous sources. Note also that, even sources with relatively that ultimately crystallise from them. However, as Stevens et al.
low fertilities are capable of supplying enough magma to produce sub- (2007) and Clemens and Stevens (2012) showed, the most significant
stantial plutons. This is because it is unnecessary for melts to source-inherited chemical heterogeneities reflect variable entrainment
 J.D. Clemens, G. Stevens / Earth-Science Reviews 160 (2016) 333–349 339

Fig. 6. Examples of source-inherited chemical and isotopic heterogeneities in granitic rock series, with various parameters plotted against SiO2 wt%, FM (FeO + MnO + MgO wt%) or ASI
(mol. Al2O3/(CaO-3.3P2O5 + Na2O + K2O)). (a) Whole-rock initial 87Sr/86Sr isotope ratios (with 2σ errors) vs FM, (b) TiO2 vs FM, (c) and (d) whole-rock initial 87Sr/86Sr isotope ratios vs
ASI and SiO2, respectively. For rocks of the S-type Mount Wombat pluton (green dots) the data are from Clemens and Phillips (2014) and for the I-type Baynton (pink dots) and Beauvallet
plutons (blue dots) the data are from Clemens et al. (2016). Note that within the all three plutons there is significant variation in the initial Sr isotope ratios recorded by individual samples.
Within the Mount Wombat pluton the initial Sr isotope ratio variation is substantial and shows only a rudimentary correlation with ASI and none with FM values. The excellent correlation
between TiO2 and FM in such rocks is interpreted to arise through peritectic assemblage entrainment. If this were a feature produced by mixing between a crustally derived granitic magma
and a mantle-derived mafic magma, there would be a strong negative correlation between FM and initial Sr isotope ratio; ASI would also decrease with FM. Thus, the variation in initial Sr
isotope ratio is due to compositional variation in the magma source. Through peritectic assemblage entrainment, sources with similar major-element but different isotopic compositions
that are subjected to the same conditions of anatexis, will produce magma compositions that define coherent trends between elements that are concentrated in the peritectic minerals
because they melt by reactions with similar stoichiometry (e.g. Clemens et al., 2011a; Clemens and Stevens, 2012).

of peritectic mineral grains into the magmas as they leave their sources. massive scale during granitic plutonism, the degree of magma mixing
This process produces primary heterogeneity in Ti, Al, Fe, Mg and Ca is evidently quite limited (Clemens et al., 2010; Clemens and Stevens,
contents of the magmas, and also influences Na and K contents, princi- 2012).
pally through dilution. Heterogeneities in the concentrations of compat-
ible trace elements are commonly transferred from the source rocks to 3. Melt connectivity and melt migration
the magmas through variable entrainment of restitic accessory phases,
such as the inherited zircons discussed above. Clemens (2014) showed 3.1. Experimental evidence
that the chemical behaviour of some minor and trace elements in suites
of granitic magmas could even encode information about the textural For small volumes of granitic melts and magmas to be able to escape
heterogeneities that existed in the source rocks, prior to magma extrac- their sources, as we propose, it is necessary that such melts be able to
tion. Farina and Stevens (2011) and Farina et al. (2014a) also showed form interconnected networks within their residual matrices. In theory,
how crystal-scale disequilibrium between biotite and feldspar in the at equilibrium in a matrix that is not undergoing deformation, melt con-
source rocks could lead to very significant variations in the initial nectivity depends on the dihedral angles between the melt and solid
87
Sr/86Sr ratios of different magma batches. The preservation of such phases, with dihedral angles b 60° allowing melt extraction at almost
isotopic variability on scales of 1 m or less (Clemens et al., 2010) pre- any volume (von Bargen and Waff, 1986). Numerous experimental
sumably reflects, at least in part, the small volumes of the individual studies (Jurewicz and Watson, 1985; Laporte, 1994; Laporte and
magma pulses that leave a source. Similarly, the preservation of Watson, 1995) have shown that, in partially molten systems relevant
176
Hf/177Hf variation within magmatic zircon crystals, even within indi- to the formation of hydrous granitic magmas, dihedral angles are con-
vidual hand samples (e.g. Villaros et al., 2012; Farina et al., 2014b), illus- sistently low (in the range 30° to 59°). This means that the formation
trates the inefficiency of any subsequent magma processes that could of an interconnected porosity should occur at very small melt fractions.
lead to homogenisation and elimination of such primary heterogene- Lupulescu and Watson (1999) performed a series of annealing and dif-
ities. Finally, we observe that the construction of large plutons from fusion-couple experiments designed to detect the degree of felsic melt
small magma batches and the preservation of source-derived heteroge- connectivity in a mafic residual matrix, at low melt fractions. They
neities must mean that, although magma mingling must operate on a found that the connectivity threshold was already crossed at melt
340 J.D. Clemens, G. Stevens / Earth-Science Reviews 160 (2016) 333–349

fractions below 5 vol.%, which agrees with the results of the Wolf and retrograde cooling of the rocks had melt not been efficiently extracted
Wyllie (1991) melting experiments on a fine-grained, solid amphibo- from the source prior to retrogression (White and Powell, 2002;
lite, in which connectivity existed even in runs with only 2 vol.% melt Brown, 2013). Leucosomes in such rocks represent quartz and plagio-
present. This may be because in such experiments in piston-cylinder ap- clase that precipitated in the near-source magma plumbing system
paratus (and in natural rocks) some deformation is inevitable and this and were concentrated through melt drainage, during prograde meta-
enhances melt connectivity. Indeed, as pointed out by Clemens and morphism (Taylor et al., 2014). Larger leucosomes commonly occupy
Mawer (1992), in deforming aggregates, wetting angles are unstable orientations that facilitated rapid upward migration of magma, while
and commonly approach 0°, resulting in high melt connectivity. Thus, smaller leucosomes commonly have shallower dips and link with the
although there are variations in melt H2O content, melt-pocket geome- larger structures to build efficient networks for upward melt escape
try and mineralogy, and in degree of deformation, the conclusion is (e.g. Taylor et al., 2014; Ward et al., 2008; Kisters et al., 2009). Secondly,
clear. As recognised by Brown et al. (1995), even at very low melt pro- within restitic anatectic granulites, relict sedimentary structures, such
portions, the melt will form an interconnected network and can then as primary compositional layering and graded bedding, are commonly
potentially flow toward lower-pressure regions. Brown et al. (1995) quite well preserved, even in rocks that have lost as much as 40 vol.%
used modelling of various mechanisms for separation of melts from melt (e.g. Vernon et al., 1990; Guernina and Sawyer, 2003). Such fea-
their matrices in such systems. They found that gravity alone is insuffi- tures could not have been preserved if such a high volume of melt had
cient to allow widespread segregation of melt from residue in the crust, been instantaneously present in the rock. Thus, efficient, progressive
and that some additional driving force, specifically tectonic deforma- melt loss must have occurred as the melting reactions developed. Third-
tion, is required. This is unsurprising since, at very low melt volumes, ly, individual sub-20-cm-scale layers within anatectic migmatites con-
the bulk properties and rheology of the system will be governed by tain distinct mineral compositions and assemblages, even in rocks in
those of the solid matrix. The most efficient mechanism for melt extrac- which the volume of preserved peritectic material indicates 30 to
tion was found to be shear-enhanced compaction of the matrix (e.g. 40 vol.% melt loss (e.g. Nicoli et al., 2015).
Rutter and Neumann, 1995). It is most unlikely that any tract of deep
crust undergoing high-grade metamorphism would exist in an isotropic 3.3. A note on early crystallisation in granitic magmas
stress field. Thus, once melting occurs, and particularly when the con-
nectivity threshold is reached, it is inevitable that deformation will trig- Silicic volcanic rocks are formed from the same sorts of magmas as
ger melt segregation, well before any high melt proportion develops, granitic plutons. However, due to their eruption, at a relatively early
and without melt having to accumulate to some particular volume pro- stage of crystallisation, they commonly preserve phenocrysts that
portion. Brown (2001) suggested that these processes could be either crystallised at near-liquidus conditions. Thermobarometry of such phe-
cyclic (pulsed) or continuous, depending on the balance between the nocryst assemblages demonstrates that they commonly contain miner-
applied differential stress and the build-up of melt overpressure. al grains that formed in the middle or deep crust, at pressures very
From the foregoing, it is clear that the concept of a rheologically crit- much higher than those that obtained in the shallow crustal reservoirs
ical melt proportion, as originally envisaged by Arzi (1978) is inapplica- from which the magmas were erupted. A fine example is the
ble, since it represents a point at which the solid matrix breaks down Toombullup Ignimbrite in the Tolmie Igneous Complex of southeastern
and the whole diatexitic mass then behaves as a magma, with the prop- Australia (Clemens et al., 2011a). This intracaldera ignimbrite contains
erties of a melt-rich mush. The theoretical, experimental and geological two suites of phenocrysts, one that formed at approximately 400 MPa
evidence suggests that melt is withdrawn from the system at very much and the other at b180 MPa. The second, lower-pressure suite evidently
lower volume proportions than those necessary for such a mush to de- formed in the subvolcanic magma chamber that produced the erup-
velop. Indeed, Rutter and Neumann (1995) did not observe such a tran- tions. However, the earlier set formed in the mid crust, at a depth
sition in their experiments. Rosenberg and Handy (2005) produced a close to that of the inferred source. Occasionally, even plutonic rocks
comprehensive review and additional experimental data that showed contain similar evidence within garnet phenocrysts (e.g. Villaros et al.,
that there is indeed a dramatic transition in behaviour of a partially mol- 2009b).
ten rock system, which occurs between 0 and 7 vol.% melt. This change The foregoing is significant in the present context of magma segre-
in rock strength (deformability) is due to the formation of an intercon- gation because the early generations of phenocrysts could be viewed
nected melt-filled porosity, and a probable change in deformation as having formed in large magma bodies, deep in the crust, and there-
mechanism (e.g. Dell'Angelo and Tullis, 1988). Thus, it is apparent that fore taken as evidence that granitic magmas do accumulate into such
everything necessary for the potentially efficient withdrawal of melt masses prior to a distinct ascent phase toward final emplacement. How-
or magma from a partially molten source is in place at melt proportions ever, evidence from granulite-facies source rocks (e.g. Taylor et al.,
considerably lower than 10 vol.%. The melt forms an interconnected 2014) demonstrates that, in the leucosomes, plagioclase and quartz
network of pores, which weakens the solid matrix enough to allow de- crystallise because of slow diffusion of cations into and out of source pla-
formation and melt or magma migration to lower-pressure regions gioclase crystals. This effect retards plagioclase participation in the melt-
(commonly upward). The efficiency of this set of processes is the most ing reactions. In addition, as magma movement begins, recrystallisation
logical explanation for the apparent ability of small melt batches to of entrained crystals is necessitated by the change in bulk composition
exit the system without having spent enough time in contact with the of the system, from that of the source to that of the new magma. Even
solid residues to permit a close approach to chemical equilibrium, as in the case of slow diffusion in entrained minerals such as garnet, this
dealt with in Section 4, below. process occurs efficiently by a dissolution-precipitation mechanism
(e.g. Taylor and Stevens, 2010). Consequently, early-formed, high-pres-
3.2. Field evidence sure phenocrysts in felsic magmas are simply the consequences of
magma formation, and their presence does not indicate the existence
Field evidence from anatectic granulites illustrates several profound of magma chambers at depth.
realities that are common to the majority of well-exposed, high-grade,
anatectic rocks. Firstly, melts are efficiently drained through vein sys- 3.4. A note on volcanic-plutonic relationships
tems that are commonly near-vertical in orientation. We know this
from a variety of related lines of evidence. The significant volumes of an- The question of a possible direct kinship between silicic ignimbrite
hydrous peritectic minerals that are commonly present in migmatitic sequences and granitic batholiths has long been debated. However, it
granulites (e.g. Nicoli et al., 2015; Taylor et al., 2014) would very likely has proven very difficult to connect volcanic rocks with their supposed
have been substantially consumed by hydration reactions during plutonic counterparts (and v.v.) and these connections are mostly
 J.D. Clemens, G. Stevens / Earth-Science Reviews 160 (2016) 333–349 341

hypothesised rather than demonstrated (e.g. Bagdonas et al., 2016). As have mineral assemblages that are chemically incompatible with each
expressed by Lundstrom and Glazner (2016, p. 91) “…it is not clear if other. A good example of this is sillimanite-bearing leucosomes that
rhyolites are formed by the extraction of melt from shallow crystal occur in proximity to orthopyroxene-bearing metapelitic granulites in
mushes that otherwise solidify to form granite plutons, or are derived the Southern Marginal Zone of the Limpopo Belt (Taylor et al., 2014;
from a greater depth in parallel with granite plutons, or are formed by Nicoli et al., 2015). In these rocks, sillimanite is commonly separated
processes separate from those which form granite plutons.” It has from orthopyroxene by ≤ 10 cm, despite the fact that the peak metamor-
been supposed that this is because erosion has either been insufficient phic temperatures were below those required for stable coexistence of
to expose the batholiths that are supposed to underly volcanic com- these two phases, and that the rocks experienced a prolonged, post-
plexes or has been too deep to preserve the mooted volcanic roofs of anatectic, high-temperature evolution (Nicoli et al., 2015). Consequent-
batholiths. These volcanic-plutonic relationships are not the subject of ly, the equilibration volume in these rocks must have been b0.001 m3
the present paper but it seems clear that all variations might apply – vol- during the peak of metamorphism.
canic complexes with no underlying batholiths, batholiths with no vol-
canic expression and possibly even the elusive volcanic complex 5. What does a granitic magma's source look like?
complete with an underlying source batholith. In any case, from chem-
ical and isotopic evidence, we are certain that many rhyolitic magmas 5.1. Theoretical considerations
are the products of crustal melting, just as granitic magmas are. Thus,
apart from their mode of emplacement, the silicic volcanic magmas White et al. (1986a,b,c) ascribed the paucity of S-type granites in the
are likely to have formed by the same physical mechanisms as the silicic western United States to a general absence of suitable source rocks at
plutonic magmas. the depths where the granitic magmas were generated. Apart from
such notable exceptions, the worldwide ubiquitous occurrence of both
4. The extent of chemical equilibration during the formation of gra- S- and I-type granites, very probably indicates that their sources are
nitic magmas both common and widespread. Indeed, contrary to common perception,
many I-type source rocks may actually be mainly sedimentary in char-
The issue of disequilibrium between zircon and melt has already acter (e.g. Clemens et al., 2011a). In any case, the likelihood that melts
been discussed, in connection with the preservation of inherited zircons can escape their sources at low proportions provides an explanation of
in granitic magmas that are chemically undersaturated in zircon. We why these two major granite types are so commonly co-produced dur-
have also mentioned the common occurrence of isotopic heterogene- ing orogenic cycles. In essence, any rock containing the assemblage
ities in granitic magmas (e.g. Figs 6c and d), in the context of a general quartz-plagioclase-biotite has the potential to produce weakly to
lack of homogenisation of the magmas at any crustal level. Duffield strongly peraluminous leucogranitic to monzogranitic magma, depend-
and Ruiz (1998) modelled the temporal evolution of isotopic zonation ing on the Al-content of the biotite and irrespective of degree of entrain-
in feldspar crystals, with the assumption that these zonation patterns ment of the peritectic assemblage (Stevens et al., 2007). Similarly, any
result from wall-rock assimilation by the magma. Though the mecha- rock containing the assemblage quartz-plagioclase-biotite-hornblende
nism and the location of their modelled process differs from ours, has the potential to produce magmas that range in composition from
their work does predict preservation of such heterogeneities. Whatever weakly peraluminous leucogranites to metaluminous granodiorites,
the cause of the isotopic zoning, its preservation in the magmas and the with increasing degrees of entrainment of the clinopyroxene-bearing
rocks (e.g. Davidson et al., 2006; Gagnevin et al., 2005; Siebel et al., peritectic assemblage (Clemens et al., 2011a). Thus, a large proportion
2005; Waight et al., 2000) results essentially from the slow rates of of the clastic metasedimentary, metamafic and metagranitic rock record
chemical diffusion in feldspar crystals (e.g. Cherniak and Watson, has the potential to produce granitic magma.
1992) in relation to the lifetimes of the magmatic systems in question. The composition of the magma is controlled by the stoichiometry of
We have these sluggish kinetics to thank for our ability to measure the the melting reaction, with variations on the basic mineralogical theme:
source-inherited heterogeneities in the Sr isotopes of granitic series. Qtz + Pl + Bt (low Al) = transitional I/S-type, Qtz + Pl + Bt (high
The evident, pulsed delivery of contrasting magma batches to grow- Al) = S-type and Qtz + Pl + Bt + Hbl = I-type. Such source rock
ing plutons also implies that there is little or no homogenisation in iso- types are commonly present in association with each other and will pro-
tope systems at source levels, even for isotopes of elements that are duce granite magmas over similar and overlapping temperature inter-
concentrated in the melts. Furthermore, we have the geometrical anal- vals. Consequently, a wide range of different source rocks could
ysis of natural vein networks, which suggests that melt transport out of contribute to the construction of any individual granitic magma body.
the sources will be so efficient that disequilibrium is virtually guaran- The incompatible-element compositions of S-type magmas are largely
teed (Valentini et al., 2007). In addition, studies such as Braun and controlled by the compositions of the plagioclase and the biotite. In
Kriegsman (2001) have demonstrated the existence of grain-scale, the case of I-type magmas, the controls will be the compositions of
major-element disequilibrium in partially melted xenoliths in mafic ig- the plagioclase, the biotite and the hornblende. The compatible-element
neous rocks, a possible analogue of the initial stages of fluid-absent par- compositions of the magmas will be controlled by the stoichiometry of
tial melting in regional terranes. Thus, we conclude that different parts the melting reaction, through the entrainment of the peritectic mineral
of a source terrane, undergoing simultaneous melt generation and ex- assemblages. Thus, different K/Na ratios in granitic rocks probably re-
traction, must exist in a state of mutual and general chemical and isoto- flect the ranges of mineral proportions and compositions in the source.
pic disequilibrium, on a variety of scales; see also McLeod et al. (2012). In contrast, the Ti/(Fe + Mg) (or TiO2/(FeO + MgO) ratios of the
Bea (1996), Bea et al. (2007), Beard (2008), Farina and Stevens (2011) magmas and rocks (e.g. Fig. 6b) will reflect the stoichiometry of the
and Farina et al. (2014a) provide primary observational data supporting melting reaction, and so will be very similar for all sources of the same
these conclusions, especially for trace-element concentrations. Howev- general type and in which a similar partial melting reaction takes
er, it seems highly likely that spatially local equilibrium is attained be- place. Thus, the proposition that a given granitic pluton has a unique
tween the main silicate minerals participating in the melting and identifiable ‘source’ composition is highly implausible.
reactions, the melts and the peritectic solids; see also Pichavant et al.
(2007). We do not yet know the actual equilibration volumes involved. 5.2. Field observations
However, as discussed above, there are indications that they may be
quite small, probably smaller than individual crystals of porphyroblasts Probably the best examples of the sources of granitic magmas are
of slow-diffusing minerals such as garnet and plagioclase. Layered granulite-facies metasediments that have undergone metamorphism
metasedimentary granulites commonly contain adjacent layers that at temperatures in excess of 850 °C. The reasons for this are threefold.
342 J.D. Clemens, G. Stevens / Earth-Science Reviews 160 (2016) 333–349

Firstly, S-type granites are derived through biotite fluid-absent partial the MASH concept has physical reality. As far as we are aware, this situ-
melting in source rocks of this type (Clemens and Wall, 1981; ation remains unchanged. For example, Miles et al. (2013) describe
Chappell and White, 2001; Clemens, 2003; Villaros et al., 2012). Second- chemical features of a granitic pluton that they suggest are “likely to
ly, the relevant reactions, modelled as Bt + Sil + Qtz + Pl = Grt + melt have occurred within a deep crustal hot zone” (p. 1, op. cit.). The de-
and Bt + Qtz + Pl = Grt + Opx + melt, can be shown to have occurred scribed composite nature of such plutons could just as likely originate
in the granulites (e.g. Stevens and van Reenen, 1992; Taylor and in the emplacement of a range of magmas derived from somewhat dif-
Stevens, 2010); The common preservation of the bulk of the peritectic ferent lithological varieties in a heterogeneous source terrane. Also in
mineral assemblage produced by the melting reactions means that the this context, experiments by Knesel and Davidson (2002) show that
granulites must have lost most of the melt volume that was produced progressive partial melting of a single source rock could produce melts
(White and Powell, 2002). Consequently, it is very likely that general with a range of isotopic characteristics. Farina and Stevens (2011) am-
observations in such metasedimentary granulites, particularly in cases plified this result, for Sr isotopes, using modelling calculations. Such al-
where they underwent metamorphism at mid- to lower-crustal pres- ternative, single-source models for magma heterogeneity are not only
sures, provide insight into the processes in the sources of S-type gran- simpler but avoid the difficulties that Miles et al. (2013) encountered
ites. In most migmatitic metasedimentary granulites, compositional in trying to explain various ‘enigmatic’ features of the rock chemistry.
layering is preserved on the dm to m scale, and is interpreted to repre- We still have no geological evidence that MASH zones are common or
sent bedding in the original sediments (e.g. Clemens and Mawer, 1992; perhaps even that they exist. Likewise, the first sentence in the abstract
Taylor et al., 2014; Taylor and Stevens, 2010; Ward et al., 2008). Com- of Annen et al. (2006, p. 505), in which DCH zones are hypothesised,
monly, evidence is preserved for sequential melting reactions that in- makes it clear that this is “A model for the generation of intermediate
volved incongruent breakdown of muscovite followed by biotite and silicic igneous rocks … based on experimental data and numerical
partial melting reactions at higher temperatures (e.g. Stevens and van modelling.” Again, there was no cited geological basis for the concept.
Reenen, 1992; Lassalle and Indares, 2014). In addition, in studies of In potential counterpoint, the Bergantz et al. (2013) AGU meeting
high-grade anatectites, evidence is commonly recorded for melt accu- abstract and Otamendi et al. (2009) describe a tilted section through
mulation in shear zones and low-pressure sites (e.g. Brown et al., part of the Early Palaeozoic Famatina arc in west-central Argentina.
1995; Kisters et al., 2009). In general, evidence of the sort noted above Here the upper crustal plutons are connected to their lower crustal
requires the source to retain its structural integrity. Consequently, it ar- sources by a series of veins and dykes. The hybridisation and mixing
gues against the melt volume in the source having reached values high phenomena, described in these works, evidently took place between
enough for the whole source to have behaved as a magma. batches of newly arrived mantle magma and the cumulates and inter-
mediate magmatic products of fractionation of earlier batches. The au-
5.3. MASH and deep crustal hot zones thors specifically state that melting of the local metasedimentary host
rocks played only a minor role and that the K2O contents of the mafic
Melting, assimilation, storage and homogenisation (MASH) are com- magmas remained unaffected. Somewhat paradoxically, they concluded
monly perceived as the birthplaces of a variety of magmas, especially in that this represents an archetypal MASH zone. If it is, then we should not
connection with collisional tectonics. Thus, it is useful to understand look to MASH zones for the genesis of ordinary S- and I-type granitic
what these postulated zones represent. These zones are envisaged to magmas but perhaps of arc tonalitic and dioritic magmas that have
occur deep in the crust, and the processes hypothesised to occur within only slight crustal signatures. Even so, the most recent and detailed
them need to be viewed as near-source phenomena. Hildreth and work on the Famatina rocks (Walker et al., 2015) indicates that they
Moorbath (1988), make it clear that MASH zones are envisaged as do not record in-situ anatexis of early-crystallised, mantle-derived
places in which there would be long-term magma storage and mixing, magmas and that no significant volumes of tonalitic magma were
to produce hybrid magmas that could have both crustal and mantle produced.
components. The implication is that the melts involved would remain Pearce (1996) puts forward the Chelan ‘migmatite’ complex in
in or near their sources, in crystal-liquid mushes, for periods long Washington state (USA) as another example of a MASH zone. This is ac-
enough for diffusion, reaction and mechanical mixing to create relative- tually a complex intrusive zone through which numerous batches of
ly homogeneous hybrid magmas. Bachmann and Bergantz (2004) esti- magma passed over a period of time. An appropriate term would be
mated the timescale required for separation of felsic liquid from a ‘magma transfer zone’, as used, for example, by Barnes et al. (2003).
cumulate crystal mush to range from 104 to 105 yr, so this would be Pearce re-interprets the MASH acronym as Magma Assimilation Segre-
the approximate time scale for magma storage and interaction in a gation and Homogenisation, describing it more as a mechanism than a
MASH zone. physical place. He notes, however, that the felsic rocks in the Chelan
The DCH designation, first used by Annen et al. (2006), is less specific complex retain the chemical and isotopic signature of the putative ba-
in what it implies about the nature of such zones and the processes saltic end-member in an hypothesised mixing between crustal melts
within them. However, the cited paper makes it clear that these zones and mantle magmas. Thus, it could be suggested that there might be
are envisaged as places in the deep crust, intruded by numerous sheets no older crustal component here. Indeed, our own observations in this
of mantle-derived magmas. Heat and fluids from the crystallising mafic complex, together with the descriptions in Hopson and Mattinson
magmas are seen as causing partial melting of the neighbouring crustal (1994), suggest that the country-rock migmatites do not contribute to
rocks, and these crustal melts, stored in or near their sources, are then the magmas in the transfer zone but occasionally form large xenoliths
assumed to interact with the differentiates from the mafic magmas to within it. The main rock types in the transfer zone are tonalites, and
produce the observed elemental and isotopic variations measured trondhjemites that fractionated from them, with some diorites and ul-
among the magmas of subduction zones, and in other tectonic settings. tramafic cumulates. Thus, it is unsurprising that a crustal signature is
The implication here is that melts from both sources (crustal melting generally absent here. In more recent years, a number of publications
and differentiation of mantle magmas) must reside in physical contact have shown diagrams that portray the chemical conception of the crust-
for time periods long enough to form large volumes of hybrid magmas. al sections beneath large ignimbrite or batholith complexes. A fairly rep-
Thus, although there are some conceptual differences between MASH resentative example is illustrated in Fig. 7 which is reproduced, slightly
and DCH zones, they are fundamentally very similar concepts. modified from Fig. 14A of Lipman and Bachmann (2015). Note that, in
When first envisaged by Hildreth and Moorbath (1988) MASH zones this vision, the entire upper lithosphere, from the shallow crust down
were a purely theoretical expedient devised to explain features of the to the Moho and beyond, is conceived as partially molten mush zones
chemistry (and particularly the isotope chemistry) of subduction-relat- that have different balances of felsic and mafic compositions. We are
ed magmas. These authors had no geological evidence to suggest that forced to question the reality of such a conception, and we are left
 J.D. Clemens, G. Stevens / Earth-Science Reviews 160 (2016) 333–349 343

pointed out that the dominant mechanism for producing heterogeneity

in granitic magmas is not any kind of differentiation process but rather
magma heterogeneity inherited from the source and transmitted to the
emplacement level by virtue of a general lack of mixing between the nu-
merous pulses of magma involved. They were careful to point out that,
at or just below the final emplacement levels, plutonic granitic magmas
commonly do undergo local crystal fractionation, to produce volatile-
enriched near-roof zones. These authors also envisaged small amounts
of magma mixing and mingling in long-lived magma systems, but
emphasised that all these other mechanisms constitute a secondary
overlay on the primary, source-derived variations that they ascribed
to a process of entrainment of variable proportions of the peritectic
phases that formed during the melting reactions in the source rocks.
By way of example, the classic rhyolitic ignimbrite sequences of the
Bishop Tuff in California have long been regarded as the products of
eruption from a stratified magma chamber. Different workers have as-
cribed the inferred stratification to a variety of differentiation mecha-
nisms but with the concept of crystal fractionation dominant and the
compositional layering of the magma chamber universally accepted
(see e.g. Hildreth, 1979; Michael, 1983; Cameron, 1984). Hildreth and
Fig. 7. Slightly modified version of Fig. 14A of Lipman and Bachmann (2015), which shows Wilson (2007) provided evidence that what they interpret as evidence
a conception of the vertically exaggerated (v/h = 3) lithospheric column beneath the of layering could not have been produced by a mechanism involving
Southern Rocky Mountain volcanic field (SRMVF), showing the volcanic edifice of the crystal-liquid separation. They ascribe this model structure of the
central and western San Juan Mountains of Colorado in the USA, and the structure of the
magma chamber to successive delivery of contrasting, largely liquid
proposed underlying San Juan batholith, as modelled from seismic surveys and
examples of inferred tilted batholith sections elsewhere. The horizontally and vertically magma batches from different source rocks. This is essentially similar
extensive mush zones shown here are meant to represent locations in which AFC and to the idea behind Clemens et al. (2010) and Clemens and Stevens
MASH processes operated, during the life of the magmatic system. (2012), with the notable difference that Hildreth and Wilson ascribed
the production of this source-inherited heterogeneity to liquid-mixing
with no credible geological example of a MASH or DCH zone in which processes in a MASH zone. In an interesting further development,
granitic magmas with crustal signatures were formed as part of a com- Gualda et al. (2012) modelled crystallisation in the late-erupted Bishop
positional continuum with a mantle-derived basaltic end-member. In- Tuff. These authors state “…derivation of Bishop pumice from two dis-
deed, if MASH or DCH zones are supposed to have the structure crete, laterally juxtaposed magma bodies of rhyolitic magma is favored
portrayed in Fig. 7, we can be reasonably certain that there is no pros- by phase equilibria … [and that their modelling] … does cast significant
pect of discovering an example in the geological record. As a final doubt on the widely held view of an archetypical vertically stratified
thought on DCH and MASH zones, it may be worth recalling that vast magma body…” (Gualda et al., 2012, p. 887). Thus, for this classic exam-
volumes of granitic rocks have formed in settings that have nothing to ple of a zoned felsic magma, there is now serious doubt not only that dif-
do with subduction, including most of the archetypal S- and I-type bath- ferentiation was involved in its construction but also on the whole idea
oliths of southeastern Australia. of a layered structure. The lateral juxtaposition of differing magma
batches, hypothesised by Gualda et al. seems to correspond more close-
6. Some magma processes and the likelihood of their occurrence ly with what is observed or inferred about the distribution of heteroge-
neities in granitic plutons, as summarised in Clemens et al. (2010) and
6.1. Crystal fractionation references cited therein. Irrespective of the exact geometrical relation-
ships, it seems that assembly of granitic magma bodies involves the ag-
In the first two decades of the 20th Century CE, the legendary Nor- glomeration of numerous batches of magma and that, in many cases,
man L. Bowen introduced the idea of crystal fractionation as a mecha- these batches separated from their partially molten sources at quite
nism for producing magma chemical variations. In the classic simple low melt fractions.
case, the settling under gravity of early-formed crystals would lead to
a vertically stratified magma body with the more fractionated (felsic) 6.2. Magma mixing and mingling
liquids at the top and the more mafic (partly cumulate) magmas at
the base. This model gained such momentum that it saw almost univer- The idea that magmas with contrasting compositions may mix to
sal acceptance and application to nearly every type of magma system. form hybrids has considerable antiquity, having been introduced by
Indeed, there are clearly demonstrated examples of its occurrence and Durocher (1857). Harker (1904) hypothesised magma mixing and min-
a sound theoretical basis for this disequilibrium process in nature. gling as the mechanism for the formation of a monzodioritic rock varie-
Using the existing literature on field relations and chemical varia- ty that was then called ‘marscoite’, on the Isle of Skye in the Scottish
tions, Clemens et al. (2010) explored the extent of the applicability of Inner Hebrides. A range of outcrop and hand-specimen features, such
crystal fractionation to the occurrence of compositional heterogeneities as ‘banded’ pumice, dark-coloured enclaves and ‘pillows’ in felsic volca-
in both volcanic and plutonic felsic magma systems. Their observations nic and plutonic rocks, and thin-section evidence for corrosion and
centred on the preservation of isotopic heterogeneities on a variety of overgrowth of phenocrysts formed the original basis of the magma-
scales, the intrinsic pulsed nature of both magma formation and mixing hypothesis (e.g. Eichelberger and Gooley, 1977; Whalen and
magma emplacement, and the incompatibility between trace-ele- Currie, 1984; Barbarin, 1990; D'Lemos, 1996), although much of this ev-
ment-based fractionation models and major-oxide variations in the idence is of magma mingling rather than mixing.
rocks. Relatively high melt viscosities and slow rates of chemical diffu- The chemical and isotopic characteristics of I-type granitic rocks are
sion retard processes such as crystal settling and magma mixing (e.g. intermediate between those of evolved crust and the mantle (e.g.
Glazner, 2014; Clemens, 2015), and result in preservation of heteroge- DePaolo and Wasserburg, 1979; Metcalfe et al., 1995; Keay et al.,
neities. Clemens et al. (2010) concluded that the bulk of the variation 1997). As observed by Clemens and Stevens (2012), this led many au-
observed cannot have been produced by crystal fractionation and thors to hypothesise a magma-mixing origin for I-type granitic magmas
344 J.D. Clemens, G. Stevens / Earth-Science Reviews 160 (2016) 333–349

in particular. Elemental and isotope-based models that involve magma involvement of mantle-derived magmas in their genesis. Melting of re-
mixing sprung up in the 1980s and continue to be refined (e.g. cent, mafic to intermediate additions to the crust can account for such
DePaolo, 1981; Castro et al., 1991; DePaolo et al., 1992; Beard et al., characteristics. In this context, we suggest that andesitic volcanic and
2005); they have proven very popular and have been widely applied. volcaniclastic sequences, as well as immature sediments derived from
These trace-element- and isotope-based models have rarely been tested such rocks, are common sources of I-type granites with relatively prim-
against constraints imposed by major-element chemistry of the rocks itive isotope characteristics. 2. The chemical range exhibited by granitic
and, as pointed out by Clemens and Stevens (2012), commonly fail rocks does not require an origin by mixing with mafic magmas. The
those mass-balance tests when they are employed (e.g. Cristofides et compositional range of common I- and S-type granites is well accounted
al., 2007; Clemens et al., 2010). Clemens et al. (2010) also pointed out for by different degrees of entrainment of the peritectic assemblages
that many granitic bodies contain evidence of considerable isotopic into the magmas. Some of the inter-element correlations defined by
and elemental heterogeneity, on a variety of scales, in the apparent ab- suites of granites seem to require this process and appear to be incom-
sence of features attributable to mingling between more mafic and felsic patible with magma mixing involving mafic end members. 3. The for-
magmas. mation of granitic rocks with high Mg#, and high contents of first-row
On the basis of chemical variations within both S- and I-type granitic transition metals (e.g. V, Cr, Co and Ni) does not require comprehensive
associations from southeastern Australia, Chappell (1996) persuasively mixing between mantle-derived magmas and crustal melts. Subduc-
argued that the observed variations are incompatible with the magmas tion-zone metasomatism of the mantle wedge, particularly in situations
having originated by the mixing of mafic and felsic end-members. He involving transport of clastic sedimentary material down the subduc-
found that chemical differences at either end of the compositional tion channel, enriches the mantle wedge in a range of components de-
ranges in granitic suites are similar, so that both the most mafic and rived from the felsic crust. Anatexis of heterogeneous, metasomatised
felsic rocks have similar relative abundances of many elements. This fea- mantle wedge materials can produce granitic, intermediate and mafic
ture cannot be reconciled with magma mixing having been the origin of magmas.
the variations. Other chemical, physical, thermodynamic and observa- Thus, it seems that magma mixing, like crystal fractionation does
tional arguments, and evidence against the general applicability of occur but that these processes cannot account for the extent and nature
magma mixing have been rehearsed in a variety of publications and of the chemical and isotopic variations observed among the products of
summarised in Clemens and Stevens (2012). Note that some chemical granitic magmatism. Nor, as noted earlier, would extensive magma
features of granitic associations, such as those displayed in Figs. 6a and mixing be compatible with the observed preservation of source-derived
b, are quite fragile, in the sense that they would be at least greatly dis- elemental and isotopic heterogeneities on quite small scales. The model
turbed and possibly completely destroyed were these magmas to have that we present – pulsed separation and ascent of compositionally con-
undergone significant chemical interaction with mantle-derived mafic trasting magma fractions from a heterogeneous source, at low melt vol-
magmas. The general preservation of such delicate chemical trends in umes – carries with it the implication that there is generally very little
granitic rocks (e.g. Clemens et al., 2011a; Clemens and Stevens, 2012) chemical interaction between mafic heat-source magmas and the crust-
therefore suggests that there is usually only minimal and local chemical al melts produced adjacent to them, except possibly to preferentially
interaction between crustal melts and mafic heat-source magmas. The modify the compositions of the mafic magmas; see for example the ex-
supposed observational evidence for magma mixing all relates to fea- periments of Johnston and Wyllie (1988). This is because the crustal
tures that present themselves at emplacement levels, so most of the ar- melts, with their cargos of entrained crystals, separate and exit from
guments involve support for or refutation of shallow-level magma their sources on time scales too short for such diffusive and reactive pro-
mixing and mingling. The conclusion is that mingling occurs but there cesses to take major effect. The likely low degree of direct surface-area
is scant evidence for true mixing (homogenisation) on any large scale. contact between the mantle- and crust-derived magmas makes mixing
The most persuasive arguments against the pervasiveness of magma interactions all the more unlikely at source depths.
mixing at deep (near-source) levels, and therefore against the model of There is no doubt that compositionally dissimilar magmas do come
DCH or MASH zones are those based on the chemistry (see earlier). into contact with each other and that they mingle physically and inter-
There are more plausible explanations for the observed chemical fea- act chemically, at least on local scales. However, there does seem to be
tures in granitic series that fit better with what we now know about in- serious doubt that wholesale magma mixing plays a significant role in
cremental and cyclic assembly of plutons and volcanic complexes. The the production of the compositional variations shown by granitic
following conclusions from Clemens et al. (2011a,) and Clemens and magmas. Having written this about the often-envisaged mixing of inter-
Stevens (2012) are critical in this regard. 1. Unevolved radiogenic iso- mediate to mafic and felsic magmas, it is worth considering the issue of
tope ratios in granitic rocks do not demand the direct chemical mixing between felsic magma batches in a growing plutonic mass or

Fig. 8. An illustration of the near-source processes involved in magma segregation from a metagreywacke source undergoing biotite fluid-absent partial melting during granulite-facies
metamorphism. The multivariant reaction involved is Al-rich Bt + Pl1 + Qz = Grt + Ilm + Pl2 + melt. Slow chemical diffusion in plagioclase prevents its equilibration during the
melting reaction, so that this feldspar reacts by dissolution, with precipitation of a more calcic, peritectic plagioclase. For panels (a), (b) and (c) black represents melt. For panels (d),
(e) and (f) black represents magma consisting of melt plus entrained peritectic garnet, ilmenite and plagioclase. At the scale of panels (d), (e) and (f), crystals of entrained peritectic
minerals are too small to portray. Pale blue represents reactant plagioclase throughout, while white represents quartz and brown biotite. Red represents peritectic garnet and the
darker blue represents peritectic plagioclase. (a) The texture of the metagreywacke close to the fluid-absent biotite solidus (~800 to 850 °C) at middle to lower crustal pressures. The
rock would also contain an accessory mineral assemblage of ilmenite, zircon and monazite, which are not illustrated. (b) With a slight temperature increase from the condition
depicted in (a), the rock has undergone a small amount of incongruent partial melting to produce about 2 vol.% melt plus a small amount of peritectic garnet, plagioclase and ilmenite
(which is not illustrated in any of the diagrams but should be considered to be present at all stages from panel (b) onward). The peritectic assemblage is present both within the melt
films and adjacent to them. (c) With further minor temperature increase above the fluid-absent solidus (~15 to 25 °C) the melt volume has grown to approximately 5 vol.%, with a
corresponding increase in the modal proportion of the co-existing peritectic minerals. At this stage the melt forms an interconnected film as well as small pockets with volumes of 20
to 50 mm3. (d) This image represents the same anatectic state as (c) but sufficiently zoomed out that a meta-arkose layer is visible below the fertile, mica-rich layer undergoing
anatexis. Extension in the direction of the foliation has resulted in the initiation of a fracture in the more competent meta-arkose layer. (e) Further extension has widened the fracture
and magma has flowed into the low-pressure region. Such magma accumulations would typically have volumes of 100 to 500 cm3, and would connect directly with shear zones (e.g.
in panel (f)) along which magma would flow out of the immediate source rock. Note that in (e) deformation has assisted drainage of the small amount of magma out of the source
and focussed it into structures that facilitate rapid upward migration. Textures observed in anatectic granulites preserve only partial, ‘fossilised’ evidence of this process. The long,
high-temperature incubation of granulites, after melting and melt withdrawal, as well as substantial mineral textural and chemical modification during decompression and cooling,
destroy the small-scale evidence. Following the stages portrayed here, larger, self-propagating, buoyancy-driven ‘hydro’ fractures convey the magmas rapidly and efficiently to upper-
crustal levels (see Fig. 9). Thus, during effective episodes of crustal differentiation, the deep crust never contains large volumes of melt because even low-proportion melts escape their
sources without significant accumulation.
 J.D. Clemens, G. Stevens / Earth-Science Reviews 160 (2016) 333–349 345

volcanic plumbing system. As has long been hypothesised, mixing texturally and mineralogically homogeneous granitic bodies (see e.g.
should most readily occur in cases where there are only small composi- Helps et al., 2003; Clemens et al., 2010). We interpret these heterogene-
tional, and therefore rheological, contrasts between the magmas in- ities as having been generated mainly through the rapid and efficient
volved (Sparks and Marshall, 1986). Nevertheless, the mixing process separation of successive low fractions of partial melt from the heteroge-
(mechanical stirring and chemical diffusion) is hampered by the rela- neous sources of the magmas during progressive heating and fluid-ab-
tively high viscosities and generally low diffusivities of non-alkali- sent hydrate breakdown.
metal components in granitic magmas (e.g. Sparks and Marshall,
1986). Thus, although there must be some mixing between the different 7. Conclusions
magma batches that are likely to exist in granitic systems, even this pro-
cess seems to be rather less than optimal in its efficiency, as is apparent In granitic systems, large, instantaneously emplaced magma cham-
from the preservation of isotopic heterogeneities in what appear to be bers form neither at shallow emplacement levels nor in near-source

(a) (b)

5mm 5mm

(c) (d)

5mm 15mm

(e) (f)

15mm 300mm
346 J.D. Clemens, G. Stevens / Earth-Science Reviews 160 (2016) 333–349

ground surface

1 km

(1)

(2)

(3)

(4)

Fig. 9. This model sketch shows our conception of the formation of a typical S-type granitic pluton emplaced in the shallow crust. The diagram represents a single point in time that is part
of the way through the assembly of the pluton. A thermal front is ascending through a deep-crustal terrane that is simultaneously undergoing deformation and high-grade regional
metamorphism. The black represents under- and intra-plated mafic magmas that have introduced substantial additional heat to the lower crust. The green represents metapelites and
the brown metagreywackes. Both rock types are compositionally layered and internally heterogeneous. The grey surrounding and between the metasediments represents tonalitic
gneisses (Hbl-Qz-Pl-Kfs rocks) with which metasediment-derived magmas would be grossly out of chemical equilibrium. White represents whatever other rock types may be present
higher in the column. Four model melting reactions are possible here, and are labelled as isotherms: (1) Qz + Pl +/− Kfs + H2O = melt, (2) Ms + Qz + Pl = Sil + melt (blue lines),
(3) Bt + Sil + Qz + Pl = Grt + melt (purple) and (4) Bt + Qz + Pl = Grt + Opx + melt (red). The numbered, dashed lines indicate the positions where each reaction becomes
possible, in appropriate bulk-rock compositions, with increasing depth. Fluid-present reaction (1) produces an insignificant melt volume due to limited supply of a free fluid phase;
the other reactions have all produced melt from some portion of the metasedimentary rock package. The tonalitic gneiss remains below its solidus. Each of the three fluid-absent
reactions has produced magmas (derived from the metapelite and metagreywacke) that have ascended in self-organised, focusing magma-fractures and have added volume to the
growing pluton. The wriggly grey lines in the lower part of the diagram are there to suggest the possible role of melt localisation in shear zones, at deep-crustal levels. Magmas are
extracted and delivered to the emplacement site directly from the source layers within which they are generated. Thus, they do not remain in contact with the infertile parts of the
source (or indeed the overlying crustal rocks) for long enough to permit significant chemical interaction with these rocks. Magma migration within the source is largely parallel to
favourably aligned anisotropies and toward points of upward magma extraction; due to scale limitations this is not shown. Magmas ascend in discrete, narrow dykes (impossible to
draw to scale), as individual batches that assemble the pluton from below and reuse the same paths as previous magma batches. This process produces a pre-heated magma plumbing
system through which the magmas rise in solitary waves, making it unlikely that individual batches mix significantly during ascent. Initial magma batches all freeze at the
emplacement site but, as this environment heats up, these early, thin sheets may partially re-melt and may also mechanically founder into subsequent magma batches, producing
some kinds of microgranular enclaves. The diagram illustrates some sills that, thus far, have survived along the pluton margins. Magma batches that arrive after a persistent magmatic
state has been achieved in the pluton will mingle with other batches on scales that depend on local thermo-mechanical conditions. However, these pulses will never mix to produce
compositionally homogenous hybrid magma. Thus, the chemical variations within the pluton largely reflect the chemical variability of the magmas produced within the source rocks.
Slow, pervasive magma percolation through the supersolidus crust, and the presence of large, partially molten magma chambers in the middle to deep crust are both specifically
excluded, for reasons detailed in the text. However, arriving magma pulses may intrude the crystal-melt mush in the pluton. Such pulses will form sills that may then be disrupted by
magma flow, largely through the deformation responsible for pluton inflation. Since there is no magma storage at depth, little or no fractionation or assimilation can occur during
ascent, and these processes are also minimal at the emplacement site, except for the late formation of highly fractionated aplitic to pegmatitic liquids extracted from the crystallising
magmas, mainly by filter pressing, and typically concentrated near the pluton roof.
 J.D. Clemens, G. Stevens / Earth-Science Reviews 160 (2016) 333–349 347

holding chambers. Crystals formed at near-source depths are simply the References
products of dissolution and re-precipitation of entrained peritectic
grains. The episodicity of magma emplacement is mirrored and perhaps Annen, C., Blundy, J.D., Sparks, R.S.J., 2006. The genesis of intermediate and silicic magmas
in deep crustal hot zones. J. Petrol. 47 (3), 505–539. http://dx.doi.org/10.1093/
governed by pulsed magma withdrawal from the source rocks, which petrology/egi084.
seems to occur at quite low melt volumes and which limits ultimate Arzi, A.A., 1978. Critical phenomena in the rheology of partially melted rocks.
source fertility, due to ongoing removal of the fluxing melt phase. The Tectonophysics 44, 173–184.
Bachmann, O., Bergantz, G.W., 2004. On the origin of crystal-poor rhyolites extracted from
main control on melt or magma withdrawal from a partially molten batholithic crystal mushes. J. Petrol. 45 (8), 1565–1582. http://dx.doi.org/10.1093/
crustal source is attainment of the melt connectivity threshold, which, petrology/egh019.
according to previous experimental results, happens at melt volumes Bagdonas, D.A., Frost, C.D., Fanning, C.M., 2016. The origin of extensive Neoarchean high-
silica batholiths and the nature of intrusive complements to silicic ignimbrites: in-
below 7%. The rapid withdrawal from the source and rapid ascent sights from the Wyoming batholith, U.S.A. Am. Mineral. 101, 1332–1347. http://dx.
rates of granitic magmas promote disequilibrium chemical behaviour doi.org/10.2138/am-2016-5512.
in the system, particularly for some trace elements, allowing preserva- Barbarin, B., 1990. Plagioclase xenocrysts and mafic magmatic enclaves in some granit-
oids of the Sierra Nevada Batholith, California. J. Geophys. Res. Solid Earth 95 (B11),
tion of some metastable accessory mineral grains and delivery of con- 17747–17756. http://dx.doi.org/10.1029/JB095iB11p17747.
trasting magma pulses to the emplacement sites. Fig. 8 presents a set Barnes, C.G., Prestvik, T., Barnes, M., Allen, C.M., 2003. Geology of a magma transfer zone:
of illustrations of various stages in the proposed near-source progres- the Hortavær Igneous Complex, north-central Norway. Nor. Geol. Tidsskr. 83 (3),
187–208.
sion from melt formation to magma extraction during granulite-facies
Bea, F., 1996. Controls on the trace element composition of crustal melts. Trans. R. Soc.
metamorphism of a model metasedimentary sequence. Edinb. Earth Sci. 87, 33–41.
There is scant geological evidence for anything resembling DCH or Bea, F., Montero, P., Gonzalez-Lodeiro, F., Talavera, C., 2007. Zircon inheritance reveals ex-
MASH zones. Indeed, our analysis of the processes by which magmas ceptionally fast crustal magma generation processes in central Iberia during the
Cambro-Ordovician. J. Petrol. 48, 2327–2339.
separate from their sources suggests that, for granitic systems, it is Beard, J.S., 2008. Crystal-melt separation and the development of isotopic heterogeneities
most unlikely that such zones exist. The observations used to infer the in hybrid magmas. J. Petrol. 49 (5), 1027–1041. http://dx.doi.org/10.1093/petrology/
physicochemical processes that could occur in these zones (mixing egn015.
Beard, J.S., Ragland, P.C., Crawford, M.L., 2005. Reactive bulk assimilation: a model for
with mantle-derived mafic magmas, assimilation and fractional crust-mantle mixing in silicic magmas. Geology 33 (8), 681. http://dx.doi.org/10.
crystallisation) all have alternative explanations that are more in tune 1130/g21470.1.
with what we know about the growth of plutons and silicic volcanic ed- Belcher, R.W., Kisters, A., 2006. Progressive adjustments of ascent and emplacement con-
trols during incremental construction of the 3.1 Ga Heerenveen batholith, South Afri-
ifices, and the origins of chemical variations therein. Indeed, one process ca. J. Struct. Geol. 28, 1406–1421.
(homogenisation) does not even seem to occur. The variations in real Bergantz, G.W., Walker, B.A., Otamendi, J., Dulcea, M.N., 2013. Petrogenetic processes in a
granitic systems are generally source-derived and commonly modified deep arc section: a MASH zone revealed. American Geophysical Union, Fall Meeting
abstract #V12B-05.
only by local fractionation and magma mingling at emplacement levels.
Bons, P.D., Arnold, J., Elburg, M.A., Kalda, J., Soesoo, A., van Milligen, B.P., 2004. Melt extrac-
Furthermore, during granite genesis, the general absence of large crys- tion and accumulation from partially molten rocks. Lithos 78 (1–2), 25–42. http://dx.
tal-melt mush zones in the crust must mean that melt-lubricated tec- doi.org/10.1016/j.lithos.2004.04.041.
Braun, I., Kriegsman, L.M., 2001. Partial melting in crustal xenoliths and anatectic
tonics is not a general feature of the crust. Such behaviour should be
migmatites: a comparison. Phys. Chem. Earth 26 (4–5), 261–266.
limited to unusual circumstances, such as magma emplacement into Brown, M., 1994. The generation, segregation, ascent and emplacement of granite
an active shear zone. magma: the migmatite-to-crustally-derived granite connection in thickened orogens.
As a consequence of the pulsed delivery of magma to growing gra- Earth Sci. Rev. 36, 83–130.
Brown, M., 2001. Crustal melting and granite magmatism: key issues. Phys. Chem. Earth
nitic plutons, these bodies inherit source-derived chemical and isotopic Solid Earth Geod. 26 (4–5), 201–212 PII: $1464-1895(01)00047-3.
heterogeneities that cannot be erased because there is no large magma Brown, M., 2007. Crustal melting and melt extraction, ascent and emplacement in
chamber in which such homogenisation could occur. Thus, traditional orogens: mechanisms and consequences. J. Geol. Soc. Lond. 164, 709–730.
Brown, M., 2013. Granite: from genesis to emplacement. Geol. Soc. Am. Bull. 125,
concepts of magma segregation, ponding, mixing, etc. retard our under- 1079–1113.
standing of the compositions of granitic magmas, the structures of felsic Brown, M., Averkin, Y.A., McLellan, E.L., Sawyer, E.W., 1995. Melt segregation in
plutons and the genesis of the magmas that formed them. There is a de- migmatites. J. Geophys. Res. Solid Earth 100, 15655–15679.
Cameron, K.L., 1984. Bishop Tuff revisited: new rare earth element data consistent with
veloping paradigm of pulsed and sometimes protracted growth of plu- crystal fractionation. Science 224, 1338–1340. http://dx.doi.org/10.1126/science.
tons, and the general absence of large, liquid-filled magma chambers 224.4655.1338.
near emplacement levels. We advocate extension of this new view to Castro, A., Moreno-Ventas, I., de la Rosa, J.D., 1991. H-type (hybrid) granitoids: a proposed
revision of the granite-type classification and nomenclature. Earth Sci. Rev. 31,
the deeper reaches of granitic magma systems, with magma withdrawal 237–253.
from source rocks in small batches, without the formation of any large Chappell, B.W., 1996. Magma mixing and the production of compositional variation with-
segregated masses at depth, and without formation of MASH or deep in granite suites: evidence from the granites of Southeastern Australia. J. Petrol. 37,
449–470.
crustal hot zones. Fig. 9 summarises our conception of a typical granitic
Chappell, B.W., White, A.J.R., 2001. Two contrasting granite types: 25 years later. Aust.
magma system, from source to emplacement depths. Magmas must es- J. Earth Sci. 48, 489–499.
cape from their residues, but segregation, as commonly conceived of, is Cherniak, D.J., Watson, E.B., 1992. A study of strontium diffusion in K-feldspar, Na-K feld-
probably rare in nature. spar and anorthite using Rutherford backscattering spectroscopy. Earth Planet. Sci.
Lett. 113 (3), 411–425. http://dx.doi.org/10.1016/0012-821X(92)90142-I.
Clemens, J., 2003. S-type granitic magmas—petrogenetic issues, models and evidence.
Earth Sci. Rev. 61 (1–2), 1–18. http://dx.doi.org/10.1016/s0012-8252(02)00107-1.
Acknowledgements Clemens, J.D., 1984. Water contents of intermediate to silicic magmas. Lithos 17, 273–287.
Clemens, J.D., 1988. Volume and composition relationships between granites and their
lower crustal source regions - an example from Central Victoria, Australia. Aust.
JC acknowledges financial support through the NRF (South J. Earth Sci. 35 (4), 445–449.
Africa) incentive funds for rated researchers, as well as the Clemens, J.D., 1998. Observations on the origins and ascent mechanisms of granitic
magmas. J. Geol. Soc. Lond. 155, 843–851.
Department of Earth Sciences at Stellenbosch University for ensuring Clemens, J.D., 2006. Melting of the continental crust: fluid regimes, melting reactions and
that all his teaching is concentrated into the first semester of the source-rock fertility. In: Brown, M., Rushmer, T. (Eds.), Evolution and Differentiation
academic year. GS acknowledges financial support through the NRF of the Continental Crust. Cambridge University Press, pp. 297–331.
Clemens, J.D., 2014. Element concentrations in granitic magmas: ghosts of textures past?
SARChI initiative. Prof. Alex Kisters (Stellenbosch) is thanked for J. Geol. Soc. Lond. 171, 13–19. http://dx.doi.org/10.1144/jgs2013-008.
many enlightening discussions and for providing the photograph Clemens, J.D., 2015. Forum comment: magmatic life at low Reynolds number. Geology
used here in Fig. 4. The positive and helpful comments and http://dx.doi.org/10.1130/G36512C.1.
Clemens, J.D., Droop, G.T.R., 1998. Fluids, P-T paths and the fates of anatectic melts in the
suggestions of two anonymous reviewers of an earlier draft of the
Earth's crust. Lithos 44, 21–36.
manuscript are acknowledged with thanks, as are detailed reviews Clemens, J.D., Mawer, C.K., 1992. Granitic magma transport by fracture propagation.
by Valentin R. Troll and Jean-Louis Vigneresse. Tectonophysics 204, 339–360.
348 J.D. Clemens, G. Stevens / Earth-Science Reviews 160 (2016) 333–349

Clemens, J.D., Phillips, G.N., 2014. Inferring a deep crustal source terrane from a high-level Harley, S.L., 1998. Ultrahigh temperature granulite metamorphism (1050 °C, 12 kbar) and
granitic pluton: the Strathbogie Batholith, Australia. Contrib. Mineral. Petrol. 168, ar- decompression in garnet (Mg70)–orthopyroxene–sillimanite gneisses from the
ticle id 1070. http://dx.doi.org/10.1007/s00410-014-1070-y. Rauer Group, East Antarctica. J. Metamorph. Geol. 16, 541–562.
Clemens, J.D., Stevens, G., 2012. What controls chemical variation in granitic magmas? Helps, P.A., Clemens, J.D., Petford, N., 2003. Source-related chemical and isotopic hetero-
Lithos 134-135, 317–329. http://dx.doi.org/10.1016/j.lithos.2012.01.001. geneities in granitoids. Geophys. Res. Abstr. 5, 02661.
Clemens, J.D., Wall, V.J., 1981. Origin and crystallization of some peraluminous (S-type) Hildreth, W., 1979. The Bishop Tuff: evidence for the origin of compositional zonation in
granitic magmas. Can. Mineral. 19, 111–131. silicic magma chambers. In: Chapin, C.E., Elston, W.E. (Eds.), Ash-flow Tuffs. The Geo-
Clemens, J.D., Birch, W.D., Dudley, R.J., 2011b. S-type ignimbrites with polybaric logical Society of America, Boulder, Colorado, pp. 43–76.
crystallisation histories: the Tolmie Igneous Complex, Central Victoria, Australia Hildreth, W., Moorbath, S., 1988. Crustal contributions to arc magmatism in the Andes of
(and erratum). Contrib. Mineral. Petrol. 162(6), 1315–1337, doi:http://dx.doi.org/ Central Chile. Contrib. Mineral. Petrol. 98, 455–489.
10.1007/s00410-011-0652-1 and doi:http://dx.doi.org/10.1007/s00410-011-0666-8 Hildreth, W., Wilson, C.J.N., 2007. Compositional zoning of the Bishop Tuff. J. Petrol. 48,
(and 1339). 951–999.
Clemens, J.D., Helps, P.A., Stevens, G., 2010. Chemical structure in granitic magmas – a sig- Holtzman, B.K., Groebner, N.J., Zimmerman, N.E., Ginsberg, S.B., Kohlstedt, D.l., 2003.
nal from the source? Earth Environ. Sci. Trans. R Soc. Edinb. 100 (1–2), 159–172. Stress-driven melt segregation in partially molten rocks. G3 4 (5), 8607. http://dx.
http://dx.doi.org/10.1017/s1755691009016053. doi.org/10.1029/2001gc000258.
Clemens, J.D., Maas, R., Waight, T.E., Kunneke, L.K., 2016. Genesis of felsic plutonic Holyoke III, C.W., Rushmer, T., 2001. An experimental study of grain scale melt segrega-
magmas and their igneous enclaves: the Cobaw batholith of SE Australia. J. Geol. 24 tion mechanisms in two common crustal rock types. J. Metamorph. Geol. 20,
(3), 293–311. http://dx.doi.org/10.1086/685509. 493–512.
Clemens, J.D., Stevens, G., Farina, F., 2011a. The enigmatic sources of I-type granites: the Hopson, C.A., Mattinson, J.M., 1994. Chelan Migmatite Complex, Washington: field evi-
peritectic connexion. Lithos 126, 174–181. dence for mafic magmatism, crustal anatexis, mixing and protodiapiric emplacement.
Coleman, D.S., Gray, W., Glazner, A.F., 2004. Rethinking the emplacement and evolution of In: Swanson, D.A., Haugerud, R.A. (Eds.), Geologic Field Trips in the Pacific Northwest.
zoned plutons: geochronologic evidence for incremental assembly of the Tuolumne Geological Society of America Annual Meeting, Seattle.
intrusive suite, California. Geology 32, 433–436. Johnston, A.D., Wyllie, P.J., 1988. Interaction of granitic and basic magmas: experimental
Cristofides, G., Perugini, D., Koroneo, A., Soldatos, T., Poli, G., Eleftheriadis, G., Del Moro, A., observations on contamination processes at 10 kbar with H2O. Contrib. Mineral. Pet-
Neiva, A.M., 2007. Interplay between geochemistry and magma dynamics during rol. 98, 352–362.
magma interaction: An example from the Sithonia Plutonic Complex (NE Greece). Jurewicz, S.R., Watson, E.B., 1985. The distribution of partial melt in a granitic system: the
Lithos 95, 243–266. application of liquid phase sintering theory. Geochim. Cosmochim. Acta 49,
D'Lemos, R.S., 1996. Mixing between granitic and dioritic crystal mushes, Guernsey, Chan- 1109–1121.
nel Islands, UK. Lithos 38, 233–257. Keay, S., Collins, W.J., McCulloch, M.T., 1997. A three-component Sr-Nd isotopic mixing
Davidson, J.P., Font, L., Charlier, B.L.A., Tepley, F.J., 2006. Mineral-scale Sr isotope variation model for granitoid genesis, Lachlan fold belt, eastern Australia. Geology 25 (4),
in plutonic rocks - a tool for unravelling the evolution of magma systems. Tranrs. R. 307. http://dx.doi.org/10.1130/0091-7613(1997)025b0307:atcsniN2.3.co;2.
Soc. Edinb. Earth Sci. 97, 357–367. Kelemen, B., Whitehead, J.A., Aharonov, E., Jordahl, K.A., 1995. Experiments on flow focus-
Dell'Angelo, L.N., Tullis, J., 1988. Experimental deformation of partially melted granitic ag- ing in soluble porous media, with applications to melt extraction from the mantle.
gregates. J. Metamorph. Geol. 6, 495–515. J. Geophys. Res. Solid Earth 100, 475–496.
DePaolo, D.J., 1981. Trace element and isotopic effects of combined wallrock assimilation Kisters, A.F.M., Ward, R.A., Anthonissen, C.J., Vietze, M.E., 2009. Melt segregation and far-
and fractional crystallization. Earth Planet. Sci. Lett. 53, 189–202. field melt transfer in the mid-crust. J. Geol. Soc. Lond. 166, 905–918.
DePaolo, D.J., Wasserburg, G.J., 1979. Petrogenetic mixing models and Nd-Sr isotopic pat- Knesel, K.M., Davidson, J.P., 2002. Insights into collisional magmatism from isotopic fin-
terns. Geochim. Cosmochim. Acta 43 (4), 615–627. gerprints of melting reactions. Science 296 (5576), 2206–2208. http://dx.doi.org/10.
DePaolo, D.J., Perry, F.V., Baldridge, W.S., 1992. Crustal versus mantle sources of granitic 1126/science.1070622.
magmas: a two-parameter model based on Nd isotopic studies. Trans. R. Soc. Edinb. Laporte, D., 1994. Wetting behaviour of crustal melts during anatexis: the distribution of
Earth Sci. 83, 439–446. hydrous silicic melts in polycrystalline aggregates of quartz. Contrib. Mineral. Petrol.
Duffield, W.A., Ruiz, J., 1998. A model that helps explain Sr-isotope disequilibrium be- 116, 486–499.
tween feldspar phenocrysts and melt in large-volume silicic magma systems. Laporte, D., Watson, E.B., 1995. Experimental and theoretical constraints on melt distribu-
J. Volcanol. Geotherm. Res. 87, 7–13 PII: S0377- 0273_98.00071-7. tion in crustal sources. Chem. Geol. 124, 161–184.
Durocher, J., 1857. Essai de pétrologie comparée, ou recherches sur la composition Lassalle, S., Indares, A., 2014. Anatectic record and contrasting P-T paths of aluminous
chimique et minéralogique des roches ignées, sur les phénomènes de leur émission gneisses from the central Grenville Province. J. Metamorph. Geol. 32, 627–646.
et sur leur classification. Ann. Min. 11, 217–259. Lipman, P.W., Bachmann, O., 2015. Ignimbrites to batholiths: integrating perspectives
Eichelberger, J.C., Gooley, R., 1977. Evolution of silicic magma chambers and their rela- from geological, geophysical, and geochronological data. Geosphere 11 (3),
tionships to basaltic volcanism. In: Heacock, J.G. (Ed.), The Earth's Crust: AGU Mono- 705–743. http://dx.doi.org/10.1130/GES01091.1.
graph 20 American Geophysical Union, Washington, DC, pp. 57–78. Lundstrom, C.C., Glazner, A.F., 2016. Silicic magmatism and the volcanic–plutonic connec-
Farina, F., Stevens, G., 2011. Source controlled 87Sr/86Sr isotope variability in granitic tion. Elements 12 (2), 91–96. http://dx.doi.org/10.2113/gselements.12.2.91.
magmas: the inevitable consequence of mineral-scale isotopic disequilibrium in the Lupulescu, A., Watson, E.B., 1999. Low melt fraction connectivity of granitic and tonalitic
protolith. Lithos 122, 189–200. http://dx.doi.org/10.1016/j.lithos.2011.01.001. melts in a mafic crustal rock at 800°C and 1 GPa. Contrib. Mineral. Petrol. 134 (2–3),
Farina, F., Dini, A., Rocchi, S., Stevens, G., 2014a. Extreme mineral-scale Sr isotopic hetero- 202–216. http://dx.doi.org/10.1007/s004100050479.
geneity in granites by disequilibrium melting of the crust. Earth Planet. Sci. Lett. 399, McKenzie, D., 1984. The generation and compaction of partially molten rock. J. Petrol. 25,
103–115. 713–765.
Farina, F., Stevens, G., Gerdes, A., Frei, D., 2014b. Small-scale Hf isotopic variability in the McKenzie, D., 1985. The extraction of magma from the crust and mantle. Earth Planet. Sci.
Peninsula pluton (South Africa): the processes that control inheritance of source Hf- Lett. 74 (1), 81–91.
176/Hf-177 diversity in S-type granites. Contrib. Mineral. Petrol. 168, 1065. McLeod, C.L., Davidson, J.P., Nowell, G.M., de Silva, S.L., 2012. Disequilibrium melting dur-
Farina, F., Stevens, G., Villaros, A., 2012. Multi-batch, incremental assembly of a dynamic ing crustal anatexis and implications for modeling open magmatic systems. Geology
magma chamber: the case of the Peninsula pluton granite (Cape Granite Suite, 40 (5), 435–438. http://dx.doi.org/10.1130/G33000.1.
South Africa). Mineral. Petrol. 106, 193–216. Memeti, V., Paterson, S., Matzel, J., Mundil, R., Okaya, D., 2010. Magmatic lobes as “snap-
Fountain, J.C., Hodge, D.S., Shaw, R.P., 1989. Melt segregation in anatectic granites: a shots” of magma chamber growth and evolution in large, composite batholiths: an
thermomechanical model. J. Volcanol. Geotherm. Res. 39, 279–296. example from the Tuolumne intrusion, Sierra Nevada, California. Geol. Soc. Am.
Frazer, R.E., Coleman, D.S., Mills, R.D., 2014. Zircon U-Pb geochronology of the Mount Bull. 122, 1912–1931.
Givens Granodiorite: implications for the genesis of large volumes of eruptible Metcalfe, R.V., Smith, E.I., Walker, J.D., Reed, R.C., Gonzales, D.A., 1995. Isotopic disequilib-
magma. J. Geophys. Res. Solid Earth 119, 2907–2924. rium among commingled hybrid magmas: evidence for two-stage magma mixing –
Gagnevin, D., Daly, J.S., Poli, G., Morgan, D., 2005. Microchemical and Sr isotopic investiga- commingling processes in the Mt Perkins pluton, Arizona. J. Geol. 103, 509–527.
tion of zoned K-feldspar megacrysts: insights into the petrogenesis of a granitic sys- Michael, P.J., 1983. Chemical differentiation of the Bishop Tuff and other high-silica
tem and disequilibrium crystal growth. J. Petrol. 46, 1689–1724. magmas through differentiation processes. Geology 11, 31–34. http://dx.doi.org/10.
Glazner, A.F., 2014. Magmatic life at low Reynolds number. Geology 42, 935–938. http:// 1130/0091-7613(1983)11b31:CDOTBTN2.0.CO;2.
dx.doi.org/10.1130/G36078.1. Miles, A.J., Graham, C.M., Hawkesworth, C.J., Gillespie, M.R., Hinton, R.W., Edinburgh Ion
Glazner, A.F., Bartley, J.M., Coleman, W.G., Taylor, R.Z., 2004. Are plutons assembled over Microprobe Facility (EIMF), 2013. Evidence for distinct stages of magma history re-
millions of years by amalgamation from small magma chambers? GSA Today 14, corded by the compositions of accessory apatite and zircon. Contrib. Mineral. Petrol.
4–11. 166, 1–19. http://dx.doi.org/10.1007/s00410-013-0862-9.
Gualda, G.A.R., Ghiorso, M.S., Lemons, R.V., Carley, T.L., 2012. Rhyolite-MELTS: a modified Miller, C.F., Watson, E.B., Harrison, T.M., 1988. Perspectives on the source, segregation and
calibration of MELTS optimized for silica-rich, fluid-bearing magmatic systems. transport of granitoid magmas. Trans. R. Soc. Edinb. Earth Sci. 79, 135–156.
J. Petrol. 53 (5), 875–890. http://dx.doi.org/10.1093/petrology/egr080. Montel, J.M., Vielzeuf, D., 1997. Partial melting of metagreywackes. 2. Compositions of
Guernina, S., Sawyer, E.W., 2003. Large-scale melt-depletion in granulite terranes: an ex- minerals and melts. Contrib. Mineral. Petrol. 128, 176–196.
ample from the Archean Ashuanipi Subprovince of Quebec. J. Metamorph. Geol. 21, Nabelek, P.I., Bartlett, C.D., 2000. Fertility of metapelites and metagraywackes during
181–201. leucogranite generation: an example from the Black Hills, U.S.A. Geol. Soc. Am.
Hall, D., Kisters, A., 2012. The stabilization of self-organised leucogranite networks Spec. Pap. 350, 1–14.
—implications for melt segregation and far-field melt transfer in the continental Nabelek, P.I., Hofmeister, A.M., Whittington, A.G., 2012. The influence of temperature-de-
crust. Earth Planet. Sci. Lett. 355-356, 1–12. http://dx.doi.org/10.1016/j.epsl. pendent thermal diffusivity on the conductive cooling rates of plutons and tempera-
2012.08.033. ture-time paths in contact aureoles. Earth Planet. Sci. Lett. 317-318, 157–164. http://
Harker, A., 1904. The Tertiary igneous rocks of Skye. Mem. Geol. Surv. UK 481 pp. dx.doi.org/10.1016/j.epsl.2011.11.009.
 J.D. Clemens, G. Stevens / Earth-Science Reviews 160 (2016) 333–349 349

Nicoli, G., Stevens, G., Moyen, J.-F., Frei, D., 2015. Rapid evolution from sediment to Valentini, L., Perugini, D., Poli, G., 2007. The ‘small-world’ nature of fracture/conduit net-
anatectic granulite in an Archaean continental collision zone: the example of the works: possible implications for disequilibrium transport of magmas beneath mid-
Bandelierkop Formation metapelites, South Marginal Zone, Limpopo Belt, South Afri- ocean ridges. J. Volcanol. Geotherm. Res. 159, 355–365. http://dx.doi.org/10.1016/j.
ca. J. Metamorph. Geol. 33, 177–202. volgeores.2006.08.002.
Ortoleva, P.J., 1994. Geochemical Self-organization. Oxford University Press, Oxford, Vanderhaeghe, O., 2001. Melt segregation, pervasive melt migration and magma mobility
411 pp. in the continental crust: the structural record from pores to orogens. Phys. Chem.
Otamendi, J.E., Ducea, M.N., Tibaldi, A.M., Bergantz, G.W., de la Rosa, J.D., Vujovich, G.I., Earth 26 (4–5), 213–223.
2009. Generation of tonalitic and dioritic magmas by coupled partial melting of gab- Vernon, R.H., Clarke, G.L., Collins, W.J., 1990. Local, mid-crustal granulite facies metamor-
broic and metasedimentary rocks within the deep crust of the Famatinian magmatic phism and melting: an example in the Mount Stafford area, central Australia. In:
arc, Argentina. J. Petrol. 50 (5), 841–887. http://dx.doi.org/10.1093/petrology/ Ashworth, J.R., Browm, M. (Eds.), High-temperature Metamorphism and Crustal
egp022. Anatexis. Springer, pp. 272–319.
Pearce, J., 1996. Sources and settings of granitic rocks. Episodes 19 (4), 120–125. Vielzeuf, D., Clemens, J.D., Pin, C., Moinet, E., 1990. Granites, granulites and crustal differ-
Petford, N., Koenders, M.A., 1998. Self-organisation and fracture connectivity in rapidly entiation. In: Vielzeuf, D., Vidal, P. (Eds.), Granulites and Crustal DifferentiationNATO
heated continental crust. J. Struct. Geol. 20, 1425–1434. ASI Series. Kluwer Academic Publishers, Dordrecht, pp. 59–86.
Petford, N., Clemens, J.D., Vigneresse, J.-L., 1997. Application of information theory to the Vigneresse, J.-L., 2006. Granitic batholiths: from pervasive and continuous melting in the
formation of granitic rocks. In: Bouchez, J.-L. (Ed.), Granite: From Segregation of Melt lower crust to discontinuous and spaced plutonism in the upper crust. Trans. R. Soc.
to Emplacement Fabrics. Kluwer Academic Publishers, Dordrecht, pp. 3–10. Edinb. Earth Sci. 97 (4), 311–324. http://dx.doi.org/10.1017/S0263593300001474.
Petford, N., Kerr, R.C., Lister, J.R., 1993. Dike transport of granitic magmas. Geology 21, Vigneresse, J.-L., Barbey, P., Cuney, M., 1996. Rheological transitions during partial melting
843–845. and crystallization with application to felsic magma segregation and transfer.
Pichavant, M., Costa, F., Burgisser, A., Scaillet, B., Martel, C., Poussineau, S., 2007. Equilibra- J. Petrol. 37, 1579–1600.
tion scales in silicic to intermediate magmas — implications for experimental studies. Villaros, A., Stevens, G., Buick, I.S., 2009b. Tracking S-type granite from source to emplace-
J. Petrol. 48 (10), 1955–1972. http://dx.doi.org/10.1093/petrology/egm045. ment: clues from garnet in the Cape Granite Suite. Lithos 112, 217–225.
Rabinowicz, M., Vigneresse, J.-L., 2004. Melt segregation under compaction and shear Villaros, A., Stevens, G., Moyen, J.-F., Buick, I.S., 2009a. The trace element compositions of
channeling: application to granitic magma segregation in a continental crust. S-type granites: evidence for disequilibrium melting and accessory phase entrain-
J. Geophys. Res. Solid Earth 109, B04407. http://dx.doi.org/10.1029/2002JB002372. ment in the source. Contrib. Mineral. Petrol. 158, 543–561.
Renner, J., Evans, B.W., Hirth, G., 2000. On the rheologically critical melt fraction. Earth Villaros, A., Buick, I.S., Stevens, G., 2012. Isotopic variations in S-type granites: an inheri-
Planet. Sci. Lett. 181, 585–594. tance from a heterogeneous source? Contrib. Mineral. Petrol. 163, 243–257.
Richards, J.P., 2003. Tectono-magmatic precursors for porphyry Cu-(Mo-Au) deposit for- von Bargen, N., Waff, H.S., 1986. Permeabilities, interfacial areas and curvatures of partial-
mation. Econ. Geol. 98 (8), 1515–1533. http://dx.doi.org/10.2113/gsecongeo.98.8. ly molten systems: results of numerical computations of equilibrium microstructures.
1515. J. Geophys. Res. Solid Earth 91, 9261–9276.
Rosenberg, C.L., Handy, M.R., 2005. Experimental deformation of partially melted granite Waight, T.E., Maas, R., Nicholls, I.A., 2000. Fingerprinting feldspar phenocrysts using crys-
revisited: implications for the continental crust. J. Metamorph. Geol. 23, 19–28. tal isotopic composition stratigraphy: implications for crystal transfer and magma
http://dx.doi.org/10.1111/j.1525-1314.2005.00555.x. mingling in S-type granites. Contrib. Mineral. Petrol. 139, 227–239.
Rushmer, T., 1995. An experimental deformation study of partially molten amphibolite: Walker, B.A., Bergantz, G.W., Otamendi, J.E., Ducea, M.N., Cristofolini, E.A., 2015. A MASH
application to low-melt fraction segregation. J. Geophys. Res. Solid Earth 100, Zone revealed: the mafic complex of the Sierra Valle Fértil. J. Petrol. 56, 1863–1896.
15681–15695. Ward, R., Stevens, G., Kisters, A., 2008. Fluid and deformation induced partial melting and
Rushmer, T., 2001. Volume change during partial melting reactions: implications for melt melt volumes in low-temperature granulite-facies metasediments, Damara Belt, Na-
extraction, melt geochemistry and crustal rheology. Tectonophysics 342, 389–405. mibia. Lithos 105, 253–271. http://dx.doi.org/10.1016/j.lithos.2008.04.001.
Rutter, E.H., Neumann, D.H.K., 1995. Experimental deformation of partially molten West- Watson, E.B., 1996. Dissolution, growth and survival of zircons during crustal fusion: ki-
erly Granite under fluid-absent conditions, with implications for the extraction of netic principles, geological models and implications for isotopic inheritance. Geol.
granitic magmas. J. Geophys. Res. Solid Earth 100, 15697–15715. Soc. Am. Spec. Pap. 315, 43–56. http://dx.doi.org/10.1130/0-8137-2315-9.43.
Sawyer, E.W., Cesare, B., Brown, M., 2011. When the continental crust melts. Elements 7 Weertman, J., 1971. Theory of water-filled crevasses in glaciers applied to vertical magma
(4), 229–234. http://dx.doi.org/10.2113/gselements.7.4.229. transport beneath ocean ridges. J. Geophys. Res. Solid Earth 76, 1171–1183.
Sengupta, P., Sen, J., Dasgupta, S., Raith, M., Bhui, U.K., Ehl, J., 1999. Ultra-high temperature Weinberg, R.F., Hasalová, P., Ward, L., Fanning, C.M., 2013. Interaction between deforma-
metamorphism of metapelitic granulites from Kondapalle, Eastern Ghats Belt: impli- tion and magma extraction in migmatites: examples from Kangaroo Island, South
cations for the Indo-Antarctic correlation. J. Petrol. 40, 1065–1087. Australia. Geol. Soc. Am. Bull. 125 (7–8), 1282–1300. http://dx.doi.org/10.1130/
Siebel, W., Reitter, E., Wenzel, T., Blaha, U., 2005. Sr isotope systematics of K-feldspars in B30781.1.
plutonic rocks revealed by the Rb–Sr microdrilling technique. Chem. Geol. 222 (3–4), Westraat, J., Kisters, A.F.M., Poujol, M., Stevens, G., 2005. Transcurrent shearing, granite
183–199. http://dx.doi.org/10.1016/j.chemgeo.2005.06.012. sheeting and the incremental construction of the tabular, 3.1 Ga Mpuluzi batholith,
Simakin, A., Talbot, C., 2001. Transfer of melt between microscopic pores and macroscopic Barberton granite-greenstone terrain, South Africa. J. Geol. Soc. Lond. 162, 373–388.
veins in migmatites. Phys. Chem. Earth 26 (4–5), 363–367. Whalen, J.B., Currie, K.L., 1984. The Topsails igneous terrane, Western Newfoundland: ev-
Solano, J.M., Jackson, M.D., Sparks, R.S.J., Blundy, J.D., Annen, C., 2012. Melt segregation in idence for magma mixing. Contrib. Mineral. Petrol. 87, 319–327.
deep crustal hot zones: a mechanism for chemical differentiation, crustal assimilation White, A.J.R., Clemens, J.D., Holloway, J.R., Silver, L.T., Chappell, B.W., Wall, V.J., 1986a. S-
and the formation of evolved magmas. J. Petrol. 53 (10), 1999–2026. http://dx.doi. type granites and their probable absence in south-western North America. Geology
org/10.1093/petrology/egs041. 14, 115–118.
Sparks, R.S.J., Marshall, L.A., 1986. Thermal and mechanical constraints on mixing be- White, A.J.R., Clemens, J.D., Holloway, J.R., Silver, L.T., Chappell, B.W., Wall, V.J., 1986b.
tween mafic and silicic magmas. J. Volcanol. Geotherm. Res. 29, 99–124. Comment and reply on “S-type granites and their probable absence in south-western
Stevens, G., van Reenen, D.D., 1992. Partial melting and the origin of the metapelitic gran- North America”. Geology 14, 805–806.
ulites in the Southern Marginal Zone of the Limpopo Belt, South Africa. Precambrian White, A.J.R., Clemens, J.D., Holloway, J.R., Silver, L.T., Chappell, B.W., Wall, V.J., 1986c.
Res. 55, 303–319. Comment and reply on “S-type granites and their probable absence in south-western
Stevens, G., Clemens, J.D., Droop, G.T.R., 1997. Migmatites, granites and granulites: the North America”. Geology 14, 895.
data from experiments on “primitive” metasedimentary protoliths. Contrib. Mineral. White, R.W., Powell, R., 2002. Melt loss and the preservation of granulite facies mineral
Petrol. 128, 352–370. assemblages. J. Metamorph. Geol. 20, 621–632.
Stevens, G., Villaros, A., Moyen, J.-F., 2007. Selective peritectic garnet entrainment as the Wickham, S.M., 1987. The segregation and emplacement of granitic magmas. J. Geol. Soc.
origin of geochemical diversity in S-type granites. Geology 35 (1), 9–12. http://dx. Lond. 144, 281–297. http://dx.doi.org/10.1144/gsjgs.144.2.0281.
doi.org/10.1130/g22959a.1. Wolf, M.B., Wyllie, P.J., 1991. Dehydration-melting of solid amphibolite at 10 kbar: textur-
Taylor, J., Stevens, G., 2010. Selective entrainment of peritectic garnet into S-type granitic al development, liquid interconnectivity and applications to the segregation of
magmas: evidence from Archaean mid-crustal anatectites. Lithos 120, 277–292. magmas. Mineral. Petrol. 44, 151–179.
Taylor, J., Nicoli, G., Stevens, G., Frei, D., Moyen, J.-F., 2014. The processes that control Yakymchuk, C., Brown, M., 2014. Consequences of open-system melting in tectonics.
leucosome compositions in metasedimentary granulites: perspectives from the J. Geol. Soc. Lond. 117, 21–40.
Southern Marginal Zone migmatites, Limpopo Belt, South Africa. J. Metamorph.
Geol. 32, 713–742.
Taylor, J., Stevens, G., Buick, I.S., Lana, C., 2012. Successive midcrustal, high-grade meta-
morphic events provide insight into Mid-Archean mountain-building along the SE
margin of the proto-Kaapvaal craton. Geol. Soc. Am. Bull. 124, 1191–1211.