Arc crustal differentiation mechanisms

Earth and Planetary Science Letters 396 (2014) 267–277

Contents lists available at ScienceDirect

Earth and Planetary Science Letters www.elsevier.com/locate/epsl

Arc crustal differentiation mechanisms Oliver Jagoutz Department of Earth, Atmospheric and Planetary Sciences, MIT, 77 Massachusetts Avenue, Cambridge, MA 02139-4307, USA

a r t i c l e

i n f o

Article history: Received 15 July 2013 Received in revised form 25 March 2014 Accepted 28 March 2014 Available online 3 May 2014 Editor: T. Elliott Keywords: chemical differentiation continental crust arc crust

a b s t r a c t The detailed vertical compositional and thermal structure of the entire Kohistan arc section is constructed using ∼63 P / T constraints and 209 plutonic whole rock analyses. The aim is to better understand the nature of the chemical differentiation in the plutonic arc crust at different crustal levels. Results indicate that a distinct difference in character exists between the upper and lower plutonic arc crust. Plutonic rocks in the lower crust (25–30 km) have whole-rock chemical characteristics that indicate mineral accumulation and residual melt loss. In these cumulate rocks there is a systematic chemical stratification of decreasing incompatible trace elements with depth (e.g. Rb, K2 O). In contrast, the plutonic rocks of the upper crust represent mainly frozen liquids and show no systematic relationship between chemical composition and intrusion depth. These results clearly indicate that magmatic differentiation in the Kohistan arc dominantly occurred in the deeper arc crust (30 km) consistent with the lower crustal ‘hot zone model’. To understand the role of density barriers for melt stagnation in the crust, and thereby for the observed chemical stratification, the detailed density structure of the plutonic arc crust is compared to melt densities calculated for plutonic rocks with near liquid compositions. Results indicate that melts are consistently less dense than rocks, and there is no evidence for a “neutral buoyancy line” controlling melt emplacement. It is speculated that melt stagnation in the lower arc crust and subsequent chemical differentiation of a felsic upper and mafic lower continental crust is dominantly controlled by temperature and to a lesser extent to density or rheological barriers. © 2014 Elsevier B.V. All rights reserved.

1. Introduction The continental crust is thought to be chemically stratified into a felsic upper crust and mafic lower crust (Rudnick and Gao, 2003) based on the observed increase of the seismic velocity of the continental crust with depth (Christensen and Mooney, 1995). The chemical stratification of the crust results in a density stratification that is thought to play a crucial role for melt stagnation and differentiation either at (1) shallow crustal levels (Ryan, 1987), (2) at mid-crustal level (Glazner and Ussler III, 1988) or (3) at the crust– mantle interface (Herzberg et al., 1983). Chemical differentiation within the continental crust occurs over a wide range of time scales and different geodynamic settings. Based on trace element systematics it is widely accepted that new (post-Archean?) continental crust (or proto-continental crust) is formed in subduction zones (Rudnick, 1995). So an important question is if the observed chemical stratification is predominantly formed during arc processes, or whether secondary processes, such as arc–continent collision, are required? To understand the importance of chemical differentiation in arcs it is essential to constrain the effect of magmatic differentiation at different crustal levels. For years it was thought that chemical differentiation occurs predominantly in the shallow arc http://dx.doi.org/10.1016/j.epsl.2014.03.060 0012-821X/© 2014 Elsevier B.V. All rights reserved.

crust (Bateman and Chapell, 1979; Glazner, 1994; Halliday et al., 1989). Support for shallow crustal differentiation comes from geophysical evidence of sizeable melt reservoirs (e.g. magma chambers) in the upper crust (∼<20 km), as well as geochemical estimates from arc magmas that generally indicate pre-eruptive storage depths of <15 km (Blundy and Cashman, 2008). Magnetotelluric and seismological observations, as well as the presence of large calderas, also indicate that large-scale magma chambers must be present in the upper crust (e.g., Jellinek and DePaolo, 2003; Schilling and Partzsch, 2001; Zandt et al., 2003). Rheological and density considerations have suggested the presence of major barriers for melts in mid- to upper-crustal levels, leading to preferential trapping of melt in this part of the crust (e.g., Glazner and Ussler III, 1988). Experiments have shown that the observed intermediate metaluminous ASI < 1; ASI = molar (Al2 O3 /CaO + Na2 O + K2 O) composition of the upper crust can be formed by hydrous, shallow-level fractionation at 2–3 kbar (Blatter et al., 2013; Grove et al., 2003; Sisson and Grove, 1993). The low ASI content of the basaltic to andesitic melts is due to early saturation of plagioclase at shallow pressures, which inhibits further Al-enrichment of the derivative liquid during differentiation. One fundamental problem with the shallow crustal differentiation model is the paucity of

268

O. Jagoutz / Earth and Planetary Science Letters 396 (2014) 267–277

mafic–ultramafic cumulates in the upper crust that would be necessary to balance the observed granites to a mafic parental melt. As a solution to this apparent paradox, it has been proposed that dense mafic to ultramafic cumulates or restites might sink down into the lower crust as so-called ‘reverse diapirs’ (Glazner, 1994). As an alternative to this model, Annen et al. (2006) proposed that magmatic differentiation in arcs dominantly occurs in the lower arc crust in so-called ‘deep crustal hot zones’, essentially an extension of the MASH model of Hildreth and Moorbath (1988). This model is attractive as it can explain the general lack of mafic to ultramafic cumulates in the crust by foundering of the lower arc crust directly back into the upper mantle without requiring them to sink through the highly viscous continental crust (Arndt and Goldstein, 1989; Behn et al., 2007; Herzberg et al., 1983; Jagoutz and Behn, 2013; Jagoutz et al., 2011; Jagoutz and Schmidt, 2013; Jull and Kelemen, 2001; Kay and Mahlburg Kay, 1993). At high pressure and elevated water content amphibole and garnet can become important early fractionating phases (AlonsoPerez et al., 2009; Grove et al., 2003; Müntener et al., 2001; Prouteau and Scaillet, 2003), leading to significant silicaenrichment of the derivative liquids over a limited fractionation interval. However, at high P (>∼5–7 kbar) and/or high aH2 O , plagioclase generally fractionates after clinopyroxene and amphibole in basaltic to basaltic–andesitic compositions and will generally produce peraluminous derivative liquids (ASI > 1, Blatter et al., 2013), not metaluminous as is observed. While, the effect of late crystallization of plagioclase at high pressure may be offset by the increased incorporation of Al2 O3 into pyroxene with increasing pressure (Müntener et al., 2001), these experiments nevertheless imply that derivative liquids from deep level crustal differentiation can become peraluminous, conflicting with the scarcity of peraluminous intermediate compositions observed in arcs (Blatter et al., 2013). It is noteworthy that these high pressure experimental studies were conducted on primitive melts with high Al-content (∼17 wt%) as are found in the Cascades Arc (e.g., Mt. Shasta). The primitive melts from the Cascades, however, have significantly higher Al-content (15–17 wt%) compared to primitive arc melts from most other subduction zones globally that are as low as 12–13 wt% Al2 O3 (Jagoutz and Schmidt, 2013). If deep crustal differentiation is the predominant process to create the granitic compositions of the upper crust, then parental melts must be either relatively Al-poor (as is more globally observed), or additional Al-rich phases such as garnet or spinel must be important early fractionating phases to explain the overall low-Al content of the upper crust. The problem in testing the deep fractionation hypothesis is that exposed lower crustal sections are rare in general, and those that exist often have a complicated and protracted geological history (e.g., Harley, 1989). Additionally, very few lower crustal sections have a complementary upper crustal section making it difficult to correlate and integrate processes occurring at different crustal levels. The Kohistan arc is the only known intact, full arc section preserved in the geological record (Tahirkheli, 1979) and is representative of juvenile continental crust formed in arcs (Jagoutz and Schmidt, 2012). Reworking of the arc crust due to later India– Eurasia collision is minor (Bouilhol et al., 2013). Therefore, the Kohistan arc provides an ideal testing ground for these two conflicting end-member models of magma differentiation in arcs. The aim of this publication is to constrain, for the first time, the detailed chemical stratification of a complete arc section by integrating the chemical variation observed at different crustal levels. This approach necessitates a statistically significant dataset of whole rock compositions for each of which the original emplacement level should be known. Constraining the emplacement depth of plutonic rocks is often difficult due to the lack of appropriate

mineral assemblage for reliable pressure estimates; however, as will be outlined in detail below, the Kohistan arc exposes a continuous crustal arc section in which major discontinuities in exposure level are absent, and thus geostatistical methods are used to interpolate the detailed exposure surface of the entire Kohistan arc from the existing 67 published reliable pressure constraints. The results of these calculations are used to approximate the intrusion depth of 209 arc-related whole rock samples, the detailed chemical stratification of the arc is discussed in terms of a statistically significant set of whole rock compositions. MELTS and PerpleX calculations are used to constrain density of rocks and melts in respect to the preserved P –T gradient. The results clearly indicate that chemical differentiation occurred predominantly in the deep arc crust during the build up of the arc. The results presented here strongly support the idea of a deep crustal ‘hot zone’ (e.g. Annen et al., 2006). 2. Geologic setting The Kohistan arc, exposed in NE Pakistan, is the best-preserved complete arc section (Bard, 1983; Tahirkheli, 1979) in the geological record, with volcanic rocks and unmetamophosed sediments overlying a predominantly felsic plutonic upper crust in the northern exposures (Fig. 1). To the south, mafic and ultramafic plutons characterize the lower crust at the base of the arc section. The Cretaceous to Tertiary oceanic arc formed in the equatorial part of the Neotethyan ocean (Khan et al., 2009; Zaman and Torii, 1999). It is composed of three main complexes from north to south (Fig. 1): (1) the Gilgit Complex, comprising the mid- to upperlevel of the arc, including the Kohistan Batholith composed of variable granitoids and its volcano-sedimentary cover sequences (Jagoutz et al., 2013; Khan et al., 2009; Petterson and Treloar, 2004; Petterson and Windley, 1985, 1991); (2) the Chilas Complex mafic– ultramafic, rift related mid- to lower-crustal intrusions (Jagoutz et al., 2006, 2007; Khan et al., 1989, 1993); (3) and the Southern Plutonic Complex (SPC), a heterogeneous sequence dominated by ultramafic to mafic plutonics, constituting the deepest exposed arc crust (Burg et al., 2005, 2006; Dhuime et al., 2007, 2009; Garrido et al., 2006; Jagoutz et al., 2011). U–Pb zircon age dating indicates that the dominant magmatic activity recorded in the SPC (∼105–85 Ma) ended with the intrusion of the Chilas Complex at ∼85 Ma (Schaltegger et al., 2002). In contrast, the magmatic activity recorded in the Kohistan Batholith ranges from at least the Cretaceous (∼110 Ma) to the Miocene (Bouilhol et al., 2013). The Kohistan Arc is separated in the north from the former southern Eurasian margin (the Karakoram) by the Shyok suture zone (or “the Northern-”, or “Karakoram–Kohistan suture zone”) and in the south is separated from India by the Indus suture zone (Fig. 1). The collision of the arc with India is well constrained at ∼50 Ma (Rowley, 1996). While the formation age of the Shyok suture zone has been discussed for decades (e.g., Bard, 1983; Brookfield and Reynolds, 1981; Petterson and Windley, 1985), it was recently shown to postdate the formation of the Indus suture by ∼10 Ma (Bouilhol et al., 2013). This indicates that the Kohistan arc, until its collision with India at ∼50 Ma, formed as an intraoceanic arc (Burg, 2011). Accordingly, with the exception of the post-collisional leucogranites (see below), the igneous rocks of the Kohistan arc formed entirely in an oceanic island arc environment (Bouilhol et al., 2013). 2.1. Estimates on the emplacement pressure recorded in the Kohistan arc Emplacement pressures of various plutons throughout Kohistan arc have been determined quantitatively (Fig. 1) using existing

O. Jagoutz / Earth and Planetary Science Letters 396 (2014) 267–277

269

Fig. 1. Geological map of the Kohistan arc. Numbers indicate pressure in kbar constrained by Al-in-Hbl barometry (brown) or net transfer reactions involving garnet (green) or pyx–plag–qtz (black) (see Table 1 for complete dataset and references). The isobars (dashed red line, number in kbar) illustrate the exhumation level of the Kohistan constrained by 67 pressure estimates that have been interpolated at unconstrained regions using geostatistical modeling as described in the text. (For interpretation of the references to color in this figure legend, the reader is referred to the web version of this article.)

barometric estimates from the literature (Jagoutz et al., 2007, 2013; Ringuette et al., 1999; Yamamoto, 1993; Yamamoto and Yoshino, 1998; Yoshino and Okudaira, 2004; Yoshino et al., 1998) (Table 1). As the mineral assemblage in the plutonic rocks varies throughout the arc, three different barometers were used (Table 1). In locations where all three different barometers were applied to nearby samples, the different pressure estimates agree within uncertainty, indicating that the inferred pressure estimates are independent of the barometer chosen. The (re-)crystallization pressures indicate an overall emplacement pressure increase from north to south, from epithermal conditions at the volcano/plutonic contact near Yasin, to ∼15 kbar at the contact between the mafic and ultramafic cumulates of the SPC. The paleo-depth levels of exposure in the Kohistan arc are complex and correlated to some degree with the subdivision into the three main complexes (Gilgit, Chilas and Southern Plutonic Complex). Importantly, the highest pressures recorded in the Gilgit Complex overlap and agree well with the lowest pressure recorded in the petrologically related Southern Plutonic Complexes (Fig. 1). The emplacement pressures in the eastern part of the Gilgit Complex increase from NW to SE from <2 kbar to ∼8 kbar and are influenced by the presence of the Nanga Parbat syntaxis (Jagoutz et al., 2013). The exhumation of the Kohistan Batholith started at ∼50 Ma, associated with the India–Kohistan collision (van der Beek et al., 2009). Present day exposure levels are dominantly controlled by the presence of the Nanga Parbat syntaxis (Treloar et al., 1989) and reflect recent Neocene exhumation (e.g. Crowley et al., 2009). In the Southern Plutonic Complex, (re-) crystallization pressures estimates exist mainly along the Indus Valley with fewer constraints on the lateral pressure differences (Fig. 1). The available pressure estimates indicate a pronounced increase in pressure from north to south (from ∼8–15 kbar) consistent with the over-

all ∼30–40 degree tilts of the units to the north (Burg et al., 2005). The observed emplacement pressures are consistent with the exhumation of the SPC during block rotation associated by intra-arc splitting and extension along a south dipping listric fault during the emplacement of the Chilas Complex at ∼85 Ma (Burg et al., 2006). This interpretation, supported by Ar–Ar cooling ages (Treloar et al., 1989), is also consistent with the absence of any significant pressure gradients across the Chilas Complex. (The Chilas Complex records formation pressures of ∼7 kbar with no significant pressure differences across it, Jagoutz et al., 2007.) The highest pressures in the study area are recorded along the Indus Valley next to the Besham/Indus Syntaxis. Throughout the arc the mineralogically-constrained pressure estimates approximate the igneous emplacement depth of the rocks (Jagoutz et al., 2013). This agreement implies that the arc grew laterally by basaltic underplating, which is in accordance with thermal models (Yoshino and Okudaira, 2004). This interpretation, however, is only correct for those rock units that were emplaced before exhumation. The arc-related magmatic activity within the SPC generally predates its exhumation event at ∼80–85 Ma (Burg et al., 2006), and SPC rocks younger than ∼80 Ma are volumetrically insignificant granitic veins (Yamamoto et al., 2005) for which no whole rock analyses exist. However, some magmatic activity within the Gilgit Complex continued even after the India/Kohistan collision at ∼50 Ma (up until at least ∼30 Ma; Bouilhol et al., 2013), producing dominantly peraluminous (ASI > 1), leucogranites that differ significantly in their mineralogy and geochemistry from the metaluminous (ASI < 1) arc related granitoids (Bouilhol et al., 2013). Based on these observations all post-collisional granitoids found within the Gilgit Complex are excluded from the subsequent discussions in this paper. The general lack of major tectonic faults in the arc (with the exception of the faults in the Dir–Kalam area, Fig. 1), in accordance

270

O. Jagoutz / Earth and Planetary Science Letters 396 (2014) 267–277

Table 1 Pressures estimates from the Kohistan arc compiled from the literature compared to the results of the Kriging calculations through cross validation. Sample no.

N

E

Geobarometric estimates Pressure [kbar]

Kriging results

P

±a

Sourceb

Barometer used

Pressure [kbar]

±

2.6 2.7 2.6 3.2 3.0 3.6 3.4 3.8 4.0 3.9 4.0 3.6 4.4 3.6 3.8 5.3 5.3 5.1 5.5 6.1 5.4 5.9 6.0 5.8 5.8 6.3 6.1 7.1 7.1 6.9 8.0 8.1 8.1 8.8 10.3

1.8 1.8 1.8 1.7 1.7 1.7 1.7 1.6 1.8 1.7 1.8 1.6 1.8 1.6 1.6 1.7 1.7 1.9 1.8 1.8 1.9 1.7 1.8 1.9 1.8 1.8 1.8 1.8 1.8 1.8 1.7 1.7 1.8 1.8 1.6

−0.2 −0.1 0.1 −0.6 0.2 −0.3 0.1 −0.2 0.2 0.1 0.3 0.8 −0.2 1.2 1.0 −0.4 −0.4 0.7 0.0 −0.3 0.6 −0.1 −0.3 0.2 0.1 0.1 0.2 0.1 0.2 0.7 −0.1 −0.5 0.6 0.6 −0.6

PYa-04 PGU-14 PYa-03 PBa-03 PGUP-12 PGak-5.3 PBa-04 PGak-18.5 1 ST-02-17 PGak-26 1 BI-02-9 PGUP-1 1 BO-02-14 PGUP-2 PGak-13.1 1 K08-7 1 K08-8 1 MD-02-7 PGLT-19 PGLT-36 2 1A-88 1 PL-02-1 1 K08-10 3 VW-16 PGLT-106 PGLT-55.2 PGLT-37.2 PJa-06 1 C75/01 1 C03-65 PCh-07 PCh-10 PCh-12 PJa-01 Pegmatite

36◦ 19 29.6 36◦ 13 55.7 36◦ 19 29.6 36◦ 02 46.8 36◦ 07 04.1 36◦ 14 55.5 36◦ 01 57.4 36◦ 12 57.9 35◦ 28 18.1 36◦ 09 24.2 35◦ 36 34.5 36◦ 11 16.0 35◦ 23 46.6 36◦ 11 14.6 36◦ 13 06.2 35◦ 44 22.8 35◦ 43 55.4 35◦ 43 21.5 36◦ 02 49.6 35◦ 59 26.2 35◦ 33 01.0 35◦ 24 39.8 35◦ 36 37.3 35◦ 24 06.0 36◦ 08 09.1 35◦ 52 39.9 35◦ 59 26.2 35◦ 49 51.7 35◦ 20 37.5 35◦ 36 56.09 35◦ 31 00.2 35◦ 28 45.3 35◦ 24 48.6 35◦ 29 39.3 35◦ 08 47.8

73◦ 22 19.1 73◦ 19 13.6 73◦ 22 19.1 73◦ 23 30.8 73◦ 19 29.3 73◦ 41 17.6 73◦ 24 28.9 73◦ 43 05.3 72◦ 15 13.5 73◦ 44 08.2 72◦ 25 30.0 73◦ 26 02.2 72◦ 03 48.9 73◦ 26 04.7 73◦ 42 56.7 73◦ 23 42.8 73◦ 23 47.8 72◦ 39 27.6 74◦ 08 27.0 74◦ 19 34.6 71◦ 52 37.1 72◦ 35 57.9 73◦ 35 00.3 71◦ 45 20.0 74◦ 17 54.5 74◦ 20 22.3 74◦ 19 34.6 74◦ 32 18.6 74◦ 08 03.6 74◦ 03 25.8 74◦ 30 48.1 74◦ 30 36.4 74◦ 17 58.5 74◦ 36 04.4 73◦ 05 22.0

2.4 2.6 2.7 2.6 3.2 3.3 3.5 3.6 4.2 4.0 4.3 4.4 4.2 4.8 4.8 4.9 4.9 5.8 5.5 5.7 6.0 5.8 5.7 6.0 6.0 6.4 6.3 7.2 7.3 7.6 7.9 7.6 8.7 9.4 9.7

1.0 1.0 1.0 1.0 1.0 1.0 1.0 1.0 1.0 1.0 1.0 1.0 1.0 1.0 1.0 1.0 1.0 1.0 1.0 1.0 1.0 1.0 1.0 1.0 1.0 1.0 1.0 1.0 1.0 1.0 1.0 1.0 1.0 1.0 1.0

1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1

Al-in Al-in Al-in Al-in Al-in Al-in Al-in Al-in Al-in Al-in Al-in Al-in Al-in Al-in Al-in Al-in Al-in Al-in Al-in Al-in Al-in Al-in Al-in Al-in Al-in Al-in Al-in Al-in Al-in Al-in Al-in Al-in Al-in Al-in Al-in

LT3 LT8 D8 PL4 P8 P5 PU5 PD23-25

35◦ 28 41.4 35◦ 26 33.5 35◦ 18 56.1 35◦ 07 59.7 35◦ 07 13.5 35◦ 07 05.2 35◦ 07 03.9 35◦ 04 30.7

73◦ 15 29.2 73◦ 12 58.3 73◦ 11 44.9 73◦ 05 11.0 73◦ 02 30.5 73◦ 01 00.0 72◦ 59 51.5 72◦ 58 13.65

8.2 8.9 9.5 11.0 10.5 9.6 10.8 15.4

1.0 1.0 1.0 1.0 1.0 1.0 1.0 1.0

2 2 2 2 2 2 2 2

(Grt)-CPX-PL-QTZ (Grt)-CPX-PL-QTZ (Grt)-CPX-PL-QTZ (Grt)-CPX-PL-QTZ (Grt)-CPX-PL-QTZ (Grt)-CPX-PL-QTZ (Grt)-CPX-PL-QTZ (Grt)-CPX-PL-QTZ

8.3 8.6 9.1 10.4 10.8 11.0 11.3 14.6

1.6 1.7 1.7 1.6 1.6 1.6 1.6 1.6

−0.1 0.3 0.4 0.6 −0.3 −1.4 −0.5 0.8

KU66 KU82 KA04

35◦ 06 23.3 35◦ 05 09.8 35◦ 28 38.8

73◦ 00 03.2 72◦ 59 19.5 72◦ 58 05.3

11.1 15.2 7.0

1.0 1.0 1.0

3 3 3

Grt-CPX-PL-QTZ Grt-CPX-PL-QTZ Grt-CPX-PL-QTZ

12.0 13.4 7. 2

1.6 1.6 1.8

−1.0 1.8 −0.2

J07

35◦ 27 06.9

74◦ 09 10.8

7.0

1.0

4

Grt-CPX-PL-QTZ

6. 8

1.7

0.2

R99

35◦ 03 29.7

72◦ 57 13.8

16.3

1.0

5

Grt-CPX-PL-QTZ

15.2

1.6

1.1

A

35◦ 31 38.5

35◦ 24 49.6 35◦ 26 58.9 35◦ 21 48.4 35◦ 17 58.1 35◦ 23 09.6 35◦ 28 44.9 35◦ 04 42.3 35◦ 12 48.4 35◦ 06 15.0 35◦ 03 31.0

74◦ 09 07.9

74◦ 08 29.8 74◦ 00 00.2 72◦ 36 32.9 73◦ 12 06.7 73◦ 13 04.6 73◦ 16 02.0 72◦ 29 00.7 72◦ 33 02.8 73◦ 00 33.2 73◦ 07 09.2

6.0 6.0 6.0 6.0 8.8 8.8 8.8 8.8 8.8 11.0 11.0

1.0 1.0 1.0 1.0 1.0 1.0 1.0 1.0 1.0 1.0 1.0

6 6 6 6 6 6 6 6 6 6 6

CPX-PL-QTZ CPX-PL-QTZ CPX-PL-QTZ CPX-PL-QTZ CPX-PL-QTZ CPX-PL-QTZ CPX-PL-QTZ CPX-PL-QTZ CPX-PL-QTZ CPX-PL-QTZ CPX-PL-QTZ

6.6 6.7 6.3 6.4 9.1 8.8 8.3 9.3 8.5 11.7 11.0

1.7 1.7 1.8 1.8 1.7 1.7 1.7 1.9 1.8 1.6 1.8

35◦ 01 47.2

72◦ 56 02.2

−0.6 −0.7 −0.3 −0.4 −0.3 0.0 0.5 −0.5 0.3 −0.7 0.0 0.42

19.4

1.0

B

C RMS K12 (fabricated)

Hb Hb Hb Hb Hb Hb Hb Hb Hb Hb Hb Hb Hb Hb Hb Hb Hb Hb Hb Hb Hb Hb Hb Hb Hb Hb Hb Hb Hb Hb Hb Hb Hb Hb Hb

a

A standard error of 1 kbar was assumed, that reflects the maximum uncertainty of the pressures estimates (when quoted in the literature). Most reported errors are <1 kbar. This was done in order to not bias the kriging towards few P estimates with low reported uncertainty. b 1 – Jagoutz et al. (2013), 2 – Yamamoto and Yoshino (1998), 3 – Yamamoto (1993), 4 – Jagoutz et al. (2007), 5 – Ringuette et al. (1999), (location approximated), 6 – Yoshino and Okudaira (2004).

O. Jagoutz / Earth and Planetary Science Letters 396 (2014) 267–277

271

( R 2 = 0.96) (Fig. 2) with an average prediction root-mean-squareerror of 0.42 kbar (Table 1). The average prediction standard error for each datum is 1.8 kbar (1σ , Table 1), which is used here as the maximum uncertainty for the error associated with the interpolation (Fig. 2). The calculated pressure prediction map can be used in combination with known field locations of plutonic rocks to approximate the intrusion pressures of these rocks. This was done for 209 representative plutonic samples from the whole-rock compilation of Jagoutz and Schmidt (2012), for which both whole-rock analyses and GPS data exist. The emplacement pressures have been converted to emplacement depth using calculated rock densities (Jagoutz and Behn, 2013). Rock samples from the Kohistan arc, with the exception of some volcanic rocks, are generally very fresh and plutonic rock analyses included in the compilation here have generally high totals (>98%). The entire dataset is presented as an electronic Appendix A. 4. Results

Fig. 2. Results of cross-validation of intrusion pressures determined by geobarometry and geostatistics (kriging). The results show an excellent correlation between measured and predicted pressures.

with the lack of significant jumps in emplacement depth over short distances (that could indicate undiscovered major fault zone), indicates that the Kohistan arc exposes a continuous crustal section without any significant gaps or repetitions due to tectonic faulting. Therefore it is possible to quantitatively model the exposure levels throughout the arc using geostatistic analyses and the existing pressure estimates as input parameters. 3. Methods ESRI ArcMap 9.3 Geostatistical Analyst was used to create an interpolated pressure surface using the ordinary Kriging method, which assumes no trend in the data. This method, based on cross validation, predicted the smallest residual mean square error and was thus favored over other interpolation methods. The 67 geobarometric estimates and their spatial locations, compiled from the literature (Table 1), were the input for the pressure prediction map and a prediction standard error map (68% confidence interval). One point (K12) for which no direct pressure estimate exist had to be estimated by calculating the pressure gradient along the southern portion of the SPC (between Kiru and Jijal) and then extrapolating it further to the southwest to the Indus suture (Table 1). This was done so that the interpolation surface included all desired bulk-rock analyses in the final extraction step, notably the ultramafic cumulates at the base of the arc for which no pressure estimates exist. UTM coordinates (zone 43N) were used for the maps and kriging, as opposed to a latitude–longitude system, as the creation of a kriging surface is based on spatial correlation and the UTM projection preserves spatial relationships more accurately. No special care was taken for trend removal or anisotropic influences. Spherical semivariogram and covariance models were employed. The search neighborhood was restricted to no less than two and no more than five neighbors within an ellipse divided into 4 sectors with 45 ◦ offset. After the kriging surface was generated, the model was validated by examining the error statistics between the known geobarometric estimates and the pressures predicted for those same locations by the interpolation surface (Table 1). The predicted and measured pressure estimates agree well

The result of the geostatistical modeling is shown as isobars in Fig. 1. In Figs. 3–5, the emplacement depth is plotted against whole-rock geochemical data and is compared to the composition of a Kohistan primitive arc melt (Jagoutz, 2010). As discussed previously, no significant variation in pressure occurs in the Chilas Complex, and accordingly no parameters vary significantly with intrusion pressures and these rocks are excluded from the further discussion of depth dependent stratification. 4.1. Geochemical structure of the Kohistan arc with depth As plutonic rocks can range from cumulates to near liquids in composition, whole rock geochemical characteristics of plutonic rocks are used to separate liquid-dominated plutonic rocks from cumulate-dominated ones as outlined in detail by Jagoutz et al. (2011). Throughout this manuscript, the term “cumulate” is defined as a magmatic rock from which some amount of residual liquid has escaped. By this definition (and according to the criteria of Jagoutz et al., 2011), cumulate rocks have SiO2 < 50 wt% and/or Si/Al <2.9% and Sr/Nd > 0.50, Eu/Eu∗ > 1 and/or Eu/Sm > 0.46. With the exception of a small exposure of residual mantle in the Sapat area (Bouilhol et al., 2009), the ultramafic rocks found in the Kohistan arc are cumulates. In contrast, the term “liquid-dominated rocks”, is hereby defined as rocks that approach frozen liquid compositions with negligible crystal accumulation. The applied criteria would also characterize ∼15% of all volcanics as cumulates, which is in accordance with field observations indicating the presence of ultra-porphyric volcanics that likely don’t represent melts compositions but rather erupted crystal mushes. Figs. 3–5 show the presence of a strong chemical stratification between the upper and lower arc crust in the Kohistan arc. Liquid-dominated plutonic rocks are volumetrically dominant in the upper part (<30 km) of the arc crust (Figs. 3–5), and remain abundant to a relatively deep level of ∼30–40 km, but are conspicuously rare in the deepest exposed levels (∼40–55 km depth). Cumulate plutonic rocks comprise the deeper part of the arc (∼40–55 km) and are rare in the dataset at intrusion depths <30 km. A broad transition zone exits (from ∼30 to 40 km) in the lower crust where cumulates and near-liquid plutonic rocks coexist. Although considerable variation in composition occurs at all depths, especially for the more incompatible elements (for both the cumulate sequence and the liquid-dominated plutonic rocks), the compositions of liquids are much more restricted (Figs. 3–5). Incompatible trace and major elements are enriched in the upper part of the arc (SiO2 , K2 O, Na2 O, Rb, Zr, Hf etc.), whereas compatible elements (e.g., MgO, FeO) are more enriched in the lower

272

O. Jagoutz / Earth and Planetary Science Letters 396 (2014) 267–277

Fig. 3. Variation of the whole rock major element composition vs. intrusion depth. Stars denote the composition of the lower (brown), bulk (red) and upper continental crust (orange) of Rudnick and Gao (2003) plotted ‘arbitrarily’ at 10 km, 20 km and 30 km depth, respectively. The blue line indicates the average composition of Kohistan primitive arc volcanics (after Jagoutz, 2010). Yellow squares indicate plutonic rocks with near liquid chemical composition, black triangles indicate cumulate rocks. (For interpretation of the references to color in this figure legend, the reader is referred to the web version of this article.)

Fig. 4. Variation of the whole rock element composition vs. intrusion depth. Symbols as in Fig. 2.

arc crust due to the dominance of high temperature mafic cumulates. Highly incompatible elements, including heat-producing elements (K, Th, U), roughly correlate exponentially with intrusion pressure across the entire arc sequence. For example, the K2 O and Rb concentration vs. pressure correlation can be fitted by a logarithmic regression line with an R 2 of ∼0.6–0.7, indicating a reasonable correlation between intrusion pressure and incompatible element concentrations of the plutonic rocks for the entire arc sequence. In detail however, little correlation between the emplacement depth and composition of the granitoids exists in the upper 30 km, indicated by a R 2 value of ∼0.1–0.2 for K2 O and Rb vs. depth for the upper most 30 km only. This indicates that the upper 30 km of the arc are essentially unstratified whereas the deeper part of the arc >30 km is (Fig. 3). Rocks of highly variable composition are observed at all depths recorded in the Gilgit complex (9–30 km) (see Figs. 3–5 and Jagoutz et al., 2013 for detailed discussion). In the deeper arc crust (>30 km) a broad correlation between chemical composition and intrusion depth is found for liquid-

dominated rocks. In cumulates rocks, only the maximum concentration of incompatible elements (e.g., K2 O) found at any given depth correlates well with intrusion depth (Fig. 3). The liquid-dominated granitoids have CaO concentration as little as <1 wt% and FeO/MgO of >2 (Fig. 4). These variations are in accordance with a standard calk-alkaline liquid line of descent (Jagoutz, 2010) and indicate that complementary cumulate compositions with high CaO concentration and/or low FeO/MgO have been fractionated. Such complementary cumulate compositions are observed in the deeper crust only (>30 km). The deepseated cumulates have geochemical trace element characteristics that indicate accumulation of magmatic garnet (low Dy/Yb) and amphibole (high Dy/Yb). Accordingly liquid compositions in the upper crust show trace element depletions corresponding to prior fractionation of garnet and amphibole (Fig. 5). The cumulate dominated lower arc crust is characterized by dominantly low Th/La ratios, with a few minor exceptions (Fig. 5), whereas the uppermiddle crust has elevated Th/La ratios supporting the suggestions of Plank (2005) that fractionation in the lower crust could be

O. Jagoutz / Earth and Planetary Science Letters 396 (2014) 267–277

273

Fig. 5. Variation of the whole rock composition vs. intrusion pressure. Symbols as in Fig. 2.

responsible for the generally elevated Th/La seen in the bulk continental crust compared with source magmas. 5. Discussion The results presented here document significant chemical stratification of an undisturbed juvenile island-arc crust. The upper arc crust is generally granitic, whereas the lower crust has an overall mafic composition. As post-collisional rocks are excluded, this observed stratification is related to intrinsic arc crustal processes and unrelated to any later stage collision and reworking events. This does not imply that secondary reworking does not occur, but it is likely not the primary control on the observed compositional stratification. The dominance of cumulate rocks in the lower arc crust indicates that fractionation and magma differentiation occurred in the lower crust in Kohistan. The results presented here are consistent with a model in which basaltic melts stagnate and differentiate in the lower arc crust. Differentiated derivative liquids sourced from such a lower crust hot zone (e.g. Annen et al., 2006) migrate to shallower crustal level and get stuck at variable depths. This implies that melt stagnation occurs at two principal locations in the arc crust: mafic melts stall in the lower arc crust and derivative felsic melts stall at higher crustal levels. The ambient thermal regime and the density structure of the arc crust are general considered key parameters that control melt stagnation. Therefore, in the following paragraph the geothermal gradient and the density structure of the arc crust compared to that of melt composition will be constrained. 5.1. The thermal gradient in the Kohistan arc Stable, conductive geothermal gradients in arcs are likely absent and the thermal regime is, to a significant degree, transient in nature due to the interaction between a background “steady state” geothermal gradient and perturbation by frequent melt infiltration events (e.g., Annen et al., 2006; Dufek and Bergantz, 2005; Kelemen et al., 2003). This results in a broad transient temperature gradient, here called the ‘arctherm’. The liquidus temperatures of the infiltrating melts, however, can range from the liquidus temperature of anhydrous primitive mantle derived

melts (∼1200–1300 ◦ C) to temperatures near the hydrous granitic solidus (∼650–750 ◦ C). Therefore the range of transient arctherms recorded in the Kohistan arc is bracketed by constraining (a) the metamorphic temperatures recorded in the plutonic rocks and (b) the minimum liquidus temperatures calculated for the compositions of near-liquidus plutonic rocks at appropriate depths. 5.1.1. Geothermometric and barometric constraints The metamorphic thermal conditions inferred for the lower Kohistan arc crust are significantly colder (600–900 ◦ C; Fig. 6; Ringuette et al., 1999; Yoshino and Okudaira, 2004) than those inferred for e.g., the lower Talkeetna arc crust (1000 ◦ C; Hacker et al., 2008; Kelemen et al., 2003), indicating that the thermal regimes in arcs might be highly variable. Within the upper crust of Kohistan, the igneous crystallization temperatures are constrained by the hornblende–plagioclase thermometer (Blundy and Holland, 1990) and range from ∼700–850 ◦ C (Jagoutz et al., 2013). This temperature range is significantly lower than the dry solidus in granitic systems and at, or slightly above, the water-saturated solidus of eutectic granite compositions indicating that the granitic melts in Kohistan had significant water contents (∼2–9 wt%, Fig. 6). The apparent ‘arctherm’ defined by the data shown in Fig. 6 is accordingly considered an artifact between magmatic near-solidus temperatures recorded in the shallower part of the arc by the hornblendeplagioclase thermometer and the metamorphic closure temperatures recorded for the lower part, and does not record a geothermal gradient. Nevertheless, the closure temperatures recorded at the base of the arc as well as the absence of eclogitic assemblages, despite the recorded high pressures (Ringuette et al., 1999), constrain the minimum background steady-state geothermal gradient to those calculated with a surface heat flow ∼60 mW/m2 . Such cold background geotherms are in accordance with results from thermal models of subduction zones that do not take into account increased heat flux due to magmatism or temperature-dependent viscosity in the sub arc mantle (Fig. 6) (Kelemen et al., 2003). 5.1.2. Liquidus temperature of intruded melts To estimate the ranges of magmatic temperatures, and thereby the influence of heat advection on the arctherm, the liquidus temperatures of all liquid-dominated plutonic rocks were calculated at their inferred intrusion pressure (assuming an H2 O

274

O. Jagoutz / Earth and Planetary Science Letters 396 (2014) 267–277

Fig. 6. P , T conditions in the Kohistan arc constrained by metamorphic exchange equilibrium (red symbols) and the near solidus temperature of felsic liquids determined by hbl–plag thermometry (yellow symbols, data source as cited in Fig. 1). Also shown (in black) are the liquidus temperatures of plutonic rocks with near liquid compositions (calculated using the MELTS program with 4 wt% H2 O). The thermal conditions in Kohistan arc compared to predicted geotherms from thermal modeling (fine lines) as compiled by Kelemen et al. (2003) and different geotherms. The stippled field indicates the range within which the ‘arctherm’ likely varied during arc activity. Conductive geotherms are calculated using the distribution of heat producing elements in Kohistan. The thermal conditions in the lower crust are best explained by a 60 mW/m2 geotherm. (For interpretation of the references to color in this figure legend, the reader is referred to the web version of this article.)

content of 4 wt%; see below) using the MELTS (at 10 kbar) and pMELTS (>10 kbar) algorithms (Ghiorso et al., 2002; Ghiorso and Sack, 1995). The minimum liquid temperature preserved at a given depth and the near-solidus temperatures recorded by the hornblende–plagioclase thermometer yield an upper estimate of the transient arctherm. In the upper crust (<30 km) a wide range of magmatic liquidus temperatures are observed (∼800–1200 ◦ C) consistent with the highly variable plutonic rocks compositions (Fig. 6). Only the coldest calculated liquidus temperatures in the shallow arc crust overlap with the temperature constraints from the hornblende–plagioclase thermometer, which probably relates to underestimation of the water content in the MELTS calculations. In the deeper arc crust, the calculated liquidus temperatures are much more restricted (∼1100–1200 ◦ C) over a significant depth interval (30–47 km), in accordance with the dominance of rather mafic liquid compositions observed. The ∼7 Myr differences between U–Pb zircon intrusion ages and Ar–Ar hbl hornblende ages for rocks form the SPC (Schaltegger et al., 2002; Treloar et al., 1989) indicate that the lower arc crust cooled slowly after the emplacement of the different units but prior to its exhumation. Accordingly, the recorded metamorphic temperature of ∼850 ◦ C at ∼40–55 km depth provides the best upper bound on the thermal conditions in the lower arc crust. Such thermal conditions are best explained by a 60–80 mW/m2 geotherms. 5.2. The existence of a neutral buoyancy line in the arc crust? The concept of a neutral buoyancy layer (NBL) in the crust in which melts are neutrally buoyant with respect to the surrounding

Fig. 7. Vertical density structure of the Kohistan arc crust along a ∼60 mW/m2 geotherm (after Jagoutz and Behn, 2013). Symbols as in Fig. 2 except for orange triangles that represent density of liquid dominated plutonic rocks calculated as hydrous melts (4 wt% H2 O) using MELTS. The dashed lines indicate the density of a Kohistan primitive arc melt with variable H2 O content. Other primitive arc melts from the Jagoutz and Schmidt (2013) compilation have similar densities (omitted to preserve clarity).

host rocks has been widely used to explain magma trapping in the upper crust (e.g. Glazner and Ussler III, 1988) or at Moho depths (Herzberg et al., 1983). In the literature, an NBL is often defined as the level where melts encounter rocks that have a density that is equal to, or less than, that of the immediately adjacent melt. More physically meaningful however, is the definition in which the NBL represents the depth at which the integrated density difference between a (partial) melt column and the adjacent rock column are equal (see e.g., Jagoutz et al., 2006 for a more detailed discussion). In this definition the NBL might be significantly shallower in the crust than the level at which melts encounter rocks that are less dense. Regardless, the density structure of the crust with respect to the infiltrating melts remains poorly constrained despite the fact that the NBL is a widely used concept. To investigate the potential control of a NBL on magma emplacement in the Kohistan arc, the density of all plutonic rocks were calculated along a 60 mW/m2 geotherm (Jagoutz and Behn, 2013). These calculated rock densities are here compared to melt densities for the liquid-dominated plutonic rocks and for primitive arc magmas (Fig. 7) using the MELTS (at 10 kbar) and pMELTS (>10 kbar) algorithms (Ghiorso et al., 2002; Ghiorso and Sack, 1995). 5.2.1. Melt densities (1) The densities of the primitive arc melts have been calculated at both their anhydrous and hydrous liquidus temperatures assuming H2 O content in the melt is 2 wt%. These water concentrations are at the lower end of what is recorded on average in arc magmas (Plank et al., 2013). As water reduces the density of the melt, melt densities reported here are likely upper estimates. (2) The melt densities of the liquid-dominated plutonic rocks have been calculated at their liquidus temperatures and emplacement pressures (see above), with both 0 wt% H2 O (anhydrous) and 4 wt% H2 O. 4 wt% H2 O is consistent with the average recorded by olivine-hosted melt inclusions in arc magmas globally (Plank et al., 2013)

O. Jagoutz / Earth and Planetary Science Letters 396 (2014) 267–277

5.2.2. Rock/crustal densities The densities of the plutonic rocks have been calculated with 1 wt% H2 O at their constrained pressure along a 60 mW/m2 geotherm and Fe2+ /Fe3+ = 0.25 (see Jagoutz and Behn, 2013 for details on the calculations). The water content assumed for the plutonic rocks is a maximum and implies that plutonic rocks are made up of a significant amount (25–50 wt%) of hydrous minerals such as amphibole or biotite, for which there is no evidence in the field. The use of a lower water content in the plutonic rocks will increase the calculated densities and the density contrast between crust and rising melts. The choice of the water contents used in the calculations aims to approximate the minimum amount of density difference between melt and rocks that could exist in the crust. 5.2.3. Results for Kohistan The results are presented in Fig. 7 and indicate a strong density contrast between the upper and lower arc crust of Kohistan. The upper-crustal rocks are characterized by densities of ∼2600 to 2950 kg/m3 , whereas the gabbroic and ultramafic rock in the lower arc crust have densities between ∼3250–3600 kg/m3 (Jagoutz and Behn, 2013). In general, calculated melt densities of the liquid-dominated plutonic rocks are significantly less dense than rocks at the same depth. Melt densities range from 2300 to 2500 kg/m3 , about 200–400 kg/m3 less dense than the corresponding rock densities. Thus, stagnation of more evolved melts in the crust due to a NBL is likely of minor importance. 5.2.4. Results for arcs globally It is frequently argued that primitive mantle derived melts pond at the base of the arc/continental crust (‘underplate’) (Herzberg et al., 1983). To test this idea, the densities of primitive arc melts (compiled in Jagoutz and Schmidt, 2013) have been calculated and compared to the densities of the crustal rocks. The calculation shows that primitive anhydrous melt compositions have densities generally below ∼2900 kg/m3 in accordance with previous results (Herzberg et al., 1983). Significant volumes of rocks with a density of <2900 kg/m3 , that could act as density filter for these anhydrous primitive melts, are only observed in the uppermost 20 km of the arc crust. While there are some rocks at 20–35 km depth with densities comparable to anhydrous primitive melts these are volumetrically minor. Additionally, primitive arc melts are generally considered hydrous. The calculations show that hydrous primitive arc melts with 2–4 wt% H2 O are significantly less dense (2400–2700 kg/m3 ) than any rock composition found in the entire arc crust (Fig. 7). Thus, density cannot be the dominant control on melt stagnation anywhere in the arc crust. Naturally, if sufficient melt volumes with an evolved granitic (s.l.) composition are present throughout a large portion of the crust at any given time, the bulk density of such a crystal mush will lie between that of solid rocks and melts (2700–2900 kg/m3 ) and could indeed act as a density filter for e.g., dry mafic arc magmas (Sisson et al., 1996). 5.3. The compositional build up of the arc crust, a simple model An important result of this study is that cumulate rocks in the Kohistan are dominantly restricted to the lower arc crust whereas the upper crust of the arc is composed largely of liquid-dominated rocks. Unlike the upper crust, systematic large-scale separation of cumulates and liquids occurred in the lower arc crust as indicated by (a) the gradual decrease in near liquid-composition rocks with depth, (b) the general compositional stratification, and (c) the lack of appropriate volumes of evolved plutonic rocks in the lower arc crust. The highly variable composition found within a cumulate sequence at any given depth is interpreted to reflect the emplacement and subsequent fractionation of different melt

275

patches at different times. This observation is in line with isotopic and geochronological data that indicate that different plutonic rocks have variable sources and were intruded over a time span of ∼30–50 Myr (Dhuime et al., 2007, 2009; Schaltegger et al., 2002). Field observations in the SPC indicate the dominance of originally horizontal layers or sills in the lower arc crust (Burg et al., 2005). This observation combined with the correlation of the incompatible elements with depth in the cumulate sequence, conflict with the theory that the cumulates were emplaced as ‘reverse diapirs’ of cumulates and restites sinking through the crust (Glazner, 1994) as this model at least locally would predict highly sheared vertical contacts and fails to explain the observed compositional stratification. The stratification is likely best explained through in-situ emplacement of variable mafic magma batches, derived from the underlying sub-arc mantle and subsequent magmatic differentiation in the lower arc crust. The preservation of the correlation between liquid-dominated rocks and intrusion depth in the lower arc crust suggests that the average minimum liquidus temperature of melts emplaced in the lower crust was controlled to a significant degree by the temperature regime in the lower arc crust, where rocks emplaced at shallower levels in the SPC could fractionate to more evolved compositions than those at deeper levels (Petford et al., 2000). While the thermal dataset presented here remains limited, the restricted melt temperature (∼1100–1200 ◦ C) observed in the lower crust could indicate that the thermal regime of the arc crust has a strong influence on the location of melt stagnation. Experimental studies show that, significant crystallization occurs over a narrow temperature interval. For example in picritic basaltic compositions fractionated under hydrous conditions and 15 kbar, the melt mass is significantly reduced, from ∼70% to ∼30% over a very limited temperature interval from 1140 ◦ C to 1110 ◦ C (Alonso-Perez, 2006). In divergent plate boundaries, it has been postulated that such drastic drop in melt mass produces low permeability layer along which melts stagnate in the lower oceanic crust (Kelemen and Aharonov, 1998). It is speculated here that melt stagnation in the lower arc crust could be similarly controlled by permeability barriers associated with significant reduction in melt mass. By analogy with the model proposed by Kelemen and Aharonov (1998) it is postulated that melts stagnate and form horizontal sills along permeability barriers once melts reach the ∼1100–1200 ◦ C temperature interval. Within these sills, melts undergo magmatic differentiation, assimilation and trigger local partial melting (Hildreth and Moorbath, 1988; Annen et al., 2006). The resulting hydrous felsic melts, with relatively low viscosities, will pool and accumulate due to compaction driven melt segregation (McKenzie, 1985). The high metamorphic temperatures and million year cooling timescales observed in the lower crust (see discussion above) are consistent with hydrous felsic composition being present as melts in the lower arc crust for sufficient times scales for melt pooling due to viscous compaction. Once the magma pressure in these melt pools exceeds the lithostatic pressure, cracking occurs and felsic melts eventually are extracted and migrate upwards to shallow crustal levels due to diking (Rubin, 1995). A related outcome of this model is that the crustal thickness of arcs, with the exception of those undergoing density sorting (e.g. Jagoutz and Behn, 2013), is controlled to first order by the thermal regime of the arc crust and mantle wedge. Independent of seismic velocities, the petrological base of the arc crust is defined by the first appearance of volumetrically significant cumulates rocks fractionated from primitive arc melts. Thus, the arc crustal thickness is controlled by the location at which primitive melts encounter an inverted geothermal gradient and begin to crystallize, typically at temperatures in the range of 1200–1250 ◦ C. Thus, while

276

O. Jagoutz / Earth and Planetary Science Letters 396 (2014) 267–277

an arc with a hotter mantle wedge may melt to greater extents, the 1200–1250 ◦ C isotherms will be encountered at a shallower depth, leading to a thinner overall arc crust. Therefore the model proposed here can be tested by future studies that better constrain the thermal regime in the lower crust of active arcs and compare the results with observed crustal thickness. 6. Conclusions This study documents the chemical stratification in a juvenile arc crust. The chemically un-stratified liquid-dominated upper crust and the dominance of cumulates in the lower crust indicate that the construction of the arc crust is largely controlled by magmatic differentiation in the lower crust. In the absence of crystal mushes, density barriers along which melts stagnate seem to play a minor role in controlling the depth of melt stagnation. Alternatively, a model is postulated that explains trapping of melts in the lower arc crust by permeability barriers, similar to the model proposed for the lower oceanic crust (e.g. Kelemen and Aharonov, 1998). If continental crust forms predominantly by amalgation of Kohistan-arc-like fragments, a significant part of the chemical differentiation seen in the continental crust must be inherited from these subduction zone processes. Acknowledgements This work was supported through funding from the NSF (EAR 1322032/692005). Alexandra Jordan, Katie Pesce and Nick van Buer are thanked for help with the ArcGIS work and the MELTS calculations. Jill VanTongeren is thanked for numerous stimulating discussions and endless challenges to the ideas presented here. The reviews of Christy Till, and three anonymous reviewers and the editorial handling of Tim Elliot are greatly appreciated. Appendix A. Supplementary material Supplementary material related to this article can be found online at http://dx.doi.org/10.1016/j.epsl.2014.03.060. References Alonso-Perez, R., 2006. The Role of Garnet in the Formation of Calc-Alkaline Laves. Institute for Mineralogy, ETH, Zurich, p. 300. Alonso-Perez, R., Müntener, O., Ulmer, P., 2009. Igneous garnet and amphibole fractionation in the roots of island arcs: experimental constraints on H2 O undersaturated andesitic liquids. Contrib. Mineral. Petrol. 157, 541–558. 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, 505–539. Arndt, N.T., Goldstein, S.L., 1989. An open boundary between lower continental crust and mantle: its role in crust formation and crustal recycling. Tectonophysics 161, 201–212. Bard, J.P., 1983. Metamorphism of an obducted island arc. Example of the Kohistan sequence (Pakistan) in the Himalayan collided range. Earth Planet. Sci. Lett. 65, 133–144. Bateman, P.C., Chapell, B.W., 1979. Crystallization, fractionation, and solidification of the Tuolumne intrusive series, Yosemite National Park, California. Geol. Soc. Am. Bull. 90, 465–482. Behn, M.D., Hirth, G., Kelemen, P.B., 2007. Trench-parallel anisotropy produced by foundering of arc lower crust. Science 317, 108–111. Blatter, D.L., Sisson, T.W., Ben Hankins, W., 2013. Crystallization of oxidized, moderately hydrous arc basalt at mid- to lower-crustal pressures: implications for andesite genesis. Contrib. Mineral. Petrol. 166, 861–886. Blundy, J., Cashman, K., 2008. Petrologic reconstruction of magmatic system variables and processes. Rev. Mineral. Geochem. 69, 179–239. Blundy, J.D., Holland, T.J.B., 1990. Calcic amphibole equilibria and a new amphibole– plagioclase geothermometer. Contrib. Mineral. Petrol. 104, 208–224. Bouilhol, P., Burg, J.P., Bodinier, J.L., Schmidt, M.W., Dawood, H., Hussain, S., 2009. Magma and fluid percolation in arc to forearc mantle: evidence from Sapat (Kohistan, Northern Pakistan). Lithos 107, 17–37. Bouilhol, P., Jagoutz, O., Hanchar, J., Dudas, F., 2013. Dating the India–Eurasia collision through arc magmatic records. Earth Planet. Sci. Lett. 366, 163–175.

Brookfield, M.E., Reynolds, P.H., 1981. Late Cretaceous emplacement of the Indus suture zone ophiolitic melanges and an Eocene–Oligocene magmatic arc on the northern edge of the Indian Plate. Earth Planet. Sci. Lett. 55, 157–162. Burg, J.P., 2011. The Asia–Kohistan–India collision. Review and discussion. In: Brown, D., Ryan, P.D. (Eds.), Arc-Continent Collision. Springer. Burg, J.P., Arbaret, L., Chaudhry, N.M., Dawood, H., Hussain, S., Zeilinger, G., 2005. Shear strain localization from the upper mantle to the middle crust of the Kohistan arc (Pakistan). In: Bruhn, D., Burlini, L. (Eds.), High-Strain Zones: Structure and Physical Properties. Geological Society, London, pp. 25–38. Burg, J.P., Jagoutz, O., Hamid, D., Hussain, S., 2006. Pre-collision tilt of crustal blocks in rifted island arcs: structural evidence from the Kohistan arc. Tectonics 25, 13. http://dx.doi.org/10.1029/2005TC001835. Christensen, N.I., Mooney, W.D., 1995. Seismic velocity structure and composition of the continental crust: a global view. J. Geophys. Res. B, Solid Earth and Planets 100, 9761–9788. Crowley, J.L., Waters, D., Searle, M., Bowring, S., 2009. Pleistocene melting and rapid exhumation of the Nanga Parbat massif, Pakistan: age and P –T conditions of accessory mineral growth in migmatite and leucogranite. Earth Planet. Sci. Lett. 288, 408–420. Dhuime, B., Bosch, D., Bodinier, J.L., Garrido, C.J., Bruguier, O., Hussain, S.S., Dawood, H., 2007. Multistage evolution of the Jijal ultramafic–mafic complex (Kohistan, N. Pakistan): implications for building the roots of island arcs. Earth Planet. Sci. Lett. 261, 179–200. Dhuime, B., Bosch, D., Garrido, C.J., Bodinier, J.L., Bruguier, O., Hussain, S.S., Dawood, H., 2009. Geochemical architecture of the lower- to middle-crustal section of a Paleo-island arc (Kohistan complex, Jijal Kamila area, Northern Pakistan): implications for the evolution of an Oceanic subduction zone. J. Petrol. 50, 531–569. Dufek, J., Bergantz, 2005. Lower crustal magma genesis and preservation: a stochastic framework for the evaluation of the basalt–crust interaction. J. Petrol. 46, 2167–2195. Garrido, C.J., Bodinier, J.L., Burg, J.P., Zeilinger, G., Hussain, S.S., Dawood, H., Chaudhry, M.N., Gervilla, F., 2006. Petrogenesis of mafic garnet granulite in the lower crust of the Kohistan paleo-arc complex (Northern Pakistan). Implications for intra-crustal differentiation of island arcs and generation of continental crust. J. Petrol. 47, 1873–1914. Ghiorso, M.S., Hirschmann, M.M., Reiners, P.W., Kress, V.C., 2002. The pMELTS: a revision of MELTS for improved calculation of phase relations and major element partitioning related to partial melting of the mantle to 3 GPa. Geochem. Geophys. Geosyst. 3. Ghiorso, M.S., Sack, R.O., 1995. Chemical mass-transfer in magmatic processes. 4. A revised and internally consistent thermodynamic model for the interpolation and extrapolation of liquid–solid equilibria in magmatic systems at elevatedtemperatures and pressures. Contrib. Mineral. Petrol. 119, 197–212. Glazner, A.F., 1994. Foundering of mafic plutons and density stratification of continental crust. Geology 22, 435. Glazner, A.F., Ussler III, W., 1988. Trapping of magma at midcrustal density discontinuities. Geophys. Res. Lett. 15, 673–675. Grove, T.L., Elkins-Tanton, L.T., Parman, S.W., Chatterjee, N., Muntener, O., Gaetani, G.A., 2003. Fractional crystallization and mantle-melting controls on calcalkaline differentiation trends. Contrib. Mineral. Petrol. 145, 515–533. Hacker, B.R., Mehl, L., Kelemen, P.B., Rioux, M., Behn, M.D., Luffi, P., 2008. Reconstruction of the Talkeetna intraoceanic arc of Alaska through thermobarometry. J. Geophys. Res., Solid Earth 113. Halliday, A., Mahood, G.A., Holden, P., Metz, J., Dempster, T., Davidson, J.P., 1989. Evidence for long residence times of rhyolitic magma in the Long Valley magmatic system: the isotopic record in precaldera lavas of Glass Mountain. Earth Planet. Sci. Lett. 94, 274–290. Harley, S., 1989. The origins of granulites: a metamorphic perspective. Geol. Mag. 126, 215–247. Herzberg, C., Fyfe, W., Carr, M., 1983. Density constraints on the formation of the continental Moho and crust. Contrib. Mineral. Petrol. 84, 1–5. Hildreth, W., Moorbath, S., 1988. Crustal contributions to arc magmatism in the Andes of Central Chile. Contrib. Mineral. Petrol. 98, 455–489. Jagoutz, O., 2010. Construction of the granitoid crust of an island arc part II: a quantitative petrogenetic model. Contrib. Mineral. Petrol. 160, 359–381. Jagoutz, O., Behn, M.D., 2013. Foundering of lower island-arc crust as an explanation for the origin of the continental Moho. Nature 504, 131–134. Jagoutz, O., Müntener, O., Schmidt, M.W., Burg, J.P., 2011. The respective roles of flux- and decompression melting and their relevant liquid lines of descent for continental crust formation: evidence from the Kohistan arc. Earth Planet. Sci. Lett. 303, 25–36. Jagoutz, O., Müntener, O., Burg, J.-P., Ulmer, P., Jagoutz, E., 2006. Lower continental crust formation through focused flow in km-scale melt conduits: the zoned ultramafic bodies of the Chilas Complex in the Kohistan Island arc (NW Pakistan). Earth Planet. Sci. Lett. 242, 320–342. Jagoutz, O., Müntener, O., Ulmer, P., Pettke, T., Burg, J.P., Dawood, H., Hussain, S., 2007. Petrology and mineral chemistry of lower crustal intrusions: the Chilas Complex, Kohistan (NW Pakistan). J. Petrol. 48, 1895–1953. Jagoutz, O., Schmidt, M.W., 2012. The formation and bulk composition of modern juvenile continental crust: the Kohistan arc. Chem. Geol. 298–299, 79–96.

O. Jagoutz / Earth and Planetary Science Letters 396 (2014) 267–277

Jagoutz, O., Schmidt, M.W., 2013. The nature and composition of the crustal delaminate in arcs. Earth Planet. Sci. Lett. 371–372, 177–190. Jagoutz, O., Schmidt, M.W., Enggist, A., Burg, J.P., Hamid, D., Hussain, S., 2013. TTGtype plutonic rocks formed in a modern arc batholith by hydrous fractionation in the lower arc crust. Contrib. Mineral. Petrol. 166, 1099–1118. Jellinek, A.M., DePaolo, D.J., 2003. A model for the origin of large silicic magma chambers: precursors of caldera-forming eruptions. Bull. Volcanol. 65, 363–381. Jull, M., Kelemen, P.B., 2001. On the conditions for lower crustal convective instability. J. Geophys. Res. B, Solid Earth and Planets 106, 6423–6446. Kay, R.W., Mahlburg Kay, S., 1993. Delamination and delamination magmatism. Tectonophysics 219, 177–189. Kelemen, P.B., Aharonov, E., 1998. Periodic formation of magma fractures and generation of layered gabbros in the lower crust beneath oceanic spreading ridges. In: Buck, W.R., Delaney Paul, T., Karson Jeffrey, A., Lagabrielle, Y. (Eds.), Faulting and Magmatism at Mid-Ocean Ridges. American Geophysical Union, Washington, DC, United States, pp. 267–289. Kelemen, P., Rilling, J.L., Parmentier, E.M., Mehl, L., Hacker, B.R., 2003. Thermal structure due to solid-state flow in the mantle wedge beneath arcs. In: Eiler, J.M. (Ed.), Inside Subduction Factory. American Geophysical Union, pp. 293–311. Khan, M.A., Jan, M.Q., Weaver, B.L., 1993. Evolution of the lower arc crust in Kohistan, N. Pakistan; temporal arc magmatism through early, mature and intra-arc rift stages. In: Treloar, P.J., Searle, M.P. (Eds.), Himalayan Tectonics. Geological Society of London, London, United Kingdom, pp. 123–138. Khan, M.A., Jan, M.Q., Windley, B.F., Tarney, J., Thirlwall, M.F., 1989. The Chilas mafic–ultramafic igneous complex; the root of the Kohistan island arc in the Himalaya of Northern Pakistan. In: Malinconico Lawrence Jr., L., Lillie, Robert J. (Eds.), Tectonics of the Western Himalayas. Geological Society of America (GSA), Boulder, CO, United States, pp. 75–94. Khan, S.D., Walker, D.J., Hall, S.A., Burke, K.C., Shah, M.T., Stockli, L., 2009. Did the Kohistan–Ladakh island arc collide first with India? Geol. Soc. Am. Bull. 121, 366–384. McKenzie, D., 1985. The extraction of magma from the crust and mantle. Earth Planet. Sci. Lett. 74, 81–91. Müntener, O., Kelemen, P.B., Grove, T.L., 2001. The role of H2 O during crystallization of primitive arc magmas under uppermost mantle conditions and genesis of igneous pyroxenites: an experimental study. Contrib. Mineral. Petrol. 141, 643–658. Petford, N., Cruden, A., McCaffrey, K., Vigneresse, J., 2000. Granite magma formation, transport and emplacement in the Earth’s crust. Nature 408, 669–673. Petterson, M.G., Treloar, P.J., 2004. Volcanostratigraphy of arc volcanic sequences in the Kohistan arc, North Pakistan: volcanism within island arc, back-arcbasin, and intra-continental tectonic settings. J. Volcanol. Geotherm. Res. 130, 147–178. Petterson, M.G., Windley, B.F., 1985. Rb–Sr dating of the Kohistan arc-batholith in the Trans-Himalaya of North Pakistan, and tectonic implications. Earth Planet. Sci. Lett. 74, 45–57. Petterson, M.G., Windley, B.F., 1991. Changing source regions of magmas and crustal growth in the Trans-Himalayas – evidence from the Chalt volcanics and Kohistan batholith, Kohistan, Northern Pakistan. Earth Planet. Sci. Lett. 102, 326–341. Plank, T., 2005. Constraints from thorium/lanthanum on sediment recycling at subduction zones and the evolution of the continents. J. Petrol. 46, 921–944. Plank, T., Kelley, K.A., Zimmer, M.M., Hauri, E.H., Wallace, P.J., 2013. Why do mafic arc magmas contain ∼4 wt% water on average? Earth Planet. Sci. Lett. 364, 168–179. Prouteau, G., Scaillet, B., 2003. Experimental constraints on the origin of the 1991 Pinatubo dacite. J. Petrol. 44, 2203.

277

Ringuette, L., Martignole, J., Windley, B.F., 1999. Magmatic crystallization, isobaric cooling, and decompression of the garnet-bearing assemblages of the Jijal Sequence (Kohistan Terrane, Western Himalayas). Geology (Boulder) 27, 139–142. Rowley, D.B., 1996. Age of initiation of collision between India and Asia: a review of stratigraphic data. Earth Planet. Sci. Lett. 145, 1–13. Rubin, A.M., 1995. Propagation of magma-filled cracks. Annu. Rev. Earth Planet. Sci. 23, 287–336. Rudnick, R.L., 1995. Making continental crust. Nature 378, 571–578. Rudnick, R.L., Gao, S., 2003. The composition of the continental crust. In: Rudnick, R.L. (Ed.), The Crust. Elsevier, Oxford, pp. 1–64. Ryan, M.P., 1987. Neutral buoyancy and the mechanical evolution of magmatic systems. In: Magmatic Processes: Physicochemical Principles, pp. 259–288. Schaltegger, U., Zeilinger, G., Frank, M., Burg, J.P., 2002. Multiple mantle sources during island arc magmatism; U–Pb and Hf isotopic evidence from the Kohistan arc complex, Pakistan. Terra Nova 14, 461–468. Schilling, F., Partzsch, G., 2001. Quantifying partial melt fraction in the crust beneath the Central Andes and the Tibetan Plateau. Phys. Chem. Earth, Part A, Solid Earth Geod. 26, 239–246. Sisson, T.W., Grove, T.L., 1993. Experimental investigations of the role of H2 O in calc-alkaline differentiation and subduction zone magmatism. Contrib. Mineral. Petrol. 113, 143–166. Sisson, T.W., Grove, T.L., Coleman, D.S., 1996. Hornblende gabbro sill complex at Onion Valley, California, and a mixing origin for the Sierra Nevada batholith. Contrib. Mineral. Petrol. 126, 81–108. Tahirkheli, R.A.K., 1979. Geology of Kohistan and adjoining Eurasia and Indio– Pakistan continents, Pakistan. Geol. Bull., Univ. Peshawar 11, 1–30. Treloar, P.J., Rex, D.C., Guise, P.G., Coward, M.P., Searle, M.P., Windley, B.F., Petterson, M.G., Jan, M.Q., Luff, I.W., 1989. K–Ar and Ar–Ar geochronology of the Himalayan collision in NW Pakistan; constraints on the timing of suturing, deformation, metamorphism and uplift. Tectonics 8, 881–909. van der Beek, P., Van Melle, J., Guillot, S., Pêcher, A., Reiners, P.W., Nicolescu, S., Latif, M., 2009. Eocene Tibetan plateau remnants preserved in the Northwest Himalaya. Nat. Geosci. 2, 364–368. Yamamoto, H., 1993. Contrasting metamorphic P –T –time paths of the Kohistan granulites and tectonics of the Western Himalayas. J. Geol. Soc. (Lond.) 150 (5), 843–856. Yamamoto, H., Kobayashi, K., Nakamura, E., Kaneko, Y., Kausar Allah, B., 2005. U–Pb zircon dating of regional deformation in the lower crust of the Kohistan arc. Int. Geol. Rev. 47, 1035–1047. Yamamoto, H., Yoshino, T., 1998. Superposition of replacements in the mafic granulites of the Jijal complex of the Kohistan arc, Northern Pakistan; dehydration rehydration within deep arc crust. Lithos 43, 219–234. Yoshino, T., Okudaira, T., 2004. Crustal growth by magmatic accretion constrained by metamorphic P –T paths and thermal models of the Kohistan arc, NW Himalayas. J. Petrol. 45, 2287–2302. Yoshino, T., Yamamoto, H., Okudaira, T., Toriumi, M., 1998. Crustal thickening of the lower crust of the Kohistan arc (N. Pakistan) deduced from Al zoning in clinopyroxene and plagioclase. J. Metamorph. Geol. 16, 729–748. Zaman, H., Torii, M., 1999. Palaeomagnetic study of Cretaceous red beds from the eastern Hindukush ranges, Northern Pakistan; palaeoreconstruction of the Kohistan–Karakoram composite unit before the India–Asia collision. Geophys. J. Int. 136, 719–738. Zandt, G., Leidig, M., Chmielowski, J., Baumont, D., Yuan, X., 2003. Seismic detection and characterization of the Altiplano–Puna magma body, Central Andes. Pure Appl. Geophys. 160, 789–807.