High-silica (>70 wt% SiO2) magmas are usually believed to form via shallow crustal–level fractional crystallization of intermediate magmas. However, the broad applicability of this model is controversial, because the required crystal-melt separation processes have rarely been documented globally up to now. The ca. 50 Ma Nyemo composite pluton of the Gangdese batholith belt in southern Tibet, which comprises intrusive rocks with intermediate- to high-silica compositions (65–78 wt%), offers a unique opportunity for substantiating the coexistence of extracted melts and complementary silicic cumulates in one of Earth's most complete transcrustal silicic magmatic systems. The Nyemo pluton intrusive rocks exhibit similar zircon Hf isotopic compositional ranges (mean εHf(t) = +5.7 to +8.3), suggesting a common, non-radiogenic magma source with crustal assimilation in the deep crust. Yet, these rocks have distinct geochemical characteristics. High-silica miarolitic and rapakivi granites are strongly depleted in Ba, Sr, and Eu, and their zircon trace elements show extremely low Eu/Eu* and Dy/Yb. In contrast, monzogranite is relatively enriched in Ba and Sr with minor Eu anomalies, and the zircon trace elements are characterized by relatively high Eu/Eu* and Dy/Yb. Therefore, we propose that the high-silica granites represent highly fractionated melt extracted from a mush reservoir at unusually low storage pressure (∼99–119 MPa), and that the monzogranite constitutes the complementary residual silicic cumulates.

Ubiquitous silicic igneous rocks, such as rhyolite and granite, are major components of the continental crust. However, understanding the processes responsible for the generation of silicic magmas is challenging due to the difficulties in identifying complementary cumulate residues predicted by the fractional crystallization and elevated heat flow required for the partial melting of the lower crust by means of underplating of mantle-derived magmas (e.g., Dufek and Bergantz, 2005; Annen et al., 2006). Recent studies suggested that high-silica melt was extracted from a rheologically locked mush through crystal-melt separation processes driven by crystal settling or compaction (Bachmann and Bergantz, 2004; Bachmann and Huber, 2019). However, this model indicates complementary cumulate residues in the mid- to upper crust (e.g., Deering and Bachmann, 2010). Some studies, therefore, focused on silica-rich rhyolites and complementary granites, thereby exploring the genetic relationship of plutonic and volcanic rocks (e.g., Deering et al., 2016; Watts et al., 2016; Yan et al., 2018). This is fundamental for the understanding of the genetic processes associated with silicic magmatic systems, which are a matter of long-lasting debate (e.g., Bachmann et al., 2007; Glazner et al., 2008; Lundstrom and Glazner, 2016), mainly due to the scarce exposure of both intrusive and extrusive suites of single magmatic systems, which occur only in regions of exceptional topographic relief or sites of structural rotation of the upper crust (Lipman, 1997). Fortunately, composite granitic plutons have the potential to document crystal-melt separation processes during the generation of silicic magmas because different parts of composite granitic plutons may record either a cumulate, an extracted melt, or a transitional composition (e.g., Bachl et al., 2001; Hartung et al., 2017; Schaen et al., 2017, 2018).

In order to shed more light on this topic, we undertook a case study of crystal-melt separation in the Nyemo composite pluton of the Gangdese batholith belt, a ∼2500-km-long giant magmatic-tectonic belt rooted in southern Tibet (Fig. 1A). The composite pluton displays a unique intermediate- to high-silica compositional stratification and a transition from equigranular and porphyritic textures at the base to miarolitic textures in the roof zones. We present in situ zircon U-Pb ages, trace element compositions, Lu-Hf isotopic data, and whole-rock geochemical data in order to document the role of crystal-melt separation in the generation of high-silica magma and associated cumulate monzogranite. The Nyemo composite pluton comprises one of Earth's most complete silicic magmatic systems and, therefore, provides an important example of in situ crystal-melt separation in composite plutons, thereby highlighting the genetic relationship of volcanic and plutonic realms.

Voluminous Triassic–Miocene plutonic rocks (Gangdese batholith belt) and Jurassic–Paleocene volcanic deposits occur in the Lhasa terrane and record the Andean-style subduction of the Neo-Tethyan Ocean and subsequent India-Asia collision (e.g., Ji et al., 2009). The Eocene Nyemo pluton, which is located in the middle part of the Gangdese batholith belt, comprises several magmatic units: high-silica (69–78 wt%) miarolitic (having irregular cavities into which crystals protrude) and rapakivi (having plagioclase-mantled alkali feldspar ovoids) granites, and intermediate-silica (65–71 wt%) monzogranite and granodiorite (Fig. 1A).

The high-silica components occur as small bodies at the roof of the Nyemo pluton (Fig. 1A). Among them, the miarolitic granite comprises ∼10 vol% of centimeter- to decimeter-diameter cavities (Fig. 1B) recording epizonal emplacement and exsolution of volatile phases from the melt (e.g., Schaen et al., 2018), consistent with fast decompression following eruption of the contemporaneous Linzizong volcanic rocks in the northern vicinity (Fig. 1A). Results of the rhyolite-MELTS geobarometer (Gualda and Ghiorso, 2014) suggest a crystallization pressure of ∼99–119 MPa for the miarolitic granite (Fig. S1 in the Supplemental Material1), indicating a very shallow depth (∼3–4 km) of the magma reservoir for the Nyemo pluton. In addition, the rapakivi granite contains few miarolitic cavities and ∼10–15 vol% of unevenly distributed alkali feldspar phenocrysts, which commonly show plagioclase overgrowth rims, forming rapakivi textures (Fig. 1C). As the main unit of the Nyemo pluton, the monzogranite is porphyritic with alkali feldspar phenocryst aggregates (Fig. 1D), suggesting feldspar accumulation (e.g., Vernon and Collins, 2011). A few mafic microgranular enclaves (MMEs) generally occur at the base of monzogranite. Rapakivi feldspar phenocrysts are present in both the monzogranite and MMEs (Figs. 1D and 1E). The granodiorite occurs at the basal contact of the Nyemo pluton (Fig. 1A) and shows a coarse-grained equigranular texture.

Laser ablation–inductively coupled plasma–mass spectrometry (LA-ICP-MS) zircon U-Pb geochronology of the different rocks composing the Nyemo pluton revealed—within error—indistinguishable ages (cf. Schaltegger et al., 2015) between 50.0 ± 0.5 Ma and 48.8 ± 0.5 Ma (Fig. 2A). Furthermore, all of these rocks exhibit similar zircon Hf isotopic compositional ranges (mean εHf(t) = +5.7 to +8.3) (Fig. 2B), indicating a common, non-radiogenic magma source. Nevertheless, their zircon εHf(t) values are lower than those of the adjacent contemporaneous mafic rocks (Figs. 1A and 2B), indicating that the felsic rocks formed in an open system including crustal contamination.

The granodiorite has the lowest Rb/Sr ratios and zircon Hf concentrations among the rocks of the Nyemo pluton, representing a composition close to that of the intermediate-silica parental magma (Figs. 24). The granodiorite further exhibits a modest heavy rare earth element (HREE) depletion with insignificant Eu and Sr anomalies, consistent with fractional crystallization dominated by amphibole and magnetite (Fig. 3). The particularly low Rb concentration in some adjacent gabbro samples may reflect accumulation of mafic minerals in response to melt separation in the deep crust (Fig. 2C). We suggest that the granitic magma was generated primarily by fractional crystallization with variable crustal assimilation from a mantle-derived magma in the deep crust. Our modeling, using assimilation-fractional and/or fractional crystallization processes with a gabbroic starting composition, suggests that the granodiorite was derived by ∼20%–40% contamination of the crust during a fractional crystallization process (Fig. 2C; see the Supplemental Material), although the compositional choice of the crustal assimilates is critical for the modeling. The granodiorite magma was subsequently transferred to and evolved in the upper crustal levels as the top part of a transcrustal magmatic system (see below).

As described above, all rock units from the Nyemo pluton were derived from a common magma source (Figs. 2A and 2B). The miarolitic granite displays “seagull”-shaped REE distribution patterns with relatively enriched HREEs and distinct negative Ba and Sr anomalies, all of which are characteristic for highly differentiated magmas (Fig. 3A; Glazner et al., 2008; Yan et al., 2018). Therefore, we suggest that the miarolitic granite composition is similar to that of melt extracted from an intermediate crystal mush in shallow-crustal silicic magmatic systems (Fig. 4E). This is consistent with Rayleigh fractionation modeling that yielded a highly fractionated melt at a crystallinity of ∼50%–80%, close to the optimal crystallinity window (∼50%–70%) for melt extraction (Figs. 3C and 3D; see the Supplemental Material; Dufek and Bachmann, 2010). This melt extraction model is roughly self-consistent with independent rhyolite-MELTS (Gualda et al., 2012) calculations constrained to a storage pressure of ∼110 MPa (Fig. S2). Although our estimated pressure is considerably lower than what has been suggested as the optimal storage pressure of subvolcanic magma reservoirs (∼200 ± 50 MPa; Huber et al., 2019), it is consistent with the presence of abundant miarolitic cavities in the roof of the Nyemo pluton (Figs. 1B and 1C).

The different lithologies of the Nyemo pluton exhibit a continuous compositional trend, yet there is an apparent gap of ∼6 wt% SiO2 between the monzogranite and the miarolitic granite. This compositional gap has also been reported from volcanic-plutonic caldera complexes and was attributed to high-silica melt extraction from crystal mushes in eruptive sequences (e.g., Dufek and Bachmann, 2010; Watts et al., 2016; Yan et al., 2018). The monzogranite samples further show steep REE patterns and minor Eu anomalies relative to the miarolitic granite (Fig. 3A), thereby indicating feldspar accumulation (Deering and Bachmann, 2010), which is also displayed by the relatively low Rb and high Eu and Ba concentrations (Figs. 3C and 3D). We further noted the absence of a distinct positive Eu anomaly in the monzogranite, indicating the presence of significant proportions of high-silica melt in the residual silicic cumulate reservoir (Fig. 3A). Therefore, in agreement with our crystal-mush model, the monzogranite shows a complementary relationship with the high-silica miarolitic granite of the Nyemo pluton, which implies that the monzogranite represents the residual silicic cumulate after extraction of the high-silica melt (Figs. 3C, 3D, and 4E). The less-evolved MMEs in monzogranite are strongly depleted in Rb with a minor Eu anomaly (Figs. 3A and 3B), and their zircon trace elements show high Eu/Eu* [Eu/Eu* = EuN / (SmN × GdN)0.5; N—chondrite normalized] and Dy/Yb ratios (Figs. 4B and 4C), which, along with the rapakivi feldspar phenocrysts in both the monzogranite and MMEs (Fig. 1), indicate a magma recharge event, which is thought to have reactivated the high-crystallinity reservoir and probably promoted the mobilization of interstitial liquid (Bachmann et al., 2002; Bachmann and Huber, 2019). Furthermore, the occurrence of both entrained crystals and miarolitic cavities in the rapakivi granite indicate that melt extraction was also driven by gas-driven filter pressing from a mush reservoir due to accumulation of a more compressible volatile phase in the melt (Pistone et al., 2015; Schaen et al., 2018).

The zircon from the monzogranite and the high-silica miarolitic and rapakivi granites can be subdivided into two types based on internal structure (Fig. 4A; see the Supplemental Material): type A has bright cathodoluminescence (CL) and clear oscillatory growth zoning, while type B is relatively CL dark with generally weak zoning and relatively high uranium contents similar to zircon described for some highly evolved volcanic, hypabyssal, and plutonic rocks (e.g., Buret et al., 2016; Troch et al., 2018). Type B zircon occurs either as the rim of the type A zircon in all rock types or as individual zircon grains in the high-silica rocks. As shown in Figure 4B, Eu/Eu* decreases steadily with increasing Hf concentration over a continuous compositional range in the type B zircon, indicating crystallization from an evolving melt mainly undergoing progressive feldspar fractionation. In addition, the Dy/Yb ratios of the type B zircon negatively correlate with Hf concentration (Fig. 4C), indicating the co-crystallization of amphibole, titanite, or apatite (Schaltegger and Davies, 2017). On the other hand, the relatively consistent low Hf concentration and the high Eu/Eu* and Dy/Yb ratios of the type A zircon indicate a less-evolved composition (Figs. 4B and 4C). Thus, we propose that the type B zircon from the monzogranite and the high-silica miarolitic and rapakivi granites crystallized subsequent to the melt extraction event, while the type A zircon from the high-silica rapakivi granite is thought to have been captured during melt extraction from the crystal mush, suggesting that zircon may be continuously mobilized into the extracting melts during the crystal-melt separation process (Deering et al., 2016; Yan et al., 2020).

Furthermore, a lattice-strain model shows a near-parabolic dependence on cation radius (Onuma et al., 1968) and is consistent with the REE partition coefficients calculated from the zircon and host-melt compositions (Fig. 4D; see the Supplemental Material). The difference between our calculated partition coefficients of the Nyemo pluton is consistent with different crystallization temperatures of the type A (∼850–750 °C) versus the type B (∼750–650 °C) zircon (Fig. 4D; Rubatto and Hermann, 2007), which are within the range obtained by the Ti-in-zircon thermometer of Ferry and Watson (2007) (647 °C to 986 °C; Table S4). The low crystallization temperature of type B zircon relative to type A zircon is also consistent with a crystal-melt separation process (Deering et al., 2016; Yan et al., 2020).

A newly discovered Eocene (ca. 50 Ma) composite granitic pluton comprising highly fractionated miarolitic and rapakivi granites and associated monzogranite and granodiorite in the Nyemo area of the Gangdese batholith belt, southern Tibet, clearly displays the coexistence of crystal mush–extracted melts and complementary silicic cumulates, as well as a mixture of them in an upper-crustal profile. Thus, the studied Nyemo complete silicic magmatic system indicates that in situ differentiation via crystal-melt separation is an effective mechanism for the generation of high-silica granitic melt and that a mush reservoir can occur at an unusually shallow crustal depth.

Careful and constructive reviews by Olivier Bachmann, Chad D. Deering, and an anonymous reviewer are gratefully acknowledged. We thank Urs Schaltegger for constructive comments and handling of the manuscript. We are also grateful to Helene Brätz for help with the LA-ICP-MS zircon trace element analyses. This study was supported by the National Natural Science Foundation of China (grants 42172070 and 41772060). This publication is a contribution to International Geoscience Programme (IGCP) 662.

1Supplemental Material. Analytical methods, modeling setup for magmatic evolution, and data. Please visit https://doi.org/10.1130/GEOL.S.16814911 to access the supplemental material, and contact editing@geosociety.org with any questions.
Gold Open Access: This paper is published under the terms of the CC-BY license.