K‐Rich Adakite‐Like Rocks in Central Tibet: Fractional Crystallization of a Hydrous, Alkaline Primitive Melt

K‐rich adakite‐like rocks (KARs) in post‐collisional settings, such as in Tibet, have been widely linked with the melting of pre‐existing thickened crust. Here, we investigate geochemical data of the Late Eocene (38–34 Ma) volcanic rocks including KARs from the southern Qiangtang terrane (SQT) of central Tibet. The data reveal that: (a) the volcanic rocks define a fractionation trend from high‐K alkaline basalt to high‐K calc‐alkaline rhyolite, with a continuous compositional range and (b) they are characterized by a narrow range of depleted Sr–Nd isotopic compositions relative to the pre‐Eocene SQT crust. We contend that the KARs in the SQT resulted from fractional crystallization of hydrous, alkaline melts derived from the lithospheric mantle where fractionation was dominated by amphibole and plagioclase. Partial melting of the lithospheric mantle beneath the SQT was possibly triggered by thermal perturbations owing to the north‐directed subduction of the Indian continental lithosphere beneath southern Tibet.

• Late Eocene K-rich adakite-like rocks (KARs) in the S. Qiangtang terrane resulted from fractional crystallization of a primitive alkaline melt • Fractionation was dominated by amphibole and plagioclase • KARs from the S. Qiantang terrane have a different origin from those of the Lhasa terrane

Supporting Information:
Supporting Information may be found in the online version of this article.  Chung et al., 2003;Hou et al., 2004;Q. Wang et al., 2005Q. Wang et al., , 2008, hybridization of crust-derived adakitic magmas with mantle-derived shoshonitic melts (e.g., Hou et al., 2004;R. Wang et al., 2018;X. Wang et al., 2019;Yang et al., 2015), fractional crystallization of medium-K basaltic melts represented by mafic microgranular enclaves , and melting of intermediate to felsic arc rocks in the stability field of garnet (Yi et al., 2022). Most of the models essentially link the KARs with the melting of pre-existing crustal rocks, and limited attention has been paid to the possibility that they result from fractional crystallization of mantle-derived, high-K basaltic melts.
In this study, we investigate the geochemistry of the Late Eocene (38-34 Ma) volcanic rocks from the southern Qiangtang terrane (SQT) of central Tibet. These rocks form a suite of high-K and high Sr/Y magmatic rocks that range from basalts to rhyolites, and most of the evolved rocks (SiO 2 ≥ 56 wt.%) are KARs. We conclude that the KARs were derived from fractionation of lithospheric mantle-derived, hydrous, high-K basaltic magmas. Our study demonstrates that not all KARs in Tibet are derived from crustal melting, and that fractional crystallization of high-K basaltic alkaline melts is an alternative source of KARs.

Geological Setting and Samples
The Tibetan Plateau is a collage of several continental fragments, and the latest amalgamation occurred when India collided with Asia at ∼55 Ma (e.g., Chung et al., 2005;Zhu et al., 2015). From south to north, it consists of the Himalaya, Lhasa terrane, Qiangtang terrane, and Songpan-Ganzi complex ( Figure 1a, Yin & Harrison, 2000). The Qiangtang terrane, located in central Tibet, is bounded by the Jinsha suture to the north and the Bangong-Nujiang suture to the south (Figure 1a). It can be further divided into northern and southern Qiangtang terranes (NQT and SQT) by a >600 km long belt of blueschist-and eclogite-bearing ophiolitic mélange ( Figure 1a). This mélange belt was interpreted to either mark the trace of the Longmu Co-Shuanghu Tethys Ocean that closed in early Mesozoic or represent rocks that have been underthrust southward beneath the NQT from the Jinsha suture during an early Mesozoic flat-slab subduction (e.g., Kapp & DeCelles, 2019). The SQT includes an ancient basement comprising the newly found Early Paleozoic sediment-derived granites ( Figure 1b, Dan et al., 2020;Hu et al., 2015). Before the India-Asia collision, north-directed subduction of the Bangong Tethyan lithosphere beneath the SQT and subsequent SQT-Lhasa collision resulted in the formation of abundant Jurassic to Cretaceous magmatic rocks (e.g., Hao et al., 2016Hao et al., , 2019Ke et al., 2021;S. M. Li et al., 2014. Magmatic rocks that post-date the India-Asia collision are widespread in Tibet, and define two stages. The early stage (55-25 Ma) is located in central Tibet (northern and southern Qiangtang terranes), and the late stage is distributed in southern Tibet (Lhasa terrane; 25-8 Ma) and northern (Songpan-Ganzi terrane; 28-0 Ma) Tibet ( Figure 1a, Chung et al., 2005;Guo & Wilson, 2019). This study focuses on the Late Eocene volcanic rocks from the early stage in the western segment of the SQT. We investigate newly sampled rocks from the Mendangle volcanic field and merge the new data with data from the literature on the Nading Co, Zougouyoucha Co, Jianshan, Yibuchaka, and Ejumaima volcanic fields ( Figure 1b and Table S1 in Supporting Information S1). The Mendangle volcanic field is located within the Bangong-Nujiang ophiolitic mélange zone (Figure 1b, L. Q. Wang et al., 2013;Zhu et al., 2016), which separates the SQT from the Lhasa terrane. The newly sampled rocks are considered to be part of the broader Late Eocene volcanic rocks in the SQT, as we show below, they were also erupted in the Late Eocene. Taken together, these volcanic rocks extend discontinuously for more than 100 km in the north-south direction and 200 km in the east-west direction ( Figure 1b).

Age of the Volcanic Rocks
One trachy-dacite (GZ12-2-7) and one rhyolite (GZ12-1-6) sample from the Mendangle region yield consistent weighted average 207 Pb-corrected 206 Pb/ 238 U ages within error (38.1 ± 0.5 and 37.5 ± 0.7 Ma; Figure S3 in Supporting Information S1). All of the analyzed zircons show prismatic to irregular crystal forms with oscillatory zoning and no inherited cores ( Figure S3 in Supporting Information S1). These features, together with their high Th/U ratios (0.58-1.60), indicate that they are magmatic zircons, and thus we interpret the Mendangle volcanic rocks were erupted at ca. 38 Ma. Amphibole phenocrysts from two Nading Co trachy-basalt samples (2003T372 and 2003T374) yield plateau ages of 36.2 ± 0.1 and 35.9 ± 0.3 Ma, respectively (Ding et al., 2007; Figure S4 in Supporting Information S1). These two ages are well defined over ≥80% cumulative 39 Ar released and statistically indistinguishable from their corresponding inverse isochron ages (Table S6 in Supporting Information S1), and thus they were interpreted to represent eruption ages of the samples (Ding et al., 2007). In addition, previous Ar-Ar studies on sanidine and zircon U-Pb dating of trachy-andesite to trachy-dacite samples suggest that intermediate to felsic volcanic rocks from the Zougouyoucha Co, Nading Co, Jianshan, Yibuchaka, and Ejumaima regions crystallized at 37-34 Ma (Table S1 in Supporting Information S1). Taken together, this suite of volcanic rocks is nearly contemporaneous and crystallized in the Late Eocene between 38 and 34 Ma.
The volcanic rocks exhibit near-parallel chondrite-normalized rare-earth element (REE) patterns and are enriched in LREE and Pb, and depleted in HREE, Nb, Ta, Ti, and P ( Figure S5 in Supporting Information S1  Figures 3g and 3h), the others display no obvious or slightly negative Eu anomalies (0.75-1.07), and define a negative trend for P 2 O 5 . See also Harker diagrams for Y, Zr, Ni, and Sc ( Figure S6 in Supporting Information S1).

Whole-Rock Sr-Nd and Zircon O Isotopes
The volcanic rocks have a much narrower range of depleted Sr-Nd isotopic compositions compared to the pre-Eocene SQT crust (the blue field in Figure 4a), with ( 87 Sr/ 86 Sr) t = 38Ma ratios from 0.70421 to 0.70715, and ε Nd(t = 38Ma) values from +1.79 to −2.58 (Figure 4a). These ranges are independent of SiO 2 contents (Figure 4b).

Petrogenesis of the KARs in the Southern Qiangtang Terrane
Among the Late Eocene volcanic samples from the SQT, samples with the highest Sr/Y ratios (white triangles in Figure 2c) are characterized by significantly positive Eu anomalies and low P 2 O 5 (Figures 3g and 3h). This is probably caused by their high plagioclase and low apatite contents, and thus they cannot represent melts and are excluded from the following discussion. In contrast, the other samples display no obvious geochemical anomaly and are considered to represent melts.
Given the present >50 km crustal thickness of Tibetan crust and the absence of subducted oceanic crust beneath Tibet since the India-Lhasa collision at ∼55 Ma, it was proposed by Chung et al. (2003) that the Miocene KARs ( Figure 2) in the Lhasa terrane resulted from partial melting of eclogites and/or garnet amphibolites in the lower part of the Tibetan crust. This and other models that essentially argue for the melting of crustal rocks were then widely used to account for the formation of the Tibetan KARs (see the "Introduction" section), including those in the SQT (Y. C. Zeng et al., 2021).
As described above, the Late Eocene volcanic rocks from the SQT, that include KARs, define a continuous spectrum of compositions, with a relatively narrow range of whole-rock Sr-Nd compositions (Figures 2, 4a, and 4b). These features suggest that they are petrogenetically related and were formed through either: (a) fractional crystallization of a common primitive melt, or (b) mixing of isotopically similar mantle-derived mafic and felsic magmas (Anderson, 1976). Magma mixing is expected to define linear arrays linking end-members in elemental variation diagrams (Keller et al., 2015). However, this is not the case for the Late Eocene volcanic rocks: (a) MgO and K 2 O define curvilinear trends with SiO 2 (Figures 2b and 3a), and (b) Al 2 O 3 and Sr first increase and then decrease with increasing SiO 2 (Figures 3e and 3f), patterns typically caused by feldspar fractionation. Similarly, the plot of Ti/Y versus Rb/Y does not define the linear trend that would be expected from magma mixing (Figure 4d). These observations indicate that the KARs are not likely products of mixing between mantle-derived mafic magmas and crust-derived felsic magmas, consistent with the absence of disequilibrium  Bas et al., 1986); the solid line is the boundary between alkaline and subalkaline series (Irvine & Baragar, 1971). (b) K 2 O versus SiO 2 diagram (Peccerillo & Taylor, 1976). The modeled liquid line of descent of the Dariv Igneous Complex, western Mongolia is from Bucholz et al. (2014). (c) Sr/Y versus Y diagram (Defant & Drummond, 1990). Adakites derived from slab melting are from Defant et al. (1991), S. M. Kay et al. (1993), Stern and Kilian (1996), Saunders et al. (1987), Morris (1995) Fujimaki et al. (1984) and Johnson (1994). Symbols are as in Figure 2.
textures in intermediate KARs (Y. C. Zeng et al., 2021). As shown in Figure 1, the Late Eocene KARs from the western SQT are spread over an area of over 100 × 200 km. The western SQT crust is made up of a basement with a wide range of isotopic compositions, with ( 87 Sr/ 86 Sr) t = 38Ma ratios ranging from 0.70475 to >0.73370, and εNd (t = 38Ma) values ranging from +2 to <−12.8 (Figure 4a). This contrasts with the KARs that are characterized by depleted Sr-Nd isotopic compositions covering a relatively small range (Figures 4a and 4b). This narrower range supports the inference that these rocks are not direct products of partial melting of the western SQT crust and instead result from fractional crystallization of basaltic magmas, represented by the coeval trachy-basalts, either derived from an isotopically heterogeneous mantle or with a small and variable input from the heterogeneous SQT crust during fractional crystallization.

Fractional Crystallization Process
The negative MgO trend and the positive trend in Al 2 O 3 at low SiO 2 in Figures 3a and 3e indicate the early fractionation of Al-poor mafic phases, such as clinopyroxene and olivine. As the silica content rises above 55 wt.% SiO 2 , there are decreases in both Al 2 O 3 and Sr (Figures 3e and 3f), corresponding to saturation and fractionation of plagioclase. The delay in plagioclase fractionation is also reflected in Figure 3g, where the Eu anomaly shows an increase followed by a decrease in values with increasing SiO 2 once plagioclase starts to fractionate. The Zr diagram follows a similar curved pattern ( Figure S6 in Supporting Information S1) and records zircon saturation followed by its fractionation in rocks with >60 wt.% SiO 2 . The continuous decrease in P 2 O 5 with increased SiO 2 suggests fractionation of apatite ( Figure 3h). Fractionation of zircon and apatite is also reflected in the plot of Y which shows a decrease with increasing SiO 2 ( Figure S6 in Supporting Information S1).
As demonstrated by Rhyolite-MELTs modeling (Gualda et al., 2012), dry and hydrous parent melts with the average composition of the studied trachy-basalts cannot produce evolved melts with SiO 2 > 65 wt.% by fractional crystallization of only olivine, pyroxene, plagioclase, apatite, and spinel at different pressures (Figures 3a-3d). This means that other low-SiO 2 phases such as amphibole, biotite, or garnet must have been involved in the fractionation process. Garnet preferentially incorporates heavy REE over middle REE, with a low D Dy/Yb value, and it would result in increasing Dy/Yb in the melt with increasing SiO 2 (Davidson et al., 2007;Macpherson et al., 2006). However, the volcanic rocks show near-parallel chondrite-normalized REE patterns ( Figure S5 in Supporting Information S1) and a gentle slope of Dy/Yb with SiO 2 (Figure 4e). This, together with the absence of garnet in the volcanic rocks, suggests that the fractionation did not involve garnet (Table S1 in Supporting Information S1). Instead, amphibole and biotite have been found in some trachy-basalts and trachy-andesites ( Figure  S1 in Supporting Information S1, Ding et al., 2007;Guo & Wilson, 2019;B. D. Wang et al., 2010), and could therefore have played an important role in silica enrichment, consistent with the negative correlation between SiO 2 and TiO 2 (Figure 3b). The important role of amphibole fractionation is also reflected in the decreasing trends of Dy/Yb and Dy/Dy* with differentiation (Figures 4e and 4f, Davidson et al., 2007Davidson et al., , 2013. The likely reason why most of the intermediate to felsic lavas lack amphibole phenocrysts is that amphibole is unstable and resorbed rapidly near the surface (Davidson et al., 2007;Rutherford & Hill, 1993).
In summary, the data suggest that the fractionation process of the Late Eocene volcanic rocks in the SQT was dominated by amphibole and plagioclase, with no garnet involvement. These volcanic rocks are similar in both fractionation processes and major-element trends (Figures 2 and 3) to those of the Dariv Igneous Complex, western Mongolia (Bucholz et al., 2014), a rare example of fractionation of hydrous, alkaline primitive magmas. Trace element modeling suggests that more than two units of trachy-basalts are required to produce one unit of trachy-andesites by amphibole-dominated fractional crystallization (Figure 4f). This ratio does not match the lower proportion of trachy-basalts than trachy-andesites in the studied volcanic field (volume ratio < 1). A possible explanation is that the exposed volumes of volcanic rocks may not represent their true proportions and more voluminous Late Eocene mafic intrusive rocks and associated cumulates exist at depth in the SQT.

Mantle Source of the Primitive Alkaline Melt
The low SiO 2 (48.5-52.0 wt.%) and high MgO (5.01-7.54 wt.%) contents of the SQT trachy-basalts indicate that they are mantle-derived rocks, but they are unlikely to represent primitive melts, because their low Mg# (52-60), Ni (73-110 ppm) and Cr (119-184 ppm) indicate crystal fractionation of olivine and pyroxene. The Sr-Nd isotopic compositions of the studied volcanic rocks are more enriched than that of mid-ocean-ridge basalts, and show a relatively narrow range compared to the local crustal rocks (Figures 4a and 4b), suggesting either an enriched heterogeneous mantle source or a small and variable crustal input during fractional crystallization. Either of these interpretations is consistent with the high δ 18 O values of zircons from the Mendangle volcanic rocks (Figure 4c, Kemp et al., 2007). Furthermore, the presence of amphibole and biotite in the trachy-basalts implies that the mantle source was hydrous (Guo & Wilson, 2019;Sisson & Grove, 1993). Given the absence of subducted oceanic crust beneath the SQT during the Late Eocene (Kapp & DeCelles, 2019), and the Sr-Nd isotopic similarities between the volcanic rocks and the ∼106 Ma lithospheric mantle-derived arc basalts in the western SQT (Figure 4a, Qi et al., 2021), we suggest that the enriched mantle is most likely the sub-continental lithospheric mantle. This mantle would have been metasomatized by slab-derived fluids and melts during earlier oceanic subduction, and remained chemically unmodified after the SQT-Lhasa collision.

Tectonic Implications
Two main geodynamic models have been proposed for the formation of the Late Eocene magmatism in the SQT: (a) intracontinental subduction (Ding et al., 2007;Guo & Wilson, 2019) and (b) removal of lithospheric mantle (e.g., Qi et al., 2021;Y. C. Zeng et al., 2021). The first model suggests that the north-directed subduction of the Indian or Lhasa continental lithosphere beneath Tibet caused perturbations of the asthenospheric flow, triggering melting of the lithospheric mantle beneath the SQT. The second model describes an extensional setting, where much of the lithospheric mantle was removed but the remaining lithospheric mantle beneath the SQT underwent melting owing to asthenospheric upwelling. Although the present study cannot provide robust constraints, extreme fractional crystallization of mantle-derived melts tends to be favored by compressive regimes, such as in the Mesozoic Gangdese Arc (e.g., Xu et al., 2019). We therefore favor the first model of thermal perturbations driven by the north-directed subduction of the Indian continental lithosphere (e.g., Nábělek et al., 2009).

Data Availability Statement
Supporting data of this study are freely available at https://doi.org/10.17605/OSF.IO/8WS69.