Lithosphere beneath the Evolving Tianshan Orogen: Constraints from Xenoliths

The Tianshan mountain range is a complex Paleozoic orogen composed of composite terrains that stretches east–west for over 2500 km across central Asia (Figure 1) [1]. The orogen is overlain by Mesozoic to Cenozoic intracratonic basins and has been affected by neotectonic activity [1–5]. Researchers have largely focused on the Paleozoic collisional history and subsequent basinal activities with little work on the nature of the lithospheric mantle beneath the region. Previous investigations of the composition of mantle xenoliths suggest the mantle beneath the Kastek in the North Tianshan is lherzolite, which is similar to those later found in the vicinities of Tuoyun and Ortosuu in the South Tianshan [6–8]. In contrast, the mantle beneath the Uchkuduk in the North Tianshan has a predominantly pyroxenite composition [7]. The aforementioned information indicated that the local mantle is laterally heterogeneous [7]. Although the Tianshan has largely experienced magmatic quiescence since late Permian, the Tuoyun basin and the central part of the Kyrgyz-Tianshan (Figure 1) preserve a volumetrically small series of basaltic rocks erupted primarily in the late Mesozoic-Paleogene [5]. Recent paleogeothermal studies using electrical conductivity indicate that the position of the Moho discontinuity beneath the Tianshan during the Cretaceous-Paleogene was at a depth of $35±5km$⁠, which corresponds to the heat flux 80–85 mW m⁻² [6]. Thus, the thickness of the crust beneath Tianshan is ∼35 km. This geologic time (70–60 Ma) is linked to plume activity and basalt volcanism in the region [9]. In this case, the lithosphere beneath Tianshan was likely hotter, weaker, and thinner than the present time. The temperature of the Moho boundary was about $800±50°C$⁠, which is higher than the temperature observed in the region today of about 720–750°C [6]. Comparatively, the current depth of the Moho boundary is $55±5km$⁠, which corresponds to a geotherm with a heat flux of 50–60 mW m⁻² [10]. The modern crustal thickness of Tianshan ranges from 45 to 75 km [11]. The current topography is associated with the India-Eurasia collision, mainly resulting from uplift in late Cenozoic time (<35 Ma) in the region. However, the nature of the lithospheric mantle beneath the Tianshan remains controversial [5, 8, 9]. Mantle-derived xenoliths in alkali basalts were collected from the Ortosuu and Uchkuduk regions of the Kyrgyz-Tianshan and provide a unique insight into the lithospheric upper mantle beneath the region. In this paper, we present a systematic study of these xenoliths, including petrographic and mineral chemistry. We use this comprehensive dataset to characterize the nature of the lithospheric mantle beneath the Tianshan orogenic belt.

The Tianshan orogenic belt is located in northwestern China, Kyrgyzstan, the southern part of Kazakhstan, Uzbekistan, and Tajikistan (Figure 1). Tectonically, it can be divided into three subunits: the North-, Middle-, and South Tianshan [12, 13] (Figure 1). The North Tianshan consists of Precambrian continental fragments, early Paleozoic ophiolites, and high-grade metamorphic rocks [14]. The basement of the North Tianshan is extensively intruded by early Paleozoic granitoids and late Paleozoic postcollisional plutons [15–17]. The North Tianshan and Middle Tianshan are separated by the Nikolaev Line (Figure 1), which is considered an early to middle Paleozoic ophiolitic suture [18]. The major part of the Middle Tianshan is characterized by middle Paleozoic passive margin sedimentary facies and a lack of early Paleozoic granitoids [19, 20]. Neoproterozoic granites and felsic volcanic rocks are common throughout the Middle Tianshan [14]. The Precambrian basement of the Middle Tianshan is overlain unconformably by late Neoproterozoic rift-related subalkaline basalts, rhyolites, sandstone, shales, cherts, carbonates, and turbidites [20]. The ophiolite-bearing Atbashi-Inylchek suture separates the Middle Tianshan from the South Tianshan [21, 22] (Figure 1). The South Tianshan is a late Paleozoic accretionary and collisional thrust-and-fold belt [23]. The belt consists mainly of middle to late Paleozoic shallow- and deep-marine sedimentary rocks, fragments of ophiolites, and metamorphic rocks [24]. Late Paleozoic postcollisional plutons occur within the South Tianshan [19, 25–27].

The Kyrgyz-Tianshan basalt province covers an area over 285,000 km² [6]. Previous studies have documented $40Ar/39Ar$ ages for the Tuoyun and Tekelik basalts of ~113 Ma and ~76–71 Ma, respectively [5, 9]. $40Ar/39Ar$ ages of the basalts yield ages of 76–61 Ma [9] are interpreted to represent within-plate magmatic systems related to mantle plume sources based on their geochemical characteristics [9]. Xenoliths of ultrabasic rocks occur in the basalts at several localities: Tuoyun, Uchkuduk, Bailamtal, and Kastek [5, 9]. In this study, we investigated mantle xenoliths sampled from the Ortosuu (N40° 27$′$ 15.20$″$⁠, E75° 49$′$ 24.90$″$⁠) and Uchkuduk (N42° 16$′$ 43.10$″$⁠, E76° 06$′$ 58.20$″$⁠) basalt sites (Figure 1). The Ortosuu site is situated near the Ortosuu River, in the South Tianshan where basalt intrudes Carboniferous limestones and shales [6]. In comparison to the neighboring Tuoyun and Tekelik sites, the $40Ar/39Ar$ age of Ortosuu basalts was estimated as 75–70 Ma [6]. The Uchkuduk site is located on the southwest of Issyk-Kul Lake in the North Tianshan, and the basalts cut Carboniferous sedimentary deposits [9]. A Rb-Sr isotopic age of ~50 Ma has been reported from the Uchkuduk basanites [28].

A suite of 18 peridotites, 3 pyroxenites, and 1 granulite xenoliths were sampled from the Ortosuu basalts, and 1 peridotite xenolith was collected from the Uchkuduk basalt outcrops for petrographic and mineral chemistry studies (Figure 2(a)). The host lavas are basanites and tephrites according to the IUGS classification [29]. The collected mantle xenoliths from the Ortosuu site are subangular to rounded and range from 3 to 7 cm in length (Figure 2(b)). Ortosuu peridotite xenoliths are spinel lherzolites, spinel wehrlites, wehrlites, and harzburgites (Figures 2(c), 2(d), and 3(a)). The spinel-bearing lherzolites commonly exhibit a protogranular texture (Figure 2(c)). Their modal compositions vary from 5% to 21% clinopyroxene, from 47% to 74.5% olivine, from 18% to 45% orthopyroxene, and 0.5% to 3% spinel. In addition to the peridotite xenoliths, pyroxenite, and granulite xenoliths are observed in the Ortosuu basalts. The composition of pyroxenite xenoliths varies from clinopyroxenite to olivine websterite (Figures 2(e) and 3(a)). Granulite xenoliths from the Ortosuu basalt site are garnet-bearing granulites. They are composed primarily of plagioclase, clinopyroxene, and garnet. Textures of garnet-bearing granulite are essentially fine- to medium-grained granoblastic. Uchkuduk peridotite xenoliths are spinel-bearing wehrlites (Figures 2(f) and 3(a)). They are composed predominantly of olivine and clinopyroxene and are orthopyroxene-poor, with accessory spinel. Like the Ortosuu peridotites, the wehrlites from this site have textures ranging from coarse granular to granuloblastic.

Mineral major element analyses were carried out at the Institute of Earth Sciences, Academia Sinica (IESAS), using a thermal emission (W-filament) electron probe micro analyzer (JEOL W-EPMA: JXA8900-R) with four channeled wavelength dispersive X-ray spectrometers (WDS). The accelerating voltage was 15 kV; a sample current of 12 nA and a beam diameter of 2 μm were used. Backscattered electron images were used to guide the analysis on target positions of minerals. Natural and synthetic standard minerals were used for calibration, and raw data were corrected online using the ZAF correction program. The measured X-ray intensities were corrected by the ZAF method using the standard calibration of synthetic (s) and natural (n) chemical-known standard minerals with various diffracting crystals in parenthesis, as follows: wollastonite for Si (TAP) and Ca (PET), rutile for Ti (PET), corundum for Al (TAP), chromium oxide for Cr (PET), fayalite for Fe (LiF), tephroite for Mn (PET), periclase for Mg (TAP), albite for Na (TAP), and adularia for K (PET). Peak counting for each element and for both upper and lower baselines is 10 and 5 s, respectively. Standards run as unknowns yielded major oxide relative standard deviations for Si, Na, and K of less than 1%, and less than 0.5% for other elements. Detection limits, based on $3σ$ of the standard calibration, were less than 600 ppm for each oxide.

Concentrations of rare earth elements (REEs) and other trace elements (Ba, Rb, Th, U, Nb, Ta, Pb, Sr, Zr, Hf, Ti, Y, Sc, V, Co, and Ni) in clinopyroxene were determined by laser ablation-inductively coupled plasma mass spectrometry (LA-ICP-MS), also at IESAS. The Agilent 7900 quadrupole ICP-MS instrument was coupled to a Photon-Machines Analyte G2 193 nm excimer laser. A beam diameter of 65 μm and a laser repetition rate of 5 Hz were used. The NIST 612 glass standard was used to calibrate relative element sensitivities. Each analysis was normalized using CaO values determined by electron microprobe. Typical theoretical detection limits are 10–20 ppb for REEs, Ba, Rb, Th, Sr, Zr, and Y; 100 ppb for Sc and V; and 2 ppm for Ti and Ni. The relative precision and accuracy for a laser analysis is 1–10%. Data reduction was performed using the software GLITTER [30].

Spinel lherzolite xenoliths are characterized by olivine (Ol) + orthopyroxene (Opx) + clinopyroxene (Cpx) + spinel (Spl) assemblages. All constituent phases are compositionally homogeneous and do not show core-to-rim chemical zonations. Major element compositions of minerals in the studied samples are variable (Table S1). The Fo content (Mg#, $Mg/Mg+Fe×100$⁠) of olivine in lherzolite samples ranges from 87 to 91, and the NiO content is up to 0.4 wt%. Olivines in the lherzolites are characterized by low CaO (0.03–0.1 wt%) content. Orthopyroxenes from Ortosuu spinel lherzolites are aluminum enstatites (Figure 3(b)) [31], which have high Mg# (87–91) and Cr₂O₃ (0.02–0.63 wt%) and low TiO₂ (0.02–0.28 wt%) contents. CaO and Al₂O₃ contents vary from 0.46 to 1.09 wt% and from 2.87 to 5.6 wt%, respectively. Clinopyroxenes from Ortosuu spinel lherzolites are aluminum chromium augites (Figure 3(b)) [31]. They can be divided into two groups based on their compositions. Group one consists of primitive clinopyroxenes, with Mg# (86–90), and Al₂O₃ (6.46–7.67 wt%) and TiO₂ (0.46–0.93 wt%) contents, whereas group two is made up of depleted clinopyroxenes, which have a high Mg# 91–92, but low contents of Al₂O₃ (3.48–4.5 wt%) and TiO₂ (0.08–0.37 wt%) (Table S1). Spinel crystals occur in the lherzolites and display Mg# values ranging from 71 to 76, and Cr# values varying from 9 to 24 (Table S1).

Wehrlite xenoliths are mostly composed of olivine and clinopyroxene with little or no orthopyroxene. Olivine compositions are more uniform in Uchkuduk peridotites (Mg# 90.5) than in Ortosuu wehrlites (Mg# 74–91). The lowest Mg# value is for olivine from wehrlite sample OS1710. Clinopyroxene displays a similar distribution (Table S1). Samples OS1729 and OS1731 are harzburgite xenoliths, which are constituted by $Ol+Opx±Cpx$⁠. The Mg# of olivine is homogeneous (~91) for Ortosuu harzburgites but shows a wider range for lherzolite (⁠$Mg#=87–91$⁠) and wehrlite (⁠$Mg#=74–91$⁠) samples (Table S1).

Sample OS1734A is orthopyroxene-bearing clinopyroxenite (modal clinopyroxene 95%, orthopyroxene 5%), whereas sample OS1734B is composed entirely of clinopyroxenes (100% modal). These samples show compositional similarities in clinopyroxenes (Table S1). Sample OS1716 is equigranular Ol-websterite with $Ol+Opx+Cpx+Spl$⁠. The Mg# of olivine crystals in this sample is 89.5. Its CaO content is rather low (0.11 wt%). Orthopyroxenes from Ol-websterites are aluminum enstatites (Figure 3(b)) [31], with Mg# 89.7, Cr₂O₃ (0.37 wt%), and low TiO₂ (0.16 wt%) contents. Clinopyroxenes in Ol-websterites are aluminum chromium augites (Figure 3(b)) [31], which are characterized by Mg# 88.7, 7.72 wt% Al₂O₃, and 0.6 wt% TiO₂. This sample contains spinel in which Mg# is 76.9, and Cr# (⁠$Cr/Cr+Al×100$⁠) is 8.73 (Table S1).

Granulite xenoliths are composed primarily of garnet, clinopyroxene, and plagioclase. Garnet from the sampled granulite xenolith (OS1736, Table S1) has Mg# 68, and with a composition of pyrope (Py₅₉), almandine (Alm₂₈), grossular (Gro₁₃), and spessartite (Sp_0.6). Clinopyroxenes in Ortosuu granulite are aluminum augite [31], which have low chromium (0.02 wt%) contents, Mg# 81, and relatively high contents of Al₂O₃ (7.63 wt%) and Na₂O (1.74 wt%). Plagioclases are homogeneous, with andesine (An₃₇).

The spinel-group end-members have been determined by using the application End-Members Generator (EMG) [32], and the results are listed in Table S2. Microchemical compositions of SiO₂, TiO₂, Al2O₃, Cr₂O₃, FeO, MnO, MgO, ZnO, and NiO were analyzed by electron microprobe on spinel-group minerals. The results showed that based on electron microprobe analyses (EMPA), the classification of the spinel-group minerals can be divided into 16 end-members of the spinel group: MgAl₂O₄ (spinel sensu stricto (s.s.)), FeAl₂O₄ (hercynite), MnAl₂O₄ (galaxite), ZnAl₂O₄ (gahnite), MgFe₂O₄ (magnesioferrite), Fe₃O₄ (magnetite), MnFe₂O₄ (jacobsite), ZnFe₂O₄ (franklinite), NiFe₂O₄ (trevorite), MgCr₂O₄ (magnesiochromite), FeCr₂O₄ (chromite), MnCr₂O₄ (manganochromite), ZnCr₂O₄ (zincochromite), NiCr₂O₄(nichromite), Mg₂TiO₄ (qandilite), and Fe₂TiO₄ (ulvöspinel). The major end-member compositions of the spinel-group minerals from the studied xenoliths are MgAl₂O₄, FeAl₂O₄, and MgCr₂O₄. The chemical analysis of the composition of the spinel-group minerals showed that they are spinel sensu stricto (MgAl₂O₄) with Fe contents (Mg : Fe atomic ratio about 4 : 1).

Clinopyroxene trace element compositions are thought to be a useful indicator of specific mantle characteristics and of chemical modifications in the mantle [6]. Their compositions are reported in Table S3. Two groups of clinopyroxene can be distinguished on the basis of their composition and relationship with Mg# (Figure 3; Tables S1 and S3). Some of the group one clinopyroxenes have similar compositions to primitive mantle, which are characterized by flat or slightly enriched LREEs (Figure 4(a)). These clinopyroxenes display negative Ba and Ti anomalies in spidergram plots (Figure 4(b)). Group two (2 and 2$∗$⁠) clinopyroxenes are more depleted in HREEs than in group one. These more depleted samples possess clinopyroxenes with LREE-enriched patterns, which range from spoon-shaped (2) to LREE/MREE-enriched convex-upward (2$∗$⁠) (Figures 4(c) and 4(e)). Clinopyroxenes of this group are highly enriched in Th and U but depleted in Nb, Zr, Hf, and Ti (Figures 3(d) and 3(f)). It is noteworthy that sample OS1710 consists of wehrlite from the Ortosuu basalt with a composition that varies from primitive to slightly depleted (Figure 4(a)). Two samples (OS1703, OS1705) in group 2 have light REE-depleted patterns that might be due to the low-concentration of the elements (Figure 4(c)).

Equilibrium temperatures and pressures obtained using the geothermometers and geobarometers are reported in Table S4. Temperatures obtained using the Ca-in orthopyroxene thermometer [33] at a pressure of 1.5 GPa fall within the range of 882–1080°C for the peridotite xenoliths and 977–1055°C for the pyroxenite xenoliths. The two-pyroxene thermometer of Wells [34] provides similar temperature ranges, 870–1089°C for the peridotite xenoliths and 910–1057°C for the pyroxenite xenoliths. Estimated equilibration temperatures based on the Cr-Al-Opx thermometer [35] range from 860°C to 1098°C for the peridotite xenoliths and from 886°C to 1025°C for the pyroxenite xenoliths. The above three thermometers give similar temperature estimates (Table S4). Equilibration temperatures obtained from the peridotite xenoliths using the spinel-orthopyroxene thermometers [35, 36] are within the range 841–1130°C and 832–1091°C, respectively. Equilibration pressures were calculated using the clinopyroxene geobarometer of Nimis [37] in the range of 1.3 to 1.5 GPa for group 1 and 0.4–0.8 GPa for group 2 lherzolites. The estimated temperatures of garnet granulite are about 772–927°C, according to the garnet-clinopyroxene geothermometers [38–40] at a pressure of 1.5 GPa. Equilibration pressures for garnet granulite are about 0.4–1.2 GPa, calculated from the garnet-clinopyroxene [41] and clinopyroxene barometers [37]. The barometer of Beyer et al. [41] can most likely be more reliably applied to garnet-clinopyroxene assemblages. The pressure is similar to those of mafic lower crustal granulites from elsewhere in the southwestern Tianshan [42] and is therefore adapted for this study.

The studied spinel lherzolites and wehrlites from the Ortosuu site can be divided into two groups based on their REE patterns in clinopyroxenes and the P-T estimates. Group 1 spinel lherzolites have higher equilibration temperatures, 1012–1130°C, and higher pressures of equilibration, 1.3–1.5 GPa. They show near-flat REE distribution patterns, which are close to those of the primitive mantle (Figure 4(a); Table S4). In contrast, group 2 is characterized by enriched LREE patterns and lower equilibration temperatures, ranging from 832°C to 932°C, and pressures, about 0.4–0.8 GPa. Likewise, group 1 wehrlites have higher equilibration temperatures and pressures than group 2 (Table S4). Uchkuduk spinel wehrlite and Ortosuu spinel wehrlite of group 2 display opposite trends in REEs, but their P-T range of equilibration is similar (Figure 4(a); Supplementary Table S4).

The Cr# of spinel can be used to estimate the degree of partial melting [43]. The spinel lherzolites found in basalts from the Ortosuu sites are suggested to be from the uppermost lithospheric mantle beneath Tianshan [6]. Modeling of partial melting for the Ortosuu upper mantle was adapted using the method of Hellebrand et al. [43]. The results indicate that the xenoliths with group 1 lherzolites contain spinels with Cr# 0.09–0.20 and correspond to a 0–8% degree of partial melting, whereas group 2 lherzolites contain spinels with Cr# 0.16–0.26, and they have experienced a higher degree of partial melting, about 6–11%. The estimated degrees of melting using Y and Yb contents in clinopyroxene [44] and the primitive mantle composition of Sun and McDonough [45] provide a range of 5–10% for group 2 and 3–5% for group 1 (Figure 5(a)). The modeled degree of partial melting for the Ortosuu xenolith estimations is 5–10%, which is similar to the Tuoyun mantle xenoliths (about 5–8% melting according to Zheng et al. [8]). Comparatively, the xenoliths from the Ortosuu and Tuoyun sites are less refractory than many of those regions in the Eastern Central Asia Orogenic Belt, North China Craton, and Eastern Siberia (Figure 5(a)) [46–48].

The LREE- and MREE-enriched patterns in Cpx and the high LILE abundances in Cpx have been associated with metasomatism by carbonatitic or silicate melts and/or H₂O–CO₂ fluids [49–51]. The spinel peridotites in this study have $CaO/Al2O3$ ratios larger than 1, indicating strong Ca-metasomatism. Experimental studies have shown that metasomatic alteration by silicate metasomatism may cause enrichment in incompatible elements (e.g., LILE, LREE, and HFSE) [52], whereas carbonatitic metasomatism would have been capable of transporting large amounts of LILE and LREE but limited HFSE [53]. However, the nature of the metasomatic agent is difficult to constrain and could be produced by any SiO₂-undersaturated melts, such as silicate melts or carbonatite melts [54]. Group two clinopyroxene from the studied xenoliths has a wide range of $Ti/Eu$ (688–9,424) and $La/Ybn$ (0.03–19.1), which may imply an agent transitional between carbonatitic and/or silicate metasomatism (Figure 5(b)) [55]. This result is consistent with the findings of Zheng et al. [8] who reported that the Tuoyun clinopyroxene has large ranges in $Ti/Eu$ (971–5,765) and $La/Ybn$ (0.19–6.33). Furthermore, it is evident that the backscattered electron (BSE) images of some samples (e.g., OS1730) from the studied xenoliths show carbonate veining that crosscuts xenolith-basalt interfaces (Figure 5(c)).

Simplified cross-section of the Kyrgyz-Tianshan lithosphere inferred from the studied xenoliths is discussed below (Figures 6 and 7). Two groups of lherzolite xenoliths from the Ortosuu show very different features. The more fertile group 1 lherzolite xenoliths contain lower Mg# and LREE-depleted patterns. In contrast, the more depleted group 2 lherzolite xenoliths have higher Mg# and LREE-enriched patterns. It is noteworthy that the REE patterns of group 2 vary from a “spoon” shape to an upward convex shape (Figure 4), suggesting that they had undergone partial melting initially and some later metasomatic processes in the originally depleted lithospheric mantle (e.g., in samples OS1701, OS1703, OS1705, OS1713, UK1708, OS1722, OS1730, OS1729, OS1731, OS1734A, and OS1734B). Combined mineral chemical compositions and the calculated P-T estimations indicate that the more fertile group 1 lherzolite xenoliths were probably derived from the uppermost mantle. These relatively fertile lherzolite xenoliths are also found in the North China Craton [56]. The more depleted group 2 lherzolite xenoliths were likely derived from a depth close to the crust mantle boundary, which may have been affected by 6–11% partial melting of the old lithospheric mantle and subsequently modified by carbonatite and/or silicate melts during metasomatism. Estimates of P-T conditions from the Ortosuu garnet granulite show that it was probably derived from a lower crustal depth and transported to the surface through the eruption of the host basalt. Notably, calculated pressures (depth) for the xenoliths are not well constrained for group 2 Ortosuu lherzolites (OS1701, OS1703, and OS1713), Ortosuu wehrlites (OS1722, OS1730), and Ortosuu granulite (OS1736), which may be due to the fact that they have experienced a higher degree of partial melting and later modification. Further electrical conductivity, seismic tomography, core-mantle boundary petrology, and thermal structure could aid in a better reconstruction of the depth of xenoliths origin.

The nature and origin of the studied xenoliths are discussed below and illustrated in Figure 7. Peridotite xenoliths from the Ortosuu basalts include lherzolite, harzburgite, and wehrlite. Lherzolites are essential members in mantle peridotite complexes, and their mineralogical compositions suggest that they can form by various mechanisms. Group 1 lherzolites from Ortosuu represent relatively fertile parts of the upper mantle which may have been affected by low degrees of partial melting. LREE-depleted patterns in these rocks are often observed in mantle xenoliths and in tectonically emplaced peridotites [57]. Spoon-shaped REE patterns in Group 2 lherzolites are similar to refertilized refractory peridotites. Thus, their compositional character cannot be ascribed solely to the depletion process caused by partial melting and likely involved a multistage history. During the first stage, the lherzolites were partially melted, then refertilized by ascending melts, and enriched in LREE. Furthermore, group 2 lherzolites have experienced a higher degree of partial melting than those in group 1. This evidence shows that spinel lherzolites can originate as secondary peridotites from more depleted harzburgites [58]. Thus, the group 2 lherzolites could have been harzburgitic in nature prior to a LREE-enriched metasomatic event [59]. It is also likely, however, that some mantle harzburgites are depleted and may have formed as the result of large degrees of partial melting from a lherzolite parent [60]. Experimental studies have also shown that liquid reacting with lherzolite can produce a harzburgite [61, 62]. Wehrlite is an ultramafic and ultrabasic rock that is a mixture of olivine and clinopyroxene. Group 1 wehrlites from Ortosuu have lower Mg# values and Ni contents (Table S1), which suggest that these xenoliths formed by high-pressure fractional crystallization of a basaltic magma [63]. However, group 2 Ortosuu wehrlite is likely the product of a reaction between a metasomatic agent and the wall rock [64]. The presence of orthopyroxene indicates that the wehrlite precursor lithology contained orthopyroxene and thus was likely a lherzolite, a characteristic of the upper mantle [65]. Previous research also indicates that wehrlites may be produced via metasomatism of harzburgites and/or dunites by carbonate melts [64]. Thus, wehrlite may be formed from lherzolite and harzburgite. We propose that the wehrlites found at Ortosuu represent a highly altered mantle that underwent extensive reaction with infiltrating magmas during 75–70 Ma. In this case, orthopyroxene broke down to olivine, clinopyroxene, and a silica-rich melt, leading to the transformation of lherzolite and harzburgite into wehrlite [66]. Pyroxenite xenoliths from the Ortosuu site include olivine websterite and clinopyroxenite. The olivine websterites have relatively high Al₂O₃ content and a positive Eu anomaly (Figure 4(b); Table S1), which are believed to involve solid-state recycling of the oceanic lower crust [67]. In addition, clinopyroxenites show convex-upward REE patterns (Figure 4(e)), typical of the fractionation products of alkaline melts [68]. These clinopyroxenites are interpreted to be cumulates formed near the Moho [69]. The Ortosuu garnet granulite is derived from the lower crust, where they may be generated granulite residues and silicic melts [70]. Furthermore, the relatively low Mg# and Ni and Cr contents of the garnet granulite xenolith suggest that it is not likely to be the reaction product of silicic magmas with peridotite [71]. This may have resulted from impregnation of mantle plume on the base of the lithosphere. Melts extracted from any of the partially molten systems emplace at crustal levels in the form of extrusive volcanic rocks and intrusive plutons. Thus, the garnet granulite xenolith sample OS1736 was likely formed by the combination of recycled ancient crust-derived melt-peridotite reaction and basaltic magma underplating. The studied xenoliths were brought to the surface by primitive alkaline magmas (basanites and tephrites basalt) that erupted during ~75–70 Ma. However, the melts may have percolated through the mantle for extended periods of time prior to the eruption of the alkaline basalts. The geodynamic mechanism of xenolith transport in the region is probably associated with continental plume-related basaltic magmatism [5, 9, 72]. It is noteworthy that the eruption of Late Cretaceous–Paleocene Tianshan basalts (76–60 Ma) was nearly coeval to the formation of the Deccan traps (∼66 Ma), implying that its origin may be related to the worldwide Late Cretaceous superplume [9, 73, 74]. Thus, it seems plausible that the distribution of magmatism was associated with continental plume-related large igneous provinces (LIPs). Although the plume head may cause both small- and large-scale vertical movements of the surface, they could exclude those models that invoke the Cretaceous subduction-related magmatism ([8]); more solid evidence would need to be sought to make such a mechanism a viable possibility.

• (1)

  Based on Mg# and trace element patterns, clinopyroxene in the sampled xenoliths is divisible into two groups. Group 1, primitive clinopyroxenes, has lower Mg# (86–90) and LREE-depleted patterns than group 2, depleted clinopyroxenes, which are characterized by a relatively high Mg#, 91–92, and LREE-enriched patterns.

• (2)

  The upper mantle beneath south Tianshan mainly consists of spinel lherzolites, which may have affected by 6–11% partial melting, and later modification by carbonatite and/or silicate melts during metasomatism.

• (3)

  Combined mineral chemical compositions and the calculated P-T estimations indicate that the more fertile group 1 xenoliths were probably derived from the uppermost mantle and the more depleted group 2 xenoliths were likely derived from a depth close to the crust mantle boundary. However, some of the estimated pressures for the group 2 xenoliths are not well constrained; further study will be needed.

The authors confirm that the data supporting the findings of this study are available within the supplementary materials.

The authors declare no competing financial interests.

This research was supported in part by the Higher Education Sprout Project, Ministry of Education to the Headquarters of University Advancement at the National Cheng Kung University (NCKU). N.H.-C. Chen acknowledges the support from the Young Scholar Fellowship Program by the Ministry of Science and Technology (MOST) in Taiwan, under grant MOST111-2636-M-006-019. P. A. Cawood acknowledges the support from the Australian Research Council, grant FL160100168. We thank Masako Usuki and F.-L. Lin for their kind assistance during experimental analysis. We also thank K.-L. Wang for assistance in the field excursion.