Introduction

The subcontinental lithospheric mantle (SCLM) plays a key role in the formation and evolution of continents over geological time (e.g., Griffin et al. 1998). In cratons, mantle roots are strongly depleted in basaltic components and have positive buoyancy and strong viscosity, which can make continents “float” on convective asthenosphere for billions of years (e.g., Frey and Green 1974; Walter 1998; Griffin et al. 1999a, b; O’Reilly et al. 2001; Bernstein et al. 2007; Lee et al. 2011; Tang et al. 2013). However, many studies show that the Archean cratonic lithospheric root could experience significant modification and even entire destruction (Foley 2008; Aulbach et al. 2017a). One well-studied example is the North China Craton (NCC) (e.g., Zheng 1999, 2009; Xu 2001; Gao et al. 2002; Zheng et al. 2007; Zhang et al. 2008; Dai et al. 2018). Mantle xenoliths and xenocrysts in Paleozoic diamondiferous kimberlites suggest that the NCC lithospheric mantle was Archean refractory mantle, which was thick (about 200 km) and cold (geotherms 36–40 mW/m2) at least until the mid-Ordovician time (Zheng 1999). By contrast, xenoliths in Cenozoic basalts show the presence of fertile lithosphere, which was thin (< 90 km) and hot (50–105 mW/m2) (Menzies et al. 1993; Zheng et al. 1998; Xu 2001; Gao et al. 2002; Zhang et al. 2008). These observations suggest great changes in the nature and thickness of the lithosphere (Gao et al. 2002; Zheng et al. 2007; Zhang et al. 2008). The mechanisms accounting for lithospheric modification have been proposed to be lithospheric delamination (Gao et al. 2004; Wu et al. 2005a) and/or asthenospheric erosion (Griffin et al. 1998; Zheng et al. 1998, 2007; Xu 2001), possibly along intra-lithospheric weak zones that are visible as mid-lithospheric discontinuities (MLD) in the neighboring intact cratonic mantle (Chen et al. 2014).

Mantle metasomatism can also cause severe chemical and mineralogical modifications in lithospheric mantle (e.g., Lee et al. 2011; O’Reilly and Griffin 2012; Tang et al. 2013; Xiong et al. 2015). Cryptic and modal types of metasomatism have been identified; the former refers to the mineral compositional alterations generated during infiltration of external fluids and/or melts (e.g., Ionov et al. 2002; Yu et al. 2005), while the latter would introduce newly crystallized minerals such as apatite, carbonates, amphibole and phlogopite (e.g., Rudnick et al. 1993; O’Reilly and Griffin 2012). The addition (and removal) of minerals that form part of the pre-existing assemblage has been termed “stealth metasomatism” and is, as the name implies, not straightforward to recognize based on mineralogy alone (O’Reilly and Griffin 2012). Identified metasomatic agents include aqueous fluids (Gorring and Kay 2000; Xu et al. 2003a), silicate melts (Zangana et al. 1999) and carbonatitic melts (Ionov et al. 1993; Rudnick et al. 1993). It is worth noting that the migration of the same fluids/melts within different domains or levels of lithospheric mantle would also form diverse metasomatic effects that could be mistaken as multiple episodes of metasomatism involving different types of fluids or melts (e.g., Navon and Stolper 1987; Vernières et al. 1997; Ionov et al. 2002; Yu et al. 2005).

The area of the northeastern NCC witnessed a complex evolution, including the closure of the Paleo-Asian Ocean from the north (Xiao et al. 2003) and subsequent subduction of the Paleo-Pacific Plate from the east (Tang et al. 2012; Zheng and Dai 2018). However, the specifics of the modification processes of the lithospheric mantle beneath this area, which resulted from its complex tectonothermal evolution, remain poorly constrained. In this contribution, we present detailed petrographic, whole-rock and in situ mineral elemental compositions of spinel-facies peridotite xenoliths from the Huinan Quaternary basalts in the northeastern NCC. Our aims are to shed new light on the nature and evolution of the lithospheric mantle beneath the northeastern part of the craton. Our findings provide new insights into the regional lithospheric architecture and reveal the complex processes of the partial replacement of the ancient SCLM in the tectonic framework of the Mesozoic-Cenozoic NCC destruction.

Geological setting

The North China Craton, with an Archean nucleus, is bounded by the Northeast China (NE China) fold belts to the north and a Pacific convergent system to the east (Fig. 1a). The craton contains three tectonic zones, i.e., the Eastern Block, the Central Zone and the Western Block (Zhao et al. 2000, 2005). The amalgamation of the Eastern Block and Western Block along the Central Zone during the Paleoproterozoic (~ 1.85 Ga) marks the final cratonization of the NCC (Zhao et al. 2005). The Eastern Block is composed mainly of Archean TTG gneisses, granitoids, granitic gneisses and supracrustal rocks (Zhao et al. 2000; Zhai and Santosh 2011) and remained tectonothermally quiescent until the middle Ordovician (Zhang and Yang 2007). During the Late Mesozoic and Cenozoic, the Eastern Block experienced multiple subductions (Windley et al. 2010) and significant decratonization (Zhu et al. 2012), as manifested by ubiquitous magmatism (Zhang et al. 2014), vigorous crustal deformation (Wang et al. 2011), widespread metallogenesis (Sun et al. 2007) and modification of the nature and thickness of the SCLM (Zheng et al. 1998, 2007; Zheng 1999), whereas the Western Block remains relatively stable with little crustal deformation and magmatism (Zhang et al. 2014).

Fig. 1
figure 1

Simplified tectonic units and distribution of Cenozoic basalts in North China Craton and Northeast China (NE China) fold belts (Fig. 1a) as well as the location of the Longgang volcanic field (Fig. 1b), revised after Xu et al. (2003b, 2012). The DTGL represents the Daxinganlin–Taihang gravity lineament

Full size image

The Longgang volcanic field (LVF) in Jilin Province is located in the northeastern corner of the NCC, related to activation and development of the northern part of the Tan-Lu fault belt (T-L FB) (Fig. 1a; Liu et al. 1992). It covers an area of 1700 km2 and consists of more than 160 Quaternary craters and calderas (Liu 1999), many of which were erupted in the early Pleistocene (0.6 Ma) (Fan et al. 2000). The eruption of LVF can be divided into three main periods, namely the Xiaoyishan Period (Early Pleistocene: 2.15–0.75 Ma), the Longgang Period (Middle-Late Pleistocene: 0.68–0.05 Ma) and the Jinglongdingzi Period (Holocene: 1600–1500 a.BP) (Fan et al. 2002). The basement of LVF consists dominantly of Archean metamorphic rocks. It is surrounded by a large number of basins that were developed in Cenozoic, the most famous of which is the Songliao Basin. These Cenozoic alkali basalts host abundant mantle-derived megacrysts and xenoliths, the latter mainly spinel-facies peridotites (Xu et al. 2003b; Wu et al. 2003; Lu and Zheng 2011; Tang et al. 2012). The studied Huinan peridotites are all collected from Dayishan (Fig. 1b).

Sample descriptions and petrography

The Huinan peridotites, part of the LVF, are very abundant and fresh with round to angular shapes and are about 1–20 cm in diameter. The studied samples are all spinel facies, including predominantly harzburgites, subordinately lherzolites and newly found dunites.

Peridotites of Group 1 have dominantly protogranular microstructures, high olivine modes of at least 84 vol.%, except one sample of 79 vol.%, and are composed of harzburgites and dunites (detailed mineral assemblage and modes can be found in Table S1). A few olivine and orthopyroxene grains exhibit kink bands and show curvilinear or irregular grain boundaries, with occasional 120° triple junctions. Spinel occurs as individual grains at the rims of large orthopyroxenes or as pyroxene–spinel clusters (intergrowths of orthopyroxene + spinel) that together mimic ovoid grain shapes (Fig. 2a, b). Exsolution lamellae and some fluid/melt inclusions are present in some large orthopyroxene porphyroclasts (Fig. 2c, d).

Fig. 2
figure 2

Petrographic features of the studied Huinan peridotite xenoliths. a Opx + sp symplectitic intergrowths in xenolith HNd-074. b Opx + Sp symplectite characterized by ovoid shape in xenolith HNd-075. c Olivine embayments partly corroding the orthopyroxene porphyroclast with melt/fluid inclusions in xenolith HNd-065. d The rims of opx in contact with olivine are riddled with tiny (~ 0.01 mm) minerals and melt/fluid inclusions in xenolith HNd-065. e Protogranular texture in xenolith HNd-083. f Porphyroclastic texture in xenolith HNd-090. g Cpx porphyroclast with exsolution lamellae in xenolith HNd-087. h 120° triple junctions between Ol grains in xenolith HNd-086. Photomicrographs for a, c, d and h are cross-polarized light. Plane-polarized light for e, f and g. Ol Olivine, Opx orthopyroxene, Cpx clinopyroxene, Sp spinel

Full size image

Group 2 peridotites show lower olivine modes (64–83 vol.%) but higher clinopyroxene modes (3–14 vol.%) compared with those of Group 1 (79–96 vol.% and < 1 vol.%, respectively) and consist of lherzolites and harzburgites with transitional textures between protogranular and porphyroclastic textures (Fig. 2e) or porphyroclastic textures (Fig. 2f). Orthopyroxene and clinopyroxene carry abundant exsolved spinel lamellae and fluid/melt inclusions (Fig. 2g). Kink bands, fractures and 120° triple junctions are common between olivine grains (Fig. 2h). Some orthopyroxenes are irregular in shape and partially enclose olivine. Spinel occurs as inclusions in olivines or as interstitial phases at grain boundaries. Rare pyroxene–spinel clusters are also observed in Group 2 peridotites.

Analytical methods

All analyses were performed in the State Key Laboratory of Geological Processes and Mineral Resources, China University of Geosciences (Wuhan). The major element compositions of bulk rocks were measured by X-ray fluorescence (XRF) using fused glass disks, by fusion of 0.6 g of rock powder, 5 g of compound flux (Li2B4O7:LiBO2 = 12:22), 0.3 g NH4NO3 and 0.4 g LiF in a high-frequency furnace for 11 min at ~ 1050 °C in 95% Pt–5% Au crucibles. The melt was swirled repeatedly to ensure complete dissolution and homogenization of the material and then poured into a mould to form a thin flat-surfaced disc (34 mm diameter). The loss-on-ignition (LOI) was measured on dried rock powder by heating in a pre-heated corundum crucible to 1000 °C for 90 min and recording the percentage weight loss. XRF analysis was carried out on a Shimadzu XRF-1800 sequential X-ray fluorescence spectrometer, using an Rh-anode X-ray tube with a voltage of 40 kV and current of 70 mA. Calibration curves used for quantification were produced by bivariate regression of data from ~ 40 reference materials encompassing a wide range of silicate compositions. Analytical precision is better than 4% for major elements. The measurement procedure and data quality were monitored by repeated analyses (one between ten samples) of USGS standard BHVO-2 and Chinese National standards GBW07102. Results for reference materials are given in Table S2-RM, suggesting that the most element contents agree within 3% (analytical accuracy) with reference values.

In situ major-element compositions of minerals were determined by a JEOL JXA-8100 Electron Probe Micro-Analyzer (EPMA), equipped with four wavelength-dispersive spectrometers (WDS), using an accelerating voltage of 15 kV and a sample current of 20 nA. The nominal diameter of the electron beam was < 1 µm. Data were corrected online using a modified ZAF (atomic number, absorption, fluorescence) correction procedure. Element peaks and backgrounds were measured for all elements with counting times of 10 s and 5 s, respectively (except for Ti, Mn, these were 20 s and 10 s, respectively). The following standards were used for quantification: Jadeite (Na), Rutile (Ti), Pyrope Garnet (Al, Fe, Mg and Si), Rhodonite (Mn), Diopside (Ca) and Sanidine (K), Nickel (Ni).

In situ quantitative analyses of clinopyroxene for trace elements were conducted on thin sections using laser ablation–inductively coupled plasma–mass spectrometry (LA-ICP-MS) (GeoLas 2005 + Agilent 7500a) with an ablation spot size of 44 µm. Helium was used as a carrier gas and nitrogen was added into the central gas flow (Ar + He) of the Ar plasma to increase the sensitivity (Hu et al. 2008). Detailed operating conditions for the laser system and the ICP-MS instrument were described in Liu et al. (2008). Each analysis included approximately 20–30 s of background acquisition (from a gas blank) followed by 50 s of data acquisition from the sample. NIST SRM 610 was used to correct the time-dependent drift of sensitivity and mass discrimination. The element contents of samples, and of USGS reference glasses (BCR-2G, BHVO-2G and BIR-1G) measured as unknowns, were determined using the USGS reference glasses as reference materials for external calibration, without applying internal standardization, and normalization of the sum of all metal oxides to 100 wt.% was applied (Liu et al. 2008). Offline selection and integration of background and analytical signals, time-drift correction, and quantitative calibration were performed by ICPMSDataCal (Lin et al. 2016; Liu et al. 2008). Results for reference materials are given in Table S4-USGS, with most element concentrations agreeing within 10% or better with accepted values.

Results

Whole-rock compositions

Compared to the Group 2 xenoliths, Group 1 peridotites have lower Al2O3 contents (0.53–1.06 wt.% vs. 1.48–3.23 wt.%) and higher MgO (Figs. 3, 4). The contents of Al2O3, CaO and TiO2 show negative correlations with MgO, and the concentrations of Al2O3 and CaO, but not FeO, are lower than those of commonly accepted primitive upper mantle compositions (Figs. 3, 4; e.g., McDonough and Sun 1995). These compositional variations are similar to those expected for residues of variable degrees of partial melting and have been observed in other worldwide mantle xenoliths suites (Frey and Green 1974; Takazawa et al. 2000). FeOT contents in both the Group 1 and Group 2 peridotites, however, do not correlate with MgO (Fig. 3d), which is not consistent with variable extraction of basaltic melts. In addition, the Mg# values (i.e., 100 × Mg/(Mg + Fe), atomic number) vary from 87.6 to 90.7 (Table S2). Specifically, the Group 1 peridotites show higher Mg# (90.1–90.7) and have an affinity with the high-Mg# group peridotites from Hebi (suggested relics of refractory cratonic mantle; Zheng et al. 2001); Group 2 xenoliths display lower Mg# (87.6–90.1), which are similar to peridotite xenoliths from Shanwang (interpreted as newly accreted fertile lithosphere; Zheng et al. 1998) (Fig. 3).

Fig. 3
figure 3

Plots of a Al2O3, b CaO, c TiO2 and d FeOT as a function of MgO in whole-rock peridotite xenoliths from Huinan. Shown for comparison are field for peridotite xenoliths from Hebi and Shanwang. The Hebi peridotites are suggested to represent remnants of the Archean lithosphere preserved locally beneath the NCC (Zheng et al. 2001, 2007), whereas the mantle beneath Shanwang has been interpreted as newly accreted fertile material (Zheng et al. 1998). Ref. refers to the Huinan samples in previous studies (Xu et al. 1998, 2003b; Wu et al. 2003; Wang et al. 2013)

Full size image
Fig. 4
figure 4

Co-variation plots of whole-rock major oxides for peridotite xenoliths from Huinan. In Fig. 4a–c, dashed blue lines show isobaric batch melting residues formed at 2, 4 and 6 GPa; solid orange lines are residues of polybaric fractional melting at 2–0, 3–0, 5–1 and 7–2 GPa (Herzberg 2004); the dashed orange curves denote the contours of partial melting and the orange numbers (15%, 20%, and 30%) represent the degrees of melt extraction. In Fig. 4d, the dashed blue and solid orange curves represent the compositional trajectories for polybaric fractional melting at 1.5–0.5 GPa and at 2.5–0.5 GPa (Niu 1997), respectively; The orange numbers show the degrees of melt extraction at the black crosses. The compositions of primitive mantle (PM) are represented by the fertile peridotite KR4003 (Walter 1998) in Fig. 4a-c and the preferred source after Niu (1997) in Fig. 4d, respectively. Data sources are the same as those for Fig. 3

Full size image

Major element compositions of minerals

Individual mineral grains (olivine, orthopyroxene, clinopyroxene and spinel) in these peridotites are commonly homogeneous by comparison of core and rim analyses by EPMA, except for spinel with sieve textures. Therefore, the average major-element compositions of minerals are used in this study (Table S3).

Olivines show a wide range in Mg# (88 to 91.5, mostly varying from 89.5 to 91.5). The highest Mg# (91.5) is found in sample HNd-68 (Group 1) and the lowest (88.0) in sample J297 (Group 2). Most olivines in Group 1 have higher Mg# than those in Group 2 (Table S3; Fig. 5a). There is a large range of TiO2 contents in olivine of Group 2 (< 0.025–0.04 wt.%), but an apparently limited variation and lower contents in that of Group 1 (< 0.025 wt.%; Fig. 5a), although this observation must be considered in the context of most values being close to or below the detection limit.

Fig. 5
figure 5

Oxide concentrations or Cr# (100Cr/(Cr + Al) molar) in olivine, spinel, orthopyroxene and clinopyroxene from Huinan xenoliths as a function of Mg# (100 Mg/(Mg + FeT) molar). The orange dotted line in a, d, e and f denotes the average detection limit (DL). Data sources as in Fig. 3

Full size image

Spinels in Group 1 peridotites have mostly higher contents of Cr2O3 (23.12–37.10 wt.% vs. 4.97–24.96 wt.%) and thus higher Cr# (26.3–44.3 vs. 5.2–28.7) than those of most samples in Group 2 (Table S3; Fig. 5b). Most samples in Group 1, with high Cr# and low Mg# in spinels, are different from the Group 2 peridotites displaying low Cr# but high Mg# in spinels (Fig. 5b). Meanwhile, the rims of sieve-textured spinels show higher Cr, Fe and Ti contents than those in the clear cores (Table S3).

Orthopyroxenes of Group 2 show large variations in Mg# (88.5–91.3), but limited range for that of Group 1 (91.3–92.1). Most opx in the Group 1 xenoliths have lower Al2O3 contents (2.10–3.21 wt.%) but higher Cr2O3 (0.43–0.48 wt.%) contents than those in the Group 2 xenoliths (3.02–4.65 wt.% and 0.17–0.60 wt.%, respectively) (Table S3; Fig. 5c, d).

All clinopyroxenes analyzed here are Cr-diopside with 0.37–1.79 wt.% of Cr2O3. Compared to clinopyroxenes in the Group 2 xenoliths, clinopyroxenes in the Group 1 peridotites have higher Mg# (91.0-92.5) but lower Al2O3 (3.59–4.97 wt.%) contents and they display large ranges of TiO2 and Na2O contents (Fig. 5e, f). A sample (HNd-065) in Group 1 has the highest TiO2 yet lowest Na2O content (Table S3).

Trace-element compositions of clinopyroxene

Clinopyroxenes can be subdivided into four types, based on their REE patterns (Fig. 6): (1) Type 1 cpx are characterized by the lowest LREE/HREE values [(La/Yb)N = 0.08–0.78], relatively flat HREE [(Gd/Yb)N = 0.72–1.67] (Table S4; Fig. 6a), and obvious negative anomalies in Nb, Zr and Ti (Fig. 6b) and are prevalent in the lherzolites dominating in Group 2 xenoliths; (2) Type 2 cpx show spoon-shaped patterns (Fig. 6c) and negative anomalies in Nb, Zr and Ti (Fig. 6d) and occur in two samples of Group 2 lherzolites; (3) Type 3 cpx display the highest LREE/HREE values [(La/Yb)N = 6.23–8.37], slightly HREE-depleted patterns [(Gd/Yb)N = 1.19–1.65] (Table S4; Fig. 6e), and negative anomalies in Nb and Ti (Fig. 6f) and are found in HNd-066 (Group 1) and J298 (Group 2); and (4) Type 4 cpx show convex-upward patterns, high MREE/HREE, but lower LREE/MREE (Fig. 6g) and negative anomalies in Nb and Ti (Fig. 6h) and occur in Group 1 sample (HNd-071) and two Group 2 samples (J230 and J297).

Fig. 6
figure 6

Chondrite-normalized REE patterns and primitive mantle-normalized trace-element spidergrams for clinopyroxenes from the studied Huinan peridotites. Group 1 peridotites display LREE-enriched (Type 3) and convex-upward (Type 4) REE patterns. Group 2 xenoliths mainly show LREE-depleted (Type 1) and spoon-shaped (Type 2) REE patterns, besides minor Type 3 and Type 4 ones. Chondrite and primitive mantle data are from McDonough and Sun (1995). Lherzolite and Harzburgite in Huinan of grey fields are from Xu et al. (2003b)

Full size image

Equilibrium temperatures

Equilibration temperatures (T) have been estimated using three thermometers based on: (1) Fe–Mg partitioning between pyroxenes (i.e., T(Wells); Wells 1977); (2) Ca content in orthopyroxene (i.e., T(BK)*) following the method of Brey and Köhler (1990) modified by Nimis and Grütter (2010), and (3) enstatite in sole clinopyroxene, which was calibrated for garnet peridotites but can be used to calculate apparent temperatures for spinel peridotites (i.e., T(NT); Nimis and Taylor 2000). Temperature estimates only use the average core compositions of minerals. Due to the lack of reliable geobarometers, all these estimates assume a pressure of 1.5 GPa, a reasonable value for spinel-facies peridotites (e.g., Ziberna et al. 2013), for consistency with peridotite xenoliths from eastern China (Table S5; Xu et al. 1993; Zheng et al. 2007). Most T(Wells) and T(NT) are higher than T(BK)*, which can generally be reconciled with the uncertainties of used thermometers (e.g., 70 °C for T(Wells), Wells 1977; 30 °C for T(NT), Nimis and Taylor 2000) and disequilibrium between the pyroxenes. The opx in the studied peridotites has low Na2O (0.03–0.31 wt.%, mainly 0.07–0.11 wt.%) whereas the cpx has a large Na2O range (0.36–1.92 wt.%; Table S3), which probably affects T estimates using cpx methods due to Na–Ca substitution (Ionov et al. 2018). Thus, we consider that T(BK)* may be a more reliable estimate. T(BK)* for Group 1 and Group 2 xenoliths studied here is 923–977 °C and 881–1110 °C, respectively. These estimates are in agreement with the temperature range previously reported for peridotites from the Longgang block (880–1197 °C; Wu et al. 2003; Wang et al. 2013).

Discussion

Depletion vs. refertilization of the lithospheric mantle beneath Huinan

Cratonic mantle, mainly composed of peridotites, underwent complex melt extraction (depletion) and metasomatic processes (refertilization) (Zheng 2009; Tang et al. 2013). Therefore, geochemical characteristics of peridotite and its minerals will record the superposition of multiple deep processes and mantle properties at the time of entrainment. Mantle partial melting should lead to the depletion of fusible phases (i.e., cpx and sp) and basaltic components (i.e., SiO2, Al2O3, CaO, TiO2 and Na2O) in residues, and the increase of MgO contents and Mg# values (Niu 1997; Herzberg 2004). This would be accompanied by mineralogical changes whereby lherzolite evolves into harzburgite and finally dunite, accompanied by increasing Mg# (Walter 2003). In contrast, refertilization by addition of variable amounts of silicate melts could induce reverse variations (Griffin et al. 2003; Zhang 2009; Tang et al. 2013; Xiong et al. 2015) that are difficult to distinguish from originally undepleted lithosphere. Thus, linear arrays in MgO vs. oxide space can be interpreted by binary mixing between depleted peridotite and a basaltic melt (e.g., Chin et al. 2014). The studied Huinan peridotites have high whole rock MgO contents and low HREE levels of clinopyroxenes (Figs. 3, 6). Moreover, Al2O3 and CaO in a few samples of Group 1 follow a partial melting trend (Fig. 4b, d), and the Group 2 peridotites mainly show LREE-depleted patterns (Fig. 6a). Prima facie, these petrochemical features suggest a variably depleted mantle source.

Partial melting trends of primitive mantle modeled according to Herzberg (2004) and Niu (1997) have been used in the literature to constrain the minimum fractions of partial melting (F) and the melting conditions (Fig. 4). The comparison shows that the majority of Group 1 peridotites could have been produced by > 18% fractional melting or batch melting, whereas most of Group 2 xenoliths apparently require < 16% partial melting in a pressure range of 0.5–2.5 GPa (Fig. 4). Melt fractions can be also estimated from the HREE in clinopyroxene at spinel-facies condition (Fig. S1) following the approach of Johnson et al. (1990), or from the Cr# values of spinel (Hellebrand et al. 2001). However, application of these models produces conflicting results: Cpx modeling indicates that Group 1 peridotites experienced 10–12% degrees of partial melting, whereas the Group 2 xenoliths were subjected to low degrees of partial melting (i.e., 0–8%). In contrast, the degrees of partial melting for the Group 1 xenoliths calculated by spinel Cr# are lower than those based on whole-rock compositions, while similar partial melting degrees for the Group 2 peridotites can be obtained by above methods (Table S3).

Nevertheless, it is also obvious in Fig. 4 that only a few Group 1 samples actually plot on expected melt depletion trends, whereas most samples have higher FeO contents and lower Mg# than primitive mantle models at a given Al2O3 content (Fig. 4a, b), e.g., sample HNd-083 with Mg# of 87.6, which significantly deviates from the partial melting trend of residual peridotites. Furthermore, whole rock compositions are characterized by impoverishment in SiO2 compared with the partial melting trends (Fig. 4c). The trend of CaO vs. Al2O3 in Huinan peridotites is more characteristic of mixing than of melt depletion (Fig. 4d). In addition, apparently linear correlations between the Mg# and Al2O3, Cr2O3, TiO2 and Na2O contents in pyroxenes of the studied peridotites (Fig. 5c-f) are the combined results of early depletion and subsequent melt infiltration (Zhang et al. 2008; Zhang 2009). Finally, various LREE enrichments of clinopyroxene (Fig. 6) and high TiO2 contents in spinels, with obvious deviation from the melting trend, suggest reaction between melts and peridotite (Pearce et al. 2000; Xu et al. 2003b; Fig. 7). These observations indicate that the peridotite xenoliths from Huinan could have experienced various degrees of depletion and profound refertilization. Refertilization not only affects the major element compositions of whole rocks, but can also introduce spinel and clinopyroxene with elevated HREE and Y concentrations. Thus, we emphasize that the estimates of melt fractions extracted from these samples must be minimum, resulting in low apparent degrees of depletion. That said, combined with a comparison of the major elements in whole rocks (Fig. 3) and minerals (Fig. 5), our results clearly indicate that the Group 1 peridotites are moderately refractory, whereas Group 2 xenoliths are more fertile, which will be further discussed in the following section.

Fig. 7
figure 7

Cr# versus TiO2 in spinel for the studied Huinan peridotites. Melting trend and parameters are from Xu et al. (2003b). Data sources are as in Fig. 3

Full size image

Diverse metasomatic effects on refractory vs. fertile SCLM

Peridotite xenoliths from Huinan show various REE patterns of clinopyroxene (Fig. 6), which cannot be explained solely by melt extraction, but must reflect subsequent modification. Mantle metasomatism may be also evidenced by petrographic features. Although the sieve texture of spinel has been identified as a result of partial melting (Su et al. 2011), the sieve-textured spinels in the studied Huinan xenoliths show Mg# decrease, Fe and Ti contents increase and Cr# increase from the cores to the sieved rims (Table S3). These features are, therefore, more consistent with melt–peridotite interaction (Shaw et al. 2006).

Cryptic and stealth metasomatic overprints

Patent metasomatism is typically evidenced by the presence of hydrous minerals (phlogopite, amphibole) and/or carbonate or other exotic minerals (Dawson 1984), which are not recognized in the xenoliths under study. Conversely, cryptic metasomatism leads to enrichment in incompatible trace elements, whereby clinopyroxene is commonly considered as the major carrier of incompatible trace elements in anhydrous garnet-free peridotites (Bedini and Bodinier 1999; Zangana et al. 1999; Xu et al. 2003a), with the exception of orthopyroxene containing some proportions of lithophile trace elements (e.g., HFSE; Byerly and Lassiter 2015) at high temperatures (Witt-Eickschen and O’Neill 2005). Bearing in mind that the trace-element distribution of some elements between the pyroxenes can be affected by subsolidus re-equilibration, clinopyroxene can, therefore, provide useful information on mantle processes, such as partial melting and metasomatic enrichment (Bodinier et al. 1990; Johnson et al. 1990; Hellebrand et al. 2002; Aulbach et al. 2017b).

The different REE patterns of clinopyroxene represent snapshots of different metasomatic agents and/or diverse metasomatic effects of percolating melts or fluids. In the latter situation, melt percolation can give rise to spatial changes in trace element contents as a result of ion-exchange chromatographic effects controlled by mineral/melt partition coefficients of elements (Ionov et al. 2002; Navon and Stolper 1987). The diffusion rates of elements through the porous peridotites are inversely proportional to their solid-melt partition coefficients, which can lead to enrichment of the highly incompatible elements (e.g., LREE) in the chromatographic front with respect to the moderately incompatible (e.g., HREE) and compatible ones (Xu et al. 2003b). Therefore, an identical initial metasomatic medium can produce a series of trace element signatures over a certain distance. At the same time, the extent of metasomatism (or enrichment) decreases with the increase of the distance between melt and rock (Yu et al. 2005).

In this model, Type 1 clinopyroxenes with LREE-depleted REE patterns are explained as the products of equilibration with LREE-depleted melts (Rivalenti et al. 2007; Rampone et al. 2010; Princivalle et al. 2014) rather than melt residues, consistent with their failure to fall on mantle residue trends (Fig. 4), as discussed in the previous section. The spoon-shaped REE patterns of Type 2 peridotites and the convex-upward REE patterns, with enrichment in LREE and humps in Nd or Eu, of Type 4 peridotites could be due to chromatographic effects related to the same metasomatic melt (e.g., Ionov et al. 2002). Conversely, the HREE and Y depletion in the single group 1 peridotite of this type could be related to an entirely different effect, namely preserved equilibration with garnet that has decomposed to spinel-pyroxene assemblages (Fig. 2a, b). This will be further discussed in a later section. The observed LREE enrichment in Type 3 cpx is difficult to reproduce via interaction between alkaline basalts and LREE-depleted ambient peridotite (Rivalenti et al. 2007), suggesting that it results from multiple metasomatic events. Indeed, multistage metasomatism has been proposed on the basis of the Li elemental and isotopic compositions of Longgang mantle xenoliths and was attributed to melts/fluids derived from the subducted Pacific plate and the asthenosphere (Tang et al. 2012).

Finally, the afore-described metasomatism not only had cryptic effects, but was likely accompanied by modal changes representing stealth effects (O’Reilly and Griffin 2012). For example, interaction with silica-undersaturated melts, such as alkaline basalts, can lead to precipitation of olivine (and cpx) at the expense of opx (Pearson and Wittig 2014), which is not necessarily evident from the microstructures or modal abundances. Indeed, Xu et al. (2003a, b) argue that the Huinan harzburgites resulted from interaction between peridotites and percolating melts, transforming lherzolite to harzburgite by the preferential dissolution of pyroxenes and precipitation of olivine. While olivine enrichment can explain the low SiO2 content at a given Al2O3 content recorded in most peridotite xenoliths from Huinan (Fig. 4c), this process alone cannot explain why Group 1 harzburgites have consistently higher Mg# than Group 2 lherzolites (Table S1), which is ascribed to melt depletion without strong refertilization.

Metasomatic agents and the effect of mineralogy

As outlined in the introduction, metasomatic agents in the SCLM include, but are not limited to, carbonatitic melts (Ionov et al. 1993; Rudnick et al. 1993), volatile-rich silicate melts (Zangana et al. 1999) and H2O-CO2 fluids (Gorring and Kay 2000; Xu et al. 2003a). Compared with silicate melts, carbonatite melts are usually enriched LILE and can fractionate HFSE and REE more efficiently (Blusztajn and Shimizu 1994). Thus, low Ti/Eu ratios, HFSE depletion and high (La/Yb)N ratios in clinopyroxenes have been widely interpreted as a key signature of carbonatite-related metasomatism (Rudnick et al. 1993; Coltorti et al. 1999). The cpx in the studied Group 2 peridotites have predominantly high Ti/Eu (2331–5947) and low (La/Yb)N (0.08–7.94) ratios (except J230) and fall in the silicate metasomatism area of the Ti/Eu-(La/Yb)N diagram (Fig. 8), similar to the fertile peridotites from Shanwang that underwent silicate-melt metasomatism (Zheng et al. 1998). Group 2 sample J230 plots outside both the silicate and the carbonatitic metasomatism area, which may reflect superposition of multiple metasomatism as discussed above. In contrast, two Group 1 samples (J298 and HNd-066), which generally show garnet-breakdown textures, probably still reflect equilibrium with former garnet and their elevated (La/Yb)N can therefore not be unambiguously ascribed to carbonatitic metasomatism, as discussed in the following section.

Fig. 8
figure 8

Modified from Coltorti et al. (1999). Data sources as in Fig. 3

(La/Yb)N vs. Ti/Eu of clinopyroxenes from the studied Huinan peridotites; subscript N denotes normalization to chondrite (McDonough and Sun 1995).

Full size image

To further constrain the nature of the metasomatic agents, hypothetical liquids in equilibrium with clinopyroxenes were calculated using the cpx/melt partition coefficients listed by Ionov et al. (2002). Some differences between the calculated melt and the host basalts may arise from the effect of temperature on trace element distribution (e.g., Lee et al. 2007), since the partition coefficient used here was determined at temperatures about 1300 °C, much higher than those of the studied peridotites (881 to 1110 °C), which are below the dry peridotite solidus at 1.5GPa (Hirschmann 2000). Calculated melts are largely similar to the alkaline basalts erupted in the Longgang region (Liu et al. 1994; Chen et al. 2007; Yan et al. 2007), except for lower Ta, Ti, La and Ce and higher Nb contents in the calculated melts (Fig. 9a). Differences in the HFSE can be explained by subsolidus redistribution into opx, whereas higher HREE and lower LREE contents may indicate that the metasomatic melt formed by a higher melt fraction than the basalts that entrained the xenoliths (Fig. 9b). However, for sample J230, the contents of most elements of the calculated melts are higher than those of the alkaline basalts (Fig. 9a), possibly due to the influence of multi-stage metasomatic overprints, which is also supported by its position at the low-Ti/Eu extreme in the Ti/Eu-(La/Yb)N diagram (Fig. 8).

Fig. 9
figure 9

Calculated compositions of melts in equilibrium with clinopyroxene from: a Type 4 samples; b selected Type 1 samples. Cpx/melt partition coefficients are from Ionov et al. (2002). The compositions of Cenozoic basalts (Chen et al. 2007; Liu et al. 1994; Yan et al. 2007) of the Huinan region are shown for comparison (gray lines)

Full size image

Of note, cpx in the two Group 1 samples and a minority of Group 2 samples have lower Ti/Eu and higher La/Yb. This may reflect chromatographic effects described above, implying a smaller degree of melt–rock interaction at a greater distance from the unreacted melt. Alternatively or additionally, mineralogy may play a role: Refractory peridotites (harzburgites, dunites) have higher solidi than lherzolites, which is permissive of interaction with a fluid or small-volume residual liquid but impedes percolation of volatile-poor silicate melts, whereas lherzolites can be in equilibrium with such melts (Stachel and Luth 2015). In this model, the diverse metasomatic effects by silicate melts on the studied Huinan peridotites can, at least partially, be explained by the difference of fertility in the SCLM.

A refined mechanism for destruction and replacement of the SCLM beneath the eastern NCC

Pseudomorphs after garnet resulting from rifting and decompression

It is conspicuous that Group 1 samples are characterized by spinel–pyroxene intergrowths, which are generally considered to be formed by garnet breakdown reactions during a transition of lithospheric mantle from the garnet stability field to the spinel stability field (Smith 1977; Rampone et al. 2010; Bhanot et al. 2017; Casagli et al. 2017), and which occupy ovoid or curvilinear shapes (Fig. 2a, b). Similar features have been described for peridotite xenoliths from the North Atlantic Craton in Greenland, where they were interpreted as pseudomorphs after garnet that was destabilized during extension (Aulbach et al. 2017b). Garnet destabilization was similarly described for mantle xenoliths from Antarctica and the East African Rift (Foley et al. 2006; Kaeser et al. 2006). Huinan is located near the Tan-Lu fault belts. All the aforementioned localities were affected by rifting and various degrees of lithosphere thinning, providing a commonality with the samples under investigation. Indeed, extension and melt–SCLM interaction have been ascribed to slab rollback of the subducting Paleo-Pacific plate (Tang et al. 2012; Zheng and Dai 2018) and can explain the decompression evidenced by the suggested pseudomorphs after garnet.

Further evidence for the former presence of garnet in Huinan peridotites comes from the low and variable Al2O3 contents over a narrow Mg# range in opx from Group 1 peridotites, which may reflect a superposition of their more refractory compositions and of preserved equilibration with garnet (now present as spinel–pyroxene intergrowths). Because these xenoliths experienced significant refertilization by silicate melt—though less intensive and complete than in the Group 2 xenoliths—it seems reasonable to suggest that this metasomatic event catalyzed garnet breakdown. Concomitant reaction of the garnet breakdown products with the infiltrating melt implies that the reaction was not isochemical. Finally, negative slope in the HREE patterns of Group 1 cpx is also suggestive of equilibration with former garnet rather than reflecting metasomatism because these samples are HREE-depleted rather than LREE-enriched relative to Group 2 cpx. This implies that after decompression, these intergrowths did not fully re-equilibrate with the matrix pyroxenes. Original depletion of the refractory Group 1 peridotites in the garnet stability field would not have fractionated Cr from Al as strongly as melting at spinel-facies conditions (Canil 2004). Residues with intense melting in the garnet field would have a relatively low Cr# value of spinel, which can explain the low estimates of melt fractions for Group 1 spinels (Hellebrand and Snow 2003), compared to estimates based on whole-rock composition. By contrast, the similar extent of melting estimated by both methods for Group 2 may indicate that the re-equilibration was facilitated and catalyzed by strong melt–rock reactions, which is consistent with the only rare occurrence of pyroxene–spinel intergrowths in this group. Therefore, these observations indicate that Group 1 and Group 2 peridotites, representing different domains of SCLM, were initially formed in the garnet stability field.

Because the refractory cratonic mantle has a high solidus temperature, it is impenetrable to refertilizing asthenospheric melts, as discussed above. This aided in the preservation of spinel-pyroxene intergrowths in Group 1 peridotites (Fig. 10d, e), which only show evidence for metasomatism by residual melts at low melt–rock ratios. Conversely, Group 2 peridotites represent decompressing fertile mantle, which was open to pervasive melt percolation, leading to the destruction of pseudomorphs after garnet at high melt–rock ratios (Fig. 10e). We suggest that both lithologies were decompressed from the garnet-stability field, but their different mineralogies and solidi resulted in different styles and intensities of interaction with metasomatic agents, which explains why the spinel–pyroxene intergrowths are rare in the fertile xenoliths, but frequent in the more refractory samples.

Fig. 10
figure 10

Schematic illustrating the possible origin of the studied Huinan xenoliths and modification of the SCLM beneath the northeastern NCC. a Inferred lithospheric mantle, which consisted of Sp and Gt peridotite and contained an MLD, beneath the Huinan area during Paleozoic time. Sp spinel, Gt garnet, MLD mid-lithospheric discontinuities. b Lithospheric extension and uplift, leading to decompression; strong refertilization and densification of the pre-existing ancient lithosphere, which was accompanied by thermochemical erosion, also occurred during asthenospheric upwelling and melt infiltration. c Partial melting of MLD due to heating; delamination of parts of the ancient mantle along a pre-existing MLD, possibly in response to multiple subduction events during the Mesozoic, besides small-scale convection (SSC) in the asthenosphere (Liu et al. 2018). d Renewed lithospheric extension during the Paleo-Pacific slab retreat led to uplift of asthenosphere, into which the ancient mantle regions had sunk, causing their re-accretion. e Both newly accreted material and blebs of relaminated ancient mantle decompressed from garnet stability field; localized melt–rock reaction of both refractory and refertilized overlying mantle portions due to continued melt infiltration, leading to complete destruction of pyroxene–spinel intergrowths after garnet in fertile mantle at high melt–rock ratios, but preservation in refractory harzburgite at low melt–rock ratios (see text for details)

Full size image

Coexistence of two distinct lithospheric mantle sources beneath Huinan

The composition of the SCLM is generally correlated with crustal age and/or tectonic setting (Griffin et al. 1999a). Wu et al. (2003) reported the ages of Longgang peridotite xenoliths suggesting minimum Mesoproterozoic model ages (1.0-–1.2 Ga) for melt depletion of the SCLM, using whole-rock Os isotopes. However, the refertilization processes could mask and even completely obliterate the Archean refractory signatures of parts of the lithosphere, and rejuvenate the Archean mantle by lowering the Re–Os model ages of refertilized peridotites (Tang et al. 2013). In addition, the Longgang Block consists mainly of Neoarchean tonalite–trondhjemite–granodiorite (TTG) rocks (Wan et al. 2005), with only minor occurrences of supracrustalrocks (Anshan Group) that have Sm-Nd isotope model ages of 2800–2500 Ma (Wu et al. 2005b). Recent studies on zircons in high-Mg andesites (Li et al. 2014) and metamorphic rocks from the Seluohe Group (Li et al. 2007) have shown that the Archean basement may be more widespread beneath this region. Therefore, such observations suggest that the age of the SCLM beneath Huinan is likely underestimated, and the SCLM possibly contains—or originally contained—more ancient materials (Fig. 10a).

A comparison of the major elements in whole rocks (Fig. 3) clearly indicates that the Group 1 peridotites are more refractory with an affinity to the refractory mantle beneath Hebi, which represent remnants of the Archean lithosphere preserved locally beneath the NCC (Zheng et al. 2001, 2007). In contrast, Group 2 xenoliths are fertile and similar to those from Shanwang which may represent newly accreted fertile material (Zheng et al. 1998). In terms of texture and degrees of partial melting, Group 1 and Group 2 peridotites are also very similar to the northern and southern SCLM domains beneath the French Massif Central, respectively (Lenoir et al. 2000). So, to some extent, Group 1 and Group 2 peridotites also have an affinity for cratonic and circumcratonic SCLM domains, respectively. The coexistence of moderately refractory and fertile lithospheric mantle beneath the study area reflects the complex lithospheric architecture and evolution.

The studied Huinan peridotites have various equilibrium temperatures. Assuming equilibration to a single conductive geotherm, in the absence of appropriate geobarometers in spinel-facies peridotites, the different equilibrium temperatures may roughly reflect different source depths. There are no systematic differences in equilibrium temperatures between the two peridotite groups. Given evidence for the former presence of garnet and the fact that Group 1 peridotites are more depleted than the more fertile group 2 peridotites, higher pressures were applied to Group 1 in order to test the effect of pressure, whereas those of Group 2 xenoliths were fixed at 1.5 GPa. This shows that the temperature estimates of Group 1 can be clearly distinguished from Group 2 only when the pressure is above 3.5 GPa, which is deeply in the garnet stability field (Ziberna et al. 2013) (Table S5). However, because garnet was clearly destabilized, probably by depressurization, and because the replacement of the lithosphere beneath the NCC occurred in the Late Cretaceous (Xu 2001), whereas the basalt erupted much later in the Quaternary (Fan et al. 2000), we consider the lower equilibrium temperatures at pressures < 2 GPa to accurately reflect the thermal state of Group 1 peridotites at the time of entrainment (Xu et al. 2003b).

In addition, there is the lack of correlation between Mg# in olivine and equilibrium temperatures (Table S5). These features suggest that the moderately refractory and fertile lithospheric mantle sources are not compositionally stratified but coexist within similar depths intervals, consistent with previous studies (Xu et al. 2003b; Lu and Zheng 2011). The juxtaposition of distinct mantle source regions is quite common in eastern China (Zheng et al. 2007; Lu et al. 2013), and also the eastern Central Asian Orogenic Belt (Pan et al. 2013). Furthermore, the two types of SCLM domains experienced a complex metasomatic process by silicate melts derived from the underlying asthenosphere, but show different metasomatic features, as discussed above (Fig. 6). Collectively, the evidence indicates that the lithospheric mantle beneath the Huinan region is heterogeneous, including fertile Phanerozoic materials and moderately refractory Proterozoic (or even Archean) mantle remnants (Fig. 10e).

Delamination of refractory lithosphere and re-accretion with fertile asthenospheric mantle

Mantle keels are usually stable beneath cratons due to their inherent buoyancy and high viscosity (Griffin et al. 1999a; Wang et al. 2015). However, modification and/or replacement of the SCLM by tectonic processes associated with continental collision, uplifting and magmatism (Griffin et al. 1998, 1999a; Aulbach 2018) can partly or even entirely transform the nature of lithospheric mantle from old, cold, thick and refractory “cratonic” lithospheric mantle to the young, hot, thin and fertile “oceanic” lithospheric mantle (Zheng et al. 1998, 2001, 2007). Refertilization by peridotite–melt reaction is considered a possible mechanism because it can account for the compositional change of the lithospheric mantle and the coexistence of young and old lithospheric mantle (Zhang 2009). Coupled with previous studies, the Sr–Nd–Li isotopic data of the peridotite xenoliths from Huinan provide further evidence that the lithospheric mantle underlying the northeastern NCC experienced multistage interactions between melts/fluids and peridotites (Tang et al. 2012; Xu et al. 2003b).

Combined, the evidence for (1) the former presence of garnet in the inherited refractory peridotites, (2) the coexistence of the latter with fertile mantle over a similar temperature and inferred depth interval, as well as (3) a role of delamination in the destruction of the mantle root beneath the eastern North China Craton based on numerical models (Liu et al. 2018) leads us to propose that portions of ancient lithospheric mantle sank into the convecting mantle after delamination some time after the Mesozoic, followed by re-accretion with fertile mantle during asthensopheric upwelling late Mesozoic-Cenozoic (Fig. 10c–e). These processes of lithosphere delamination and re-accretion were probably related to the closure of the Paleo-Asian Ocean (Xiao et al. 2003; Dai et al. 2018) and the slab rollback of the subducting Paleo-Pacific plate (Tang et al. 2012; Zheng and Dai 2018). During continued upwelling of asthenosphere and extension of the overlying lithosphere, the mixed SCLM (including moderately refractory relics and newly accreted fertile mantle) was diversely metasomatized by silicate melts.

Conclusions

We studied the textures and major and trace element compositions of spinel-facies peridotite xenoliths from the decratonized northeastern part of the North China Craton in order to better understand their petrogenesis and the insights they may provide on the response of the subcontinental lithospheric mantle to the regional complex tectonothermal evolution. Several salient observations lead us to propose a refined model for the destruction and replacement of SCLM beneath the North China Craton:

  1. (1)

    The samples can be subdivided into two groups: Group 1 peridotites show protogranular textures and represent moderately refractory SCLM (bulk-rock Mg# 90.1–90.7), whereas Group 2 xenoliths display porphyroclastic to protogranular textures and have more fertile mantle composition (bulk-rock Mg# 87.6–90.1).

  2. (2)

    Clinopyroxenes in the Group 1 xenoliths display LREE-enriched and convex-upward REE patterns, whereas those in Group 2 mainly show LREE-depleted and spoon-shaped REE patterns, with minor LREE-enriched and convex-upward ones. These different metasomatic overprints are ascribed to differences in mineralogy and composition, which results in different solidi and permeability to silicate melts.

  3. (3)

    The former presence of garnet is supported by low Al2O3 contents in some opx and HREE-Y depletion in some cpx and by spinel-pyroxene intergrowths with ovoid shapes apparently occupying former coarse grains. These intergrowths, which suggest decompression from the garnet stability field, are frequent in Group 1, whereas they are rare in Group 2. This difference is again ascribed to the different permeability of fertile vs. refractory peridotite to silicate melts, which catalyze recrystallization and chemical re-equilibration.

  4. (4)

    Despite these mineralogical and compositional differences, peridotites of both groups have similar equilibration temperatures (i.e., 923–977 °C and 881–1110 °C, respectively), which are not correlated with Mg# in olivines, suggesting that they coexist over a range of depths.

Combined, these observations can be explained if ancient cratonic relics sank into the convecting mantle after delamination and were re-accreted together with more fertile mantle during renewed asthenospheric upwelling and decompression, followed by melt–rock interaction, all of which occurred in response to lithosphere extension caused by rollback of the subducting Pacific plate.