| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
Journal of Petrology Volume 41 Number 1 Pages 111-148 2000
© Oxford University Press 2000
Genesis of Young Lithospheric Mantle in Southeastern China: an LAM–ICPMS Trace Element Study
1DEPARTMENT OF EARTH SCIENCES, NANJING UNIVERSITY, NANJING 210093, P.R. CHINA
2GEMOC NATIONAL KEY CENTRE, DEPARTMENT OF EARTH AND PLANETARY SCIENCES, MACQUARIE UNIVERSITY, SYDNEY, N.S.W. 2109, AUSTRALIA
3CSIRO EXPLORATION AND MINING, NORTH RYDE, SYDNEY, N.S.W. 2113, AUSTRALIA
Received July 16, 1998; Revised typescript accepted July 5, 1999
 |
ABSTRACT |
|---|
 TOP ABSTRACT INTRODUCTIONSAMPLE DESCRIPTIONSANALYTICAL METHODSGEOCHEMICAL CHARACTERISTICS OF...DISCUSSIONCONCLUSIONSREFERENCES |
|---|
Geological and geophysical evidence indicates that at least 100 km of Archaean to Proterozoic lithospheric mantle has been removed from beneath large areas of eastern and southeastern China during late Mesozoic to Cenozoic time. Mantle-derived xenoliths in Tertiary basalts from several localities across this region have been studied by X-ray fluorescence, electron microprobe and laser ablation microprobe–inductively coupled plasma-mass spectrometry to characterize this thinner lithosphere. Trace element patterns of clinopyroxenes in the peridotites from southeastern China can be divided into four groups: fertile garnet lherzolites, fertile spinel (± garnet) lherzolites, and depleted and enriched peridotites. The addition of Nb, Sr, light rare earth elements, but not of Ti and Zr, suggests a metasomatizing agent containing both H2O and CO2. This study also demonstrates that the negative Ti anomaly commonly observed in clinopyroxene from mantle peridotites cannot be balanced by the Ti in coexisting orthopyroxene, but can be explained by small degrees of partial melting, using appropriate distribution coefficients. Most of the peridotites from southeastern China, whether spinel or garnet facies, are highly fertile in terms of Al2O3 and CaO contents and mg-number; many resemble commonly used primitive mantle compositions. Modelling of trace element patterns in clinopyroxene indicates that most spinel and garnet peridotites from the Nushan, Mingxi and Niutoushan localities experienced less than 5%, and many less than 2%, partial melting. A few depleted spinel peridotites from Nushan, and all spinel peridotites from Mingxi, require 10–25% fractional partial melting; almost all spinel peridotites from the Qilin locality show evidence of higher degrees (6–25%) of fractional partial melting. At both Nushan and Mingxi, the more depleted compositions occur in the upper part of the lithospheric mantle, which now is
100 km thick. Garnet peridotites are essentially undepleted, and Y–Ga–Zr relationships of the garnets are typical of Phanerozoic mantle. The overall highly fertile nature of the existing lithosphere requires that the Archaean and Proterozoic mantle that existed beneath the region in Palaeozoic times has been largely or completely removed, and replaced by younger, hotter and more fertile material. This probably occurred by upwelling of asthenospheric material during late Mesozoic to Cenozoic time, underplating to form new lithosphere. The occurrence of rare depleted xenoliths may show that some older mantle material is residual and coexists with younger material beneath southeastern China. KEY WORDS: clinopyroxene trace elements; fertile mantle; mantle trace elements; lithospheric mantle genesis; mantle metasomatism
|
INTRODUCTION |
|---|
|
TOPABSTRACTINTRODUCTIONSAMPLE DESCRIPTIONSANALYTICAL METHODSGEOCHEMICAL CHARACTERISTICS OF...DISCUSSIONCONCLUSIONSREFERENCES |
|---|
Mantle-derived xenoliths brought to the surface by basaltic rocks provide direct information on the nature of the lithospheric mantle and on mantle processes. Studies of the composition of ultramafic xenoliths from continental areas indicate that the mantle lithosphere has undergone a complex history of depletion and enrichment of various elements in response to extensive partial melting and metasomatism (e.g. Downes & Dupuy, 1987
; Downes et al., 1992; Carignan et al., 1996). Recent studies also indicate that the composition of subcontinental lithospheric mantle has changed episodically and irreversibly from Archaean to Phanerozoic time (Griffin et al., 1998a, 1998b).
In eastern China, Cenozoic basaltic rocks are extensive and widespread and contain a wide variety of deep-seated xenoliths. Much petrological work has been undertaken on these xenoliths (Zhou & Chen, 1980, 1984; Zhang & Cong, 1983, 1985; Cao & Zhu, 1987; Fan & Hooper, 1989; Basu et al., 1991; Xu et al., 1995, 1996, 1998). However, only a few detailed studies of trace element geochemistry have been carried out, and even fewer studies have integrated major and trace element geochemistry to deal quantitatively with metasomatism and partial melting in the lithospheric mantle (e.g. Song & Frey, 1989; Tatsumoto et al., 1992; Liu et al., 1996).
Southeastern China is particularly interesting as a laboratory for the study of lithospheric processes, because ancient Archaean and Proterozoic lithospheric mantle apparently has been replaced by new material during Mesozoic–Tertiary time [Griffin et al. (1998a) and references therein]. The widespread eruption of young xenolith-bearing basalts in eastern China makes it possible to study the composition of this young subcontinental mantle over a wide area, and to evaluate both its heterogeneity and the processes that produced it.
This study uses xenoliths from Cenozoic basaltic rocks from southeastern China, spanning several large-scale tectonic blocks, for this purpose. We have especially studied the key localities of Nushan in the Sino-Korean Block, and Mingxi, Qilin, and Niutoushan in the Cathaysia Block (Fig. 1), and have analysed fresh mantle-derived xenoliths including garnet peridotite, spinel–garnet peridotite and spinel peridotite. In situ analysis of trace elements in minerals by laser-ablation inductively coupled plasma-mass spectrometry (ICPMS) microprobe (LAM-ICPMS), and major element analysis by electron microprobe, have been used to explore the trace element geochemistry of minerals in the xenoliths and to understand the processes of partial melting and metasomatism of the lithospheric mantle.
|
|
SAMPLE DESCRIPTIONS |
|---|
|
TOPABSTRACTINTRODUCTIONSAMPLE DESCRIPTIONSANALYTICAL METHODSGEOCHEMICAL CHARACTERISTICS OF...DISCUSSIONCONCLUSIONSREFERENCES |
|---|
Nushan xenoliths
The Nushan volcano lies at the southern edge of the Sino-Korean Block (Fig. 1), at the juncture of Jiangsu and Anhui Province comprising alkali olivine basalt, basanite, and nephelinite, dated to the Middle Pleistocene (0·53–0·73 Ma; Chen & Peng, 1988
).
Spinel peridotite xenoliths
Spinel peridotite is the most common xenolith type in Nushan, as at other basaltic localities in eastern China and world-wide (e.g. Zhang & Cong, 1985; Cao & Zhu, 1987; O’Reilly & Griffin, 1987; O’Reilly et al., 1989). Most of the spinel peridotites have fine- to coarse-grained porphyroclastic or equigranular microstructures, but some are strongly foliated, commonly on a millimetre to centimetre scale, defined largely by tabular olivine crystals of 1–2 mm grain-size, and more rarely by discontinuous layers of pyroxene and/or spinel grains.
The peridotites range from relatively fertile compositions, with 15–25 vol. % clinopyroxene (cpx), to more depleted compositions with 2–4 vol. % cpx (Table 1). The depleted samples with cpx <5% were classified as harzburgite by Streckeisen (1976), but all samples contain the lherzolite assemblage (olivine, orthopyroxene and clinopyroxene ± spinel).
|
Garnet-bearing xenoliths
Numerous garnet-bearing xenoliths including spinel-free garnet peridotite, spinel–garnet peridotite and spinel–garnet pyroxenite were found during this work, and have been used to construct a palaeogeotherm for the Nushan lithospheric column (Xu et al., 1998). Most of the garnet-bearing peridotites have coarse-grained porphyroclastic or equigranular microstructures, and high (>15%) modal proportions of cpx.
Garnets in the xenoliths have kelyphitic rims, most commonly appearing as a mass of fine radiating fibres that become more coarsely crystalline towards the outer edges. The kelyphitic rim may consist of two layers; the inner rim is composed of garnet and glass, whereas the outer rim is made up of orthopyroxene, spinel, clinopyroxene, olivine and glass, similar to kelyphites described elsewhere (Hunter & Taylor, 1982; Qi et al., 1995). Fine-grained orthopyroxene and spinel commonly form symplectitic intergrowths.
Mingxi, Qilin and Niutoushan xenoliths
The Mingxi basalt locality lies within the Cathaysia Block (Fig. 1) in northwest Fujian Province, and is of Plio-Pleistocene age (<5 Ma, K–Ar dating; Chen & Zhang, 1992). Qilin and Niutoushan in southeastern Fujian Province are Tertiary basalt localities within the younger fold belt of the southeastern coastal area, which has a Cathaysia Precambrian basement (Fig. 1). No reliable isotopic dating has been published for the Qilin volcanic pipe; the Niutoushan volcano is dated by K–Ar at 16·6–19·2 Ma (Chen & Zhang, 1992).
Compared with the xenoliths from Nushan, those from Mingxi, Qilin and Niutoushan show a relatively restricted range of rock type variation:
- numerous garnet, garnet–spinel and spinel peridotites have been collected from the Mingxi basalts. The garnet-bearing xenoliths show strong foliations, usually have high garnet contents (4·7–10·8%) and high cpx contents (>10%), and typically contain primary spinel as well. We emphasize the ubiquitous presence of cpx even in the harzburgites and rare dunite. One garnet peridotite (MX3SY) contains interstitial biotite adjacent to garnet. The spinel peridotites from Mingxi may have either high (>10%) or low (<5%) cpx contents.
- Spinel lherzolites from Qilin have weakly porphyroclastic to coarse equant microstructures, with grain-sizes in the 1–6 mm range, but a few are foliated and finer grained. The typical mineral assemblage is olivine (60–70%), orthopyroxene (20–30%), clinopyroxene (5–10%) and spinel (2%), ± amphibole (2%) and ± carbonate (<1%; Xu et al., 1996).
- Spinel lherzolites collected from Niutoushan are smaller (<6 cm in diameter) than those from other localities, but all have coarse equant (1–5 mm) microstructures. The mineral assemblage is olivine (53–56%), orthopyroxene (29–34%), clinopyroxene (14–21%) and spinel (1%).
The peridotites from southeastern China show a large range in clinopyroxene contents; a few of them have higher modal clinopyroxene than most estimated primitive mantle compositions (Jagoutz et al., 1979; Hart & Zindler, 1986; Falloon & Green, 1987; Hirose & Kushiro, 1993; Baker & Stolper, 1994; McDonough & Sun, 1995).
The spinel lherzolite samples from Qilin commonly contain amphibole. Amphibole grains are usually smaller than 1 mm and yellow to red–brown in thin section; most are interstitial, some occur as reaction products around the spinel and clinopyroxene or as small veins between other phases, and some form pyroxene + amphibole intergrowths, or very fine lamellae and blebs within pyroxenes, similar to the occurrence of amphibole in some southeastern Australia xenoliths (O’Reilly, 1987). Nushan spinel peridotites may also contain amphibole, but not as commonly as the Qilin spinel lherzolites. In addition, some Nushan spinel peridotites contain mica and apatite. The biotite found in two Nushan spinel lherzolites (Nu9619 and Nu9629) occurs at the xenolith margins; apatite occurs as short rounded-prismatic or polygonal dispersed grains, coexisting with amphibole, which are clouded, similar to those described for southeastern Australia (O’Reilly, 1987).
The presence of amphibole in mantle xenoliths has been interpreted as the product of modal metasomatism of the lithospheric mantle (Wass, 1979a; Dawson, 1980; O’Reilly, 1987). However, cryptic metasomatism may have occurred even where there are no volatile-bearing minerals present (Dawson, 1980; O’Reilly & Griffin, 1988; O’Reilly et al., 1991). Amphibole-free lherzolite xenoliths commonly display strong evidence for metasomatic enrichment in the form of steep rare earth element (REE) patterns (Carignan et al., 1996), suggesting that these samples were cryptically metasomatized (Griffin et al., 1988; O’Reilly & Griffin, 1988; Zindler & Jagoutz, 1988; O’Reilly et al., 1991). The sieved textures or spongy margins of clinopyroxene and amphibole grains in some of the peridotites may be the mineralogical indicators of partial melting. To understand better the detailed processes of partial melting and metasomatism, we have analysed both whole rocks and minerals for major element compositions, and carried out detailed microanalysis of selected minerals by LAM-ICPMS.
|
ANALYTICAL METHODS |
|---|
|
TOPABSTRACTINTRODUCTIONSAMPLE DESCRIPTIONSANALYTICAL METHODSGEOCHEMICAL CHARACTERISTICS OF...DISCUSSIONCONCLUSIONSREFERENCES |
|---|
The major elements of whole-rock samples were analysed by X-ray fluorescence (XRF), using methods described by O’Reilly & Griffin (1988)
. During sample preparation, we selected large (>15 cm in diameter) fresh xenoliths, cut slabs from their centre portions and then disaggregated the slabs carefully between plastic sheets. Clinopyroxene and garnet separates were hand-picked from part of this coarsely crushed material, and rock powders were prepared from a split of this material.
For most of these xenoliths, hand-picked clinopyroxene and garnet were mounted and polished. For the smaller xenoliths, polished thin sections of 100 µm thickness were made. Major elements of the minerals, in both polished grain mounts and thin sections, were analysed using a Cameca SX50 electron microprobe operated at 15 kV accelerating voltage, 20 nA beam current, and using a fully focused beam 2–3 µm across. Standards were natural minerals and matrix corrections were made by the Pouchou & Pichoir (1984) method. Counting times were 10 s for peaks and 5 s for background on either side of the peak.
Trace element compositions of the polished samples were determined by laser ablation microprobe–inductively coupled plasma-mass spectrometry (LAM–ICPMS), using methods described by Norman et al. (1996). A UV (266 nm) laser beam is focused onto the surface of a solid sample. The intense energy of the beam ablates a small amount of the sample, which is swept by high-purity Ar into an inductively coupled plasma-mass spectrometer for determination of analyte mass abundances. The laser used for this study is a Continuum Surelite I-20 Q-switched and frequency quadrupled Nd:YAG laser operated at 4 Hz and 1–3 mJ per pulse. The ICPMS instrument is a Perkin–Elmer Sciex ELAN 5100 operated at 1040 W and adjusted to produce 248ThO/232Th < 1% on the NIST 610 glass. A typical analysis consists of 110 replicates, with each replicate representing one sweep of the mass range at a dwell time of 50–100 ms per mass. For each sample, 35–40 replicates were counted on the carrier gas alone to establish the background, followed by 70–80 replicates for ablation. Each analysis was normalized to a major element as an internal standard; Ca was used for clinopyroxene, garnet, amphibole and apatite. The LAMTRACE program developed by S. E. Jackson (e.g. Longerich et al., 1995) was used for data reduction. Each analysis used in this paper is the average of at least 4–5 grains where grain homogeneity was established.
|
GEOCHEMICAL CHARACTERISTICS OF THE PERIDOTITE XENOLITHS |
|---|
|
TOPABSTRACTINTRODUCTIONSAMPLE DESCRIPTIONSANALYTICAL METHODSGEOCHEMICAL CHARACTERISTICS OF...DISCUSSIONCONCLUSIONSREFERENCES |
|---|
Whole-rock geochemistry
Major element (Si, Al, Fe, Ca, etc.) data may preserve primary information on melting processes in the mantle, as the behaviour of these elements is governed largely by stoichiometry and phase equilibria both under subsolidus conditions and during melting (e.g. Niu et al., 1996
; Niu, 1997). However, the major element composition also may be influenced by metasomatism, especially if the metasomatic agent is a silicate melt. We therefore emphasize the importance of integration of major and trace element geochemistry.
Because some samples (Nushan garnet lherzolites and some Mingxi garnet-bearing lherzolites) are too small for whole-rock analysis, we have calculated compositions using modal analysis and mineral analyses. The reliability of this procedure has been evaluated using samples for which whole-rock analyses are available. A least-squares mixing program has been used to calculate modes for the Nushan xenoliths, using the known mineral compositions and whole-rock analyses. This approach requires careful attention to the Cr2O3 content, which will strongly affect the calculated contents of spinel and garnet; the Cr2O3 contents of the Nushan whole-rock samples were measured by XRF. The agreement between calculated and observed modes is excellent (Table 1; Fig. 2). Qi et al. (1995) presented both whole-rock and mineral compositions for some Chinese xenoliths, and detailed modal compositions determined by point counting (>5000 points for each sample). Calculated modes for these samples are very consistent with the measured modes of Qi et al. (1995) (Table 1, Fig. 2). Therefore, in the case of samples for which whole-rock chemical compositions are not available because of sample size limitations, we use major element rock compositions calculated from modal analysis and mineral compositions (Table 2). The calculated compositions for Nu9604 and Nu9621 in Table 2 are sufficiently similar to the analysed compositions in Table 1 to justify this approach.
|
|
Major elements
Al vs Ca. As shown in Tables 1 and 2 and Fig. 3, the Nushan garnet lherzolites, spinel–garnet lherzolites and most of the spinel lherzolites have high Al2O3 and CaO contents, reflecting a low degree of depletion in basaltic components. These fertile samples have high (>14%) modal proportions of clinopyroxene. A few spinel peridotites, with lower Al2O3 and CaO contents, are more depleted.
|
Mingxi garnet-bearing lherzolites and a few spinel lherzolites have high Al2O3 and CaO contents, like most Nushan samples. Most of the spinel peridotites, however, have lower Al2O3 and CaO contents. The Qilin spinel lherzolites are moderately depleted, whereas Niutoushan spinel lherzolites are all fertile (Fig. 3).
Many of the fertile samples from southeastern China have higher Al2O3 and CaO contents than commonly accepted primitive mantle compositions (Jagoutz et al., 1979; Hart & Zindler, 1986; Falloon & Green, 1987; Hirose & Kushiro, 1993; Baker & Stolper, 1994; McDonough & Sun, 1995). These fertile xenoliths plot in the upper part of the ‘oceanic trend’ of Boyd (1989) and only a few samples plot in the area typical of depleted cratonic peridotite xenoliths (Fig. 3).
mg-number, cr-number vs Ti, Al. The variations of mg-number vs Ti, Al and cr-number are shown in Fig. 4. The strong correlation between mg-number and cr-number suggests that all of the xenoliths are linked by a common depletion process. The negative correlations of Ti and Al with mg-number imply that this process has removed Ti and Al together with Fe. This is consistent with removal of cpx and spinel with increasing melting (i.e. removal of basaltic components from the residue). Most of the Nushan xenoliths lie within a small range of mg-number (89–90), whereas a small number of spinel peridotites are significantly more depleted (higher mg-number and cr-number). The garnet peridotites tend to be less depleted than the spinel peridotites, although the two groups overlap; amphibole-bearing peridotites have on average more depleted major element compositions than other spinel peridotites. The spinel peridotites from Mingxi are significantly more depleted than the garnet peridotites from the same locality. The Niutoushan peridotites have lower mg-number than the most fertile Nushan and Mingxi peridotites, but similar cr-number.
|
Trace elements
mg-number vs V, Ni, Sr, Zr, and Y. Trace element analyses of whole rocks (Fig. 5) show that Ni is positively correlated, and V negatively correlated, with mg-number. Y shows a weak negative correlation with mg-number except for some depleted Nushan samples with mg-number > 91. However, the fertile Niutoushan samples fall away from all these trends, suggesting that they have not been affected by the same processes. Sr and Zr are negatively correlated with mg-number in most of the Nushan amphibole-free suite, but not correlated with mg-number in the other xenolith suites.
|
Mineral geochemistry
Major element data on the minerals of the Nushan, and Qilin xenoliths have been given by Xu et al. (1996, 1998). Data on the minerals of Mingxi xenoliths have been presented by Qi et al. (1995), and new data are given in Table 3.
|
mg-number and mode of olivine
Modal and chemical data for variably depleted peridotites of oceanic origin, together with estimates of the composition of fertile parental peridotite, have been used by Boyd (1989, 1997) to define an ‘oceanic trend’ representing the overall effects of progressive depletion by partial melting at low pressures (Fig. 6). This trend is significantly different from the field occupied by peridotite xenoliths from kimberlites in Archaean terrains; the few known xenoliths from Proterozoic terrains may lie intermediate between these two trends (Griffin et al., 1998b). Many of the peridotites from southeastern China, including most of those from Nushan and Niutoushan, and the Mingxi garnet lherzolites, have low mg-number and plot along the ‘oceanic’ peridotite trend (Fig. 6). However, some samples, especially from Qilin and Mingxi, have more magnesian olivine, and plot away from this trend toward the field of Proterozoic peridotites. Garnet-bearing samples from both Nushan and Mingxi tend to have low mg-number and modal olivine percent, and plot near the generalized ‘pyrolite’ composition. Some amphibole-bearing samples from Nushan plot below the ‘oceanic trend’, suggesting that the metasomatic process associated with the introduction of amphibole has lowered the Fo content of the olivine.
|
Major and trace elements of clinopyroxene in the peridotites
Major elements. Al and Ti contents of clinopyroxenes from all the localities decrease with increasing mg-number and cr-number (Fig. 7). These compositional variations are consistent with primary control by partial melting. However, the plots of Na vs mg-number and cr-number show more scatter than those for Al, which may indicate disturbance of Na by metasomatic processes, as the Nushan amphibole-bearing lherzolites obviously plot outside the general trend. Clinopyroxenes in fertile samples have high TiO2 and Al2O3 contents and are lower in mg-number and cr-number (Table 3), consistent with the whole-rock data.
|
Trace elements. The trace element (Table 4) patterns of clinopyroxenes from southeastern China xenoliths show considerable variation, but can be divided into four distinct groups, illustrated by the Nushan samples in Fig. 8. Group A, represented by cpx from fertile garnet lherzolites, is strongly convex-upward with small negative Ti and Zr anomalies. Group B, represented by cpx from fertile spinel–garnet lherzolites and most of the spinel lherzolites, has relatively flat heavy REE (HREE), small negative Ti (± Zr) anomalies and depleted light REE (LREE). Group C, typical of modally metasomatized, mainly amphibole-bearing lherzolites (except Nu9601, which has a flat to upward-concave HREE), has high LREE contents and large negative Nb and Ti anomalies. Group D, typical of cpx-poor, depleted spinel peridotites, has low HREE contents, high LREE, and very low contents of Zr, Hf and Ti.
|
|
These four groups of cpx can be uniquely discriminated using their trace element signatures (Fig. 9). Specifically, plots of (La/Lu)n against (La/Sm)n (normalized to primitive mantle) separate the fertile garnet lherzolites (Group A) from the fertile spinel (± garnet) (Group B) lherzolites and delineate a discrete field for the amphibole-bearing (Group C) and cpx-poor, depleted (Group D) spinel peridotites combined. Plots of Zrn or Yn against (La/Sm)n separate fields distinctive of Groups C and D, and the Yn plot also discriminates between Groups A and B.
|
Cpx from the Mingxi spinel–garnet peridotites mostly shows Group A patterns, reflecting the higher modal garnet/spinel ratio in these rocks compared with the Nushan spinel–garnet peridotites (Table 1). Cpx in biotite-bearing garnet lherzolite MX3SY has the highest trace element content. Most Mingxi spinel peridotites have cpx with Group D patterns (Fig. 10), reflecting a generally higher degree of depletion than is observed at Nushan. The cpx in Niutoushan spinel lherzolites has consistent Group B trace element patterns, reflecting fertile samples, although two have Group C patterns (Fig. 10). Most of the Qilin cpx are Group D (Fig. 11), with concave-downward HREE to middle REE (MREE), and elevated LREE and Th contents. As with the Nushan samples, the cpx groups can be distinguished using plots of (La/Lu)n against (La/Sm)n and Yn against (La/Sm)n (Fig. 9).
|
|
Cpx from spinel–garnet peridotites with low garnet contents typically has trace element patterns similar to those of cpx from fertile spinel lherzolites, whereas cpx from garnet-rich peridotites is depleted in HREE but not in MREE. Cpx in depleted samples has low MREE and HREE contents, but with variably high LREE, whereas modal metasomatism typically is accompanied by introduction of LREE, Sr, but not Ti and Zr. These correlations are similar to those described in many other xenolith suites world-wide (e.g. Stosch & Lugmair, 1986; O’Reilly & Griffin, 1988; Bodinier et al., 1988; Hauri & Hart, 1994; Rivalenti et al., 1996). The distinctive trace element signatures of the depleted, fertile and metasomatized clinopyroxenes are fingerprints that can be used to identify mantle process type using the trace element analyses of the clinopyroxenes.
Major and trace elements in garnet
Garnets from the peridotites from Nushan and Mingxi peridotites show a small range in major element composition and are very similar to each another (Xu et al., 1998). They are low-Cr pyropes, with CaO contents of 4·5–5·5% and Cr2O3 contents from 0·5 to 2·2%, typical of many from Phanerozoic regions. Similar garnets occur in xenoliths from the Vitim area in Russia (Ionov et al., 1993).
Garnets from the Nushan garnet peridotites show a very narrow range of trace element patterns (Fig. 12), whereas those in Mingxi peridotites show more diversity. The garnet in MX3SY (a mica-bearing lherzolite) has higher contents of REE, Zr and Hf than the other samples (Fig. 12), whereas MX14 is more depleted in LREE than the others. The high HREE contents and positive Zr anomalies of the garnets are complementary to the low HREE and negative Zr anomalies of the cpx in garnet-rich rocks (Fig. 8a). Both garnet and cpx show negative Ti anomalies.
|
The trace element compositions of garnets from the lithospheric mantle record secular changes in the composition of lithospheric mantle (Griffin et al., 1998b). Lherzolitic garnets from Archaean sections have high mean Zr/Y (
5) and low mean Y/Ga (<3), whereas Phanerozoic lherzolitic garnets have low mean Zr/Y (<1) and high mean Y/Ga (4); garnets from Proterozoic sections have intermediate values. These variations in trace element composition can be modelled in terms of higher fertility, expressed as higher (cpx + gnt) and cpx/gnt, in the Phanerozoic mantle (Griffin et al., 1998b). The garnets plotting above the Phanerozoic (Tecton) field in Fig. 13 extend the originally defined compositional range. The mean values of Y/Ga and Zr/Y for garnets from the peridotites from southeastern China plot in the field typical of garnets from fertile Phanerozoic mantle (Fig. 13), indicating that none of the garnet-bearing peridotites represents relict Archaean or Proterozoic mantle.
|
Trace elements in amphibole and apatite
Amphibole analyses from the Nushan xenoliths have been given by Xu et al. (1998). The amphiboles are mildly titanian pargasites with Na2O/K2O = 2·5, similar to many in xenoliths from eastern Australia (Griffin et al., 1984; O’Reilly et al., 1991; Ionov et al., 1997).
The whole-rock REE patterns of most mantle rocks in regions of melting are controlled by clinopyroxene and garnet if present. Garnet contains most of the HREE (Fig. 12), whereas clinopyroxene contains most of the LREE. However, amphibole contains higher REE than clinopyroxene (Fig. 12), and may be an important site for REE and other trace elements, in samples with high modal amphibole (Nu9602 and Nu9628). Amphibole in amphibole-poor peridotite (Nu9601) has a distinctive trace element pattern with a flat slope and positive Ti anomaly (Fig. 12), in contrast to the steeper pattern and negative Zr–Hf and Ti anomalies of the more common disseminated amphibole (Nu9602).
The apatites contain significant amounts of SiO2 (0·13 wt %), FeO (0·33 wt %), MgO (1·08 wt %), Na (1·48 wt %), F (0·23 wt %) and Cl (0·90 wt %), similar to apatites from the southeastern Australia peridotites (O’Reilly et al., 1991; Matsumoto et al., 1997). They are inferred to be CO2 bearing on the basis of this similarity (O’Reilly et al., 1997). Compared with other mineral phases, apatite is enriched in all trace elements except Nb and Zr (Fig. 12). Apatite has been found only in amphibole-rich samples, similar to those from western Victoria, Australia and Alaska (O’Reilly & Griffin, 1988; O’Reilly et al., 1991). This apatite commonly is associated with abundant fluid inclusions, which are dominantly CO2 (O’Reilly et al., 1997). Xia et al. (1993) analysed the primary fluid inclusions in mantle-derived lherzolite xenoliths from Nushan by laser Raman microprobe, and found their composition to be: CO2 72–99%; CO 19·0%; H2O 9·2%; SO2 1·0–9·0%; CH4 2·2%.
These volatile-bearing minerals along selvedges may suggest the presence of mantle fractures filled with metasomatic veins (O’Reilly, 1989; Wilshire & Kirby, 1989). Amphibole, mica and apatite in these peridotite xenoliths are geochemically significant because they are reservoirs of abundant large ion lithophile elements (LILE) such as K, Ba, Sr, U and LREE, and thus are important as records of mantle metasomatism and mantle heat production (Wilshire, 1987; O’Reilly et al., 1991, 1997).
|
DISCUSSION |
|---|
|
TOPABSTRACTINTRODUCTIONSAMPLE DESCRIPTIONSANALYTICAL METHODSGEOCHEMICAL CHARACTERISTICS OF...DISCUSSIONCONCLUSIONSREFERENCES |
|---|
Partial melting: clinopyroxene trace element modelling
The major element compositional trends among the peridotite xenoliths from southeastern China are similar to those found from experimental partial melting studies, and Qi et al. (1995)
therefore interpreted these peridotite xenoliths to be residues from different degrees of partial melting of a hypothetical primitive mantle. However, the cpx trace element patterns suggest that the mantle heterogeneity revealed in these xenoliths also may reflect the interplay between partial melting and metasomatism in the upper mantle. These trace element patterns can be used to place quantitative constraints on the relative importance of these processes, by modelling the degree of partial melting. Exercises of this type have been carried out on oceanic peridotites by Johnson et al. (1990) and Niu (1997). As pointed out by Norman (1998), modelling melting in continental peridotites is more difficult, because metasomatism may affect elements such as Ti and Zr [used by Johnson et al. (1990)] and LREE. However, the generally flat LREE–MREE patterns of the clinopyroxenes studied here suggest that the contents of these elements may be controlled primarily by partial melting processes (see Norman, 1998). We therefore have adapted the techniques of Johnson et al. (1990) and Norman (1998) to model partial-melting trends in the xenoliths studied here. Two approaches have been used; in one the cpx compositions are taken as analysed; in the other, the cpx compositions are corrected for the elements held in garnet, if present.
Uncorrected-cpx modelling
Johnson et al. (1990) considered the behaviour of batch melting and fractional melting for spinel lherzolite and garnet lherzolite assemblages. Because the compositions of other phases have already been considered, no composition correction is required for clinopyroxene in this modelling. The Ti vs Zr modelling used by Johnson et al. (1990) is not used here, because many southeastern China peridotites fall off the modelling trend on a Ti vs Zr plot (not shown), indicating disturbance of Zr or unsuitable distribution coefficients for Ti and Zr (see below).
Results of modelling using Ti and Yb are shown in Fig. 14a and b, revealing two trends for residual cpx compositions depending on the modal proportions of cpx and garnet. Starting and melting modes are those from the appendices of Johnson et al. (1990); these starting modes overlap those determined for the eastern China mantle peridotites. In the spinel peridotites and garnet-poor spinel–garnet peridotites, Yb in cpx decreases with increasing degrees of partial melting as Ti decreases. However, in the garnet lherzolites and garnet-rich spinel–garnet lherzolite (Fig. 14), Yb increases in the cpx until most of the garnet is consumed, after which the cpx composition trend parallels that of the garnet-free assemblage. We have taken 15% as the melting fraction at which garnet is consumed, although this will vary with bulk composition. Using the distribution coefficients of Johnson et al. (1990), we find that the evolution trend of cpx during melting of spinel–garnet lherzolite with <2 modal % garnet is similar to that of cpx in spinel lherzolite (Fig. 14).
|
Analysed Ti and Yb contents of clinopyroxenes from the southeastern China localities are shown in Fig. 14c and d. Cpx from the spinel peridotites from all localities closely parallels the calculated trend with a small offset towards higher Yb values. Cpx from the garnet peridotites from Nushan (Fig. 14c) plots along our calculated garnet trend, except for the garnet-poor one (1·2% modal garnet), which plots near the most fertile (least melted) spinel peridotite. The garnet-bearing spinel peridotites all contain <2 modal % garnet and the cpx from these plots along the spinel peridotite trend. Cpx from the Mingxi garnet peridotites parallels the calculated trend for the garnet-bearing assemblages and shows the trend towards the spinel lherzolite cpx melting trend.
Y and Yb modelling
Commonly used cpx melting models are for melting within the spinel lherzolite stability field (Johnson et al., 1990; Niu, 1997; Norman, 1998). These models are not suitable for garnet-bearing lherzolites where garnet is the dominant reservoir of HREE. Therefore, we have redistributed modal garnet to pyroxene and spinel (Table 4), using the reaction 0·17 Olivine + Garnet 
0·47 Clinopyroxene + 0·53 Orthopyroxene + 0·17 Spinel (Hauri & Hart, 1994). Using this equation,, the clinopyroxene REE composition can be corrected for the garnet component before modelling.
Figure 15 shows the clinopyroxene modelling results using these corrected clinopyroxene compositions. The starting compositions for Y and Yb are 4·68 and 0·4875 ppm, and DYCpx/melt = 0·42 and DYbCpx/melt = 0·40. The results of both corrected-Cpx modelling and uncorrected-Cpx modelling (Figs 14 and 15) suggest that the Nushan garnet-bearing lherzolites and most of the Nushan spinel lherzolites experienced <5% partial melting, in either batch or fractional melting regimes, and many have experienced <2% melting. The metasomatized amphibole-bearing lherzolites may also be residues of low degrees of partial melting, assuming that the HREE were not significantly affected by metasomatism. The more depleted Nushan spinel peridotites can be modelled with 10–25% fractional partial melting; these low levels of Ti, Yb and Y would require at least 30–50% of batch melting, which is not likely.
|
Clinopyroxenes from Niutoushan spinel peridotites, Mingxi garnet-bearing peridotites and a few Mingxi spinel peridotites have high Y and Yb contents, consistent with <2% partial melting. Some of the Mingxi spinel peridotites with lower Y and Yb contents require 10–30% fractional melting, like similar rocks from Nushan. Almost all the Qilin spinel peridotites are intermediate between these two groups, suggesting that they have experienced moderate to high degrees (6–25%) of fractional partial melting.
Metasomatism: trace element fingerprints
Clinopyroxenes with LREE-enriched trace element patterns are found in modally metasomatized samples, coexisting with amphibole and/or apatite. Clinopyroxenes with U-shaped trace element patterns may be formed by the combination of at least two events; the most likely scenario includes an earlier stage of partial melting to deplete the rock in LREE and MREE, and a later stage of metasomatism that enriched the rock in LREE and other LILE (Frey & Green, 1974; O’Reilly et al., 1991).
Processes that have been proposed to account for the enrichment of mantle peridotites include fluid or melt percolation in veins and/or porous systems, resulting in modal and chemical variation in the wall-rock peridotite by chromatographic fractionation of infiltrated metasomatic components, or melt retention in the peridotite and re-equilibration with the peridotite minerals (e.g. Navon & Stolper, 1987; O’Reilly & Griffin, 1988; Bodinier et al., 1990; O’Reilly et al., 1991). Proposed metasomatic agents include CO2 ± H2O-rich fluids and alkali-rich silicate to carbonate melts (e.g. Green & Wallace, 1988; O’Reilly & Griffin, 1988; O’Reilly et al., 1991).
Comparison of trace element variation diagrams for Group B clinopyroxenes with those of Groups C and D clinopyroxenes (Fig. 8) shows that in the southeastern China peridotite xenoliths, metasomatic enrichment in LREE is accompanied in most cases by enrichment in Sr, and, to a lesser extent, in Nb. The correlation of Sr with La is well defined in all four localities, whereas the correlations of Nb with La are reasonably well defined for the amphibole-free lherzolite at Nushan, but less so in the other localities (Fig. 16). Ti and Zr show no correlation with LREE enrichment, suggesting that Ti and Zr have not been significantly affected by metasomatism. Rivalenti et al. (1996) have proposed that the apparent immobility of Ti during the metasomatic process in mantle lherzolites may be due to several causes, such as reduced Ti solubility in hydrous fluids, fractionation of Ti-rich phases from percolating silicate melts, or reaction with carbonate-rich melts formerly equilibrated with amphibole peridotite.
|
Ti and Zr anomalies
Negative Ti and Zr anomalies (when plotted in trace diagrams like Fig. 8) are found in clinopyoxenes from many peridotites, and have generated considerable discussion and controversy (e.g. Salters & Shimizu, 1988; Rampone et al., 1991). Rampone et al. (1991) proposed that positive Ti and Zr anomalies in orthopyroxene are complementary to the negative anomalies in clinopyroxene, producing whole-rock patterns without anomalies. We have analysed the trace element contents of orthopyroxenes coexisting with clinopyroxene in Nushan peridotites, and used the modal data to evaluate whether the negative Ti anomalies observed in the clinopyroxenes are balanced by Ti in the orthopyroxenes. However, the Ti positives of orthopyroxene could not match the negatives of clinopyroxene here (Table 5).
|
We have modelled the Ti concentrations at various degrees of partial melting using a range of distribution coefficients and compared them with previous work. Norman (1998) used DTiCpx/melt = 0·35, which is similar to experimental determinations at high temperatures and pressures (e.g. Hart & Dunne, 1993; Jenner et al., 1993), and showed that it was not possible to model the negative Ti anomalies by the same partial melting processes as used for the HREE. However, the negative Zr anomalies in the same clinopyroxenes could be successfully modelled as the result of partial melting, using DZrCpx/melt = 0·12, a value also consistent with several experimental studies. For the present samples we find that both Ti and Zr anomalies in the clinopyroxenes can be roughly accounted for as the result of partial melting processes, using values for DTi and DZr of 0·20–0·25. As Fig. 17 shows, the modelling results are consistent with that of Fig. 15.
|
The experimental studies cited above have measured partitioning between basaltic melts and phenocryst clinopyroxenes (augites), whereas clinopyroxenes in spinel peridotite xenoliths are magnesian Cr-diopsides. The differences in composition and Al site occupancy between these two types of pyroxene could affect partitioning, particularly of high field strength elements (HFSE), which require charge-balancing substitutions. In the phenocryst pyroxenes, Al substitution is divided roughly equally between octahedral and tetrahedral sites [Tschermak’s molecule and titanpyroxene molecule (Yagi & Onuma, 1967; Wass, 1973; Wass, 1979b)] whereas in the xenolith pyroxenes it is primarily in the octahedral sites, as the jadeite molecule, and there is little substitution in the tetrahedral sites. This situation favours Ti substitution in the phenocryst-type pyroxenes, and may produce a high D value in experiments, which is not directly applicable to partial melting processes in peridotites. In contrast, our modelling suggests that the experimental D value for Zr may be lower than the realistic value in peridotites.
We therefore suggest that the values of DTiCpx/melt = 0·20 and DZrCpx/melt = 0·25 are more applicable to partial melting processes in the shallow mantle, and that the negative Ti and Zr anomalies observed in the clinopyroxenes of these xenoliths are simply explained by low degrees of partial melting and were not significantly affected by subsequent metasomatism.
Wiechert et al. (1997) studied the trace element and Sr–Nd isotope compositions of the spinel peridotites from the Atsagin-Dush volcanic centre, southeastern Mongolia. They suggested that the absence of hydrous minerals, ubiquitous CO2-rich micro-inclusions in the enriched samples and negative anomalies of Nb, Hf, Zr, and Ti in primitive mantle-normalized trace element patterns of whole rocks and clinopyroxenes indicate that carbonate melts may have been responsible for the metasomatic enrichment. Experimental studies of the peridotite–H2O system (Green, 1973) constrained models of mantle metasomatism by hydrous silicate melts which may be very enriched in incompatible elements (LILE and HFSE) in the mantle environment. In contrast, experiments on amphibole-bearing peridotites with carbonatite (carbonate-rich) melt metasomatism show that carbonate-rich melts are capable of transporting considerable amounts of LILE, REE, U, Th, P, etc. but have low contents of HFSE (Green & Wallace, 1988).
On the basis of petrographic and geochemical characteristics, we suggest that the metasomatic agent involved in the metasomatism of southeastern China mantle may be a silicate-rich fluid with both hydrous and carbonate components. This fluid was rich in LREE, Sr, and Nb but was probably low in Ti and Zr as it did not obliterate the negative Ti and Zr anomalies produced from the partial melting discussed above. This suggestion is consistent with the common occurrence of both carbonate and glass in Qilin amphibole-bearing lherzolites (Xu et al., 1996), and the predominance of CO2 in primary fluid inclusions in mantle-derived lherzolite xenoliths from Nushan (Xia et al., 1993).
Fertile primitive mantle
Many of the fertile lherzolites from southeastern China have higher Ca and Al contents than commonly accepted estimates of the primitive mantle composition (e.g. McDonough & Sun, 1995). This implies either that these xenoliths represent unusually primitive mantle compositions, or that they have experienced some form of refertilization processes. However, these samples show no obvious modal metasomatism, and the trace element patterns of their clinopyroxenes do not show the enrichment in LREE, Sr, Th and other incompatible elements that is typical of samples showing obvious metasomatism. We therefore regard these compositions as reflecting very low degrees of depletion of primitive mantle.
The most fertile garnet-bearing lherzolites from southeastern China may provide a realistic estimate for a primitive mantle composition. We have estimated this primitive mantle composition beneath southeastern China based on the mean of analytical data for the 10 most fertile, fresh, large xenoliths in Table 1, excluding lherzolites containing modal metasomatic minerals; the estimate is given in Table 6. All of these rocks have Al2O3 and CaO contents >3·2 wt %, mg-number <89.0, and modal clinopyroxene contents >15%. Compared with other fertile mantle compositions, this mantle composition is higher in Al2O3, CaO and FeO, but lower in MgO.
|
Origin of the Cenozoic lithosphere: southeastern China
The Sino-Korean Block of eastern China represents an Archaean craton, with regions reworked during Proterozoic time. Ordovician kimberlites that penetrated this block carry xenocrysts and xenoliths of typical Archaean mantle, and document the existence of a cool continental root extending to depths of 200 km (Griffin et al., 1998a). However, geophysical data show that the lithosphere beneath this region now is thin (<100 km) and has a high geothermal gradient, implying the removal of at least 100 km of the Archaean lithosphere. This lithosphere erosion is believed to have been initiated in late Mesozoic time (Griffin et al., 1998a).
The eastern part of the Cathaysia Block (Fig. 1) where sampled by the Mingxi, Qilin and Niutoushan volcanoes, represents material accreted to the continent during Mesozoic time; much of the crust is as old as Proterozoic, but the lithospheric mantle underlies a region heavily affected by Mesozoic magmatism. Calculations of equilibrium temperature and pressure (Xu et al., 1998) show that the garnet-bearing peridotites studied in this paper are derived from 60–75 km depth, and the spinel peridotites come from 27–60 km depth. These xenoliths therefore provide a sample of the thinner and hotter younger lithosphere.
Most of the spinel peridotite and garnet peridotite xenoliths from southeastern China have relatively low olivine contents and low mg-number, so that they plot within the ‘oceanic trend’ of Boyd (1997; Fig. 6). None fall within the field of Archaean peridotites; this strongly implies that no Archaean lithosphere has been sampled from Nushan, nor from the other localities. The garnet peridotite xenoliths from Nushan and Mingxi are highly fertile rocks, very unlike the strongly depleted garnet peridotites characteristic of Archaean mantle sections (Griffin et al., 1998b). The Zr/Y–Y/Ga data on the garnets (Fig. 13) also indicate that no garnet-bearing Precambrian mantle has been sampled beneath southeastern China since the Palaeozoic. This strongly suggests that the Archaean and Proterozoic lithosphere beneath the Sino-Korean and South China blocks has not simply been thinned, but has been replaced by younger, more fertile material consistent with the seismic tomography data of Yuan (1996a, 1996b).
Both Nushan and Mingxi contain a small proportion of more strongly depleted xenoliths, which have high mg-number combined with relatively low olivine contents, so that they fall off the ‘oceanic trend’ toward the field of Proterozoic xenoliths. At both localities, these more depleted samples are in the shallow part of the section (spinel peridotites). Most of the Qilin samples also are strongly depleted, although all show metasomatic enrichment in LREE and Sr. These more depleted peridotites may suggest that some older lithospheric mantle material is embedded in the Cenozoic lithosphere beneath the southeastern China region.
The fact that many xenoliths from Nushan, Mingxi and Niutoushan plot on or near the ‘oceanic trend’ in Fig. 6 might suggest that they represent oceanic peridotites thrust under the continental margins. However, a detailed comparison with data from abyssal peridotites (Johnson et al., 1990; Niu & Hekinian, 1997; Griffin et al., 1999; Poudjom Djomani et al., 2000); shows that few oceanic peridotites show the low degrees of depletion characteristic of these suites. In addition, most oceanic peridotites have suffered extreme modification of their Ca and Al contents during ocean-floor metamorphism so that in detail they do not resemble the fertile xenoliths studied here. Griffin et al. (1999) have suggested that the ‘oceanic’ trend reflects generalized Phanerozoic processes that have formed subcontinental as well as suboceanic mantle. The very low degrees of depletion observed in these xenoliths suggest to us that they represent little-modified asthenospheric mantle, which has been emplaced ben