Introduction

Syn- and post-collisional igneous activities in the Tibetan Plateau record the evolution of the underlying mantle and lower crust. Previous studies have documented extensive magmatism related to the subduction of Neo-Tethyan oceanic lithosphere and the Indian continental crust (e.g., Turner et al. 1996; Ding et al. 2003; Chung et al. 2005; Mo et al. 2006; 2007). In SW China and Burma, along the southeastern margin of the plateau, recent volcanism is associated with strike-slip fault systems. Two volcanic fields are recognized: one to the north of Wuntho, Central Burma (Stephenson 1984) and one in Tengchong, western Yunnan, SW China (Fig. 1). The lavas include high-K, calc-alkaline to weakly shoshonitic, volcanic rocks, superficially similar to those of arc volcanoes. However, they have very complex geochemical and isotopic compositions, suggesting derivation from a highly variable and composite source.

Fig. 1
figure 1

A simplified geological map showing major tectonic units in the collisional zone between the Indian and Eurasian plates (after Tapponnier et al. 1982; McCaffrey 2009) and the location of the Tengchong volcanic field in SW China. Also shown is the location of the Quaternary volcanic field north of Wuntho, Burma

Full size image

A major Mesozoic suture zone, considered to be the extension of the Yarlung-Zangbo suture in Tibet, lies between these two volcanic fields (Fig. 1), and the Kohima suture to the west probably marks an east-dipping, continental subduction zone. Concurrent, ongoing northward subduction of the Indian Ocean lithosphere has created the active Burma-Sunda volcanic arc to the south (Fig. 1). The convergence of these systems in southwest China and Burma has produced a region with a very complex tectonic history. Little detailed information is available for either the Tengchong or Wuntho lavas, and nothing is known about the nature of the underlying mantle.

Pliocene to Recent volcanic activity in the Tengchong area is recorded in numerous small cinder cones and associated lava flows (Fig. 2). The most recent eruption was in 1609 (Xu 1939), and the volcanoes are considered to be active based on on-going geothermal and earthquake activity. The lavas at Tengchong comprise three units of mafic and intermediate rocks (trachybasalt, basaltic trachyandesite, and trachyandesite) and one unit of dacite. The earliest description of the volcanoes is in a report by Xiake Xu, a noted Chinese geographer, published in 1639 (Xu 1939; also see Liu 1999). The volcanoes were also described by Anderson (1871, 1876), Széchenyi (1893), Brown (1913), Zhang (1924) and Chang (1948). Modern scientific investigation of these rocks by Zhu et al. (1983) suggested that the most voluminous lavas were produced by partial melting of a metasomatized mantle source. Mu et al. (1987a) argued that the isotopic and geochemical characteristics of the rocks point to a continental component derived either by assimilation or by recycling of crustal material into the magma source. Cong et al. (1994) and Wang et al. (2006) suggested that the rocks were derived by partial melting of a metasomatized mantle source previously modified by oceanic subduction at ~65 Ma. Wang et al. (2006) recognized two source domains for the lavas; one that produced a LREE-enriched, high Rb/Sr end-member marked by 230Th enrichment and the other that produced a typical arc-type melt with lower 230Th. These studies demonstrated that the volcanic rocks in the southeast most margin of the Tibetan Plateau have arc-like characteristics but that they are more K-rich than typical arc assemblages and are significantly different from volcanic rocks elsewhere in the plateau.

Fig. 2
figure 2

Distribution of Quaternary lavas in the Tengchong volcanic field (a) and sampling locations (b). Four eruption units are recognized (1–4), from oldest to youngest (adopted from an unpublished geological map of 1:50,000 scale by Geological Survey of Yunnan 2000)

Full size image

The eruption of the Tengchong volcanoes north of the Burma arc raises many questions regarding the nature of the mantle source, composition of parental magma(s), the roles of crustal contamination and fractional crystallization in the magma evolution, the source of the heat that triggered melting and the nature of the overall tectonic environment in which the volcanoes formed. Answers to these questions are obscure because no integrated study of the petrology, mineralogy and geochemistry of these rocks is available.

In this paper, we describe the most primitive lavas that erupted during three stages of volcanic activity. The dacite unit involved crustal melting and is not considered in this paper. The new data set including mineral chemistry and whole-rock major oxide, trace element and Sr–Nd–Pb–Hf isotopic compositions is used to document some unique features of the volcanic rocks compared with typical arc rocks and to identify a highly heterogeneous mantle source that was modified by variable amounts of subduction components. We propose that the underlying mantle was originally modified by subduction of Neo-Tethyan Oceanic lithosphere followed by subduction of the Indian continental lithosphere (~40 Ma). The modern activity in Tengchong records recent modification of the mantle by subduction of present-day Indian Ocean lithosphere, perhaps including the Ninety East Ridge.

Geological background

The Tibetan Plateau consists of the Kunlun, Songpan-Ganzi, Qiangtang, and Lhasa terranes to the north and the Indian sub-continent to the south. These terranes were successively accreted to the Asian continent since the Paleozoic. Sinistral strike-slip motion along several major faults accommodated southeastward escape of the Tibetan Plateau since ~27 Ma (Tapponnier et al. 1982; Socquet and Pubellier 2005) and continues to the present time. The major terranes extend to the east and southeast, partially due to this extrusion of the plateau.

In the southeastern margin of the plateau, from west to east, these terranes include the Indian continent, Tibet-Yunnan Fold System, Indo-China Fold System and Yangtze Block, all of which are separated by major faults (Fig. 1; Yan et al. 2006). The Tibet-Yunnan Fold System is separated from the Yangtze Block by the Red-River fault to the east and from the Burma Central Lowlands to the west by the Mali Hka-Mandalay suture (Fig. 1). This suture extends to the Indus-Yarlung Zangbo suture zone in the northwest and is marked by ophiolites formed in the Neo-Tethyan Ocean. Farther to the west, the Burma Central Lowlands are separated from the Indian continent by the Kohima suture, which extends to the main boundary thrust to the northwest (Fig. 1).

The collision between the Asian and Indian continents was associated with extensive magmatism. The subduction of Neo-Tethyan Oceanic lithosphere produced the Gangdese granitoid batholiths and the 5-km-thick Linzizong volcanic succession (Chung et al. 2005; Mo et al. 2007). In the Tengchong area, the 68–53 Ma granites emplaced at are considered to be the continuation of the Gangdese magmatic arc of the Lhasa terrane.

Post-collisional volcanism began immediately after the collision and has continued to the present (Turner et al. 1996; Ding et al. 2003; Nomade et al. 2004; Williams et al. 2004; Chung et al. 2005; Guo et al. 2006; Mo et al. 2006; Zhao et al. 2009). This magmatism generated potassic to ultrapotassic volcanic rocks in the Lhasa and Qingtang terranes. High-potassium igneous rocks are also widespread along the Red River fault, just north of the Tengchong region (Wang et al. 2001). They show Sr–Nd isotopic compositions that resemble those of the Indian crustal basement. This has been taken as evidence for subduction of the Indian continental crust underneath the Lhasa terrane since the Oligocene (Zhao et al. 2003; 2009; Guo et al. 2006).

The Tengchong volcanic field in the Tibet-Yunnan Fold System is located within pull-apart basins produced in a stress field of NNE-NE-compression and WNW-NW-extension (Wang et al. 2007). The volcanoes occur along the Gaoligong dextral strike-slip fault near its intersection with the Ruili-Mandalay sinistral strike-slip fault (Fig. 1). The Gaoligong metamorphic complex is exposed in the Gaoligong mountain range to the east (YBGMR 1979) and is thought to form the basement in the region. The basement complex is intruded by numerous Late Mesozoic to early Cenozoic arc granites (YBGMR 1979) and overlain by pre-Quaternary sedimentary strata.

Field relations and petrography of volcanic rocks

In the Tengchong volcanic field, the volcanoes are distributed in an area ~90 km long from north to south and ~30 km wide (Fig. 2). More than 90 individual cones are recognized (Fig. 2), and many of these are associated with short, stubby flows, up to about 35 m thick, which typically show well-developed columnar jointing (Fig. 2).

Based on 230Th–238U, K/Ar and thermoluminescence dating, the Tengchong volcanic rocks range in age from 5.5 Ma to present (Mu et al. 1987a, b; Li et al. 2000; Yin and Li 2000; Wang et al. 2006). However, many of these dates are of unknown quality, and most of the rocks are very young. The most recent eruption was in 1609, and the field is considered to still be active.

According to rock types, four principal lithologic units are recognized in the Tengchong area (Fig. 2). Unit 1, generally considered the oldest, consists chiefly of olivine trachybasalt and basaltic trachyandesite, whereas a distinct sequence of hornblende dacite makes up Unit 2. Relatively uniform basaltic trachyandesite comprises Unit 3, and Unit 4, the youngest, consists of trachyandesite.

Units 1 and 3 lavas are fine-grained, moderately phyric, sparsely vesicular, lavas (see Electronic Appendix 1). Phenocrysts of olivine, plagioclase and sparse augite make up ~10–15 modal% and are set in fine-grained, intersertal to intergranular groundmass. The olivine phenocrysts exhibit both oscillatory and reverse zoning with Fo values ranging from 84 to 75, whereas the groundmass olivine has Fo values ranging from 64 to 59 (Appendix 2). Some of the olivine phenocrysts contain small chromite inclusions with Cr#s (100Cr/(Cr + Al)) of 23–35 and Mg#s (100 Mg/(Mg + Fe2+)) of 45–67 (Appendix 3). On a plot of TiO2 versus Al2O3, the chromite grains all plot in, or near, the MORB field and have much higher Al2O3 than chromites from island arc, OIB and LIP mafic–ultramafic rocks and much higher TiO2 than island arc volcanic rocks and depleted mantle peridotites (Fig. 3). Plagioclase compositions are similar in both Units 1 and 3, generally ranging from An68 to An59 in the large phenocrysts and from An63 to An55 in the groundmass (Appendix 4). Some very large plagioclase phenocrysts in Unit 1 exhibit reverse zoning. Clinopyroxene is abundant in the groundmass of Units 1 and 3 and also forms small phenocrysts (Appendix 1). It has a relatively uniform composition, ranging from En48Fs10Wo42 to En44Fs15Wo41 (Appendix 5). Minute crystals of magnetite and ilmenite are also common in the groundmass (Appendix 3).

Fig. 3
figure 3

Plots of Al2O3 versus TiO2 of chromite from the lavas at Tengchong. Reference fields after Kamenetsky et al. (2001). LIP large igneous provinces, OIB intra-plate ocean islands, ARC island arcs, MORB mid-ocean ridge basalts, SSZ peridotite spinel from supra-subduction zone, and abyssal peridotite-spinel from mid-ocean ridge

Full size image

The rocks of Unit 2 are fine-grained, moderately phyric, hornblende dacite (Appendix 1). They contain 20–25 modal% phenocrysts, chiefly plagioclase and hornblende, with minor amounts of biotite and pyroxene, all set in a glassy groundmass. Small plagioclase phenocrysts are accompanied by relatively large corroded plagioclase xenocrysts with reverse zoning. The hypersthene laths are commonly corroded and rimmed, or partly replaced by, clinopyroxene or hornblende. These features suggest a complex origin of the dacite involving crustal melting, and this unit will not be discussed further in this paper.

Unit 4 consists of very fine-grained, moderately to weakly porphyritic, trachyandesites with a glassy groundmass. Phenocrysts make up ~5–12 modal% and consist chiefly of pyroxene with lesser amounts of plagioclase (Appendix 1). The plagioclase phenocrysts range from An63 to An49, with the large crystals being more calcic than the smaller ones. Plagioclase in the groundmass has a narrow range of composition (An55–52; Appendix 4). Most of the pyroxene is augite (En44–48), but some laths of hypersthene (En74–77; Appendix 4) are also present and many of them have narrow rims of augite. A few of larger hypersthene grains have narrow exsolution lamellae of clinopyroxene. All of the rocks are very fresh with no observable secondary minerals. A single olivine xenocryst in Unit 4 is normally zoned from Fo77 in the core to Fo73 in the rim.

The rocks of Units 1, 3 and 4 all contain 1–5 modal% of xenocrysts, chiefly plagioclase and quartz (Appendix 1). The plagioclase xenocrysts occur as anhedral to subhedral laths up to 5 mm long, with fritted margins, some of which are surrounded by clear overgrowth rims. A few of the grains also have strongly pitted cores. Quartz forms small, subrounded grains, 0.3–0.5 mm across, which are commonly rimmed with clinopyroxene (Appendix 1). One sample (TC-6) has numerous clots of fine-grained pyroxene, 0.2–0.3 mm across, that were probably originally small quartz xenocrysts.

Analytical techniques

Major and trace elemental analyses

Whole-rock major oxides were analyzed by X-ray fluorescence on fused glass beads at The University of Hong Kong, following the analytical procedures of Zhou et al. (2006). Analytical precision is better than 1% for SiO2 and better than 2% for the other oxides. Trace elements were analyzed by inductively coupled plasma-mass spectrometry (ICP-MS) also at The University of Hong Kong, according to procedures described by Zhou et al. (2006). Analytical precision for most elements is better than 5%.

Sr–Nd–Pb–Hf isotope analyses

Whole-rock Sr–Nd–Pb–Hf isotopes were measured at the Institute of Geology and Geophysics, Chinese Academy of Sciences. Sr–Nd isotopes were analyzed with a Finnigan MAT 262 thermal ionization mass spectrometer. Measured 87Sr/86Sr and 143Nd/144Nd ratios were normalized to 86Sr/88Sr = 0.1194 and 146Nd/144Nd = 0.7219, respectively. The reported 87Sr/86Sr and 143Nd/144Nd ratios were adjusted to the NBS SRM 987 standard 87Sr/86Sr = 0.71025 and the Shin Etsu JNdi-1 standard 143Nd/144Nd = 0.512115.

Pb isotopes were separated by anion exchange techniques with diluted HBr as an eluant. Repeated analyses of standard NBS 981 yielded average values of 204Pb/206Pb = 0.05897 ± 15, 207Pb/206Pb = 0.91445 ± 80 and 208Pb/206Pb = 2.16170 ± 180.

Whole-rock Lu–Hf isotopic analyses were conducted using a Thermo Electron Neptune multicollector (MC)-ICP-MS. The detailed procedure for sample preparation is available in Yang et al. (2007). 176Hf/177Hf measurements were normalized to 179Hf/177Hf = 0.7325. Repeated analysis of the reference material, BHVO-1, used to monitor the precision, yielded an average 176Hf/177Hf ratio of 0.283105 ± 6 (2σ), in agreement with the recommended value of 0.283100 ± 3 (Weis et al. 2005).

Analytical results

Major element oxides and trace elements

All of the rocks from Units 1, 3 and 4 at Tengchong contain <1 wt% loss on ignition (LOI; Table 1 and also Appendix 6 for complete data set). In the TAS classification diagram, the Tengchong lavas range from basalt, trachybasalt, basaltic trachyandesite to trachyandesite (Fig. 4a). All of the rocks have higher Na2O than K2O and plot on the boundary between the high-K calc-alkaline and shoshonitic fields (Fig. 4b, c). K2O has a well-defined positive correlation with SiO2 (Fig. 4b).

Table 1 Major and trace elements of representative lavas from Tengchong, SE Tibetan Plateau
Full size table
Fig. 4
figure 4

Total alkali versus silica (TAS) (a), K2O versus SiO2 (b) and Na2O versus K2O (c) (Le Maitre et al. 1989) for lavas at Tengchong

Full size image

Olivine trachybasalts and basaltic trachyandesites from Units 1 and 3 have SiO2 contents of 48.5–55.8 wt% and total alkalis ranging from 4.9 to 7.8 wt%. They have a relatively narrow range of MgO (4.2–6.2 wt%), Al2O3 16.2–17.9 wt%, Fe2O3T (7.0–10.8 wt%) and TiO2 (1.1–1.8 wt%; Table 1). On Harker diagrams, Al2O3, MgO, Fe2O3t, CaO and TiO2 show moderately good negative correlations with SiO2.

The trachyandesites of Unit 4 have relatively high SiO2 (56.9–61.0 wt%) and in most cases, lie on the projection of fractionation trends defined by rocks of Units 1 and 3 (Fig. 4a–c). However, these rocks together with some Unit 1 lavas are significantly enriched in P2O5 and form a distinct trend on the Harker diagram (Fig. 5d).

Fig. 5
figure 5

Harker diagrams of SiO2 versus Fe2O3total, MgO, CaO, P2O5, Ta and Zr for the lavas at Tengchong

Full size image

Like the major oxides, trace elements of Units 1 and 3 lavas show generally linear trends with limited scatter. MgO shows a generally negative correlation with Ta and Zr (Fig. 5e, f). The rocks of Unit 4 form distinct trace element clusters having higher Rb, and Zr than the lavas of Units 1 and 3 and moderate Sr and Ta. All the lavas of Units 1, 3 and 4 display variable LREE enrichment (La/YbN = 8.2–22) and weak negative Eu anomalies (Eu/Eu* = 0.98–0.70) in the chondrite-normalized REE patterns (Fig. 6a–c). In the primitive mantle-normalized trace element diagram, the rocks are characterized by enrichment of LILE and LREE and depletion of Sr and HFSE (e.g., Nb and Ta) and show negative Ti anomalies (Fig. 6d–f).

Fig. 6
figure 6

Chondrite-normalized REE patterns (a–c) and primitive mantle-normalized trace element patterns (d–f) for the lavas at Tengchong. Normalization values after Sun and McDonough (1989)

Full size image

Sr–Nd–Pb–Hf isotopes

The rocks from Units 1 and 3 have approximately similar Sr–Nd–Pb–Hf isotopes (Fig. 7) with 87Sr/86Sr ratios ranging from 0.7058 to 0.7073, εNd values from −3.1 to −6.1 (Table 2), 206Pb/204Pb from 18.0 to 18.3, 207Pb/204Pb from 15.6 to 15.7 and 208Pb/204Pb from 39.0 to 39.4 (Table 3). The trachyandesites of Unit 4 have essentially constant and more radiogenic 87Sr/86Sr ratios and lower εNd values than those of the rocks of Units 1 and 3 (87Sr/86Sr = ~0.7085 and εNd = −8.4 to −8.7). However, the Pb isotopes are similar to those of Units 1 and 3 (Fig. 7b, c).

Fig. 7
figure 7

Plots of Sr–Nd-Pd-Hf isotopic compositions for the lavas at Tengchong. a 206Pb/204Pb versus 87Sr/86Sr(t); b 206Pb/204Pb versus initial 143Nd/144Nd; c 206Pb/204Pb versus 208Pb/204Pb; and d εHf versus εNd. Data sources: reference fields in a–c are from Zindler and Hart (1986), Hart (1988) and Hart and Zindler (1989); Reference fields for MORB and OIB in d are after Vervoort et al. (1999). The mantle array (εHf = 1.33; εNd + 3.19) is adapted from Vervoort et al. (1999) and the igneous rock array (εHf = 1.37; εNd + 2.89) from Bayon et al. (2009)

Full size image
Table 2 Sr–Nd–Hf isotopes of representative lavas from Tengchong, SE Tibetan Plateau
Full size table
Table 3 Pb isotopic data of representative lavas from Tengchong, SE Tibetan Plateau
Full size table

The isotopic signatures of rocks from Units 1 and 3 are similar to those of an EMI mantle source (Fig. 7a–c). The Pb isotopic compositions of all the rocks plot to the left of the northern hemisphere reference line (NHRL; Fig. 7c).

The rocks of Units 1 and 3 have εHf values from +4.8 to +0.8, whereas the rocks of Unit 4 have lower values, ranging from −6.2 to −6.4 (Table 2). The Nd–Hf isotope data for the Tengchong volcanic rocks are not distributed along the mantle array (Vervoort et al. 1999) but plot along the ‘zircon-free sediment array’, which is between the ‘seawater array” and ‘igneous rock array’ (Bayon et al. 2009; Fig. 7d).

Discussion

Magma differentiation and crustal contamination

Magma differentiation

The chemical trends for all the rocks (Fig. 5) are generally compatible with fractionation of olivine, plagioclase, pyroxene, apatite and magnetite, the observed phases in the lavas. The chromite grains in the lavas of Units 1 and 3 likely crystallized as an early phase from magmas with MORB-like compositions. The most magnesian olivine of the lavas at Tengchong has a composition of Fo84.8, yielding a Mg# of 0.63 and an initial MgO/FeO ratio of 0.93 using the equation of Roeder and Emslie (1970). These values are higher than the Mg#s and MgO/FeO ratios of the olivine-bearing rocks of Units 1 and 3 (Table 1), suggesting significant fractional crystallization before eruption. Olivine phenocrysts in rocks of Units 1 and 3 have Fo values ranging from 84.8 to 71.2, whereas the groundmass olivine has Fo values as low as 59.2. The wide range of Fo values indicates continued crystallization during transport and eruption of the lavas.

It has been traditionally thought that all the rocks at Tengchong have continuous chemical trends and formed simply by differentiation of a single parental magma (Zhu et al. 1983; Cong et al. 1994; Wang et al. 2006). However, a close examination of the Harker diagrams (Figs. 4, 5) shows several differentiation trends. For example, in the plot of CaO versus SiO2, rocks of Units 3 and 4 show slightly different trends (Fig. 5c). Plots of P2O5 versus SiO2 reveal two parallel trends, one including some of the basaltic trachyandesites of Unit 1 and the trachyandesites of Unit 4 and the other including the trachybasalts of Unit 1 and trachyandesites of Units 1 and 3 (Fig. 6e). These trends are difficult to link by fractional crystallization of a single magma and suggest the presence of several slightly different parental magmas. Most incompatible trace elements (e.g., Ta and Zr) in the rocks of Units 1, 3 and 4 correlate negatively with MgO (Fig. 5e, f), demonstrating some assimilation-fractional crystallization (AFC) processes in their evolution.

Crustal contamination

Earlier isotopic studies suggested that the Tengchong lavas underwent very little contamination (Zhu et al. 1983; Chen et al. 2002; Zou et al. 2010), despite the presence of abundant crustal xenocrysts (Fig. 3e, f). The rocks of Units 1 and 3 are mafic in composition and have similar, relatively low 87Sr/86Sr ratios, and εNd and εHf values, indicating derivation from a common source and ruling out involvement of significant crustal material. Rocks of Unit 4 have much higher 87Sr/86Sr ratios and much lower εNd and εHf values than those of Units 1 and 3 (Table 2), and follow clear crustal contamination trends (Fig. 7b). Thus, we conclude that the rocks of Units 1 and 3 have undergone very little crustal contamination but that those of Unit 4 have been significantly modified.

Staging magma chambers

The Tengchong region is considered an active volcanic field because magma still exists at shallow levels. Active geothermal systems at Tengchong are well known (Tong and Zhang 1989; Huangpu and Jiang 2000; Du et al. 2005), and a low-velocity layer at a depth of 3–10 km, identified by seismic sounding data, and earthquake results (Qin et al. 2000; Bai et al. 2002; Zhao et al. 2006) are consistent with the presence of magma chambers in the upper crust.

Crystal fractionation recorded in the lavas at Tengchong may have occurred during and after magma emplacement into crustal chambers or on the way from the chambers to the surface. The majority of differentiation probably took place in one or more staging magma chambers in the middle or lower crust. Slightly different chemical trends of the different lava groups can best be explained by eruption of batches of new, variably evolved and/or contaminated magma from staging chambers.

Nature of the mantle source and its heterogeneity

The source of the lavas at Tengchong would have been a MORB-like mantle as demonstrated by Cr-spinel compositions, which are similar to those of MORB (Fig. 3). Thus, we suggest that the magmas were formed at a level deeper than normal arc volcanic rocks. This difference is further reflected in the higher TiO2 contents and Ti/V ratios of the Tengchong lavas than typical arc rocks (Table 1, Fig. 8a). In the plot of La/Nb versus Ba/Nb, the Tengchong rocks fall between the arc volcanic and Dupal OIB fields (Fig. 8b). They have higher Nb/Y ratios than arc basalts (Fig. 8c) and plot in the field of mixed asthenospheric mantle sources, having variable but lower Nb/La ratios than average oceanic island basalts (OIB; Fig. 8d).

Fig. 8
figure 8

Plots of elements and elemental ratios for the lavas at Tengchong: a Ti versus V (after Shervais 1982); b Nb/Y versus Ti/Y (Pearce 1982); c La/Yb versus Nb/La (after Watson 1993); and d La/Nb versus Ba/Nb (after Jahn et al. 1999) with data sources: PM primitive mantle (Sun and McDonough 1989); CC average continental crust (Taylor and McLennan 1985; Condie 1993); Clastic sediment average (Condie 1993); MORB, OIB and Dupal OIB (Le Roux 1986)

Full size image

The least contaminated lavas from Units 1 and 3 are enriched in LREE (Fig. 6a, b), consistent with a garnet-bearing mantle source (c.f. van Westrenen et al. 2000). Rocks from Unit 1 have higher Sr/Y ratios (>15) than normal arc rocks, consistent with melting at the garnet stable depth. The lavas at Tengchong have Sm/Yb ratios higher than the spinel lherzolite melting curve, but lower than the garnet-lherzolite melting trend (Fig. 9a). The rocks all plot near the garnet + spinel lherzolite mantle source in the Sm/Yb versus La/Sm diagram (Fig. 9b).

Fig. 9
figure 9

Plots of a Sm/Yb versus Sm and b Sm/Yb versus La/Sm for the lavas at Tengchong. Mantle array (heavy line) defined by depleted MORB mantle (DMM, McKenzie and O’Nions 1991) and primitive mantle (PM, Sun and McDonough 1989). Melting curves for spinel lherzolite (Ol53 + Opx27 + Cpx17 +Sp11) and garnet peridotite (Ol60 + Opx20 + Cpx10 +Gt10) with both DMM and PM compositions are after Aldanmaz et al. (2000). Numbers along lines represent the degree of the partial melting. c Sr/Ce versus Th showing a negative correlation; d Th/Ba versus Th/La (after Gasperini et al. 2000); e Th/Yb versus Ta/Yb (Pearce 1983); f Ba/Zr versus Th/Zr, the dashed arrow line is the mixing line between the mantle wedge and AOC fluid, and the solid arrow line is the mixing trend between the mantle wedge and bulk sediment (Ishizuka et al. 2003); g Th/Zr versus Nb/Zr; and h Nb/Y versus Rb/Y. Reference fields in both d and e are from Kepezhinskas et al. (1997)

Full size image

The mantle source may have been heterogeneous, because the rocks of Units 1 and 3 showing the least crustal contamination have variable isotopic compositions (Fig. 7). They show two different trends in particular in the plot of SiO2 versus P2O5 (Fig. 5d). They also have different Nb/Y, La/Yb and Nb/La ratios (Fig. 8c, d), clearly reflecting variable compositions of the mantle sources because these elements have similar partition coefficients during mantle melting and fractionation.

We conclude that the rocks at Tengchong were generated from a mantle source that was isotopically and chemically heterogeneous due to various degrees of enrichment of LREE and depletion of HFSE (Ti, Nb and Ta; c.f. Arndt et al. 1998). Melting of this highly heterogeneous mantle produced a range of primary magmas, which were injected into shallower staging magma chambers where they experienced variable fractionation and crustal assimilation before final eruption.

Subduction-modified mantle source

Lavas at Tengchong have an arc-like geochemical affinity with an EMI-like mantle source as demonstrated by their isotopic compositions (Fig. 9). Such a source is traditionally explained to be subduction-related and likely modified by slab-related fluids and crustal materials (Alabaster et al. 1982). Because Sr is more compatible than Ce in phlogopite (LaTourrette et al. 1995), the negative correlation between Sr/Ce and Th in the Tengchong lavas (Fig. 9c) may reflect the presence of phlogopite in the mantle (c.f. Yang et al. 2003) due to fluid-related metasomatism. Such a fluid would likely have been derived from subducted altered oceanic crust and associated sediments.

The enrichment of Th over LREE and LILE provides a possible key to the origin of these rocks and their mantle source. The abundance of highly incompatible elements (LILE, e.g., Rb, Sr and Ba) in magma sources is generally controlled by fluid influx (Stolper and Newman 1994). Compared to LILE, Th and HFSE (Zr, Hf, Nb and Ta) are relatively immobile in aqueous fluids (Tatsumi 1986). Therefore, enrichment of LILE and LREE is indicative of involvement of slab fluids, whereas enrichment of REE and HFSE in a mantle wedge indicates an introduction of slab melts (Elliott et al. 1997; Plank and Langmuir 1998). Although high Th/Ba and Th/La ratios of the lavas at Tengchong clearly point to melt-related enrichment (Fig. 9c), relatively low (Th/Ba)N and (Th/La)N ratios of olivine trachybasalts of Unit 1 may indicate that the source was modified by slab-derived fluids.

LILE enrichment can be examined using a plot of Th/Yb versus Ta/Yb, where Yb-normalization is used to eliminate the effects of partial melting and fractional crystallization (Pearce 1982). Melt-related enrichment processes affect Th and Ta equally because of similar partition coefficients during mantle melting and would produce magmas which lie along the “within-plate” trend in the plot. However, lavas at Tengchong all plot above the “within-plate” array with much higher Th/Yb ratios, even higher than many subduction-related rocks (Fig. 9e). This selective enrichment of Th can be best explained by greater mobility of LILE in hydrous fluids relative to HREE and HFSE.

Hydrous fluids from subducted slabs can be derived either from altered oceanic crust (AOC) or from subducted sediments, and these sources can be clearly distinguished in plots of Ba/Zr versus Th/Zr (Ishizuka et al. 2003). The lavas at Tengchong have high Ba/Zr and Th/Zr ratios relatively to Pacific Ocean and Philippine Sea MORB and clearly plot between the fields of AOC fluids and bulk sediment (Fig. 9f), confirming that large amounts of fluid from altered oceanic crust and sediment were added to the mantle source. Melt-related enrichment of the source region is also demonstrated in Fig. 9g, h.

The high 208Pb/204Pb ratios of the lavas at Tengchong relative to MORB support involvement of a sediment component in the genesis of these rocks. The interpretation of a mantle source modified by sediment-derived Pb is favored over the introduction of radiogenic Pb into the ascending magma at upper-crustal levels, because of the similar Pb isotopes of the three units. They have relatively high εHf and εNd values and therefore plot above the igneous rock array, lying between the seawater and zircon-free sediment arrays (Fig. 7d), suggesting that the sediment component in the source regions was mud rather than sand (Chauvel and Blichert-Toft 2001; Eisele et al. 2002; Chauvel et al. 2008).

An integrated model for the Tengchong volcanoes

The lavas at Tengchong have relatively high K2O, plotting in the field of K-rich calc-alkaline lavas and shoshonites (Fig. 4b), but are distinctly different from the post-collisional potassic and ultrapotassic volcanic rocks in Tibet that were generated by subduction of continental crust. Although they have an affinity to rocks produced by active subduction, their TiO2 and Nb contents are higher than those of typical intra-oceanic arc basalts (TiO2 < 1wt% and Nb < 2 ppm) and of subduction-related calc-alkaline andesites. Their high TiO2 and P2O5 contents are more comparable to ocean island lavas. Thus, the unusual compositions indicate derivation from a variably modified mantle source.

The Mali-Hka Mandalay suture zone west of Tengchong contains ophiolites like those of the Yarlung-Zangbo suture in Tibet and the two are thought to be part of the same zone, marked by subduction of Neo-Tethyan oceanic subduction prior to the collision of India and Eurasia (Fig. 1). Subduction of Neo-Tethyan lithosphere produced the Gangdese arc rocks, such as the 70–40 Ma Linzizong volcanic succession (Chung et al. 2005, Mo et al. 2007), and the Mesozoic granitoids in the Tengchong area. Closure of the Neo-Tethyan Ocean at either Late Cretaceous (~65 Ma) or Oligocene (~34 Ma; Yin and Harrison 2000; Aitchison et al. 2007) was followed by subduction of Indian continental crust beneath Eurasia. In the Gaoligong area, north of Tengchong, ~40-Ma mafic dykes and lavas are thought to be related to slab break-off in the early stages of continental collision and thus record the final closure of the ocean (Xu et al. 2008, Gao et al. 2010).

Post-collisional magmatism associated with this subduction generated the potassic and ultrapotassic volcanic rocks of the Tibetan Plateau (Mo et al. 2006; Zhao et al. 2009). Potassic and ultrapotassic rocks along the Red River Fault show Sr–Nd isotopic compositions similar to those of the Indian continental basement, providing evidence for subduction of the Indian continental crust underneath Tibet and SW China (Wang et al. 2001; Zhao et al. 2003; Guo et al. 2006). Seismological data indicate the presence of a modern, eastward-dipping zone of earthquake foci beneath the Burma Central Lowlands (Huan et al. 1981; Lei et al. 2009).

In contrast to other parts of the Tibetan Plateau, subduction of the Indian continental crust beneath Burma and SW China was later replaced by the subduction of the Indian Oceanic plate. This plate continues to move northward and is currently being subducted beneath Burma and SW China (Fig. 10a). A hydrous zone under Tengchong, defined by geophysical data (Bai et al. 2002), likely reflects dewatering of this slab. High-resolution P-wave tomography of the Tengchong region indicates a broad low-velocity zone about 100 km wide that extends down to the 410-km discontinuity (Lei et al. 2009; Zhao and Liu 2010). The modern volcanism at Tengchong was recently explained to be probably related to deep subduction of the Indian Ocean plate along a discontinuity at ~660 km (Zhao and Liu 2010; Fig. 10b). However, the oceanic slab at this depth would have already dehydrated so that it cannot explain the low-velocity zone at the 410-km discontinuity described by Zhao and Liu (2010). This portion of the Indian Ocean plate contains the northern end of the Ninety East Ridge (Figs. 1, 10a). This ridge was formed by eruptions from the Kerguelen hotspot during the last 130 million years (Muller et al. 1993) and drifted northward and then northeastward near the Tengchong region at 110 Ma (Storey et al. 1989). Thus, an older portion of the Indian Ocean crust and the Ninety East Ridge is likely somewhere beneath the Tengchong area. The plateau ridge would be more buoyant than the adjacent normal oceanic lithosphere, causing a kink, or tear, in the downgoing slab, for a much flatter subduction (Fig. 10c). Therefore, we propose that the volcanism in Tengchong was due to the much flatter Ninety East Ridge subduction likely along the ~410 km discontinuity (Fig. 10d).

Fig. 10
figure 10

a Plate tectonic setting of SE Asia (After McCaffrey 2009). The Ninety East Ridge is marked by volcanic rocks related to the Kerguelen hotspot (Seno and Rehman 2011). The hotspot traces on the Indian plate over the Kerguelen hotspots younger than 100 Ma are from Muller et al. (1993) and are shown by the blue lines with ages; it is noted that the traces of 110 Ma are from Storey et al. (1989); b Interpretation of P-wave tomography passing through the Tengchong volcanic field (After Zhao and Liu 2010). c A schematic diagram showing the subduction of the buoyant Ninety East Ridge. d New interpretation of the geophysical data

Full size image

The present-day ocean plate subduction was preceded by subduction of Indian continental crust, which presumably added a continental component to the source region that was responsible for the evolved isotopic compositions observed in the lavas. This modification by the Indian continental crust may also have produced the K2O enrichment of the lavas at Tengchong. The ridge has an OIB component (Storey et al. 1989) which could explain some of unusual features of the Tengchong lavas, including the OIB and arc characteristics of the lavas. The OIB-type volcanic rocks of the ridge may be more resistant to melting during subduction than MORB, thus metasomatizing the mantle wedge at a depth deeper than the source required for normal arc volcanoes. The ridge subduction would have brought altered basalts, water and sediment to the wedge to form a heterogeneous mantle source.

Thus, the unusual compositions of the lavas at Tengchong reflect long-lived and repeated modification of the mantle source by subduction of oceanic and continental lithosphere. The generation and eruption of the magmas are likely linked to transtensional faulting in the region. Both the Tengchong and Wuntho volcanic fields lie along major, right-lateral, transtensional faults, which are known to extend into the lower crust and probably the upper mantle (Qin et al. 2000). Movement on these two faults may have produced local extension, increased crustal permeability and enhanced the delivery of mantle-derived melts to the overlying crust to allow magma accumulation and eruption.

Conclusions

Lavas at Tengchong include olivine trachybasalt, basaltic trachyandesite, trachyandesite and hornblende dacite, typical of an arc-like assemblage. However, they are unique in terms of their high TiO2 and P2O5 contents and extremely high Th relative to typical arc volcanic rocks. Olivine trachybasaltic and trachyandesitic rocks were derived from variably differentiated magmas, but their compositions were primarily controlled by a heterogeneous mantle source and variable degrees of crustal contamination. The mantle source was metasomatized by the subduction of the earlier Indian continental crust and then the Indian Oceanic plate, probably the Ninety East Ridge. The current modification was produced by the introduction of slab fluids and melts from the subducted ridge. Strike-slip faulting may have initiated the present-day magmatism and facilitated the upward transport of mantle-derived melts. Subsequent processes controlled the compositional variations within individual lava units. These include minor contamination by continental crust and variable degrees of silicate and oxide fractionation in staging magma chambers.