Dicheng Zhu a,b,c, Xuanxue Mo b,, Guitang Pan a, Zhidan Zhao b, Guochen Dong b, Yuruo Shi c, Zhongli Liao a, Liquan Wang a, Changyong Zhou a

Transcription

1 Available online at Lithos 100 (2008) Petrogenesis of the earliest Early Cretaceous mafic rocks from the Cona area of the eastern Tethyan Himalaya in south Tibet: Interaction between the incubating Kerguelen plume and the eastern Greater India lithosphere? Dicheng Zhu a,b,c, Xuanxue Mo b,, Guitang Pan a, Zhidan Zhao b, Guochen Dong b, Yuruo Shi c, Zhongli Liao a, Liquan Wang a, Changyong Zhou a a Chengdu Institute of Geology and Mineral Resources, Chengdu, China b State Key Laboratory of Geological Processes and Mineral Resources, China University of Geosciences, Beijing, China c Beijing SHRIMP II Center, Institute of Geology, Chinese Academy of Geological Sciences, Beijing, China Received 1 April 2006; accepted 8 June 2007 Available online 8 August 2007 Abstract The relationship between the breakup of eastern Gondwanaland and the Kerguelen plume activity is a subject of debate. The Cona mafic rocks are widely exposed in the Cona area of the eastern Himalaya of south Tibet, and are studied in order to evaluate this relationship. Cona mafic rocks consist predominantly of massive basaltic flows and diabase sills or dikes, and are divided into three groups. Group 1 is composed of basaltic flows and diabase sills or dikes and is characterized by higher TiO 2 and P 2 O 5 content and OIB-like trace element patterns with a relatively large range of ɛnd(t) values (+1.84 to +4.67). A Group 1 diabase sill has been dated at 144.7±2.4 Ma. Group 2 consists of gabbroic sills or crosscutting gabbroic intrusions characterized by lower TiO 2 and P 2 O 5 content and depleted N-MORB-like trace element patterns with relatively higher, homogeneous ɛnd(t) values (+5.68 to +6.37). A Group 2 gabbroic diabase dike has been dated at 131.1±6.1 Ma. Group 3 basaltic lavas are interbedded with the Late Jurassic Early Cretaceous pelitic sediments; they have compositions transitional between Groups 1 and 2 and flat to slightly enriched trace element patterns. Sr Nd isotopic data and REE modeling indicate that variable degrees of partial melting of distinct mantle source compositions (enriched garnet clinopyroxene peridotite for Group 1 and spinel-lherzolite for Group 2, respectively) could account for the chemical diversity of the Cona mafic rocks. Geochemical similarities between the Cona mafic rocks and the basalts probably created by the Kerguelen plume based on spatial temporal constraints seem to indicate that an incubating Kerguelen plume model is more plausible than a model of normal rifting (nonplume) for the generation of the Cona mafic rocks. Group 1 is interpreted as being related to the incubating Kerguelen plume lithosphere interaction; Group 2 is likely related to an interaction between anhydrous lithosphere and rising depleted asthenosphere enriched by a droplet originating from the Kerguelen plume, while Group 3 may be attributed to thermal erosion resulting in the partial melting of lithosphere during the longterm incubation of a magma chamber/pond at a shallow crustal level. The Cona mafic rocks are probably related to a progressively lithospheric thinning beneath eastern Gondwanaland from Ma to 130 Ma. Our new observations seem to indicate that the Corresponding author. addresses: dchengzhu@163.com (D. Zhu), moxuanxue@hotmail.com (X. Mo) /$ - see front matter 2007 Elsevier B.V. All rights reserved. doi: /j.lithos

2 148 D. Zhu et al. / Lithos 100 (2008) Kerguelen plume may have started its incubation as early as the latest Jurassic or earliest Cretaceous period and that the incubating Kerguelen plume may play an active role in the breakup of Greater India, eastern India, and northwestern Australia Elsevier B.V. All rights reserved. Keywords: Geochemistry; SHRIMP zircon dating; Kerguelen plume; Cona mafic rocks; Tethyan Himalaya 1. Introduction The Kerguelen Plateau in the South Indian Ocean is one of the best-known Large Igneous Provinces (LIPs) on earth (Larson, 1991; Coffin and Eldholm, 1994). This LIP has been linked to the presence of the longlived Kerguelen mantle plume, situated at present beneath the northwestern margin of the Kerguelen Plateau (Storey et al., 1992; Kent et al., 1996; Coffin et al., 2002). Cretaceous and Cenozoic igneous rocks related to the Kerguelen plume in the eastern Indian Ocean region and neighboring continental margin (Fig. 1a) have been widely dispersed from their original sites of emplacement due to the changing motions of the Antarctic, Australian, and Indian plates from the Early Cretaceous period to the present (Mahoney et al., 1983; Storey et al., 1989; Davies et al., 1989; Storey et al., 1992; Müller et al., 1993; Frey et al., 1996; Kent et al., 1997; Kent et al., 2002; Ingle et al., 2002, 2003, 2004; Srivastava et al., 2005). The first manifestation in an oceanic environment probably attributable to the Kerguelen plume is exposed on the Southern Kerguelen Plateau ( Ma, Coffin et al., 2002). On the neighboring continental margins (Fig. 1a), the Rajmahal Sylhet Traps (118 Ma; Sarkar et al., 1996; Kent et al., 1997, 2002), igneous complexes in the Shillong Plateau ( Ma, Ray and Pande, 2001; Srivastava et al., 2005) from eastern India and alnoites from Antarctica (110 Ma, Ghose et al., 1996) have been widely accepted as the result of the Kerguelen plume for their geochemical affinity with the contemporaneous construction of the oldest portion of the Southern Kerguelen Plateau (Storey et al., 1992; Ghose et al., 1996; Ray and Pande, 2001; Coffin et al., 2002; Kent et al., 2002; Ingle et al., 2002, 2003, 2004; Kumar et al., 2003; Srivastava and Sinha, 2004; Srivastava et al., 2005). However, for the older uncontaminated Bunbury Casuarina basalts of southwestern Australia (132 Ma), Frey et al. (1996) proposed that their geochemical similarity with younger oceanic lavas from Ninetyeast Ridge and the Kerguelen Archipelago that have been related to the Kerguelen plume might be fortuitous, whereas other authors have argued that the Bunbury Casuarina basalts may be the first manifestation in a continental marginal environment potentially attributable to the Kerguelen plume (Davies et al., 1989; Duncan and Storey, 1992; Kent et al., 2002; Ingle et al., 2002, 2004). The differing statements above on the Bunbury basalts resulted in a contentious issue regarding the relationship between the breakup of eastern Gondwanaland and the Kerguelen plume activity. Some investigators believe that the oldest ( Ma) known lavas from the Southern Kerguelen Plateau and Rajmahal Sylhet Traps associated with the Kerguelen plume (Mahoney et al., 1983; Class et al., 1993; Baksi, 1995) are considerably younger than the initial breakup ( 132 Ma, Powell et al., 1988) of southwest Australia and Greater India, and have thus concluded that the plume was not a major factor in the breakup of eastern Gondwanaland (Frey et al., 1996). Others, however, believe that the first manifestation of the Kerguelen plume around 132 Ma, exposed in southwestern Australia (Bunbury Casuarina basalt), is significantly coeval with the initial seafloor magnetic anomaly of the eastern Indian Ocean at 132 Ma (Powell et al., 1988). They have therefore suggested that the Kerguelen plume played a significant role in the breakup of eastern Gondwanaland (Ingle et al., 2002; Coffin et al., 2002). The reason for this contentious issue is that it is a more complicated question, where three important concerns need to be considered. These are: (1) the large distance between the proposed Kerguelen plume location and southwest Australia during the eruption of the Bunbury basalts is often cited as a possible argument against a plume origin (Frey et al., 1996; Coffin et al., 2002); however, the distance of km between the Kerguelen plume and southwestern Australia at 120 Ma is not believed to be a convincing argument against a common plume source for the Bunbury basalts and the Kerguelen Plateau basalts (Ingle et al., 2004); (2) the older Bunbury Casuarina basalts ( 132 Ma) preceded the presumed initiation of volcanism on the Kerguelen Plateau ( Ma) by Ma; and (3) the small volume of Bunbury Casuarina basalts ( 10 3 km 3, Coffin et al., 2002) is incompatible with the traditional plume model (Frey et al., 1996; Coffin et al., 2002). Therefore, an appealing question is whether we can find the older magma manifestations potentially related to the early activity of the Kerguelen plume on

3 D. Zhu et al. / Lithos 100 (2008) Fig. 1. (a). Physiographic map of the Indian Ocean and surrounding continents, showing spatial temporal locations of the Cona mafic rocks and products created by the Kerguelen plume (including the various provinces: Southern Kerguelen Plateau (SKP), Central Kerguelen Plateau (CKP), Northern Kerguelen Plateau (NKP), Elan Bank and Skiff Bank, Broken Ridge and Ninetyeast Ridge, Kerguelen Archipelago (KA), Heard and MacDonald Island (HI), Ingle et al., 2003). Locations for continental tholeiites on the margins of the eastern India (Rajmahal Traps) and the southwestern Australia (Bunbury basalts) are also illustrated (Ingle et al., 2003; these may be related to the Kerguelen plume activity (Frey et al., 1996; Ingle et al., 2002). Age-constraints for various provinces and continental volcanics are from Coffin et al. (2002), Duncan (2002), Srivastava et al. (2005) and Zhu et al. (2005). (b). Sketch tectonic map of the east central Tethyan Himalaya and the eastern India showing the present locations of the Cona mafic rocks, the two exposures of Sangxiu volcanic rocks (Duojiu, Rimowa village), the Rajmahal, Sylhet Traps and the carbonatite alkaline complexes on the Shillong Plateau of eastern India. Tectonic subdivision in Himalaya (Pan et al., 2004), the igneous rocks in the eastern India (Ray and Pande, 2001).

4 150 D. Zhu et al. / Lithos 100 (2008) Fig. 2. Geological map for the studied area in the Cona area (Yin et al., 2005, unpublished). the associated oceanic plateau and continental margin? This answer awaits two matters in progress: one is a new deep drilling program on the Southern Kerguelen Plateau (Ingle et al., 2004; personal communication with Frey F., 2006); the other is a detailed study on the little known continental margin (e.g., south Tibet). Zhu et al. (2005, 2007) first reported the chronology and geochemical data of the Sangxiu Formation volcanic rocks from the southeast of Yangzuoyong Co. of the central segment of the Tethyan Himalaya, south Tibet. The Sangxiu felsic rocks from Rimowa village (Fig. 1b) with features of A-type granite have been dated at 133 Ma (Zhu et al., 2005), coeval with the initial breakup of southwest Australia and Greater India. The underlying Sangxiu basalts characterized by OIB-type geochemical characteristics that are comparable to those of the Kerguelen basalts were interpreted as a consequence of an interaction between the Kerguelen hotspot and the lithosphere of the northeastern margin of Greater India (Zhu et al., 2007). On the other hand, it is well known that the isotopic characteristics of lavas erupted on the Cenozoic Islands (Kerguelen and Heard) are quite diverse. This is equally true of basalts recovered by drilling from the Kerguelen Plateau and Broken Ridge (Weis et al., 1993, 1998; Neal et al., 2002; Ingle et al., 2003). Hence, the genetic relationship between the Sangxiu volcanic rocks and the early activity of the Kerguelen plume requires further testing because only a small amount of geochemical data from two exposures (Duojiu, Rimowa village, Fig. 1b) are used for such a hypothesis (Zhu et al., 2007). New geological mapping shows that the contemporaneous mafic rocks are also well exposed in the Cona area of the eastern Himalaya (Yin et al., 2005, unpublished). In this paper, we present new geochemical, Sr, Nd isotopic results and SHRIMP U Pb zircon chronology for the widely exposed mafic rocks around the Cona area of the eastern Tethyan Himalaya in south Tibet. With these data we hope to further constrain the nature of the mantle source and to test the relationship to the Kerguelen plume, and also to attempt to discuss the implication of our new results for the breakup of eastern Gondwanaland. 2. Field occurrence and petrography The Tethyan Himalaya, which is located between the Indus-Yarlung Zangbo Suture Zone to the north and the Greater Himalayan crystalline assemblages to the south, belongs paleogeographically to the northern margin of Greater India. It comprises four subsequences: (1) a Proterozoic to Devonian sequence characterized by laterally persistent lithologic units deposited in an epicratonal setting; (2) a Carboniferous Lower Jurassic sequence that shows dramatic northward changes in thickness and lithofacies; (3) a Jurassic Cretaceous passive continental margin sequence; and (4) the

5 D. Zhu et al. / Lithos 100 (2008) Fig. 3. Photomicrographs of the Cona mafic rocks showing the different texture features. (a). Moderately phyric texture consists mainly of plagioclase phenocrysts with groundmasses of fine-grained plagioclase and clinopyroxene, basalt location CN17-3; (b). Intergranular and doleritic textures, note that the plagioclase and clinopyroxene phenocrysts are relative fresh, basalt location CN22-4; (c). Medium-grained with ophitic texture of diabase dike margin, dike location: CN20-1; (d). Coarse-grained with typical poikilitic texture of dike middle portion, dike location: CN20-2. uppermost Cretaceous Eocene sequence (Yin, 2006 and references therein). The Cretaceous mafic rocks are widely exposed in the Jurassic Cretaceous passive continental margin sequences, eastern Tethyan Himalaya (Fig. 1b) and paleogeographically located in northeastern Greater India. In the present work, Cona mafic rocks well exposed in the Lakang Formation have been studied (Fig. 2). The Lakang Formation sequence experienced low-grade metamorphism and deformation, and is characterized by fault contacts with the underlying Early Jurassic Ridang Formation, the Late Triassic Qulonggongba and Nieru Formation (Fig. 2), and by the absence of an overlying strata (Yin et al., 2005, unpublished), resulting in a disagreement in age constraint that is attributed to Late Triassic (Team of Tibet Regional Geological Survey, 1979) or Early Cretaceous time (Zhong et al., 2003, unpublished; Yin et al., 2005, unpublished). Lithology in the Lakang Formation is composed mostly of slates, limestones, and minor siliceous rocks, with a few siltstones observed in the upper part. In terms of lithologic features, the Lakang Formation is broadly comparable to the Sangxiu Formation. Combined with the results of zircon SHRIMP dating (shown below), we believe that the Lakang Formation belongs to Late Jurassic to Early Cretaceous in age, coeval with the Sangxiu Formation exposed on the southeast and east of Yangzuoyong Co. (Zhu et al., 2007), km north of the Cona area (Fig. 1b). The Cona mafic rocks in the Lakang Formation consist of basalts and associated diabase, as well as gabbroic diabase. Basaltic lavas with massive or amygdaloidal structures occur in the middle lower parts of the Lakang Formation. The thicknesses of the basalts range from several meters to 614 m (e.g., the Niangzhong section, Yin et al., 2005, unpublished). Approximately E W trending diabases are well exposed and have intruded

6 Table 1 Zircon SHRIMP analysis data of the Cona mafic rocks of the eastern Tethyan Himalaya in south Tibet Spot 206 Pb c (%) U (ppm) Th (ppm) Th/U 206 Pb a (ppm) 206 Pb/ 238 U (Ma±1σ) 207 Pb a / 206 Pb a ±% 207 Pb a / 235 U ±% 206 Pb a / 238 U ±% Sample CN8-1: diabase dike, ophitic texture, average weighted 206 Pb/ 238 U age of 14 spots=144.7±2.4 Ma (MSWD=0.76) without spot CN8-1-4 CN ± CN ± CN ± CN ± CN ± CN ± CN ± CN ± CN ± CN ± CN ± CN ± CN ± CN ± CN ± Sample CN20-2: gabbroic diabase dike, poikilitic texture, average weighted 206 Pb/ 238 U age of 10 spots=131.1±6.1 Ma (MSWD=3.6) without spot CN20-2-6, 9, 13, 14, 15 CN ± CN ± CN ± CN ± CN ± CN ± CN ± CN ± CN ± CN ± CN ± CN ± CN ± CN ± CN ± Pb c (%) represents the percentage of common 206 Pb in total 206 Pb. a Denotes radioactivity lead. 152 D. Zhu et al. / Lithos 100 (2008)

7 D. Zhu et al. / Lithos 100 (2008) limestones and slates of the Lakang Formation (Fig. 2). These diabases show predominantly concordant intrusive contact (as sills) with wall rocks and scarcely cut through the detrital layers as dikes. The same is true for the approximately E W trending gabbroic diabases except for two outcrops (locations CN8, CN20, Fig. 2), where the gabbroic diabase is emplaced within the middle portion of a wide diabase sill. Although the basalts from the Cona area experienced variable alteration, original textures can still be observed. In hand specimens, most of the basalts are phyric with the exception of samples CN17-1 and CN24-1 that contain microvesicles filled with chlorite. In thin sections, a few basalts present small plagioclase phenocrysts (samples CN17-1, CN24-1). Most of the basalts have abundant phenocrysts consisting of plagioclase ( 5 8%) and clinopyroxene (2 3%) with groundmasses of fine-grained plagioclase, clinopyroxene (Fig. 3a), and secondary altered minerals (e.g., chlorite, epidote), as well as ilmenite, magnetite plus minor titanite. Pseudomorphs of altered olivine are observed in some samples. The basalts from Kada Gorge are composed of these same minerals, but exhibit distinct doleritic and intergranular textures (Fig. 3b) with relatively fresh plagioclase and clinopyroxene. The diabase samples from the Cona area experienced higher alteration relative to the basalts. A variety of textures are seen in these diabase samples, but the most common texture is ophitic (Fig. 3c). The major mineral compositions of these rocks are altered plagioclase ( 40%) and clinopyroxene ( 30%) with secondary chlorite (10 15%). Other, minor minerals are titanite (2 5%), calcite (3%), magnetite (3%), and quartz (2%). Abundant needleshaped ilmenite (5 7%) is observed in the samples. The gabbroic diabase samples are relatively fresh compared to the diabase samples. These samples are composed of medium- to coarse-grained plagioclase ( 60%), chloritized clinopyroxene ( 35%) with minor ilmenite (2%), quartz (1 2%), and magnetite (b1%), and show a dominant ophitic gabbro texture. The coarse-grained variety exhibits a typical poikilitic texture (Fig. 3d). 3. Analytical procedures For U Pb zircon dating, zircons were separated from rock samples using standard density and magnetic separation techniques at the Special Laboratory of the Geological Team of Hebei Province, China. Zircon grains, together with the zircon U Pb standard TEMORA (Black et al., 2003), were cast in an epoxy mount, which was then polished to section the crystals in half for analysis. Zircons were documented with transmitted and reflected light micrographs as well as cathodoluminescence (CL) images to reveal their internal structures, and the mount was vacuum-coated with a 500-nm layer of high-purity gold. Under the guidance of zircon CL images, the zircons were analyzed for U Pb isotopes and U, Th, and Pb concentrations using a SHRIMP II ion microprobe at the Beijing SHRIMP Center, Chinese Academy of Geological Sciences. U Th Pb ratios were determined relative to the TEMORA standard zircon corresponding to 417 Ma 206 Pb/ 238 U= (Black et al. 2003), and the absolute abundances were calibrated to the standard zircon SL13. Analyses of the TEMORA standard zircon were interspersed with those of unknown grains, following operating and data processing procedures similar to those described by Williams (1998). The reference zircon was analyzed after every fourth analysis. Measured compositions were corrected for common Pb using the 204 Pb method, and data processing was carried out using Isoplot (Ludwig, 2001). Uncertainties on individual analyses are reported at the 1σ level; mean ages for pooled 206 Pb/ 238 U results are quoted at the 95% confidence level. The U Pb zircon data are presented in Table 1. Major element data were collected using X-ray fluorescence (XRF) on fused glass beads using a Rigaku ZSX100e spectrometer in the Analytical Center, Chengdu Institute of Geology and Mineral Resources. The analytical uncertainty is usually b 5%. The same whole-rock powders were used for trace element concentrations and Sr and Nd isotopic ratios. Trace element concentrations were determined using a Perkin Elmer Elan 6000 for acid dissolution inductively coupled plasma mass spectrometer (ICP-MS) at the National Geological Analytical Center, Chinese Academy of Geological Sciences, Beijing. The analytical procedures were similar to those described by Li (1997). The analytical precision is generally within 5%. Wholerock major and trace element data for the analyzed samples are given in Table 2. Sr and Nd isotopic ratios were measured on a Finnigan MAT-262 thermal ionization mass spectrometer (TIMS) at the Laboratory for Radiogenic Isotope Geochemistry, Institute of Geology and Geophysics, Chinese Academy of Sciences, Beijing. About mg of whole rock powder was completely decomposed in a mixture of HF HClO 4 for Sr and Nd isotopic analysis. Sr and REE were separated on quartz columns with a 5 ml resin bed of AG 50 W-X12, mesh. Nd was separated from other REEs on quartz columns using a 1.7 ml Teflon powder as a cation exchange medium. Procedural blanks were b200 pg for Sr and

8 154 D. Zhu et al. / Lithos 100 (2008) Table 2 Major (wt.%) and trace element (ppm) chemistry of the Cona mafic rocks of the eastern Tethyan Himalaya in south Tibet Sample no. CN1-1 CN1-2 CN5-1 CN6-1 CN8-1 CN9-1 CN12-1 CN13-1 CN14-1 CN19-1 CN20-1 CN2-1 CN2-2 CN7-1 Attitude DS DS DS DS DD DD DD DD DS DD DS Highly altered DSs Grouping Major element (wt.%) SiO TiO Al 2 O Fe 2 O MnO MgO CaO Na 2 O K 2 O P 2 O LOI Trace element (ppm) Sc V Cr Co Ni Rb Sr Y Zr Nb Cs Ba La Ce Pr Nd Sm Eu Gd Tb Dy Ho Er Tm Yb Lu Hf Ta Pb Th U Eu/Eu CN17-1 Attitude: DS CN17-2 = diabase CN17-3 sill, DD CN23-1 = diabasecn24-1 dike, GDS CN25-1 = gabbroic CN8-3 diabase CN10-1 sill, GDD CN16-1 = gabbroic CN16-2 diabase CN20-2 dike; Grouping: CN CN26-2 = Group 1, CN = Group CN22-4 2, 3 = Group 3; LOI = loss on ignition; Eu/Eu*=Eu Basalt Basalt Basalt Basalt Basalt N /(Sm Basalt N Gd N ) 1/2, subscript N denotes normalized to Chondrite (Sun and McDonough, 1989). GDD GDD GDS GDS GDD GDS GDS Basalt Basalt Major element (wt.%) b50 pg for Nd. For the measurements of isotopic composition, Sr was loaded with a Ta-Hf activator on a single W filament, and Nd was loaded as phosphates and measured in a Re-double-filament configuration. The concentrations of Rb, Sr, Sm, and Nd were analyzed by isotopic dilution. The NBS-987 standard measured

9 D. Zhu et al. / Lithos 100 (2008) Table 2 (continued) CN17-1 CN17-2 CN17-3 CN23-1 CN24-1 CN25-1 CN8-3 CN10-1 CN16-1 CN16-2 CN20-2 CN26-1 CN26-2 CN22-3 CN22-4 Basalt Basalt Basalt Basalt Basalt Basalt GDD GDD GDS GDS GDD GDS GDS Basalt Basalt Major element (wt.%) Trace element (ppm) during the course of analyses gave values of ± 3(2σ) for the 143 Nd/ 144 Nd ratio and ±4 (2σ) for the 87 Sr/ 86 Sr ratio. 143 Nd/ 144 Nd ratios were normalized to 146 Nd/ 144 Nd= and 87 Sr/ 86 Sr ratios to 86 Sr/ 88 Sr = Raw data obtained were calculated using the Isoplot program (Ludwig, 2001), giving a 2σ

10 156 D. Zhu et al. / Lithos 100 (2008) error. Technical details on chemical separation and measurement are described in Chen et al. (2002). 4. Results 4.1. U Pb zircon geochronology Cathodoluminescence (CL) images and concordia plots for the analyzed zircons are shown in Fig. 4. Zircons from the diabase (sample CN8-1) are characterized by high length/width ratios, straight rhythmic stripes (Fig. 4a), and high Th/U ratios ( 0.98, Table 1), corresponding to the features of magmatic zircons (Hoskin and Black, 2000). Fourteen spot analyses yielded 206 Pb/ 238 U ages ranging from to 150 Ma with a weighted mean age of 144.7±2.4 Ma (MSWD = 0.76) (Fig. 4b), which represents the magma emplacement age of the Cona diabase with OIB-type geochemical features (shown below). Zircons from the gabbroic diabase (sample CN20-2) show different crystal faces with variable length/width ratios ranging from 1:1 to 2:1. Three of fourteen zircons (spots 1, 4, 5, 11) have subtle magmatic oscillatory zoning in CL images (Fig. 4c). In combination with all the analyzed spots have ratios of Th/U larger than 1.0 (Table 1), indicating an affinity of magmatic crystallization zircons (Hoskin and Black, 2000). Note that some zircons display an inherited core and overgrowth rim. Nine zircon spots and one overgrowth rim spot (spot 5) yielded 206 Pb/ 238 U ages ranging from to Ma with a weighted mean age of 131.1±6.1 Ma (MSWD = 3.6) (Fig. 4d), which is taken to represent the magma emplacement age of the Cona gabbroic diabase with N-MORB-type geochemical features (shown below). Fig. 4. Cathodoluminescence (CL) images and concordia plots of zircon SHRIMP dating for the Cona mafic rocks of the eastern Tethyan Himalaya in south Tibet.

11 D. Zhu et al. / Lithos 100 (2008) and variable MgO contents. Group 2 is observed in the same area as Group 1, and consists of gabbroic sills or crosscutting gabbroic intrusions, presenting lower TiO 2 and P 2 O 5 contents and a relatively restricted MgO content. Group 3 is composed of basaltic lavas interbedded with the Late Jurassic Early Cretaceous pelitic sediments (Lakang Formation) in the Kada Gorge (Fig. 2). It shows compositions transitional between Groups 1 and 2 (Fig. 5a and b). In terms of field occurrence, there is no spatial difference between Groups 1 and 2 (Fig. 2) around the Cona area, except for the emplacement age of Group 2 intruded 14 Ma later than Group 1. It should be noted that this geochemical classification is rather preliminary and could be refined when more data are available. Nevertheless, it will be justified further in later sections because Groups 1, 2, and 3 have distinct abundances of trace elements and are believed to have originated from different mantle sources under various melting depths. Fig. 5. Geochemical classification for the Cona mafic rocks. (a) TiO 2 vs. MgO diagram; (b) P 2 O 5 MgO diagram. Note that the three groups in the Cona mafic rocks are seen on these two diagrams, and that the highly altered samples (CN2-1, CN2-2, CN7-1) did not have distinctive abundances of TiO 2 and P 2 O 5 contents. See text for details Geochemistry Major elements and geochemical classification Although they are of basalt composition, the Cona mafic rocks are compositionally diverse (Table 2). For example, they have different TiO 2 and P 2 O 5 contents at comparable MgO contents (Fig. 5a and b). One highly altered sample (CN7-1 with LOI =13.86 wt.%) had an anomalously low SiO 2 content (40 wt.%). When normalized on a volatile-free basis, the sample showed an anomalously high total iron content (= wt.%). These characteristics in major element composition are well-known features of weathering. This sample, however, did not have distinctive abundances of other major and trace elements (Figs. 5 and 6). Consequently, we infer that most of the geochemical characteristics (except for Sr isotopes) of Cona mafic rocks reflect the nature of the mantle source and/or magmatic processes. The Cona mafic rocks form three compositionally distinct groups (Fig. 5a and b). Group 1 is exposed over the whole sampling area, consisting of basalts, diabase sills or dikes and having higher TiO 2 and P 2 O 5 contents Fig. 6. Ni vs. MgO (a) and Ni vs. Cr (b) diagrams of the Cona mafic rocks. MgO is in wt.%, and compatible elements are in ppm. Broad positive correlations are observed, suggesting a fractionation crystallization of mafic mineral (e.g., olivine, clinopyroxene, etc.). Symbols as in legend for Fig. 5.

12 158 D. Zhu et al. / Lithos 100 (2008) Trace elements Cr and Ni concentrations in the analyzed samples are generally low (with averages around 119 ppm and 77 ppm, respectively), indicating that the Cona mafic lavas do not have the Cr and Ni contents expected for partial melts of peridotite. This is also observed on Ni vs. MgO and Ni vs. Cr plots (Fig. 6a and b), which exhibit broad positive correlations as expected for a fractionation crystallization of mafic mineral (e.g., olivine, clinopyroxene, etc.). All samples in Group 1 have light rare earth element (LREE) enriched (with variable chondrite-normalized (La/Yb) N ratios ranging from 5.35 to 10.98) subparallel patterns on chondrite-normalized rare earth element (REE) plots (Fig. 7a and b). Such LREE-enriched patterns are typical of OIB. Eu anomalies (Eu/Eu = ) are negligible or slightly positive in the samples of this group. Group 2 samples are characterized by depleted, subparallel REE patterns (Fig. 7c) with (La/ Yb) N = (average=0.88). Eu anomalies (Eu/ Eu = , averaging 1.1) are slightly or discernibly positive. Group 3 basalts are characterized by flat to slightly enriched REE patterns (Fig. 7d) with (La/ Yb) N = (average=2.01). Eu anomalies (Eu/ Eu = ) are absent or slightly negative. On primitive mantle-normalized trace element plots, all the samples of Group 1 have humped subparallel, OIB-like spectra (Fig. 8a and b). The Nb and Ta exhibit slightly negative to positive anomalies for the basalts (Fig. 8a) and positive anomalies for the diabases (Fig. 8b) relative to the neighboring elements. Group 2 samples share most of the geochemical features of N- MORB, characterized by a depleted to slightly flat spectra of high-field strength elements (HFSEs) (Fig. 8c). Discernable Nb and Ta negative anomalies relative to Th and La are presented in this group. In addition, a slight negative Ti anomaly is also noted in some samples of the group (Fig. 8c). Unlike the Group 1 and 2 samples, Group 3 basalts have relatively flat patterns of immobile trace elements, which are significantly higher than those of the Group 2 samples in absolute abundance. Group 3 lavas are generally characterized by a significant depletion of Nb and Ta relative to Th and La (Fig. 8d). Depletions of P and Ti and strong enrichment of Th relative to the neighboring elements are observed in sample CN22-3 (Fig. 8d). It is obvious that the basalts and diabases in Group 1 have similar geochemical features, such as coherent REE patterns (Fig. 7a and b) and trace element spidergrams (Fig. 8a and b). We therefore infer that the basalts and diabases in Group 1 are probably cogenetic. It should also be noted that the dated samples Fig. 7. Chondrite-normalized REE patterns for the Cona mafic rocks. Data for chondrite and N-MORB are from Sun and McDonough (1989).

13 D. Zhu et al. / Lithos 100 (2008) CN8-1 and CN20-2 compare closely with the diabases in Group 1 and the gabbroic diabases in Group 2 (Fig. 7a and b; Fig. 8a and b), respectively. In this case, the SHRIMP zircon ages of Ma for the sample CN8-1 and Ma for the sample CN20-2 can be used to represent the ages of magmatic events of Groups 1 and 2, respectively Sr and Nd isotopic data The measured and age-corrected 87 Sr/ 86 Sr and 143 Nd/ 144 Nd ratios of selected samples are listed in Table 3 and shown in Fig. 9. The initial isotopic ratios were corrected to Ma for Group 1, to Ma for Group 2 samples according to the SHRIMP zircon dating, and to 140 Ma for sample CN22-3 in Group 3 estimated from field observations and geochemical features. The initial Sr isotopic ratios of Group 1 samples span a large range, from to , with limited ɛnd(t) values varying between 1.8 and 4.7. Within the samples of Group 1, initial Sr isotopic data form two suites (Fig. 9): one is characterized by low initial 87 Sr/ 86 Sr ratios ( ) designated as the Low Initial Strontium (LIS) suite, the other presents high initial 87 Sr/ 86 Sr ratios ( ) and is designated as the High Initial Strontium (HIS) suite. Group 2 samples have limited initial 87 Sr/ 86 Sr ratios ( ) with elevated positive ɛnd(t) values ( ) relative to Group 1 samples. Only one sample (CN22-3) in Group 3 has been analyzed for Sr Nd isotopic composition. This sample is characterized by a moderate positive ɛnd(t) value (2.0) and the highest initial Sr isotopic ratios ( ) in all of the analyzed Cona mafic rocks. 5. Discussion 5.1. Effects of alteration and low-grade metamorphism on elemental mobility The Cona mafic rocks are Ma old and have been altered to various degrees after eruption/ emplacement, judging from the petrographic observation and variable LOI listed in Table 2. Relatively fresh samples and highly altered samples (e.g., samples CN2-1, CN2-2, CN7-1) have been analyzed to evaluate the Fig. 8. Primitive mantle-normalized trace element patterns for the Cona mafic rocks. Normalizing values and plotting order are from Sun and McDonough (1989). Group P lavas (P indicating plume-derived) from Mont Bureau and Mont Rabouillère on the northern part of the Kerguelen Archipelago (Yang et al., 1998), Sangxiu basalts from the southeast of Yangzuoyong Co. in south Tibet (Zhu et al., 2007), Bunbury Casuarina basalts from the southwestern Australia (132 Ma, Frey et al., 1996), Rajmahal Group II basalts from the eastern India (Kent et al., 1997), and Site 758 basalts on Ninetyeast Ridge ( de/georoc/entry.html) are shown for comparison.

14 160 D. Zhu et al. / Lithos 100 (2008) Table 3 Rb Sr and Sm Nd isotopic compositions of the Cona mafic rocks of the eastern Tethyan Himalaya in south Tibet Sample no. Grouping Sm (ppm) Nd (ppm) ( 147 Sm/ 144 Nd) m ( 143 Nd/ 144 Nd) m (±2σ) Rb (ppm) Sr (ppm) ( 87 Rb/ 86 Sr) m ( 87 Sr/ 86 Sr) m (±2σ) Agecorrected (T=Ma) ( 87 Sr/ ɛnd 86 Sr) T (T) ( 143 Nd/ 144 Nd) T CN1-1 Group ± ± CN2-1 Group ± ± CN6-1 Group ± ± CN7-1 Group ± ± CN8-1 Group ± ± CN9-1 Group ± ± CN13-1 Group ± ± CN14-1 Group ± ± CN19-1 Group ± ± CN20-1 Group ± ± CN17-3 Group ± ± CN23-1 Group ± ± CN24-1 Group ± ± CN10-1 Group ± ± CN16-1 Group ± ± CN20-2 Group ± ± CN22-3 Group ± ± m=measured isotopic ratios; T=age-corrected initial isotopic ratios. ɛnd(t) are initial values, calculated using present day ( 143 Nd/ 144 Nd) CHUR = and ( 147 Sm/ 144 Sm) CHUR = (CHUR = chondritic uniform reservoir). Corrected formula as follows: ( 87 Sr/ 86 Sr) T =( 87 Sr/ 86 Sr) m + 87 Rb/ 86 Sr (e λt 1), λ= a 1 ;( 143 Nd/ 144 Nd) T =( 143 Nd/ 144 Nd) m +( 147 Sm/ 144 Nd) m (e λt 1), ɛnd(t)=[( 143 Nd/ 144 Nd) m /( 143 Nd/ 144 Nd) CHUR (T) 1] 10 4,( 143 Nd/ 144 Nd) CHUR (T)=( 143 Nd/ 144 Nd) CHUR ( 147 Sm/ 144 Sm) CHUR (e λt 1), λ Sm Nd = a 1. effects of alteration and low-grade metamorphism on elemental mobility. A general consensus exists that large-ion lithophile elements (LILEs, e.g., K, Na, Rb, Ba, Sr) are mobile, whereas transition metals (e.g., Cr, Ni), rare earth elements (REEs), high-field strength elements (HFSEs), and Th and Ti in mafic rocks are relatively immobile during low-temperature alteration and low-grade metamorphism (Ludden and Thompson, 1978; Bienvenu et al., 1990; Staudigel et al., 1996). Indeed, as shown above, both the relatively fresh samples and the highly altered samples in Group 1 have subparallel patterns of REE and HFSE concentrations, indicating that these mafic rocks still preserve their original REE and HFSE signatures. The same will be true of the samples in Group 2, as seen from their uniform patterns of REE and HFSE concentrations. Thus, only immobile elements such as the HFSEs (Ti, Zr, Y, Nb, Ta, Hf), Th, and REE are used in the following discussion to identify the magmatic affinity and petrogenesis of these mafic rocks. Seawater has a high Sr abundance ( 8 ppm) and can easily modify the Sr isotopic composition of the oceanic crust (McCulloch et al., 1981). However, Nd concentrations in seawater are extremely low ( ppm). Hence, the Nd isotopic ratios of the oceanic crusts are not easily changed by alteration (McCulloch et al., 1981). It could be concluded that the LIS suite in Group 1 for selected samples is a reflection of the magmatic source. We note that two highly altered samples (CN2-1 with LOI=10.12 wt.% and CN7-1 with LOI=13.86 wt.%, Table 2) in the HIS suite of Group 1 have anomalously high initial Sr isotopic ratios ( and , respectively, Table 3)atrelativelyconstant 143 Nd/ 144 Nd (T) values (Fig. 9). As Sr abundance and Sr isotope ratio are highly sensitive to alteration and low-grade metamorphism, we therefore prefer to interpret the anomalously high initial Sr isotopic ratios observed in the HIS suite of Fig. 9. ( 143 Nd/ 144 Nd) T vs. ( 87 Sr/ 86 Sr) T diagram of the Cona mafic rocks. Note that the initial Sr isotopic compositions of Group 1 samples clearly form the Low Initial Strontium (LIS) suite and the High Initial Strontium (HIS) suite. Symbols as in legend for Fig. 5.

15 D. Zhu et al. / Lithos 100 (2008) Group 1 as likely postmagmatic processes rather than a magmatic source. The lower initial 87 Sr/ 86 Sr ratios ( ) and higher positive ɛnd(t) values ( ) of the samples in Group 2 indicate that the effects of alteration and low-grade metamorphism on Sr Nd isotopes could be disregarded. One relatively fresh sample (CN22-3 with LOI=4.3 wt.%) in Group 3 also shows anomalously high initial Sr isotopic ratios ( ), which can be ascribed to crustal contamination (discussed below) Crustal contamination Although not diagnostic, on the primitive mantlenormalized multi-element diagrams (Fig. 8), Ti troughs are typical for crustal rocks, especially for the Upper Crust (Rudnick and Gao, 2003); they may therefore be present in rocks extensively contaminated by continental crust. Ti troughs are lacking in the patterns of the lavas in Group 1 (Fig. 8a and b); an upper crustal contribution to the magmas can therefore be excluded. However, some of the samples in Groups 2 and 3 present discernible Ti troughs (Fig. 8c and d), suggesting that small amounts of assimilation of crustal material cannot be ruled out. A more diagnostic signature of continental crust is a relative depletion in Nb and Ta (Thompson et al., 1984; Frey et al., 2002). As seen in Fig. 8c and d, the samples in Groups 2 and Group 3 are relatively depleted in Nb and Ta compared to the neighboring Th and La. To evaluate this depletion, we use a (Th/Nb) PM vs. (La/ Nb) PM plot (Fig. 10), which not only shows the effects of continental crust assimilation, but in some cases may distinguish between upper and lower continental crust (Fitton et al., 1998; Frey et al., 2002). In this plot, Group 1 data plot within or near the field of most oceanic basalts and within the field defined by Bunbury Casuarina basalts from SW Australia and Rajmahal Group I basalts from NE India (Fig. 10), suggesting that crustal components involved in these mafic lavas are minor or negligible. Group 2 data show slight differences by their higher (La/Nb) PM and (Th/Nb) PM ratios and define a trend projecting toward the addition of lower crust materials (Fig. 10), indicating an input of lower crustal contaminant in their generation. In contrast, the effects of sialic continental contamination are very evident in Group 3 basalts as these basaltic lavas exhibit a distinct trend of increasing (Th/Nb) PM at constant (La/Nb) PM ratios (Fig. 10). Therefore, we predominantly use the least-altered and least-contaminated Cona mafic rocks (e.g., the LIS suite in Group 1, and Group 2) to infer their origin and geodynamic inference, as the HIS suite in Group 1 and the basalts in Group 3 are probably modified by postmelting crustal contamination, alteration, and metamorphism. Fig. 10. (Th/Nb) PM vs. (La/Nb) PM diagram for testing crustal contamination of the Cona mafic rocks. Subscripts PM indicate ratios normalized to primitive mantle values of Sun and McDonough (1989). Compositional fields of the sites 747, 738, and most oceanic basalt are from Frey et al. (2002), middle crust (MC), lower crust (LC) data are from Rudnick and Gao (2003), Bunbury Gosselin basalts data are from Frey et al. (1996), Rajmahal Group II basalts are from Kent et al. (1997), other data sources are same as the Fig. 8. Note that there is an unappreciable crustal contamination for the samples in Group 1, while some of Group 2 samples may have contaminated by lower crust, Group 3 samples define a trend toward sialic crustal contamination. Symbols as in legend for Fig. 5.

16 162 D. Zhu et al. / Lithos 100 (2008) Fig. 11. Selected diagrams for testing the nature of mantle source and magmatic process for the Cona mafic rocks. (a) Th/Ta vs. Th/Tb diagram, showing at least two distinct mantle sources (sources 1 and 2) identified by different correlation lines (Saunders et al., 1988; Rollinson, 1993) involved in the generation of the Cona mafic rocks; (b) Th/Nb vs. Th diagram, showing two trends of partial melting (trends 1 and 2) judged from the good linear correlations for Groups 1 and 2 samples according to Allègre and Minster (1978); (c) Sm/Yb vs. La/Sm diagram showing melt curves obtained using the nonmodal batch melting equations of Shaw (1970). Melt curves for spinel-lherzolite (with mode and melt mode of Ol.53% + Opx.27% +Cpx.17% +Sp.3%, Kinzler, 1997) and garnet clinopyroxene peridotite (with mode and melt mode of Ol.53.3% + Cpx.35.7% +Gt.11.0%; Walter, 1998) were drawn following the approach of Aldanmaz et al. (2000). Mineral/matrix partition coefficients are from McKenzie and O'Nions (1991), Depleted MORB Mantle (DMM) compositions are from Workman and Hart (2005), PM compositions are from Sun and McDonough (1989) and the sample s749c-16r-6 data are from Frey et al. (2000). The open circles denote the Sangxiu basalts. The dashed curves represent the melting trend defined using sample s749c-16r-6 compositions; the solid and dot curves are the melting trends from PM and DMM, respectively. Tick marks on each curve correspond to degrees of partial melting (%) for a given mantle source. Note that the data for the effects of continental contamination (e.g., Group 3) have been filtered. See text for details. Symbols as in legend for Fig Plagioclase accumulation vs. partial melting of distinct mantle sources As discussed earlier, a fractionation crystallization of mafic mineral (e.g., olivine, clinopyroxene, etc.) is obvious in the Cona mafic rocks. This process of separation of a mafic mineral is generally accompanied by plagioclase crystallization, resulting in a negative Eu anomaly. However, if plagioclase separation from the residual melts is inefficient, abundant plagioclase phenocrysts or microcrystals (accumulation) would be left as the residual mineral in the samples. Therefore, more plagioclase in the rock gives a higher Al 2 O 3 content while bulk rock has a lower MgO content i.e., the absence of a negative Eu anomaly may very well result from plagioclase phenocryst accumulation. In this regard, the discernibly positive Eu anomalies (Eu/ Eu = ) of the basalt samples in Group 1 can be accounted for by accumulation of plagioclase (phenocryst or microcrystal), corresponding to the petrographic observations in the basalt samples of Group 1 (Fig. 3a). A similar accumulation of plagioclase

17 D. Zhu et al. / Lithos 100 (2008) probably occurred in both the diabases of Group 1 (Eu/ Eu = ) and Group 2 samples (Eu/Eu = ), as indicated by their discernible positive Eu anomaly and abundant plagioclase crystals in thin section (Fig. 3c and d). Although fractionation crystallization of early mafic mineral followed by late plagioclase accumulation exists, the compositional variations of the Cona mafic rocks are likely controlled by melting processes of distinct mantle sources. Some lines of evidence for this suggestion include: (1) the considerable differences in absolute abundances and patterns of incompatible trace elements (Table 2, Figs. 7 and 8); (2) Group 2 magmas have elevated Nd isotopic ratios (Table 3, Fig. 9), which would not be attributable to fractionation crystallization; (3) ratio ratio plots of highly incompatible elements (e.g., Th/Ta vs. Th/Tb diagram), which can minimize the effects of fractionation and provide effective diagnostic signatures to examine the character of the mantle source, which can be identified from different correlation lines on these kind of plots (Saunders et al., 1988; Rollinson, 1993), showing that at least two possible mantle sources existed in the analyzed Cona mafic samples (Fig. 11a). Allègre and Minster (1978) showed that the ratio of the concentration of a highly incompatible element to a moderately incompatible element can be used to identify partial melting trends. The Th/Nb vs. Th diagram (Fig. 11b) illustrates that both Groups 1 and Group 2 magmas were probably generated by partial melting of distinct mantle sources. According to Aldanmaz et al. (2000), partial melting of spinel-lherzolite produces magmas with Sm/Yb ratios similar to the source and La/Sm ratios that decrease as melting increases. Consequently, partial melts from spinel-lherzolite source would be expected to produce melting trends that lies within, or close to, a mantle array defined by depleted-morb mantle (DMM) and primitive mantle (PM) compositions. However, melts generated by small (or moderate) degrees of partial melting of a garnet-lherzolite source (with garnet residue) would have Sm/Yb ratios significantly higher than Sm/Yb values in the mantle source. Consequently, the garnet-lherzolite melting trend is displaced from the DMM-PM mantle array to higher Sm/Yb (Aldanmaz et al., 2000). Therefore, after filtering the data for the effects of continental contamination (Lassiter and DePaolo, 2000), the La/Sm vs. Sm/Yb diagram allows monitoring of the source characteristics and also the degree of partial melting (Ernst and Buchan, 2003). In Fig. 11c, we modeled REE ratios to constrain the source characteristics of the Cona mafic lavas in terms of REE concentrations, source mineralogy, and degree of partial melting following the approach of Aldanmaz et al. (2000). The modeling uses the nonmodal batch melting equations of Shaw (1970) and the REE partition coefficient compilation of McKenzie and O'Nions (1991). Three different reference compositions are selected to define the likely mantle array of the Cona mafic lavas: (1) depleted MORB mantle (DMM), the source reservoir of MORB, which is assumed here to represent the convecting asthenospheric mantle with the compositions proposed by Workman and Hart (2005); (2) primitive mantle (PM, Sun and McDonough (1989)), which is representative of the initial mantle composition prior to MORB formation and depletion; and (3) the most depleted Kerguelen plume-derived magmas (sample s749c-16r-16 from Site 749, Frey et al., 2000), which are assumed here to represent the most depleted mantle component originating from the Kerguelen plume. The results of modeling show that: (1) Group 1 mafic rocks and the Sangxiu basalts are displaced from the spinel-lherzolite melting trend to higher Sm/Yb ratios and plot between the melting trajectories (for the proposed source composition; PM and s749c-16r-16) drawn for garnet clinopyroxene peridotite and spinellherzolite (Fig. 11c). This indicates the presence of a garnet residue in their source region. (2) The mafic samples in Group 1 and the Sangxiu basalts have La/Yb ratios greater than those that could be generated by a single-stage melting of PM when the degrees of partial melting are 4 10%. It seems likely that the REE ratios of the mafic samples in Group 1 and the Sangxiu basalts could not be entirely produced from the one-stage melting of PM. This interpretation for the generation of Cona mafic rocks is similar to the conclusions of Aldanmaz et al. (2000), who suggested that the mantle source of the Late Miocene alkaline rocks could have been enriched in LREE with respect to DMM composition prior to produce the alkaline magma in western Anatolia, Turkey. In this case, a source more enriched than PM is required to produce the compositions observed in these mafic rocks. This inference is possible because these mafic rocks plot below and align well with the melting trend of garnet clinopyroxene peridotite derived from the most depleted mantle component originating from the Kerguelen plume (Fig. 11c). (3) Group 2 mafic lavas have comparatively constant Sm/Yb ratios and variable La/Sm ratios, and hence create a relatively horizontal trend on the Sm/Yb vs. La/Sm diagram (Fig. 11c), indicating a spinellherzolite source. Group 2 samples have La/Yb ratios greater than those that could be generated by a singlestage melting of asthenospheric mantle; it can thus be

18 164 D. Zhu et al. / Lithos 100 (2008) argued that the one-stage melting of a normal convecting asthenospheric mantle source with a DMM-like composition cannot account for the observed REE ratios in these mafic rocks, rather an enrichment event (either source- or process-related or both) is required. In summary, it is probable that a common plagioclase accumulation exists in the Cona mafic magmas during late magma fractionation. Variable degrees of partial melting of distinct mantle source compositions (related to enriched garnet clinopyroxene peridotite for Group 1 and spinel-lherzolite for Group 2, respectively) appear to be very important, although fractional crystallization seems to be required prior to emplacement in the generation of the Cona mafic magmas. This suggestion is similar to the conclusions of Davis et al. (1994) and Moll-Stalcup (1994), who proposed that the compositional variation of basaltic magmas erupted at a given site in the Bering Sea basalt province reflects varying degrees of melting of mantle source material. Archipelago. Moreover, the available Sr and Nd isotopic compositions of the LIS suite in Group 1 overlap with Bunbury Casuarina basalts, Rajmahal Group I basalts, and Group P lavas in the 143 Nd/ 144 Nd(T) vs. 87 Sr/ 86 Sr (T) diagram (Fig. 12b). It should be mentioned that two basalt samples in Group 1 (LIS suite) are closely comparable to the basalts derived from the Cretaceous Kerguelen plume head identified by Ingle et al. (2003, 2004). Geochemical comparison between many pairs of incompatible trace elements and Sr Nd isotopic ratios of Group 1 mafic rocks from the Cona area shows similar features to the basalts probably created by the Kerguelen plume seems to indicate a similar magma source and origin Could the Cona mafic rocks be caused by the Kerguelen plume? The OIB-type geochemical features for Group 1 samples presented above suggest a plume origin. An important issue for the mafic magmatism in the Cona area is which plume was involved in their generation during the Early Cretaceous time Geochemical comparison with the basalts probably created by the Kerguelen plume It has been suggested that Group P lavas (P indicating plume-derived) with low MgO content (b6% MgO), relatively high 87 Sr/ 86 Sr, and low 143 Nd/ 144 Nd ratios from Mont Bureau and Mont Rabouillère in the northern part of the Kerguelen Archipelago are probably intrinsic to the Kerguelen plume (Yang et al., 1998). A good match of the primitive mantle-normalized multi-element diagrams (Fig. 8a and b) between the mafic rocks in Group 1 and Group P lavas and Sangxiu basalts is observed, suggesting a similar source for these mafic lavas. This inference can also be observed in the Lu/Hf vs. Sm/Nd diagram (Fig. 12a), which offers a good opportunity to address the effects of partial melting and the chemical composition of the source on the petrogenesis of the Cona mafic rocks as these ratios can minimize the effects of magmatic differentiation. It has been noted that the lower limited Lu/Hf ( ) and variable Sm/Nd ( ) ratios (Fig. 12a) of Group 1 samples from the Cona area are similar to those of the Sangxiu basalts and Group P lavas from Mont Bureau and Mont Rabouillère on the Northern Kerguelen Fig. 12. Sm/Nd vs. Lu/Hf (a) and 143 Nd/ 144 Nd vs. 87 Sr/ 86 Sr (b) diagrams of the Cona mafic rocks for geochemical comparison with the basalts probably created by the Kerguelen plume. Data sources: RM82-8 (Mahoney et al., 1983), Cretaceous Kerguelen plume head (Ingle et al., 2003, 2004), Kerguelen Archipelago basalts, Ninetyeast Ridge basalts, and Sites 750, 738 (Frey et al., 2000), others are same as the Fig. 8. Note that Group 1 mafic rocks exhibit similar Sm/Nd, Lu/Hf ratios to the Sangxiu basalts and Group P lavas, Group 2 mafic rocks overlap with the sample RM82-8 and Rajmahal Group I basalts, and that the available Sr and Nd isotopic compositions of the LIS suite in Group 1 are closely comparable to the basalts derived from the Cretaceous Kerguelen plume head, Bunbury Casuarina basalts and Group P lavas. See text for details. Symbols as in legend for Fig. 5.

19 D. Zhu et al. / Lithos 100 (2008) The primitive mantle-normalized multi-element patterns of the mafic rocks in Group 2 are similar to N-MORB (Sun and McDonough, 1989) and lie within the patterns of the lavas from Site 758 on the Ninetyeast Ridge (Fig. 8c), except for negative Nb and Ta anomalies which can be attributed to the input of continental crust or lithosphere. Group 2 rocks also present similar highfield strength element (HFSEs) patterns (Fig. 8c) to the Bunbury Casuarina basalts and Rajmahal Group II basalts. Group 2 samples have high, limited Lu/Hf ( ) and Sm/Nd ( ) ratios that overlap with sample RM82-8, which is believed to be an uncontaminated sample from the Rajmahal area (Mahoney et al., 1983), and both plot within or near the fields of the Rajmahal Group I basalts and Bunbury Casuarina basalts (Fig. 12a). The Sr Nd isotopic ratios of Group 2 samples plot near the boundary of or within the field of the Ninetyeast Ridge basalts and close to the field of Site 750 basalts on the Ninetyeast Ridge in the 143 Nd/ 144 Nd(T) vs. 87 Sr/ 86 Sr(T) diagram (Fig. 12b). These observations suggest that the mafic lavas of Group 2 from the Cona area have geochemical similarities to some Ninetyeast Ridge and Rajmahal and Bunbury lavas, indicating a similarity in the nature of mantle source. It is notable that the Group 3 basalts exhibit similar trace element patterns (Fig. 8d) to the Bunbury Casuarina basalts and Rajmahal Group II basalts. The Lu/Hf ( ) and Sm/Nd ( ) ratios of the two basalts in Group 3 are intermediate between those of Group 1 and Group 2 samples, and close to the fields defined by Bunbury basalts and Rajmahal Group II basalts (Fig. 12a). These geochemical features may also suggest that they share a similar origin Age and tectonic constraints for the Cona mafic rocks Many pairs of incompatible trace elements and Sr Nd isotopic ratios presented above suggest a similar source for the Cona mafic lavas and the basalts probably created by the Kerguelen plume. Were the Kerguelen plume implicated, it would be relatively straightforward to explain the common geochemical characteristics of the Cona mafic rocks in south Tibet and continental basalts of the eastern Indian and southwestern Australian margins and those of the Kerguelen Plateau. This hypothesis appears to be plausible, but three important concerns should be considered further: (1) the location of the Kerguelen plume head before 120 Ma; (2) the large age gap between Group 1 rocks and the older Fig. 13. (a b) Generalized plate tectonic reconstruction (ca. 140 Ma, Boger et al., 2001) and the distributions of magnetic anomalies (ca. 130 Ma, Heine and Müller, 2005) of the eastern Gondwanaland. The major Proterozoic terranes for each continental block and the known locations of the Shillong Plateau (SP), Rajmahal Traps (RT) and Bunbury basalt (BB) are labeled. The location of the Jurassic province along the northwestern Australia margin and the area influenced by Kerguelen igneous and tectonic activity are from Segev (2002). The locations of the Cona mafic rocks (large red star) and the Sangxiu volcanic rocks (small black star) in the northeastern Greater India are shown in view of present distance and Cenozoic shortening relative to the Rajmahal Traps province. Note that the Cona area can match well with the northwestern Australia in Fig. 13b. See text for details. (For interpretation of the references to colour in this figure legend, the reader is referred to the web version of this article.)

20 166 D. Zhu et al. / Lithos 100 (2008) Bunbury Casuarina basalts of 14 Myr and the age of Group 1 considerably preceded the extensive effusion on the Southern Kerguelen Plateau ( Ma) by Myr; and (3) the small volume of the Cona mafic rocks is inconsistent with a mantle plume head model characterized by a huge volume of volcanism. In order to evaluate the relationship of the Cona mafic rocks to the Kerguelen plume, it is first necessary to know the spatial relationship of these rocks to the Kerguelen plume between Ma. Early papers suggested that at 130 Ma the Kerguelen plume was beneath the triple junction between Australia, Antarctica, and Greater India (Davies et al., 1989; Curray and Munasinghe, 1991). Kent et al. (2002) showed that the calculated Kerguelen plume is closer to northeastern India at Ma. New paleomagnetic results indeed indicate that the Kerguelen plume has moved southward 3 10 since 120 Ma (Antretter et al., 2002). The Cretaceous Cona mafic rocks are now exposed 400 km north of Rajmahal Traps province (Fig. 1b), and lie near the northwestern margin of Australia on the plate tectonic reconstruction of eastern Gondwana ca. 140 Ma (Fig. 13a). A km shortening (Hodges, 2000) occurred in the region between the foreland and the Indus-Yarlung Zangbo suture during the Cenozoic collision between India and Eurasia; hence the Cona area is located km north of the Cretaceous Rajmahal Traps province. This distance is actually compatible with the location of the Cretaceous Kerguelen plume as indicated by new paleomagnetic results. Therefore, the spatial distance between the Cona mafic rocks and the Kerguelen plume before 120 Myr is not a convincing argument against the hypothesis that the Cona mafic rocks in the eastern Himalaya could be affected by the Kerguelen plume at that time. The major difficulty with this interpretation is that thinning of the lithosphere during the breakup of a continent overlying a plume should have led to the formation of a large igneous province between Ma. White and McKenzie (1989) proposed that flood basalts are due to the buildup of plume material beneath the continental lithosphere, followed by lithospheric extension associated with continental breakup. Kent (1991) and Kent et al. (1992) argued that long-term lithospheric uplift in eastern Gondwana resulted from a hot plume residing beneath the continental lithosphere. In these scenarios the Cona magmatism could be attributed to the effects of plume incubation, i.e., mixing of components derived from the plume and incubated continental lithosphere. Coffin et al. (2002) suggested that the initial Kerguelen plume head might have broken into droplets of variable sizes within the rapidly convecting upper mantle, resulting in spatially and temporally displaced magmatic events of differing volumes, all ultimately related to the Kerguelen plume. In this scenario, it is probable that more than one droplet could be recorded on neighboring continental margins. Regarding the Cona mafic magmatism, at least four observations should be considered in any genetic model of their generation: (1) the contemporary mafic rocks (including extrusive and intrusive components) are distributed only in the eastern Himalaya with an areal extent of about 240 km 135 km, which is comparable to that of the Bunbury basalts; (2) the LIS suite of Group 1 rocks from the Cona area is consistent with the Kerguelen plume head-related magmatism during Cretaceous time (Fig. 12b); (3) the later magmatism in the Cona area is coeval with the basaltic magmatism in southwestern Australia (Bunbury basalt) and the anatexis of ensialic crust in southeastern Yangzuoyong Co. (Sangxiu dacite); and (4) the common existence of a magma chamber/pond in the continental lithosphere during the Cretaceous time as indicated by a similar plagioclase accumulation from the Cona mafic rocks, the Sangxiu basalts (Zhu et al., 2007), and the Rajmahal Traps (Ghose et al., 1996; Srivastava and Sinha, 2004). Considering the geochemical similarities and spatial temporal constraints presented above, we propose here an incubating Kerguelen plume (i.e., incubating during the earliest Early Cretaceous time) model combined with the genetic explanations proposed by Kent (1991), Kent et al. (1992), andcoffin et al. (2002), which seem applicable to the Cona mafic magmatism. If this ad hoc suggestion has some plausibility, it seems to imply that (1) the Kerguelen plume may have started its incubation as early as the latest Jurassic or earliest Cretaceous period; (2) Group 1 magma may relate to one early droplet originating from the initial Kerguelen plume head, and the later synchronal mafic magmas in the Cona area and southwestern Australia and felsic magma in the Sangxiu Formation may be associated with the later two droplets derived from the Kerguelen plume in terms of material and/or heat; and (3) the large effusion of magma will have occurred when the magma chamber/pond was sufficiently large, sufficiently hot, and the channel of the Kerguelen plume magma was unblocked. Of course, it should be noted that a lesser-known Jurassic igneous province along the northwestern Australian margin (Fig. 13a and b; Hopper et al., 1992; Crawford and von Rad, 1994; Boger et al., 2001) attributed to a mantle plume (White and McKenzie, 1989) is another possible plume source candidate for the generation of the Cona mafic magmatism. However, this province is mainly characterized by evolved trachytic to

21 D. Zhu et al. / Lithos 100 (2008) rhyolitic lavas that have been dated to Early Jurassic (von Rad et al., 1992) and ended at 155 Ma (K-Ar date, near the Scott Plateau and the Exmouth Plateau, Ludden, 1992). According to plate reconstructions at 140 Ma and 130 Ma (Fig. 13a and b), the distance from this Jurassic province was too large to account for the Cona mafic magmatism, although the total distance between these two locations in the late Jurassic time have not been determined. We therefore argue against the hypothesis that the Cona mafic magmatism can be sourced from the Jurassic plume recovered from the northwestern Australia margin Petrogenetic explanation for the Cona mafic rocks A feature of many flood basalt provinces is the existence of high-ti and low-ti basaltic magmas that were approximately contemporaneously erupted in distinct geographic distributions (such as the Parana, the Karoo, and the Emeishan basalts). Condie (2001) argued that the high-ti source is a plume head, and that the low-ti might be sourced from a component in the subcontinental lithosphere rather than from a shallower level in a plume head according to the distinct geographical distribution of the high-ti and low-ti basalts. Although Group 1 magma with high-ti features and Group 2 magma with low-ti features are observed in the Cona area, they are significantly different from the high- Ti and low-ti basaltic magmas in some flood basalt provinces. This is mainly due to the fact that (1) Group 1 diabases were intruded by some Group 2 diabases with a 14 Ma age gap and (2) no spatial difference is observed between Groups 1 and 2. These differences may indicate a distinct petrogenesis relative to the high-ti and low-ti basaltic magmas from flood basalt provinces. As presented above, major and trace element data and Sr Nd isotopic compositions suggest that the lavas in Group 1 may have derived from an OIB-type source associated with partial melting of a garnet-bearing mantle source. Considering the broad geochemical similarities between Group 1 samples and the basalts probably created by the Kerguelen plume based on the spatial temporal constraints, we propose an attractive model of the incubating Kerguelen plume lithosphere interaction for the generation of Group 1 magma in the Cona area of the eastern Himalaya. Group 2 magmas cannot be derived from Group 1 magma by fractional crystallization because they have similar SiO 2 content and different Sr and Nd isotopic ratios (Tables 1 and 2). As discussed earlier, Group 2 magmas appear to be related to partial melting of a spinel-bearing mantle source and some are probably contaminated by lower crust. As the bulk of the lithospheric mantle (i.e., that above the thermal boundary layer) is substantially cooler than the underlying convective mantle, it therefore requires additional heat to induce melting (McKenzie and Bickle, 1988). It is generally agreed that an underlying plume may supply the additional heat required to melt the anhydrous lithosphere and produce a small volume of melt, whereas partial melting of a hydrous mantle lithosphere, which is trace element enriched due to previous subduction events, can result in the large volumes observed in continental flood basalt provinces (Arndt and Christensen, 1992; Reichow et al., 2005). In addition, a depleted N-MORB-type REE pattern and higher ɛnd(t) values (+5.68 to +6.37) observed in Group 2 samples indicate that a depleted asthenosphere material is also required in their generation. Combined with the fact that Group 2 shows geochemical similarities to some Ninetyeast Ridge and Rajmahal and Bunbury lavas and based on the spatial temporal constraints, the Cona mafic lavas in Group 2 are interpreted as originating from an interaction between anhydrous lithosphere and rising depleted asthenosphere enriched by a droplet originating from the Kerguelen plume. Amagmatic intervals or a few sporadic crust-contaminated magmatisms (e.g., Group 3 magmas) may be attributed to thermal erosion and result in the partial melting of lithosphere (Johnston and Thorkelson, 2000) during the long-term incubation of a magma chamber/ pond at a shallow crustal level, in which the heat and materials can be supplied by capturing discontinuous droplets originating from the Kerguelen plume Could a nonplume origin be a possible alternative explanation for the Cona mafic rocks? It is in fact difficult to recognize plume involvement based on geochemistry because (1) there are multiple mantle reservoirs that can contribute and (2) there is uncertainty about which of these are located in the deep mantle (Ernst et al., 2005). In the case of the Cona mafic lavas, the geochemical overlap with the basalts probably created by the Kerguelen plume may be fortuitous, and the incubating Kerguelen plume origin may not be a unique possible explanation for their generation. In such case, a nonplume origin could be considered. In recent years, many models of nonplume origin for LIP formation have been proposed, as summarized by Ernst et al. (2005). An alternative possible explanation for the generation of the Cona mafic lavas involves normal continental rifting (nonplume). A widely accepted view is that the mixing of melts from depleted lherzolite and

22 168 D. Zhu et al. / Lithos 100 (2008) chemically enriched mafic veins (e.g., garnet pyroxenites) embedded in a peridotite matrix (without involved plume) can account for the incompatible element enrichment observed in OIB-type magmas (Lassiter et al., 2000). Such an explanation probably fits the generation of the OIB-type trace element signatures of Group 1 lavas, as these lavas are likely to be connected with the partial melting of garnetbearing mantle (Fig. 11c). White (1992) proposed that if the rifting process prior to breakup is prolonged, extending over periods of 10 Myr or more, the mantle rising beneath the rift may cool by conduction, generating even less melt than would be expected for instantaneous rifting above normal temperature. This may allow stretching of the continental crust to as thin as 3 4 km, even thinner than normal oceanic crust; it may also cause the initial oceanic crust immediately adjacent to the rifted continent to be thinner than its normal 7 km. In relation to the Neo-Tethyan spreading during Cretaceous time, this geodynamic scenario seems likely for the generation of the small volume of the Cona mafic lavas, as the Tethyan Himalaya experienced the long-term rifting event that is commonly linked to the separation of the Lhasa terrane from northern India and the eventual opening of Neo-Tethys (Yin, 2006). However, if we scan the Cona area more broadly, it appears that the Cona mafic lavas are more likely associated with the initial opening of the eastern Indian Ocean between northeastern Greater India and northwestern Australia relative to the Neo-Tethyan spreading, as this area matches well with northwestern Australia in the tectonic reconstruction of eastern Gondwana ca Ma (Fig. 13a and b). It has been shown that, in most cases, initial large igneous province (LIP) magmatism precedes rifting (Campbell, 1998; Courtillot et al., 1999; Menzies et al., 2002), favoring a plume origin (Ernst et al., 2005). The fact that the OIB-type mafic magmatism (144.7 Ma) recorded in the Cona area precedes the oldest magnetic anomaly between northeastern Greater India and northwestern Australia (130.9 Ma, M10N in Fig. 13b, Heine and Müller, 2005) by 15 Ma may support a plume origin for the generation of the Cona mafic rocks. It has been proposed that continental LIPs can generally be divided into two distinct pulses, a prerift pulse and a postrift (or syn-rift) pulse (Campbell, 1998; Courtillot et al., 1999). The prerift pulse is linked to the arrival of a new mantle plume, and is the first eruptive phase from that plume. The postrift pulse, which includes seaward-dipping reflectors (White and McKenzie, 1989) and zones of high-velocity lower crust (Menzies et al., 2002), is produced during rifting associated with the continental breakup and can be interpreted as resulting from hot mantle in the plume head being drawn into the zone of rifting. The time gap between the two pulses varies from case to case, but is typically a few million to tens of millions of years (Ernst et al., 2005). The volume of magma produced during the second pulse may exceed to that produced during the first (Campbell, 1998; Courtillot et al., 1999). Such a relationship between continental LIPs and rifting is likely applicable to the generation of the Cona mafic lavas and the breakup of eastern Gondwanaland. The Cona mafic magmatism could be interpreted as the prerift pulse and as the arrival of one droplet originating from the incubating Kerguelen mantle plume, and as the first eruptive phase from this plume. In summary, therefore, an incubating Kerguelen plume model is more attractive than a model of normal continental rifting (nonplume) for the generation of the Cona mafic rocks. However, the criteria for interpreting an incubating Kerguelen plume origin for the Cona mafic magmatism and other related areas such as a link to surface uplift, compositional geochemical and geochronological arguments (including the presence of high Mg rocks, a high 3 He/ 4 He ratio, and Argon Argon dating or SHRIMP zircon dating, etc.), a link with an age-progressive hotspot track with the Kerguelen plume and other criteria (Ernst et al., 2005) will be tested in future work in order to support or reject such a hypothesis Implication for the Kerguelen plume and the breakup of eastern Gondwana Thompson et al. (1980) noted that the early transitional alkali basalts from Skye, NW Scotland, had depleted heavy REE patterns, whereas the later more tholeiitic lavas generally possessed much flatter heavy REE patterns. Ellam (1992) interpreted the observations of Thompson et al. (1980) in terms of a progressive lithosphere thinning beneath Skye during the development of the lava succession. Kerr (1994) proposed that the observed degree of lithospheric thinning cannot be explained by extension alone, and it appears that some relatively rapid lower lithospheric erosion by the plume head is also required. To assess the influence of an asthenospheric source region on the chemistry of the Cona mafic rocks and other associated rocks in southwestern Australia and eastern India, the altered and contaminated Cona mafic rocks (e.g., the HIS suite in Group 1 and Group 3 samples) are filtered. Both Ce/Yb and Sm/Yb ratios

23 D. Zhu et al. / Lithos 100 (2008) Fig. 14. Ce/Yb vs. Sm/Yb plots comparing the Cona mafic rocks, the Sangxiu basalts, the Bunbury Casuarina and Rajmahal Group I lavas with model melts generated by progressive lithospheric extension (Ellam, 1992). Model curve is a path of accumulated melts derived by fractional melting assuming melting of a source with uniform Ce/Yb and Sm/Yb ratios, beginning at a depth of 125 km, and using parameters given by Ellam (1992). This figure presents results for a range of lithospheric thickness between 110 and 0 km. Tick marks on model curve indicate depth of final melt segregation in 10 km increments. Data for Skye are from Ellam (1992); others are same as the Fig. 8. Note that the lithospheric thickness decreased from Group 1 samples to the Bunbury Casuarina basalts in 15 Ma. Symbols as in legend for Fig. 5. offer sensitive indicators of changing lithospheric thickness that can usefully be applied to basalts because they will not be radically affected by fractional crystallization (Ellam, 1992). Plots comparing the mafic rocks in the Cona area, the Sangxiu basalts, the Bunbury Casuarina-type and Rajmahal Group I lavas with model melts generated by progressive lithospheric extension (Ellam, 1992) are illustrated in Fig. 14. Obviously, the high Ce/Yb and Sm/Yb ratios in Group 1 samples indicate a km depth of final melt segregation, whereas the 132 Ma magmatisms define a deepest depth of 45 km of final melt segregation (Bunbury Casuarina-type basalt). If this analysis were at least qualitatively correct, then it would seem that by the time the 132 Ma magmas were emplaced, lithospheric thinning of at least 15 km in 15 Ma attributed to extension of the lithosphere and/or erosion of subcontinental lithosphere mantle by the Kerguelen plume head had already occurred beneath eastern Gondwanaland. These observations are similar to those for Skye basalts (Thompson et al., 1980) and the plume-related tertiary lavas from the Isle of Mull, Scotland (Kerr, 1994), which seems to indicate that a progressively thinning lithosphere probably existed beneath eastern Gondwanaland from Ma to 130 Ma. In addition, Yuen and Fleitout (1985) calculated that small-scale secondary convection associated with extra heat and extension could cause thinning of a 100-km-thick lithosphere by 15 km in 2 Mys. This rate of lithospheric thinning is considerably higher than the rate ( 1 km/ma) estimated from Group 1 samples in the Cona area and the Bunbury Casuarina-type basalt (Fig. 14), implying that lithospheric erosion is a relatively slow process, consistent with the plume-incubation model proposed by Kent et al. (1992). The older basaltic and associated diabase magmatism (Table 1) in the Cona area is actually approximately coeval with the old known magnetic anomalies ranging from Ma (M25) to Ma (M16) in the Indian Ocean (Veevers et al., 1991; Heine and Müller, 2005) off the northwest coast of Australia (Fig. 13b). These were interpreted as representing the activity of the spreading Argo ridge around the northern margin of Greater India (Heine and Müller, 2005). A southward ridge jump occurred at M14 (135.8 Ma) in the western spreading segment along the western Australian margin (Powell and Luyendyk, 1982; Heine and Müller, 2005), and then transferred along the western coast of Australia. The geochemical and spatial constraints shown in the current paper allow us to assume that the incubating Kerguelen plume may have been present during the early rifting activity off the northeast coast of Greater India. The notably thinned lithosphere attributed to extension and erosion by the incubating Kerguelen plume from 150 Ma to 132 Ma will lead to enhanced magmatic events, as documented from Group 2 mafic magmatism in Cona ( Ma), the anatexis of ensialic continental crust recovered from the Sangxiu felsic rocks ( 133 Ma), and the Bunbury Casuarina magmatism ( Ma) in southwestern Australia. These increasing magmatic activities may indicate the initial separation of northeast Greater India and northwest Australia, as indicated by the contemporaneous oldest magnetic anomaly between northeast Greater India and northwest Australia, which occurred at Ma (M10 N, Heine and Müller, 2005) along the Cape Range fracture zone (CRFz) and Wallaby Zenith fracture zone (WZFz), which were close to the location of the Cona mafic rocks and near the location of the Bunbury basalts at 130 Ma (Fig. 13b). Subsequently, separation of northern India and western Australia may have proceeded as an unzipping from north to south over a time interval spanning more than 20 Mys (Ingle et al., 2002). In this case, it is probable that the incubating Kerguelen plume played an active role in the breakup of Greater India, eastern India, and northwestern Australia.