Petrogenesis of Cenozoic Basalts from Vietnam: Implication for Origins of a ‘Diffuse Igneous Province’

after the early Tertiary India‐Asia collision. This activity does not conform to the ‘Large Igneous Province’ model in view of lower eruption and melt production rates, wide dispersal of centres and the apparent absence of deep mantle upwelling. Age data for Vietnamese plateau basalts reflect spatial‐temporal KEY WORDS: Vietnam; basalt; Cenozoic; geochemistry patterns consistent with a rotating stress field rather than suprahotspot lithosphere migration. For most of the volcanic centres there are two eruptive episodes: an early series formed by highSiO2, low-FeO* quartz and olivine tholeiites—large melt fractions of refractory (lithosphere-like) mantle—and a later series made INTRODUCTION up of low-SiO2, high-FeO* olivine tholeiites, alkali basalts and Neogene‐Quaternary intraplate volcanism is widespread basanites—smaller melt fractions of more fertile (asthenosphere- in east and southeast Asia (Fig. 1a) forming basalt plateaux like) mantle. Comparison of Mg-15 normalized basalt com- associated with pull-apart, extensional rifts (Barr & positions with parameterized anhydrous and hydrous experimental McDonald, 1981; Whitford-Stark, 1987). Although melt compositions allowed calculation of melt segregation pressures widely dispersed the activity shares common source isoand temperatures. Computed for anhydrous conditions these range topic and lithosphere structural character with intraplate from <4 GPa and ~1470∞C (for alkali basalts) to <0·5 GPa and back-arc volcanism in the western Pacific and has and ~1400∞C (quartz tholeiites), and for H2O-undersaturated been referred to as a ‘diVuse’ igneous province (Hoang conditions, from <3·5 GPa and ~1450∞C to ~1·5 GPa and et al., 1996). The activity post-dates the early Tertiary 1350‐1400∞C, respectively. Hydrous conditions are more realistic India‐Asia collision and may be related to asthenospheric in view of high measured basalt H2O + contents, pressure and lithospheric tectonic extrusion processes (Tapponnier estimates consistent with melting below a thinned mechanical et al., 1982, 1986). The province is bounded to the east boundary layer (MBL) and interpolated mantle adiabats of and southeast by active subduction at the Izu‐Bonin, 2‐3∞C/km (compared with <1∞C/km for anhydrous conditions), Mariana and Indonesian archipelagos, and to the northconsistent with fluid dynamic constraints and a 1440∞C potential west by the collision-thickened Tibet plateau (Flower et temperature. After collision-induced ‘extrusion’ of east and al., 1998a). The Indochina and China plates appear to southeast Asia, the lithosphere was probably thinned during have been tectonically extruded along regional strike-slip faults, with concomitant opening of the South China

Although widely dispersed, east Asian Cenozoic basalt cussed by Hoang et al. (1996) with reference to Vietnamese basalts, and by Flower et al. (1998a) with respect to east magmatism does not conform to the 'Large Igneous Province' model (Coffin & Eldholm, 1994) in view of the Asia in general.
The causes of dispersed, relatively sudden mantle melt-lower rates of eruption and melt production, dispersal of eruptive centres and apparent absence of deep mantle ing events have not been extensively discussed in the literature although there has been progress in under-upwelling ( Su et al., 1994). In common with other 'diffuse' provinces (e.g. Ormerod et al., 1988;Hoernle et al., 1995) standing the constraints of pressure, temperature, P (H 2 O) and mantle fertility on melt composition and volume the east Asian activity reflects contemporaneous, rapid appearance of dispersed basalt centres, a transtensional (e.g. McKenzie & Bickle, 1988;Latin & White, 1990;Wilson, 1993). Here, we advance the idea that 'diffuse' setting, and proximity to major continent-continent plate collisions. In addition, east Asian continental and western volcanic provinces may reflect the combined effects of lithosphere transtension and asthenosphere Pacific back-arc volcanics appear to share a common asthenosphere isotopic signature, sometimes referred to decompression concomitant with collision-induced  (Tung & Tri, 1992). I, Northern accretionary belt (Precambrian to Palaeozoic); II, Central-Kontum Massif (Archaean, Proterozoic, Palaeozoic); III, Southwest-Khorat Plateau (Precambrian) and surrounding Palaeozoic and Mesozoic belts; IV, Southeast-undifferentiated Precambrian overlain by Mesozoic. (b). Map of south-central Vietnam (insert from a). Basalt centres shaded, drill site locations numbered, and surface or dredge sample site numbers italicized. Lithospheric sector boundaries shown by dashed lines, lithospheric sub-sector boundaries by dotted lines (IIa-d, Archaean, Proterozoic, Cambrian and Permo-Triassic, respectively). extrusion of thermally anomalous mantle. Assuming that Quoc & Giao, 1980;Hoang & Han, 1990;Thi, 1991; the volumes, supply rates and compositions of mantle and references given by Whitford-Stark (1987)]. Most of melts are simple functions of mantle composition, thermal these centres are associated with pull-apart structures structure and the extent of lithosphere stretching [folcomprising short extensional rifts bounded by strike-slip lowing McKenzie & Bickle (1988) and Latin & White faults (Rangin et al., 1995a). The centres appear to have (1990)], we develop a petrogenetic model for the Vietinvolved at least two eruptive episodes, referred to here namese basalts as a basis for understanding possible as 'early' and 'late' eruptive series, thick palaeosols markasthenospheric thermal responses during the closure of ing the intervening quiescent periods (Hoang, 1996). eastern Tethys.
Early episodes usually produced quartz and olivine tholeiite flows, with rare alkali basalt, whereas later episodes erupted olivine tholeiite, alkali basalt, basanite and (rarely) nephelinite. This bi-episodal pattern is recognized at the

VIETNAM
and Re Island centres, and probably other offshore

Ages and eruption rates of basalts
localities, although at Buon Ma Thuot the compositional trend is inverted (see below). Tholeiitic eruptives are the Basalt plateaux in southern and central Vietnam often most voluminous, with flows up to 30 m thick erupted exceed 100 km in diameter, are up to several hundred from axial rifts. Alkali basalts, basanites, and rare nemetres thick, and cover a total area of~23 000 km 2 ( Fig. 1a) (Lacroix, 1933;Carbonnel & Saurin, 1975; phelinites formed thinner, more sporadic flows erupted VOLUME 39 NUMBER 3 MARCH 1998 mostly from small central volcanoes aligned on conjugate rates of up to~2700 km 3 /my, assuming a total of at strike-slip faults (Quoc & Giao, 1980;Hoang & Han, least 8000 km 3 basalt were erupted at the major centres. 1990;Thi, 1991;Hoang, 1996). However, such estimates underplay the effects of frac- Barr & Macdonald (1981) reviewed new and published tional crystallization and also ignore the presence of K-Ar age data for southeast Asian Cenozoic basalts and trapped melt consolidated at depth (see Latin et al., 1993). concluded that basalt activity in Indochina appeared by at least 12 Ma, after the cessation of South China Sea opening, and peaked in the last 3 my with an equally Structural and dynamic controls on rapid abatement. More recent K-Ar data (Novikov et al., volcanism 1989; Arva- Sös et al., 1990;Rangin et al., 1995a) and 24 The Indochina peninsula comprises fragments of Gondhigh-precision Ar-Ar age dates for our stratigraphically wana, made up of Precambrian, Palaeozoic and Mesozoic selected core samples (  suggest the Vietcrust, that migrated northwards during the Palaeozoic namese centres were active over the following intervals: and accreted to pre-Tethyan Eurasia (e.g. Gatinski et al., Dalat (17·6-7·9 Ma), Phuoc Long (straddling the border 1984;Hutchison, 1989;Tung & Tri, 1992). These are with Cambodia) (<8-3·4 Ma), Buon Ma Thuot (5·8-1·67 preserved as distinct lithospheric sectors separated by Ma), Pleiku (4·3-0·8 Ma), Xuan Loc (0·83-0·44 Ma) and tectonic sutures of known age ( Fig. 1a and b) (e.g. Tung the Île des Cendres (0·8-0 Ma), and confirm the bi-& Tri, 1992). A northern sector (I) comprises an accreted episodal eruptive pattern (Fig. 1b).
Archaean, Proterozoic and Palaeozoic complex in north-Although the bulk of volcanism in Indochina postern Vietnam and Laos, and southern China through dates South China Sea spreading, Paleogene activity has which Cenozoic basalts were erupted at Dien Bien Phu, been recorded by drilling on the southern Chinese and Phu Quy, Con Co island and Khe Sanh (Fig. 1a), Vietnamese continental shelves (e.g. Zhang Qi Ming, and also localities in Hainan, Yunnan and Guangdong Nanhai West Co., personal communication, 1989) and provinces (South China), and the South China Sea (e.g. appears in Thailand, NW Vietnam, Yunnan and Sichuan Flower et al., 1992;Tu et al., 1992). A central sector (II) close to the Ailao Shan-Red River (ASRR) and other comprises the Kontum Massif, a quasi-cratonic block strike-slip shear zones (Flower et al., 1998b). Historic with a 2·8 Ga core (Archaean) (Tung & Tri, 1992) activity in Vietnam is confined to the offshore Con Son surrounded by concentric 'sub-sectors' separated by Protswell (Île des Cendres, Fig. 1a), sporadic ash eruptions erozoic, Cambrian and Permo-Triassic sutures (Fig. 1b). in the central highlands (e.g. Pleiku, April, 1993), and Cenozoic basalts were erupted within this sector at Song submarine activity along the eastern seaboard (Koloskov Cau (Archaean), Buon Ma Thuot, Pleiku and Kong et al., 1986). Late series undersaturated lavas carry mantle Plong (Proterozoic), and at smaller centres in Quang xenoliths, including garnet lherzolites, spinel lherzolites Ngai and Re Island, and offshore north of latitude 15°N and harzburgite, and megacrysts of pyroxene, olivine, (Cambrian) ( Fig. 1a and b). A southwestern sector (III) plagioclase, garnet, zircon and corundum (e.g. Saincludes the eastern part of the Khorat Plateau, which pozhinkov et al., 1979;Han & Hoang, 1985; probably has a Precambrian core (in Thailand) enclosed Han, 1990). The latter characterize centres in southern by accretionary Palaeozoic and Mesozoic belts (in Cam-Vietnam (especially Pleiku, Xuan Loc, Buon Ma Thuot bodia and southwest Vietnam) (Hutchison, 1989). This and Île des Cendres), Cambodia and Thailand (Lacombe, sector contains the largest single basalt complex in Indo-1967; Barr & MacDonald, 1981), and Hainan Island, china (Phuoc Long; 200 km across and up to 500 m Mingxi and other southern Chinese localities [Flower et thick) which straddles the Cambodia- Vietnam border. al., 1992;compare Sutherland (1983) Mukasa & Zhou, 1994;Intasopa et al., 1995;volumes erupted in Vietnam: first, the large number of Mukasa et al., 1996) (Fig. 1a). A southeastern sector (IV) cored hydrologic sections drilled through the basalts to probably includes most of southern Vietnam and consists basement, second, the well-documented areal extent of of accreted Proterozoic, Palaeozoic and Mesozoic basethe eruptives, and third, their essentially basaltic charment (Tung & Tri, 1992). With estimated crustal thickacter. Provisional estimates of magmatic volume for the nesses of~30 km (Tien, personal communication, 1993) principal centres are 1500 km 3 for Dalat, 2200 km 3 for this sector includes basalts of both the oldest (Dalat) and Phuoc Long, 1500 km 3 for Buon Ma Thuot, 2000 km 3 youngest (Xuan Loc) onshore complexes, and the active for Pleiku and 500 km 3 for Xuan Loc, and for smaller offshore Île des Cendres and Katwit centres ( Fig. 1a centres to the north are, for example, 180 km 3 for Song and b). Cau and 160 km 3 for Kong Plong ( Fig. 1a and b). Despite Tapponnier et al. (1982Tapponnier et al. ( , 1986 proposed that the Indoa lack of age data for some smaller centres, the known age and volume relationships suggest magma production china plate was extruded southeastwards and rotated anticlockwise after the India-Asia collision. Developed moderately phyric (<10% phenocrysts) with plagioclase (An 83-72 ), olivine (Fo 78-74 ) and augite (Wo 47-49 En  ). on the basis of scaled experiments, the extrusion model has dominated post-Mesozoic plate tectonic re-Most are petrographically similar to mid-ocean ridge basalt (MORB), with plagioclase preceding clino-constructions in the region and provides a simple explanation for the opening of contiguous marginal basins. pyroxene at the low-pressure liquidus. Quartz tholeiites from early Phuoc Long and Dalat series and Dien Bien Although the extent of extrusion has been questioned in the light of additional experiments (e.g. Jolivet et al., Phu, however, may contain small, unreacted phenocrysts 1990), and palaeomagnetic ( McCabe et al., 1993), gravity of orthopyroxene (~En 82-78 ), in which respect they re-  and seismic data ( Rangin et al., semble volcanic arc rather than MORB tholeiite (e.g. 1995b), metamorphic thermochronology ( Leloup et al., Kushiro, 1990). Orthopyroxene-phyric tholeiites are 1995) and igneous age dates  associated highly unusual in intraplate tectonic settings and their with the ASRR shear zone confirm that 500-600 km petrogenesis may reflect high P(H 2 O) in the mantle source left-lateral motion occurred between~30 and 17 Ma, (see below). Olivine tholeiites are aphyric to sparsely followed by a few tens of kilometres of right-lateral phyric and are dominant at most other centres, usually motion. Whether South China Sea opening was wholly interlayered with quartz tholeiite and lesser amounts of a response to extrusion (Tapponnier et al., 1986; Briais alkali basalt. Phenocrysts rarely exceed 10-15% by vol et al., 1993) or represents Pacific plate-induced back-arc ume and are mostly olivine (Fo 82-78 ) and augite (Wo 44spreading (Taylor & Hayes, 1983) is unresolved , with lesser amounts of plagioclase et al., 1995b). In either case, the migration of magmatism (An 85-68 ). Alkali basalts and basanites are common in late from the South China Sea spreading axis to Indochina series of Xuan Loc and Re Island centres, whereas alkali marks a major mid-Miocene shift in the locus of lithobasalts are also prominent in the Pleiku late series and sphere extension (Le Pichon et al., 1995).
Buon Ma Thuot early series. These are moderately On the basis of palaeostress measurements in Vietnam, phyric, with 7-15% olivine phenocrysts (Fo 89-70 ), several Rangin et al. (1995a) argued that faulting during the generations of which may be distinguished on the basis Paleogene and Neogene was determined by two superof morphology and composition, together with lesser imposed stress systems. An older system compatible with amounts of augite. an east-west maximum compressional axis predominates Mantle xenoliths include garnet lherzolite, spinel lherin northern and central Vietnam and produced NW-SE zolite and harzburgite, along with eclogite of unknown left-lateral strike-slip faults (parallel to the ASRR shear provenance and cumulate xenoliths comprising wehrlite, zone) with conjugate SW-NE right-lateral faults (Fig. 1a). websterite and pyroxenite. Megacrysts include olivine, A younger system compatible with a north-south max-Al-rich clinopyroxene, orthopyroxene, Ti-amphibole, imum compressional axis produced dominantly NNWanorthoclase, phlogopite, sapphire and zircon (see Flower SSE to north-south right-lateral faults. This pattern of et al., 1992). Spinel lherzolites are abundant at Phu Quy stress redistribution is confirmed by our basalt Ar-Ar and Con Co island (sector I), Quang Ngai, Re Island, ages ( , which record a clockwise rotation Buon Ma Thuot and Pleiku (sector II), Dalat, Xuan Loc of transtensional fractures, an initial NE-SW trend (Dalat, and Île des Cendres (sector IV) along with cumulates 18-8 Ma) superceded by NW-SE (Buon Ma Thuot and and megacrysts. Eclogites were found in Pleiku basalts Pleiku, 8-2 Ma) and north-south (Pleiku, Xuan Loc and comprise idiomorphic garnet (40%) and euhedral to and offshore, 4-0 Ma) trends. Palaeomagnetic data for subhedral clinopyroxene (60%), with or without or-Vietnamese Mesozoic to Quaternary eruptives (McCabe thopyroxene (see Halton & Gurney, 1987). Feldspar et al., 1993;Chi et al., 1998) suggest little or no tectonic megacrysts were also found in Pleiku and range from rotation of Indochina occurred since the India-Asia sanidine to anorthoclase in composition. Clinopyroxene, collision, apparently at variance with predictions of the orthopyroxene and Ti-amphibole megacrysts were enextrusion model (Tapponnier et al., 1986). However, this countered mostly in Pleiku and Dalat centres, often with can be reconciled with extrusion if Indochina behaved amphibole-free lherzolites, and are common in Île des as a non-rigid plate (Rangin et al., 1995a(Rangin et al., , 1995b in which Cendres basalts. case basalt magmatism may reflect stretching associated with the change from left-to right-lateral ASRR shear zone motion ( Leloup et al., 1995).

GEOCHEMISTRY OF BASALTS AND XENOLITHS PETROGRAPHY OF BASALTS AND
Geochemical data for basalts and xenoliths are used to XENOLITHS define compositions and stratigraphic relations of the erupted magmas, identify compositions of their un-Quartz tholeiites are dominant in Dalat and Phuoc Long and are generally aphyric (<3% phenocrysts) to fractionated (mantle-equilibrated) parent melts, and from VOLUME 39 NUMBER 3 MARCH 1998 this basis estimate melt segregation conditions and ther-dominated by quartz tholeiite with subsidiary olivine tholeiite, basalts from Pleiku and Buon Ma Thuot include mal state of the asthenosphere. Assuming the entrained lherzolites and harzburgites represent lithospheric mantle quartz and olivine tholeiites and alkali basalts in approximately equal amounts, and Xuan Loc and offshore fragments and that high-pressure undersaturated melts are probably generated in the asthenosphere, estimated seamounts comprise olivine tholeiite and alkali basalt (Fig. 2). melt segregation conditions help define the 'thermal boundary layer' (TBL) between convecting as-'Chemical types' were established as a basis for stratigraphic correlation between drill sites (Hoang, 1996) thenosphere and a rigid 'mechanical boundary layer' (MBL). Sr, Nd and Pb isotopic data for representative and for comparing primitive magma types within and basalts-summarized below and discussed in detail by between centres (this work). Where possible, chemical Hoang et al. (1996)-confirm the presence of anomalous types were taken to include stratigraphic intervals of (DUPAL-like) asthenosphere beneath Indochina and resimilar composition-cooling unit batches identified from cord evidence for the interaction of asthenospheric melts drill core stratigraphy, or coherent groupings identified with enriched lithospheric mantle and continental crust. from surface sampling of mapped flows-whose internal Thermobarometric estimates of xenolith equilibration, a variation is consistent with crystal-liquid redistribution basis for estimating conductive geotherms in the lior mixing. Distinctions were based on major element thospheric mantle, are being published separately along oxides and CIPW normative character, although surface with chronologic interpretations of xenolith Re-Os and sample groups were further constrained by element ratios Sm-Nd isotopic decay systematics .
such as K/Na, Rb/Sr and Ba/Zr. This approach is exclusive rather than inclusive such that chemical types identified at different sites may represent different parts of the same erupted magma batch (e.g. types XL-B and

Sampling and analysis
-D at Xuan Loc; types DT-A, -B and -C at Dalat) Most samples analysed were selected from fresh, unwhereas a single type is unlikely to include compositions weathered drill-core from the Dalat, Phuoc Long, Buon from distinct eruptive episodes. Type averages are given Ma Thuot, Pleiku and Xuan Loc plateaux, surface outin Table 1, with representative incompatible element crops in these and other basalt plateaux, and dredge ratios, ages (where known) and melt segregation pressures hauls from the South China Sea. Stratigraphic sections and temperatures-the last computed from simulated were developed from hydrologic drill records of the primitive magma compositions (see below). Vietnamese Geological Survey and are accessible with Figure 3a-f shows chemical type variation in plots of geochemical data at the Journal of Petrology website (http:// SiO 2 and TiO 2 vs MgO (wt %) annotated for age where www.oup.co.uk/jnls/list/petroj). Dredge samples were possible, with early and late series outlined by continuous collected by Vietnamese and Soviet scientists during and dashed lines, respectively. Dalat centre basalts range cruises of the R.V. Vulkanolog between 1981 and 1987 between 6 and 9 wt % MgO and are exclusively tholeiitic ( Koloskov et al., 1986) (Fig. 1a and b).
( Fig. 3a), including four chemical types from drill sites Representative major and trace element analyses were 711, 736 and 756. Early series eruptives (10·5-14·0 Ma) published with isotopic data by Hoang et al. (1996) include low-Ti olivine and quartz tholeiite types DT-A together with descriptions, precisions and accuracies of and -B, whereas late series (1·8-2·6 and <~1·8 Ma) the analytical techniques employed. Whole-rock major comprise low-and high-Ti olivine tholeiite types DT-C and trace elements were determined using X-ray fluorand -D. Basalts from Phuoc Long section 804 also range escence spectrometry (Michigan State University) and between 6 and 9 wt % MgO and show a quartz tholeiite instrumental neutron activation analysis (University of type (PH-A) and two olivine tholeiite types (PH-B and Michigan), H 2 O + and CO 2 using a CHN analyser (Ar--C), types PH-A and -B forming a 15-15·1 Ma early gonne National Laboratory), and phenocryst comseries and PH-C a late series of~5 Ma (Fig. 3a). positions using a scanning electron microscope with Pleiku basalts are more variable, ranging between 5 energy-dispersive attachment (UIC). and 12 wt % MgO and showing a broad range of SiO 2 and TiO 2 contents (Fig. 3b). Early series eruptives (4·8-2·6 Ma) include quartz tholeiite (chemical types PL-A and

Major elements
-B), olivine tholeiite (PL-D and -E) and low-Ti alkali basalt (PL-F), whereas the late series type PL-C (<~2·5 CIPW normative variation defines a range of quartz Ma), a high-Ti alkali basalt, succeeds type A at site 90. tholeiite (QT) (qz-normative), olivine tholeiite (OT) (ol The early series types appear to comprise two groups, + hy-), alkali basalt (AB) (ne-up to 5%) and basanite with 5-8 wt % MgO (types PL-A and -B) and 8-12 wt (BA) (ne-normative >5%) at most centres, reflecting an % MgO (PL-C, -D and -F), types PL-F and -B, and PLoverall increase in undersaturation from older to younger centres (Fig. 2). The Phuoc Long and Dalat plateaux are D and -A representing possible cogenetic pairs. Type PL-C, the youngest of the early series (2·6 Ma), resembles and late alkali basalt series (type RE-B) (1·2-0·4 Ma) are recognized on Re Island, and analogous distinctions exist the late series type PL-E in SiO 2 but not TiO 2 content and appears to be unique. Buon Ma Thuot shows a for Île des Cendres (IC-A and -B) and Katwit Island seamounts. Thus, with the important exception of Buon similar compositional range, including quartz tholeiite (type BMT-B), olivine tholeiite (BMT-A and -C) and Ma Thuot, early eruptive series comprise quartz and olivine tholeiite with only rare alkali basalt (e.g. type PL-alkali basalt (BMT-D, -E and -F) (Fig. 3b). However, in contrast to the other centres, SiO 2 -undersaturated D in Pleiku), whereas late series are generally formed by olivine tholeiite, alkali basalt and basanite. Reasons for the eruptives are confined to the early BMT series (4·6-3·2 Ma) with late series made up exclusively of quartz tholeiite 'inverted' Buon Ma Thuot pattern are unclear, although it is possible these eruptives represent late and early re-(1·9-0·3 Ma). Although Buon Ma Thuot chemical types were identified from several sections, this 'inverted' spective phases of distinct magmatic cycles.
In general, the alkali basalts show higher contents of sequence is best seen at site 45, where quartz tholeiite (type BMT-B) overlies olivine tholeiite (BMT-A) and TiO 2 and FeO * than quartz and olivine tholeiites at equivalent MgO contents, such that a rough inverse alkali basalt (BMT-D).
Xuan Loc early and late series are distinguished at site correlation exists between SiO 2 undersaturation and TiO 2 or FeO * . Stratigraphic successions from low-to high-507, where olivine and quartz tholeiites (types XL-A and -C) (2·42-2·2 Ma) precede basanites (types XL-B TiO 2 basalts are common in intraplate flood basalts and have been interpreted to reflect variable source rather and -D) (1·1-0·4 Ma) (Fig. 3c). Offshore islands and seamounts show similar associations of tholeiite and alkali than fractional crystallization or contamination effects (e.g. Hawkesworth et al., 1988). However, as noted, this basalt although stratigraphic control is poor or lacking (Fig. 3c). An early tholeiite series (type RE-A) (12·0 Ma) pattern is not ubiquitous in Vietnam, as both Pleiku and VOLUME 39 NUMBER 3 MARCH 1998       Ar-39 Ar age data on these samples from . Chemical types named alphabetically with respect to drill site number. FeO * , total Fe oxide. VOLUME 39 NUMBER 3 MARCH 1998

Fig. 3.
Buon Ma Thuot high-Ti alkali basalts precede lower-Ti range 90·1-95·3. These data suggest that Indochinese lithospheric mantle is variably refractory and similar to variants in the sequence. The existence of both coupled and decoupled variation between TiO 2 and SiO 2 may Phanerozoic subcontinental mantle elsewhere (e.g. Song Hawkesworth et al., 1990). result from the interplay of variable melt fraction, source fertility and P(H 2 O) in generating primitive melts. We suggest therefore that parental magmas reflect a spectrum between large melt fractions, low pressures and a (low-Ti, Fe) refractory source (on the one hand), and small Trace elements and isotopes melt fractions, high pressures and a (high-Ti, Fe), fertile source (on the other).
Crust addition, whether resulting from wallrock re-tasomatism, rather than wallrock contamination of ascending melt. action or mantle metasomatism, is thus significant at Buon Ma Thuot and Xuan Loc, and is further supported Ti/Zr typically decreases from early to late series eruptions (except in Buon Ma Thuot) and is generally by the presence of isotopic EM2 (Hoang et al., 1996).
However, two points should be noted. The presence matched by increasing ratios of more-to less-incompatible elements (e.g. Fig. 6a-c). In addition to source hetero-of systematic isotopic differences between basalts from distinct lithospheric sectors suggests that the effects of geneities these arrays reflect a decrease in melt fraction from tholeiite to alkali basalt (indicated by dashed parallel enriched lithospheric mantle predominate over wallrock reaction. Second, at most centres there is a secular lines with decreasing Ti/Zr) and contrast with divergent trends towards higher Ti/Zr, reflecting the addition of change from EM2-rich early series to EM1-rich late series compositions. Hoang et al. (1996) proposed that, crust components (CC). For example, negative covariation bands of Rb/Zr and Ba/Zr vs Ti/Zr include irrespective of whether EM2 is incorporated by rising melt or derives from lithospheric mantle metasomes, members of both early and late series at each centre ( Fig. 6a and b), whereas sharp increases in Rb/Zr and EM1-like components must have been present in the convecting asthenosphere and were perhaps delaminated Ba/Zr with increasing or near-constant Ti/Zr (shown by DT-I, BMT-I, BMT-II and XL-II series) ( Fig. 6a and from the Sino-Korean craton. As samples of sub-Indochina lithospheric mantle, the b) are consistent with the addition of crustal material. In contrast, again with the exception of Buon Ma Thuot, spinel lherzolite xenoliths show low Rb/Sr (0·001-0·008) and high Sm/Nd (0·3-0·65) values, with 87 Sr/ 86 Sr and Sr/Zr ratios show little overall variation with Ti/Zr and VOLUME 39 NUMBER 3 MARCH 1998  Table 1). These are labelled alphabetically with drill site numbers in parentheses (shown in Fig. 1b) and age ranges based where possible on Ar-Ar data ( . 'Early series' types are shown by continuous outlines, 'late series' by dashed outlines. Tholeiites are shown by fine lines and alkali basalts and basanites by bold lines. 143 Nd/ 144 Nd ratios corresponding to N-MORB ± EM2 averages of chemical types based on samples with >7·5 ( Hoang et al., 1998). Although these differ from the MgO wt %. The rationale for this procedure (following EM1-rich host alkali basalts, the (decoupled) negative Scarrow & Cox, 1995) was that: (1) olivine (Fo 89-83 ) is covariance of Sm/Nd and mg-number suggests cryptic the only significant phenocryst phase in basalts with EM2-rich metasomatism of the refractory lithospheric >7·5 MgO wt % which shows negligible effects of mantle (Fig. 8) (see Frey & Green, 1974). clinopyroxene and plagioclase fractionation, (2) parent magmas are almost certainly more magnesian than the most Mg-rich erupted lavas, (3) liquids with >13 wt % MgO are likely to have equilibrated with Fo 85-90 (Roeder & Emslie, 1970), and (4) experimental peridotite To interpret basaltic melt segregation conditions, we melts fall in the range 12-17 wt % MgO. These have considered the compositions of primitive, mantlecriteria apply to both anhydrous and hydrous conditions equilibrated, melts to be a simple function of lithospheric (Kushiro, 1990;Hirose & Kushiro, 1993;Hirose & stretching ( ) and asthenosphere potential temperature Kawamoto, 1995). The effect of adding olivine is (T p ), in turn reflecting melt segregation pressure, melt relatively trivial for most oxides but provides a basis fraction and bulk source composition (after McKenzie & for establishing the variation of primitive melts, and Bickle, 1988;Latin & White, 1990).

PRIMITIVE MELT COMPOSITIONS
by comparison with published experimental results, estimating melt segregation pressures and temperatures.

Variation of Mg-15 normalized eruptive
Olivine ( Hoang et al. (1998) (1) (olivine tholeiite, alkali basalt or basanite) end-members, replicating the overall pattern of regional variation. Third, (Al 2 O 3 /FeO) p ·(1 -K) (2) at several centres (e.g. Dalat, Phuoc Long, Pleiku, Xuan Loc and Re Island) the low-and high-FeO * suites corwhere p denotes the primary melt composition, i denotes respond, respectively, to early and late eruptive series. the initial (i.e. erupted) liquid composition, and K is the Whereas the covariation of Si-15 with Fe-15 is negative, distribution coefficient K Mg/Fe" ol/liq (from Pearce, 1978). that of Ti-15 and P-15 is mostly positive, and K-15 Figure 9a-d shows Mg-15-normalized variation from appears to vary independently of Fe-15 (Fig. 9a-d). which three distinctive features are evident. First, discrete Covariance of Si-15 and K-15 (not shown) is mostly 'suites' of chemical type averages are recognized in terms negative within magmatic suites, except for the Xuan Loc and Pleiku upper series basalts, which show a positive of FeO * , SiO 2 , TiO 2 , K 2 O and P 2 O 5 at most of the VOLUME 39 NUMBER 3 MARCH 1998 deflection consistent with isotopic and trace element in mantle fertility as reflected by the bulk source mgnumber, and contents of CaO, Al 2 O 3 and Na 2 O. These indications for wallrock reaction (Hoang et al., 1996).
contrasting vectors are analogous to those recognized in A simple model was developed as a basis for petrooceanic (Wilkinson, 1991;Nicolson & Latin, 1992) and genetic interpretation in terms of relevant published continental (Turner & Hawkesworth, 1995) intraplate experimental data (in this paper), and Sr, Nd and Pb basalts. isotopic variation (Hoang et al., 1996). The within-suite variation of SiO 2 saturation and incompatible element contents represents a spectrum of melts generated from a discrete volume of decompressing mantle, reflecting Source fertility and melt segregation minor isotopic and trace element source heterogeneities, conditions and the combined effects of variable melt fraction, segregation pressure and temperature (Langmuir et al., Establishing pressure and temperature conditions of prim-1992;Scarrow & Cox, 1995). In contrast, the betweenitive melt segregation, although problematic, can be approached in at least two ways. The first involves suite differences in Fe-15 were attributed to differences (e.g. Albarède, 1992;Scarrow & Cox, 1995;Turner & Hawkesworth, 1995). A complementary approach involves mathematical inversion of erupted melt compositions with respect to an assumed mantle source and known subsolidus assemblages determined experimentally as a function of pressure and temperature, assuming fractional melting of a polybaric melt column comparing natural melt compositions with those produced experimentally at known pressures, temperatures and P(H 2 O), from a source of known composition, assuming experimental melt compositions are realistic, that batch equilibrium melting is a valid analogue for mantle melting (even if fractional melting is more realistic) and that an independent measure of P(H 2 O) is available VOLUME 39 NUMBER 3 MARCH 1998 Hirose & Kushiro, 1993;Baker & Stolper, 1994) whose compositions correlate with equilibration pressures and temperatures. Melts produced under H 2 O-undersaturated conditions show similar compositional variation with respect to pressure and temperature, although their oxide covariation slopes differ significantly and SiO 2 contents are consistently higher than those of the anhydrous melts (Kushiro, 1990;Hirose & Kawamoto, 1995).

Pressure and temperature calibrations
Covariation of FeO * in melts with both temperature and pressure has been recognized in experimental studies (e.g. Jaques & Green, 1980;Falloon et al., 1988) although simple temperature-dependent functions have been elusive as a result of problems such as Fe loss from experimental capsules. For a given bulk composition the temperature dependence of Fe is pronounced at low pressures (0·5-1 GPa) ( Jaques & Green, 1980) but less so at higher pressures where garnet appears at the subsolidus. Scarrow & Cox (1995) discussed this relationship with respect to Hirose & Kushiro's (1993) data and projected isopleths for Fe and Mg in pressuretemperature space for melts equilibrated with fertile lherzolite HK-66. For melts of restricted MgO range (e.g. 14-16 wt %) over a pressure range of 1-3 GPa (implying decreasing melt fraction with increasing pressure) the Fe-temperature relationship is quasi-linear for both fertile and refractory bulk compositions ( Fig. 11a; see below). Equivalent melt fractions formed in hydrous 'sandwich' and 'diamond aggregate' experiments also show this relationship over the pressure range 1·2-2·5 GPa, although temperatures are 50-100°C less for melts of equivalent FeO * content (Fig. 11a). The relationship between SiO 2 and pressure in experimental melts has been quantified by Albarède (1992) and Scarrow & Cox (1995), and is shown in Fig. 11b (see below) for anhydrous and hydrous conditions. Compositions of experimental melts from fertile peridotite (HK-66) were therefore calibrated for pressure and temperature according to the reported experimental conditions of Kushiro (1990) and Fig. 6. Plots of (a) Rb/Zr, (b) Ba/Zr and (c) Sr/Zr vs Ti/Zr ratios for Hirose & Kushiro (1993). It seems reasonable to apply samples from Dalat (DT), Phuoc Long (PH), Pleiku (PL), Buon Ma these relationships to the natural melt compositions, Thuot (BMT), Xuan Loc (XL) and Re Island (RE), distinguishing early assuming the latter were derived from sources of similar (I) and late (II) eruptive series. fertility, segregated over a similar pressure interval, and represent a similar range of melt fractions to those of the (McKenzie & O'Nions, 1991;Watson & McKenzie, experiments. 1991).
We adopted the first approach and used experimental melts produced in 'sandwich' capsules (avoiding reaction

Comparison of natural and experimental
between experimental charge and capsule) (e.g. Stolper, melts Kushiro, 1990) and using 'diamond aggregate' methods (allowing equilibrium between melt and solid In general, the Vietnamese melts are closer in composition to FeO * -rich melts generated from HK-66 than to those while avoiding problems of quench crystallization) (e.g.  Hoang et al. (1996) for details]. from refractory KBL-1 (Fig. 10). However, the inverse of~2-3°C/km (Table 1). The prominence of quartz tholeiite, shallow Fe-15 vs Si-15 slope, and high juvenile covariance of Fe-15 and Si-15 in the natural melts is less steep than that of anhydrous HK-66 melts and resembles H 2 O contents in Vietnamese basalts, are not typical of oceanic and continental intraplate basalts and strongly those of melts equilibrated with fertile peridotite under hydrous conditions (Fig. 10). According to the anhydrous suggest a hydrous source. Hydrous melt segregation depths of between~45 and 110 km correspond closely calibrations, melt segregation pressures range from nearly 4 GPa and temperatures of~1470°C (e.g. alkali basalts to estimates for basalts in Iceland (Nicolson & Latin, 1992), Skye (Scotland) (Scarrow & Cox, 1995), and the from Xuan Loc) to <0·5 GPa and~1400°C (quartz tholeiites from Dalat and Phuoc Long), representing a East African Rift (Latin et al., 1993). Mantle adiabats interpolated from H 2 O-undersaturated melt segregation pressure-temperature slope of~0·75°C/km. In contrast, segregation conditions based on the H 2 O-undersaturated conditions may thus offer useful insights concerning ambient mantle T p , topology of the MBL-TBL interface, calibrations range from <3·5 GPa pressure and temperatures of~1450°C to~1·5 GPa and and the relationship between convecting asthenosphere and extended, conductively cooled lithosphere. 1350-1400°C, representing a pressure-temperature slope  (Hirose & Kushiro, 1993). H 2 O-undersaturated experimental melts were equilibrated with 'sandwiches' of olivine, orthopyroxene and clinopyroxene from HK-66 (Kushiro, 1990). the peridotite solidus (Gallagher & Hawkesworth, 1992)

MELTING DYNAMICS AND MANTLE
or mobilization of relict melt fractions ('mafic com-

BOUNDARY LAYER MODEL
ponents') (Harry & Leeman, 1995). Harry & Leeman A petrogenetic model for Vietnamese basalts needs to (1995) have contended that H 2 O does not play a sigexplain the relatively high volume basalt plateaux ocnificant role in generating magma, as they believe that cupying extensional pull-apart nodes, the common produring initial stages of extension volatile fractions are gression from tholeiite-dominated early series, tapping quickly exhausted by early-formed melts. They suggested, refractory lithosphere-like sources, to later alkali basaltrather, that significant melt fractions are produced at dominated series, tapping relatively fertile, asthenospheric the solidi of 'mafic components' present in the lower sources, a T p of~1440°C and adiabatic gradient of lithosphere, depending on the amount and duration of >2°C/km, and LILE-and H 2 O-rich asthenosphere. In lithospheric extension. However, despite the presence of the absence of data for lithosphere thickness, we depyroxenite (etc.) in peridotite restites, geothermal graveloped a simple boundary layer model for Indochina dients based on xenolith thermobarometry (e.g. Ionov et that is consistent with the petrologic and geochemical al., 1998) differ significantly from adiabats of the type data and general considerations of rheology and thermal indicated by Vietnamese melt segregation conditions, state.
suggesting non-ductile MBL is not decompressed to the The progression from FeO * -poor to FeO * -rich (OIBextent needed to produce magma. In contrast, conlike) basalts in continental settings is commonly insideration of mantle dehydration reactions supports a terpreted to reflect a change from lithospheric to askey role for H 2 O in melting refractory lithospheric mantle. thenospheric mantle sources (e.g. Perry et al., 1987; Water is held in amphibole or mica at most lithospheric DePaolo, 1988; Kempton et al., 1991;Gallagher & pressures (Lambert & Wyllie, 1970;Gallagher & Hawke-Hawkesworth, 1992;Turner & Hawkesworth, 1995). sworth, 1992. Beyond the stabilities of these phases, However, the problem of melting relatively non-ductile, small melt fractions saturated with H 2 O and fusible mafic refractory MBL has been raised as an objection to this components will be able to form if temperatures exceed the H 2 O-saturated solidus. However, even where such interpretation and led to models invoking H 2 O to lower VOLUME 39 NUMBER 3 MARCH 1998 As a framework for reconciling petrologic, thermal and geodynamic constraints, we adopt a lithospheric MBL whose mantle component is isotopically depleted and refractory, and rigidly attached to the crust (McKenzie & Bickle, 1988;Hawkesworth et al., 1990). Below this, the TBL is hypothetically characterized by attenuated shear-wave velocities and taken to mark the rheologic transition from dominantly convective to conductive heat transfer. It also marks the closest approach of the inflected thermal gradient to the peridotite solidus, allowing the highest potential for melting, and is assumed to be weak and isotopically enriched (Anderson, 1995). The asthenosphere is generally held to be isotopically depleted, unless lithospheric components have been added-either recycled via plumes or delaminated directly from above (e.g. Storey et al., 1989;Hart et al., 1992)-and residual to the extraction of primordial crust (Hofmann, 1988).
Hydrous phase breakdown would also produce changes in mantle rheology at similar depths (with or without melting) such that the TBL also marks the intersection of geotherm and dehydration curves (Lambert & Wyllie, 1970;Anderson, 1995). It is thus helpful to look at the combined effects of T p and H 2 O on peridotite solidi, which for a particular thermal gradient determine the depth and extent of melting. These relationships are shown in Fig. 12 along with the mantle adiabat interpolated for Vietnamese melts. The depth of phlogopite dehydration ranges from~3·8 GPa (~110 km depth) at normal T p to <3 GPa (<90 km depth) at T p >1400°C (Fig. 12). The interpolated Indochina T p (~1440°C) is significantly higher than normal for subcontinental asthenosphere (e.g. McKenzie & Bickle, 1988;Wilson, 1993) suggesting an ambient MBL thickness of~80 km (~2·6 GPa) (Fig. 12). A conductive type 1 geotherm is believed to be typical for small or negligible lithospheric stretching factors ( ) and is consistent with xenolith thermobarometric data from analogous settings (e.g. Ionov et al., 1998). It also agrees with thermal gradients Fig. 11. Plots of (a) FeO * (wt %) vs temperature (°C), and (b) SiO 2 computed as a function of uniform stretching (Latin & (wt %) vs pressure (GPa) for experimental partial melts produced under hydrous and anhydrous conditions. Open circles, anhydrous melts of White, 1990), which suggest type 2 geotherms (Fig. 12) fertile peridotite HK-66 (Hirose & Kushiro, 1993); crossed circles, H 2 Oresult at values of 2-3 (Fig. 13). undersaturated melts equilibrated with HK-66 'sandwiches' (Kushiro, An important implication of the model is that uniform 1990). For H 2 O-undersaturated experimental results (e.g. Kushiro, 1990) only runs with <5 wt % added H 2 O and K D(Fe/Mg) OL-MELT 3 [0·27 stretching causes changes in MBL bulk composition as were used. The regressions used in the melt segregation P-T calculations a result of its partial conversion to TBL. This effect is are shown (see text). illustrated in Fig. 12, where stretching (producing type 2 geotherms) causes upward migration of the rheologic boundary, separating solid from partially molten regions, melts remain in situ (i.e. in equilibrium) the extent of relative to the pre-stretching (type 1 geotherm) boundary melting necessarily remains small, buffered by the small which was both rheologic and compositional. In other amount of H 2 O present. Similarly, fractional melting words, the upward migration of dehydration depth reresulting from decompression could not advance beyond sulting from a local increase in converts the lowermost initial increments and pooled fractions would be enriched lithospheric mantle into rheologically weak TBL. Inin SiO 2 , LILE, HFSE and LREE, whereas in contrast, corporation of refractory peridotite into the ductile region significant (magma-producing) equilibrium melts could only form close to dry or H 2 O-undersaturated solidi.
produces a 'macro'-heterogeneous column, which, with  (Hirose & Kushiro, 1993), stability fields of phlogopite and amphibole in peridotite (Modreski & Boettcher, 1973;Milhollen et al., 1974) continued stretching and decompression, will yield ad-& Hai, 1991) and high heat flow (Duchkov et al., 1992;Uyeda & Nagao, 1994) in south-central Vietnam. vanced partial melts. According to Fig. 12, the column ranges from garnet lherzolite to spinel lherzolite, and presumably undergoes continuous subsolidus reaction during diapiric uprise. Although simplified, this model CONCLUSIONS avoids the requirement for magma production within the MBL, and reconciles the apparent conflict between low (1) Cenozoic basalt plateaux in southern and central melt segregation pressures (<1·5 GPa) and MBL Vietnam appeared over a total area of~23 000 km 2 as thicknesses >~80 km. part of a widespread regional volcanic episode. Eruptions We therefore propose that ambient lithospheric thickat discrete centres appear to have involved at least two nesses of 80-100 km are reasonable at T p of~1440°C, episodes separated by thick palaeosols, referred to as and that thinning at transtensional 'nodes' leads to the 'early' and 'late' series. The bi-episodal pattern is repenetration by and advanced polybaric melting of lowcognized at the Dalat, Phuoc Long, Pleiku, Buon Ma viscosity TBL and asthenosphere. Decompression melting Thuot, Xuan Loc and Re Island centres, and probably would yield plateau basalt sequences of low-pressure, other offshore localities, although at Buon Ma Thuot the large-fraction tholeiites and high-pressure, low-fraction compositional trend is 'inverted'. alkali basalts and basanites. This conclusion is consistent (2) K-Ar and Ar-Ar age data indicate activity occurred with element inversions on the Vietnamese basalt data over the following intervals: Dalat (17·6-7·9 Ma), Phuoc conducted by D. McKenzie at Cambridge University Long (<8-3·4 Ma), Buon Ma Thuot (5·8-1·67 Ma), which indicate a T p of~1450°C and partial melt column Pleiku (4·3-0·8 Ma), Xuan Loc (0·83-0·44 Ma) and Île extending between garnet lherzolite and spinel lherzolite des Cendres (0·8-0 Ma), and reflects clockwise rotation facies. Our conclusion that a major thermal anomaly of transtensional fractures. Palaeomagnetic data indicate exists beneath Indochina is supported by geomorphologic little or no tectonic rotation since the India-Asia collision, although this can be reconciled with the extrusion model evidence for~600 m uplift since the late Neogene (Bao VOLUME 39 NUMBER 3 MARCH 1998 Fig. 13. Adiabatic upwelling as a result of differential stretching of a convective geotherm generated from a 100 km thick MBL for an interior potential temperature 1440°C (after Latin & White, 1990). Geotherms for respective stretching factor values ( ) (continuous lines) and the locus of melt fraction (F ) of 0·25 (dot-dashed lines) are shown together with the dry peridotite solidus (bold dashed line) and typical H 2 O-undersaturated solidus (fine dashed line) [adapted from McKenzie & Bickle (1988) and Latin & White (1990)]. Xenolith P-T equilibration estimates reflecting conductive thermal gradients are expected to correspond to values between 1 and~1·5 (type 1 geotherm in Fig. 12). Melt segregation P-T estimates based on H 2 O-undersaturated experiments (Fig. 11) indicate adiabats consistent with values between 2 and 3 (type 2 geotherm in Fig. 12) for T p = 1440°C. No melting would be expected for this range at T p = 1280°C (see Latin et al., 1993).
if Indochina behaved as a non-rigid plate. The basalts result from collision-extruded asthenosphere rather than a deep plume. probably reflect stretching associated with the change (5) Phlogopite stability provides a possible model for from left-to right-lateral motion on the Ailao Shan-Red the base of the lithosphere MBL. Given the P-T shape River shear zone.