Journal of Petrology Advance Access originally published online on April 22, 2005
Journal of Petrology 2005 46(8):1603-1643; doi:10.1093/petrology/egi028
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
Source of the Quaternary Alkalic Basalts, Picrites and Basanites of the Potrillo Volcanic Field, New Mexico, USA: Lithosphere or Convecting Mantle?
1 DEPARTMENT OF GEOLOGICAL SCIENCES, UNIVERSITY OF DURHAM, SOUTH ROAD, DURHAM DH1 3LE, UK
2 DIVISION OF GEOLOGICAL AND PLANETARY SCIENCES, CALIFORNIA INSTITUTE OF TECHNOLOGY, MC 170–25, PASADENA, CA 91125, USA
3 DEPARTMENT OF GEOLOGY, McMASTER UNIVERSITY, 1280 MAIN STREET WEST, HAMILTON, ONTARIO, CANADA L8S 4M1
4 23 NIGEL AVENUE, NORTHFIELD, BIRMINGHAM B31 1LL, UK
5 BRITISH ANTARCTIC SURVEY, MADINGLEY ROAD, CAMBRIDGE CB3 0ET, UK
6 DEPARTMENT OF EARTH SCIENCES, UNIVERSITY OF CAMBRIDGE, DOWNING STREET, CAMBRIDGE CB2 3EQ, UK
RECEIVED SEPTEMBER 15, 2003; ACCEPTED FEBRUARY 15, 2005
 |
ABSTRACT |
|---|
TOP
ABSTRACT
INTRODUCTIONThe <80 ka basalts–basanites of the Potrillo Volcanic Field (PVF) form scattered scoria cones, lava flows and maars adjacent to the New Mexico–Mexico border. MgO ranges up to 12·5%; lavas with MgO < 10·7% have fractionated both olivine and clinopyroxene. Cumulate fragments are common in the lavas, as are subhedral megacrysts of aluminous clinopyroxene (with pleonaste inclusions) and kaersutitic amphibole. REE modelling indicates that these megacrysts could be in equilibrium with the PVF melts at
1·6–1·7 GPa pressure. The lavas fall into two geochemical groups: the Main Series (85% of lavas) have major- and trace-element abundances and ratios closely resembling those of worldwide ocean-island alkali basalts and basanites (OIB); the Low-K Series (15%) differ principally by having relatively low K2O and Rb contents. Otherwise, they are chemically indistinguishable from the Main Series lavas. Sr- and Nd-isotopic ratios in the two series are identical and vary by scarcely more than analytical error, averaging 87Sr/86Sr = 0·70308 (SD = 0·00004) and 143Nd/144Nd = 0·512952 (SD=0·000025). Such compositions would be expected if both series originated from the same mantle source, with Low-K melts generated when amphibole remained in the residuum. Three PVF lavas have very low Os contents (<14 ppt) and appear to have become contaminated by crustal Os. One Main Series picrite has 209 ppt Os and has a 
Os value of +13·6, typical for OIB. This contrasts with published 187Os/188Os ratios for Kilbourne Hole peridotite mantle xenoliths, which give mostly negative Os values and show that Proterozoic lithospheric mantle forms a thick Mechanical Boundary Layer (MBL) that extends to 70 km depth beneath the PVF area. The calculated mean primary magma, in equilibrium with Fo89, has Na2O and FeO contents that give a lherzolite decompression melting trajectory from 2·8 GPa (95 km depth) to 2·2 GPa (70 km depth). Inverse modelling of REE abundances in Main Series Mg-rich lavas is successful for a model invoking decompression melting of convecting sub-lithospheric lherzolite mantle (
Nd = 6·4; Tp 1400°C) between 90 and 70 km. Nevertheless, such a one-stage model cannot account for the genesis of the Low-K Series because amphibole would not be stable within convecting mantle at Tf 1400°C. These magmas can only be accommodated by a three-stage model that envisages a Thermal Boundary Layer (TBL) freezing conductively onto the 70 km base of the Proterozoic MBL during the 20 Myr tectonomagmatic quiescence before PVF eruptions. As it grew, this was veined by hydrous small-fraction melts from below. The geologically recent arrival of hotter-than-ambient (Tp 1400°C) convecting mantle beneath the Potrillo area re-melted the TBL and caused the magmatism. KEY WORDS: western USA; picrites; Sr–Nd–Os isotopes; petrogenetic modelling; thermal boundary layer
|
INTRODUCTION |
|---|
Small-volume basic alkalic magmatism characterizes many current and geologically recent areas of continental extension. The geochemistry of the most primitive magmatic rocks can be used to model mantle partial-melting conditions and, in geologically young examples, these results may be compared with such features as: (1) geophysical data bearing on lithospheric thickness and the state of the subjacent convecting mantle; and (2) the petrology and geochemistry of the local lithospheric mantle, in cases where this is sampled by xenolith suites in the magmatic rocks. Hence, the relative contributions of melts from lithospheric and convecting mantle can be assessed.
Wang et al. (2002)
focused on the major-element compositions of <10 Ma basic lavas, many of them alkaline, from throughout the Basin and Range province and adjacent regions of the southwest USA. They used techniques originally developed for understanding MORB petrogenesis to deduce that the sources of these continental magmas were essentially entirely within the convecting mantle, beneath a lithospheric lid of varying thickness. Wang et al. (2002) showed that their modelled depths to the base of the lithosphere in western USA agreed well with geophysical data for the region.
In contrast, other authors have noted geochemical features of worldwide alkaline basic lavas that have seemed to them to be incompatible with an origin simply by partial melting of convecting mantle. They have developed alternative magma genesis models in which the lithospheric mantle also plays a role. For instance, invocation of a geologically young metasomatic event affecting older lithospheric mantle, which then contributes to the melts, is the most widely favoured method of explaining the characteristic elemental and isotopic geochemistry of such magmatism occurring on both continental and old oceanic lithosphere (e.g. Wilson & Downes, 1991; McKenzie & O'Nions, 1995; Wilson et al., 1995; Class & Goldstein, 1997; Späth et al., 2001; Yang et al., 2003).
This study concerns the <80 ka Potrillo Volcanic Field (PVF), southern New Mexico, which occurs in the region where the Rio Grande rift ceases to be a distinct physiographic feature and joins the regional Basin and Range extension zone of the southwestern USA and northern Mexico. The chemical and Sr–Nd–Os isotopic compositions of the lavas are used to attempt to define the depth and conditions of their genesis, and of the subsequent processes that affected them. Such an approach is assisted in this area by the availability of extensive published data on both the regional geophysics and the petrology and geochemistry of suites of lithospheric mantle xenoliths sampled by some of the lavas.
|
POTRILLO VOLCANIC FIELD |
|---|
The PVF lies in the desert that extends for
60 km west of the segment of the Rio Grande rift between El Paso and Las Cruces. It encroaches slightly into Mexico but falls mostly within New Mexico. The magmatism is entirely basaltic (s.l.), forming scattered scoria cones (Fig. 1). Up to several short lava flows originate from each cone and there are a few locations (e.g. Aden Crater) that appear to be centres of more persistent magmatism. The region is well known for the spectacular maar and suite of mantle xenoliths at Kilbourne Hole (Padovani & Reid, 1989). Most of the scoria cones lie along the broad approximately N–S trending ridge of the West Potrillo Mountains (Fig. 1). Another small line of cones forms the Black Hills, between Kilbourne Hole and the Rio Grande. The ages of the lavas, measured by 3He surface exposure dating, range from 80 to 17 ka (Anthony & Poths, 1992).
|
The varied basement rock types beneath the Potrillo volcanics, and locally protruding through them, range from Proterozoic granites through a Phanerozoic sedimentary succession to basalt–andesite volcanics of the southern fringes of the Sierra de las Uvas volcanic field—an
30 Ma igneous succession that lies mostly to the north of Fig. 1 (Seager et al., 1984; unpublished work by Thompson and Leat). The Potrillo area is generally classified as part of the southern Rio Grande rift and shows the Tertiary tectonic evolution of that structure. Extension took place in an intense 30–20 Ma phase, involving low-angle normal faults, and a less intense <10 Ma phase, involving high-angle normal faults (Olsen et al., 1987; Keller et al., 1990). Mack & Seager (1995) argued that the Quaternary magmatism reached the surface via a transfer zone linking two adjacent N–S-trending, long-lived, extensional structures—the West Robledo and Camel Mountain faults. In order to decide whether the PVF magmas originated within the subcontinental lithospheric mantle, the underlying convecting mantle or both, it is first necessary to summarize what is known about the lithospheric mantle in this area.
Previous geophysical estimates of lithosphere thickness
Previous attempts to estimate lithospheric thickness beneath the southern Rio Grande rift have been made by modelling seismic, heat flow and gravity data, and based on thermo-barometric studies of mantle xenoliths in the PVF lavas, notably at Kilbourne Hole. Sinno et al. (1986) established seismic refraction profiles crossing the Potrillo area, both E–W (reversed) and N–S (unreversed). They interpreted these data as showing that the Rio Grande rift at 32–33°N is a distinct structure, separating the Great Plains and Basin and Range tectonic provinces. The low value of Pn (P-wave velocity for the ray path reflected from immediately below the Moho) beneath the rift (7·7 km/s) led them to suggest that ‘the asthenosphere could be in direct contact with the crust’ (Sinno et al., 1986, p. 6154), i.e. that no lithospheric mantle exists beneath the Potrillo area.
Morgan et al. (1986) suggested that the very high heat flow values across the axis of the southern Rio Grande rift ‘indicate a convective heat transfer regime in the upper mantle’ (Morgan et al., 1986, p. 6269). Keller et al. (1990) refined this view by suggesting that below 35 km depth in the southern Rio Grande rift, ‘the geotherm appears to become non-conductive and is probably controlled by convection of basaltic magmas from depths of 50 to 70 km up into the lower crust’ (Keller et al., 1990, p. 31). It is not clear whether they envisaged this process as melt generation within upwelling convecting mantle (asthenosphere) or melt migration advecting heat from below 50 km into overlying lithosphere. Cordell et al. (1991) derived a map of total lithospheric thickness beneath the southern Rio Grande rift by inversion of gravity anomaly data and concluded that beneath the Potrillo area ‘the lithosphere may be as thin as 40 km’ (Cordell et al., 1991, p. 6566).
Achauer & Masson (2002) summarized previous extensive teleseismic travel-time delay tomographic studies of the rift north of the Potrillo area. They reached two main conclusions: (1) Below 185 km depth, there is an extensive zone of seismically relatively slow mantle (interpreted as relatively hot). This is centred on the physiographic rift but extends for 100–200 km on either side and widens downwards. (2) Between 35 km (Moho) and 185 km there are several areas along the physiographic rift axis and its flanks where seismically slow (hot) mantle extends up to the base of the crust. They interpreted the laterally extensive, deep, hot mantle as sourced by a convective plume and suggested that asthenospheric diapirism may have carried plume mantle up to near-Moho depths locally. They emphasized that in these seismic features, the Rio Grande rift resembles the East African rift and contrasts with the Rhinegraben and Baikal rifts, which both lack widespread deep seismically slow mantle. The zone of upwelling convecting mantle appears to be slightly oblique to the trend of the rift axis, such that the Potrillo area overlies the eastern flank rather than the crest of the present-day upwarp of low-velocity mantle. The geophysically defined lithospheric thickness beneath Kilbourne Hole is 80 km.
Petrological features of the lithospheric mantle beneath Kilbourne Hole
The extensive and well-studied suite of mantle xenoliths that occur at Kilbourne Hole (Fig. 1) gives an important perspective on the nature of the lithospheric mantle beneath the PVF (Padovani & Reid, 1989). Most of the mafic–ultramafic xenoliths are medium- to coarse-grained so-called Group I peridotites, with protogranular textures and green Cr diopside. Thin bands of pyroxenite cutting this peridotite are rare. About 10% of the xenoliths are more Fe-rich, finer-grained, tabular-textured so-called Group II peridotites, frequently veined by pyroxenite and wehrlite (Irving, 1980; Padovani & Reid, 1989; Bussod & Williams, 1991). All of these authors stress that hydrous minerals, such as amphibole and mica, are rare in the Kilbourne Hole xenolith suite.
Bussod & Williams (1991) provided crucial evidence by showing that calculated P–T equilibration conditions for an extensive suite of Kilbourne Hole peridotite xenoliths defined a continuous array from 850°C at 35 km depth to 1050°C at 67 km depth. This is far below the temperature of 1320°C in this depth range attained by mantle convecting at a potential temperature (Tp) of 1300°C (McKenzie & Bickle, 1988). Potential temperatures significantly lower than this value are not usually suggested for convecting mantle. Thus, the Kilbourne Hole peridotite xenoliths appear to provide tangible evidence for stable (Mechanical Boundary Layer) lithospheric mantle extending to at least 70 km beneath the area. Furthermore, Bussod & Williams (1991, fig. 1) noted that within the set of 18 xenoliths that they studied, the depth distribution of Group I and II peridotites was anything but random; the fine-grained tabular-textured Group II peridotites and their pyroxenite-veined composite associates are confined to the shallower (35–55 km) part of the lithospheric mantle, whilst Group I coarse peridotites are confined to the 50–70 km depth range and are the only type sampled below 55 km.
|
PETROGRAPHY AND MINERALOGY OF THE PVF LAVAS |
|---|
Sampling was designed to give regional coverage of the volcanics; sample localities are given in Electronic Appendix A (which may be downloaded from http://www.petrology.oupjournals.org). Either lava flows or large spindle bombs provided suitable material. In addition, small sets of the megacrysts and coarsely crystalline ultramafic inclusions that occur at many places in the field were collected; Kilbourne Hole is the best-known location for these suites (Padovani & Reid, 1989
). A few of the localities at the northern end of the West Potrillo Mountains (Fig. 1) provided samples of the older Uvas volcanics (Seager et al., 1984; Olsen et al., 1987). The locality and geochemical data for West Potrillo Mountains Uvas lavas sampled during this study are included in Electronic Appendices A and B.
All the lavas are porphyritic, with up to 20% (mostly <10%) of euhedral to subhedral phenocrysts, rarely exceeding 2 mm, in fine-grained holocrystalline groundmasses of olivine, clinopyroxene, plagioclase and opaques, together with sparse patches of brown glass. The phenocrysts are of olivine alone, olivine plus clinopyroxene, or olivine plus clinopyroxene and plagioclase (occasionally). One lava (8104) contains sparse opaque-granule-rich pseudomorphs, probably after amphibole, plus trace amounts of <1 mm equant opaque and apatite microphenocrysts.
Apart from sample 8104, no hydrous minerals were observed in thin sections as either phenocrysts or groundmass phases. Nevertheless, large amphibole megacrysts (<5 cm) occur quite commonly throughout the West Potrillo Mountains area, mostly in scoria but also as rare fragments within lavas. Other abundant megacrysts are clinopyroxene and feldspar; Ortiz (1980) mentioned rare spinel. Crystals of each mineral, found loose in scoria, are euhedral but characterized by polished and pitted surfaces.
Coarse-grained ultramafic inclusions, up to 10 cm in size, are locally common throughout the West Potrillo Mountains area. Spinel lherzolites are bright green, with a coarse equigranular texture, curved grain boundaries, olivine with deformation lamellae and brown spinel. They predominate at Kilbourne Hole (Roden et al., 1988; Padovani & Reid, 1989; Bussod & Williams, 1991; Burton et al., 1999) but are a minor part of the West Potrillo Mountains suites. One West Potrillo Mountains green xenolith is a granular olivine websterite, with some brown spinel and coarse orthopyroxene exsolution lamellae in its clinopyroxenes. All the remaining ultramafic inclusions form a series between clinopyroxenite (± olivine) and hornblendite. Lherzolite-like granular textures characterize the hornblendites but members of the clinopyroxenite to amphibole clinopyroxenite suite have igneous cumulate textures, with euhedral to subhedral minerals surrounded by a network of brown interstitial glass. Some clinopyroxenites lack olivine (except as tiny euhedra accompanying plagioclase in the glass) and amphibole (except as flecks along clinopyroxene cleavage planes). As amphibole increases in abundance, it forms poikilocrysts and eventually equant crystals. One olivine clinopyroxenite with a few percent of brown amphibole shows small euhedra of this phase within pools of interstitial glass. Spinel forms rounded <0·5 mm grains in members of the clinopyroxenite–hornblendite suite. In fresh samples, the spinel is green in the clinopyroxenites, including some with poikilitic amphibole, and opaque in the hornblendites. Many inclusions are variably oxidized and these have opaque spinels. Plagioclase is common in the interstitial glass patches and also occurs as a minor part of the main mineral assemblage in two of the hornblende clinopyroxenites.
The olivine–clinopyroxene-phyric lavas commonly contain rounded-fritted olivine, pyroxene and feldspar macrocrysts, in addition to their euhedral phenocrysts. Electron microprobe study of the olivine and pyroxene macrocrysts (described below) shows that neither is from the same spinel lherzolite source as the mantle xenoliths. The 11 lavas used for electron microprobe studies have MgO contents between 9·1 and 12·5%. Analyses were made using the Geoscan (mostly) and Cameca SX100 instruments at Manchester University (Table 1). The analyses concentrated on phenocrysts, macrocrysts, megacrysts and other minerals included in them, in order to shed light on the post-genesis history of the lavas. The Geoscan settings were as follows: accelerating potential 15 kV; incident beam current 6 nA; take-off angle 75°. Standards used were: Si, Ca, wollastonite; Ti, rutile (synthetic); Al, corundum (synthetic); Fe, fayalite (synthetic); Mn, tephroite (synthetic); Mg, periclase (synthetic); Na, jadeite; K, orthoclase; Cr, Cr2O3 (synthetic); Ni, NiO (synthetic). The analysis system was a PGT (Princeton Gamma Tech) energy-dispersive Excalibur system. All data were ZAF corrected.
|
Olivine
Representative olivine analyses are given in Table 1. A plot of olivine forsterite content versus whole-rock Mg-number (Fig. 2a) summarizes the core compositions and marginal zoning of euhedral phenocrysts, plus groundmass compositions. If a value of Kd for distribution of Fe2+ and Mg between olivine and melt of 0·30 is used (Ulmer, 1989
), it appears that all the phenocryst cores in bulk rocks with Mg-number <68 (Table 2) are approximately in equilibrium with their bulk-rock compositions. The two most Mg-rich samples—8159 and 8164 (Table 1)—plot below this curve. This relationship can be caused by olivine phenocryst accumulation in a melt and such an explanation may be appropriate because several percent of >2 mm olivine phenocrysts occur in each lava. Nevertheless, this conclusion is substantially model-dependent because the bulk-rock Mg-number values used in Fig. 2 were calculated with 10% of the Fe oxide in each analysis allocated to Fe2O3. If all the Fe is taken as FeO, the phenocryst cores in 8159 and 8164 lie on the Kd = 0·30 line and those of the lower Mg-number samples lie above it. Thus, Fig. 2a can in no sense be taken to ‘prove’ that samples 8159 and 8164 either are, or are not, olivine cumulates.
|
|
An unambiguous feature of many of these lavas is evident in Fig. 2b, which illustrates the compositions of the cores of the rounded macrocrysts. Although some of these are approximately in equilibrium with the bulk-rock compositions, most are much too Fe-rich for this to be so. Ulmer (1989)
showed that Kd might rise to 0·34 if the olivine formed at 15 kbar, but Fig. 2b emphasizes that this cannot explain the most abundant macrocryst composition, around Fo75, let alone the Fo65 in sample 8140. These widespread, partly dissolved macrocrysts are relatively Fe-rich olivine xenocrysts within these lavas. The significance of this point will be discussed below.
Clinopyroxene
Euhedral clinopyroxene phenocrysts in the PVF lavas have a substantial range of compositions, generated by sector, radial and oscillatory zoning. Compositions range from: SiO2 = 41·3–50·1 wt %; Al2O3 = 4·1–11·4 wt %; TiO2 = 0·8–6·0 wt %; Na2O = 0·3–1·6 wt % (Table 1). The highest Ti and Al values characterize pink titanaugite phenocryst rims and groundmass grains. A section through a large clinopyroxene megacryst (683a) from the Kilbourne Hole breccia was also analysed. Its mean composition (Table 1) has 8·8 wt % Al2O3 but only 1·5 wt % TiO2 and 1·4 wt % Na2O. When calculated into pyroxene components, using the scheme proposed by Kushiro (1962), this megacryst has small amounts of acmite, Ti-Tschermak's (Ti-Ts) and Fe3+-Tschermak's (Fe-Ts; Fe3+ estimated by charge balance) components. No jadeite is calculated but the megacryst is rich in Ca-Tschermak's component (Ca-Ts). In contrast, the euhedral clinopyroxene phenocrysts in the typical Potrillo lava (683) pierced by the Kilbourne Hole maar have core compositions with little Ca-Ts and almost twice the Ti-Ts component of the megacryst. Ti-Ts rises to >15% in groundmass titanaugites within this lava. In alkalic lavas with similar bulk compositions, especially Si-saturation, Ca-Ts is sensitive to crystallization pressure (Thompson, 1974; Soesoo, 1997).
The Kilbourne Hole clinopyroxene megacryst has a lower CaO content than those of the Potrillo lava euhedral phenocryst or groundmass clinopyroxenes (Table 1). This is clear on a pyroxene Di–Hed–En–Fs quadrilateral plot (Fig. 3). If they both formed from similar basic alkalic melts, the megacryst crystallized at higher temperatures than the euhedral phenocrysts because it falls closer to the two-pyroxene solvus in the pyroxene system. In some lavas, the phenocrysts have rounded-irregular cores of clinopyroxene, with a greenish tint, that show signs of having been partly redissolved by the melt, producing a ‘fingerprint’ texture. Occasional <5 mm irregular to rounded fragments of clinopyroxene macrocrysts also occur. As apparent in Fig. 3, analyses of these disequilibrium clinopyroxene remnants overlap both the megacryst and phenocryst data. We therefore infer that these are remnants of partly dissolved sub-calcic clinopyroxene high-pressure megacrysts or other clinopyroxenes that also formed at relatively high pressures (e.g. precipitated onto the walls of magma conduits). Figure 3 also shows that the Kilbourne Hole megacryst and Ca-Ts-rich sub-calcic remnants are appreciably richer in Fe than the cores of euhedral Potrillo clinopyroxenes.
|
Amphibole
The amphibole megacrysts in PVF scoria were not subjected to microprobe analysis because Irving & Frey (1984)
have already studied this mineral in detail and identified it as kaersutite. Their mean PVF amphibole megacryst major-element analysis is as follows: SiO2, 40·0; TiO2, 5·1; Al2O3, 14·5; FeO, 9·6; MnO, 0·11; MgO, 13·9; CaO, 11·1; Na2O, 2·4; K2O, 1·7 (all values in wt %).
Spinels
Three types of spinel occur in these samples (Table 1). Dark-green Mg–Fe–Al spinel (pleonaste) plots close to the spinel–hercynite join in Fig. 4b. It forms inclusions up to 5 mm in diameter within the Kilbourne Hole clinopyroxene megacryst. It also occurs as very small equant inclusions within a subhedral clinopyroxene phenocryst in sample 8134 and within a Fo74·5 olivine macrocryst in 8101. A loose cluster of subhedral olivine and clinopyroxene phenocrysts in 8134 includes olivines that appear to be a liquidus phase in this composition (marked as crosses in Fig. 2a). The greenish core of a coexisting clinopyroxene in the cluster contains a pleonaste grain. Brown chrome spinel occurs within the euhedral olivine phenocrysts of the lavas and titanomagnetite in their groundmasses (Table 1).
|
Plagioclase
Rounded megacrysts of plagioclase, up to 4 mm in size, occur in three of the analysed samples. Their compositions are in the range An27 to An48 (Table 1).
|
LAVA GEOCHEMISTRY |
|---|
All lava samples with low loss-on-ignition (below 1·2 wt % in all but one instance and positive in about 30% of samples) were analysed by XRF at Birmingham University for a range of trace elements. From this set of 145 samples, subsets were analysed for major elements (XRF), at Birmingham University, and additional trace elements (INAA), at Durham University, to describe the full range of elemental compositions in greater detail. Subsequently, a subset of 60 samples was analysed by ICP-MS at Durham, using the methodology of Ottley et al. (2003)
, so that modelling of the rare earth elements (REE) could use high-quality data, with the shapes of the chondrite-normalized patterns fully delineated. These samples were mostly those with the highest abundances of MgO and Ni. One clinopyroxene and two amphibole megacrysts were also analysed by ICP-MS (Table 3).
|
Analytical methods
The trace elements analysed by XRF (Birmingham) were determined on a Philips PW 1400 spectrometer, using a Rh anode. Appropriate corrections were made for overlap of Rb Kß on Y K
, of Sr Kß on Zr K and target contamination on Ni. The Rh Compton-scatter peak was used for absorption corrections. All samples were crushed in an agate ‘swing mill’ to 50 µm, and pelleted with a small amount of 7% aqueous PVA. Major-element analyses (Birmingham) were analysed using glass discs—a minimum of two for each sample. Synthetic standards were used (Leake & Hendry, 1969).
Rare earth elements Cr, Hf, Sc, Ta, Th and U were determined by instrumental neutron-activation analysis (INAA) at Durham. Samples were irradiated for five consecutive days at the University of London Reactor Centre at Silwood Park, in batches of nine samples interspaced with three standards to achieve satisfactory correction for variations in neutron flux. Standards were made by drying multi-element chemical solutions onto filter papers. For counting, samples were loaded onto a horizontal autochanger that accurately reproduced a sample-to-detector distance of 1·5–2·0 cm. Gamma-ray photons were collected by an EG&G® Ortec planar high-purity Ge low-energy photon detector (25 mm in diameter by 10 mm in thickness). Resolution of this instrument was excellent [nominally 550 eV full width at half maximum (FWHM) at 122 keV]. Gamma-rays were collected in the range 50–900 keV and processed as 4096-channel spectra. The efficiency of the detector was good up to 300–400 keV but declined dramatically in the energy range 400–920 keV. Nevertheless, good results were obtained using the following peaks with energies >300 keV: La (328·8 and 487·1 keV); Th (311·8 keV); Sc (889·4 keV). Leat et al. (1990) give results for international standards.
The current Durham technique for preparing and analysing samples by ICP-MS (Ottley et al., 2003) is not yet published in an international journal and therefore an abbreviated version is given here. The powdered samples, together with several blanks and a suitable standard, were prepared using a HF–HNO3 mixture (Romil SPA-grade) in a Savillex 22 ml PFA vial followed by evaporation to near dryness and further addition of HNO3. After driving off the initial HF–HNO3 mixture, samples were taken up in 1 ml of concentrated HNO3, ‘dried’ again and then a further 2·5 ml of concentrated HNO3 was added together with 10–15 ml of high-purity water, and then samples were left overnight on a hot plate. Following cooling, samples were further diluted to exactly 50 ml, to yield a solution in approximately 3·5% HNO3·
Sample and standard weight was always 0·1000 ± 0·001 g, in order to attempt to equalize the dissolved solids load and hence potential matrix effects. Any sample weight deviations were corrected for later. The final sample dilution was 1:5000. Re and Rh internal ‘spikes’ were added during the final dilution, to yield 20 ppb in solution. The use of Re and Rh as internal standards compensated for: (1) calibration drift because of changes in instrument sensitivity; (2) matrix suppression; (3) dilution errors.
Calibration of the Perkin Elmer Sciex Elan 6000 ICP-MS was achieved via the use of in-house standards and international reference materials (e.g. W-2, BHVO-1 and AGV1), together with procedural blanks. A blank and standard were routinely run every 10 samples to ascertain the magnitude of calibration drift. On the very few occasions that the RSD of the drift monitor exceeded 5% of the measured value, the samples were either re-analysed or a drift correction algorithm was incorporated into the data.
The isotopes selected for analysis were chosen to be free from isobaric interference wherever practical. Nevertheless, some of the REE in particular suffer from molecular oxide interferences. The magnitude of the interference correction is directly related to the plasma conditions and sample introduction system. All correction factors have been calculated relative to a CeO/Ce ratio of 2·6%. The Elan 6000 ICP-MS was optimized each day to obtain a 2·6% CeO/Ce ratio by changing the nebulizer gas flow.
Using these procedures, the elements determined showed good to excellent detection limits with all REE except Ce, being better than 5 ppb in the rock (Ottley et al., 2003). Routine detection limits (in the rock) of better than 10 ppb were obtained for most other trace elements of geological interest (LILE, HFSE). The reproducibility of elemental concentrations assessed by replicate runs of the same solution during an analysis period was usually better than 5% and frequently 3%. The external reproducibility of samples prepared from multiple separate dissolutions and run in separate analytical sessions were mostly better than 10% (RSD) and frequently better than 5% (2 SD) in typical basaltic rocks (Ottley et al., 2003).
Zr/Hf values for two international standards (NBS688 and BE-N) are between 37 and 38, and within error of the less precisely determined ‘accepted’ values. More importantly, the reproducibility (for separate dissolutions) of these values is excellent in each standard (1·2–1·7% RSD). Nb/Ta values given in the literature for Primitive Mantle and chondrites show more agreement than Zr/Hf, and are typically 17·3–17·6. External reproducibility for the two international standards (NBS688 and BE-N) was excellent (4·4 and 2·7% RSD, respectively) and shows good agreement with ‘accepted’ values. Both our values and the ‘accepted’ values are different from chondritic, but the good reproducibility indicates that this is probably because of a mineralogical factor rather than analytical problems (Ottley et al., 2003).
Sr and Nd isotopes were determined at McMaster University. Samples were leached with 6 M HCl overnight, then rinsed twice with high-purity de-ionized water before dissolution. The residues were dissolved in concentrated HF and HNO3 in Savillex ‘bomb’ vessels at 125°C for 3 days, before conversion to the chloride form for ion exchange. Standard cation and reverse-phase column separation methods were used. Analytical blanks were less than 0·02 ng for Nd (Sr not measured). Sr and Nd isotope analyses were performed on a VG isomass 354 mass-spectrometer. Sr was analysed by dynamic multi-collection, using a three-collector algorithm, and was normalized to a ratio of 0·1194 for 86Sr/88Sr. Nd was analysed using double filaments and a four-collector peak switching algorithm, and was normalized to a 146Nd/144Nd ratio of 0·7219. During this work, an average value of 0·710287 ± 35 (2
population) was determined for the NBS 987 standard (n = 10), and an average value of 0·511860 ± 15 (2 population) was determined for the La Jolla standard (n = 9). Sr isotope data were normalized to a value of 0·71024 for the NBS 987 standard.
Os isotopes were determined at the Arthur Holmes Isotope Geology Laboratory, Durham University. Whole-rock Re–Os chemical procedures followed the methods of Pearson & Woodland (2000). Digestions were performed in Carius tubes spiked for Re and Os. Total procedural Os blanks were 0·5 pg; all samples were blank corrected. Os samples were analysed by N-TIMS on a Thermo Electron Triton mass spectrometer. All analyses were carried out on the secondary electron multiplier, via ion-counting, in peak-hopping mode using a Ba(OH)2 activator solution. Using this procedure, our long-term mean 187Os/188Os value for 161 runs of the University of Maryland College Park standard, at signal sizes equivalent to those of the samples, was 0·11383 ± 32 (2 SD; 2·8 per mil) and is within error of the value of 0·113791 ± 3 produced from static Faraday runs of large loads by Walker et al. (1997). The mean 189Os/188Os over this period is 1·21976 ± 192 (2 SD; equates to 1·6 per mil). Re was analysed on a Thermo Electron Neptune mass spectrometer on the SEM, in peak-hoping mode, using Ir as a mass bias correction.
Element and oxide variation
Analyses of the lavas with ICP-MS data are given in Table 2; the remainder are in Electronic Appendix B. On a plot of total alkalis versus silica (TAS; Fig. 5), the Potrillo lavas classify as basanites, basalts and hawaiites (Le Maitre, 2002), in agreement with earlier analyses (Anthony et al., 1992). MgO in the Potrillo suite varies from 5·6 to 12·5 wt %, with three samples classifying as picrites (IUGS; MgO > 12 wt %; Le Bas, 2000). Our analyses of PVF lavas with MgO < 6·5 wt % are both from the Santo Tomas–Black Mountain area, east of the main PVF (Electronic Appendix A) and we have not studied them in detail (Hoffer, 1971; Anthony et al., 1992).
|
Clues to the nature of the subdivisions and fractionating assemblages in the PVF magmas are given by a plot of MgO versus various elements and oxides (Fig. 6). Strong positive correlations between MgO and both Ni and Cr (Fig. 6a) show that one or more ferromagnesian phases fractionated from this suite (the marked trend-lines are linear least-squares fits through non-inflected data arrays). In contrast, a plot of MgO versus K2O (Fig. 6b) shows that the overall PVF suite contains two series. About 85% of the lavas form the Main Series, in which K2O behaves incompatibly throughout the sampled MgO range (Fig. 6b). The remainder all have variable but distinctly lower K2O abundances (<1·3%), with a mean of
1·0% that does not vary as a function of MgO. This Low-K Series is distinguished from the Main Series in Fig. 5 and subsequent diagrams. Figure 6 is discussed in more detail below.
|
Isotopes
There are no published isotopic studies of PVF lavas where the elemental compositions of the samples are also recorded and MgO exceeds 9 wt %. As the Mg-rich lavas are the main focus of this study, nine samples were analysed for Sr and Nd isotopes, and four of these were analysed for Os isotopes. The results are given in Tables 4 and 5.
|
|
Sr and Nd isotopes
Five Main Series and four Low-K Series Mg-rich lavas were analysed for Sr and Nd isotopes (Table 4). Their 87Sr/86Sr and 143Nd/144Nd ratios define a range that is scarcely larger than analytical precision but, nevertheless, shows a weak negative correlation between 87Sr/86Sr and 143Nd/144Nd (Fig. 7a, inset). The elemental differences between the Main and Low-K series are not reflected in their Sr–Nd isotopes, which occupy the same small field (Fig. 7a).
|
Os isotopes
The most striking feature of the results (Table 5) is the wide range of Os contents within lavas having a relatively narrow range of MgO contents (11·1–12·5 wt %). Thus, one lava (6110; Main Series) has an Os content (209 ppt) that might be expected for a picrite elsewhere (e.g. Hauri, 2002
), whereas the Os contents of the other lavas are much lower (4–14 ppt). Figure 7b and c show that the low-Os lavas have a much higher 187Os/188Os than sample 6110. As with Sr–Nd isotopes, the Low-K Series sample (870) has a similar value of 187Os/188Os to the two Main Series lavas with low Os contents. |
DISCUSSION OF PVF LAVA GEOCHEMISTRY |
|---|
What minerals fractionated from the magmas: the phenocrysts, megacrysts or both?
At the outset of this discussion, it is necessary to establish what process was predominant in causing the large range of chemical variation in the PVF lavas. Because the Low-K lavas are only a small proportion of the sample set, their genesis will be considered separately below. It is first necessary to identify what process can cause MgO to vary continuously from 12·5 to 5·9 wt % (Table 2) with equally wide ranges of Ni (332–47 ppm) and Cr (409–55 ppm). Although both mantle melting and fractional crystallization processes could cause such variation in Mg-rich lavas, the Mg–Ni–Cr-poor PVF lavas clearly require fractional crystallization from more magnesian parent magmas to be the main factor in their genesis. Furthermore, all the PVF lavas are in equilibrium with olivine far less forsteritic than any plausible mantle composition (Fig. 2). The distinctive patterns of variation in Fig. 6, where some trends are straight and others inflected, gives clues as to which phases separated from the magmas.
Phenocrysts. Olivine is a euhedral phenocryst in all the porphyritic PVF lavas; alone in samples with >10·7 wt % MgO and accompanied by euhedral, sector-zoned clinopyroxene in those with 7·0–10·7 wt % MgO. If olivine alone fractionated from PVF lavas with >10·7 wt % MgO, and was joined by clinopyroxene subsequently, the types of plot forming Fig. 6 should show inflections in trends of CaO and Sc vs MgO, as they do (Fig. 6c and d). Separation of plagioclase in the low-Mg lavas is probably marked by the slight inflection in the Al2O3 trend at 7–8 wt % MgO but there are insufficient data for this to be clear. Euhedral plagioclase phenocrysts occur only in lavas with MgO < 7 wt %. Rounded, relatively sodic plagioclase macrocrysts occur in more magnesian lavas (Table 1), but these are unlikely to have precipitated from such basic magmas (Irving & Frey, 1984). Their significance will be discussed below.
Megacrysts. The reduction in both CaO and Sc contents of the magmas, with falling MgO, requires that a major fractionating phase was either clinopyroxene or amphibole (or both). The 11% MgO Main Series lavas contain 10·0–11·4% CaO and 34 ppm Sc (Table 2). These values cannot be compared quantitatively with those of the small euhedral PVF clinopyroxene phenocrysts because the latter are sector-zoned, with extremely internally variable compositions. Apart from olivine, the other obvious candidates for fractionating phases are the clinopyroxene and amphibole megacrysts (Tables 1 and 3).The megacryst clinopyroxene contains several percent of green Ti-poor Mg–Al–Fe spinel (Table 1) but none has been found within megacryst amphibole. The coarse, cumulate-textured ultramafic xenoliths (see above) show textural relationships, suggesting that small amounts of both olivine and Mg–Al–Fe spinel co-precipitated with clinopyroxene, whereas amphibole crystallized at lower temperatures and alone. Extensive removal of amphibole alone from 11 wt % MgO PVF magmas could reduce CaO and Sc (Table 3) in the residua but would not permit a concomitant increase in Al2O3 because the Al2O3 content of the amphibole (14·5%; Irving & Frey, 1984) and the 11 wt % MgO melts are approximately the same (Tables 2 and 3). Amphibole separation (mean TiO2 content by ICP-MS 6·1 wt %, Table 3) would also strongly deplete TiO2 in the residual melt. Vectors for removal of 10 wt % megacryst clinopyroxene and amphibole from a typical PVF Mg-rich basalt are marked on the Al2O3 and TiO2 diagrams of Fig. 6.
Both the lava mineralogy and chemistry, and the ultramafic inclusions described above closely resemble those of the Geronimo Volcanic Field, southern Arizona, about 200 km west of the PVF (Kempton et al., 1987; Kempton & Dungan, 1989). Kempton et al. (1987) modelled the fractionation of the Geronimo magmas in detail, using a major-element mass-balance approach. Their preferred model for hawaiite derivation from alkali basalt required fractionation of a 5:1 assemblage of megacryst aluminous clinopyroxene and Mg–Al–Fe spinel, plus minor olivine. Two features of the PVF lavas rule out a similar quantified fractional crystallization model for this suite: (1) the trends on such plots as Fig. 6 are all scattered and this is a characteristic of melts where repetitive episodes of magma mixing have taken place; (2) petrographic evidence for such mixing is clear from the abundant rounded/fritted macrocryst fragments in many thin sections. In each sample studied in detail (e.g. Fig. 2b), the rounded olivine macrocrysts are more Fe-rich than coexisting euhedral phenocrysts of this phase. This is consistent with magmatic plumbing beneath the PVF in which uprising magma batches frequently intercepted and mixed with the fractionated residua of previous batches. Thus, the composition of any specific sample is a function of three post-genesis processes: (1) fractional crystallization; (2) mixing between such liquids and others encountered during their uprise; (3) re-fusion of unknown amounts of previous crystal cumulates from earlier magma batches.
Kempton (1987; Kempton & Dungan, 1989) considered that the Geronimo aluminous clinopyroxene and Mg–Al–Fe spinel megacrysts precipitated at ‘moderate pressures’. Two thermobarometers have subsequently been proposed to calculate the P and T of Ca-rich clinopyroxene formation from its elemental composition (Putirka et al., 1996; Soesoo, 1997). The thermodynamic approach of Putirka et al. (1996) requires the equilibrium compositions of both clinopyroxene and coexisting melt. Therefore, it cannot be applied to the PVF because these euhedral–subhedral phenocrysts are sector-zoned and therefore not equilibrium phases (Thompson, 1972).
The empirical approach of Soesoo (1997) uses principal component analysis of the major-element compositions of clinopyroxenes crystallized from melting experiments to construct eigenvector grids for P and T. The eigenvectors for the mean composition of the Kilbourne Hole clinopyroxene megacryst 683a (Table 1) are XPT = 32·35 and YPT = –27·55. On Soesoo's plots, these correspond to P = 1·1–1·5 GPa and T 1210°C. Greenwood (2001) compared Soesoo's results with the experimental data of Thompson (1974) for alkalic melts and suggested that the Soesoo thermobarometer might underestimate P by 0·4 GPa and T by 40°C. Therefore, the Ca-Ts-rich clinopyroxene megacryst 683a may have formed at 1·7 GPa pressure. This corresponds to a depth of 58 km, if a 30 km sialic crustal thickness is used (Sinno et al., 1986; Bussod & Williams, 1991). Perhaps 50–60 km is a sufficiently cautious estimate; i.e. 10–20 km above the base of the lithosphere, as defined by Bussod & Williams (1991).
The green Mg–Al–Fe spinel forming <5 mm rounded inclusions within the 5 cm clinopyroxene megacryst 683a appears, on textural grounds, to have crystallized either before or together with the latter. The cumulus-textured ultramafic inclusions within the PVF volcanics likewise give textural evidence of a clinopyroxene–green spinel–olivine assemblage. Thus, we conclude that PVF magmas with MgO < 10·7 wt % have undergone fractional crystallization dominated by removal of aluminous sub-calcic augite, accompanied by Mg–Al–Fe spinel and olivine at a pressure of 1·7 GPa. This pressure agrees well with the value of 1·6 GPa calculated by Huckenholz et al. (1992) for equilibrium between clinopyroxene and amphibole, based on the PVF megacrysts analysed by Irving & Frey (1984).
The REE data for clinopyroxene and amphibole megacrysts (Table 3) may be used to test the hypothesis that these phases crystallized from melts of the same composition as the PVF magmas. On a chondrite-normalized REE plot (Fig. 8), the patterns of all the PVF lavas analysed using ICP-MS (Table 2) plot in a compact group. The REE patterns for the average Potrillo clinopyroxene and amphibole megacrysts (Table 3) are also shown. Bottazzi et al. (1999) investigated trace-element partitioning between a kaersutitic amphibole (TiO2 = 4·06 wt %) and an alkali basalt melt—a situation very similar to the Potrillo example. The melt calculated to be in equilibrium with the Potrillo amphibole megacryst (Fig. 8), using the kaersutite/melt REE partition coefficients of Bottazzi et al. (1999), has similar REE abundances to the PVF lavas.
|
The modelling of melts in equilibrium with a clinopyroxene megacryst is more complicated. Recent experimental studies of clinopyroxene/melt partition coefficients, using ion microprobe data, have produced very variable values, especially for the heavy REE (e.g. Hart & Dunn, 1993
; Blundy et al., 1998). Therefore, we used the predictive thermodynamic model of Wood & Blundy (1997) to calculate the values of REE partition coefficients appropriate to the crystallization of the Kilbourne Hole clinopyroxene megacryst (Table 3) from a PVF basaltic melt. Figure 8 shows the calculated equilibrium melt REE pattern appropriate to the P and T estimated above using the method of Soesoo (1997). Additional patterns are given for equilibria at 1·7 GPa—our preferred pressure for the clinopyroxene megacryst formation—and both 1200 and 1250°C. All three patterns are very close to each other and also to the result obtained from the Potrillo amphibole megacrysts. It is clear from Fig. 8 that at the depths relevant to this study, the effects of P are much smaller that those of T in the Wood & Blundy model. It is concluded that the calculations shown in Fig. 8 support the hypothesis that the amphibole and clinopyroxene megacrysts in the PVF lavas are cognate.
Isotope geochemistry
Sr–Nd isotopes. Figure 7a shows that all the analysed PVF lavas have 87Sr/86Sr higher and 143Nd/144Nd lower than MORB, with values of these ratios falling in the same range as OIB. Recent alkali basalts and basanites in the western USA have similar Sr–Nd isotopic ratios (e.g. Wang et al., 2002, and references therein). The complete overlap between both 87Sr/86Sr and 143Nd/144Nd values in the Main and Low-K series samples (Fig. 7a, inset) removes the simplest explanation for these two suites, namely that they originated from different mantle sources. Possibly, they originated from the same mantle source and the differences between them developed subsequently. Alternatively, the mantle source might have varied mineralogically, with a K-bearing mineral (such as amphibole) remaining in the residue after partial melting in some places but not in others (see below). Furthermore, crustal contamination can also be ruled out as the cause of the Low-K Series because any such process, on a scale sufficient to make changes to their elemental compositions as large as those seen in Fig. 6, would be bound also to affect their Sr–Nd isotopic ratios. The 1·6 Ga acid crust sampled as xenoliths in the lava at Kilbourne Hole has values of 87Sr/86Sr up to 0·79 and 143Nd/144Nd down to 0·5119 (Padovani & Reid, 1989). The Sr–Nd isotopic compositions of the lithospheric mantle beneath the PVF are represented in Fig. 7a by analyses of peridotite xenolith samples (some with pyroxenite veins) from Kilbourne Hole (Roden et al., 1988). The majority of these have lower 87Sr/86Sr and higher 143Nd/144Nd than the PVF lavas. Nevertheless, two of these xenoliths have more radiogenic compositions than the Potrillo lavas and therefore Sr–Nd isotopic ratios alone cannot completely exclude the possibility that the sub-PVF lithospheric mantle has an isotopically variable composition that was averaged when the PVF magmas melted from it.
Os isotopes. Such a possibility can be evaluated in the light of the Os isotope results. Osmium is an ideal element for discriminating between the melting products of convecting and lithospheric mantle (e.g. Hauri, 2002). During the melting associated with lithosphere formation, Re is partitioned preferentially into the melt and Os into the residuum, thus creating large differences in Re/Os. Over geological time, as 187Re decays to 187Os, the ratio 187Os/188Os in a given volume of mantle peridotite rises at very variable rates, depending on its melting history and whether or not it has avoided the constant remixing associated with convection. Thus, geologically old lithospheric mantle characteristically has relatively low Re/Os and 187Os/188Os (e.g. Pearson et al., 2004). If this were to melt to produce basaltic magmas, they would inherit low 187Os/188Os ratios.
If the lithospheric mantle contains abundant geologically old pyroxenite veins, the situation is more complicated. In peridotite massifs such as Beni Bousera, Morocco, the pyroxenite veins have high enough Os contents and 187Os/188Os ratios that a subsequent basaltic melt from both the peridotite and its included pyroxenite veins could have significantly greater 187Os/188Os than a simple peridotite melt (Pearson & Nowell, 2004). Nevertheless, before speculating about a hypothesis of pyroxenite-veined lithospheric mantle fusion for PVF magmatism, it would seem prudent to find evidence that such radiogenic pyroxenite veins actually occur within the lithospheric mantle sampled at Kilbourne Hole. This suite of mantle xenoliths has been subjected to one of the most extensive Os-isotope studies of any non-cratonic, basalt-hosted example to date (Burton et al., 1999; Meisel et al., 2001), but, unfortunately, no pyroxenites were analysed. Nevertheless, even if such radiogenic pyroxenites from the Kilbourne Hole xenolith suite are reported in the future, it is important to remember that Bussod & Williams (1991) showed that such veins mostly came from the shallowest parts of the sub-PVF lithospheric mantle, in peridotites with pre-eruption pyroxene equilibration temperatures between 850 and 1000°C, in the 35–55 km depth range—well above the depths at which the Potrillo magmas appear to have originated (see below). Therefore, their relevance to PVF magma genesis would be minimal.
Figure 7b shows the ranges of 187Os/188Os (expressed as Os) and the published data for Kilbourne Hole lithospheric mantle peridotites. The parameter Os is used because this logarithmic-scale plot clarifies relationships in sample suites with 187Os/188Os close to Bulk Earth [using the value of Pearson & Nowell (2004)]. The Os range of the PVF lavas is clearly higher than that of the Kilbourne Hole lithospheric mantle xenoliths, with no overlap between the groups. Meisel et al. (2001) calculated a model rhenium depletion age for the Kilbourne Hole lherzolite xenoliths of 1·6 Ga, comparable with the basement age of the local crust. This suggests that both crust and mantle formation in this area were results of a single 1·6 Ga event, and that no subsequent process in the lithospheric mantle disturbed the Os-isotope systematics of the latter.
The PVF samples (Table 5) include three from the Main Series and one from the Low-K Series, all with MgO contents among the highest found in our sample set. Three of the samples (two Main; one Low-K) have extremely low Os contents (4·1–14·1 ppt), far lower than could occur in an unmodified basic magma (MgO > 12 wt %) that had originated in equilibrium with mantle peridotite (45 ppt Os or more; Hauri, 2002). Clearly, despite their high MgO contents, these magmas have not proceeded without delay from their sources to the surface. Their low Os contents are combined with extremely high 187Os/188Os. This suggests that they paused in crustal reservoirs during their uprise and that this delay allowed two processes to take place: (1) a sulphide phase precipitated from the melts, so that most of the original Os in the melt was lost; (2) as a result, the Os-depleted melts became hypersensitive to contamination by continental crust with 187Os/188Os 0·5–2·5 (Saal et al., 1998). These three samples may be ‘failures’ in any attempt to decipher the original sources of the PVF magmas but they confirm our interpretation, detailed above, that even the MgO-rich members of this suite are affected by complex crystal–liquid differentiation and subtle crustal contamination.
The fourth sample, 6110, is the lava that erupted to form Potrillo maar (Fig. 1). This has exhumed a suite of mantle xenoliths, testifying to its uninterrupted uprise from mantle depths. Its Os content of 209 ppt is high enough to make it strongly resistant to crustal contamination and its 187Os/188Os value of 0·145 places it firmly in the OIB field (Fig. 7c) close to the maximum value of 0·150 (Hauri, 2002). A mixing curve is plotted in Fig. 7c between sample 6110 and average Kilbourne Hole lithospheric mantle peridotite, using an arithmetic mean of the Os isotope compositions published for Kilbourne Hole peridotites (Burton et al., 1999; Meisel et al., 2001). The 187Os/188Os range of these xenol