V21A-2073
Laser Ablation ICP-MS of Mixed Pumices from Hekla Volcano, Iceland
Despite its more recent effusive behavior in historic times, Hekla Volcano has experienced at least 5 major Plinian eruptions in the past 7000 years (H5, H4, the Selsund pumice, and H3), the most recent in 1104 AD (H1). As first observed by Thorarinsson, the explosive phases of the Plinian eruptions are characterized by rhyolitic to dacitic pumice (up to 75 wt.% SiO2) that grades progressively to more mafic compositions as the eruptions proceed. Most of the pumices collected throughout the H3 deposit appear to be compositionally homogeneous. However, when the individual pumices are examined closely, some exhibit swirled textures of lighter and darker colored glasses. Larsen and Thorarinsson (1977) also found veined pumices from the H4 eruption. Outstanding unresolved questions regarding Hekla's eruptive behavior include the origin of the different liquids, the compositional variation that the liquids exhibited immediately before eruption, and the amount of time the liquids have been in contact with each other. To address these questions, we are microanalyzing mingled pumices from the middle part of Hekla Plinian fallout deposits, along with matrix glass of homogeneous clasts. The Colgate University facility couples a Varian 820-MS with a New Wave UP-213 laser ablation system. Sample areas down to 5 microns in diameter and up to several hundred square millimeters have been analyzed in a single acquisition, permitting analysis of individual crystals and inclusions, as well as bulk analysis of matrix glass. With our apparatus, typical detection limits for analysis of a sample range from parts per trillion to tens of parts per billion for a laser spot diameter of 100 microns.
V21A-2074
Composition of Melt Inclusions in High-3He/4He Picrites From Baffin Island and West Greenland
High 3He/4He ratios in Baffin Island and West Greenland picrites, measured by in vacuo crushing of olivine phenocrysts, are associated with a wide range of whole-rock incompatible trace element ratios and Nd and Sr isotopic compositions. To establish that this range reflects heterogeneity of the mantle source we must rule out the possibility that it is due to crustal contamination of depleted primitive magma. Contamination would most likely occur during olivine crystallisation and this should result in differences in composition between bulk rocks and the melt-inclusion hosts of the helium. This study represents an essential step in linking high 3He/4He to specific mantle sources. We present major and trace element compositions of olivine-hosted melt inclusions from a set of high- 3He/4He (40-50Ra) Baffin Island and West Greenland picrites that cover a range of bulk-rock incompatible trace element and Nd- and Sr-isotope ratios. Only naturally quenched glassy melt inclusions were selected for electron- and ion microprobe analyses. Melt inclusions show a much wider range in incompatible-element concentrations than do the host rocks, as previously shown at other locations by many authors, but incompatible-element ratios are comparable. The conformity of whole-rock and melt inclusion rare earth element (REE) profiles suggests that the Baffin Island-West Greenland magmas were not affected by crustal contamination either during or after olivine crystallisation. This is supported by modelling using incompatible trace element ratios. Importantly, melt inclusion incompatible trace element ratios correlate with whole-rock 143Nd/144Nd. We conclude that whole-rock incompatible-element and Nd-isotope variation is not due to crustal contamination but is directly linked to source variation. It therefore follows that high 3He/4He is a feature of both depleted and relatively enriched mantle sources.
V21A-2075
Co-variation of Boron Concentrations and Isotope Ratios in Glasses From the Laki- Grimsvotn Fissure Eruption of 1783-1784
The Laki-Grimsvotn fissure, located in the Eastern Volcanic Zone in southeast Iceland, erupted in 1783-1784 producing over 15 km3 of basaltic lava and associated aerosols. We have collected proximal tephra samples from the fissure, north and south of Laki (hyaloclastite mountain). Trace element concentrations and boron isotope analyses were performed on matrix glasses and plagioclase-hosted melt inclusions from various eruptive phases from this eruption using SIMS (Cameca IMS 3F and 1280) at the Woods Hole Oceanographic Institution. Preliminary data from SIMS boron isotope analyses shows variation in δ11B from -5 to -19‰ (2σ<±4‰ at B concentration level <1.5 ppm). A negative correlation between δ11B and boron concentrations in the melt inclusions demonstrates that as boron concentrations increase in the residual melt, the melt is also increasingly more enriched in 10B relative to 11B. Scoria lapilli glasses form a relatively tight cluster, defining the lowest δ11B values (-10 to -19‰) and highest boron concentrations (1.1-1.4 ppm). Typical δ11B values for MORBs are around -5‰ and may represent the primitive end member of observed results from the Laki melt inclusions. Hervig et al. (2002) have shown experimentally that 11B is partitioned into the fluid phase and 10B into a silicate melt when the two are held at equilibrium, which could account for some of the variation observed here (Hervig et al. observed 3‰ fractionation between the fluid and basaltic melt). They have also shown that B is compatible in basaltic melt during degassing. Thus simultaneous degassing and fractional crystallization could explain the increase in B during degassing, even as δ11B decreases. Based on our current understanding, however, it seems unlikely that degassing alone can account for the >10‰fractionation observed in the Laki melt inclusions. Isotopically light boron may also be introduced into the system by shallow-level assimilation of preexisting altered hyaloclastite material prior to eruption, as has been suggested by Bindeman et al. (2008) on the basis of oxygen isotope data. R. Hervig et al. (2002), Am. Min. 87, 769-774 I. Bindeman et al. (2008), Geochim. Cosmochim. Acta, 72, 4397-4420
V21A-2076
Concurrent mixing and cooling of melts under Iceland
Closely spaced eruptions of similar age in Iceland have significantly different trace element and isotopic compositions. It is widely accepted that these differences reflect variation in the composition of mantle melts generated under Iceland. Olivine-hosted melt inclusions show an even greater range of trace element compositions, with inclusions from single eruptions showing a 40-fold variation in the concentration of highly incompatible elements (Nb, La) and a factor of 5 for less incompatible elements (Yb, Y). While the origin of this extreme trace element variation was originally linked to mantle source heterogeneity and fractional melting, more recent studies have interpreted the inclusion trace element variability as a product of crustal magma chamber processes. Indeed, the presence of large positive Sr anomalies in ~5% of the 360 inclusions considered in this study indicates that some of the inclusions have compositions that have been influenced by interaction with plagioclase-bearing rocks. However, inclusions from a single hand-specimen of picrite exhibit significant Pb-isotope variation which correlates with incompatible element concentrations. These data indicate that an important part of the melt inclusion trace element variation reflects variation in the composition of melts supplied from the mantle. This conclusion is supported by the coincidence of arrays of whole-rock and melt inclusion data in trace element-isotope space and by the inability of simple models of magma chamber processes to match key aspects of the trace element systematics. The 11 eruptions studied exhibit a close correspondence between the mean composition of whole-rock samples of the carrier lava and the mean composition of the olivine hosted melt inclusion for trace element ratios that are not sensitive to crustal processes (e.g. Sm/Yb). The compositions of the olivines that host the inclusions are variable and 90% are too forsteritic to be in equilibrium with the melt that carried them to the surface. In two well-studied flows there is a statistically robust drop in the variation in trace element ratios in the inclusions with decreasing forsterite content of the host olivine. A similar record of concurrent mixing and crystallisation is preserved in the composition of clinopyroxene crystals found in one of the flows. The inclusions trapped in olivines close to Mg-Fe equilibrium with their carrier melt have trace element contents similar to the carrier. The entire Icelandic inclusion dataset exhibits a robust drop in the deviation of inclusion composition from carrier melt composition while the forsterite content of the host olivine drops. These observations are consistent with a model where compositionally variable melts are supplied from the mantle and trapped in forsteritic olivines. These melts undergo concurrent mixing and crystallisation until the well-mixed melt carries the olivines to the surface.
V21A-2077
The infidelity of melt inclusions?
Melt inclusions provide important information about magmatic systems and represent unique records of magma composition and evolution. However, it is also clear that melt inclusions do not necessarily constitute a petrological 'magic bullet', and potential exists for trapped melt compositions to be modified by a range of inclusion-specific processes. These include trapping of diffusional boundary layers, crystallization of the host mineral after trapping and dissolution of co-trapped minerals during homogenization, diffusional exchange between trapped liquid and the host mineral and external melt, and cryptic alteration of trapped material during weathering or hydrothermal alteration. It clearly important to identify when melt inclusions are unmodified, and which compositional indices represent the most robust sources of petrogenetic information. In this presentation I review and discuss various approaches for evaluating compositions and compositional variations in inclusion suites. An overriding principle is that the variations evident in melt inclusions should be able to be understood in terms of petrological processes that are known, or can be reasonably inferred to also effect bulk magma compositions. One common approach is to base petrological conclusions on species that should be more robust, and many workers use variations in incompatible trace elements for this purpose. However important information may also be obtained from a comparison of variations in melt inclusions and the lavas that host them, and in most cases this comparison is the key to identifying inclusions and suites that are potentially suspect. Comparisons can be made between individual inclusions and lavas, although comparison of average inclusion composition and the host lava, after correction for differences in crystal fractionation, may also be valuable. An important extension of this is the comparison of the variability of different species in inclusions and host lavas. This also provides a means to directly test for effects of inclusion-specific processes through comparison between variance and diffusivity, partition coefficient or other parameters believed to drive compositional changes. Another test that is becoming more accessible is the direct comparison of trace element compositions of inclusions and host minerals, coupled with known element partitioning behavior.
V21A-2078
Timing of crustal contamination from melt inclusions in the Ice Springs Flow, Utah
This study aims to better understand crustal assimilation processes by focusing on relatively simple monogenetic flows and by looking at trace element variations in whole rock and mineral hosted-melt inclusions. The 660 year B.P. Ice Springs flow in central Utah is a small (20 km2) primitive basalt with less than 1% phenocrysts of olivine, plagioclase and rare orthopyroxene. The flow has two main lobes: a low silica lobe (49% SiO2, K/Ti: ~0.6) and a high silica lobe (51% SiO2, K/Ti: ~1.0) for 7-8 wt. % MgO. We use the ratio K/Ti to track assimilation as this ratio should not change during differentiation. The flow contains partially melted crustal xenoliths which have a very high K/Ti ratio, quartz xenocrysts, and has within flow trace element variations which can be explained by assimilation of crustal material. We present major and trace element data on melt inclusions hosted in olivine, plagioclase, and orthopyroxene as well as on the partially melted xenoliths and local rhyolites. Olivine-hosted melt inclusion composition for both lobes have the same SiO2 (~49%) and K/Ti (~0.5), suggesting that assimilation has taken place after olivine crystallization. Orthopyroxene-hosted melt inclusion compositions show a less differentiated magma (SiO2 (~47%-48% and K/Ti 0.3) than inclusions in olivine. The simplest hypothesis is that the chemical variations seen in the flow are the result assimilation of crust – likely to be represented by the partially melted xenoliths - after olivine and orthopyroxene crystallization. Major and trace element data from xenoliths will be utilized to test our late stage assimilation hypothesis and to calculate mass balances for melting and residual phases.
V21A-2079
Volatile Evolution of Magma Associated with the Solchiaro Eruption in the Phlegrean Volcanic District (Italy)
The Phlegrean volcanic district (PVD) in southern Italy is one of the best known volcanic hazard areas in the world. More than 1.5 million people live in close proximity to the volcanic centers. The PVD comprises three volcanic fields: the Campi Flegrei caldera and the islands of Ischia and Procida. The area contains many volcanic centers (cinder cones, tuff rings, calderas) and has been the site of episodic volcanic activity for more than 60 ka. Some of these eruptions have been extremely violent. We studied volatiles in the magma associated with the Solchiaro eruption on the Island of Procida, Italy, to gain a better understanding of the relationship between eruptive style and intensity and the volatile content of the magma. The Solchiaro eruption is one of the more primitive products erupted in the PVD and provides information on the source of later more evolved magmas associated with this volcanic system. The composition of the magma before eruption was determined by analyzing melt inclusions (MIs) in forsteritic olivine and diopside. Several different types of MIs were observed in both phases. Some MI contained only glass, others contained glass plus one or more bubbles, and some contained glass plus bubbles plus crystals. We analyzed MIs containing only glass and those containing one or two bubbles. The composition of MIs ranges from basaltic to trachy-basaltic. Among major elements potassium shows the highest variability, from 1.48 to 3.73 wt %. Laser ablation-inductively coupled plasma mass spectroscopy (LA-ICPMS) analysis of MIs combined with major element analysis suggest assimilation of LILE and halogens from wallrock. Diopside-hosted MI have a higher H2O content than olivine-hosted MI, based on Secondary Ion Mass Spectrometric (SIMS) analysis. MI in olivine contain from 1.13 to 1.38 wt % H2O. F, Cl, S, CO2 contents are highly variable in both olivine and diopside-hosted MIs. The trend in H2O versus CO2 suggests emplacement of an originally CO2-saturated magma at >5 km that rises through the crust as it is undergoing crystallization. The magma eventually ponds at shallow depths (<1 km) before eruption. The inferred emplacement and crystallization history of magma at Campi Flegrei is similar to what was observed at Mt. St. Helens in the months immediately preceding the May 1980 eruption.
V21A-2080
Magma Evolution at Vulsini Volcanoes (Roman Province) in the Light of Phenocryst Melt Inclusions, Thermochemical Modeling and HP-HT Laboratory Experiments
Volcanic terrains south of Bolsena lake (Vulsini Volcanoes, central Italy) include a large variety of lava and pyroclastic rock-types of potassic affinity from Campi Vulsini, Latera and Montefiascone volcanic complexes in the 0.5-0.1 Ma span time. By tradition, the erupted compositions, ranging continuously from trachybasalts and basanites-tephrites to trachytes and phonolites in the TAS diagram, have been related to different magmatic series - i.e: KS, HKS -, originated from different primary magma sources. Here we propose, instead, a common parental magma for the typical compositional trends, based on several lines of evidence: i) the close association in time and space, even at the scale of individual eruptive centers, of rock-types attributed to different series; ii) the occurrence of trachybasalt (Al2O3=12 wt%, MgO=9 wt%, CaO=16 wt%) as the most primitive composition among the erupted products, as also found from melt inclusions in early olivine phenocrysts through the different volcanic complexes. The eruption of this parental (possibly primary) magma was favoured by the peculiar volcano-tectonic setting, related to major faults bordering Bolsena, Latera and Montefiascone caldera structures. Phase relations at different P, T, %H2O(melt) and fO2, from MELTS code thermochemical calculations, as also supported by HP-HT laboratory experiments performed on the most primitive rock-types, may explain consistently the observed variety of erupted magma compositions.
V21A-2081
Evidence for Boninite Genesis in the Eastern Manus Basin
The eastern part of the Manus Back-Arc Basin features a wide compositional range of neovolcanic rocks, ranging from arc picrite to rhyolite. Preliminary inspection of rocks recovered by ROV Jason during an RV Melville cruise in 2006 by electron microprobe provides hints for boninite genesis hitherto unknown from this area. Porphyritic rocks from the areas of Susu Knolls and Umbo Ridge in the easternmost part of the Manus Basin show olivine xenocrysts (Mg 87-94) hosting Cr-spinel inclusions (Cr up to 92). Mixed rocks from the submarine North Su volcano are nominally andesites that are composed of olivine, clinopyroxene, orthopyroxene and Ca-rich plagioclase in a dacitic matrix. Olivine has uniform composition and is separated by a thin rim (>10-15μm) of orthopyroxene (Mg 73) from the matrix. The Mg-rich olivine and the Cr- spinel inclusions plot in the most depleted end of the olivine-spinel mantle array (Arai, JGR, 1994), indicating extremely high degrees of melt depletion (>40 wt.%). The most chromian spinels, however, have high TiO2 contents (up to 0.2 wt.%), indicating a subsequent stage of pronounced melt-peridotite interaction (Pearce et al., CMP, 2000). Uncommon for olivine with Mg 93 is the high CaO (0.26 wt.%) and low NiO content (0.24 wt.%). Several melt inclusions in olivine from North Su have boninite-like major element composition with 55 wt.% SiO2 and high contents of CaO and MgO. Arc picrites from the Umbo Ridge (20 km west of North Su) show a range of olivine compositions and zonation patterns. Some olivines are Mg- rich in the core (Mg 94) with more ferroan rims (Mg =87). Cr-spinel inclusions in the Mg-rich olivine cores have compositions that are identical to those of Cr-spinel in North Su andesites. Other olivine has a more complex zonation with a low-Mg core (Mg 87), followed by a zone of higher XMg (Mg 90), again decreasing towards the rim (Mg 88). The trace element patterns of the matrix glasses from both locations are typical for fluid-related signatures of arc volcanics extremely depleted in HFSE and enriched in LILE. MORB- normalized REE patterns are flat and slightly LREE-enriched. Most samples from the eastern Manus Basin show U excesses, implying that fluid-mediated mass transfer from a subducting slab into the magma source is likely. There are different explanations for the origin of boninites in the Manus Basin. (1) Both Umbo Ridge and North Su are neovolcanic structures within a pull-apart zone between the Djaul fault and the Weitin fault. Extreme crustal thinning in this area may have facilitated the ascent of hot asthenosphere that could melt preexisting arc lithosphere. (2) Subduction of the Solomon Microplate since the Miocene has the tip of the downgoing slab just arriving below the neovolcanic zone of the eastern Manus Basin. Slab melting typical for the initial stage of subduction could hence have played a role in boninite genesis, although the whole rock trace element chemistry does not reveal a garnet signature. (3) A transient pulse of volatiles released from the downgoing slab may have flux-melted the mantle wedge beneath the Manus basin. Major element compositions of olivine and spinel strongly suggest that extremely depleted mantle wedge material interacted with ascending melts. Whether those melts were created by slab melting, flux-supported pressure release melting or by melting of pre-existing arc lithosphere is presently unclear. We hope to resolve this issue by trace element microanalyses of different phases and by whole-rock isotope geochemistry.
V21A-2082
Melt Inclusions in Olivine Phenocrysts From Primitive Submarine Subduction-Related Lavas From the Hunter Ridge (North Fiji Backarc Basin, SW Pacific)
The southern end of the North Fiji backarc basin is a volcanically active submarine boundary between the backarc basin spreading centre, the Vanuatu trench and the Hunter Ridge. This area has been mapped and sampled during the SS10/2004 and SS08/2006 voyages of the R/V Southern Surveyor. The processes occurring during southward propagation of the backarc basin involve sequential formation of 50-100 km long rift zones that split the crust of the Hunter Ridge along its strike. Initially such rift zones form as narrow (2-3 km wide) well-defined grabens with incipient magmatic activity. Then the graben widens to form a robust magmatic rift with clear volcanic centres and incipient spreading. Finally, the rift undergoes rotation and evolves into a N-S aligned robust spreading segment. Volcanic rocks erupted during rifting within each segment include a large spectrum of subduction-related magma compositions which vary from incompatible element-depleted to strongly enriched, with boninitic, tholeiitic and adakitic affinities. The rift lavas often display clear petrographic and geochemical evidence for extensive mixing between high-Mg adakitic and backarc basin magmas. Many of the samples are very primitive high-MgO olivine-phyric lavas representing the entire spectrum of the geochemical magma types. Olivine phenocrysts in most samples have a range of compositions from Fo> 92 to Fo < 85 and contain abundant primary melt inclusions. Olivine grains with melt inclusions have been experimentally reheated in a Vernadsky-type heating stage under visual control and inclusions quenched to glass and exposed for chemical analysis. Inclusions and host olivines were analysed by electron microprobe, FTIR spectroscopy and LA-ICPMS for major elements, H2O and trace elements including base metals. The compositions obtained will be used to constrain magma generation and evolution process in this complex tectonic setting.
V21A-2083
Evolution processes of primitive adakite magmas revealed by melt inclusions in olivine phenocrysts from Kadavu adakites, Fiji
The 0.5-3.5 Ma adakite magmatic suite forming the Kadavu Island Group., Fiji, contains primitive high-Mg lavas [1]. These lavas have two distinctive compositions, both strongly enriched in Sr and La, with high Sr/Y (>120) and La/Yb (>30) values. The east Kadavu high-Mg adakite is relatively enriched in K2O and depleted in TiO2 and Nb, compared to the Ngaloa high-Mg adakite. The east Kadavu primitive lavas contain phenocrysts of magnesian olivine (Fo 93-84) and clinopyroxene (Mg# 92-86). We present here the results of experimental study of primary melt inclusions in olivine phenocrysts from the east Kandavu adakites. Olivine grains with melt inclusions have been experimentally reheated in a Vernadsky-type heating stage under visual control and inclusions quenched to glass and exposed for chemical analysis. Inclusions, their host olivines and clinopyroxene phenocrysts were analysed by electron microprobe and LA-ICPMS for major and trace elements. The trace element compositions of clinopyroxene indicate that they crystallised from an adakite melt and are not xenocrystic. The major and trace element compositions of melt inclusions in olivine also have typical adakite compositions demonstrating that olivine is not xenocrystic either. The results show that the extent of the adakite geochemical signature (eg, Sr/Y values) is the highest in the most magnesian compositions (300) and continuously decreases in more evolved compositions (100). This effect cannot result from olivine+clinopyroxene crystallisation. The major element compositions of melt inclusions are also inconsistent with primitive melts that evolved to typical Kadavu adakites via olivine+clinopyroxene crystallisation. The results suggest that the primitive Kadavu adakite melts studied evolved via the dissolution-reaction-mixing processes [2] within the Kadavu volcanic plumbing system. These processes are likely to occur at the margins of primitive magma bodies, where magma is in contact with cooler semi-solidified more evolved adakitic crystal mush zones The existence of such processes is confirmed by the common presence of primitive enclaves in more evolved adakite lavas. [1] Danyushevsky L.V., Falloon T.J., Crawford A.J., Tetroeva S.A., Leslie R., Verbeeten A. (2008) Geology 36, 499-502. [2] Danyushevsky L.V., Leslie R., Crawford A.J.,.Durance P. (2004) JPetrology, 45, 2531-2553.
V21A-2084
Volatiles in Olivine Hosted Melt Inclusions from Small Eruptive Centers of the Central Southern Volcanic Zone (CSVZ), Chile
This study is a geochemical analysis of basaltic scoria cones of the CSVZ, where bulk rock analyses have revealed two predominant lava types. We consider the relationship of cones to nearby stratovolcanos and the proximity of both to the Liquine-Ofqui Fault Zone to understand how primitive melts with wide ranging compositions are generated. We characterize melt inclusions to see if they show similar trends as bulk rocks and to evaluate degassing of melts during ascent and crystallization. We report new CO2, H2O, F, S, and Cl data collected by SIMS and FTIR. Type 1 and Type 2 lavas have been identified in prior studies. Type 1 rocks have significant 238U/232Th excesses and higher abundances of fluid mobile elements. Type 2 lavas and scorias have higher LREE/HREE ratios, lower abundances of fluid mobile elements, lower 87Sr/86Sr and U- series isotopes near secular equilibrium. Type 1, usually associated with composite volcanic centers, erupt at one cone in our field area, while Type 2 is found only at other individual cones aligned with the LOFZ. We characterize olivine-hosted melt inclusions with emphasis on volatile element analyses to test the hypotheses of flux melting beneath Type 1 centers versus a lesser extent of decompression melting beneath Type 2 centers. Melt inclusions show the following overall ranges of volatiles: CO2: below detection-1208 ppm; H2O(Total): 0.11-3.9 wt percent: F: 52-1442 ppm; S: 34-2884 ppm, and Cl: 177-1606 ppm. Inclusions from lavas are generally more degassed when compared to those from scoria for CO2 and H2O. Some Cl and F values are lower in inclusions from lavas, but not all. Sulfur is variably degassed in melt inclusions from both lavas and scoria. Melt inclusions from Type 2 Caburgua scoria show evidence for variable degassing and crystallization pressures. One subset of the inclusions were trapped at low pressures (P < 500 bars, CO2 < 200 ppm, H2O < 1wt %), whereas the other subset were trapped at higher pressure (1000-2000 bars), yet have variable H2O and CO2 (1.5-4.0 wt % and 300-800 ppm and all under 1100 bars). Type I San Jorge melt inclusions also show evidence of variable degassing and trapping pressures. Some inclusions trend higher in CO2 at constant H2O values up to 541 ppm CO2 and 0.5 wt % H2O. Other inclusions trend towards higher H2O with high CO2 and have almost linear progression running from a low of 2.2 wt % H2O and 143 ppm CO2 to 3.9 wt % H2O and 648 ppm CO2. All lie between 1-2 kbar. These degassing phenomena may complicate testing of our hypotheses, however major, minor and trace element data for the melt inclusions are also under consideration.
V21A-2085
Pilot study of melt inclusions in chromites of the Merensky Reef reflecting melt compositions at Karee Mine, western Bushveld Complex, South Africa
Various melt inclusions in chrome spinel crystals of tempered samples from the lower chromitite stringer of the 'Merensky Reef' at Karee Mine in the western Bushveld Complex, were investigated to obtain an indication of the initial melt composition of the upper Critical Zone. The material was heated to 1300°C in a furnace under a graphite buffered atmosphere for at least 12 h and was rapidly quenched cold air thereafter. High resolution microscopy and SEM-EDX reveals that melt inclusions consist essentially of a glass matrix, sulphide phases and silicate phases of predominantly olivine composition. Both minerals are embedded into an amorphous melt phase. The mineral inclusions are chemically highly variable but the composition of the quenched glass phase is homogeneous. The rock matrix surrounding the chrome spinel crystals consists of bytownitic plagioclase and melt of basaltic composition. During the rapid cooling by quenching 'blade textured' plagioclases were formed. Chrome spinels display chemical variability with rim positions having higher Al2O3 and lower Cr2O3 contents than grain centres as optically evident by change in colour. The reason could be a exchange of the chromite crystal with the silicate melt in the process of reaching a diffusive equilibrium, resulting in a substitution of Cr3+ by Al3+. The sulfide phase is very heterogeneous containing mineral phases such as pyrrhotite, pentlandite, bornite, trevorite, daubreelite, taenite with variable elemental proportions. The finding of sulfide next to sulfur-undersaturated phases in the silicate inclusions does not allow any conclusion on origin of the sulfur neither on the state of equilibrium of the melt. SEM-EDX analyses prove the existence of two compositionally distinct groups of melt inclusions. The first are part of a melt batch which exsolved sulfidic, silicate and oxidic melt phases, as evident from the co-existence of sulfidic next to silicate inclusions in chromites. The second group has a basaltic composition and constitutes the matrix to the chromite crystals; it probably formed from a homogenized melt after magma mixing. Both melt compositions show a chemical affinity to Bushveld B1 and B2 magmas respectively.