Edmond A. Mathez

Sulfide melt-silicate melt distribution coefficients for noble metals and other chalcophile elements as deduced from MORB: Implications for partial melting

Abstract Analyses of chalcophile elements in coexisting sulfide and glass of basalts from the FAMOUS area of the mid-Atlantic ridge have been obtained by combined instrumental and radiochemical neutron activation analysis and directly coupled plasma spectrometry of hand-picked separates. In one sulfiderich sample, 526-1, both the glass and sulfides were analyzed for the noble metals. The sulfide meltsilicate melt Nernst partition coefficients determined for Au and Ir are 1.5–1.9 × 10 4 and 1.2–1.6 × 10 4 , respectively. That for Pd is estimated to be 3.5 × 10 4 . Partition coefficients for Ni are within the range 500–900, those for Co are between 26–51 and a single determination for Cu yields a value of 1383. The distribution coefficients for Ni, and possibly Cu and Co, are shown to be compositionally dependent. The data have been used to model the behavior of chalcophile elements during generation of MORB in its source region. The model assumes that all the platinum group elements, Se, Cu and S in the mantle reside in sulfides and that sulfide-saturated melt in the mantle contains 800 ppm S. Starting with a bulk upper mantle composition based on data from mantle xenoliths, 20–23% batch partial melting yields a silicate melt and a residual mantle sulfide that, with the exception of Ir, have compositions similar to sample 526-1 glass and sulfide, respectively. This implies that sulfide similar to that in 526-1 is left as a residue of partial melting to produce MORB and that primitive MORB compositions are determined by their equilibration with that sulfide. In contrast to MORB glass, the relative abundances of the platinum group elements, Se, Cu, and S in boninites closely resemble those of the sulfide that separated from MORB. This supports a model which holds that while the relative PGE abundances of boninite are controlled by mantle sulfide, no sulfide residue remains after the generation of boninite magma. That the modelling fails to account for Ir abundances supports the possibility that the distribution of Ir is controlled by a phase(s) in addition to sulfide.

Chloride and water solubility in basalt and andesite melts and implications for magmatic degassing

Abstract The solubilities of chloride (Cl − ) and H 2 O in aluminosilicate melts of basalt, andesite, and latite compositions saturated in aqueous vapor and/or hydrosaline liquid were determined at 2000 bars and ≈1 bar by melting mixtures of NaCl, KCl, H 2 O, and natural and synthetic rock powders and by measuring Cl − and H 2 O in the run product glasses. The abundances of Cl − in several of the aqueous run product liquids were also measured, and the partitioning of Cl − between aqueous vapor and silicate melt was determined for these experiments. Chloride is highly soluble in H 2 O-poor melts. Maximum Cl − solubilities range from 2.9 wt.% in molten basalt to 1 wt.% in molten latite at relatively high oxygen fugacities, 1040°C to 1210°C, and 2000 bars. The solubility of Cl − varies directly with pressure and the molar ((Al+Na+Ca+Mg)/Si) ratio of aluminosilicate melts. Chloride solubility in basalt melt is an order of magnitude greater than that in silicic melts, so the role of Cl − in driving the exsolution of vapor and/or liquid from magma will increase dramatically as mafic, H 2 O- and CO 2 -undersaturated magmas fractionate and evolve to more silicic compositions. The solubility of H 2 O in silicate melts saturated in aqueous vapor and/or hydrosaline liquid varies inversely with Cl − content. Chloride has little effect on H 2 O solubility with up to about 1.9 wt.% Cl − in melt because the coexisting vapor phase contains little Cl − . Hydrosaline liquid is stable with higher Cl − contents in melt, and H 2 O solubility is highly sensitive to Cl − content at these conditions. This relationship is a result of highly nonideal mixing of H 2 O and Cl − at magmatic temperatures; in several Cl − -enriched andesite experiments, immiscible vapor and hydrosaline liquid are apparently stable instead of a single Cl − -bearing volatile phase. At 2000 bars, Cl − -bearing aqueous vapor exsolves with − in the andesite melt, vapor and hydrosaline liquid exsolve with 1 to 2 wt.% Cl − and 2 O in melt, and only hydrosaline liquid exsolves if the andesite melt contains deletion ≥2 wt.% Cl − and 2 O. At 2000 bars and temperatures near 1100°C, the distribution coefficients [D Cl = (wt.% Cl − in aqueous vapor/wt.% Cl − in silicate melt)] for basalt and andesite range from 0.9 to 6 for coexisting aqueous vapors containing 1 to 11 wt.% Cl − , respectively. Silicate melt inclusions in phenocrysts from most basalts and andesites contain − implying that, at these conditions, only Cl-bearing vapor (not vapor and hydrosaline liquid) will exsolve from most basalt and andesite magmas and that the Cl − contents of the aqueous vapors will be

Experimentally determined sulfide melt-silicate melt partition coefficients for iridium and palladium

Abstract Sulfide melt-silicate melt partition coefficients for Ir and Pd have been determined from experiments run in a piston-cylinder apparatus at 1450°C, 8 kbar and under ƒ O 2 ƒ S 2 conditions appropriate for mafic magmas. Preferred values of DIr and DPd [ D = ( wt % silicate melt ) ( wt % silicate melt ) ] are 3.5·104 and 3.4·104, respectively. This validates the common assumption that sulfide melt dominates the geochemistry of the platinum-group elements in igneous processes. It also indicates that in magmatic differentiation processes in which sulfide alone controls precious metal abundance, Pd and Ir should not significantly fractionate from each other. The Pd results are consistent with those deduced from analyses of coexisting MORB sulfide and glass and with other published experiments. No dependence of DPd on ƒ O 2 ƒ S 2 was observed. For Ir, disparate results were obtained depending on the nature of the starting material. Experiments using natural silicate glass as a starting material yielded consistent values of DIr in the range 2.5·104–5.4·104. The experiments included a compositional convergence, in which the Ir-bearing sulfide produced in one experiment was used as the starting material for the second experiment. The results using natural starting compositions are similar to the value of DIr deduced from analyses of MORB. In contrast, all experiments in which synthetic silicate starting compositions were used yielded erratic values of DIr from 13·104 to 152·104. There are several possibilities to account for these high and variable values: (1) an unidentified trace or minor element present in the natural compositions but absent in the synthetic ones might complex with Ir in the silicate melt and enhance its solubility, (2) this element might enhance reaction rate, or (3) an Ir-rich quench phase might have formed in the experiments using synthetic compositions but not in the natural ones and been removed during preparation of the charges for analysis. All of these possibilities suggest that the experiments using synthetic compositions are not appropriate for determining the behavior of Ir in nature. The preferred, experimentally-determined values of DIr and DPd do not explain the fractionation of these elements observed in natural systems. The observed relative and absolute abundances of Ir and Pd reflect either concentration of Ir or Pd in a phase other than sulfide or redistribution of the metals subsequent to their initial concentration.

Magmatic metasomatism and formation of the Merensky reef, Bushveld Complex.

The rare earth element (REE) contents of pyroxenes and other minerals from the Merensky reef and stratigraphically adjacent rocks of the Atok section, Bushveld Complex, have been determined with the ion microprobe. Merensky reef clinopyroxene and orthopyroxene contain much higher and more variable concentrations of the REE than their cumulus counterparts in rocks several meters below the reef. Chondrite-normalized Merensky clinopyroxene Ce contents vary from ≈10 to 90 for Ce and from ≈4 to 17 for Yb. They also possess deep, negative Eu anomalies, the Eu anomalics being deeper for crystals having high REE contents and relatively shallow for pyroxenes with low REE contents. Similar compositional characteristics are displayed by Cl-rich apatite, which is an accessory phase in the rocks. Interstitial pyroxene in cumulates above and below the reef also tends to have elevated REE contents and in general is not in equilibrium with coexisting cumulus minerals. The melt from which the cumulus minerals crystallized falls within the compositional range of continental basalts; that from which Merensky and postcumulus pyroxenes crystallized is inferred to be much more highly enriched in REE than any normal tholeiitic or alkalic basalt. Despite their highly evolved nature in terms of the REE, the Merensky reef pyroxenes are not evolved in terms of major elements. The decoupling of incompatible trace and major elements is best explained by a metasomatic process. It is speculated that metasomatism involved upward percolation of hydrated silicate melt through and its reaction with the crystalline cumulate pile. The fact that the rocks enriched in the platinum group elements are also those that show evidence for metasomatism suggests that these elements were also metasomatically redistributed.

Petrogenesis of the Merensky Reef in the Rustenburg section of the Bushveld Complex

A petrogenetic model for the Merensky Reef in the Rustenburg section of the Bushveld Complex has been developed based on detailed field and petrographic observations and electron microprobe data. The model maintains that the reef formed by reaction of hydrous melt and a partially molten cumulate assemblage. The model is devised to account for several key observations: (1) Although the dominant rock type in the Rusterburg sections is pegmatoidal feldspathic pyroxenite, there is a continous range of reef lithology from pyroxenite to pegmatoidal harzburgite and dunite, and small amounts of olivine are present in nearly all pegmatoids. (2) The pegmatoid is usually bounded above and below by chromitite seams and the basal chromitite separated from underlying norite by a centimeter-thick layer of anorthosite. The thicknesses of the two layers exhibit a well-defined, positive correlation. (3) Inclusion of pyroxenite identical to the hanging wall and of leuconorite identical to the footwall are present in the pegmatoid. The leuconorite inclusions are surrounded by thin anorthosite and chromitite layers in the same sequence as that at the base of the reef. (4) Chromite in seams adjacent to plagioclase-rich rocks is characterized by higher Mg/Mg+Fe and Al/R3 and lower Cr/R3 than that in seams adjacent to pyroxene-rich rocks. Similar variations in mineral compositions are observed across individual chromitite seams where the underlying and overlying rock types differ. The chromite compositional variations cannot be rationalized in terms of either fractional crystallization or reequilibration with surrounding silicates. It is proposed that the present reef was originally a melt-rich horizon in norite immediately overlain by relatively crystallized pyroxenite. Magmatic vapor generated by crystallization of intercumulus melt migrated upward through fractures in the cumulate pile below the protoreef. The melt-rich protoreef became hydrated because fractures were unable to propagate through it and because the melt itself was water-undersaturated. Hydration of the intercumulus melt was accompanied by melting, and the hydration/melting front migrated downward into the footwall and upward into the hanging wall. In the footwall melting resulted first in the dissolution of orthopyroxene and then of plagioclase. With continued hydration chromite was stabilized as melt alumina content increased. The regular variations in chromite compositions reflect the original gradients in melt composition at the hydration front. The stratigraphic sequence downward through the base of the reef or pegmatoid (melt)-chromitite-anorthosite-norite represents the sequence of stable mineral assemblages across the hydration/melting front. The sequence is shown to be consistent with knowledge gained from experiments on melting of hydrous mafic systems at crustal pressures. With cooling the hydrated mixture from partial melting of norite footwall and more mafic hanging wall crystallized in the sequence chromite-olivine-pyroxene-plagioclase, with peritectic loss of some olivine. Calculations of mass balance indicate that a significant proportion of the melt was lost from the melt-rich horizon. Variations in the development of the pegmatoid and associated lithologies and amount of modal olivine in the pegmatoids along the strike of the Merensky Reef resulted because the processes of hydration, melting and melt loss operated to varying extents.

Sulfide melt-silicate melt distribution coefficients for nickel and iron and implications for the distribution of other chalcophile elements

Distribution coefficients for Fe and Ni between sulfide and silicate melts have been determined as functions of f;O2 and f;S2. Experiments were carried out in a piston-cylinder apparatus at 1450°C and 8 kbars. In a system at sulfide saturation the f;S2, f;O2 and aFeO are interdependent; therefore, it was possible to vary f;O2/f;S2 systematically by changing the bulk Fe content of the starting materials. This was accomplished in two ways, first by varying the sulfide to silicate mass ratio and second by synthesizing silicate starting materials of differing FeO contents. The f;O2/f;S2 was further constrained by the choice of graphite capsule material combined with pre-oxidation of the starting silicate. This produced melts which were saturated or nearly saturated in C-O-S vapor. The f;O2 and f;S2 were computed from the independent relationships of f;O2/f;S2, with vapor and silicate melt compositions. The molar distribution coefficients of Ni and Fe, D(Nimol) (=XsulfNiSXsilNio) and D(Femol), exhibit strong dependencies on the molar S concentration of the sulfide melt and FeO content of the silicate melt. D(Nimol) ranges from 1458 to 8857 for sulfide melt S concentrations of 45.8 to 51.1 mol%, respectively. Negative correlations of log D(Nimol) and log D(Femol) with log (f;O2/f;S2) are consistent with reactions between sulfide and oxide components of these elements in both melts and indicate that the controlling equilibrium is MOsil + 1/2 S2 = MSsulf + 1/2 O2, (M = metal). The different values of D(Niwt) reported in the literature can be rationalized in terms of differences in f;O2/f;S2 of the experiments. It is suggested that all chalcophile elements, including PGEs, behave similarly. In particular, it is demonstrated that D(Irwt) and D(Pdwt) vary directly with the S content of the sulfide melt and indirectly with f;O2 in published experiments. For the possible range in f;O2/f;S2 of sulfide-saturated magmas in nature, variations in Ds for Ni and the other chalcophile elements of more than one order of magnitude may exist.

Lead isotopic disequilibrium between sulfide and plagioclase in the bushveld complex and the chemical evolution of large layered intrusions

Abstract The Pb isotopic compositions of coexisting plagioclase and sulfide from the Bushveld Complex were determined by laser ablation multi-collector ICPMS (LA MC-ICPMS). The samples are of the upper Critical Zone in the northeast corner of the Complex and were collected from drill core and underground mine exposures. All the rocks are fresh and exhibit no evidence for alteration, weathering, or disruption of the Pb isotope systematics subsequent to the initial cooling of the intrusion. Furthermore, individual plagioclase and sulfide crystals do not contain enough U to warrant correction for radiogenic in-growth. For these reasons, the measured Pb isotope ratios approximate the initial ones. For plagioclase, 207Pb/206Pb ranges from 0.98 to 1.02 and 208Pb/206Pb from 2.26 to 2.35. Low 207Pb/206Pb and 208Pb/206Pb ratios characterize grain boundaries and partially annealed microcracks, some of which contain minute fragments of sulfide and other phases, and this accounts for most, if not all, the heterogeneity exhibited by individual samples. Real compositional differences exist, however, in plagioclase from different lithologic layers. For example, plagioclase 207Pb/206Pb values vary from 1.004 in norite beneath the Merensky pyroxenite to 1.009 in the mineralized pyroxenite, and 0.997 in overlying norite. In most samples in which sulfide and plagioclase coexist, the sulfide 207Pb/206Pb ratio is lower and 208Pb/206Pb ratio higher than the corresponding ones in plagioclase. For example, in a mineralized Merensky reef sample, average sulfide 207Pb/206Pb and 208Pb/206Pb ratios are 0.993 and 2.313, respectively, while those in plagioclase are 1.000 and 2.292. In one sample, the sulfide is extremely heterogeneous, with 207Pb/206Pb and 208Pb/206Pb ratios as low as 0.84 and 2.12. In this particular sample, the compositions must represent an isolated occurrence of addition of a young Pb component. The array of sulfide and plagioclase compositions requires multiple sources of Pb at the time of crystallization or soon thereafter. The disequilibrium between plagioclase and sulfide implies that some of the Pb originated from the isotopically distinct country rocks and was introduced at temperatures at which the composition of sulfide but not plagioclase could be modified. Thus, Bushveld sulfide, and to some extent plagioclase, do not reliably record the initial Pb isotopic composition(s) of the parent magma(s).

Carbon isotopic composition and origin of SiC from kimberlites of Yakutia, Russia

Abstract The stability of moissanite (SiC) has been computed for upper mantle conditions using the internally optimized thermodynamic dataset for the MgSiO compounds of Fei et al. (1990). The computations consider the effects of pressure and temperature on the elastic properties of phases involved in the reactions. The maximum stability of moissanite throughout the upper mantle is typically five to six orders of magnitude lower in oxygen fugacity (fO2) than the Fe metal-wustite oxygen buffer at equivalent temperature and pressure, in agreement with previous calculations. Under conditions of SiC stability, silicates will be Fe-free, Fe metal will contain substantial amounts of Si but little C in solution, and Mg-rich sulfides will be stable. Moissanite from the heavy mineral concentrate of the Mir and Aikhal kimberlite pipes, Yakutia, has been studied. Moissanite crystals are gemmy and vary in color from a characteristic blue-green to pale green to nearly colorless to blue-black. Most exhibit crystallographic faces and are in the size range 0.5 to 1 mm in long dimension. Their compositions include small quantities of Fe, which is ubiquitous, Al, Ca, V, Cr, and Mn, all of which may be present in concentrations > 100 ppmwt. Mineral inclusions are present in some crystals. Silicon metal is the most common; inclusions of ferrosilicite (Fe3Si7), FeTi silicides, REE silicate, and sinoite (Si2N2O) have also been observed. The carbon isotopic compositions of individual moissanite grains have been determined by ion microprobe. The nine analyzed crystals from Aikhal and fourteen from Mir are characterized by a narrow range in δ 13C values of −22 to −29‰; the majority of crystals fall within a more restricted range of −24 to −27‰. Two grains were analyzed for N and found to have a δ 15N of +9.7 ± 4.0 and +5.6 ± 2.0‰. Five mechanisms for the formation of moissanite are considered. Moissanite may be a relict of a reduced, primordial Earth and now present only as a trace phase in an otherwise oxidized mantle. Alternatively, there may be present-day global regions of the Earth that are both highly reduced and characterized by light carbon isotopic compositions. Although these possiblities cannot be disproved, they are not supported by observations. Two other possibilities, namely that moissanite stability extends to more oxidized conditions at pressures of the lower mantle or that it may form metastably, cannot be evaluated with present knowledge. The possibility most consistent with, although not proven by, the isotopic data is that moissanite formed by metamorphism of reduced, carbonaceous sediments during subduction.

Carbonaceous matter in peridotites and basalts studied by XPS, SALI, and LEED

Abstract Carbonaceous matter in peridotite xenoliths and basalt from the Hualalai Volcano, in a basalt glass collected directly from an active lava lake on the east rift of Kilauea, in garnet and diopside megacrysts from the Jagersfontein kimberlite, and in gabbros from the Stillwater and Bushveld Complexes has been studied by X-ray photoelectron spectroscopy (XPS), thermal-desorption surface analysis by laser ionization (SALI), and low-energy electron diffraction (LEED). The basalt and two of the four xenoliths from Hualalai and both Jagersfontein megacrysts yielded trace quantities (≤ 10 nanomoles) of organic compounds on heating to 700°C. Organics were not detected in the rocks from the layered intrusions, and neither carbonaceous matter nor organics were detected in the glass from the lava lake. Where detected, organics appear to be associated with carbonaceous films on microcrack surfaces. Carbonaceous matter exists as films less than a few nm thick and particles up to 20 μm across, both of which contain elements expected to be present in significant quantities in magmatic vapors, namely Si, alkalis, halogens, N, and transition metals. LEED studies suggest that the carbonaceous films are amorphous. The data suggest two possible mechanisms for the formation of the organics. One is that they are a product of abiotic heterogeneous catalysis of volcanic gas on new, chemically active mineral surfaces formed by fracturing during cooling. Alternatively, organics may have been assimilated into the volcanic gases prior to eruption and then deposited on cracks formed during eruption and cooling. In any case, there is no evidence to suggest that the organics represent laboratory or field biogenic contamination.

Carbon isotopes in xenoliths from the Hualalai Volcano, Hawaii, and the generation of isotopic variability

Abstract The isotopic composition of carbon has been determined in a suite of xenoliths from lava of the 1800–1801 Kaupulehu eruption of Hualalai Volcano, Hawaii. Several lithologies are represented in the suite, including websterite, dunite, wehrlite, pyroxenite, and gabbro. In addition, there are composite xenoliths in which contacts between lithologies are preserved. Most of the xenoliths represent deformed cumulates. The contact relations in the composite samples indicate that the lithologies originated from the same source region, which, based on pressures determined from fluid inclusions, is estimated to be at a depth of ≈20 km, or near the crust-mantle boundary. Samples were heated in steps from 200 to 1475°C to obtain separation of the different carbonaceous phases, and the isotopic composition of carbon released at each step was determined. Grossular glass was found to be a suitable flux to fuse refractory samples. Upon heating, carbon exhibits the typical bimodal evolution behavior observed in other studies of xenoliths and basalts. Carbon extracted from all samples at temperatures below 900°C is characterized by a δ 13 C of about −25%. vs PDB and is thought to be composed dominantly of graphitic and organic material, which is known to be present on virtually all cracks. The δ 13 C of the carbon fraction extracted at 1200°C and above from wehrlite and dunite is in the range−1.5%. to 5.2%., whereas that extracted from websterite ranges from −22%. to −26‰ Similarly, in one composite sample, the compositions of dunite and websterite were found to be −2.4%. and −7.0%., respectively. The large difference can be associated with specific petrographic features unique to each lithology. In wehrlite and dunite, carbon exists mostly as CO 2 -rich inclusions in arrays representing partially annealed microcracks. The websterite xenoliths contain megascopic zones of large, irregularly-shaped inclusions. The zones traverse entire thinsections and are interpreted to represent fractures annealed at depth. Most of the carbon is believed to exist in the inclusion-rich zones and to consist of carbonaceous material precipitated from fluid. The observations and isotopic results demonstrate that isotopic variability can be generated by multistage fractionation processes such as degassing of CO 2 from magma and precipitation of CO 2 -rich fluids to form graphitic compounds. Such processes operated over regions the scales of which were determined by style and intensity of deformation and by lithology.