Jon D Blundy

Non-ideal interactions in calcic amphiboles and their bearing on amphibole-plagioclase thermometry

AbstractAmphibole thermodynamics are approximated with the symmetric formalism (regular solution model for within-site non-ideality and a reciprocal solution model for cross-site-terms) in order to formulate improved thermometers for amphibole-plagioclase assemblages. This approximation provides a convenient framework with which to account for composition-dependence of the ideal (mixing-on-sites) equilibrium constants for the equilibria:A)edenite+4quartz=tremolite+albiteB)edenite+albite=richterite+anorthite For A and B all possible within-site and cross-site interactions among the species □−K−Na−Ca−Mg−Fe2+−Fe3+−Al−Si on the A, M4, M1, M3, M2 and T1 amphibole crystallographic sites were examined. Of the 36 possible interaction energy terms, application of the symmetric formalism results in a dramatic simplification to eight independent parameters. Plagioclase nonideality is modelled using Darkens quadratic formalism. We have supplemented an experimental data set of 92 amphibole-plagioclase pairs with 215 natural pairs from igneous and metamorphic rocks in which the pressure and temperature of equilibration are well constrained. Regression of the combined dataset yields values for the eight interaction parameters as well as for apparent enthalpy, entropy and volume changes for each reaction. These parameters are used to formulate two new thermometers, which perform well (±40°C) in the range 400–1000°C and 1–15 kbar over a broad range of bulk compositions, including tschermakitic amphiboles from garnet amphibolites which caused problems for the simple thermometer of Blundy and Holland (1990). For silica-saturated rocks both thermometers may be applied: in silica-undersaturated rocks or magmas thermometer B alone can be applied. An improved procedure for estimation of ferric iron in calcic amphiboles is presented in the appendix.

SIMS determination of trace element partition coefficients between garnet, clinopyroxene and hydrous basaltic liquids at 2–7.5 GPa and 1080–1200°C

Abstract Trace element partition coefficients (Ds) for up to 13 REE, Nb, Ta, Zr, Hf, Sr and Y have been determined by SIMS analysis of seven garnets, four clinopyroxenes, one orthopyroxene and one phlogopite crystallized from an undoped basanite and a lightly doped (200 ppm Nb, Ta and Hf) quartz tholeiite. Experiments were conducted at 2–7.5 GPa, achieving near-liquidus crystallization at relatively low temperatures of 1080–1200°C under strongly hydrous conditions (5–27 wt.% added water). Garnet and pyroxene DREE show a parabolic pattern when plotted against ionic radius, and conform closely to the lattice strain model of Blundy and Wood (Blundy, J.D., Wood, B.J., 1994. Prediction of crystal–melt partition coefficients from elastic moduli. Nature 372, 452–454). Comparison, at constant pressure, between hydrous and anhydrous values of the strain-free partition coefficient (D0) for the large cation sites of garnet and clinopyroxene reveals the relative importance of temperature and melt water content on partitioning. In the case of garnet, the effect of lower temperature, which serves to increase D0, and higher water content, which serves to decrease D0, counteract each other to the extent that water has little effect on garnet–melt D0 values. In contrast, the effect of water on clinopyroxene–melt D0 overwhelms the effect of temperature, such that D0 is significantly lower under hydrous conditions. For both minerals, however, the lower temperature of the hydrous experiments tends to tighten the partitioning parabolas, increasing fractionation of light from heavy REE compared to anhydrous experiments. Three sets of near-liquidus clinopyroxene–garnet two-mineral D values increase the range of published experimental determinations, but show significant differences from natural two-mineral Ds determined for subsolidus mineral pairs. Similar behaviour is observed for the first experimental data for orthopyroxene–clinopyroxene two-mineral Ds when compared with natural data. These differences are in large part of a consequence of the subsolidus equilibration temperatures and compositions of natural mineral pairs. Great care should therefore be taken when using natural mineral–mineral partition coefficients to interpret magmatic processes. The new data for strongly hydrous compositions suggest that fractionation of Zr–Hf–Sm by garnet decreases with increasing depth. Thus, melts leaving a garnet-dominated residuum at depths of about 200 km or greater may preserve source Zr/Hf and Hf/Sm. This contrasts with melting at shallower depths where both garnet and clinopyroxene will cause Zr–Hf–Sm fractionation. Also, at shallower depths, clinopyroxene-dominated fractionation may produce a positive Sr spike in melts from spinel lherzolite, but for garnet lherzolite melting, no Sr spike will result. Conversely, clinopyroxene megacrysts with negative Sr spikes may crystallize from magmas without anomalous Sr contents when plotted on mantle compatibility diagrams. Because the characteristics of strongly hydrous silicate melt and solute-rich aqueous fluid converge at high pressure, the hydrous data presented here are particularly pertinent to modelling processes in subduction zones, where aqueous fluids may have an important metasomatic role.

Heavy REE are compatible in clinopyroxene on the spinel lherzolite solidus

Abstract Trace element partitioning between clinopyroxene and melt was investigated experimentally under conditions appropriate to near-solidus melting of spinel lherzolite in the upper mantle. Starting material was a high-Na, Al basalt glass previously shown to be a very low degree (∼1%) partial melt of spinel lherzolite at 1.5 GPa, 1269°C [Robinson et al., Earth Planet. Sci. Lett., in press]. The experiment was run with a spinel seed under sub-liquidus conditions (1255°C) to ensure clinopyroxene crystallisation. The experimental clinopyroxene composition is consistent with equilibrium close to the solidus of fertile mantle lherzolite, most notably in its high contents of Ca-Tschermaks (CaTs) molecule (∼22 mol %) and Na2O (1.4 wt%). Clinopyroxene–melt partition coefficients (D) for a wide range of trace elements, determined by SIMS analysis of run products, differ markedly from those reported in other studies under conditions less appropriate to mantle melting. In particular partition coefficients for the heavy rare earth elements (Gd–Lu) are greater than unity (e.g. DLu=1.45), and the critical partitioning parameter, (DSm/DNd)×(DHf/DLu), is 0.68. These features, which arise due to the high CaTs content of the clinopyroxene, dramatically reduce the required involvement of garnet in the melting region beneath mid-ocean ridges.

Partitioning of trace elements between crystals and melts

Abstract Advances in analytical geochemistry have made it possible to determine precisely the concentration of many trace elements and their isotopes in rocks. These data provide the cornerstone for geochemical models of the Earth and terrestrial planets. However, our understanding of how trace elements behave has not kept pace with the analytical advances. As a result, geochemists are often hampered in their interpretation of geochemical data by an incomplete knowledge of trace element partitioning under the conditions of interest. Through advances in trace element microbeam analysis it is now possible to determine partition coefficients experimentally under important conditions, such as during melting of the crust and mantle. This large body of experimental data can be used to investigate the fundamental controls on element partitioning. Simple continuum theories of elastic strain and point charges in crystal lattices, that account, respectively, for mismatch in ionic radius and ionic charge between the substituent trace ion and the lattice site on which it is accommodated, provide a very useful theoretical framework. This approach can be used as the basis for quantitative models of trace element partitioning, in terms of pressure, temperature, redox state and composition, and as a means of predicting partition coefficients for elements not routinely analysed. Experimental studies of partitioning are supported by atomistic computer simulations at zero K. Developments in computational techniques that enable direct simulation of high temperature and pressure mineral-melt partitioning will revolutionise the field in the near future. Use of novel, spectroscopic techniques to probe the structural environment of trace elements in crystals and glasses will provide valuable new data for computational and theoretical models. Extension of high temperature partitioning theory to ambient conditions is an essential step in understanding current climate change proxies, and tackling a host of environmental problems.

Crystal-chemical controls on trace element partitioning between garnet and anhydrous silicate melt

We report new findings related to sample 14161,7373, the first analyzed lunar intrusive rock to show textural evidence for the formation of granitic material by silicate liquid immiscibility (SLI) and geochemical evidence for separation of that material, at least on a small scale, from mafic residua. Ion microprobe analyses of rare-earth element (REE) concentrations in silicates record the compositional evolution of this late-stage assemblage. For a sample whose bulk assemblage, that of whitlockite monzogabbro, has REE concentrations at ~5–6× KREEP levels, the silicates have only moderately high REE concentrations. The calculated REE concentrations of melt in equilibrium with pigeonite, after re-integration of pigeonite host and augite lamellae, is approximately 2× high-K KREEP. The high bulk REE concentrations of the assemblage are the result of a high proportion of whitlockite, probably an excess, i.e., the assemblage is a whitlockite cumulate. Variations in silicate REE concentrations reflect the co-crystallization of whitlockite, which we interpret to have begun crystallizing prior to the onset of late-stage immiscibility of the felsic melt fraction. Calculated cooling rates, based on a quantitative model of pyroxene exsolution, reveal a two-stage thermal history, which involved initial crystallization at a minimum depth of 700–800 m during which the primary augite and pigeonite grains unmixed to form the observed host-lamella pairs. Unusual compositional profiles close to the host-lamella interface are consistent with a second stage of mild reheating followed by rapid cooling at a depth of ~30 m, most likely in an ejecta blanket. The shallow depth of origin indicated for 14161,7373 is consistent with other studies of evolved lunar intrusive rocks, supporting the interpretation that lunar QMD and granite, as found among the Apollo samples, are late-stage differentiates of high-level intrusions and are not related to differentiation of deep crustal KREEP reservoirs.

The beginning of melting of fertile and depleted peridotite at 1.5 GPa

We have determined the solidus temperatures and liquid compositions for batch melts of fertile and depleted lherzolite in the 0–15% melting interval at 1.5 GPa. Because establishment of equilibrium is difficult at very low melt fractions we have used an iterative technique in which the liquids from sandwich-type experiments are synthesized and tested for multiple saturation at the same pressure, temperature conditions as those of the sandwich experiment. Only when all 4 mantle phases are stable and of correct composition is the result accepted as being at equilibrium. This technique permits accurate determination of solidus temperatures to ±10°C and provides the compositions of low-degree mantle melts. We find that at degrees of batch melting above 3%, liquids produced from a fertile peridotite (MORB-Pyrolite) contain ∼49% SiO2, have <4% Na2O and have Fe–Mg partitioning relationships with the solid phases which are typical of basalt. As the degree of melting drops to 0% in fertile peridotite (MORB-Pyrolite), SiO2 content of the melt increases to 53% and the Na2O content to 8%. The melts remain strongly nepheline-normative and Fe favours the melt relative to olivine even more strongly than normal. We ascribe this latter observation to Na–Fe3+ coupling in the melt at high Na2O content. These high-alkali and high-silica melts are not observed at the solidus of depleted peridotite (Tinaquillo Lherzolite). These experimental results provide the first direct tests of fractional melting models which invoke the successive extraction of low-degree partial melts. No melting model currently available provides a good description of the compositions of low-degree melts from a fertile peridotite at 1.5 GPa.

Magma heating by decompression-driven crystallization beneath andesite volcanoes

Explosive volcanic eruptions are driven by exsolution of H2O-rich vapour from silicic magma. Eruption dynamics involve a complex interplay between nucleation and growth of vapour bubbles and crystallization, generating highly nonlinear variation in the physical properties of magma as it ascends beneath a volcano. This makes explosive volcanism difficult to model and, ultimately, to predict. A key unknown is the temperature variation in magma rising through the sub-volcanic system, as it loses gas and crystallizes en route. Thermodynamic modelling of magma that degasses, but does not crystallize, indicates that both cooling and heating are possible. Hitherto it has not been possible to evaluate such alternatives because of the difficulty of tracking temperature variations in moving magma several kilometres below the surface. Here we extend recent work on glassy melt inclusions trapped in plagioclase crystals to develop a method for tracking pressure–temperature–crystallinity paths in magma beneath two active andesite volcanoes. We use dissolved H2O in melt inclusions to constrain the pressure of H2O at the time an inclusion became sealed, incompatible trace element concentrations to calculate the corresponding magma crystallinity and plagioclase–melt geothermometry to determine the temperature. These data are allied to ilmenite–magnetite geothermometry to show that the temperature of ascending magma increases by up to 100 °C, owing to the release of latent heat of crystallization. This heating can account for several common textural features of andesitic magmas, which might otherwise be erroneously attributed to pre-eruptive magma mixing.

Degassing and crystallization of ascending andesite and dacite

The prevalence of andesitic and dacitic volcanic eruptions over the past 20 years has led to a new appreciation of processes typical of magmas of intermediate composition. Extensive syn-eruptive crystallization, driven by decompression and volatile exsolution, is one such process. A water-saturated melt that is decompressed isothermally from its liquidus must crystallize in response to the diminishing capacity of the melt to retain volatiles (particularly H2O). Only rapid magma ascent allows such a melt to reach the Earths surface without crystallizing. Intermediate rates of ascent permit varying amounts of syn-eruptive crystallization, which in turn changes magma rheology and affects continued magma progress toward the surface. Feedback among magma decompression, vesiculation, and crystallization is poorly understood, particularly with regard to the kinetics of crystallization. Here we present two complementary approaches to the study of syn-eruptive, degassing-induced crystallization. The first involves projection of matrix glass compositions onto the well-understood Qz-Ab-Or ternary, which allows relative (quartz-undersaturated melt) or absolute (quartz-saturated melt) determination of magma equilibration (or ‘closure’) pressure. We show that glass composition (groundmass crystallinity) changes as a function of decompression rate, and that either very slow ascent or rapid ascent followed by arrest and shallow cooling can lead to extensive cotectic precipitation of quartz + feldspar. The second approach involves quantification of plagioclase textures, which provides a direct measurement of the relative importance of crystal nucleation and growth (J/G). This parameter can, in turn, be linked to the effective undercooling (supersaturation) experienced during decompression. Finally, we use phenocryst melt inclusion data to suggest that a substantial amount of phenocryst crystallization may also be explained by decompression of water-saturated melt.

A dearth of intermediate melts at subduction zone volcanoes and the petrogenesis of arc andesites

Andesites represent a large proportion of the magmas erupted at continental arc volcanoes and are regarded as a major component in the formation of continental crust. Andesite petrogenesis is therefore fundamental in terms of both volcanic hazard and differentiation of the Earth. Andesites typically contain a significant proportion of crystals showing disequilibrium petrographic characteristics indicative of mixing or mingling between silicic and mafic magmas, which fuels a long-standing debate regarding the significance of these processes in andesite petrogenesis and ultimately questions the abundance of true liquids with andesitic composition. Central to this debate is the distinction between liquids (or melts) and magmas, mixtures of liquids with crystals, which may or may not be co-genetic. With this distinction comes the realization that bulk-rock chemical analyses of petrologically complex andesites can lead to a blurred picture of the fundamental processes behind arc magmatism. Here we present an alternative view of andesite petrogenesis, based on a review of quenched glassy melt inclusions trapped in phenocrysts, whole-rock chemistry, and high-pressure and high-temperature experiments. We argue that true liquids of intermediate composition (59 to 66 wt% SiO2) are far less common in the sub-volcanic reservoirs of arc volcanoes than is suggested by the abundance of erupted magma within this compositional range. Effective mingling within upper crustal magmatic reservoirs obscures a compositional bimodality of melts ascending from the lower crust, and masks the fundamental role of silicic melts (≥66 wt% SiO2) beneath intermediate arc volcanoes. This alternative view resolves several puzzling aspects of arc volcanism and provides important clues to the integration of plutonic and volcanic records.

Rapid decompression-driven crystallization recorded by melt inclusions from Mount St. Helens volcano

Crystals in hydrous magmas can form in response to falling temperature (magma cooling) or degassing (magma decompression). It remains unclear which process dominates beneath explosive silicic volcanoes. Because decompression and cooling operate on very different time scales, resolving the driving force behind crystallization is of fundamental importance for determining magma dynamics and eruption hazard. Here we use ion-microprobe measurements of dissolved H 2 O in phenocryst-hosted melt inclusions from pumices erupted between May and October 1980 at Mount St. Helens volcano to show that all microlites and a significant proportion of phenocrysts were formed by near isothermal decompression. Magmas erupted after 18 May show evidence for subsequent crystallization of both phenocrysts and microlites, indicating that the time scales of crystal nucleation and growth are on the order of months or less.