Melt Inclusion Entrapment Research Articles

Phenocryst-hosted melt inclusions record stalling of magma during ascent in the conduit and upper magma reservoir prior to vulcanian explosions, Soufrière Hills volcano, Montserrat, West Indies

The mechanics of explosive eruptions influence magma ascent pathways. Vulcanian explosions involve a stop–start mechanism that recurs on various timescales, evacuating the uppermost portions of the conduit. During the repose time between explosions, magma rises from depth and refills the conduit and stalls until the overpressure is sufficient to generate another explosion. We have analyzed major elements, Cl, S, H2O, and CO2 in plagioclase-hosted melt inclusions, sampled from pumice erupted during four vulcanian events at Soufriere Hills volcano, Montserrat, to determine melt compositions prior to eruption. Using Fourier transform infrared spectroscopy, we measured values up to 6.7 wt.% H2O and 80 ppm CO2. Of 42 melt inclusions, 81 % cluster between 2.8 and 5.4 wt.% H2O (57 to 173 MPa or 2–7 km), suggesting lower conduit to upper magma reservoir conditions. We propose two models to explain the magmatic conditions prior to eruption. In Model 1, melt inclusions were trapped during crystal growth in magma that was stalled in the lower conduit to upper magma reservoir, and during trapping, the magma was undergoing closed-system degassing with up to 1 wt.% free vapor. This model can explain the melt inclusions with higher H2O contents since these have sampled the upper parts of the magma reservoir. However, the model cannot explain the melt inclusions with lower H2O because the timescale for plagioclase crystallization and melt inclusion entrapment is longer than the magma residence time in the conduit. In Model 2, melt inclusions were originally trapped at deeper levels of the magma chamber, but then lost hydrogen by diffusion through the plagioclase host during periodic stalling of the magma in the lower conduit system. In this second scenario, which we favor, the melt inclusions record re-equilibration depths within the lower conduit to upper magma reservoir.

Integrated field, satellite and petrological observations of the November 2010 eruption of Erta Ale

Erta Ale volcano, Ethiopia, erupted in November 2010, emplacing new lava flows on the main crater floor, the first such eruption from the southern pit into the main crater since 1973, and the first eruption at this remote volcano in the modern satellite age. For many decades, Erta Ale has contained a persistently active lava lake which is ordinarily confined, several tens of metres below the level of the main crater, within the southern pit. We combine on-the-ground field observations with multispectral imaging from the SEVIRI satellite to reconstruct the entire eruptive episode beginning on 11 November and ending prior to 14 December 2010. A period of quiescence occurred between 14 and 19 November. The main eruptive activity developed between 19 and 22 November, finally subsiding to pre-eruptive levels between 8 and 15 December. The estimated total volume of lava erupted is ∼0.006 km3. The mineralogy of the 2010 lava is plagioclase + clinopyroxene + olivine. Geochemically, the lava is slightly more mafic than previously erupted lava lining the caldera floor, but lies within the range of historical lavas from Erta Ale. SIMS analysis of olivine-hosted melt inclusions shows the Erta Ale lavas to be relatively volatile-poor, with H2O contents ≤1,300 ppm and CO2 contents of ≤200 ppm. Incompatible trace and volatile element systematics of melt inclusions show, however, that the November 2010 lavas were volatile-saturated, and that degassing and crystallisation occurred concomitantly. Volatile saturation pressures are in the range 7–42 MPa, indicating shallow crystallisation. Calculated pre-eruption and melt inclusion entrapment temperatures from mineral/liquid thermometers are ∼1,150 °C, consistent with previously published field measurements.

A comment on: ‘TitaniQ under pressure: the effect of pressure and temperature on the solubility of Ti in quartz’, by Jay B. Thomas, E. Bruce Watson, Frank S. Spear, Philip T. Shemella, Saroj K. Nayak and Antonio Lanzirotti

Trace concentrations of Ti in quartz are used to indicate the pressure and temperature conditions of crystallization in the ‘TitaniQ’ geothermobarometer of Thomas et al. (Contrib Miner Petrol 160:743–759, 2010). It utilises the partitioning of Ti into quartz as an indicator of the pressures and/or temperatures of crystal growth. For a given value of TiO2 activity in the system, if temperatures are inferred to ±20 °C, pressure is constrained to ±1 kbar and vice versa. There are significant contrasts, however, between the conclusions from TitaniQ and those for natural quartz (as well as other mineral phases) in volcanic rocks. Application of the TitaniQ model to quartz from the 27 ka Oruanui and 760 ka Bishop high-silica rhyolites, where the values of T, P and TiO2 activity are constrained by other means (Fe–Ti oxide equilibria, melt inclusion entrapment pressures in gas-saturated melts, melt and amphibole compositions), yields inconsistent results. If realistic values are given to any two of these three parameters, then the value of the third is wholly unrealistic. The model yields growth temperatures at or below the granite solidus, pressures in the lower crust or upper mantle, or TiO2 activities inconsistent with the mineralogical and chemical compositions of the magmas. CL imagery and measurements of Ti (and other elements) in quartz are of great value in showing the growth histories and changes in conditions experienced by crystals, but direct linkages to P, T conditions during crystal growth cannot be achieved.

The volatile contents of the Galapagos plume; evidence for H 2O and F open system behavior in melt inclusions

We investigated the geochemistry of naturally quenched, olivine-hosted melt inclusions from lavas corresponding to two of the four end-member mantle components from the Galápagos Archipelago: the high 3He/ 4He (Fernandina Is.) and the MORB-like (Santiago Is.) mantle components. Melt inclusions from both islands have a large range in chemical compositions and are trapped within olivine phenocrysts that are in FeO–MgO disequilibrium with the matrix glasses. The majority of the inclusions have major and trace element compositions similar to the matrix glasses. The volatile contents (H 2O, CO 2, F, S, Cl) and volatile/refractory trace element ratios (H 2O/Ce, CO 2/Nb, F/Nd, S/Dy, Cl/K) in these inclusions agree with previous values reported for oceanic basalt. Surprisingly, a subset of inclusions with major and trace elements contents different from the matrix glasses have water and fluorine contents that are remarkably constant and comparable to those of the matrix glasses. This produces unusual H 2O/Ce and F/Nd ratios in those inclusions. These results indicate that the inclusions have not remained closed system for H 2O or F. Models of diffusive re-equilibration for H + and FeO–MgO suggest that the inclusions and their host olivines spent days to months within the surrounding magma. These timescales, combined with the depths of melt inclusion entrapment, suggest minimum melt percolation velocities of ~ 10 − 2 to 10 − 4 m/s within the oceanic crust.

Partitioning behavior of chlorine and fluorine in the system apatite–melt–fluid. II: Felsic silicate systems at 200 MPa

Hydrothermal experiments were conducted to determine the partitioning of Cl between rhyolitic to rhyodacitic melts, apatite, and aqueous fluid(s) and the partitioning of F between apatite and these melts at ca. 200 MPa and 900–924 °C. The number of fluid phases in our experiments is unknown; they may have involved a single fluid or vapor plus saline liquid. The partitioning behavior of Cl between apatite and melt is non-Nernstian and is a complex function of melt composition and the Cl concentration of the system. Values of D Cl apat/melt (wt. fraction of: Cl in apatite/Cl in melt) vary from 1 to 4.5 and are largest when the Cl concentrations of the melt are at or near the Cl-saturation value of the melt. The Cl-saturation concentrations of silicate melts are lowest in evolved, silica-rich melts, so with elevated Cl concentrations in a system and with all else equal, the maximum values of D Cl apat/melt occur with the most felsic melt. In contrast, values of D F apat/melt range from 11 to 40 for these felsic melts, and many of these are an order of magnitude greater than those applying to basaltic melts at 200 MPa and 1066–1150 °C. The Cl concentration of apatite is a simple and linear function of the concentration of Cl in fluid. Values of D Cl fluid/apat for these experiments range from 9 to 43, and some values are an order of magnitude greater than those determined in 200-MPa experiments involving basaltic melts at 1066–1150 °C. In order to determine the concentrations and interpret the behavior of volatile components in magmas, the experimental data have been applied to the halogen concentrations of apatite grains from chemically evolved rocks of Augustine volcano, Alaska; Krakatau volcano, Indonesia; Mt. Pinatubo, Philippines; Mt. St. Helens, Washington; Mt. Mazama, Oregon; Lascar volcano, Chile; Santorini volcano, Greece, and the Bishop Tuff, California. The F concentrations of these magmas estimated from apatite–melt equilibria range from 0.06 to 0.12 wt% and are generally equivalent to the concentrations of F determined in the melt inclusions. In contrast, the Cl concentrations of the magmas estimated from apatite–melt equilibria (e.g., ca. 0.3–0.9 wt%) greatly exceed those determined in the melt inclusions from all of these volcanic systems except for the Bishop Tuff where the agreement is good. This discrepancy in estimated Cl concentrations of melt could result from several processes, including the hypothesis that the composition of apatite represents a comparatively Cl-enriched stage of magma evolution that precedes melt inclusion entrapment prior to the sequestration of Cl by coexisting magmatic aqueous and/or saline fluid(s).

Halogen diffusion in magmatic systems: Our current state of knowledge

Diffusion of halogens has the potential to influence petrogenetic processes in magma chambers and conduit degassing processes. This contribution reviews our current state of knowledge concerning halogen diffusion and the influence of halogens on the diffusion of major elements in silicate melts. The addition of halogens to silicate melts at common, natural concentration levels will have little effect on the diffusion of major elements. However, the differences between the diffusivity of water, the diffusivities of halogens, and the diffusivity of sulfur are significant enough that during melt inclusion entrapment, or during rapid bubble or crystal growth, diffusive fractionation between water and the halogens, and between halogens and sulfur, are expected to occur and can influence the compositions of melt inclusions, crystals and volcanic gases.

Pb isotopic variability in melt inclusions from the EMI–EMII–HIMU mantle end-members and the role of the oceanic lithosphere

Melt inclusions from four individual lava samples representing the HIMU (Mangaia Island), EMI (Pitcairn Island) and EMII (Tahaa Island) end member components, have heterogeneous Pb isotopic composition larger than that defined by the erupted lavas in each island. The broad linear trend in 207Pb/206Pb–208Pb/206Pb space produced by the melt inclusions from Mangaia, Tahaa and fPitcairn samples reproduces the entire trend defined by the Austral chain, the Society islands and the Pitcairn island and seamount groups. The inclusions preserve a record of melt composition of far greater isotopic diversity than that sampled in whole rock basalts. These results can be explained by mixing of a common depleted component with the HIMU, EMI and EMII lavas, respectively. We favor a model that considers the oceanic lithosphere to be that common component. We suggest that the Pb isotopic compositions of the melt inclusions reflect wall rock reaction of HIMU, EMI and EMII melts during their percolation through the oceanic lithosphere. Under these conditions, the localized rapid crystallization of olivine from primitive basalt near the reaction zone would allow the entrapment of melt inclusions with different isotopic composition.

Volatile abundances in the sub-arc mantle: insights from volcanic and hydrothermal gas discharges

Volatile emissions from arc-related volcanic and hydrothermal systems are characterized by elevated CO 2/He, N 2/He and N 2/Ar ratios compared to MORB. The source of CO 2 and N 2 is thought to be subducted sediments and oceanic crust, whereas helium is mainly derived from the mantle wedge and Ar is contributed from the atmosphere. The CO 2 and H 2O concentrations of arc-related magmas have been measured in melt inclusions from explosively erupted tephra and suggest that arc-magmas have CO 2 contents of up to ∼2000 ppm and H 2O contents up to ∼6 wt.%, supporting the idea of elevated volatile contents in the sub-arc mantle compared to MORB. We present here a new approach to estimating the volatile content of primary undegassed sub-arc magmas taking advantage of the fact that 3He originates solely from the asthenospheric mantle wedge. Recent estimates of 3He flux from arcs, 3He concentration in the crystalline mantle, and arc S, CO 2, N 2, and Ar fluxes allow to estimate concentrations of these volatiles in primitive undegassed sub-arc melts. Our results suggest ∼8 to 16 wt.% H 2O, ∼500 to 2200 ppm S, ∼3500 to ∼7600 ppm CO 2, ∼11 to 107 ppm N 2, 0.1 to 0.3 ppm Ar for undifferentiated melts produced by 20% partial melting of the crystalline mantle wedge. Although the uncertainties are large, the results for H 2O and CO 2 are generally higher than what has previously been measured in olivine-hosted melt inclusions from arcs. We reconcile this difference being due to deep volatile degassing of sub-arc magmas prior to melt inclusion entrapment at shallower levels in the crust.

Accurate quantification of melt inclusion chemistry by LA-ICPMS: a comparison with EMP and SIMS and advantages and possible limitations of these methods

Laser ablation inductively coupled plasma mass spectrometry (LA-ICPMS) has recently emerged as a powerful in situ microanalytical technique for major to trace elements in heterogeneous samples such as fluid and melt inclusions. Here, a rigorous comparison of melt inclusion (MI) data acquired by electron microprobe (EMP), ion microprobe (the secondary ion mass spectrometry, SIMS) and LA-ICPMS is used to evaluate the applicability and advantages/drawbacks of these approaches. We are specifically interested in determining if LA-ICPMS data on entire, unexposed, crystallized MI that cannot be homogenized in the lab are accurate and of a useful precision. Quantification of LA-ICPMS MI signals requires the use of an internal standard, i.e., the concentration of one element, or an element ratio, at the time of MI entrapment must be known independently, in order to derive the pure MI composition from the MI plus host mixed signal. Analysis of plagioclase-hosted glassy MI of a mid-ocean ridge basalt (MORB) sample from the East Pacific Rise illustrates that melt inclusion chemistry can be accurately quantified by LA-ICPMS, including the correction for postentrapment sidewall crystallisation of the host mineral without prior reheating in the lab. The LA-ICPMS data obtained on crystallized MI demonstrate agreement with the EMP and SIMS data on exposed glassy MI at the 1 standard deviation uncertainty level except for a few elements close to their limits of detection. LA-ICPMS data reduction schemes include the quantification of analytical uncertainty on each element of single MI. Therefore, weighted average element concentrations can be obtained for MI assemblages, at precisions that compare well with those of average element concentrations obtained by EMP and SIMS. Simple sample preparation minimizing inclusion loss through polishing combined with the analytical efficiency of 50 inclusions plus neighbouring host mineral at up to 40 elements per day enable the collection of statistically relevant datasets by LA-ICPMS. These allow to recognize nonrepresentative MI (e.g., heterogeneous entrapment). Application to individual clinopyroxene crystals from the AD79 pumice horizon of Mt. Somma-Vesuvius reveals chemical variability that exceeds the analytical precision on single melt inclusions. This variability was not obvious from the limited data set obtained by SIMS and EMP. The largest source of nonquantifiable error for EMP and SIMS data stems from the requirement of reheating the melt inclusions in the lab in order to reverse postentrapment crystallisation onto inclusion walls or growth of crystallites. For LA-ICPMS analysis of unexposed MI, the reliability with which the internal standard (IS) element concentration is known determines the quality of the data. LA-ICPMS, however, cannot analyse H 2O, F, S and Cl reliably, has higher limits of detection (LODs) than SIMS for some elements for MI below ∼25 μm, has lower spatial resolution than both EMP and SIMS and consumes much more sample per analysis. Therefore, EMP, SIMS and LA-ICPMS are complementary in MI research, and the type of application will determine the analytical method or methods of choice.

Melt Inclusions in Primitive Olivine Phenocrysts: the Role of Localized Reaction Processes in the Origin of Anomalous Compositions

Melt inclusions are small portions of liquid trapped by growing crystals during magma evolution. Recent studies of melt inclusions have revealed a large range of unusual major and trace element compositions in phenocrysts from primitive mantle-derived magmatic rocks [e.g. in high-Fo olivine (Fo > 85 mol %), spinel, high-An plagioclase]. Inclusions in phenocrysts crystallized from more evolved magmas (e.g. olivine Fo < 85 mol %), are usually compositionally similar to the host lavas. This paper reviews the chemistry of melt inclusions in high-Fo olivine phenocrysts focusing on those with anomalous major and trace element contents from mid-ocean ridge and subduction-related basalts.We suggest that a significant portion of the anomalous inclusion compositions reflects localized, grainscale dissolution-reaction-mixing (DRM) processes within the magmatic plumbing system. The DRM processes occur at the margins of primitive magma bodies, where magma is in contact with cooler wall rocks and/or pre-existing semi-solidified crystal mush zones (depending on the specific environment). Injection of hotter, more primitive magma causes partial dissolution (incongruent melting) of the mush-zone phases, which are not in equilibrium with the primitive melt, and mixing of the reaction products with the primitive magma. Localized rapid crystallization of high-Fo olivines from the primitive magma may lead to entrapment of numerous large melt inclusions, which record the DRM processes in progress. In some magmatic suites melt inclusions in primitive phenocrysts may be naturally biased towards the anomalous compositions. The occurrence of melt inclusions with unusual compositions does not necessarily imply the existence of new geologically significant magma types and/or melt-generation processes, and caution should be exercised in their interpretation.

Gradients in H2O, CO2, and exsolved gas in a large‐volume silicic magma system: Interpreting the record preserved in melt inclusions from the Bishop Tuff

Infrared spectroscopic analyses of ∼140 melt inclusions in quartz phenocrysts from the zoned Bishop rhyolitic tuff demonstrate that systematic gradients in dissolved magmatic H2O and CO2 concentrations were present during preemptive crystallization of the magma body. Melt inclusions from the earliest erupted samples contain lower H2O (5.3±0.4 wt %) and CO2 (62±37 ppm) than inclusions from the middle of the eruption (5.7±0.2 wt % H2O; 120±60 ppm CO2). Melt inclusions from late erupted samples have much lower H2O (4.1±0.3 wt %) and higher and variable CO2 (150–1085 ppm). Trace element analyses of melt inclusions by ion microprobe show that inclusions within single pumice clasts from the early and middle Bishop Tuff have an inverse correlation between CO2 and incompatible elements. This pattern indicates that the magma was gas‐saturated during crystallization, with CO2 partitioning into a coexisting gas phase. Quantitative modeling using H2O‐CO2 solubility relations reveals a preeruptive gradient in exsolved gas, with gas contents varying from ∼1 wt % in the deeper regions of the magma body to nearly 6 wt % near the top. Dissolved Cl, B, Li, and Be in melt inclusions correlate negatively with CO2. Mass balance modeling of Cl loss to exsolving H2O‐rich gas during crystallization provides strong corroborating evidence for the mass fractions of exsolved gas estimated from H2O, CO2, and trace element data. Pressures of quartz crystallization and melt inclusion entrapment calculated from inclusion H2O‐CO2 data are consistent with progressive downward tapping of a zoned magma body during the eruption. Melt inclusion gas saturation pressures, magma volume estimates, and time‐stratigraphic‐compositional relations suggest that early erupted magma was stored at the top of a downward widening magma body. Melt inclusion data and the inferred gradients in dissolved H2O, CO2 and exsolved gas in the Bishop magma body suggest that gas saturation plays an important role in the formation and subsequent preservation of compositional gradients in silicic magma reservoirs.

A study of melt inclusions at Vulcano (Aeolian Islands, Italy): insights on the primitive magmas and on the volcanic feeding system

This work presents the results of a microthermometric and EPMA-SIMS study of melt inclusions in phenocrysts of rocks of the shoshonitic eruptive complex of Vulcano (Aeolian Islands, Italy). Different primitive magmas related to two different evolutionary series, an older one (50–25 ka) and a younger one (15 ka to 1890 A.D.), were identified as melt inclusions in olivine Fo88–91 crystals. Both are characterized by high Ca/Al ratio and present very similar Rb/Sr, B/Be and patterns of trace elements, with Nb and Ti anomalies typical of a subduction zone. The two basalts present the same temperature of crystallization (1180±20 °C) and similar volatile abundances. The H2O, S and Cl contents are relatively high, whereas magmatic CO2 concentrations are very low, probably due to CO2 loss before low-pressure crystallization and entrapment of melt inclusions. The mineral chemistry of the basaltic assemblages and the high Ca/Al ratio of melt inclusions indicate an origin from a depleted, metasomatized clinopyroxene-rich peridotitic mantle. The younger primitive melt is characterized with respect to the older one by higher K2O and incompatible element abundances, by lower Zr/Nb and La/Nb, and by higher Ba/Rb and LREE enrichment. A different degree of partial melting of the same source can explain the chemical differences between the two magmas. However, some anomalies in Sr, Rb and K contents suggest either a slightly different source for the two magmas or differing extents of crustal contamination. Low-pressure degassing and cooling of the basaltic magmas produce shoshonitic liquids. The melt inclusions indicate evolutionary paths via fractional crystallization, leading to trachytic compositions during the older activity and to rhyolitic compositions during the recent one. The bulk-rock compositions record a more complex history than do the melt inclusions, due to the syneruptive mixing processes commonly affecting the magmas erupted at Vulcano. The composition and temperature data on melt inclusions suggest that in the older period of activity several shallow magmatic reservoirs existed; in the younger one a relatively homogeneous feeding system is active. The shallow magmatic reservoir feeding the recent eruptive activity probably has a vertical configuration, with basaltic magma in the deeper zones and differentiated magmas in shallower, low-volume, dike-like reservoirs.

Chloride Solubility in Felsic Melts and the Role of Chloride in Magmatic Degassing

The degassing of Cl− and H2O-bearing magmas has been investigated experimentally at ∼ 1 bar and 2000 bars and 800–1075°C by determining the solubility of H2O and Cl− in felsic to intermediate liquids variably enriched in F, P, B, C, and excess alkalis or aluminum. Chloride solubility in H2O-undersaturated silicate liquids is low and increases with increasing values of the molar [(Al+Na+Ca+Mg)/Si] of silicate liquids, F content, the abundance of network-modifying Al, and pressure. The presence of B or P in peraluminous felsic liquids has no discernible influence on Cl− solubility. The experimental solubility data are compared with H2O and Cl− in silicate melt inclusions from seven high-silica rhyolites and tin and topaz rhyolites of western North America. Small ranges in Cl− concentration, minimum Cl− contents of 600–800 ppm, large ranges in H2O content and steeply negative to near-infinite slopes in plots of H2O vs Cl− for many of the melt inclusions are similar to abundances and trends of H2O vs Cl− for volatile phase-saturated felsic liquids, suggesting that some portions of each of these magmas might have been saturated in volatile phases before melt inclusion entrapment and eruption.

Metasomatism of Cascades subarc mantle: Evidence from a rare phlogopite orthopyroxenite xenolith

Ultramafic xenoliths in the southern Washington Cascades lavas provide a rare view of the upper mantle beneath a convergent margin volcanic arc. An unusual olivine orthopyroxenite from this xenolith suite shows evidence of extreme multistage metasomatism involving infiltration of high-K 2 O silicic melts and CO 2 fluid, replacement of olivine by orthopyroxene, and entrapment of fluid, melt, and phlogopite inclusions within orthopyroxene—all prior to its entrainment in the host basaltic magma. Analyses of minerals, glass, and fluid inclusions show that the mantle source of this rock was ≥30 km below the surface and reached a temperature of at least 1000 °C—possibly in response to basaltic underplating. This region of the mantle was previously enriched in silica and potassium, but not necessarily due to subduction processes. The metasomatic melts may be derived from an accreted slab of oceanic lithosphere, as opposed to a subducting oceanic plate.

Chemical evolution of basalts from 23°N along the mid-Atlantic ridge: evidence from melt inclusions

The chemical compositions of melt inclusions in a primitive and an evolved basalt recovered from the mid-Atlantic ridge south of the Kane Fracture Zone (23°–24°N) are determined. The melt inclusions are primitive in composition (0.633–0.747 molar Mg/(Mg+Fe2+), 1.01–0.68 wt% TiO2) and are comparable to other proposed parental magmas except in having higher Al2O3 and lower CaO. The primitive melt inclusion compositions indicate that the most primitive magmas erupted in this region are not near primary magma compositions. Olivine and plagioclase microphenocrysts are close to exchange equilibrium with their respective basalt glasses, whose compositions are displaced toward olivine from 1 atm three phase saturation. The most primitive melt inclusion compositions are close to exchange equilibrium with the anorthitic cores of zoned plagioclases (An78.3-An83.1; the hosts for the melt inclusions in plagioclase) and with olivines more forsteritic (Fo89-Fo91) than the olivine microphenocrysts (the hosts for the melt inclusions in olivine). Xenocrystic olivine analyzed is Fo89 but contains no melt inclusions. These observations indicate that olivines have exchanged components with the melt after melt inclusion entrapment, whereas plagioclase compositions have remained the same since melt inclusion entrapment. Common denominator element ratio diagrams and oxide versus oxide variation diagrams show that the melt inclusion compositions, which represent liquids higher along the liquid line of descent, are related to the glass compositions by the fractionation of olivine, plagioclase and clinopyroxene (absent from the mincral assemblage), probably occurring at elevated pressures. A model is proposed whereby clinopyroxene segregates from the melt at elevated pressures (to account for its absence in the erupted lavas that have the chemical imprint of clinopyroxene fractionation). Zoned plagioclases in the erupted lavas are thought to be survivors of decompressional melting during magma ascent. Since similar primitive melt inclusions occur in olivine microphenocrysts and in the cores of zoned plagioclases, any model must account for all phases present.