Some of the stuff we are doing with melt inclusions and crystals:


Melt inclusions are bits of melt trapped inside crystals.  They are valuable magma aliquots -- in most cases they are the only eruptive products that have escaped degassing.  Ion microprobe analyses of the now glassy inclusions provide concentration data on several volatile species (H2O, CO2, S, F, Cl).  The fidelity of the melt inclusion record, especially for water, depends criticallly on the crystal cooling rate, size and diffusivity.  Much of our recent work is aimed at studying volatile diffusion in crystals and glass to get at timescales of magma ascent and cooling.  And we continue to mine melt inclusions in well-quenched tephra to obtain first-order information on the volatile contents of magmas.

Fidelity of Melt Inclusions, Magma Ascent Rate,

Water Diffusion and Water in Arc Magmas

Volatile loss from melt inclusions in pyroclasts of differing sizes.

Alex Lloyd, Terry Plank, Philipp Ruprecht, Erik Hauri and Bill Rose

Contributions to Mineralogy and Petrology, (2013) DOI 10.1007/s00410-012-0800-2

We have investigated the loss of H2O from olivine-hosted melt inclusions (MIs) by designing an experiment using tephra samples that cooled at different rates owing to their different sizes: ash, lapilli, and bomb samples that were deposited on the same day (10/17/74) of the sub-Plinian eruption of Volca´n de Fuego in Guatemala. Ion microprobe, laser ablation-ICPMS, and electron probe analyses show that MIs from ash and lapilli record the highest H2O contents, up to 4.4 wt%. On the other hand, MIs from bombs indicate up to 30 % lower H2O contents (loss of *1 wt% H2O) and 10 % post-entrapment crystallization

of olivine. This evidence is consistent with the longer cooling time available for a bomb-sized clast, up to 10 min for a 3–4-cm radius bomb, assuming conductive cooling and the fastest H diffusivities measured in olivine (D*10-9 to 10-10 m2/s). On the other hand, several lines of evidence point to some water loss prior to eruption, during magma ascent and degassing in the conduit. Thus, results point to both slower post-eruptive cooling and

slower magma ascent affecting MIs from bombs, leading to H2O loss over the timescale of minutes to hours. The important implication of this study is that a significant

portion of the published data on H2O concentrations in olivine-hosted MIs may reflect unrecognized H2O loss via diffusion. This work highlights the importance of reporting clast and MI sizes in order to assess diffusive effects and the potential benefit of using water loss as a chronometer of magma ascent.

The whole-block approach to measuring hydrogen diffusivity in nominally anhydrous minerals

Elizabeth Ferriss, Terry Plank, David Walker and Meredith Nettles

American Mineralogist, DOI: http: // (2015)

A method is developed for determining the diffusivity of infrared-active species by transmission Fourier transform infrared spectroscopy (FTIR) in samples prepared as rectangular prisms without cutting the sample. The primary application of this “whole-block” or “3D-WB” method is in measuring the diffusion of hydrogen (colloquially referred to as “water”) in nominally anhydrous minerals, but the approach is applicable to any IR-active species. The whole-block method requires developing a three-dimensional model that includes the integration of the beam signal through the sample, from rim to core to opposite rim. The analysis is carried out using both forward and tomographic inverse modeling techniques. Measurements collected from central slices cut from the whole block are simpler to interpret than whole-block measurements, but slicing requires destructive sample analysis. Because the whole-block method is nondestructive, this approach allows a time-series of diffusion experiments on the same sample.

    The potential pitfalls of evaluating whole-block measurements without correcting for path integra- tion effects are explored using simulations. The simulations demonstrate that diffusivities determined from whole-block measurements without considering path-averaging may be up to half an order of magnitude too fast. The largest errors are in fast and/or short directions, in which the diffusion pro- files are best developed. A key characteristic of whole-block measurements is that the central values in whole-block traverses always change before the concentration of the IR-active species changes in the block’s center because of signal integration that includes concentrations in the sample rims. The resulting plateau in the measurements is difficult to fit correctly without considering path integration effects, ideally by using 3D whole-block models. However, for early stages of diffusion with <50% progress, diffusivities can be accurately determined within 0.5 log units using a 1D approximation and the whole-block central plateau values because diffusivities are more dependent on profile shape than absolute concentrations.

     To test the whole-block method, a dehydration experiment was performed on an oriented piece of diopside from the Kunlun Mts with minimal zoning, cracks, or inclusions. The experiment was performed in a gas mixing furnace for 3 days at a temperature of 1000 ∞C and oxygen fugacity of 10–11.1 bar (QFM). First, whole-block analysis was performed by taking FTIR traverses in three orthogonal directions. Then, a slice was cut from the center of the sample, and hydrogen profiles were measured by FTIR and secondary ion mass spectrometry (SIMS). The results of FTIR and SIMS measurements on the slice are in good agreement both with each other and with diffusion profiles calculated based on the results of forward and inverse models of the whole-block FTIR measurements. Finally, the new method is applied to previous whole-block measurements of hydrogen diffusion in San Carlos olivine using both the forward and inverse approaches.

Why do Mafic Arc Magmas Contain

~ 4 wt% Water on Average?

Terry Plank, Katherine Kelley, Mindy Zimmer, Erik Hauri and Paul Wallace

EPSL (2013) Frontiers Article


Over the last fifteen years there has been an explosion in data on the volatile contents of magmas parental to arc volcanoes. This has occurred due to the intense study of melt inclusions, trapped primarily within olivine, as aliquots of magma that have escaped degassing during eruption. The surprising first-order result is the narrow range in H2O concentrations of the least degassed melt inclusions when averaged by volcano (based on 7 arcs for which such data exist for > 5 volcanoes:  Central America, Mexico, Kamchatka, Marianas, Cascades, Tonga and the Aleutians). Nearly all arc volcanoes are sourced with mafic magmas that contain 2-6 wt% H2O. Moreover, the average for each arc varies even less, from 3.2 (for the Cascades) to 4.5 (for the Marianas), with an average for all seven arcs of 3.8 +/- 0.5 wt% H2O.  The narrow range and common average value for H2O is in stark contrast to that for most other subduction tracers, such as Nb or Ba, which vary by orders of magnitude. 

A modulating process, either in the crust or mantle, is likely responsible for the restricted range in the H2O contents of melt inclusions. One possibility is that melt inclusions reflects vapor saturation at the last storage depth prior to eruption. Magmas rise from the mantle with variable H2O contents (> 4 wt%), start degassing at the depth of H2O-saturation, and continue to degas up until the depth at which they stall.  If the stalling depths were ~6 km, not atypical for storage depths beneath volcanoes, magmas would be saturated at ~4 wt% H2O, and melt inclusions, which become sealed during ascent, would thus record ≤ 4 wt% H2O. Another possibility is that the melting process modulates water content in the melt such that magmas rise out of the mantle with ~4 wt% H2O. A strong relationship between the water content of the source (H2Oo) and the degree of melting (F) maintains nearly constant water contents in the melt for a restricted range in mantle temperature. Magmas with 3-4 wt% H2O can be generated at 1230-1280°C and 1.5 GPa for a wide range in F and H2Oo. Crust and mantle controls may dominate in different regions and may be distinguished from coupled trace element or CO2 variations.


NanoSIMS results from olivine-hosted melt embayments: Magma ascent rate during explosive basaltic eruptions

Alex Lloyd, Terry Plank, Philipp Ruprecht, Erik Hauri, Helge Gonnermann and Bill Rose

Journal of Volcanology and Geothermal Research, 283: 1-18 (2014)

The explosivity of volcanic eruptions is governed in part by the rate at which magma ascends and degasses. Because the time scales of eruptive processes can be exceptionally fast relative to standard geochronometers, magma ascent rate remains difficult to quantify. Here we use as a chronometer concentration gradients of volatile species along open melt embayments within olivine crystals. Continuous degassing of the external melt during magma ascent results in diffusion of volatile species from embayment interiors to the bubble located at their outlets. The novel aspect of this study is the measurement of concentration gradients in five volatile elements (CO2, H2O, S, Cl, F) at fine-scale (510 μm) using the NanoSIMS. The wide range in diffusivity and solubility of these different volatiles provides multiple constraints on ascent timescales over a range of depths. We focus on four 100200 μm, olivine-hosted embayments erupted on October 17, 1974 during the sub-Plinian eruption of Volcán de Fuego. H2O, CO2, and S all decrease toward the embayment outlet bubble, while F and Cl increase or remain roughly constant. Compared to an extensive melt inclusion suite from the same day of the eruption, the embay- ments have lost both H2O and CO2 throughout the entire length of the embayment. We fit the profiles with a 1-D numerical diffusion model that allows varying diffusivities and external melt concentrations as a function of pressure. Assuming a constant decompression rate from the magma storage region at approximately 220 MPa to the surface, H2O, CO2 and S profiles for all embayments can be fit with a relatively narrow range in decompres- sion rates of 0.30.5 MPa/s, equivalent to 1117 m/s ascent velocity and an 8 to 12 minute duration of magma ascent from ~10 km depth. A two stage decompression model takes advantage of the different depth ranges over which CO2 and H2O degas, and produces good fits given an initial stage of slow decompression (0.050.3 MPa/s) at high pressure (N145 MPa), with similar decompression rates to the single-stage model for the shallower stage. The magma ascent rates reported here are among the first for explosive basaltic eruptions and demonstrate the potential of the embayment method for quantifying magmatic timescales associated with eruptions of different vigor.